Rapid divergence of the copulation proteins in the Drosophila dunni group is associated with hybrid post-mating-prezygotic incompatibilities

Proteins involved in post-copulatory interactions between males and females are among the fastest evolving genes in many species and this has been attributed to reproductive conflict. Likely as a result, these proteins are frequently involved in cases of post-mating-prezygotic isolation between species. The Drosophila dunni subgroup consists of a dozen recently diverged species found across the Caribbean islands with varying levels of hybrid incompatibility. We sought to examine how post-mating-prezygotic factors are involved in isolation among members of this species group. We performed experimental crosses between species in the dunni group and find evidence of hybrid inviability. We also find an insemination reaction-like response preventing egg laying and leading to reduced female survival post-mating. To identify that genes may be involved in these incompatibilities, we sequenced and assembled the genomes of four species in the dunni subgroup and looked for signals of rapid evolution between species. Despite low levels of divergence, we found evidence of rapid evolution and divergence of some reproductive proteins, specifically the seminal fluid proteins. This suggests post-mating-prezygotic isolation as a barrier for gene flow between even the most closely related species in this group and seminal fluid proteins as a possible culprit.


Introduction 15
Numerous groups of recently diverged species have been used to study speciation across multicellular taxa 16 crosses producing only female offspring, or sterile male offspring (HEED 1962). 51 Here we perform experimental crosses in the dunni group and find that in some crosses, 52 heterospecific matings reduces female survival compared to conspecific matings, potentially caused by an 53 insemination reaction-like effect (PATTERSON 1947). Using a combination of long-read and short-read 54 sequencing, we assembled the genomes of four species in the dunni group to identify proteins driving this 55 incompatibility. We find these genomes are of similar quality and composition as other higher quality We set the reference level as the conspecific cross (e.g. D. arawakana ♂ x D. arawakana ♀) and looked 102 for significant differences from these for interaction terms to determine if unmated females (e.g. D. 103 arawakana ♀ not mated) or heterospecifically crossed females (e.g. D. arawakana ♂ x D. nigrodunni ♀) 104 show significant differences from the conspecific cross. To consider the effect of Wolbachia infection on 105 these crosses, we repeated these initial crosses alongside the same crosses with Wolbachia cured flies (cured 106 as described above) and a Cox's Hazard Ratio was used to determine the effect of Wolbachia on survival, 107 and to test for differences in survival between sets of crosses after accounting for Wolbachia: 108 Post-mating dissection of the female reproductive tract 111 We collected virgin males and females for tetracycline-cured D. arawakana and D. nigrodunni as described 112 above and aged them 2-3 days. We then established conspecific and heterospecific experimental crosses 113 for 6 replicates of 10 males and 10 females at 10AM central time, as well as virgin control females for 6 114 replicates of 10 females. Following 24 hours of cohabitation, for 3 replicates of each cross, we separated 115 the females for each cross and dissected the reproductive tract. Based on previous work describing the 116 insemination reaction (PATTERSON 1947;GRANT 1983; MARKOW AND ANKNEY 1988), we scored the 117 reproductive tract for each female, identifying if the female had mated (by the presence of sperm), if the 118 reproductive tract appeared to be swollen (relative to the unmated virgin females) or if the reproductive 119 tract was destroyed or damaged (alongside a swollen tract, if possible to tell). We repeated this scoring for 120 the remaining 3 replicates of each cross 24 hours later (48 hours total). We then compared conspecific and 121 heterospecific crosses for rates of mating and rates of insemination reaction occurrence. 122

Genome sequencing, assembly and annotation 123
We extracted DNA following the protocol described in (  short-read data for each species, which we used to make a second map of reference genome repetitive 168 regions using RepeatMasker. For both sets of repeat content assemblies we identified which TE families 169 were shared between species and which were unique to species using blastn (e-value < 10e-5, hsps = 1, 170 alignments = 1). We then identified what proportion of the genome each TE family constituted across 171

species. 172
Placing the dunni group in the Drosophila phylogeny 173 To find the consensus species tree despite the differing evolutionary histories of different genes (MENDES 174 AND HAHN 2016), we randomly sampled 100 genes conserved across Drosophila and humans from and 175 extracted these from our four focal species, as well as from several other Drosophila species, taken from 176 We then took outlier genes (e.g. genes above the 97.5 th percentile in each category) and looked for 204 enrichments in gene ontology categories compared to non-outlier genes using GOrilla (EDEN et al. 2009). 205 For GO categories of interest, such as those enriched for duplications or for high levels of dN/dS, we 206 compared dN/dS of genes in these categories to the nearby genomic background. For each gene we 207 extracted nearby genes (within 100kbp up or downstream on the same chromosome), of similar divergence 208 levels on each branch (within 0.01 dS), we then found the difference in dN/dS between the median of the 209 background genes and the focal gene. We then used a Wilcoxon-Rank Sum test to identify GO categories 210 on each branch with significantly higher (or lower) dN/dS than the background. 211 Using the annotations of all species and D. innubila, we identified genes with more than one copy 212 in one species, relative to all other species. We confirmed this by estimating copy numbers of genes in each 213 species using short read information and dudeML (following the tutorial pipeline for N = 1) with the short 214 read information mapped to the genome of the sister species (HILL AND UNCKLESS 2019). We then used 215 GOrilla (EDEN et al. 2009) to identify Gene ontology categories that are enriched for duplicates on specific 216 branches, which we confirmed using PANTHER (THOMAS et al. 2003). 217

Statistics 218
We used R for all statistics in this analysis (R-CORE-TEAM 2013), and ggplot2 for data visualization and 219 figure production (WICKHAM 2009). 220

Results 221
The Drosophila dunni group shows varying levels of hybrid compatibility 222 The Drosophila dunni group is a species group endemic to islands in the Caribbean, with each island 223  = -2.948, p-value = 0.00319). In several mated females when compared to virgin females, we find a swelling 295 of the reproductive tract consistent with the insemination reaction ( Figure 3C). Exclusively in several 296 heterospecifically crossed females, we also saw damaged and destroyed reproductive tracts ( Figure 3D).

Genes involved in copulation and immune defense have high rates of divergence between species 313
We reasoned that these incompatibilities between species could be caused by a divergence in copulation 314 proteins. Previous work has suggested that females may be susceptible to bad reactions following hybrid 315 matings due to no protection from the other species accessory gland proteins (MARKOW AND ANKNEY 316 1988; KNOWLES AND MARKOW 2001). Specifically, that there is an arms race between sexes to 317 block/unblock the female reproductive tract and that females of other species have not evolved to suppress 318 these reactions. Based on this, we sought to examine the levels of divergence and identify rapidly evolving 319 genes between species. We sequenced, assembled and annotated the genomes of each species involved (see 320 Materials and Methods), producing two high quality genomes with high synteny to each other and to D.  Table 6, p-value < 0.05 after multiple testing 351 correction). This is consistent with rapid evolution occurring in genes involved in the reproductive conflict 352 between the sexes (Figure 4) (HAERTY et al. 2007). While not significant outliers, we also find that immune 353 recognition proteins, antiviral RNA and piRNA pathways are also rapidly evolving in some species, 354 consistent with arms races between the species and their parasites (Supplementary Table 6). 355 Here, we assessed the extent of hybrid incompatibilities between species of the dunni subgroup, 417 focusing on post-mating-prezygotic incompatibilities. We then sequenced and assembled the species 418 genomes to identify highly divergent and rapidly evolving genes. Between D. nigrodunni and D. 419 arawakana, we find elevated divergence of several immune system pathways, as well as divergence in 420 genes involved in copulation. This divergence fits with the hybrid male inviability between these two 421 species, as well as the reduced survival of females following insemination by a heterospecific male. The functional annotation of the more diverged genes may also provide us with clues as to how 434 these species are diverging. As we find premating-behavior proteins are divergent between D. arawakana 435 and D. nigrodunni, this may result in a divergence in premating behavior, resulting in the reduced rate of 436 hybrid matings scored (Figure 3). We also see no difference in the proportion of hybrid matings after 24 437 hours and 48 hours, suggesting that in these cases, if a female has rejected all males, she may not change due to this species requiring a well-adapted stress response pathway, given its negative reaction to 453 heterospecific matings (Figures 1-3). 454 Several of the functional gene categories identified in this study as highly divergent between 455 species are also promising regions for future study, particularly when focusing on immune evolution. Our  (Figures 4 and 5). 467 The repetitive content also appears to be diverging rapidly across this species complex 468 (Supplementary Figure 5). This is commonly seen between species, given the elevated mutation Overall, our findings suggest that the rapid divergence of reproductive genes has led to 485 incompatibilities between species in the dunni group, including inviable male offspring and the 486 insemination reaction associated with reduced female survival. We also find multiple areas for further 487 investigation in the D. dunni group, either in immune evolution of continuing to investigate the speciation 488 in this species group, suggesting promise in the future of research for this group. 489