Divergence and Functional Degradation of a Sex Chromosome-like Supergene

SUMMARY

A major challenge in biology is to understand the genetic basis of adaptation. One compelling idea is that groups of tightly linked genes (i.e. ‘supergenes’ [1, 2]) facilitate adaptation in suites of traits that determine fitness. Despite their likely importance, little is known about how alternate supergene alleles arise and become differentiated, nor their ultimate fate within species. Herein we address these questions by investigating the evolutionary history of a supergene in white-throated sparrows, Zonotrichia albicollis. This species comprises two morphs, tan and white, that differ in pigmentation and components of social behavior [3–5]. Morph is determined by alternative alleles at a balanced >100Mb inversion-based supergene, providing a unique system for studying gene-behavior relationships. Using over two decades of field data we document near-perfect disassortative mating among morphs, as well as the fitness consequences of rare assortative mating. We use de novo whole genome sequencing coupled with population- and phylo-genomic data, to show that alternate supergene alleles are highly divergent at over 1,000 genes, that these alleles originated prior to the split of Z. albicollis from its sister species, and may be polymorphic in Z. albicollis due to a past hybridization event. We provide evidence that the ‘white' allele may be degrading, similar to neo-Y/Wsex chromosomes. We further show that the ‘tan’ allele has surprisingly low levels of genetic diversity, yet does not show several canonical signatures of recurrent positive selection. We discuss these results in the context of the origin, molecular evolution, and possible fate of this remarkable polymorphism.

Graphical Abstract

RESULTS AND DISCUSSION

Understanding how suites of adaptive characters remain linked despite the disruptive forces of recombination has been a major challenge in evolutionary biology. ‘Supergenes’ are linked clusters of coevolved genes that, like sex chromosomes, give rise to divergent fitness-related traits that are variable within species [1, 2]. Nonetheless, how supergenes originate, how they generate adaptive variation, and how they are maintained and evolve within a species remains unclear. To examine these basic questions, we analyzed supergene evolution in the white-throated sparrow (Zonotrichia albicollis). The two plumage morphs (Figure 1A) of the white-throated sparrow differ dramatically in reproductive behavior [3, 5], representing two extremes in reproductive tradeoffs: white males are promiscuous and invest heavily in securing additional matings at the expense of paternal care [3] whereas tan males are monogamous [3] and contribute more to parental care [6]; females exhibit similar trade-offs between investment in parental care versus mating effort [3].

Figure 1. White-throated Sparrows comprise two morphs with an ancient origin.

(A) Tan (T) and white (W) morph sparrows (Note differences in plumage color in the head stripe and throat) differ in chromosome 2 (sometimes referred to as ZAL2) genotype and mate almost exclusively with the opposite morph, maintaining polymorphism in the species. (B) Tan (2) and white (2^m) chromosomes are highly divergent within the inverted portion of chromosome, suggesting an ancient origin of the chromosomes 2 and 2^m. All nodes are supported by 100% bootstrap values and the depicted topology for the inverted region is a significantly better fit to the data than trees constrained to the monophyly of 2 and 2^m (p < 0.001, see Supplementary Materials). Taxa marked with an asterisk are those for which whole genomes were sequenced. The arrows highlights the ancestral node for the Zonotrichia genus. See also Table S2.

Our long-term genotypic analysis builds on previous work [6, 7], and through extensive genotyping of thousands of individuals over more than two decades, confirms that white morphs are almost always heterozygous for alternative chromosome 2 alleles (2^m/2). We find that 99.7% of white morphs are heterozygous (n=1,014, Table S1), whereas tan morphs are always homozygous (2/2; 100%, n=978). Thus, chromosomes 2 and 2^m segregate at ~75 and ~25% frequency, respectively, and the two plumage morphs occur in approximately equal numbers. Chromosome 2 remains polymorphic via near-obligate mating between tan and white morphs (i.e., disassortative mating) [8]. Only 17 out of 1,116 pairs (1.5%) observed in this study mated with an individual of the same morph (i.e. homotypic or assortative mating). Homotypic pairs are even less common in primary pairings (9/1106, 0.8%), but occur more commonly in secondary pairings (8/10, 80%; Table S2) following the disappearance of an opposite-morph mate and when mate-choice options are limited. As a consequence of obligate disassortative mating the species effectively has four sexes, wherein any individual can mate with only 1/4 of the individuals in the population.

Given the myriad phenotypic and behavioral differences controlled by this inversion polymorphism [3], and the unusual dynamics of the mating system, the white-throated sparrow harbors a classic example of a balanced supergene [9]. To resolve the evolutionary history of this supergene, we first sequenced and assembled the genome of a single tan male (homozygous 2/2 and Z/Z; Figure 1), resulting in a draft genome 1.03Gb long, comprising 6,018 scaffolds, 13,811 protein-coding genes, 1,104 non-coding transcripts, and 205 pseudogenes. The assembled genome has a contig N50 of 113kb and a scaffold N50 of 4.9 megabases, comparable or better than most recent short read-based bird assemblies [10].

Pooled whole-genome resequencing of 24 tan and 25 white males revealed a bimodal distribution of scaffold-wide average F_ST between morphs estimated from 18.4 million biallelic SNPs and polymorphic indels (collectively, "variants”; Figure S1). Previous studies [11] revealed high differentiation between 2 and 2^m, and we reasoned that high F_ST scaffolds would correspond to regions within the chromosome 2 inversion. We confirmed this by associating a subset of assembled scaffolds with FISH-mapped bacterial artificial chromosomes (BACs). Scaffolds associated with BACs known to map within the inversion had correspondingly high F_ST (median = 0.3; Table S3). The scaffolds that map to the region near the distal end of the inverted region on 2 have decreased average F_ST (~0.2, Figure 2A, Table S3) and a significantly higher rate of shared polymorphism between 2 and 2^m (odds-ratio=5.73, Fisher’s Exact Test p-value < 1e-10) suggesting that limited recombination or gene conversion occurs between 2 and 2^m in this region. However, the extent of genetic exchange in this region or the rest of the inverison is presently unknown. High F_ST (≥ 0.2) scaffolds include 1,137 genes that comprise the supergene. F_ST is expected to be 0.33 at variants that are fixed between 2 and 2^m in comparisons between tan (2/2) and white (2/2^m) birds. Approximately 78.6% of the 1.5M variants (or ~1/100 bp) within the inversion have F_ST of ~0.33 (Supplementary Figure 1B), demonstrating that the majority of variants in the inversion are heterozygous among all the white birds we sampled. To confirm fixed differences between 2^m and 2 identified in pooled genotypes, we resequenced the genomes of three individual white morph birds sampled from a different population than the pooled sample. We find that 75% of variants within the inversion are heterozygous in all three birds; note, given limited sequencing depth of individual birds (~10X) we expect that a fraction of truly polymorphic variants to appear homozygous thus explaining the discrepancy between the rate of fixed differences (75% versus 78.6%) observed in our individual and pooled based resequencing, respectively. In addition, we estimated linkage disequilibrium (r²) within the pooled samples. As expected, r² is ~ 1 within the inverted region among the pool of white birds (Fig. 2C).

Figure 2. Contrasting divergence, diversity and linkage disequilibrium inside (pink area) and outside the inverted region.

Sliding window estimate across FISH-mapped scaffolds of (A) F_ST between white and tan birds and (B) genetic diversity along the tan allele, 2. The grey bar represents the 95% confidence band (standard deviation) of neutral expectation for genetic diversity . (C) Linkage disequilibrium decays rapidly outside of the inversion, but is high, especially in 2/2^m heterozygotes, within the inversion. Using golden-crowned or Harris’ sparrow as outgroups, (D) Fay and Wu’s H estimates do not show a strong signal of recent selective sweeps in the inverted region of chromosome 2. See also Figures S1, S3 and Table S3.

The inversion contains a number of well-studied genes related to the neurophysiological control of behavior, including multiple steroid hormone receptors (Table 1). ESR1 has already been implicated as contributing to behavioral differences between white-throated sparrow morphs [12], and VIP has been shown to mediate variation in aggressive behavior in other songbirds [13, 14]. Within this gene complex are also two genes with annotated roles in pigmentation (FIG4 and LYST, [15]). We note, however, that none of these genes show signatures of recent selection (see below) or unusual patterns of divergence relative to others genes within the inversion. Indeed the high levels of divergence throughout the inversion provide a challenge in resolving the relative contributions of specific genes to the suite of phenotypic differences among morphs.

Table 1.

Candidate genes for behavior and pigmentation

Gene	Gene Name	Exon (NS/S)	Intron	UTR (5’/3’)	UP	DOWN
CGA	Glycprotein Hormones, Alpha Polypeptide	0/1	34	0/4	51	71
ESR1	Estrogen Receptor 1	2/4	2565	0/14	53	54
ESRRG	Estrogen Related Receptor Gamma	1/3	2465	0/17	52	68
FSHR	Follicle Stimulating Hormone Receptor	6/11	1291	0/8	112	5
HTR1B	Serotonin Receptor 1B	0/4	1	7/3	46	59
HTR1E	Serotonin Receptor 1E	2/7	0	1/4	78	65
LHCGR	Luteinizing Hormone/Choriogonadropin Receptor	5/14	285	0/0	103	44
VIP	Vasoactive Intestinal Peptide	0/0	79	4/1	99	104
FIG4	FIG4 homologue	3/11	470	3/	154	67
LYST	Lysosomal Trafficking Regulator	19/48	931	2/19	56	0

Fixed differences between chromosomes 2 and 2^m in candidate genes for behavior and pigmentation (FIG4 and LYST). Polymorphisms are divided by site class including exons (nonsynonymous/synonymous substitutions), introns, UTR (5’ UTR/3’UTR), and 1 kb UP and DOWN-stream of start and stop codons, respectively.

The high level of nucleotide divergence between chromosomes 2 and 2^m suggests that the divergence of these chromosomes predates divergence of white-throated sparrow and its sister species. To investigate the long-term evolutionary history of chromosomes 2 and 2^m, we examined genome-wide patterns of divergence between the white-throated sparrow and its close relatives. For comparison, we analyzed closely-related taxa with previously [16, 17] or newly sequenced transcriptomes or genomes (Figure 1B). As in previous studies [11], we find that the relationship of chromosomes 2 and 2^m conflicts with the species tree and suggests more recent common ancestry of chromosome 2 with other taxa than with 2^m. Our genome-scale analyses and complete taxon sampling of the Zonotrichia genus revealed that outside the inversion white-throated sparrow is sister to Harris’ sparrow. This same relationship holds for the tan (2) allele within the inversion. Inclusion of rufous-collared sparrow in our analysis allowed us to resolve that the white 2^m allele arose after the origin of Zonotrichia and falls sister to the rest of the clade.

Because 2^m exists almost exclusively in a non-recombining heterozygous state (Table S1), it has been hypothesized that 2^m may be degrading in a manner that is similar to neo-Y/W chromosomes [18, 19]. Hyunh and colleagues [18], using a smaller set of markers, failed to find an enrichment of non-synonymous changes on 2^m, and thus concluded that the chromosome was not degrading. In contrast, here we provide evidence consistent with functional degradation of 2^m by using genome-wide estimates of polymorphism and divergence at non-synonymous and synonymous sites (Figure 3A). We calculated the Direction of Selection (DoS) statistic [20], which is conceptually related to the McDonald-Kreitman test, to compare the relative rates of non-synonymous and synonymous polymorphism to non-synonymous and synonymous divergence. On average, genes inside the inverted region that are linked to 2^m have a significantly negative DoS (p_{Golden-crowned} < 0.001; p_Harris’ < 0.001) indicating an excess of non-synonymous polymorphism and consistent with functional degradation of 2^m. Outside the inversion, DoS is slightly positive, indicating a small excess of non-synonymous fixed differences between the white-throated sparrow and either of two out-group species (p_{Golden-crowned} = 0.059; p_Harris’ < 0.001). Genes inside the inverted region that are linked to 2 do not show a significant deviation from expectation based on a randomization test (p > 0.05 for both outgroups). Intriguingly, fixed differences between 2 and 2^m are only slightly more likely to be non-synonymous relative to expectations from interspecific divergence (p_{Golden-crowned} < 0.001; p_Harris’ < 0.001; see Figure 3A, ‘Inv. (2/2m)’). This latter result suggests that chromosome 2 has relatively recently become polymorphic in the white-throated sparrow: long-term heterozygosity of 2^m would likely lead to fixation of putatively deleterious non-synonymous mutations giving rise to a DoS between 2 and 2^m that is similar in magnitude to the DoS statistic based on polymorphisms segregating on 2^m.

Figure 3. Functional degradation of 2^m.

(A) Using two different outgroups (diamond = golden-crowned sparrow, circle = Harris’ sparrow), the direction of selection statistic is strongly negative on 2^m, indicating an increased level of non-sysnonymous polymorphism. (B) For genes within the inversion, gene expression in higher for tan birds than heterozygous (2/2^m) white-morph birds and (C) in white birds, the white (2^m) allele tends to be under-expressed relative to the tan (2) allele (**: p < 0.01; ***: p < 0.001). Error bars represent 95% confidence intervals (standard error). See also Figure S2, S4, and Table S1.

In neo-sex chromosomes, functional degradation is accompanied by decreased average gene expression of Y/W linked genes [21, 22]. We contrasted gene expression levels between three tan and three white birds and also measured patterns of allele specific expression within white birds. On average, the expression level of genes within the inverted region is lower in white birds than tan birds (Figure 3B) and expression of the allele linked to 2^m is lower than the allele linked to 2 within heterozygous white birds (Figure 3C). Differences in gene expression between 2- and 2^m-linked genes are subtle (Figure S2) and may reflect tissue-specific effects or a relatively recent degradation of 2^m. Decreased expression of 2^m linked genes is not an artifact of increased general expression of 2 linked genes, as seen by a comparison of log-transformed expression levels of genes inside the inverted region versus the rest of the genome within tan birds (t = 0.0324, p = 0.97). While differences in expression levels of some genes inside the inverted region likely contributes to adaptive differentiation between morphs, our results suggest that 2^m is degrading.

As in sex chromosomes, we hypothesized that pairwise genetic diversity (Θ_π) on 2 and 2^m would be 75 and 25% of that in the rest of the genome [23]. In contrast, Θ_π on 2 is only ~30% of background diversity (0.0007 vs 0.0023; Figure 2B) and Θ_π on 2^m is only ~20% of background (0.00047 vs. 0.0023; Supplemental Figure S3), indicating that diversity on these chromosomes is not at the expected equilibrium under a simple neutral model. In sex chromosomes, departures from equilibrium diversity ratios have been attributed to multiple forces including differences in life-history between sexes [24] and differences in how natural selection and genetic drift operate. One component of life history that could influence patterns of diversity is differences in the variance of reproductive success between morphs. In white-throated sparrows, if tan birds have a higher variance in reproductive success the ratio of diversity between 2 and the genomic background will be >75% whereas if white birds have a higher variance in reproductive success the ratio will be <75% [24]. To assess whether the observed deviations from expected patterns of genetic diversity could be attributed to life-history variation among morphs, we calculated lifetime reproductive success among 255 white and tan adult birds. We find no significant difference in the variance of reproductive success among morphs (Levene Test, p-value = 0.84). Thus life-history variation among morphs may play a limited role in shaping patterns of genetic variation between the inverted and non-inverted regions of the genome, however we cannot reject more complicated demographic scenarios at the present time [25].

The differential action of natural selection is another factor that can influence the ratio of genetic diversity between the X/Z and autosomes and, by extension, chromosomes 2 and 2^m. On the white allele, 2^m, which is analogous to Y/W chromosomes, the lack of recombination makes discerning signatures of selection challenging, so we focused analyses of selection on the tan allele, 2. On 2, like X/Z chromosomes, beneficial mutations may be more likely to go to fixation (positive selection) and deleterious mutations may be more likely to be purged (purifying selection) because of functional hemizygosity in white 2/2^m birds [24]. Either positive or negative selection would lead to a decrease in genetic diversity on 2 and could possibly explain the substantial reduction of diversity and elevated linkage disequilibrium observed on this chromosome (Figure 2B). Under a scenario of wide-spread positive selection along the inverted region of 2, we would expect a substantial increase in the proportion of high-frequency derived alleles which would be revealed by negative values of Fay and Wu’s H [27]. We calculated Fay and Wu’s H using either Harris’ or golden-crowned sparrow as outgroups. H is slightly more negative inside the inverted region of 2 than outside but does not show an excess of extremely negative H values (Figure 2D). Wide-spread positive selection along the inverted region of 2 may also lead to a slight excess of fixed non-synonymous differences however no such pattern is observed (Figure 3A). While positive selection has likely shaped patterns of genetic variation at certain loci the inverted region of 2, current tests fail to resolve systematic evidence for increased rates of positive selection on 2. The reduction in diversity along 2 relative to the genomic background could therefore be influenced by strengthened purifying selection arising as a consequence of functional degradation of chromosome 2^m, strengthened background selection arising a consequence of reduced recombination, as well as occasional selective sweeps at specific loci.

Given the strength of disassortative mating and evidence for degradation of chromosome 2^m, we hypothesized that there are fitness consequences associated with homotypic mating. White male × white female (W×W) homotypic pairs are expected to produce, on average, 25% “superwhite” 2^m/2^m homozygotes, potentially exposing deleterious recessive alleles that have arisen on 2^m. In 27 years of study, we have found only three superwhite 2^m/2^m birds out of 1,989 (0.15%) genotyped birds. Based on the observed frequency of mating between white morph males and females, both via social pairing and extra pair mating, we expected 12 super whites in our sample (see Supplementary Experimental Procedures). The lower than expected number of observed superwhites (Fisher’s Exact Test p = 0.035) suggests that the 2^m/2^m genotype is deleterious (Table S1), that parents reduce investment in such offspring [27], or both. While the two observed female superwhites appeared normal in size, the one superwhite male we found was over 2.5 standard deviations smaller than all other age-matched chicks, and is the smallest male nestling sampled to date (Figure S4), suggesting that 2^m/2^m might be more deleterious in males. Occasional recombination on 2^m when superwhites do survive and reproduce may also mitigate the degradation of this chromosome.

Homotypic pairings of both types (W×W and T×T) suffer additional fitness costs. For example, different pairings (W×W, T×T, T×W, W×T) suffer different rates of extra-pair paternity in their nests (F_3,380 = 20.10, p < 0.0001, Table S4). Like white male × tan female (W×T) pairs, both W×W and T×T pairs experience a significant (~25%) increase in extra-pair paternity in their nests relative to tan male × white female (T×W) pairs (Tukey-Kramer post hoc tests, p < 0.05). We also find that homotypic pairs show a tendency towards producing female-biased clutches (homotypic pairs (N=12) = 37 ± 9 % males, heterotypic pairs (N=284) = 52 ± 1 % males, F_1,294 = 2.66, p = 0.10; Table S4). Since many species adaptively adjust offspring sex ratios to produce a higher proportion of offspring of the smaller sex (females, in the case of the white-throated sparrow) when they are in poor condition or under stress [28], such data suggest that homotypic pairs are at a disadvantage. Lastly, although chicks raised by T×T pairs are similar in mass to disassortative pairs, chicks raised by W×W pairs are significantly smaller than those raised by T×T pairs (F_1,30 = 11.57, p = 0.0019), likely a result of reduced parental care by white morph birds. This pattern is primarily driven by mass differences in male chicks, as sons from W×W pairs are significantly smaller than sons from T×T pairs (F_1,12=7.45, p = 0.01), whereas the mass of daughters does not differ by pair type (F_1,16 = 2.85, p = 0.11). These costs associated with homotypic pairing may contribute to the maintenance of the chromosomal polymorphism in the species by favoring disassortative pairing.

Two plausible models explain the presence of both 2 and 2^m in white-throated sparrows subsequent to their ancient divergence: either these chromosomes were polymorphic in an ancestral Zonotrichia species but remain so only in white-throated sparrows, or one of the chromosomes entered white-throated sparrow via hybridization followed by adaptive introgression (Figure 1B). Previous studies favored the former scenario [11] because no potential hybridizing taxon was found [11]. As we discuss above, however, patterns of polymorphism on chromosome 2^m (Figure 3A) suggest that this chromosome may have become polymorphic in the white-throated sparrow relatively recently. While the lack of a clear sister species to 2^m is challenging for this interpretation, complete genus-level taxon sampling suggests that the donor species of 2^m may be extinct. In support of the model that 2^m is polymorphic in the white-throated sparrow via introgression, we note that hybridization between various Zonotrichia species have been observed in the wild [29] and inferred from discordance between mitochondrial and nuclear phylogenies [30]. Origination of supergene alleles in separate species provides a simple mechanism by which coadapted gene complexes can be formed, and under which recombination suppression between alleles would be expected when heterokaryotypes arise via hybridization [9].

Chromosomes 2 and 2^m in white-throated sparrows are independent of sex chromosomes, but show striking parallels with patterns of sex chromosome divergence and degradation [31, 32]. Indeed, because of disassortative mating based on both chromosomes 2 and 2^m and the W and Z sex chromosomes, the species operates as if there are four sexes. Sexual systems with more than two sexes are exceedingly rare among animals and theory predicts that they are unlikely to persist for long periods of time [33]. In part, this instability arises because four sex systems have a two-fold increase in some aspects reproductive effort, such as finding a mate. If such four-sex systems are truly unstable, the persistence of the inversion based plumage morph polymorphism found in the white-throated sparrow may be transient, despite the observed fitness benefits of disassortative mating.