Clonal Dynamics in Barrett's Oesophagus and Oesophageal Adenocarcinoma

Abstract

Introduction Outcomes in oesophageal adenocarcinoma (OAC) are poor, largely driven by most patients being diagnosed at a late stage. OAC has a metaplastic pre-cursor lesion, Barrett’s oesophagus (BO), which can be monitored and effectively treated if dysplasia develops. Screening for BO could improve outcomes but needs to be balanced with the consequences of over-diagnosis and over-treatment since most BO will not progress. BO is a mosaic epithelium, and the precise definition of BO varies in clinical guidance. In other countries, only intestinal-type metaplasia (IM) BO undergoes monitoring, but in the UK gastric-type metaplasia (GM) is also considered. Both IM and OAC are known to be highly heterogeneous, but while this heterogeneity has been well studied for IM, little is known about the clonal dynamics in GM and OAC. Clonal dynamics can be explored through longitudinal or multi-sample genomic studies, or experimentally. There is no physiologically relevant in vivo model for BO/OAC, but patient derived 3D organoids may provide an appropriate in vitro system for studying clonal dynamics, including competition between different stages of the disease. The evolution of IM through to OAC is still being explored. To date we know that progression is characterised by the acquisition of driver genes, increasing mutational burden, and increasing copy number alterations. The earliest, most frequent changes are loss of heterozygosity (LOH) events, which start to be observed in non-dysplastic IM. The most frequently mutated driver gene is TP53, with this mutation typically occurring in dysplastic BO. Hypothesis A deeper understanding of clonal dynamics and evolution within BO and OAC could give a better understanding of prognostic differences including progression potential of premalignant states and survival of patients with invasive tumours. This could inform clinical management both of patients under BO surveillance and patients with OAC. Aims • Assess the malignant potential of GM compared with IM in the evolution to OAC • Investigate feasibility of assessing clonal dynamics in BO and OAC using organoid models • Examine the role of LOH in the evolution of BO and OAC • Analyse intra-tumour heterogeneity and evolution within OAC Materials and methods A comparative genomic analysis was performed on 160 GM and IM specimens, using whole genome and whole exome sequencing. This was combined with an assessment of clinical outcomes following GM or IM diagnosis, using a thorough pathology and endoscopy review. A novel clinical outcomes cohort comprised of 244 GM and IM patients, across 1854 person years of follow-up. The growth and morphology of three different organoid lines were compared: one non-dysplastic BO, one dysplastic BO, and one OAC line. These had been grown in in Matrigel, using IntestiCult Organoid Growth Medium, and imaged using the Zeiss live cell microscope. The organoids were measured using the Fiji distribution of ImageJ. An in silico model of organoid growth and morphology development was built using CompuCell3D. LOH progression was assessed using whole genome sequencing on a cohort of 100 pre-OAC samples, encompassing IM, dysplasia, and IMC, as well as a cohort of 81 pairs of BO (IM and dysplasia) and OAC samples. Intra-tumour heterogeneity and evolution were assessed using whole exome sequencing of multiple samples, from 93 chemo-naïve OAC tumours. Each tumour had been macro-dissected to yield multiple samples. Mutations across the samples were clustered and the resulting clusters were used to derive phylogenetic trees. Heterogeneity and evolution measures were calculated from the clonality of mutations and copy number alterations, as well as from the clusters and trees. Results Clinical outcome analysis revealed that progression to dysplasia or OAC was very low in GM, and significantly lower than in IM (p=0.03). Genomic analysis showed that unlike IM, GM did not bear mutational hallmarks of OAC, with a significantly lower mutational burden (p<0.01), typically no detectable mutations in OAC driver genes and little evidence of Signature 17a/b. Image analysis of organoids revealed a spectrum of morphologies from hollow in IM through to solid in OAC. In silico modelling demonstrated that these morphology differences could be explained by the disruption of polarity that is characteristic of progression of this disease. A small set of genomic loci were identified with LOH consistently present in IM. By contrast, LOH prevalence progressed steadily across dysplasia to OAC on 18q21 (SMAD4) and a small set of other cytobands, with mutation rates in relevant driver genes seeming to lag behind LOH prevalence. Moreover, LOH on 18q21 was associated with worse outcomes in OAC, even in the absence of a SMAD4 mutation. Mutational and copy number intra-tumour heterogeneity was highest in Stage I tumours, driven primarily by fewer clonal events than observed in later stage tumours. For example, Stage I tumours typically only had one clonal mutation in an OAC driver gene, with this increasing to two in Stage II- III, and three in Stage IV. Heterogeneity was not found to be prognostic, beyond its association with stage. Discussion Genomic analysis and clinical outcomes demonstrated that the malignant potential of GM is low, and lower than that of IM. As such, it is my recommendation that short-segment GM is not included in BO surveillance. This would lead to a change in UK clinical guidelines and bring them in line with other international guidance. The strikingly different morphologies across BO and OAC organoid models coupled with their polyclonal nature need consideration in clonal dynamics experiments. I recommend working with monoclonal organoids, and flow-sorting and counting to accurately assess clonal dynamics. The role of LOH in driving progression is not resolved, but it has potential for detecting and measuring progression of the disease. My work highlights genomic locations where either haploinsufficiency, methylation or mutations in unidentified tumour suppressor genes are being selected for, and these locations should be explored further. Finally, OAC is a highly heterogeneous disease, with the level of heterogeneity varying in a non-linear manner through the disease course. The high heterogeneity observed in early tumours suggests that further study of Stage I tumours could reveal more about fitness of different combinations of driver gene mutations. In conclusion, clonal dynamics are in play from IM through the progression to OAC and beyond. These dynamics could be studied further through organoid experiments, but genomic analysis remains most informative at this time. There is a lot of potential for further study of the role LOH is playing in disease progression and of fitness of clones in early OAC.