PATERNAL GENETIC EFFECTS ON OFFSPRING FITNESS ARE CONTEXT DEPENDENT WITHIN THE EXTRAPAIR MATING SYSTEM OF A SOCIALLY MONOGAMOUS PASSERINE

Transcription

1 Evolution, 59(3), 2005, pp PATERNAL GENETIC EFFECTS ON OFFSPRING FITNESS ARE CONTEXT DEPENDENT WITHIN THE EXTRAPAIR MATING SYSTEM OF A SOCIALLY MONOGAMOUS PASSERINE TIM SCHMOLL, 1,2 VERENA DIETRICH, 3 WOLFGANG WINKEL, 4 JÖRG T. EPPLEN, 5 FRANK SCHURR, 6 AND THOMAS LUBJUHN 1 1 Institute for Evolutionary Biology and Ecology, University of Bonn, An der Immenburg 1, D Bonn, Germany 2 tschmoll@evolution.uni-bonn.de 3 Zoological Institute, TU Braunschweig, Fasanenstrasse 3, D Braunschweig, Germany 4 Institute of Avian Research Vogelwarte Helgoland, Working Group Population Ecology, Bauernstrasse 14, D Cremlingen, Germany 5 Human Genetics, Ruhr-University Bochum, D Bochum, Germany 6 Department of Ecological Modelling, UFZ Centre for Environmental Research Leipzig-Halle, Permoserstrasse 15, D Leipzig, Germany Abstract. Avian extrapair mating systems provide an interesting model to assess the role of genetic benefits in the evolution of female multiple mating behavior, as potentially confounding nongenetic benefits of extrapair mate choice are seen to be of minor importance. Genetic benefit models of extrapair mating behavior predict that females engage in extrapair copulations with males of higher genetic quality compared to their social mates, thereby improving offspring reproductive value. The most straightforward test of such good genes models of extrapair mating implies pairwise comparisons of maternal half-siblings raised in the same environment, which permits direct assessment of paternal genetic effects on offspring traits. But genetic benefits of mate choice may be difficult to detect. Furthermore, the extent of genetic benefits (in terms of increased offspring viability or fecundity) may depend on the environmental context such that the proposed differences between extrapair offspring (EPO) and within-pair offspring (WPO) only appear under comparatively poor environmental conditions. We tested the hypothesis that genetic benefits of female extrapair mate choice are context dependent by analyzing offspring fitness-related traits in the coal tit (Parus ater) in relation to seasonal variation in environmental conditions. Paternal genetic effects on offspring fitness were context dependent, as shown by a significant interaction effect of differential paternal genetic contribution and offspring hatching date. EPO showed a higher local recruitment probability than their maternal half-siblings if born comparatively late in the season (i.e., when overall performance had significantly declined), while WPO performed better early in the season. The same general pattern of context dependence was evident when using the number of grandchildren born to a cuckolding female via her female WPO or EPO progeny as the respective fitness measure. However, we were unable to demonstrate that cuckolding females obtained a general genetic fitness benefit from extrapair fertilizations in terms of offspring viability or fecundity. Thus, another type of benefit could be responsible for maintaining female extrapair mating preferences in the study population. Our results suggest that more than a single selective pressure may have shaped the evolution of female extrapair mating behavior in socially monogamous passerines. Key words. Extrapair paternity, genotype-by-environment interaction, good genes, half-sibling comparison, mating preference, Parus ater, reproductive success. Hundreds of molecular studies across a wide range of animal taxa have clearly demonstrated that multiple paternity as a result of female multiple mating is the rule rather than an exception in natural populations, even in many socially monogamous species that establish pair bonds (Birkhead and Møller 1998; Jennions and Petrie 2000; Griffith et al. 2002). But why do females of so many species mate multiply? One hypothesis proposes that genetic (i.e., indirect) fitness benefits outweigh the direct costs of searching and performing copulations with multiple males and thus select for female multiple matings ( good genes hypothesis of multiple mating, reviewed in Jennions and Petrie 2000). This idea is difficult to test, however, whenever females also accrue nongenetic (i.e., direct) fitness benefits (like nuptial gifts or male parental care) when mating with multiple males. Within this context, avian extrapair mating systems (e.g., Petrie and Kempenaers 1998; Griffith et al. 2002) do offer an interesting model system to study and assess the role of genetic benefits in the evolution of female multiple mating behavior (and female mate choice in general), since potentially confounding nongenetic benefits of extrapair mate choice are considered 2005 The Society for the Study of Evolution. All rights reserved. Received February 6, Accepted December 11, to be of very limited importance (Jennions and Petrie 2000; Griffith et al. 2002). The costs and benefits of female extrapair mating behavior remain unclear and remarkably difficult to uncover (Griffith et al. 2002), despite substantial research effort (reviewed by Westneat et al. 1990; Birkhead and Møller 1992; Petrie and Kempenaers 1998; Jennions and Petrie 2000; Griffith et al. 2002; Westneat and Stewart 2003). The most popular hypothesis to explain the evolution and maintenance of female extrapair mating behavior suggests that females are able to enhance the genetic quality and thus the reproductive value of (part of) their offspring by engaging in extrapair copulations with males of higher genetic quality compared to their respective social mates (good genes hypothesis of extrapair mating, Westneat et al. 1990; Birkhead and Møller 1992; Griffith et al. 2002). A straightforward and crucial test of good genes models of female extrapair mating is a direct, pairwise comparison of groups of maternal half-siblings, which are raised by the same social parents under the same environmental conditions. Here, potentially confounding nongenetic common environ-

2 646 TIM SCHMOLL ET AL. ment as well as maternal effects on offspring fitness-related traits are controlled for and any systematic differences in phenotypic traits between the two maternal half-sibships can only be attributed to the differential genetic contributions of the respective genetic sires (Sheldon et al. 1997; Griffith et al. 2002). Therefore, a consistently better performance of extrapair offspring (EPO) compared to their within-pair offspring (WPO) maternal half-siblings can be seen as strong evidence in favor of good genes models of extrapair mating (Sheldon et al. 1997; Griffith et al. 2002). This is true, regardless of whether the proposed genetic benefits are assumed to be caused by absolute (good genes as viability or attractiveness genes, cf. Kokko et al. 2002) or relative (good genes as compatible genes, cf. Tregenza and Wedell 2000) genetic quality of the respective extrapair sires. A potential problem of the maternal half-sibling test design is that differential maternal investment with respect to paternity may confound paternal genetic effects on offspring fitness (Møller and Thornhill 1998; Mousseau and Fox 1998). In addition to the lack of evidence for female ability to discriminate between half-sibling offspring (Sheldon 2000a), a biased investment in favor of EPO may be seen as a kind of an indirect paternal genetic effect, since female differential investment exclusively depends on differential paternal genetic contribution (Sheldon 2000a). Maternal half-sibship comparisons have successfully been used to demonstrate paternal genetic effects on offspring fitness in amphibians (Welch et al. 1998; Doty and Welch 2001; Sheldon et al. 2003; Welch 2003) and fish (Barber et al. 2001; Wedekind et al. 2001). In birds, some of the few empirical studies supported good genes models of female extrapair mating (Kempenaers et al. 1997; Sheldon et al. 1997; Johnsen et al. 2000; Foerster et al. 2003), while others failed to do so, even in closely related or the same species (Krokene et al. 1998; Strohbach et al. 1998; Lubjuhn et al. 1999a; Schmoll et al. 2003; see also Whittingham and Dunn 2001; Kleven and Lifjeld 2004). However, since genetic fitness benefits of mate choice are generally small (Møller and Alatalo 1999), even studies with reasonable sample sizes may fail to detect good genes effects due to a lack of statistical power (Griffith et al. 2002; Schmoll et al. 2003). Another explanation for the apparently contradictory results may be that genetic benefits of female extrapair mate choice are context dependent (see Sheldon 1999, 2000b; Schmoll et al. 2003). For example, in environments with favorable conditions, the expected differences in half-sibling performance may be masked and, hence, remain undetected, because a beneficial environment is likely to decrease the significance of heritable paternal genetic variation for offspring fitness. However, when environmental conditions are comparatively poor, the proposed differences between EPO and WPO progeny may appear with EPO showing detectably better performance compared to their WPO maternal half-siblings (see Sheldon 1999). If this is the case, the presumed interaction of differential paternal genetic contribution and environmental context represents an interesting type of a genotype-by-environment interaction (Falconer and Mackay 1996). An intriguing case of contextdependent differential expression of genetic variation has been reported for stalk-eyed flies (Cyrtodiopsis dalmanni) in which differences in male eye span (a condition-dependent, sexually selected ornament thought to indicate male genetic quality) were most pronounced under the least favorable environmental conditions (as represented by food quality, David et al. 2000). This implies that genetic benefits of female mate choice of wide-stalked fly males are highest under the poorest conditions. A related phenomenon has been described for an insular population of the cactus finch (Geospiza scandens), where the effect of inbreeding depression on survival probability was pronounced (or even only detectable), when conspecific competition was high and/or rainfall-dependent food availability was low, demonstrating that the magnitude of inbreeding depression was dependent on environmental context (Keller et al. 2002; see also Coltman et al. 1999). Furthermore, a growing body of evidence in different taxa suggests that paternal genetic effects on offspring traits can indeed show context dependence (Jia and Greenfield 1997; Qvarnström 2001 and references therein; Sheldon et al. 2003; Welch 2003). We tested the hypothesis that genetic fitness benefits of female extrapair mate choice (in terms of differential offspring viability or fecundity), or their magnitude, are context dependent. We analyzed nestling body condition, nestling local recruitment, and recruit reproductive success for coal tit (Parus ater) offspring originating from multiply sired broods that experienced differential environmental conditions (as represented by year of birth and hatching date). If genetic fitness benefits of female extrapair mate choice are context dependent, we expect significant interaction effects of differential paternal genetic contribution (i.e., a nestling s identity as WPO or EPO) with year of birth and/or hatching date on offspring traits. Furthermore, by analyzing whether the mean number of grandchildren (hatchlings) born to a cuckolding female via her WPO progeny differed from that produced via her EPO progeny, we provide the most highly integrated measure of the fitness consequences of female extrapair mating behavior revealed so far. With a total of 2126 nestlings originating from 287 broods with multiple paternity, our analyses are based on the largest sample of genotyped maternal half-siblings tested up to date. MATERIALS AND METHODS Study Species, Study Population, and General Field Methods Coal tits are small, territorial, altricial cavity-nesting passerine birds with biparental care (Glutz von Blotzheim and Bauer 1993). They are socially monogamous (Glutz von Blotzheim and Bauer 1993), but show comparatively high rates of extrapair paternity (see appendix in Griffith et al. 2002). Within the well-investigated genus Parus (chickadees and titmice), these are the highest rates by far (Lubjuhn et al. 1999b; Dietrich et al. 2004). We studied an established nest-box population of coal tits in a mixed coniferous forest near Lingen/Emsland (Lower Saxony, Germany, N, 7 15 E) from 2000 to The 325-ha study area contained about 560 nest-boxes, harboring 132 coal tit breeding pairs in 2000, 184 in 2001, 177 in 2002, and 174 in The proportion of second broods in the population varies from 0% to 100% between years (Winkel and Winkel 1997) and amounted to 63.5% of females with

3 CONTEXT-DEPENDENT PATERNAL GENETIC EFFECTS 647 a successful first brood in 2000, 1.2% in 2001, 31.3% in 2002, and 4.7% in During the breeding seasons (April July) all nest-boxes were monitored at least weekly to record breeding phenology (laying and hatching date), parameters of reproductive performance (clutch and brood size, hatching, fledging, and recruitment success), and the identity of adult birds. Adults captured while feeding nestlings days old were regarded the social (i.e., putative) parents of the respective broods. Both adults and nestlings were banded with uniquely numbered metal rings of the Institute of Avian Research Vogelwarte Helgoland (Wilhelmshaven, Germany). In , blood samples ( 50 l) were taken from the ulnar vein (license no. 509f ), diluted in 250 l APS buffer (Arctander 1988), and stored at 20 C until further use. Blood sampling had no detectable effect on fledging success or nestling local recruitment probability into the study population (Schmoll et al. 2004). In some of the broods, nestling body mass and tarsus length were taken at nestling day 14 post-hatch to a precision of 0.1 g and 0.1 mm using an electronic balance or a caliper, respectively. Distances between nest-boxes were measured by means of standardized nest-site maps (accuracy 10 m). Parentage Exclusion Analysis Details of the basic DNA fingerprinting procedures employed for parentage exclusion analysis have been described elsewhere (Lubjuhn et al. 1999b; Dietrich 2001), hence the fundamental method is outlined only briefly. DNA was isolated according to a modified standard protocol (Lubjuhn and Sauer 1999) and digested with the restriction enzyme HaeIII. After separation by horizontal agarose gel electrophoresis, gels were dried followed by in gel hybridization using the 32 P-labeled oligonucleotide (CA) 8. The banding patterns were visualized by scanning with a phospho-imager (Storm 860, Amersham Biosciences, Freiburg, Germany). They were highly informative and analyzed according to Westneat (1990). The probability of falsely assigning one putative parent to an offspring was as low as (Dietrich 2001). Molecular Sex Determination For molecular determination of offspring sex we used a method based on a primer pair (RG 11 and RG 12, for sequence information see King and Griffiths 1994) that amplifies a W-chromosomal (i.e., female-specific) locus in the coal tit (King and Griffiths 1994). To check the amplificability of each individual DNA sample (positive control) we added another primer pair (2 KM-5 and 2 KM-3), which was originally developed for microsatellite analyses in the great tit (Parus major), but amplifies an anonymous locus in coal tits as well (Gerken 2001). After polymerase chain reaction (for specific conditions see Dietrich et al. 2003), agarose gel electrophoresis (gel size 7 10 cm, 2% agarose, 130 V/cm), and staining with ethidium bromide, this procedure yields only one visible fragment for males and two distinct fragments for females. Statistical Analysis We used the software R (Ihaka and Gentleman 1996) and SPSS 11.0 (SPSS Inc., Chicago, IL) for statistical analyses. All statistical tests were two-tailed and the null hypothesis was rejected at P Nestling condition near fledging Nestling condition near fledging was defined as nestling body mass corrected for tarsus length at day 14 post-hatch. We used linear mixed-effects (LME) models (R-package NLME, Pinheiro and Bates 2000) and likelihood ratio tests to test for the effects of differential paternal genetic contribution (i.e., nestling identity as WPO or EPO), environmental conditions (year of nestling birth or hatching date), and tarsus length on nestling body mass (correcting mass for body size with an ANCOVA model is advantageous to the use of residuals, cf. Garcia-Berthou 2001). Because mass appeared to increase asymptotically with tarsus length, we included both linear and quadratic terms for tarsus length in the analysis. The model also included all three-way interactions involving differential paternal genetic contribution. We included brood period (i.e., first or second brood period) nested within year of nestling birth nested within identity of social parents (hereafter pair) as random effects. The model therefore accounts for random variation between breeding pairs, between different years in which the same pair bred, and between different broods of the same pair in the same year. Nestling local recruitment and recruit reproductive performance We regarded all nestlings as locally recruited that have been found breeding within the period in which nestling local recruitment had been recorded (i.e., ). This included a total of 41 nestlings that had not been found breeding in the year subsequent to their year of birth but in the secondnext (34 recruits) or third-next (seven recruits) year (most probably these individuals had before been nesting just outside the study area or in natural cavities). We analyzed effects of differential paternal genetic contribution, year of nestling birth, hatching date, and their two-way interactions on nestling local recruitment probability using generalized linear mixed models (GLMM, R function glmm-pql in R-package MASS, Venables and Ripley 2002) with binomial error structure and logit link function. These models included the same random effects as the analysis of nestling condition (see above). As it has been shown that local recruitment probability within the study population is not related to offspring sex (Dietrich et al. 2003), we did not consider nestling sex in the model. To analyze long-term reproductive performance (i.e., the number of grandchildren born to a cuckolding female), the same basic model was employed as for local recruitment. However, in this case analyses were carried out separately for the sexes (see below) and a quasi-poisson error structure with a log link function was used. Quasi-Poisson errors provide a way to model over-dispersed count data, that is, count data with variance greater than expected for a Poisson distribution (Venables and Ripley 2002). In the reproductive performance data, overdispersion may arise from the

4 648 TIM SCHMOLL ET AL. fact that only recruited offspring produce grandchildren. Since the function glmmpql fits GLMMs using a penalized quasi-likelihood approach, it is impossible to directly test for significance of model terms and to perform model selection (Venables and Ripley 2002). We therefore first fitted a GLMM appropriate to address a given hypothesis and then judged the significance of effects based on the parameter estimates obtained for this model. In cases where the respective hypothesis addressed solely within-brood differences (i.e., overall mean differences between WPO and EPO maternal half-sibships from the same brood) we included the proportion of EPO as a covariate. This serves to control potentially confounding effects that may arise because paternal genetic contribution varies within as well as between broods (see Snijders and Bosker 1999). For GLMMs, we subsequently refer to an effect as significant if at least one of the parameters associated with the effect is significantly different from zero (at P 0.05). RESULTS To test whether effects of differential paternal genetic contribution (i.e., nestling identity as WPO or EPO) on fitnessrelated offspring traits depend on environmental context, we analyzed interaction effects of differential paternal genetic contribution with the year of nestling birth (2000, 2001, or 2002, reflecting possible interannual variation in environmental conditions) and hatching date within every respective year (reflecting possible intraannual variation in environmental conditions). We used nestling condition near fledging as a short-term measure and nestling local recruitment probability as well as recruit reproductive success as long-term measures for the fitness consequences of extrapair matings experienced by cuckolding female coal tits. Occurrence of Extrapair Paternity Analyses were based on a sample of 431 coal tit broods (involving 257 different females, 264 different males, and 307 different, uniquely composed social pairs) from three consecutive years comprising 3143 offspring. From these, 996 (31.7%) were shown to be EPO. Multiple paternity (i.e., the presence of at least one EPO and one WPO) was detected in 287 (66.6%) broods (involving 197 different females, 198 different males, and 228 uniquely composed social pairs) comprising 2126 offspring. From these, 857 (40.3%) were shown to be EPO, and the mean ( SD) proportion of EPO per multiply sired brood amounted to %. The distribution of the 431 broods across years and breeding periods is summarized in Table 1. Nestling Condition Near Fledging We obtained data on nestling condition for a total of 981 nestlings from 139 broods in the years 2001 (first broods only) and 2002 (first and second broods). For broods with multiple paternity (MP broods), data for 665 nestlings from 93 broods were available. In a first step, we restricted the analysis to MP broods, as we aimed to conduct pairwise comparisons of the performance of maternal half-siblings within every respective brood. In a second step, we extended the analysis to all broods. We tested for effects of differential paternal genetic contribution, year of nestling birth, hatching date, and their interactions on nestling condition using a LME model with nestling mass as the response variable and tarsus length as a covariate (see Materials and Methods). As significant predictors of nestling mass, we identified hatching date (MP broods: , df 1, P ; all broods: , df 1, P ), and the linear and quadratic terms of tarsus length (MP broods: linear term , df 1, P 0.05; quadratic term , df 1, P 0.05; all broods: linear term , df 1, P 0.01; quadratic term , df 1, P 0.05). Predicted nestling mass increased over the range of tarsus lengths measured and decreased with progressing hatching date (data not shown). Yet none of the two- and three-way interactions involving differential paternal genetic contribution, environmental conditions (year of birth or hatching date), and tarsus length had a significant effect on nestling mass (P 0.05 for all interaction terms). Hence, effects of differential paternal genetic contribution on nestling mass did not differ when comparing WPO versus EPO under different environmental conditions. Furthermore, nestling mass was not significantly affected by the main effects of differential paternal genetic contribution (MP broods: , df 1, P 0.72; 95% confidence interval for the parameter estimate for differential paternal genetic contribution when added to the minimal adequate model for nestling mass: 0.07 to 0.1 g; all broods: , df 1, P 0.74) and year of nestling birth (MP broods: , df 1, P 0.18; all broods: , df 1, P 0.16). Nestling Local Recruitment Probability In total, recruitment probability of EPO from MP broods did not differ significantly from that of their maternal WPO TABLE 1. Distribution of 431 coal tit (Parus ater) broods across breeding periods with respect to their paternity status as broods with multiple paternity, broods containing 100% within-pair offspring (WPO) or broods containing 100% extrapair offspring (EPO). Numbers in parentheses refer to the number of offspring genotyped within every respective category of broods Second brood Second brood 2002 Others 1 Total Multiple paternity 60 (503) 47 (303) 102 (747) 41 (314) 35 (245) 2 (14) 287 (2126) 100% WPO 30 (250) 10 (58) 55 (373) 20 (157) 4 (23) 3 (17) 122 (878) 100% EPO 3 (23) 4 (26) 6 (35) 2 (11) 7 (44) 0 (0) 22 (139) Total 93 (776) 61 (387) 163 (1155) 63 (482) 46 (312) 5 (31) 431 (3143) 1 Includes two replacement clutches and one third brood from 2000 and two second broods from 2001.

5 CONTEXT-DEPENDENT PATERNAL GENETIC EFFECTS 649 TABLE 2. Absolute number and proportion of locally recruited within-pair offspring (WPO) and extrapair offspring (EPO), with respect to the breeding periods recruits originated from, for (a) 287 broods with multiple paternity; and (b) all 431 broods genotyped. a) 2000 Second brood Second brood 2002 Others 1 Total EPO recruits 24/187 13/144 25/291 9/123 4/110 0/2 75/ % 9.0% 8.6% 7.3% 3.6% 0.0% 8.8% WPO recruits 48/316 8/159 29/456 17/191 3/135 1/12 106/ % 5.0% 6.4% 8.9% 2.2% 8.3% 8.4% Total 72/503 21/303 54/747 26/314 7/245 1/14 181/ % 6.9% 7.2% 8.3% 2.9% 7.1% 8.5% b) 2000 Second brood Second brood 2002 Others 2 Total EPO recruits 27/210 16/170 26/326 9/134 7/154 0/2 85/ % 9.4% 8.0% 6.7% 4.5% 0.0% 8.5% WPO recruits 83/566 16/217 56/829 27/348 3/158 1/29 186/ % 7.3% 6.6% 7.8% 1.9% 3.4% 8.7% Total 110/776 32/387 82/ /482 10/312 1/31 271/ % 8.3% 7.1% 7.5% 3.2% 3.2% 8.6% 1 Includes one replacement clutch from 2000 and one second brood from Includes two replacement clutches and one third brood from 2000 and two second broods from FIG. 1. Pairwise comparisons of the proportion of within-pair offspring (WPO) and the proportion of their extrapair offspring (EPO) half-siblings recruited from the same brood for 287 coal tit (Parus ater) broods with multiple paternity. Circle size indicates number of multiple datapoints and the number refers to the number of zerozero differences (i.e., broods that produced no recruits at all). Datapoints on the dashed 1:1 isoline represent cases where the proportions of recruited WPO and EPO were identical. half-siblings (GLMM with differential paternal genetic contribution and proportion of EPO per brood as fixed effects: parameter estimate for paternal genetic contribution not significantly different from zero, P 0.53; see Table 2a, Fig. 1). When splitting the sample into first and second brood periods, no overall effect was revealed for first brood periods (P 0.98), but EPO performed significantly better than their WPO maternal half-siblings in second brood periods (P 0.036). When testing for effects of differential paternal genetic contribution, year of nestling birth, hatching date, and their two-way interactions on nestling local recruitment probability, we revealed a significant interaction of differential paternal genetic contribution and hatching date (Table 3a). While local recruitment probability decreased with progressing hatching date for all nestlings, this decrease was less pronounced for EPO than for WPO (Table 3a, Fig. 2). This result reflects the fact that recruited WPO nestlings mainly originated from early (i.e., first) broods, whereas recruited EPO nestlings originated from nests initiated throughout the season in 2000 and 2002, including late (i.e., second) broods (Fig. 3). Overall, nestling local recruitment from 2000 was approximately twice as high compared to recruitment rates from the two other years (Table 2), resulting in a significant effect of year of birth on nestling local recruitment probability (Table 3a). This effect was not an artifact of an a priori differential recapture probability for offspring born in different years due to differential sampling effort, as the differences were also evident when performing the analysis with recruitment in the year subsequent to year of birth as the response variable (data not shown). However, none of the interactions involving year of nestling birth were significant. Extending the analyses to the total of 431 broods genotyped (Table 2b) revealed very similar results (Table 3b). Essentially the same results were obtained when replacing the predictor variable hatching date with breeding period (i.e., first or second brood period) and thus regarding the detected calendar effect as a discrete rather than a continuous phenomenon (i.e., assuming that within-brood period differences in environmental conditions are small compared to betweenbrood period differences). In addition to significant differences between years (parameter estimates significantly different from zero, P for 2001 and P for 2002 when compared to 2000), recruitment from second broods was significantly lower than from first broods (P 0.003, Table 2a), and a significant interaction effect of differential paternal genetic contribution and breeding period was revealed (P 0.040). Since differential dispersal of nestlings may have con-

6 650 TIM SCHMOLL ET AL. TABLE 3. Model parameters for effects of differential paternal genetic contribution (Pat Gen), year of nestling birth, hatching date, and their (two-way) interactions on the probability of nestling local recruitment as estimated from a generalized linear mixed model for (a) 2126 nestlings from 287 broods with multiple paternity; and (b) 3143 nestlings from the total of 431 broods analyzed (SE, standard error; parameter estimates given are on the logit scale; note that positive/negative parameter estimates indicate positive/negative effects relative to the intercept, which represents performance of within-pair offspring born on May 1, 2000). Parameter estimate SE df t P a) Intercept Pat Gen (EPO) Year (2001) Year (2002) Hatching date Pat Gen (EPO) year (2001) Pat Gen (EPO) year (2002) Year (2001) hatching date Year (2002) hatching date Pat Gen (EPO) hatching date b) Intercept Pat Gen (EPO) Year (2001) Year (2002) Hatching date Pat Gen (EPO) year (2001) Pat Gen (EPO) year (2002) Year (2001) hatching date Year (2002) hatching date Pat Gen (EPO) hatching date founded our analysis of local recruitment, we analyzed dispersal distances of the 181 nestlings recruited into the study area. Using LME models, we found that only year of nestling birth had a significant effect on local dispersal distance (mean SD dispersal distances for recruitment from 2000: m; from 2001: m; from 2002: m; LME model using the same random effects as in the analysis of local recruitment: P 0.028), while differential paternal genetic contribution, hatching date, and any two- as well as three-way interactions did not (P 0.23 for all terms). The same results were obtained when extending the analysis to the total of 271 recruits originating from all 431 broods genotyped (data not shown). Recruit Reproductive Performance The hypothesized effects of differential paternal genetic contribution may not only affect nestling survival and subsequent recruitment, but could also be responsible for improved reproductive performance of EPO compared to their WPO maternal half-siblings. We therefore analyzed whether the mean number of grandchildren per offspring born to a cuckolding female via her WPO progeny differed from that produced via her EPO progeny. For all those nestlings that actually recruited, we integrated over all the broods recorded for any given recruit within those three years of the study, in which recruit reproductive performance had been measured (i.e., for female recruits and for male recruits, see below). Offspring that had not been recaptured at all entered the analysis with a value of zero. In 2002 and 2003, cross-foster experiments had been performed involving part of the broods of the respective first brood periods. Since nestlings had always been cross-fostered at day 2 post-hatch, the number of fledglings produced within experimentally manipulated broods cannot be used as the response variable in the current analysis and we used the number of hatchlings instead. Because the number of fledglings is positively and highly significantly correlated with the number of hatchlings in unmanipulated broods (Pearson s product moment correlation: r 0.78, P 0.001, n 362), the number of hatchlings provides a suitable measure of recruit reproductive performance. In a first step, we restricted the analysis to female progeny, as male realized reproductive success cannot be reliably estimated from the number of hatchlings due to the extraordinarily high rates of extrapair paternity. Since exclusion of genetic maternity has been proved to be extremely rare (T. Schmoll and T. Lubjuhn, unpubl. data), we assumed that the social mother was also the genetic mother in every brood of female progeny in the year 2003, in which no parentage analyses had been performed. This enabled us to include a total of 761 female offspring originating from 182 broods, in which at least one female WPO and one female EPO were present (the restriction to broods with sex-specific multiple paternity is necessary to properly conduct a maternal halfsibling comparison). Seventy-one of these 761 female offspring recruited and produced a total of 106 broods. In total, the mean number of hatched grandchildren born to a cuckolding female per offspring via her female WPO progeny did not differ from that obtained via her female EPO progeny (GLMM with differential paternal genetic contribution and proportion of extrapair daughters per brood as fixed effects: parameter estimate for paternal genetic contribution not significantly different from zero, P 0.99; see also Table 4a). The number of hatched grandchildren de-

7 CONTEXT-DEPENDENT PATERNAL GENETIC EFFECTS 651 FIG. 2. Predicted local recruitment probability of nestling coal tits (Parus ater) in relation to hatching date displayed separately for within-pair offspring (WPO) and extrapair offspring (EPO) born in (A) 2000, (B) 2001, and (C) Values are fixed effect predictions of a generalized linear mixed model with year of nestling birth, hatching date, and nestling identity (as WPO or EPO) as fixed predictor variables (see text for details). FIG. 3. Distributions of hatching dates for coal tit (Parus ater) within-pair offspring (WPO) and extrapair offspring (EPO) that recruited (dark bars) compared to all WPO or EPO nestlings analyzed (light bars) for (A) 2000, (B) 2001, and (C) Box-whisker plots show 25th and 75th percentiles (box), median (line within the box), 10th and 90th percentiles (whiskers), and datapoints outside the 10th and 90th percentiles.

8 652 TIM SCHMOLL ET AL. TABLE 4. (a) Mean ( SD) number of grandchildren (hatchlings) per offspring born to cuckolding females via their within-pair offspring (WPO) or extrapair offspring (EPO) female progeny, respectively, for 761 female offspring originating from 182 broods with multiple paternity, in which at least one female WPO and one female EPO were present. (b) Mean ( SD) number of grandchildren (WPO nestlings) per offspring born to cuckolding females via their WPO or EPO male progeny, respectively, for 601 male offspring originating from 147 broods with multiple paternity, in which at least one male WPO and one male EPO were present. Note that, in contrast to the analyses for female progeny, broods from 2003 had to be omitted completely, because no data on parentage were available for this year. a) Year of offspring birth WPO female progeny Mean number of hatchlings produced via EPO female progeny ( 4.75) 1.64 ( 5.24) ( 2.97) 0.93 ( 3.22) ( 1.67) 0.48 ( 1.81) b) Year of offspring birth WPO male progeny Mean number of WPO nestlings produced via EPO male progeny ( 2.27) 0.49 ( 1.96) ( 1.17) 0.44 ( 1.80) creased significantly with progressing offspring hatching date in all the years (Table 5a, Fig. 4). As in the analysis of nestling local recruitment (see above), we found a significant interaction of differential paternal genetic contribution with hatching date, but not with year of offspring birth (Table 5a, Fig. 4). In a second step, we performed the corresponding analyses for a cuckolding female s male progeny using the number of WPO nestlings from broods of male recruits present at blood sampling as the response variable. Although this measure disregards nestling mortality between hatching and bloodsampling, it can be used as a minimum estimate for a male s number of WPO hatchlings. Note, however, that extrapair mating success of male recruits had to be ignored here, because the identity of extrapair sires was known only in part in 2001 and completely unknown in Because we did not perform parentage analyses in 2003, only data from 2001 and 2002 could be considered in these analyses. We were able to include a total of 601 male offspring originating from 147 broods, in which at least one male WPO and one male EPO were present. Forty-five of these 601 male offspring recruited and produced a total of 59 broods, for which data on paternity were available. In total, the mean number of WPO nestling grandchildren born to a cuckolding female per offspring via her male WPO progeny did not differ from that obtained via her male EPO TABLE 5. Model parameters for effects of differential paternal genetic contribution (Pat Gen), year of offspring birth, hatching date and their (two-way) interactions on (a) the number of grandchildren (hatchlings) per offspring born to cuckolding females via their female progeny and (b) the number of grandchildren (within-pair offspring nestlings) per offspring born to cuckolding females via their male progeny. Parameters were estimated from a generalized linear mixed model for (a) 761 female nestlings from 182 broods with multiple paternity, in which at least one female within-pair offspring (WPO) and one female extrapair offspring (EPO) were present and (b) for 601 male nestlings from 147 broods with multiple paternity, in which at least one male WPO and one male EPO were present (SE, standard error; parameter estimates given are on the log scale; note that positive/negative parameter estimates indicate positive/ negative effects relative to the intercept which represents performance of within-pair offspring born on May 1, 2000). Parameter estimate SE df t P a) Intercept Pat Gen (EPO) Year (2001) Year (2002) Hatching date Pat Gen (EPO) year (2001) Pat Gen (EPO) year (2002) Year (2001) hatching date Year (2002) hatching date Pat Gen (EPO) hatching date b) Intercept Pat Gen (EPO) Year (2001) Hatching date Pat Gen (EPO) year (2001) Year (2001) hatching date Pat Gen (EPO) hatching date

9 CONTEXT-DEPENDENT PATERNAL GENETIC EFFECTS 653 FIG. 5. Predicted mean number of grandchildren (WPO nestlings) per offspring born to a cuckolding coal tit (Parus ater) female via her male progeny in relation to offspring hatching date. Predictions are displayed separately for within-pair offspring (WPO) and extrapair offspring (EPO) born in (A) 2000 and (B) Values are fixed effects predictions of a generalized linear mixed model with year of offspring birth, hatching date, and nestling identity (as WPO or EPO) as fixed predictor variables (see text for details). FIG. 4. Predicted mean number of grandchildren (hatchlings) per offspring born to a cuckolding coal tit (Parus ater) female via her female progeny in relation to offspring hatching date. Predictions are displayed separately for within-pair offspring (WPO) and extrapair offspring (EPO) born in (A) 2000, (B) 2001, and (C) Values are fixed effects predictions of a generalized linear mixed progeny (GLMM with differential paternal genetic contribution and proportion of extrapair sons per brood as fixed effects: parameter estimate for paternal genetic contribution not significantly different from zero, P 0.33; see also Table 4b). As for female progeny, the number of WPO nestling grandchildren decreased significantly with progressing offspring hatching date in both years (Table 5b, Fig. 5). For male progeny, differential paternal genetic contribution showed a significant interaction with year of offspring birth but not with hatching date (Table 5b, Fig. 5). model with year of offspring birth, hatching date, and nestling identity (as WPO or EPO) as fixed predictor variables (see text for details).

10 654 TIM SCHMOLL ET AL. Finally, we tested for effects of differential paternal genetic contribution, hatching date, year of recruit birth, and their interactions on the reproductive performance of only those female and male offspring that had actually recruited. Even though these analyses did not include random effects (and their power therefore tended to be exaggerated), we found significant effects on the total number of hatchlings produced per female recruit only for year of birth (P 0.001; P 0.26 for all other model terms). For male recruits, none of the predictor variables had a significant effect on the total number of WPO nestlings produced (P 0.10 for all model terms). DISCUSSION To test whether genetic benefits of female extrapair mate choice depend on environmental context, we analyzed offspring performance for 2126 nestlings from 287 multiply sired coal tit broods in relation to differing environmental conditions. We found a strong negative relationship between progressing hatching date and nestling condition near fledging and nestling local recruitment probability (Fig. 2) within all the years covered by our study. Environmental conditions also differed between years as indicated by significant year effects on local recruitment probability. These are wellknown patterns regularly observed in passerines inhabiting seasonal environments (see Verboven and Visser 1998; Naef- Daenzer et al. 2001). Furthermore, the number of grandchildren (hatchlings) per individual offspring born to a female via her female progeny and the number of grandchildren (WPO nestlings) per individual offspring born to a female via her male progeny were negatively related to offspring hatching date (Figs. 4, 5). If good genes effects or their magnitude indeed depended on environmental context, we would predict EPO nestlings to perform better (in terms of viability and/or fecundity) compared to their maternal half-siblings, when environmental conditions were comparatively poor (i.e., late in season, when overall nestling performance had significantly declined or in the year 2001, from which recruitment was significantly lower compared to 2000). In accordance with this prediction, we found a significant interaction effect of differential paternal genetic contribution and hatching date on local recruitment probability, albeit not on nestling condition. This interaction was caused by the fact that EPO recruitment probability declined less strongly with progressing hatching date than did WPO recruitment probability (Fig. 2). Basically the same results emerged, when the predictor variable hatching date was replaced with breeding period (i.e., first or second brood). This indicates that a substantial proportion of the variation in environmental conditions occurs between rather than within distinct breeding periods. The same general pattern of context dependence was evident when using the number of grandchildren per offspring born to a cuckolding female via her female progeny as the dependent variable (Fig. 4). Together with the fact that reproductive success did not differ between those WPO and EPO that had actually recruited, this result suggests that differences in offspring recruitment, rather than offspring fecundity, are crucial to a cuckolding female s fitness. For the number of WPO grandchildren per offspring born to a cuckolding female via her male progeny, however, a different pattern emerged. In contrast to female progeny, no interaction between differential paternal genetic contribution and hatching date was evident, while an interaction between differential paternity and year was significant (Fig. 5). In line with our prediction, in the year that gave rise to significantly lower overall recruitment (2001 compared to 2000) male EPO performed better than their male WPO half-siblings, while the opposite was evident for recruitment from However, our results regarding male progeny reproductive performance need to be treated with caution, as the number of WPO nestlings produced by male WPO or EPO progeny reflects only a part of the fitness consequences experienced by a cuckolding female, because extrapair fertilization success of male progeny in other broods could not be assessed. In three consecutive years, the two groups of maternal halfsiblings showed consistent differences in the hatching date related change in recruitment probability (Fig. 2) and in the hatching date related change in the number of grandchildren born to a female via her female progeny (Fig. 4) as a consequence of the differences between maternal half-sibships in recruitment probability. This suggests that the patterns observed may indeed represent a general phenomenon rather than being caused by particular conditions within a single year. Our results add to a small number of studies that have demonstrated environmental context dependence of paternal genetic effects on offspring fitness (Jia and Greenfield 1997; Jia et al. 2000; Sheldon et al. 2003; Welch 2003; see also Qvarnström 2001). The fact that genotype-by-environment interactions of this kind can render genetic benefits of mate choice context dependent also in natural populations of birds may have important implications for studying the evolution of female (extrapair) mating preferences (cf. also Qvarnström et al. 2000; Qvarnström 2001; Badyaev and Qvarnström 2002), the evolution of male secondary sexual characters subject to female (extrapair) mate choice (cf. David et al. 2000; Badyaev and Qvarnström 2002; Sheldon et al. 2003), and the maintenance of additive genetic variance in sexually selected traits (cf. Jia et al. 2000; Badyaev and Qvarnström 2002). Therefore, an important consequence of our results is that the potential context dependence of paternal genetic effects on offspring fitness needs to be explicitly accounted for in future studies on the adaptive significance of extrapair matings, either by measuring relevant environmental variables and integrating them into statistical models or by modifying experimental designs a priori. With respect to the shape of the benefit functions calculated from our data (Figs. 2, 4), we predict female preferences for extrapair mates to be plastic (Qvarnström et al. 2000; Qvarnström 2001; Badyaev and Qvarnström 2002), such that females breeding late (when expected benefits are greatest or even exclusively present) show stronger preferences for extrapair mates, while females breeding early should be more reluctant to avoid the associated costs. This prediction is supported by the fact that both, the proportion of females successfully engaging in extrapair copulations and the proportion of EPO, increased from first to second broods in the study population in 2000 as well as 2002 (Dietrich et al.

11 CONTEXT-DEPENDENT PATERNAL GENETIC EFFECTS , see also Table 1). However, more than two thirds of first broods within each of the three years surveyed contained EPO (see Table 1), also in 2000 and 2002, when this seemed to be rather maladaptive (Figs. 2, 4). Additionally, we found no significant overall effect of differential paternal genetic contribution, indicating that there was no net genetic fitness benefit of extrapair copulations for cuckolding females (Fig. 1). Thus, although late-breeding females seem to be able to benefit from extrapair copulations, indirect fitness benefits in terms of increased offspring viability or fecundity alone are unlikely to have maintained female multiple mating behavior within the three years covered by the study. The degree to which such a type of benefit can contribute to selection for female extrapair matings in general may then depend on the frequency with which environmental conditions comparable to those in second brood periods are encountered and on the fitness costs associated with extrapair mating behavior. If costs are low (e.g., Kempenaers et al. 1998; MacDougall- Shackleton and Robertson 1998), even occasional situations with suitable environmental conditions may reward cuckolding females with benefits great enough to outweigh these minimal costs. If costs are substantial (e.g., Lubjuhn et al. 1993; Dixon et al. 1994; Valera et al. 2003), these benefits may not suffice and a further type of benefit could be involved contributing to maintain extrapair mating behavior in our study population in such an extraordinarily high frequency. This postulated further type of reward may also represent a genetic benefit. For example, females may be able to increase attractiveness and, accordingly, extrapair mating success of their male progeny (good genes as attractiveness genes, Kokko et al. 2002). By involving more than a single genetic sire, females may also perform a kind of genetic bethedging if reliable assessment of male genetic quality is difficult or environmental conditions experienced by offspring vary in an unpredictable manner (Jennions and Petrie 2000). In this case, we would expect differences in the long-term fitness rewards of females producing multiply sired broods compared to females not producing them than differences between categories of maternal half-siblings within broods. Furthermore, nongenetic benefits (cf. compilation in Griffith et al. 2002) and the possibility that extrapair matings are not (solely) a female driven behavioral strategy (Westneat and Stewart 2003) deserve more attention. A further alternative explanation for females engaging in extrapair copulations also during early broods may be that females indeed behaved maladaptively in these cases, due to an evolutionary trap (Schlaepfer et al. 2002). If the ecological conditions for breeding within the study area have not always been as beneficial as encountered in contemporary first brood periods, current female extrapair mating decisions may represent a ghost-of-past (genetic) benefit. This implies that selection has not yet had sufficient time to adaptively fine-tune this complex behavioral trait. This interpretation is to some extent supported by the fact that both population density (Winkel and Winkel 1997 and unpublished data) and breeding success (mean number of offspring fledged, Winkel et al. 2004) during first brood periods have significantly increased over the last three decades within the study area. Conclusions In conclusion, we have shown that paternal genetic effects on offspring fitness were context dependent when analyzing the extrapair mating system within a natural population of the socially monogamous coal tit. If genotype-by-environment interactions of this kind are widespread in nature, this has important implications for the evolution of mate preferences, the evolution of secondary sexual characters, and the maintenance of additive genetic variation in sexually selected traits (cf. Qvarnström 2001; Badyaev and Qvarnström 2002). However, we were not able to demonstrate that cuckolding females did obtain a net genetic fitness benefit from extrapair fertilizations in terms of offspring viability or fecundity. Therefore, if female extrapair mating behavior is costly (as suggested by results revealed in some passerine birds though not in others) as well as adaptive, a further type of benefit may be involved maintaining the high frequency of extrapair matings in our study population. With respect to results obtained from maternal half-sibship comparisons in other passerine birds (Kempenaers et al. 1997; Sheldon et al. 1997; Johnsen et al. 2000; Foerster et al. 2003), our results suggest the possibility that more than a single selective pressure may have shaped the evolution of female extrapair mating behavior in passerine birds. Even between closely related species and species with similar ecology, the relative importance of these selective pressures could vary (cf. also Foerster et al. 2003; Kleven and Lifjeld 2004). Both, the potential context dependence of genetic benefits of extrapair mate choice and the possibility that different selective forces have shaped female extrapair mating behavior, need to be considered and accounted for in future studies. ACKNOWLEDGMENTS We thank S. Bleidissel, A. Kalt, M. Orland, A. Petzold, T. Meißner, and C. Wallnisch for their help in the laboratory; J. Brün, J. Delingat, J. Eggert, T. Gerken, D. Stiels, J. Welcker, and especially V. Janzon, A. Quellmalz, and D. Winkel for support in the field; and K. Körner and H. Körner for their hospitality during field work. K. Reinhold, M. Webster, B. Sheldon, and two anonymous reviewers provided helpful comments. S. Verhulst pointed our attention to potential pitfalls in mixed model analyses. This project was supported by the Deutsche Forschungsgemeinschaft (Lu 572/2-3,4). LITERATURE CITED Arctander, P Comparative studies on avian DNA restriction fragment length polymorphism analysis: convenient procedures based on blood samples from live birds. J. Ornithol. 129: Badyaev, A. V., and A. Qvarnström Putting sexual traits into the context of an organism: a life-history perspective in studies of sexual selection. Auk 119: Barber, I., S. A. Arnott, V. A. Braithwaite, J. Andrew, and F. A. Huntingford Indirect fitness consequences of mate choice in sticklebacks: offspring of brighter males grow slowly but resist parasitic infections. Proc. R. Soc. Lond. B 268: Birkhead, T. R., and A. P. Møller Sperm competition in birds: evolutionary causes and consequences. Academic Press, New York Sperm competition and sexual selection. Academic Press, San Diego, CA.