Population structure
The final LD-filtered SNP set contained 25,532 markers, of which 24,574 were located on the autosomes, and 958 were located on the X chromosome. Pairwise IBS values between samples averaged 91.52% (range 88.41%-97.01%). By contrast, IBS values estimated for DNA replicates approached 100% (average 99.91%; range 99.84%-99.96%), as expected if the genotyping error rate is low (i.e., below 0.16%, as estimated using the nine replicates). The STRUCTURE analyses performed using the 24,574 autosomal SNPs confirmed our previous results on genomic variation in these populations (Bock et al., 2021). The delta K criterion indicated that two genetic clusters (K = 2) are the best fit for our data (Figure S5a-c). As expected, all individuals were assigned ancestry from both genetically distinct clusters in varying proportions. Based on information from mtDNA sequences (Kolbe et al., 2004, 2007a) and genome-wide SNP data from the native range (Bock et al., 2021), these clusters can be interpreted as representative of ancestry from native lineages in Western and Central-eastern Cuba. Southern Florida populations have higher frequencies of Western Cuba ancestry, whereas central to northwestern Florida populations are highly admixed and contain higher frequencies of Central-eastern Cuba ancestry (Figure 2). Note, however, that DAPC indicated that a clustering ofK = 3 could be a marginally better fit (Figures S6-S7).
Associations between the dewlap, genetic ancestry, and the environment
We found significant correlations between dewlap traits, genetic ancestry, and environmental variables. Dewlap variation as represented by PC1 and PC2 showed significant positive correlations with Western Cuba ancestry (P < 0.001, Table 1). As the frequency of Western Cuba ancestry increased, lizards exhibited brighter and redder dewlaps (Table 1, Figure 1b, Figure S8a,b). Additionally, PC1 was negatively correlated with canopy openness (P = 0.001; Table 1). As canopy openness increased (i.e., less canopy cover and lighter conditions), lizards exhibited darker dewlaps.
To further investigate the effects of genetic ancestry and environmental variation on distinct aspects of the dewlap and to follow-up on the PC-based analyses, we built separate models for each dewlap trait (i.e., univariate tests), including UV reflectance, total brightness, hue (cut-on wavelength), color composition (red, orange, yellow), area, and perimeter. Spectral colorimetric variables differed across dewlap positions; therefore, further analyses were also done by position (i.e., P1, P2 and P3). UV reflectance was negatively correlated with Western Cuba ancestry across dewlap positions (P1–edge, P = 0.003; P2–edge, P = 0.017; P3–center, P = 0.003; Table 1). Center UV reflectance showed significant positive correlations with canopy openness (P2, P < 0.001; P3, P = 0.021) and annual mean precipitation (P2, P = 0.032). Total brightness was positively correlated with Western Cuba ancestry (P1, P2 and P3, P < 0.001), negatively correlated with canopy openness (P1, P < 0.001; P2, P = 0.028), and negatively correlated with temperature (P3, P = 0.028). These univariate results were consistent with PC analyses, such that PC1 can be interpreted as brightness. Both PC1 (Figure S8a,l) and total brightness (Figure S8f-h,o,p) resulted in significant positive correlations with Western Cuba ancestry and negative correlations with canopy openness across dewlap positions (Table 1). PC2 and yellow composition were correlated with Western Cuba ancestry (P < 0.01; Figure S8); lizards with low PC2 values had yellow dewlaps (i.e., greater yellow composition) and low frequencies of Western Cuba ancestry. As for size, dewlap area and perimeter resulted in significant positive correlations with Western Cuba ancestry (Figure S8j-k) and negative correlations with canopy openness (Figure S8q,r; Table 1). The colorimetric variable hue (cut-on wavelength) was inconclusive due to a lack of model convergence.
Binomial logistic regression indicated that the probability of a lizard having a solid dewlap increased with Western Cuba ancestry (P = 0.01; Figure S9c). Although we observed variation in dewlap pattern across populations (Figs. S9-S10), this trait was not significantly correlated with any environmental variable (Table 2). Furthermore, we found no significant correlation between red color composition, genetic ancestry, and environmental variables. However, yellow color composition was negatively correlated with Western Cuba ancestry (P = 0.004, Figure S8i). These results are consistent with analyses of PC2. Lizards with low PC2 values had yellow dewlaps (Figure 1b) and low frequencies of Western Cuba ancestry (Figure 1b; Figure S8b).
Genetic architecture and selection tests for dewlap traits
Analysis of correlations among traits (Figure S11) revealed strong and significant correlations between our measurements of dewlap size (i.e., area and perimeter; r = 0.91) and between our measurements of dewlap brightness (i.e., mean and total dewlap brightness; r = 1). Moderate, albeit still significant, positive correlations were further revealed between dewlap hue, yellow chroma, and brightness, as well as between dewlap size and dewlap red chroma. Finally, dewlap UV chroma was negatively correlated with most dewlap traits (Figure S11).
The linear mixed model implemented in GEMMA (i.e., standard GWAS), identified seven loci distributed on chromosomes 2, 4, 6, 7, and 11 as significantly associated with the percent of red color (Figure 3, Table S2) at the Bonferroni-adjusted significance level. These associations were of moderate effect as estimated using percent variance explained (PVE; 3.5 to 9.6%; Figure 3). All other analyses using the GEMMA model did not reveal any associations with dewlap traits at the Bonferroni-adjusted significance level (Table S2). There were, however, SNPs that passed the suggestive genome-wide significance threshold. These suggestive associations were identified for nine of the dewlap traits, and were distributed on all macrochromosomes except chromosome 3, as well as on microchromosomes 9, 11, and 12 (Figure 3, Table S2).
The ancestry-specific GWAS identified one significant locus at the Bonferroni-adjusted level (Figure 4, Table S2), located on macrochromosome 2. This locus was associated with total and mean brightness with a moderate effect in samples of Western Cuba ancestry (PVE=3%; Figure 4), and a much smaller effect for samples in the Central-eastern Cuba ancestry group (PVE = 0.7%; Figure 4). Other ancestry-specific associations (N = 37 loci) were identified at the suggestive genome-wide threshold for 10 of the dewlap traits, distributed on chromosomes 1,2,4,5,6 and 11 (Figure 4; Table S2).
The F ST values estimated for the 45 dewlap-associated SNPs (i.e., loci with both significant and suggestive associations with dewlap traits; Table S2) ranged from 0 to 0.62 (average 0.04), whereas F ST for the 45 random SNPs ranged from 0 to 0.43 (average 0.03). The difference inF ST between the dewlap-associated SNPs and the random SNPs was not, however, significant (P = 0.08; Figure S12a). As well, most dewlap-associated SNPs (i.e., 89%, or 40/45 SNPs) were not classified as F ST outliers (Figure S12b). The remaining five dewlap-associated SNPs wereF ST outliers in one or two population pairs at most (Figure S12b).
The GEA analysis identified one SNP on macrochromosome 2 at coordinate 22,431,024 that was significantly associated with canopy openness
(q = 0.009; Figure S13a). However, this SNP was not associated with any dewlap traits (Figure S13b; Table S2), and it was not classified as an F ST outlier, displayingF ST values that ranged from 0 to 0.33 (average 0.05).