Ancestry dynamics and trait selection in a designer cat breed

Abstract

The Bengal cat breed was developed from intercrosses between the Asian leopard cat, Prionailurus bengalensis, and the domestic cat, Felis catus, with a last common ancestor approximately 6 million years ago. Predicted to derive ∼94% of their genome from domestic cats, regions of the leopard cat genome are thought to account for the unique pelage traits and ornate color patterns of the Bengal breed, which are similar to those of ocelots and jaguars. We explore ancestry distribution and selection signatures in the Bengal breed by using reduced representation and whole-genome sequencing from 947 cats. The mean proportion of leopard cat DNA in the Bengal breed is 3.48%, lower than predicted from breed history, and is broadly distributed, covering 93% of the Bengal genome. Overall, leopard cat introgressions do not show strong signatures of selection across the Bengal breed. However, two popular color traits in Bengal cats, charcoal and pheomelanin intensity, are explained by selection of leopard cat genes whose expression is reduced in a domestic cat background, consistent with genetic incompatibility resulting from hybridization. We characterize several selective sweeps in the Bengal genome that harbor candidate genes for pelage and color pattern and that are associated with domestic, rather than leopard, cat haplotypes. We identify the molecular and phenotypic basis of one selective sweep as reduced expression of the Fgfr2 gene, which underlies glitter, a trait desired by breeders that affects hair texture and light reflectivity.

Keywords: genetic incompatibility; genome-wide association; interspecies hybridization; introgression; pigmentation pathways.

Figure 1.. Bengal breed formation and leopard cat admixture

(A) A species tree of felid lineages that gave rise to the Asian leopard cat (Prionailurus bengalensis, red) the domestic cat (Felis catus, black). (B) A breeding paradigm for introducing leopard cat ancestry into the Bengal cat breed. (C) Distinct sets of RADseq SNVs distinguishing genetic variation within or between leopard cats and domestic cats. Variants present in multiple marker sets (grey) were removed from all sets. (D) Distribution and mean (red bar) of leopard ancestry fraction in Bengal cats, measured from ancestry informative RADseq markers. (E) Leopard ancestry fraction in hybrids from an experimental cross, including domestic cat-leopard cat intercross offspring (F1) and subsequent backcross offspring (BC1-3). Red bars indicate mean leopard ancestry fraction per group. See also Figures S1 and S2, and Data S1.

Figure 2.. The genome distribution of leopard cat ancestry in Bengal cats

(A) A diplotype plot from a representative Bengal cat, showing species ancestry inferred from lcWGS in 48,208 50kb windows across chromosomes (in rows), colored by the number of predicted leopard cat haplotypes. (B) Ancestry diplotype plot for 200 Bengal cats (in rows) across chromosome F2. (C) The distribution and quartile metrics for distinct leopard cat haplotypes per Bengal cat (n=673 cats). (D) The length distribution and quartile metrics of 15,949 leopard cat haplotypes inferred from 673 Bengal cats. (E) Leopard cat ancestry frequencies (n=673 SBT Bengal cats), measured from lcWGS in 45,673 50kb windows across autosomes (felCat9). Windows in 15 regions (red, DataS1F) are significantly enriched for leopard cat ancestry based on genome-wide distribution (Z-score > 3, leopard cat allele frequency > 0.172, red broken line). See also Figures S1, S2, S3, and S4, and Data S1.

Figure 3.. Genetic and functional characterization of charcoal

(A) Images of non-charcoal (left panel) and charcoal (center and right panels) Bengal cats, with corresponding genotypes at Asip and Mc1r. lc and dom superscripts refer to the normal leopard cat and normal domestic cat alleles, respectively, and a refers to the Asip nonagouti allele. (B) Case-control GWAS for charcoal using either domestic cat markers imputed from lcWGS (top) or 50kb species-informative ancestry windows (bottom). Association peaks are highlighted in red (domestic markers) or blue (ancestry windows). (C) High resolution view of the charcoal association interval in (B) as an overlay of the two marker sets (top) or with domestic cat markers colored by correlation with the lead variant (chrA3:26,745,409, Wald test p = 1.91x10⁻³¹, bottom). (D) Five distinct charcoal-associated leopard cat haplotypes (H1-H5), inferred from 33 leopard cat species informative RADseq SNVs. Alleles are colored as ancestral (yellow) or derivative (blue). (E) Leopard cat (Asip^lc) and domestic cat (Asip^dom) allele ratios measured by cDNA pyrosequencing from skin biopsies of four Asip^dom/lc Bengal cats. Cat age at time of biopsy is indicated. See also Data S1.

Figure 4.. Genetic characterization of pheomelanin intensity in Bengal cats

(A) Representative images of five coat color categories, ranging from grey-brown (category 1) to red-orange (category 5), used to classify pheomelanin intensity in 289 Bengal cats. (B) GWAS for pheomelanin intensity based on the scoring system in (A), using 48,208 species-informative ancestry windows. Red and blue points correspond to suggestive association peaks shown at higher resolution in the bottom panels. The red and blue lines in the upper panel indicate Bonferroni (p-value < 0.05) and Benajmini (q-value < 0.05) adjusted significance thresholds, respectively. (C) Q-Q plot for the GWAS in (B) with red and blue points representing Asip and Corin ancestry windows, respectively. (D and E) Top panels - GWAS for pheomelanin intensity using ~1.5 million domestic cat markers imputed from lcWGS, with genotypes at association peaks on chrE2 (D) or chrB1 (E) used reciprocally as covariates. Bottom panels - association peaks at higher resolution and with denser markers, colored by correlation with the lead variants - chrB1:167,683,391 (Wald test p = 3.41x10⁻¹⁵) and chrE2:63,823,500 (Wald test p = 8.60x10⁻¹⁰). Positions of candidate genes, Corin and Mc1r, are indicated with grey bars. See also Figure S5 and Data S1.

Figure 5.. Evidence for selection in the Bengal cat genome

The frequency distributions of (A) pooled heterozygosity (${\mathrm{H}}_{\mathrm{p}}$) from 387 Bengal cats sequenced by lcWGS and (B) the fixation index (FST) between Bengal cats and other domestic cats, measured in (C) 11,468 200 kb windows across the autosomes. The colored circles indicate cluster of windows with both low ${\mathrm{H}}_{\mathrm{p}}$ (Z-score < −3) and high FST (Z-score >3). See also Data S1.

Figure 6.. Genetic characterization of glitter in Bengal cats

(A) The iridescent sheen apparent in the coat of glitter cats. (B) Top panel - a case-control GWAS to identify a glitter association interval (red), using ~1.5 million lcWGS-imputed SNVs, with an association peak at chrD2:79,620,465 (Wald test p = 1.66x10⁻⁴⁰). Bottom panel – the glitter association interval at higher resolution and with additional markers that were LD pruned from the GWAS marker set. SNVs located in Fgfr2 are highlighted in red. (C) Recombinant glitter haplotypes (in rows) inferred within a 2.57Mb interval, which define a 513 kb genetic interval (red lines). SNV loci are colored by reference (yellow) or alternate (blue) allele in the felCat9 assembly. (D) Glitter genomic interval refinement, using inferred ancestry recombinant breakpoints from both RADseq and lcWGS (see methods) to identify 14,229 bp and 10,493 bp genomic intervals (grey bars), with a glitter specific LTR insertion in Fgfr2 intron 5 at chrD2:79,707,475-79,707,478. See also Figure S6 and Data S1.

Figure 7.. The functional basis of glitter in Bengal cats

(A) Measures of hair length and medulla width, in guard (left panels, n=6 hairs per group) and down hairs (right panels, n=15 hairs per group), from three non-glitter (black) and three glitter (grey) Bengal cats, with mean values per group (red lines). Representative images for each hair type are shown. (B) Cartoon representation and photographic images at different positions along the length of down hairs from non-glitter (left, black circle) and glitter (right, grey circle) Bengal cats. Measures of hair tip medulla fragmentation (top right) and hair length (n=15 hairs from 3 Bengal cats in each group) in down hairs. (C) Allele count ratios at Fgfr2 polymorphic sites in transcripts from whole skin RNAseq, intended to directly compare the expression levels of Fgfr2⁺ to Fgfr2^gl transcripts. Each row provides an allele count ratio at a different polymorphic site or from a different cat. Symbols represent different cats, 2 Fgfr2^gl/+ (grey) and 3 Fgfr2^+/+ (black). Ratios in Fgfr2^+/+ samples are included to reflect a normal range of Fgfr2 allelic variation. See also Data S1.