Potential fitness benefits from mate selection in the Atlantic cod (Gadus morhua)

Abstract

Little evidence of benefits from female mate choice has been found when males provide no parental care or resources. Yet, good genes models of sexual selection suggest that elaborated male sexual characters are reliable signals of mate quality and that the offspring of males with elaborate sexual ornaments will perform better than those of males with less elaborate ornaments. We used cod (Gadus morhua L.), an externally fertilizing species where males provide nothing but sperm, to examine the potential of optimal mate selection with respect to offspring survival. By applying in vitro fertilizations, we crossed eight females with nine males in all possible combinations and reared each of the 72 sib groups. We found that offspring survival was dependent on which female was mated with which male and that optimal mate selection has the potential to increase mean offspring survival from 31.9 to 55.6% (a 74% increase). However, the size of the male sexual ornaments and sperm quality (i.e. sperm velocity and sperm density) could not predict offspring survival. Thus, even if there may be large fitness benefits of mate selection, we might not yet have identified the male characteristics generating high offspring survival.

Introduction

Sexually selected characters in males are assumed to reveal information about individual qualities (Hamilton & Zuk, 1982; Andersson & Iwasa, 1996). Consequently, nonrandom mating and expression of female preference are commonly observed when males provide parental care or other resources for the female (Milinski & Bakker, 1990, reviewed in Andersson, 1994; Iyengar & Eisner, 1999). Further, females could potentially allocate egg numbers, egg quality or parental care according to perceived mate quality (Reynolds & Gross, 1992; Norris, 1993; Petrie & Williams, 1993; Moore, 1994; Møller, 1994; Petrie, 1994; Von Schantz et al., 1994; Hasselquist et al., 1996; Jia & Greenfield, 1997; reviewed in Sheldon et al., 1997; Alatalo et al., 1998; Hoikkala et al., 1998; Watson, 1998; Wedekind et al., 1998; Gil et al., 1999; Cunningham & Russell, 2000, 2001). However, less evidence of fitness benefits from mate choice has been found in species were such potential confounding factors are eliminated (Andersson, 1994; Sheldon, 2000). Recent studies that experimentally control for nongenetic benefits provide empirical evidence for a genetic basis of offspring viability (Welch et al., 1998; Wedekind et al., 2001; Welch, 2003). Moreover, variation in male characters known to be preferred by females in general explains only 1.5% of the variance in offspring viability (Møller & Alatalo, 1999). However, in the alpine whitefish (Coregonus sp.) where males provide nothing but sperm, variation in the expression of male sexual ornamentation could explain up to 32% of the variance in offspring survival (Wedekind et al., 2001). Additionally, average offspring survival differed with more than 10% depending on which male the females were crossed with (Wedekind et al., 2001). Mate choice may therefore have a large effect on reproductive success, even when no parental care is provided.

An alternative hypothesis to mate choice for superior genes is mate choice for genetic compatibility (Tregenza & Wedell, 2000; Zeh & Zeh, 2001) and for example polyandry has been suggested to enhance female reproductive success by reducing genetic incompatibility in offspring (reviewed in Jennions & Petrie, 2000 and Tregenza & Wedell, 2000). Offspring viability would in this scenario be related to particular combinations of genotypes and consequently the most suitable male for one female may not be the best for another. Two recent studies have shown that genes within the major histocompatibility complex (MHC) might influence mate compatibility. Atlantic salmon (Salmo salar) choose their mates in order to increase offspring MHC heterozygosity and thereby provide them with better defense against parasites and pathogens (Landry et al., 2001). Moreover, relationships through mate choice have been found in female sticklebacks (Gasterosteus aculeatus), which try to achieve an optimum number of MHC alleles for their offspring (Aeschlimann et al., 2003).

The Atlantic cod (Gadus morhua L.) aggregate at spawning areas in spring and their mating behaviour resembles that of other lekking species (Nordeide & Folstad, 2000). In the last decades, dramatic population declines of cod of up to 99% have been reported (Hutchings, 2003 in Rowe & Hutchings, 2003). As the cod is most heavily exploited during the annual mating aggregations, male skewed catches, disturbance of male hierarchies and selective removal of dominant males might negatively affect the availability of high quality males for females (Morgan & Trippel, 1996; Nordeide, 1998; Rowe & Hutchings, 2003). This may in turn disturb or hinder mate choice and reduce female reproductive success, resulting in reduced recruitment and lowered population size (Salvanes & Balino, 1998; Reynolds & Jennings, 2000; Rowe & Hutchings, 2003). However, effects because of mate choice disruptions are hard to quantify and have so far not been incorporated in population models (Rowe & Hutchings, 2003).

Cod spawn repeatedly and each female produces millions of small eggs that are externally fertilized; no parental care is provided (Kjesbu, 1989; Kjesbu et al., 1991, 1996). Although the natural mating behaviour of cod is not well known, male–male competition, male display involving both auditory and visual signals, and female mate selection are observed in experimental studies (Brawn, 1961; Hutchings et al., 1999; Bekkevold et al., 2002). The males’ acoustic signals, used during display behaviour, are produced by contractions of paired, striated drumming muscles surrounding two to four external lobes of the swimming bladder wall (Brawn, 1961; Hawkins, 1993; own observations). Males have larger drumming muscles than females and the sound they produce is suggested to be important in mate assessment (Engen & Folstad, 1999; Hutchings et al., 1999; Nordeide & Kjellsby, 1999). Drumming muscle size is also related to male fertilization potential, measured as sperm density (Engen & Folstad, 1999). Before engaging in spawning, males display their fin size by lowering and erecting both the dorsal and ventral fins during a circling courtship dance (Brawn, 1961). When milt and eggs are released, males swim up-side down beneath the females in a ‘ventral mount’ position. Here the male has to match the females’ swimming speed while grasping her with his pelvic fins (Brawn, 1961; Hutchings et al., 1999; Rakitin et al., 2001). Thus, the size of fins and fin muscles may be of importance to ensure that the urogenital openings of both fish are aligned opposite one another during the simultaneous release of milt and eggs. Consequently, the courtship behaviour may be indicative of fertilization potential and thus be of importance for males’ reproductive success (Hutchings et al., 1999).

Spawning pairs are often joined by several satellite males that shed their sperm among the newly released eggs (Brawn, 1961) and, when mating in tanks, eggs from a single bout are often fertilized by more than one male (Hutchings et al., 1999; Rakitin et al., 2001; Bekkevold et al., 2002). Although sperm competition is only documented in captivity, large ejaculate size (Stockley et al., 1997) and the up to 1 h longevity of both sperm and unfertilized eggs (Trippel & Morgan, 1994) suggest that sperm competition also occurs under natural conditions. As sperm density and sperm velocity could predict male reproductive success (Froman & Feltmann, 1998; Birkhead et al., 1999; Rakitin et al., 1999a; Gage et al., 2002, 2004), these ejaculate characteristics are likely to be under strong sexual selection.

In our experiment, we fertilized eggs in vitro and females could thus not bias reproductive investment in response to the quality of the males. By cross-fertilizing all possible parental combination we were able to examine the potential benefits of mate selection and to relate male sex traits (i.e. both primary and secondary sex traits) to offspring survival. According to the good gene hypotheses, we predicted that males with large secondary sexual traits (i.e. drumming muscle mass, fin size and fin muscles mass) and high milt quality (i.e. high sperm velocity and high sperm density) should be males of high quality and consequently sire offspring with enhanced survival.

Materials and methods

Parental fish and gamete collection

On 3 April 2003, sexually mature adult Atlantic cod were randomly selected from trawl captures at Henningsværstrømmen, (68°11′N, 14°7′E), Northern Norway. The fish were kept in outdoor tanks with flowing seawater until handling. Ten ripe males and 10 ripe females were initially selected for the breeding experiment. We collected their gametes in separate plastic beakers by gently pressing the abdomen after drying the fish surface to avoid contaminating the samples. On the basis of their otoliths and synaptophysin (Syp I) locus, as respectively described by Rollefsen (1934) and Fevolden & Pogson (1997), nine of the males and eight of the females were identified as belonging to the north-east Arctic population of Atlantic cod. The three remaining parental fish belonged to the coastal population and their offspring were not included in later analyses. The gametes were stored at approximately seawater temperature until video recordings of sperm motility and fertilization. To be able to estimate the effect of gamete storing on offspring survival, time from stripping until fertilization was recorded. The effect of storage time on survival of both eggs (334–573 min) and milt (78–400 min) was evaluated with a multiple regression (type I sums of square). Storage time of eggs was found to have a significant effect on offspring survival (F_1,141= 28.83, P < 0.001). Storage time of milt on the contrary, did not have any significant effect on offspring survival, neither alone (F_1,141 = 0.01, n.s.) nor after removing the effect of egg storage time (F_1,141 = 0.25, n.s.). Storage time of eggs was therefore included as a covariate in the later analysis.

Fertilization and breeding

As fertilization success depends on sperm density (Rakitin et al., 1999a), milt volume used in fertilization was adjusted to sperm density so that approximately the same number of sperm cells was used in all 72 parental combinations. The milt samples used for fertilization varied from 0.37 to 0.86 mL (volume adjusted for sperm density) and were diluted in 50 mL seawater immediately before each fertilization. This sperm : seawater dilution should, according to Trippel & Neilson (1992) give more than 95% fertilization success. For each sib group, the sperm : seawater solutions were poured over 2 mL of eggs in ovarian fluid, and the fertilized eggs were immediately rinsed with seawater to avoid polyspermy (Ginzburg, 1972; Styan & Butler, 2000; Yund, 2000; Franke et al., 2002). The freshly fertilized eggs were then transported to the hatchery in bottles filled with unfiltered seawater at seawater temperature (approximately 6 °C). In the hatchery each sib group was divided over two beakers that were positioned in different tanks. Each beaker was randomly positioned in each tank, giving two replicates of each sib group (mean egg number in each beaker = 244, SD = 59.53). Prior to the survival analyses, we examined if there were effects of experimental setup. There was no difference in mean egg survival between the two replicates sired in the two different tanks (paired t-test, t₇₁ = 0.103, n.s.). Further, we examined the effect of beaker position by entering the beaker position from both axis of the tank as a fixed factors, with offspring survival as the depend variable in an anova. There was no significant effect of neither the axis, nor any significant interaction (F_10,47 = 0.5980, n.s., F_10,47 = 0.5109, n.s., F_76,47 = 0.8745, n.s., respectively).

Seawater, taken from 50 m depth and maintained at approximately 6 °C, was continuously flowing through each beaker and was not rinsed or filtered. During the 19 days long hatching period, we checked the beakers at day 3, 5, 8, 10 and 12 to count and remove possible unfertilized, dead or misdeveloped eggs, eyelings or fry. Mortality of these eggs, eyeling and fry were immediately ascertained under a microscope. Survival was estimated as the number of eyelings or fry alive at day 19 over the initial number of eggs. At day 19 the larvae were still feeding on their egg yolks and mortality because of starvation had not yet occurred (Yin & Blaxter, 1986). Eyelings and fry were killed with benzocain.

Ejaculate characteristics

Sperm motility analyses were conducted by using an aliquot (<0.12 μL) of undiluted milt placed on a cooled (approximately 5 °C) microscope slide with a fixed cover glass and a 20 μm deep chamber (volume of ∼5 μL, LEJA Products BV, Nieuw-Vennep, the Netherlands). Sperm motility was induced in a one-step procedure by adding 4.5 μL seawater. Sperm activity was recorded using a Sony CCD black and white video camera (XC-ST50CE PAL, Sony, Tokyo, Japan) at 50 Hz vertical frequency, mounted on an external negative phase-contrast microscope (Olympus CH30, Olympus, Tokyo, Japan) with a 10× objective. The recordings were stored on DV tapes and later analysed using a HTM-CEROS sperm tracker (CEROS version 12; Hamilton Thorne Research, Beverly, MA, USA). Computer assisted sperm analysis (CASA) is an objective tool for examining sperm motility in fish (Kime et al., 1996, 2001; Elofsson et al., 2003). The sperm analyzer was set as follows: frame rate 50 Hz; No. of frames 25; minimum contrast 6; minimum cell size 11 pixels. The motility parameters assessed were: average path velocity (VAP), straight line velocity (VSL) and curvilinear velocity (VCL). To remove the potential effect of drift and Brownian movement, cells having VAP < 20 μm s⁻¹ and VSL < 10 μm s⁻¹ were considered as static and were excluded from the motility analysis. All recordings were analysed 20 s after activation. VSL, VAP and VCL measurements do not differ within males (repeated measurement anova: male* sperm motility (VCL;VAP;VSL), F_16,34 = 1.2184, P = 0.304). Moreover, the repeatability was highest for VSL, measured as the intraclass correlation coefficient (r = 0.38, F_8,17 = 2.748, P < 0.05, Lessells & Boag, 1987). Consequently, only mean VSL, weighted according to cell numbers, was used in the analysis. Additionally, as spermatocrit is known to be a good estimate of sperm density in cod (Rakitin et al., 1999b), a sample of 10 μL of homogenized milt was collected in a capillary tube and centrifuged for 195 s at 11 500 rpm with a Compur mini-centrifuge (Compur-Electronic Gmbh, Munich, Germany).

Biometry of the parental fish and their secondary sexual traits

In the laboratory, each fish was weighted, both gutted and ungutted, to the nearest 10 g and fork length, which is the length from the nose tip to the cleft of the caudal fin, was measured to the nearest mm. The drumming muscles were dissected and weighted to the nearest 0.01 g. Estimates of fin size were obtained by the following measurements: the length of the third fin ray in the foremost and the second foremost dorsal fins; the length of the forth ray in the third and last dorsal fin; the mean length of right and left third rays in the pelvic fins. All fin rays were measured to the nearest 0.1 mm with a digital calliper. To estimate the mass of fin muscles of the pelvic fins, the fish were simmered, so that the different fin muscles could easily be separated. Fin muscle mass was measured to the nearest 0.01 g and mean fin muscle mass from the right and left side were used.

Statistical analysis

We analysed the parental effects by using an ancova, with offspring survival as the dependent variable, male identity and female identity as random factors. Each sib specific beaker was the unit of analysis in the statistics on offspring survival. As there was a significant effect of time elapsed from stripping the eggs until fertilization, egg storage time was entered first in the model as a covariate.

We followed the method used by Wedekind et al. (2001) when calculating the potential benefits of optimal mate selection on offspring survival, i.e. the mean offspring survival from complete random mating vs. the mean survival from optimal mate choice. Mean offspring survival from random mating was calculated from 1000 random assignments. The optimal male for each female (i.e. the male giving highest offspring survival) was calculated according to the following procedure: for a randomly selected female, a given number of males (1–9) were selected randomly and only the male giving the highest offspring survival was used to calculate offspring survival for the optimal male. Mean values from 1000 assignments, each from 1 to 9 males were used. The survival measurements were adjusted for the effect of storage time on eggs and mean values from both replicates were used.

Following Wedekind et al. (2001), we examined the relationship between male sexual characters and offspring survival with a directional heterogeneity test (Rice & Gaines, 1994a, b, c). We consequently tested for an a priori expectancy of difference in offspring survival between males by combining an ancova (male as the main factor and egg storage time as covariate) with a Spearman rank order correlation. Test statistics r_sP_c become increasingly large as the data increasingly refute the null hypothesis in the direction of the alternative hypothesis (Rice & Gaines, 1994a). In the Spearman rank order correlation we used mean offspring survival adjusted for egg storage time.

Results

In total, the survival of the 34 972 eggs from the 72 different sib groups was monitored for 19 days after fertilization. We do not, like Wedekind et al. (2001), report our results from independent analysis of early (during the first 5 days) and late mortality, since early and late mortality were correlated (r = 0.56, F_1,142 = 64.78, P < 0.0001) and may not stem from different sources of mortality.

We found a significant parental interaction and a maternal effect on offspring survival, yet no significant paternal effect (Table 1). The full factorial design allows us to determine which pair of parents would be optimal with respect to offspring survival. Random mating gave a mean survival of 31.9%. Yet, when the optimal male of all nine males was selected for each female, mean survival was 55.6% (Fig. 1). Thus the most successful pairings, when selecting the best of nine males vs. random mating, resulted in a 74.3% increase of mean offspring survival (Fig. 1).

Table 1.  Type I sums of square and test statistics from an ancova with parental (random factors) effects on offspring survival. Egg storage time is used as covariate.

	SS	d.f.	F	P-value
Egg storage time	1.59	1	14.29	0.002
Female	3.75	7	11.19	<0.0001
Male	0.37	8	0.96	0.48
Female × male	2.64	55	3.35	<0.0001
Error	1.03	72

Figure 1. Mean offspring survival (y-axis) when the optimal male is selected for each female from 1 to 9 males (x-axis). Filled circles give the means of 1000 random assignments each. The hatched line gives the mean outcome of random mating, i.e. the mean observed survival rates randomly selected 1000 times from all the 72 sib combinations.

Males’ drumming muscle mass did not predict offspring survival, nor did the relative drumming muscle size adjusted for gutted weight (directed ancova, F_8,134 = 1.16, r_sP_c = −0.158, n.s., F_8,134 = 1.16, r_sP_c = 0.214, n.s., respectively). As there was a significant correlation between pelvic fin size and fish fork length (r_s = 0.93, n = 9, P < 0.001) we analysed the effect of relative pelvic fin size on survival. However, neither relative pelvic fin size, nor relative pelvic fin muscle mass (adjusted for gutted fish weight) was related to offspring survival (directed ancova, F_8,134 = 1.16, r_sP_c = −0.18, n.s., F_8,134 = 1.16, r_sP_c = −0.044, n.s., respectively). In order to examine the correlation between the sizes of the three dorsal fins and offspring survival, the three fin measurements were entered into a principal component analysis. PC1 explained 95.8% (eigenvalue = 2.87) and was not correlated to fish fork length (r_s = 0.095, n = 9, P = 0.823). PC1 did not predict offspring survival (directed ancova, F_8,14 = 1.16, r_sP_c = −0.177, n.s.). Finally, none of the milt characters examined predicted the offspring survival (directed ancova, F_8,134 = 1.16, r_sP_c = 0.079, n.s., F_8,134 = 1.16, r_sP_c = 0.056, n.s., sperm velocity (VSL) and spermatocrit respectively).

Discussion

Under our experimental conditions, eggs being fertilized by the optimal mate had a large effect on offspring survival compared with random mating, and we found a significant parental interaction. There was also a significant effect of females, but no main effect of males on offspring survival. Moreover, our results failed to support the theory that the expression of secondary sexual traits in males, or their milt quality, could predict offspring survival.

The significant main effect of females, which is also the strongest effect in our study, is likely due to variation in egg quality between females (Chambers & Waiwood, 1996; Nissling et al., 1998; Trippel, 1998). Variation in egg quality may have several proximate explanations as resources invested in egg production may differ between females. For example, lectin activity in cod eggs may play an important role in nonspecific pathogen defence at an early life stage (reviewed in Hansen & Olafsen, 1999). As we did not attempt to remove pathogens from the hatching environment, variation in lectin activity or other immune active substances in the eggs may have influenced the different survival of egg and larvae observed between females. Moreover, cod may spawn 17–19 batches of eggs during a single spawning season and dry weight of eggs from the same female has been found to decrease by 20–30% from early to late batches (Kjesbu, 1989), suggesting that differences in reproductive timing between females may also have influenced egg survival.

The potential effect of female mate selection is large in the present study and the effect sizes exceed other available data that suggest an increase in mean fitness between 1 and 10% (Burt, 1995). However, as with the majority of studies, our measurement of offspring survival is before adulthood and may consequently not reflect fitness (Andersson, 1994). Yet, strong selection on survival is found at an early life stage in cod, as the number of hatching larvae is estimated to be only 10% of the number of eggs spawned in wild populations (Sundby et al., 1989). With an increase in offspring survival of up to 74% when selecting the optimal male, the study suggests a large effect of mate selection in natural populations of cod.

A significant interaction and lack of main male effect suggest that no single male is best for all females, but rather that there is individual compatibility resulting in enhanced offspring survival. Consequently, in order to maximize offspring survival, females should assess male qualities relative to their own individual qualities. For example, genetic compatibility at the MHC loci has been suggested to be important for mate complementarity (Penn & Potts, 1999) also in externally fertilizing species (Bakker & Zbinden, 2001; Landry et al., 2001; Reusch et al., 2001). As MHC-genes are important for immune recognition, they may also have been important for egg survival during our experimental hatching and caused the observed interaction effect.

There is ample evidence for skewness in male reproductive success and nonrandom mate choice in cod (Hutchings et al., 1999; Skjæraasen, 2003). However, offspring viability was not related to the expression of visual and sound producing male traits likely to be important for female mate choice. In whitefish variation in male ornamentation explained up to 32% of the variance in offspring viability (Wedekind et al., 2001). The complex mating behaviour of cod may suggest that multiple cues are involved in mating preferences. Multiple displays may evolve due to multiple female preferences (Brooks, 2002) and the lack of correlation between male trait expression and offspring survival could result from our inability to identify the traits involved. Additionally, the parental fish in our study were randomly selected from large trawl catches, and consequently we may have sampled individuals from different spawning shoals or leks (Rose, 1993; Nordeide, 1998; Nordeide & Folstad, 2000). Consequently, this may have rendered our males less comparable and thus influenced our results.

In general, milt quality could indicate the quality of males, that is, predict the males’ fertilization ability (Wishart, 1984; Froman et al., 1999; Al-Qarawi et al., 2002). However, in our study neither sperm density nor sperm velocity did predict offspring survival. No studies have yet examined the relationship between mating role and ejaculate characteristics in cod. However, a negative correlation has been found between drumming muscle mass and sperm density, suggesting that dominant males spawn more frequently or mature less sperm, and consequently have lower spermatocrit (Engen & Folstad, 1999). Additionally, in fish, ovarian fluid is known to influence sperm motility (Litvak & Trippel, 1998; Turner & Montgomerie, 2002; Elofsson et al., 2003) and may enhance sperm motility from certain males over others (D. Urbach, I. Folstad & G. Rudolfsen, unpublished). Thus, our sperm velocity measurements may not represent sperm velocity under conditions where natural gamete interaction occurs. Consequently, when examining sperm velocity, maternal factors and maternal–paternal interactions could be essential (Litvak & Trippel, 1998).

Variation in mating preferences, polyandry and avoidance of genetic incompatibility as a mechanism in mate choice have received much attention (Zeh & Zeh, 1996, 1997, 2001, 2003; Tregenza & Wedell, 1998, 2000, 2002; Newcomer et al., 1999; Widemo & Sæther, 1999). Our study shows that appropriate mate selection, even when nothing but sperm is provided, could have larger fitness consequences than previously expected. Large variation in fitness caused by interaction between males and females can complicate or even undermine directional sexual selection on males (Zeh & Zeh, 2003). This may partly explain why male traits, in general, only account for 1.5% of the variance in offspring viability (Møller & Alatalo, 1999) and why there is no main male effect in our result. Despite ample evidence of female preference for males with elaborated secondary traits, gene compatibility may be a substantial predictor of fitness and our study demonstrates that knowledge of fitness consequences because of parental interaction may have far-reaching implications for understanding the evolution of mate choice, as well as for breeding programmes, aquaculture, management and conservation. Finally, our study supports the concern that disturbance of mate choice could be important at the population level in cod, as offspring survival is dependent on which mate is selected (Morgan & Trippel, 1996; Salvanes & Balino, 1998; Marteinsdottir et al., 2000; Rowe & Hutchings, 2003). This may have implications for the commercial exploitation of cod.

Acknowledgments

We would like to thank captain Karl Angelsen and crew of R/V Oscar Sund for catching the fish, T.E. Bjørnsen J.L., Finstad and D. Urbach for assistance, Institute of Marine Research, Norway for analysing the otoliths, S.E. Fevolden for synaptophysin locus analyses and S. Kaino for assistance measuring the secondary sexual characters. We would also like to thank R. Rødven and A. Stien for their unique help with the statistics, S. Liljedal, D. Urbach and two anonymous referees for constructive comments on the manuscript. The study was supported by Fiskeriforskningsfondet and Bodø Regional University.