Strong inbreeding depression in male mating behaviour in a poeciliid fish


Outi Ala-Honkola, Department of Biological and Environmental Sciences, PO Box 65, University of Helsinki, Helsinki 00014, Finland.
Tel.: +358 400 815764; fax: +358 9 191 57694; e-mail:


The magnitude of inbreeding depression is often larger in traits closely related to fitness, such as survival and fecundity, compared to morphological traits. Reproductive behaviour is also closely associated with fitness, and therefore expected to show strong inbreeding depression. Despite this, little is known about how reproductive behaviour is affected by inbreeding. Here we show that one generation of full-sib mating results in a decrease in male reproductive performance in the least killifish (Heterandria formosa). Inbred males performed less gonopodial thrusts and thrust attempts than outbred males (δ = 0.38). We show that this behaviour is closely linked with fitness as gonopodial performance correlates with paternity success. Other traits that show inbreeding depression are offspring viability (δ = 0.06) and maturation time of males (δ = 0.19) and females (δ = 0.14). Outbred matings produced a female biased sex ratio whereas inbred matings produced an even sex ratio.


Mating between close relatives often leads to a decrease in fitness known as inbreeding depression. The magnitude of inbreeding depression varies among species, populations, environments and traits (Keller & Waller, 2002). It has been suggested that inbreeding depression may be more severe in stressful environments (Roff, 1997), but this is not a general rule (Armbruster & Reed, 2005). Inbreeding depression can even be strong enough to drive small populations to extinction (Saccheri et al., 1998).

Inbreeding depression is widely accepted to be a consequence of dominance and therefore it can not occur in characters with a purely additive genetic basis (Lynch & Walsh, 1998). There are two rival hypotheses, which aim to explain inbreeding depression (Charlesworth & Charlesworth, 1987). According to the overdominance hypothesis, inbreeding depression is due to heterozygote advantage. Inbreeding increases homozygosity and therefore decreases the frequency of heterozygotes that are assumed to be superior to homozygotes. This leads to a decline in fitness. Alternatively, the partial dominance hypothesis proposes that it is the unmasking of deleterious recessive alleles that leads to a fitness decline.

The magnitude of inbreeding depression is often larger in traits closely related to fitness, such as survival and fecundity, compared with morphological traits (Falconer, 1989; DeRose & Roff, 1999). This is because the relative proportion of dominance variance is expected to be greater in such traits, as additive genetic variance has been decreased by directional selection (Mousseau & Roff, 1987; Roff, 1997). Reproductive behaviour or traits under sexual selection are closely related to fitness and therefore should be sensitive to inbreeding depression. This is what Sheridan & Pomiankowski (1997) found in Poecilia reticulata (the guppy): there was inbreeding depression in the area of carotenoid pigment displayed by males, a character important in female choice, but morphological characters showed no inbreeding depression. In Drosophila montana, male courtship song frequency, which is closely associated with male courtship success and offspring survival, showed strong inbreeding depression, whereas other song characters unimportant to female choice showed much less of it (Aspi, 2000). In Melospiza melodia (song sparrow), male song repertoire size is under sexual selection (Reid et al., 2004), and a male’s repertoire size declines with his inbreeding level (Reid et al., 2005). Only a few other studies have looked at the effects of inbreeding on reproductive behaviour. In Drosophila (Sharp, 1984; Miller et al., 1993), Musca domestica (the housefly) (Meffert & Bryant, 1991) and the Bicyclus anynana butterfly (Joron & Brakefield, 2003) inbreeding decreased male mating success. In the guppy, both van Oosterhout et al. (2003) and Mariette et al. (2006) found that inbreeding decreased male mating behaviour. More studies are needed to be able to generalize the findings that reproductive behaviour is susceptible to strong inbreeding depression and in particular, it is important to verify that inbreeding depression in reproductive behaviour also leads to impaired reproductive success of inbred individuals.

In this study, we measured inbreeding depression in a number of reproductive traits after one generation of full-sib mating in the least killifish, Heterandria formosa Agassiz. This H. formosa is a small live-bearing poeciliid fish found in a variety of habitats in the coastal plain of the southeastern US (Martin, 1980). Fertilization is internal and females are matrotrophic, meaning that they provide embryos with resources for development through a placenta-like structure (Fraser & Renton, 1940; Scrimshaw, 1944; Grove & Wourms, 1991), resembling mammalian offspring provisioning. Because females superfetate (simultaneously carry broods of embryos at different stages of development, Fraser & Renton, 1940), they give birth to several young every few days. Heterandria formosa has a resource free mating system and the reproductive behaviour of males consists mainly of forced matings, so called gonopodial thrusts (Farr, 1989; Bisazza & Pilastro, 1997). During reproduction, males provide sperm only (Farr, 1989). We expected that inbred males would perform less gonopodial thrusts, spend less time following the females and have lower paternity success as compared with outbred males. We also measured inbreeding depression in the following life history traits: gestation time, number of offspring born, offspring viability, offspring size at birth and at maturity, offspring age at maturity and sex ratio.

Materials and methods

The experimental fish were aquarium born offspring of fish collected from two locations, the Saint Johns River system and the Otter Creek river in Florida (USA). The original sizes of the laboratory populations were about 600 fish for the Saint Johns population and 800 fish for the Otter Creek population. Families for this experiment were created by allowing randomly selected pairs of virgin females and stock tank males to mate. Their offspring were raised as family groups in 6-L plastic tanks (Geo Large, Ferplast SpA, Italy). When the offspring started to mature, they were separated into different aquaria according to sex. At maturity, females develop a black spot on their anal fin and around their gonopore (Fraser & Renton, 1940). In males, the anal fin starts to elongate to form a gonopodium. Virgin females of a family were kept together in one aquarium. All the males of a family were housed together with two unrelated females in one aquarium. This ensured that the males were sexually active and producing sperm when used in the experiment. In a related species, the guppy, the presence of females increases sperm production of males (Bozynski & Liley, 2003). The experiments were conducted between May 2006 and January 2007. The fish were maintained at 14 : 10 light : dark photoperiod at 28 °C and fed ad lib twice a day with frozen Artemia in the morning and frozen mosquito larvae in the afternoon.

Experiment 1: comparison of life history traits of inbred and outbred offspring

From each family, we selected two virgin females approximately of the same size and two males that showed sexual activity in their storage tank. Females were then randomly allocated to either the inbreeding treatment, where they mated with their brother (randomly selected from the two males chosen from that family), or the outbreeding treatment, where they mated with an unrelated male from another randomly chosen family. Female/male pairs were left in their tanks to breed. The fish were housed in plastic tanks (6 L) with a plastic plant that served as cover for the newborn offspring. Males were removed from the tanks after 60 days. Dead and live offspring were collected from the tanks every Monday, Wednesday and Friday for 90 days. The dead-born offspring were in most cases embryos that were not fully developed (see Schrader & Travis, 2008). The first eight live-born offspring of each female were photographed and placed individually in plastic containers (0.5 L) and raised to maturity in three Sanyo (Tokyo, Japan) MLR-350 growth chambers. The temperature of the chambers was set to 28 °C and the lighting was on for 12 h per day. The position of each container was changed every Monday, Wednesday and Friday to avoid any position effect in the growth chamber. The containers were also changed between growth chambers. Each offspring was checked for maturity once a week, on the same weekday as it was collected (Monday, Wednesday or Friday). These fish were fed twice a day with small live nematodes (Panagrellus sp.), so called ‘microworms’, grown on oatmeal, until they were 3 weeks old. Feeding was done by diluting 0.25 g of microworm mass with 100 mL of water. One teaspoon of well-mixed solution was given to each fish. When the fish reached the age of 3 weeks, they were fed with Tetra Min baby flakes in the mornings (0.20 g of flakes mixed with 25 mL of water, one drop of well-mixed solution from a plastic 1.5-mL pipette to each fish) and microworms (as before) in the afternoon. Females were considered to be mature when they had developed a black spot around their gonopore and on their anal fin (Fraser & Renton, 1940) and males were considered to be mature when the hook in the tip of the gonopodium was developed (Constantz, 1989).

For logistical reasons, we had to run the experiment in two sets, the first one starting in mid-May and the other one in mid-July. In the first set, we had 10 families from the Otter Creek population and eight families from the Saint Johns population. In the second set, we had six families from both populations. Only the families in which both females produced offspring in the same set were included in the analyses. Some families were present in both sets to ensure we collected enough data. There were two families in which both females produced offspring in both sets. Only the results from one set (chosen randomly) were included in the analysis. Thus, in our final data set, we have five families from the Otter creek population in set one and four in set two. From the Saint Johns population, we have eight families in set one and two in set two. Thus, our dataset is not orthogonal. When analysing the data, only those families that had measures of a particular trait in question in both treatments were used because family was a blocking factor in our experiment. This decreased our sample size because of missing values, as some families did not produce both female and male offspring in both treatments.

For size measurements (standard length, SL), the fish were photographed with a Canon (Tokyo, Japan) EOS 10D digital camera fitted with a 28- to 135-mm lens. SL was measured from these photographs using software for image analysis (sigma scan pro 5, SPSS Inc., Chicago, IL, USA).

Data analysis of experiment 1

We used general linear models, generalized least square models and general linear mixed models to identify factors that explained significant proportions of variation in our dependent variables. Proportional data was analysed with generalized linear models (GLM) with binomial errors and a logit link function with sample sizes as weights. The dependent variables were gestation time, the number of offspring born in 90 days, the proportion of live offspring born, offspring size at birth, offspring age and size at maturity, sex ratio and the proportion of offspring that survived until they could be sexed (of the maximum eight offspring per female that were raised until maturity). We analysed the age and size at maturity separately for males and females as the sexes differ in these traits.

We entered the following fixed factors into a full model: population, set, treatment, population by treatment interaction and set by treatment interaction. Sex was added to the model for offspring size at birth and thus only those offspring that were sexed were included in the analysis. The SL of the mother (measured at the beginning of the experiment) was correlated with the number of offspring born. Therefore, we used it as a covariate in the analysis of the number of offspring born. Mother (or mother nested within family, if it improved the model based on likelihood ratio test) was treated as a random factor in the analysis of offspring size at birth and male and female age and size at maturity because the number of offspring per mother differed from one to eight. Family was treated as a random factor in the analyses of gestation time and the number of offspring born in 90 days. The significance of the random term was assessed using likelihood ratio tests [restricted maximum likelihood (REML) estimation]. To be conservative, we kept the random term ‘mother’ in the analysis of female age at maturity even though it was not significant. This was because the number of offspring varied between one to several per mother. The two sets differed in the variance in gestation time and male maturation time. Having ‘set’ as a variance covariate (varIdent function in R) significantly improved those models based on likelihood ratio tests. We simplified the fixed part of the models using backward selection with likelihood ratio tests based on maximum likelihood estimation. The final model was refitted with REML estimation. We validated the final models by checking the homogeneity and independence of errors. For GLM’s, we used F-tests instead of the χ2-tests to correct for overdispersion. Family was always retained in these models as it is a blocking factor in the design. The statistical package used in the analysis was R 2.7.0 (R Development Core Team 2007).

To compare inbreeding depression between different traits, we calculated the standardized coefficient of inbreeding, δ, by dividing the difference in mean trait values between outbred and inbred individuals by the mean trait value of outbred individuals (Lande & Schemske, 1985). We only calculated δ when the difference between inbreeding and outbreeding treatment was significant.

Experiment 2: comparison of reproductive behaviour and paternity success of inbred and outbred males

In this experiment, we compared the reproductive performance of inbred and outbred males in terms of both behaviour and paternity success. Only fish originating from Saint Johns population were used in this experiment. The females and males that matured in the growth chambers (see above) were moved to larger tanks according to inbreeding and sex when they reached maturity. Males were again housed with two females to provide sexual stimulus. Of these fish, 15 inbred (‘ibmale’) and 15 outbred (‘obmale’) males were chosen for the paternity success experiment. We only used males that were sexually active toward their storage tank females, and we size matched male pairs as closely as possible. Mating trials were formed by two virgin outbred females and one inbred and one outbred male. Each female was first placed together with one of the males in individual tanks. After 2 days, the two males were switched between the females and another 2 days of mating was allowed. We switched the males between females four more times, such that each female spent a total of 6 days with the inbred male and another 6 days with the outbred male. Hence, a replicate lasted a total of 12 days, after which the males were removed. This switching of males was done in order to allow for sperm mixing, but to prevent any physical competition between the males. Two days is a time long enough for a mating encounter to happen and short enough to maximize sperm mixing, avoiding any clear first or second male precedence (O. Ala-Honkola, personal observation). Because of this procedure, the females in a pair experienced a different mating order. While one female was mated to the outbred male first and to the inbred male second, the ‘obib’ treatment, the other female mated first with the inbred male followed by the outbred male, the ‘ibob’ treatment.

During a mating trial, six 10-min behavioural observations were carried out on each female–male pair to assess potential differences in mating behaviour between inbred and outbred males. During the 2 days each pair was together, the observations were performed twice, one right after the pair was formed (initial observation) and the other just before the males were switched between females (final observation). Thus, each male was observed for a total of 120 min. During observations, the number of gonopodial thrusts (male approaches the female from behind and thrusts his gonopodium toward the female’s genital pore) and thrust attempts (an initiation of a gonopodial thrust that does not reach the female) was counted. The sum of these was used as a measure of gonopodial activity (GPA), as in practice it is very difficult to judge whether the gonopodium of a male actually touches the female gonopore or not. We also recorded the time a male spent following the female (following time), which is when the male follows the female as she moves around the tank or, if the pair is stationary, when the male orients himself toward the female. After the mating trials, the females were left in their tanks until they had given birth to 15 offspring. These offspring were collected and preserved in 70% ethanol for paternity tests.

Paternity tests

All DNA analyses were performed in the MES-laboratory of the Department of Biological and Environmental Sciences at the University of Helsinki between January and March 2007. To determine paternity, DNA was extracted from the caudal fin tissue of the adults and half of the body of the offspring of each family. The tissue was digested in 220 μL of salt extraction buffer (Aljanabi & Martinez, 1997) and 20 μL of 20% sodium dodecyl sulfate by 9 μL of proteinase K (30 U mg−1). Samples were vortexed and then incubated in 60 °C over night. Genomic DNA was extracted using a protocol modified from Elphinstone et al. (2003). On a 96-well microtitre filter tray 10 μL of glass beads (1 : 1 silica beads : milliQ water) and 140 μL of binding buffer were added to 50 μL of sample. After vacuuming, the samples were washed twice with 200 μL of wash solution. Finally, DNA was eluted into a 96-well tray with 60 μL of 60 °C milliQ-water.

For the polymerase chain reaction (PCR), a Qiagen (Germantown, MD, USA) Multiplex PCR Kit was used. Each 10 μL PCR reaction mixture contained 5 μL of Qiagen Mastermix 2×, 1 μL Q-solution, 1.4 μL RNase-free water, 1.6 μL of primer mix (i.e. a mixture of 0.2 μL of 10 mm primers of the four loci, both forward and reverse chains) and 1 μL of DNA template. PCR amplifications were carried out on a Peltier Thermal Cycler (PTC-100; MJ Research, Inc., Waltham, MA, USA). Thermal cycles consisted of an initial activation step of 15 min at 95 °C followed by 28 cycles: at 94 °C for 30 s, 55 °C for 90 s and 72 °C for 60 s; and a final elongation step of 20 min at 60 °C. The final step was 2 min at 20 °C. PCR products were diluted (1 : 75 μL, PCR product : milliQ water) and then genotyped using a MegaBACE 1000 Genotyping System (GE Healthcare, Chalfont St. Giles, UK). The injection mixture consisted of 2 μL of the diluted sample, 0.25 μL of the MegaBACE ET400-ROX (GE Healthcare, Chalfont St. Giles, UK) and 12.75 μL of milliQ water.

Four loci were used to assess the paternity of the offspring. The four pairs of primers used were TSS005–TSS006 (Seckinger et al., 2002; Soucy & Travis, 2003), TSS013–TSS014 and TSS051–TSS052 (Nakamura, 2001; Soucy & Travis, 2003) and SLS045–SLS046 (Nakamura, 2001, forward: 5′-TG GAA ATT GTA AAT CTG TGT TC-3′, reverse: 5′-CC GGG AAC TTC ATT GTC AGT-3′). The first two forward primers were labelled with fluorescent HEX label and the last two with 6-carboxyfluorescein (6-FAM). A GTTT-tail was added to the 5′-end of the reverse primers. Fragments were assayed using the software MegaBACE Genetic Profiler Analyzer (version 2.2, GE Healthcare, Chalfont St. Giles, UK). The allele sizes of offspring were compared with those of the adults, manually assigning each offspring to a sire according to allele sharing between the putative fathers and the mother.

Analysis of male reproductive behaviour and paternity success

To consider a possible time trend in male behaviour, we used a repeated measures anova. For this, the first four observations were summed together (2 days with the first female and 2 days with the second female) to form the first data point, the next four observations formed the second data point and the last four observations formed the third data point. Male pair and male type were used as factors and the dependent variables (GPA and following time) were square root transformed. These analyses were performed with systat 9 (SPSS Inc., Chicago, IL, USA).

The error distribution of paternity success of outbred males could not be fitted to any common error distribution. Therefore, we used a nonparametric Wilcoxon’s paired sample test to test whether the proportion of offspring sired by the outbred male (POB) differed depending on whether it was the first or the second to mate with the two females (ibob and obib treatment). There are two POB values for each outbred male as each male mated with two females (and two values for the proportion of offspring sired by the inbred male, PIB). Therefore, the paternity success of outbred and inbred males was compared after summing up the number of offspring sired by a male for both females he mated with. A paired t-test was then used to compare POB and PIB. We counted the proportional GPA for each outbred male by dividing the number of GPA the male performed by the sum of GPA performed by males in that pair. We similarly counted proportional following time for each outbred male and tested for a correlation between proportional GPA (and proportional following time) and POB using Spearman rank correlation.


Experiment 1: comparison of life history traits of inbred and outbred offspring

There was no difference in overall pregnancy success between treatments, as 25 and 27 of 30 females produced offspring in the inbreeding and in the outbreeding treatment, respectively. There was no difference between treatments in female SL at the beginning or at the end of the experiment, but female SL differed between populations, sets and families in the beginning of the experiment (final model: population effect F1,19 = 15.3, P = 0.001, set effect F1,19 = 21.0, P < 0.001, family effect F16,19 = 51.7, P < 0.001) and between sets and families at the end of the experiment (final model: set effect F1,19 = 4.5, P < 0.05, family effect F17,19 = 10.3, P < 0.001). Summary statistics for the effects of inbreeding on life history traits studied are presented in Tables 1 and 2. There was no difference in gestation time between treatments (Table 1). Also, there was no difference between treatments in the total number of offspring born during the 90 days of the experiment, but larger females produced more offspring (Table 1).

Table 1.   Final models of factors explaining variance in life history traits and mean trait values of inbred and outbred fish.
TraitSourced.f.tP-valueMean (SD)Random termVariance covariate
  1. SL, standard length.

Gestation time (days)Set362.700.01043.3 (12.5)38.8 (9.7)NoneSet
Number of offspring bornSL of the mother353.660.000822.9 (4.3)27.8 (15.4)NoneNone
SL at birth (mm)Set34−2.420.0216.08 (0.42)6.12 (0.41)MotherNone
Male development to maturity (days)Inbreeding24−2.830.00985.6 (23.7)72.1 (13.8)MotherSet
Male SL at maturity (mm)Intercept only   12.8 (0.46)13.1 (0.43)MotherNone
Female development to maturity (days)Inbreeding16−2.540.02234.4 (9.7)30.1 (7.4)Mother % in % familyNone
Female SL at maturity (mm)Inbreeding32−3.220.00310.8 (1.1)10.2 (0.62)MotherNone
Table 2.   Effects of inbreeding on traits measured as proportions.
Factord.f.DevianceResid d.f.Resid devianceFP-value
  1. Final models (generalized linear models with binomial error distribution) are presented.

Proportion of live offspring at birth
 Null  35145.02  
 Population112.4534132.5612.45< 0.001
 Family1685.841846.735.36< 0.001
 Treatment112.231734.4912.23< 0.001
Proportion of offspring surviving growth experiment until sexed
 Null  3563.07  
 Population × treatment17.121630.307.120.008
Proportion of offspring surviving growth experiment until sexed in Otter Creek population
 Null  1532.76  
Proportion of offspring surviving growth experiment until sexed in Saint Johns population
 Null  1930.28  
Sex ratio (proportion of females)
 Null  3550.57  

The proportion of live offspring at birth was lower in the inbreeding treatment than in the outbreeding treatment, δ = 0.06 (Table 2, Fig. 1). Also our study populations differed in this trait; the proportion of live offspring at birth was higher in the Otter Creek population than in the Saint Johns population (Table 2, Fig. 1). When the viability of those offspring that were raised individually until maturity was compared, we found that there was an interaction between population and treatment (Table 2). The mean proportion of offspring that survived until they could be sexed was 0.69 (SD = 0.23, N = 16) in the Otter Creek population and 0.71 (SD = 0.20, N = 20) in the Saint Johns population, but only in the Otter Creek population there was a difference in viability between treatments. In the Otter Creek population, 59% (SD = 0.23, N = 8) of the inbred offspring but 79% (SD = 0.20, N = 8) of the outbred offspring survived until they could be sexed. Our two populations did not differ from each other in any other trait studied.

Figure 1.

 Boxplots of the proportion of live offspring at birth in the outbreeding and the inbreeding treatments of the two study populations (in both Otter Creek treatments N = 8, Saint Johns N = 10).

Females in the outbreeding treatment produced a more female-biased sex ratios than females in the inbreeding treatment (Table 2, Fig. 2). The sex ratio differed from an even sex ratio in the outbreeding treatment (N females = 60, N males = 39, χ2 = 4.45, d.f. = 1, P < 0.05), but not in the inbreeding treatment (N females = 41, N males = 47, χ2 = 0.45, d.f. = 1, P = 0.52).

Figure 2.

 Boxplots of the sex ratio (proportion of females) in the outbreeding and the inbreeding treatment (N = 18 in both treatments).

Male offspring were larger at birth than female offspring [mean (mm) ± SD, males: 6.21 ± 0.38, n = 85; females: 6.01 ± 0.43, n = 100). There was no difference in the size of the newborn offspring between treatments (Table 1). The inbred males and females matured later than the outbred ones (Table 1), δ = 0.19 for males and δ = 0.14 for females. There was no difference in the size at maturity between inbred and outbred males but inbred females were larger than outbred females at maturity (Table 1). Female age and size at maturity were highly correlated (Pearson correlation for inbred females: r = 0.63, P < 0.001, n = 41, outbred females: r = 0.69, P < 0.001, n = 52). To study whether the relationship between female age and size at maturity differed between inbred and outbred females, we performed an ancova with female size at maturity as a dependent variable and female age at maturity as a covariate. Also treatment and the interaction between treatment and female age at maturity were included in the full model (Final model: female age at maturity: t58 = 9.2, P < 0.001, ‘mother’ as a random factor). There was no interaction between treatment and covariate, which suggests that inbred females were larger at maturity because it took 4 days longer for them to reach maturity.

Experiment 2: comparison of reproductive behaviour and patenity success of inbred and outbred males

In this experiment, outbred males (obmales) were larger than inbred males (ibmales) even though our aim was to size match male pairs [paired t-test, t = −2.47, d.f. = 12, P < 0.05 (two missing values), mean SL of obmales 12.9 mm ± 0.74 (mean ± SD), ibmales 12.2 mm ± 0.51]. There was, however, no correlation between male SL and male GPA or male SL and following time (Pearson correlation for ibmale SL and GPA: r = 0.05, P = 0.88, N = 14, obmales: r = −0.17, P = 0.56, N = 14, ibmale SL and following time r = 0.11, P = 0.72, N = 14, obmales: r = 0.12, P = 0.68, N = 14). Inbred males showed less GPA than outbred males (Table 3) resulting in an inbreeding depression of δ = 0.38. Inbred and outbred males also differed in their response over time (Table 3, Fig. 3a). Inbred males clearly decreased their GPA over time whereas that was not the case with outbred males. There was no difference in following time between inbred and outbred males (Table 3), but again inbred males decreased their activity over time (Table 3, Fig. 3b).

Table 3.   The effect of inbreeding (treatment) on male reproductive behaviour (repeated measures anova).
TraitSourced.f.FP-valueMean (SD) (untransformed data)
  1. GPA, gonopodial activity; ib, inbreeding; ob, outbreeding.

Sqrt [male gonopodial activity (GPA)] (number of attempts)Between subjects     16.4 (13.7)  26.4 (10.2)
Treatment (ib/ob)1, 1410.000.007  
Male pair1, 141.500.230  
Within subjects     
Time2, 2810.100.001  
Time × treatment2, 283.400.048  
Time × male pair28, 281.200.320  
Huyhn-Feldt epsilon = 1.0     
Sqrt (following time) (s)Between subjects   1079 (997)1353 (1134)
Treatment (ib/ob)1, 140.700.430  
Male pair1, 142.300.060  
Within subjects     
Time2, 2810.000.001  
Time × treatment2, 286.600.004  
Time × male pair28, 282.600.007  
Huyhn-Feldt epsilon = 1.0     
Figure 3.

 (a) The gonopodial activity (GPA) (mean ± SE) of outbred (closed bars) and inbred males (open bars) over time. N = 15 in both treatments. (b) The following behaviour (mean ± SE) of outbred (closed bars) and inbred males (open bars) over time. N = 15 in both treatments.

There were nine cases where both females in a mating trial produced offspring. Not all females produced the attempted 15 offspring, and some offspring could not be assigned to the sires because of low resolution of the microsatellites, or in case of dead-born offspring, because of tissue decay. A total of 198 offspring were assigned to the putative fathers. The average number of offspring genotyped per female was 11 (SD = 4.6, N = 18).

The paternity success of the outbred male was not affected by whether he was initially the first one or the second one to mate (ibob and obib treatments) (Wilcoxon paired-sample test Z = −0.43, P = 0.67, N = 9). The average POB male was 0.60, SD = 0.43, N = 9 and the average of inbred males (PIB) was 0.40. These proportions are not significantly different (paired t-test, t = 0.71, d.f. = 8, P = 0.5). Within pairs, it was always the male with the higher GPA value that got the larger share of paternity (binomial test, P < 0.05, N = 9). This is shown as a correlation between the proportion of the outbred male’s GPA (of total GPA of males within a pair) and POB (Spearman r = 0.78, P < 0.05, N = 9) (Fig. 4a). In eight cases of nine, the male that followed more had higher paternity success (binomial test, P < 0.05, N = 9). The correlation between the proportion of the outbred male’s following time (of total following time of males within a pair) and POB was not significant (Spearman r = 0.69, ns, N = 9) (Fig. 4b). Thus, both the GPA of a male and the activity of his competitor affected paternity success.

Figure 4.

 (a) The relationship between the proportional GPA of the outbred male (of total GPA of males within a pair) and his share of paternity. The data points on the upper right section of the graph are cases where the outbred male was more active and also fathered more than half of the offspring. The data points on the lower left section of the graph are cases where the inbred male was more active and also fathered more than half of the offspring. (b) The relationship between the proportional following time of the outbred male (of total following time of males within a pair) and his share of paternity. For interpretation see (a).


We have shown that there is strong inbreeding depression in three important life history traits in H. formosa. Inbred females gave birth to a lower proportion of live offspring (δ = 0.06) and inbred offspring took longer to mature (δ = 0.19 for males and δ = 0.14 for females). According to Stearns (1992), changes in age at maturity and in juvenile survival have large impacts on fitness, compared with other life history traits. Also male reproductive behaviour showed inbreeding depression as inbred males performed less gonopodial thrusts and thrust attempts toward females than outbred males (δ = 0.38). This trait is closely related to fitness as GPA correlated with paternity success. The values of the standardized coefficient of inbreeding, δ, measured in this study, are within the range reported by Crnokrak & Roff (1999) for wild animals. We found no inbreeding depression in gestation time, number of offspring born in 90 days, offspring size at birth and at maturity and male following behaviour.

Inbreeding effects on male reproductive behaviour

We found a clear difference in the reproductive behaviour of inbred and outbred males. Outbred males performed more gonopodial thrusts and thrust attempts and they did not decrease their activity over time as inbred males did. GPA clearly is a trait very closely related to fitness as the male with the higher GPA within a pair always (nine cases) fathered more offspring than the less active one. There was a correlation between the outbred males’ proportional GPA and paternity success. Thus, the paternity of a male depended both on his activity and the activity of his competitor. Following behaviour, on the other hand, showed no inbreeding depression and was not as closely related to fitness as GPA. In eight cases of nine, the male that followed more also fathered more offspring, but the correlation between the outbred males’ proportional following time and paternity success was not significant. This result is similar to that of Aspi (2000) where the Drosophila courtship song component most strongly related to fitness showed more inbreeding depression than other courtship song components that were not directly related to fitness. It is also in accordance with the general pattern that primary fitness characters tend to show high levels of inbreeding depression (Lynch & Walsh, 1998). In a related species, the guppy, gonopodial thrusting behaviour showed no inbreeding depression, but courtship behaviour and following behaviour did (van Oosterhout et al., 2003; Mariette et al., 2006). Mating tactics of these species differ, as in the guppy males use both courtship and gonopodial thrusting interchangeably (Houde, 1997). Mariette et al. (2006) suggested that the more costly courtship behaviour is more prone to inbreeding depression compared with the relatively cheap gonopodial thrusting in the guppy.

Many studies on poeciliid fishes assume that reproductive behaviour correctly predicts mating and paternity success (e.g. Aspbury & Basolo, 2002 in their study on H. formosa), even though there is only one other very recent study (Deaton, 2008) that has looked at the relationship between the amount of gonopodial thrusts and paternity success. Our study shows that GPA predicts paternity success irrespective of inbreeding. Our results and the results of Deaton (2008) on Gambusia affinis, suggest that GPA can be used as a proxy for paternity success in poeciliid species that use gonopodial thrusts as their most common or only mating tactic.

Our data showed no direct difference in paternity success of inbred and outbred males, but our sample size of nine is too small to draw conclusions from. As gonopodial thrusting is the most common mating tactic in this species (Bisazza & Pilastro, 1997), the more active outbred males should generally do better than inbred males.

Inbreeding and sex ratio

There appears to be no clear predictions on how inbreeding should affect sex ratios, except in the case of haplodiploid Hymenoptera with single locus complementary sex determination. Under inbreeding, these species produce a male-biased sex ratio as the number of homozygotes (diploid males) increases (van Wilgenburg et al., 2006). We found a bias toward females in outbred fish whereas fish in the inbreeding treatment produced an equal sex ratio. One explanation for sex ratio difference between treatments in our study could be that inbred females suffer more from early mortality than males do. As the proportion of live offspring was lower in the inbreeding treatment, it is possible that there were disproportionately more females than males among dead-born offspring. Our study is the first to report a difference in size between the sexes of newborn offspring in this species, and to our knowledge in poeciliids. Males were larger than females at birth.

In nature, populations of H. formosa tend to be female-biased (Leips & Travis, 1999), as is common for many poeciliids (Snelson, 1989). Leips et al. (2000) reported similar adult mortality rates for both sexes of H. formosa, but higher female recruitment rates in low densities in an experimental study on population dynamics. As the least killifish males do not seem to be under any higher predation pressure than females (Richardson et al., 2006), a female bias in nature may well be explained by uneven sex ratios at birth. Another explanation for a female biased sex ratio in nature could be that females are more likely to survive until maturation merely because their maturation time is shorter than that of males (Bell, 1980; Leips et al., 2000).

Inbreeding depression in maturation time

It took on average almost 2 weeks longer for inbred males to mature compared with outbred males and the maturation time of inbred females was 4 days longer than that of outbred females. Male maturation time shows inbreeding depression also in the guppy (Pitcher et al., 2008). As there was no difference in the size at maturity between inbred and outbred males despite the average 2 weeks difference in age, inbred males most likely grew at a slower rate than outbred males after birth. Also the fact that we could not size-match inbred and outbred males in our behavioural experiment, but the inbred males were smaller than outbred males, suggests this. Heterandria formosa males grow at a reduced rate after they reach maturity, as is common in poeciliid fishes (Snelson, 1989). At maturity, inbred females were actually larger than outbred females, but we suggest that this is because inbred females were on the average 4 days older at maturity than the outbred ones. The later the females matured, the larger they were at maturity. As the relationship between female age and size at maturity was similar between treatments (no interaction), we suggest that inbred and outbred females were growing at about the same rate.

Probability of inbreeding in the wild

Heterandria formosa often have high population densities in nature (Leips & Travis, 1999). However, they could still suffer from inbreeding as population sizes have been shown to collapse during droughts and recover slowly (Ruetz et al., 2005). As re-colonization of sites after a drought is primarily from reproduction of surviving individuals and not of migrants (Ruetz et al., 2005), mating between relatives is quite likely to occur. Inbreeding would most likely be harmful to the fitness of parents, as the survival of their offspring would be impaired. Also, their male offspring would most likely have low reproductive success for at least three reasons. First, they would have lower probability to reach maturity because of their longer development time (this also applies to female offspring). Second, their lower mating activity would likely lead to poor paternity success. Third, their smaller size compared with outbred males may make them less desirable mating partners as females of H. formosa have been shown to prefer larger males (Aspbury & Basolo, 2002). All these factors make it likely that H. formosa would suffer from inbreeding depression also in the wild, even though predicting the response to inbreeding in different environments is very difficult (Armbruster & Reed, 2005). The amount of inbreeding depression we found in this study might be strong enough to select for the evolution of inbreeding avoidance mechanisms. Theoretically, if inbreeding depression exceeds δ = 1/3, females should prefer unrelated males over brothers as their mates (Parker, 1979; Kokko & Ots, 2006). Inbreeding depression in male GPA exceeds this value (δ = 0.38) and inbreeding depression in male and female maturation times come close to it (δ = 0.19 and 0.14). It is difficult to estimate from these values what the actual fitness loss a female mating with a brother is, but it certainly comes close to the threshold where she should avoid mating with her brother.

In conclusion, this study provides some unique insights into how reproductive behaviour is susceptible to inbreeding depression. We have shown that male reproductive behaviour, namely the number of gonopodial thrusts and thrust attempts, shows very strong inbreeding depression. We also showed that this trait is very closely related to fitness as it is correlated with paternity success.


The authors thank J.E. Brommer and H. Kokko for valuable discussions, J. Blyth, J.E. Brommer, L. Sundström and ‘The Fish Group’ for comments on earlier versions of the manuscript, L. Tuominen for help with fish maintenance and Finnish Cultural Foundation (O.A.-H.), Ella and Georg Ehrnrooth’s Foundation (O. A.-H.) and the Academy of Finland (K.L.) for funding.