Content uploaded by Lorenzo Zane
Author content
All content in this area was uploaded by Lorenzo Zane on May 11, 2017
Content may be subject to copyright.
1
Evans JP et al. (2002) Nature 421, 360-363 doi:10.1038/nature01367
This is the authors’ post-review copy.
The full text is available on the publisher website at
https://www.nature.com/nature/journal/v421/n6921/full/nature01367.html
Directional postcopulatory sexual selection revealed
by artificial insemination
Jonathan P. Evans, Lorenzo Zane, Samuela Francescato & Andrea Pilastro
Department of Biology, University of Padova, via U. Bassi 58/B, Padova, I-35131,
Italy
Postcopulatory sexual selection comprises both sperm competition, where the
sperm from different males compete for fertilization1, and cryptic female choice,
where females bias sperm utilization in favour of particular males2. Despite
intense current interest in both processes as potential agents of directional sexual
selection3, few studies have attributed the success of attractive males to events
that occur exclusively after insemination. This is because the interactions
between pre- and post-insemination episodes of sexual selection can be important
sources of variation in paternity4. The use of artificial insemination overcomes
this difficulty because it controls for variation in male fertilization success
attributable to the female’s perception of male quality, as well as effects due to
mating order and the relative contribution of sperm from competing males5.
Here, we adopt this technique and show that in guppies, when equal numbers of
sperm from two males compete for fertilization, relatively colourful individuals
achieve greater parentage than their less ornamented counterparts. This finding
indicates that precopulatory female mating preferences can be reinforced
exclusively through postcopulatory processes occurring at a physiological level.
Our analysis also revealed that relatively small individuals were advantaged in
sperm competition, suggesting a possible trade-off between sperm competitive
ability and body growth.
Postcopulatory sexual selection, arising from polyandry, is a pervasive evolutionary
force that permits directional selection in the form of sperm competition1 and cryptic
female choice2,3. In experimental studies, the outcome of postcopulatory sexual
selection is traditionally described in terms of the proportion of offspring sired by the
second male to copulate with a female, formally defined as the P2-value4. However,
values of P2 are typically characterised by extreme and often unexplained variation6
which can obscure the relative importance of pre- and post-insemination processes of
sexual selection. The use of artificial insemination overcomes this difficulty because it
uncouples precopulatory sexual selection (notably mate choice) from postcopulatory
selective processes. The technique therefore offers unrivalled opportunities for
distinguishing between the fertilizing capacities of different males7 and makes it
possible to test explicitly whether directional sexual selection can proceed exclusively
through postcopulatory processes at a physiological level5. Here, we adopt this
technique and examine the importance of postcopulatory sexual selection on preferred
male traits in the guppy Poecilia reticulata, an internally fertilizing species of
freshwater fish. Recent evidence reveals that attractive male guppies (those that were
accepted as mates more quickly) achieve greater parentage following natural double
2
matings8. Our aim is to determine whether the success of attractive males can be
attributed to events occurring exclusively during post-insemination episodes of
selection, as predicted by a positive association between male sexual attractiveness
and sperm quality9-11.
Guppies have a promiscuous, non-resource-based mating system12 in which
female choice and sexual coercion by males both play a role13. During precopulatory
mate choice, females typically prefer relatively colourful males exhibiting high rates
of courtship13. In particular, the area of carotenoid coloration in males (including
orange, red and yellow) consistently influences female mating decisions13,14. We
therefore specifically focused our analysis of male phenotype on body coloration and
mating behaviour, but also included body size since this trait has been shown to
influence female choice in some15, but not all studies (see ref. 14). In each trial, the
relative phenotype scores of two males, arbitrarily labelled A and B, were related to
each male’s subsequent share of paternity following the artificial insemination of
equal sized ejaculates from both males. For each of the broods we calculated the
proportion of offspring sired by the focal male B (hereafter termed PB) and related this
parameter to differences in the behavioural and morphological phenotypes of the two
putative sires. Thirty-five inseminations were performed yielding n = 27 broods
(mean number of offspring per brood = 11.1 ± 5.5 SD; range 3-25), which is a similar
rate of success to that observed following natural double matings in this population8.
Mean body size and the extent of body coloration (mean trait measurements for the
two males in each trial) did not differ significantly between males involved in
unsuccessful (n = 8) and successful trials (n = 27) (student t-test: body size t8,27 =
0.39, P = 0.70; total body coloration t8,27 = 1.71, P = 0.11; all individual colour
components n.s.).
The resultant paternity distribution ranged from 0 to 100% (mean = 0.53 ± 0.33
SD). The variance in PB was significantly larger than the expected binomial variance
(James test statistic = 15.1, P < 0.0001)16, which is assumed under the null hypothesis
that sperm from each male have equal chances of fertilizing ova – the so-called ‘fair
raffle’ model of sperm competition1 (Fig. 1). Furthermore, a logistic regression
analysis revealed that differences in the phenotype of competing males accounted for
significant deviance in PB (Table 1). Within pairs of males, individuals with relatively
more orange sired a greater proportion of offspring than predicted by chance, whereas
the extension of black and blue did not predict male fertilization success (Fig. 2a–c).
We suggest that this finding is unlikely to be explained by the differential survival of
embryos from attractive and unattractive males since such an interpretation requires
that broods sired exclusively by the more colourful male contain more offspring than
those with shared paternity (and these more than families sired entirely by the drab
male). Our results revealed that brood size was not correlated with the proportion of
offspring sired by the most colourful male in each pair (r = 0.23, n = 27, P = 0.26)
and furthermore that brood size was not a function of the mean extension of orange of
the two males in each trial (r = 0.15, P = 0.44, n = 27).
We also found that body size was related to male success, with relatively small
males achieving greater paternity (Fig. 2d). Body size was not significantly correlated
with either absolute or relative orange body coloration (data pooled to include all n =
54 males: r = 0.18, P = 0.21; r = -0.04, P = 0.78, respectively). Again, there was no
evidence that the success of relatively small males was due to the enhanced survival
of their embryos or offspring. Indeed, brood size tended to be positively, rather than
negatively associated with the mean body size of the two males in each trial (mean
body length, r = 0.35, P = 0.07, n = 27), although this trend was less apparent when
3
one family, a statistical outlier with respect to the others, was excluded from the
analysis (r = 0.27, P = 0.19, n = 26). Although the resorption of embryos has never
been documented in poeciliids17, further work is required to determine the extent of
embryo mortality in guppies, and furthermore whether this phenomenon is non-
random with respect to male phenotype. Until this information is obtained, the
possibility that the variation in brood size observed in this study, and its possible
influence on paternity share, was due to natural selection for enhanced embryo
survival cannot be fully discounted. Finally, male courtship behaviour, estimated for
each male prior to the artificial insemination trials, did not predict male B paternity
(Table 1).
To our knowledge, these results provide the first evidence that when
behavioural interactions between males and females are prevented, postcopulatory
selection can favour the same traits that are preferred by females during precopulatory
mate choice. The colour components measured in this study, and in particular the
relative area of orange, consistently predict female mating preferences in guppies (e.g.
refs 14,18,19), including the population chosen for this study (A. Pilastro et al.
unpublished data). In previous studies, such congruence between pre- and
postcopulatory sexual selection has been reported but may have arisen because
attractive males inseminate more sperm20, or because female behaviour can favour
particular males in sperm competition2,21-23. In our experimental design, however,
ejaculate size was controlled and females were denied the opportunity of assessing
male quality, and therefore such effects cannot explain the finding that relatively
colourful males were more successful. Additionally, our results are unlikely to be
explained by female sperm selection since this mechanism requires that the haploid
expression of sperm reflect male attractiveness. Although haploid gene expression
does occur in nature, the examples to date are confined to cases involving
compatibility-based discrimination against particular sperm genotypes rather than
directional postcopulatory sexual selection (see refs 3,24 and references therein). Such
compatibility effects, in addition to other non-directional selective processes (e.g.
poor sperm mixing25) cannot account for the success of colourful males, although it is
possible that both factors contributed towards the observed variation in PB in this
study.
Our results suggest instead that males exhibiting high levels of coloration
produce competitively superior ejaculates9. Whether this effect is due to condition-
dependence (see below) or the enhanced genetic quality of colourful males remains to
be investigated. However, our results also reveal that relatively small males achieve
greater parentage than larger individuals. Unlike colour, the influence of male body
size on female preferences is less clear, with some studies reporting female
preferences for large males15 and others revealing no discrimination by females for
this trait14. Irrespective of whether body size is under selection through precopulatory
female choice, for the reasons articulated above it seems unlikely that sperm choice
by females accounted for the small male advantage uncovered by our study. Thus, to
the extent that sperm competition, rather than sperm choice accounts for our findings,
the two phenotypic traits identified in this study (body size and coloration) appear to
covary with sperm competitive ability in opposing directions. Such effects would be
predicted, for example, if sperm quality were to depend on an adequate intake of
carotenoids in the diet10 as well as energy traded-off against body growth26. An
important direction for future research is to identify the specific sperm traits that
confer reproductive advantages on males and examine how they interact with the
phenotypic traits identified in our study.
4
Methods
Origin of fish and assessment of male mating behaviour
Guppies, which were first-generation descendents of wild-caught fish from the
Tacarigua River, Trinidad, were maintained as previously described8. Behavioural
trials took place between 09.00 and 14.00. On the evening prior to each trial an adult,
non-virgin female and two sexually immature juvenile fish were placed into an
observation tank (46cm x 19cm x 27cm). On the following morning a test male (aged
six months), unfamiliar to the female and juvenile fish, was placed in the tank to settle
for 30 minutes. The test males were fully mature and were selected only if they were
in good condition and displayed sexual behaviour (as evident from their behaviour in
the stock tank prior to capture). Juveniles were included to facilitate the settlement of
the adult fish. The rate of male courtship (sigmoid displays) was measured over a 10-
minute period following an established protocol27. After each trial the female and
juveniles were replaced by different fish for the following day’s trial. The males were
isolated for 3 days prior to the phenotype measurements and the extraction of sperm
for artificial insemination.
Male ornamentation
Test males were paired at random, anaesthetized (MS222) and photographed with a
digital camera. Image analysis software (Scion Corporation,
http://www.scioncorp.com) was used to measure total length and analyse body
coloration. Briefly, three main components of these colour patterns were considered:
the area of (1) carotenoid pigmentation (including orange, red and yellow), (2)
melanistic black spots, and (3) the iridescent structural colours, which include blue
and green. These three components are termed orange, black and blue respectively in
the main text. There were no significant differences between paired males (labelled A
and B) in body size (paired t-tests: t26 = 0.75, P = 0.46), courtship behaviour (t23 <
0.01, P > 0.99; behaviour not obtained from n = 3 males) or overall body coloration
(t26 = 0.48, P = 0.63; all individual components n.s.). The analysis of body coloration
was done blind of paternity assignment (see below).
Artificial insemination
Male guppies produce sperm packaged in bundles (spermatozeugmata). In
preliminary trials, 23 males were repeatedly stripped to obtain three independent
measures for the number of sperm per bundle for each male (for two males only two
counts were obtained). Analysis of variance (with males as random factors) revealed
that between-male variation in the number of sperm per bundle did not exceed that
observed within individuals (ANOVA: F22,44=0.76, P=0.76). The number of sperm
per bundle was not significantly correlated with any of the morphological traits
measured in this study (orange: r = 0.20, P = 0.35; blue: r = 0.14, P = 0.52; black: r =
0.03, P = 0.91; body size: r = 0.03, P = 0.88, n = 23). In each trial, equal numbers of
bundles were obtained from each of the two males27, gently mixed and inseminated
simultaneously into an anaesthetised female using a machine pulled micropipette
(penetration depth approximately 2mm). The number of sperm inseminated was based
on the size of natural ejaculates in the study population (~0.5x106 spermatozoa) (see
also ref. 20). Following AI, females were revived and isolated until they produced
their first brood. Tissue samples were obtained from all fish (mother, two putative
sires and offspring) for the subsequent paternity analysis.
Paternity analysis
We observed no offspring mortality prior to paternity assignment. Two microsatellite
markers (accession numbers: AF164205 and AF026459) were used to estimate each
5
male’s relative share of paternity (PB). The PCR protocol followed previous work8
with the exception that one primer from each pair was end-labelled with a fluorescent
phosphoramidite dye. Amplified fragments were separated by electrophoresis on an
ABI 3100 sequencer (ABI PRISM), using 400 HD ROX (Perkin-Elmer) as a size
standard. PCR products were visualised using Genographer software (v. 1.6.0) and
paternity was assigned to offspring according to allele sharing between putative sires,
mother and offspring. Both loci were highly variable (18 and 33 alleles, with expected
heterozygosities of 0.901 and 0.941 respectively) and exhibited a global exclusion
probability of 0.92. Where one parent (the mother) is known, paternity assignment
with two putative sires (as in our study) is calculated to be 100% (Cervus v. 2.0)28. In
practice, paternity was unambiguously assigned to all offspring (n = 300 from 27
broods).
Data analysis
The James method16 was used to test whether the proportion of offspring sired by the
male labelled B fluctuates randomly between families (with different brood sizes)29,
and thus whether the observed PB variance deviates from binomial expectation. A
generalized linear model (GLM) with binomial errors and logit link function was then
used to determine whether male phenotype accounted for deviance in PB. For each
family, male B success (the number of offspring sired by male B) and the total
number of offspring were entered as the binomial response variables. Predictor
variables, representing differences in the phenotypic trait measurements taken from
the two males per family (male B trait minus male A trait), were fitted into the model.
Initially, the full model included all possible explanatory variables; the term with the
least significant probability was then excluded in a stepwise procedure. The change in
deviance in the GLM resulting from the removal of each term was tested against an F-
distribution. We removed all terms whose exclusion did not cause a significant change
in the deviation of the model (Table 1).
1. Parker, G. A. in Sperm Competition and Sexual Selection (eds. Birkhead, T. R. &
Møller, A. P.) 3-54 (Academic Press, San Diego, 1998).
2. Eberhard, W. G. Female Control: Sexual Selection by Cryptic Female Choice.
(Princeton University Press, Princeton, 1996).
3. Birkhead, T. R. & Pizzari, T. Postcopulatory sexual selection Nature Rev. Gen. 3,
262-273 (2002).
4. Birkhead, T. R. & Møller, A. P. (eds.) Sperm Competition and Sexual Selection
(Academic Press, San Diego, 1998).
5. Birkhead, T. R. Defining and demonstrating postcopulatory female choice - again.
Evolution 54, 1057-1060 (2000).
6. Simmons, L. W. Sperm Competition and Its Evolutionary Consequences in the
Insects. (Princeton Unversity Press, Princeton NJ, 2001).
7. Dziuk, P. J. Factors that influence the proportion of offspring sired by a male
following heterospermic insemination. Anim. Reprod. Sci. 43, 65-68 (1996).
8. Evans, J. P. & Magurran, A. E. Patterns of sperm precedence and predictors of
paternity in the Trinidadian guppy. Proc. R. Soc. Lond. B 268, 719-724 (2001).
9. Sheldon, B. C. Male phenotype, fertility, and the pursuit of extra-pair copulations by
female birds. Proc. R. Soc. Lond. B 257, 25-30 (1994).
10. Blount, J. D., Møller, A. P. & Houston, D. C. Antioxidants, showy males and sperm
quality Ecol. Lett. 4, 393-396 (2001).
11. Jennions, M. D. & Petrie, M. Why do females mate multiply? a review of the genetic
benefits. Biol. Rev. 75, 21-64 (2000).
12. Evans, J. P. & Magurran, A. E. Multiple benefits of multiple mating in guppies. Proc.
Natl. Acad. Sci. USA 97, 10074-10076 (2000).
6
13. Houde, A. E. Sex, Color, and Mate Choice in Guppies. (Princeton University Press,
Princeton NJ, 1997).
14. Endler, J. A. & Houde, A. E. Geographical variation in female preferences for male
traits in Poecilia reticulata. Evolution 49, 456-468 (1995).
15. Reynolds, J. D. & Gross, M. R. Female mate preference enhances offspring growth
and reproduction in a fish, Poecilia reticulata. Proc. R. Soc. Lond. B 264, 57-62
(1992).
16. James, W. H. The distribution of the combinations of the sexes in mammalian litters.
Genetical Research 26, 45-53 (1975).
17. Constantz, G. D. in Ecology and Evolution of Livebearing Fishes. (eds. Meffe, G. K.
& Snelson, J., F.F.) 33-50 (Prentice-Hall, New Jersey, 1989).
18. Endler, J. A. Natural and sexual selection on color patterns in poeciliid fishes.
Environmental Biology of Fishes 9, 173-190 (1983).
19. Houde, A. E. Mate choice based on naturally occurring color-pattern variation in a
guppy population. Evolution 41, 1-10 (1987).
20. Pilastro, A., Evans, J. P., Sartorelli, S. & Bisazza, A. Male phenotype predicts
insemination success in guppies. Proc. R. Soc. Lond. B 269, 1325-1330 (2002).
21. Pizzari, T. & Birkhead, T. R. Female feral fowl eject sperm of subdominant males
Nature 405, 787-789 (2000).
22. Lewis, S. M. & Austad, S. N. Sexual selection in flour beetles: the relationship
between sperm precedence and male olfactory attractiveness. Behav. Ecol. 5, 219-224
(1994).
23. Edvardsson, M. & Arnqvist, G. Copulatory courtship and cryptic female choice in red
flour beetles Tribolium castaneum. Proc. R. Soc. Lond. B 267, 559-563 (2000).
24. Zeh, J. A. & Zeh, D. W. The evolution of polyandry II: post-copulatory defences
against genetic incompatibilty. Proc. R. Soc. Lond. B 264, 69-75 (1997).
25. Harvey, I. F. & Parker, G. A. 'Sloppy' sperm mixing and intraspecific variation in
sperm precedence (P2) patterns. Proc. R. Soc. Lond. B 267, 2537-2542 (2000).
26. Neems, R. M., Lazarus, J. & McLachlan, A. J. Lifetime reproductive success in a
swarming midge: trade-offs and stabilizing selection for male body size. Behav. Ecol.
9, 279-286 (1998).
27. Matthews, I. M., Evans, J. P. & Magurran, A. E. Male display rate reveals ejaculate
characteristics in the Trinidadian guppy Poecilia reticulata. Proc. R. Soc. Lond. B
264, 695-700 (1997).
28. Marshall, T. C., Slate, J., Kruuk, L. E. B. & Pemberton, J. M. Statistical confidence
for likelihood-based paternity inference in natural populations. Mol. Ecol. 7, 639-655
(1998).
29. Krackow, S., Meelis, E. & Hardy, I. C. W. in Sex Ratios: Concepts and Research
Methods. (ed. Hardy, I. C. W.) 112-131 (Cambridge University Press, Cambridge,
2002).
30. Williams, D. A. Extra-binomial variation in logistic linear models Appl. Stat. 31, 144-
148 (1982).
Acknowledgements
We thank Tim Birkhead, Jennifer Kelley, Bart Kempenaers, Francis Neat, Tom Pizzari, Leigh
Simmons and Andy Skinner for comments on an earlier draft of the manuscript, Chiara
Romualdi for statistical advice and Anna Ludlow for assistance with the preliminary artificial
insemination trials. This research was supported by a Marie Curie Independent Research
Fellowship from the EU and was carried out in conformity with the relevant Italian laws
governing the care of animals in research.
Competing interests statement
The authors declare that they have no competing financial interests.
Correspondence and requests for materials should be addressed to J.P.E.
7
Table 1 Proportion of offspring sired in relation to male phenotype
Generalized linear model with binomial errors*
d.f.
deviance
mean
deviance
ratio
P
Regression
2
7.84
3.918
3.92
0.02
Residual
24
29.23
1.218
Total
26
37.07
1.426
Parameters included in the model
estimate
s.e.
t
P
Constant
0.372
0.278
1.34
0.180
Difference in orange†
0.089
0.040
2.23
0.026
Difference in body size (TL)
–0.430
0.177
–2.43
0.015
Parameters excluded from the model
t
P
Difference in sigmoid rate
0.85
ns
Difference in black
0.67
ns
Difference in blue
1.20
ns
* Final model: response variate, offspring sired by male B; binomial totals, brood size;
fitted terms (logit transformation), difference in absolute orange coloration and difference
in body size (TL). Overdispersion was corrected using the Williams procedure30 (phi =
0.306, model II). † We obtained the same results when the three colour components
were combined using principal components analysis: two components were extracted
from the original variables (difference in colour area); the first component (PC1)
explained 62.7% of the variance and was correlated with the differences in orange (r =
0.962); the second component (PC2) explained 25.3% of the variance and was
negatively correlated with blue (r = –0.956) and positively with black (r = 0.261). Once
entered in the GLM, only the first component was a significant predictor of paternity
share (PC1, t = 2.48, P = 0.013; difference in body length, t = –2.54, P = 0.011; PC2 t =
–0.83, P = 0.41). Qualitatively similar results were obtained when percentage, rather
than absolute colour was used (analysis not shown).
8
Figure 1. The expected paternity distribution, based on the assumption of
random sperm utilization (line), is compared with the observed values of PB
(bars). The expected PB curve is the mean of 10,000 randomised distributions
(Monte Carlo simulation), where both putative sires per family were assumed
to have an equal chance of siring each individual offspring.
9
Figure 2. The relationship between the proportion of offspring sired by male B
(PB) and the differences in phenotype scores (male B trait minus male A trait)
between competing males. a, differences in orange; b, black; c, blue (area of
colour spots, mm2); d, differences in body size (TL, mm). Points correspond to
the partial residuals of the observed data with respect to the fitted curves (see
Table 1).