The pdm3 Locus Is a Hotspot for Recurrent Evolution of Female-Limited Color Dimorphism in Drosophila

Abstract

Sex-limited polymorphisms are an intriguing form of sexual dimorphism that offer unique opportunities to reconstruct the evolutionary changes that decouple male and female traits encoded by a shared genome. We investigated the genetic basis of a Mendelian female-limited color dimorphism (FLCD) that segregates in natural populations of more than 20 species of the Drosophila montium subgroup. In these species, females have alternative abdominal color morphs, light and dark, whereas males have only one color morph in each species. A comprehensive molecular phylogeny of the montium subgroup supports multiple origins of FLCD. Despite this, we mapped FLCD to the same locus in four distantly related species-the transcription factor POU domain motif 3 (pdm3), which acts as a repressor of abdominal pigmentation in D. melanogaster. In D. serrata, FLCD maps to a structural variant in the first intron of pdm3; however, this variant is not found in the three other species-D. kikkawai, D. leontia, and D. burlai-and sequence analysis strongly suggests the pdm3 alleles responsible for FLCD originated independently at least three times. We propose that cis-regulatory changes in pdm3 form sexually dimorphic and monomorphic alleles that segregate within species and are preserved, at least in one species, by structural variation. Surprisingly, pdm3 has not been implicated in the evolution of sex-specific pigmentation outside the montium subgroup, suggesting that the genetic paths to sexual dimorphism may be constrained within a clade but variable across clades.

Keywords: abdominal pigmentation; parallel evolution; pdm3; sex-limited polymorphism; sexual dimorphism.

Figure 1

(See also Table S1) Female-limited color dimorphism in four Drosophila montium subgroup species. Females have light or dark posterior abdominal segments (A6 and A7), but males of each species are invariant for pigmentation of their last abdominal segment (A6).

Figure 2

(See also Table S1) Phylogenetic analysis of FLCD in the Drosophila montium subgroup. (A) Bayesian tree for 44 species inferred from four genes (Adh, Amy1, Amyrel, and COII). Male and female abdominal pigmentation is represented by colored dots (brown = dark morph, yellow = light morph, and red = dimorphic). Gray bars at nodes show the 95% confidence intervals for the estimated divergence times assuming D. melanogaster and montium species diverged 28 my ago [68]. The four species in our mapping analysis are underlined in blue. (B) Probabilities of ancestral pigmentation for the montium subgroup (bottom) and each major clade. (C) Inferred transition rates (the number of times a continuous-time Markov chain moves among states relative to branch lengths) between different male and female abdominal pigmentation phenotypes. A transition from one female state to another (dark → light or light → dark) requires a transitional state where light and dark alleles segregate within species so that the females are color-dimorphic until the new allele is fixed; this is represented by the FLCD axis. Arrow thickness reflects the frequency of transitions.

Figure 3

(See also Figures S1, S2, S3) Pigmentation-associated variants in D. serrata map to the first intron of pdm3. (A) Association results for pigmentation in 97 inbred lines in the 708 kb interval identified through backcross and introgression mapping. One SNP (red dot) was significantly associated with female color and another SNP 480 bases away (blue dot) was weakly associated. (B) The top two SNPs fall within the first predicted intron of pdm3. Putative pdm3 coding and non-coding exon boundaries in D. serrata were determined by reciprocal BLAST between the D. serrata pdm3-containing scaffold and D. melanogaster exons. Non-coding exons are represented with gray boxes and coding exons are represented with black boxes. pdm3 is 69 kb long in D. melanogaster and is predicted to span approximately 75 kb in the D. serrata genome assembly (1000 bp scale bar for reference). (C) Cartoon of the structural fragment polymorphism between light (d) and Dark (D) alleles relative to the two FLCD-associated SNPs (red and blue lines). PCR amplification primers are shown. (D) PCR reveals two fragments perfectly correlated with genotype. From left to right, we PCR amplified the haplotypes from females from two light strains (dd: 681.3L and light FORS4), three heterozygotes (Dd: 681.3L (x2640) × 681.5D (x2642), dark FORS4 (x2640) × light FORS4 (x2642), and light FORS4 (x2640)x dark FORS4 (x2642)), and two dark strains (DD: 681.5D and dark FORS4). The light allele is shorter (453 bp) and the dark allele is longer (786 bp). (E) A dot plot of de novo assembled light (horizontal) and dark (vertical) haplotypes shows that flanking sequences are syntenic but the two haplotypes lack homology in the intervening region. Arrows show repeated sequence motifs in the dark haplotype. (F) A schematic representation of repeated sequence motifs found in the dark haplotype.

Figure 4

(See also Figures S4, S6) FLCD in three other montium species maps to the pdm3 locus. Manhattan plot for the Cochran-Mantel-Haenszel (CMH) test comparing SNP allele frequencies between introgression lines and their corresponding light parental strains for (A) D. kikkawai (two intraspecific introgression experiments), (B) D. leontia (one intraspecific introgression and one interspecific experiment), and (C) D. burlai (one intraspecific experiment). Scaffolds are arranged in descending order by size and colored according to the corresponding D. melanogaster chromosome arm, except for KB459631 and KB459527, which were concatenated according to their synteny.

Figure 5

(See also Figure S2) Different variants are associated with FLCD in four montium subgroup species. (A) Allele tree for a ∼1 kb region surrounding the D. serrata structural variant in the first intron of pdm3. Sequences from light (shaded in yellow) and dark (shaded in brown) strains were amplified by PCR or extracted from reference genomes for D. serrata (ser), D. kikkawai (kik), D. leontia (leo), D. burlai (bur), D. bunnanda (bun), and D. birchii (bir). The D. serrata alleles form two clades by color morph, and do not group with sequences from other species. (B) FLCD-associated SNPs shared between D. kikkawai and D. leontia (red triangles) occur in the intergenic region upstream of pdm3, in contrast to the FLCD-associated SNPs in D. serrata that occur in the first intron of pdm3 (blue asterisks). Positive and negative coordinates refer to scaffolds KB459527 and KB459631, respectively.

Figure 6

(See also Figure S5) pdm3 acts as a pigmentation repressor in D. melanogaster. (C,D) RNAi knockdown of pdm3 (Bloomington Drosophila Stock Center #53887) using a pnr-GAL4 driver increased pigmentation along the dorsal midline (arrowheads) in females (n = 3) and males (n = 7) relative to control females (n = 3) and males (n = 4) lacking pnr-GAL4 (A,B). (G,H) Hypomorphic mutants where pdm3 is disrupted by a piggyBac transposon insertion between exons 8 and 9 (PBac {WH}pdm3^f00828) have increased dark pigmentation throughout all abdominal segments (arrowheads) in females (n = 1) and males (n = 6) relative to females (n = 8) and males (n = 5) with a precise excision of the piggyBac transposon (E, F). (I,J) Ectopic expression of a short pdm3 isoform (without exon 9) driven with AbdB-GAL4 and tub-GAL80^ts results in patchy, decreased pigmentation of posterior segments (arrowheads) in females (n = 3) and males (n = 6). Expression of a long pdm3 isoform (including exon 9) produces similar phenotypes (not shown).