Expression of a carotenoid‐modifying gene and evolution of red coloration in weaverbirds (Ploceidae)

Red carotenoid colours in birds are widely assumed to be sexually selected quality indicators, but this rests on a very incomplete understanding of genetic mechanisms and honesty‐mediating costs. Recent progress was made by the implication of the gene CYP2J19 as an avian carotenoid ketolase, catalysing the synthesis of red C4‐ketocarotenoids from yellow dietary precursors, and potentially a major mechanism behind red coloration in birds. Here, we investigate the role of CYP2J19 in the spectacular colour diversification of African weaverbirds (Ploceidae), represented by five genera and 16 species: eight red, seven yellow and one without carotenoid coloration. All species had a single copy of CYP2J19, unlike the duplication found in the zebra finch, with high expression in the retina, confirming its function in colouring red oil droplets. Expression was weak or undetected in skin and follicles of pigment‐depositing feather buds, as well as in beaks and tarsi, including those of the red‐billed quelea. In contrast, the hepatic (liver) expression of CYP2J19 was consistently higher (>14‐fold) in seven species with C4‐ketocarotenoid coloration than in species without (including one red species), an association strongly supported by a phylogenetic comparative analysis. The results suggest a critical role of the candidate ketolase, CYP2J19, in the evolution of red C4‐ketocarotenoid colour variation in ploceids. As ancestral state reconstruction suggests that ketocarotenoid coloration has evolved twice in this group (once in Euplectes and once in the Quelea/Foudia clade), we argue that while CYP2J19 has retained its ancestral role in the retina, it has likely been co‐opted for red coloration independently in the two lineages, via increased hepatic expression.

C4 carbon position of one or both end rings of the carotenoid molecule. This results in a monoketo-or diketo-carotenoid with peak absorptance shifted towards longer wavelengths (and "redder" hue).
The ability to perform carotenoid ketolation is thus likely an important innovation in the evolution and diversification of carotenoid pigments and coloration in vertebrates. In birds, fish and lizards, there is much evidence for sexual or social signal selection for red coloration (Hill & McGraw, 2006;Ibanez, Polo-Cavia, Lopez, & Martin, 2014;Milinski & Bakker, 1990;Svensson & Wong, 2011) and even pre-existing receiver biases (Ninnes, Webb, & Andersson, 2017). Despite this, intensely red-coloured integument (plumage, beak, skin) has a surprisingly limited and patchy distribution across birds (Aves). Even in clades where red carotenoid coloration is common, for example widowbirds and bishops (Prager & Andersson, 2010), New World blackbirds (Friedman, McGraw, & Omland, 2014) and cardueline finches (Ligon, Simpson, Mason, Hill, & McGraw, 2016), its absence in several lineages is not associated with any obvious and relevant ecological or behavioural differences from their red-coloured relatives. This suggests that some genetic or physiologically "hard-wired" constraint is at play and that C4-ketolation of integumentary carotenoids is likely to be a major hurdle for the evolution of red pigmentation.
A vertebrate C4 ketolase was proposed decades ago (V€ olker, 1962), but its genetic basis has remained unknown. Recently, however, progress was made when the locus CYP2J19, of the cytochrome P450 family of monooxygenases, was described as a putative ketolase associated with ketocarotenoid pigmentation in two independent studies of aberrantly coloured cage birds: the "yellowbeak" zebra finch mutant  and the "red factor" breed of canary (Lopes et al., 2016), in which red coloration was introgressed from the red siskin. The generality of this mechanism, however, has yet to be evaluated as these are the only cases where red integumentary coloration has been linked to CYP2J19, in a mutant and a hybrid breeding line, respectively. In addition, they differ in a) CYP2J19 gene copy number (two in zebra finch and one in the red factor canary) and b) tissue expression ("peripherally" in the integument in zebra finch and both "peripherally and centrally" in the feather follicles and liver in the "red factor" canary). There are thus many remaining questions concerning the generality, nature and location of the CYP2J19 mechanism and function in red coloration. More broadly, CYP2J19 appears to be conserved for retinal red oil droplet pigmentation within the turtles (the only other group of tetrapods to possess red oil droplets apart from the birds), from which it has been recruited for red integumentary coloration independently within certain turtle and avian lineages (Twyman, Valenzuela, Literman, Andersson, & Mundy, 2016). In particular, it remains unknown whether differential CYP2J19 expression can account for variation in red coloration across and between avian clades. Weaverbirds (Aves: Ploceidae, 116 species) are a clade of predominantly African, seed-eating passerines, which are ideal for studying the mechanisms and evolution of carotenoid coloration.
Whereas conspicuous yellow plumage colours dominate, especially in the most speciose lineages of "true weavers" (Ploceus spp), red carotenoid coloration occurs in several genera, and a few lineages lack integumentary carotenoid pigmentation altogether. The underlying mechanisms (e.g., dietary vs. metabolically modified pigments) have been established for several species, notably the brilliant yellow or red plumage displays of widowbirds and bishops (Euplectes) (Andersson et al., 2007;Prager, Johansson, & Andersson, 2009). In this clade, red colour hues are agonistic (threat) signals used in male contest competition and appear to have evolved at least twice from a yellow ancestor (Prager & Andersson, 2010) and apparently due to a pre-existing receiver bias for redder (longer wavelength) hues (Ninnes et al., 2017). Outside Euplectes, weaverbird colour signalling functions are largely unexplored, except in the red-billed quelea (Quelea quelea), where the red beak likely is sexually selected (through either female mate choice or male-male contests) whereas the polymorphic red plumage coloration may be involved in individual recognition (Dale, 2000).
In most ploceids where it has been analysed, red colour patches contain red C4-ketocarotenoids, primarily a-doradexanthin and canthaxanthin, codeposited with the dietary yellow precursor pigments (Andersson et al., 2007;unpublished results). By comparison with phylogenetically, socially and ecologically closely related yellowcoloured species, this provides an excellent opportunity to test the significance of CYP2J19 for red carotenoid coloration. Moreover, the fantailed widowbird (Euplectes axillaris) has been found to achieve its striking red wing patch coloration without C4-ketocarotenoids (Andersson et al., 2007;Prager et al., 2009), which provides an additional test of the proposed function (C4-ketolation) of

CYP2J19.
In this study of 16 red or yellow weaverbird species, we investigate the role of CYP2J19 in the evolution of carotenoid pigmentation in weaverbirds. First, we establish whether the gene is present in ploceids and, if so, in how many copies. Second, we identify the anatomical site(s) of CYP2J19 expression in this group. Finally, using a phylogenetic comparative analysis, we test whether CYP2J19 expression is associated with the occurrence of red C4-ketocarotenoid pigmentation across the ploceids.

| Samples
Feathers (for HPLC) and tissue samples (for qRT-PCR) from male ploceids in breeding plumage were largely obtained from natural populations in Africa, complemented with a few samples from aviary birds in Spain and Sweden (Table 1), under all applicable national and international permits. Five to ten feathers were plucked with flat-tipped tweezers and stored in dark envelopes until analysis.
Euthanized birds were freshly dissected and tissues placed in RNAlater (Qiagen) or DNA/RNA-shield (Zymo) until DNA/RNA extraction. Follicles for gene expression analysis were sampled from growing, carotenoid-depositing feather buds.

| C4-ketocarotenoid pigmentation
The presence of integumentary C4-ketocarotenoid pigments ( Figure 1)   | 451 liquid chromatography) analyses of feathers and beak tissue from six of the included species: Euplectes ardens, E. axillaris and Euplectes macroura (Andersson et al., 2007), Euplectes afer and Euplectes orix (Prager et al., 2009) and Q. quelea (Walsh, Dale, McGraw, Pointer, & Mundy, 2012). For the remaining species in this study, C4-ketocarotenoid presence or absence was determined from unpublished HPLC analyses performed in conjunction with the above studies, using identical or very similar methods (see Supporting Methods,

| CYP2J19 expression
Total RNA was extracted from all tissue samples using RNeasy Mini kits (Qiagen). Dissected tissues were manually homogenized using an Eppendorf homogenizer prior to addition of buffer RLT. The lysate was centrifuged for 2 min at 11,000 g in QIAshredder spin columns before proceeding with subsequent full speed centrifugation step for 3 min. DNase digestion was performed using Qiagen RNase-free DNase set.
First-strand synthesis was performed with 10 ll total RNA and N6 primer ( Hepatic expression of CYP2J19 of 16 weaverbirds, in relation to phylogeny, coloration and ketocarotenoid presence. Gene expression was normalized against b-actin, GAPDH and HPRT1, and log 10 -transformed. Expression levels of three species (E. aureus, E. axillaris and E. macroura) were undetectable after 50 PCR cycles. The phylogeny is a 50% majority-rule consensus (MRC) tree constructed in Mesquite 3.03 based on 10,000 trees downloaded from birdtree.org, numbers showing clade credibility (Bayesian posterior probability) in per cent. Discrete scores of hepatic CYP2J19 expression level (CYP h : 0 = "low," 1 = "high," comprising white and grey dots on the continuous scale, respectively) and red ketocarotenoid pigmentation (KC: 0 = "absent," 1 = "present") used in evolutionary association tests are shown. The carotenoid-based coloration of the species (red: "R," yellow: "Y," carotenoid absent: "N") is also shown Normalization following Pfaffl (2001)
Log-transformed normalized values of liver CYP2J19 expression were first discretized using k-means clustering in R version 3.3.1 (R Core Team 2016) with two cluster centres ("high" and "low"), excluding Euplectes aureus, E. axillaris and E. macrourus where expression was undetectable after 50 PCR cycles (Figure 2). Based also on results from the b-actin-normalized analyses ( Figure S1), the latter were manually scored as "low." In BAYESTRAITS V2, an "independence model" estimating four separate evolutionary rates (gain and loss for each trait) was compared to a "dependence model" allowing for a maximum of eight separate rates ( Figure 3). Assuming, however, that red ketocarotenoid pigmentation is contingent on high hepatic CYP2J19 expression, the rates of all (four) changes involving a state of ketocarotenoid presence at low CYP expression were set to zero in the final depen-

| Integumentary carotenoid pigmentation
Based on published or hitherto unpublished HPLC analyses of carotenoids in coloured feathers or beak tissue of all 16 ploceids (Table S3), each species was categorized as either "KC present" or "KC absent" depending on whether any integumentary C4-ketocarotenoids were detected (see Table S3). Whereas only the presence/absence of C4-ketocarotenoids ("modified red") was analysed in relation to CYP2J19 expression (below), it can be noted that in all seven "KC present" species, it is the same set of five C4-ketocarotenoids (a-doradexanthin, b-doradexanthin, adonirubin, canthaxanthin and astaxanthin) but in variable absolute and relative amounts. Only one species, Q. quelea, seems to lack one of the C4-ketocarotenoids, b-doradexanthin. Also noteworthy, as pointed out in Andersson et al. (2007), the one species with red coloration without any C4-ketocarotenoids, E. axillaris, has two to three times as high total concentration of carotenoids and is also the only species with the "modified yellow" carotenoid anhydrolutein. See Table S1 for further details.

| CYP2J19 presence, copy number and variation
Tissue samples from 16 species of weaverbirds (seven Euplectes, five Ploceus, two Quelea, one Foudia, one Philetairus) were analysed (see Table 1). Based on a long-range PCR assay on genomic DNA, all 16 species were found to have a single CYP2J19 gene copy of~10-15 kb, which was confirmed by Illumina Miseq sequencing in two species (E. orix and Q. quelea). Given the possibility of differential expression of copies in different tissues (as in the zebra finch), we further confirmed a single copy in ploceids by showing that fulllength sequences of CYP2J19 cDNA from different tissues (retina and liver) of the same individual were identical, in three species (E. ardens, Foudia madagascariensis and Q. quelea).
Full-length CYP2J19 cDNA sequences revealed that there were no amino acid substitutions shared among species with C4-ketocarotenoid coloration that were not present among species without C4-ketocarotenoids.

| Patterns of CYP2J19 expression
Analysis of expression using qualitative RT-PCR showed strong  Table 2). Using Q. quelea as an example as it is the only sampled species with red bare body parts (beak and tarsus), significantly higher expression of CYP2J19 was found in the liver and retina compared to the beak and tarsus ( Figure 2, Table 3).
Initial qRT-PCR quantification of hepatic expression of CYP2J19 using a single control locus (b-actin) and samples of one to three breeding males across the 16 species showed high levels of CYP2J19 in four members of the Euplectes clade (E. orix, Euplectes hordeaceus, Euplectes nigroventris and Euplectes ardens), two queleas and a fody, with levels > 100-fold greater than all other species ( Figure S1). We confirmed these findings by performing qRT-PCRs using three control loci on a randomly chosen subset of samples (one per species) ( Figure 1). These gave similar results, with the same seven species showing high (0.1-8.6) levels of hepatic CYP2J19 compared to the remaining species (<0.007) (>14-fold difference).

| Association between CYP2J19 and red ketocarotenoid pigmentation
There is a perfect association between high hepatic CYP2J19 expression and the presence of red C4-ketocarotenoids: breeding males of the seven species with high liver CYP2J19 all have red coloration due to red C4-ketocarotenoid pigments (Figure 1). In contrast, the nine species without C4-ketocarotenoids (eight of which have yellow here that produces a red colour hue based on "yellow" carotenoids alone, that is without using C4-ketocarotenoids (Andersson et al., 2007).
Phylogenetic comparative tests of correlated evolution between hepatic CYP2J19 expression and red C4-ketocarotenoid pigmentation were performed in BAYESTRAITS V2. Estimated marginal likelihoods, based on five different prior assumptions of transition rates (Table 4), consistently support a "dependence model," where the evolution of red C4-ketocarotenoid pigmentation is contingent on high hepatic expression of CYP2J19, over an "independence model," where the rate of change in one trait is unaffected by the state of the other trait ( Figure 3). Even with the most conservative priors (i.e., in favour of the "independence model"), Bayes factor test statistics (calculated as 2*[lnL(dependent model) -lnL(independent model)] exceeded 10 which is usually interpreted as very strong support for an association (Kass & Raftery, 1995).

| DISCUSSION
Our results suggest that hepatic expression of CYP2J19, a candidate carotenoid ketolase, constitutes a principal mechanism and evolutionary innovation behind red carotenoid coloration in weaverbirds (Ploceidae). As the interspecific association between high CYP2J19 expression and presence of red C4-ketocarotenoid pigments could be due to phylogenetic nonindependence (shared ancestry), the relationship was tested in a Bayesian phylogenetic comparative analysis and found to be very strong. Our results strengthen CYP2J19 as the prime candidate for the long-sought avian C4-ketolase.
We have furthermore established that weaverbirds consistently seem to have a single copy of CYP2J19. In contrast, the zebra finch, an estrildid finch belonging to the nearest outgroup clade to Ploceidae (De Silva et al., 2017), has two copies, CYP2J19A and CYP2J19B, seemingly specialized for retinal oil droplet pigmentation and integumentary coloration, respectively . It therefore appears that the estrildid CYP2J19 duplication occurred after the split between ploceids and estrildids. More broadly, a single copy of CYP2J19 was reported also in the red factor canary (Lopes et al., 2016), as well as in chicken and ostrich  and GenBank searches reveal only a single copy in the vast majority of available avian genomes (Emerling, 2018, Twyman, Andersson, & Mundy, 2018, which means that a single CYP2J19 copy probably is the "normal" situation for birds. The tissue-specific expression data for CYP2J19 strongly implicate the liver as the main site of carotenoid ketolation. As earlier suggested by high plasma concentration of red ketocarotenoids (Prager et al., 2009;unpublished results), ploceids thus seem to be "central" ketoconverters. Notably in this context, the hepatic CYP2J19 expression was very low in E. axillaris, a species with red carotenoid coloration that does not involve C4-ketocarotenoids (Andersson et al., 2007). Apart from implying an intriguing alternative "redness mechanism" (possibly related to the in birds unusual presence of "an- In contrast to the liver, CYP2J19 expression was very low or undetectable in peripheral tissues (skin, feather follicles, beak, tarsus), including the red and ketocarotenoid-pigmented beak and legs of the red-billed quelea. Nevertheless, more extensive and careful sampling, covering a broader range of feather growth stages, will be required to rule out the possibility of a "peripheral" (integumentary) role of CYP2J19 in feather follicles, as implicated in the red factory canary (Lopes et al., 2016).
There was substantial variation in hepatic CYP2J19 expression overall, not least among the ketocarotenoid-coloured species, where by far the highest expression was found in the red-billed quelea. As this is the only of our study species that has a red-coloured beak (and tarsi), we speculate that, compared to plumage, this continually renewing tissue may require a more constant and larger supply of ketocarotenoids to maintain its red colour. Most of the variability of CYP2J19 expression among red species, however, had no such obvious association with phenotype and probably relates to timing of sampling in relation to the prenuptial plumage moult, or to some other genetic, social or environmental factor. For example, given that cytochrome P450 enzymes often are regulated by substrate availability (Zanger & Schwab, 2013), CYP2J19 expression is likely affected by both amount and composition of carotenoids in the diet.
Further studies of inter-as well as intraspecific variation in CYP2J19 expression, with carefully controlled and standardized sampling, are needed to explore whether some of this variation is biologically meaningful, for example by suggesting physiological costs or tradeoffs with detoxification (see Mundy et al., 2016) that may mediate honest signalling.
Historically, there has been considerable debate over the anatomical site of ketolation (Del Val et al., 2009;McGraw, 2004) and even with a few examples, it is now apparent that there is substantial variation in the strategy employed by different passerine species. The contrast between the red-billed quelea and zebra finch, which both have red beak and tarsus, is particularly striking: the former has high CYP2J19 expression in liver and low/absent expression in beak and tarsus, while the zebra finch shows the opposite pattern.
T A B L E 4 Estimated marginal log-likelihoods (lnL) of "dependency" (dep) versus "independency" (indep) models, and the Bayes factor test statistic (2lnBF), given different prior assumptions of evolutionary transition rates

Rate distribution prior lnL (dep) lnL (indep) 2lnBF
Uniform ( Unlike the situation with two copies of CYP2J19 in the zebra finch , an estrildid that uses peripheral ketoconversion to colour its bill and tarsi, the 16 ploceids in this study all appear to have a single copy of CYP2J19 and central (liver) ketoconversion, supplying either plumage or, in the red-billed quelea, beak and bare part coloration. The red factor canary (a hybrid fringillid) likewise has a single CYP2J19 copy with both central expression and peripheral expression (Lopes et al., 2016), although this needs to be confirmed in a natural fringillid species. Broader sampling and further study of interspecific variation in CYP2J19 copy number and site(s) of action may yield interesting evolutionary implications as regards micro-and macroevolutionary constraints on colour and pattern diversification.
Based on a previous ancestral character state reconstruction (Prager & Andersson, 2010), the two clades with high hepatic genase" (i.e., the ketolase), mediated by cis-regulatory elements, was also suggested to explain the evolution of C4-ketocarotenoid pigmentation in Colaptes woodpeckers (Hudon, Wiebe, Pini, & Stradi, 2015). Given the relatively rare but phylogenetically widespread occurrence of red carotenoid coloration, the co-option of CYP2J19 seems to have occurred several times independently in birds and also in turtles , but a scenario with early gains and multiple subsequent losses may well emerge as further lineages are investigated. It is also important to note that we have only considered a single aspect of carotenoid coloration, the presence of integumentary C4-ketocarotenoids; several other mechanisms, for example uptake, metabolism, transport and deposition, may also be key factors behind interspecific colour variation.
Cis-regulatory evolution is often regarded as a major motor behind adaptive change (Stern & Orgogozo, 2008 Andersson, 2010). This may indicate that other locus-specific factors contribute to the constraint, which may include coordination of expression in relation to age, sex, body condition and season, potentially requiring the evolution of multiple cis-regulatory modules for CYP2J19. Moreover, red ketocarotenoid-based coloration also has a sparse distribution among turtles (the only nonavian group shown to possess CYP2J19), and whereas less is known about selection for red coloration in this group, locus-specific genetic constraints may also explain some of the patterns of interspecific colour variation in the turtles. Hence, elucidating and disentangling potential constraints on the evolution of carotenoid coloration in animals will require detailed investigation of the genetic and environmental causes and consequences of co-opting the CYP2J19 for integumentary pigmentation.
Rapid progress has recently been made in documenting the genetic basis of convergent evolution of naturally selected traits (Stern, 2013). For example, in birds, evolution of melanin-based coloration in birds is frequently due to two loci, MC1R and ASIP (Mundy, 2005;Toews et al., 2017;Uy et al., 2016). Here, we have uncovered one of the first examples in vertebrates where a locus is involved in convergent evolution of a sexually selected trait. Future work on CYP2J19 promises many novel insights into both function and evolution of carotenoid coloration in birds, as well as general questions regarding diversification due to differential selection or differential constraints.