Genetic dissection of assortative mating behavior

Abstract

The evolution of new species is made easier when traits under divergent ecological selection are also mating cues. Such ecological mating cues are now considered more common than previously thought, but we still know little about the genetic changes underlying their evolution or more generally about the genetic basis for assortative mating behaviors. Both tight physical linkage and the existence of large-effect preference loci will strengthen genetic associations between behavioral and ecological barriers, promoting the evolution of assortative mating. The warning patterns of Heliconius melpomene and H. cydno are under disruptive selection due to increased predation of nonmimetic hybrids and are used during mate recognition. We carried out a genome-wide quantitative trait locus (QTL) analysis of preference behaviors between these species and showed that divergent male preference has a simple genetic basis. We identify three QTLs that together explain a large proportion (approximately 60%) of the difference in preference behavior observed between the parental species. One of these QTLs is just 1.2 (0-4.8) centiMorgans (cM) from the major color pattern gene optix, and, individually, all three have a large effect on the preference phenotype. Genomic divergence between H. cydno and H. melpomene is high but broadly heterogenous, and admixture is reduced at the preference-optix color pattern locus but not the other preference QTLs. The simple genetic architecture we reveal will facilitate the evolution and maintenance of new species despite ongoing gene flow by coupling behavioral and ecological aspects of reproductive isolation.

Fig 1. Divergence in warning pattern cue and corresponding preference in sympatric Heliconius butterflies.

(A) Wing pattern phenotypes of top, H. cydno chioneus (left), H. melpomene rosina (right), their nonmimetic first-generation hybrid (center); and bottom, their sympatric comimics H. sapho sapho (left) and H. erato demophoon (right). (B) Distribution of H. cydno (blue) and H. melpomene (orange). Individuals were collected and experiments performed in Panama (black circle), where the two species co-occur in Central and northern South America. (C) Overview of crossing design. Colored boxes represent segregating H. cydno (blue) and H. melpomene (orange) alleles; Z and W refer to the alleles on the sex chromosomes and A to those on autosomes. (D) Proportion of courtships directed toward H. melpomene (as opposed to H. cydno) females for CYD, MEL, their F1, and BC and BM. Values in parentheses indicate total number of individuals with behavioral data. Solid colored boxes represent expected average genome contribution of each generation. Note that a further 11 BC individuals were tested but performed no courtship behaviors. Underlying data can be found in S1 Table. BC, backcross hybrid to H. cydno; BM, backcross hybrid to H. melpomene; CYD, H. cydno; MEL, H. melpomene; F1, first-generation hybrid.

Fig 2. QTL analysis of variation in mate preference.

(A) QTLs for relative time males court H. melpomene (as opposed to H. cydno) females on Chromosomes 1, 17, and 18 (n = 139). Scale on right axis depicts genome-wide significance, determined through permutation, corresponding to the LOD score as shown on the left axis. Dotted red line represents LOD significance threshold (genome-wide alpha = 0.05, LOD = 2.99). Dashes indicate position of genetic markers (SNPs), and red arrows indicate the position of the max LOD score for each QTL (used in B). Vertical blue lines represent the position of major color pattern loci and their phenotypic effects. Note that the K locus only has limited phenotypic effects in crosses between H. cydno chioneus and H. melpomene rosina but is responsible for the switch from yellow to white color pattern elements between other taxa within the melpomene–cydno clade. Underlying data can be found in the online Dryad repository doi.10.5061/dryad.4b240j4, specifically raw_data/data_for_Rqtl.csv, derived_data/genome_scan_lod_score.csv, and derived_data/permutations_courtship_prop.csv. (B) Proportion of time males court H. melpomene (as opposed to H. cydno) females for each of the two genotypes for respective QTLs (homozygous = CYD:CYD and heterozygous = CYD:MEL). Error bars represent 95% confidence intervals. Lower dashed blue and upper orange bars represent mean phenotypes measured in H. cydno and H. melpomene, respectively. Circle size depicts total number of “courtship minutes” for each male. Vertical black bars indicate the percentage of the difference measured in the parental species explained. Underlying data can be found in the online Dryad repository doi.10.5061/dryad.4b240j4, specifically raw_data/data_for_Rqtl.csv and derived_data/qtl_data.csv. LOD, log odds ratio; QTL, quantitative trait locus, CYD, cydno allele; MEL, melpomene allele.

Fig 3. Genetic and physical positions of behavioral QTL and the warning pattern loci and localized levels of admixture (f_d).

Vertical blue lines represent the position of major color pattern loci and orange lines represent the position of peak LOD score for each behavioral QTL. Gray boxes indicate the 1.5-LOD support interval for each QTL. Top panel: Dashes along the x-axis indicate position of genetic markers (SNPs). Bottom panel: Blue points represent f_d values for 100-kb windows. f_d was measured between H. melpomene rosina and H. cydno chioneus individuals from population samples in Panama; H. melpomene melpomene from French Guiana, which is allopatric with respect to H. cydno, was the “control” population. Underlying data can be found in the online Dryad repository doi.10.5061/dryad.4b240j4, specifically raw_data/data_for_Rqtl.csv, derived_data/genome_scan_lod_score.csv, and raw_data/fd_values_SM_2MAY18/ /bar92.DP8MP4BIMAC2HET75.fourPopPol_melG_melW_cyd_num.w100m1s20.merged.csv. LOD, log odds ratio; QTL, quantitative trait locus.

Fig 4. QTL effects in consideration of the Beavis effect.

(A) Proportion of “significant” simulations that would be smaller than our empirically measured effects for preference QTLs on Chromosome 1 (blue), Chromosome 17 (black), and Chromosome 18 (orange). 10,000 simulations were run for effect sizes corresponding to between 5% and 40% of the difference in male preference behavior between the parental species. In each case, the distribution of sample effect sizes was determined for those simulations that reached the genome-wide significance threshold determined through permutation (Fig 2). Underlying data can be found in the online Dryad repository doi.10.5061/dryad.4b240j4, specifically derived_data/QTL_1_sim_results.csv, derived_data/QTL_17_sim_results.csv, and derived_data/QTL_18_sim_results.csv. (B) Proportion of time males court H. melpomene (as opposed to H. cydno) females for each of the two genotypes for white (homozygous = CYD:CYD) and red (heterozygous = CYD:MEL) hybrid males for which we were unable to generate RAD-seq data (so which were not included in our initial QTL analysis). Error bars represent 95% confidence intervals. Lower dashed blue and upper orange bars represent mean phenotypes measured in H. cydno and H. melpomene, respectively. Circle size depicts total number of “courtship minutes” for each male. Vertical black bars indicate the percentage of the difference measured in the parental species explained. Underlying data can be found in the online Dryad repository doi.10.5061/dryad.4b240j4, specifically raw_data/IDs_with_pheno_no_RAD.csv. CYD, cydno allele; MEL, melpomene allele; QTL, quantitative trait locus; RAD-seq, restriction site–associated DNA sequencing.

Fig 5. Different QTLs affect different aspects of behavior.

The QTL on chromosome 1 influences courtship toward H. cydno but not H. melpomene females. The opposite is the case for the QTLs on Chromosomes 17 and 18, for which there is little evidence that either QTLs influence courtships directed toward H. cydno females. Data presented are for number of courtship events corrected by the total number of trials. Blue circles and boxplots represent data for individuals homozygous at each QTL (i.e., CYD:CYD); orange circles and boxplots represent data for individuals heterozygous at each QTL (i.e., CYD:MEL). Underlying data can be found in the online Dryad repository doi.10.5061/dryad.4b240j4, specifically derived_data/qtl_data.csv. QTL, quantitative trait locus.