The Consequences of Feminization in Breeding Groups of Wild Fish

Background The feminization of nature by endocrine-disrupting chemicals (EDCs) is a key environmental issue affecting both terrestrial and aquatic wildlife. A crucial and as yet unanswered question is whether EDCs have adverse impacts on the sustainability of wildlife populations. There is widespread concern that intersex fish are reproductively compromised, with potential population-level consequences. However, to date, only in vitro sperm quality data are available in support of this hypothesis. Objective The aim of this study was to examine whether wild endocrine-disrupted fish can compete successfully in a realistic breeding scenario. Methods In two competitive breeding experiments using wild roach (Rutilus rutilus), we used DNA microsatellites to assign parentage and thus determine reproductive success of the adults. Results In both studies, the majority of intersex fish were able to breed, albeit with varying degrees of success. In the first study, where most intersex fish were only mildly feminized, body length was the only factor correlated with reproductive success. In the second study, which included a higher number of more severely intersex fish, reproductive performance was negatively correlated with severity of intersex. The intersex condition reduced reproductive performance by up to 76% for the most feminized individuals in this study, demonstrating a significant adverse effect of intersex on reproductive performance. Conclusion Feminization of male fish is likely to be an important determinant of reproductive performance in rivers where there is a high prevalence of moderately to severely feminized males.


Research
There is international concern regarding the effects of natural and man-made chemicals on the health of humans and wildlife. Estrogenic and anti androgenic chemicals [so-called endocrine-disrupting chemicals (EDCs)] are of particular concern to aquatic eco systems, because these compounds are present in almost all treated sewage effluents and in lowland rivers receiving these effluents throughout Europe, Asia, and the United States (Desbrow et al. 1998;Hinck et al. 2009;Hotchkiss et al. 2008;Sumpter and Johnson 2008;Sun et al. 2008). Steroid estrogens in particular can be biologically active at very low concentrations (in the low nanogram per liter range) and are known to cause altered sex ratios (Lange et al. 2008;Länge et al. 2001) and feminiza tion of male fish (Hahlbeck et al. 2004). Feminized pheno types include the presence of vitellogenin in the blood of male fish (Purdom et al. 1994) and the presence of developing eggs (oocytes) and/or female reproductive ducts (oviducts) in the testes of other wise male fish (the inter sex condition) (Jobling et al. 1998). Although these conditions have been widely reported in both fresh water and marine fish species (Allen et al. 1999;Bjerregaard et al. 2006;Blazer et al. 2007;De Metrio et al. 2003;Hinck et al. 2009;Jobling et al. 1998;Penaz et al. 2005;Stentiford and Feist 2005), there is no evidence that their existence directly affects numbers in wild fish populations.
Some of the most compelling evidence for inter sex in wild fish comes from studies on roach (Rutilus rutilus) inhabiting U.K. rivers contaminated with sewage effluents. Intersex roach (male fish with developing eggs in their testes) have been found at 86% (of a total of 51) of U.K. river locations (Jobling et al. 2006). Severely intersex fish produce less milt (semen), the milt has a lower sperm density, and the sperm have reduced motility, compared with apparently normal male fish from less contaminated sites (Jobling et al. 2002). Fertilization success-as meas ured through in vitro studies that have determined the proportion of eggs successfully fertilized by intersex fish and the number of these fertilized eggs capable of giving rise to live offspring-is also reduced with increasing severity of the intersex condition (Jobling et al. 2002). Data such as those described above naturally lead to concern that EDCs may have detrimental effects on fish populations. A limited number of studies have been under taken to investigate this possibility. Long-term studies in an experimental lake in north western Ontario, Canada, showed that exposure to the potent estrogen ethinyl estradiol, at 4-6 ng/L over a period of 3 years resulted in the collapse of the population of the fathead minnow (Pimephales promelas) (Kidd et al. 2007), but no adverse effects have yet been found on populations of the longer-lived fish species pearl dace (Margariscus margarita) and lake trout (Salvelinus namaycush) Werner et al. 2006). Modeling approaches, using information from experimental exposures, have predicted that concentrations of EDCs found in the environment could lead to population declines, mainly resulting from reduced female, rather than male, fecundity (Grist et al. 2003;Gutjahr-Gobell et al. 2006;Miller and Ankley 2004). Recently, another modeling study, using data obtained from wild fish (Jobling et al. 2002), predicted intersex on the whole to have a mini mal effect on the population growth rate in roach but suggested that it may increase the risk to local roach populations when present in combination with selective fishing practices .
A critical gap in the knowledge and information that have fed population models published to date is how successfully intersex fish reproduce when competing with apparently normal males in breeding populations. In the wild, roach compete in groups to fertilize the eggs of spawning females (Diamond 1985;Wedekind 1996), so it is important to know whether intersex fish might be at a disadvantage in these competitive spawning situations. In the present study, we attempted to fill this knowledge gap by allowing groups of roachrepresentative of a naturally spawning shoal containing both normal and intersex fish-to spawn naturally in large tanks and using DNA micro satellites to assign the offspring to the parents. In this way, we assessed the abilities of intersex fish to compete with one another (and with other male fish) and to contribute to the next generation.

Materials and Methods
Experimental design. We conducted two experiments, the first in 2006 (study 1) and the second in 2008 (study 2). These studies were designed specifically to examine the effect of long-term exposure of wild male fish to endocrine disruptors. The effects of these chemicals on male gonadal histology are well known, whereas there is currently little evidence that long-term EDC exposure affects female histology in the wild. For each of the two experiments, we collected adult roach using standard electro-fishing methods from wild populations living in effluent-contaminated rivers in the United Kingdom. The sites were chosen based on data from previous national surveys that indicated where roach with varying degrees of intersex were likely to occur (Jobling et al. 2006).
We collected fish in late April, shortly before the natural spawning season, in cooperation with the U.K. Environment Agency. We selected the larger adult fish from those that were caught in order to avoid the inclusion of sexually immature fish and to maximize the chances of including more severely intersex individuals in the study because prevalence and severity of the intersex condition increase with age (Jobling et al. 2006). We then transported the fish to a holding facility (the Environment Agency's fish breeding unit at Calverton), where we separated them into groups of males and females, based on their body morpho type and the presence of secondary sex charac teristics (Kortet et al. 2003). We placed the fish into large fiber glass tanks (each containing kakabans, aquatic weed-like spawning substrates) receiving water via a recirculating system, which was topped up with bore hole water, as necessary. Water temperatures were initially set to match ambient temperatures in the river where the fish were collected and were subsequently increased to 15°C over 2 days to encourage ovulation of the females.
The experimental design involved placing six males with three females to create competition between the males (and intersex fish) for the females. We first allocated males to the spawning tanks and allowed them to acclimatize. Spermiation in male roach can occur in captivity, and males can produce sperm over a period of a few weeks without the need for any intervention. For females, synchronicity of spawning was required to enable removal of the adults as soon as possible after spawning (to minimize fish eating the spawned eggs). This was achieved by injection of females with carp pituitary extract at a time just prior to natural ovulation. We then placed three females into each tank with the males to allow spawning. Subsequently, some tanks were found to contain five males and four females, because the external features used for classification of sex can be affected by exposure to EDCs contained in effluent discharges.
All animals used in this research were treated humanely and with regard for the alleviation of suffering; all procedures were subject to approval by the local ethical review process as required under the U.K. Animals (Scientific Procedures) Act (1986).
Study 1: 2006 breeding experiment. We collected fish on 24 April 2006 from a 200-m stretch immediately downstream of Chertsey sewage treatment works on the River Bourne (Surrey, UK; 51°24´08˝N; 0°32´07˝W). Seven breeding tanks were used, with a total of 63 fish, including 38 males (Table 1).
Study 2: 2008 breeding experiment. The 2008 experiment included a greater number of fish-and more severely intersex fishthan were included in 2006, and was designed to reduce the confounding effect of size on reproductive success. Fish were obtained from the River Arun (Sussex, UK), downstream of Horsham sewage treatment works (51°03´19˝N; 0°22´02˝W) on 21 April 2008. We sorted the males by size (by measuring to the nearest centimeter) before introducing them to the spawning tanks, so that each tank contained a restricted size range of male fish. Thirteen breeding tanks were used, with a total of 117 fish, including 75 males ( Table 1).
Sampling of fish and morphometric analyses. We removed adult fish from the breeding tanks within 24 hr of spawning (5 days after collection from the wild), anesthetized them with 1:10,000 benzo caine, and measured the following end points for each adult male/intersex fish: a) length; b) weight; c) age (using counts of scale annuli); d) gonad histology; e) roughness of skin [on a scale of 0 (smooth) to 3 (large, visible tubercles all over skin of fish)]; f) sperm density; and g) sperm viability (assessed using trypan blue exclusion) (Hackett and MacPherson 1965). In addition, we collected fin clips from all adult fish and preserved them in 100% ethanol for genetic profiling and parentage analysis.
Sampling of fry. The fry started to hatch approximately 1 week after spawning. We randomly sampled fry from the tanks 3-4 days post hatch and terminally anesthetized them before preserving them in absolute ethanol for DNA microsatellite analyses.
Quantification of intersex index. We assessed gonadal histology using standard histological techniques as described by Jobling et al. (2006). We assigned each male fish a numerical score (intersex index) to classify the level of gonadal disruption based on the number of oocytes present in the testes (Jobling et al. 2006). Each of six sections was scored separately, and a mean score was calculated for each fish. Here we define intersex severity using intersex index values as follows: 0, nonintersex (normal male); > 0 but < 2, mildly intersex; ≥ 2 but < 4, moderately inter sex; ≥ 4, severely intersex. Collectively, we refer to both male and intersex fish as "males." The presence of female-like reproductive ducts (ovarian cavities) (Nolan et al. 2001) was also recorded but not included in the intersex index. Ovarian cavities in males can occur as a consequence of exposure to estrogen during early life. Oocytes in the testis, however, is a progressive condition, increasing with the level of EDC exposure and age of the fish. Our approach is consistent with that previously described for intersex indices in roach (Jobling et al. 2006).
Microsatellite genotyping. We extracted DNA from the fin tissues of the adult fish, as well as from the fry, using the Chelex proto col (Estoup et al. 1996). We then genotyped all adult fish and selected fry from each tank using seven variable micro satellite loci (Hamilton and Tyler 2008). We geno typed each adult using an additional five microsatellites (CypG3, Lid1, Lid8, Lid11, Rru4) for individual genetic diversity calculations. We also genotyped four fry using these extra five genotypes, when the seven were unable to resolve parentage. We used the PROBMAX program (version 1.3; Danzmann 1997) for A single triploid adult male was identified in study 2. This fish did not reproduce, and no milt was obtained from it. Because triploid fish are sterile, this individual was excluded from analyses of reproductive success.
Individual genetic diversity. We calculated two indices of individual genetic diversity (measures of how "inbred" each individual is): homozygosity by loci (HL) (Aparicio et al. 2006) and mean standardized d 2 (Amos et al. 2001;Coulson et al. 1998), from the parental micro satellite geno types using the IRmacroN3 (Amos 2010). HL is calculated using allele frequencies, whereas d 2 is calculated using size differences of alleles for each micro satellite locus within an individual.
Statistical analysis. To deduce the influence of intersex on reproductive performance of male fish, we fitted linear mixed-effect (LME) models using the proportion of offspring sired per male as the response variable. To control for differences between tanks, we added a random tank effect to each model to capture the assumption that the reproductive performance of each fish depends on the other fish in the same tank. A weighted LME model was used to correct for the hetero skedasticity observed in the data (i.e., the greater variance in reproductive success observed at low values of intersex index), and we used an exponential variance function structure for the weights (Pinheiro and Bates 2000). We included all explanatory variables known to influence, or suspected of influencing, reproductive performance in the maxi mal model. These were intersex index, presence of ovarian cavities, length, HL, d 2 , roughness, sperm viability, and sperm density. We ran analyses both including and excluding fish for which sperm density and/or sperm viability data were unavailable (one fish from study 1 and six from study 2). Full models showing the results of the analysis including and excluding sperm parameters are presented in Supplemental Material, Tables 1 and 2, respectively (doi:10.1289/ehp.1002555). For both studies, weight and age were highly correlated with length (correlation coefficients ≥ 0.80), so these were excluded from the analyses. Thus, the full model used in the analysis was as follows: where IS is intersex index, LEN is length, R is roughness, OC is presence of ovarian cavities, D2 is d 2 , SV is sperm viability, and SD is sperm density. The parameters β i , for i = 1, . . . 8, and the tank effects were estimated using the LME function in R (version 2.8.1) (Ihaka and Gentlemen 1996). The likelihood ratio test (LRT) p-value for comparing the model with and without tank effects was > 0.99, suggesting that no tank effects are present in the data. To identify the key factors associated with reproductive performance, we obtained minimal adequate models using the stepwise backward procedure, until all selected variables had p-values < 0.1 calculated using an LRT. p-Values presented here were derived using an LRT based on model simplification of maximum-likelihood versions of the mixed models. In addition, we used permutation tests to verify the significance of terms with LRT p-values < 0.05 in the minimal adequate mixed-effects models (Table 1). Empirical distributions of the slope of the relationship between proportion paternity and each of the fixed effects in our minimal adequate models were created by shuffling randomly the paternities of fish within tanks, while maintaining the tank structure of the trials. p-Values for the permutation tests were calculated as the proportion of permutation slopes, based on 1,000 permutations, that were at least as extreme as the observed slope in the minimal adequate model.

Results and Discussion
Gonadal disruption in males. We found evidence of feminization of male fish in both the rivers Bourne (study 1) and Arun (study 2), although the prevalence and severity of the condition differed between studies [see Supplemental Material, Figure 1 (doi:10.1289/ ehp.1002555)]. Two of the 38 males (5.2%) from the River Bourne had ovarian cavities, and 15 (39%) had oocytes in their testes. Most of these intersex fish (13 of 15) were mildly feminized (had a low intersex index, with few oocytes in their testes). Only 2 were moderately affected, and we observed no severely intersex fish (high intersex index) in this study.
In contrast, almost all of the 76 males from the River Arun used in study 2 had disrupted gonads: 88% had ovarian cavities, and 41% had oocytes in their testes. Of the intersex males used in this study, 18 were mildly disrupted, 9 were moderately intersex, and 4 were severely intersex. Examples of gonadal sections from males with different intersex indices are shown in Figure 1.

Male reproductive success.
In both studies, we observed considerable variations in male reproductive success. For example, in two tanks from study 1 and three tanks from study 2, a single male (out of either five or six present) sired > 50% of the offspring. Nevertheless, in both studies, most males (95% in study 1 and 91% in study 2), including intersex fish of all severities, sired offspring, demon strating that in the scenarios represented by these experiments, most intersex fish were able to participate in spawning [see Supplemental Material, Figure 2 (doi:10.1289/ehp.1002555)]. Several factors significantly influenced male reproductive success (Table 1).

Study 1: factors influencing reproduction in a mildly intersex population.
In study 1, males of greater length were significantly more successful at siring offspring (LME model coefficient = 0.0036, p = 0.036; Table 1, Figure 2). Other studies have also found larger male fish to be more successful in competitive breeding scenarios (Fessehaye et al. 2006; Jacob et al. ; the gonad consisted of large numbers of primary oocytes, as well as some oocytes in more advanced stages and some that were degenerating (do) and/or vacuolated, all set among male tissue, most of which (in this fish) consisted of mature spermatozoa. Bars = 100 μm.

A B C
2009), and across various taxa, dominance is related to body size (Qvarnström and Forsgren 1998). Roach have a "lek-like" breeding system (Wedekind 1996) in which a number of males occupy the spawning site, and females enter the area specifically to spawn with the competing males. Larger males of the European minnow (Phoxinus phoxinus), which also has a lek-like spawning strategy, are better able to defend spawning territory and have greater reproductive success (Jacob et al. 2009); this may have been the case in our study. We observed that neither intersex index nor the presence of ovarian cavities was significantly correlated with male reproductive performance in study 1 [p-values from LME models were 0.36 and 0.17, respectively; see Figure 3A and Supplemental Material,

Study 2: factors influencing reproduction in a population containing severely intersex males.
We found a significant negative relation ship between intersex index and reproductive success in study 2 (LME model coefficient = -0.029, p < 0.0001; Table 1, Figure 3B); more severely feminized fish had reduced success. Within the limits of our model (between intersex indices 0 and 5), the intersex condition decreased the proportion of offspring sired by affected individuals in each tank by 2.9% per unit increase in intersex index, when all other variables in the model are held constant (Table 1). Thus, the intersex condition reduced the average contribution to the offspring within each tank from 19% for non intersex males (intercept for model with success and intersex only) to 4.5% for fish with an intersex index of 5. This represents a relative decrease in reproductive performance of 76%, or 15% per intersex unit, assuming a linear relationship (as shown in Figure 3B). These slopes should not be used to extrapolate beyond the group sizes used in these trials, but they none the less demon strate a significant adverse effect of intersex on reproductive performance.
The difference in the consequence of intersexuality on reproductive success between studies 1 and 2 is likely due to the increased range of intersex indices observed in the fish used in study 2: Study 1 only included 2 fish that were moderately intersex, whereas study 2 included 13 fish with an intersex index > 2. This fact, together with the increase in the total number of males used in the second study, afforded a wider and more powerful analysis of the relation ship between inter sexuality and reproductive success. Moreover, males in the second study were size sorted (reducing the mean maximum size difference within each tank from 40 mm in study 1 to 12 mm in study 2) in order to reduce the potential confounding effect of size on reproductive performance. Indeed, the small size differences that were still present in the second study did not significantly affect the reproductive hierarchy (p = 0.076; Table 1). We did, however, observe a statistically significant relationship between the presence of ovarian cavities and reproductive success (LME model coefficient = 0.045, p = 0.05; Table 1), albeit the trend was opposite to that expected, with males with ovarian cavities performing better than those without.
The observed reduction in reproductive success of intersex fish could have resulted from a combination of factors, including a reduced ability to release milt due to blocked or obstructed sperm ducts, reduced sperm quality, reduced hatching success or survival of fry, and/or reduced ability to compete with other males or attract females. Using in vitro techniques, Jobling et al. (2002) found that intersex fish have reduced sperm quality. This may have been the case in our study, because we found that sperm viability was significantly correlated with reproductive success [LME model coefficient = 0.0023, p = 0.001; see Supplemental Material, We also demonstrated a significant association between the internal genetic diversity measure HL and reproductive success in study 2. However, the trend was opposite to that expected, with the more "inbred" fish being more successful (LME model coefficient = 0.14, p = 0.019; Table 1). "Outbred" individuals of several vertebrate species have been shown to have greater reproductive success (Amos et al. 2001); however, the influence of inbreeding on reproductive success of fish is less clear. Although experimentally generated inbred male and female tilapia (Oreochromis niloticus) have been found to have reduced reproductive success (Fessehaye et al. 2009), another study found no effects of parental internal genetic diversity on the reproductive output of Atlantic salmon (Salmo salar) (Garant et al. 2005).
Analysis of the combined results from both studies. A statistically signifi cant negative relationship between severity of intersex and reproductive success was also apparent in the analysis of the combined data from the two experiments (LME model coefficient = -0.029, p = 0.0001; Table 1). Across all tanks, the percentage of intersex fish not reproducing was higher (13%) than the percentage of non intersex males not reproducing (4.4%), although this apparent consequence of intersexuality was not statistically significant at the 95% level (Fisher's exact test, p = 0.15).
In contrast to the analysis of study 2 data alone, we found no significant association between the presence of ovarian cavities and reproductive success when the data were combined [p = 0.35; see Supplemental Material,  In addition, after analysis of the combined data, we observed that fish with more prominent secondary sexual characteristics (breeding tubercles causing roughness of the body surface) were more successful in fathering offspring (LME model coefficient = 0.031, p = 0.022; Table 1). Although it is unclear whether there is any female choice in roach, it has been suggested that breeding tubercles are used by females as a cue for choosing a highquality mate, and males with large breeding tubercles exhibit significantly more active courtship behavior (Kortet et al. 2004). The presence of tubercles has also been found to be important in the reproductive success of minnows (Jacob et al. 2009).
Other significant trends observed after analysis of the separate data sets (length and sperm viability in studies 1 and 2, respectively) were also reflected in analy ses of the combined data; reproductive success was significantly positively correlated both with body length (p = 0.035; Implications and limitations. Our results, although demonstrating that intersex fish are able to reproduce in a realistic competitive breeding scenario, nonetheless indicate a significant reduction in reproductive capability of severely intersex fish, as well as an effect on parentage outcome in the popu lation as was suggested (but not proved) by our previous work using in vitro fertilization techniques (Jobling et al. 2002). The implications of the reduced reproductive capacity of a proportion of fish in a given population are largely unknown but could potentially include diminished recruitment. Likewise, effective population sizes, which are important for long-term maintenance of genetic variability, could be affected, especially in combination with other documented effects of laboratory-based EDC exposure, such as sex reversal, reduced female fecundity, and alterations in timing of reproduction. Finally, if susceptibility to EDC exposure has a genetic basis, a skew in success toward less feminized individuals provides a mechanism for the evolution of tolerance to the feminizing effects of EDCs.
It is difficult to extrapolate the data presented here directly to the wild because too much is unknown about the actual spawning scenario of wild roach. For instance, as far as we are aware, the size of breeding colonies in the wild is unknown and may vary widely. Likewise, the sex ratio at spawning sites is largely unknown but potentially could be more male biased in the wild than the 2:1 male:female ratio used here, because males congregate at spawning sites, whereas females enter only to spawn (Diamond 1985;Wedekind 1996).

Conclusion
We conclude that if severely inter sex fish are present in a river, there could be implications for the fish population concerned, whereas if only mildly intersex fish are present, it seems likely that the effects (if there are any) will be less severe. This conclusion needs to be considered in the context of available data concerning the prevalence of intersex fish in rivers. For instance, in U.K. rivers the proportion of male roach that are moderately to severely intersex (intersex index ≥ 2) is generally < 10%, and the proportion of severely intersex fish alone (intersex index ≥ 4) is < 4%, even in highly contami nated sites, based on the combined data from surveys undertaken previously (Jobling et al. 1998(Jobling et al. , 2002(Jobling et al. , 2006. However, data specifying the severity of gonadal disruption in individual fish (as opposed to the overall proportion of intersex fish in a river) are rare; those studies that have provided such detail suggest a low proportion of severely intersex fish in river systems in Europe and the United States (Bjerregaard et al. 2006;Blazer et al. 2007;Jobling et al 1998;Minier et al. 2000). Locally high prevalences of intersex have been reported by Hinck et al. (2009) for other fish species in the United States (91% for largemouth bass and 73% for smallmouth bass), although the authors did not describe the severity of intersex observed in individual fish. Therefore, currently available information would suggest that, because the majority of feminized males at effluent-impacted sites are mildly intersex, it seems unlikely that the intersex condition alone would result in short-term population crashes on the scale of those observed in previous long-term (albeit high concentration) EDC exposure studies (e.g., Kidd et al. 2007). It is nevertheless clear that further knowledge of roach populations and the application of population modeling to these collective data would be helpful in determining the long-term consequences of the intersex condition.