University of Cambridge, Board of Graduate Studies, 4 Mill Lane, Cambridge CB2 1RZ · Tel: 01223 338389 ·  
email: student.registry@admin.cam.ac.uk · http://www.admin.cam.ac.uk/students/studentregistry/exams/submission/phd/submitting.html 
Deposit & Copying of 
Dissertation Declaration
  Board of Graduate Studies 
Please note that you will also need to bind a copy of this Declaration into your final, hardbound copy of 
thesis - this has to be the very first page of the hardbound thesis.
1 Surname (Family Name) Forenames(s) Title 
2 Title of Dissertation as approved by the Degree Committee 
In accordance with the University Regulations in Statutes and Ordinances for the PhD, MSc and MLitt Degrees, I 
agree to deposit one print copy of my dissertation entitled above and one print copy of the summary with the Secretary 
of the Board of Graduate Studies who shall deposit the dissertation and summary in the University Library under the 
following terms and conditions: 
1. Dissertation Author Declaration
I am the author of this dissertation and hereby give the University the right to make my dissertation available in print 
form as described in 2. below. 
My dissertation is my original work and a product of my own research endeavours and includes nothing which is the 
outcome of work done in collaboration with others except as declared in the Preface and specified in the text.  I hereby 
assert my moral right to be identified as the author of the dissertation. 
The deposit and dissemination of my dissertation by the University does not constitute a breach of any other 
agreement, publishing or otherwise, including any confidentiality or publication restriction provisions in sponsorship or 
collaboration agreements governing my research or work at the University or elsewhere. 
2. Access to Dissertation
I understand that one print copy of my dissertation will be deposited in the University Library for archival and 
preservation purposes, and that, unless upon my application restricted access to my dissertation for a specified period 
of time has been granted by the Board of Graduate Studies prior to this deposit, the dissertation will be made available 
by the University Library for consultation by readers in accordance with University Library Regulations and copies of 
my dissertation may be provided to readers in accordance with applicable legislation. 
3 Signature Date 
Corresponding Regulation 
Before being admitted to a degree, a student shall deposit with the Secretary of the Board one copy of his or her hard-
bound dissertation and one copy of the summary (bearing student’s name and thesis title), both the dissertation and 
the summary in a form approved by the Board.  The Secretary shall deposit the copy of the dissertation together with 
the copy of the summary in the University Library where, subject to restricted access to the dissertation for a specified 
period of time having been granted by the Board of Graduate Studies, they shall be made available for consultation by 
readers in accordance with University Library Regulations and copies of the dissertation provided to readers in 
accordance with applicable legislation. 
Antimicrobial resistance gene monitoring in aquatic 
environments 
 
 
 
 
 
 
By 
Will Rowe 
 
 
 
 
 
 
September 2015 
This dissertation is submitted for the degree of 
Doctor of Philosophy 
 
 
 
 
 
 
Magdalene College 
University of Cambridge 
 
 ii 
For the three girls I met during my PhD 
 
 iii 
Preface and declaration 
The work described in this dissertation has been carried out as part of a BBSRC Industrial 
CASE PhD studentship, sponsored by GlaxoSmithKline and the Centre for Environment, 
Fisheries and Aquaculture Science. This work was primarily carried out in the Bacterial 
Infection Group laboratory at the University of Cambridge, however several months were 
spent in the collaborating laboratories of the Centre for Environment, Fisheries and 
Aquaculture Science in Weymouth, UK and also at the Wellcome Trust Sanger Institute, UK. 
This dissertation is the result of my own work and includes nothing that is the outcome of 
work done in collaboration except as declared in the preface and specified in the formal 
acknowledgements. It is not substantially the same as any that I have submitted, or, is being 
concurrently submitted for a degree or diploma or other qualification at the University of 
Cambridge or any other University or similar institution except as declared in the preface and 
specified in the text. I further state that no substantial part of my dissertation has already 
been submitted, or, is being concurrently submitted for any such degree, diploma or other 
qualification at the University of Cambridge or any other University of similar institution except 
as declared in the preface and specified in the text. The word count of this dissertation does 
not exceed the prescribed word limit set by the Department of Veterinary Medicine, University 
of Cambridge. 
 
 
 
Will Rowe ________________________________ 
 
 iv 
Formal acknowledgements  
This table shows the contributions of others to the work in this dissertation. 
Function Person Institute 
Doctoral supervisors Gareth Pearce University of Cambridge 
- "guidance, editing and laboratory 
resources" 
Duncan Maskell University of Cambridge 
 
David Verner-Jeffreys 
Centre for Environment, Fisheries and 
Aquaculture Science 
 
Craig Baker-Austin 
Centre for Environment, Fisheries and 
Aquaculture Science 
 
Jim Ryan GlaxoSmithKline 
Publication co-authors Kate Baker Wellcome Trust Sanger Institute 
 
David Verner-Jeffreys 
Centre for Environment, Fisheries and 
Aquaculture Science 
 
Craig Baker-Austin 
Centre for Environment, Fisheries and 
Aquaculture Science 
 
Jim Ryan GlaxoSmithKline 
 
Duncan Maskell University of Cambridge 
 
Gareth Pearce University of Cambridge 
Facilitated sample collection Hedley Skelton Anglian Water 
 
Gavin Hughes Cambridge University dairy farm 
 
Gordon Cullingford Writtle College Pig Farm 
 
Nick Brown Addenbrooke's Hospital 
Provided computing support Jenny Barna University of Cambridge 
Provided laboratory and training Trevor Lawley Wellcome Trust Sanger Institute 
Assisted with method development Chris Coward University of Cambridge 
 
Mark Stares Wellcome Trust Sanger Institute 
Provided antimicrobial usage data Christianne Micallef Addenbrooke's Hospital 
Provided sequencing services Sophie Broughton Eastern Sequence and Informatics Hub 
 
Konrad Paszkiewicz Exeter Sequencing Service 
Performed LCMS analysis of effluent 
samples 
David Ratcliffe RPS Mountainheath 
Performed ribotyping of C. difficile 
isolates 
Cornelis Knetsch Leiden University 
 
 
 v 
Abstract 
This dissertation documents the development of an environmental framework for monitoring 
antimicrobial resistance gene (ARG) dissemination in the aquatic environment. The work 
opens with a review of the relevant literature and outlines the importance of an environmental 
framework for monitoring ARG dissemination as part of antimicrobial resistance risk 
assessments.  
The ability to interrogate sequencing data quickly and easily for the presence of ARGs is 
crucial in order to facilitate their monitoring in the environment. As current laboratory methods 
for the detection and surveillance of antimicrobial resistant bacteria in the environment were 
limited in their effectiveness and scope, the dissertation begins by describing the design and 
implementation of a Search Engine for Antimicrobial Resistance (SEAR), a pipeline and web 
interface for detection of horizontally-acquired ARGs in raw sequencing data. 
The suitability of metagenomic methods for monitoring the ARG content of effluents from 
faecal sources was then assessed via a pilot study of a river catchment. Novel metagenomes 
generated from effluents entering the catchment were interrogated for ARGs. The relative 
abundance of ARGs in effluents were determined to be higher relative to the background 
environment, as were sequences relating to human and animal pathogens and mobile 
genetic elements. Thus, effluents were implicated in the dissemination of ARGs throughout 
the aquatic environment. 
To determine if ARGs were potentially in use in the environment, the expression of ARGs 
within effluents was then evaluated across a series of longitudinal samples through the use of 
metatranscriptomics, and the presence of potential environmental antimicrobial selection 
pressures was examined. This demonstrated that the abundance of ARGs, as well as 
antimicrobial usage at the effluent source, was correlated with the transcription of ARGs in 
aquatic environments.  
The work described in this dissertation has also found that horizontally transmitted ARGs 
were present in pathogenic endospore-forming bacteria commonly found across the aquatic 
environment, potentially providing a mechanism for ARG persistence in the environment.  
Finally, these findings were integrated into a universal framework for monitoring ARG 
dissemination in aquatic environments and used to highlight the developments required to 
incorporate this framework into future environmental ARG research and to facilitate 
antimicrobial resistance risk assessments. 
 vi 
Table of contents 
Preface and declaration .......................................................................................................... iii 
Formal acknowledgements ..................................................................................................... iv 
Abstract .................................................................................................................................... v 
Table of contents .................................................................................................................... vi 
List of tables ............................................................................................................................. x 
List of figures .......................................................................................................................... xi 
List of appendices .................................................................................................................. xiii 
List of abbreviations .............................................................................................................. xiv 
Chapter 1. General introduction ............................................................................................ 1 
1.1. Preface ....................................................................................................................... 1 
1.2. Antimicrobial resistance .............................................................................................. 1 
1.2.1. Antimicrobial resistance genes ............................................................................. 2 
1.2.1.1. Target alteration ............................................................................................ 2 
1.2.1.1. Target bypass ................................................................................................ 2 
1.2.1.2. Barrier to antimicrobial entry .......................................................................... 3 
1.2.1.3. Antimicrobial efflux ........................................................................................ 3 
1.2.1.4. Enzyme-based modification ........................................................................... 3 
1.2.1.5. Non-enzymatic protection proteins ................................................................ 3 
1.2.2. Transfer of ARGs ................................................................................................. 3 
1.2.2.1. Plasmids and other conjugative transfer elements ........................................ 4 
1.2.2.2. Transposons, integrons and phages ............................................................. 4 
1.3. The resistome ............................................................................................................. 5 
1.3.1. Monitoring ARGs in the environment ................................................................... 7 
1.3.1.1. Bacterial cell culture ...................................................................................... 7 
1.3.1.2. Metagenomics ............................................................................................... 8 
1.3.2. Dissemination of ARGs in the environment .......................................................... 9 
1.4. Antimicrobial resistance risk assessment ................................................................. 10 
1.4.1. Monitoring ARG dissemination in the aquatic environment ................................ 12 
Chapter 2. Search Engine for Antimicrobial Resistance ...................................................... 13 
2.1. Preface ..................................................................................................................... 13 
2.2. Introduction ............................................................................................................... 13 
2.3. Materials and methods ............................................................................................. 15 
2.3.1. SEAR requirements ........................................................................................... 15 
 vii 
2.3.1.1. Reference databases .................................................................................. 15 
2.3.1.2. Hardware ..................................................................................................... 16 
2.3.2. SEAR ................................................................................................................. 16 
2.3.2.1. The pipeline ................................................................................................. 16 
2.3.2.2. Pipeline outputs ........................................................................................... 19 
2.3.3. Demonstrating SEAR utility ................................................................................ 19 
2.3.3.1. Data sets and parameters used in this study ............................................... 19 
2.3.3.2. Novel environmental metagenomes ............................................................ 20 
2.3.3.3. Pre-existing metagenomic and clinical isolate data ..................................... 21 
2.4. Results ...................................................................................................................... 21 
2.4.1. Discrimination of ARG presence and abundance between environmental 
metagenomes ................................................................................................................. 22 
2.4.2. Efficacy of SEAR for detecting ARGs in human faecal microbiomes ................. 24 
2.4.3. Accuracy of SEAR ARG detection using clinical isolate sequencing data .......... 24 
2.5. Discussion ................................................................................................................ 25 
2.6. Conclusion ................................................................................................................ 27 
Chapter 3. Comparative metagenomics reveals a diverse range of antimicrobial resistance 
genes in effluents entering a river catchment ........................................................................ 28 
3.1. Preface ..................................................................................................................... 28 
3.2. Introduction ............................................................................................................... 28 
3.3. Methods .................................................................................................................... 30 
3.3.1. Sample collection and DNA sequencing ............................................................ 30 
3.3.2. Bioinformatic analyses ....................................................................................... 30 
3.3.2.1. Identification of ARGs .................................................................................. 30 
3.3.2.2. Identification of mobile genetic elements ..................................................... 30 
3.3.2.3. Abundance analysis .................................................................................... 31 
3.3.2.4. Taxonomic profiling and pathogen detection ............................................... 31 
3.4. Results ...................................................................................................................... 32 
3.4.1. Metagenome analysis ........................................................................................ 32 
3.4.2. Identification of antimicrobial resistance genes .................................................. 32 
3.4.3. Identification of mobile genetic elements ........................................................... 36 
3.4.4. Taxonomic profiling and pathogen detection ...................................................... 40 
3.5. Discussion ................................................................................................................ 44 
3.6. Conclusion ................................................................................................................ 46 
Chapter 4. Expression of ARGs in effluents ........................................................................ 47 
 viii 
4.1. Preface ..................................................................................................................... 47 
4.2. Introduction ............................................................................................................... 47 
4.3. Methods .................................................................................................................... 49 
4.3.1. Sample collection, DNA and RNA sequencing ................................................... 49 
4.3.1.1. Antimicrobial residue testing ........................................................................ 50 
4.3.2. Bioinformatic analyses ....................................................................................... 50 
4.3.2.1. Identification of ARGs, MGEs and abundance analysis .............................. 50 
4.3.2.2. Antimicrobial usage and statistics ................................................................ 50 
4.3.2.3. Taxonomic profiling and pathogen detection ............................................... 51 
4.4. Results ...................................................................................................................... 51 
4.4.1. Metagenome and metatranscriptome analysis ................................................... 51 
4.4.2. Correlation of ARG abundance to ARG transcript abundance ........................... 53 
4.4.3. Antimicrobial residues ........................................................................................ 55 
4.4.4. Expression of ARGs in response to hospital antimicrobial usage ...................... 58 
4.4.5. Taxonomic profiling and pathogen detection ...................................................... 61 
4.4.6. MGEs ................................................................................................................. 63 
4.4.1. Abundance of ARGs over time ........................................................................... 66 
4.5. Discussion ................................................................................................................ 68 
4.6. Conclusion ................................................................................................................ 72 
Chapter 5. Bacterial endospores present in the environment harbour antimicrobial 
resistance genes .................................................................................................................... 73 
5.1. Preface ..................................................................................................................... 73 
5.2. Introduction ............................................................................................................... 73 
5.2.1. Endospore-forming bacteria ............................................................................... 74 
5.2.1.1. Clostridium difficile ....................................................................................... 75 
5.2.1.2. Clostridium perfringens ................................................................................ 76 
5.2.2. Chapter hypothesis ............................................................................................ 76 
5.3. Materials and methods ............................................................................................. 77 
5.3.1. Sample collection ............................................................................................... 77 
5.3.2. Isolate culture ..................................................................................................... 77 
5.3.3. DNA extraction and 16S rRNA sequencing ........................................................ 78 
5.3.4. DNA extraction for whole genome sequencing .................................................. 78 
5.3.5. Ribotyping .......................................................................................................... 79 
5.3.6. Metagenome sequencing ................................................................................... 79 
5.3.7. Construction of high-quality draft genomes ........................................................ 79 
 ix 
5.3.8. Bioinformatic analysis ........................................................................................ 80 
5.4. Results ...................................................................................................................... 80 
5.4.1. Isolation and identification of endospore-forming bacteria ................................. 80 
5.4.1. Unclassified environmental isolates ................................................................... 85 
5.4.1. Clostridium difficile isolates ................................................................................ 88 
5.4.1.1. Phylogenetic analysis .................................................................................. 88 
5.4.1.2. Antimicrobial resistance genes .................................................................... 92 
5.4.1.2.1. Aminoglycoside resistance genes ......................................................... 94 
5.4.1.2.2. Tetracycline resistance genes ............................................................... 96 
5.4.1.2.3. Other types of antimicrobial resistance genes ...................................... 98 
5.4.2. Clostridium perfringens isolates ......................................................................... 98 
5.4.2.1. Phylogenetic analysis .................................................................................. 98 
5.4.2.2. Antimicrobial resistance genes .................................................................. 102 
5.4.2.2.1. Macrolide resistance genes ................................................................ 104 
5.4.2.2.2. Tetracycline resistance genes ............................................................. 106 
5.4.3. Annotating endospore-formers from metagenomes ......................................... 108 
5.5. Discussion .............................................................................................................. 109 
5.6. Conclusion .............................................................................................................. 112 
Chapter 6. General discussion .......................................................................................... 113 
6.1. Preface ................................................................................................................... 113 
6.2. Summary of findings and dissertation outcomes .................................................... 113 
6.2.1. Establishing an environmental framework for monitoring ARG dissemination in 
the aquatic environment ............................................................................................... 114 
6.2.1.1. Summary of findings .................................................................................. 114 
6.2.1.2. The environmental framework ................................................................... 116 
6.2.2. Future work on an environmental framework for monitoring ARG dissemination
 118 
6.2.2.1. Assigning significance to antimicrobial resistance potential....................... 118 
6.2.2.2. Understanding ARG dynamics .................................................................. 120 
6.2.2.3. Incorporating additional factors and technologies into the framework ....... 122 
6.2.2.4. Problems remaining ................................................................................... 124 
6.3. Overall conclusions of this work.............................................................................. 125 
References .......................................................................................................................... 127 
 x 
List of tables 
Table 2.1 List of parameters and default settings for SEAR. ................................................. 20 
Table 2.2 Example runtimes for SEAR. ................................................................................. 22 
Table 2.3 SEAR detection of ARGs in human faecal microbiomes. ....................................... 24 
Table 2.4 Accuracy of SEAR ARG detection using clinical isolate sequencing data ............. 25 
Table 3.1 Summary of the metagenomes used in the work of this chapter. .......................... 32 
Table 3.2 Antimicrobial resistance gene analysis. ................................................................. 34 
Table 3.3 Mobile genetic element analysis. ........................................................................... 38 
Table 3.4 Pathogen analysis of effluents. .............................................................................. 42 
Table 4.1 Summary of metagenomes used in this work. ....................................................... 52 
Table 4.2 Summary of metatranscriptomes used in this work ................................................ 52 
Table 4.3 Correlation of ARG and transcript abundance for all hospital effluent samples, 
according to antimicrobial class ...................................................................................... 55 
Table 4.4 Correlation of ARG and transcript abundance for all farm effluent samples, 
according to antimicrobial class. ..................................................................................... 55 
Table 4.5 LCMS results for hospital and farm effluent samples ............................................. 57 
Table 5.1 All endospore-forming bacteria successfully isolated from environmental samples 
and processed for whole genome sequencing. .............................................................. 82 
Table 5.2 Unclassified environmental isolates and their identified ARGs. ............................. 86 
Table 5.3. Clostridium difficile isolates and their identified ARGs. ......................................... 93 
Table 5.4 Clostridium perfringens isolates and their identified ARGs. ................................. 103 
 
 
 xi 
List of figures 
Figure 1.1 Graphical overview of potential routes of ARG dissemination .............................. 11 
Figure 2.1 Screen shot of SEAR web interface. ..................................................................... 17 
Figure 2.2 SEAR results for environmental metagenomes. ................................................... 23 
Figure 3.1 Abundance of ARGs found in each effluent sample ............................................. 33 
Figure 3.2 Abundance of MGEs found in each effluent sample ............................................. 37 
Figure 3.3 Metagenomic phylogenetic analysis and annotation of potential bacterial 
pathogens. ...................................................................................................................... 41 
Figure 4.1 Linear regression analysis of ARG abundance against corresponding transcript 
abundance, for ARGs detected in all hospital effluent samples ...................................... 53 
Figure 4.2 Linear regression analysis of ARG abundance against corresponding transcript 
abundance, for ARGs detected in all farm effluent samples ........................................... 54 
Figure 4.3 Linear regression analysis of ARG transcript abundance against hospital 
antimicrobial usage ......................................................................................................... 59 
Figure 4.4 Scree plot showing the variance observed for the three principal components. ... 60 
Figure 4.5 Biplot showing how the initial variables contribute to principal components. ........ 61 
Figure 4.6 Graph showing the number of pathogenic bacteria species and the total species 
number, for both hospital and farm effluents, over time. ................................................. 62 
Figure 4.7 A. Column chart showing the ten most commonly identified pathogen species in 
the hospital and farm effluent metagenomes. B. Venn diagram depicting the number 
pathogen species found in the combined hospital and farm effluent datasets. ............... 63 
Figure 4.8 A. Graph showing the abundance of ARGs and MGEs, in relation to total species 
number, for hospital effluent over time. B. Graph showing the abundance of ARGs and 
MGEs, in relation to total species number, for farm effluent over time............................ 65 
Figure 4.9 A. Graph showing the abundance of ARGs, grouped by antimicrobial class, in 
hospital effluent metagenomes. B. Graph showing the abundance of ARGs, grouped by 
antimicrobial class, in farm effluent metagenomes. ........................................................ 67 
Figure 5.1 Genome synteny of potentially novel bacterial isolates, related to C. butyricum... 87 
Figure 5.2 Whole genome phylogenetic tree of Clostridium difficile isolates.......................... 89 
Figure 5.3 Rarefaction curves for C. difficile core genome analysis ...................................... 90 
Figure 5.4 Tanglegram comparing two phylogenetic trees for the C. difficile isolates ............ 91 
Figure 5.5 Genome synteny between the C. difficile reference 630 genome and contig 5 of 
isolate #26 ...................................................................................................................... 95 
 xii 
Figure 5.6. A. Genome synteny among the C. difficile reference 630 genome, the conjugative 
transposon Tn5397, and contig 18 of isolate #26. B. Genome synteny between the C. 
difficile reference 630 genome and contig 18 of isolate #26. .......................................... 97 
Figure 5.7 Phylogenetic tree of Clostridium perfringens isolates. .......................................... 99 
Figure 5.8 Rarefaction curves for C. perfringens core genome analysis.............................. 100 
Figure 5.9 Tanglegram comparing two phylogenetic trees for the C. perfringens isolates ... 101 
Figure 5.10 Comparison between the C. perfringens reference ATCC13124 genome and 
contig 9 of isolate #15. .................................................................................................. 104 
Figure 5.11 Genome synteny between the C. perfringens reference ATCC13124 genome, a 
construct of contigs 42, 31 and 28 of isolate #15 and contig 31 of isolate #15. ............ 105 
Figure 5.12 Comparison between the C. perfringens reference ATCC13124 genome and the 
tetPA-containing contig of isolate #23. .......................................................................... 106 
Figure 5.13 A. Genome synteny between the C. perfringens plasmid pCW3 and the tetPB 
and tetPA containing contig of isolate #14. B. Circularised representation of tetPB-
containing plasmid. ....................................................................................................... 107 
Figure 5.14 Genome synteny between the Clostridium sticklandii reference genome (GI: 
310657316) and the metagenome contigs of the pig farm lagoon effluent. .................. 108 
Figure 6.1 Proposed environmental framework for monitoring ARG dissemination and 
assigning antimicrobial resistance potential to environmental samples. ....................... 117 
 
 
 xiii 
List of appendices 
Appendix 1. Manuscript and source code for SEAR. ........................................................... 146 
Appendix 2. River Cam catchment area. ............................................................................. 147 
Appendix 3. List of sequencing data generated and used in these studies. ........................ 148 
Appendix 4. Accession numbers of sequences used in these studies. ................................ 153 
Appendix 5. SEAR ARG detection using clinical isolate sequence data. ............................. 162 
Appendix 6. Evaluation of metagenome preparation methodologies for aquatic samples. .. 170 
 
 
  xiv 
List of abbreviations 
ABACAS Algorithm-Based Automatic Contiguation of Assembled Sequences 
AMD Advanced Molecular Detection 
ARDB Antibiotic Resistance Genes Database 
ARG Antimicrobial Resistance Gene 
ARP Antimicrobial Resistance Potential 
BHI Brain-Heart Infusion 
BLAST Basic Local Alignment Search Tool 
BWA Burrows-Wheeler Aligner 
CARD Comprehensive Antimicrobial Resistance Database 
CDAD Clostridium difficile Associated Disease 
CDI Clostridium difficile Infection 
contig contiguous sequence 
ENA European Nucleotide Archive 
ESBL Extended Spectrum Beta Lactamase 
Gbp Giga base-pairs 
HMP Human Microbiome Project 
MGE Mobile Genetic Element 
MIC Minimum Inhibitory Concentration 
MLS Macrolide / Lincosamide / Streptogramin 
PBS Phosphate Buffered Saline 
PCA Principal Component Analysis 
PCR Polymerase Chain Reaction 
RAC Repository of Antibiotic resistance Cassettes 
RDP Ribosomal Database Project 
RPKM Reads Per Kilobase per Million 
SEAR  Search Engine for Antimicrobial Resistance 
SNP Single Nucleotide Polymorphism 
WWTP Wastewater Treatment Plant 
  
  1 
Chapter 1. General introduction 
1.1. Preface 
This dissertation documents the development of an environmental framework for monitoring 
antimicrobial resistance gene (ARG) dissemination in the aquatic environment. The work 
described within includes the development of methodology and analysis tools, the 
identification of factors that may influence antimicrobial resistance gene dissemination and 
mechanisms by which they may persist in the environment. 
This chapter contains a review of the current literature concerning what is known about ARGs 
in the environment; identifying appropriate technologies to facilitate ARG monitoring, 
assessing methodology and analysis tools requiring development and determining factors 
that may influence ARG dissemination in the aquatic environment. 
 
1.2. Antimicrobial resistance 
Since the inception of applying antimicrobial compounds as therapeutic and prophylactic 
agents in human and veterinary medicine, antimicrobial resistance has remained a persistent 
issue in the effective treatment of bacterial infections and the sustainability of antimicrobial 
drugs as therapeutic agents. The ever increasing demand for antimicrobials coupled with 
their over, under and misuse has resulted in bacteria evolving a plethora of mechanisms for 
rendering antimicrobial compounds ineffective and has propagated the occurrence of 
resistance within bacterial communities (Davies and Davies, 2010b). Multiple antimicrobial 
resistances present major health and economic concerns as the decreasing efficacy of 
antimicrobial compounds require treatment regimens to utilise alternative drugs and 
increased dosages, which have the potential for more costly or unsuccessful treatment of 
bacterial infections (Eliopoulos et al., 2003). 
There has been considerable research effort to address the clinical ramifications of 
antimicrobial resistance and its mitigation. Current preventative measures against 
antimicrobial resistance development are based on infection control, altered or restricted use 
of specific antimicrobials and also through ensuring optimum clinical practice (i.e. correct 
dosages, durations and drug choice according to sensitivity testing). However, there remains 
a significant lack of information concerning the ecological risks from antimicrobial resistance 
  2 
present in the environment and the dissemination of causative agents, with limited 
regulations in place to mitigate these risks.  
This review of the literature will focus on evaluating the causative agents of antimicrobial 
resistance, i.e. antimicrobial resistance genes. The various resistance mechanisms of ARGs 
will be considered, along with details on their propagation and the concept and importance of 
the reservoir of ARGs in the environment - the resistome. The review will then progress to 
consider the methodology available to detect ARGs and how this can be used to determine 
the extent of ARG dissemination in the environment. Finally, the literature review will examine 
the impact of ARGs in the environment, highlighting the risks to human and animal health 
from their dissemination and the shortfalls in current monitoring approaches that this 
dissertation primarily aims to address. 
 
1.2.1. Antimicrobial resistance genes 
ARGs encompass genes that confer a resistant phenotype in bacteria to one or more 
antimicrobial compounds. There are several mechanisms by which an ARG can confer 
resistance: 
 
1.2.1.1. Target alteration 
An ARG can be a mutated version of a gene that encodes the protein target of an 
antimicrobial, thus decreasing the efficacy of the antimicrobial through a reduction in it’s 
ability to bind to the protein target. An example of this mechanism is the single nucleotide 
polymorphism (SNP) mutation of the DNA gyrase gene (gyrA), which alters the gyrase 
protein that is the target of fluoroquinolone antimicrobials. Consequently the mutated gyrase 
confers resistance by reduced fluoroquinolone binding (Hooper and Jacoby, 2015). 
 
1.2.1.1. Target bypass 
In addition to target alteration, an ARG can also confer resistance by bypassing the 
antimicrobial target completely. This is exemplified by a cluster of glycopeptide ARGs 
(vanHAX) that result in a different bacterial peptidoglycan cell wall structure to which 
glycopeptides bind with approximately 1000-fold lower affinity (Henson et al., 2015). 
 
  3 
1.2.1.2. Barrier to antimicrobial entry 
Antimicrobial compounds can be prevented from entering a bacterial cell entirely via ARGs 
that encode outer membrane proteins. This is a form of intrinsic resistance (Cox and Wright, 
2013). 
 
1.2.1.3. Antimicrobial efflux 
Antimicrobial efflux from a cell, via transporters such as multiple trans-membrane spanning 
proteins and resistance nodulation cell division efflux systems, can prevent certain 
antimicrobials reaching their targets and results in antimicrobial resistance (Li et al., 2015b).  
 
1.2.1.4. Enzyme-based modification 
ARGs can encode enzymes that modify the antimicrobial target (e.g. via methylation) or the 
antimicrobial itself. A common example of an ARG that encodes an antimicrobial-modifying 
enzyme is the betalactamase group of hydrolases that target betalactam antimicrobials (Tang 
et al., 2014). 
 
1.2.1.5. Non-enzymatic protection proteins 
Major sources of antimicrobial resistance are accounted for by ARGs that encode target-
binding protection proteins. An example of this mechanism is the class of tetracycline ARGs 
that encode ribosomal protection proteins that bind to the ribosome and prevent tetracycline 
antimicrobials from inhibiting protein synthesis (Nguyen et al., 2014a). 
 
1.2.2. Transfer of ARGs 
Bacteria can obtain ARGs spontaneously via gene mutations or they can acquire them, either 
vertically or horizontally. Vertical inheritance via cell division allows bacteria to transfer ARGs 
to their progeny. In addition to this, and of greater concern due to the rapid emergence of 
antimicrobial resistant pathogens, horizontal gene transfer can facilitate the acquisition of 
ARGs by unrelated bacteria. 
  4 
Mobile genetic elements (MGEs) are a group of genetic constructs that facilitate the 
movement of DNA within genomes or between bacterial cells (horizontal gene transfer). 
Horizontal gene transfer can occur via several processes (transformation, transduction or 
conjugation), depending on the type of MGE (Frost et al., 2005). The major types of MGE 
that are implicated in the mobilisation of ARGs are: 
 
1.2.2.1. Plasmids and other conjugative transfer elements 
Conjugative plasmids are stable, mobile elements that are able to replicate with autonomy. 
Plasmids can be grouped on the basis of their replication machinery; plasmids with the same 
type of machinery are unable to co-habit the same bacterial cell and this “incompatibility type” 
is the basis of a major form of plasmid classification (Frost et al., 2005). Due to the replication 
autonomy of plasmids, their ability to transfer to new hosts via conjugation and their ability to 
carry a wide array of genes, plasmids are key factors in the horizontal transfer of ARGs 
(Svara and Rankin, 2011). Non-conjugative plasmids can also harbour ARGs but require a 
conjugative plasmid to enable their transfer (O'Brien et al., 2015).  
 
1.2.2.2. Transposons, integrons and phages 
Transposons are MGEs that contain gene cassettes, a transposase enzyme and terminal 
recombination regions that allow transposons to incorporate into and out of host genomes 
(Alekshun and Levy, 2007). Integrons are similar in structure to transposons and are genetic 
elements that can capture and rearrange gene cassettes (including ARGs) through the use of 
a site-specific tyrosine recombinase (integrase). Integrons are also able to promote 
transcription of their gene cassettes and they acquire their mobility through association with 
conjugative plasmids and transposons (Stalder et al., 2012, Escudero et al., 2015). 
Phages (bacteriophages) are MGEs that facilitate horizontal gene transfer via transduction; 
phages can infect a bacterial cell and incorporate their genetic material into the host genome, 
then replicating as part of the host (lysogenic cycle) or multiplying within the cell (lytic cycle) 
(Balcazar, 2014). ARGs have been found in the virome of environmental samples and 
phages have been implicated in the transfer of ARGs among bacteria (Colomer-Lluch et al., 
2011). 
 
  5 
There is concern that MGEs are a driving force behind the acquisition of antimicrobial 
resistances by pathogenic bacteria in a clinical setting (Canica et al., 2015). Analyses of 
isolates that pre-date the “antibiotic era” reveal a paucity of ARGs (Hughes and Datta, 1983), 
however analyses of modern day clinical isolates reveal a high incidence of ARGs and MGEs 
(Sunde et al., 2015). Indeed, the risks associated with MGEs and the horizontal acquisition of 
antimicrobial resistance is illustrated by the frequent finding that a pathogenic species of 
bacteria has incorporated one or more ARGs into its genome. For example, an isolate of the 
pathogen Yersinia pestis (the causative agent of plague) was found to have acquired a MGE 
that conferred resistance to six classes of antimicrobial drugs, including the two drugs that 
are frequently used to treat Y. pestis infection (tetracycline and chloramphenicol) (Galimand 
et al., 1997, Pan et al., 2008). Another example of a pathogen acquiring antimicrobial 
resistance through MGEs and that has resulted in great clinical concern is the 
enterobacteriaceae, Klebsiella pneumoniae. K. pneumoniae  has been known to carry MGE-
associated ARGs and has recently been found to have acquired carbapenamase ARGs via 
horizontal gene transfer  (Robilotti and Deresinski, 2014). 
 
1.3. The resistome 
The concept of the “antimicrobial resistome” was proposed by D’Costa et al. (2006) and is 
designed to encompass all ARGs existing in pathogenic bacteria and non-pathogenic 
bacteria present in both the clinical setting and environment, as well as the genes that can 
give rise to antimicrobial resistance (precursor elements) (D'Costa et al., 2006). The 
resistome is an important concept as it does not focus solely on resistant pathogens or 
instances of clinical resistance, due in part to the growing evidence that environmental 
bacteria are originating a large amount of the ARGs that are emerging in the clinical setting 
(Allen et al., 2010). The concept of the resistome provides a useful way of linking a reservoir 
of ARGs to the clinical incidence of antimicrobial resistance. 
Research into the resistome as well as the role of environmental bacteria in the emergence 
of antimicrobial resistance has shown instances of antimicrobial resistance in non-pathogenic 
bacteria, offering explanation as to the origin of certain ARGs. Studies of environmental 
bacteria found in soil (which are the originators of many clinically used antimicrobials) have 
isolated bacteria that can subsist on clinically relevant antimicrobials (some of these resistant 
bacteria were also closely related to human pathogens) (Dantas et al., 2008). It is understood 
that many anti-producing bacteria may have evolved ARGs to enable self-resistance to the 
  6 
antimicrobial compounds that they produce. For instance, the glycopeptide resistance cluster 
vanHAX, a mechanism that plays a role in the widespread clinical cases of vancomycin 
resistant enterococci (Courvalin, 2006), appears to have originated in glycopeptide-producing 
bacteria for self-resistance. The vanHAX ARGs result in altered cell wall biochemistry, 
causing glycopeptide antimicrobials to have a low affinity for the peptidoglycan chain 
(Marshall and Wright, 1997, Marshall et al., 1999). It should be stated that the production of 
antimicrobial compounds by bacteria, and consequently the possession of ARGs for self-
resistance, is thought to allow the bacteria to be more competitive for resources in a given 
environment. However, it has also been postulated that antimicrobial secretion by bacteria is 
used as an inter-cellular signalling mechanism (Yim et al., 2007). 
Given the wealth of ARGs in the resistome of a particular environment, as well as the 
presence of MGEs that carry ARGs, the environment is postulated to serve as a reservoir for 
ARGs that could be acquired by clinically relevant pathogens. An example of the interplay 
between environmental reservoirs of ARGs and antimicrobial resistant pathogenic bacteria in 
the clinical setting is that of the extended spectrum beta-lactamase (ESBL) blaCTX-M family of 
ARGs (Lahlaoui et al., 2014). The blaCTX-M family of ARGs has emerged as a major clinical 
problem as they confer resistance to third-generation cephalosporins. The blaCTX-M ARGs 
share low identity to other beta-lactamase ARGs and it was proposed that blaCTX-M ARGs 
emerging in the clinical setting were a new class of beta-lactamase. However, later studies 
found that the blaCTX-M family of beta-lactamases share a high level of similarity to a beta-
lactamase ARG identified in a species of environmental bacteria closely related to the 
Kluyvera genus (Decousser et al., 2001, Olson et al., 2005). Importantly, the flanking regions 
of the blaCTX-M ARGs that have been found in pathogens appear to originate from Kluyvera 
chromosomes, suggesting the transfer of a clinically relevant ARG from environmental 
bacteria to clinically relevant pathogenic bacteria (Canton and Coque, 2006). 
It should be noted that despite a shared resistome between pathogens and environmental 
bacteria as well as evidence that indicates transfer of ARGs between these two groups, far 
greater research is needed to explore the link between these groups and assess the role of 
ARG dissemination in the environment. 
 
  7 
1.3.1. Monitoring ARGs in the environment 
With an increasing interest in the role of environmental sources of ARGs in the development 
of clinically relevant antimicrobial resistance, the methodology to facilitate the monitoring of 
ARGs in environmental samples has seen increased development over recent years. 
 
1.3.1.1. Bacterial cell culture 
Bacterial cell culture can be used to identify bacteria within a given sample through use of 
phenotypic testing (e.g. Gram staining, other biochemical testing, and colony morphology) 
and downstream molecular testing such as both polymerase chain reaction (PCR) testing 
and whole genome sequencing. Successfully cultured bacterial isolates can then be 
interrogated for ARG carriage through techniques such as antimicrobial susceptibility testing 
and PCR testing. 
However, there are several issues concerning the use of bacterial cell culture for bacterial 
identification or the downstream testing of ARGs. Importantly, only a relatively small 
proportion of bacteria can be cultured using a given type of laboratory media, therefore 
culturing is not suitable for large-scale monitoring of ARGs in diverse bacterial communities 
(Gasc et al., 2015). In addition to this, phenotypic testing of cultured bacterial isolates may 
reveal unknown combinations of characteristics that mean cultured bacteria cannot be 
identified (Stewart, 2012). Finally, in terms of ARG carriage, phenotypic testing cannot 
determine the exact ARG responsible for any identified resistance and relies on the ARG 
being expressed. Also, the use of PCR to identify ARGs in isolates can only test for a pre-
determined ARG panel, is limited in its multiplexing capability and cannot generally identify 
novel ARGs. 
In light of the disadvantages of culture-based approaches, the use of specifically designed 
experimental protocols such as tailored growth media (Rasmussen et al., 2008) can be used 
to apply bacterial cell culture to test specific hypotheses or be used in conjunction with 
downstream testing such as antimicrobial sensitivity testing to form part of a wider 
assessment (Caplin et al., 2008). 
 
  8 
1.3.1.2. Metagenomics 
A major advancement in the fields of microbiology and molecular ecology came with the 
advent of next generation sequencing technologies. This technology has allowed for the 
study of whole bacterial communities in a given environment using a technique termed 
metagenomics (National Center for Biotechnology Information, 2007). 
Metagenomics involves the sequencing of the collective DNA isolated from an environmental 
sample. The term originally encompassed 16S phylogenetic community profiling, whole 
genome shotgun metagenomics and functional metagenomics. 16S community profiling 
involves sequencing amplicons of the small ribosomal subunit (16S) of the collective bacteria 
in a sample to assess phylogenetic diversity (Degnan and Ochman, 2012). Whole genome 
shotgun metagenomics instead uses the total DNA from a given sample and can be used to, 
amongst other applications, identify and compare specific genes in microbial communities 
(Tringe et al., 2005). Functional metagenomics, which involves the cloning of metagenomic 
DNA into vectors (such as E.coli) and subsequent selective testing, is a powerful tool that can 
be used in the identification of novel ARGs in environmental samples as it does not require 
prior knowledge of ARGs thought to be present and can be applied to un-cultivable bacteria 
(Mullany, 2014). Sequencing the DNA insert of any vector that grows on the selective media 
identifies novel ARGs; this DNA sequence is then associated with resistance to the antibiotic 
in the selection media and is interrogated for the gene or genes that may be responsible for 
the observed phenotype (either by homology searching or de novo gene prediction). 
However, functional metagenomics is a resource-intensive tool that requires a large amount 
of time to create appropriate cloning libraries, test susceptibility to a wide-range of 
antimicrobials and characterise any identified ARGs (Perron et al., 2015). 
There has been a gradual reduction in the number of 16S community profiling metagenomic 
studies and a transition to whole genome shotgun metagenomics (hereafter referred to as 
just metagenomics) (Davenport and Tummler, 2013). The development of metagenomic 
analysis methodology such as species-specific biomarkers to facilitate the taxonomic binning 
of metagenomic sequencing (Segata et al., 2011) has resulted in metagenomics becoming a 
powerful tool in the analysis of environmental samples. The constant augmentation of 
sequence databases and the development of rapid metagenomic analysis pipelines such as 
MG-RAST (Meyer et al., 2008) and QIIME (Caporaso et al., 2010), have resulted in 
metagenomics becoming a viable tool for microbial monitoring applications. Indeed, 
  9 
metagenomics is being used to help answer specific questions that are being asked of 
microbial communities in the environment (Mason et al., 2012). 
 
1.3.2. Dissemination of ARGs in the environment 
Antimicrobial resistant bacteria, as well as un-metabolised antimicrobials and their residues, 
have been widely reported in the terrestrial and aquatic environments and are likely to have 
been distributed via a variety of sources. Effluents from wastewater treatment plants 
(WWTPs), as well as from hospital, farm, aquaculture and pharmaceutical manufacturing 
waste, have been shown to harbour significant levels of antimicrobial compounds (Hirsch et 
al., 1999, Golet et al., 2001, Dolliver and Gupta, 2008, Charpentier et al., 2011, Kristiansson 
et al., 2011, Kümmerer and Henninger, 2003) and antimicrobial resistant bacteria (Schwartz 
et al., 2003, Fuentefria et al., 2011, Heuer et al., 2002, Caplin et al., 2008) that may be a 
direct result of anthropogenic activity. Also, the agricultural application of effluents from 
medicated animals (Jørgensen and Halling-Sørensen, 2000) and the use of antimicrobials in 
crop protection (McManus et al., 2002) has been shown to lead to the terrestrial 
accumulation of antimicrobial compounds and antimicrobial resistant bacteria (Forsberg et 
al., 2012). These in turn can enter watercourses subsequent to rainfall, drainage and 
leaching (Dolliver and Gupta, 2008, Kemper, 2008). This widespread distribution of 
antimicrobial resistant bacteria and antimicrobial selection pressures in the environment is 
likely to have resulted in the augmentation of environmental reservoirs of ARGs and the 
resistome. 
Indeed, several studies have demonstrated the presence of ARGs in a wide-ranging 
selection of environmental samples using different techniques e.g. many experiments have 
used PCR to identify clinically relevant ARGs in environmental samples, such as 
glycopeptide ARGs in agricultural and garden soil samples (Guardabassi and Agerso, 2006) 
and tetracycline ARGs in effluents of urban residential areas (Li et al., 2015a). Other studies 
have utilised metagenomics to interrogate environments such as polluted lakes for ARGs that 
may have arisen due to the identified pollution sources (such as pharmaceutical 
manufacturing waste effluents) (Bengtsson-Palme et al., 2014). There is also an ever-
increasing body of scientific evidence that ARGs are present in a plethora of aquatic 
environments and there are numerous sources that are likely to contribute to these 
environmental reservoirs (Zhang et al., 2009). However, there are no studies that have 
  10 
assessed the factors contributing to ARG dissemination in a single defined aquatic 
environment over a period of time. 
Environmental reservoirs of ARGs, in conjunction with the global contamination of the 
biosphere with antimicrobial products, may allow for the propagation of antimicrobial 
resistances that could potentially impact on human and animal health. Several studies have 
speculated transmission routes of ARGs from the environment to humans and animals, with 
ARGs found to be present in treated drinking water and food sources (Schwartz et al., 2003, 
Pruden et al., 2006, Rodriguez et al., 2006) suggesting important clinical relevance to the fate 
of ARGs in the environment. 
The impact of the environmental release of ARGs on human health and the evolution of 
environmental bacteria is beginning to be highlighted (Martinez, 2009). As a result of this 
there is an increased awareness that the characterisation of ARGs in the environment is 
required to enable a detailed understanding of antimicrobial resistance. With the recent 
recognition of ARGs as being environmental contaminants (Pruden et al., 2006, Auerbach et 
al., 2007), it is vital that the dissemination of ARGs is monitored in order to assess the 
potential risk regarding the emergence and transmission of antimicrobial resistance. 
 
1.4.  Antimicrobial resistance risk assessment 
Recent reviews of the emergence and spread of antimicrobial resistance have highlighted 
current research aims that can be summarised into three broad components: i. the 
emergence of resistance in the environment, ii. antimicrobial resistance dissemination and iii. 
antimicrobial resistance transmission to humans (Berendonk et al., 2015). There are 
increasing calls to incorporate these components into overarching antimicrobial resistance 
risk assessments that detail potential routes dissemination, such as those represented in 
Figure 1.1. The purpose of a risk assessment being to identify and mitigate the factors that 
are contributing to the emergence and spread of antimicrobial resistance. 
 
  11 
 
Figure 1.1 Graphical overview of potential routes of ARG dissemination. The figure is comprised of 
three components, which are linked by ARG transfer. The coloured arrows indicate transfer of ARGs: 
purple represents transfer within human populations, green represents transfer of ARGs within the 
environment and non-human animal species and orange represents ARG dissemination. 
A core principal behind a unified antimicrobial resistance risk assessment is that the three 
components can be linked by ARG transfer; depicted in the figure by coloured arrows that 
correspond to the ARG source and direction of ARG transfer (Figure 1.1). As illustrated by 
the figure, the dissemination of ARGs plays a pivotal role in linking the emergence of 
antimicrobial resistance in the environment and the transmission of antimicrobial resistance 
to humans. As outlined above, ARGs have been found extensively in effluents that link faecal 
sources (both human and animal) to the wider aquatic environment. However, these studies 
have not monitored the dissemination of ARGs over multiple time points or used a single 
aquatic catchment to investigate the relative importance of multiple effluent inputs 
contributing to a single aquatic environment. It has been proposed that an environmental 
framework to monitor ARG dissemination is required in order to achieve antimicrobial 
resistance risk assessments (Port et al., 2014). With this in mind, the aim of the current 
  12 
programme of research is to develop an environmental framework that will allow the 
dissemination of ARGs to be monitored in effluents that are entering the aquatic environment. 
 
1.4.1. Monitoring ARG dissemination in the aquatic environment 
In relation to the proposal that an environmental framework to monitor ARG dissemination is 
required, it has been suggested that sequencing technologies such as metagenomics are 
good candidates for facilitating quantitative microbial risk assessment as they can target a 
large array of ARGs both quickly and semi-quantitatively (Brul et al., 2012). The present 
review has highlighted that metagenomics is an ideal tool for the study of bacterial 
communities in environmental samples due to non-reliance on bacterial cell culture and the 
availability of reference databases and analysis tools. Metagenomics therefore offers 
considerable potential for facilitating the monitoring of ARG dissemination and with the 
development of ARG-specific metagenomic methodology and analysis tools it is likely to form 
the basis of the environmental framework that the work described in this dissertation will help 
to develop. 
In addition to developing a metagenomics-based environmental framework to monitor ARG 
dissemination, the work in this dissertation investigates additional approaches that can be 
utilised to facilitate monitoring and will examine factors that may contribute to ARG 
dissemination and the emergence and persistence of resistance. 
 
  13 
Chapter 2. Search Engine for Antimicrobial Resistance 
2.1. Preface 
As outlined in Chapter 1, current laboratory methods for the detection and surveillance of 
antimicrobial resistant bacteria in the environment are limited in their effectiveness and 
scope. As an alternative to culture and polymerase chain reaction approaches, 
metagenomics is able to identify known genes within the genetic content of environmental 
samples. Therefore, metagenomics has been proposed as a viable technology to assess the 
resistome. However, the ability to interrogate sequencing data quickly and easily for the 
presence of antimicrobial resistance genes (ARGs) is crucial in order to utilise metagenomics 
as part of an environmental framework to monitor ARG dissemination. 
This chapter presents the Search Engine for Antimicrobial Resistance (SEAR), a pipeline and 
web interface for detection of horizontally acquired antimicrobial resistance genes in raw 
sequencing data. This study has since been published in PLoS ONE (manuscript and source 
code available in Appendix 1). 
 
2.2. Introduction 
The global threat of antimicrobial resistance is growing at an alarming rate; infections that 
were once easily treatable now constitute public health crises (Sack et al., 2001). This has 
lead to the consensus that more must be done to monitor and combat the occurrence and 
spread of antimicrobial resistance (World Health Organisation, 2012, Laxminarayan et al., 
2013). Current diagnostic laboratory practice for the detection of antimicrobial resistance 
relies on isolate culturing, followed by growth inhibition assays for the identification of 
resistant phenotypes and determination of Minimum Inhibitory Concentrations against a 
range of antimicrobials (MICs) (Public Health England, 2013). Alternatively, ARGs can be 
identified using polymerase chain reaction (PCR) and quantified using real-time PCR, 
requiring specific primers for the amplification of target sequences (Espy et al., 2006). These 
approaches take time, consume resources, and have limitations that may result in clinically 
relevant resistances being undetected e.g. phenotypic testing will miss non-culturable 
bacteria and non-expressed ARGs, whereas limitations of multiplex composition and size in 
  14 
molecular testing complicates the detection of ARGs (Diekema and Pfaller, 2013, Heiman et 
al., 2014).  
Perhaps not surprisingly, the Centers for Disease Control and Prevention (CDC) identified 
one of the current downfalls in the approach to combatting antimicrobial resistance as the 
poor use of advanced molecular detection (AMD) technologies (Centers for Disease Control 
and Prevention, 2013). AMD technologies, such as the whole genome sequencing of 
bacterial isolates as well as uncultured bacteria (metagenomic sequencing), have the 
potential to identify antimicrobial resistance more quickly and effectively than conventional 
laboratory assays (Centers for Disease Control and Prevention, 2013). In addition to these 
well-understood advantages, AMD technologies can also be applied to circumvent the 
requirement of prior knowledge of causative agents and provide clinically relevant information 
for the treatment and surveillance of pathogens as well as antimicrobial resistance (Miller et 
al., 2013). Upon receipt of a metagenomic (e.g. environmental or faecal microbiome) or 
isolate sample, DNA can be extracted, compiled into a library and sequenced within hours 
(Illumina, 2013). Indeed, AMD approaches to pathogen detection are currently being 
developed and seek to identify pathogens directly from metagenomic samples within clinically 
relevant timeframes (Naccache et al., 2014). Recent studies have also shown AMD to be 
effective in the epidemiological tracking of pathogens, as well as the detection of ARGs 
present in their genomes (Koser et al., 2012, Harrison et al., 2013). AMD offers an alternative 
screening tool that may be quicker than traditional culture-based techniques. For example, 
the detection of Legionella species in clinical pneumonia cases requires the inoculated 
isolation media to be incubated for several days in order to diagnose infection and 
characterise antimicrobial resistance, as there is no commercially available molecular assay 
for Legionella species (Reller et al., 2003). This highlights the potential for developing more 
efficient diagnostic tests and the utilisation of AMD technologies to create more rapid 
alternatives for ARG detection.  
In addition to these direct clinical applications, AMD technologies are also beginning to 
become a common tool in the detection of ARGs in the environment, which is vital for 
identifying reservoirs of ARGs (Lewin et al., 2013, Mason et al., 2012, Oh et al., 2013). 
However, there is need to establish a metagenomic framework for use in the monitoring of 
ARGs within the environment in order to influence public health decisions and the growing 
concern over antimicrobial resistance (Port et al., 2014). This must include the development 
of reliable surveillance methods and tools for risk assessment (Berendonk et al., 2015). 
When designing metagenomic tools for the environmental monitoring of ARGs, it is therefore 
  15 
necessary to provide context in terms of the relative abundance of ARGs, so that these can 
be correlated with environmental variables (e.g. such as antimicrobial concentrations, etc.) as 
well as to obtain information on the mobile genetic elements (MGE) and pathogens that they 
are associated with. 
Currently published resources available for ARG detection are online databases that use the 
Basic Local Alignment Search Tool (BLAST) algorithm to find possible matches between the 
database and query sequences (e.g. ARDB, CARD, ResFinder) (Liu and Pop, 2009, 
McArthur et al., 2013, Zankari et al., 2012, Altschul et al., 1990). To the author’s knowledge, 
no existing tools give an ARG abundance measure or simultaneously provide MGE 
information. The targeting of full-length gene matches using BLAST requires a sequence 
assembly step, adding time, infrastructure requirements, and complexity to the analysis. 
Furthermore, full-length gene assembly is often difficult to achieve in metagenomic samples 
where coverage is frequently low and uneven across the sample. Ideally, raw sequencing 
data would be used directly to rapidly identify and quantify ARGs of interest. Although 
mapping-based approaches have been used for individual studies (Hu et al., 2013, Forslund 
et al., 2013) and tools that work directly with reads (though on non-ARG databases) such as 
the SEED subsystems and SRST2 can be applied to work to this aim (Inouye et al., 2014), 
there is as yet no such ARG-detection algorithm. This chapter documents an automated 
pipeline, the Search Engine for Antimicrobial Resistance (SEAR), which quickly and 
accurately identifies antimicrobial resistance information from biological samples. 
Furthermore, it also provides abundance estimates and returns the true sample full-length 
reconstructed gene sequence. To demonstrate efficacy, the pipeline was applied to a range 
of sequencing data types including novel environmental metagenomes, human faecal 
metagenomes and clinical isolates of pathogenic enteric bacteria (Shigella sonnei). 
 
2.3. Materials and methods 
2.3.1. SEAR requirements  
2.3.1.1. Reference databases 
SEAR requires reference databases for read subtraction (optional) and read clustering. The 
user can supply a read subtraction reference database in the form of a BWA index, such as 
the human genome (HG19 build, (Genome Reference Consortium, 2009)) or Escherichia coli 
  16 
K12 genome (these are not supplied in the SEAR package to reduce file size but can be 
readily downloaded (https://genome.ucsc.edu/)). SEAR requires a reference database for the 
read-clustering step, the default supplied is the ARGannot database (Gupta et al., 2014) but 
other options are available (such as CARD (McArthur et al., 2013)) and the user can supply 
any multiFASTA file as a database. The supplied ARGannot database was customized as 
follows: ARGannot ARGs were clustered at 97% identity using USEARCH (Edgar, 2010) and 
the representative sequence for each cluster was added to the pipeline’s ARG database. 
Each cluster and representative sequence is annotated with gene type and the class of 
antimicrobial to which the gene confers resistance. The Shigella reference database for the 
benchmarking test was made by downloading the FASTA files for each ARG tested (n=19) in 
the Holt et al. study, removing duplicate entries and creating a multiFASTA file (n=16). 
 
2.3.1.2. Hardware 
Minimum hardware requirements for SEAR comprise a Unix server (tested using Ubuntu 
10.04) with ~2 GB of disk space for reference data and software dependencies. Whilst 
running, SEAR requires up to 2X the input FASTQ file size (bytes) in both RAM and disk 
space for temporary file storage. 
 
2.3.2. SEAR 
2.3.2.1. The pipeline 
SEAR is a pipeline consisting of Perl, Shell and R scripts that call on several pieces of open 
source software and utilise a customisable reference database to annotate ARGs direct from 
short-read sequencing data. SEAR is downloadable from 
http://computing.bio.cam.ac.uk/sear/SEAR_WEB_PAGE/SEAR.html, in stand-alone 
command-line and web-based versions (Figure 2.1).  
  17 
 
Figure 2.1 Screen shot of SEAR web interface including homepage (A) and quick start settings (B). 
  18 
The pipeline follows five main steps in the annotation of ARGs: (1) processing of input files, 
(2) clustering of sequence reads to known ARGs in user-defined (or pre-loaded) database, 
(3) mapping of reads to reference sequences, (4) ARG annotation and calculation of relative 
abundance and (5) local alignment of annotated ARGs to online databases. 
(1) Processing of input files 
The pipeline accepts raw or compressed (.gz) FASTQ files (either 33 or 64 ASCII encoding) 
from metagenomic, metatranscriptomic or isolate sequencing. Where more than one input file 
(e.g. paired-end data) is provided, these files are merged to give a single input file (pair-end 
information is not currently utilised in the pipeline). The pipeline has the optional step of pre-
filtering reads, by removing those that map against a user-defined reference, such as the 
human genome or a bacterial strain. FASTQ files are quality checked using user-defined cut 
offs and converted to FASTA formatted reads. 
(2) Clustering of sequence reads to ARG database 
The pipeline is supplied with a custom ARG database that has been built by clustering and 
annotating the ARGs held in the ARGannot-database (Gupta et al., 2014). Notably however, 
other ARG databases can be used or the user can use a custom FASTA file. Reads are 
clustered to the ARG database by global alignment with USEARCH (version 7.0.959) using a 
default identity cut-off of 99% (Edgar, 2010). Where multiple matches occur, the read is 
clustered with the highest identity match. SEAR parses the clusters by grouping reads to 
each matched reference gene and retrieving corresponding FASTQ information for each 
matched read. 
(3) Mapping of clustered sequence reads to ARG references  
The Burrows-Wheeler Aligner (BWA-mem version 0.7.8) (Li and Durbin, 2009) is used for 
read mapping each cluster of FASTQ reads to the corresponding reference gene. Samtools 
is then used to analyse the BWA alignment and generate a consensus sequence using 
mpileup (Li, 2011). 
(4) ARG annotation and relative abundance 
The consensus sequences are used to annotate ARGs and calculate relative abundance 
values; an ARG is present in the sample if sequence reads can be mapped to the ARG 
reference sequence above the defined coverage cut-off (coverage is the percentage length of 
reference ARG with mapped reads). For relative abundance calculation, SEAR uses a similar 
method to the reads per kilobase/million reads (RPKM) method that is commonly used in 
transcriptome studies (Mortazavi et al., 2008). When calculating relative abundance, the total 
number (n) of ARGs that have been annotated are used to calculate a relative abundance 
  19 
(RA) percentage for each ARG. Firstly, an abundance value (A) is calculated for each gene 
according to: , where X denotes the number of reads that successfully mapped, Y 
denotes the total number of reads in the input file/s and L denotes the length of the reference 
gene (in bases). Relative abundance is then calculated using:  
In this way, the relative abundance measure describes the proportion of sequence reads that 
have built the consensus sequence of each annotated ARG from a single pipeline run. 
(5) Local alignment 
The consensus sequences for annotated ARGs are aligned to the NCBI nucleotide and 
protein databases using command line BLAST (Altschul et al., 1990) (using the –remote 
BLAST service by default, see documentation to utilise local database versions). In addition, 
sequences are also aligned to the current Repository of Antimicrobial Resistance Cassettes 
(RAC) (Tsafnat et al., 2011) and Antibiotic Resistance Database (ARDB) (Liu and Pop, 2009) 
databases using BLAST (though ARDB has not recently been curated). 
 
2.3.2.2. Pipeline outputs 
In both command-line and web versions of SEAR, output includes: graphical overview, ARG 
annotations, relative abundance scores, consensus sequences, flat files (html, csv, blast 
files) and links to further gene information and homologues found in online databases (such 
as the repository of antimicrobial resistance cassettes, NCBI non-redundant nucleotide and 
protein databases). 
 
2.3.3. Demonstrating SEAR utility 
2.3.3.1. Data sets and parameters used in this study  
Several datasets were used to demonstrate the utility of this pipeline across broad data 
categories. All datasets were analysed using a UNIX server (Ubuntu 10.04) running SEAR 
with default parameters (99% clustering identity and 90% coverage cut-off for ARG 
annotation, full default parameter list found in Table 2.1).  
  
A =
(X
Y
)
L
RA = ( A
i=1
n
 
)*100
  20 
Table 2.1 List of parameters and default settings for SEAR. 
Parameter Default Explanation 
--fqformat 
(-ff) 33 
ASCII offset for the input fastq files. 
Accepts either 33 or 64. 
--lengthcutoff 
(-lc) 70 Discard sequences with length < lc. 
--qualitycutoff 
(-qc) 20 Quality score cutoff for input fastq files. 
--filter 
(-f) N 
Filter reads by mapping to a reference 
database (default db = Human Genome) 
and discarding mapped reads. Accepts 
either Y or N. 
--coveragecutoff 
(-cc) 90% 
The coverage cut-off parameter dictates 
what proportion of the reference 
sequence must be covered by reads for a 
successful annotation. In this way, the 
annotation stringency is controlled and 
customisable. 
--clusteringident 
(-ci) 0.99 Identity value for usearch clustering. 
--references 
(-r) arg_annot_database.fa The reference gene dataset to use. 
--threads 
(-t) 1 
The number of threads to use in steps 
that allow multi-threading.  
 
2.3.3.2. Novel environmental metagenomes 
Samples were collected from two effluent sources within the River Cam Catchment, 
Cambridge, UK on the 21st June 2012. A map of the river catchment is included in Appendix 
2. The waste effluent of the University of Cambridge dairy farm (latitude: 52.22259, longitude: 
0.02603) was sampled prior to it being applied to the surrounding fields as fertiliser, where it 
subsequently enters drainage ditches that drain into the River Cam. The effluent of the 
municipal wastewater treatment plant (WWTP) (latitude: 52.234469, longitude: 0.154614) 
was collected from the effluent discharge pipe that enters the River Cam. Samples were 
collected in 10 L sterile polypropylene containers. Sample volumes were based on the 
microbial abundances, as previously determined for these sites using a DNA extraction 
series (data not shown). Samples were transported at 4°C to the laboratory and processed 
within 2 hours. 
Similarly as in Dancer et al. (Dancer et al., 2014), samples were filtered under pressure at 
approximately 2 bar using a pressure vessel system (10 L SM 1753, Sartorius). Samples 
were first pre-filtered through 3.0 µm membranes (Millipore) at 2 Bar to remove eukaryotic 
  21 
cells and debris. The filtrate was subsequently filtered through 0.22 µm membranes 
(Millipore) to capture the prokaryotic cells, metagenomic DNA was then extracted by washing 
and vortexing the membranes in phosphate buffered saline with Tween20 (2%) before 
enzymatic lysis (Meta-G-Nome DNA isolation kit; Epicentre). Assessment of DNA quality and 
concentration was made by TBE agarose (2%) gel electrophoresis and spectrophotometry 
(Nanodrop ND-1000; ThermoScientific). 
For each sample, 2 µg of DNA was sequenced by the Eastern Sequence and Informatics 
Hub, Cambridge, UK. Two libraries (seventy-five base pair, paired-end reads) were prepared 
from the samples and were sequenced using an Illumina HiSeq2000. The FASTQ files for the 
WWTP and farm effluent metagenomes are available via the European Nucleotide Archive 
(ENA) (study: ERP003955). Sample accession numbers are as follows: farm effluent 
(ERS786322), WWTP effluent (ERS781558) (Appendix 3). 
 
2.3.3.3. Pre-existing metagenomic and clinical isolate data  
Human Microbiome Project (HMP) data for 32 Spanish human faecal microbiomes (for which 
the ARGs have previously been characterised in an in silico study by Forslund et al. 
(Forslund et al., 2013)) was downloaded from the ENA website (study: PRJEB1220) 
(accessed: 02.03.2015) (Human Microbiome Project, 2012). Additionally, SEAR was used to 
detect ARGs in a global dataset of 126 clinical isolates of the pathogenic bacteria Shigella 
sonnei (SRA Study ERP000182) (Holt et al., 2012). The FASTQ files for the 126 isolates 
were downloaded from the Sanger FTP site (study: PRJEB2128) (accessed: 02.03.2015). In 
the case of the clinical isolates, SEAR ARG detection was compared with the published ARG 
content of the isolates, with SEAR being run with default parameters on a custom reference 
database of ARGs originally detected by 100% mapping (Holt et al., 2012). Further details on 
sequencing datasets are provided in Appendix 4. 
 
2.4. Results 
To test the utility of SEAR the pipeline was run using a variety of sample types 
(environmental metagenomes, human faecal microbiome and bacterial clinical isolate), 
recorded pipeline run times (Table 2.2) and then investigated the presence and abundance of 
ARGs in all samples. 
  
  22 
Table 2.2 Example runtimes for SEAR using default parameters and server settings. 
Name Sample ID Type File size (MB) 
Time 
(mins) 
ShIB1976 ERR025684 Clinical isolate 434 6.1 
O2.UC1-0 ERR209529 HMP metagenome 3200 36 
WWTP effluent ERS781558 Environmental metagenome 14000 194 
 
2.4.1. Discrimination of ARG presence and abundance between 
environmental metagenomes 
A total of 28 (15 in each) ARGs were identified among the environmental metagenomes from 
WWTP effluent and farm waste effluent (Figure 2.2). Only two genes, strA and strB (both 
conferring aminoglycoside resistance), were common between the metagenomes and each 
gene found in both sets was five times more abundant in the WWTP effluent compared to the 
farm effluent when using the normalised abundance values for the combined datasets. The 
WWTP effluent had ARGs conferring resistance to a total of four antimicrobial classes with 
the most diverse (i.e. greatest number of ARGs) being the aminoglycoside class and the 
greatest abundance being ARGs conferring tetracycline resistance. In contrast, the farm 
effluent had ARGs conferring resistance to five antimicrobial classes with the most diverse 
being the beta lactam class and the most abundant also being tetracycline resistance (Figure 
2.2). The most abundant ARGs in the metagenome datasets were tetracycline resistance 
genes; tetC (41.6%) in the farm effluent and tet39 (15.3%) in the WWTP effluent. A subset of 
ARGs identified by SEAR (tetA, qnrB and bla-ACT; chosen to encompass clinically relevant 
resistances, drugs with both a long and short history of resistance and chemically diverse 
antimicrobials) was confirmed in the original farm effluent DNA sample using PCR. Briefly, 
primers were designed using Primer3 (Rozen and Skaletsky, 1988) and were amplified using 
GoTaq DNA polymerase (Promgega) (not shown). 
 
  23 
 
Figure 2.2 SEAR results for environmental metagenomes. The column chart in A shows the breakdown of 
the number of ARGs in each effluent, grouped by antimicrobial class. The column chart in B shows the relative 
abundance of ARGs found in each metagenome (coloured according to the key). The MLS class of 
antimicrobial represents macrolides, lincosamides and streptogramins. 
  24 
2.4.2. Efficacy of SEAR for detecting ARGs in human faecal 
microbiomes 
To assess the efficacy of SEAR for detecting ARGs in microbiome data, SEAR was tested on 
32 faecal microbiome samples (Appendix 4). ARGs were detected in 31 of the samples and a 
total of 295 genes conferring resistance to 6 classes of antimicrobials were identified across 
the samples (Table 2.3). Genes conferring resistance to tetracyclines were again the most 
common ARGs identified (39% of total ARGs detected). 
Table 2.3 SEAR detection of ARGs across antimicrobial classes in human faecal microbiomes. The table 
shows the number of genes identified in each antimicrobial class for the combined dataset of HMP samples. 
Antimicrobial class Number of ARGs 
Aminoglycosides 54 
Beta lactams 38 
Quinolones 0 
Glycopeptides 0 
MLS 82 
Phenicols 1 
Rifampicin 0 
Sulfonamides 5 
Tetracyclines 115 
Trimethoprims 0 
 
2.4.3. Accuracy of SEAR ARG detection using clinical isolate 
sequencing data 
To evaluate SEAR’s efficacy in detecting ARGs in clinical isolate sequencing data, SEAR 
was run on sequencing data from 126 isolates of the enteric pathogen Shigella sonnei. To 
evaluate SEAR’s performance, the results were compared to the ARG detection data 
presented in the original publication (Holt et al., 2012). Of the 231 detection events (see 
methods for criteria) originally presented in the publication, SEAR identified 221 of these, and 
a further 20 ARGs (Table 2.4, full results shown in Appendix 5). 
  
  25 
Table 2.4 Accuracy of SEAR ARG detection using clinical isolate sequencing data. The contingency table 
compares the detection and non-detection of ARGs by SEAR relative to the published ARG detection data for 
126 S. sonnei isolates. 
 
Reported in Holt et al. (Holt et 
al., 2012)  
detected not-detected Total 
SEAR results detected 221 20 241 not-detected 10 0 10 
 Total 231 20  
 
2.5. Discussion 
SEAR is an ARG annotation tool that is freely available and may be downloaded as a cloud 
compatible web interface or a stand-alone command line program. It offers advantages over 
currently available ARG annotation tools as it provides ARG annotations, relative abundance 
values, gene sequence and gene information from raw sequencing data without requiring any 
sequence assembly. In contrast to tools based on BLAST comparison of de novo 
assemblies, the clustering and mapping approach used by SEAR, combined with the 
customisable database and annotation parameters, allows the user to detect putative ARGs 
in incomplete or low coverage sequencing data that is common in metagenomic analyses. 
SEAR successfully identified ARGs in sequencing datasets that were generated from novel 
environmental metagenomic samples, human microbiomes and clinical isolates of Shigella 
sonnei. 
SEAR was able to detect the ARGs present in two novel environmental metagenomes 
allowing direct comparison between two different wastewater effluent samples. SEAR 
identified meaningful differences among ARGs of clinical interest, for example the presence 
of quinolone resistance genes (qnrB and qnrS) exclusively in the wastewater effluent from 
the farm source. It also showed that the two sources had different qualitative ARG 
characteristics (with either aminoglycosides or beta lactams being the most diverse 
antimicrobial resistance class) and in both sources tetracycline resistance genes were 
present in the greatest abundance. In addition to detecting important differences among 
these sample types, the confirmation of a subset of identified ARGs by PCR demonstrated 
the robustness of the pipeline.  
Similarly, SEAR was effective for identifying ARGs from clinical samples. ARGs were 
detected in human microbiomes demonstrating the potential of using metagenomic analyses 
for the surveillance and management antimicrobial resistance. Additionally, SEAR 
  26 
successfully identified ARGs in a global dataset of 126 clinical isolates of an important enteric 
pathogen. There were a few discrepancies, which were consistent with a given isolate or 
gene family, however the results were overwhelmingly consistent. Furthermore, the 
congruence of ARG detection results from SEAR with the published ARG content of the 
isolates further highlighted the effectiveness of the pipeline, providing further compelling 
argument for the application of high-throughput AMD into clinical microbiology. 
SEAR offers increased functionality over existing bioinformatic tools by providing a 
consensus sequence of annotated ARGs, links to online resources containing information on 
the ARGs (and gene homologs) and a relative abundance estimate for each ARG detected. 
Each ARG consensus sequence is generated using reads that clustered to a reference 
sequence and consequently any variability in the consensus sequence in a metagenomic 
sample may be due to either sequencing noise or the presence of multiple bona fide 
sequence variants. The relative abundance estimate is relative within an individual sample, 
however the SEAR output features the information required to calculate relative abundance 
across multiple samples. Due to possible large variations in user file size and upload speed, 
the SEAR interface and command line tool are available for use as downloadable packages.  
SEAR is designed for detecting ARGs that are horizontally acquired, not antimicrobial 
resistance that is caused (or inactivated) by single nucleotide polymorphisms (SNPs) e.g. 
SNPs in the gyrA gyrase gene that result in quinolone resistance. SNPs are not currently 
tested for due to the annotation parameters being calibrated for detecting partial ARG 
matches to compensate for low sequencing coverage. Hence, such SNPs may be missed by 
SEAR due to the number of mismatches permitted or by a low coverage cut-off (though these 
are both customisable settings). For these reasons, it is not recommended to include SNP-
based resistances in reference databases used with SEAR as they may lead to false 
positives. The detection of SNP-based resistances in metagenomic samples represents a 
significant future challenge that needs to be addressed. It should also be stressed that the 
default SEAR parameters, which are based on high-stringency read clustering and mapping, 
result in an analysis that finds ARGs that are known in the reference data and it is not suited 
for discovery of emergent ARGs. The high-stringency settings are designed to exclude the 
possibility of non-competitive read mapping causing false positive results by ensuring that 
annotated ARGs have a high sequence identity compared to the reference database. 
 
  27 
2.6. Conclusion 
This chapter has presented a bioinformatic pipeline that is highly effective at detecting ARGs 
directly from raw sequencing reads, in addition it also provides relative abundance estimation 
and sequences of identified genes. The work in this chapter has illustrated the application of 
SEAR on sequence data from metagenomic datasets and bacterial isolates. The work in this 
chapter has also demonstrated the application of SEAR in potential clinical and 
environmental monitoring applications, highlighting the advantages of automated 
interpretation of sequencing data for generating timely and informative reports for informing 
public health and potentially clinical decision-making. With the increasing drive to integrate 
AMD technology and existing laboratory assays in order to combat antimicrobial resistance, 
this pipeline has been presented as a valuable step towards this important goal.  
 
  28 
Chapter 3. Comparative metagenomics reveals a diverse range 
of antimicrobial resistance genes in effluents entering a river 
catchment 
3.1. Preface 
The work in this chapter has been designed as a pilot study to test the suitability of 
metagenomics for monitoring antimicrobial resistance genes (ARGs) and to address the 
underlying assumption of this dissertation; effluents entering the aquatic environment are 
disseminating ARGs. 
Chapter 2 described the design and implementation of SEAR; a bioinformatic pipeline to 
detect horizontally acquired ARGs in sequencing data. This study uses SEAR to interrogate 
novel metagenomes generated from effluents entering a single river catchment, determining 
the relative abundance of ARGs in relation to the environment they are entering. This study 
has been accepted for publication in Water Science and Technology. 
 
3.2. Introduction 
Antimicrobial resistance remains a significant and growing concern for both human and 
veterinary clinical practice (Levy and Marshall, 2004, Davies and Davies, 2010a), with 
infections that were once readily treated now being resilient to antimicrobial therapy (World 
Health Organisation, 2012). The use of antimicrobial compounds exerts selection pressures 
on bacteria, leading to the fixation of gene mutations, selection of resistant precursors and 
the up-regulation and lateral transfer of antimicrobial resistance genes (ARGs) within 
prokaryotic communities (Gillings, 2013). The maintenance and transfer of ARGs is 
responsible in part for the rising threat of antimicrobial resistance (Laxminarayan et al., 
2013). 
The collective pool of ARGs in a given environment is termed the resistome (D'Costa et al., 
2006, Wright, 2007). Although a proportion of these ARGs are genes that have evolved to 
utilise antimicrobial compounds for functions other than defence, such as signalling 
molecules or constituents of metabolic pathways (Linares et al., 2006, Dantas et al., 2008), 
the resistome may also serve as a reservoir for ARGs that can be transferred to clinically 
  29 
significant pathogens (Forsberg et al., 2012, Wellington et al., 2013). Indeed, ARGs are 
commonly associated with mobile genetic elements (MGEs) that facilitate the transfer of 
ARGs between bacteria and enable their entry into the accessory genome of pathogenic 
bacteria (Gaze et al., 2013). 
There is growing evidence showing that aquatic environments harbour ARGs, MGEs and 
pathogenic bacteria (Chen et al., 2013, Lu et al., 2015, Devarajan et al., 2015). It is also likely 
that these environments may host many uncharacterised and novel ARGs that may be 
selected for under sufficient selection pressures (Bengtsson-Palme et al., 2014). Effluents 
that feed into the aquatic environment have also been shown to contain ARGs, such as the 
effluents of urban residential areas and hospitals (Li et al., 2015a), as well as other 
wastewater and faecal sources (Li et al., 2012, Pruden et al., 2006, Zhang et al., 2009) but 
the abundance and diversity of these genes relative to background samples needs to be 
clarified. It is therefore crucial to establish whether effluents entering the aquatic environment 
are carrying ARGs, along with MGEs and pathogenic bacteria, thus contributing to the 
reservoirs of resistance genes that may be utilised by pathogenic bacteria and subsequently 
re-enter human and animal populations (Berendonk et al., 2015). 
Previous studies into the presence of ARGs within the aquatic environment have utilised 
techniques such as bacterial culture and polymerase chain reaction (PCR) (Tao et al., 2010, 
Zhang and Zhang, 2011, Lu et al., 2015). These techniques offer the ability to detect 
phenotypic resistance (culture), or a panel of ARGs (PCR), but they are limited by culturing 
bias or inadequate detection panels. Next generation sequencing techniques, such as 
metagenomics, offer the ability to circumvent these limitations and identify all known ARGs 
within a sample (if suitable reference sequences are available), providing a new approach for 
the environmental monitoring of antibiotic resistance (Port et al., 2014). 
In this chapter, two distinct effluents that enter a single river catchment were identified. Both 
effluents originated from faecal sources and were sampled several times, immediately prior 
to them entering the environment. Using a comparative metagenomic approach, the work in 
this chapter described the ARG content of these effluents, characterised the MGEs and 
pathogenic bacteria present, and related the abundance of these features to a background 
sample of the river source water, taken from upstream of the effluent entry points. 
 
  30 
3.3. Methods 
3.3.1. Sample collection and DNA sequencing 
Water samples were collected from three sources within the River Cam Catchment, 
Cambridge, UK by grab sampling. A pilot collection was made on 21st June 2012 and is 
described in 2.3.3.2. Further collections from the municipal wastewater treatment plant 
(WWTP) and the University of Cambridge dairy farm (both detailed in 2.3.3.2) were made on 
the 2nd May 2013 and 4th August 2014. The river source water of the River Cam was 
collected at Ashwell Spring (latitude: 52.0421, longitude: 0.1497) once on the 2nd May 2013. 
Samples were collected in 10 L sterile polypropylene containers, transported at 4°C to the 
laboratory and processed using the same methodology described in 2.3.3.2. For each 
sample, 2 µg of DNA was used to generate Illumina paired-end libraries that were sequenced 
using an Illumina HiSeq2500 (Exeter Sequencing Service, UK). A map of the River Cam 
catchment is included as Appendix 2. A full description of the metagenomic samples used in 
this chapter is available in Appendix 3. 
 
3.3.2. Bioinformatic analyses 
3.3.2.1. Identification of ARGs 
ARGs were identified using the Search Engine for Antimicrobial Resistance (SEAR) with 
default parameters (Rowe et al., 2015). A full description of SEAR is available in Chapter 2. 
In brief, SEAR quality checked and filtered metagenomic reads, clustered the filtered reads to 
the ARG-annot (Gupta et al., 2014) database of horizontally acquired ARGs and used the 
resulting clusters to map the reads and generated a consensus sequence for each ARG in 
the query metagenome. Consensus sequences were then aligned to online databases (NCBI 
genbank, RAC, ARDB), annotated and given an abundance value based on the Reads Per 
Kilobase per Million (RPKM) value from the read-mapping stage. 
 
3.3.2.2. Identification of mobile genetic elements 
Identification of MGEs was performed by mapping metagenomic reads to a custom MGE 
database using BWA-mem (default options) (Li and Durbin, 2009). The MGE database was 
  31 
built from the NCBI Refseq plasmid genomes dataset, combined with the representative 
sequences generated from clustering the Integrall dataset (Moura et al., 2009) at 97% 
identity using USEARCH (Edgar, 2010). MGE mapping results with less than 90% coverage 
of the reference sequence were discarded from the analysis. Successfully mapped 
sequences where then binned into class I and class II integrons, transposons and mobilisable 
plasmids. 
 
3.3.2.3. Abundance analysis 
The ARG and MGE abundance data was normalised to the number of 16S rRNA sequences 
as in Bengtsson-Palme et al. (Bengtsson-Palme et al., 2014). In brief, bacterial 16S rRNA 
sequences were extracted from each metagenome using Metaxa 2.0 (Bengtsson-Palme et 
al., 2015) using default settings and then grafted to sequences from the SILVA RNA 
database using Megraft (Bengtsson et al., 2012) and subsequently clustered using 
USEARCH (Edgar, 2010). ARG abundance values were normalised to 16S sequences by 
dividing the number of extracted 16S sequences by the length of the 16S gene (Bengtsson-
Palme et al., 2014). 
 
3.3.2.4. Taxonomic profiling and pathogen detection 
Taxonomic profiling of metagenomes was carried out by mapping sequencing reads to clade-
specific marker genes using the Metaphlan package (Segata et al., 2012) (default 
parameters). Metaphlan output was then cross-referenced to the PATRIC database of 
pathogenic bacteria (Gillespie et al., 2011) to annotate potential human-specific bacterial 
pathogens. Biomarker discovery and identification of differentially abundant features between 
metagenomes from 2012, 2013 and 2014 was performed using LEfSe (Segata et al., 2011). 
Taxonomic profiling and pathogen data was then combined and presented using the 
Graphlan package (Segata, 2014). 
 
  32 
3.4. Results 
3.4.1. Metagenome analysis 
In this work, 29.52 Giga base-pairs (Gbp) of data was generated across all samples, with the 
number of reads produced from the total farm effluent samples being approximately double 
that produced from the total WWTP effluent samples (Table 3.1). 
Table 3.1 Summary of the metagenomes used in the work of this chapter. 
Sample Total reads Gbp Total ARG reads % ARGs 
Farm effluent 2012 88674294 4.4337 7715 0.0087 
Farm effluent 2013 66120642 3.3060 2317 0.0035 
Farm effluent 2014 184149408 9.2075 13094 0.0071 
WWTP effluent 2012 57392478 2.8696 4205 0.0073 
WWTP effluent 2013 65960602 3.2980 250 0.0004 
WWTP effluent 2014 73273516 3.6637 3767 0.0051 
River source 2013 54799282 2.7400 181 0.0003 
 
3.4.2. Identification of antimicrobial resistance genes 
In the effluent from the dairy farm an average of 7709 reads (0.007%) matching ARGs were 
found across the three samples. An average of 2740 reads (0.004%) matching ARGs were 
found across the three WWTP effluent samples. Only 181 reads (0.0003%) were found to 
match ARGs from the river source water. A significant diversity of ARGs was observed 
across the samples, with 53 different ARGs found in total, conferring resistance to seven 
antimicrobial classes (Figure 3.1, Table 3.2). There were 18 ARGs common between the 
farm and the WWTP effluent samples. The river source water contained the lowest diversity 
of ARGs (five ARGs, conferring resistance to two antimicrobial classes). When normalised to 
the number of 16S sequences in each sample, the most abundant ARG across all the 
samples was found to be sul2 (sulfonamide resistance) in the WWTP effluent 2014 (0.097 
copies per 16S sequence) and the least abundant ARG was catB4 (phenicol resistance), 
found in the farm effluent 2014 (0.0001 copies per 16S sequence). When looking at the 
effluents individually, tetracycline resistance genes tetC (farm effluent 2012) and tetW (farm 
effluent 2013 and 2014) were the most abundant genes within the farm effluent samples. In 
comparison, the aminoglycoside resistance genes strA/strB (WWTP effluent 2012) and the 
sulfonamide resistance genes sul1/sul2 (WWTP effluent 2013 and 2014) were the most 
  33 
abundant ARGs within the WWTP effluent samples. On average, the abundance of ARGs in 
the farm effluents was three times that of the river source water. Similarly, the average 
abundance of ARGs in the WWTP effluents was double that found in the river source water. 
In terms of the diversity of ARGs relative to the river source water, the farm effluent had an 
average of five different ARGs for each ARG found in the river source water, whereas the 
WWTP effluent had two different ARGs for each ARG present in the source water. 
When comparing samples over the three years, the abundance of ARGs was found to 
decrease year on year in the WWTP effluent for all but sulfonamide resistance genes, which 
were found to increase over time (11% average change in abundance of sulfonamide 
resistance genes over three years). The largest change over time for the farm effluent was 
the 10% increase in the abundance of aminoglycoside resistance genes observed between 
2012-2013. 
 
 
Figure 3.1 Abundance of ARGs found in each effluent sample, binned by antimicrobial class. Abundance 
of antimicrobial resistance genes is normalised to the number of 16S sequences per sample. The MLS class of 
antimicrobial represents marcolides, lincosamides and streptogramins. 
  
  34 
Table 3.2 Antimicrobial resistance gene analysis. The table lists the antimicrobial resistance genes found in 
the effluents entering the river catchments. 
Sample Gene name Antimicrobial class Relative ARG abundance for sample 
Farm effluent 2012 tetC tetracycline 73.4753 
Farm effluent 2012 sulI sulfonamide 6.0742 
Farm effluent 2012 tetA tetracycline 3.7590 
Farm effluent 2012 tetR tetracycline 3.1282 
Farm effluent 2012 qnrS2 quinolone 2.4774 
Farm effluent 2012 blaOXA-320 beta lactam 1.9940 
Farm effluent 2012 aph3-Ia aminoglycoside 1.6270 
Farm effluent 2012 ACT-12 beta lactam 1.4771 
Farm effluent 2012 blaOXY6-4 beta lactam 1.3378 
Farm effluent 2012 strA aminoglycoside 1.0920 
Farm effluent 2012 strB aminoglycoside 1.0209 
Farm effluent 2012 qnrB7 quinolone 0.8548 
Farm effluent 2012 blaOXY5-2 beta lactam 0.6711 
Farm effluent 2012 blaCMY-95 beta lactam 0.5512 
Farm effluent 2012 ampC1 beta lactam 0.4600 
Farm effluent 2013 tetW tetracycline 17.1873 
Farm effluent 2013 blaCFX-A5 beta lactam 12.6614 
Farm effluent 2013 Str aminoglycoside 9.5407 
Farm effluent 2013 lnuB MLS 7.2596 
Farm effluent 2013 ermB MLS 6.8475 
Farm effluent 2013 ant6-Ia aminoglycoside 6.4809 
Farm effluent 2013 tetM tetracycline 5.8033 
Farm effluent 2013 blaGES-22 beta lactam 3.9702 
Farm effluent 2013 strB aminoglycoside 3.4575 
Farm effluent 2013 Spc aminoglycoside 3.3259 
Farm effluent 2013 sulI sulfonamide 3.0942 
Farm effluent 2013 ermG MLS 2.8150 
Farm effluent 2013 strA aminoglycoside 2.6841 
Farm effluent 2013 tetO tetracycline 2.6719 
Farm effluent 2013 sat4A aminoglycoside 2.5109 
Farm effluent 2013 aph3-III aminoglycoside 2.1123 
Farm effluent 2013 ermF MLS 2.1098 
Farm effluent 2013 lnuA MLS 2.0399 
Farm effluent 2013 tet-36 tetracycline 1.9703 
Farm effluent 2013 blaOXA-210 beta lactam 1.4571 
Farm effluent 2014 tetW tetracycline 24.3143 
Farm effluent 2014 lnuB MLS 6.8709 
Farm effluent 2014 tetM tetracycline 4.8201 
Farm effluent 2014 ermF MLS 4.5820 
Farm effluent 2014 tet-44 tetracycline 4.2108 
Farm effluent 2014 ant6-Ia aminoglycoside 4.0593 
Farm effluent 2014 ant6-Ib aminoglycoside 3.8291 
  35 
Farm effluent 2014 tet-39 tetracycline 3.5128 
Farm effluent 2014 aph3-III aminoglycoside 3.3706 
Farm effluent 2014 strB aminoglycoside 3.3693 
Farm effluent 2014 tetO tetracycline 3.2634 
Farm effluent 2014 ermB MLS 3.1951 
Farm effluent 2014 strA aminoglycoside 2.8502 
Farm effluent 2014 Str aminoglycoside 2.5190 
Farm effluent 2014 ermG MLS 2.4419 
Farm effluent 2014 tet-36 tetracycline 2.4248 
Farm effluent 2014 blaGES-22 beta lactam 2.1196 
Farm effluent 2014 sulI sulfonamide 2.0911 
Farm effluent 2014 sat4A aminoglycoside 2.0250 
Farm effluent 2014 erm35 MLS 1.5781 
Farm effluent 2014 tet-32 tetracycline 1.4578 
Farm effluent 2014 sulII sulfonamide 1.4248 
Farm effluent 2014 blaCFX-A5 beta lactam 1.3575 
Farm effluent 2014 Spc aminoglycoside 1.3163 
Farm effluent 2014 blaOXA-320 beta lactam 1.2794 
Farm effluent 2014 tetT tetracycline 0.9209 
Farm effluent 2014 tet-40 tetracycline 0.6100 
Farm effluent 2014 aadA8b aminoglycoside 0.5367 
Farm effluent 2014 blaTEM-201 beta lactam 0.4977 
Farm effluent 2014 lnuF MLS 0.4964 
Farm effluent 2014 ermA MLS 0.4632 
Farm effluent 2014 blaOXA-210 beta lactam 0.4465 
Farm effluent 2014 aph3-Ia aminoglycoside 0.3546 
Farm effluent 2014 catA1 phenicol 0.3505 
Farm effluent 2014 Cmr phenicol 0.3430 
Farm effluent 2014 msrE MLS 0.3213 
Farm effluent 2014 tetZ tetracycline 0.2206 
Farm effluent 2014 catB4 phenicol 0.1553 
WWTP effluent 2012 strB aminoglycoside 18.3772 
WWTP effluent 2012 strA aminoglycoside 16.2801 
WWTP effluent 2012 tetW tetracycline 14.9175 
WWTP effluent 2012 tet-44 tetracycline 8.9014 
WWTP effluent 2012 ant6-Ib aminoglycoside 7.2859 
WWTP effluent 2012 ermB MLS 7.0075 
WWTP effluent 2012 tetM tetracycline 5.6956 
WWTP effluent 2012 lnuB MLS 5.6885 
WWTP effluent 2012 tet-39 tetracycline 5.2060 
WWTP effluent 2012 Str aminoglycoside 4.0791 
WWTP effluent 2012 tetO tetracycline 1.8454 
WWTP effluent 2012 sulII sulfonamide 1.4551 
WWTP effluent 2012 lnuC MLS 1.3073 
WWTP effluent 2012 Spc aminoglycoside 1.2387 
WWTP effluent 2012 ermG MLS 0.7147 
WWTP effluent 2013 sulI sulfonamide 29.0255 
  36 
WWTP effluent 2013 ereA MLS 24.7971 
WWTP effluent 2013 strB aminoglycoside 18.3611 
WWTP effluent 2013 tetC tetracycline 17.0953 
WWTP effluent 2013 strA aminoglycoside 10.7210 
WWTP effluent 2014 sulII sulfonamide 86.4888 
WWTP effluent 2014 blaCARB-10 beta lactam 5.0967 
WWTP effluent 2014 blaGES-22 beta lactam 3.1151 
WWTP effluent 2014 sulI sulfonamide 1.4774 
WWTP effluent 2014 strB aminoglycoside 0.9286 
WWTP effluent 2014 strA aminoglycoside 0.9062 
WWTP effluent 2014 msrE MLS 0.8102 
WWTP effluent 2014 tetW tetracycline 0.6260 
WWTP effluent 2014 blaOXA-183 beta lactam 0.5512 
River source 2013 aph3-III aminoglycoside 29.7367 
River source 2013 strB aminoglycoside 25.7637 
River source 2013 strA aminoglycoside 22.3138 
River source 2013 catA1 phenicol 12.5585 
River source 2013 aph3-Ia aminoglycoside 9.6274 
 
3.4.3. Identification of mobile genetic elements 
In conjunction with determining the abundance and diversity of ARGs, the effluents were also 
interrogated for MGEs (Figure 3.2, Table 3.3).  No MGEs were found to be present in the 
river source water. Mobilisable plasmids were the most abundant class of MGE found out of 
the combined metagenomic datasets, although no mobilisable plasmids were identified in the 
WWTP effluent 2012 or farm effluent 2014 samples. Class I and class II integrons, as well as 
transposon sequences, were found in all effluent samples. Class I integrons were more 
abundant in the collective farm effluent samples, compared to class II integrons that were 
more abundant in the collective WWTP effluent samples. 
 
  37 
 
Figure 3.2 Abundance of MGEs found in each effluent sample, binned by MGE type. Plasmids were 
binned as mobilisation plasmids if they contained conjugation genes (tra, mob etc.) and integrons were binned 
as class I or II depending on the Integrall annotation. Relative abundance of MGEs is normalised to the number 
of 16S sequences per sample. 
 
  
38 
Table 3.3 Mobile genetic element analysis. The table lists the mobile genetic elements found in the effluents entering the river catchments. 
Sample MGE reference ID MGE description Database 
Farm effluent 2012 gi|15983520 Aeromonas salmonicida plasmid pRAS3.2 REFSEQ 
Farm effluent 2012 gi|15983531 salmonicida plasmid pRAS3.1 REFSEQ 
Farm effluent 2012 gi|198286625 Aeromonas hydrophila plasmid pBRST7.6 REFSEQ 
Farm effluent 2012 gi|20514397 KCL-2 plasmid pMGD2 REFSEQ 
Farm effluent 2012 gi|209947514 Klebsiella pneumoniae plasmid pIGMS31 REFSEQ 
Farm effluent 2012 gi|209947788 Escherichia coli plasmid pEC278 REFSEQ 
Farm effluent 2012 gi|255929160 Endophytic bacterium LOB-07 plasmid pLK39 REFSEQ 
Farm effluent 2012 gi|305678726 Klebsiella pneumoniae plasmid unnamed REFSEQ 
Farm effluent 2012 gi|435855445 Enterobacter cloacae strain BB1092 plasmid pB1023 REFSEQ 
Farm effluent 2012 gi|435855463 Klebsiella pneumoniae strain BB1088 plasmid pB1019 REFSEQ 
Farm effluent 2012 gi|482907348 U288 plasmid pSTU288-3 REFSEQ 
Farm effluent 2012 gi|507579660 Cronobacter sakazakii strain ATCC 29544 plasmid pCSA2 REFSEQ 
Farm effluent 2012 gi|690630974 Escherichia coli strain K317 plasmid ColE7-K317 REFSEQ 
Farm effluent 2012 gi|746219889 Enterobacter cloacae strain 34998 plasmid p34998-4.921kb REFSEQ 
Farm effluent 2012 gi|749202706 Enterobacter cloacae strain 34983 plasmid p34983-328.905kb REFSEQ 
Farm effluent 2012 gi|749293681 Klebsiella oxytoca strain M1 plasmid pKOXM1D REFSEQ 
Farm effluent 2012 gi|749296055 pneumoniae Kp13 plasmid pKP13b REFSEQ 
Farm effluent 2012 gi|765030385 Enterobacter cloacae strain 34978 plasmid p34978-4.938kb REFSEQ 
Farm effluent 2012 gi|817657570 Cronobacter sakazakii strain ATCC 29544 plasmid CSK29544_2p REFSEQ 
Farm effluent 2012 gi|57635337  plasmid pSEM integron INTEGRALL 
Farm effluent 2012 gi|20530945  integron-derived beta-lactamase (ampC) INTEGRALL 
Farm effluent 2012 gi|94442253  class 1 integron IntI1 INTEGRALL 
Farm effluent 2013 gi|763126141 Lactobacillus salivarius strain JCM 1046 plasmid pCTN1046 REFSEQ 
Farm effluent 2013 gi|587656492  class 1 integron IntI1 INTEGRALL 
Farm effluent 2014 gi|13345249 class 2 integron IntI2 INTEGRALL 
Farm effluent 2014 gi|788265642 class 2 integron IntI2 INTEGRALL 
Farm effluent 2014 gi|215397925 ICE integron putative INTEGRALL 
  
39 
WWTP effluent 2012 gi|13345249  class 2 integron IntI2 INTEGRALL 
WWTP effluent 2012 gi|563324892  class 2 integron IntI2 INTEGRALL 
WWTP effluent 2012 gi|710572  mercury resistance transposon INTEGRALL 
WWTP effluent 2012 gi|215397925  ICE integron putative INTEGRALL 
WWTP effluent 2013 gi|194442162 Bacteroides fragilis plasmid pBFP35 REFSEQ 
WWTP effluent 2013 gi|661525289 Bacteroides cellulosilyticus WH2 plasmid pBWH2A REFSEQ 
WWTP effluent 2013 gi|659224469  class 1 integron IntI1 INTEGRALL 
WWTP effluent 2014 gi|194442162 Bacteroides fragilis plasmid pBFP35 REFSEQ 
WWTP effluent 2014 gi|294057975 Sphingobium japonicum UT26S plasmid pUT2 DNA REFSEQ 
WWTP effluent 2014 gi|661525289 Bacteroides cellulosilyticus WH2 plasmid pBWH2A REFSEQ 
WWTP effluent 2014 gi|766626985 Aeromonas hydrophila strain AL06-06 plasmid pAH06-06-2 REFSEQ 
WWTP effluent 2014 gi|242876676  plasmid class 1 integron INTEGRALL 
WWTP effluent 2014 gi|215397925  ICE integron putative INTEGRALL 
  40 
3.4.4. Taxonomic profiling and pathogen detection 
Finally, the effluent metagenomes were subjected to taxonomic profiling. At genus level, the 
most abundant prokaryotes in the farm samples were Pseudomonas  (farm effluent 2012) 
and Butyrivibrio (farm effluent 2013 and 2014). The most abundant prokaryotes at genus 
level in the WWTP samples were Acinetobacter (WWTP effluent 2012), Thiomonas (WWTP 
effluent 2013) and Proteus (WWTP effluent 2014). For the river source water, the most 
abundant prokaryotic genus was Sphingobium. After cross-referencing the identified species 
level, clade-specific marker genes for all the metagenomes to the PATRIC pathogen 
database (Gillespie et al., 2011), a total of 35 species of potential bacterial pathogens were 
identified (Figure 3.3, Table 3.4). The most commonly identified species were E.coli, 
A.butzleri, E.rectale, R.bromii and S.enterica. The WWTP effluent 2014 contained the 
greatest diversity of potential bacterial pathogens, whereas the river source water and the 
WWTP effluent 2012 were found to contain the lowest diversity. 
 
  41 
 
Figure 3.3 Metagenomic phylogenetic analysis and annotation of potential bacterial pathogens. The 
phylogenetic tree was built using Graphlan from the merged Metaphlan and LEfSe output for the effluent 
metagenomes. The PATRIC pathogens are highlighted as red stars and the external rings denote species 
prevalence in each metagenome. 
  
  42 
Table 3.4 Pathogen analysis of effluents. The table lists the potential bacterial pathogens (according to 
Metaphlan and PATRIC analysis) found in the effluents entering the river catchments. 
Sample Bacterial pathogen 
Farm effluent 2012 Bifidobacterium adolescentis 
Farm effluent 2012 Enterobacter cloacae 
Farm effluent 2012 Escherichia coli 
Farm effluent 2012 Klebsiella oxytoca 
Farm effluent 2012 Klebsiella pneumoniae 
Farm effluent 2012 Pseudomonas aeruginosa 
Farm effluent 2012 Pseudomonas putida 
Farm effluent 2012 Ruminococcus bromii 
Farm effluent 2012 Salmonella enterica 
Farm effluent 2013 Arcobacter butzleri 
Farm effluent 2013 Bifidobacterium adolescentis 
Farm effluent 2013 Bifidobacterium longum 
Farm effluent 2013 Enterococcus faecium 
Farm effluent 2013 Escherichia coli 
Farm effluent 2013 Eubacterium rectale 
Farm effluent 2013 Faecalibacterium prausnitzii 
Farm effluent 2013 Lactobacillus fermentum 
Farm effluent 2013 Lactobacillus gasseri 
Farm effluent 2013 Lactobacillus plantarum 
Farm effluent 2013 Lactobacillus rhamnosus 
Farm effluent 2013 Ruminococcus bromii 
Farm effluent 2013 Salmonella enterica 
Farm effluent 2014 Arcobacter butzleri 
Farm effluent 2014 Bacteroides fragilis 
Farm effluent 2014 Bifidobacterium adolescentis 
Farm effluent 2014 Bifidobacterium longum 
Farm effluent 2014 Bifidobacterium pseudolongum 
Farm effluent 2014 Butyrivibrio fibrisolvens 
Farm effluent 2014 Campylobacter jejuni 
Farm effluent 2014 Corynebacterium aurimucosum 
Farm effluent 2014 Escherichia coli 
Farm effluent 2014 Eubacterium rectale 
Farm effluent 2014 Faecalibacterium prausnitzii 
Farm effluent 2014 Propionibacterium acnes 
Farm effluent 2014 Ruminococcus bromii 
Farm effluent 2014 Ruminococcus obeum 
Farm effluent 2014 Ruminococcus torques 
Farm effluent 2014 Salmonella enterica 
River source 2013 Enterococcus faecium 
River source 2013 Eubacterium rectale 
River source 2013 Lactobacillus gasseri 
River source 2013 Lactobacillus rhamnosus 
  43 
River source 2013 Propionibacterium acnes 
River source 2013 Salmonella enterica 
WWTP effluent 2012 Arcobacter butzleri 
WWTP effluent 2012 Bifidobacterium adolescentis 
WWTP effluent 2012 Escherichia coli 
WWTP effluent 2012 Propionibacterium acnes 
WWTP effluent 2012 Ruminococcus torques 
WWTP effluent 2012 Salmonella enterica 
WWTP effluent 2013 Arcobacter butzleri 
WWTP effluent 2013 Bifidobacterium adolescentis 
WWTP effluent 2013 Bifidobacterium longum 
WWTP effluent 2013 Escherichia coli 
WWTP effluent 2013 Eubacterium rectale 
WWTP effluent 2013 Faecalibacterium prausnitzii 
WWTP effluent 2013 Gordonibacter pamelaeae 
WWTP effluent 2013 Ruminococcus bromii 
WWTP effluent 2013 Ruminococcus obeum 
WWTP effluent 2013 Ruminococcus torques 
WWTP effluent 2013 Streptococcus thermophilus 
WWTP effluent 2014 Arcobacter butzleri 
WWTP effluent 2014 Bacteroides fragilis 
WWTP effluent 2014 Bacteroides xylanisolvens 
WWTP effluent 2014 Bifidobacterium adolescentis 
WWTP effluent 2014 Bifidobacterium longum 
WWTP effluent 2014 Eggerthella lenta 
WWTP effluent 2014 Enterobacter cloacae 
WWTP effluent 2014 Enterococcus faecalis 
WWTP effluent 2014 Escherichia coli 
WWTP effluent 2014 Eubacterium cylindroides 
WWTP effluent 2014 Eubacterium rectale 
WWTP effluent 2014 Eubacterium siraeum 
WWTP effluent 2014 Faecalibacterium prausnitzii 
WWTP effluent 2014 Gordonibacter pamelaeae 
WWTP effluent 2014 Klebsiella pneumoniae 
WWTP effluent 2014 Proteus mirabilis 
WWTP effluent 2014 Pseudomonas aeruginosa 
WWTP effluent 2014 Ruminococcus bromii 
WWTP effluent 2014 Ruminococcus obeum 
WWTP effluent 2014 Ruminococcus torques 
WWTP effluent 2014 Streptococcus thermophilus 
WWTP effluent 2014 Vibrio parahaemolyticus 
 
  44 
3.5. Discussion 
The comparative metagenomic approach used in this chapter has shown that two types of 
effluent entering a shared river catchment contain ARGs and MGEs at higher average 
abundances than were present in a background sample of the river source water. This would 
suggest that effluents such as these are likely to serve as sources of ARGs and thus 
contribute to the resistome of river catchments and other aquatic environments. Because of 
this, it may be appropriate to routinely monitor such effluents as sources of ARGs, particularly 
when considering the current view of ARGs as environmental contaminants (Pruden et al., 
2006) and the call for an environmental framework to tackle antimicrobial resistance 
(Berendonk et al., 2015). 
One such reason for the high abundance of ARGs in effluents may be the presence of 
antimicrobial compounds that could consequently provide a selective pressure for the 
maintenance of ARGs. There have been several studies that document the presence of 
antimicrobial compounds, from both human and veterinary medicine, in the environment 
(Kemper, 2008, Hu et al., 2010). Although these compounds are often present at relatively 
low concentrations, some studies have shown therapeutic concentrations of antimicrobials 
being discharged into the environment, such as the effluent from Indian drug manufacturers 
containing toxic concentrations of antimicrobial compounds (Larsson et al., 2007). 
Subsequent studies by Larsson et al. found a high abundance of ARGs downstream of the 
effluent discharge point relative to upstream of the manufacturers and when compared to a 
Swedish WWTP (Kristiansson et al., 2011). While the environmental release of antimicrobial 
compounds at therapeutic concentrations is largely prevented in the UK, Europe and US 
through proper wastewater management and controls, clinically important antimicrobials can 
be found in the environment at sub-inhibitory concentrations and it is possible that these very 
low antimicrobial concentrations could be enriching for resistant bacteria and promote 
increased persistence of ARGs (Gullberg et al., 2011). Thus, it may be pertinent to couple 
future environmental ARG monitoring frameworks and antimicrobial resistance risk 
assessments with information on antimicrobial usage and the antimicrobial concentrations in 
the effluents being investigated. 
Interestingly, the average abundance of ARGs was found to be greater in the farm effluents 
than in the WWTP effluents (Figure 3.1). Although these two effluents are from differently 
treated faecal sources, one being a treated effluent (sedimentation treatment) from a 
municipal WWTP (i.e. predominantly human faecal source) and the other being an untreated 
  45 
effluent from a farm (predominantly bovine faecal source), this finding does offer some insight 
into the debate surrounding the relative impact of human and animal contributions to the 
development of antimicrobial resistance (Phillips et al., 2004, Mather et al., 2013). The fact 
that WWTP effluent had undergone a form of water treatment prior to being released into the 
river catchment, whereas the farm effluent did not, may suggest that some form of water 
treatment could reduce the abundance or diversity of ARGs. A comparison of WWTP crude 
influent to the effluent could elaborate on the effectiveness of sedimentation treatment on the 
abundance of ARGs. Studies have shown that wastewater treatment processes do not 
completely remove ARGs (Wang et al., 2015) and that some WWTP processing can result in 
an increase in the proportion of antimicrobial resistant bacteria in WWTP effluents (Harris et 
al., 2012). Considering that effluents may also disseminate antimicrobial compounds, it raises 
the question as to whether the combination of ARGs and antimicrobial compounds within 
effluents is resulting in the expression of ARGs and the occurrence of phenotypic 
antimicrobial resistance. This will be addressed in the work of the next chapter with the aim of 
identifying additional factors that contribute to the dissemination of ARGs. 
In terms of the mobility of genes within the effluents, an array of mobilisable plasmids, 
integrons and transposons were present in the metagenomes (Figure 3.2) and many of the 
ARGs identified aligned to the Repository of Antibiotic resistance Cassettes (RAC) (Tsafnat 
et al., 2011). This raises the possibility that the ARGs within the effluents could be readily 
mobilised into other bacteria, including both directly into pathogens also discharged into the 
environment and environmental bacteria. These environmental bacteria in turn could pose a 
risk as potential bacterial intermediaries, harbouring these ARGs in the environment prior to 
transferring them into other pathogens. Indeed, the taxonomic analysis revealed a diverse 
array of bacterial species present across all of the effluent samples (Figure 3.3), including 
several pathogenic species (Table 3.4). 
Based on the observations in this study, it is recommended that future ARG monitoring 
frameworks and antimicrobial resistance risk assessments should incorporate direct MGE 
and pathogen detection with metagenomic assessments of effluents entering river 
catchments, especially considering the absence of MGEs and the lower diversity of 
pathogens found in the river source water. However, this study did identify five resistance 
genes in the river source water conferring resistance to two classes of antimicrobials. When 
normalised to 16S sequences the river source water were found to be accountable for the 
most abundant phenicol resistance gene and the third most abundant aminoglycoside 
resistance genes out of all the metagenome libraries examined. However, when using the 
  46 
raw SEAR abundance metric, that does not include normalisation to the 16S sequences 
within the sample, the relative abundance of ARGs from the river source water were reduced 
relative to the other effluent samples. This raises the question as to whether 16S 
normalisation is the most appropriate approach to metagenomic abundance estimates as 
factors such as variation in 16S copy number can skew the data generated as well as 
interpretation (Case et al., 2007). An alternative could be to use the RPKM value generated 
as part of the SEAR analysis and featured in Table 3.2. 
The metagenomic approach used was relatively less sensitive than more direct-targeted 
measures of known ARG abundance (e.g. qPCR-based detections (LaPara et al., 2011)). 
However it had the advantages that it was relatively unbiased and semi-quantitative, giving a 
good estimation of relative key ARG and MGE abundance and diversity across bacterial 
populations. It was also potentially able to detect novel ARGs that would otherwise not be 
found using these more targeted approaches.  
 
3.6. Conclusion 
The work in this chapter delivers a detailed metagenomic analysis of effluents entering a river 
catchment. Effluents were found to contain an array of ARGs, MGEs and pathogenic bacteria 
that, when compared to a background sample, were found to be more diverse and abundant 
than in the river source water. This study has shown that the discharge of effluents into river 
catchments contributes to the dissemination of ARGs, MGEs and pathogenic bacteria, and 
that they may play an important role in the propagation of environmental reservoirs of ARGs. 
This work has demonstrated the suitability of metagenomics as a key component of an 
environmental framework to monitor ARG dissemination. When used in conjunction with 
SEAR, metagenomics is able to quickly and efficiently detect horizontally acquired ARGs in 
samples from the aquatic environment. 
 
  47 
Chapter 4. Expression of ARGs in effluents 
4.1. Preface 
The work described in the previous chapters has begun to develop a framework for 
monitoring antimicrobial resistance genes (ARGs) in effluents that are entering the aquatic 
environment. Using this framework, effluents entering a single river catchment were shown to 
contain ARGs in greater abundance and diversity than the environment they were entering. 
This finding raises the question as to whether or not the ARGs in effluents are being actively 
maintained and used in the environment, thus providing a potential explanation for the higher 
abundance and diversity of ARGs in effluents than compared to background levels. 
The work of this chapter assesses whether ARGs within effluents are being transcribed 
across a series of longitudinal samples. In addition to this, the presence of potential 
antimicrobial selection pressures was explored and the occurrence of pathogenic species of 
bacteria in effluents was investigated. 
 
4.2. Introduction 
The continued scientific interest in the effects of anthropogenic activities on environmental 
microbial communities has led to the impact of these activities on the development of 
antimicrobial resistance in the environment being questioned and consequently, the possible 
implications they may have for public and animal health being debated (Martinez, 2009, 
Wellington et al., 2013, Port et al., 2014, Berendonk et al., 2015) (see Chapter 6). 
Previous chapters have described a diverse array of ARGs in effluents entering an aquatic 
environment, adding to the wealth of research that has demonstrated the global 
dissemination of ARGs within a variety of environmental biomes (Pruden et al., 2006, Zhang 
and Zhang, 2011, Devarajan et al., 2015, Li et al., 2015a). The fact that the diversity (and 
abundance) of ARGs was found to be greater in effluents than in the receiving environment 
(see Chapter 3) may be due to a greater concentration of ARG-containing bacteria present in 
effluents from faecal sources. Additionally, it may be the case that there are selective forces 
impacting the maintenance of ARGs in these effluents that are either not present or diluted in 
the wider environment. To explore these possibilities, effluents should be tested for 
anthropogenic factors that may constitute selective forces for ARG uptake and maintenance. 
  48 
One such anthropogenic factor likely to act as a selective force, impacting ARG abundance 
and diversity, is the presence of antimicrobial compounds that may select for specific ARGs 
in bacterial communities. Several recent studies have documented the presence of 
pharmaceutical residues in the aquatic environment, usually as a result of the manufacturing 
process and wastewater treatment plant (WWTP) treatment (Khan et al., 2013, McEneff et 
al., 2014). However farming practices, particularly in aquaculture and fishery management, 
also lead to the release of antimicrobial compounds into the environment (Kemper, 2008, 
Cabello et al., 2013); as do human sources such as hospital effluents (Martinez, 2009, 
Kleywegt et al., 2015) and WWTP effluents (Li et al., 2008). Studies have also identified 
antimicrobial compounds in river and estuary environments that are not in direct receipt of 
industrial pharmaceutical run-off or domestic waste (Lu et al., 2015, McEneff et al., 2014). As 
antimicrobial compounds are designed to kill susceptible bacteria or inhibit their growth, and 
there are reports of sub-inhibitory concentrations of sulfonamide antimicrobials in aquatic 
environments associated with expression of sulfonamide ARG sul1 (Bruchmann et al., 2013), 
it is reasonable to assume that the environmental release of antimicrobials might select for 
the presence and maintenance of ARGs. Indeed studies have interrogated environmental 
samples for both ARGs and antimicrobial compounds, however these studies have been 
limited to specific ARGs, and not performed using multiple effluent sites over multiple time 
points or have not taken into account the antimicrobial usage at the effluent source (Luo et 
al., 2010, Li et al., 2012, Lu et al., 2015). 
In this chapter, metagenomics and a combination of liquid chromatography mass 
spectrometry (LCMS) combined with antimicrobial usage data were used to determine if 
antimicrobial usage at effluent source points resulted in an enrichment of ARGs (and 
pathogenic species of bacteria) that are entering a river catchment. The work of this chapter 
aimed to establish whether the ARGs present in the environment are being expressed, thus 
offering insight into whether ARGs are actively utilised by the members of the bacterial 
communities within the effluents, thus exploring their maintenance in the environment. To 
assess both the ARGs and the ARG transcripts of bacterial communities within effluent 
samples, a combination of metagenomic and metatranscriptomic approaches were utilised. 
This study also aimed to relate the expression of ARGs to factors such as the abundance of 
ARGs, as well as selective pressures such as antimicrobial presence and usage. Indeed, 
studies that have investigated both antimicrobial residue concentration and ARG abundance 
in effluents (of a pharmaceutical WWTP) have shown significant correlations between 
antimicrobial concentrations and the associated relative ARG abundance (Luo et al., 2010, 
  49 
Wang et al., 2015). However, no research has been done that aims to correlate ARG 
abundance and antimicrobial resistance selection pressures to the abundance of ARG 
transcripts in effluents and thus show that ARGs are expressed and utilised by active 
members of bacterial communities in the environment. Finally, through the collection of a 
series of longitudinal samples, this work documented the continued dissemination and 
dynamic trends in ARGs and pathogenic species of bacteria entering the environment via 
effluents. 
 
4.3. Methods 
4.3.1. Sample collection, DNA and RNA sequencing 
Samples were collected from three sources within the River Cam Catchment, Cambridge, UK 
on 02.05.2013 and over a five-month period (August 2014 – December 2014), an average of 
five weeks apart (n=6 sampling times, Appendix 3). Collections were made from the 
combined wastewater effluent of the main wards of Addenbrooke’s hospital, Cambridge, UK, 
via a combined sewage pit/drain access (latitude: 52.174343, longitude: 0.139346) prior to 
the effluent entering the municipal sewers. The other two sources from which collections 
were made were the effluent lagoon of the University of Cambridge dairy farm (prior to the 
effluent being distributed to surrounding fields as fertiliser, detailed in 2.3.3.2) and the River 
Cam source water (collected at Ashwell Spring, detailed in 3.3.1). The river source water 
served as a background sample for the environment that both effluents were entering. 
Samples for antimicrobial residue testing were collected in 1 L sterile glass containers and 
transported at 4°C to the laboratory. Samples for metagenome and metatranscriptome 
preparation were collected in 10 L sterile polypropylene containers, transported at 4°C to the 
laboratory and prokaryotic cells were isolated as described in 2.3.3.2. Each sample of 
prokaryotic cells was then split in two, for separate DNA and RNA extractions to generate a 
metagenome and metatranscriptome per sample. The metagenome preparation followed the 
same DNA extraction methodology described in 2.3.3.2. For each metagenome, 2 µg of DNA 
was used to generate Illumina paired-end libraries (100 bp). For each metatranscriptome, the 
prokaryotic cells were washed in phosphate buffered saline solution before being treated with 
Max Bacterial Enhancement reagent (ThermoFisher Scientific, UK) to denature bacterial 
proteins and deactivate RNases. Bacterial cell lysis and RNA extraction was then performed 
using TRIzol reagent (ThermoFisher Scientific, UK). For each metatranscriptome 2µg of RNA 
  50 
was subjected to ribosomal RNA depletion (Ribo-Zero Gold, Epicentre, UK), quality checked 
using a BioAnalyzer (Agilent Technologies, US) and used to generate Illumina TruSeq RNA 
libraries (100 bp). All metagenome and metatranscriptome libraries were sequenced using an 
Illumina HiSeq2500 (Exeter Sequencing Service, UK). A full description of the metagenomic 
and metatranscriptomic samples used in this chapter is available in Appendix 3. 
 
4.3.1.1. Antimicrobial residue testing 
Antimicrobial residues were quantified in effluent samples using LCMS. Quantification 
standards were created for the three most used compounds in each class of antimicrobials 
prescribed at Addenbrooke’s hospital in 2013. No standards could be generated for 
aminoglycosides or trimethoprims. LCMS antimicrobial residue testing was performed by 
RPS Mountainheath, Hertfordshire, UK. 
 
4.3.2. Bioinformatic analyses 
4.3.2.1. Identification of ARGs, MGEs and abundance analysis 
ARGs were identified in metagenomes and metatranscriptomes using the Search Engine for 
Antimicrobial Resistance (SEAR) with default parameters (Rowe et al., 2015). A full 
description of SEAR is available in Chapter 2. MGE detection was carried out on all 
metagenomes using the methodology described in Chapter 3. For each metagenome ARG 
and MGE data was normalised to the number 16S rRNA sequences as in 3.3.2.3.  
 
4.3.2.2. Antimicrobial usage and statistics 
Monthly antimicrobial usage data for periods overlapping sample collection were obtained for 
Addenbrooke’s hospital. Usage data was generated via the EPIC health record system for 
each class of antimicrobial used by the hospital that month. Hospital antimicrobial usage is 
recorded as Defined Daily Dose (DDD), which is the assumed average maintenance dose 
per day for a given drug and is used as a statistical measure of drug consumption.  
 
  51 
4.3.2.3. Taxonomic profiling and pathogen detection 
Taxonomic profiling and pathogen detection was performed on all metagenomes according to 
the methodology described in 3.3.2.4. 
 
4.4. Results 
4.4.1. Metagenome and metatranscriptome analysis 
In this chapter, 102 Giga base-pairs (Gbp) of sequencing data was generated across all 
samples. Metagenomes from three samples of the river source water (AS:M:4, AS:M:5 and 
AS:M:6) failed sequencing library quality checking and were removed from the study. A total 
of 15 metagenomes were successfully sequenced, passed quality checking and were found 
to contain ARGs at varying levels (Table 4.1). The percentage of reads matching ARGs was 
an average 10-fold greater in the hospital effluent samples, compared to the farm effluent 
samples, and approximately 70-fold greater than the background samples of river source 
water. The percentage of reads matching ARGs was an average 8-fold greater in the farm 
effluent compared to the background samples of river source water, however, one 
metagenome from the background sample (AS:M:2) was found to have a greater percentage 
of ARG reads than a metagenome from the farm effluent (DF:M:2). 
  
  52 
Table 4.1 Summary of metagenomes used in this work (metadata available in Appendix 3). 
Sample Total reads Gbp Total ARG reads % ARG reads 
AH:M:1 64659230 3.2330 97698 0.1511 
AH:M:2 52355416 2.6178 122164 0.2333 
AH:M:3 109795652 5.4898 207767 0.1892 
AH:M:4 61573380 3.0787 125019 0.2030 
AH:M:5 50845128 2.5423 25987 0.0511 
AH:M:6 53928494 2.6964 28629 0.0531 
DF:M:2 66120642 3.3060 2317 0.0035 
DF:M:3 184149408 9.2075 13094 0.0071 
DF:M:4 262823622 13.1412 29006 0.0110 
DF:M:5 58179398 2.9090 31518 0.0542 
DF:M:6 53192154 2.6596 6999 0.0132 
DF:M:7 49516248 2.4758 4072 0.0082 
AS:M:1 54799282 2.7400 181 0.0003 
AS:M:2 150787198 7.5394 7226 0.0048 
AS:M:3 128125534 6.4063 1199 0.0009 
AH = hospital effluent (Addenbrooke’s hospital) 
DF = farm effluent (dairy farm) 
AS = river source water (Ashwell spring) 
Numbers in sample name represent sequential sampling 
 
Of the RNA samples, all six samples of the river Cam source water (AS:T:1 - AS:T:6), two 
samples of hospital effluent (AH:T:2 and AH:T:3) and two samples of farm effluent (DF:T:2 
and DF:T:3) did not yield sufficient RNA for metatranscriptome sequencing and were 
consequently removed from the analysis. A total of eight metatranscriptomes were 
successfully sequenced, passed quality checking and were found to contain transcripts of 
ARGs (Table 4.2).  
 
Table 4.2 Summary of metatranscriptomes used in this work (metadata available in Appendix 3). 
Sample Associated metagenome Total reads Total ARG reads % ARG reads 
AH:T:1 AH:M:1 152298536 308848 0.2028 
AH:T:4 AH:M:4 74411930 948890 1.2752 
AH:T:5 AH:M:5 61143518 23765 0.0389 
AH:T:6 AH:M:6 51640378 40379 0.0782 
DF:T:1 DF:M:2 123559962 8017 0.0065 
DF:T:4 DF:M:5 49293728 4447 0.0090 
DF:T:5 DF:M:6 64102402 7057 0.0110 
DF:T:6 DF:M:7 64850756 1022 0.0016 
 
  53 
4.4.2. Correlation of ARG abundance to ARG transcript abundance 
A Pearson product-moment correlation was performed to determine the relationship between 
the abundance of ARGs in the metagenome and the abundance of ARG transcripts in the 
metatranscriptome detected in each type of environmental effluent (hospital and farm). A 
strong, positive correlation was observed between ARG abundance and corresponding 
transcript abundance across all hospital effluent samples, which was statistically significant 
(r=0.8550, p<0.0005, n=648 ARGs) (Figure 4.1). However, only a weak, positive correlation 
was observed between ARG abundance and corresponding transcript abundance across all 
farm effluent samples, which was statistically significant (r=0.1467, p<0.005, n=368 ARGs) 
(Figure 4.2). 
 
 
Figure 4.1 Linear regression analysis of ARG abundance against corresponding transcript abundance, 
for ARGs detected in all hospital effluent samples. The coefficient of determination (r2) is included as a 
measure of how well the data fits the linear regression model. 
  54 
 
 
Figure 4.2 Linear regression analysis of ARG abundance against corresponding transcript abundance, 
for ARGs detected in all farm effluent samples. 
 
Pearson product-moment correlations were also performed for ARG abundance and ARG 
transcript abundance for ARGs grouped by antimicrobial class. For the hospital effluent 
samples, a strong, positive correlation between ARG abundance and corresponding 
transcript abundance was observed for nine of ten classes of antimicrobial (Table 4.3). 
Abundance of sulfonamide ARGs and transcripts was the only antimicrobial class with a 
weak correlation, for which the null hypothesis (ARG transcript abundance is not associated 
with ARG abundance at significance level 0.05) was accepted. For the farm effluent samples, 
a weak positive correlation was observed for tetracycline ARGs and transcripts, however no 
correlation was found for all other antimicrobial classes (Table 4.4). 
  
  55 
Table 4.3 Correlation of ARG and transcript abundance for all hospital effluent samples, according to 
antimicrobial class. The null hypothesis, that ARG transcript abundance is not associated with ARG 
abundance, was rejected for all classes of antimicrobials except sulfonamides (for which the P-value was 
>0.05). The MLS class of antimicrobial represents marcolides, lincosamides and streptogramins. 
Antimicrobial class r value p value Number of ARGs H0 (5%) 
aminoglycosides 0.8444 <0.0005 144 rejected 
beta lactams 0.9165 <0.0005 188 rejected 
glycopeptides 0.7611 <0.0005 52 rejected 
MLS 0.6802 <0.0005 68 rejected 
phenicols 0.9900 <0.0005 32 rejected 
quinolones 0.9891 <0.0005 16 rejected 
rifamycins 0.9842 <0.05 4 rejected 
sulfonamides 0.5290 <0.5 12 accepted 
tetracyclines 0.7207 <0.0005 88 rejected 
trimethoprims 0.9592 <0.0005 44 rejected 
 
Table 4.4 Correlation of ARG and transcript abundance for all farm effluent samples, according to 
antimicrobial class. Correlations could not be performed for glycopeptides, phenicols, quinolones or rifamycins 
due to insufficient data points. The null hypothesis, that ARG transcript abundance is not associated with ARG 
abundance, was only rejected for tetracycline ARGs and transcripts. NA denotes not analysed as insufficient 
data points available. 
Antimicrobial class r value p value Number of ARGs H0 (5%) 
aminoglycosides 0.0187 >0.5 96 accepted 
beta lactams 0.14419 <0.5 71 accepted 
glycopeptides NA NA 1 NA 
MLS 0.09823 <0.5 59 accepted 
phenicols NA NA 6 NA 
quinolones NA NA 1 NA 
rifamycins NA NA 1 NA 
sulfonamides 0.1989 >0.5 11 accepted 
tetracyclines 0.48158 <0.0005 79 rejected 
trimethoprims -0.11174 >0.5 11 accepted 
 
4.4.3. Antimicrobial residues 
Concentrations of antimicrobial residues in all samples were determined by LCMS (Table 
4.5). The reporting limit for LCMS had to be raised from 0.1µg/ L on several samples due to 
matrix interference, low sensitivity of target analyte, or poor recovery from matrix spikes. For 
the hospital effluent samples, eight antimicrobial compounds spanning six classes of 
antimicrobials were identified at least once in the samples. The antibiotics vancomycin 
  56 
(glycopeptide), clarithromycin (MLS) and ciprofloxacin (quinolone) were the only antimicrobial 
compounds to be identified in every hospital effluent sample. Of the antimicrobial compounds 
tested for, the only classes of antimicrobial not identified in any hospital effluent samples 
were phenicols and tetracyclines. The only antibiotic found in the farm effluent was the 
sulfonamide antibiotic sulfadiazine; which was not found in any hospital effluent samples. No 
traces of antimicrobial residues were found in the river source water samples (not shown). 
  
  
57 
Table 4.5 LCMS results for hospital and farm effluent samples (associated with metagenome samples). A ‘<’ denotes a result of less than the reporting limit (listed 
after ‘<’), bold type indicates detected concentrations. 
Class Drug Concentration of antimicrobial present in sample (µg/litre) 
AH:M:2 AH:M:3 AH:M:4 AH:M:5 AH:M:6 DF:M:3 DF:M:4 DF:M:5 DF:M:6 DF:M:7 
beta lactam amoxicillin < 10.0 < 10000.0 < 10000.0 failed < 10000.0 < 1000.0 < 50.0 < 1000.0 failed < 1000.0 
beta lactam flucloxacillin < 0.1 < 1.0 < 0.1 < 0.5 24.5 < 100.0 < 1.0 < 2.0 < 5.0 < 100.0 
beta lactam piperacillin < 200.0 < 100.0 < 100.0 < 200.0 < 1000.0 < 100.0 < 10.0 < 5.0 < 10.0 < 100.0 
glycopeptide teicoplanin < 500.0 < 5000.0 < 100.0 < 1000.0 < 100.0 < 10000.0 < 5000.0 < 5000.0 < 10000.0 < 10000.0 
glycopeptide vancomycin 18.2 3160 29.1 58.1 90.2 < 1000.0 < 500.0 < 100.0 < 1000.0 < 1000.0 
MLS azithromycin 57.4 < 0.1 failed < 0.5 < 10.0 < 1000.0 < 5.0 failed < 50.0 < 1000.0 
MLS clarithromycin 13.4 23.6 17.4 31.1 7.3 < 10.0 < 5.0 < 2.0 < 10.0 < 10.0 
MLS erythromycin < 0.1 < 1.0 < 1.0 < 1.0 < 0.1 < 10.0 < 1.0 < 5.0 < 10.0 < 10.0 
phenicol chloramphenicol < 0.1 < 1.0 < 0.1 < 1.0 < 1.0 < 100.0 < 10.0 < 10.0 < 100.0 < 100.0 
quinolone ciprofloxacin 59.3 10 82.9 10.1 6.5 < 200.0 < 100.0 < 100.0 < 100.0 < 200.0 
quinolone moxifloxacin 2 < 2.0 < 10.0 < 1.0 < 20.0 < 200.0 < 20.0 < 100.0 < 100.0 < 200.0 
quinolone ofloxacin < 0.5 < 2.0 < 5.0 < 2.0 < 0.2 < 200.0 < 20.0 < 100.0 < 50.0 < 200.0 
rifamycin rifabutin < 0.1 < 1.0 < 1.0 < 5.0 < 0.1 < 10.0 < 10.0 < 1.0 < 50.0 < 10.0 
rifamycin rifampicin < 0.5 < 2.0 < 50.0 < 5.0 1.3 < 10.0 < 2.0 < 50.0 < 10.0 < 10.0 
sulfonamide sulfadiazine < 0.1 < 0.1 < 0.1 < 0.1 < 0.1 < 10.0 1.7 2.7 < 10.0 < 10.0 
sulfonamide sulfamethoxazole 257 1220 < 1.0 173 15.4 < 10.0 < 10.0 < 5.0 < 10.0 < 10.0 
tetracycline demeclocycline < 0.5 < 100.0 < 100.0 < 10.0 < 10.0 < 100.0 < 100.0 < 1000.0 < 100.0 < 100.0 
tetracycline doxycycline < 1.0 < 5.0 < 1000.0 < 5.0 < 5.0 < 100.0 < 500.0 < 1000.0 < 50.0 < 100.0 
tetracycline tigecycline < 1.0 < 10000.0 < 1000.0 < 10.0 < 100.0 < 1000.0 failed < 1000.0 < 100.0 < 1000.0 
 
  58 
4.4.4. Expression of ARGs in response to hospital antimicrobial 
usage 
Based on the strong, positive correlation observed between ARG abundance and ARG 
transcript abundance in hospital effluent across the four sampling dates, in addition to the 
presence of antimicrobial residues in hospital effluent samples, the hospital was the focus of 
further analysis. The abundance of ARG transcripts in each hospital effluent 
metatranscriptome was correlated to the monthly hospital antimicrobial usage records to test 
if antimicrobial usage was associated with ARG transcript abundance (Figure 4.3). As no 
correlation was found between sulfonamide ARGs and ARG transcripts (4.4.2), they were not 
included in this analysis. Also, due to hospitals recording sulfonamide usage together with 
trimethoprim usage, trimethoprim ARG transcripts were also not included in this analysis. 
Overall a weak, positive correlation was observed between ARG transcript abundance and 
antimicrobial usage in all hospital effluent samples, which was statistically significant 
(r=0.5051, p<0.005, n=32). However, as noted in the figure key (Figure 4.3), the association 
between transcript abundance and antimicrobial usage varied greatly between antimicrobial 
classes. The strongest, positive correlation was observed for the tetracycline class of 
antimicrobials (r=0.9163, r2=0.8397). Rifamycins also showed a positive correlation 
(r=0.5527, r2=0.3054) between transcript abundance and rifamycin usage. Negative 
correlations were observed for the glycopeptide and phenicol classes of antimicrobials. 
 
  59 
 
Figure 4.3 Linear regression analysis of ARG transcript abundance against hospital antimicrobial 
usage, grouped by antimicrobial class for each hospital effluent sample. The Pearson product-moment 
correlation (r) and coefficient of determination (r2) values for each antimicrobial class are included in the figure 
key. 
 
In order to investigate the effect of both ARG abundance and antimicrobial usage on the 
transcript abundance of ARGs in hospital effluent, and to test how much the data for these 
three variables differ, a multiple linear regression and ANOVA test was performed. The 
correlation coefficient of the multiple linear regression revealed a strong positive correlation 
between all the three variables, for which ARG abundance and antimicrobial usage can 
account for 79% of the variation in ARG expression  (r=0.8903, adjusted r2=0.7857, 
p<0.0005, n=32). Despite this model being true at significance level 0.05 (p=5.6x10-11) the 
two independent variables differed greatly, ARG abundance (coefficient=8.06, p<0.0005) was 
a significant variable whereas antimicrobial usage (coefficient=0.00, p<0.5) was insignificant 
  60 
and thus the null hypothesis (ARG usage is not associated with ARG expression at 
significance level 0.05) was accepted for the antimicrobial usage variable. 
Finally, to further explore, and better visualise the data (ARG abundance, ARG transcript 
abundance and antimicrobial usage), a Principal Component Analysis (PCA) was performed 
on the three variables. Of the three resulting principal components (PC), PC1 and PC2 were 
able to explain the most variance between the variables (Figure 4.4).  
 
Figure 4.4 Scree plot showing the variance observed for the three principal components.  
 
When using a biplot to see how the initial variables contribute to these two principal 
components (and grouping observations by antimicrobial class), the least variation was found 
within the tetracycline, rifamycin, phenicol and glycopeptide groups (Figure 4.5). ARG 
transcript abundance, ARG abundance and antimicrobial usage can be negatively explained 
by PC1 (r=-0.95, -0.96, -0.64 respectively). ARG transcript and ARG abundance can be 
positively explained by PC2 (r=0.28 and 0.23 respectively), whereas antimicrobial usage is 
negatively explained by PC2 (r=-0.77). 
 
  61 
 
Figure 4.5 Biplot showing how the initial variables (ARG abundance, transcript abundance and 
antimicrobial usage) contribute to principal components 1 and 2. 
 
4.4.5. Taxonomic profiling and pathogen detection 
The number of different pathogenic bacterial species, in relation to the overall number of 
different species, was determined for each effluent metagenome (Figure 4.6). The hospital 
effluent sample from sampling point 2 (AH:M:2) contained the greatest number of pathogen 
species of all samples, whereas the farm effluent sample from sampling point 3 (DF:M:4) 
contained the lowest number of pathogen species of all samples. Sampling point 4 (AH:M:4, 
DF:M:5) was found to have the greatest number of pathogen species of all farm effluent 
samples and the lowest number of pathogen species of all hospital effluent samples. Overall, 
the number of pathogen species identified in the hospital and farm effluent samples over the 
six sampling points did not follow the same trend observed for the total species number in the 
same samples. 
 
  62 
 
Figure 4.6 Graph showing the number of pathogenic bacteria species and the total species number, for 
both hospital and farm effluents, over time. 
 
In terms of the frequency of isolation of particular pathogen species, only Escherichia coli 
was identified in all hospital and farm effluent samples (Figure 4.7 A). A total of 68 different 
species of bacterial pathogen were identified across all effluent samples, 27 of which were 
unique to hospital effluent and 5 were unique to farm effluent (Figure 4.7, B). Of the 36 
pathogen species shared by hospital and farm effluents, 47% (n=17) of species belonged to 
the phylum Firmicutes (predominantly Gram-positive, endospore forming bacteria). 
  63 
 
Figure 4.7 A. Column chart showing the ten most commonly identified pathogen species (according to 
Metaphlan and PATRIC analysis) in the hospital and farm effluent metagenomes. B. Venn diagram 
depicting the number pathogen species found in the combined hospital and farm effluent datasets. 
There were 36 species of pathogen that were common between hospital and farm effluent. 
 
4.4.6. MGEs 
The abundance of ARGs and MGEs was plotted for each effluent metagenome and 
compared against species number (Figure 4.8). The number of MGEs was consistently 
  64 
higher than the number of ARGs in all metagenomes. The hospital effluent metagenomes 
had a greater abundance of MGEs than the farm effluent metagenomes. The hospital effluent 
metagenomes also had a greater abundance of ARGs than all but one of the farm effluent 
metagenomes; DF:M:5 had a greater abundance of ARGs than the hospital effluent samples 
AH:M:5 and AH:M:6. In terms of species number, all hospital effluent metagenomes had a 
greater number of bacterial species present than in the farm effluent metagenomes. The 
greatest number of species was observed in the AH:M:6 hospital effluent metagenome, 
which also featured the lowest number of ARGs and MGEs. 
 
  65 
 
Figure 4.8 A. Graph showing the abundance of ARGs and MGEs, in relation to total species number, for 
hospital effluent over time. B. Graph showing the abundance of ARGs and MGEs, in relation to total 
species number, for farm effluent over time. 
  66 
4.4.1. Abundance of ARGs over time 
Finally, to investigate the abundance of ARGs over time, all ARGs identified in the effluent 
samples were grouped by antimicrobial class and plotted by metagenome (Figure 4.9). The 
beta lactam class contained the highest abundance of ARGs for both hospital and farm 
effluent metagenomes; the beta lactam class was found to contain the highest abundance of 
ARGs in every farm effluent metagenome. Out of the farm effluent metagenomes, one 
sample (DF:M:4) had the highest abundance of ARGs across all classes of antimicrobials. 
For both hospital and farm effluent metagenomes, no trend was observed in ARG expression 
across antimicrobial classes over sampling time. 
 
  67 
 
Figure 4.9 A. Graph showing the abundance of ARGs, grouped by antimicrobial class, in hospital 
effluent metagenomes. B. Graph showing the abundance of ARGs, grouped by antimicrobial class, in 
farm effluent metagenomes. 
  68 
4.5. Discussion 
This chapter has confirmed that ARGs are expressed in effluents that are entering the 
environment. In addition, strong correlations between transcript abundance and antimicrobial 
usage for several classes of antimicrobial have been observed. Finally, pathogenic species of 
bacteria, as well as MGEs and ARGs, have been found in all effluent samples, taken over a 
period of several months.  
The initial analysis of all the metagenomes in this chapter revealed that the hospital effluent 
samples had a far greater abundance of ARGs than both the farm effluent samples (10-fold 
more ARG reads) and the background samples of river source water (70-fold more ARG 
reads) (Table 4.1). There are several factors that may contribute to this difference. Firstly, the 
hospital has a greater number of individuals contributing to the hospital effluent (1000 patient 
beds in the main hospital, in addition to staff members and visitors) than the individuals 
contributing to the farm effluent (an approximate herd size of 200). This difference in 
population size may result in a greater number and diversity of bacteria in the hospital 
effluent (as reflected in the Metaphlan species analysis, Figure 4.6), leading to increased 
competition within the effluent biome and thus, a greater number of ARGs. Secondly, the 
monthly antimicrobial usage of the hospital is much greater than the farm (usage data not 
shown for farm), as reflected by comparative antimicrobial residues detected in the farm and 
hospital effluents (Table 4.5). It is likely that proximate antimicrobial usage and the presence 
of antimicrobial residues may exert a selective force on the ARGs in the effluents, and that 
the variation in this selective force may explain the differences in ARG abundance exhibited 
in the two effluent types. 
In terms of the expression of ARG transcripts in relation to the abundance of ARGs, it was 
found that the hospital effluent had a far stronger positive correlation between ARG transcript 
abundance and ARG abundance than was observed for the farm effluent (Figure 4.1). On 
closer inspection, it was noted that certain classes of ARGs might be mainly responsible for 
the overall correlation observed between ARG and ARG transcripts (Table 4.3, Table 4.4). 
 For the hospital effluent, only one class (sulfonamide ARGs) was found to have a correlation 
between ARGs and ARG transcripts that was deemed to be insignificant (at significance level 
0.05), all other classes of ARGs were found to have a significant positive correlation. 
However for the farm effluent, only one class (tetracycline ARGs) was found to have a 
significant correlation between ARGs and ARG transcripts (at significance level 0.05). The 
fact that a significant, strong correlation between ARG transcript and ARG abundance was 
  69 
found for the majority of ARG classes in the hospital effluent but not for the farm may be due 
to the far greater abundance of ARGs in the hospital effluent than the farm effluent (10-fold 
greater) that was discussed earlier. Another explanation may be the respective presence of 
selection pressures in the effluents, causing ARGs to be expressed at a greater level than in 
the presence of no selection pressures (e.g. the ARG abundance observed in the farm 
effluent may have predominately been related to background expression, while the hospital 
effluent data more truly represents the response to environmental pressures). The presence 
of possible selection pressures was confirmed by LCMS, which showed eight antimicrobial 
compounds (spanning six classes of antimicrobials) to be present in the combined hospital 
effluent samples and only one antimicrobial compound to be present in the combined farm 
effluent (Table 4.5). As the antimicrobial compounds tested for were selected based on 
hospital usage data, not all of these compounds will have been used on the farm and this 
may account for the fact that fewer compounds were detected in the farm effluent. 
Additionally, the half-life or the mode of excretion of each of these antimicrobial compounds 
may have impacted their detection (e.g. bile excretion may have a different impact on the 
amount of compound that can be detected, compared to urinary excretion).  
It should be noted that the LCMS experiments failed for aminoglycosides and trimethoprims, 
due to the matrix suppression and the inability to optimise the mass spectrometry for these 
antimicrobial standards. Another caveat that should be noted for the LCMS results is that the 
effluent samples were very impure water samples; consequently it was difficult to detect 
antimicrobial residues due to the solid matter causing very poor chromatography results. The 
difficulties experienced whilst using LCMS to detect antimicrobial residues in effluent samples 
suggest that it may not be the most appropriate methodology to use in future studies 
requiring antimicrobial residue testing. 
For the hospital effluent, it was hypothesised that the antimicrobial usage of the hospital for 
the month of sample collection would serve as a positive ARG selection pressure and impact 
the expression of ARGs in the effluent. This hypothesis was found to be true; overall a 
positive correlation was observed between ARG transcript abundance and antimicrobial 
usage in all hospital effluent samples, which was statistically significant (Figure 4.3). 
However, this overall correlation was weak and may have been largely due to a strong 
correlation for tetracycline ARG transcripts and tetracycline usage (in addition to correlations 
for rifamycins, glycopeptides and phenicols). 
In light of finding that both ARGs and antimicrobial usage could be correlated to ARG 
transcripts in the hospital effluent, a multiple linear regression, ANOVA and principal 
  70 
components analysis was performed to determine the relative contribution of these two 
variables to ARG transcript abundance in the hospital data. A strong positive correlation 
between the three variables was found, for which ARG abundance and antimicrobial usage 
could account for 79% of the variation in ARG expression. Due to the fact that not all 
variation could be accounted for by ARG abundance and antimicrobial usage, as well as the 
large variation found within certain classes of antimicrobial (Figure 4.5), there are likely to be 
one or more variables that are unaccounted for in order to explain the expression of ARGs in 
effluents that are entering the environment (e.g. genetic context factors such as ARG 
promoters, or drug factors such as the half life of antimicrobial compounds, etc.). Another 
factor that should be taken into account when discussing the correlations found in this study 
is the small number of data points used when looking at expression according to 
antimicrobial class. Due to two metatranscriptomes failing sequencing for hospital and farm 
effluent samples, only four metatranscriptomes were available for analysis for each sample 
site. Four ARG transcript data points per antimicrobial class may be insufficient to draw 
meaningful correlations, and additional work should be carried out to augment these 
datasets. 
In addition to the effect of selection pressures (antimicrobial residues and antimicrobial 
usage) and ARG abundance, the function of the protein that the ARG encodes may play a 
role in ARG expression. For example, an ARG that encodes a tetracycline efflux pump may 
be expressed more readily than a non-efflux ARG, particularly if the efflux pump conferred an 
advantage other than just tetracycline resistance (for instance, multidrug efflux pumps, which 
may be particularly true in an environmental context) (Van Bambeke et al., 2003). Another 
case where gene type or protein function could play a role in ARG expression is in the case 
of linked or co-inherited genes, such as the sulfonamide resistance gene, sul1, which is 
typically co-inherited as a core component of class 1 integrons and thus might be expressed 
for reasons other than antimicrobial selection pressures (Mazel, 2006). With this in mind, 
future work to determine factors that contribute to ARG expression in the environment need 
to take into account ARG mechanism (and possibly gene mobility or genome location), rather 
than just antimicrobial resistance class. It may be necessary to weight each ARG with values 
that account for how readily expressed they are, the level of gene expression required to 
evoke a phenotypic response, or the specificity of the resistance mechanism etc. 
It should be noted that an in silico analysis limitation that may have impacted the findings in 
this chapter is the inability of the bioinformatics methodology to detect antimicrobial 
resistances that are caused by single nucleotide polymorphisms (SNPs), such as gyrase-
  71 
based quinolone resistance (Jacoby, 2005). This could explain instances where a low 
abundance of quinolone ARGs were detected yet a high level of quinolone usage, in addition 
to the presence of quinolone drug compounds, were recorded. In addition to the inability to 
detect SNP-based resistances, the methodology may also be missing novel ARGs that the 
ARG database cannot account for. 
A possible caveat of the wet-laboratory component of the metatranscriptomic methodology 
used in this study (that may contribute to differences between ARG and corresponding ARG 
transcript abundance) is the nucleic acid extraction approach. This approach involved 
dividing the isolated cells from each effluent sample in two and performing separate 
extractions, resulting in the RNA and DNA from each sample not being isolated from the 
same individual cells but the same bacterial community as a whole. Consequently an ARG 
detected in a metagenome could have been expressed by the bacteria of that metagenome 
but the same ARG may not have been expressed by the bacteria in the half of the sample 
that constitute the corresponding metatranscriptome. To circumvent this caveat, an 
alternative option would be to use the acidic phenol phase (containing DNA) of the RNA 
extraction to extract DNA, thus allowing RNA and DNA to be extracted from the same group 
of cells sampled. However, this single phenol-chloroform extraction approach would require 
considerable calibration and refinement in order to successfully isolate high-quality RNA and 
DNA suitable for high-throughput sequencing. 
The presence of pathogenic species of bacteria, as well as the number of MGEs and ARGs, 
was determined in effluents over time. A total of 68 different species of pathogenic bacteria 
were identified across all effluent samples and 36 pathogen species were present in both 
hospital and farm effluents. Not only does this illustrate that both effluents have clinical 
significance in terms of the pathogenic species of bacteria that they are contributing to the 
environment, but it also offers possible mechanisms by which ARGs could be maintained and 
disseminated in the environment. One such mechanism could be via endospore-forming 
bacteria that could harbour ARGs and protect them from degradation by the formation of 
stable endospores (Galperin, 2013). As 47% of the shared pathogen species identified in the 
effluents belonged to the Firmicute phylum, known for their ability to form endospores, the 
hypothesis that endospore-forming bacteria can harbour and disseminate ARGs appears to 
be a valid one and is explored in the next chapter (Chapter 5). 
No discernible trend was observed between pathogen species number and total species 
number over the sampling time points, neither was there any trend observed in ARG or MGE 
abundance over time. Despite no obvious trend, there was a large amount of temporal 
  72 
variation in pathogen species number and total species number, as well as in ARG and MGE 
abundance. This may be due to the nature of random sampling, or it could be a result of 
seasonal variations in pathogen abundance (e.g. disease outbreaks) and environmental 
variables (e.g. rainfall) across the sampling dates. It is possible that the species diversity 
across the metagenomes may affect the correlations that have been reported between ARG 
transcript, ARG abundance and antimicrobial usage. An important direction for future study 
would be to investigate the effect of such seasonal variation on the abundance of ARGs, in 
addition to pathogens and MGEs, in order to determine possible temporal points that may 
lead to increased release of ARGs. This information would greatly contribute to models of 
ARG dissemination that are beginning to be utilised in surveillance and risk management 
programmes. 
 
4.6. Conclusion 
The work of this chapter has shown that both ARGs and ARG transcripts are present in 
effluents entering the aquatic environment. A strong positive correlation was found between 
the abundance of ARGs and ARG transcripts across the samples. A correlation was also 
observed between ARG transcript abundance and antimicrobial usage in hospital effluent. 
Although correlation does not verify causation, this study demonstrates that ARGs are being 
expressed in effluents and that antimicrobial usage may influence the fate of ARGs in 
effluents entering the environment, thus suggesting a microbial response to the 
anthropogenic release of antimicrobials into the aquatic environment. In terms of monitoring 
ARG dissemination, selection pressures that are known to illicit a microbial response and 
effect ARG load are important factors that should be incorporated into environmental 
frameworks. 
 
  73 
Chapter 5. Bacterial endospores present in the environment 
harbour antimicrobial resistance genes 
5.1. Preface 
The metagenomic analyses documented in previous chapters have identified pathogenic 
bacteria and antimicrobial resistance genes (ARGs) in environmental effluents. I have shown 
that the genes of these pathogens and ARGs are expressed in the environment, apparently 
in response to the presence of antimicrobial selection pressures. However I have yet to 
definitely link the presence of ARGs with pathogenic species, and to demonstrate possible 
mechanisms of persistence of these ARGs in the environment. Validating the finding that 
ARGs are persisting in the environment would give significant insight into the fate of ARGs in 
the environment, particularly in the context of pathogenic bacteria. 
Taxonomic analysis of the environmental metagenomes in previous chapters has shown the 
phylum Firmicutes, in particular the genus Clostridia (several of which are known pathogens), 
to be consistently present across the sample sites. Many Firmicutes produce endospores, 
which are able to survive extreme conditions. In this way, Firmicutes could facilitate the 
persistence of ARGs in the environment. In the present study I explore this possibility through 
the isolation and genomic analysis of endospore-forming bacteria from environmental 
samples.  
 
5.2. Introduction 
Thus far, metagenomics has been used to identify ARGs, mobile genetic elements (MGEs) 
and pathogenic bacteria within effluents that are entering a river catchment. Metagenomics 
has allowed the relative abundance of ARGs to be estimated for each sample and has been 
used to correlate ARG abundance to antimicrobial usage data. However, there are several 
challenges to microbial community analysis that metagenomics cannot address when used in 
isolation. One such challenge is the inability to assemble and model the whole genome of 
individual bacteria, which is often a result of insufficient metagenomic sequencing depth 
(Kuczynski et al., 2012, Zengler and Palsson, 2012). This means that potential functions 
identified within microbial communities, such as antimicrobial resistance, cannot be linked 
with phylogeny using a solely metagenomic approach (Clingenpeel et al., 2015).  
  74 
The advent of multi-omic strategies for microbial community profiling has allowed researchers 
to apply multiple sequencing techniques to single environmental samples in order to address 
key questions, such as the use of metagenomics, metatranscriptomics and single-cell 
sequencing to determine microbial response to an oil spill (Mason et al., 2012). These 
approaches can be augmented by the use of whole genome sequencing (WGS) data from 
isolates of specific environmental bacteria and used in conjunction with metagenomics to 
offer insight into the dynamics of ARG carriage and the relationship to ARG diversity in the 
wider microbial community. 
In conjunction with identifying bacterial isolates that carry ARGs, culture of environmental 
bacteria could be used to target specific pathogens, thus testing whether clinically important 
bacteria are those harbouring ARGs and on a larger scale, observing any trends or 
relationships among isolates from different geographically-proximate sample sites (i.e. within 
the river catchment in this study). Finally, to address whether ARGs are persisting in the 
environment (as opposed to metagenomic approaches capturing the transient occurrence of 
ARGs), candidate bacterial species that are able to preserve their genetic content whilst 
withstanding the changes in environment often associated with effluent biomes should be 
targeted. 
 
5.2.1. Endospore-forming bacteria 
Certain species of bacteria that belong to the phylum Firmicutes (mainly Gram-positive, low 
GC-content bacteria) feature a set of core sporulation proteins, allowing them to form highly 
resistant endospores (Galperin, 2013). Sporulation enables the bacteria to survive adverse 
environmental conditions by transitioning from a vegetative form, which allows maintenance 
of metabolic and reproductive activities, to a dormant endospore form that remains 
metabolically inactive (whilst preserving genetic content) until the environment becomes 
more favourable (Talukdar et al., 2015). Endospores have been found to withstand extreme 
environmental conditions, such as those found in the Arctic permafrost (Suetin et al., 2009) 
and outer space (Horneck et al., 2012), and endospore-forming bacteria are known to inhabit 
most aquatic and terrestrial environments (Galperin, 2013). Thus endospore-forming bacteria 
represent an important bacterial group to target in order to confirm the persistence of ARGs 
in effluents. 
Several species of endospore-forming bacteria, in particular of the Clostridium genus, are 
pathogenic and are known to acquire ARGs (Huang et al., 2009, Kouassi et al., 2014). 
  75 
Members of the Clostridium genus have been identified in metagenomes from wastewater 
treatment plant (WWTP), dairy farm and hospital effluent sample sites, as described in 
Chapters 2, 3 and 4. Therefore, the Clostridium genus was chosen for investigating the 
persistence of ARGs in pathogenic species present in effluents. 
 
5.2.1.1. Clostridium difficile 
Clostridium difficile (C. difficile) is an obligate anaerobe that is a major pathogen and 
frequently implicated in hospital-acquired infections (Rupnik et al., 2009). C. difficile infection 
(CDI) requires the disruption of the normal gut flora (which often results from antimicrobial 
therapy) and occurs once the bacterium is acquired exogenously. CDI results in Clostridium 
difficile-associated disease (CDAD). CDAD ranges in its severity, from diarrhoea, dehydration 
and metabolic changes through to pseudomembranous colitis and haemorrhaging. A number 
of factors, including the virulence of the C. difficile strain and the antimicrobial therapy a 
patient receives, influence the development of CDAD (Johnson and Gerding, 1998, Brouwer 
et al., 2013, Tonna and Welsby, 2005). 
C. difficile (along with other members of the Firmicute phylum) produce highly resistant 
endospores that are capable of withstanding extremes in environmental conditions and 
facilitate disease transmission (Lawley et al., 2009). Asymptomatic carriage of C. difficile 
causes low-level shedding of endospores, however antimicrobial therapy can result in a 
contagious super-shedder state in which the normal gut flora is disrupted, the reduction in 
microbiota diversity leading to an overgrowth of C. difficile and the shedding of high levels of 
endospores (Lawley et al., 2009). Owing to its significance as a major nosocomial pathogen, 
mandatory surveillance has been established in the UK that requires all cases of CDI in 
patients over the age of two years old to be reported to Public Health England (Public Health 
England, 2013). In addition to this, antimicrobial usage and resistance is also monitored as 
part of the England Stewardship of Antimicrobial Utilization and Resistance program (Ashiru-
Oredope and Hopkins, 2013).  
Although considered a nosocomial pathogen, C. difficile can also be ‘community-acquired’, 
i.e. arise in individuals who have not been hospitalised or exposed to antimicrobial therapy 
(Wilcox et al., 2008). Such cases have led to questions being raised as to other transmission 
routes for C. difficile and the associated risk factors. In particular, animals have been 
suggested as possible sources for food-borne transmission of C. difficile (Gould and 
Limbago, 2010) as C. difficile is also a pathogen of several domestic and food animals, 
  76 
including neonatal pigs, horses, cattle and companion animals (Hammitt et al., 2008). 
Recently several studies have found cases of CDI in animals and humans that share the 
same C. difficile strains, suggesting that the C. difficile may have originated from a common 
source, or have been involved in zoonotic transmission (Jhung et al., 2008, Debast et al., 
2009). C. difficile can also be isolated from environmental sources, including soil, river and 
sea water samples (Saif and Brazier, 1996). 
C. difficile is known to harbour antimicrobial resistance genes and strains have been found to 
be phenotypically resistant to beta-lactams, lincosamides, quinolones and tetracyclines 
(Keessen et al., 2013). However the role of antimicrobial resistance in the epidemiology of 
CDI is complex, due in part to antimicrobial treatment resulting in favourable conditions for 
CDI development and also, the ARGs harboured by C. difficile often confer resistance to 
antimicrobials that are not used to treat CDI (Coia, 2009). The high number of MGEs 
associated with C. difficile means that C. difficile genome has the potential to be highly plastic 
(Brouwer et al., 2011). It is hoped that surveillance schemes, antimicrobial resistance data 
and WGS can be used to offer insight into the global spread of C. difficile (He et al., 2013). 
 
5.2.1.2. Clostridium perfringens 
Clostridium perfringens (C. perfringens) is an obligate anaerobe that is found in many 
environments, including faeces, sewage, soils and food (Li et al., 2013), and is the causative 
agent of several diseases such as food poisoning and gas gangrene (Rood and Cole, 1991). 
Strains of C. perfringens are known to vary in their virulence and phenotypic characteristics 
due in part to the incorporation of MGEs, containing genes encoding toxins, sporulation 
factors and other secreted enzymes, into the genome (Myers et al., 2006). Strains of C. 
perfringens have also been found to carry to conjugative plasmids containing ARGs and 
toxin-encoding genes (Bannam et al., 2011).  
 
5.2.2. Chapter hypothesis 
The Firmicute phylum, in particular the genus Clostridium, contains bacterial species that are 
clinically relevant pathogens in both human and veterinary medicine. These species are 
readily transmissible and can exhibit resistance to antimicrobials. Endospore formation and 
faecal shedding aids their widespread dissemination in the environment, and the multitude of 
effluents that are entering the aquatic environment are likely to be key sources in the 
  77 
dispersal of endospore-forming bacteria. Therefore endospore-forming bacteria, as 
exemplified by Clostridium sp., were hypothesised to be contributors of ARGs to the 
environmental resistome and responsible for their persistence. The work in this chapter 
explored the link between endospore-forming, pathogenic bacteria and antimicrobial 
resistance in the environment by reactivating endospores from environmental samples and 
subjecting them to WGS in order to establish their phylogenetic and genomic relationships 
with respect to ARGs. 
 
5.3. Materials and methods 
5.3.1. Sample collection 
Environmental samples of effluents and faeces were collected over a four-month time period 
from all the sampling sites used in the previous chapters (hospital effluent, WWTP influent 
and effluent, dairy farm). In addition, effluent from a pig farm (latitude: 51.736053, longitude: 
0.408824) was sampled on the 15.07.2014 and samples were also taken from the River Cam 
on 01.10.2014. Samples were collected in a 50 mL sterile polypropylene containers and 
transported at 4°C to the laboratory. Metagenomic sampling from the pig farm was also 
undertaken according to methods described in Chapter 2.  
 
5.3.2. Isolate culture 
Environmental samples were subjected to ethanol (70%) shock for 24 hours to kill vegetative 
microorganisms before being spun down and re-suspended in sterile Phosphate Buffered 
Saline (PBS) solution. An aliquot of 1 mL was added to pre-reduced Brain-Heart Infusion 
(BHI) broth (containing 0.025 g sodium taurocholate hydrate) and grown for 24 hours at 37°C 
under anaerobic conditions to reactivate endospores. Aliquots of the 24-hour culture were 
serially diluted (by factors of 100, 101, and 102) in sterile PBS and streaked onto pre-reduced 
Brazier’s cefoxitin cycloserine egg yolk (CCEY) agar plates (containing 0.05 g sodium 
taurocholate hydrate) and grown for 48 hours at 37oC under anaerobic conditions. 
 
  78 
5.3.3. DNA extraction and 16S rRNA sequencing 
Isolates of interest were identified using a similar 16S rRNA sequencing methodology as 
described by Lawley et al. (Lawley et al., 2012). Briefly, colonies that had distinct morphology 
were isolated, sub-cultured on pre-reduced CCEY agar plates and grown overnight at 37°C 
under anaerobic conditions. From each plate, a single colony was picked, added to 1 mL of 
PBS, subjected to bead beating for 30 seconds and centrifuged (13000 rpm, 10 minutes) to 
obtain the supernatant. The supernatant (containing isolate DNA) was used as template in 
the 16S PCR reaction with broad range primers. The 16S product was sequenced using 
capillary sequencing (Sanger Institute, UK). To identify bacterial species of each isolate, the 
16S rRNA sequence was compared against the Ribosomal Database Project (RDP) and 
GenBank databases. 
 
5.3.4. DNA extraction for whole genome sequencing 
Isolates for WGS sequencing were picked and grown overnight in pre-reduced BHI broth at 
37°C under anaerobic conditions. For each isolate, 500 µL of culture was added to pre-
reduced glycerol and stored at -80°C. The remaining culture was spun down for 10 minutes 
at 6500 RPM. The resulting pellet was washed with 10 mL of PBS solution, re-spun at 6500 
RPM and incubated at -20°C for 24 hours. Genomic DNA was extracted from each pellet by 
cell lysis, phenol-chloroform extraction and ethanol precipitation. Briefly, each pellet was re-
suspended in 2 mL of sucrose (25%) in TE buffer (10 mM Tris pH8 and 1 mM EDTA pH 8) 
and incubated at 37°C for 1 hour with 50 µL Lysozyme (100mg/mL; Sigma-Aldrich). To 
complete lysis, 100 µL of Proteinase K (18 mg/mL; Sigma-Aldrich), 30 µL RNase A (10 
mg/µL; Invitrogen), 400 µL EDTA (0.5M) and 250 µL 10% Sarkosyl solution were added to 
the cell suspension and left on ice for 2 hours before being incubated at 50°C for 12 hours. 
Genomic DNA was extracted by phenol:chloroform:isoamyl alcohol (Sigma-Aldrich) washes 
and chloroform:isoamyl alcohol (Sigma-Aldrich) washes. Genomic DNA was then precipitated 
using 100% ethanol and purified with a wash of 70% ethanol. The purity of the DNA was 
assessed and quantified using Quibit Fluorometric Quantitation (Life Technologies). 
 
  79 
5.3.5. Ribotyping 
Ribotyping was performed on potential C. difficile isolates by Wilco Knetsch at the Leiden 
University Medical Center, Netherlands. The methodology used was as described in 
(Knetsch et al., 2012). 
 
5.3.6. Metagenome sequencing 
At the same time as isolate sample collection, a sample of the pig farm effluent from the 
lagoon was collected in a 10 L sterile polypropylene container, transported at 4°C to the 
laboratory and processed using the same methodology described in 2.3.3.2. A total of 2 µg of 
DNA was used to generate an Illumina sequencing library (100 bp, paired-end) that was 
sequenced using an Illumina HiSeq2500 (Exeter Sequencing Service, UK). A full description 
of the metagenomic samples used in this chapter is available in Appendix 3. 
 
5.3.7. Construction of high-quality draft genomes  
Illumina sequencing libraries (150 bp, paired-end) were generated for all isolates with 
sufficient DNA for WGS and sequenced (Illumina MiSeq, San Diego, CA, USA) according to 
in-house protocols at the Wellcome Trust Sanger Institute (Quail et al., 2012). Assembly of 
each isolate was performed using the VelvetOptimiser script (Zerbino and Birney, 2008). 
Provisional classification of isolates was performed using Kraken (Wood and Salzberg, 
2014); isolates that were unclassified by Kraken were classified using Metaphlan2 (Segata et 
al., 2012) and then aligned to the top species hit using BLAST (Altschul et al., 1990) to 
confirm classification. Isolates were then separated into species groups (difficile, perfringens 
etc.) for downstream analysis, which was done according to classification (as above) and 
features of genome assembly (total draft genome size, number of contiguous sequences 
(contigs)). Comparative genomics of assembled genome data was facilitated by comparison 
with reference genomes of C. difficile and C. perfringens (Appendix 4). Draft genomes from 
isolates in this study were annotated using Prokka (Seemann, 2014), which was also used to 
re-annotate reference genomes including in core genome analyses; annotations were used 
to explore genome features, as well as define core and accessory genomes. 
 
  80 
5.3.8. Bioinformatic analysis 
For phylogenetic analysis of isolates, a multiple sequence alignment was generated by 
mapping isolate FASTQ data (and in-silico generated FASTQ files of reference isolates) to a 
suitable reference genome for each species (C. difficile 630 and C. perfringens ATCC13124) 
using SMALT (Ponstingl, 2014). For core genome analysis of each species group, core 
genes (genes present once in every isolate) were identified using the Roary pipeline (Page et 
al., 2015), aligned using SMALT (Ponstingl, 2014). For each alignment, the variable sites 
were used to construct a maximum likelihood tree using RAxML 7.8.6 (Stamatakis, 2006). 
Phylogenetic trees were visualised using Figtree (Rambaut, 2007) and Dendroscope (Huson 
and Scornavacca, 2012). 
Antimicrobial resistance gene annotation was performed for each isolate by comparing 
predicted protein coding sequences to the Antibiotic Resistance Database (ARDB) using the 
annotation tool ARDBanno.pl (Liu and Pop, 2009). 
For comparative genomics work, ABACAS (Algorithm-Based Automatic Contiguation of 
Assembled Sequences), Artemis, the Artemis Comparison Tool and DNAPlotter were used to 
identify coordinates generate image files (Carver et al., 2005). Where re-arrangements (i.e. 
insertions or deletions) were identified, de-novo assemblies where checked by mapping 
reads to a concatenated construct of the contiguous sequences (self-mapping) (BWA-mem) 
(Li, 2013). 
Taxonomic profiling of the metagenomic reads was performed using Metaphlan2 (Segata et 
al., 2012) and the ARG content of the metagenome was assessed using SEAR (Rowe et al., 
2015) with default settings. Metagenomes were assembled using the VelvetOptimiser script 
and Metavelvet (Zerbino and Birney, 2008) and compared to isolate contigs using ABACAS. 
 
5.4. Results 
5.4.1. Isolation and identification of endospore-forming bacteria 
Over a period of 4 months a total of 100 samples were collected from various effluents in the 
environment, including hospital, WWTP, pig farm and dairy farm effluents (Appendix 3). Each 
sample was enriched for endospores and plated in a three dilution series to give 300 plates 
of environmental endospore cultures. Colony numbers varied (from one colony to entire 
lawns) on each plate. A total of 113 colonies were successfully sub-cultured and subjected to 
  81 
16S sequencing. Based on sample location, colony morphology and 16S annotation, 68 of 
the 113 colonies were selected for isolation (aiming to select diverse isolates that included 
several different species of Clostridia), DNA extraction and WGS. Of the 68 isolates of 
environmental, endospore-forming bacteria prepared for WGS, a total of 59 isolates were 
successfully sequenced (Table 5.1).  
  
82 
Table 5.1 All endospore-forming bacteria successfully isolated from environmental samples and processed for whole genome sequencing. 
Sampling DNA extraction  Assembly statistics Kraken annotations 
Is
ol
at
e 
nu
m
be
r 
Lo
ca
tio
n 
D
at
e 
S
am
pl
e 
ty
pe
 
16
S
 R
N
A 
se
qu
en
ci
ng
 
C
on
c.
 D
N
A 
(n
g/
µL
) 
To
ta
l W
G
S
 
yi
el
d 
(k
b)
 
To
ta
l l
en
gt
h 
N
o.
 c
on
tig
s 
A
vg
. c
on
tig
 
le
ng
th
 (b
p)
 
K
ra
ke
n 
to
p 
hi
t 
un
cl
as
si
fie
d 
C.
 d
iffi
cil
e 
C.
 p
er
fri
ng
en
s 
C.
 b
ot
uli
nu
m
 
1 Pig farm 15.07.2014 Combined farrow house effluent Clostridium perfringens 60 396967 7859847 979 8028 Unclassified 80.66 0.04 17.45 0.01 
2 Pig farm 15.07.2014 Combined farrow house effluent Clostridium perfringens >120 384744 6919008 296 23375 C. perfringens 10.73 24.83 63.89 0.03 
3 Pig farm 15.07.2014 Piglet crate effluent Clostridium difficile 91.2 513766 9196469 214 42974 C. difficile 45.75 52.85 0.03 0.01 
4 Pig farm 15.07.2014 Piglet crate effluent Clostridium difficile 118 583950 3828920 56 68374 C. difficile 15.69 83.86 0.02 0.01 
5 Pig farm 15.07.2014 Combined farrow house effluent Clostridium difficile 120 613668 3826452 57 67131 C. difficile 14.81 84.69 0.08 0.01 
6 Pig farm 15.07.2014 Piglet crate effluent Clostridium difficile >120 553922 4054739 57 71136 C. difficile 5.23 93.57 0.02 0 
7 Pig farm 15.07.2014 Piglet crate effluent Clostridium perfringens 120 622796 4760721 2576 1848 C. difficile 28.4 70.16 0.74 0.01 
8 Pig farm 15.07.2014 Sow faeces Clostridium difficile 68.8 667938 3830384 45 85120 C. difficile 15.33 84.16 0.02 0.01 
9 Pig farm 15.07.2014 Combined farrow house effluent Clostridium perfringens 112 477872 6806631 1474 4618 C. perfringens 10.75 13.44 75.13 0.03 
10 Pig farm 15.07.2014 Fertiliser/lagoon effluent* Clostridium colicanis >120 453049 3394496 44 77148 Unclassified 95.17 0.18 0.7 0.65 
12 Pig farm 15.07.2014 Fertiliser/lagoon effluent* Clostridium butyricum 94.6 433858 4534048 71 63860 Unclassified 92.76 0.06 0.18 1.05 
13 Pig farm 15.07.2014 Fertiliser/lagoon effluent* Clostridium colicanis >120 521535 3455264 69 50076 Unclassified 94.46 0.08 0.82 0.71 
14 Pig farm 15.07.2014 Manure heap Clostridium perfringens 112 447642 3394466 19 178656 C. perfringens 8.57 0.03 90.92 0.01 
15 Pig farm 15.07.2014 Manure heap Clostridium perfringens 68.6 457313 3141691 84 37401 C. perfringens 9.96 0.03 89.38 0.04 
16 Pig farm 15.07.2014 Combined weaner shed effluent Clostridium butyricum >120 528191 4453401 76 58597 Unclassified 92.55 0.04 0.16 1.04 
17 Pig farm 15.07.2014 Combined weaner shed effluent Clostridium difficile >120 431443 3834726 56 68477 C. difficile 15.01 84.54 0.05 0.01 
18 Pig farm 15.07.2014 Soil (ammended with effluent) Clostridium perfringens 106 443344 4093716 1820 2249 C. perfringens 9.58 0.02 89.83 0.04 
19 Pig farm 15.07.2014 Soil (ammended with effluent) Clostridium perfringens 68 398784 3101644 112 27693 C. perfringens 8.84 0.04 90.38 0.06 
20 Pig farm 15.07.2014 Soil (ammended with effluent) Clostridium perfringens 87 353803 3107330 93 33412 C. perfringens 9.94 0.03 89.43 0.02 
21 Pig farm 15.07.2014 Soil (ammended with effluent) Clostridium perfringens >120 494861 3116792 93 33514 C. perfringens 10.79 0.04 88.53 0.02 
22 WWTP 04.08.2014 WWTP influent Clostridium butyricum 85 391122 4516539 59 76552 Unclassified 92.62 0.04 0.18 1.01 
  
83 
23 WWTP 04.08.2014 WWTP influent Clostridium perfringens 104 476564 3307749 25 132310 C. perfringens 10.97 0.02 88.54 0.01 
24 WWTP 04.08.2014 WWTP influent Clostridium difficile 114 502071 4339582 43 100921 C. difficile 9.11 88.83 0.02 0 
25 WWTP 04.08.2014 WWTP influent Clostridium perfringens 81.6 527463 3266305 54 60487 C. perfringens 10.53 0.04 88.64 0.03 
26 WWTP 04.08.2014 WWTP effluent Clostridium difficile 83.4 534890 4303354 55 78243 C. difficile 4.66 91.69 0.03 0 
27 WWTP 04.08.2014 WWTP effluent Clostridium difficile 59.4 457216 4296951 42 102308 C. difficile 4.64 91.62 0.02 0 
28 Hospital 04.08.2014 Hospital wastewater effluent Clostridium difficile >120 498694 4323406 60 72057 C. difficile 6.23 89.35 0.03 0.01 
29 Hospital 04.08.2014 Hospital wastewater effluent Clostridium difficile 79 413581 4144436 40 103611 C. difficile 4.94 93.56 0.03 0 
30 Hospital 04.08.2014 Hospital wastewater effluent Clostridium difficile 78.6 384058 4323683 58 74546 C. difficile 6.25 89.57 0.02 0 
31 Hospital 04.08.2014 Hospital wastewater effluent Clostridium difficile 102 475056 4064553 39 104219 C. difficile 4.95 94.52 0.02 0 
32 Hospital 04.08.2014 Hospital wastewater effluent Clostridium difficile 73.6 524438 4068536 41 99233 C. difficile 5.15 94.23 0.02 0 
33 Hospital 04.08.2014 Hospital wastewater effluent Clostridium difficile 58 550368 4141364 40 103534 C. difficile 3.64 92.62 0.04 0 
34 Hospital 04.08.2014 Hospital wastewater effluent Clostridium difficile 102 429473 4066105 33 123215 C. difficile 5.13 94.3 0.02 0 
35 Hospital 04.08.2014 Hospital wastewater effluent Clostridium difficile 89.2 426528 4574826 297 15403 C. difficile 3.66 95.1 0.03 0 
36 Dairy farm 06.08.2014 Combined milking shed effluent Terrisporobacter glycolicus 97.6 419652 4112310 54 76154 Unclassified 97.11 1.15 0.1 0.07 
37 Dairy farm 06.08.2014 Combined milking shed effluent Terrisporobacter glycolicus 88.2 438254 4106488 46 89271 Unclassified 96.82 1.29 0.1 0.07 
38 Dairy farm 06.08.2014 Calf shed effluent Terrisporobacter glycolicus 49.8 421671 4132142 48 86086 Unclassified 96.59 1.38 0.12 0.08 
39 Dairy farm 06.08.2014 Calf shed effluent Terrisporobacter glycolicus 80.8 498836 4129645 43 96038 Unclassified 97.2 1.14 0.09 0.07 
40 Dairy farm 06.08.2014 Calf shed effluent Clostridium perfringens 116 556516 3214184 14 229585 C. perfringens 5.66 0.03 93.89 0.01 
41 Dairy farm 06.08.2014 Calf shed effluent Clostridium perfringens 98 627832 3215299 14 229664 C. perfringens 5.8 0.03 93.76 0.01 
42 Dairy farm 06.08.2014 Calf shed effluent Clostridium perfringens 102 539625 3972568 1075 3695 C. perfringens 23.4 0.67 74.6 0.01 
43 Hospital 15.09.2014 Hospital wastewater effluent Clostridium difficile 83.8 466178 4098800 37 110778 C. difficile 2.79 95.05 0.02 0 
44 Hospital 15.09.2014 Hospital wastewater effluent Clostridium difficile 29.8 434080 4102890 36 113969 C. difficile 2.68 95.39 0.02 0 
45 Hospital 15.09.2014 Hospital wastewater effluent Clostridium difficile 30 414952 4100726 35 117164 C. difficile 2.52 95.56 0.02 0 
46 Hospital 29.09.2014 Hospital wastewater effluent Clostridium perfringens 20.4 466087 3272482 50 65450 C. perfringens 8.75 0.03 90.65 0 
47 Hospital 29.09.2014 Hospital wastewater effluent Clostridium perfringens 28.4 432151 3272685 32 102271 C. perfringens 8.46 0.03 91 0.01 
48 Hospital 27.10.2014 Hospital wastewater effluent Clostridium difficile 20.6 460378 6424279 47 136687 Unclassified 94.51 0.14 0.07 0 
49 Hospital 27.10.2014 Hospital wastewater effluent Streptococcus lutetiensis 86.4 503362 4537607 30 151254 Unclassified 96.15 0.31 0.06 0 
50 Dairy farm 15.09.2014 Fertiliser/lagoon effluent Clostridium beijerinckii 81.2 534596 4553648 57 79889 Unclassified 93.04 0.07 0.2 0.9 
51 Dairy farm 29.09.2014 Fertiliser/lagoon effluent Clostridium beijerinckii 39.8 495111 4665555 87 53627 Unclassified 93.26 0.09 0.19 1.05 
  
84 
52 River 01.10.2014 River water Clostridium beijerinckii 19.1 388996 6542271 55 118950 Unclassified 96.63 0.21 0.05 0 
53 River 01.10.2014 River water Clostridium difficile 77.4 497315 3737181 30 124573 C. botulinum 11.41 0.03 0.02 87.34 
54 River 01.10.2014 River water Clostridium beijerinckii 80.6 484115 3572217 22 162374 C. botulinum 8.12 0.03 0.02 90.72 
55 Dairy farm 29.09.2014 Fertiliser/lagoon effluent Clostridium perfringens 59 489987 3524380 17 207316 Unclassified 76.6 0.05 22.01 0.23 
56 Hospital 29.09.2014 Hospital wastewater effluent Clostridium perfringens 15.7 506820 3276555 30 109219 C. perfringens 9.69 0.14 89.69 0.01 
57 Hospital 29.09.2014 Hospital wastewater effluent Clostridium perfringens >120 582916 3280735 25 131229 C. perfringens 8.52 0.03 91.01 0.01 
58 Hospital 29.09.2014 Hospital wastewater effluent Clostridium perfringens 16.6 587255 7502072 129 58156 C. perfringens 9.21 21.14 68.79 0 
59 Hospital 29.09.2014 Hospital wastewater effluent Clostridium perfringens >120 523965 3281160 22 149144 C. perfringens 8.8 0.06 90.69 0.01 
* fertiliser/lagoon effluent sample split and used for pig farm metagenome 
  85 
The Kraken annotation of WGS data classified all isolates as being either a Clostridium 
species (43 isolates) or unclassified (16 isolates).  
 
5.4.1. Unclassified environmental isolates 
Subsequent attempts to classify the unclassified isolates using Metaphlan2 resulted in the 
taxonomic assignment of reads to the Firmicutes phylum, (mainly into the 
Peptostreptococcaceae, Erysipelotrichaceae or Clostridium genera). Based on the 
assembled draft genome of each isolate (including size and number of contigs) and the 
annotation of reads matching Metaphlan2 taxonomic marker genes, 15 of the 16 unclassified 
isolates were putatively classified as Firmicutes (Table 5.2). Draft genomes of the isolates 
were compared to the NCBI Refseq database using BLAST; each isolate matched the 
corresponding ‘WGS species classification’ reference sequence with an average of 80% 
coverage (at 99% identity, data not shown), potentially indicating novel bacterial species. 
 
  
86 
Table 5.2 Unclassified environmental isolates and their identified ARGs. 
Isolate Effluent type WGS species classification 16S RNA Assembly statistics Identified ARGs (ARDB) 
  
(Metaphlan2)  
To
ta
l 
Le
ng
th
 
No
. 
Co
nt
igs
 
Av
g.
 
Co
nt
ig 
Le
ng
th
 
an
t6
ia 
ba
cA
 
m
ef
A 
m
ph
b 
te
tO
 
te
tP
A 
te
tP
B 
va
tB
 
#10 Pig farm Clostridium colicanis Clostridium colicanis 3394496 44 77147.64  1    1   
#12 Pig farm Clostridium butyricum Clostridium butyricum 4534048 71 63859.83  1       
#13 Pig farm Clostridium colicanis Clostridium colicanis 3455264 69 50076.29  1    1 1  
#16 Pig farm Clostridium butyricum Clostridium butyricum 4453401 76 58597.38  1 1 1     
#22 WWTP* Clostridium butyricum Clostridium butyricum 4516539 59 76551.51  1       
#36 Dairy farm Peptostreptococcaceae sp. Terrisporobacter glycolicus 4112310 54 76153.89  1    1 1  
#37 Dairy farm Peptostreptococcaceae sp. Terrisporobacter glycolicus 4106488 46 89271.48  1    1 1  
#38 Dairy farm Peptostreptococcaceae sp. Terrisporobacter glycolicus 4132142 48 86086.29  1    1 1  
#39 Dairy farm Peptostreptococcaceae sp. Terrisporobacter glycolicus 4129645 43 96038.26  1    1 1  
#49 Hospital Erysipelotrichaceae bacterium Streptococcus lutetiensis 4537607 30 151253.57 1 1   1    
#50 Dairy farm Clostridium butyricum Clostridium beijerinckii 4553648 57 79888.56  1       
#51 Dairy farm Clostridium butyricum Clostridium beijerinckii 4665555 87 53627.07  1       
#52 River Clostridium bolteae Clostridium beijerinckii 6542271 55 118950.38  1   1    
#53 River Clostridium botulinum Clostridium difficile 3737181 30 124572.7  1      1 
#54 River Clostridium botulinum Clostridium beijerinckii 3572217 22 162373.5  1      1 
* denotes WWTP influent 
  87 
The draft genomes of each potentially novel bacterial species were ordered to the 
corresponding ‘WGS species classification’ reference sequence (as retrieved from the 
Metaphlan2 database), and then the ordered genomes of potential novel bacterial isolates 
were compared among themselves and revealed incomplete, but high identity, synteny in 
some areas (C. butyricum isolates shown as representative in Figure 5.1), potentially 
indicating some evolutionary relationship among novel bacterial isolates. 
 
 
Figure 5.1 Genome synteny (displayed using ACT) of potentially novel bacterial isolates, related to C. 
butyricum. 
 
Analysis of the ARG content of each of these potentially novel bacterial isolates revealed a 
total of 33 ARGs. All Peptostreptococcaceae-like isolates (dairy farm) contained tetracycline 
resistance genes tetPA and tetPB. All Clostridium botulinum-like isolates (river) contained 
streptogramin resistance gene vatB. The Erysipelotrichaceae-like isolate (hospital) was the 
only non-difficile/non-perfringens isolate to contain an aminoglycoside resistance gene 
(ant6ia). 
 
  88 
5.4.1. Clostridium difficile isolates 
Based on the assembled draft genome and the percentage of reads matching taxonomic 
marker genes (Kraken annotations, Table 5.1), a total of 19 isolates had WGS data of 
sufficient quantity and quality to putatively annotate the isolate as C. difficile. 
 
5.4.1.1. Phylogenetic analysis 
To assess the relatedness of the 19 C. difficile isolates obtained from the three different 
environmental effluents (hospital, WWTP and pig farm), a whole genome phylogeny was 
constructed alongside a set of references. The phylogeny showed that the majority of 
environmental C. difficile isolates fell into four distinct clades (Clades A-D, Figure 5.2). All 11 
hospital C. difficile isolates fell in to clades A, B and C. The two isolates from the WWTP 
effluent fell in to clade C, whilst the isolate from the WWTP influent (isolate #24) was most 
closely related to the C. difficile M68 reference. For the pig farm, four of the isolates fell in to 
clade D (clustering with the C. difficile M120 reference) and the remaining isolate (#6) fell 
outside any of the defined clades. Notably clade B had isolates from two collection dates, a 
month and a half apart. Isolate #29 from hospital wastewater effluent (collected 04.08.2014) 
was 3295, 3283 and 3278 SNPs apart from isolates #43, #44 and #45 respectively, all of 
which were collected from hospital wastewater effluent (collected 15.09.2014). 
 
  89 
 
Figure 5.2 Whole genome phylogenetic tree of Clostridium difficile isolates. The phylogenetic tree shows 
the evolutionary relationships among the C. difficile isolates and reference genomes. Colour coding represents 
the origin of the isolates. Bootstraps less than 90 are shown. Clades containing multiple novel samples are 
highlighted by lettered orange lines. The scale bar is in expected number of nucleotide substitutions per site (n= 
188278). 
  90 
To ensure the robustness of the results from the whole genome phylogeny of the isolates, 
and to explore the genetic content of the isolates, a core genome analysis of the C. difficile 
isolates was performed. The core genome was found to consist of 2521 genes (aligned to 
2400468 sites), the total number of accessory genes was 4798 and the average number of 
genes per isolate was 3703 (Figure 5.3). 
 
 
Figure 5.3 Rarefaction curves for Clostridium difficile core genome analysis. A. Rarefaction curve showing 
the number of conserved genes for the isolates, i.e. the core genome. B. Rarefaction curve showing the number 
of genes in the pan-genome, i.e. the combined core and accessory genes. 
 
The core genes were aligned and the SNPs were used to build a core genome-based 
phylogeny. When the core genome-based tree was compared to the earlier mapping-based 
tree, the phylogenetic trees had a consistent topology (Figure 5.4). The consistent topology 
suggests that the previously defined clades (see Figure 5.2) are robust and can be used to 
group the C. difficile isolates for subsequent ARG analysis. 
  91 
 
Figure 5.4 Tanglegram comparing two phylogenetic trees for the Clostridium difficile isolates (generated 
using different approaches). The left phylogeny (mapping-based tree) from Figure 5.2 is compared with the 
right phylogeny (core genome-based tree) was generated using 93953 SNPs present in the core genome of the 
isolates. Horizontal lines between trees connect cognate isolates. 
  92 
5.4.1.2. Antimicrobial resistance genes 
For the 19 isolates that were identified as C. difficile, a total of 48 ARGs were identified using 
the ARDB isolate annotation tool (ARDBanno.pl, (Liu and Pop, 2009)) (Table 5.3). The 48 
ARGs encompassed resistance to five classes of antimicrobial (aminoglycoside, bacitracin, 
macrolide, tetracycline and glycopeptide). In relation to the clades identified in the 
phylogenetic analysis (Figure 5.2), the aminoglycoside, macrolide and tetracycline ARGs 
were only found in clades A and C. 
  
93 
Table 5.3. Clostridium difficile isolates and their identified ARGs. 
Isolate Effluent type Assembly statistics 
Kraken results (percentage identified 
reads) Ribotype Clade Identified ARGs (ARDB) 
Toxin genes 
(core 
genome) 
  
To
ta
l le
ng
th
 
No
. c
on
tig
s 
Av
g.
 co
nt
ig 
len
gt
h 
un
cla
ss
ifie
d 
C.
 d
iffi
cil
e 
C.
 p
er
fri
ng
en
s 
C.
 b
ot
uli
nu
m
 
 
 
an
t6
ia 
ba
cA
 
er
m
B 
te
tM
 
va
nR
G 
tcd
A 
tcd
B 
#28 Hospital 4323406 60 72057 6.23 89.35 0.03 0.01 39 C 1 1 1 1 1   
#29 Hospital 4144436 40 103611 4.94 93.56 0.03 0 14 B  1   1 1 1 
#30 Hospital 4323683 58 74546 6.25 89.57 0.02 0 39 C 1 1 1 1 1   
#31 Hospital 4064553 39 104219 4.95 94.52 0.02 0 11 A  1   1 1 1 
#32 Hospital 4068536 41 99233 5.15 94.23 0.02 0 11 A  1   1 1 1 
#33 Hospital 4141364 40 103534 3.64 92.62 0.04 0 10 C 1 1 1  1   
#34 Hospital 4066105 33 123215 5.13 94.3 0.02 0 11 A  1   1 1 1 
#35 Hospital 4574826 297 15403 3.66 95.1 0.03 0 unknown A 1 1 1 1 1  1 
#43 Hospital 4098800 37 110778 2.79 95.05 0.02 0 untested B  1   1 1 1 
#44 Hospital 4102890 36 113969 2.68 95.39 0.02 0 untested B  1   1  1 
#45 Hospital 4100726 35 117164 2.52 95.56 0.02 0 untested B  1   1 1 1 
#17 Pig farm 3834726 56 68477 15.01 84.54 0.05 0.01 78 D  1    1 1 
#4 Pig farm 3828920 56 68374 15.69 83.86 0.02 0.01 78 D  1     1 
#5 Pig farm 3826452 57 67131 14.81 84.69 0.08 0.01 78 D  1     1 
#6 Pig farm 4054739 57 71136 5.23 93.57 0.02 0 26 -  1   1 1 1 
#8 Pig farm 3830384 45 85120 15.33 84.16 0.02 0.01 78 D  1     1 
#24 WWTP* 4339582 43 100921 9.11 88.83 0.02 0 unknown -  1      
#26 WWTP 4303354 55 78243 4.66 91.69 0.03 0 39 C 1 1  1 1   
#27 WWTP 4296951 42 102308 4.64 91.62 0.02 0 140 C 1 1  1 1   
* denotes WWTP influent 
  94 
5.4.1.2.1. Aminoglycoside resistance genes 
All five isolates in clade C and one isolate (#35) in clade A had an identical copy of the 
aminoglycoside resistance gene ant6ia, which was not present in any of the reference 
genomes. Comparative genomics of these six isolates to the C. difficile 630 reference 
genome (the most closely related reference genome in the phylogeny) revealed that ant6ia 
was contained within an insertion at base 1821844 of the reference genome (shown for 
isolate #26, Figure 5.5). The ant6ia-containing insertion was 8090 bases long and had an 
average GC content of 47.53%, 1.6 times greater than the average GC content of C. difficile 
(29.06%, Figure 5.5, (Sebaihia et al., 2006)). Notably, the ant6ia-containing insertion was 
located adjacent to another insertion containing phage-related genes (Figure 5.5, phage-like 
insertion). 
 
  95 
 
Figure 5.5 Genome synteny between the Clostridium difficile reference 630 genome and contig 5 of 
isolate #26 (featuring ant6ia-containing insertion and a phage-like insertion). Aqua-coloured boxes 
represent putative coding sequences of each reading frame.  
 
  96 
5.4.1.2.2. Tetracycline resistance genes 
In addition to ant6ia, four of the C. difficile isolates from clade C were found to contain the 
tetracycline resistance gene tetM. Comparative genomics of the identified tetM genes 
showed that they were 100% identical among all four isolates and, when compared to the 
non-redundant nucleotide collection, the tetM gene was found to share 100% identity with the 
tetM gene of conjugative transposon Tn5397 (GI: 669664115). Comparison of the tetM-
containing contigs of the clade C isolates showed that they were identical to each other, and 
had collinear synteny with the C. difficile 630 reference genome and the conjugative 
transposon Tn5397 (isolate #26 shown as representative in Figure 5.6, A). The reference 
genome and Tn5397 shared 100% identity over the length of the transposon, however the 
four isolates from clade C all featured a deletion of 2795 bases from the transposon 
(henceforth referred to as the Tn5397-like element, Figure 5.6). Alignment of tetM-containing 
contigs to the C. difficile 630 reference genome revealed that the Tn5397-like element has 
changed location within the clade C isolates, relative to the reference genome (isolate #26 
shown as representative in Figure 5.6, B). This was not attributable to a mis-assembly as 
read mapping was consistent across contigs (data not shown). Inspection of the Tn5397-like 
element in each clade C isolate showed an average GC content of 38.61%. In addition to the 
transposition of the tetM gene in the clade C isolates, the isolates were also found to be 
missing the nearby toxin-encoding genes tcdA and tcdB (confirmed in the core genome 
analysis, Figure 5.3). 
  97 
 
Figure 5.6. A. Genome synteny among the Clostridium difficile reference 630 genome, the conjugative 
transposon Tn5397, and contig 18 of isolate #26. B. Genome synteny between the C. difficile reference 
630 genome and contig 18 of isolate #26, showing varied location of the Tn5397 transposon and the 
Tn5397-like element. In addition, there has been a deletion of tcdA and tcdB, downstream of the element, in 
isolate #26 (as for all tetM-containing clade C isolates, not shown). 
  98 
Isolate #35 (clade A) had a tetM gene divergent to the tetM gene found in the clade C 
isolates (henceforth referred to as tetM-type II), with which it shared 89% identity. When 
compared to the non-redundant nucleotide collection, the tetM-type II gene was found to 
share 100% identity with the tetM gene of Tn5801-like tetM gene (GI: 220898661). 
 
5.4.1.2.3. Other types of antimicrobial resistance genes 
For the 19 C. difficile isolates, ARG annotation with ARDB identified 14 isolates that 
contained the glycopeptide resistance gene vanRG. On closer inspection, these 14 isolates 
were found to contain the entire vanG-type glycopeptide resistance locus described by 
Sebaihia et al. (Sebaihia et al., 2006). The five isolates not to feature the vanG-type 
glycopeptide resistance locus were all clade D isolates (pig farm) and the clade-less isolate 
#24 (WWTP influent). 
Three isolates from clade C (#28, #30, #33) and one isolate from clade A (#35) were found to 
contain the macrolide resistance gene ermB, encoding an adenine methyltransferase. ARG 
annotation with ARDB also revealed all isolates to contain the bacitracin resistance gene 
bacA. However, bacA is only known to confer antimicrobial resistance if overexpressed 
(Ghachi et al., 2005). 
 
5.4.2. Clostridium perfringens isolates 
Based on the assembled draft genome and the percentage of reads matching taxonomic 
marker genes (Table 5.1), a total of 14 isolates had WGS data of sufficient quantity and 
quality to putatively annotate the isolate as C. perfringens. 
 
5.4.2.1. Phylogenetic analysis 
To assess the relatedness of the 14 C. perfringens isolates from the four different 
environmental effluents (dairy farm, hospital, pig farm and WWTP), a whole genome 
phylogeny was performed alongside a set of references. The phylogeny resulted in the 
majority of environmental C. perfringens isolates falling in to 4 distinct clades (Clades A-D, 
Figure 5.7). 
 
  99 
 
Figure 5.7 Phylogenetic tree of Clostridium perfringens isolates. The phylogenetic tree shows the 
evolutionary relationships between the C. perfringens isolates and reference genomes. Colour coding 
represents the origin of the isolates. Bootstraps less than 90 are shown. Clades containing multiple novel 
isolates are highlighted by lettered orange lines (A – D). The scale bar is in expected number of nucleotide 
substitutions per site (n= 139845).  
  100 
To ensure the robustness of the results from the whole genome phylogeny of the isolates, 
and to explore the gene diversity of the isolates, a core genome analysis of the C. 
perfringens isolates was performed. The core genome was found to consist of 1842 genes 
(aligned to 1774527 sites), the total number of was 4982 and the average number of genes 
per isolate was 2850 (Figure 5.8). 
 
 
Figure 5.8 Rarefaction curves for Clostridium perfringens core genome analysis. A. Rarefaction curve 
showing the number of conserved genes for the isolates, i.e. the core genome. B. Rarefaction curve showing 
the number of genes in the pan-genome, i.e. the combined core and accessory genes. 
 
The core genes were aligned and the SNPs were used to build a core genome-based 
phylogeny. When the core genome-based tree was compared to the earlier mapping-based 
tree, the phylogenetic trees had consistent topology (Figure 5.9). The consistent topology 
illustrated by the tanglegram suggests that the previously defined clades (see Figure 5.7) are 
robust and could be used to group the C. perfringens isolates for subsequent ARG analysis. 
 
  101 
 
Figure 5.9 Tanglegram comparing two phylogenetic trees for the Clostridium perfringens isolates 
(generated using different approaches). The left phylogeny (mapping-based tree) from Figure 5.7 is 
compared with the right phylogeny (core genome-based tree) generated using 70982 SNPs present in the core 
genome of the isolates. Horizontal lines between trees connect the isolates. 
  102 
5.4.2.2. Antimicrobial resistance genes 
For the 14 isolates that were identified as C. perfringens, a total of 33 ARGs were identified 
using the ARDB isolate annotation tool (ARDBanno, (Liu and Pop, 2009)) (Table 5.4). The 33 
ARGs encompassed resistance to three classes of antimicrobial (bacitracin, macrolide and 
tetracycline). In relation to the clades identified in the phylogenetic analysis (Figure 5.7), the 
macrolide ARGs were only found in clade D and the tetracycline ARGs were only found in 
clades A, C and D. The two dairy farm isolates (clade B) were the only C. perfringens isolates 
that did not contain tetracycline resistance genes. 
 
  
103 
Table 5.4 Clostridium perfringens isolates and their identified ARGs. 
Isolate Effluent type Assembly statistics 
Kraken results 
(percentage identified reads) Clade Identified ARGs (ARDB) 
  
To
ta
l le
ng
th
 
No
. c
on
tig
s 
Av
g.
 co
nt
ig 
len
gt
h 
un
cla
ss
ifie
d 
C.
 d
iffi
cil
e 
C.
 p
er
fri
ng
en
s 
C.
 b
ot
uli
nu
m
 
 
ba
cA
 
er
m
Q 
m
ph
B 
te
tP
A 
te
tP
B 
#40 Dairy farm 3214184 14 229585 5.66 0.03 93.89 0.01 B 1     
#41 Dairy farm 3215299 14 229664 5.8 0.03 93.76 0.01 B 1     
#47 Hospital 3272685 32 102271 8.46 0.03 91 0.01 C 1   1  
#57 Hospital 3280735 25 131229 8.52 0.03 91.01 0.01 C 1   1  
#46 Hospital 3272482 50 65450 8.75 0.03 90.65 0 C 1   1  
#59 Hospital 3281160 22 149144 8.8 0.06 90.69 0.01 C 1   1  
#56 Hospital 3276555 30 109219 9.69 0.14 89.69 0.01 C 1   1  
#14 Pig farm 3394466 19 178656 8.57 0.03 90.92 0.01 A 1   1 1 
#19 Pig farm 3101644 112 27693 8.84 0.04 90.38 0.06 D 1  1 1  
#20 Pig farm 3107330 93 33412 9.94 0.03 89.43 0.02 D 1   1 1 
#15 Pig farm 3141691 84 37401 9.96 0.03 89.38 0.04 D 1 1 1 1  
#21 Pig farm 3116792 93 33514 10.79 0.04 88.53 0.02 D 1   1 1 
#25 WWTP* 3266305 54 60487 10.53 0.04 88.64 0.03 A 1   1 1 
#23 WWTP* 3307749 25 132310 10.97 0.02 88.54 0.01 C 1   1  
* denotes WWTP influent 
  104 
5.4.2.2.1. Macrolide resistance genes 
Two of the C. perfringens isolates contained ARGs conferring resistance to macrolides. 
Isolate #19 contained the macrolide resistance gene mphB and isolate #15 contained both 
ermQ and mphB. Comparative genomics of isolate #15 to the C. perfringens ATCC13124 
reference genome (the most closely related reference genome in the phylogenetic tree) 
revealed that ermQ was contained within an insertion at base 621786 of the reference 
genome (Figure 5.10). The ermQ-containing insertion was 4438 bases long and had an 
average GC content of 26.89% (consistent with the broader C. perfringens genome). 
 
 
Figure 5.10 Comparison (displayed using ACT) between the Clostridium perfringens reference 
ATCC13124 genome and contig 9 of isolate #15 (featuring ermQ-containing insertion). 
  105 
The mphB gene present in both isolate #15 and #19 was found to be identical. In both 
isolates, the mphB gene was located on the end of a contig, therefore a construct was made 
with adjacent contigs (adjacent relative to the reference genome) for each isolate. When the 
mphB construct was compared to the ATCC13124 reference genome it was found to contain 
an insertion of 4246 bases at base 428411 of the reference genome (Figure 5.11). The 
average GC content of the mphB-containing insertion was 35.81%.  
 
 
Figure 5.11 Genome synteny between the Clostridium perfringens reference ATCC13124 genome, a 
construct of contigs 42, 31 and 28 of isolate #15 and contig 31 (containing mphB) of isolate #15.  
  106 
5.4.2.2.2. Tetracycline resistance genes 
Of the 14 isolates, 12 contained the tetracycline resistance gene tetPA (encoding a 
tetracycline efflux pump). The tetPA gene shared an average of 98% identity among the 12 
isolates, however it was present in different genomic locations amongst isolates. Where 
isolates were closely related the tetPA gene was found in the same genomic context, for 
example in Clade C the tetPA gene was present in an insertion at base 3203264 of the 
reference (isolate #15 shown as representative in Figure 5.12). The tetPA insertion is 
10904bp long, has a GC content of 26.90% and also contains phage-like proteins on a single 
strand. 
 
 
Figure 5.12 Comparison (displayed using ACT) between the Clostridium perfringens reference 
ATCC13124 genome and the tetPA-containing contig of isolate #23. 
  107 
In addition to tetPA, four isolates also contained the tetracycline resistance gene tetPB 
(encoding a ribosomal protection protein) adjacent to a second copy of tetPA. The tetPB 
gene was conserved between isolates but was not present in the reference genome. Using 
isolate #14 as a representative, the contig containing tetPB was compared the non-redundant 
nucleotide database using NCBI blast and was found to be similar to the C. perfringens 
plasmid pCW3 (GenBank: DQ366035.1). The plasmid pCW3 covered 63% of the contig with 
99% identity. Comparative genomics between the tetPB-containing contig had a large (21777 
bases) insertion relative to the pCW3 plasmid reference (Figure 5.13). 
 
 
Figure 5.13 A. Genome synteny between the Clostridium perfringens plasmid pCW3 and the tetPB and 
tetPA containing contig of isolate #14. B. Circularised representation of tetPB-containing plasmid 
(isolate #14 tetPB-containing plasmid shown). 
  108 
 
5.4.3. Annotating endospore-formers from metagenomes 
To compare the taxonomic results of the metagenomic analysis and bacterial isolation from 
environmental samples, a split sample was used. A sample taken from the pig farm was used 
to prepare a metagenomic analysis (using methods described in Chapters 2-4), which was 
compared to three isolates obtained from that same sample (see footnote in Table 5.1). 
Metagenomic analysis binned 2% of reads in the Firmicute phylum; 1% in the Clostridia class 
and 0.3% into Clostridium sticklandii (C. sticklandii). No metagenomic reads were annotated 
as C. colicanis or C. butyricum; the most closely related species to the bacteria isolated from 
the sample (see Table 5.1, section 5.4.1). Further attempts were made to identify sequences 
related to C. sticklandii (GenBank: NC_014614) and the draft genome of one of the unknown 
isolates (#10, C. colicanis-like) from the metagenomic data. BLAST comparison of the 
assembled metagenome (resulting in a total of 14689361 contigs with an N50 of 153 bases) 
yielded an alignment that covered 6% of the C. sticklandii reference (Figure 5.14); whereas 
only 2% of the draft genome of isolate #10 was covered. However, mapping coverage of 
metagenomic reads against both genomes was equivocal (26% for C. sticklandii and 25% for 
isolate #10). 
 
 
Figure 5.14 Genome synteny between the Clostridium sticklandii reference genome (GI: 310657316) and 
the metagenome contigs of the pig farm lagoon effluent. 
 
  109 
5.5. Discussion 
This chapter has shown that endospores of bacteria can be successfully reactivated to 
culture clinically relevant pathogens from all effluent sampling locations. Furthermore, the 
endospore-forming bacteria have been found to contain ARGs and also show evidence of 
being horizontally transmitted within their genomes. 
Samples of various environmental effluents were collected over a period of four months and 
the endospores in each sample were selected, reactivated and used to culture anaerobic 
bacteria. Of the cultured bacteria, those with Clostridium sp. colony morphology were 
subsampled and subjected to 16S sequencing for preliminary identification. A total of 59 
isolates were then subjected to WGS. Although all isolates received preliminary identification 
by 16S sequencing, some isolates were unclassified after Kraken classification of the WGS 
data. Some WGS data also created bad assemblies. This was addressed by quality checking 
the WGS data and removing isolates from further analysis if the number of contigs exceeded 
200, or if the assembled draft genome size was unusually large, and then re-classifying 
remaining isolates with Metaphlan2. This approach resulted in 48 isolates; all successfully 
classified as Firmicutes and used in the next stages of the analysis. The fact that Metaphlan2 
successfully classified isolates when Kraken was unable to may be due to the fact that, unlike 
Kraken, Metaphlan2 does not attempt to assign every read to a taxonomic level (Segata et 
al., 2012), but uses taxonomic markers to bin reads and identify taxa. Consequently, the 
isolates that Kraken was unable to classify but passed WGS data quality checking and 
Metaphlan2 classification are likely divergent or novel species of bacteria that are related to 
endospore-forming Firmicutes. Therefore these isolates could be hitherto unknown clinically 
relevant pathogens. Considering that 15 of the 48 sequenced and quality-checked isolates 
were unclassified environmental isolates (Table 5.2), the issue of database limitations in the 
identification of bacterial species from WGS data is an important one that should be 
addressed. Further work is required to annotate the genomes of unclassified environmental 
isolates, in order to augment the available databases and facilitate in silico identification of 
bacterial isolates. More work is also needed to characterise the nature and pathogenic 
potential of these apparently novel bacterial species.  
Of the 48 isolates that were sequenced and passed quality checking, 19 isolates were 
classified as Clostridium difficile and 14 were classified as Clostridium perfringens, both of 
which are pathogenic bacteria that cause significant clinical disease in both human and 
veterinary medicine. Indeed, for the months that C. difficile was isolated from the hospital 
  110 
effluent, cases of C. difficile infection were record by Public Health England’s data capture 
system for the hospital being sampled (Public Health England, 2013).Clade B of the C. 
difficile phylogenetic tree contained isolates from two collection dates, a month and a half 
apart. Additionally, clade C of the same phylogeny featured C. difficile isolates that were from 
different sampling locations (hospital and WWTP effluent samples) (Figure 5.2). Further work 
is required to see if identical isolates can be recovered over prolonged periods of time, thus 
potentially demonstrating the prolonged persistence of specific pathogenic bacteria in the 
environment. 
These isolates of pathogenic bacteria were found to harbour ARGs that conferred resistance 
to several classes of commonly used antimicrobials. The similarity of ARG-containing contigs 
to known MGEs and the abnormal GC-content of ARG-containing elements, in addition to the 
transposase and recombinase genes found in close proximity to the ARGs, suggest that 
these ARGs were likely to be horizontal acquisitions (Baran and Ko, 2008). This indicates the 
potential for bacterial endospores that are present in the environment to harbour mobilisable 
ARGs. 
In addition to ARGs, some of the C. difficile and C. perfringens isolates also contained toxin-
encoding genes. In the case of C. difficile, eight isolates had both C. difficile toxin-encoding 
genes (tcdA and tcdB) and five isolates had the toxin-encoding gene tcdB only. It has been 
reported that only toxin B is required for C. difficile virulence (Lyras et al., 2009), meaning 
that 68% of C. difficile isolates were likely to be virulent. The presence of toxin-encoding 
genes, in addition to ARGs, in the genomes of these isolates of pathogenic bacteria suggests 
that environmental instances of these bacteria may have clinical significance if they were to 
re-enter human or animal populations. This is also suggested by the phylogenetic 
relatedness of these isolates to known pathogenic reference genomes. More extensive 
monitoring of effluents and effluent destinations may help elucidate the fate of pathogenic 
bacteria that are entering the environment. 
Due to the fact that all sampled effluents on the pig farm entered a single effluent lagoon 
(from where the lagoon effluent is subsequently used as a fertiliser), a metagenome was 
generated from the lagoon effluent for use as background sample. Although both C. difficile 
and C. perfringens were cultured from endospores collected from various pig farm effluents 
(but not the lagoon itself), no trace of C. difficile or C. perfringens were found in the 
metagenome, either by taxonomic assignment of reads or the alignment of metagenomic 
contigs to reference databases. It is possible that no C. difficile or C. perfringens were 
present in the lagoon; however, C. perfringens was cultured from lagoon effluent that was 
  111 
being applied as fertiliser to a neighbouring field. It is likely that the metagenome sample was 
not sequenced at a sufficient sequencing depth, as exemplified by the fact that 473120 
(0.3%) reads were classified as C. sticklandii yet only 6% of the C. sticklandii genome was 
assembled (26% of the genome was covered by mapping metagenomic reads). Alternatively, 
it may be the case that C. difficile and C. perfringens were only present in the metagenome 
sample as endospores (awaiting reactivation to a vegetative state when environmental 
conditions were appropriate), which the metagenomic methodology may have been unable to 
extract DNA from. In terms of isolates that were cultured from the split metagenome sample, 
isolate #10 (C. colicanis-like) was cultured yet only 2% of the C. colicanis genome was 
covered in the metagenome assembly (25% of the genome was covered by mapping 
metagenomic reads). 
While metagenomics is a powerful tool that has allowed for the characterisation of ARGs in 
environmental samples, it is as yet unable to facilitate the linking of function with taxonomy, 
specifically the identification of species of bacteria that are carrying ARGs. This chapter has 
shown that WGS of environmental isolates can be used to address this limitation of 
metagenomic analysis. At the same time it has also shown that the metagenomic 
identification of ARGs in the environment is corroborated by the genomic analysis of 
environmental isolates of clinically relevant pathogens, revealing them to harbour ARGs often 
carried on horizontally transmissible MGEs. The fact that ARGs were also present in 
potentially novel environmental bacteria suggests that there is more research required, both 
in terms of metagenomics and environmental culture, in order to characterise environmental 
bacteria and monitor ARGs from the environment. 
Finally, this chapter has revealed that endospores of environmental bacteria are harbouring 
ARGs. Considering that endospores can withstand extremes in environmental conditions and 
can remain dormant for many years until appropriate conditions for reactivation are 
encountered (Cano and Borucki, 1995), this finding suggests that endospore-forming bacteria 
are facilitating the persistence of ARGs in the environment. Taken in conjunction with the fact 
that clinically relevant pathogens can form endospores and the fact that the ARGs identified 
in this chapter appear to be mobile, this work offers insight into the dynamics of ARGs in the 
environment and suggests that the monitoring of bacterial endospores would be a valuable 
addition to the environmental framework to tackle antimicrobial resistance (Berendonk et al., 
2015). 
 
  112 
5.6. Conclusion 
The work in this chapter has contributed to the development of an environmental framework 
for monitoring ARG dissemination by showing that ARGs are harboured by bacterial 
endospores isolated from effluents, thus presenting a possible mechanism through which 
ARGs could persist in the environment. Furthermore, many of the bacterial endospores that 
have been reactivated and cultured in this study are known pathogenic species that cause 
significant human and animal disease. 
In summary, the presence of endospores in the environment that can be reactivated to yield 
viable pathogenic bacteria harbouring apparently mobile ARGs suggests that environmental 
reservoirs of ARGs pose a risk to human and animal health and the longevity of 
antimicrobials. 
 
  113 
Chapter 6. General discussion 
6.1. Preface 
The purpose of the work described in this dissertation has been to help establish an 
environmental framework to monitor antimicrobial resistance gene (ARG) dissemination in 
the aquatic environment. This framework offers the potential to quantify the risk of 
antimicrobial resistance in environmental samples and contribute toward antimicrobial 
resistance risk assessments. The establishment of this environmental framework has 
required the design and development of appropriate experimental methodology and 
analytical tools, as well as the identification of factors likely to influence ARG dissemination in 
the environment. In addition, this dissertation has described the persistence of ARGs in the 
environment through endospore-forming bacteria. 
This discussion examines the research findings of the doctoral research detailed in this 
dissertation and summarises the framework that has been developed. In addition, an outline 
of the work that would be required in order to incorporate this framework into future ARG 
research projects and environmental ARG monitoring programmes is discussed. 
 
6.2. Summary of findings and dissertation outcomes 
In order to develop the basis for an environmental framework to monitor ARG dissemination 
in the aquatic environment, work outlined in this dissertation involved the development of a 
bioinformatics tool to identify ARGs in sequencing data, the implementation of a pilot study to 
confirm effluents entering the aquatic environment contain ARGs and the demonstration that 
a single river catchment contains a diverse array of bacterial endospores that can be re-
activated and interrogated to reveal ARG presence. 
 
  114 
6.2.1. Establishing an environmental framework for monitoring ARG 
dissemination in the aquatic environment 
6.2.1.1. Summary of findings 
The first step in establishing an environmental framework for monitoring ARG dissemination 
was the development of suitable experimental methodology to detect and quantify ARG 
presence in environmental samples. A review of the literature suggested that metagenomics 
was likely to be the most appropriate technology to form the basis of the framework. 
However, at the time of planning the work, there were insufficient analytical tools available to 
identify ARGs within metagenomic sequencing data in a quick, efficient and quantitative 
manner, which were capable of taking into account the frequently changing ARG databases 
available. Therefore the first results chapter (Chapter 2) described the necessary 
development of the Search Engine for Antimicrobial Resistance (SEAR), a bioinformatic tool 
for the detection of horizontally acquired ARGs in raw sequencing data (Rowe et al., 2015). 
Firstly, a wet-laboratory methodology for preparing metagenome quality DNA from aquatic 
samples was developed. This involved determining the suitability of centrifugation and 
vacuum filtration for the isolation of bacteria from aquatic samples, as well as the assessment 
of DNA extraction protocols to determine the most appropriate method for preparing high 
quality DNA suitable for metagenomic sequencing. The results of this methodology 
development are detailed in Appendix 6 and the finalised wet-laboratory methodology for 
metagenome sequencing was implemented in Chapter 2.  
SEAR was then designed to utilise sequencing data straight from the sequencing machine; it 
begins by quality checking the data and subtracting unnecessary reads before reconstructing 
full-length ARG sequences. These full-length ARG sequences are subsequently annotated, 
compared to online databases and presented in a concise report. SEAR therefore enabled 
metagenomics to be used as a viable approach for rapidly detecting all known horizontally 
acquired ARGs in a given sample (according to available databases and not including SNP-
based ARGs), whilst giving important information such as the relative abundance of each 
ARG and associated mobile genetic elements (MGEs). SEAR also served as a proof of 
principal demonstrating that sequencing applications such as metagenomics can be 
developed for rapid diagnostic testing; the diagnostic application of such technologies is 
currently being proposed in the scientific literature (Miller et al., 2013). Finally, in addition to 
using SEAR to analyse metagenomes for ARG content, additional software and databases to 
  115 
facilitate the interrogation of metagenomes for pathogenic bacteria, mobile genetic elements 
and taxonomic composition (such as the PATRIC pathogen database) were trialled and 
subsequently implemented in the methodology described in results chapters 3 and 4. 
Once appropriate methodology had been developed and the bioinformatic tools for analysing 
metagenomic data were in place, the next step in establishing the environmental framework 
for monitoring ARG dissemination was to confirm the underlying assumption of this 
dissertation; namely that effluents entering the aquatic environment were disseminating 
ARGs. Chapter 3 served as a pilot study that used a set of longitudinal samples from two 
effluents entering a single river catchment and compared them to a background sample of 
river source water in order to confirm this underlying assumption. The work reported in 
Chapter 3 found that not only did effluents entering the aquatic environment from both human 
and animal faecal sources contain a diverse range of ARGs, but that the abundance of these 
genes was consistently higher than in the background sample of river water. In addition to 
abundance, the diversity of pathogens and MGEs were also greater in the effluent samples 
compared to the river water. Consequently, the underlying assumption that effluents 
disseminate ARGs was confirmed and the methodology used in the Chapter 3 work deemed 
suitable to serve as the foundation of analysis within the environmental framework to be 
developed.  Specifically, this entailed the use of metagenomics and SEAR to assess ARG 
load of effluent samples entering the aquatic environment in addition to the use of 
background sampling to provide context for ARG load. 
The next step in establishing the environmental framework was to determine additional 
factors that may affect ARG dissemination in the effluents being monitored, identifying key 
factors that may need to be incorporated into the framework. To address this, the work 
detailed in Chapter 4 aimed to assess the impact of ARG load and potential selection 
pressures on the expression of ARGs in environmental samples. The results reported in 
Chapter 4 showed that in addition to hospital and farm effluents having a consistently high 
load of ARGs in relation to background samples across a series of monthly samples, many 
different ARGs were also being expressed and a strong positive correlation was observed 
between the abundance of ARGs and corresponding ARG transcripts. Chapter 4 also 
demonstrated a positive correlation between hospital antimicrobial usage and ARG transcript 
abundance in hospital effluent. Finally, Chapter 4 also demonstrated the presence of 
antimicrobial compounds in both hospital and farm effluents. Collectively, these results 
suggested that ARG load and potential antimicrobial selection pressures were impacting on 
the level of expression of ARGs in environmental samples. Thus, the work reported in 
  116 
Chapter 4 developed the environmental framework through the application of 
metatranscriptomics, antimicrobial usage data and LCMS measurement of antimicrobial 
compounds to identify additional factors likely to be playing an important role in the dynamics 
of ARG abundance within effluents.  
In addition to the metabolically active bacteria that were assessed through use of 
metatranscriptomics in the work reported in Chapter 4, the potential impact of non-
metabolically active bacteria on ARG dissemination in the environment was investigated in 
the work reported in Chapter 5. Chapter 5 described the identification of ARGs within 
bacterial endospores that were isolated and re-activated from a variety of environmental 
samples, including the effluents that had been studied in the previous chapters and other 
environmental sources. The majority of these endospores were dormant Clostridium difficile 
and Clostridium perfringens bacteria, which are clinically relevant pathogens (Tonna and 
Welsby, 2005, Rood and Cole, 1991). Several of the ARGs identified within the genomes of 
these bacteria showed evidence of mobility and recent horizontal transfer highlighting that 
bacterial endospores present a mechanism through which ARGs could persist in the 
environment and contribute to the resistome. 
 
6.2.1.2. The environmental framework 
In summary, the results presented within this dissertation describe work that forms a basis for 
an environmental framework to monitor the dissemination of ARGs, via effluents in the 
aquatic environment (Figure 6.1). By placing strategic effluent monitoring points in the 
antimicrobial resistance risk assessment model introduced in Chapter 1, the environmental 
framework for monitoring ARG dissemination would generate an antimicrobial resistance 
potential (ARP) categorisation for a given effluent that would contribute to the ongoing overall 
antimicrobial resistance risk assessment for a given catchment.  
 
  117 
 
Figure 6.1 Proposed environmental framework for monitoring ARG dissemination and assigning 
antimicrobial resistance potential to environmental samples. Effluent samples are taken within a catchment 
that is being subjected to an antimicrobial resistance risk assessment, which takes into account transmission of 
resistance to humans, ARG dissemination and emergence of resistance in the environment (upper half of figure, 
augmented from Chapter 1 to show points of effluent monitoring used in this work). Effluent samples are run 
through the environmental framework to monitor ARG dissemination (lower half of figure), giving an 
antimicrobial resistance potential categorisation that feeds into the overall antimicrobial resistance risk 
assessment. Flowchart key: dark blue represents an experimental process, green represents an analysis, aqua 
represents data handling and storage, orange represents a categorisation. 
  118 
In the proposed environmental framework represented in Figure 6.1, an effluent sample is 
split and subjected to several experimental processes in order to determine the ARG and 
pathogen load of the sample, as well as the presence of selection pressures, virulence 
factors and metabolic activity. After normalising data and comparing to a suitable background 
sample, a primary ARP categorisation is made. Effluent samples with high ARG and 
pathogen load, in addition to the presence of potential antimicrobial selection pressures such 
as high antimicrobial usage at the effluent origin or the presence of antimicrobial residues in 
situ, would be deemed to have a high ARP and are subjected to additional experimental 
processes, such as bacterial culture and whole genome sequencing of isolates. These 
additional experimental processes (including culture, susceptibility testing, WGS and analysis 
of stored effluent samples) take considerably more time and effort than the initial 
experimental process and analysis components of the framework, so the use of a stored 
sample for isolate work is only introduced if the primary ARP outcome is high. This therefore 
ensures that the monitoring of ARG dissemination can be performed rapidly yet additional 
risk factors can be identified in the event of effluent samples with high ARP. The use of 
culturing and susceptibility testing also allows for the discovery of novel bacteria and ARGs, 
which in turn can be used to augment the existing monitoring databases. 
The ARP categorisation is designed to reflect the level of ARG dissemination that an effluent 
could be responsible for. It is envisaged that this environmental framework can be 
incorporated into antimicrobial risk assessments and facilitate a standardised approach for 
the regular monitoring of ARG dissemination in the aquatic environment. 
 
6.2.2. Future work on an environmental framework for monitoring 
ARG dissemination 
6.2.2.1. Assigning significance to antimicrobial resistance potential 
As part of the future work to develop the environmental framework, it is crucial that a 
measure of significance can be applied to the ARP categorisation, or to the factors that 
contribute to this categorisation. Indeed, in parallel to the work described in this thesis, efforts 
to develop an ARP categorisation for antimicrobial resistance determinants based on 
metagenomic analysis of ARGs, MGEs and pathogens in existing environmental samples 
has been reported (Port et al., 2014). In this study, the authors used their ARP (termed an 
antimicrobial resistance determinants index) and evaluated the differences between aquatic 
  119 
environments with respect to their proximity to likely human impact. However, both the ARP 
categorisation used in Port et al (2014) (and indeed the expanded ARP categorisation 
outlined in the framework in Figure 6.1) use an arbitrary system that does not take into 
account the potential difference of effect among individual genes, pathogens or antimicrobial 
classes and also restricts the ARP categorisation to simplified discrete descriptive values 
(high, low etc.). It is therefore crucial that future work on the framework must formalise ARP 
categorisation, possibly by incorporating the continuous values generated from quantitative 
experimental analyses of the framework into the ARP categorisation and further exploring the 
differential impacts of the qualitative analysis of contributory data.  
To elaborate further on the value of assigning significance to ARP categorisation, it is 
currently possible that, despite a high ARP, an effluent may actually have a relatively low risk 
of contributing to the emergence of antimicrobial resistance in the environment or to the 
transmission of antimicrobial resistance to humans. This is due to the fact that the 
categorisation system described here looks at each analysis result as a whole, for example 
the total ARG load of the effluent, and does not take into account if an individual ARG is 
driving this, or whether the highly abundant ARGs are frequently implicated in clinical cases 
of antimicrobial resistance. This is demonstrated by the results described in Chapter 4, where 
particular classes of ARGs had high ARG abundance but did not correlate well with ARG 
transcript abundance or antimicrobial usage (e.g. sulfonamide ARGs in the hospital effluent 
samples). It is also clear that other factors may have a varied effect depending on ARG type. 
For instance, it was found that tetracycline ARG abundance had a strong positive correlation 
with tetracycline ARG transcript abundance and it was also found that tetracycline ARG 
transcript abundance in turn had a strong positive correlation with tetracycline usage. 
However, despite a strong positive correlation between beta lactam ARG abundance and 
ARG transcript abundance, very poor correlation was observed between beta lactam ARG 
transcript abundance and beta lactam usage. This illustrates the point that despite an overall 
trend such as total ARG abundance correlating to ARG transcript abundance, there are likely 
to be differential key forces driving this trend. These key forces are most likely at the level of 
an antimicrobial class, ARG mechanism or even individual ARGs. The differential impact 
discussed here in relation to antimicrobial class is equally applicable to the great diversity of 
MGEs (e.g. different plasmid Incompatibility groups) and pathogenic bacterial species in a 
sample, and much further work is needed to fully understand these differential effects.   
In conclusion, in order to assign significance to ARP categorisation it is likely that it will be 
necessary to look further than just the analysis scores (e.g. ARG load or pathogen load) and 
  120 
give appropriate weighting to specific classes, types or individual components (e.g. a high 
abundance of tetracycline ARGs may result in a more severe ARP than an equivalent 
abundance of sulfonamides). 
However, an argument against the categorisation of ARP through weighting each component 
is that there may not be enough information available to be able to determine the effect of a 
particular component on the overall ARP. In light of this, a viable alternative would be to 
create a selection of sentinel ARGs and antimicrobial resistant bacteria that could be tested 
for within effluents and used to determine the ARP category, as suggested recently by 
Berendonk et al. (2015) in their opinion piece on tackling antimicrobial resistance (Berendonk 
et al., 2015). This approach would rely on the creation of a set of maximum admissible levels 
for the sentinel ARGs and bacteria. However, although offering considerable advantages, this 
approach would require a large amount of validation work to determine suitable sentinel 
candidates before it could be implemented. 
 
6.2.2.2.  Understanding ARG dynamics 
Following on from the previous points that it is necessary to go beyond ARG load as a whole 
and weight individual ARGs according to their own ARP, it is vital to the development of the 
framework that a greater understanding of ARG dynamics is achieved, particularly within the 
environment. A good first step towards this goal within the scientific community has been to 
view ARGs as emerging environmental contaminants in studies that examine the occurrence 
of ARGs in different environments (Pruden et al., 2006, Gillings, 2013). This has allowed for 
greater emphasis on the importance of the environment in the emergence of antimicrobial 
resistance, resulting in a great volume of studies in recent years that document the 
widespread dissemination of ARGs in the environment from human and animal sources 
(Berglund et al., 2015, Bengtsson-Palme et al., 2014). Consequently, environmental ARG 
research now incorporates additional factors that may contribute to the dissemination and 
persistence of ARGs that have been released into the environment, such as heavy metals 
and antibiotics (Lu et al., 2015, Khan et al., 2013). 
A key area of research concerning ARG dynamics, highlighted by the work in this dissertation 
and the current body of literature as requiring further understanding, is the effect of clinically 
used antimicrobials on ARGs in the environment. The work in this dissertation has shown 
that clinically used antimicrobials are present in effluents that also contain a high load of 
ARGs, MGEs and bacterial species, and it has also shown that antimicrobial usage at the 
  121 
effluent source influences the occurrence of ARGs within the effluent. However, the work 
described in this dissertation has not determined if the presence (or concentration) of 
antimicrobial compounds in effluents plays a direct role on the abundance of specific ARGs 
(or their transcripts). Studies have shown that even sub-inhibitory concentrations of 
antimicrobials can influence bacterial behaviours in the environment (Bruchmann et al., 2013) 
and another recent study into environmental antimicrobial concentrations has shown that 
sub-inhibitory concentrations of antimicrobials can select for resistant bacteria (Gullberg et 
al., 2011). The fact that hospital effluents contain detectable levels of antimicrobial 
compounds that are in very high use at the effluent source may result in increased 
abundance of ARGs, as was observed in a study of pharmaceutical manufacturer effluents 
that found high concentrations of antimicrobials impacting ARG levels in the aquatic 
environment (Larsson et al., 2007, Kristiansson et al., 2011). With an improved 
understanding of the effect of clinically used antimicrobials on bacteria in the environment, 
the contribution of the presence of antimicrobial compounds in the environment to the 
dissemination of ARGs and the emergence of antimicrobial resistance can be assessed. 
Additional factors that may affect ARG dynamics include the genetic context of ARGs (e.g. 
ARGs located within MGEs or downstream of constitutively-active promoters) and the impact 
of non-antimicrobial selection pressures. The work presented in Chapter 4 describes the 
possible effect of antimicrobial selection pressures on ARG abundance, however non-
antimicrobial selection pressures may also impact ARG abundance. For example, the 
presence of heavy metal contamination in the environment may result in the increased 
abundance of ARGs due to the co-selection that can occur if ARGs exist on the same MGE 
as heavy metal resistance genes (Baker-Austin et al., 2006, Knapp et al., 2011). It may also 
be a possibility that ARGs may be located on MGEs with varying horizontal transfer potential, 
with certain MGEs responding to specific environmental stimuli with an increase in 
abundance and relative gene transfer potential (Wright et al., 2008). Further work is required 
into the effects of genetic linkage and co-selection on the abundance of ARGs and the role 
that these might play in their dissemination before the importance of such influences can be 
fully understood. 
A final note on the areas of ARG dynamics to be addressed concerns the interplay between 
the environmental resistome and that found in the clinical setting. Specifically, which ARGs 
are present in both the environment and in pathogenic bacteria that are found in the clinical 
setting? Additionally, can these core ARGs be used as sentinel ARGs and form part of 
antimicrobial resistance risk assessments? There are a plethora of studies documenting 
  122 
antimicrobial resistant pathogenic bacteria isolated from clinical samples and there are also 
numerous studies documenting the environmental release of similar bacteria from human 
and animal sources (Schwartz et al., 2003, Caplin et al., 2008, Fuentefria et al., 2011, Harris 
et al., 2012, Li et al., 2015a). Studies are now increasingly documenting antimicrobial 
resistant bacteria that are also present in the wider environment. Environments such as soil 
have been found to contain environmental bacteria that can subsist on clinically relevant 
antibiotics, some of these resistant bacteria being closely related to human pathogens 
(Dantas et al., 2008). However, a recent study of soil bacteria has shown that the transfer of 
ARGs between environmental bacteria may not occur as readily as between pathogenic 
bacteria and that the composition of bacterial communities may play a greater role in ARG 
load than compared to horizontal gene transfer (Forsberg et al., 2014). Despite this, it is 
conceivable that a common core set of ARGs (and also pathogens and MGEs) exists in both 
the environment and the clinic setting and that these ARGs are likely to be able to be 
transferred between taxonomic groups. Indeed the work reported in this dissertation has 
demonstrated that there is definite overlap between the ARGs and pathogens identified in 
numerous different environmental samples. If it is the case that a core set of ARGs exist, it 
would be extremely beneficial to the monitoring of ARG dissemination if these core ARGs 
were incorporated into environmental frameworks as a set of sentinel ARGs. For example, 
the wide range of ARGs identified within the human gut microbiota may contain good 
candidates for sentinel ARGs (Hu et al., 2013); which could be cross-referenced to those 
identified in environmental studies and used to create a set of sentinel ARGs that are likely to 
pose a risk to the emergence of resistance. 
 
6.2.2.3. Incorporating additional factors and technologies into the 
framework 
There are several key aspects of the environmental framework for monitoring ARG 
dissemination in the environment that will require further development in future iterations of 
the framework. Firstly, the sequencing and bioinformatics technologies utilised in the 
framework are advancing rapidly, resulting in faster and more sophisticated analysis 
techniques. For example, single molecule sequencing technologies are being considered for 
isolate diagnostic testing and ARG detection due to the relatively high speed and low cost, 
although they are currently unsuitable for detecting SNP based resistances (Judge et al., 
2015). The arrival of new technologies will need to be continually assessed for their 
  123 
relevance to ARG monitoring and incorporated as necessary. This increase in technological 
power will also result in an increase in the quantity of data being generated, this too will have 
to be addressed and appropriate data handling and storage capabilities will be vital. 
With these changes in mind, the bioinformatic ARG detection tool (SEAR) that was 
developed in this work was designed to be scalable in terms of its server requirement and the 
ARG databases it utilises; the initial clustering database is static but can be upgraded 
whereas the ARG cross-referencing databases are regularly updated and SEAR utilises the 
current versions every search instance. Building on previous discussion points in this chapter 
and Chapter 2, future developments of SEAR that would improve the monitoring of ARGs 
include the inclusion of subtraction databases and regular database curation. Subtraction 
databases could be used to remove ARGs that do not have clinical importance or to separate 
ARGs based on their associated MGEs. This would contribute to assigning significance to the 
ARP of samples. 
The current drawbacks to the implementation of SEAR are centred on the restriction to only 
identify horizontally acquired ARGs. The inability to identify SNP based ARGs may have a 
large impact on the ARP categorisation of samples due to significant causes of antimicrobial 
resistance being omitted. For instance, the levels of quinolone ARGs and ARG transcripts in 
the hospital effluent of Chapter 4 do not seem to reflect the high quinolone usage and the 
presence of quinolone antimicrobials; the fact that a significant source of quinolone 
resistance is a result of SNP-based gyrase mutations (Jacoby, 2005) that are not detected by 
SEAR may thus account for this discrepancy. 
It is possible for SEAR to be upgraded to search for SNP-based ARGs, through the use of 
variant calling and incorporation of a SNP-based ARG database but this would require 
sufficient sequencing depth to ensure that SNPs are genuine and not a result of sequencing 
errors or poor coverage, which is a goal particularly difficult to achieve within metagenomic 
samples. SEAR also does not identify novel ARGs, although it can identify genes that are 
divergent from known ARGs through lower annotation stringency settings but these may not 
be bona fide ARGs. This raises the point that environmental samples may contain novel 
ARGs with clinical significance and thus should be monitored. To investigate the occurrence 
of novel ARGs, techniques such as functional metagenomics could be used to identify novel 
ARGs that could subsequently be incorporated into existing databases (Perron et al., 2015). 
It is also hoped that the culture and sensitivity testing proposed in the framework might add to 
this possibility. Metagenomic techniques could also be augmented with other technologies 
  124 
such as proteomics in order to characterise ARGs and determine if the encoded proteins 
actually confer phenotypic resistance (Fouhy et al., 2015). 
 
6.2.2.4. Problems remaining 
Despite the massive advancements that sequencing technologies allow for in the field of 
ARG dissemination and tracking, there are some questions that are not possible to 
investigate with these technologies. As exemplified in the work of this dissertation, 
metatranscriptomics is a powerful technique to identify ARGs that are expressed in the 
environment. However, metabolically inactive bacteria would not be accounted for by 
metatranscriptomics and to investigate what role inactive bacteria may have in ARG 
dissemination, techniques such as the re-activation and culturing of bacterial endospores are 
appropriate. Thus, a combination of technologies and experimental approaches is required to 
form a clear idea of what is contributing to ARG dissemination. This is especially true in the 
use of metagenomics as the sole technology in ARG monitoring. Metagenomics is a field that 
has received massive advancements in recent years (National Center for Biotechnology 
Information, 2007, Davenport and Tummler, 2013), moving from 16S amplicon-based 
community profiling to being able to extract whole genomes from metagenomic sequencing 
data (Albertsen et al., 2013). However, there are also shortfalls to metagenomics that are yet 
to be addressed. For instance, metagenomes may underrepresent certain groups of bacteria, 
such as the under-detection of Firmicutes observed in a recent study that attributes 
methodological biases to the low retrieval of endospore-forming bacteria (Filippidou et al., 
2015). This is an important finding in light of the potential role of endospore-forming bacteria 
as persistent reservoirs of horizontally mobilisable ARGs in these sample types, as shown in 
the results reported in Chapter 5. In addition to methodological biases, other metagenomic 
studies have postulated database limitations as a possible cause for the underrepresentation 
of endospore-forming bacteria (For example, sporulation genes may be evolutionary distinct 
and not accounted for, or strategies for long-term cell survival may differ from known 
endospore pathways) (Kawai et al., 2015). 
The main advantage of the environmental framework proposed in this dissertation shown in 
Figure 6.1, as well as its improvement on other proposed environmental frameworks (Port et 
al., 2014), is that it does not rely on just one technology or experimental technique to assign 
ARP to samples.  
  125 
Following on from the suggested inclusion of additional technologies in environmental 
frameworks for monitoring ARG dissemination, another valuable inclusion for the future 
development of this framework is the use of mathematical modelling to understand 
antimicrobial resistance (Opatowski et al., 2011). For instance, mathematical modelling can 
be used to assess the impact of antimicrobial exposure and treatment duration on the 
excretion level of antimicrobial resistant bacteria (Nguyen et al., 2014b); this research could 
be used to provide insight into ARG dissemination and be scaled to apply to effluents as a 
whole. 
 
6.3. Overall conclusions of this work 
To conclude, this dissertation has established the basis for an environmental framework to 
monitor ARG dissemination and suggested the means to facilitate the quantification of 
antimicrobial resistance potential in environmental samples. This work offers an important 
step toward the routine monitoring of effluents entering the environment and the provision of 
antimicrobial resistance risk assessments. 
On a more specific level, this work has contributed a freely available analysis tool for the 
detection of ARGs to the scientific community. The work has also contributed several high 
quality linked metagenomes and metatranscriptomes to publically available databases 
complete with extensive metadata and ARG annotations, thus providing robust data for the 
scientific community to utilise and contribute to further understanding of microbial 
communities in the environment. 
In terms of the scientific findings produced by the work described in this dissertation, it has 
been shown that ARGs are present in several different effluents that are entering a single 
aquatic environment. These ARGs are more abundant than in background samples of the 
aquatic environment and there is apparent seasonal variation in the abundance of ARGs 
across the multiple sampling points. In addition to ARGs, the effluents have also been shown 
to contain MGEs and pathogenic bacteria as well as antimicrobial compounds. Furthermore, 
the ARGs were shown to be expressed in the effluent samples and a possible factor driving 
this expression in the effluent of a hospital was demonstrated to be the antimicrobial usage of 
the hospital. Finally, the work also found ARGs to be persisting in the environment as a result 
of carriage by endospore-forming bacteria.  
The work described in this dissertation has contributed to the potential for the development of 
an environmental framework to monitor ARG dissemination in the environment. This 
  126 
framework utilises the experimental techniques employed in the results chapters to determine 
the antimicrobial resistance potential of effluent samples with a view to contributing to 
antimicrobial resistance risk assessments. With the current drive to incorporate 
environmental research findings into applications with clinical benefit, it is hoped that this 
dissertation provides a useful step toward achieving this goal. 
 
  127 
References 
ALBERTSEN, M., HUGENHOLTZ, P., SKARSHEWSKI, A., NIELSEN, K. L., TYSON, G. W. 
& NIELSEN, P. H. 2013. Genome sequences of rare, uncultured bacteria obtained by 
differential coverage binning of multiple metagenomes. Nature biotechnology, 31, 533-8. 
ALEKSHUN, M. N. & LEVY, S. B. 2007. Molecular Mechanisms of Antibacterial Multidrug 
Resistance. Cell, 128, 1037-1050. 
ALLEN, H. K., DONATO, J., WANG, H. H., CLOUD-HANSEN, K. A., DAVIES, J. & 
HANDELSMAN, J. 2010. Call of the wild: antibiotic resistance genes in natural environments. 
Nature reviews. Microbiology, 8, 251-9. 
ALTSCHUL, S. F., GISH, W., MILLER, W., MYERS, E. W. & LIPMAN, D. J. 1990. Basic local 
alignment search tool. Journal of Molecular Biology, 215, 403-410. 
ASHIRU-OREDOPE, D. & HOPKINS, S. 2013. Antimicrobial stewardship: English 
Surveillance Programme for Antimicrobial Utilization and Resistance (ESPAUR). J Antimicrob 
Chemother, 68, 2421-3. 
AUERBACH, E. A., SEYFRIED, E. E. & MCMAHON, K. D. 2007. Tetracycline resistance 
genes in activated sludge wastewater treatment plants. Water research, 41, 1143-51. 
BAKER-AUSTIN, C., WRIGHT, M. S., STEPANAUSKAS, R. & MCARTHUR, J. V. 2006. Co-
selection of antibiotic and metal resistance. Trends in microbiology, 14, 176-82. 
BALCAZAR, J. L. 2014. Bacteriophages as Vehicles for Antibiotic Resistance Genes in the 
Environment. PLoS Pathogens, 10, e1004219. 
BANNAM, T. L., YAN, X.-X., HARRISON, P. F., SEEMANN, T., KEYBURN, A. L., 
STUBENRAUCH, C., WEERAMANTRI, L. H., CHEUNG, J. K., MCCLANE, B. A., BOYCE, J. 
D., MOORE, R. J. & ROOD, J. I. 2011. Necrotic Enteritis-Derived Clostridium perfringens 
Strain with Three Closely Related Independently Conjugative Toxin and Antibiotic Resistance 
Plasmids. mBio, 2, e00190-11. 
BARAN, R. H. & KO, H. 2008. Detecting Horizontally Transferred and Essential Genes 
Based on Dinucleotide Relative Abundance. DNA Research: An International Journal for 
Rapid Publication of Reports on Genes and Genomes, 15, 267-276. 
BENGTSSON, J., HARTMANN, M., UNTERSEHER, M., VAISHAMPAYAN, P., 
ABARENKOV, K., DURSO, L., BIK, E. M., GAREY, J. R., ERIKSSON, K. M. & NILSSON, R. 
H. 2012. Megraft: a software package to graft ribosomal small subunit (16S/18S) fragments 
onto full-length sequences for accurate species richness and sequencing depth analysis in 
  128 
pyrosequencing-length metagenomes and similar environmental datasets. Res Microbiol, 
163, 407-12. 
BENGTSSON-PALME, J., BOULUND, F., FICK, J., KRISTIANSSON, E. & LARSSON, J. 
2014. Shotgun metagenomics reveals a wide array of antibiotic resistance genes and mobile 
elements in a polluted lake in India. Frontiers in Microbiology, 5. 
BENGTSSON-PALME, J., HARTMANN, M., ERIKSSON, K. M., PAL, C., THORELL, K., 
LARSSON, D. G. & NILSSON, R. H. 2015. Metaxa2: improved identification and taxonomic 
classification of small and large subunit rRNA in metagenomic data. Mol Ecol Resour. 
BERENDONK, T. U., MANAIA, C. M., MERLIN, C., FATTA-KASSINOS, D., CYTRYN, E., 
WALSH, F., BURGMANN, H., SORUM, H., NORSTROM, M., PONS, M.-N., KREUZINGER, 
N., HUOVINEN, P., STEFANI, S., SCHWARTZ, T., KISAND, V., BAQUERO, F. & 
MARTINEZ, J. L. 2015. Tackling antibiotic resistance: the environmental framework. Nat Rev 
Micro, 13, 310-317. 
BERGLUND, B., FICK, J. & LINDGREN, P. E. 2015. Urban wastewater effluent increases 
antibiotic resistance gene concentrations in a receiving northern European river. Environ 
Toxicol Chem, 34, 192-6. 
BROUWER, M. S. M., ROBERTS, A. P., HUSSAIN, H., WILLIAMS, R. J., ALLAN, E. & 
MULLANY, P. 2013. Horizontal gene transfer converts non-toxigenic Clostridium difficile 
strains into toxin producers. Nat Commun, 4. 
BROUWER, M. S. M., WARBURTON, P. J., ROBERTS, A. P., MULLANY, P. & ALLAN, E. 
2011. Genetic Organisation, Mobility and Predicted Functions of Genes on Integrated, Mobile 
Genetic Elements in Sequenced Strains of Clostridium difficile. PLoS ONE, 6, e23014. 
BRUCHMANN, J., KIRCHEN, S. & SCHWARTZ, T. 2013. Sub-inhibitory concentrations of 
antibiotics and wastewater influencing biofilm formation and gene expression of multi-
resistant Pseudomonas aeruginosa wastewater isolates. Environ Sci Pollut Res Int, 20, 3539-
49. 
BRUL, S., BASSETT, J., COOK, P., KATHARIOU, S., MCCLURE, P., JASTI, P. R. & BETTS, 
R. 2012. ‘Omics’ technologies in quantitative microbial risk assessment. Trends in Food 
Science & Technology. 
CABELLO, F. C., GODFREY, H. P., TOMOVA, A., IVANOVA, L., DOLZ, H., MILLANAO, A. & 
BUSCHMANN, A. H. 2013. Antimicrobial use in aquaculture re-examined: its relevance to 
antimicrobial resistance and to animal and human health. Environmental microbiology, 15, 
1917-42. 
  129 
CANICA, M., MANAGEIRO, V., JONES-DIAS, D., CLEMENTE, L., GOMES-NEVES, E., 
POETA, P., DIAS, E. & FERREIRA, E. 2015. Current perspectives on the dynamics of 
antibiotic resistance in different reservoirs. Res Microbiol, 166, 594-600. 
CANO, R. & BORUCKI, M. 1995. Revival and identification of bacterial spores in 25- to 40-
million-year-old Dominican amber. Science, 268, 1060-1064. 
CANTON, R. & COQUE, T. M. 2006. The CTX-M beta-lactamase pandemic. Curr Opin 
Microbiol, 9, 466-75. 
CAPLIN, J. L., HANLON, G. W. & TAYLOR, H. D. 2008. Presence of vancomycin and 
ampicillin-resistant Enterococcus faecium of epidemic clonal complex-17 in wastewaters from 
the south coast of England. Environmental microbiology, 10, 885-92. 
CAPORASO, J. G., KUCZYNSKI, J., STOMBAUGH, J., BITTINGER, K., BUSHMAN, F. D., 
COSTELLO, E. K., FIERER, N., PENA, A. G., GOODRICH, J. K., GORDON, J. I., HUTTLEY, 
G. A., KELLEY, S. T., KNIGHTS, D., KOENIG, J. E., LEY, R. E., LOZUPONE, C. A., 
MCDONALD, D., MUEGGE, B. D., PIRRUNG, M., REEDER, J., SEVINSKY, J. R., 
TURNBAUGH, P. J., WALTERS, W. A., WIDMANN, J., YATSUNENKO, T., ZANEVELD, J. & 
KNIGHT, R. 2010. QIIME allows analysis of high-throughput community sequencing data. 
Nat Methods, 7, 335-6. 
CARVER, T. J., RUTHERFORD, K. M., BERRIMAN, M., RAJANDREAM, M. A., BARRELL, 
B. G. & PARKHILL, J. 2005. ACT: the Artemis Comparison Tool. Bioinformatics, 21, 3422-3. 
CASE, R. J., BOUCHER, Y., DAHLLÖF, I., HOLMSTRÖM, C., DOOLITTLE, W. F. & 
KJELLEBERG, S. 2007. Use of 16S rRNA and rpoB Genes as Molecular Markers for 
Microbial Ecology Studies. Applied and Environmental Microbiology, 73, 278-288. 
CENTERS FOR DISEASE CONTROL AND PREVENTION. 2013. Antibiotic Resistance 
Threats in the United States, 2013. 
CHARPENTIER, X., KAY, E., SCHNEIDER, D. & SHUMAN, H. A. 2011. Antibiotics and UV 
radiation induce competence for natural transformation in Legionella pneumophila. Journal of 
bacteriology, 193, 1114-21. 
CHEN, B., YANG, Y., LIANG, X., YU, K., ZHANG, T. & LI, X. 2013. Metagenomic profiles of 
antibiotic resistance genes (ARGs) between human impacted estuary and deep ocean 
sediments. Environ Sci Technol, 47, 12753-60. 
CLINGENPEEL, S., CLUM, A., SCHWIENTEK, P., RINKE, C. & WOYKE, T. 2015. 
Reconstructing each cell’s genome within complex microbial communities - dream or reality? 
Frontiers in Microbiology, 5. 
  130 
COIA, J. E. 2009. What is the role of antimicrobial resistance in the new epidemic of 
Clostridium difficile? Int J Antimicrob Agents, 33 Suppl 1, S9-12. 
COLOMER-LLUCH, M., JOFRE, J. & MUNIESA, M. 2011. Antibiotic Resistance Genes in the 
Bacteriophage DNA Fraction of Environmental Samples. PLoS ONE, 6, e17549. 
COURVALIN, P. 2006. Vancomycin resistance in gram-positive cocci. Clin Infect Dis, 42 
Suppl 1, S25-34. 
COX, G. & WRIGHT, G. D. 2013. Intrinsic antibiotic resistance: mechanisms, origins, 
challenges and solutions. Int J Med Microbiol, 303, 287-92. 
D'COSTA, V. M., MCGRANN, K. M., HUGHES, D. W. & WRIGHT, G. D. 2006. Sampling the 
antibiotic resistome. Science, 311, 374-7. 
DANCER, D., BAKER-AUSTIN, C., LOWTHER, J. A., HARTNELL, R. E., LEES, D. N. & 
ROBERTS, L. O. 2014. Development and Integration of Quantitative Real-Time PCR 
Methods for Detection of Mitochondrial DNA and Methanobrevibacter smithii nifH Gene as 
Novel Microbial Source Tracking Tools. Environmental Forensics, 15, 256-264. 
DANTAS, G., SOMMER, M. O., OLUWASEGUN, R. D. & CHURCH, G. M. 2008. Bacteria 
subsisting on antibiotics. Science, 320, 100-3. 
DAVENPORT, C. F. & TUMMLER, B. 2013. Advances in computational analysis of 
metagenome sequences. Environmental microbiology, 15, 1-5. 
DAVIES, J. & DAVIES, D. 2010a. Origins and evolution of antibiotic resistance. Microbiology 
and molecular biology reviews : MMBR, 74, 417-33. 
DAVIES, J. & DAVIES, D. 2010b. Origins and evolution of antibiotic resistance. Microbiology 
and molecular biology reviews, 74, 417-433. 
DEBAST, S. B., VAN LEENGOED, L. A., GOORHUIS, A., HARMANUS, C., KUIJPER, E. J. 
& BERGWERFF, A. A. 2009. Clostridium difficile PCR ribotype 078 toxinotype V found in 
diarrhoeal pigs identical to isolates from affected humans. Environ Microbiol, 11, 505-11. 
DECOUSSER, J. W., POIREL, L. & NORDMANN, P. 2001. Characterization of a 
chromosomally encoded extended-spectrum class A beta-lactamase from Kluyvera 
cryocrescens. Antimicrob Agents Chemother, 45, 3595-8. 
DEGNAN, P. H. & OCHMAN, H. 2012. Illumina-based analysis of microbial community 
diversity. ISME J, 6, 183-194. 
DEVARAJAN, N., LAFFITE, A., GRAHAM, N. D., MEIJER, M., PRABAKAR, K., MUBEDI, J. 
I., ELONGO, V., MPIANA, P. T., IBELINGS, B. W., WILDI, W. & POTE, J. 2015. 
Accumulation of clinically relevant antibiotic-resistance genes, bacterial load, and metals in 
freshwater lake sediments in central europe. Environ Sci Technol, 49, 6528-37. 
  131 
DIEKEMA, D. J. & PFALLER, M. A. 2013. Rapid Detection of Antibiotic-Resistant Organism 
Carriage for Infection Prevention. Clinical Infectious Diseases, 56, 1614-1620. 
DOLLIVER, H. & GUPTA, S. 2008. Antibiotic Losses in Leaching and Surface Runoff from 
Manure-Amended Agricultural Land All rights reserved. No part of this periodical may be 
reproduced or transmitted in any form or by any means, electronic or mechanical, including 
photocopying, recording, or any information storage and retrieval system, without permission 
in writing from the publisher. J. Environ. Qual., 37, 1227-1237. 
EDGAR, R. C. 2010. Search and clustering orders of magnitude faster than BLAST. 
Bioinformatics, 26, 2460-1. 
ELIOPOULOS, G. M., COSGROVE, S. E. & CARMELI, Y. 2003. The Impact of Antimicrobial 
Resistance on Health and Economic Outcomes. Clinical Infectious Diseases, 36, 1433-1437. 
ESCUDERO, J. A., LOOT, C., NIVINA, A. & MAZEL, D. 2015. The Integron: Adaptation On 
Demand. Microbiol Spectr, 3, Mdna3-0019-2014. 
ESPY, M. J., UHL, J. R., SLOAN, L. M., BUCKWALTER, S. P., JONES, M. F., VETTER, E. 
A., YAO, J. D., WENGENACK, N. L., ROSENBLATT, J. E., COCKERILL, F. R., 3RD & 
SMITH, T. F. 2006. Real-time PCR in clinical microbiology: applications for routine laboratory 
testing. Clinical microbiology reviews, 19, 165-256. 
FILIPPIDOU, S., JUNIER, T., WUNDERLIN, T., LO, C.-C., LI, P.-E., CHAIN, P. S. & JUNIER, 
P. 2015. Under-detection of endospore-forming Firmicutes in metagenomic data. 
Computational and Structural Biotechnology Journal, 13, 299-306. 
FORSBERG, K. J., PATEL, S., GIBSON, M. K., LAUBER, C. L., KNIGHT, R., FIERER, N. & 
DANTAS, G. 2014. Bacterial phylogeny structures soil resistomes across habitats. Nature, 
509, 612-616. 
FORSBERG, K. J., REYES, A., WANG, B., SELLECK, E. M., SOMMER, M. O. A. & 
DANTAS, G. 2012. The shared antibiotic resistome of soil bacteria and human pathogens. 
Science, 337, 1107-1111. 
FORSLUND, K., SUNAGAWA, S., KULTIMA, J. R., MENDE, D. R., ARUMUGAM, M., 
TYPAS, A. & BORK, P. 2013. Country-specific antibiotic use practices impact the human gut 
resistome. Genome Research, 23, 1163-1169. 
FOUHY, F., STANTON, C., COTTER, P. D., HILL, C. & WALSH, F. 2015. Proteomics as the 
final step in the functional metagenomics study of antimicrobial resistance. Front Microbiol, 6, 
172. 
FROST, L. S., LEPLAE, R., SUMMERS, A. O. & TOUSSAINT, A. 2005. Mobile genetic 
elements: the agents of open source evolution. Nat Rev Micro, 3, 722-732. 
  132 
FUENTEFRIA, D. B., FERREIRA, A. E. & COR√S√£O, G. 2011. Antibiotic-resistant 
Pseudomonas aeruginosa from hospital wastewater and superficial water: Are they 
genetically related? Journal of environmental management, 92, 250-255. 
GALIMAND, M., GUIYOULE, A., GERBAUD, G., RASOAMANANA, B., CHANTEAU, S., 
CARNIEL, E. & COURVALIN, P. 1997. Multidrug resistance in Yersinia pestis mediated by a 
transferable plasmid. N Engl J Med, 337, 677-80. 
GALPERIN, M. Y. 2013. Genome Diversity of Spore-Forming Firmicutes. Microbiology 
spectrum, 1, TBS-0015-2012. 
GASC, C., RIBIERE, C., PARISOT, N., BEUGNOT, R., DEFOIS, C., PETIT-BIDERRE, C., 
BOUCHER, D., PEYRETAILLADE, E. & PEYRET, P. 2015. Capturing prokaryotic dark 
matter genomes. Res Microbiol. 
GAZE, W. H., KRONE, S. M., LARSSON, J. D. G., LI, X.-Z., ROBINSON, J. A., SIMONET, 
P., SMALLA, K., TIMINOUNI, M., TOPP, E., WELLINGTON, E. M. H., WRIGHT, G. D. & 
ZHU, Y.-G. 2013. Influence of Humans on Evolution and Mobilization of Environmental 
Antibiotic Resistome. Emerging Infectious Disease journal, 19. 
GENOME REFERENCE CONSORTIUM 2009. hg19. In: GENOME REFERENCE 
CONSORTIUM (ed.). UCSC. 
GHACHI, M. E., DERBISE, A., BOUHSS, A. & MENGIN-LECREULX, D. 2005. Identification 
of Multiple Genes Encoding Membrane Proteins with Undecaprenyl Pyrophosphate 
Phosphatase (UppP) Activity in Escherichia coli. Journal of Biological Chemistry, 280, 18689-
18695. 
GILLESPIE, J. J., WATTAM, A. R., CAMMER, S. A., GABBARD, J. L., SHUKLA, M. P., 
DALAY, O., DRISCOLL, T., HIX, D., MANE, S. P., MAO, C., NORDBERG, E. K., SCOTT, M., 
SCHULMAN, J. R., SNYDER, E. E., SULLIVAN, D. E., WANG, C., WARREN, A., WILLIAMS, 
K. P., XUE, T., SEUNG YOO, H., ZHANG, C., ZHANG, Y., WILL, R., KENYON, R. W. & 
SOBRAL, B. W. 2011. PATRIC: the Comprehensive Bacterial Bioinformatics Resource with a 
Focus on Human Pathogenic Species. Infection and Immunity, 79, 4286-4298. 
GILLINGS, M. R. 2013. Evolutionary consequences of antibiotic use for the resistome, 
mobilome and microbial pangenome. Frontiers in Microbiology, 4. 
GOLET, E. M., ALDER, A. C., HARTMANN, A., TERNES, T. A. & GIGER, W. 2001. Trace 
Determination of Fluoroquinolone Antibacterial Agents in Urban Wastewater by Solid-Phase 
Extraction and Liquid Chromatography with Fluorescence Detection. Analytical Chemistry, 
73, 3632-3638. 
  133 
GOULD, L. H. & LIMBAGO, B. 2010. Clostridium difficile in food and domestic animals: a 
new foodborne pathogen? Clin Infect Dis, 51, 577-82. 
GUARDABASSI, L. & AGERSO, Y. 2006. Genes homologous to glycopeptide resistance 
vanA are widespread in soil microbial communities. FEMS microbiology letters, 259, 221-5. 
GULLBERG, E., CAO, S., BERG, O. G., ILBÄCK, C., SANDEGREN, L., HUGHES, D. & 
ANDERSSON, D. I. 2011. Selection of resistant bacteria at very low antibiotic concentrations. 
PLoS Pathog, 7, e1002158. 
GUPTA, S. K., PADMANABHAN, B. R., DIENE, S. M., LOPEZ-ROJAS, R., KEMPF, M., 
LANDRAUD, L. & ROLAIN, J.-M. 2014. ARG-ANNOT, a New Bioinformatic Tool To Discover 
Antibiotic Resistance Genes in Bacterial Genomes. Antimicrobial Agents and Chemotherapy, 
58, 212-220. 
HAMMITT, M. C., BUESCHEL, D. M., KEEL, M. K., GLOCK, R. D., CUNEO, P., DEYOUNG, 
D. W., REGGIARDO, C., TRINH, H. T. & SONGER, J. G. 2008. A possible role for 
Clostridium difficile in the etiology of calf enteritis. Vet Microbiol, 127, 343-52. 
HARRIS, S., CORMICAN, M. & CUMMINS, E. 2012. The effect of conventional wastewater 
treatment on the levels of antimicrobial-resistant bacteria in effluent: a meta-analysis of 
current studies. Environmental geochemistry and health, 34, 749-62. 
HARRISON, E. M., PATERSON, G. K., HOLDEN, M. T. G., LARSEN, J., STEGGER, M., 
LARSEN, A. R., PETERSEN, A., SKOV, R. L., CHRISTENSEN, J. M., BAK ZEUTHEN, A., 
HELTBERG, O., HARRIS, S. R., ZADOKS, R. N., PARKHILL, J., PEACOCK, S. J. & 
HOLMES, M. A. 2013. Whole genome sequencing identifies zoonotic transmission of MRSA 
isolates with the novel mecA homologue mecC. EMBO molecular medicine, 5, 509-515. 
HE, M., MIYAJIMA, F., ROBERTS, P., ELLISON, L., PICKARD, D. J., MARTIN, M. J., 
CONNOR, T. R., HARRIS, S. R., FAIRLEY, D., BAMFORD, K. B., D'ARC, S., BRAZIER, J., 
BROWN, D., COIA, J. E., DOUCE, G., GERDING, D., KIM, H. J., KOH, T. H., KATO, H., 
SENOH, M., LOUIE, T., MICHELL, S., BUTT, E., PEACOCK, S. J., BROWN, N. M., RILEY, 
T., SONGER, G., WILCOX, M., PIRMOHAMED, M., KUIJPER, E., HAWKEY, P., WREN, B. 
W., DOUGAN, G., PARKHILL, J. & LAWLEY, T. D. 2013. Emergence and global spread of 
epidemic healthcare-associated Clostridium difficile. Nat Genet, 45, 109-13. 
HEIMAN, K. E., KARLSSON, M., GRASS, J., HOWIE, B., KIRKCALDY, R. D., MAHON, B., 
BROOKS, J. T. & BOWEN, A. 2014. Shigella with Decreased Susceptibility to Azithromycin 
Among Men Who Have Sex with Men — United States, 2002–2013. Morbidity and Mortality 
Weekly Report - CDC, 63, 132. 
  134 
HENSON, K. E., LEVINE, M. T., WONG, E. A. & LEVINE, D. P. 2015. Glycopeptide 
antibiotics: evolving resistance, pharmacology and adverse event profile. Expert Rev Anti 
Infect Ther, 13, 1265-78. 
HEUER, H., KROGERRECKLENFORT, E., WELLINGTON, E. M. H., EGAN, S., ELSAS, J. 
D., OVERBEEK, L., COLLARD, J.-M., GUILLAUME, G., KARAGOUNI, A. D., 
NIKOLAKOPOULOU, T. L. & SMALLA, K. 2002. Gentamicin resistance genes in 
environmental bacteria: Prevalence and transfer. Microbiology Ecology, 42, 289-302. 
HIRSCH, R., TERNES, T., HABERER, K. & KRATZ, K.-L. 1999. Occurrence of antibiotics in 
the aquatic environment. Science of The Total Environment, 225, 109-118. 
HOLT, K. E., BAKER, S., WEILL, F. X., HOLMES, E. C., KITCHEN, A., YU, J., SANGAL, V., 
BROWN, D. J., COIA, J. E., KIM, D. W., CHOI, S. Y., KIM, S. H., DA SILVEIRA, W. D., 
PICKARD, D. J., FARRAR, J. J., PARKHILL, J., DOUGAN, G. & THOMSON, N. R. 2012. 
Shigella sonnei genome sequencing and phylogenetic analysis indicate recent global 
dissemination from Europe. Nature genetics, 44, 1056-9. 
HOOPER, D. C. & JACOBY, G. A. 2015. Mechanisms of drug resistance: quinolone 
resistance. Ann N Y Acad Sci. 
HORNECK, G., MOELLER, R., CADET, J., DOUKI, T., MANCINELLI, R. L., NICHOLSON, 
W. L., PANITZ, C., RABBOW, E., RETTBERG, P., SPRY, A., STACKEBRANDT, E., 
VAISHAMPAYAN, P. & VENKATESWARAN, K. J. 2012. Resistance of Bacterial Endospores 
to Outer Space for Planetary Protection Purposes—Experiment PROTECT of the EXPOSE-E 
Mission. Astrobiology, 12, 445-456. 
HU, X., ZHOU, Q. & LUO, Y. 2010. Occurrence and source analysis of typical veterinary 
antibiotics in manure, soil, vegetables and groundwater from organic vegetable bases, 
northern China. Environmental pollution, 158, 2992-8. 
HU, Y., YANG, X., QIN, J., LU, N., CHENG, G., WU, N., PAN, Y., LI, J., ZHU, L., WANG, X., 
MENG, Z., ZHAO, F., LIU, D., MA, J., QIN, N., XIANG, C., XIAO, Y., LI, L., YANG, H., 
WANG, J., YANG, R., GAO, G. F. & ZHU, B. 2013. Metagenome-wide analysis of antibiotic 
resistance genes in a large cohort of human gut microbiota. Nature communications, 4, 2151. 
HUANG, H., WEINTRAUB, A., FANG, H. & NORD, C. E. 2009. Antimicrobial resistance in 
Clostridium difficile. International Journal of Antimicrobial Agents, 34, 516-522. 
HUGHES, V. M. & DATTA, N. 1983. Conjugative plasmids in bacteria of the 'pre-antibiotic' 
era. Nature, 302, 725-6. 
HUMAN MICROBIOME PROJECT 2012. Structure, function and diversity of the healthy 
human microbiome. Nature, 486, 207-14. 
  135 
HUSON, D. H. & SCORNAVACCA, C. 2012. Dendroscope 3: An Interactive Tool for Rooted 
Phylogenetic Trees and Networks. Systematic Biology. 
ILLUMINA. 2013. Illumina MySeq Benchtop Sequencer [Online]. Available: 
http://www.illumina.com. 
INOUYE, M., DASHNOW, H., RAVEN, L.-A., SCHULTZ, M., POPE, B., TOMITA, T., ZOBEL, 
J. & HOLT, K. 2014. SRST2: Rapid genomic surveillance for public health and hospital 
microbiology labs. Genome Medicine, 6, 90. 
JACOBY, G. A. 2005. Mechanisms of Resistance to Quinolones. Clinical Infectious Diseases, 
41, S120-S126. 
JHUNG, M. A., THOMPSON, A. D., KILLGORE, G. E., ZUKOWSKI, W. E., SONGER, G., 
WARNY, M., JOHNSON, S., GERDING, D. N., MCDONALD, L. C. & LIMBAGO, B. M. 2008. 
Toxinotype V Clostridium difficile in humans and food animals. Emerg Infect Dis, 14, 1039-
45. 
JOHNSON, S. & GERDING, D. N. 1998. Clostridium difficile-Associated Diarrhea. Clin Infect 
Dis., 26, 1027-1034. 
JØRGENSEN, S. E. & HALLING-SØRENSEN, B. 2000. Drugs in the environment. 
Chemosphere, 40, 691-699. 
JUDGE, K., HARRIS, S. R., REUTER, S., PARKHILL, J. & PEACOCK, S. J. 2015. Early 
insights into the potential of the Oxford Nanopore MinION for the detection of antimicrobial 
resistance genes. J Antimicrob Chemother, 70, 2775-8. 
KAWAI, M., UCHIYAMA, I., TAKAMI, H. & INAGAKI, F. 2015. Low frequency of endospore-
specific genes in subseafloor sedimentary metagenomes. Environmental Microbiology 
Reports, 7, 341-350. 
KEESSEN, E. C., HENSGENS, M. P., SPIGAGLIA, P., BARBANTI, F., SANDERS, I. M., 
KUIJPER, E. J. & LIPMAN, L. J. 2013. Antimicrobial susceptibility profiles of human and 
piglet Clostridium difficile PCR-ribotype 078. Antimicrob Resist Infect Control, 2, 14. 
KEMPER, N. 2008. Veterinary antibiotics in the aquatic and terrestrial environment. 
Ecological Indicators, 8, 1-13. 
KHAN, G. A., BERGLUND, B., KHAN, K. M., LINDGREN, P. E. & FICK, J. 2013. Occurrence 
and abundance of antibiotics and resistance genes in rivers, canal and near drug formulation 
facilities--a study in Pakistan. PLoS One, 8, e62712. 
KLEYWEGT, S., PILEGGI, V., LAM, Y. M., ELISES, A., PUDDICOMB, A., PURBA, G., DI 
CARO, J. & FLETCHER, T. 2015. The contribution of pharmaceutically-active compounds 
  136 
from healthcare facilities to a receiving sewage treatment plant in Canada. Environ Toxicol 
Chem. 
KNAPP, C. W., MCCLUSKEY, S. N. N. M., SINGH, B. K., CAMPBELL, C. D., HUDSON, G. & 
GRAHAM, D. W. 2011. Antibiotic Resistance Gene Abundances Correlate with Metal and 
Geochemical Conditions in Archived Scottish Soils. PloS one, 6, e27300. 
KNETSCH, C. W., TERVEER, E. M., LAUBER, C., GORBALENYA, A. E., HARMANUS, C., 
KUIJPER, E. J., CORVER, J. & VAN LEEUWEN, H. C. 2012. Comparative analysis of an 
expanded Clostridium difficile reference strain collection reveals genetic diversity and 
evolution through six lineages. Infection, Genetics and Evolution, 12, 1577-1585. 
KOSER, C. U., HOLDEN, M. T., ELLINGTON, M. J., CARTWRIGHT, E. J., BROWN, N. M., 
OGILVY-STUART, A. L., HSU, L. Y., CHEWAPREECHA, C., CROUCHER, N. J., HARRIS, 
S. R., SANDERS, M., ENRIGHT, M. C., DOUGAN, G., BENTLEY, S. D., PARKHILL, J., 
FRASER, L. J., BETLEY, J. R., SCHULZ-TRIEGLAFF, O. B., SMITH, G. P. & PEACOCK, S. 
J. 2012. Rapid whole-genome sequencing for investigation of a neonatal MRSA outbreak. 
The New England journal of medicine, 366, 2267-75. 
KOUASSI, K. A., DADIE, A. T., N'GUESSAN, K. F., DJE, K. M. & LOUKOU, Y. G. 2014. 
Clostridium perfringens and Clostridium difficile in cooked beef sold in Côte d'Ivoire and their 
antimicrobial susceptibility. Anaerobe, 28, 90-94. 
KRISTIANSSON, E., FICK, J., JANZON, A., GRABIC, R., RUTGERSSON, C., 
WEIJDEGÅRD, B., SÖDERSTRÖM, H. & LARSSON, D. G. J. 2011. Pyrosequencing of 
Antibiotic-Contaminated River Sediments Reveals High Levels of Resistance and Gene 
Transfer Elements. PloS one, 6, e17038. 
KUCZYNSKI, J., LAUBER, C. L., WALTERS, W. A., PARFREY, L. W., CLEMENTE, J. C., 
GEVERS, D. & KNIGHT, R. 2012. Experimental and analytical tools for studying the human 
microbiome. Nat Rev Genet, 13, 47-58. 
KÜMMERER, K. & HENNINGER, A. 2003. Promoting resistance by the emission of 
antibiotics from hospitals and households into effluent. Clinical Microbiology and Infection, 9, 
1203-1214. 
LAHLAOUI, H., BEN HAJ KHALIFA, A. & BEN MOUSSA, M. 2014. Epidemiology of 
Enterobacteriaceae producing CTX-M type extended spectrum beta-lactamase (ESBL). Med 
Mal Infect, 44, 400-4. 
LAPARA, T. M., BURCH, T. R., MCNAMARA, P. J., TAN, D. T., YAN, M. & EICHMILLER, J. 
J. 2011. Tertiary-Treated Municipal Wastewater is a Significant Point Source of Antibiotic 
  137 
Resistance Genes into Duluth-Superior Harbor. Environmental science & technology, 45, 
9543-9549. 
LARSSON, D. G., DE PEDRO, C. & PAXEUS, N. 2007. Effluent from drug manufactures 
contains extremely high levels of pharmaceuticals. J Hazard Mater, 148, 751-5. 
LAWLEY, T. D., CLARE, S., WALKER, A. W., GOULDING, D., STABLER, R. A., 
CROUCHER, N., MASTROENI, P., SCOTT, P., RAISEN, C., MOTTRAM, L., 
FAIRWEATHER, N. F., WREN, B. W., PARKHILL, J. & DOUGAN, G. 2009. Antibiotic 
Treatment of Clostridium difficile Carrier Mice Triggers a Supershedder State, Spore-
Mediated Transmission, and Severe Disease in Immunocompromised Hosts. Infection and 
Immunity, 77, 3661-3669. 
LAWLEY, T. D., CLARE, S., WALKER, A. W., STARES, M. D., CONNOR, T. R., RAISEN, C., 
GOULDING, D., RAD, R., SCHREIBER, F., BRANDT, C., DEAKIN, L. J., PICKARD, D. J., 
DUNCAN, S. H., FLINT, H. J., CLARK, T. G., PARKHILL, J. & DOUGAN, G. 2012. Targeted 
Restoration of the Intestinal Microbiota with a Simple, Defined Bacteriotherapy Resolves 
Relapsing Clostridium difficile Disease in Mice. PLoS Pathog, 8, e1002995. 
LAXMINARAYAN, R., DUSE, A., WATTAL, C., ZAIDI, A. K. M., WERTHEIM, H. F. L., 
SUMPRADIT, N., VLIEGHE, E., HARA, G. L., GOULD, I. M., GOOSSENS, H., GREKO, C., 
SO, A. D., BIGDELI, M., TOMSON, G., WOODHOUSE, W., OMBAKA, E., PERALTA, A. Q., 
QAMAR, F. N., MIR, F., KARIUKI, S., BHUTTA, Z. A., COATES, A., BERGSTROM, R., 
WRIGHT, G. D., BROWN, E. D. & CARS, O. 2013. Antibiotic resistance—the need for global 
solutions. The Lancet Infectious Diseases, 13, 1057-1098. 
LEVY, S. B. & MARSHALL, B. 2004. Antibacterial resistance worldwide: causes, challenges 
and responses. Nature medicine, 10, S122-9. 
LEWIN, A., JOHANSEN, J., WENTZEL, A., KOTLAR, H. K., DRABLOS, F. & VALLA, S. 
2013. The microbial communities in two apparently physically separated deep subsurface oil 
reservoirs show extensive DNA sequence similarities. Environmental microbiology. 
LI, D., YANG, M., HU, J., REN, L., ZHANG, Y. & LI, K. 2008. Determination and fate of 
oxytetracycline and related compounds in oxytetracycline production wastewater and the 
receiving river. Environ Toxicol Chem, 27, 80-6. 
LI, H. 2011. A statistical framework for SNP calling, mutation discovery, association mapping 
and population genetical parameter estimation from sequencing data. Bioinformatics, 27, 
2987-93. 
LI, H. 2013. Aligning sequence reads, clone sequences and assembly contigs with BWA-
MEM. arXiv. 
  138 
LI, H. & DURBIN, R. 2009. Fast and accurate short read alignment with Burrows-Wheeler 
transform. Bioinformatics, 25, 1754-60. 
LI, J., ADAMS, V., BANNAM, T. L., MIYAMOTO, K., GARCIA, J. P., UZAL, F. A., ROOD, J. I. 
& MCCLANE, B. A. 2013. Toxin Plasmids of Clostridium perfringens. Microbiology and 
Molecular Biology Reviews : MMBR, 77, 208-233. 
LI, J., CHENG, W., XU, L., STRONG, P. J. & CHEN, H. 2015a. Antibiotic-resistant genes and 
antibiotic-resistant bacteria in the effluent of urban residential areas, hospitals, and a 
municipal wastewater treatment plant system. Environmental Science and Pollution 
Research, 22, 4587-4596. 
LI, J., WANG, T., SHAO, B., SHEN, J., WANG, S. & WU, Y. 2012. Plasmid-mediated 
quinolone resistance genes and antibiotic residues in wastewater and soil adjacent to swine 
feedlots: potential transfer to agricultural lands. Environmental health perspectives, 120, 
1144-9. 
LI, X. Z., PLESIAT, P. & NIKAIDO, H. 2015b. The challenge of efflux-mediated antibiotic 
resistance in Gram-negative bacteria. Clin Microbiol Rev, 28, 337-418. 
LINARES, J. F., GUSTAFSSON, I., BAQUERO, F. & MARTINEZ, J. L. 2006. Antibiotics as 
intermicrobial signaling agents instead of weapons. Proceedings of the National Academy of 
Sciences of the United States of America, 103, 19484-9. 
LIU, B. & POP, M. 2009. ARDB‚ Antibiotic Resistance Genes Database. Nucleic Acids 
Research, 37, D443-D447. 
LU, Z., NA, G., GAO, H., WANG, L., BAO, C. & YAO, Z. 2015. Fate of sulfonamide 
resistance genes in estuary environment and effect of anthropogenic activities. Sci Total 
Environ, 527-528C, 429-438. 
LUO, Y., MAO, D., RYSZ, M., ZHOU, Q., ZHANG, H., XU, L. & J. J. ALVAREZ, P. 2010. 
Trends in Antibiotic Resistance Genes Occurrence in the Haihe River, China. Environmental 
science & technology, 44, 7220-7225. 
LYRAS, D., O/'CONNOR, J. R., HOWARTH, P. M., SAMBOL, S. P., CARTER, G. P., 
PHUMOONNA, T., POON, R., ADAMS, V., VEDANTAM, G., JOHNSON, S., GERDING, D. 
N. & ROOD, J. I. 2009. Toxin B is essential for virulence of Clostridium difficile. Nature, 458, 
1176-1179. 
MARSHALL, C. G. & WRIGHT, G. D. 1997. The glycopeptide antibiotic producer 
Streptomyces toyocaensis NRRL 15009 has both D-alanyl-D-alanine and D-alanyl-D-lactate 
ligases. FEMS Microbiol Lett, 157, 295-9. 
  139 
MARSHALL, C. G., ZOLLI, M. & WRIGHT, G. D. 1999. Molecular mechanism of VanHst, an 
alpha-ketoacid dehydrogenase required for glycopeptide antibiotic resistance from a 
glycopeptide producing organism. Biochemistry, 38, 8485-91. 
MARTINEZ, J. L. 2009. Environmental pollution by antibiotics and by antibiotic resistance 
determinants. Environmental pollution, 157, 2893-902. 
MASON, O. U., HAZEN, T. C., BORGLIN, S., CHAIN, P. S., DUBINSKY, E. A., FORTNEY, J. 
L., HAN, J., HOLMAN, H. Y., HULTMAN, J., LAMENDELLA, R., MACKELPRANG, R., 
MALFATTI, S., TOM, L. M., TRINGE, S. G., WOYKE, T., ZHOU, J., RUBIN, E. M. & 
JANSSON, J. K. 2012. Metagenome, metatranscriptome and single-cell sequencing reveal 
microbial response to Deepwater Horizon oil spill. The ISME journal, 6, 1715-27. 
MATHER, A. E., REID, S. W. J., MASKELL, D. J., PARKHILL, J., FOOKES, M. C., HARRIS, 
S. R., BROWN, D. J., COIA, J. E., MULVEY, M. R., GILMOUR, M. W., PETROVSKA, L., DE 
PINNA, E., KURODA, M., AKIBA, M., IZUMIYA, H., CONNOR, T. R., SUCHARD, M. A., 
LEMEY, P., MELLOR, D. J., HAYDON, D. T. & THOMSON, N. R. 2013. Distinguishable 
Epidemics of Multidrug-Resistant Salmonella Typhimurium DT104 in Different Hosts. 
Science, 341, 1514-1517. 
MAZEL, D. 2006. Integrons: agents of bacterial evolution. Nature reviews. Microbiology, 4, 
608-20. 
MCARTHUR, A. G., WAGLECHNER, N., NIZAM, F., YAN, A., AZAD, M. A., BAYLAY, A. J., 
BHULLAR, K., CANOVA, M. J., DE PASCALE, G., EJIM, L., KALAN, L., KING, A. M., 
KOTEVA, K., MORAR, M., MULVEY, M. R., O'BRIEN, J. S., PAWLOWSKI, A. C., PIDDOCK, 
L. J., SPANOGIANNOPOULOS, P., SUTHERLAND, A. D., TANG, I., TAYLOR, P. L., 
THAKER, M., WANG, W., YAN, M., YU, T. & WRIGHT, G. D. 2013. The comprehensive 
antibiotic resistance database. Antimicrobial agents and chemotherapy, 57, 3348-57. 
MCENEFF, G., BARRON, L., KELLEHER, B., PAULL, B. & QUINN, B. 2014. A year-long 
study of the spatial occurrence and relative distribution of pharmaceutical residues in sewage 
effluent, receiving marine waters and marine bivalves. Sci Total Environ, 476-477, 317-26. 
MCMANUS, P. S., STOCKWELL, V. O., SUNDIN, G. W. & JONES, A. L. 2002. Antibiotic use 
in plant agriculture. Annual review of phytopathology, 40, 443-65. 
MEYER, F., PAARMANN, D., D'SOUZA, M., OLSON, R., GLASS, E. M., KUBAL, M., 
PACZIAN, T., RODRIGUEZ, A., STEVENS, R., WILKE, A., WILKENING, J. & EDWARDS, R. 
A. 2008. The metagenomics RAST server - a public resource for the automatic phylogenetic 
and functional analysis of metagenomes. BMC bioinformatics, 9, 386. 
  140 
MILLER, R., MONTOYA, V., GARDY, J., PATRICK, D. & TANG, P. 2013. Metagenomics for 
pathogen detection in public health. Genome Medicine, 5, 81. 
MORTAZAVI, A., WILLIAMS, B. A., MCCUE, K., SCHAEFFER, L. & WOLD, B. 2008. 
Mapping and quantifying mammalian transcriptomes by RNA-Seq. Nature methods, 5, 621-8. 
MOURA, A., SOARES, M., PEREIRA, C., LEITAO, N., HENRIQUES, I. & CORREIA, A. N. 
2009. INTEGRALL: a database and search engine for integrons, integrases and gene 
cassettes. Bioinformatics, 25, 1096-1098. 
MULLANY, P. 2014. Functional metagenomics for the investigation of antibiotic resistance. 
Virulence, 5, 443-7. 
MYERS, G. S. A., RASKO, D. A., CHEUNG, J. K., RAVEL, J., SESHADRI, R., DEBOY, R. 
T., REN, Q., VARGA, J., AWAD, M. M., BRINKAC, L. M., DAUGHERTY, S. C., HAFT, D. H., 
DODSON, R. J., MADUPU, R., NELSON, W. C., ROSOVITZ, M. J., SULLIVAN, S. A., 
KHOURI, H., DIMITROV, G. I., WATKINS, K. L., MULLIGAN, S., BENTON, J., RADUNE, D., 
FISHER, D. J., ATKINS, H. S., HISCOX, T., JOST, B. H., BILLINGTON, S. J., SONGER, J. 
G., MCCLANE, B. A., TITBALL, R. W., ROOD, J. I., MELVILLE, S. B. & PAULSEN, I. T. 
2006. Skewed genomic variability in strains of the toxigenic bacterial pathogen, Clostridium 
perfringens. Genome Research, 16, 1031-1040. 
NACCACHE, S. N., FEDERMAN, S., VEERARAGHAVAN, N., ZAHARIA, M., LEE, D., 
SAMAYOA, E., BOUQUET, J., GRENINGER, A. L., LUK, K.-C., ENGE, B., WADFORD, D. 
A., MESSENGER, S. L., GENRICH, G. L., PELLEGRINO, K., GRARD, G., LEROY, E., 
SCHNEIDER, B. S., FAIR, J. N., MARTÍNEZ, M. A., ISA, P., CRUMP, J. A., DERISI, J. L., 
SITTLER, T., HACKETT, J., MILLER, S. & CHIU, C. Y. 2014. A cloud-compatible 
bioinformatics pipeline for ultrarapid pathogen identification from next-generation sequencing 
of clinical samples. Genome Research, 24, 1180-1192. 
NATIONAL CENTER FOR BIOTECHNOLOGY INFORMATION 2007. The New Science of 
Metagenomics. NCBI. 
NGUYEN, F., STAROSTA, A. L., ARENZ, S., SOHMEN, D., DONHOFER, A. & WILSON, D. 
N. 2014a. Tetracycline antibiotics and resistance mechanisms. Biol Chem, 395, 559-75. 
NGUYEN, T. T., GUEDJ, J., CHACHATY, E., DE GUNZBURG, J., ANDREMONT, A. & 
MENTRÉ, F. 2014b. Mathematical Modeling of Bacterial Kinetics to Predict the Impact of 
Antibiotic Colonic Exposure and Treatment Duration on the Amount of Resistant 
Enterobacteria Excreted. PLoS Comput Biol, 10, e1003840. 
O'BRIEN, F. G., YUI ETO, K., MURPHY, RILEY J. T., FAIRHURST, HEATHER M., 
COOMBS, G. W., GRUBB, W. B. & RAMSAY, J. P. 2015. Origin-of-transfer sequences 
  141 
facilitate mobilisation of non-conjugative antimicrobial-resistance plasmids in Staphylococcus 
aureus. Nucleic Acids Research. 
OH, S., TANDUKAR, M., PAVLOSTATHIS, S. G., CHAIN, P. S. & KONSTANTINIDIS, K. T. 
2013. Microbial community adaptation to quaternary ammonium biocides as revealed by 
metagenomics. Environmental microbiology. 
OLSON, A. B., SILVERMAN, M., BOYD, D. A., MCGEER, A., WILLEY, B. M., PONG-
PORTER, V., DANEMAN, N. & MULVEY, M. R. 2005. Identification of a progenitor of the 
CTX-M-9 group of extended-spectrum beta-lactamases from Kluyvera georgiana isolated in 
Guyana. Antimicrob Agents Chemother, 49, 2112-5. 
OPATOWSKI, L., GUILLEMOT, D., BOELLE, P. Y. & TEMIME, L. 2011. Contribution of 
mathematical modeling to the fight against bacterial antibiotic resistance. Curr Opin Infect 
Dis, 24, 279-87. 
PAGE, A. J., CUMMINS, C. A., HUNT, M., WONG, V. K., REUTER, S., HOLDEN, M. T. G., 
FOOKES, M., FALUSH, D., KEANE, J. A. & PARKHILL, J. 2015. Roary: Rapid large-scale 
prokaryote pan genome analysis. Bioinformatics. 
PAN, J. C., YE, R., WANG, H. Q., XIANG, H. Q., ZHANG, W., YU, X. F., MENG, D. M. & HE, 
Z. S. 2008. Vibrio cholerae O139 multiple-drug resistance mediated by Yersinia pestis 
pIP1202-like conjugative plasmids. Antimicrob Agents Chemother, 52, 3829-36. 
PERRON, G. G., WHYTE, L., TURNBAUGH, P. J., GOORDIAL, J., HANAGE, W. P., 
DANTAS, G. & DESAI, M. M. 2015. Functional characterization of bacteria isolated from 
ancient arctic soil exposes diverse resistance mechanisms to modern antibiotics. PLoS One, 
10, e0069533. 
PHILLIPS, I., CASEWELL, M., COX, T., DE GROOT, B., FRIIS, C., JONES, R., 
NIGHTINGALE, C., PRESTON, R. & WADDELL, J. 2004. Does the use of antibiotics in food 
animals pose a risk to human health? A critical review of published data. The Journal of 
antimicrobial chemotherapy, 53, 28-52. 
PONSTINGL, H. 2014. SMALT. http://www.sanger.ac.uk/resources/software/smalt/ Sanger 
Institute, WelcomeTrust. 
PORT, J. A., CULLEN, A. C., WALLACE, J. C., SMITH, M. N. & FAUSTMAN, E. M. 2014. 
Metagenomic Frameworks for Monitoring Antibiotic Resistance in Aquatic Environments. 
Environmental Health Perspectives, 122, 222-228. 
PRUDEN, A., PEI, R., STORTEBOOM, H. & CARLSON, K. H. 2006. Antibiotic resistance 
genes as emerging contaminants: Studies in northern Colorado. Environmental Science and 
Technology, 40, 7445-7450. 
  142 
PUBLIC HEALTH ENGLAND. 2013. Antibiotic Resistance Monitoring & Reference 
Laboratory (ARMRL) [Online]. Available: http://www.hpa.org.uk/. 
QUAIL, M. A., OTTO, T. D., GU, Y., HARRIS, S. R., SKELLY, T. F., MCQUILLAN, J. A., 
SWERDLOW, H. P. & OYOLA, S. O. 2012. Optimal enzymes for amplifying sequencing 
libraries. Nat Meth, 9, 10-11. 
RAMBAUT, A. 2007. FigTree. http://tree.bio.ed.ac.uk/software/figtree/. 
RASMUSSEN, L. D., ZAWADSKY, C., BINNERUP, S. J., ØREGAARD, G., SØRENSEN, S. 
J. & KROER, N. 2008. Cultivation of Hard-To-Culture Subsurface Mercury-Resistant Bacteria 
and Discovery of New merA Gene Sequences. Applied and Environmental Microbiology, 74, 
3795-3803. 
RELLER, L. B., WEINSTEIN, M. P. & MURDOCH, D. R. 2003. Diagnosis of Legionella 
Infection. Clinical Infectious Diseases, 36, 64-69. 
ROBILOTTI, E. & DERESINSKI, S. 2014. Carbapenemase-producing Klebsiella 
pneumoniae. F1000Prime Rep, 6, 80. 
RODRIGUEZ, C., LANG, L., WANG, A., ALTENDORF, K., GARCIA, F. & LIPSKI, A. 2006. 
Lettuce for human consumption collected in Costa Rica contains complex communities of 
culturable oxytetracycline- and gentamicin-resistant bacteria. Applied and environmental 
microbiology, 72, 5870-6. 
ROOD, J. I. & COLE, S. T. 1991. Molecular genetics and pathogenesis of Clostridium 
perfringens. Microbiological Reviews, 55, 621-648. 
ROWE, W., BAKER, K. S., VERNER-JEFFREYS, D., BAKER-AUSTIN, C., RYAN, J. J., 
MASKELL, D. & PEARCE, G. 2015. Search Engine for Antimicrobial Resistance: A Cloud 
Compatible Pipeline and Web Interface for Rapidly Detecting Antimicrobial Resistance 
Genes Directly from Sequence Data. PLoS ONE, 10, e0133492. 
ROZEN, S. & SKALETSKY, H. J. 1988. Primer3. 
RUPNIK, M., WILCOX, M. H. & GERDING, D. N. 2009. Clostridium difficile infection: new 
developments in epidemiology and pathogenesis. Nat Rev Microbiol, 7, 526-36. 
SACK, D., LYKE, C., MCLAUGHLIN, C. & SUWANVANICHKIJ, V. 2001. Antimicrobial 
resistance in shigellosis, cholera and campylobacteriosis. World Health Organization: 
Department of Communicable Disease Surveillance and Response. 
SAIF, N. & BRAZIER, J. S. 1996. The distribution of Clostridium difficile in the environment of 
South Wales. J Med Microbiol, 45, 133-7. 
  143 
SCHWARTZ, T., KOHNEN, W., JANSEN, B. & OBST, U. 2003. Detection of antibiotic-
resistant bacteria and their resistance genes in wastewater, surface water, and drinking water 
biofilms. FEMS microbiology ecology, 43, 325-335. 
SEBAIHIA, M., WREN, B. W., MULLANY, P., FAIRWEATHER, N. F., MINTON, N., 
STABLER, R., THOMSON, N. R., ROBERTS, A. P., CERDENO-TARRAGA, A. M., WANG, 
H., HOLDEN, M. T., WRIGHT, A., CHURCHER, C., QUAIL, M. A., BAKER, S., BASON, N., 
BROOKS, K., CHILLINGWORTH, T., CRONIN, A., DAVIS, P., DOWD, L., FRASER, A., 
FELTWELL, T., HANCE, Z., HOLROYD, S., JAGELS, K., MOULE, S., MUNGALL, K., 
PRICE, C., RABBINOWITSCH, E., SHARP, S., SIMMONDS, M., STEVENS, K., UNWIN, L., 
WHITHEAD, S., DUPUY, B., DOUGAN, G., BARRELL, B. & PARKHILL, J. 2006. The 
multidrug-resistant human pathogen Clostridium difficile has a highly mobile, mosaic 
genome. Nat Genet, 38, 779-86. 
SEEMANN, T. 2014. Prokka: rapid prokaryotic genome annotation. Bioinformatics, 30, 2068-
9. 
SEGATA, N. 2014. GraPhlAn. https://bitbucket.org/nsegata/graphlan/wiki/Home. 
SEGATA, N., IZARD, J., WALDRON, L., GEVERS, D., MIROPOLSKY, L., GARRETT, W. & 
HUTTENHOWER, C. 2011. Metagenomic biomarker discovery and explanation. Genome 
Biology, 12, R60. 
SEGATA, N., WALDRON, L., BALLARINI, A., NARASIMHAN, V., JOUSSON, O. & 
HUTTENHOWER, C. 2012. Metagenomic microbial community profiling using unique clade-
specific marker genes. Nat Meth, 9, 811-814. 
STALDER, T., BARRAUD, O., CASELLAS, M., DAGOT, C. & PLOY, M. C. 2012. Integron 
involvement in environmental spread of antibiotic resistance. Front Microbiol, 3, 119. 
STAMATAKIS, A. 2006. RAxML-VI-HPC: maximum likelihood-based phylogenetic analyses 
with thousands of taxa and mixed models. Bioinformatics, 22, 2688-90. 
STEWART, E. J. 2012. Growing Unculturable Bacteria. Journal of Bacteriology, 194, 4151-
4160. 
SUETIN, S. V., SHCHERBAKOVA, V. A., CHUVILSKAYA, N. A., RIVKINA, E. M., SUZINA, 
N. E., LYSENKO, A. M. & GILICHINSKY, D. A. 2009. Clostridium tagluense sp. nov., a 
psychrotolerant, anaerobic, spore-forming bacterium from permafrost. International Journal of 
Systematic and Evolutionary Microbiology, 59, 1421-1426. 
SUNDE, M., SIMONSEN, G. S., SLETTEMEAS, J. S., BOCKERMAN, I. & NORSTROM, M. 
2015. Integron, Plasmid and Host Strain Characteristics of Escherichia coli from Humans and 
  144 
Food Included in the Norwegian Antimicrobial Resistance Monitoring Programs. PLoS One, 
10, e0128797. 
SVARA, F. & RANKIN, D. J. 2011. The evolution of plasmid-carried antibiotic resistance. 
BMC Evol Biol, 11, 130. 
TALUKDAR, P. K., OLGUÍN-ARANEDA, V., ALNOMAN, M., PAREDES-SABJA, D. & 
SARKER, M. R. 2015. Updates on the sporulation process in Clostridium species. Research 
in Microbiology, 166, 225-235. 
TANG, S. S., APISARNTHANARAK, A. & HSU, L. Y. 2014. Mechanisms of beta-lactam 
antimicrobial resistance and epidemiology of major community- and healthcare-associated 
multidrug-resistant bacteria. Adv Drug Deliv Rev, 78, 3-13. 
TAO, R., YING, G. G., SU, H. C., ZHOU, H. W. & SIDHU, J. P. 2010. Detection of antibiotic 
resistance and tetracycline resistance genes in Enterobacteriaceae isolated from the Pearl 
rivers in South China. Environmental pollution, 158, 2101-9. 
TONNA, I. & WELSBY, P. D. 2005. Pathogenesis and treatment of Clostridium difficile 
infection. Postgraduate Medical Journal, 81, 367-369. 
TRINGE, S. G., VON MERING, C., KOBAYASHI, A., SALAMOV, A. A., CHEN, K., CHANG, 
H. W., PODAR, M., SHORT, J. M., MATHUR, E. J., DETTER, J. C., BORK, P., 
HUGENHOLTZ, P. & RUBIN, E. M. 2005. Comparative metagenomics of microbial 
communities. Science, 308, 554-7. 
TSAFNAT, G., COPTY, J. & PARTRIDGE, S. R. 2011. RAC: Repository of Antibiotic 
resistance Cassettes. Database, 2011. 
VAN BAMBEKE, F., GLUPCZYNSKI, Y., PLESIAT, P., PECHERE, J. C. & TULKENS, P. M. 
2003. Antibiotic efflux pumps in prokaryotic cells: occurrence, impact on resistance and 
strategies for the future of antimicrobial therapy. The Journal of antimicrobial chemotherapy, 
51, 1055-65. 
WANG, J., MAO, D., MU, Q. & LUO, Y. 2015. Fate and proliferation of typical antibiotic 
resistance genes in five full-scale pharmaceutical wastewater treatment plants. Sci Total 
Environ, 526, 366-73. 
WELLINGTON, E. M. H., BOXALL, A. B. A., CROSS, P., FEIL, E. J., GAZE, W. H., 
HAWKEY, P. M., JOHNSON-ROLLINGS, A. S., JONES, D. L., LEE, N. M., OTTEN, W., 
THOMAS, C. M. & WILLIAMS, A. P. 2013. The role of the natural environment in the 
emergence of antibiotic resistance in Gram-negative bacteria. The Lancet Infectious 
Diseases, 13, 155-165. 
  145 
WILCOX, M. H., MOONEY, L., BENDALL, R., SETTLE, C. D. & FAWLEY, W. N. 2008. A 
case-control study of community-associated Clostridium difficile infection. J Antimicrob 
Chemother, 62, 388-96. 
WOOD, D. E. & SALZBERG, S. L. 2014. Kraken: ultrafast metagenomic sequence 
classification using exact alignments. Genome Biol, 15, R46. 
WORLD HEALTH ORGANISATION 2012. The evolving threat of antimicrobial resistance - 
options for action. In: WHO (ed.). Geneva. 
WRIGHT, G. D. 2007. The antibiotic resistome: the nexus of chemical and genetic diversity. 
Nature reviews. Microbiology, 5, 175-86. 
WRIGHT, M. S., BAKER-AUSTIN, C., LINDELL, A. H., STEPANAUSKAS, R., STOKES, H. 
W. & MCARTHUR, J. V. 2008. Influence of industrial contamination on mobile genetic 
elements: class 1 integron abundance and gene cassette structure in aquatic bacterial 
communities. The ISME journal, 2, 417-28. 
YIM, G., WANG, H. H. & DAVIES, J. 2007. Antibiotics as signalling molecules. Philos Trans 
R Soc Lond B Biol Sci, 362, 1195-200. 
ZANKARI, E., HASMAN, H., COSENTINO, S., VESTERGAARD, M., RASMUSSEN, S., 
LUND, O., AARESTRUP, F. M. & LARSEN, M. V. 2012. Identification of acquired 
antimicrobial resistance genes. The Journal of antimicrobial chemotherapy, 67, 2640-4. 
ZENGLER, K. & PALSSON, B. O. 2012. A road map for the development of community 
systems (CoSy) biology. Nat Rev Microbiol, 10, 366-72. 
ZERBINO, D. R. & BIRNEY, E. 2008. Velvet: Algorithms for de novo short read assembly 
using de Bruijn graphs. Genome Research, 18, 821-829. 
ZHANG, X. X. & ZHANG, T. 2011. Occurrence, abundance, and diversity of tetracycline 
resistance genes in 15 sewage treatment plants across China and other global locations. 
Environmental science & technology, 45, 2598-604. 
ZHANG, X. X., ZHANG, T. & FANG, H. H. 2009. Antibiotic resistance genes in water 
environment. Applied microbiology and biotechnology, 82, 397-414. 
  146 
Appendix 1. Manuscript and source code for SEAR. 
ROWE, W., BAKER, K. S., VERNER-JEFFREYS, D., BAKER-AUSTIN, C., RYAN, J. J., 
MASKELL, D. & PEARCE, G. 2015. Search Engine for Antimicrobial Resistance: A 
Cloud Compatible Pipeline and Web Interface for Rapidly Detecting Antimicrobial 
Resistance Genes Directly from Sequence Data. PLoS ONE, 10, e0133492. 
 
RESEARCH ARTICLE
Search Engine for Antimicrobial Resistance: A
Cloud Compatible Pipeline and Web
Interface for Rapidly Detecting Antimicrobial
Resistance Genes Directly from Sequence
Data
Will Rowe1*, Kate S. Baker2, David Verner-Jeffreys3, Craig Baker-Austin3, Jim J. Ryan4,
Duncan Maskell1, Gareth Pearce1
1 Department of Veterinary Medicine, University of Cambridge, Cambridge, United Kingdom, 2 Wellcome
Trust Sanger Institute, Cambridge, United Kingdom, 3 Centre for Environment, Fisheries and Aquaculture
Science, Weymouth, United Kingdom, 4 Environment, Health and Safety, GlaxoSmithKline, Ware, United
Kingdom
* wpmr2@cam.ac.uk
Abstract
Background
Antimicrobial resistance remains a growing and significant concern in human and veterinary
medicine. Current laboratory methods for the detection and surveillance of antimicrobial
resistant bacteria are limited in their effectiveness and scope. With the rapidly developing
field of whole genome sequencing beginning to be utilised in clinical practice, the ability to
interrogate sequencing data quickly and easily for the presence of antimicrobial resistance
genes will become increasingly important and useful for informing clinical decisions. Addi-
tionally, use of such tools will provide insight into the dynamics of antimicrobial resistance
genes in metagenomic samples such as those used in environmental monitoring.
Results
Here we present the Search Engine for Antimicrobial Resistance (SEAR), a pipeline and
web interface for detection of horizontally acquired antimicrobial resistance genes in raw
sequencing data. The pipeline provides gene information, abundance estimation and the
reconstructed sequence of antimicrobial resistance genes; it also provides web links to
additional information on each gene. The pipeline utilises clustering and read mapping to
annotate full-length genes relative to a user-defined database. It also uses local alignment
of annotated genes to a range of online databases to provide additional information. We
demonstrate SEAR’s application in the detection and abundance estimation of antimicrobial
resistance genes in two novel environmental metagenomes, 32 human faecal microbiome
datasets and 126 clinical isolates of Shigella sonnei.
PLOS ONE | DOI:10.1371/journal.pone.0133492 July 21, 2015 1 / 12
a11111
OPEN ACCESS
Citation: Rowe W, Baker KS, Verner-Jeffreys D,
Baker-Austin C, Ryan JJ, Maskell D, et al. (2015)
Search Engine for Antimicrobial Resistance: A Cloud
Compatible Pipeline and Web Interface for Rapidly
Detecting Antimicrobial Resistance Genes Directly
from Sequence Data. PLoS ONE 10(7): e0133492.
doi:10.1371/journal.pone.0133492
Editor:Willem van Schaik, University Medical Center
Utrecht, NETHERLANDS
Received: May 13, 2015
Accepted: June 27, 2015
Published: July 21, 2015
Copyright: © 2015 Rowe et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any
medium, provided the original author and source are
credited.
Data Availability Statement: All novel environmental
metagenomes are available from the European
Nucleotide Archive (Metagenomics) database. ENA
project numbers and dataset accession numbers are
available in supplemental methods, S3 and S4 files.
Funding: This research was funded by
GlaxoSmithKline, the Centre for Environment,
Fisheries and Aquaculture Science and the
Biotechnology and Biological Sciences Research
Council under an industrial CASE studentship. The
funder Centre for Environment, Fisheries and
Conclusions
We have developed a pipeline that contributes to the improved capacity for antimicrobial
resistance detection afforded by next generation sequencing technologies, allowing for
rapid detection of antimicrobial resistance genes directly from sequencing data. SEAR uses
raw sequencing data via an intuitive interface so can be run rapidly without requiring
advanced bioinformatic skills or resources. Finally, we show that SEAR is effective in
detecting antimicrobial resistance genes in metagenomic and isolate sequencing data from
both environmental metagenomes and sequencing data from clinical isolates.
Introduction
The global threat of antimicrobial resistance is growing at an alarming rate; infections that
were once easily treatable now constitute public health crises [1]. This has lead to the consensus
that more must be done to monitor and combat the occurrence and spread of antimicrobial
resistance [2, 3]. Current diagnostic laboratory practice for the detection of antimicrobial resis-
tance relies on isolate culturing, followed by growth inhibition assays for the identification of
resistant phenotypes and determination of Minimum Inhibitory Concentrations against a
range of antimicrobials (MICs) [4]. Alternatively, antimicrobial resistance genes (ARGs) can
be identified using polymerase chain reaction (PCR) and quantified using real-time PCR,
requiring specific primers for the amplification of target sequences [5]. These approaches take
time, consume resources, and have limitations that may result in clinically relevant resistances
being undetected e.g. phenotypic testing will miss non-culturable bacteria and non-expressed
ARGs, whereas limitations of multiplex composition and size in molecular testing complicates
the detection of ARGs [6, 7].
Perhaps not surprisingly, the Centers for Disease Control and Prevention (CDC) identified
one of the current downfalls in the approach to combatting antimicrobial resistance as the
poor use of advanced molecular detection (AMD) technologies [8]. AMD technologies, such as
the whole genome sequencing of bacterial isolates as well as uncultured bacteria (metagenomic
sequencing), have the potential to identify antimicrobial resistance more quickly and effectively
than conventional laboratory assays [8]. In addition to these well-understood advantages,
AMD technologies can also be applied to circumvent the requirement of prior knowledge of
causative agents and provide clinically relevant information for the treatment and surveillance
of pathogens as well as antimicrobial resistance [9]. Upon receipt of a metagenomic (e.g. envi-
ronmental or faecal microbiome) or isolate sample, DNA can be extracted, compiled into a
library and sequenced within hours [10]. Indeed, AMD approaches to pathogen detection are
currently being developed and seek to identify pathogens directly from metagenomic samples
within clinically relevant timeframes [11]. Recent studies have also shown AMD to be effective
in the epidemiological tracking of pathogens, as well as the detection of ARGs present in their
genomes [12, 13]. AMD offers an alternative screening tool that may be quicker than tradi-
tional culture-based techniques. For example, the detection ofMycobacterium tuberculosis
requires inoculated isolation media to be incubated for several days in order to diagnose infec-
tion and additional time for phenotypic characterisation of antimicrobial resistance [14]. This
highlights the potential for developing more efficient diagnostic tests and the utilisation of
AMD technologies to create more rapid alternatives for ARG detection.
Search Engine for Antimicrobial Resistance
PLOS ONE | DOI:10.1371/journal.pone.0133492 July 21, 2015 2 / 12
Aquaculture Science provided support in the form of
salaries, research materials and facilities for authors
DVJ and CBA, but did not have any additional role in
the study design, data collection and analysis,
decision to publish, or preparation of the manuscript.
The funder GlaxoSmithKline provided support in the
form of salaries for author JR, but did not have any
additional role in the study design, data collection and
analysis, decision to publish, or preparation of the
manuscript. The specific roles of the authors are
articulated in the ‘author contributions’ section.
Competing Interests: The authors would like to
declare a commercial affiliation with GSK, who partly
funded the research and employs JR. This does not
alter the authors' adherence to PLOS ONE policies
on sharing data and materials.
In addition to these direct clinical applications, AMD technologies are also beginning to
become a common tool in the detection of ARGs in the environment, which is vital for identify-
ing reservoirs of ARGs [15–17]. However, there is need to establish a metagenomic framework
for use in the monitoring of ARGs within the environment in order to influence public health
decisions and the growing concern over antimicrobial resistance [18]. This must include the
development of reliable surveillance methods and tools for risk assessment [19]. When design-
ing metagenomic tools for the environmental monitoring of ARGs, it is therefore necessary to
provide context in terms of the relative abundance of ARGs, so that these can be correlated with
environmental variables (e.g. such as antimicrobial concentrations, etc.) as well as to obtain
information on the mobile genetic elements (MGE) and pathogens that they are associated with.
Currently published resources available for ARG detection are online databases that use the
Basic Local Alignment Search Tool (BLAST) algorithm to find possible matches between the
database and query sequences (e.g. ARDB, CARD, ResFinder) [20–23]. To our knowledge, no
existing tools give an ARG abundance measure or simultaneously provide MGE information.
The targeting of full-length gene matches using BLAST requires a sequence assembly step, add-
ing time, infrastructure requirements, and complexity to the analysis. Furthermore, full-length
gene assembly is often difficult to achieve in metagenomic samples where coverage is fre-
quently low and uneven across the sample. Ideally, raw sequencing data would be used directly
to rapidly identify and quantify ARGs of interest. Although mapping-based approaches have
been used for individual studies [24, 25] and tools that work directly with reads (though on
non-ARG databases) such as the SEED subsystems and SRST2 can be applied to work to this
aim [26], there is as yet no such ARG-detection algorithm. Here, we present an automated
pipeline, the Search Engine for Antimicrobial Resistance (SEAR), which quickly and accurately
identifies antimicrobial resistance information from biological samples. Furthermore, it also
provides abundance estimates and returns the true sample full-length reconstructed gene
sequence. To demonstrate efficacy, we present the application of the pipeline to a range of
sequencing data types including novel environmental metagenomes, human faecal metagen-
omes and clinical isolates of pathogenic enteric bacteria (Shigella sonnei).
Materials and Methods
SEAR requirements
Reference databases. SEAR requires reference databases for read subtraction and read
clustering. Details of the supplied databases and how the user can supply their own custom
databases are given in supplemental methods (Supplemental methods A in S1 File). The default
databases supplied for read subtraction and read clustering are the human genome (HG19
build) and the ARG-annot database [27].
Hardware. Minimum hardware requirements for SEAR comprise a Unix server (tested
using Ubuntu 10.04) with ~2 GB of disk space for reference data and software dependencies
(see S1 Table). Whilst running, SEAR requires up to 2X the input FASTQ file size (bytes) in
both RAM and disk space for temporary file storage.
SEAR
The pipeline. SEAR is a pipeline consisting of Perl, Shell and R scripts that call on several
pieces of open source software and utilise a customisable reference database to annotate ARGs
direct from short-read sequencing data. SEAR is downloadable from http://computing.bio.
cam.ac.uk/sear/SEAR_WEB_PAGE/SEAR.html, in stand-alone command-line and web-based
versions (Fig 1).
Search Engine for Antimicrobial Resistance
PLOS ONE | DOI:10.1371/journal.pone.0133492 July 21, 2015 3 / 12
Fig 1. Screen shot of SEARwebinterface including homepage (A) and quick start settings (B).
doi:10.1371/journal.pone.0133492.g001
Search Engine for Antimicrobial Resistance
PLOS ONE | DOI:10.1371/journal.pone.0133492 July 21, 2015 4 / 12
The pipeline follows five main steps in the annotation of ARGs: (1) processing of input files,
(2) clustering of sequence reads to known ARGs in user-defined (or pre-loaded) database, (3)
mapping of reads to reference sequences, (4) ARG annotation and calculation of relative abun-
dance and (5) local alignment of annotated ARGs to online databases.
(1) Processing of input files
The pipeline accepts raw or compressed (.gz) FASTQ files (either 33 or 64 ASCII encoding)
from metagenomic, metatranscriptomic or isolate sequencing. Where more than one input file
(e.g. paired-end data) is provided, these files are merged to give a single input file (pair-end
information is not currently utilised in the pipeline). The pipeline has the optional step of pre-
filtering reads, by removing those that map against a user-defined reference, such as the
human genome or a bacterial strain. FASTQ files are quality checked using user-defined cut
offs and converted to FASTA formatted reads.
(2) Clustering of sequence reads to ARG database
The pipeline is supplied with a custom ARG database that has been built by clustering and
annotating the ARGs held in the ARGannot-database [27]. Notably however, other ARG data-
bases can be used or the user can use a custom FASTA file (Supplemental methods B in S1
File). Reads are clustered to the ARG database by global alignment with USEARCH (version
7.0.959) using a default identity cut-off of 99% [28]. Where multiple matches occur, the read is
clustered with the highest identity match. SEAR parses the clusters by grouping reads to each
matched reference gene and retrieving corresponding FASTQ information for each matched
read.
(3) Mapping of clustered sequence reads to ARG references
The Burrows-Wheeler Aligner (BWA-mem version 0.7.8) [29] is used for read mapping each
cluster of FASTQ reads to the corresponding reference gene. Samtools is then used to analyse
the BWA alignment and generate a consensus sequence using mpileup [30].
(4) ARG annotation and relative abundance
The consensus sequences are used to annotate ARGs and calculate relative abundance values;
an ARG is present in the sample if sequence reads can be mapped to the ARG reference
sequence above the defined coverage cut-off (coverage is the percentage length of reference
ARG with mapped reads). For relative abundance calculation, SEAR uses a similar method to
the reads per kilobase/million reads (RPKM) method that is commonly used in transcriptome
studies [31]. Full details on cut-off values and abundance calculation are given in supplemental
methods (Supplemental methods C in S1 File).
(5) Local alignment
The consensus sequences for annotated ARGs are aligned to the NCBI nucleotide and protein
databases using commandline BLAST [23] (using the–remote BLAST service by default, see
documentation to utilise local database versions). In addition, sequences are also aligned to the
current Repository of Antimicrobial Resistance Cassettes (RAC) [32] and Antibiotic Resistance
Database (ARDB) [20] databases using BLAST (though ARDB has not recently been curated).
Pipeline outputs. In both command-line and web versions of SEAR, output includes:
graphical overview, ARG annotations, relative abundance scores, consensus sequences, flat
files (html, csv, blast files) and links to further gene information and homologues found in
Search Engine for Antimicrobial Resistance
PLOS ONE | DOI:10.1371/journal.pone.0133492 July 21, 2015 5 / 12
online databases (such as the repository of antimicrobial resistance cassettes, NCBI non-redun-
dant nucleotide and protein databases).
Demonstrating SEAR utility
Data sets used in this study. Several datasets were used to demonstrate the utility of this
pipeline across broad data categories. All datasets were analysed using a UNIX server (Ubuntu
10.04) running SEAR with default parameters (99% clustering identity and 90% coverage cut-
off for ARG annotation, full default parameter list found in S2 Table).
Novel environmental metagenomes. Information on metagenome sample collection,
library construction and sequencing are provided in supplemental methods (Supplemental
methods D, E F in S1 File). Briefly, faecal wastewater effluent samples were taken from a dairy
farm (latitude: 52.22259, longitude: 0.02603) and a metropolitan (human) wastewater treat-
ment works (WWTW) (latitude: 52.234469, longitude: 0.154614). Samples were vacuum fil-
tered through 0.22μmmembranes, DNA extracted and sequenced using the Illumina HiSeq
2000 platform.
Pre-existing metagenomic and clinical isolate data. Human Microbiome Project (HMP)
data for 32 Spanish human faecal microbiomes (for which the ARGs have previously been
characterised in an in silico study by Forslund et al. [25]) were used (SRA Study ERP002061).
Additionally, SEAR was used to detect ARGs in a global dataset of 126 clinical isolates of the
pathogenic bacteria Shigella sonnei (SRA Study ERP000182) [33]. In the case of the clinical iso-
lates, SEAR ARG detection was compared with the published ARG content of the isolates, with
SEAR being run with default parameters on a custom reference database of ARGs originally
detected by 100% mapping [33]. Further details on datasets are provided in S3 Table.
Results
To test the utility of SEAR we ran the pipeline using a variety of sample types (environmental
metagenomes, human faecal microbiome and bacterial clinical isolate), recorded pipeline run
times (S4 Table) and then investigated the presence and abundance of ARGs in all samples.
Discrimination of ARG presence and abundance between
environmental metagenomes
A total of 28 (15 in each) ARGs were identified among the environmental metagenomes from
WWTW effluent and farm waste effluent (Fig 2). Only two genes, strA and strB (both confer-
ring aminoglycoside resistance), were common between the metagenomes and each gene
found in both sets was five times more abundant in the WWTW effluent compared to the farm
effluent when using the normalised abundance values for the combined datasets. The WWTW
effluent had ARGs conferring resistance to a total of four antimicrobial resistance profiles with
the most diverse (i.e. greatest number of ARGs) being the aminoglycoside resistance profile
and the greatest abundance being ARGs conferring tetracycline resistance. In contrast, the
farm effluent had ARGs conferring resistance to five resistance profiles with the most diverse
being the beta lactam resistance profile and the most abundant also being tetracycline resis-
tance (Fig 2). The most abundant ARGs in the metagenome datasets were tetracycline resis-
tance genes; tetC (41.6%) in the farm effluent, and tet39 (15.3%) in the WWTW effluent. A
subset of ARGs identified by SEAR (tetA, qnrB and bla-ACT; chosen to encompass clinically
relevant resistances, drugs with both a long and short history of resistance and chemically
diverse antimicrobials) was confirmed in the original farm effluent DNA sample using PCR.
Briefly, primers were designed using Primer3 [34] and were amplified using GoTaq DNA poly-
merase (Promgega) (not shown).
Search Engine for Antimicrobial Resistance
PLOS ONE | DOI:10.1371/journal.pone.0133492 July 21, 2015 6 / 12
Fig 2. SEAR results for environmental metagenomes. The column chart in A shows the breakdown of the number of ARGs in each effluent, grouped by
antimicrobial resistance profile. The column chart in B shows the relative abundance of ARGs found in each metagenome (coloured according to the key).
doi:10.1371/journal.pone.0133492.g002
Search Engine for Antimicrobial Resistance
PLOS ONE | DOI:10.1371/journal.pone.0133492 July 21, 2015 7 / 12
Efficacy of SEAR for detecting ARGs in human faecal microbiomes
To assess the efficacy of SEAR for detecting ARGs in microbiome data, SEAR was tested on 32
faecal microbiome samples (S5 Table). ARGs were detected in 31 of the samples and a total of
295 genes conferring resistance to 6 classes of antimicrobials were identified across the samples
(Table 1). Genes conferring resistance to tetracyclines were again the most common ARGs
identified (39% of total ARGs detected).
Accuracy of SEAR ARG detection using clinical isolate sequencing data
To evaluate SEAR’s efficacy in detecting ARGs in clinical isolate sequencing data, SEAR was
run on sequencing data from 126 isolates of the enteric pathogen Shigella sonnei. To evaluate
SEAR’s performance, the results were compared to the ARG detection data presented in the
original publication [33]. Of the 231 detection events (see methods for criteria) originally pre-
sented in the publication, SEAR identified 221 of these, and a further 20 ARGs (Table 2, full
results shown in S6 Table).
Discussion
SEAR is an ARG annotation tool that is freely available and may be downloaded as a cloud
compatible web interface or a stand-alone command line program. It offers advantages over
currently available ARG annotation tools as it provides ARG annotations, relative abundance
Table 1. SEAR detection of ARGs across antimicrobial profile/classes in human faecal microbiomes.
Antimicrobial resistance profile Number of ARGs
Aminoglycosides 54
Beta lactams 38
Quinolones 0
Glycopeptides 0
Macrolides/Lincosamides/Streptogramins 82
Phenicols 1
Rifampicin 0
Sulfonamides 5
Tetracyclines 115
Trimethoprims 0
The table shows the number of genes identified in each antimicrobial resistance profile for the combined
dataset of HMP samples.
doi:10.1371/journal.pone.0133492.t001
Table 2. Accuracy of SEAR ARG detection using clinical isolate sequencing data.
Reported in Holt et al. [33]
detected not-detected TOTAL
SEAR results detected 221 20 241
not-detected 10 0 10
TOTAL 231 20
The contingency table compares the detection and non-detection of ARGs by SEAR relative to the published ARG detection data for 126 S. sonnei
isolates.
doi:10.1371/journal.pone.0133492.t002
Search Engine for Antimicrobial Resistance
PLOS ONE | DOI:10.1371/journal.pone.0133492 July 21, 2015 8 / 12
values, gene sequence and gene information from raw sequencing data without requiring any
sequence assembly. In contrast to tools based on BLAST comparison of de novo assemblies, the
clustering and mapping approach used by SEAR, combined with the customisable database
and annotation parameters, allows the user to detect putative ARGs in incomplete or low cov-
erage sequencing data that is common in metagenomic analyses. SEAR successfully identified
ARGs in sequencing datasets that were generated from novel environmental metagenomic
samples, human microbiomes and clinical isolates of Shigella sonnei.
SEAR was able to detect the ARGs present in two novel environmental metagenomes allow-
ing direct comparison between two different wastewater effluent samples. SEAR identified
meaningful differences among ARGs of clinical interest, for example the presence of quinolone
resistance genes (qnrB and qnrS) exclusively in the wastewater effluent from the farm source. It
also showed that while the two sources had different qualitative ARG characteristics (with
either aminoglycosides or beta-lactams being the most diverse resistance profiles) and in both
tetracycline resistance genes were present in the greatest abundance. In addition to detecting
important differences among these sample types, the confirmation of a subset of identified
ARGs by PCR demonstrated the robustness of the pipeline.
Similarly, SEAR was effective for identifying ARGs from clinical samples. ARGs were
detected in human microbiomes demonstrating the potential of using metagenomic analyses
for the surveillance and management antimicrobial resistance. Additionally, SEAR successfully
identified ARGs in a global dataset of 126 clinical isolates of an important enteric pathogen.
There were a few discrepancies, which were consistent with a given isolate or gene family, how-
ever the results were overwhelmingly consistent. Furthermore, the congruence of ARG detec-
tion results from SEAR with the published ARG content of the isolates further highlighted the
effectiveness of the pipeline, providing further compelling argument for the application of
high-throughput AMD into clinical microbiology.
Limitations and future improvements
SEAR offers increased functionality over existing bioinformatic tools by providing a consensus
sequence of annotated ARGs, links to online resources containing information on the ARGs
(and gene homologs) and a relative abundance estimate for each ARG detected. Each ARG
consensus sequence is generated using reads that clustered to a reference sequence and conse-
quently any variability in the consensus sequence in a metagenomic sample may be due to
either sequencing noise or the presence of multiple bona fide sequence variants. The relative
abundance estimate is relative within an individual sample, however the SEAR output features
the information required to calculate relative abundance across multiple samples. Due to possi-
ble large variations in user file size and upload speed, the SEAR interface and command line
tool are available for use as downloadable packages.
SEAR is designed for detecting ARGs that are horizontally acquired, not antimicrobial resis-
tance that is caused (or inactivated) by single nucleotide polymorphisms (SNPs) e.g. SNPs in
the gyrA gyrase gene that result in quinolone resistance. SNPs are not currently tested for due
to the annotation parameters being calibrated for detecting partial ARG matches to compen-
sate for low sequencing coverage. Hence, such SNPs may be missed by SEAR due to the num-
ber of mismatches permitted or by a low coverage cut-off (though these are both customisable
settings). For these reasons, it is not recommended to include SNP-based resistances in refer-
ence databases used with SEAR as they may lead to false positives. The detection of SNP-based
resistances in metagenomic samples represents a significant future challenge that needs to be
addressed. It should also be stressed that the default SEAR parameters, which are based on
high-stringency read clustering and mapping, result in an analysis that finds ARGs that are
Search Engine for Antimicrobial Resistance
PLOS ONE | DOI:10.1371/journal.pone.0133492 July 21, 2015 9 / 12
known in the reference data and it is not suited for discovery of emergent ARGs. The high-
stringency settings are designed to exclude the possibility of non-competitive read mapping
causing false positive results by ensuring that annotated ARGs have a high sequence identity
compared to the reference database.
Conclusion
We have presented a bioinformatic pipeline that is highly effective for detecting ARGs directly
from raw sequencing reads that also provides relative abundance estimation and sequences of
identified genes. We have shown its application on sequence data from metagenomic datasets
and bacterial isolates. We have demonstrated the application of SEAR in potential clinical and
environmental monitoring applications, highlighting the advantages of automated interpreta-
tion of sequencing data for generating timely and informative reports for informing public
health and potentially clinical decision-making. With the increasing drive to integrate AMD
technology and existing laboratory assays in order to combat antimicrobial resistance, we pres-
ent this pipeline as a valuable step towards this important goal.
Availability and requirements
Project name: Search Engine for Antimicrobial Resistance (SEAR)
Project home page:
http://computing.bio.cam.ac.uk/sear/SEAR_WEB_PAGE/SEAR.html
Operating system(s): UNIX
Programming language: Perl
Other requirements: Usearch (v.7), BWA, samtools, R
License: GNU GPL (version 3)
Any restrictions to use by non-academics: na
Supporting Information
S1 File. Supplemental methods.
(PDF)
S1 Table. A list of dependencies required by SEAR.
(PDF)
S2 Table. SEAR parameters.
(PDF)
S3 Table. NGS datasets.
(PDF)
S4 Table. Example runtimes for SEAR.
(PDF)
S5 Table. SEAR ARG detection for HMP sequence data.
(PDF)
S6 Table. SEAR ARG detection using clinical isolate sequence data.
(PDF)
Acknowledgments
The authors would like to thank Dr. Jenny Barna for computing support.
Search Engine for Antimicrobial Resistance
PLOS ONE | DOI:10.1371/journal.pone.0133492 July 21, 2015 10 / 12
Author Contributions
Conceived and designed the experiments: WR GP. Performed the experiments: WR KB. Ana-
lyzed the data: WR. Contributed reagents/materials/analysis tools: WR KB. Wrote the paper:
WR KB DVJ CBA JR DM GP.
References
1. Sack D., Lyke C., McLaughlin C., Suwanvanichkij V. Antimicrobial resistance in shigellosis, cholera
and campylobacteriosis. World Health Organization: Department of Communicable Disease Surveil-
lance and Response, 2001.
2. WHO. The evolving threat of antimicrobial resistance—options for action. Geneva2012.
3. Laxminarayan R, Duse A, Wattal C, Zaidi AKM,Wertheim HFL, Sumpradit N, et al. Antibiotic resistance
—the need for global solutions. The Lancet Infectious Diseases. 2013; 13(12):1057–98. doi: 10.1016/
S1473-3099(13)70318-9 PMID: 24252483
4. PHE. Antibiotic Resistance Monitoring & Reference Laboratory (ARMRL) 2013. Available from: http://
www.hpa.org.uk/.
5. Espy MJ, Uhl JR, Sloan LM, Buckwalter SP, Jones MF, Vetter EA, et al. Real-time PCR in clinical micro-
biology: applications for routine laboratory testing. Clin Microbiol Rev. 2006; 19(1):165–256. Epub
2006/01/19. doi: 10.1128/CMR.19.1.165-256.2006 PMID: 16418529; PubMed Central PMCID:
PMC1360278.
6. Diekema DJ, Pfaller MA. Rapid Detection of Antibiotic-Resistant Organism Carriage for Infection Pre-
vention. Clinical Infectious Diseases. 2013; 56(11):1614–20. doi: 10.1093/cid/cit038 PMID: 23362298
7. Katherine E. Heiman, Karlsson Maria, Julian Grass, Becca Howie, Robert D. Kirkcaldy, Barbara
Mahon, et al. Shigella with Decreased Susceptibility to Azithromycin Among MenWho Have Sex with
Men—United States, 2002–2013. Morbidity and Mortality Weekly Report—CDC. 2014; 63(6):132.
8. CDC CfDCaP. Antibiotic Resistance Threats in the United States, 20132013.
9. Miller R, Montoya V, Gardy J, Patrick D, Tang P. Metagenomics for pathogen detection in public health.
GenomeMedicine. 2013; 5(9):81. doi: 10.1186/gm485 PMID: 24050114
10. Illumina. Illumina MySeq Benchtop Sequencer 2013. Available from: http://www.illumina.com.
11. Naccache SN, Federman S, Veeraraghavan N, Zaharia M, Lee D, Samayoa E, et al. A cloud-compati-
ble bioinformatics pipeline for ultrarapid pathogen identification from next-generation sequencing of
clinical samples. Genome Research. 2014; 24(7):1180–92. doi: 10.1101/gr.171934.113 PMC4079973.
PMID: 24899342
12. Koser CU, Holden MT, Ellington MJ, Cartwright EJ, Brown NM, Ogilvy-Stuart AL, et al. Rapid whole-
genome sequencing for investigation of a neonatal MRSA outbreak. N Engl J Med. 2012; 366
(24):2267–75. Epub 2012/06/15. doi: 10.1056/NEJMoa1109910 PMID: 22693998; PubMed Central
PMCID: PMC3715836.
13. Harrison EM, Paterson GK, Holden MTG, Larsen J, Stegger M, Larsen AR, et al. Whole genome
sequencing identifies zoonotic transmission of MRSA isolates with the novelmecA homologuemecC.
EMBOMol Med. 2013; 5(4):509–15. doi: 10.1002/emmm.201202413 PMID: 23526809
14. Kidenya BR, Kabangila R, Peck RN, Mshana SE, Webster LE, Koenig SP, et al. Early and Efficient
Detection ofMycobacterium tuberculosis in Sputum by Microscopic Observation of Broth Cultures.
PLoS One. 2013; 8(2):e57527. doi: 10.1371/journal.pone.0057527 PMID: 23469014
15. Lewin A, Johansen J, Wentzel A, Kotlar HK, Drablos F, Valla S. The microbial communities in two
apparently physically separated deep subsurface oil reservoirs show extensive DNA sequence similari-
ties. Environ Microbiol. 2013. Epub 2013/07/06. doi: 10.1111/1462-2920.12181 PMID: 23827055.
16. Mason OU, Hazen TC, Borglin S, Chain PS, Dubinsky EA, Fortney JL, et al. Metagenome, metatran-
scriptome and single-cell sequencing reveal microbial response to Deepwater Horizon oil spill. Isme J.
2012; 6(9):1715–27. Epub 2012/06/22. doi: 10.1038/ismej.2012.59 PMID: 22717885; PubMed Central
PMCID: PMC3498917.
17. Oh S, Tandukar M, Pavlostathis SG, Chain PS, Konstantinidis KT. Microbial community adaptation to
quaternary ammonium biocides as revealed by metagenomics. Environ Microbiol. 2013. Epub 2013/
06/05. doi: 10.1111/1462-2920.12154 PMID: 23731340.
18. Port JA, Cullen AC, Wallace JC, Smith MN, Faustman EM. Metagenomic Frameworks for Monitoring
Antibiotic Resistance in Aquatic Environments. Environ Health Perspect. 2014; 122(3):222–8. doi: 10.
1289/ehp.1307009 PMC3948035. PMID: 24334622
Search Engine for Antimicrobial Resistance
PLOS ONE | DOI:10.1371/journal.pone.0133492 July 21, 2015 11 / 12
19. Berendonk TU, Manaia CM, Merlin C, Fatta-Kassinos D, Cytryn E, Walsh F, et al. Tackling antibiotic
resistance: the environmental framework. Nat Rev Micro. 2015;advance online publication. doi: 10.
1038/nrmicro3439
20. Liu B, Pop M. ARDB, Antibiotic Resistance Genes Database. Nucleic Acids Research. 2009; 37(suppl
1):D443–D7. doi: 10.1093/nar/gkn656
21. McArthur AG,Waglechner N, Nizam F, Yan A, Azad MA, Baylay AJ, et al. The comprehensive antibiotic
resistance database. Antimicrob Agents Chemother. 2013; 57(7):3348–57. Epub 2013/05/08. doi: 10.
1128/AAC.00419-13 PMID: 23650175; PubMed Central PMCID: PMC3697360.
22. Zankari E, Hasman H, Cosentino S, Vestergaard M, Rasmussen S, Lund O, et al. Identification of
acquired antimicrobial resistance genes. J Antimicrob Chemother. 2012; 67(11):2640–4. Epub 2012/
07/12. doi: 10.1093/jac/dks261 PMID: 22782487; PubMed Central PMCID: PMC3468078.
23. Altschul SF, GishW, Miller W, Myers EW, Lipman DJ. Basic local alignment search tool. Journal of
Molecular Biology. 1990; 215(3):403–10. doi: 10.1016/S0022-2836(05)80360-2 PMID: 2231712
24. Hu Y, Yang X, Qin J, Lu N, Cheng G, Wu N, et al. Metagenome-wide analysis of antibiotic resistance
genes in a large cohort of human gut microbiota. Nat Commun. 2013; 4:2151. Epub 2013/07/24. doi:
10.1038/ncomms3151 PMID: 23877117.
25. Forslund K, Sunagawa S, Kultima JR, Mende DR, ArumugamM, Typas A, et al. Country-specific antibi-
otic use practices impact the human gut resistome. Genome Research. 2013; 23(7):1163–9. doi: 10.
1101/gr.155465.113 PMID: 23568836
26. Inouye M, Dashnow H, Raven L-A, Schultz M, Pope B, Tomita T, et al. SRST2: Rapid genomic surveil-
lance for public health and hospital microbiology labs. GenomeMedicine. 2014; 6(11):90. doi: 10.1186/
s13073-014-0090-6 PMID: 25422674
27. Gupta SK, Padmanabhan BR, Diene SM, Lopez-Rojas R, Kempf M, Landraud L, et al. ARG-ANNOT, a
New Bioinformatic Tool To Discover Antibiotic Resistance Genes in Bacterial Genomes. Antimicrob
Agents Chemother. 2014; 58(1):212–20. doi: 10.1128/aac.01310-13 PMID: 24145532
28. Edgar RC. Search and clustering orders of magnitude faster than BLAST. Bioinformatics. 2010; 26
(19):2460–1. Epub 2010/08/17. doi: 10.1093/bioinformatics/btq461 PMID: 20709691.
29. Li H, Durbin R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformat-
ics. 2009; 25(14):1754–60. Epub 2009/05/20. doi: 10.1093/bioinformatics/btp324 PMID: 19451168;
PubMed Central PMCID: PMC2705234.
30. Li H. A statistical framework for SNP calling, mutation discovery, association mapping and population
genetical parameter estimation from sequencing data. Bioinformatics. 2011; 27(21):2987–93. Epub
2011/09/10. doi: 10.1093/bioinformatics/btr509 PMID: 21903627; PubMed Central PMCID:
PMC3198575.
31. Mortazavi A, Williams BA, McCue K, Schaeffer L, Wold B. Mapping and quantifying mammalian tran-
scriptomes by RNA-Seq. Nat Methods. 2008; 5(7):621–8. Epub 2008/06/03. doi: 10.1038/nmeth.1226
PMID: 18516045.
32. Tsafnat G, Copty J, Partridge SR. RAC: Repository of Antibiotic resistance Cassettes. Database.
2011;2011. doi: 10.1093/database/bar054
33. Holt KE, Baker S, Weill FX, Holmes EC, Kitchen A, Yu J, et al. Shigella sonnei genome sequencing and
phylogenetic analysis indicate recent global dissemination from Europe. Nat Genet. 2012; 44(9):1056–
9. Epub 2012/08/07. doi: 10.1038/ng.2369 PMID: 22863732; PubMed Central PMCID: PMC3442231.
34. Rozen S, Skaletsky HJ. Primer3. 1988.
Search Engine for Antimicrobial Resistance
PLOS ONE | DOI:10.1371/journal.pone.0133492 July 21, 2015 12 / 12
  
147 
Appendix 2. River Cam catchment area. 
 
  
148 
Appendix 3. List of sequencing data generated and used in these studies. 
Section Sample ID Location collected Data type Biome Date collected 
Chapter 2 DF:M:1 Cambridge University dairy farm Metagenome Fertiliser/lagoon effluent 22.06.2012 
 WWTP:M:1 City of Cambridge Wastewater Treatment Plant Metagenome Treated wastewater effluent 21.06.2012 
Chapter 3 AS:M:1 Ashwell spring Metagenome River Cam source water 02.05.2013 
 DF:M:1 Cambridge University dairy farm Metagenome Fertiliser/lagoon effluent 22.06.2012 
 DF:M:2 Cambridge University dairy farm Metagenome Fertiliser/lagoon effluent 02.05.2013 
 DF:M:3 Cambridge University dairy farm Metagenome Fertiliser/lagoon effluent 06.08.2014 
 WWTP:M:1 City of Cambridge Wastewater Treatment Plant Metagenome Treated wastewater effluent 21.06.2012 
 WWTP:M:2 City of Cambridge Wastewater Treatment Plant Metagenome Treated wastewater effluent 02.05.2013 
 WWTP:M:3 City of Cambridge Wastewater Treatment Plant Metagenome Treated wastewater effluent 04.08.2014 
Chapter 4 AH:M:1 Addenbrooke's hospital Metagenome Hospital wastewater effluent 02.05.2013 
 AH:M:2 Addenbrooke's hospital Metagenome Hospital wastewater effluent 04.08.2014 
 AH:M:3 Addenbrooke's hospital Metagenome Hospital wastewater effluent 15.09.2014 
 AH:M:4 Addenbrooke's hospital Metagenome Hospital wastewater effluent 29.09.2014 
 AH:M:5 Addenbrooke's hospital Metagenome Hospital wastewater effluent 27.10.2014 
 AH:M:6 Addenbrooke's hospital Metagenome Hospital wastewater effluent 24.11.2014 
 AS:M:1 Ashwell spring Metagenome River Cam source water 02.05.2013 
 AS:M:2 Ashwell spring Metagenome River Cam source water 04.08.2014 
 AS:M:3 Ashwell spring Metagenome River Cam source water 15.09.2014 
 AS:M:4 Ashwell spring Metagenome River Cam source water 29.09.2014 
 AS:M:5 Ashwell spring Metagenome River Cam source water 27.10.2014 
 AS:M:6 Ashwell spring Metagenome River Cam source water 24.11.2014 
  
149 
 DF:M:2 Cambridge University dairy farm Metagenome Fertiliser/lagoon effluent 02.05.2013 
 DF:M:3 Cambridge University dairy farm Metagenome Fertiliser/lagoon effluent 06.08.2014 
 DF:M:4 Cambridge University dairy farm Metagenome Fertiliser/lagoon effluent 15.09.2014 
 DF:M:5 Cambridge University dairy farm Metagenome Fertiliser/lagoon effluent 29.09.2014 
 DF:M:6 Cambridge University dairy farm Metagenome Fertiliser/lagoon effluent 27.10.2014 
 DF:M:7 Cambridge University dairy farm Metagenome Fertiliser/lagoon effluent 24.11.2014 
 AH:T:1 Addenbrooke's hospital Metatranscriptome Hospital wastewater effluent 02.05.2013 
 AH:T:2 Addenbrooke's hospital Metatranscriptome Hospital wastewater effluent 04.08.2014 
 AH:T:3 Addenbrooke's hospital Metatranscriptome Hospital wastewater effluent 15.09.2014 
 AH:T:4 Addenbrooke's hospital Metatranscriptome Hospital wastewater effluent 29.09.2014 
 AH:T:5 Addenbrooke's hospital Metatranscriptome Hospital wastewater effluent 27.10.2014 
 AH:T:6 Addenbrooke's hospital Metatranscriptome Hospital wastewater effluent 24.11.2014 
 AS:T:1 Ashwell spring Metatranscriptome River source water 02.05.2013 
 AS:T:2 Ashwell spring Metatranscriptome River source water 04.08.2014 
 AS:T:3 Ashwell spring Metatranscriptome River source water 15.09.2014 
 AS:T:4 Ashwell spring Metatranscriptome River source water 29.09.2014 
 AS:T:5 Ashwell spring Metatranscriptome River source water 27.10.2014 
 AS:T:6 Ashwell spring Metatranscriptome River source water 24.11.2014 
 DF:T:1 Cambridge University dairy farm Metatranscriptome Fertiliser/lagoon effluent 02.05.2013 
 DF:T:2 Cambridge University dairy farm Metatranscriptome Fertiliser/lagoon effluent 06.08.2014 
 DF:T:3 Cambridge University dairy farm Metatranscriptome Fertiliser/lagoon effluent 15.09.2014 
 DF:T:4 Cambridge University dairy farm Metatranscriptome Fertiliser/lagoon effluent 29.09.2014 
 DF:T:5 Cambridge University dairy farm Metatranscriptome Fertiliser/lagoon effluent 27.10.2014 
 DF:T:6 Cambridge University dairy farm Metatranscriptome Fertiliser/lagoon effluent 24.11.2014 
  
150 
Chapter 5 PF:M:1 Writtle College pig farm Metagenome Fertiliser/lagoon effluent 15.07.2014 
 #1 Writtle College pig farm Isolate WGS Combined farrow house effluent 15.07.2014 
 #2 Writtle College pig farm Isolate WGS Combined farrow house effluent 15.07.2014 
 #3 Writtle College pig farm Isolate WGS Piglet crate effluent 15.07.2014 
 #4 Writtle College pig farm Isolate WGS Piglet crate effluent 15.07.2014 
 #5 Writtle College pig farm Isolate WGS Combined farrow house effluent 15.07.2014 
 #6 Writtle College pig farm Isolate WGS Piglet crate effluent 15.07.2014 
 #7 Writtle College pig farm Isolate WGS Piglet crate effluent 15.07.2014 
 #8 Writtle College pig farm Isolate WGS Sow faeces 15.07.2014 
 #9 Writtle College pig farm Isolate WGS Combined farrow house effluent 15.07.2014 
 #10 Writtle College pig farm Isolate WGS Fertiliser/lagoon effluent 15.07.2014 
 #12 Writtle College pig farm Isolate WGS Fertiliser/lagoon effluent 15.07.2014 
 #13 Writtle College pig farm Isolate WGS Fertiliser/lagoon effluent 15.07.2014 
 #14 Writtle College pig farm Isolate WGS Manure heap 15.07.2014 
 #15 Writtle College pig farm Isolate WGS Manure heap 15.07.2014 
 #16 Writtle College pig farm Isolate WGS Combined weaner shed effluent 15.07.2014 
 #17 Writtle College pig farm Isolate WGS Combined weaner shed effluent 15.07.2014 
 #18 Writtle College pig farm Isolate WGS Soil (ammended with lagoon effluent) 15.07.2014 
 #19 Writtle College pig farm Isolate WGS Soil (ammended with lagoon effluent) 15.07.2014 
 #20 Writtle College pig farm Isolate WGS Soil (ammended with lagoon effluent) 15.07.2014 
 #21 Writtle College pig farm Isolate WGS Soil (ammended with lagoon effluent) 15.07.2014 
 #22 City of Cambridge Wastewater Treatment Plant Isolate WGS Untreated wastewater influent 04.08.2014 
 #23 City of Cambridge Wastewater Treatment Plant Isolate WGS Untreated wastewater influent 04.08.2014 
 #24 City of Cambridge Wastewater Treatment Plant Isolate WGS Untreated wastewater influent 04.08.2014 
  
151 
 #25 City of Cambridge Wastewater Treatment Plant Isolate WGS Untreated wastewater influent 04.08.2014 
 #26 City of Cambridge Wastewater Treatment Plant Isolate WGS Treated wastewater effluent 04.08.2014 
 #27 City of Cambridge Wastewater Treatment Plant Isolate WGS Treated wastewater effluent 04.08.2014 
 #28 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 04.08.2014 
 #29 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 04.08.2014 
 #30 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 04.08.2014 
 #31 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 04.08.2014 
 #32 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 04.08.2014 
 #33 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 04.08.2014 
 #34 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 04.08.2014 
 #35 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 04.08.2014 
 #36 Cambridge University dairy farm Isolate WGS Combined milking shed effluent 06.08.2014 
 #37 Cambridge University dairy farm Isolate WGS Combined milking shed effluent 06.08.2014 
 #38 Cambridge University dairy farm Isolate WGS Calf shed effluent 06.08.2014 
 #39 Cambridge University dairy farm Isolate WGS Calf shed effluent 06.08.2014 
 #40 Cambridge University dairy farm Isolate WGS Calf shed effluent 06.08.2014 
 #41 Cambridge University dairy farm Isolate WGS Calf shed effluent 06.08.2014 
 #42 Cambridge University dairy farm Isolate WGS Calf shed effluent 06.08.2014 
 #43 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 15.09.2014 
 #44 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 15.09.2014 
 #45 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 15.09.2014 
 #46 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 29.09.2014 
 #47 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 29.09.2014 
 #48 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 27.10.2014 
  
152 
 #49 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 27.10.2014 
 #50 Cambridge University dairy farm Isolate WGS Fertiliser/lagoon effluent 15.09.2014 
 #51 Cambridge University dairy farm Isolate WGS Fertiliser/lagoon effluent 29.09.2014 
 #52 River Cam Isolate WGS River water 01.10.2014 
 #53 River Cam Isolate WGS River water 01.10.2014 
 #54 River Cam Isolate WGS River water 01.10.2014 
 #55 Cambridge University dairy farm Isolate WGS Fertiliser/lagoon effluent 29.09.2014 
 #56 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 29.09.2014 
 #57 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 29.09.2014 
 #58 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 29.09.2014 
 #59 Addenbrooke's hospital Isolate WGS Hospital wastewater effluent 29.09.2014 
 
  
153 
Appendix 4. Accession numbers of sequences used in these studies. 
Section Sequence source Sequence reference ID Source Accession 
Chapter 2 Faecal microbiomes used to test SEAR efficacy O2.UC1-0 ENA ERR209529 
  O2.UC4-0 ENA ERR209614 
  O2.UC11-0 ENA ERR209531 
  O2.UC12-0 ENA ERR209534 
  O2.UC13-0 ENA ERR209538 
  O2.UC16-0 ENA ERR209544 
  O2.UC17-0 ENA ERR209547 
  O2.UC18-0 ENA ERR209550 
  O2.UC19-0 ENA ERR209554 
  O2.UC21-0 ENA ERR209564 
  O2.UC22-0 ENA ERR209567 
  O2.UC23-0 ENA ERR209572 
  O2.UC24-0 ENA ERR209576 
  V1.CD2-0 ENA ERR209704 
  V1.CD3-0 ENA ERR209723 
  V1.CD6-0 ENA ERR209743 
  V1.CD8-0 ENA ERR209750 
  V1.CD9-0 ENA ERR209752 
  V1.CD11-0 ENA ERR209682 
  V1.CD12-0 ENA ERR209684 
  V1.CD13-0 ENA ERR209686 
  
154 
  V1.CD14-0 ENA ERR209688 
  V1.UC7-0 ENA ERR209901 
  V1.UC8-0 ENA ERR209903 
  V1.UC9-0 ENA ERR209905 
  V1.UC10-0 ENA ERR209754 
  V1.UC13-0 ENA ERR209766 
  V1.UC14-0 ENA ERR209770 
  V1.UC15-0 ENA ERR209774 
  V1.UC17-0 ENA ERR209780 
  V1.UC19-0 ENA ERR209786 
  V1.UC21-0 ENA ERR209789 
 
S. sonnei reference sequences used to test SEAR 
accuracy Sh41191 ENA ERR024604 
  Sh66470 ENA ERR024605 
  Sh74369 ENA ERR024606 
  Sh55623 ENA ERR024607 
  Sh60108 ENA ERR024608 
  Sh62542 ENA ERR024609 
  Sh65179 ENA ERR024610 
  Sh65387 ENA ERR024611 
  Sh65623 ENA ERR024612 
  ShIB1 ENA ERR024616 
  ShIB690 ENA ERR024617 
  ShIB691 ENA ERR024618 
  ShIB694 ENA ERR024619 
  
155 
  ShIB2 ENA ERR024620 
  ShIB3 ENA ERR024621 
  ShIB10 ENA ERR024622 
  ShIB681 ENA ERR024625 
  ShIB683 ENA ERR024626 
  ShIB687 ENA ERR024627 
  ShIB695 ENA ERR025682 
  ShIB1970 ENA ERR025683 
  ShIB1976 ENA ERR025684 
  ShIB1980 ENA ERR025685 
  ShIB696 ENA ERR025686 
  ShIB697 ENA ERR025687 
  ShIB698 ENA ERR025688 
  ShIB713 ENA ERR025689 
  ShIB716 ENA ERR025690 
  ShIB717 ENA ERR025691 
  ShIB739 ENA ERR025692 
  ShIB748 ENA ERR025693 
  ShIB1985 ENA ERR025695 
  ShIB2009 ENA ERR025696 
  ShIB2012 ENA ERR025697 
  ShIB2013 ENA ERR025698 
  ShIB1987 ENA ERR025699 
  ShIB1990 ENA ERR025700 
  
156 
  ShIB1993 ENA ERR025701 
  ShIB1995 ENA ERR025702 
  ShIB1997 ENA ERR025703 
  ShIB2000 ENA ERR025704 
  ShIB2004 ENA ERR025705 
  ShIB2008 ENA ERR025706 
  ShIB2015 ENA ERR025708 
  ShIB3488 ENA ERR025709 
  ShIB3507 ENA ERR025710 
  ShIB3580 ENA ERR025711 
  ShIB2018 ENA ERR025712 
  ShIB2024 ENA ERR025713 
  ShIB2026 ENA ERR025714 
  ShIB2493 ENA ERR025715 
  ShIB48279 ENA ERR025716 
  ShIB3277 ENA ERR025717 
  ShIB3300 ENA ERR025718 
  ShIB3374 ENA ERR025719 
  ShIB3599 ENA ERR025721 
  Sh54213 ENA ERR025722 
  Sh54228 ENA ERR025724 
  PWR105 ENA ERR025725 
  Sh54178 ENA ERR025726 
  Sh54179 ENA ERR025727 
  
157 
  Sh54184 ENA ERR025729 
  Sh54185 ENA ERR025730 
  Sh54190 ENA ERR025731 
  Sh54210 ENA ERR025732 
  Sh658 ENA ERR025734 
  Sh1267 ENA ERR025735 
  Sh1567 ENA ERR025736 
  Sh259 ENA ERR025737 
  Sh1460 ENA ERR025738 
  Sh1461 ENA ERR025739 
  Sh1263 ENA ERR025741 
  Sh1265 ENA ERR025742 
  Sh1166 ENA ERR025743 
  Sh1167 ENA ERR025744 
  Sh1173 ENA ERR025746 
  Sh8883 ENA ERR025747 
  Sh970044 ENA ERR025748 
  Sh2073 ENA ERR025749 
  Sh273 ENA ERR025750 
  Sh373 ENA ERR025751 
  Sh1274 ENA ERR025752 
  Sh2574 ENA ERR025753 
  Sh4374 ENA ERR025754 
  Sh4474 ENA ERR025755 
  
158 
  Sh476 ENA ERR025756 
  Sh988743 ENA ERR025758 
  Sh36224 ENA ERR025759 
  Sh989560 ENA ERR025761 
  Sh9810267 ENA ERR025762 
  Sh998911 ENA ERR025763 
  Sh2225 ENA ERR025764 
  Sh5827 ENA ERR025765 
  Sh31382 ENA ERR025767 
  Sh32222 ENA ERR025768 
  20051272 ENA ERR028671 
  20061758 ENA ERR028672 
  19911483 ENA ERR028673 
  19920319 ENA ERR028674 
  19910761 ENA ERR028675 
  20041367 ENA ERR028676 
  20060018 ENA ERR028677 
  20081885 ENA ERR028678 
  20040880 ENA ERR028679 
  CS2 ENA ERR028680 
  20051541 ENA ERR028681 
  20061309 ENA ERR028684 
  CS20 ENA ERR028685 
  CS6 ENA ERR028686 
  
159 
  CS14 ENA ERR028687 
  20003593 ENA ERR028688 
  20021122 ENA ERR028689 
  20031275 ENA ERR028690 
  19984123 ENA ERR028691 
  20040489 ENA ERR028692 
  20052631 ENA ERR028693 
  19904011 ENA ERR028694 
  20011685 ENA ERR028695 
  20040924 ENA ERR028697 
  20062087 ENA ERR028699 
  20071599 ENA ERR028700 
  20010007 ENA ERR028702 
  20062313 ENA ERR028703 
  CS7 ENA ERR028704 
  CS8 ENA ERR028705 
  CS1 ENA ERR028706 
Chapter 3 Mobile genetic element reference sequences detected in metagenomes Aeromonas salmonicida plasmid pRAS3.2 REFSEQ gi|15983520 
  salmonicida plasmid pRAS3.1 REFSEQ gi|15983531 
  Aeromonas hydrophila plasmid pBRST7.6 REFSEQ gi|198286625 
  KCL-2 plasmid pMGD2 REFSEQ gi|20514397 
  Klebsiella pneumoniae plasmid pIGMS31 REFSEQ gi|209947514 
  Escherichia coli plasmid pEC278 REFSEQ gi|209947788 
  Endophytic bacterium LOB-07 plasmid pLK39 REFSEQ gi|255929160 
  
160 
  Klebsiella pneumoniae plasmid unnamed REFSEQ gi|305678726 
  Enterobacter cloacae strain BB1092 plasmid pB1023 REFSEQ gi|435855445 
  Klebsiella pneumoniae strain BB1088 plasmid pB1019 REFSEQ gi|435855463 
  U288 plasmid pSTU288-3 REFSEQ gi|482907348 
  Cronobacter sakazakii strain ATCC 29544 plasmid pCSA2 REFSEQ gi|507579660 
  Escherichia coli strain K317 plasmid ColE7-K317 REFSEQ gi|690630974 
  Enterobacter cloacae strain 34998 plasmid p34998-4.921kb REFSEQ gi|746219889 
  Enterobacter cloacae strain 34983 plasmid  REFSEQ gi|749202706 
  Klebsiella oxytoca strain M1 plasmid pKOXM1D REFSEQ gi|749293681 
  pneumoniae Kp13 plasmid pKP13b REFSEQ gi|749296055 
  Enterobacter cloacae strain 34978 plasmid p34978-4.938kb REFSEQ gi|765030385 
  Cronobacter sakazakii strain ATCC 29544 plasmid  REFSEQ gi|817657570 
  plasmid pSEM integron INTEGRALL gi|57635337 
  integron-derived beta-lactamase (ampC) INTEGRALL gi|20530945 
  class 1 integron IntI1 INTEGRALL gi|94442253 
  Lactobacillus salivarius strain JCM 1046 plasmid pCTN1046 REFSEQ gi|763126141 
  class 1 integron IntI1 INTEGRALL gi|587656492 
  class 2 integron IntI2 INTEGRALL gi|13345249 
  class 2 integron IntI2 INTEGRALL gi|788265642 
  ICE integron putative INTEGRALL gi|215397925 
  class 2 integron IntI2 INTEGRALL gi|13345249 
  class 2 integron IntI2 INTEGRALL gi|563324892 
  mercury resistance transposon INTEGRALL gi|710572 
  ICE integron putative INTEGRALL gi|215397925 
  
161 
  Bacteroides fragilis plasmid pBFP35 REFSEQ gi|194442162 
  Bacteroides cellulosilyticus WH2 plasmid pBWH2A REFSEQ gi|661525289 
  class 1 integron IntI1 INTEGRALL gi|659224469 
  Bacteroides fragilis plasmid pBFP35 REFSEQ gi|194442162 
  Sphingobium japonicum UT26S plasmid pUT2 DNA REFSEQ gi|294057975 
  Bacteroides cellulosilyticus WH2 plasmid pBWH2A REFSEQ gi|661525289 
  Aeromonas hydrophila strain AL06-06 plasmid pAH06-06-2 REFSEQ gi|766626985 
  plasmid class 1 integron INTEGRALL gi|242876676 
  ICE integron putative INTEGRALL gi|215397925 
Chapter 5 Reference sequences used in C. difficile phylogenetic tree C. difficile 630 GENBANK AM180355 
  C. difficile M68 GENBANK FN668375 
  C. difficile CF5 GENBANK FN665652 
  C. difficile M120 GENBANK FN665653 
  C. difficile BI-9 GENBANK FN668944 
  C. difficile BI-1 GENBANK FN668941 
  C. difficile 2007855 GENBANK FN665654 
  C. difficile CD196 GENBANK FN538970 
  C. difficile R20291 GENBANK FN545816 
 
Reference sequences used in C. perfringens phylogenetic 
tree C. perfringens 13124 GENBANK CP000246 
  C. perfringens SM101 GENBANK CP000312 
  C. perfringens str. 13 GENBANK BA000016 
  
162 
Appendix 5. SEAR ARG detection using clinical isolate sequence data. This table shows the ARGs detected in clinical isolate data using SEAR, a value of 1 indicates 
gene presence using the defined cut-offs and a value of 0 indicates gene absence. 
Accession Method aadA1 strB tetA tetR TEM sul1 dhfr3b catB3 dfrA5 dfrA8 dfrA12 dfrA14 aadA2 CTXM-15 OXA-1 OXA-10 
ERR024604 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR024605 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR024606 Holt et al., published 0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 
ERR024607 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR024608 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR024609 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR024610 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR024611 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR024612 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR024616 Holt et al., published 0 1 1 1 1 1 0 0 0 0 0 1 0 0 0 0 
 SEAR 1 1 1 1 1 1 0 0 0 0 1 0 1 0 0 0 
ERR024617 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR024618 Holt et al., published 1 1 1 1 0 1 0 0 0 1 0 1 0 0 0 0 
 SEAR 1 1 1 1 0 1 0 0 0 0 1 0 0 0 0 0 
ERR024619 Holt et al., published 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 
ERR024620 Holt et al., published 1 1 1 1 0 1 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 1 0 0 1 0 0 0 0 0 0 0 
  
163 
ERR024621 Holt et al., published 1 0 0 0 0 1 0 1 0 0 0 0 0 0 1 1 
 SEAR 1 0 0 0 1 1 0 1 0 0 0 0 0 0 0 0 
ERR024622 Holt et al., published 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR024625 Holt et al., published 1 0 0 0 1 1 0 0 0 0 0 1 0 0 0 0 
 SEAR 1 0 0 0 1 1 0 0 0 0 1 0 1 0 0 0 
ERR024626 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR024627 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025682 Holt et al., published 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 
ERR025683 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025684 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025685 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025686 Holt et al., published 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 
ERR025687 Holt et al., published 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 
ERR025688 Holt et al., published 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 
ERR025689 Holt et al., published 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 
ERR025690 Holt et al., published 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 
ERR025691 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025692 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
  
164 
ERR025693 Holt et al., published 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 
ERR025695 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025696 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025697 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025698 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025699 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025700 Holt et al., published 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 
ERR025701 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025702 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025703 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025704 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025705 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025706 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025708 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025709 Holt et al., published 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025710 Holt et al., published 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
  
165 
ERR025711 Holt et al., published 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025712 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025713 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025714 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025715 Holt et al., published 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025716 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025717 Holt et al., published 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025718 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025719 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025721 Holt et al., published 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025722 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025724 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025725 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025726 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025727 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 
ERR025729 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
  
166 
ERR025730 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025731 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025732 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025734 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025735 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025736 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025737 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025738 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025739 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025741 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025742 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025743 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025744 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025746 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025747 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025748 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
  
167 
ERR025749 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025750 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025751 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025752 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025753 Holt et al., published 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025754 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025755 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025756 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025758 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025759 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025761 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025762 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025763 Holt et al., published 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 
ERR025764 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR025765 Holt et al., published 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 1 0 0 1 0 0 0 0 0 0 0 
ERR025767 Holt et al., published 1 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 
  
168 
ERR025768 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028671 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028672 Holt et al., published 0 1 0 0 1 0 0 0 1 0 0 0 0 0 0 0 
 SEAR 0 1 0 0 1 0 0 0 1 0 0 0 0 0 0 0 
ERR028673 Holt et al., published 0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 1 0 1 1 0 0 0 0 0 0 0 0 0 0 0 
ERR028674 Holt et al., published 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 
ERR028675 Holt et al., published 0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 
ERR028676 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028677 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028678 Holt et al., published 1 1 1 1 1 0 0 0 0 0 0 0 1 0 0 0 
 SEAR 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 
ERR028679 Holt et al., published 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 
ERR028680 Holt et al., published 0 1 0 0 1 0 0 0 1 0 0 0 0 0 0 0 
 SEAR 0 1 0 0 1 0 0 0 1 0 0 0 0 0 0 0 
ERR028681 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028684 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028685 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028686 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028687 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
  
169 
ERR028688 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028689 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028690 Holt et al., published 1 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028691 Holt et al., published 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 
ERR028692 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028693 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028694 Holt et al., published 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 
ERR028695 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028697 Holt et al., published 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028699 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028700 Holt et al., published 0 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 
ERR028702 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028703 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028704 Holt et al., published 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028705 Holt et al., published 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
ERR028706 Holt et al., published 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 SEAR 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
  170 
Appendix 6. Evaluation of metagenome preparation methodologies for aquatic samples. 
Assessment of filtering methodology: 
To confirm that vacuum filtration of aquatic samples would capture bacteria that 
could then be subjected to DNA extraction, a 100mL suspension of an overnight 
culture of Salmonella typhimurium was vacuum filtered (approximately 2 bar 
pressure) using a 0.22 µm filter membrane. Aliquots of culture (100μL), as well as the 
filter membrane, were plated on agar pre- and post-filtration and incubated at 37oC 
for 24 hours. Plate comparison showed the unfiltered culture and filter membrane 
had grown a bacterial lawn; the filtrate plate produced no colonies, demonstrating the 
successful retention of bacterial cells from liquid suspension using vacuum filtration. 
The Meta-G-Nome DNA extraction kit was then used to examine whether it was 
possible to extract DNA from the bacterial cells retained on the filter membrane. 
Successful extraction of DNA was confirmed by spectrophotometry (DNA 
concentration = 788.5ng/μl, A260:A280 = 1.98). 
 
Comparison of DNA extraction protocols: 
DNA extraction protocols were compared to determine DNA yield and type (plasmid 
and genomic). Manufacturer protocols were followed and the extracted DNA was 
eluted in volumes and buffers specified. DNA concentrations were measured using 
spectrophotometry, recorded in triplicate and averaged (Appendix table 1). For the 
MiniPrep kit the extraction of pUC19 was confirmed by restriction digest and gel 
electrophoresis. 
 
Appendix table 1 DNA concentrations from Escherichia coli DH5α DNA extractions. 
Extraction kit Manufacturer DNA concentration (ng/µl) 
Total DNA quantity 
(ng) 
Meta-G-Nome Epicentre 48.42 2421 
PowerWater MoBio 12.02 1202 
DNeasy Qiagen 12.16 2432 
MiniPrep Qiagen 4.10 820 
 
Relative yields of genomic and plasmid DNA from the extraction protocols were 
determined using qPCR (Appendix table 2). Genomic primers amplified a region of 
leuA on DH5α chromosomal DNA; plasmid primers amplified a region of bla on 
  171 
pUC19. Standard curves generated from genomic DNA and plasmid copy numbers 
were used to calculate the ratio of genomic to plasmid DNA produced by each 
extraction protocol. 
 
Appendix table 2 Ratio of genomic to plasmid DNA copy number for DNA extractions. 
Extraction kit 
Concentration 
Genomic DNA 
(copies/μL) 
Concentration 
Plasmid DNA 
(copies/μL) 
Ratio Genomic:Plasmid 
DNA copy number 
Meta-G-Nome 3.26x104 5.82x106 1:178 
PowerWater 4.73x103 1.59x106 1:336 
DNeasy 2.68x105 6.85x106 1:26 
  
The Meta-G-Nome and DNeasy DNA extraction kits were concluded to be suitable 
for metagenome DNA extraction as they yielded sufficient genomic and plasmid 
DNA, the kits were cost effective and the protocols were quick and efficient. These 
kits were used on subsequent environmental samples. 
 
Determination of sampling requirements: 
Due to the bacterial load in aquatic samples being undetermined, several different 
approaches were trialled in order to estimate the sample volume, filtering 
methodology and extraction kit required to extract sufficient DNA for commercial 
sequencing (>2μg). 
 
Appendix table 3 Amount of DNA extracted using different sample preparation methods. 
Sample Volume (mL) 
Filter 
membrane 
size (μm) 
Extraction kit 
Concentration 
of extracted 
DNA (ng/μL) 
Total DNA 
quantity 
(μg) 
1 100 0.22 Meta-G-Nome 25.87 1.294 
2 100 0.22 DNeasy 0.21 0.042 
3 1000 0.22 Meta-G-Nome 139.42 6.971 
4 1000 0.22 DNeasy 18.53 3.706 
5 1000 3.0, 0.22 Meta-G-Nome 36.24 1.812 
6 1000 3.0, 0.22 DNeasy 3.05 0.610 
  
The results obtained from a set of samples collected on 16.04.2012 are shown in 
Appendix table 3. The Meta-G-Nome protocol consistently produced a greater yield 
than the DNeasy protocol. The use of 3.0 μm pre-filter resulted in a reduction in DNA 
  172 
yield, however gel electrophoresis and A260:A280 indicated the DNA extraction was 
cleaner (less degradation and larger fragment size); indicating the pre-filtration was 
preferable for metagenome preparation. Subsequent sample processing was 
performed using a sample volume of 10000mL (not shown).