Multilocus heterozygosity, parental relatedness and individual fitness components in a wild mountain goat, Oreamnos americanus population

Molecular Ecology (2009) 18, 2297–2306 doi: 10.1111/j.1365-294X.2009.04197.x

© 2009 Blackwell Publishing Ltd

Blackwell Publishing LtdMultilocus heterozygosity, parental relatedness and individual fitness components in a wild mountain goat, Oreamnos americanus population

JULIEN MAINGUY,* STEEVE D. CÔTÉ * and DAVID W. COLTMAN†*Département de biologie et Centre d’études nordiques, Université Laval, Québec, QC, Canada G1V 0A6, †Department of Biological Sciences, University of Alberta, Edmonton, AB, Canada T6G 2E9

Abstract

Matings between relatives lead to a decrease in offspring genetic diversity which can reducefitness, a phenomenon known as inbreeding depression. Because alpine ungulates gener-ally live in small structured populations and often exhibit a polygynous mating system,they are susceptible to inbreeding. Here, we used marker-based measures of pairwisegenetic relatedness and inbreeding to investigate the fitness consequences of matingsbetween relatives in a long-term study population of mountain goats (Oreamnos americanus)at Caw Ridge, Alberta, Canada. We first assessed whether individuals avoided mating withkin by comparing actual and random mating pairs according to their estimated geneticrelatedness, which was derived from 25 unlinked polymorphic microsatellite markers andreflected pedigree relatedness. We then examined whether individual multilocus heterozygosityH, used as a measure of inbreeding, was predicted by parental relatedness and associatedwith yearling survival and the annual probability of giving birth to a kid in adult females.Breeding pairs identified by genetic parentage analyses of offspring that survived to 1 year ofage were less genetically related than expected under random matings. Parental relatednesswas negatively correlated with offspring H, and more heterozygous yearlings had highersurvival to 2 years of age. The probability of giving birth was not affected by H in adultfemales. Because kids that survived to yearling age were mainly produced by less geneticallyrelated parents, our results suggest that some individuals experienced inbreeding depressionin early life. Future research will be required to quantify the levels of gene flow betweendifferent herds, and evaluate their effects on population genetic diversity and dynamics.

Keywords: heterozygosity–fitness correlations, inbreeding, mate choice, pairwise genetic related-ness, ungulates, yearling survival

Received 23 August 2008; revision received 14 March 2009; accepted 17 March 2009

Introduction

Matings between relatives can lead to a decline in offspringfitness through the increased expression of homozygousdeleterious recessive alleles inherited from a commonancestor, a phenomenon known as inbreeding depression(Charlesworth & Charlesworth 1987). Inbred offspringmay exhibit lowered fitness as a result of both the fixationof partially recessive alleles in inbred lines (partial dominance

hypothesis) and a heterozygote advantage at fitness-linkedloci favouring outbred offspring (overdominance hypothesis;Charlesworth & Charlesworth 1999). The detrimentaleffects of inbreeding on fitness have been studied forover a century, and reported in many captive species andan increasing number of wild populations (Hedrick &Kalinowski 2000; Keller & Waller 2002). Thus, inbreedingdepression is now regarded as a widespread phenomenonthat can negatively impact individual fitness and thedemography of populations (Spielman et al. 2004).

The avoidance of inbreeding is expected to occur throughdifferent mechanisms within a population (Pusey & Wolf1996), such as dispersal (Rosenfield & Bielefeldt 1992;Bollinger et al. 1993) or kin recognition (Gerlach & Lysiak

Correspondence: Julien Mainguy, Ministère des Ressourcesnaturelles et de la Faune, Direction de l’expertise sur la faune et seshabitats, 880, Chemin Ste-Foy, 2e étage, Québec, QC, Canada G1S4X4. Fax: 418-646-6863; E-mail: [email protected]

2298 J . M A I N G U Y, S . D . C Ô T É and D . W. C O LT M A N

© 2009 Blackwell Publishing Ltd

2006; Hoffman et al. 2007). For instance, kin recognitionmay occur when dispersal does not completely eliminate therisk of mating with relatives (Archie et al. 2007). In somewild populations, however, individuals do not appearto avoid inbreeding (Duarte et al. 2003; Lane et al. 2007)despite sometimes apparent signs of inbreeding depressionin offspring (Keller & Arcese 1998; Hansson et al. 2007).This pattern could possibly be the result of a limited abilityto discriminate between more or less related partners, orbecause the costs of inbreeding are potentially outweighedby other benefits, such as increased fitness through thespread of the same alleles (Kokko & Ots 2006).

Studying inbreeding in the wild is not trivial, as it requiresdetailed pedigree information over at least a few generationsto determine inbreeding levels within individuals throughthe calculation of inbreeding coefficients F (Keller 1998;Overall et al. 2005). When a pedigree is lacking or incomplete,an alternative is to use marker-based measures of geneticdiversity such as multilocus heterozygosity H and otherderived metrics to infer inbreeding (Coulson et al. 1998;Amos et al. 2001). Depending on the demographic historyand mating system of a population, however, the correlationbetween H and F can vary, but is generally weak (Ballouxet al. 2004; Markert et al. 2004; Slate et al. 2004). Our capacityto detect the detrimental effects of inbreeding throughmarker-based methods can thus be limited and, whensignificant effects are found, they are also generally weak(reviewed in Coltman & Slate 2003). Despite these limita-tions, new marker-based methods are still being developedto estimate levels of inbreeding (e.g. Aparicio et al. 2006), ashomozygosity at a set of neutral markers partly reflectsgenome-wide homozygosity and, therefore, represents asurrogate of inbreeding levels (Aparicio et al. 2007). Inaddition, as inbreeding depression is often more severe inthe wild than in captivity (Crnokrak & Roff 1999), generallydue to the harsher environmental conditions occurring innature (Jiménez et al. 1994; Halverson et al. 2006), studiesof wild populations can enhance our capacity to detectinbreeding depression with H (Jensen et al. 2007).

Detection of heterozygosity-fitness correlations (HFC)using ‘neutral’ loci such as microsatellites does not necessarilyimply that inbreeding depression is the sole underlyingmechanism originating from a ‘genome-wide’ effect(reviewed by Hansson & Westerberg 2002). This is becauseHFC may also be generated by linkage disequilibriumbetween some of the neutral markers and functional genes(‘local’ effect; Hansson & Westerberg 2002). Evidence forthis second mechanism has been reported in some wildpopulations (Hansson et al. 2004; von Hardenberg et al. 2007),but some HFC are still best explained by genome-widerather than local effects (Hoffman et al. 2004; Charpentieret al. 2005; Lesbarrères et al. 2005). Thus, there is likely aspectrum of explanations for HFC in nature (Balloux et al.2004), with genome-wide effects having nonetheless a local

cause, that is, functional genes of varying positive andnegative effects (Lieutnant-Gosselin & Bernatchez 2006).In populations subjected to inbreeding, heterozygosity atcoding loci is expected to be positively correlated to thatfound at neutral markers (Aparicio et al. 2007).

The mountain goat (Oreamnos americanus) is an alpineungulate generally found in small populations withapparently limited but unknown levels of gene flowbetween them (Côté & Festa-Bianchet 2003). Therefore,mountain goats are at risk of inbreeding because they arethought to exhibit strong population structure such as inother alpine ungulates (e.g. Amills et al. 2004; Worley et al.2006). The polygynous mating system of mountain goats(Mainguy et al. 2008) may also exacerbate the risks ofinbreeding, as only a few males achieve high reproductivesuccess each year (Mainguy 2008). This may explain whymountain goats exhibit low levels of genetic diversity atneutral (Mainguy et al. 2005) and even some functionalgenes (Mainguy et al. 2007). Altogether, the characteristicsof this alpine ungulate suggest that genetic diversity at aset of neutral markers should correlate with F (Balloux et al.2004). Additionally, the occasional genetic contributionof immigrant males (Mainguy et al., unpublished data)combined with close inbreeding may increase the variancein inbreeding and favour genetic diversity levels that canpromote the detection of HFC (Da Silva et al. 2006).

Here we examined whether actual mating pairs deter-mined from genetic parentage analyses in a long-term studyof mountain goats were less genetically related than whatwould be expected under random matings. Among matingpairs, we further tested whether age and social rank of femalesplayed a role in the degree of relatedness to their matingpartner, expecting more experienced and higher-rankingfemales to be less related to their mate than younger andsubordinate females. To assess the effect of inbred matings, weexamined the relationship between parental marker-basedrelatedness and offspring genetic diversity (H), and the effectof individual genetic diversity on two life-history traits:yearling survival and the annual probability of giving birthin adult (≥ 3 years old) females. We predicted that mostmatings should occur between less-related individualsbecause of the generally negative effects of inbreeding (Keller& Waller 2002). Finally, assuming that H would partiallyreflect F in our study population, we predicted that yearlingsurvival and the probability of giving birth to a kid infemales should decrease with decreasing H under a genome-wide effect, while also testing for potential local effects.

Materials and methods

Study area

Our study was conducted at Caw Ridge (54°N, 119°W),west-central Alberta, Canada, in the front range of the

I N B R E E D I N G A N D F I T N E S S I N M O U N TA I N GO AT S 2299

© 2009 Blackwell Publishing Ltd

Rocky Mountains. The native mountain goat population ofCaw Ridge uses about 28 km2 of alpine grassy slopes, shortcliffs, and open subalpine forest that range from 1750 to2170 m in altitude. The population, studied intensively since1989, has ranged from 81 to 159 individuals, with ≥ 98% ofindividuals aged ≥ 1 year old marked since 1993. The CawRidge population is one of the largest in Alberta and isalso geographically isolated from other mountain goatpopulations, the closest large herd being located about 20 kmto the southeast on Mount Hammel (Hamel et al. 2006). A fewimmigrant males are however observed each year on CawRidge, and successful emigration to other nearby populationsby males born at Caw Ridge has also been documented(Festa-Bianchet & Côté 2008). Further description of thestudy area can be found in Festa-Bianchet & Côté (2008).

Captures, measurements and genetic sampling

Goats were captured from late May to mid-September1986–2007 (n = 756 captures of 398 individuals) in remotelycontrolled box traps and self-tripping nylon mesh Clovertraps baited with salt (Côté et al. 1998a). All individuals weremarked with plastic ear tags, visual or radiocollars. Since1994, an ear tissue sample was also taken for DNA analysesusing a punch during marking. Ear tissues were kept in1.5 mL tubes filled a solution of 20% dimethyl sulfoxidesaturated with sodium chloride at –20 °C. Goats that wereadults (≥ 3 years old) at their first capture were agedaccording to the number of horn annuli (Côté et al. 1998b).Goats were weighed to the nearest 0.5 kg with a spring scale.Many goats were also weighed without handling (n = 1331masses from 172 individuals aged ≥ 1 year during 2001–2007) using one to three electronic platform scales baitedwith salt (Bassano et al. 2003). All body masses were adjustedto 15 July according to the sex-specific growth rate of threeage classes (yearlings, 2-year-olds and adults; Côté et al. 1998a;Mainguy & Côté 2008). Because the capture of a female witha kid at heel increases the risk of kid abandonment (Côtéet al. 1998a), we did not capture kids starting in 1998 andtherefore few data were available for the study of neonatalfitness traits in relation with genetic diversity. Côté et al.(1998a, b) provide further details of capture procedures.

Field observations

We used spotting scopes (15–45×) to sample goat behaviourat distances generally ranging from 200 to 700 m.Observations were conducted almost daily from May toSeptember. In each year, we searched the study areaintensively from mid-May to early June and attempted to findas many adult females as possible each day to determinewhich ones had given birth (Côté & Festa-Bianchet 2001a).Females produce only one kid annually generally startingat 4 or 5 years of age, with very few females primiparous

at age 3 (Côté & Festa-Bianchet 2001b). The proportionof adult females giving birth at Caw Ridge generallyincreases rapidly until 6 years of age, peaks at 8 to 12 years,then declines slightly for females ≥ 13 years, yielding anaverage reproductive lifespan of about 9 to 12 years(Festa-Bianchet & Côté 2008). During summer, we notedthe identity of individuals present and determined whichmarked females were nursing a kid until weaning in mid-September (Côté & Festa-Bianchet 2001a). We measuredoverwinter survival by determining which individualssurvived to 1 June the following year. From 1994 to 2007, wealso used all-occurrences sampling and focal observationsto record agonistic encounters between adult females todetermine social rank (Côté 2000), as dominance statusinfluences kid production in females, especially at a youngage (Côté & Festa-Bianchet 2001b). Individual femaleswere ordered in annual dominance hierarchies using themethodology described in Côté (2000). Because annualmatrix size varied from 38 to 61 adult females, wetransformed social ranks according to the formula 1 – rank/Ni, where Ni is the number of adult females during year i(Côté 2000). Standardized social ranks therefore variedfrom 0 (subordinate) to 1 (dominant). As social rank isstrongly correlated with age in adult females (r > 0.9), theresiduals of standardized rank on age were used as ameasure of social rank (Côté 2000).

Microsatellite genotyping

A total of 296 individuals, or 74% of the individuals markedand monitored since the beginning of the study, were sampledfor genetic analyses. Genomic DNA was extracted fromear tissues with QIAGEN DNeasy extraction kits andthen polymerase chain reaction-amplified using a set of 28polymorphic microsatellite loci that were in Hardy–Weinberg equilibrium: ARO28, BL6, BM1225, BM1818,BM4025, BM4513, BM4630, BM6444, BR3510, HEL10,HUJ616, HUJ1177, ILSTS058, MAF36, MAF64, McM64,McM152, McM527, OarCP26, OarHH35, OarHH62,OarJMP29, OarJMP58, RT9, RT27, TGLA10, TGLA122, andURB038. We followed methods described in Mainguy et al.(2005) and obtained an overall genotyping success of 99.9%.These 28 loci were used for parentage analyses, whereasonly a subsample of 25 loci were used for the estimationof molecular measures of relatedness and inbreeding,as some pairs of loci were in linkage disequilibrium(i.e. ARO28, MAF64 and OarHH62 were removed, seeMainguy et al. 2005).

Parentage analyses

A total of 194 goats had a mother identified throughbehavioural associations in the field (Gendreau et al. 2005). Wetested 111 of these relationships by comparing microsatellite

2300 J . M A I N G U Y, S . D . C Ô T É and D . W. C O LT M A N

© 2009 Blackwell Publishing Ltd

genotypes and verified all but one relationship based onsimple genetic exclusion using the parentage-assignmentsoftware Cervus 3.0 (Kalinowski et al. 2007). Because notall individuals were assigned to a mother, however, weattempted to determine genetic maternity using thelikelihood-based approach implemented in Cervus.Briefly, this programme calculates a logarithm-of-the-odds (LOD) score for each candidate mother based onsimulations of offspring genotypes, and then assigns eachoffspring to the most likely mother at 95% (strict) or 80%(relaxed) statistical confidence. Simulations were carried outfor each year separately to estimate the critical differencein LOD scores for assignments based on the number ofcandidate mothers and the proportion of them sampled inthat year. Input parameters common to all years includedsimulation of 100 000 offspring genotypes based on allelefrequencies of the whole population, a typing error of 0.2%according to the mean observed error rate across loci ofknown mother–offspring pairs, and the overall proportionof loci typed (99.9%). Candidate mothers considered musthave been observed with a kid at heel in the year of birthof the offspring tested, and their kid must have survived toweaning in September. Information about kid sex was alsoused in assigning the mother, as sex determination from fieldobservations was highly accurate (Côté & Festa-Bianchet2001c). Then, to maximize the confidence level of paternityassignments, only kids whose mothers had been genotypedand confirmed with Cervus were tested (n = 191). Based onfield observations conducted during the rut (Mainguy &Côté 2008; Mainguy et al. 2008), only males aged ≥ 3 yearsthat were observed at least once during the year precedingthe kid birth were considered as candidate fathers. Becausea few (three to five) adult immigrant males were alsoobserved to rut on Caw Ridge (Mainguy et al. 2008), weadded five unsampled males in the simulations forpaternal assignments in each year to provide us with aconservative proportion of sampled males. Only paternitiesassigned under strict statistical confidence in Cervus andwith no more than one mismatch in the trio ‘kid-mother-father’ entered the pedigree. Because some individuals(mainly males) were not sampled for DNA in the earlyyears of the study, the pedigree was incomplete and ourability to calculate inbreeding coefficients restricted. Thepedigree of 399 individuals included 278 maternal and 100paternal links (see Mainguy 2008).

Molecular measures of pairwise relatedness

Because the accuracy of relatedness estimators generallyvaries between different data sets (Van De Casteele et al.2001), we tested two commonly used estimators of pairwisegenetic relatedness (Queller & Goodnight 1989; Lynch &Ritland 1999) and a more recent one for structuredpopulations (Oliehoek et al. 2006) to estimate true genetic

relatedness in our study population. The first two estimatorswere computed using SPAGeDi 1.2 (Hardy & Vekemans 2002)and can theoretically vary between –1 and 1, with negativevalues suggesting non-kin and positive values kin. Themost recent estimator (Oliehoek et al. 2006) was computedusing rea 0.2 available at www.geneticdiversity.net/estimators.html and which can vary on a scale from 0 to1 (none to highly related). We determined the ‘minimal’true genetic relatedness among all possible pairs of indi-viduals in the Caw Ridge pedigree (RPED) using cfc 1.0(by M. Sarolzaei, H. Iwaisaki, & J.-J. Colleau; available atwww.agr.niigata-u.ac.jp/~iwsk/cfc.html). Then, we com-pared the three estimators of pairwise genetic relatednessto RPED once individuals with values of 0 were excluded(range of RPED = 0.016 to 0.625; n = 2431 pairwise com-parisons). Based on these comparisons (Table 1), we chosethe estimator of Lynch & Ritland (1999), as it explained themost variance in known relatedness. This is consistent withfindings of Csilléry et al. (2006) who reported that thisestimator performed generally better than others in fivedifferent wild populations, including one alpine ungulate.

Molecular measures of inbreeding

In addition to H, which varied from 0.24 to 0.80 in the CawRidge population (mean ± SD = 0.50 ± 0.09), we computedtwo molecular metrics of inbreeding: internal relatedness IR(Amos et al. 2001) and homozygosity by loci hL (Aparicioet al. 2006). Pairwise comparisons of these three metricsrevealed, however, that H measured the same extent ofgenetic diversity as IR and hL, (all r’s > 0.96). Thus, we onlypresent the results obtained with H.

Data analyses

We examined whether mating pairs that produced a kidsurviving to 1 year of age (i.e. age at first capture) were less

Table 1 Comparisons of the performance of three estimators ofpairwise genetic relatedness in reflecting genetic relatednessdetermined from the pedigree (RPED) in a population of mountaingoats (n = 296 genotyped individuals out of 399) at Caw Ridge,Alberta, Canada. Because the pedigree was incomplete, RPED

sometimes represented a minimal value of true geneticrelatedness. All correlations are significant at P < 0.0001

Estimators of pairwise genetic relatedness

Variance explained (r2) in RPED

Queller & Goodnight (1989) 0.35Lynch & Ritland (1999) 0.42Oliehoek et al. (2006)* 0.30

*The weighted equal drift similarity (WEDS) estimator was applied with a β2 correction (see Oliehoek et al. 2006).

I N B R E E D I N G A N D F I T N E S S I N M O U N TA I N GO AT S 2301

© 2009 Blackwell Publishing Ltd

genetically related than random mating pairs followingmethods described in Holand et al. (2007). Briefly, wecompared the estimated pairwise genetic relatedness ofactual mating pairs to that of all potential nonmating pairsusing a permutation test based on 10 000 randomizations.We considered all individuals that were adults duringruts 1994 to 2005 in the simulations to compare them to theactual mating pairs that produced a kid caught at 1 year ofage in 1996–2007. In addition, we used a Cochran-Armitagepermutation test (Agresti 1990) with 10 000 randomizationsto test whether actual mating pairs had more often thanexpected negative (non-kin) rather than positive (kin)pairwise genetic relatednesses compared to all potentialnonmating pairs. Because we expected individuals to avoidinbreeding due to its detrimental effects, we used one-tailed tests for these analyses. We also examined the effectof females’ characteristics on the degree of pairwise geneticrelatedness with their mating partner using a linear mixedmodel (LMM) with ‘female identity’ and ‘year’ fitted asrandom terms. The estimator of genetic relatedness, whichvaried between –0.27 and 0.50, was normalized before theanalysis by performing a log-transformation [loge(x + 1)].As both random terms did not explain variance in thedegree of genetic relatedness between mating partners(covariance parameter estimate of ‘female identity’ was0.004 ± 0.004, Z = 0.93, P = 0.18, whereas estimate for ‘year’was 0), we present the results of a generalized linear model(GLM).

Before testing for HFC, we evaluated whether pairwisegenetic relatedness between individuals of actual matingpairs predicted H in their offspring using a GLM. We thenexplored non-genetic sources of variation in yearlingsurvival, defined as the probability of surviving from 1to 2 years of age, and kid production in adult females,expressed as the annual probability of giving birth to a kid,by fitting these dependent variables as binary responses inlogistic regressions. Explanatory variables considered toexamine variance in yearling survival were yearling sexand mass because Festa-Bianchet & Côté (2008) reported anearly significant 11% difference in survival from 1 to 2years of age favouring females, and kid mass was known toaffect positively survival to 1 year (Côté & Festa-Bianchet2001a). Because the ability to provide maternal care increaseswith age in female mountain goats (Côté & Festa-Bianchet2001b, c) and this could have influenced survival later, wefitted maternal age at birth as a third explanatory variable.For kid production, we considered female age, age2, socialrank, and an interaction between age and social rank, asthese factors have all been previously reported to influencethe annual probability of giving birth (Côté & Festa-Bianchet2001b). For both yearling survival and kid production,population density was also added as an explanatoryvariable expressed as the number of adult females on 1June, because density nearly doubled since the beginning

of the study (Festa-Bianchet & Côté 2008) and recentanalyses have revealed density-dependent effects, such asincreased costs of reproduction in females at high density(Hamel 2008). For both analyses, we used generalizedestimating equations (GEE; Liang & Zeger 1986) with‘year’ fitted as a repeated term to account for variation infitness components that could be attributable to between-year differences.

All statistical analyses were conducted in sas 9.1 (SASInstitute 2003). Nonsignificant variables were removedfrom the full models of non-genetic terms based on theirP values (i.e. by removing the least significant term) using abackward stepwise procedure with statistical significanceset at α = 0.05. We then added H or single-locus heterozygosityto the reduced models to examine if they explained significantadditional variance. Inspection of residuals and collinearitydiagnostics indicated no violation of the assumptions ofnormality and homoscedasticity when required, or ofmulticollinearity among explanatory variables. ‘Dispersionparameter’ values were close to 1 in the modelling of binaryresponse variables, suggesting no problems of under-or overdispersion. Probabilities were corrected using thesequential Bonferroni method for single-locus analysesto account for multiple tests (Rice 1989). All means arepresented ± SE.

Results

Pairwise genetic relatedness of actual vs. random mating pairs

There were 95 mating pairs that produced a kid survivingto 1 year of age and in which both parents were adultsduring the rut. The relatedness of these mating pairswas compared to that of 8222 possible nonmating pairs.Mean pairwise genetic relatedness of actual mating pairs(–0.03 ± 0.01) was more than three times lower than that ofrandom mating pairs (–0.009 ± 0.002), but the differencewas not statistically significant (P = 0.09; Fig. 1). The valuesof pairwise genetic relatedness of actual mating pairs,however, were significantly more often negative thanpositive compared to that of all possible nonmating pairs(68% vs. 59%, respectively) under a Cochran–Armitagepermutation test (P = 0.04; Fig. 1). The pairwise geneticrelatedness of mating pairs was not affected by female age(F = 1.25, d.f. = 1, 71, P = 0.27) nor social rank (F = 1.87,d.f. = 1, 72, P = 0.07). As expected, offspring H wasnegatively related to the pairwise genetic relatedness oftheir parents (F1,98 = 19.4, d.f. = 1, 98, P < 0.0001, r2 = 0.16).

Yearling survival

Based on a sample of 155 yearlings monitored over 15 years,yearling sex, mass, and linear effects of maternal age did

2302 J . M A I N G U Y, S . D . C Ô T É and D . W. C O LT M A N

© 2009 Blackwell Publishing Ltd

not affect the probability of survival from age 1 to 2,whereas population density had a positive effect on survival(Table 2). After accounting for this non-genetic effect,yearling H was associated with survival (Table 2), as moreheterozygous individuals had higher survival probabilitiesthan less heterozygous ones (Fig. 2). When considered alone,H still marginally affected yearling survival positively(β = 3.68 ± 1.79, χ2 = 3.61, d.f. = 1, 153, P = 0.057). Whencomparing the first (most homozygous) and last (mostheterozygous) quartiles, the difference in yearling survivalwas 10% (82% vs. 92%, respectively). Because morethan half of the mothers (59%, n = 64) contributed morethan 1 yearling in this analysis (range = 2–5; 18 of the

155 yearlings used in this analysis were not assigned to amother) and some of these mothers could have been ofhigher ‘quality’, this could have also affected yearlingsurvival positively. To test for such a possible confoundingeffect of ‘mother identity’ on yearling survival, we repeatedthe GEE analysis on a reduced data set by randomlyselecting a single yearling from each mother (see Hoffmanet al. 2006 for details). We then repeated this analysis onfour other randomly selected subsets of yearlings. TheP values of the five data sets varied by less than 0.03, arguingagainst a confounding influence of ‘mother identity’ on therelationship between yearling survival and H. No singlelocus had a significant effect on yearling survival aftersequential Bonferroni correction (critical P value = 0.05/25loci = 0.002) when accounting for population density.However, as Bonferroni correction can sometimes beregarded as too stringent (Nakagawa 2004), we verifiedwhether removing one of the two loci with 0.002 < Pvalues < 0.05 (HEL10 and McM152, both loci with apositive effect of heterozygosity on yearling survival) inthe calculation of H changed the relationship. The removalof HEL10 reduced the positive association between H andyearling survival to a trend (P = 0.06), whereas the removalof McM152 did not change the results (P = 0.03). When weremoved both loci in the calculation of H, however, thepositive relationship was no longer significant (P = 0.16).When single-locus heterozygosity was fitted as a uniqueexplanatory variable, we found 14 positive and 11 negativeassociations with yearling survival, a ratio not differentfrom unity (Fig. 3a, χ2 = 0.18, d.f. = 1, P = 0.67).

Annual probability of giving birth in adult females

As in Côté & Festa-Bianchet (2001b), the annual probabilityof giving birth to a kid was influenced by age, age2, social

Fig. 1 Distribution of estimates of pairwise genetic relatedness(Lynch & Ritland 1999) of actual mating pairs (n = 95) and allpossible nonmating pairs (n = 8222) of adult (≥ 3 years old)mountain goats during ruts 1994–2005 at Caw Ridge, Alberta,Canada. Actual mating pairs were determined by genetic parent-age analyses of yearlings.

Table 2 GEE of the effects of sex (females relative to males),adjusted body mass on 15 July, maternal age, population den-sity (number of adult females on 1 June), and multilocusheterozygosity H, on the probability of survival from 1 to 2 yearsof age in mountain goats at Caw Ridge, Alberta, Canada. Statisticsfor H are reported when H is added to a reduced model ofnon-genetic terms. Year of observation was included as a repeatedterm

Variables β SE* χ2 d.f. P

Population density 0.09 0.03 4.11 1, 153 0.04H 4.67 1.49 5.61 1, 152 0.02

Not includedMaternal age –0.02 0.06 0.12 1, 132 0.72Sex (female) 0.24 0.54 0.21 1, 151 0.65Adjusted mass –0.04 0.03 1.93 1, 152 0.17

*Estimates of empirical standard error (SE) are shown.

Fig. 2 Percentage of yearling mountain goats surviving to 2 yearsof age in relation to multilocus heterozygosity H at Caw Ridge,Alberta, Canada. Numbers on top of bars indicate sample size.

I N B R E E D I N G A N D F I T N E S S I N M O U N TA I N GO AT S 2303

© 2009 Blackwell Publishing Ltd

rank, and an interaction between age and social rank(Table 3) in a sample of 120 individual females monitoredover 12 years, whereas population density did not affectthe probability of producing a kid (Table 3). Fitting H in areduced model of the non-genetic terms did not explainadditional variance for this trait (Table 3). Using only onerandomly selected observation from each female led tosimilar results. When tested alone, no single locus hada significant effect on the probability of giving birth to akid after sequential Bonferroni correction. Althoughthere were almost twice more positive (n = 16) than negative(n = 9) associations between single-locus heterozygosityand the probability of giving birth in adult females

(Fig. 3b), the difference was not statistically significant(χ2 = 1.00, d.f. = 1, P = 0.32).

Discussion

In the Caw Ridge mountain goat population, actualmating pairs had more often negative than positive valuesof pairwise genetic relatedness compared to all possiblenonmating pairs. The degree of genetic relatedness offemales with their mate was independent of maternal ageand dominance status. Estimated parental genetic relatednesswas negatively correlated with offspring genetic diversity,which was in turn positively related to survival from 1 to 2years of age. As opposed to yearling survival, the annualprobability of giving birth in adult females was not affectedby individual genetic diversity. Our study provides evidencethat inbreeding depression occurs at an early life stage inmountain goats from the Caw Ridge population.

The results of the analysis of mate choice based on esti-mated pairwise genetic relatedness are likely to originatefrom inbreeding depression rather than indicating inbreedingavoidance, or could possibly be a combination of these twoprocesses. This is because we studied mating pairs thatproduced a kid that survived to 1 year of age and thatinbreeding often affects survival and other fitness traitsmore strongly during early life stages than later in life(Pujolar et al. 2006; Fessehaye et al. 2007; Rijks et al. 2008,but see Szulkin et al. 2007). As a consequence, it is possiblethat kids that died before 1 year old were more inbred thanthose that survived, potentially decreasing our estimate ofparental genetic relatedness. In other words, mating pairsthat produced an inbred offspring which did not survive toyearling age would be considered as nonmating pairs inthe analyses. Our results support this hypothesis since Hwas positively related with survival from 1 to 2 years of ageafter accounting for a positive effect of population density

Fig. 3 Difference (β ± SE) in the probability of (a) survival from 1to 2 years of age, and (b) giving birth to a kid in adult (≥ 3 years old)females on an annual basis, between heterozygotes andhomozygotes (the latter used as reference) at 25 microsatellite lociin mountain goats (Oreamnos americanus) inhabiting Caw Ridge,Alberta, Canada. Single-locus heterozygosity was fitted as a soleexplanatory variable. None of the logistic regression coefficientswere significant after sequential Bonferroni corrections, but thosewith 0.002 < P-values < 0.05 are indicated with an asterisk.

Table 3 GEE of the effects of female’s characteristics, populationdensity (number of adult females on 1 June), and multilocusheterozygosity H on the annual probability of giving birth to a kidin mountain goats at Caw Ridge, Alberta, Canada. Statistics for Hare reported when H is added to a reduced model of non-geneticterms. Year of observation was included as a repeated term

Variables β SE* χ2 d.f. P

Age 1.24 0.12 9.51 1, 537 0.002Age2 –0.06 0.01 9.41 1, 537 0.002Social rank 9.41 3.06 6.50 1, 537 0.01Age × social rank –1.12 0.37 7.70 1, 537 0.006H 0.83 1.10 0.37 1, 440 0.54

Not includedPopulation density –0.02 0.03 0.53 1, 536 0.47

*Estimates of empirical standard error (SE) are shown.

2304 J . M A I N G U Y, S . D . C Ô T É and D . W. C O LT M A N

© 2009 Blackwell Publishing Ltd

on survival, which was somewhat puzzling, but could haveoriginated from a benefit of gaining higher probabilitiesof detecting predators in larger than in smaller groups,similarly to what was reported in a gregarious bird(Watson et al. 2007). This is possible as predation is acommon cause of mortality in kids and yearlings (Côté &Beaudoin 1997; Côté et al. 1997; Festa-Bianchet & Côté 2008).Although speculative, it is also possible that populationdensity was partly correlated with another environmentalfactor (e.g. the length of the vegetation growth period),that was also positively correlated with yearling survival.Altogether, our results suggest that matings betweenrelatives occur in our study population, but may be moredifficult to detect if inbred offspring die before beingsampled at 1 year of age (Marshall et al. 2002). Nevertheless,our results may still be indicative of inbreeding avoidance.For instance, we did not find close inbreeding (F = 0.25) inthe pedigree. Therefore, mountain goats may be able topartly discriminate among more or less related kins formatings, unless all highly inbred offspring are beingremoved from the population by natural selection. On theother hand, low to moderate inbreeding, which canaccount for most of the overall inbreeding in a population(Marshall et al. 2002), may be responsible for the signs ofinbreeding depression that we found.

We interpreted the decrease in yearling survival withincreasing levels of homozygosity as evidence of inbreedingdepression under a genome-wide effect (Hansson &Westerberg 2002). We advocate that this is likely for threemain reasons. First, we found no significant single-locuseffects (as in Hoffman et al. 2004; Charpentier et al. 2005;Ortego et al. 2007), although heterozygosity at some of theloci examined had stronger positive effects on juvenilesurvival than others. This result is not surprising, however,given that we used a fairly large number of loci of whichnone had a significant effect on survival after correcting formultiple tests. Second, our estimates of pairwise geneticrelatedness were correlated with pedigree relatedness andparental estimated relatedness was positively correlatedwith offspring homozygosity, as expected for inbredmatings. Third, our study population met the conditions inwhich H should be correlated with F, as mountain goatsare found in small structured populations and exhibit apolygynous mating system (Balloux et al. 2004). At CawRidge, five males out of 57 sampled sired more than half ofthe yearlings genetically assigned over 12 years (Mainguy2008), suggesting that many individuals could be relatedpaternally. Thus, inbred matings stemming from a commonpaternal ancestor are likely, whereas father–daughtermatings are much less likely. This is because most malesstart to reproduce intensely at 7–9 years of age only (Mainguy2008), but rarely survive past 10 years (Festa-Bianchet &Côté 2008), whereas females are generally primiparous at4–5 years of age (sometimes as late as 6 or 7 years),

precluding them from mating with their fathers who areunlikely to be alive by that age. On the other hand, femalesare more likely to reproduce with a paternally relatedhalf-sib, as observed in our pedigree, due to the numerousoffspring sired by a few mature dominant males (Mainguy2008). Because of this polygynous mating system andlikely limited genetic exchanges among populations, F isthus expected to correlate moderately to strongly with H inthis alpine ungulate. Continued monitoring of the individualsof the Caw Ridge population will allow us to assess the extentof this relationship, as well as increasing our statisticalpower to study inbreeding depression.

To conclude, we used marker-based measures of related-ness and inbreeding to show an apparent lack or weakevidences for the presence of inbreeding avoidance, andsubsequent inbreeding depression on yearling survival ina long-term study population of mountain goats. Thesefindings suggest that despite generations of matingsbetween relatives, mountain goats are still showing signsof inbreeding depression. This may thus represent anevolutionary stable state for this alpine ungulate, similarlyto that found in insular metapopulations of birds which alsoexhibit signs of inbreeding depression partly generatedby strong population structure (e.g. Jensen et al. 2007). Afuture and important area of research will be to quantify thegenetic exchanges among different mountain goat popula-tions, and assess their effects on population genetic diversityand dynamics.

Acknowledgements

Many people helped with fieldwork at Caw Ridge over the yearsand we thank all of them. M. Festa-Bianchet at the Université deSherbrooke and K. G. Smith from the Alberta Natural ResourcesService (ANRS) were pivotal in the inception of the Caw Ridgestudy and provided valuable logistical help, as did M. Ewald andD. Hobson from ANRS. We also thank A. S. Llewellyn and K. Worleyfor genotyping mountain goats captured in 1994–2003. We aregrateful to C. Barrette, L. Bernatchez, L. E. B. Kruuk, and J. M.Pemberton for valuable discussions, and L. Bernatchez, M. Festa-Bianchet, D. Fortin, J.-M. Gaillard, A. G. McElligott, J. Poissant andan anonymous reviewer for comments that improved the qualityof the manuscript. Funding was provided by the Natural Sciencesand Engineering Research Council of Canada (NSERC), the NaturalEnvironment Research Council (UK), ANRS, the Alberta Con-servation Association (ACA), and Alberta Ingenuity. J. Mainguywas supported by NSERC, Fonds québécois de la recherche sur lanature et les technologies, Fonds Richard-Bernard, FondationJ.-Arthur-Vincent, and ACA grants in biodiversity scholarships.

References

Agresti A (1990) Categorical Data Analysis. John Wiley & Sons, NewYork.

Amills M, Jiménez N, Jordana J et al. (2004) Low diversity in themajor histocompatibility complex class II DRB1 gene of theSpanish ibex, Capra pyrenaica. Heredity, 93, 266–272.

I N B R E E D I N G A N D F I T N E S S I N M O U N TA I N GO AT S 2305

© 2009 Blackwell Publishing Ltd

Amos W, Wilmer JW, Fullard K et al. (2001) The influence of parentalrelatedness on reproductive success. Proceedings of the RoyalSociety B: Biological Sciences, 268, 2021–2027.

Aparicio JM, Ortego J, Cordero PJ (2006) What should we weigh toestimate heterozygosity, alleles or loci? Molecular Ecology, 15,4659–4665.

Aparicio JM, Ortego J, Cordero PJ (2007) Can a simple algebraicanalysis predict markers-genome heterozygosity correlations?Journal of Heredity, 98, 93–96.

Archie EA, Hollister-Smith JA, Poole JH et al. (2007) Behaviouralinbreeding avoidance in wild African elephants. MolecularEcology, 16, 4138–4148.

Balloux F, Amos W, Coulson T (2004) Does heterozygosity estimateinbreeding in real populations? Molecular Ecology, 13, 3021–3031.

Bassano B, von Hardenberg A, Pelletier F, Gobbi G (2003) Amethod to weigh free-ranging ungulates without handling.Wildlife Society Bulletin, 31, 1205–1209.

Bollinger EK, Harper SJ, Barrett GW (1993) Inbreeding avoidanceincreases dispersal movements of the meadow vole. Ecology, 74,1153–1156.

Charlesworth D, Charlesworth B (1987) Inbreeding depressionand its evolutionary consequences. Annual Review of Ecology andSystematics, 18, 237–268.

Charlesworth B, Charlesworth D (1999) The genetic basis ofinbreeding depression. Genetical Research, 74, 329–340.

Charpentier M, Setchell JM, Prugnolle F et al. (2005) Geneticdiversity and reproductive success in mandrills (Mandrillussphinx). Proceedings of the National Academy of Sciences, USA, 102,16723–16728.

Coltman DW, Slate J (2003) Microsatellite measures of inbreeding:a meta-analysis. Evolution, 57, 971–983.

Côté SD (2000) Dominance hierarchies in female mountain goats:stability, aggressiveness and determinants of rank. Behaviour,137, 1541–1566.

Côté SD, Beaudoin C (1997) Grizzly bear (Ursus arctos) attacks andnanny-kid separation on mountain goats (Oreamnos americanus).Mammalia, 61, 614–617.

Côté SD, Festa-Bianchet M (2001a) Birthdate, mass and survival inmountain goat kids: effects of maternal characteristics andforage quality. Oecologia, 127, 230–238.

Côté SD, Festa-Bianchet M (2001b) Reproductive success in femalemountain goats: the influence of age and social rank. AnimalBehaviour, 62, 173–181.

Côté SD, Festa-Bianchet M (2001c) Offspring sex ratio in relation tomaternal age and social rank in mountain goats (Oreamnosamericanus). Behavioral Ecology and Sociobiology, 49, 260–265.

Côté SD, Festa-Bianchet M (2003) Mountain goat Oreamnos ameri-canus. In: Wild Mammals of North America: Biology, Managementand Conservation (eds Feldhamer GA, Thompson BC, ChapmanJA), pp. 1061–1075. John Hopkins University Press, Baltimore,Maryland.

Côté SD, Peracino A, Simard G (1997) Wolf, Canis lupus, predationand maternal defensive behavior in mountain goats, Oreamnosamericanus. Canadian Field Naturalist, 111, 389–392.

Côté SD, Festa-Bianchet M, Fournier F (1998a) Life-history effectsof chemical immobilization and radiocollars on mountain goats.Journal of Wildlife Management, 62, 745–752.

Côté SD, Festa-Bianchet M, Smith KG (1998b) Horn growth in mountaingoats (Oreamnos americanus). Journal of Mammalogy, 79, 406–414.

Coulson TN, Pemberton JM, Albon SD et al. (1998) Microsatellitesreveal heterosis in red deer. Proceedings of the Royal Society B:Biological Sciences, 265, 489–495.

Crnokrak P, Roff DA (1999) Inbreeding depression in the wild.Heredity, 83, 260–270.

Csilléry K, Johnson T, Beraldi D et al. (2006) Performance ofmarker-based relatedness estimators in natural populations ofoutbred vertebrates. Genetics, 173, 2091–2101.

Da Silva A, Luikart G, Yoccoz NG, Cohas A, Allainé D (2006)Genetic diversity-fitness correlation revealed by microsatelliteanalyses in European alpine marmots (Marmota marmota).Conservation Genetics, 7, 371–382.

Duarte LC, Bouteiller C, Fontanillas P, Petit E, Perrin N (2003)Inbreeding in the greater white-toothed shrew, Crocidura russula.Evolution, 57, 638–645.

Fessehaye Y, Komen H, Rezk MA, van Arendonk JAM, BovenhuisH (2007) Effects of inbreeding on survival, body weight andfluctuating asymmetry (FA) in Nile tilapia, Oreochromis niloticus.Aquaculture, 264, 27–35.

Festa-Bianchet M, Côté SD (2008) Mountain Goats: Ecology, Behavior,and Conservation of an Alpine Ungulate. Island Press, Washington, DC.

Gendreau Y, Côté SD, Festa-Bianchet M (2005) Maternal effects onpost-weaning physical and social development in juvenilemountain goats (Oreamnos americanus). Behavioral Ecology andSociobiology, 58, 237–246.

Gerlach G, Lysiak N (2006) Kin recognition and inbreeding avoidancein zebrafish, Danio rerio, is based on phenotype matching. AnimalBehaviour, 71, 1371–1377.

Halverson MA, Skelly DK, Caccone A (2006) Inbreeding linked toamphibian survival in the wild but not in the laboratory. Journalof Heredity, 97, 499–507.

Hamel S (2008) Les compromis à court et à long terme associés à lareproduction chez les ongulés femelles: l’exemple de la chèvre demontagne. PhD thesis (chapters are in English). Université Laval,Québec, Canada.

Hamel S, Côté SD, Smith KG, Festa-Bianchet M (2006) Populationdynamics and harvest potential of mountain goat herds inAlberta. Journal of Wildlife Management, 70, 1044–1053.

Hansson B, Westerberg L (2002) On the correlation betweenheterozygosity and fitness in natural populations. MolecularEcology, 11, 2467–2474.

Hansson B, Westerdahl H, Hasselquist D, Åkesson M, Bensch S(2004) Does linkage disequilibrium generate heterozygosity-fitness correlations in great reed warbler? Evolution, 58, 870–879.

Hansson B, Jack L, Christians JK et al. (2007) No evidence forinbreeding avoidance in a great reed warbler population.Behavioral Ecology, 18, 157–164.

von Hardenberg A, Bassano B, Festa-Bianchet M et al. (2007)Age-dependent genetic effects on a secondary sexual trait inmale Alpine ibex, Capra ibex. Molecular Ecology, 16, 1969–1980.

Hardy OJ, Vekemans X (2002) SPAGeDi: a versatile computerprogram to analyse spatial genetic structure at the individual orpopulation levels. Molecular Ecology Notes, 2, 618–620.

Hedrick PW, Kalinowski ST (2000) Inbreeding depression inconservation biology. Annual Review of Ecology and Systematics,31, 139–162.

Hoffman JI, Boyd IL, Amos W (2004) Exploring the relationshipbetween parental relatedness and male reproductive success in theAntarctic fur seal Arctocephalus gazella. Evolution, 58, 2087–2099.

Hoffman JI, Forcada J, Amos W (2006) No relationship betweenmicrosatellite variation and neonatal fitness in Antarctic furseals, Arctocephalus gazella. Molecular Ecology, 15, 1995–2005.

Hoffman JI, Forcada J, Trathan PN, Amos W (2007) Female furseals show active choice for males that are heterozygous andunrelated. Nature, 445, 912–914.

2306 J . M A I N G U Y, S . D . C Ô T É and D . W. C O LT M A N

© 2009 Blackwell Publishing Ltd

Holand Ø, Askim KR, Røed KH et al. (2007) No evidence ofinbreeding avoidance in a polygynous ungulate: the reindeer(Rangifer tarandus). Biology Letters, 3, 36–39.

Jensen H, Bremset EM, Ringsby TH, Sæther B-E (2007) Multilocusheterozygosity and inbreeding depression in an insular housesparrow metapopulation. Molecular Ecology, 16, 4066–4078.

Jiménez JA, Hughes KA, Alaks G, Graham L, Lacy RC (1994) Anexperimental study of inbreeding depression in a naturalhabitat. Science, 266, 271–273.

Kalinowski ST, Taper ML, Marshall TC (2007) Revising how thecomputer program Cervus accommodates genotyping errorincreases success in paternity assignment. Molecular Ecology, 16,1099–1106.

Keller LF (1998) Inbreeding and its fitness effects in an insularpopulation of song sparrows (Melospiza melodia). Evolution, 52,240–250.

Keller LF, Arcese P (1998) No evidence for inbreeding avoidancein a natural population of song sparrows (Melospiza melodia).American Naturalist, 152, 380–392.

Keller LF, Waller DM (2002) Inbreeding effects in wild populations.Trends in Ecology & Evolution, 17, 230–241.

Kokko H, Ots I (2006) When not to avoid inbreeding. Evolution, 60,467–475.

Lane JE, Boutin S, Gunn MR, Slate J, Coltman DW (2007) Geneticrelatedness of mates does not predict patterns of parentage inNorth American red squirrels. Animal Behaviour, 74, 611–619.

Lesbarrères D, Primmer CR, Laurila A, Merilä J (2005) Environ-mental and population dependency of genetic variability-fitnesscorrelations in Rana temporaria. Molecular Ecology, 14, 311–323.

Liang K-Y, Zeger SL (1986) Longitudinal data analysis usinggeneralized linear models. Biometrika, 73, 13–22.

Lieutnant-Gosselin M, Bernatchez L (2006) Local heterozygosity-fitness correlations with global positive effects on fitness inthreespine stickleback. Evolution, 60, 1658–1668.

Lynch M, Ritland K (1999) Estimation of pairwise relatedness withmolecular markers. Genetics, 152, 1753–1766.

Mainguy J (2008) Variabilité génétique et reproduction des mâles chezla chèvre de montagne. PhD thesis (chapters are in English).Université Laval, Québec, Canada.

Mainguy J, Côté SD (2008) Age- and state-dependent reproductiveeffort in male mountain goats, Oreamnos americanus. BehavioralEcology and Sociobiology, 62, 935–943.

Mainguy J, Llewellyn AS, Worley K, Côté SD, Coltman DW (2005)Characterization of 29 polymorphic artiodactyl microsatellitemarkers for the mountain goat (Oreamnos americanus). MolecularEcology Notes, 5, 809–811.

Mainguy J, Worley K, Côté SD, Coltman DW (2007) Low MHCDRB class II diversity in the mountain goat: past bottlenecks andpossible role of pathogens and parasites. Conservation Genetics,8, 885–891.

Mainguy J, Côté SD, Cardinal E, Houle M (2008) Mating tacticsand mate choice in relation to age and social rank in male mountaingoats. Journal of Mammalogy, 89, 626–635.

Markert JA, Grant PR, Grant BR et al. (2004) Neutral locus hetero-zygosity, inbreeding, and survival in Darwin’s ground finches(Geospiza fortis and G. scandens). Heredity, 92, 306–315.

Marshall TC, Coltman DW, Pemberton JM et al. (2002) Estimatingthe prevalence of inbreeding from incomplete pedigrees. Pro-ceedings of the Royal Society B: Biological Sciences, 269, 1533–1539.

Nakagawa S (2004) A farewell to Bonferroni: the problems of lowstatistical power and publication bias. Behavioral Ecology, 15,1044–1045.

Oliehoek PA, Windig JJ, van Arendonk JAM, Bijma P (2006)Estimating relatedness between individuals in general popula-tions with focus on their use in conservation programs. Genetics,173, 483–496.

Ortego J, Aparicio JM, Calabuig G, Cordero PJ (2007) Risk ofectoparasitism and genetic diversity in a wild lesser kestrelpopulation. Molecular Ecology, 16, 3712–3720.

Overall ADJ, Byrne KA, Pilkington JG, Pemberton JM (2005)Heterozygosity, inbreeding and neonatal traits in Soay sheep onSt Kilda. Molecular Ecology, 14, 3383–3393.

Pujolar JM, Maes GE, Vancoillie C, Volckaert FAM (2006) Environ-mental stress and life-stage dependence on the detection ofheterozygosity-fitness correlations in the European eel, Anguillaanguilla. Genome, 49, 1428–1437.

Pusey A, Wolf M (1996) Inbreeding avoidance in animals. Trends inEcology & Evolution, 11, 201–206.

Queller DC, Goodnight KF (1989) Estimating relatedness usinggenetic markers. Evolution, 43, 258–275.

Rice WR (1989) Analyzing tables of statistical tests. Evolution, 43,223–225.

Rijks JM, Hoffman JI, Kuiken T, Osterhaus ADME, Amos W (2008)Heterozygosity and lungworm burden in harbour seals (Phocavitulina). Heredity, 100, 587–593.

Rosenfield RN, Bielefeldt J (1992) Natal dispersal and inbreedingin the Cooper’s Hawk. Wilson Bulletin, 104, 182–184.

SAS Institute (2003) SAS/STAT User’s Guide, Version 9.1. SAS InstituteInc, Cary, North Carolina.

Slate J, David P, Dodds KG et al. (2004) Understanding therelationship between the inbreeding coefficient and multilocusheterozygosity: theoretical expectations and empirical data.Heredity, 93, 255–265.

Spielman D, Brook BW, Frankham R (2004) Most species are notdriven to extinction before genetic factors impact them. Proceedingsof the National Academy of Sciences, USA, 101, 15261–15264.

Szulkin M, Garant D, McCleery RH, Sheldon BC (2007) Inbreedingdepression along a life-history continuum in the great tit. Journalof Evolutionary Biology, 20, 1531–1543.

Van De Casteele T, Galbusera P, Matthysen E (2001) A comparisonof microsatellite-based pairwise relatedness estimators. MolecularEcology, 10, 1539–1549.

Watson M, Aebischer NJ, Cresswell W (2007) Vigilance and fitnessin grey partridges Perdix perdix: the effects of group size andforaging-vigilance trade-offs on predation mortality. Journal ofAnimal Ecology, 76, 211–221.

Worley K, Carey J, Veitch A, Coltman DW (2006) Detecting thesignature of selection on immune genes in highly structured popu-lations of wild sheep (Ovis dalli). Molecular Ecology, 15, 623–637.

J.M. conducted his PhD on mountain goats at Université Laval(Québec) with S.C. and D.C. He is now working as a wildlifebiologist for the Québec Ministry of Natural Resources andWildlife. His research interests include behavioural and molecularecology. S.C. is a professor in the Biology Department of UniversitéLaval and senior scientist at the Centre for Northern Studies. Hisresearch interests include behavioural ecology of large herbivores,evolution of life-history strategies, wildlife management, conservationbiology and population genetics. D.C. is a professor in theDepartment of Biological Sciences of University of Alberta. Hisresearch centers on conservation genetics and ecological genomicsof handsome ungulates and other wildlife.