Multiple mechanisms of cryptic female choice act on intraspecific male variation in Drosophila simulans

Postcopulatory sexual selection can arise when females mate with multiple males and is usually mediated by an interaction between the sexes. Cryptic female choice (CFC) is one form of postcopulatory sexual selection that occurs when female morphology, physiology, or behavior generates a bias in fertilization success. However, its importance in nonrandom reproductive success is poorly resolved due to challenges distinguishing the roles of females and males in generating patterns of fertilization bias. Nevertheless, two CFC mechanisms have recently been documented and characterized in Drosophila simulans within the context of gametic isolation in competitive hybrid matings with Drosophila mauritiana: sperm ejection and nonrandom use of sperm storage organs for fertilization. Here, we explore if and how female D. simulans employ these two mechanisms of CFC in response to intraspecific male size variation. We used transgenic males expressing green (GFP) or red fluorescent protein (RFP) in sperm heads to document postcopulatory processes, in conjunction with a probabilistic analytical model. We unexpectedly found that differential reproductive success was also a function of male population (GFP or RFP), suggesting that females use different CFC mechanisms to select for different male traits. Moreover, concordance of selection at the precopulatory (as measured by mating latency) and postcopulatory stages depends on both the male trait and the CFC mechanism examined. Larger males were more successful both before and after mating, but we unexpectedly found that females also mated more quickly with males with GFP-labeled sperm, while fertilization bias favored RFP-labeled sperm.


Introduction
storage are evolutionarily labile across the Drosophila lineage, but they are always both present 141 (Pitnick et al. 1999). Although relatively complex reproductive tracts can provide females with 142 an opportunity to engage in CFC, they may not always evolve mechanisms to do so. Using a 143 probabilistic analytical model (Manier et al. 2013c), we estimated three parameters of 144 fertilization bias between sperm from two males: 1) bias in first-or second-male sperm used 145 from the spermathecae, 2) bias in first-or second-male sperm used from the SR, and 3) an 146 "organ use bias" that described which (if either) of the two sperm storage organ types are used 147 preferentially. This approach was applied in D. melanogaster, D. simulans, and D. mauritiana 148 and revealed different modes of fertilization bias in all three species. In D. melanogaster, the SR 149 was favored over the spermathecae, while D. simulans exhibited second-male bias from the 150 spermathecae but first-male bias from the SR. We found no fertilization bias at all in D. precedence. 160 Here, we examine the degree to which CFC at both stages is employed when male quality 161 (body size) is systematically varied in conspecific matings. We also evaluate any associations among precopulatory female choice (mating latency) and both stages of CFC. Evidence for 163 consistent selection across precopulatory and postcopulatory stages of sexual selection is mixed, 164 but more studies have found that attractive males enjoy a competitive advantage during sperm  Consistent with this pattern, Hosken et al. (2008) found that in D. simulans, attractive males 169 (those that mated faster) sire more offspring, with a significant correlation between second-male 170 mating latency and second-male paternity success (P 2 ). Part of this study's elegance lies in 171 allowing experimental females to define male attractiveness rather than the researchers, but other 172 studies found that female D. simulans mate more frequently with larger males both in natural 173 (Markow and Ricker 1992) and laboratory populations (Taylor et al. 2008). We manipulated 174 male size in a fully factorial double mating design, in which wild type females were randomly 175 assigned to one of four mating treatments: two large males (LL), two small males (SS), a large 176 male followed by a small male (LS), and a small male followed by a large male (SL). We 177 previously found evidence that D. simulans females use their first mate as a basis for evaluating 178 their second mate in competitive heterospecific and conspecific matings, such that females 179 ejected heterospecific sperm much sooner if the heterospecific male was her second mate than if 180 he was her first mate (Manier et al. 2013b). We thus predicted that females exposed to a second 181 male that was much larger (SL) or much smaller (LS) than her first mate would yield the 182 strongest evidence for both precopulatory and postcopulatory female choice.

187
For all four treatment groups (LL, SS, SL, LS) across three concurrent experiments described 188 below, we quantified the following parameters for both first and second matings: mating latency 189 and copulation duration (first copulation from Experiments 1-3, second copulation from  For all matings, we used D. simulans males from lines with GFP-or RFP-labelled sperm heads 199 (henceforth "GFP" or "RFP"; Manier et al 2013a). GFP lines also carried a fluorescent GFP eye 200 marker that is physically linked to the sperm label, which allowed paternity assignment of 201 offspring sired by GFP males. All females were derived from the same wild type population into 202 which the transgenic populations were backcrossed for five generations. All stocks were 203 maintained at ambient room temperature (23-25°C) and light regime in half-pint milk bottles on 204 standard corn meal-agar-yeast-molasses medium sprinkled with live yeast grains.
Males of two distinct body size classes were generated by transferring first-instar larvae 206 to vials at densities of 300 individuals with 0.25 cm 3 medium and 50 individuals with 1.5 cm 3 207 medium. Larval density has previously been shown to influence larval development and adult (ESM)). We generated L and S males from both GFP and RFP lines to allow sperm competition 214 between an L or S GFP male against an L or S RFP male in a fully factorial mating design (Table   215   S1). Large and small males differed in thorax length (all Tukey p-values < 0.001 in pairwise 216 comparisons of L and S males) and L-GFP males were a little bit larger than L-RFP males, while 217 S-GFP and S-RFP males did not differ in thorax length (see ESM and Fig. S1 for further details).

218
Experimental males and females were collected as virgins within 6 h of emergence under 219 CO 2 anesthesia and maintained in plastic vials (10 flies per vial) with 1.5 cm 3 medium 220 supplemented with live yeast. Females were mated at 2-3 days, and males at 3-7 days post-    Table S1). Sample size for LS matings was higher than in the other treatments, because we 242 expected females to be less willing to remate with a male that was smaller than her first mate.

243
Females were provided a 4-hr opportunity to remate with a second male 2, 3, and 4 days after 244 their first mating. For first and second matings, we recorded mating latency and copulation 245 duration. For second matings, mating latency was measured in number of minutes of interaction, 246 accumulated over days, if applicable. For example, a female that remated after one hour on the 247 second day of remating showed a mating latency of 300 minutes (5 hrs), which also accounts for 248 the previous day's 4-hr opportunity to remate. Immediately after remating, females were frozen, dissected, and sperm in the reproductive tract counted as described above. All males were frozen 250 after mating and thoraxes measured for body size. RFP and GFP sperm were counted at 400X or 251 630X on a Nikon Eclipse Ni-U compound microscope. For all experiments, sperm counts were 252 performed blind to treatment and male mating order.  Table S1) and mated a third group of flies as described above. We quantified the 259 timing of ejection by gently aspirating females immediately after copulation into individual wells 260 of glass 3-well spot plates (Pyrex) and covered with glass coverslips secured with spots of clay.

261
Females were monitored for ejection under a stereoscope every 5-15 min, until either an ejected 262 mass (slightly smaller than an egg) was observed or a 4-hour time limit was reached. with either RFP labeled sperm (RFP sire) or unlabeled sperm (GFP sire). 279 We applied the paternity and sperm count data to estimate fertilization bias using an 280 analytical model that calculates first-or second-male bias from the spermathecae (x) and the SR 281 (y), as well as bias in sperm use from the different types of sperm storage organ (z; Manier et al.  organ, X or Y. x or y = 0.5 represents equal probabilities for first-and second-male sperm to be used for fertilization (i.e., no fertilization bias, or sperm from competing males are used in 293 proportion to their relative abundance in storage).

294
It is important to note that this model estimates bias in how sperm are used for 295 fertilizations above and beyond the relative proportions of first-male and second-male sperm in 296 storage. The null hypothesis for this test is that there is no fertilization bias; that is, sperm are 297 used from the first and second male in direct proportion to their relative abundance in storage. 298 Thus, even if the second male's sperm outnumber first-male sperm in storage, fertilization bias 299 may still favor the first male's sperm. This bias may not result in an overall P 2 less than 0.5 (the 300 first male's sperm "wins" overall), but it would yield a lower P 2 than under no fertilization bias. For all statistical analyses, we used R 3.0.2 (R Development Core Team 2013). We used t-test to 305 detect the difference between large and small males, and general linear models (lm) with male 306 type (L-GFP, S-GFP, L-RFP, S-RFP) as a factor to detect size differences among male larval 307 density treatments in the separate experiments. To detect size differences among male 308 treatments, multiple comparisons were performed with Tukey's tests using the "multcomp" We used lms to analyse the effect of our size treatments on sperm transfer, number of 315 progeny produced before and after remating, mating latency in the first and second mating 316 (log 10 -transformed), copulation duration in the first and second mating, and ejection latency 317 (log 10 -transformed). Full models for mating latency and copulation duration in the first mating of first-male sperm in storage at the end of second copulation, which did not include the genotype of the GFP male, because these data were from Experiment 2 for which we did not 337 have that information. 338 We applied model selection using the AIC (Akaike 1973) for lms and GLMs with a 339 negative binomial distribution. We applied a forward selection algorithm using R's AIC statistic  Table S4 for model selection process).

344
Proportional data, i.e., P 2 and S 2 in the SR and spermathecae, were analysed with GLMs 345 with quasibinomial error distribution (binomial models were overdispersed) and a logit link 346 function with sample sizes as weights. The full model included first-male size × second-male 347 size interaction, second-male line, GFP genotype, female thorax length, and the number of 348 progeny produced before remating. The number of progeny produced before remating can affect 349 P 2 values, because the more first-male sperm that are used for fertilization prior to remating, the 350 fewer that will be available for fertilization after remating, potentially increasing P 2 (see

361
Heterozygosity for the GFP marker did not differ among treatments (χ 2 = 5.2, df = 3, p = 362 0.16). Homozygous males tended to have a lower proportion of second male sperm in the SR (S 2 363 SR) than heterozygous males (t 82 = 2.0, p = 0.052), but this difference did not result in 364 significantly lower P 2 (Table 5). In matings with virgin females, small males (t 485 = 4.2, p = 0.002) copulated longer than large 369 males (Table 1, Table 3). In rematings, copulations with GFP males lasted on average 1.3 ± 0.7 370 minutes longer than with RFP males (t 326 = 1.9, p = 0.063; Table 4), and small males copulated 371 on average 1.2 ± 0.7 minutes longer than large males (t 326 = 1.8, p = 0.068) but these differences 372 were not significant at the α = 0.05 level. In addition to male size, females unexpectedly also mated more quickly with males according to 385 line, with a shorter mating latency with GFP males. Virgin females were equally likely to mate 386 with GFP or RFP males (deviance = -0.04, p = 0.84, df = 1; Fig 1a) but were quicker to mate 387 with GFP males (t 487 = 4.2, p < 0.001, Fig. 2a; Table 3). Mated females were both more likely to 388 mate with GFP males (deviance = -12.3, p < 0.001, df =1; Fig. 1b) and had a shorter mating 389 latency with them (t 325 = 4.5, p = 0.027; Fig. 2b, Table 4). These results suggest that both virgin 390 and mated females overall choose to mate more quickly with GFP males.  There are a number of factors that could explain this low P 2 for the LS treatment. We 400 found that small males transfer fewer sperm upon remating than large males (t 108 = -6.4, p < 0.001; Fig. 4; Table 2, Table 4), with no effect of the first male's size on the number of sperm 402 transferred by the second male (Table S4). We also found that all four male types (L-GFP, S-403 GFP, L-RFP, S-RFP) had no differences in number of sperm transferred to virgin females (Table   404 1, Table 3, Table S4). In D. simulans and related species, the number of sperm in an ejaculate is the SS treatment to also have a lower P 2 , but it does not (Fig. 3).

409
There is some evidence that D. simulans females use their first mate as a basis by which  In this case, timing of ejection as a cryptic female choice mechanism best explains the low P 2 of 413 the LS treatment, because females ejected sperm sooner if their second mate was small (t 117 = -414 2.2, p = 0.028; Table 2, Table 4), as well as if their first male was small (t 117 = -2.5, p = 0.015; 415 Table 2, Table 4). Ejection time is affected most by a decrease in quality (size) from the first 416 male to the second (LS) than an increase in quality (SL). Using planned orthogonal contrasts, we 417 found that females that mate first with a large male eject sperm of small males sooner than those 418 that mate with a second large male (LS vs. LL; t = 2.23, p = 0.026). However, females whose 419 first mate is small relative to their second mate have no difference in ejection time (SL vs. SS; t = 420 1.08, p = 0.28), suggesting that a step down in male quality has more of an effect on ejection 421 time than a step up. Overall, females mate more quickly with larger males and eject sperm more  Across all treatments, GFP males had higher paternity success than RFP males (t 123 = 7.8, p < 430 0.001; Fig. 3; Table 5). At the same time, GFP males transferred more sperm to mated females 431 than RFP males (t 108 = 6.3, p < 0.001; Fig. 4, Table 4). Larger ejaculates of GFP males are 432 predicted to more effectively displace first-male RFP sperm when in the offensive second-male 433 role, allowing them to achieve a higher P 2 . Furthermore, GFP first males had more sperm 434 remaining in storage upon remating than RFP first males (z 109 = -4.5, p < 0.001; Table S3), 435 giving them a greater advantage in resisting displacement in the defensive role.  (Table S4), suggesting that higher P 2 of GFP males is best explained by their superior sperm 438 numbers rather than female ejection. In the egg-laying phase (Stage 2 CFC), we predicted that 439 females would shift use of their sperm storage organs between the second-male biased 440 spermathecae (z < 0.5) , and the first-male biased SR (z > 0.5) based on whether their preferred 441 male was first or second. Here, we define fertilization bias favoring first-male or second-male 442 sperm as disproportionate use of first-male or second-male sperm beyond relative proportions in 443 the sperm storage organ(s). In other words, sperm that are heavily outnumbered in storage may 444 still be used for fertilization in a biased manner if they are used disproportionately more often than the numerically dominant sperm. In this experiment, we found persistent second-male bias 446 in the spermathecae (x > 0.5; Fig. 5a, b), consistent with Manier et al. (2013b, c). On the other 447 hand, the SR switched from first-male biased (y < 0.5) to second-male biased (y > 0.5), 448 depending on the line of the second male, to favor RFP sperm (Fig. 5c, d; with the exception of 449 SS with GFP as second male). When RFP was the second male, sperm storage organ use favored 450 the second-male biased spermathecae (z < 0.5), with no clear pattern of organ bias when GFP 451 males were second (z = 0.5; Fig. 5e, f). Despite this cryptic female preference for RFP sperm, 452 GFP males have a higher P 2 , likely due to their larger ejaculates. We therefore found evidence 453 that females mate more quickly with GFP males but may select against them at the egg-laying 454 phase. Nevertheless, GFP males have higher P 2 due to superior sperm numbers over RFP males.  Regardless of its adaptive significance, female selection of an easily distinguishable male marker 533 nevertheless provides a valuable opportunity to dissect precopulatory and postcopulatory female 534 preference of multiple male traits with unprecedented resolution. 535 We found evidence that females adjust the timing of sperm ejection based on the 536 comparison between the quality of their first and second mates. Females had shorter ejection 537 times after remating when their second male was of lower quality than their first male (smaller), 538 but no increase in ejection time when their second male was larger than their first male. In other 539 words, ejection time only changed when the second male was perceived as a "step down". This  There are few studies that have examined how a female's first mate influences her 546 remating behavior or preference for a second mate, but the available evidence shows no clear 547 patterns (e.g., Byrne and Rice 2005). Pitnick (1991) tested several hypotheses explaining female 548 remating pattern in D. melanogaster, including "Mate improvement", in which females remate 549 more quickly with a higher quality (larger) male, and "Mate diversity", in which females will 550 remate more quickly with a phenotypically different male. His results supported a "Courtship 551 threshold hypothesis", in which females remate more quickly with larger males, presumably due 552 to their more vigorous courtship (Partridge et al. 1987). Our results support an entirely different 553 hypothesis that we call the "Poor male avoidance hypothesis", in which females adjust remating behavior (ejection time) in response to a decrease, but not an increase (or no change), in male 555 quality from her first mate to her second mate.

556
In summary, we found that different male traits are favored via different mechanisms at 557 both the precopulatory and postcopulatory stages. Large males had a consistent advantage over 558 both stages, with timing of sperm ejection the primary mechanism of cryptic female choice. In      Tables  Table 1 Percentages