Polyandry enhances offspring viability with survival costs to mothers only when mating exclusively with virgin males in Drosophila melanogaster

Abstract A prominent hypothesis for polyandry says that male–male competitive drivers induce males to coerce already‐mated females to copulate, suggesting that females are more likely to be harassed in the presence of multiple males. This early sociobiological idea of male competitive drive seemed to explain why sperm‐storing females mate multiply. Here, we describe an experiment eliminating all opportunities for male–male behavioral competition, while varying females’ opportunities to mate or not with the same male many times, or with many other males only one time each. We limited each female subject's exposure to no more than one male per day over her entire lifespan starting at the age at which copulations usually commence. We tested a priori predictions about relative lifespan and daily components of RS of female Drosophila melanogaster in experimental social situations producing lifelong virgins, once‐mated females, lifelong monogamous, and lifelong polyandrous females, using a matched‐treatments design. Results included that (1) a single copulation enhanced female survival compared to survival of lifelong virgins, (2) multiple copulations enhanced the number of offspring for both monogamous and polyandrous females, (3) compared to females in lifelong monogamy, polyandrous females paired daily with a novel, age‐matched experienced male produced offspring of enhanced viability, and (4) female survival was unchallenged when monogamous and polyandrous females could re‐mate with age‐ and experienced‐matched males. (5) Polyandrous females daily paired with novel virgin males had significantly reduced lifespans compared to polyandrous females with novel, age‐matched, and experienced males. (6) Polyandrous mating enhanced offspring viability and thereby weakened support for the random mating hypothesis for female multiple mating. Analyzes of nonequivalence of variances revealed opportunities for within‐sex selection among females. Results support the idea that females able to avoid constraints on their behavior from simultaneous exposure to multiple males can affect both RS and survival of females and offspring.


| INTRODUCTION
Classical ideas (Bateman, 1948) say females are "passive" and choosy with limited reproductive capacities predicting few benefits for polyandrous females. Yet, polyandry is common (Anderson, 1974;Gowaty, 2006Gowaty, , 2012Gowaty, , 2013Gowaty & Hubbell, 2013;Taylor, Price, & Wedell, 2014), despite potential fitness costs to females (Otti, 2015). Given classical assumptions, several functional hypotheses may explain why females multiply mate including that female multiple mating may be a result of male-male competitive drive (Trivers, 1972), occurring when males coerce females to mate. Male manipulation ("gifts" or "lures") may affect females' nutritional status enhancing direct benefits for females whenever males provide females with resources transferred during courtship or mating (Arnqvist & Nilsson, 2000), but male coercive mechanisms by definition also create female fitness costs. Female multiple mating may also arise as a correlate to selection on males to mate multiply (Halliday & Arnold, 1987), and it is possible that endosymbionts manipulate females to mate multiply (Wedell, 2013).
An alternative (Anderson, Kim, & Gowaty, 2007;Gowaty, 1996Gowaty, , 2008Gowaty et al., 2007b) to classical ideas that females are coy and choosy assumes that females have evolved resistance mechanisms to coercion, whenever coercion is costly to female fitness or the fitness of their offspring. For example, whenever wild-living females are able to escape or avoid the behavioral effects of male harassment (Gowaty, 1996), polyandry may evolve with few costs to breeding females and with health or other viability benefits for offspring (Gowaty, 2008;Gowaty, Kim, Rawlings, & Anderson, 2010;Lively, 1996;Simmons & Holley, 2011). The freedom of females' movements in wild flies is notable to watchers in the wild (Markow & O'Grady, 2005), particularly so for females that first arrive at new feeding sites (SC pers obs). In the wild, flying females may be able to escape or avoid coercive males, just because they can fly away.
However, in the general discussion of potential mechanisms affecting female reproductive decisions-including to mate or not with multiple males under coercion-investigators seldom focus on females' options to avoid coercion, which is what we have attempted to do here, while testing a variety of potential explanations for female multiple mating.
Given the diversity of the hypotheses explaining multiple mating and recently reviewed in Gowaty (2012;, we used an experimental design (Table 1) allowing simultaneous tests of alternative predictions of multiple hypotheses, while reducing opportunities for male behavioral coercion of females. Using captive Drosophila melanogaster free of Wolbachia and Spiroplasma endosymbionts, we controlled females' exposure to conspecific males so that no female in any treatment was with more than one male in a single day, providing some leveling of the ecological playing field of subjects in a way that seldom occurs in captive studies (Billeter, Jagadeesh, Stepek, Azanchi, & Levine, 2012;Maklakov, Immler, Løvlie, Flis, & Friberg, 2013).
We report variation in components of fitness of female subjects in two by two matched sets of treatments testing a priori predictions (Leek & Peng, 2015) of hypotheses (Gowaty, 2012(Gowaty, , 2013Gowaty et al., 2010) about the fitness costs/benefits of mating opportunities available to female subjects. Predictions include the following: 1. Females may gain direct fitness benefits from exposure to males in which case virgins may die faster than mated females.

2.
Limits to the number of, or the viability of sperm in a single ejaculate, probably occasionally occur in most organisms, especially those without sperm-storing tissues or organs, but flies do have "sperm management" organs (Markow & O'Grady, 2005) suggesting that one copulation for many organisms is enough to fertilize a females' lifelong production of eggs. Nonetheless, sperm limitation occurs in some Drosophila species (Turner & Anderson, 1983, 1984 and may favor female multiple mating in D. melanogaster. If so, females achieving only a single copulation may oviposit fewer eggs than females with multiple copulations.

3.
If multiple mating increases female exposure to pathogens or parasites (Lively, 1996), female lifespan may be reduced.

4.
However, even if polyandry extracts costs decreasing female survival, mating with multiple males may allow them to increase reproductive success (RS) through sorting of male haplotypes in a lottery P E Females (n = 30) with a novel, experienced male (n = 30), each age-matched-tofemales with copulation ad libitum. We round-robin rotated males daily. For example, on day 2 female 1 was with the male who was with female 2 on day 1. On day 3, female 1 was with the male who was with female 2 on day 2, and so-forth. Because of death day variation, we held some females or males for a single day without exposure to the opposite sex.

5.
If 4) is so, polyandrous females may have shorter lifespans, but higher reproductive success than lifelong monogamous females.
6. Female multiple mating may occur because of male-male competitive drive resulting in behavioral coercive polyandry. If so, when females can escape or otherwise mitigate coercive polyandry, females may manage their re-mating schedules to reap fitness rewards that may accrue without incurring costs.

7.
Under the assumption that experienced males, age-matched to females are less "eager" and less coercive than young, virgin males (Hoffmann, 1990) (which has been attributed to conditioning after male exposure to already-mated females), polyandrous females with age-matched males may live longer than polyandrous females exposed daily to young, virgin males, and perhaps as long as females in lifelong monogamy.  Figure 1 shows results comparing lifelong virgins (V LF ) to females exposed to one male on only 1 day (M OC ). Figure 2 shows results comparing M OC females-to-females in lifelong monogamy (M L ). Figure 3 shows results comparing M L females-to-P E females in lifelong polyandry who were exposed to a novel, age-matched male each day. Figure 4 has results comparing P E females to P V females exposed to a young virgin male each day. Table 2  (c) of between-treatment variances in female RS and survival (Table 3) facilitating a discussion of the opportunity for selection on females in the absence of behavioral sexual selection in either males or females.

| Notes on natural history
Drosophila melanogaster are common, human commensals. Females sometimes fail to re-mate for about 5 days (Markow & O'Grady, 2005), but some females also may copulate several times in a single day before becoming refractory (SC pers obs).

| Capture of flies and testing for endosymbionts
Our subjects were from a multifemale stock LA1206 set up in December, 2011 that included only individuals drawn from endosymbiont-free isofemale lines that we collected between September and October 2010 from locations in Los Angeles, CA (Castrezana, Faircloth, & Gowaty, 2010). Before constituting LA1206, we tested 277 isolines (Braig, Zhou, Dobson, & O'Neill, 1998;O'Neill, Giordano, Colbert, Karr, & Robertson, 1992;Pool, Wong, & Aquadro, 2006) from over 200 locations in the Los Angeles basin. Fewer than 1% (nine isolines) were free of Spiroplasma and Wolbachia, although these lines where each collected at geographically distinct places in the Los Angeles area.

| Culturing of subjects
We maintained the endosymbiont-free isolines on cornmeal transferring them every 10-15 days for 19 months. In June 2012, we popu- We never used CO 2 when handling flies. We expected that our culturing methods, including the expansion of the population, had allowed for a near-natural level of genetic diversity among our subjects with limited opportunities for selection at least compared to other studies of polyandry in other captive insects.

| Environmental conditions of the experimental room
We ran our experiments in a controlled-temperature (21°C) room with 12/12-hr light/dark period. UCLA's Drosophila Kitchen provided cornmeal fly food, which we modified: Using a BPA-free container, we put 450 g of solidified cornmeal food and 100 ml ddH 2 O, which we melted in a microwave (high for 4.5 min), and then added 10 ml of alcohol (Everclear, 190-proof, 95% ABV) as an extra mold inhibitor (SC pers. obser.). We used a 100 cc syringe to set 7.5 ml of cornmeal food in each vial.

| Experimental controls
• Technicians and laboratory helpers were blind to our hypotheses and predictions.
• Once adult subjects were in vials, they were never able to encounter another same sex individual. Thus, no behavioral sexual selection was possible either among males or among females.
• A key to our experimental treatments was the elimination of coercion of females from simultaneous interactions with multiple males, that is, in all treatments except for lifelong virgins, females' exposures to males were limited daily to only one male, so that no female ever saw more than one male in a 24-hr period thereby eliminating opportunities for male-male behavioral competitive effects on subject females' reproductive decisions.
• We randomly placed females in five treatments and males in one treatment at the start of the experiment. Each subject was the same age, and each treatment set began on the same day, so that the ages of all individuals in all treatments were the same.
• There were five treatments for females in the experiment and one for males, which we labeled arbitrarily as "A," "B," "C," "D," "E," and "F" to mask the manipulation from helpers (Table 1).
• We then matched 150 females-30 per treatment-labeling each female subject with an ID # from 1 to 30, and 30 virgin males subjects.
• We then sorted vials into 30 matching sets by ID numbers containing a single vial from each treatment. Thus, each matching set had six vials, one from each treatment but having the same ID #.
Matching by ID number across treatments controlled for bench effects as we rotated the orientation of vials daily in boxes and on shelves. This matching of subjects between treatments also allowed us to do robust day by day comparisons of components of fitness among treatment females and of lifespans of females and males, similar to other published studies Turner & Anderson, 1983) • Our analyzes are of two types. Most tested a priori planned predictions of hypotheses (Leek & Peng, 2015), and thus, the results are capable of rejecting or confirming a priori predictions. In addition, as an explicit control, we characterize some of our analyzes as descriptive and exploratory. We define exploratory analyzes following (Leek & Peng, 2015/p 1314 as "data interpretation that builds on a descriptive analysis by searching for discoveries, trends, correlations, or relationships between the measurements to generate ideas or hypotheses."

| Treatments
See Table 1. On 9/8/2012, using flies 6-day posteclosion, we randomly put 150 female subjects (30 individuals in each of five treatments) and 120 males (30 individuals used in four of the female treatments). We also put 30 additional males into a male only treatment (Table 1). We emphasize again that no posteclosion female or male saw more than one opposite sex conspecific on any day, eliminating any sexual selection from female-female and male-male behavioral competition, similar to an earlier study using D. pseudoobscura (Gowaty et al., 2010).

| Behavioral observations
On day 1, we scanned all vials for 3 hr after placing a male in a vial with a female to record if the pairs copulated. Only nine of 150 female subjects each of whom were with a single male failed to copulate in the first three hrs of day 1: M OC females: C14, C15, C26, C27, C30; and M L females: D4, D17, E1, and E9. However, all females copulated on day 1, as offspring eclosed from each day-1 vial. On following days, after moving females each day to new vials we scanned each vial for copulations. We emphasize that after day-1, our attention was only on whether a subject had additional copulations. Our haphazard observations of copulations after the first day indicate that additional copulations occurred in all treatment groups in which males were present, however, we did not continue to watch vials throughout the 24 hr that females had access to males simply because it would have been impossible given the size of the experiment, not to mention extraordinarily costly. We emphasize that despite the interests of others in the numbers of copulations that polyandrous females might have, our interest was rather on the likelihood that females flexibly take or resist options for re-mating (Gowaty, 2013). We designed the experiment to enhance females' abilities to manage or avoid coercion from males that can arise under male-male behavioral sexual selection (Trivers, 1972): No female subject ever was with more than one male a day except for M OC females who saw one male on the first day of the experiment and V F females, who never encountered a male. Otherwise females had ad libitum access to interactions with a known male (M L ) or a novel male each day (P E or P V ). from which one could randomly draw without replacement a set of unique females for each day of life. Such an experiment would be difficult requiring extraordinary resources, especially given the many controls we used.

| Components of fitness and numbers of observations
Unplanned analyzes allowed exploration (Leek & Peng, 2015) of lifespan and RS variation as well as between-treatment variances of fitness components.

| V L versus M OC
Lifespan variation of V L versus M OC females tested the hypothesis that copulation enhances female survival, a conclusion in a study of wild-living D. melanogaster (Markow, 2011): previously mated females lived longer than never-mated females, a surprising result because mated individuals often die faster than virgins (Partridge, 1987). Markow speculatively attributed her unexpected finding either to (1) the enhanced feeding opportunities of already-mated females, who were presumably older than unmated females and/or to (2) male-derived benefits delivered at copulation. In the current experiment, subjects entered the experiment at the same age, yet M OC females lived significantly longer than V L females ( Figure 1a) and produced significantly more eggs. Our experimentally controlled food availability plus the fact that all subjects were the same age put the differential feeding time idea off the table as an explanation of longer life in M OC compared with V L females. Some may argue that in our captive flies mated females were hungrier and ate more than V L females enhancing the health of M OC females, but others would expect that enhanced eating would decrease female lifespan (Grandison, Piper, & Partridge, 2009). Nevertheless, our results agree with Markow's, 2011) observations of wild flies. The significant enhancements to lifespan and egg number (Figure 1b) for M OC compared with V L females are consistent with ejaculate contributions nourishing zygotes and females (Gillott, 2003) and/or mating-induced female resource contributions and/or immunity (Morrow & Innocenti, 2012;Zhong et al., 2013), but of course, our results cannot discern between these alternatives. The possibility of male-derived benefits from copulation implies between-sex physiological cooperation that may enhance mother's health, in contrast to male manipulation/coercion of females that may decrease female survival (Wigby & Chapman, 2005). The fact that a single copulation enhances female lifespan compared to lifespan of virgin females is consistent with the idea that male-derived benefits may favor female multiple mating.

| M OC versus M L
M OC and M L (Figure 2) differences evaluated the cost of multiple copulations and tested female RS variation associated with possible sperm limitations (not enough or nonviable sperm), which occurs in some species (Turner & Anderson, 1983, 1984. Multiple copulations may be energy and time taxing for females, predicting that compared to M OC females, M L females die faster. Indeed, M OC females lived significantly longer than M L females (Figure 2a). Despite shorter lifespans, M L females laid significantly more eggs, and produced significantly more eclosed offspring with significantly enhanced offspring viability (Figure 2b), all results consistent with the hypothesis that multiple copulations provide material benefits to females that enhance all components of female RS.
Studies of D. pseudoobscura Turner & Anderson, 1983, 1984 had similar results. Despite the survival costs to females of more than one copulation, the reproductive benefits to females of multiple copulations are suggestive of similar benefits from copulation with multiple mates, not just of multiple copulations, an idea which the next comparisons between M L and P E females directly tests.

| M L versus P E females
Polyandry costs may not be offset by any benefits if encounters are random and mate choice is absent. Microbes are common, so that females mating with multiple partners are likely to have greater exposure to pathogenic fungi, viruses, and bacteria (Otti, 2015), which can affect female health and induce perhaps costly upregulation of immune responses in females (Knell & Webberley, 2004;Lockhart, Thrall, & Antonovics, 1996;Zhong et al., 2013) permitting the prediction that polyandrous females have great mortality risk than females with multiple copulations in lifelong monogamy (Figure 3). Frequently predicted benefits (Lively, 1996;Williams, 1975) of polyandry compared to lifelong monogamy include enhanced offspring viability from diversification of progeny genes (see summary, Table 2).
M L and P E females had statistically similar lifespans (Figure 3a), not a unique result Simmons & Holley, 2011), but perhaps unexpected given that P E females were with an unfamiliar, novel male each day, who had also been previously exposed to other females (except for the first day of the experiment) presumably increasing pathogen exposure risk. Offspring viability was significantly greater for P E than M L females (Figure 3b): Polyandry compared to monogamy enhances lineage success, reducing extinction risk (Price et al., 2010). There were no statistical differences in the number of eclosed offspring (2nd panel Figure 3b), but the distribution of daily differences showed P E females had late life RS advantage over M L females that seemed to have early life advantage over P E females.
Differences in offspring viability (3rd panel, Figure 3c) occurred because M L females laid significantly more eggs, fewer of which sur- 2007a) saying that females breeding under constraints compensate for expected deficits in health of offspring by increasing egg number.
The next comparison between polyandrous females with exposure to age-matched experienced males P E versus those exposed each day to virgin males P V , under the assumption that experienced males are less coercive than virgin males, directly tests the prediction that reduction in coercion benefits females.

| P V versus P E
Assuming inexperienced, young, virgin males are more sexually assertive (Hoffmann, 1990) than experienced, older, previously mated males, a comparison of polyandrous females daily with a new, young virgin male (P V ) versus polyandrous females with an experienced, previously mated male age-matched to the females (P E ) may illustrate how behavioral variation of males to females (Long, Markow, & Yaeger, 1980) may affect female survival and RS. Further, given the classical assumption that females are unlikely to increase RS when they mate with multiple males, RS components should be no different for P V versus P E females. P V lifespan began declining when other females still had half or more of their lives in front of them. P E females live significantly longer than P V females (Figure 4a), and every RS component was significantly enhanced for P E versus P V females (Figure 4b, all three panels). Constant exposure to young, virgin males extracted costs to females and lowered offspring viability. As in D. bipectinata (Krishna, Santhosh, & Hegde, 2012), polyandrous D. melanogaster females mating older males had more offspring and healthier offspring than females mating younger males. Coercive attention from young males may explain female preferences for older (Avent, Price, & Wedell, 2008;Brooks & Kemp, 2001;Hansen & Price, 1995;Somashekar & Krishna, 2011), perhaps more sedate males, and not be just a function of male fertility that increases as males age (Long et al., 1980). We speculate that (1) differences will be revealed with comparisons of behavior of virgin versus already-mated males to virgin and mated females (one on one to control for withinsex behavioral competition) and (2) mechanistic studies will reveal F I G U R E 4 P V versus P E comparisons of female survival (a) and components of RS (b). (a) Product-limit survival fit of P V versus P E shows statistically significant differences in female lifespan (Log-Rank X 2 = 27.2171, df 1, P > X 2 = 0.0001; Wilcoxon X 2 = 18.6104, df 1, P > X 2 = 0.0001). (b) P E -P V matched pairs mean differences over female age (N = 27) in components of RS. Top panel: number of eggs: P E oviposited 0.94 ± 0.47 (SE) more eggs/day than P V females (Wilcoxon Signed Rank S = 69.00, P < S = 0.0490). Middle panel: number of eclosed offspring. P E females had 2.27 ± 0.56 (SE) more eclosed offspring/day than P V females (Wilcoxon Signed Rank S = .000137, P > |S| < 0.0003; Prob > S = 0.0001). Bottom panel: arcsine fraction egg-to-adult survival. P E females' average/day arcsine egg-to-adult survival was 1.19766 and P V females was 1.09 (Wilcoxon Signed Rank S = 117.000, P > |S| = 0.0030 and P < S = 0.0015)

(a) (b)
physiological loads with impacts on female's RS when they interact exclusively with eager, and perhaps lower "fertility" virgin males. Table 2 summarizes the a priori predictions and the results in Figures 1, 2, 3, and 4.

| Exploratory analyzes
The experiment facilitated several exploratory analyzes, including across treatment comparison of the lifespans among females ( Figure 5a) and among males (Figure 5b). Behavioral sexual selection could not explain the variation among females (Figure 5a) or among males (Figure 5b), because no subject was ever with a same sex conspecific in this experiment.
That P V females (Figure 5a) died significantly faster than females in other treatments suggests usually over-looked benefits to females who-in the wild-may be able to escape virgin males, just by flying away.
Despite absence of male-male behavioral competition, multiplymated males (those that rotated between P E females) died significantly faster than males in other treatments (Figure 5b). We discuss the comparative lifespan variation between subject females and males elsewhere (Gowaty et al. in prep.).
The mean daily variation in RS components with the time course for P V females started to decline perhaps due to behavioral "eagarness" T A B L E 2 A priori planned tests of predictions (second column) and results of tests (third column) of hypotheses of adaptive significance of multiple copulations and polyandry Male-male competitive drive produces polyandry with greater sexual conflict reducing female survival females experienced from virgin males. The rapid decline in offspring viability for M L females compared to P E females also suggests a potential interaction effect with males' reproductive capacities, when with one female for life versus with several females over their lifespans, suggesting significant mating costs to both sexes in monogamy.
Significant variance differences between treatments (Table 3) included female survival (p < .001), egg number (p < .0153), and offspring viability (p < .0001). Variance differences between females in lifespan, egg number, and offspring viability are consistent with some social environments having greater potential for producing evolutionary responses in females, an observation begging for more investigation of selection on females.

| CONCLUSIONS
The results of these experiments provided no support for nonadaptive, by-product polyandry ( M L females oviposited significantly more eggs than P E females, but there were no statistical differences in the numbers of eclosed offspring. Thus P E females had statistically greater offspring viability indicating that polyandry fosters lineage success (Price et al., 2010). P E females with lifelong random exposure to many males-one at a time-had opportunities to manage sperm use, which also begs greater mechanistic attention.
The comparisons between P V and P E females seem completely consistent with the idea that virgin males are more "eager" compared with more "sedate" experienced males: P V females have significantly lower fitness than P E females, a result consistent with previous observations of eager virgins versus more sedate experienced males (Greenspan & Ferveur, 2000). An experiment that would potentially buttress that conclusion would include detailed behavioral observations of subject females and the males they are exposed to every day of their lives. Even without such labor-intensive observations our results also are consistent with the earlier observations of (Hoffman 1990) indicating the enhanced "eagerness" of virgin males compared with experienced males. Our results are consistent too with the idea that females who are able to escape male harassment may have longer lives and greater reproductive success at all ages than females unable to escape male coercion.
The evidence here suggests that greater female control via reduced exposures to simultaneous male-male behavioral competition is adaptive for D. melanogaster females. Wild females who can fly away may seek coercion-free social situations, flexibly favoring their own continued survival and the viabilities of their offspring (Gowaty, 2013).
Polyandry may have fewer benefits for laboratory-living than wildliving flies simply because opportunities for avoidance of coercion are fewer inside of fly-filled jars than outside.

| Differences with an earlier study of D. melanogaster polyandry
In contrast to the results here, Brown, Bjork, Schneider, and Pitnick (2004) found no benefits for polyandrous females compared to monogamous females with multiple copulations. The differences in scope and methods of the two studies probably completely account for the differences in results. Brown et al.'s source populations had been in laboratory culture for 200 + generations allowing for significant selection in captivity, ours for <20. Ours were free of the endosymbionts, Spiroplasma and Wolbachia, something not reported in Brown et al.
Their experiment lasted 14 days when their subjects were 19-20 days old; ours lasted 69 days including the natural lifespan of each of our subjects who each entered our experiment when they were 6 days old.
Their monogamous subjects were exposed to the same male at 48-hr interval for 4 hr each time, and their polyandrous females saw a new unfamiliar male every other day (without notes about ages or experience of the males). In contrast, our matched pairs-M L and P E -were constantly exposed to males (but only one at a time). Our repeated measures allowed comparisons of daily variation in RS of all subjects, and controlled uninteresting variance due to bench effects. Last, our study used only wild-type flies, while Brown et al. used flies with visible heritable mutations to evaluate sperm competition, a part of their study with no counterpart in ours. Any of the many differences in source of flies, endosymbiont-load variations, culturing techniques, handling protocols, time in captivity, use of mutants versus wild-type flies, methods including timing of mating, handling variation, etc. could account for the differing conclusions between Brown et al. (2004) and the current study.

ACKNOWLEDGMENTS
We thank Stephen P. Hubbell for comments on a previous version of the manuscript; Roland Azurin, UCLA undergraduate, and Miles Meyer, a UCLA postgraduate, for help in the lab. NSF IOS Grant # 1121797 to PAG supported the research.

CONFLICT OF INTEREST
None declared.

AUTHOR CONTRIBUTIONS
PAG designed experimental treatments. SC captured, cultured, and maintained populations, carried out all treatments, managing helpers, daily data acquisition, and computerization of data. SC and PAG checked computer records against manuscript records. BCF designed and ran genetic tests for endosymbionts. PAG and WCB analyzed the data and PAG wrote the manuscript. PAG was corresponding author.