Environmental and population dependency of genetic variability-fitness correlations in Rana temporaria

Molecular Ecology (2005) 14, 311–323

doi: 10.1111/j.1365-294X.2004.02394.x

Environmental and population dependency of genetic variability-fitness correlations in Rana temporaria

Blackwell Publishing, Ltd.

D A V I D L E S B A R R È R E S ,*§ C R A I G R . P R I M M E R ,†¶ A N S S I L A U R I L A ‡ and J U H A M E R I L Ä * *Ecological Genetics Research Unit, Department of Biological and Environmental Sciences, PO BOX 65, FI-00014 University of Helsinki, Finland, §Present Address: Department of Biology, Laurentian University, Sudbury, Ontario P3E 266, Canada, †Department of Biological and Environmental Sciences, PO BOX 65, FI-00014 University of Helsinki, Finland, ¶Present Address: Department of Biology, 20014, University of Turku, Finland, ‡Department of Population Biology, Evolutionary Biology Centre, Uppsala University, SE-752 36 Uppsala, Sweden

Abstract Considerable effort has been invested in studying the relationship between fitness and genetic variability. While evidence exists both for and against positive genetic variability– fitness correlations (GFC), the possible environment and population-dependency of GFCs has seldom been tested. We investigated GFCs in common frog (Rana temporaria) tadpoles reared under different temperatures and feeding regimes in four replicate populations. Genetic variability in eight microsatellite loci in 238 parents was used to estimate heterozygosity (H) and mean expected d2 in 158-sibships (4515 offspring). Generalized linear mixed model analyses of offspring fitness traits (survival to metamorphosis, developmental and growth rate) revealed that offspring survival probability was positively correlated with H, and that relationships were similar in all four populations tested. However, significant interaction between other genetic variability measures (d 2, relatedness) and treatment conditions indicated that GFCs were detectable in some, but not in all environments. Interestingly, GFCs between survival and both heterozygosity and relatedness were most pronounced in stressful environments (i.e. limited food). Developmental and growth rates were significantly associated with d 2 but less with H and relatedness. Furthermore, many of these GFCs were population-specific. These results suggest — in line with the contention that expression of inbreeding depression can be environment dependent — that GFCs can also be highly sensitive to the environmental conditions under which they are measured. The results further suggest that the observed positive correlation between H and survival probability is likely to be explainable by the ‘general’, rather than by the ‘local’ or ‘direct’ effect hypotheses. Keywords: Amphibians, d2, fitness, genotype –environment interaction, heterozygosity, microsatellites Received 30 June 2004; revision received 29 September 2004; accepted 29 September 2004

Introduction The relationship between fitness and genetic variability (genetic variability–fitness correlation or GFC) has been investigated for decades (Allendorf & Leary 1986; Mitton 1997; David 1998; Hansson & Westerberg 2002). Inferences Correspondence: David Lesbarrères, §Present Address: Department of Biology, Laurentian University, Sudbury, Ontario PZE 266, Canada, Fax: + 358-9-191 57694; E-mail: [email protected] ¶Present Address: Department of Biology, 20014, University of Turku, Finland © 2004 Blackwell Publishing Ltd

about this relationship in natural populations have commonly been made from associations between markerbased measures of genetic variability, such as average heterozygosity at molecular markers, and fitness-related traits such as survival, rate of development, growth or fluctuating asymmetry (Allendorf & Leary 1986; Britten 1996; David 1998). Both positive and negative correlations have been found and the association between average heterozygosity and fitness seems to be generally weak (Britten 1996). In a recent meta-analysis, Coltman & Slate (2003) found that life-history trait-heterozygosity correlation was significantly greater than zero whereas

312 D . L E S B A R R È R E S E T A L . morphometric trait-heterozygosity correlation was not. This is consistent with the fact that life-history traits exhibit greater inbreeding depression than morphometric traits (DeRose & Roff 1999). The relationship between genetic variability and fitness may arise resulting from several reasons, inbreeding depression being one of them. Mating between relatives increases the proportion of homozygous loci in offspring, therefore leading to increased probability of expression of recessive deleterious alleles, and thereby decreased fitness. This phenomenon is often evidenced through increased homozygosity (or decreased heterozygosity) observed at genetic markers such as allozyme or microsatellite loci (Frankham & Ralls 1998). In contrast, mating between genetically divergent individuals can result in offspring with higher than average fitness, a phenomenon known as a ‘heterosis’. It is often invoked to explain higher fitness of individuals with higher levels of genetic variability, the cause for this being either a reduced occurrence of deleterious alleles in the homozygous state, and/or increased heterozygosity of loci displaying associative overdominance (Lynch & Gabriel 1990). Three hypotheses have been put forth to explain GFCs (Hansson & Westerberg 2002). Under the direct effect hypothesis, GFCs result from selection acting directly on the analysed loci. Under the local effect hypothesis, linkage disequilibrium causes GFCs between the neutral markers and the fitness loci in their chromosomal vicinity. Under the general effect hypothesis, GFCs result from effects of homozygosity at genome-wide distributed loci reflected to the marker loci (Hansson & Westerberg 2002). Despite numerous evaluations of multilocus GFCs during the last decades, there is still controversy over the underlying mechanisms of such correlations both in natural and experimental populations (David 1998; Hansson & Westerberg 2002). Part of the problem is that average multilocus heterozygosity provides relatively limited information as only the identity or nonidentity of allelic phenotype at each locus is considered. Moreover, allozyme loci have commonly low heterozygosity and few alleles segregating at each locus so that even outbred individuals will appear homozygous at many loci. Hence, individual heterozygosities tend to fall in a relatively narrow range. Markers with higher levels of polymorphism, such as microsatellites, can offer a more sensitive measure of genetic variability. First, the range of heterozygosity values is typically wider, and second, they also provide genetic distance information beyond allelic identity vs. nonidentity. Regarding the latter, the squared difference in repeat units between two alleles at a locus averaged over all typed loci, known as mean d2 (Coulson et al. 1998), provides an alternative measure to individual multilocus heterozygosity. Under the stepwise mutation model, it is expected that d2 is a linear function of time as coalescence of the two alleles, therefore mean d2 is expected

to reflect the genetic distance between the two parental gametes (Valdés et al. 1993; Goldstein et al. 1995; Coulson et al. 1998). Thus, a positive correlation between mean d2 and fitness would suggest that individuals with dissimilar parents have greatest fitness. It has been suggested that heterozygosity is suited for detecting recent inbreeding events, whereas mean d2 may detect events (e.g. population admixture) deeper in an individual’s ancestry (Coulson et al. 1998). Mean d2 is also a convenient metric for assessing the relative position of individuals within a population on the inbred-outbred continuum (Pemberton et al. 1999), but seems to perform better than multilocus heterozygosity only following admixture of two large divergent populations (Hedrick et al. 2001; Tsitrone et al. 2001). Inbred individuals are thought to be less stress tolerant (Wright 1922), but the relationship between the magnitude of inbreeding depression and environmental stress is not always clear (Keller & Waller 2002). However, there is some evidence to suggest that at least sometimes, GFCs may be environmentally dependant and possibly related to the harshness of the environment in which they are measured. For example, variation in GFCs may be enhanced by laboratory-induced (Danzmann et al. 1988; Audo & Diehl 1995) or natural (Borsa et al. 1992) levels of environmental stress. However, Scott & Koehn (1990) observed positive GFC in coot clams subjected to either a thermal stress or a salinity stress but cumulating both stresses did not reveal such a correlation. Similarly, Audo & Diehl (1995) observed varying GFCs for earthworm growth under certain conditions but not in the ones they expected to be the most stressful. Further, Borsa et al. (1992) observed strong GFCs for survival under naturally high stress but failed to observe similar relationship under experimental conditions. Hence, while these results hint about the importance of environmental effects, they are not always clear-cut, and also, suffer from various limitations in terms of population number (but see Mcelroy & Diehl 2001), low numbers of alternative environmental conditions, small numbers of individuals and/or lack of replication. The aim of this paper was to investigate environmental dependency of GFCs using common frog (Rana temporaria) tadpoles as models. In contrast with previous studies, we investigated the environmental dependency of GFCs in a multiple environment and population context. We did this by relating genetic variability — as assessed from eight microsatellite loci — with variation in three fitness-related traits (survival probability, developmental and growth rate) in families of tadpoles originating from controlled crosses (4515 offspring) and subject to different experimental treatments (varying temperature and food leading to six environmental conditions) known to influence tadpole development and survival probability (e.g. Merilä et al. 2000a; Laugen et al. 2003a). Under the assumptions that (1) genetic variability in microsatellite loci reflects variability © 2004 Blackwell Publishing Ltd, Molecular Ecology, 14, 311–323

G E N E T I C V A R I A B I L I T Y - F I T N E S S C O R R E L A T I O N S 313 in genome-wide heterozygosity as postulated by the general effect hypothesis (cf. Hansson & Westerberg 2002), and that (2) environmental stress enhances expression of recessive deleterious alleles (Crnokrak & Roff 1999; Hedrick & Kalinowski 2000) and thereby also GFCs, we predicted that GFCs would be more pronounced under stressful than nonstressful environmental conditions. These predictions were tested using data from four different R. temporaria populations to see whether relationships were consistent over different geographically widely separated populations.

Materials and methods The study species and populations The common frog — the most widespread anuran of Europe (Fog et al. 1997) — is a medium-sized (c. 30 g) pond-breeding frog with aquatic larval and terrestrial adult stage. Adults are site tenacious (Haapanen 1970), but little is known about juvenile dispersal. However, population genetic studies indicate that dispersal is limited: genetic differentiation in microsatellite loci is extensive (Palo et al. 2003), even over relatively short distances (2–3 km; Hitchings & Beebee 1997). Furthermore, both larval and adult life history traits display extensive geographical variation (Fog et al. 1997; Miaud et al. 1999; Miaud & Merilä 2001). While evidence for genetic differentiation in adult traits is lacking, genetic differentiation in larval traits over relatively short (c. 5– 200 km; Laurila 2000; Loman 2002; Räsänen et al. 2002) and longer (200–1600 km; Merilä et al. 2000a; Laurila et al. 2002; Pahkala et al. 2002; Laugen et al. 2003b; Palo et al. 2003) distances is common. Although evidence for occurrence of inbreeding is mostly circumstantial (Hitchings & Beebee 1997; Pakkasmaa et al. 2003), the size of local breeding populations in Scandinavia is typically quite small (Loman 2001; Laugen et al. 2003b). In 81 localities surveyed during 1989–2001 in southern Sweden, the average female population size was 79.2 (± 123.1; Loman 2001) females. In Kilpisjärvi in northern Finland, the average female population size in 61 ponds was 14.0 (± 0.1; J. Merilä, unpublished). The four populations included in this paper were situated along a 1600 km latitudinal gradient from southern Sweden to northern Finland (Fig. 1; Table 1). The populations were the same as used in earlier studies (e.g. Pahkala et al. 2002; Laugen et al. 2003a; Laugen et al. 2003b; Palo et al. 2003), and all of them had female population size less than 100 individuals. More information about populations and the climatic conditions in these localities is given in Laugen et al. (2003a,b).

Estimation of fitness components Tadpoles used in common garden experiments were obtained from laboratory crosses of adults collected from © 2004 Blackwell Publishing Ltd, Molecular Ecology, 14, 311–323

spawning sites at the onset of the breeding season in 1998. This procedure ensured that all tadpoles were of the same age at the start of the experiment and subject to the same early environmental effects, and allowed controlled crosses according to a North Carolina II design (e.g. Lynch & Walsh 1998) to be made. In Kilpisjärvi and Uppsala populations, 32 full-sib families (i.e. 16 half-sib families) were created where eggs from each of eight females were fertilized by sperm from four of the 16 males. The Lund tadpoles stem from 30 full-sib families (eight females and 15 males used). The Umeå tadpoles stem from 64 full-sib families (16 females and 32 males used). Because of the large difference in the time of spawning among the populations (Merilä et al. 2000b), the starting dates for the experiment also differed. In the case of the southernmost population (Lund), the fertilizations were performed on 9 April 1998, whereas in the case of the northernmost population (Kilpisjärvi), the corresponding date was 4 June 1998. However, the rearing conditions were identical for all populations (Laugen et al. 2003b). The crosses were made as outlined in Laugen et al. (2003b). The eggs were divided into three different temperature treatments (14, 18 and 22°C ± 1°C, two bowls per cross in each temperature) at which they were kept until hatching. Water was changed every third day during embryonic development. When most of the embryos in a given temperature treatment had reached developmental stage 25 (Gosner 1960), eight randomly chosen tadpoles from each cross were placed individually in 0.9 L opaque plastic containers at each of the two food levels (restricted and ad libitum). This procedure was repeated for each population in the three temperature treatments, resulting in 48 experimental tadpoles per cross. However, as a result of mortality during the experiment, the final number of tadpoles per family was typically fewer than 48. Every seventh day, the tadpoles were fed a finely ground 1: 3 mixture of fish flakes (TetraMin, Ulrich Baensch GmbH, Germany) and rodent pellets (AB Joh. Hansson, Uppsala, Sweden). The amount of food given to each tadpole was 15 mg (restricted) or 45 mg (ad libitum) for the first week, 30 or 90 mg for the second week, and 60 or 180 mg per week thereafter until metamorphosis. The tadpoles were raised in dechlorinated tap water that was aerated and aged for at least 24 h before use. The water was changed every seventh day in conjunction with feeding. The light rhythm was 16 L:8D. Close to metamorphosis, the tadpoles were checked every day, and individuals that had reached Gosner stage 42 were noted. Developmental rate (age at metamorphosis) was defined as the number of days elapsed between reaching Gosner stages 25 and 42. Growth rate was estimated as the (fresh) weight (to closest milligram) at metamorphosis corrected for the time (in days) elapsed between Gosner stages 25 and 42. In effect, this estimates the weight the tadpoles would have obtained

314 D . L E S B A R R È R E S E T A L . Fig. 1 Map of northern Europe showing the location of the four study populations in Sweden and Finland.

Table 1 Descriptive information about observed (parents) and estimated (offspring) genetic variability measures (± SE) in four Rana temporaria populations Parents

Offspring

Population

Coordinates

Np*

A†

H O‡

d 2§

rxy¶

FIS**

No††

H EST‡‡

2 d EST §§

Kilpisjärvi Umeå Uppsala Lund

69°03′ N, 20°47′ E 63°49′ N, 20°14′ E 59°51′ N, 17°14′ E 55°42′ N, 13°23′ E

24 48 24 23

5.4 6.5 6.5 8

0.61 (0.18) 0.54 (0.17) 0.52 (0.19) 0.53 (0.23)

33.8 (31.9) 57.5 (36.5) 50.9 (47.1) 63.4 (45.4)

− 0.03 (0.26) − 0.11 (0.21) − 0.07 (0.31) − 0.04 (0.19)

− 0.040 − 0.043 0.143 0.186

1116 (32) 1508 (64) 478 (32) 1417 (30)

0.62 (0.12) 0.55 (0.11) 0.65 (0.18) 0.68 (0.13)

51.4 (28.1) 65.5 (19.8) 70.1 (32.7) 89.0 (41.1)

*Np = number of parents genotyped; †A = average number of alleles per microsatellite locus; ‡HO = observed heterozygosity for the parents; §d2 = mean d2 for the parents; ¶rxy = mean pairwise relatedness of parents (see methods); **FIS = fixation index indicating deviations from HW expectations (in all cases P > 0.05); ††No = number of offspring in the experiments (number of full-sib families); 2 ‡‡HEST = mean estimated heterozygosity for the offspring (see methods); and §§ d EST = mean estimated d 2 for the offspring (see Methods). © 2004 Blackwell Publishing Ltd, Molecular Ecology, 14, 311–323

G E N E T I C V A R I A B I L I T Y - F I T N E S S C O R R E L A T I O N S 315 had they all reached metamorphosis on the same date. This is more flexible than dividing the weight by time, which carries a strong assumption that growth is constant over the whole period. Fresh, rather than dry weights were used as the measures are strongly correlated (Laugen et al. 2003b). Survival probability was estimated as described in succeeding discussions.

Genetic variability analyses To obtain estimates of genetic variability, allelic variation was assessed for both sires and dams from each of the four populations in eight presumably neutral microsatellite loci: Rt2Ca2–22 (Trent Garner unpublished: F-CGGCTTACAAGAGGTGGAG; R-AGACTCCCTTACAGGCATGG), Rt2Ca25 (Trent Garner unpublished: F-GCCAGGGTATGTAAACTTATGAGC; R-CAAATGTATATTATTGGTGCAATGG), RRD590 (Vos et al. 2001), RtµH (Pidancier et al. 2002), RtSB03 (Berlin et al. 2000), Rtempµ4, Rtempµ5 and Rtempµ7 (Rowe & Beebee 2001a). DNA extraction, polymerase chain reaction (PCR) amplifications and gel electrophoresis were performed as in Palo et al. (2003). The following estimators of genetic variability were calculated for each individual with fstat 2.8 (Goudet 1999) and kinship 1.2 (Goodnight & Queller 1999): observed heterozygosity (HO), i.e. the number of heterozygous loci divided by the total number of loci analysed in an individual; mean d2, the squared distance (in repeat units) between the two alleles within a locus, averaged over all loci analysed in an individual (Coulson et al. 1998); and rxy, a pairwise relatedness estimator (Queller & Goodnight 1989), which measures the relatedness of the parents used in the common garden experiment. The dinucleotide RtµH and RtSB03 were excluded from the data set for mean d2 calculations because of the common occurrence of 1 bp size differences, suggesting that a stepwise mutation model was not appropriate for these loci. Tests for deviation from Hardy–Weinberg equilibrium and linkage disequilibrium were implemented using genepop vs. 3.1 (Raymond & Rousset 1995). The average number of successfully genotyped parents for each locus was 23.9, 47.3, 23.9 and 22 for Kilpisjärvi, Umeå, Uppsala and Lund populations, respectively (Table 1). The average number of successfully genotyped loci for each parent was 7.96, 7.88, 7.96 and 7.65 for Kilpisjärvi, Umeå, Uppsala and Lund populations, respectively.

Estimation of offspring genetic variability based on parental genotype data Average within family offspring genetic variability was estimated from parental genotype information (Primmer et al. 2003). This method was chosen as it enabled fitness © 2004 Blackwell Publishing Ltd, Molecular Ecology, 14, 311–323

estimates to be based on a large sample size (4515 offspring) without requiring genotyping of these individuals. It has also been shown to be highly accurate for estimating genetic diversity–fitness correlations compared to estimates where observed offspring genotypes were used (Primmer et al. 2003). Briefly, multilocus microsatellite genotypes for all possible offspring genotypic combinations in a family were estimated using the parental genotypes (4n combinations where potential parents have been analysed for n loci). Mean d2 and HO values were then calculated for each of the simulated multilocus genotypes and these values then averaged over all 4n genotypic combinations (65536 combinations for eight loci) resulting in one value per male-female pair for each estimator. This was done using the r2-d2 computer program (available from http:// www.helsinki.fi/∼primmer) which assumed Mendelian inheritance, no mutation and no linkage between loci. These estimated offspring genetic diversity indices based on parental genotypic data are hereafter referred to as 2 mean d EST and HEST, respectively.

Statistical analyses of GFC correlations Separate analyses were carried out for survival probability, developmental and growth rates. In addition, in order to disentangle the three main hypotheses proposed to explain the cause of genetic variability–fitness correlations, each locus was also analysed separately. The mean value of each full-sib family in each treatment combination was used as a response variable in the analyses. Hence, the sample size per population for each treatment combination was 32, except for the Umeå population for which it was 64. Developmental rate was log-transformed before the analyses to meet the assumptions of normality and homogeneity of variances. Associations between fitness measures and genetic diversity were analysed using general and generalized linear mixed models (GLMMs; McCullagh & Nelder 1989). GLMMs were appropriate for these analyses as they allow analysis of data where the response variable is determined by both random and fixed effects (McCullagh & Nelder 1989; Merilä et al. 2001). In our data set, the response variable results from families sharing one parent either dam or sire. Therefore ‘sire’ and ‘dam’ were included in all models as random effects within populations. However, for simplicity of presentation, the random effects are not shown in Table 2. This was done using glimmix macro (for survival analysis) and proc mixed (for developmental and growth rates) of the sas statistical package according to (Littell et al. 1996). To detect whether GFCs resulted from a general trend across loci or from some particular loci, we also estimated GFCs for each locus separately and looked at their sign and significance after applying Bonferroni correction for multiple tests.

316 D . L E S B A R R È R E S E T A L . Table 2 General linear models of survival, development rate and growth rate of Rana temporaria tadpoles from four populations accounting 2 for effects estimated multilocus genetic variability (Gvar). Results are shown for heterozygosity d EST and relatedness (rxy) as estimators of genetic variability. Pop. = population, Temp. = temperature, d.f. = degrees of freedom Survival

Developmental rate

Growth rate

Source

d.f.

HEST

2 d EST

rxy

HEST

2 d EST

rxy

HEST

2 d EST

rxy

Pop. Temp. Food Pop. × temp. Pop. × food Temp. × food Pop. × temp. × food Gvar Gvar × pop. Gvar × temp. Gvar × food Gvar × pop. × temp. Gvar × pop. × food Gvar × temp. × food Gvar × pop. × temp. × food

3,35 2587 1587 6587 3587 2587 6587 1587 3587 2587 1587 6587 3587 2587 6587

0.97 1.13 0.03 0.96 0.37 1.36 1.15 4.99* 1.1 1.47 3.18˚ 0.61 0.87 1.12 0.9

1.39 3.47 10.51** 1.41 1.97 2.54 2.21* 3.72˚ 2.08 4.28* 4.14* 0.53 2.33˚ 1.71 1.15

0.31 0.61 104.27*** 9.11*** 1.5 0.72 5.02*** 7.94** 1.74 1.05 4.20 0.54 0.27 3.56* 2.25*

14.25*** 467.3*** 24.51*** 3.67** 0.83 11.68*** 1.28 2.58˚ 1.66 1.63 0.08 2.51* 0.69 3.32* 1.06

46.69*** 1395*** 41.63*** 11.32*** 2.67* 19.86*** 1.22 3.91* 2.28˚ 1.11 4.06* 1.93˚ 3.85** 2.09 0.70

93.38*** 8117*** 415.46*** 47.04*** 2.70* 110.7*** 4.30** 2.04 0.66 2.15 0.02 2.66* 0.13 1.43 0.70

1.19 1.37 140.4*** 1.02 0.29 2.38˚ 1.49 0.95 0.45 0.66 0.01 0.50 1.44 0.11 0.77

3.75* 3.96* 408.6*** 0.92 0.57 4.25* 1.95˚ 5.16* 2.70* 2.83˚ 2.21 2.53* 1.61 0.11 0.61

32.68*** 76.05*** 2673*** 13.20*** 12.04*** 30.36*** 7.96*** 1.22 1.50 0.72 0.81 0.53 2.28˚ 0.07 0.82

˚P < 0.1, *P < 0.05, **P < 0.01, ***P < 0.001. Underlined are the results significant after applying Bonferroni correction.

In order to allow comparisons with other studies, we estimated the weight of our correlations according to the following formula (Sokal & Rohlf 1995): [(A − 2)N]0.5 where A is the number of populations and N is the mean number of individuals used for the fitness assay per population. In our case (A = 4 and N = 1128.75), the formula leads to a weight of 47.51. The degrees of freedom for a correlation are determined solely by the number of populations (A), but the actual power of the correlation to reveal the underlying relationship also depends on the standard errors surrounding the point estimates used in the correlation. Thus, this approach was appropriate for comparison with other studies (Reed & Frankham 2001). We also converted the result of our analyses to r, the equivalent of the Pearson product moment correlation coefficient, for use as a common metric of effect size. We chose r because it is has been reported or converted from other studies on the subject and its interpretation is quite simple (i.e. r2 represents the variance explained). As we reported results from F-tests with a single numerator degree of freedom and d as the denominator degree of freedom, the conversion was (Rosenthal 1991): r = √[F(1, d)/(F1, d + dferror)] We estimated r for each statistical test based on each genetic variability estimator separately.

Results Genetic variability — Descriptives The genotype frequencies did not deviate significantly from expected Hardy–Weinberg proportions in any of the four populations after correcting for multiple tests using Bonferroni procedure (Rice 1989), and FIS values did not deviate significantly from zero in any of the populations (Table 1). Statistically significant linkage disequilibrium was observed only between loci Rtempµ4 and Rtempµ5 in one population (Umeå). Although populations did not differ in their levels of genetic variability (for both heterozygosity, d2 and the degree of relatedness), the crosses led to differences in the estimated genetic variability of the offspring (Table 1). Average predicted heterozygosity (HEST) 2 ) differed among and average predicted d2-values ( d EST populations (anova: F3,154 = 8.28, P < 0.0001; F3,154 = 8.6, 2 , respectively). Correlations P < 0.0001 for HEST and d EST between the genetic variability estimators for both parents and offspring were significant (HO and d2: r = 0.391, n = 2 : r = 0.447, n = 158, P < 0.001). 119, P < 0.001; HEST and d EST

Genetic variability and survival The probability to survive until metamorphosis was positively associated with all the three measures of genetic 2 variability, albeit marginally nonsignificantly with d EST (Table 2; Fig. 2). While the HEST — survival relationships were similar in all populations and temperature treatments © 2004 Blackwell Publishing Ltd, Molecular Ecology, 14, 311–323

Fig. 2 A depiction of genetic variability-fitness relationships. X-axes display Pearson product moment correlation coefficients between fitness components (survival [a–c], development rate [d–f] and growth rate [g–i]) and different measures of genetic variability (heterozygosity [a, d, g], d2 [b, e, h] and relatedness [c, f, i]) for tadpoles in different populations and temperature-food treatment conditions (low = restricted food; high = ad libitum).

G E N E T I C V A R I A B I L I T Y - F I T N E S S C O R R E L A T I O N S 317

© 2004 Blackwell Publishing Ltd, Molecular Ecology, 14, 311–323

318 D . L E S B A R R È R E S E T A L . as revealed by nonsignificant HEST — population and HEST — temperature treatment interactions (Table 2), a nearly significant HEST × food interaction was observed (Table 2). Separate analyses for restricted and ad libitum food treatments revealed that increased HEST enhanced survival probability only in restricted food treatment (F1,371 = 7.83; P = 0.005) but not in ad libitum food treatment (F1,370 = 0.01; P = 0.91). In restricted food treatments, 11 out of 12 of the associations between survival and HEST were positive (Fig. 2). Regarding relatedness (rxy), the probability to survive until metamorphosis was higher for tadpoles with less related parents (Table 2). In the case of a significant rxy × temperature × food interaction, further analyses revealed that it was the result of the most stressful treatment (low temperature and restricted food) where an increase in parental relatedness decreased offspring survival probability (F1,131 = 5.51; P = 0.02, nonsignificant after Bonferroni correction). A significant four-way interaction was also observed between all the factors resulting from the stronger association between rxy and survival in Uppsala tadpoles in the most stressful treatment (low temperature and restricted food; F1,17 = 6.15; P = 0.024, nonsignificant 2 , after Bonferroni correction; Fig. 2c). In the case of d EST 2 2 both significant d EST × temperature and d EST × food interactions were observed (Table 2). Separate analyses for different temperature treatments revealed that increased 2 enhanced survival probability in low temperature d EST treatment (F1,263 = 7.9; P = 0.001), had no effect in medium temperature treatment (F1,250 = 0.06; P = 0.81) and decreased survival in warm temperature treatments (F1,226 = 4.09; P = 0.044, nonsignificant after Bonferroni correction). For the different food treatments, separate analyses revealed 2 that d EST enhanced survival probability in restricted food treatment (F1,371 = 7.83; P = 0.004) but not in ad libitum food treatment (F1,370 = 0.01; P = 0.91). In the case of significant 2 × populations × food interaction, further analyses d EST showed that it was because of two particular populations in restricted food treatment (Kilpisjärvi, F1,90 = 4.52, P = 0.036, nonsignificant after Bonferroni correction; Uppsala, F1,50 = 4.63, P = 0.036, nonsignificant after Bonferroni correction; 2 Fig. 2b), where d EST enhanced survival probability.

Genetic variability and developmental rate 2 While d EST was significantly and positively associated with developmental rate (i.e. the higher the genetic variability, the longer the development; Table 2), HEST was only marginally associated and relatedness was not associated with developmental rate (Table 2). Significant three–way interactions were observed with all the genetic variability estimators (Table 2). Separate analyses revealed that the significant HEST × population × temperature interaction resulted from the deviating behaviour of southernmost Lund tadpoles in cold temperature treatment (F1,58 = 9.19;

P = 0.004) and that the significant HEST × temperature × food interaction resulted from the medium temperature and ad libitum food treatment (F1,113 = 6.25; P = 0.005, nonsignificant after Bonferroni correction). Likewise, the 2 significant d EST × population × temperature interaction was resulting from the deviating behaviour of northernmost Kilpisjärvi tadpoles in cold temperature treatment (F1,60 = 5.67; P = 0.021, nonsignificant after Bonferroni correction). In the case of a significant rxy × population × temperature interaction, further analyses revealed that it was the result of one particular population in medium temperature treatment whose longer developmental rate was marginally associated with relatedness (Lund: F1,57 = 2.58; P = 0.10).

Genetic variability and growth rate 2 ) was significantly One estimator of genetic variability ( d EST associated with the growth rate of the tadpoles (increased genetic variability decreased the growth rate, Table 2). Inter2 2 actions with other factors such as d EST × population, d EST × 2 temperature, d EST × population × temperature and rxy × population × food were analysed separately. The significant 2 × population interaction was brought about by one pard EST ticular population (Umeå: F1,253 = 2.84; P = 0.093, non2 significant after Bonferroni correction). The significant d EST × temperature interaction was the result of one particular temperature treatment (warm: F1,213 = 4.51; P = 0.035, non2 significant after Bonferroni correction). The significant d EST × population × temperature interaction was because of the deviating behaviour of Umeå tadpoles in cold temperature treatment (F 1,92 = 3.03; P = 0.085, nonsignificant after Bonferroni correction), and the significant rxy × population × food interaction was resulting from the deviating behaviour of Lund tadpoles in ad libitum food treatment (F1,86 = 5.21; P = 0.025, nonsignificant after Bonferroni correction).

Locus-specific effects Considering individual loci, the same trends were observed for each significant correlation between the traits and all genetic variability measures (e.g. survival was positively associated with the heterozygosity of each individual locus). The genetic variability of only two loci (RRD590 and Rt2Ca25) was significantly correlated with survival, but these did not remain significant after applying sequential Bonferroni method. Similarly, two loci (Rt2Ca25 and Rtempµ4) exhibited significant association between genetic variability and developmental rate. Despite being nonsignificant after correction for multiple tests, these correlations were in the direction of the multilocus genetic variability (i.e. the higher the genetic variability, the longer the development). There was no significant association between genetic variability and growth rate at any single locus. However, there was again a sign-similarity between © 2004 Blackwell Publishing Ltd, Molecular Ecology, 14, 311–323

G E N E T I C V A R I A B I L I T Y - F I T N E S S C O R R E L A T I O N S 319 Table 3 Effect size for the association between measure of genetic variation and three fitness traits. Genetic variability estimators are 2 heterozygosity (HEST), d2 ( d EST ) and relatedness (rxy) Fitness trait

H EST

dEST

rxy

Survival Development rate Growth rate

0.0918* 0.0686˚ 0.0417

0.0793˚ 0.0843* 0.0968*

0.1154** 0.061 0.0472

˚P < 0.1, *P < 0.05, **P < 0.01.

single locus and multilocus correlations at the three genetic variability estimators. If GFCs are caused by inbreeding depression, then locusspecific heterozygosity must be positively correlated across the genome. Therefore, we compared heterozygosity at four randomly selected loci with heterozygosity at the remaining four loci for each pair of parents within each population. Repeating this procedure 10 times, we observed positive significant associations (P ≤ 0.05) in all the combinations (except one for Lund population) for all the populations (r2 = 0.1– 0.15, 0.07– 0.12, 0.18 – 0.26 and 0.08– 0.13 for Kilpisjärvi, Umeå, Uppsala and Lund populations, respectively).

Effect size of the GFCs We observed a relatively small but significant effect size for the association between survival and all the genetic variability estimators, albeit marginally nonsignificantly 2 with d EST (Table 3). However, for both developmental and growth rates, there was a small but significant effect size 2 (Table 3). with d EST

Discussion A noteworthy finding of this paper was significant positive correlations observed between different genetic variability estimates and individual survival probability. Our analyses revealed also that the levels of expression of GFCs were highly environment dependent, the positive effect of genetic variability on survival being more obvious under more stressful (restricted food) conditions. Interestingly, correlations between survival probability and genetic variability were similar across different populations, and the usage of different measures of genetic variability (heterozygosity and d2) gave qualitatively concordant results. The observed positive correlation between survival probability and genetic variability is intriguing as our results do not suffer from a small sample (and effect) size typical for GFCs studies (Coltman & Slate 2003; Reed & Frankham 2003). The weight of our GFCs (47.51) is higher than any of those reported by Reed & Frankham (2003). Furthermore, © 2004 Blackwell Publishing Ltd, Molecular Ecology, 14, 311–323

the effect sizes of the different associations between genetic variability measures and phenotypic traits (Table 3) fall in the upper range of the estimates reviewed by Coltman & Slate (2003). In addition, it is worth noting that the fitness data are based on at least 10 times more individuals than in any previous study. Hence, considering the statistically powerful nature of general linear mixed model analyses, as well as the robustness of the results, it is unlikely that the association between genetic variability and survival probability was observed as a result of chance. In line with observations of a few earlier studies (Danzmann et al. 1988; Scott & Koehn 1990; Borsa et al. 1992; Audo & Diehl 1995), we also observed significant interactions involving genetic variability measures and different treatment effects for some of the traits. In fact, significant three-way interactions between population, environmental treatment and genetic variability estimates revealed that detection of effect of genetic variability can be both environment and population dependent. However, as the interaction involving population effects were not significant after multiple-test corrections, caution is needed in interpretation of these effects. As to the significantly stronger association between genetic variability and fitness traits in some populations, our results suggest population dependence of GFCs for some traits. This is understandable in light of the fact that these populations are genetically divergent both in microsatellite allele frequencies (Palo et al. 2003; Palo et al. 2004) and in genes coding quantitative traits (Palo et al. 2003). Further, genetic variability in presumably neutral microsatellite loci in these populations declines with increasing of latitude, suggesting differences in long-term effective population size (Palo et al. 2003, 2004). Hence, the genetic background and levels of inbreeding are likely to differ between populations, as is the effect of inbreeding on different traits. In general, life-history traits should exhibit high levels of dominance variance and should therefore be more strongly affected by inbreeding depression than more weakly selected morphometric traits (Roff 1997; Merilä & Sheldon 1999). In line with this, the GFCs were clearer for the most important fitness trait (survival) than for traits less closely associated with fitness (growth and development rate). Our results also evidenced the environmental dependency of GFCs. In fact, local adaptation and countergradient variation might have differently shaped the responses of the populations under investigation (Laugen et al. 2003b; Palo et al. 2003). In contrast with most other studies typically looking at few populations and only one set of alternative environmental conditions, the extensive factorial design allowed us to detect the environmental dependency of GFCs. In a previous study, Scott & Koehn (1990) observed that the combined effects of two stress variables (fluctuating salinity and temperature) negated the advantage of heterozygosity perhaps by producing a less stressful environment

320 D . L E S B A R R È R E S E T A L . than either factor alone. Here, we observed the strongest GFCs under restricted food and/or low temperature (14 °C) treatments known to be stressful for the Scandinavian common frogs (Merilä et al. 2000a; Laugen et al. 2003b; Merilä et al. 2004). This suggests that GFCs are more readily detected under stressful than nonstressful environmental conditions, and might partly explain why earlier amphibian studies on GFC (McAlpine 1993; Wright & Guttman 1995; Rowe & Beebee 2001b; but see Rowe et al. 1999) have failed to find these associations. Interestingly, although GFCs for developmental and growth rates were environment and population specific — and also depended on the genetic metric used — the significant associations were to somewhat unexpected directions. 2 ), the For instance, the higher the genetic variability ( d EST lower the growth rate and the longer the development. While one would a priori expect the opposite relationships, with regards to the high variance usually associated with mean d2 (discussed below), these results may be explainable by the fact that phenotypic correlation between developmental rate and size at metamorphosis is usually negative. In other words, the longer the development, the larger the size at metamorphosis. Hence, if high genome-wide (see below) heterozygosity leads to delayed development, it will indirectly increase the size at metamorphosis. Size at metamorphosis on the other hand is usually under positive directional selection in amphibians (review in Altwegg & Reyer 2003). While the lack of environmental and population dependency of GFCs for survival might be explained by the ultimate consequence of this trait, fitness traits such as developmental and growth rates are known to be plastic and subject to local adaptation (Laugen et al. 2003b; Palo et al. 2003). Hence, differences in local selection pressures might explain the observed heterogeneity of GFCs.

GFCs result from a general effect? One main unresolved problem in most GFC studies is the question about which of the three main hypotheses (see Introduction) best explains the observed correlations (Hansson & Westerberg 2002). In this paper, the observed positive association between survival probability and genetic variability appears to be explainable by the general effect hypothesis. First, it is unlikely that the apparent heterozygote advantage at the used markers result of effects of homozygosity at closely-linked fitness loci because survival in the different treatment conditions is a highly important life history trait per se, involving probably a large number of loci. Moreover, analyses performed with individual loci gave the same picture with no discrepancies between the sign of the GFCs at each locus and the multilocus GFCs. Therefore the local effect hypothesis is an unlikely explanation for the GFC observed in Rana temporaria populations. Second, microsatellite markers are supposed

to be less informative to detect a direct effect hypothesis because they are considered selectively neutral (Queller et al. 1993; Jarne & Lagoda 1996) and presumably not displaying functional overdominance. Additionally, we found significant positive associations between heterozygosity at two randomly chosen set of markers and little support for linkage disequilibrium (one population, two loci), giving some indication that marker heterozygosity reflects a genome-wide metric and is unlikely to be a consequence of selection on fitness loci in the local chromosomal vicinity of the markers (Hansson et al. 2004). We propose that the GFCs observed for survival (with multilocus heterozygosity, mean d2 and the relatedness of the parents) and in age at metamorphosis (with the multilocus heterozygosity and mean d2) most likely result from effects of homozygosity at genome-wide distributed fitness loci. While the results provide support for the general effect hypothesis (Hansson & Westerberg 2002), a higher resolution approach (along with finer analytical methods and knowledge of the actual genetic architecture of the traits) would be required to firmly discard the local effect hypothesis.

Use of d2 and heterozygosity in capturing GFCs The question regarding which of the measures of genetic diversity — d2 or heterozygosity — is a better indicator of the level of inbreeding has been the focus of recent discussions (Hedrick et al. 2001; Tsitrone et al. 2001; Goudet & Keller 2002). Unlike in previous studies, d2 performed relatively well in our case, even if survival probability was less 2 closely associated with d EST than with HEST. The complexity of the results for developmental rate and growth rate (although d2 is significantly associated with these two traits) is of little help to discard d2 as a genetic variability estimator to detect GFCs in these two traits. Although both heterozygosity and d2 should be informative, Coulson et al. (1998) suggested that the latter contains both the information from heterozygosity as well as additional information on genetic divergence between the parent individuals that may indicate the extent of outbreeding. Several significant interactions between treatment effects and d2 may suggest that d2 is revealing something that is not captured by heterozygosity. However, these interactions were in the same direction with both heterozygosity and d2. The same holds for both genetic estimators as main factors. Moreover, the significant correlations between the two estimators are as high as those often reported (r = 0.391 and r = 0.447 for observed and estimated coefficients, respectively; Coltman et al. 1998; Coulson et al. 1998; Rowe & Beebee 2001b; Borrell et al. 2004) suggesting that they indeed are alike, and that the differences in level of significance might be the result of chance. The high sample size and the intrinsically high variance of d2 might have helped in that direction. On the other hand, it is worth mentioning that across all 72 tests © 2004 Blackwell Publishing Ltd, Molecular Ecology, 14, 311–323

G E N E T I C V A R I A B I L I T Y - F I T N E S S C O R R E L A T I O N S 321 (three traits × four populations × three temperatures × two food treatments), mean d2 revealed as many positive as negative associations whereas heterozygosity and relatedness both returned more associations in the expected direction. While it could suggest that mean d2 is a less robust measure, the results cannot confirm both theoretical (Tsitrone et al. 2001) and empirical (Hedrick et al. 2001) work suggesting that heterozygosity estimate may be a better metric than d2 for capturing inbreeding depression most of the time.

Conclusions Our results revealed a general trend for a positive association between genetic variability and the survival probability of the tadpoles from different populations under different treatment conditions. Furthermore, the significant interactions between genetic variability estimators and other factors in the models with developmental rate and growth rate highlight the complexity of the GFCs. There are biological reasons to expect such heterogeneity. For instance, different populations have different demographic histories and are thus expected to have undergone different levels of inbreeding and differ in their genetic architectures (i.e. the number, magnitude, and interactions of genetic factors that contribute to variation in a trait). At any rate, our results give support for the ‘general effect hypothesis’, i.e. that the genetic variability in microsatellite loci is indicative of genome-wide heterozygosity, which is apparently positively correlated with survival probability of tadpoles in a wide range of environmental conditions. To this end, the results support the contention that life history traits — such as survival — are the best candidates to detect GFCs (David 1998; Coltman & Slate 2003).

Acknowledgements We thank Satu Karttunen, Niclas Kolm, Ane Timenes Laugen and Katja Räsänen for help in laboratory, as well as Trenton W. J. Garner for providing information about unpublished microsatellite primers. Cano Arias and Pierre-Alexandre Landry helped at various stages of the analyses, and Kilpisjärvi Biological Station provided support in obtaining the samples from the north. Our research was funded by CIMO (DL), the Swedish Research Council (AL) and the Academy of Finland (CP, JM).

References Allendorf FW, Leary RF (1986) Heterozygosity and fitness in natural populations of animals. In: Conservation Biology (ed. Soule ME), pp. 57–76. Sinauer, Sunderland, M.A. Altwegg R, Reyer H-U (2003) Patterns of natural selection on size at metamorphosis in water frogs. Evolution, 57, 872– 882. Audo MC, Diehl WJ (1995) Effect of quantity and quality of environmental stress on multilocus heterozygosity — growth relationships in the earthworm, Eisenia fetida. Heredity, 75, 98–105. © 2004 Blackwell Publishing Ltd, Molecular Ecology, 14, 311–323

Berlin S, Merilä J, Ellegren H (2000) Isolation and characterization of polymorphic microsatellite loci in the common frog, Rana temporaria. Molecular Ecology, 9, 1938–1939. Borrell YJ, Pineda H, McCarthy I, Vázquez E, Sánchez JA, Lizana GB (2004) Correlations between fitness and heterozygosity at allozyme and microsatellite loci in the Atlantic salmon, Salmo salar L. Heredity, 92, 585–593. Borsa P, Jousselin Y, Delay B (1992) Relationships between allozymic heterozygosity, body size, and survival to natural anoxic stress in the palourde Ruditapes decussatus L. (Bivalvia: Veneridae). Journal of Experimental Marine Biology and Ecology, 155, 169–181. Britten HB (1996) Meta-analyses of the association between multilocus heterozygosity and fitness. Evolution, 50, 2158– 2164. Coltman DW, Bowen WD, Wright JM (1998) Birth weight and neonatal survival of harbour seal pups are positively correlated with genetic variation measured by microsatellites. Proceedings of the Royal Society of London B Biology Sciences, 265, 803–809. Coltman DW, Slate J (2003) Microsatellite measures of inbreeding: a meta-analysis. Evolution, 57, 971–983. Coulson TN, Pemberton JM, Albon SD, et al. (1998) Microsatellites reveal heterosis in red deer. Proceedings of the Royal Society of London B Biology Sciences, 265, 489–495. Crnokrak P, Roff DA (1999) Inbreeding depression in the wild. Heredity, 83, 260–270. Danzmann RG, Ferguson MM, Allendorf FW (1988) Heterozygosity and components of fitness in a strain of rainbow trout. Biology Journal of the Linnean Society, 33, 285–304. David P (1998) Heterozygosity–fitness correlations: new perspectives on old problems. Heredity, 80, 531–537. DeRose MA, Roff DA (1999) A comparison of inbreeding depression in life history and morphological traits in mammals. Evolution, 53, 1288–1292. Fog K, Schmedes A, de Lasson DR (1997) Nordens Padder Og Krybdyr. G.E.C. Gad, Copenhagen. Frankham R, Ralls K (1998) Inbreeding leads to extinction. Nature, 392, 441–442. Goldstein DB, Linares AR, Cavalli-Sforza LL, Feldman MW (1995) An evaluation of genetic distances for use with microsatellite loci. Genetics, 139, 463–471. Goodnight KF, Queller DC (1999) Computer software for performing likelihood tests of pedigree relationship using genetic markers. Molecular Ecology, 8, 1231–1234. Gosner KL (1960) A simplified table for staging anuran embryos and larvae with notes on identification. Herpetologica, 16, 183 –190. Goudet J (1999) FSTAT, a Program to Estimate and Test Gene Diversities and Fixation Indices. 2.8. Université de Lausanne, Switzerland. Goudet J, Keller L (2002) The correlation between inbreeding and fitness: does allele size matter? Trends in Ecology and Evolution, 17, 201–202. Haapanen A (1970) Site tenacity of the common frog (Rana temporaria L.) and the moor frog (Rana arvalis Nilss.). Annales Zoologici Fennici, 7, 61–66. Hansson B, Westerberg L (2002) On the correlation between heterozygosity and fitness in natural populations. Molecular Ecology, 11, 2467–2474. Hansson B, Westerdahl H, Hasselquist D, Åkesson M, Bensch S (2004) Does linkage disequilibrium generate heterozygosity– fitness correlations in great reed warblers? Evolution, 58, 870 – 879.

322 D . L E S B A R R È R E S E T A L . Hedrick P, Fredrickson R, Ellegren H (2001) Evaluation of d2, a microsatellite measure of inbreeding and outbreeding, in wolves with a known pedigree. Evolution, 55, 1256 –1260. Hedrick PW, Kalinowski ST (2000) Inbreeding depression in conservation biology. Annual Review of Ecology and Systematics, 31, 139–162. Hitchings SP, Beebee TJC (1997) Genetic substructuring as a result of barriers to gene flow in urban Rana temporaria (common frog) populations: implications for biodiversity conservation. Heredity, 79, 117–127. Jarne P, Lagoda PJL (1996) Microsatellites, from molecules to populations and back. Trends in Ecology and Evolution, 11, 424– 429. Keller LF, Waller DM (2002) Inbreeding effects in wild populations. Trends in Ecology and Evolution, 17, 230 –241. Laugen AT, Laurila A, Merilä J (2003a) Latitudinal and temperature-dependent variation in embryonic development and growth in Rana temporaria. Oecologia, 135, 548 – 554. Laugen AT, Laurila A, Räsänen K, Merilä J (2003b) Latitudinal countergradient variation in the common frog (Rana temporaria) developmental rates — evidence for local adaptation. Journal of Evolutionary Biology, 16, 996 –1005. Laurila A (2000) Responses to predator chemical cues and local variation in antipredator behaviour of Rana temporaria tadpoles. Oikos, 88, 159–168. Laurila A, Karttunen S, Merilä J (2002) Adaptive phenotypic plasticity and genetics of larval life histories in two Rana temporaria populations. Evolution, 56, 617– 627. Littell RC, Milliken GA, Stroup WW, Wolfinger RD (1996) SAS System for Mixed Models. SAS Institute Inc, Cary, N.C. Loman J (2001) Inventering Av Vanlig Groda Och Åkergroda I Skåne 2001. Lund University Press, Lund, Sweden. Loman J (2002) Microevolution and maternal effects on tadpole Rana temporaria growth and development rate. Journal of Zoology (London), 257, 93– 99. Lynch M, Gabriel W (1990) Mutation load and the survival of small populations. Evolution, 44, 1725 –1737. Lynch M, Walsh B (1998) Genetics and Analysis of Quantitative Traits. Sinauer Associates, Sunderland, M.A. McAlpine S (1993) Genetic heterozygosity and reproductive success in the green treefrog, Hyla cinerea. Heredity, 70, 553–558. McCullagh P, Nelder JA (1989) Generalized Linear Models, 2nd edn Chapman & Hall, London. Mcelroy TC, Diehl WJ (2001) Heterosis in two closely related species of earthworm (Eisenia fetida and E. andrei). Heredity, 87, 598–608. Merilä J, Kruuk LEB, Sheldon BC (2001) Cryptic evolution in a wild bird population. Nature, 412, 76 –79. Merilä J, Laurila A, Laugen AT, Räsänen K (2004) Heads or tails? Variation in tadpole body proportions in response to temperature and food stress. Evolutionary Ecology Research, 6, 727–738. Merilä J, Laurila A, Timenes Laugen A, Räsänen K, Pahkala M (2000a) Plasticity in age and size at metamorphosis in Rana temporaria — comparison of high and low latitude populations. Ecography, 23, 457– 465. Merilä J, Pahkala M, Johanson U (2000b) A frog perspective on increased ultraviolet-B radiation, climate change and latitudinal adaptation. Annales Zoologici Fennici, 37, 129 –134. Merilä J, Sheldon BC (1999) Genetic architecture of fitness and nonfitness traits: empirical patterns and development of ideas. Heredity, 83, 103–109.

Miaud C, Guyétant R, Elmberg J (1999) Variations in life history traits in the common frog Rana temporaria (Amphibia: Anura): a literature review and new data from the French Alps. Journal of Zoolgy, London, 249, 61–73. Miaud C, Merilä J (2001) Local adaptation or environmental induction? Causes of population differentiation in Alpine amphibians. Biota, 2, 31–50. Mitton JB (1997) Selection in Natural Populations. Oxford University Press, Oxford, U.K. Pahkala M, Laurila A, Merilä J (2002) Effects of ultraviolet-B radiation on common frog, Rana temporaria from along a latitudinal gradient. Oecologia, 133, 458–465. Pakkasmaa S, Merilä J, O’Hara RB (2003) Genetic and maternal effect influences on viability of common frog tadpoles under different environmental conditions. Heredity, 91, 117–124. Palo J, O’Hara RB, Laugen AT, Laurila A, Primmer CR, Merilä J (2003) Latitudinal divergence of common frog (Rana temporaria) life history traits by natural selection: evidence from a comparison of molecular and quantitative genetic data. Molecular Ecology, 12, 1963–1978. Palo J, Schmeller D, Laurila A, Primmer CR, Kuzmin SL, Merilä J (2004) High degree of population subdivision in a widespread amphibian. Molecular Ecology, 13, 2631–2644. Pemberton JM, Coltman DW, Coulson TN, Slate J (1999) Using microsatellites to measure the fitness consequences of inbreeding and outbreeding. In: Microsatellites: Evolution and Applications (eds Goldstein B, Schlötterer C), pp. 151–164. Oxford University Press, New York. Pidancier N, Gauthier P, Miquel C, Pompanon F (2002) Polymorphic microsatellite DNA loci identified in the common frog (Rana temporaria, Amphibia, Ranidae). Molecular Ecology Notes, 2, 304–305. Primmer CR, Landry P-A, Ranta E, et al. (2003) Prediction of offspring fitness based on parental genetic diversity in endangered salmonid populations. Journal of Fish Biology, 63, 909–927. Queller DC, Goodnight KF (1989) Estimating relatedness using genetic markers. Evolution, 43, 258–275. Queller DC, Strassmann JE, Hughes CR (1993) Microsatellites and kinship. Trends in Ecology and Evolution, 8, 285–288. Räsänen K, Laurila A, Merilä J (2002) Carry-over effects of embryonic acid conditions on development and growth of Rana temporaria tadpoles. Freshwater Biology, 47, 19–30. Raymond M, Rousset F (1995) genepop (v. 1.2): population genetics software for exact tests and ecumenism. Journal of Heredity, 83, 239. Reed DH, Frankham R (2001) How closely correlated are molecular and quantitative measures of genetic variation? A meta-analysis. Evolution, 55, 1095–1103. Reed DH, Frankham R (2003) Correlation between fitness and genetic diversity. Conservation Biology, 17, 230–237. Rice WR (1989) Analysing tables of statistical tests. Evolution, 43, 223–225. Roff DA (1997) Evolutionary Quantitative Genetics. Chapman & Hall, New York. Rosenthal R (1991) Meta-Analytic Procedures for Social Research. Sage Publications, Newbury Park, CA. Rowe G, Beebee TJC (2001a) Polymerase chain reaction primers for microsatellite loci in the common frog Rana temporaria. Molecular Ecology Notes, 10, 6–7. Rowe G, Beebee TJC (2001b) Fitness and microsatellite diversity estimates were not correlated in two outbred anuran populations. Heredity, 87, 558–565.

© 2004 Blackwell Publishing Ltd, Molecular Ecology, 14, 311–323

G E N E T I C V A R I A B I L I T Y - F I T N E S S C O R R E L A T I O N S 323 Rowe G, Beebee TJC, Burke T (1999) Microsatellite heterozygosity, fitness and demography in natterjack toads, Bufo calamita. Animal Conservation, 2, 85 – 92. Scott TM, Koehn RK (1990) The effect of environmental stress on the relationship of heterozygosity to growth rate in the coot clam Mulinia lateralis (Say). Journal of Experimental Marine Biology and Ecology, 135, 109 –116. Sokal RR, Rohlf FJ (1995) Biometry. The Principles and Practice of Statistics in Biological Research. Freeman W.H. & Co, New York. Tsitrone A, Rousset F, David P (2001) Heterosis, marker mutational processes and population inbreeding history. Genetics, 159, 1845–1859. Valdés AM, Slatkin M, Freimer NB (1993) Allele frequencies at microsatellite loci: the stepwise mutation model revisited. Genetics, 159, 1845–1859.

© 2004 Blackwell Publishing Ltd, Molecular Ecology, 14, 311–323

Vos CC, Antonisse-De Jong AG, Goedhart PW, Smulders MJM (2001) Genetic similarity as a measure for connectivity between fragmented populations of the moor frog (Rana arvalis). Heredity, 86, 598–608. Wright S (1922) The effects of inbreeding and crossbreeding on guinea pigs. III. Crosses between highly inbred families. US Department of Agriculture Bulletin, 1121, 1–60. Wright MF, Guttman SI (1995) Lack of an association between heterozygosity and growth rate in the wood frog, Rana sylvatica. Canadian Journal of Zoology, 73, 569–575.

This study is the result of a collaboration between researchers interested in utilising molecular methods to study population history and microevolutionary processes in wild vertebrates.