Inbreeding depression along a life-history continuum in the great tit

Authors


Marta Szulkin, Edward Grey Institute of Field Ornithology, Department of Zoology, South Parks Road, OX1 3PS Oxford, UK.
Tel.: +44 1865 271255; fax: +44 1865 271168;
e-mail: marta.szulkin@zoo.ox.ac.uk

Abstract

Inbreeding resulting from the mating of two related individuals can reduce the fitness of their progeny. However, quantifying inbreeding depression in wild populations is challenging, requiring large sample sizes, detailed knowledge of life histories and study over many generations. Here we report analyses of the effects of close inbreeding, based on observations of mating between relatives, in a large, free-living noninsular great tit (Parus major) population monitored over 41 years. Although mating between close relatives (f ≥ 0.125) was rare (1.0–2.6% of matings, depending on data set restrictiveness), we found pronounced inbreeding depression, which translated into reduced hatching success, fledging success, recruitment to the breeding population and production of grand offspring. An inbred mating at f = 0.25 had a 39% reduction in fitness relative to that of an outbred nest, when calculated in terms of recruitment success, and a 55% reduction in the number of fledged grand offspring. Our data show that inbreeding depression acts independently at each life-history stage in this population, and hence suggest that estimates of the fitness costs of inbreeding must focus on the entire life cycle.

Introduction

Inbreeding occurs when the parents share a common ancestor (Crow & Kimura, 1970; Malécot, 1948). Inbred individuals will more frequently inherit the same allele from each of their parents, which will increase their homozygosity, lead to the expression of deleterious recessive alleles and ultimately cause inbreeding depression (Lynch & Walsh, 1998). Deleterious consequences of inbreeding are thus likely to be responsible for the evolution of various aspects of inbreeding avoidance mechanisms in plant and animal mating systems (Charlesworth & Charlesworth, 1987; Pusey, 1988; Pusey & Wolf, 1996). The negative effects of inbreeding in animals have been well-studied under laboratory conditions (Charlesworth & Charlesworth, 1999) and through controlled breeding of animals, such as poultry (Abplanalp, 1990), dairy cattle (Smith et al., 1998) and salmon (Wang et al., 2001). Similarly, records of inbreeding depression are found in wild species kept in captivity or semi-wild conditions (Hedrick & Kalinowski, 2000; Cassinello, 2005; Charpentier et al., 2006). However, captivity can mask the effects of inbreeding on fitness (Joron & Brakefield, 2003) and Crnokrak & Roff (1999) found that estimates of inbreeding depression in wild populations were generally higher than those found in captive populations. Investigations of the effects of inbreeding on individual fitness in a natural setting, where environmental variation interacts with the expression of deleterious recessive alleles, are therefore of great relevance to our understanding of the importance of inbreeding in evolution.

Keller & Waller (2002) reviewed evidence for inbreeding depression in the wild and found that, overall, this process can have severe effects on fitness, which are particularly often documented in the juvenile stage (Keller & Waller, 2002 and references therein). However, there is relatively little detail available about which particular life-history stages are affected by inbreeding depression. More importantly, most studies focus on the effects of increased homozygosity on early developmental stages, i.e. till offspring reach adulthood (e.g. Daniels & Walters, 2000; Kruuk et al., 2002; Liberg et al., 2005); very little is known about the extent to which inbreeding depression occurs in adult life-history stages (although see Keller, 1998; Keller et al., 2002, for examples from two insular bird populations). Thus, to estimate the importance of inbreeding for fitness accurately, and hence to gain an insight as to the strength of selection on inbreeding avoidance mechanisms (Kokko & Ots, 2006), estimates of fitness costs caused by inbreeding should be quantified in a natural setting, and over an individual's entire life cycle.

Establishing the relatedness between individuals in a breeding pair – which determines the inbreeding coefficient of their offspring – might be done either using genetic markers or a pedigree. In pedigree-based inbreeding estimates, an individual is considered inbred when the parents share ancestors (see Keller & Waller, 2002). The extent of inbreeding is then related to the amount of ancestry that is shared by the parents of an inbred individual. In principle, genetic markers can be used to infer levels of inbreeding and inbreeding depression directly from sampled individuals. However, recent evidence suggests that the correlation between multi-locus heterozygosity measured using genetic markers and realized inbreeding in pedigrees (Balloux et al., 2004; Slate et al., 2004) will often be weak. This suggests that pedigrees derived from long-term monitoring of populations remain the best tool available for investigating individual and population inbreeding levels (Pemberton, 2004).

In contrast to laboratory and captive investigations, relatively few studies have reported inbreeding depression in wild vertebrates using pedigrees (see Keller & Waller, 2002 for a review). The paucity of such studies of inbreeding in the wild can be explained by the fact that in many populations inbreeding is infrequent, vertebrate species have relatively long generation times, and because intensive population marking and following is required to generate the pedigree and the fitness data. Many of the studies of inbreeding that have been conducted have been carried out on geographically isolated populations, for example, on islands (Van Noordwijk & Scharloo, 1981; Gibbs & Grant, 1989; Keller, 1998; Kruuk et al., 2002; Kruuk, 2004). Such systems often experience reduced gene flow and may be more efficient at purging deleterious alleles (Keller & Waller, 2002), thus potentially influencing the strength of inbreeding. As inbreeding depression is thought to be a process that has had evolutionary consequences across a very wide range of systems, more studies of inbreeding depression in nonisolated populations are therefore needed to better understand the likely consequences of inbreeding depression.

Many ‘outbred’ populations, typified by those found on continental landmasses or large islands, have, inevitably, a rather low frequency of inbreeding (Van Noordwijk & Scharloo, 1981; Daniels & Walters, 2000; Kruuk et al., 2002). This means that large sample sizes are needed to assess the importance of inbreeding depression in such populations. Even if the frequency of inbreeding is low, this may be the result of effective inbreeding avoidance mechanisms, and assessing the fitness cost of inbreeding is of importance for understanding the potential selection causing such behaviours. Our aim in this study was to document the occurrence of inbreeding and quantify inbreeding depression in a population of a passerine bird, the great tit Parus major, using a long-term data set spanning more than four decades. We aimed to quantify the effects of inbreeding on a range of morphological and life-history traits, with a particular emphasis on determining the sequential effect of inbreeding depression at each life-history stage from egg-laying to adulthood. Two previous studies have addressed some aspects of the occurrence and consequences of inbreeding in this population (Bulmer, 1973; Greenwood et al., 1978): the data available for analysis have increased by a factor of five since the most recent of these papers, which greatly increases potential resolution given that, as we show here, inbreeding events are relatively rare in this population.

Materials and methods

Study population

Great tits P. major have been studied at Wytham Woods, Oxfordshire, UK (1°20′W 51°46′N) since 1947 (Perrins, 1979). The study site is a semi-deciduous forest of c. 380 ha within excess of 1000 nestboxes scattered throughout the forest, at variable densities. From 1964 onwards, the number of nestboxes available to great tits, their location and the data collection protocols have been standardized. All boxes are checked at weekly intervals at the beginning of each breeding season and throughout the hatching period; clutch size, egg-laying date, hatching date and brood size are recorded or estimated from these checks. Young are ringed with individually numbered metal rings between days 7 and 15 (hatching = day 1) of the c. 20-day nestling period, and all young are weighed on day 15. Parents are captured at the nest during the mid- to late-nestling period: their age, mass, wing length and ring number recorded, and immigrant parents not carrying a ring are ringed. This approach, repeated year after year, allows us to construct a pedigree of the Wytham great tit population. Rates of immigration into the population are quite high: on average, 47% of females and 40% of males breeding in any year within Wytham were born outside the wood (McCleery et al., 2004). More details on field data collection can be found in McCleery et al. (2004) and associated references. Any nest where experimental manipulations such as cross-fostering, brood reduction or egg swaps had been carried out resulted in its exclusion from our analyses (see below for effects on pedigree construction). In addition, predated broods were also excluded from brood-level analyses, as it is not clear whether predation results from the behaviour of the parents, or possibly from human interference. Results are unchanged as regards the inference drawn if such nests are included.

Pedigree building and inbreeding estimation

We built a pedigree using information on all great tit breeding events occurring in Wytham and its vicinity between 1958 and 2004. We first selected breeding events where both parents could be linked to the previous breeding generation. In cases where one or both of the breeding parents could not be identified, the breeding event was still included in the pedigree, and the unknown parent(s) assigned a unique identification number specific to the breeding event in which it took part (hence we could identify siblings even when both parents were unknown). Additionally, for approximately 1.6% of all breeding events (n = 8842) recorded during the time of this study, cross-fostering of young or eggs occurred, but the biological parents could not be identified; those breeding events were excluded from the final pedigree. Their descendants, however, were included and assumed to have unknown and unrelated parents.

An individual is considered inbred when the parents share ancestors. The standard measure of an individual's degree of inbreeding from a pedigree is the coefficient of inbreeding f (Wright, 1922; Keller & Waller, 2002). Originally described as a correlation between uniting gametes (Wright, 1922), f is now interpreted as the probability of identity by descent of two alleles at a locus in an individual (Crow & Kimura, 1970; Malécot, 1948). We estimated the coefficient of inbreeding f from pedigree data using Pedigree Viewer version 5.2 (available at: http://www-personal.une.edu.au/bkinghor/pedigree.htm). Inbreeding coefficients are estimated relative to the founder population, and it is assumed that the founding population and all immigrant birds are unrelated. The kinship coefficient of two individuals breeding with each other, thereafter indicated as fij, is the same as the inbreeding coefficient assigned to their offspring, referred in the text as fi. (Crow & Kimura, 1970).

As pointed out by Marshall et al. (2002), it is important not only to identify inbreeding events themselves, but also to identify cases in which a particular type of inbreeding event could have been detected. Thus, the lower the inbreeding coefficient to be considered, the more must be known about an individual's ancestors. For example, it is only by knowing both parents and all four grandparents of each individual in the data set that all inbreeding events at fij = 0.125 could be identified. For inbreeding coefficients lower than 0.125, data set restrictiveness – which accounts for the number of known ancestors – would need to be further increased (Marshall et al., 2002). However, in a population with a high immigration rate (see above), and where inbreeding is relatively rare, increased data set restrictiveness is likely to be at the expense of statistical power. For example, in the present study, by restricting the data set from individuals with both parents known to one containing only individuals that have all four grandparents known, the number of pairs with fij = 0.25 is reduced by 38%, and the number of pairs assumed to be unrelated is reduced by 73%; Table 1 illustrates the change in sample sizes resulting from different levels of data set restrictiveness. However, it is likely that the vast majority of pairs for which one or more grandparents are not known include one or more immigrant individuals in their recent ancestry, and that such pairs are not closely inbred at all. Occasional errors, in which inbred pairs are wrongly assigned to the ‘outbred’ category will have a very small effect on the mean of this category, which is by far the largest group, as they are likely to involve < 1% of cases. Thus, to maximize the data set of inbred events while in the same time removing individuals with no known ancestors, we followed Kruuk et al.’s (2002) approach and performed our analyses on a data set for which inbreeding was estimated for individuals with both parents and at least one grandparent known. We also explored the effect of increasing, and decreasing the stringency with which relatives are identified with respect to the main effects of inbreeding in this population. Because the detection of inbreeding at fij < 0.125 requires severe restriction of our breeding events data set, we decided to fit f (parental relatedness fij or individual inbreeding fi– depending on the type of analysis made) as a continuous variable with only three levels of inbreeding: f = 0, 0.125, 0.25 in all our analyses (but see Fig. 1).

Table 1.   Number of breeding events in the data sets depending on the degree of restrictiveness applied.
 Number of breeding pairs with
Both parents knownAt least one grandparent knownAll four grandparents known
  1. The ‘number of grandparents known’, defining data set restrictiveness, refers to the point of view of the offspring of any given breeding event. Population inbreeding calculated with fij = 0, 0.125, 0.25 and overall population inbreeding coefficient in the three restricted data sets are shown.

f = 0.0557744651481
f = 0.125131312
f = 0.25454528
Total563545231521
Overall f0.002730.003400.00715
f = 0, 0.125, 0.250.002280.002850.00556
Figure 1.

 Mean population inbreeding for each year of study in Wytham great tits. Black bars represent mean population inbreeding based on all fij values available. Grey bars represent yearly mean population inbreeding based on fij values where only coefficients of inbreeding of fij = 0, 0.125 and 0.25 were included. Mean and median population inbreeding for all 41 years of study are indicated by the plain and dotted horizontal lines respectively.

Statistical analysis

We were particularly interested in quantifying inbreeding depression independently at each life-history stage that we could separate. Thus, we carried out sequential analyses testing for inbreeding depression at the brood level controlling for any effects of inbreeding at the previous stages, by analysing the proportion of offspring from one stage that reached the next stage. For example, we analysed the effects of inbreeding on hatching success using generalized linear models (GLMs) with binomial errors and a logit link, using brood size as the numerator and clutch size as the denominator (see for more details) in the model. In addition, we estimated the cumulative effect of inbreeding up to each life-history stage using a general linear mixed model (GLMM) with normal error structure, with male and female parental identity fitted as random effects in the analysis.

Because of the nonindependence of siblings in terms of their inbreeding coefficient fij (and environmental and maternal effects, Lynch & Walsh, 1998), hatching, fledging and recruitment success, as well as fledging mass, were all analysed at the brood level. All models were initially fitted with the full number of variables potentially explaining variation in the investigated trait; standard deletion-testing techniques were then used (Crawley, 2005), where the significance of each term was tested by comparing the change in deviance before and after the term was dropped from the full model; least significant terms were dropped first, and all terms were tested for significance after each step of reduction from the full model. The minimum adequate model included the inbreeding variable f and all terms where P < 0.2.

We split our analyses into five major sections, namely: (i) parentally determined traits; (ii) morphometric traits; (iii) brood level analyses of life-history traits where each life-history stage was investigated independently of each other; (iv) individual-level analyses of life-history traits in recruited inbred and outbred birds, and finally (v) analyses of the cumulative effects of inbreeding, starting with brood size and following a breeding event's success up to estimating the number of recruited grand-offspring (see for a definition of all traits and variables used in our analyses).

Parentally determined traits (i) and offspring mean fledging weight (ii) were analysed in linear mixed models with male and female (parental) identity fitted as random effects to control for multiple breeding from the same individual; inbreeding fij was fitted as explanatory variable, among other continuous and categorical fixed effects. Recruited offspring morphometric traits were analysed using a linear mixed model with individual identity and birth nest (unique record denoting each breeding event, identical to all individuals in a brood) fitted as random effects. Independent life-history traits (iii) were analysed in GLMs with binomial proportions, a logit link and an estimate of the dispersion parameter (see for more details).

Recruitment values for individuals born in 2004 were supplemented by information on their breeding status as found in 2005, thus enabling the inclusion of recruitment success for year 2004. Although not all offspring that survive the winter breed in their first year, we decided to keep 2004 data: potential bias in recruitment success caused by birds that did not recruit in the population yet (but would breed the following year) should not interfere with our analyses of inbreeding effects on recruitment. Analyses of recruitment were thus carried out on all 41 years of data.

All analyses on recruited offspring in terms of their own reproductive success (iv) or morphometrics (such as mass or wing length) were carried out at an individual level, as at this stage we considered these characters to be independent units. In the data set of individual level analyses, we included all birds born in nonmanipulated broods that recruited in Wytham or its vicinity for which at least one grandparent was known. While for all traits detailed above (sections i–iii, with the exception of recruited adult morphometrics) we tested the effects of inbreeding fij on the brood (i.e. to test whether the reproductive events of parents with different degrees of kinship differ), section (iv) investigates the effects of fi, that is the effects of a parent being inbred, and how that affects its reproductive success given that the offspring it is raising is outbred (none of the offspring raised by inbred individuals had an inbreeding coefficient greater than 0.125). We additionally included in all our models the variable named ‘interference’, reflecting disturbance caused to any breeding event experienced by a recruited individual (i.e. cross-fostering, brood reduction, egg swaps; see). In this section, we did not exclude predated breeding events from the analyses as predation may potentially be limited if parental care is efficient.

Two approaches were used to test the effect of offspring inbreeding coefficient fi on the subsequent fitness of its own offspring: (1) we investigated the effects of parental inbreeding fi on the following traits: egg-laying date, clutch size, mean egg weight, brood size (with clutch size fitted as a covariate), the number of fledged individuals (with brood size fitted as a covariate), the number of recruited individuals (with the number of fledged individuals fitted as a covariate) and offspring mean fledging weight. Fitting the preceding life-history stage in the models allowed us to test the effect of parental inbreeding fi at each offspring life-history stage separately. The analyses were performed for male offspring, female offspring and for both sexes combined. All traits were tested using GLMMs with normal error structure, where parental identity and parental birth nest were fitted as random effects to control for: (i) repeated breeding of one individual in time and (ii) common genetic background, given that often more than one individual per brood recruited to the population. The explanatory variable in all models were inbreeding fi and interference (for egg-laying date, only inbreeding fi was used as explanatory variable). (2) Offspring lifetime reproductive success (LRSO) was calculated as the log-transformed ‘lifetime number of offspring + 1’ a recruited individual reared till fledging. We performed two types of analyses on LRSO using a linear mixed model with normal errors where parental birth nest was fitted as a random effect, testing the effects of: (i) inbreeding fi on LRSO; (ii) lifetime clutch size and lifespan, sex and their interactions with inbreeding fi on LRSO.

To estimate the cumulative effects of inbreeding fij (i.e. parental kinship), we ran a model with normal errors where the effects of year, woodland sector, male and female age, egg-laying date and inbreeding fij were tested on the following traits: brood size, number of fledged individuals, number of recruits, number of fledged and recruited grand offspring; any explanatory variables where P < 0.2 were dropped from the model. The variable ‘interference’ was additionally included in models of the number of fledged and recruited grand offspring. The cumulative decrease in fitness caused by inbreeding was illustrated graphically using fitted values where the fij = 0.25 mean fitted value was scaled relative to an fij = 0 mean fitted value of 1. All statistical analyses were carried out using GenStat Version 9.1. (VSN International Ltd, Hemel Hempstead, UK).

Estimation of the number of lethal equivalents

The decline in fitness caused by inbreeding can be standardized into an estimate of lethal equivalents per haploid genome, expressed in units where one lethal equivalent would cause death in a homozygote (Morton & Crow, 1956). This allows comparisons of the effect size of inbreeding depression across studies. We estimated the number of lethal equivalents B present in successive life-history stages using a single class of inbred individuals, f = 0.25, and compared survivorship at a brood level with outbred individuals (f = 0) where

image

where Sf is the probability of survival at inbreeding level f = 0.25, and S0 is the probability of survival at inbreeding level f = 0 (Lynch & Walsh, 1998). The number of lethal equivalents found in a diploid genome is twice the rate of increase in mortality caused by inbreeding, and is equivalent to 2B.

Results

Pedigree and occurrence of inbreeding

We built a pedigree consisting of 71 008 individual great tits; the average pedigree depth was 7.7 generations with a median of 4. Close inbreeding (fij ≥ 0.125) was rare, and occurred in only 1.0% of all breeding events; the occurrence of inbreeding increased (while remaining rare) with greater data set restrictiveness, and reached values of 1.3% and 2.6% in the data sets with at least one grandparent known and all four grandparents known respectively (Table 1). In the data set where both parents and at least one grandparent were known, consisting of 4523 breeding events, 13 events were inbred at fij = 0.125, and 45 had a coefficient of inbreeding of fij = 0.25; these comprised 27 brother–sister matings, six father–daughter and 12 mother–son matings.

By comparing inbreeding coefficients obtained from pedigree analyses and our synthetic variable for which only three degrees of inbreeding were used, namely fij = 0, 0.125 and 0.25, we found that the latter explained 84% of the entire population inbreeding (calculated as the proportion of overall population inbreeding using the synthetic variable divided by overall population inbreeding using ‘raw’ inbreeding coefficients). Annual population levels of inbreeding fluctuated around overall mean and median levels of inbreeding of fij = 0.00385 and 0.00326 respectively (Fig. 1). Interestingly, population mean inbreeding levels varied with respect to time, with mean inbreeding levels being highest at the beginning and at the end of the study (quadratic model: F2,38 = 4.64, P = 0.016; see Fig. 1). The causes of this fluctuation are not clear; if anything, an increase over the course of the study might be expected as the capture rate of parents has increased, but we do not yet know whether dispersal behaviour has also changed.

Inbreeding and parental traits

There was no evidence suggesting that individuals involved in breeding events with close kin differed with respect to life-history traits compared with individuals breeding with an unrelated partner. There was no effect of inbreeding fij on clutch size (Waldd.f.=1 = 0.04, P = 0.836, parameter estimate (SE): 0.193 (0.93); mean clutch size was 8.7 (n = 4459), 9.1 (n = 13) and 9.0 (n = 45) for fij = 0. fij = 0.125 and 0.25 breeding events respectively). Nor did we find any effect of inbreeding fij on mean egg weight [Waldd.f.=1 = 0.00, P = 0.996, n = 3733, parameter estimate (SE): 0.00041 (0.082)] or egg-laying date [Waldd.f.=1 = 0.11, P = 0.741, n = 4310, parameter estimate (SE): −1.25 (3.79)]. This was expected, given that the expression of deleterious recessive alleles should be expressed in traits determined by the offspring's genotype, but provides evidence that individuals who inbreed are not a subset of individuals of poor phenotypic quality, to the extent that these traits are a reliable index of phenotypic quality.

Inbreeding and offspring fitness

Although close inbreeding events were rare, their effects on offspring fitness were substantial, with independent negative effects on hatching success, fledging success and recruitment success (Table 2; Fig. 2). Although the effect of inbreeding fij on recruitment success was not statistically significant, the effect was large, in the expected direction and consistent across the different data sets (Tables 2 and 3). A comparison of the effect size caused by inbreeding across data sets with different degrees of restrictiveness can be found in Table 3: overall, although the effect size of inbreeding apparently increased with data set restrictiveness, it was also accompanied by greater standard errors around the estimates.

Table 2.   Minimal adequate models with binomial proportions investigating the effect of inbreeding on hatching success (n = 3777), fledging success (n = 3766) and recruitment success (n = 4250).
Minimum adequate modelHatching successFledging successRecruitment success
d.f.DevianceP-valueParameter estimate (SE)d.f.DevianceP-valueParameter estimate (SE)d.f.DevianceP-valueParameter estimate (SE)
  1. Each trait is corrected for the trait preceding it in development (hatching success is modelled as a binomial proportion of brood size over clutch size, fledging success is a binomial proportion of number of fledged young over brood size, and recruitment success is a binomial proportion of number recruited offspring over the number of fledged offspring). Note that the parameter estimates are dependent on the logit link function used in the GLM. The estimated number of lethal equivalents (B) is presented for each life-history stage.

Year39485< 0.001 39768< 0.001 40609< 0.001 
Sector    8310.017 841< 0.001 
Female age869< 0.001         
Male age7290.003         
Mean egg weight175< 0.0011.576 (0.208)1170.0010.759 (0.234)    
Egg-laying date    161< 0.0010.031 (0.005)1110< 0.001−0.036 (0.004)
Inbreeding136< 0.0014.069 (0.703)123< 0.0013.443 (0.865)130.0931.536 (0.957)
Brood size        118< 0.001−0.051 (0.013)
Mean fledging weight        164< 0.0010.015 (0.002)
Residual deviance37765639  37657303  42496013  
Lethal equivalents (B)0.410.421.30
Lethals expressed till adulthood2.12
Figure 2.

 Inbreeding at each life-history stage in great tits. To visualize the effect of inbreeding at each life-history stage, figures show residuals from simple models where (a) brood size was corrected for clutch size, (b) the number of fledged offspring was corrected for brood size, and (c) the number of recruited offspring was corrected for number of fledged offspring.

Table 3.   Effects of inbreeding fij in data sets with varying degrees of restrictiveness: minimum adequate models from Table 2 were tested in data sets where a varying number of ancestors were known.
 Hatching successFledging successRecruitment success
DevianceP-valueParameter estimate (SE)DevianceP-valueParameter estimate (SE)DevianceP-valueParameter estimate (SE)
  1. The first value for deviance indicates the amount of deviance explained by inbreeding; Δtot is the total deviance explained by the model.

Both parents known37
Δtot = 6962
n = 4685
< 0.001−4.085 (0.699)23
Δtot = 9222
n = 4685
< 0.001−3.422 (0.866)3
Δtot = 7365
n = 5277
0.150−1.316 (0.951)
At least 1 grandparent known36
Δtot = 5639
n = 3777
< 0.001−4.069 (0.703)23
Δtot = 7303
n = 3766
< 0.001−3.443 (0.865)3
Δtot = 6013
n = 4250
0.093−1.536 (0.957)
All 4 grandparents known38
Δtot = 1887
n = 1277
< 0.001−4.994 (0.834)17
Δtot = 2526
n = 1265
0.002−4.11 (1.19)4
Δtot = 2001
n = 1423
0.085−1.90 (1.16)

The overall number of lethal equivalents carried by the population until adulthood was 2.12 per haploid genome (Table 2), which is equivalent to 2B = 4.2 lethal equivalents. The number of lethal equivalents was similar for two relatively short juvenile life-history stages, namely survival to hatching and to fledging, where B = 0.4, but the value increased for recruitment success (B = 1.3), which was measured over the course of almost a full year (i.e. between fledging and breeding the following year).

In contrast to offspring life-history traits, mean fledging weight was not affected by inbreeding fij [Waldd.f.=1 = 1.88, P = 0.170, n = 3650, parameter estimate (SE): −10.3 (7.32)], and similarly, there was no effect of inbreeding fi on wing length or mass in recruited birds [effect of inbreeding on wing length: Waldd.f.=1 = 0.43, P = 0.511, n = 3961, parameter estimate (SE): 11.14 (17.0); effect of inbreeding on mass: Waldd.f.=1 = 0.71, P = 0.398, n = 3966, parameter estimate (SE): −7.81 (9.2)]. Hence, there was no evidence that any of the morphological traits available for analysis were influenced by parental relatedness.

Life-history traits of recruited inbred and outbred birds

Over 41 years of breeding events (1964–2004) where both parents and at least one grandparent were known, a total of 3770 offspring recruited, of which nine had an inbreeding coefficient of fi = 0.125 (seven males, two females), and 23 of fi = 0.25 (11 males and 12 females). There was a nonsignificant tendency for inbred males and females to breed later, and for females to produce a smaller clutch size (Table 4), all of which are consistent with continued effects of inbreeding. Although the sample was limited in size, a significant reduction in the reproductive success of recruited inbred individuals was detected: brood size (with clutch size fitted as a covariate) was significantly smaller in both inbred males and females (Table 4), suggesting that hatching success and/or parental care (till the time when brood size was recorded) was reduced in inbred breeding birds. The reduced brood size of inbred birds ultimately translated into a reduced lifetime reproductive success, as measured by their production of fledged young (Table 5a,b; Fig. 3). The effect of inbreeding fi on reproductive success may be because of lower rates of production of offspring, as the effect of inbreeding is still marked (although marginally nonsignificant) even when lifespan was added to the model (Table 5b). There was no evidence of any sex-specific effect of differential lifetime reproductive success in inbred males and females (f * sex interaction: P > 0.2).

Table 4.   Linear mixed models testing the effects of parental inbreeding fi on offspring traits, with parental identity and parental birth nest fitted as random effects.
Egg-laying dateClutch sizeMean egg weightMean fledging weightBrood size (clutch size corrected)No. of individuals fledged (brood size corrected)Recruited individuals (corrected for the nr of individuals fledged)
  1. The test-statistics and parameter estimates (+SE) for the effects of inbreeding are presented for males (n = 2870), females (n = 3071) and both sexes altogether (n = 5941). Significant values are presented in bold.

Males
Wald1 = 0.54
P = 0.464
6.81 (9.31)
Wald1 = 0.84
P = 0.359
1.62 (1.76)
Wald1 = 1.22
P = 0.269
0.17 (0.16)
Wald1 = 0.37
P = 0.543
8.37 (13.78)
Wald1 = 4.66
P = 0.031
2.86(1.32)
Wald1 = 0.70
P = 0.403
−1.75 (2.09)
Wald1 = 0.21
P = 0.644
−0.46 (1.00)
Females
Wald1 = 0.33
P = 0.565
5.50 (9.56)
Wald1 = 1.25
P = 0.263
−2.07 (1.85)
Wald1 = 0.08
P = 0.772
0.04 (0.15)
Wald1 = 0.59
P = 0.442
11.02 (14.35)
Wald1 = 3.75
P = 0.053
−3.65 (1.88)
Wald1 = 0.10
P = 0.754
0.67 (2.15)
Wald1 = 0.07
P = 0.792
−0.23 (0.88)
Combined
Wald1 = 0.92
P = 0.338
6.74 (7.03)
Wald1 = 0.03
P = 0.863
−0.22 (1.30)
Wald1 = 0.64
P = 0.423
0.087 (0.11)
Wald1 = 1.01
P = 0.316
10.19 (10.16)
Wald1 = 8.03
P = 0.005
3.37 (1.19)
Wald1 = 0.07
P = 0.786
−0.40 (1.50)
Wald1 = 0.25
P = 0.614
−0.34 (0.68)
Table 5.   Linear mixed model of the effect of inbreeding fi on lifetime reproductive success without (a) and with (b) lifespan and lifetime clutch size fitted as a covariates (n = 3770).
 d.f.(a) Lifetime no. of fledged offspring(b) Lifetime no. of fledged offspring
Wald/d.f.P-valueParameter estimate (SE)Wald/d.f.P-valueParameter estimate (SE)
  1. Parental birth nest is fitted as random effect to control for the nonindependence of breeding individuals; interference was dropped from (b) as P > 0.2.

Interference246.18< 0.001    
Inbreeding13.710.0541.274 (0.662)3.140.0760.90 (0.51)
Sex119.57< 0.001 46.83< 0.001 
Lifespan1   7.350.0070.04 (0.015)
Lifetime clutch size1   766.1< 0.0010.056 (0.002)
Figure 3.

 Lifetime reproductive success of recruited offspring, calculated in terms of total number of nestlings reared till fledging during an individual's lifetime (n = 3738, 9 and 23 for fi = 0.0, 0.125 and 0.25 respectively). To maximize the sample size of raw data, individuals that encountered ‘interference’ in their lifetime were not excluded from the mean calculations that were used to create the graph above; however, a variable ‘interference’ was fitted in all models.

Cumulative effects of inbreeding

There was a significant effect of parental relatedness fij on every life-history stage investigated (Table 6), and as an increasing number of life-history stages were taken into account, the effect size of inbreeding increased (Fig. 4). While inbreeding at fij = 0.25 decreased fledging success by 16%, it decreased recruitment success by 39% and the number of fledged grand offspring by 55% (Fig. 4). Hence, the effect size of inbreeding is probably best estimated over the entire lifespan, as our analyses suggest that there are independent and additive effects at each of the major life-history stages of an individual's life.

Table 6.   Linear mixed model of the cumulative effects of inbreeding fij, with male and female identity fitted as random effects.
 d.f. Brood size (n = 4310) Fledged (n = 4307) Recruited (n = 4328)Fledged grand offspring (n = 4368)Recruited grand offspring (n = 4368)
Wald/d.f.P-valueWald/d.f.P-valueWald/d.f.P-valueWald/d.f.P-valueWald/d.f.P-value
  1. Terms where P < 0.2 were dropped from the model (see text for details). Parameter estimates for inbreeding (+SE) are presented below the test-statistic.

Year4021.27< 0.00118.39< 0.00117.17< 0.00112.88< 0.0018.97< 0.001
Sector88.21< 0.0017.35< 0.0015.44< 0.0014.42< 0.0014.38< 0.001
Female age94.19< 0.0012.560.006      
Male age72.580.0122.770.0071.730.098    
Egg-laying date1204.52< 0.001256.58< 0.001133.47< 0.00192.51< 0.00157.23< 0.001
Inbreeding16.57
2.56 (1.00)
0.01018.74
4.93 (1.14)
< 0.0016.95
1.51 (0.57)
0.0087.36
19.01 (7.00)
0.0075.20
2.38 (1.04)
0.023
Interference6      105.91< 0.00152.02< 0.001
Figure 4.

 Cumulative effect size of inbreeding in great tits. Effect size of inbreeding increases with the number of life-history stages that are included. Grey (fij = 0.0) and black bars (fij = 0.25) represent fitted trait values based on the models shown in Table 6. Outbred fitted values were scaled to 1.

Discussion

Our study of the effects of inbreeding in a wild population of great tits reveals that inbreeding causes a substantial decline in fitness. Inbreeding depression independently affected hatching, fledging and recruitment success; it also affected the breeding performance of recruited inbred offspring, which were found to have reduced brood size and lifetime reproductive success. Although the mean population levels of inbreeding were low, individual effects were strong and halved the production of grand offspring relative to outbred nests. Our results also show that the effect size of inbreeding depression varied depending on the life-history framework considered, and suggest that strong selection on inbreeding avoidance mechanisms should be operating.

The low level of inbreeding we found contrasts with some other studies of wild populations where the mean inbreeding coefficient has accumulated over time, mainly because of geographic isolation, limited population size and reduced gene flow (Keller, 1998; Kalinowski et al., 2000). Interestingly, population inbreeding levels dropped at least twofold relative to the first decade of study, which may have been caused by higher immigration rates into the population, or changes in natal dispersal distances. Although close inbreeding (i.e. mating with close kin where fij ≥ 0.125) was recorded in only 1.3% of all breeding events, the fitness consequences of such matings were substantial, as inbreeding depression caused a 39% difference in recruitment success between breeding events with fij = 0.25 relative to an outbred mating, and a 55% decrease in the mean number of fledged grand offspring relative to an outbred nest. The effect size of inbreeding observed in the Wytham great tit population does not fully follow the trends set by other studies of inbreeding in wild birds. In an insular population of song sparrows, Keller (1998) found that inbreeding at fij = 0.25 reduced survival from egg to breeding age by 49%, whereas in a collared flycatcher population on the island of Gotland Kruuk et al. (2002) found a reduction in fitness of 95% over the same life-history stages. Van Noordwijk & Scharloo (1981) studied inbreeding in an island great tit population characterized by high levels of gene flow (Postma & van Noordwijk, 2005), but despite a relatively large sample size (n = 783 breeding pairs) they did not document any reduction in recruitment success caused by inbreeding.

Such discrepancies in estimates of the genetic load in different wild populations might be explained by limitations of sample size, although each of these studies had relatively large samples for analysis. Although the genetic basis of inbreeding depression are dominance interactions (Lynch & Walsh, 1998), little is known about the extent of differences in nonadditive interactions between different populations of the same species, or closely related species (e.g. passerine birds). It is to be expected that dominance effects would differ in populations with different gene pools and levels of gene flow (Moreno, 1994; Liberg et al., 2005). Further studies investigating the genetic architecture of subdivided populations (as in Glemin et al., 2003; Theodorou & Couvet, 2006), or of closely related species (as in Schiffer et al., 2006), are needed to improve our understanding of forces shaping dominance interactions in different environments and in populations with varying gene pools.

The field methods used, the assumptions underlying the analyses, but also some characteristics of the population may potentially affect quantitative aspects of inbreeding and inbreeding depression for any study. In the present case, first, there may be bias in parental recaptures. Because parents are only identified in the mid- to late nestling period, there is a bias towards recapture of most successful parents (McCleery et al., 2004), leading to an over-inclusion of successful parents in the pedigree. It is thus possible that the inbreeding levels are higher than our estimates, but because of parental desertion in cases of complete nest failure, this possibility is difficult to assess. Second, extra-pair fertilizations in birds result in some offspring being reared by their nonbiological fathers, which will ultimately lead to the false categorization of some outbred offspring as inbred. In the context of our variable fij with three levels of inbreeding, it is unlikely that pedigree errors will have a substantial effect on our results. The occurrence of extra-pair paternity will not influence the inbreeding coefficient for mother–son matings; in the case of brother–sister matings, siblings would still share maternal genes, and in this case still be related as fij = 0.125. In this population, estimates of the proportion of extra-pair young are in the order of 15% (Blakey, 1994; S. Patrick, unpublished work); hence, the underestimation of the effect size of inbreeding depression at the brood level should not be substantial. Finally, the categorical treatment of the inbreeding coefficient (all inbreeding coefficients < 0.125 were categorized as fij = 0, as we are not able to identify many cases of less close inbreeding without losing statistical power) may lead to an underestimate of the effect size of inbreeding caused by low coefficients of fij. Indeed, analyses with the original coefficients of inbreeding (ranging from fij = 0.000 to 0.258) sometimes yielded stronger effects of inbreeding in the traits examined than the synthetic variable which was used in all final analyses of inbreeding depression (see Table 4).

Overall, the potential biases in our analyses (outlined above) are most likely to lead to a downward bias in the estimate of the occurrence of inbreeding and in the inbreeding depression effect size. However, we do not expect that those biases would cause substantial changes in our estimates, because the inbreeding events that are least affected (fewest pedigree links involved) are those between first-order relatives, which have the biggest effect on the estimates of inbreeding depression. But as with any field study, the potential for bias must be kept in mind, as not only can it affect our inference as to the true costs of inbreeding depression, but it may also influence theoretical predictions as to whether a given population should be expected to express inbreeding avoidance behaviour (see Bengtsson, 1978; May, 1979; Packer, 1979; Kokko & Ots, 2006).

There was little evidence that birds involved in matings with close kin differed from individuals mating with unrelated kin, as assessed by egg-laying date, egg mass or clutch size. We have good evidence that the effect of inbreeding estimated within individuals is of similar magnitude to that estimated across individuals (as in this study) (M. Szulkin and B.C. Sheldon, unpublished work). Hence, the estimates of inbreeding depression presented here are not likely to be confounded by the effects of individual quality (see Kruuk et al., 2002; Reid et al., 2006 for potential examples). In contrast to the absence of differences in parental traits (i.e. individuals mating with related and unrelated partners), we observed marked effects of inbreeding on a range of life-history traits, from hatching through to the production of independent offspring. It is important to emphasize that we have approached the question of whether inbreeding depression occurs independently at each life-history stage from a statistical perspective. Our data do not, however, allow us to test whether inbreeding effects at successive life-history stages, might be because of pleiotropic effects of genes causing inbreeding depression.

Although the effect of parental relatedness fij on recruitment was nonsignificant when inbreeding depression in other life-history components was controlled for, and the effect of inbreeding fj on lifetime reproductive success marginally nonsignificant, we suggest that this is mainly the result of low statistical power given the low frequency of matings between relatives, as the estimated effect sizes were large, in the expected direction, and consistent across analyses. The situation where estimated effect sizes of inbreeding are large, but where confidence in these effect sizes is low, highlights the difficulties of accurately detecting inbreeding depression in large populations in the wild where inbreeding events are rare (Kruuk et al., 2002; M. Szulkin and B.C. Sheldon, unpublished data). The case of three studies of inbreeding in Wytham great tits, carried out in different years with varying sample size, illustrates this problem. Bulmer (1973) reported a decrease in recruitment success in inbred birds based on a sample of 397 matings, although the result was not subject to any formal statistical analysis; Greenwood et al. (1978) found no effect of inbreeding on recruitment (n = 885), whereas our study (n = 4523) found a marginal, yet biologically important effect of inbreeding on recruitment, when controlling for the effects of inbreeding at earlier life-history stages.

Pedigrees are not always available to infer relatedness and estimate inbreeding depression; our study differs in some respects with results obtained from a Belgian great tit population, in which Van de Casteele et al. (2003) used microsatellite data to estimate coefficients of relatedness between breeding partners, and found a negative relationship between microsatellite-inferred relatedness and hatching rates and nestling survival, but no effect on fledging and recruitment success. Because there was no difference in terms of the number of recruits originating from matings with different estimated coefficients of relatedness, the authors concluded that inbreeding depression does not seem to affect fitness in a considerable way, and that compensatory mechanisms (such as divorce) may additionally alleviate inbreeding depression experienced in previous breeding events. Our results suggest, in contrast, that there are substantial fitness effects of inbreeding. Irrespective of the factors causing differences between the two studies (e.g. pedigree vs. microsatellites-based estimation of relatedness, genetic characteristics of the population), we suggest that it is not easy to compensate for such large fitness costs experienced while inbreeding: not only are the fitness costs in our population visible throughout an individual's life, but also compensation mechanisms, such as divorce, may have limited scope given that great tits have low annual survival (typically < 50%).

The absence of an effect of parental relatedness fij on offspring fledging mass and on adult size measures is also consistent with the theoretical expectation that the magnitude of inbreeding depression will be small in morphological traits, as those have little dominance variance (Lynch & Walsh, 1998; DeRose & Roff, 1999). Interestingly, our results contrast with analyses of weight in some laboratory and controlled breeding studies where significant effects of inbreeding were found on mass in the fruit fly (Roff, 2002), sheep (Wiener et al., 1992) and cattle (Smith et al., 1998). Our results may suggest that there are no dominance interactions in genes coding for mass in great tits, and are consistent with the suggestion that body mass is subject to different forms of natural selection in volant birds with determinate growth as opposed to some other taxa.

There is a paucity of data on inbreeding depression in adult morphological and life-history traits (Keller, 1998; Hoglund et al., 2002), mainly caused by the difficulty of measuring them in wild and unmanaged populations. A lack of evidence for inbreeding depression in post-juvenile life-history stages has been found in many studies of pedigreed animal populations (Greenwood et al., 1978; Packer, 1979; Van Noordwijk & Scharloo, 1981; Gibbs & Grant, 1989; Hoogland, 1992; Grant & Grant, 1995; Charpentier et al., 2006), in contrast to Jimenez et al. (1994), Keller (1998), Daniels & Walters (2000) and Kruuk et al. (2002) who have found clear evidence of inbreeding depression for these traits. There are no theoretical grounds which would argue that inbreeding depression should not be observed in later life-history stages. On the contrary, evolutionary theories of senescence predict that deleterious mutations would be expressed in later life-history stages because of weaker selection against late-acting mutations (Stearns, 1992; Keller, 1998). It is thus important to emphasize that in our study population, the expression of deleterious recessive alleles was detected in early, juvenile and adult life-history traits. Our study also showed an effect of inbreeding on adult reproductive traits and lifetime reproductive success, the latter most likely being independent of lifespan. The decrease in lifetime reproductive success in inbred birds was not caused by inbreeding in their offspring, as all were mated with unrelated partners. Our results thus clearly show that deleterious recessives affect every life-history stage investigated. Altogether our results show that the more life-history traits were included in an analysis of inbreeding effects on fitness, the greater the effect size of inbreeding was established (see Fig. 4). Consequently, studies investigating the costs of inbreeding based on only a part of the life history are likely to underestimate the effects of inbreeding (see also Van Noordwijk & Scharloo, 1981).

A recent theoretical paper (Kokko & Ots, 2006) showed that the costs incurred through inbreeding may potentially be exceeded by the kin-selected benefits associated with producing offspring of increased mean relatedness. Being able to quantify the overall costs of inbreeding will allow a more complete picture to be drawn regarding the processes that act to increase and reduce the frequency of inbreeding. Our observation that inbreeding depression acts throughout the life history of inbred individuals also has implications for captive breeding programs where reintroductions into the wild are planned. If recessive deleterious alleles are continuously expressed, as suggested by our results, then one cannot assume that individuals who survive a certain life stage will be free of inbreeding depression. This in turn decreases chances of successful reintroductions of inbred endangered species back into the wild. To reliably predict population dynamics and demographic parameters of species with small population size, the effects of inbreeding over the entire lifespan should be taken into consideration.

Acknowledgments

This research was supported by scholarships to M.S. from the Christopher Welch Trust and The Queen's College; D.G. was supported by the Natural Sciences and Engineering Research Council of Canada (NSERC) and a grant from the BBSRC to B.C.S. and L.E.B. Kruuk, and B.C.S. by a Royal Society University Research Fellowship during the initial stages of the work. We are very grateful to the many generations of fieldworkers without whom this study would have not been possible. Finally, we thank L. Keller and one anonymous referee for their detailed comments, which substantially improved the quality of this manuscript.

Ancillary