The genetics of male pheromone preference difference between Drosophila melanogaster and D. simulans

Species of flies in the genus Drosophila differ dramatically in their preferences for mates, but little is known about the genetic or neurological underpinnings of this evolution. Recent advances have been made to our understanding of one case: pheromone preference evolution between the species D. melanogaster and D. simulans. Males of both species are very sensitive to the pheromone 7,11-HD that is present only on the cuticle of female D. melanogaster. In one species this cue activates courtship, and in the other it represses it. This change in valence was recently shown to result from the modification of central processing neurons, rather than changes in peripherally expressed receptors, but nothing is known about the genetic changes that are responsible. In the current study, we show that a 1.35 Mb locus on the X chromosome has a major effect on male 7,11-HD preference. Unfortunately, when this locus is divided, the effect is largely lost. We instead attempt to filter the 159 genes within this region using our newfound understanding of the neuronal underpinnings of this phenotype to identify and test candidate genes. We present the results of these tests, and discuss the difficulty of identifying the genetic architecture of behavioral traits and the potential of connecting these genetic changes to the neuronal modifications that elicit different behaviors.


42
Understanding the proximate mechanisms of phenotypic divergence has long been a goal of 43 evolutionary biologists (Stern et al., 2009;Stern & Orgogozo, 2008). Advances in genome 44 sequencing have led to a recent boom in genotype-phenotype association studies, the large 45 majority being for morphological traits   we have fewer studies of the proximate mechanisms underlying behavioral divergence with 51 which to draw broad conclusions. This is unfortunate, because behaviors are important 52 phenotypes, particularly with respect to speciation and biodiversity. For example, differences in 53 host, habitat, and mating behaviors can form strong reproductive barriers between species 54 (Coyne & Orr, 1997. 55 Behaviors holistically involve the detection of stimuli via the peripheral nervous system, 56 sensory integration via central nervous system processing, and the coordinated production of a 57 behavioral output. It is therefore surprising that a large proportion of genes known to cause 58 behavioral divergence between species affect sensory perception at the periphery, rather than 59 the other molecular determinants of behavior (Auer et  that this pattern is due mainly to ascertainment bias (Rockman, 2012). Indeed, many case 65 studies that have successfully mapped causal genes explaining behavioral divergence have used 66 a candidate-gene approach targeting sensory receptors, so it is plausible that other types of 67 changes are more common but have been overlooked. For example, recent studies that take like an on/off switch for male courtship (Auer & Benton, 2016). A recent study found that the 95 evolution of the interactions among these neurons in the central nervous system causes the 96 difference in 7,11-HD preference, rather than the evolution of the peripheral nervous system. 97 (Seeholzer et al., 2018). In D. melanogaster, 7,11-HD promotes courtship because the excitation 98 of P1 neurons is greater than the inhibition; in D. simulans, the opposite seems to be the case 99 (Seeholzer et al., 2018). However, the molecular changes underlying these differences in 100 neuronal interactions remain unknown. This detailed, if still incomplete, understanding of the 101 cellular basis of behavioral evolution presents an excellent opportunity to map the causal genes 102 and link evolution in behavior at the genetic, cellular, and organismal levels. 103 Here we present the results of a genotype-phenotype association study where we make 104 considerable progress toward identifying the loci underlying shifts in 7,11-HD preference 105 behavior between D. simulans and D. melanogaster. First, we confirm a previous result 106 demonstrating that a portion of the male preference phenotype maps to the X chromosome 107 (Kawanishi & Watanabe, 1981). We then show that the preference for 7,11-HD can be 108

Fly stocks and maintenance 116
We maintained all fly strains in 25 mm diameter vials on standard cornmeal/molasses/yeast 117 medium at 25°C under a 12 h:12 h light/dark cycle. Under these conditions, we established non-118 overlapping two-week lifecycles. Every 14 days, we transferred all of the emerged male and 119 female adult flies into vials containing fresh food, where they were allowed to oviposit for 1-3 120 days before being discarded. To test for species differences in male preference, we used two D. 121 simulans strains: simC167.4 (obtained from the UC San Diego Drosophila Stock Center; Stock #: 122 14021-0251.99) and Lhr (Brideau et al., 2006;Watanabe, 1979

Hybrid crosses 128
To confirm the role of the X chromosome in male mate discrimination, we needed to make F1 129 hybrid males by crossing D. melanogaster females to D. simulans males (hybrid offspring would 130 then have a D. melanogaster X) and by the reciprocal cross (hybrids would have the same 131 autosomal genotype, but the D. simulans X). D. melanogaster/D. simulans hybrid males 132 normally die during development, while hybrid females are infertile (Watanabe, 1979). The Lhr 133 strain of D. simulans rescues male viability, allowing us to collect living male and female hybrid 134 offspring ( Figure 1A). However, crossing females from this D. simulans hybrid male rescue strain 135 to D. melanogaster males never yielded offspring, probably because of very strong pre-mating 136 isolation. Instead, we used genetic tools (see below) to make these same genotypes while only 137 crossing D. melanogaster females to D. simulans males. In this crossing direction, we found 138 hybrids could be made using the following steps. First, we collected 20 D. simulans males as 139 virgins and aged them for 7-12 days. We collected 10 D. melanogaster females as very young 140 virgins, just 2-4 hours after eclosion. We immediately combined 10 very young D. melanogaster 141 virgins with 20 aged D. simulans males in a vial with food media. We pushed a foam plug into 142 the vial, leaving only 1-2 cm of space above the food surface. We held these hybrid cross vials in 143 this manner for 2-3 days before transferring the flies to a new vial with fresh food to oviposit. 144 To create male hybrids with the D. melanogaster X chromosome (melX), we set up the

Male courtship assays 156
For all courtship assays, we aged virgin males and females in single-sex vials for 4 days at 25°C 157 in densities of 10 and 20, respectively. We gently aspirated a single experimental male into a 25 158 mm diameter vial with standard cornmeal/molasses/yeast medium 24 hours before 159 observation. On the morning of observation, we aspirated a single female into the vial and 160 pushed a foam plug down into the vial, leaving a space of 1-3 cm above the food surface. This 161 limited space ensures that flies interact within the observation time period. We observed flies 162 for 30 minutes, collecting minute-by-minute courtship data by manually scoring each pair for 163 three easily observed stages of courtship: singing (single wing extensions and vibration), 164 attempted copulation, and successful copulation. We scored pairs that exhibited multiple 165 stages within a single minute once within that minute. We conducted 2 observations per day at 166 room temperature between the hours of 9 and 11 AM (0-2 hours after fly incubator lights turn 167 on). All observers were blind to both male and female genotype (see below). 168 We tested the male courtship behavior of the following strains: D. melanogaster (LH M , 169 Canton-S, RNAi strains), D. simulans (Lhr and simC167.4), and various melX and simX F1 hybrids. 170 We measured male courtship towards three types of females, D. melanogaster, D. simulans, 171 and F1 hybrid females, using no-choice (single female) assays. All D. simulans female courtship 172 objects were of the simC167.4 strain. Female courtship objects for D. melanogaster were of the 173 DGRP-380 strain when testing males of the LH M , Lhr, melX, simX, and duplication hybrid 174 genotypes (i.e. when testing for an effect of the X chromosome), but we used females from the 175 Canton-S strain with Canton-S males, RNAi males and controls, and gene aberration melX 176 hybrids (see below), because these additional data were collected during a separate follow-up 177 experiment. We made F1 hybrid females using the methods described above (i.e. from a cross 178 between LH M and Lhr; Figure 1A). 179 180

Duplication hybrid crosses 181
Mapping the genes responsible for this evolved behavior is very challenging for several reasons. 182 QTL analysis is not possible using these species, as hybrid males are inviable and hybrid females 183 are infertile. One way around this problem is to use large engineered deletions to make loci 184 hemizygous rather than heterozygous in F1 hybrids, exposing recessive or additive alleles from  Table 1A). This method has been used to map hybrid incompatibility loci (Cattani & Presgraves,192 2012), but to our knowledge has not been used to map differences in morphology or behavior. 193 The primary caveat to this method is the inability to detect D. melanogaster loci that are 194 recessive to their D. simulans counterparts. Because F1 hybrid males are hemizygous, it is 195 impossible to assess dominance a priori. Despite this limitation, the primary advantage to this 196 method is our ability to assay a large portion of the X chromosome (80%) using just 16 DP(1;Y) 197 hybrid strains. To create these hybrids, we crossed D. melanogaster males from a DP(1;Y) strain 198 to D. melanogaster females from the C(1)DX-LH M strain ( Figure 1C). We then took the resulting 199 female offspring, which carry a D. melanogaster compound X chromosome and a D. 200 melanogaster Y chromosome that has a translocated segment of the X chromosome, and 201 crossed them to D. simulans Lhr males using the hybrid cross methods described above. The 202 DP(1;Y) Y chromosome is marked with the dominant visible Bar mutation, so inheritance of this 203 chromosome is easy to track. After assaying our original 16 duplication hybrid strains, we tested 204 additional strains with duplications that partially overlap a region we identified in the initial 205 screen in an attempt to fine-map loci within DP(1;Y) segments of interest (Table 1B).  We then vortexed the vials on the highest setting for three 20-second intervals separated by 20 214 seconds of recovery. We allowed females to recover from vortex mixing for 30 minutes before 215 loading them into courtship vials. We conducted courtship assays as described above. We 216 detected no difference in courtship between our 200 and 400 g perfuming treatments ( 2 = 217 0.083, df = 1, p = 0.773), so we combined them into a single perfume treatment in our final 218 analysis. 219 220

Selecting and testing candidate genes for validation 221
The BSC100 duplication region that has a significant effect on male courtship behavior in a 222 hybrid background (see results) contains 159 genes (Table S1). In order to select appropriate were able to procure a non-lethal aberration (Table 2A). We crossed females of these strains to 233 D. simulans Lhr males to create melX hybrids with individual gene knockouts. Two of these 234 knockouts (Smr and pot) are held over a balancer chromosome in females, and thus produce 235 two types of melX hybrid males: those carrying a balancer (intact) X chromosome, and those 236 carrying a defective X-linked allele. Unfortunately, pot defective hybrid males were not viable, 237 and thus could not be observed (Table S3). For Smr, we compared both balancer hybrids and 238 Smrhybrids paired with D. melanogaster and D. simulans females in the courtship assays 239 Pde9 hybrids presented an extra challenge, as males have white eyes (Table 2A), and thus, 242 difficulty tracking a female courtship target. To remedy this challenge, we observed these males 243 as above, but in a smaller reduced arena to increase interaction between visually impaired 244 males and females. To modify the courtship arena for these males, we inserted a single piece of 245 plastic vertically into the media, dividing the vial in half, before pushing a foam plug down into 246 the vial, resting 1-2 cm from the food surface. 247 Because the two remaining genes were either lethal in males (cacophony), or had no 248 aberration available at all (Ir11a), we knocked down expression in D. melanogaster flies using 249 RNAi under the control of the Gal4/UAS system (Perkins et al., 2015). For RNAi knock-down 250 strains, we compared progeny of RNAi lines crossed with the pan-neuronal elav-Gal4 stock 251 (Table 2B). The elav-Gal4 insertion is maintained in heterozygotes over a balancer

Data analysis 266
From the minute-by-minute courtship data, for each male-female combination (see below), we 267 collected binomial data (court/did not court) to determine the courtship frequency (CF) of male 268 genotypes. We only considered males that spent 10% or more of the total assay time (i.e., >3 269 min) in one of the three courtship stages as successfully displaying courtship. For male 270 genotypes, unless otherwise stated, we used Fisher's exact test to compare the proportion of 271 males that courted a given female type followed by posthoc analysis with sequential Bonferroni 272 tests (Holm, 1979). 273 For each male that displayed courtship towards a female target, we also calculated the 274 total percent of assay time (30 minutes) that a male spent courting as a proxy for male 275 courtship effort (CE). For males that mated with females, we calculated CE as the percent of 276 time a male spent courting from the start of the assay until the time of copulation. Unlike with 277 CF, we used no minimum threshold for CE. CE is representative of male investment in any given 278 female, another indication of male choice (Edward & Chapman, 2011). Because the courtship 279 effort distributions are highly skewed, we report the median values for comparison across 280 strains. For each male genotype, we compared courtship effort between female genotypes 281 using the Mann-Whitney U test followed by posthoc analysis with sequential Bonferroni tests 282 (Holm, 1979). 283 284

A significant conspecific courtship preference between D. melanogaster and D. simulans 286
For both D. melanogaster strains that we assayed, we detected significantly higher courtship 287 frequencies when males were paired with females of their own species ( Figure 2A, Table S2). (Lhr: p = 5.34E-09; simC167.4: p = 5.47E-12). There were no differences in the courtship 292 frequencies among strains of the same species (p = 1 for both D. melanogaster and D. 293 simulans). Unsurprisingly, when we calculated a consensus p-value (Rice, 1990), which tests the 294 combined effect of independent tests of the same hypothesis, for the two D. melanogaster 295 strains and the two D. simulans strains, conspecific courtship preferences remained highly 296 significant (p = 1.61E-5 and p < 1.00E-25, respectively), suggesting this is indeed a species-level, 297 rather than strain-specific difference. 298 We also observed each D. melanogaster and D. simulans strain with F1 hybrid females 299 ( Figure 2A, Table S2). These females still produce 7,11-HD, due to a single functioning copy of 300 desatF (Shirangi et al., 2009), and concordantly, D. melanogaster and D. simulans males court 301 them similarly to D. melanogaster females. We found that our D. melanogaster strains were 302 just as likely to court F1 hybrid females as they were to court D. melanogaster (p = 0.6486 for 303 LH M , and p = 6724 for Canton-S), and these comparisons remained non-significant when we 304 calculated a consensus p-value (p = 0.7980). Conversely, for one of our D. simulans strains, 305 simC167.4, we found that males court F1 hybrid females significantly less often than D. 306 simulans females (p = 5.59E-07). Lhr males had a much lower courtship frequency with F1 307 hybrid females compared to D. simulans females (20% vs. 65%, respectively), but this difference 308 was not significant, likely due to small sample size (p = 0.1003 and N = 10). Supporting this, a 309 consensus p-value for both D. simulans strains found that D. simulans overall had a significantly 310 higher courtship frequency with D. simulans females than with F1 females (p = 9.93E-07). We find less striking differences when we calculate the courtship effort of those males 315 that did court ( Figure 2B, Table S2). For example, once they began courting, Canton-S males 316 courted all three female types with equal vigor (p = 1 for all comparisons). LH M males, however, 317 courted D. melanogaster with much higher vigor than D. simulans females (p = 0.0011). Thus, 318 unlike courtship frequency, courtship effort appears to have strain-specific effects within the D. 319 melanogaster strains we surveyed. For D. simulans, we detected a nearly significant increase in 320 courtship effort for simC167.4 males that courted D. simulans females compared to males that 321 courted F1 females (p = 0.0506). We are unable to detect significant differences in courtship 322 effort for Lhr males among any comparisons, or for simC4 male comparisons involving D. indistinguishably (all p > 0.8541, Figure 2B). 338 The behavior of simX males more closely reproduces that of the D. simulans parent 339 strain (Lhr). Like Lhr, simX males court D. simulans females at much higher frequencies than D. 340 melanogaster females (p = 1.32E-15, Figure 2A) and F1 females (p = 4.03E-14). Unlike the Lhr 341 parent strain, simX males court F1 females significantly more frequently than D. melanogaster 342 females (p = 1.57E-05). Courtship towards D. melanogaster was still too rare to detect 343 differences in courtship effort compared to D. simulans or F1 females (both p = 1, Figure 2B), 344 but simX males courted D. simulans with significantly higher effort than F1 females (p = 0.0005, 345 Figure 2B). Quantitatively, simX males behave very similarly to their D. simulans parents. simX 346 males court D. melanogaster and F1 females with frequencies equivalent to that of Lhr males (p 347 = 0.5876 and p = 0.7190, respectively), but they court D. simulans females at a higher frequency 348 (p = 0.0373). We suspect this latter result is a byproduct of increased heterozygosity relative to 349 the inbred Lhr parent strain. To test specific regions of the X chromosome for their role in courtship preference differences, 354 we measured the courtship behavior of 16 duplication hybrids (Table 1A, Figure 1C). These 355 duplication hybrids are simX hybrids made heterozygous for one stretch of the D. melanogaster 356 X chromosome. We observed 15 of these strains with D. melanogaster females, and 357 interestingly, none displayed courtship (N = 6-20 for each, N = 179 total,  Figure 3A, Table S2); three of these lines showed a significant preference for D. simulans 372 females after correction for multiple tests (p = 0.0415 for BSC296, p = 0.0061 for BSC277, and p 373 = 0.0437, for BSC200). Only one duplication hybrid strain, BSC100, courted F1 hybrids with 374 higher frequency than D. simulans hybrids (p = 0.0136). In general, the duplication hybrid 375 strains courted D. simulans females with greater effort than F1 hybrid females (grand median 376 CE = 17% for D. simulans females, and the grand median CE = 10% for F1 females, p = 8.78E-05). 377 Again, BSC100 duplication hybrids were the only hybrids to display higher courtship effort 378 toward F1 females (CE = 30%) than toward D. simulans females (CE = 13.33%), although this 379 difference was not significant after correcting for multiple comparisons (p = 0.0976, Figure 3B, 380 Table S4). 381

Testing candidate genes 428
The BSC100 region is known to contain 159 genes (Table S1). To reduce this list to a testable 429 number, we focused on genes with neurological functions and fruitless binding sites (see 430 Methods). This produced 6 candidate genes: papillote (pot), cacophony (cac), Tenascin-a (Ten-431 a), Smrtr (Smr), Ionotropic receptor 11a (Ir11a), and Phosphodiesterase 9 (Pde9, Figure 5). We 432 used loss-of-function mutations or RNAi to investigate 5 of these 6 genes further 433 (unfortunately, pot loss-of-function hybrid males were inviable). 434 For Smr, the hybrid cross between the D. melanogaster knockout strain produced 435 balancer hybrid (melX males with the X chromosome intact) and Smrhybrid males (melX males 436 with the X chromosome lacking a functional Smr gene). Smr balancer hybrids courted D. 437 simulans and D. melanogaster at similar frequencies (p = 1, Figure 6A, Table S2), and with 438 similar efforts (p = 0.1136, Figure 6B). The same is true for Smrhybrid males (p = 1 for both CF 439 and CE). However, these males courted both females at significantly higher frequencies than preference when compared to intact balancer hybrid males. 444 In contrast, Ten-a and Pde9 are not held over a balancer, so hybrid crosses yield only 445 knockout males. Ten-ahybrids court both D. simulans and D. melanogaster at equal 446 frequencies (p = 1), and with similar effort (p = 1). Pde9hybrids court D. melanogaster at non-447 significantly higher frequencies than D. simulans females (p = 0.2124), and court both females 448 with equal effort (p = 1). In contrast, melX males with functioning copies of these genes court D. 449 melanogaster females more frequently (and with non-significantly higher courtship effort). 450 Although these knockout hybrids differ in some aspects from melX males, it is difficult to 451 discern whether these differences are due to the different D. melanogaster strains with which 452 these hybrids were made (see Discussion). 453

An important difference in male courtship preference between D. melanogaster and D. 474 simulans partially maps to the X chromosome 475
Our observation of two D. simulans and two D. melanogaster strains confirms a previously 476 described species difference in male courtship preference, where each species dramatically 477 prefers their own females (Manning, 1959). While we did observe some variation in the 478 quantitative amount of courtship among our lines (Figure 2A), the valence of male preference 479 was always consistent within species. Because we also used females from two different lines, 480 this variation in male behavior could be due to individual variation in female CHC quantity 481 behavior, but also with respect to the evolution of reproductively isolating barriers. 487 The courtship data we collected using reciprocal D. melanogaster/D. simulans hybrids 488 (melX and simX males) created in a homogenous background and controlled for cytoplasmic 489 inheritance also confirms the significant role of the X chromosome in male courtship preference 490 differences between these species (Kawanishi et al., 1981). Though we didn't strictly control for 491 an effect of the Y chromosome, it is unlikely to explain our results because hybrids with the D. than the X chromosome, as both hybrids behave more similarly to D. simulans males. It remains 506 to be seen whether the same loci affect 7,11-HD response between these species, but the 507 effects/interactions of loci are undoubtedly different. However, D. simulans males still strongly preferred D. simulans females over F1 hybrid females, 529 likely due to the presence of 7,11-HD on the F1 female cuticle (Coyne, 1996), which suppresses 530 courtship in D. simulans males (Billeter et al., 2009). This is also likely the reason that D. 531 melanogaster males court F1 females comparably to D. melanogaster females. As with the male 532 preference difference between D. melanogaster and D. simulans females, the preference 533 difference between D. simulans and F1 hybrid females still maps to the X chromosome: simX 534 males behave like D. simulans males -courting F1 females infrequently, and significantly less 535 frequently and with less effort, than D. simulans females (Figure 2). Likewise, melX males 536 behave like D. melanogaster, courting F1 females with a similar frequency as D. melanogaster 537 females, and significantly more frequently than D. simulans females. Because the overall 538 preference patterns between D. melanogaster and D. simulans were replicated when we 539 compared melX and simX male courtship towards F1 and D. simulans females, we additionally 540 observed each of the 16 DP(1;Y) hybrid strains with F1 hybrid females. Again, we expect the 541 courtship preferences of all DP(1;Y) hybrid strains to resemble the simX hybrid strain unless the 542 duplicated D. melanogaster X chromosome segment harbors male preference loci, in which 543 case we expect them to court F1 hybrid females at higher frequencies than D. simulans females, 544 as we see for melX males. 545 When we observed hybrid males from the 16 DP(1;Y) genotypes, we found that some 546 males from every strain courted F1 females. Most genotypes courted F1s at low levels, as we 547 saw for simX hybrids. BSC100 hybrids were the only duplication hybrids that displayed higher 548 courtship frequency and effort with F1 hybrids than with D. simulans females, replicating the 549 pattern seen in melX hybrid males. However, like simX hybrids, BSC100 hybrids showed no 550 courtship towards D. melanogaster females. The fact that BSC100 hybrid males prefer F1 551 females to D. simulans females, but are still unwilling to court D. melanogaster females, 552 suggests that the D. melanogaster variants at this locus are insufficient to completely mask the 553 effects of the D. simulans X genome, which is also present in the BSC100 hybrid. We 554 hypothesize that the greater courtship we see towards F1 hybrid females is not seen towards D. are multiple male preferences and female cues that have evolved, the BSC100 duplication may 558 recover the D. melanogaster preference towards one signal, but still be insufficient to activate 559 the P1 courtship command neurons because of inhibition on other sensory channels by the D. 560 simulans genome (Clowney et al., 2015). 561 This hypothesis is supported by our perfuming data. Although past work has shown that 562 the male preference difference between D. melanogaster and D. simulans is primarily dictated 563 by female pheromones (Manning, 1959) -specifically 7,11-HD (Billeter et al., 2009) -we could 564 not be sure that that BSC100 hybrid males were responding to this cue, especially because they 565 were unwilling to court D. melanogaster females that also express this pheromone. However, 566 we found that BSC100 hybrid males significantly preferred D. simulans females perfumed with 567 7,11-HD to sham-perfumed females, and courted them at a frequency and effort comparable to 568 what we saw with F1 hybrid females, which also have 7,11-HD on their cuticle. These results 569 confirm the role of this X region in 7,11-HD response. Taken together, our findings demonstrate 570 that a single 1.35 Mb segment of the X chromosome has a specific effect on the evolved 7,11-571 HD preference differences between D. simulans and D. melanogaster. 572 573

Subdividing this region for fine-mapping results in the loss of the significant preference 574 difference 575
In order to further fine map the X chromosome region duplicated in BSC100 hybrids, we 576 created 6 additional hybrid genotypes with partially overlapping duplicated segments (Table  577 1B). When we observed these overlapping duplication hybrid strains, none showed the same 578 pattern we observed for BSC100. Five of these had higher courtship frequencies with D. 579 simulans females than with F1 females, just like simX males, although two of these differences 580 were not significant. One strain, BSC101, did have a marginally higher CF and CE with F1 581 females than D. simulans females, albeit non-significantly. This segment has the largest overlap, 582 covering 98% of the segment in BSC100 hybrids ( Figure 3C). We hypothesize 2 potential 583 explanations for the loss of significant preference when this region was subdivided. 584 (1) The genetic architecture of male courtship preference within this region is polygenic. 585 It is possible that male courtship preference differences are polygenic-even if these genes are 586 constrained within a single 1.35 Mb segment. In this case, subdivision of this locus may reduce 587 the behavioral effect if these loci are additive, or result in its loss altogether if these loci have 588 epistatic interactions. This possibility fits somewhat with the pattern we observe with our 589 overlapping duplications: hybrids with the smaller overlapping segments have lost the 590 phenotype entirely, while the largest overlap appears to, at least partially, reproduce the 591 phenotype. The simplest model that fits this scenario consists of at least two interacting loci, at 592 either end of BSC100, such that all loci are never captured by any of the overlapping duplication 593 hybrid strains. BSC101 overlaps 98% of BSC100. If this is indeed the case, then at least one locus 594 must reside within the 2% not covered by BSC101. This type of genetic architecture is not 595 uncommon. Many loci contributing to a single phenotype constrained within a single region 596 have similarly been discovered for morphological traits in Drosophila and other organisms 597 (2) Hybrid males heterozygous for regions of the X-chromosome behave differently than 599 typical hybrids. Overall, we observed reduced levels of courtship among our duplication hybrid 600 strains compared to simX or melX hybrids, suggesting that the duplication hybrids behave 601 differently from typical hybrids. The consistent reduction in courtship by duplication hybrids 602 also reduces our statistical ability to detect a significant effect. We feel that abnormal behavior 603 of duplication hybrids -particularly those carrying smaller subdivisions of the initial 16 604 duplication segments, is the most likely explanation for why our attempts to fine-map the 605 BSC100 region were unsuccessful. 606 Duplication hybrid males may behave differently for a variety of reasons; the most likely 607 are those that stem from the Y-translocated X-duplication segments themselves. Males made 608 heterozygous for regions of the X chromosome that are typically hemizygous may have 609 abnormal behavior due to epistatic interactions between X chromosome loci. In the melX and 610 simX hybrids, D. melanogaster and D. simulans X loci are not present in the same genetic 611 background, but they are in duplication hybrids. These loci may interact in unpredictable, non-612 additive ways, having unforeseen effects on behavior. Alternatively, genes translocated from 613 the X to the Y chromosome may have altered expression patterns due to their new genomic 614 environment, producing a similar effect. Smaller segments, like those we used for fine-615 mapping, may be more susceptible to this problem, as genes contained within smaller 616 translocated segments are more likely to be surrounded by a foreign genomic environment. 617 The panel of Y-linked X duplications we used to create duplication hybrids was also 618 created using irradiation (Cook et al., 2010). In fact, each breakpoint was induced by irradiating 619 males, originally creating strains that contained large subdivisions of the X chromosome, like 620 BSC100. Further subdivision of these regions (to create strains like BSC101 and all of the smaller 621 subdivisions that we used for fine-mapping) required additional irradiation. This additional 622 irradiation likely introduced new mutations to the genetic background of these flies, making 623 comparisons between BSC100 and its subdividing strains, like BSC101, imperfect. 624 Finally, these duplication segments are marked with a dominant visible eye mutation, 625 Bar, that substantially reduces the shape of the eye to a small sliver in males. These males likely 626 have restricted fields of vision, and may have difficulty tracking females in the courtship arena, 627 as has been shown for mutations affecting eye pigmentation (Connolly & Cook, 1973;Spiess & 628 Schwer, 1978). Indeed, our data collectors noted during courtship observations that males 629 often seemed to lose track of the females they were courting, and courtship would cease. This, 630 too, likely contributed to the reduced courtship frequency and effort we observed, but is 631 constant across all duplication hybrids. Regardless of the cause(s) of atypical behavior in our 632 duplication hybrids, our failure to fine-map the BSC100 region must be considered in light of 633 the above caveats. 634 635

Testing five candidate genes yields inconclusive results 636
We were able to test five candidate genes we identified within the BSC100 region, 637 either through the use of gene aberrations or RNAi knockdown. Qualitatively, our experiments 638 using an Smr gene aberration suggest that the loss of Smr expression in a melX hybrid 639 background has no effect on courtship behavior, as Smrhybrids behaved like Smr balancer 640 hybrids in that they court both D. simulans and D. melanogaster females at equal frequency 641 and with equal effort. Quantitatively, however, Smrhybrids showed higher overall courtship 642 frequencies and efforts towards both females compared to balancer hybrids. This result 643 suggests that males harboring an X chromosome balancer are less vigorous, and may behave 644 atypically due to the presences of large inversions on a hemizygous sex chromosome. Thus, 645 balancer hybrids are not an ideal comparison. 646 The results of our comparisons of Ten-aand Pde9hybrids are congruent with that of 647 Smrhybrids. Ten-aand Pde9hybrids also court both female types with equal frequency and 648 effort. There are, however, quantitative differences in the courtship frequencies of each hybrid, 649 suggesting that D. melanogaster genetic background also influences male behavior (each has 650 the same D. simulans background), making comparisons between strains imperfect. 651 Nonetheless, in courting indiscriminately, all three of these strains display a different 652 overall pattern of courtship than intact melX males, which court D. melanogaster females at 653 significantly higher frequencies than D. simulans females (melX males also display non-654 significantly higher efforts with D. melanogaster females). This may simply be because melX 655 males were created using a different D. melanogaster background (LH M ), and other D. 656 melanogaster backgrounds may not discriminate as strongly (consistent with the Smr balancer 657 hybrid data and RNAi/balancer data discussed below). In the case of Ten-ahybrids, the D. 658 melanogaster background is Canton-S, which also courts D. melanogaster more frequently than 659 D. simulans females, and does not differ from LH M in this respect ( Figure 2A). Thus, it may also 660 be because each of these genes plays a small roll in reducing male preference for D. 661 melanogaster females, and additively produce a much larger effect, like that seen with BSC100 662

hybrids. 663
The results of our RNAi knockdown screen did identify one gene that, when knocked 664 should yield a clear result is where the difference in behavior is attributable to a loss of function 681 or expression in D. simulans. However, differences in phenotype can also be due to coding 682 differences among genes, or differences in the amount, location, or timing of gene expression. 683 In these cases, it is unclear what phenotypic change to expect from males completely lacking a 684 gene altogether. While comparing gene aberrations from a D. simulans male may provide the 685 reciprocal test (if phenotypic change is due to a loss in D. melanogaster), this test is still subject 686 to all of the same problems discussed above. Additionally, it would require significant time and 687 effort to create these aberrations, as none are currently available in D. simulans strains. This 688 difficulty is specific to mapping male phenotypes on the X chromosome, as quantitative 689 complementation (Turner, 2014) and reciprocal hemizygosity tests are not possible (Stern, 690 2014). 691 It is important to note that we only tested five candidate genes. We selected these 692 genes because they met specific criteria (see methods) that we believed made them likely 693 candidates, and because screening 159 knockouts is too large an undertaking (we observed 694 nearly 500 pairs of courting flies to test our 5 candidates). It is quite possible, then, that the 695 gene(s) responsible are among those that did not meet our strict criteria. For instance, perhaps 696 the gene(s) are expressed in the developing larvae, rather than the adult CNS. Alternatively, 697 perhaps the gene(s) are not specific to the nervous system, and instead have a more general 698 function. Finally, the gene(s) may not directly interact with Fru, and instead act downstream or 699 independently of Fru. 700 701

Conclusions 702
Our results demonstrate that male courtship preference differences between D. melanogaster 703 and D. simulans is at least partially explained by 1.35 Mb region of the X chromosome. We 704 further show that this region responds to the presence of the D. melanogaster cuticular 705 hydrocarbon pheromone, 7,11-HD. Unfortunately, attempts to fine-map this region were 706 unsuccessful using the DP(1;Y) hybrid method and a candidate gene approach. Because hybrid 707 offspring of these species are sterile, we cannot pursue other avenues to map the causal loci, 708 such as QTL mapping. Similarly, because males are hemizygous for the X chromosome, we Still, our findings contribute to our understanding of the 7,11-HD preference phenotype. 712 Although the neuronal circuitry required for 7,11-HD response in D. melanogaster has been 713 known for some time (Clowney et al., 2015), it was just recently found that the same circuitry 714 detects and responds to 7,11-HD in D. simulans (Seeholzer et al., 2018). While the anatomy of 715 this circuit has remained constant during the divergence of these species, the valence of male 716 response has undoubtedly changed -in large part due to changes in the interactions between 717 these neurons, rather than their physical connections. It is still unclear what genetic changes 718 are required to modify the interactions of these neurons, but our results provide a narrowed 719 region of the genome with which to identify and continue to test candidates. Our results also 720 highlight the difficulty of dissecting such a complex phenotype using a purely mapping 721 approach. It is our hope that these data, when paired with functional dissection of the nervous 722 system, can contribute to the identification of alleles explaining behavioral evolution. This is a 723 necessary goal if we wish to understand the common patterns of genetic change underlying 724 behavioral divergence. 725 Table 1. DP(1;Y) D. melanogaster stocks. 1A. The 16 Y-linked X duplication strains used to create duplication hybrids and the reported breakpoints of their Y-linked X duplication segment (if multiple segments, base pairs are indicated for each). Ranges are shown for breakpoints where precise estimation is not available. 1B. The 6 Y-linked X duplication strains used to create duplication hybrids overlapping the region covered by BSC100 and the reported breakpoints of their Y-linked X duplication segment. Note, all translocations contain a basal segment of the X chromosome (X:1;X:493529), and many also contain a region from the end of the X chromosome (ex: X:21572099-22456281). Coordinates of translocated segments are from D. melanogaster genome release 6 (Dos Santos et al., 2015).  Tenascin-a (Ten-a) Ten-a[cbd-KS171] introgressed into a CantonS background   Courtship effort is shown for the same male and female combinations. Courtship effort is calculated as the percent of time that males spent courting (only males that courted were included in these calculations). Boxplots show the median (bold black line), interquartile range (box) and full extent of the data excluding outliers (whiskers).