CESAM, Department of Biology, Universidade de Aveiro, Campus de Santiago, Aveiro, 3810, Portugal

ARC Centre of Excellence for Coral Reef Studies, James Cook University, Townsville, 4811, QLD, Australia

Scottish Oceans Institute, School of Biology, University of St Andrews, East Sands, St Andrews, KY16 8LB, Fife, United Kingdom

Abstract

Background

The observation that females mate multiply when males provide nothing but sperm - which sexual selection theory suggests is unlikely to be limiting - continues to puzzle evolutionary biologists. Here we test the hypothesis that multiple mating is prevalent under such circumstances because it enhances female fitness. We do this by allowing female Trinidadian guppies to mate with either a single male or with multiple males, and then tracking the consequences of these matings across two generations.

Results

Overall, multiply mated females produced 67% more F2 grand-offspring than singly mated females. These offspring, however, did not grow or mature faster, nor were they larger at birth, than F2 grand-offspring of singly mated females. Our results, however, show that multiple mating yields benefits to females in the form of an increase in the production of F1. The higher fecundity among multiply mated mothers was driven by greater production of sons but not daughters. However, contrary to expectation, individually, the offspring of multiply mated females do not grow at different rates than offspring of singly mated females, nor do any indirect fitness benefits or costs accrue to second-generation offspring.

Conclusions

The study provides strong evidence that multiple mating is advantageous to females, even when males contribute only sperm. This benefit is achieved through an increase in fecundity in the first generation, rather than through other fitness correlates such as size at birth, growth rate, time to sexual maturation and survival. Considered alongside previous work that female guppies can choose to mate with multiple partners, our results provide compelling evidence that direct fitness benefits underpin these mating decisions.

Background

Female multiple mating is prevalent in nature, even when males provide no material benefits such as food or parental care to females

The adaptive significance of multiple mating has been extensively debated and the general idea is that, to be adaptive, the costs of multiple mating must be offset by benefits that enhance female fitness. Two types of benefits are commonly used to explain the adaptive value of multiple mating: non-genetic benefits (direct/first generation)

Although, direct benefits play a crucial role in the adaptiveness of female multiple mating

Despite evidence that females gain indirect benefits (i.e., second-generation benefits) from multiple mating, two difficulties are often identified. First, it is difficult to disentangle direct and indirect benefits. The mechanism by which females obtain direct benefits, such as increased fecundity, may also affect offspring viability, thereby obscuring any evidence of indirect benefits

In order to circumvent these difficulties, a stronger test of the adaptiveness of multiple mating would be one that: 1) tracks the fate of offspring across two generations

Quantifying the benefits of multiple mating solely on the basis of the number of offspring produced, however, may produce biased estimates of fitness. Fitness is a function of the number of viable descendants produced, as well as the influence that other life history traits have on the performance of offspring in particular contexts, or at a given point in the life cycle

Here, we test the hypothesis that multiple mating results in increased F1 fitness, by examining the number of ‘grand-offspring’ produced. We develop and apply a multi-generational test that allows us to disentangle the contributions of first generation (direct) and second-generation (indirect) effects of this outcome. Specifically, using the Trinidadian guppy (

We first tested the prediction that multiple mated females obtain indirect benefits by producing more grand-offspring (F2). We did this by comparing the number of F2s generated via single and multiple mated F0 treatments. We then examined how mating success and brood size of F0 females and F1 offspring contributed to this effect. First, the number of F1s produced was compared between the two mating treatments. Second, we tested whether the F1 offspring of multiply mated females were more viable than those of singly-mated females, when all F1s were paired with randomly selected mates under a common garden experimental design. This allowed us to attribute any overall differences in the number of grand-offspring produced to; having more F1 offspring, to having F1 offspring that were more viable, or to a combination of both. In addition, to assess the extent to which these fitness measures may be biased by differences in offspring characteristics, we measured size at birth and growth rates, which have been previously described as important fitness correlates. Differences between the offspring of single and multiple mated females in these quantities could, in nature, offset differences in the number of F1 or F2 offspring produced.

Results

F0 to F2

The estimated probability of breeding success for F0 females (i.e., of successfully producing a first brood) was higher for multiple (0.67) than for single (0.55) matings, but the difference was not statistically significant (likelihood ratio test:

**Fitness Variable**

**Mating treatment**

**N**

**Mean**

**SEM**

**F0**

**F1**

**F2**

**F0**

**F1**

**F2**

**F0**

**F1**

**F2**

Sample size (N), Mean and Standard Error of the Mean for single and multiple mating treatments at F0, F1 and F2, for each component of fitness. Mean brood size for F2 was calculated using only viable F1s (i.e., individuals surviving to 12 weeks). Breeding success in brackets is the fitted probability of producing a brood, based on a negative binomial distribution for F0 and a binomial probability distribution for F1. The means and standard error of means for breeding success were calculated using these two probability distributions. For all other variables arithmetic means are presented. Growth rate was calculated as the rate of weekly growth (cm) over 12 weeks. Maturation is the number of days from birth until sexual maturation. Mortality is the number of individuals that died per mating treatment before producing a first brood.

Brood Size

Single

40

121

335

3.02

2.83

0.28

0.20

Multiple

39

154

436

4.00

3.00

0.36

0.16

Breeding Success

Single

73(.55)

121(.70)

32.7

82.2

0.55

0.46

Multiple

58(.69)

154(.75)

18.0

116

0.43

0.40

Growth Rate

Single

121

144

0.12

0.12

.003

.002

Multiple

148

160

0.11

0.13

.002

.003

Size at Birth

Single

121

144

0.85

0.86

0.06

0.07

Multiple

147

148

0.86

0.86

0.06

.004

Sexual Maturation

Single

58

158

45.6

40

1.65

0.77

Multiple

91

187

47.8

42

1.43

0.76

Mortality

Single

3

8

15

Multiple

2

10

14

Indirect Fitness (F0 to F2)

**Indirect Fitness (F0 to F2).** Mean number of viable grand-offspring (F2) produced by a singly or multiply mated F0 females via **(A)** all F1, **(B)** male F1, and **(C)** female F1. Values in **(B)** and **(C)** do not sum to the values in **(A)**, because not all F0s produced mixed sex broods. Whiskers indicate 95% bootstrap percentile confidence intervals. Sample sizes used to calculate the means are shown in Supporting Information 1.

F0 to F1

Multiply mated females produced 60% more viable F1 offspring, on average, than singly mated females (Figure

**Supporting Information 1.** Total number of individuals and the descendants produced, which were used to calculate the means and confidence intervals for Figures

Click here for file

Direct Fitness (F0 to F1)

**Direct Fitness (F0 to F1).** Mean number of viable offspring (F1) produced by singly or multiply mated F0 females, counting **(A)** all F1, **(B)** male F1, and **(C)** female F1. Means were calculated using only those F1s that reached sexual maturity. Whiskers indicate 95% bootstrap confidence intervals. Sample sizes used to calculate the means are shown in Supporting Information 1.

F1 to F2

In the next generation, there were no significant differences in the reproductive success of individual F1 that had been produced from multiple versus single matings (Figure _{Fo singly}; SR = 0.94, _{F0 multiply}; SR = 0.93,

Partitioning Fitness (F1 to F2)

**Partitioning Fitness (F1 to F2).** Mean number of viable offspring (F2) produced by **(A)** all F1, **(B)** only male F1, and **(C)** only female F1. Whiskers indicate 95% bootstrap confidence intervals. Sample sizes used to calculate the means are shown in Supporting Information 1.

Effect of multiple mating on size at birth, growth rate, time to sexual maturation and survival

Model selection revealed that for all traits, the estimated best model included a random effect due to tank of origin (i.e., random variation among F0 females), but no consistent difference between females in different treatments. There was also some support for a mixed effect model, which included treatment and tank (Table

**Supporting Information 2.** Model selection using values of ΔAIC (Akaike weights). k: Number of parameters of the model. The estimated best fitting model is shaded in grey.

Click here for file

**Size at birth**

Effect of mating treatment on size at birth, growth rate and time to sexual maturation in F1s and F2s using linear mixed-effects models. Tank of origin was used as a random effect nested within mating treatment. Parameters are only shown for best-fitted model. Selection of best-fitted model and Akaike weights are shown in Additional file

**F1**

Estimate

Std. error

DF

Fixed effect

Intercept

−0.160

0.008

191

<0.001

Random effects

Std. Dev.

Tank (Intercept)

0.068

Residual

0.044

**F2**

Estimate

Std. error

DF

Fixed effect

Intercept

0.859

0.006

262

<0.001

Random effects

Std. Dev.

Tank (Intercept)

0.020

Residual

0.057

Growth rate

**F1**

Estimate

Std. error

DF

Fixed effect

Intercept

−2.192

0.024

190

<0.001

Random effects

Std. Dev.

Tank (Intercept)

0.184

Residual

0.162

**F2**

Estimate

Std. error

DF

Fixed effect

Intercept

−2.120

0.041

274

<0.001

Random effects

Std. Dev.

Tank (Intercept)

0.204

Residual

0.190

Time to sexual maturation

**F1**

Fixed effect

Estimate

Std. error

DF

Intercept

3.825

0.034

84

<0.001

Random effects

Std. Dev.

Tank (Intercept)

0.253

Residual

0.130

**F2**

Fixed effect

Estimate

Std. error

DF

Intercept

3.720

0.027

242

<0.001

Random effects

Std. Dev.

Tank (Intercept)

0.174

Residual

0.201

Discussion

Despite growing evidence that females obtain reproductive benefits from mating multiply, the extent to which these benefits are partitioned between first- and second-generations remains controversial

Our results strongly support the hypothesis that multiple mating is adaptive, as manifested in an increase in female fecundity. We found that multiply mated females produce substantially more grand-offspring than singly mated females. However, because the reproductive output (F2) of progeny from multiply and singly mated females was not significantly different, we also showed that this fitness advantage is driven by the production of more offspring in the first generation (F1), rather than by elevating the fitness of offspring (second-generation effects).

Our results do not preclude the possibility that selection for indirect benefits exists, because direct and indirect benefits may operate simultaneously

Fitness can be defined as a measure of the proportion of individuals that are propagated into the following generations

It is possible that singly mated F0 females produced fewer F1s as result of brood retention, but we think this possibility is unlikely. One advantage of multiple mating over single mating is that it enables mechanisms of post-copulatory sexual selection to operate to maximize fitness

When the environment is controlled, as it was in this experiment, it is the increased quantity of F1 offspring, rather than the quality of those offspring, that is the key determinant of fitness. Previous work found the offspring of multiply mated female to be larger at birth, and phenotypically more diverse, and to have enhanced schooling, and predator avoidance skills

Similar increases in fecundity associated with multiple mating have been reported across different taxa

The greater fecundity in the multiple mating treatment was driven by the over-production of viable sons (> 87%). Likewise, multiply mated females of house wren produce a surplus of male offspring

One possible mechanism for the over-production of sons is the existence of segregation distortion genes or sex ratio meiotic drive. During spermatogenesis, the sex ratio distortion gene links to one of the sex chromosomes and prevents the production of functional gametes bearing the other sex chromosomes

Conclusions

For most of the twentieth century, studies of sexual selection assumed that female fitness can be maximized by mating with a single male

Methods

Experimental design

We used descendants of wild caught guppies from the Lower Tacarigua River, Trinidad, to generate virgin females and males that were later used to generate singly and multiply mating broods. Sixty pregnant females were haphazardly selected and transferred to single 10 L tanks and allowed to give birth. Of the 60 females, 51 produced broods that provided the first generation of fish (F0) used in our experiment. After birth each offspring was allocated to a single 10 L tank for 12 weeks at which point sex could be unambiguously determined. All individuals were kept in identical laboratory conditions. Tanks were filled with de-chlorinated tap water, contained clean natural gravel and maintained at approximately 20–24°C under a 12-hour light/dark regime. Each tank had its own filter. All individuals were fed ad libitum daily with live artemia.

At three months old, F0 females and F0 males were haphazardly allocated to either a single or a multiple mating treatment. In both mating treatments, each female had access to only one male at any given time. After the first day, all males were removed and allocated to individual tanks for 24 hours. On the second day, in the single mating treatment the same male was introduced to the same female as in day 1, whereas in the multiple mating treatment a novel male was introduced to the female. This process was repeated for the next two days, with the same male introduced to the singly mated female and a new male introduced to the multiply mated female. F0 males allocated to the multiple mating treatment were not rotated among different replicates (i.e., each group of four males was only used in one replicate). In both mating treatments, F0 females were allowed to settle for 24 hours before mating trials began. F0 males were introduced the following day at 0700 and removed at 1700. We adopted a similar experimental design to that used by Tregenza and Wedell (1998)

All tanks were inspected for newborns twice a day (morning and afternoon). After the birth of the first brood the F0 female was allocated to a stock tank and not used again. F0 females that failed to successfully produce a brood were removed and replaced by a new female. Therefore, to obtain a total of 40 broods for each treatment, 73 and 78 F0 females were mated for the single and multiple mating treatment, respectively (Table

After 12 weeks, each F1 offspring was presented with either a virgin female or male (these individuals were reared in individual tanks and used only as pairs for F1s), according to its sex, of similar size, and allowed to mate freely until a first F2 brood was produced. In contrast with the F0s, F1s were allowed to mate indeterminately until either a first brood of F2s was produced, or one of the F1 fish died. After the birth of the first F2 brood, a random sample of F2s had their size at birth, growth rate and time to sexual maturation recorded (Table

All behavioural observations were carried out at the School of Biology at the University of St Andrews. The premises where the observations were carried out comply with the ASAB Guidelines for the treatment of animal in behavioural Research and Teaching, set by UK Home Office (PCD 60/2609).

Statistical design

We first tested whether multiply mated females had a significantly higher probability of producing a first brood than singly mated ones (i.e., breeding success). Because the response variable for calculating breeding success is dichotomous (i.e., a females either produces a brood, or not), and matings were conducted until 40 replicates per mating treatment were obtained, the appropriate probability distribution is the negative binomial (i.e., the probability distribution for the number of “trials” required to obtain a pre-determined number of “successes”

where the subscripts _{
i
} is the number of “trials” (i.e., number of F0 females for whom mating was attempted) in treatment _{
i
} is the pre-specified number of “successes” (F0 females that produced a brood: 40 for each treatment, in our case) in that treatment, and _{
i
} is the probability that a randomly chosen F0 female will successfully produce a brood in that treatment. _{
i
} and _{
i
} are the data, and the _{
i
} are the parameters that must be estimated. To test for differences in breeding success between mating treatments, we fitted two versions of the likelihood in eq. (1): one in which breeding success differed between mating treatments (i.e., _{s} and _{m} were estimated as distinct parameters (_{s} ≠ _{
m
}), and a second in which the multiply and singly mated females had the same probability of success (_{
s
}
_{
m
}
_{
s
}
_{
m
}
_{
s
} ≠ _{
m
}), and we determined whether the null hypothesis of equal mating success could be rejected with 95% confidence by comparing the two models with a likelihood ratio test.

Second, we tested whether the average number of viable grand-offspring produced (i.e., those surviving to adulthood; ~12 weeks) was greater for multiply mated females than for singly mated ones. No standard parametric distribution provided a satisfactory characterization of the number of grand-offspring per successful brood. Therefore, we used non-parametric bootstrapping, which makes no distributional assumptions about the data _{
s
} and _{
m
}, we randomly drew a probability of breeding success from the uncertainty distribution around our negative binomial maximum likelihood estimate (MLE) of this quantity (which we obtained from the inverse of the second partial derivative of the likelihood function, according to standard likelihood theory). Second, we used non-parametric bootstrapping to produce an uncertainty distribution for the mean number of viable grand-offspring produced per F0 female (i.e., number of F1 that survive 12 weeks). By randomly drawing a value of

In addition, to gain more insight into the proximate mechanisms by which any differences in reproductive success arose, we conducted two further sets of analyses. We used a bootstrap analysis similar to that described above (except that viable sons and daughters were counted, rather than grand-offspring) to estimate the difference in average number of viable F1 offspring produced from multiple versus single matings. We also estimated the difference in the average number of F2 produced between F1 offspring of multiply mated mothers and F1 offspring of singly mated mothers. In F1s, however, a fixed number of individuals were mated (rather than mating occurring until a fixed number of successes occurred), so the appropriate likelihood for estimating the probability of breeding success for F1 individuals was the binomial distribution, rather than the negative binomial distribution.

Additionally, we also compared sizes at birth, growth rates and time to male sexual maturation between mating treatments, to determine whether any differences in brood size were being traded off against any other fitness related traits. Both of these response variables followed an approximately Gaussian distribution after log-transformation, allowing application of a more conventional statistical analysis. In the analysis of brood size, there is only one response variable value per parent (number of progeny). For birth, growth, and time to sexual maturation, however, we have replication within parents (i.e., each offspring contributes a value). Because parental effects on these traits are likely, we treated each parent as a random effect nested within mating treatment, and fitted the data with linear mixed-effects models (function glmmPQL in R)

For all analyses involving brood size, reported

Competing interests

The authors declare that they have no competing interests.

Authors’ contributions

MB, AEM conceived the idea and designed the experiment. MB carried out the experiment. MB, SRC, MH, MD, performed the statistical analyses. MB, SRC, MH, MD, AEM, interpreted the results and wrote the manuscript. All authors read and approved the final manuscript.

Acknowledgments

This study was supported by a PhD fellowship to MB by Fundação para a Ciência e a Tecnologia (FCT), Portugal. MD and AEM acknowledge the European Research Council (advanced grant BioTIME 250189) for funding. We also thank Carl Smith, Alfredo Ojanguren, David Shuker, Christine Dreye and the Biodiversity and Behaviour Group at University of St Andrews, for comments and discussions. All behavioural observations complied with current UK animal welfare and health and safety regulations, and met ASAB Guidelines for the treatment of animals in behavioural research and teaching.