You are on page 1of 7

Opinion

Inbreeding avoidance, tolerance,


or preference in animals?
Marta Szulkin1,2*, Katie V. Stopher3, Josephine M. Pemberton3, and Jane M. Reid4*
1
Centre d’Ecologie Fonctionnelle et Evolutive, UMR 5175 Campus CNRS, 1919 Route de Mende, 34293 Montpellier cedex 5, France
2
Edward Grey Institute, Zoology Department, University of Oxford, South Parks Road, Oxford, OX1 3PS, UK
3
Institute of Evolutionary Biology, School of Biological Sciences, University of Edinburgh, West Mains Road, Edinburgh,
EH9 3JT, UK
4
Institute of Biological & Environmental Sciences, School of Biological Sciences, University of Aberdeen, Zoology Building,
Tillydrone Avenue, Aberdeen AB24 2TZ, UK

Animal ecologists commonly assume that the reduced testing whether animals should or do avoid, tolerate, or
fitness that often afflicts inbred offspring will inevitably prefer relatives as mates is challenging. Few, if any, em-
cause selection for inbreeding avoidance. Although ear- pirical tests have an explicit theoretical basis or fulfil all
ly empirical studies often reported inbreeding avoid- key requirements to quantify and interpret departures
ance, recent studies suggest that animals sometimes from random inbreeding. Consequently, general conclu-
show no avoidance or even prefer to mate with relatives. sions regarding inbreeding strategies (see Glossary) can-
However, current theory is insufficient to predict wheth- not yet be drawn.
er animals should avoid, tolerate, or prefer inbreeding In many populations, dispersal reduces the probability
and hence to understand overall inbreeding strategy. that inbreeding will occur and thereby causes inbreeding
Furthermore, quantifying inbreeding strategy is chal- avoidance without the need for mating bias against rela-
lenging, requiring relatedness among unbiased sets of tives during post-dispersal life stages [19–21]. This out-
actual and potential mates to be accurately estimated. come arises irrespective of whether dispersal evolved due
Here, we highlight key limitations of current theory and to selection for inbreeding avoidance or due to other forces
empirical tests, and summarise the advances required such as local resource availability or kin competition
to predict, quantify, and understand animal inbreeding [13,22]. However, when dispersal is limited, inbreeding
strategies. strategy, consisting of inbreeding avoidance, tolerance,
or preference, could be expressed through mate choice
Inbreeding depression and inbreeding avoidance among philopatric relatives [3,11,23]. Here, we highlight
Inbreeding depression (the reduced fitness of inbred indi- limitations of existing theory and data that impede the
viduals resulting from reproduction among relatives) is clear prediction and analysis of inbreeding strategies in
widespread in wild, domesticated, and experimental popu- philopatric animals, and outline theoretical and empirical
lations [1,2]. Animal ecologists commonly assume that
inbreeding depression will drive the evolution of reproduc-
tive strategies that result in inbreeding avoidance, includ-
ing sex-biased dispersal, specific choice of unrelated mates, Glossary
and associated mechanisms of kin recognition and sperm Coefficient of inbreeding f: the probability that two alleles at any locus in an
selection [3–5]. The assumption that inbreeding avoidance individual are identical by descent (i.e., that they are copies of the same alleles
in a common ancestor). Inbreeding, or mating between relatives, produces
must be adaptive because inbreeding reduces offspring offspring with f > 0.
fitness pervades numerous hypotheses explaining repro- Coefficient of kinship k: the probability that homologous alleles sampled from
ductive strategy and the design and interpretation of two individuals are identical by descent. Usually equal to half the coefficient of
relatedness between the two individuals, and to the coefficient of inbreeding of
empirical tests [4,6–9]. By contrast, long-standing theory offspring that would be produced if the two individuals mated.
shows that the basic assumption that inbreeding depres- Coefficient of relatedness r: the proportion of all alleles that two individuals
share that are identical by descent due to recent common ancestry (Box 2).
sion will inevitably cause selection for inbreeding avoid-
Identity by descent: alleles that are identical because they are copies of the
ance is simplistic because it ignores the potential benefits same ancestral allele.
of inbreeding and costs of inbreeding avoidance [10–13]. Identity by state: identical alleles that are not necessarily copies of the same
ancestral allele.
Against this conflicting backdrop, there has been a recent Inbreeding strategy: the overall system of inbreeding, comprising inbreeding
surge of interest in testing for departures from the distri- avoidance (avoidance of mating with relatives, resulting in negative deviation
bution of inbreeding expected given random mating in from relatedness expected under random mating), inbreeding tolerance
(absence of inbreeding preference or avoidance, resulting in inbreeding
animals, and empirical studies report diverse patterns occurring to the degree that would be expected given random mating among
(e.g., [6,11,14–18]) (Figure 1). However, predicting and population members), or inbreeding preference (preference for mating with
relatives, resulting in positive deviation from relatedness expected under
random mating).
Corresponding author: Szulkin, M. (marta.szulkin@zoo.ox.ac.uk) Inclusive fitness: the relative number of gene copies an individual leaves in the
Keywords: inbreeding avoidance; inbreeding preference; inbreeding strategy; next generation through its own direct reproduction and through reproduction
inclusive fitness; mate choice; relatedness. of relatives.
*
These authors contributed equally to this work.

0169-5347/$ – see front matter ß 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tree.2012.10.016 Trends in Ecology and Evolution, April 2013, Vol. 28, No. 4 205
Opinion Trends in Ecology and Evolution April 2013, Vol. 28, No. 4

(a) (b) (c) (d)

(e) (f) (g) (h)

TRENDS in Ecology & Evolution

Figure 1. Empirical studies of inbreeding strategy have reported inbreeding avoidance in (a) red-winged fairy wrens Malurus elegans [6]; (b) bank voles Myodes glareolus
[15]; (c) zebra finches Taeniopygia guttata [14]; (d) grey wolves Canis lupus [30]; inbreeding tolerance in (e) New Zealand robins Petroica australis [11]; (f) bighorn sheep
Ovis canadensis [16]; (g) great tits Parus major [17] (but see [66,67]); and inbreeding preference in (h) cichlid fish Pelvicachromis taeniatus [18]. Photographs reproduced,
with permission, from L. Brouwer (a), M.D. Thom (b), J. Rutkowska (c), S. Kabessa (d), I. Jamieson (e), A.M. Martin (f), T. Veen (g), and H. Kullmann (h).

advances that are required to understand the occurrence, They also highlighted that selection depends on the fitness
ecology, and evolution of such strategies. costs of avoiding inbreeding, for example through dispersal
or searching for unrelated mates, and on the degree to
Theory of inbreeding strategy which mating with a relative impedes mating with other
Inbreeding and inclusive fitness non-relatives (Box 1) [34]. Recent extensions consider si-
The widespread presumption that inbreeding depression multaneous versus sequential mate choice and relative
will inevitably cause net selection for inbreeding avoidance investment in parental care, and show that the magnitude
ignores the inclusive fitness benefit of inbreeding that exists of inbreeding depression under which inbreeding could be
because parents are more closely related to inbred offspring adaptive is even greater than previously predicted [12].
than to outbred offspring. This increased relatedness arises Links between costs of inbreeding avoidance, life history
because inbred offspring, by definition, inherit alleles from and ecology have also been further emphasised, but such
one parent that are identical by descent to those carried by, discussions have not always incorporated the inclusive
and potentially inherited from, the other parent. Inbreeding fitness benefit of inbreeding [11,23].
thereby increases the expected proportion of alleles that any However, key limitations of existing theory inhibit
one parent shares with its offspring, including alleles prediction of inbreeding strategies. First, basic models
favouring inbreeding [10,24,25]. This inclusive fitness ben- predict sexual conflict over inbreeding; females should
efit underpins extensive theory predicting the evolution of tolerate less inbreeding than males under certain cir-
self-fertilisation in plants [10,25–27]. That this same benefit cumstances (Box 1) [28,34]. However, these ‘battle-
could also cause the evolution of biparental inbreeding was ground’ models do not consider the resolution of
suggested three decades ago (Box 1) [28,29], but remained sexual conflict or predict what sex-specific inbreeding
widely ignored by animal ecologists. The assumption that strategies might consequently evolve [35]. One notable
inbreeding depression inevitably causes selection for in- model showed that female mate choice for inbreeding
breeding avoidance continued to pervade hypotheses avoidance could drive male-biased dispersal, but attrib-
explaining dispersal, mate choice, and polyandry uted mating decisions entirely to singly mated females
[3,5,6,11,30–32]. Recent theory has re-emphasised that bi- and did not consider selection on inbreeding in males
parental inbreeding could be adaptive, even given inbreed- [13]. A recent analysis of optimal inbreeding also consid-
ing depression, with the inclusive fitness benefit phrased as ered female fitness only [33].
‘helping relatives to breed’ and hence kin selection Second, existing models consider inbreeding between
[12,13,33]. However, existing theory does not predict what two relatives within an otherwise outbred population
sex-specific inbreeding strategies should evolve in popula- [12,28,34], or consider mean population-level relatedness
tions of potentially inbreeding animals. and inbreeding [13,33]. They do not explicitly consider
populations comprising individuals of varying relatedness,
Limitations of current theory in which different degrees of biparental inbreeding are
Early models of biparental inbreeding demonstrated the likely and selection on inbreeding strategy could be strong.
fundamental conceptual point that there will only be se- This restriction is important because selection on inbreed-
lection for inbreeding avoidance if the costs of inbreeding ing avoidance might depend on the probability that a
exceed the inclusive fitness benefit (Box 1) [28,29,34,35]. potential mate that a focal individual rejects to avoid
206
Opinion Trends in Ecology and Evolution April 2013, Vol. 28, No. 4

Box 1. Predicting conditions favouring sex-specific inbreeding avoidance or preference


A basic model predicting conditions under which females and males weighs inbreeding depression. By contrast, male fitness is maximised
should prefer or avoid inbreeding balances the inclusive fitness by inbreeding with a sister if d < 2/3. This basic model therefore
benefit of inbreeding (which arises because parents are more closely predicts sexual conflict over inbreeding; females should avoid
related to inbred offspring) against fitness costs of avoiding inbreed- inbreeding given smaller d than males [34,35] (Figure I).
ing, reduced fitness in inbred offspring (inbreeding depression), and This basic ‘battleground’ model demonstrates that inbreeding
any reduction in further matings due to inbreeding [34]. Assuming an should not necessarily be avoided even if d > 0, and identifies key
otherwise outbred population, females and males should mate with a parameters on which sex-specific thresholds for inbreeding avoid-
relative if (Equations [I] and [II]): ance depend. However, it does not predict equilibrium inbreeding
strategies that might evolve in animals with biparental inbreeding.
Existing models of selfing are not directly applicable because they
d < 1 þ ½DNf ðcf  1Þ þ rDNm ðcm  1Þ=ð1 þ rÞ; [I]
consider the special case of r = 1, where there is no sexual conflict
d < 1 þ ½DNm ðcm  1Þ þ rDNf ðcf  1Þ=ð1 þ rÞ; [II] (Figure I).

respectively, where d is the magnitude of inbreeding depression (the


reduction in reproductive success through inbreeding relative to 1.0

Magnitude of inbreeding depression


outbreeding), DNf and DNm are the degrees to which females and 0.9
males forfeit effective outbred matings by inbreeding, respectively, cf 0.8
and cm are the costs of avoiding inbreeding for females and males, 0.7
respectively, and r is the relatedness between prospective mates.
0.6
If there is zero cost of inbreeding avoidance (cf = cm = 0) and both

(δ)
0.5
sexes forfeit all other reproductive opportunities by inbreeding (DNf =
0.4
DNm = 1, i.e., strict monogamy), both expressions reduce to d < 0 and
both sexes should avoid inbreeding unless inbreeding depression is 0.3
negative (i.e., reproductive success is greater through inbreeding than 0.2
outbreeding). 0.1
However, if mating with a related female does not restrict the 0.0
mating opportunity of a male (DNm = 0, i.e., polygyny), then d < r/(1+r) 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
and d <1/(1+r) for females and males, respectively (Figure I). Conse- Relatedness to potenal mate (r)
quently, fitness can be maximised by inbreeding rather than by TRENDS in Ecology & Evolution
outbreeding, even with inbreeding depression (d > 0). Furthermore,
the value of d at which fitness is maximised by outbreeding rather than Figure I. Under polygyny, inbreeding is favoured in both sexes in the grey zone of
by inbreeding differs between males and females. Given the option to the parameter space, males only (black zone) or neither sex (white zone). There is
mate with a brother (r = ½) or outbreed, female fitness is maximised therefore sexual conflict over inbreeding in the black zone. Adapted, with
by inbreeding if d < 1/3, because the inclusive fitness benefit out- permission, from [34].

inbreeding will subsequently mate with a different relative Quantifying inbreeding strategy: approaches and
and hence inbreed anyway. challenges
Third, existing models of biparental inbreeding in ani- Both observational and experimental studies will be re-
mals do not consider coevolution of inbreeding and inbreed- quired to quantify whether animals avoid, tolerate, or
ing depression, or their coevolution with other prefer inbreeding across a range of biologically relevant
reproductive strategies, such as polyandry, that could conditions and, hence, to test evolutionary predictions.
potentially facilitate inbreeding preference or avoidance Such studies should comprise five key components: (i) an
[31]. Predictions of plant selfing theory change substan- appropriate system; identification of (ii) actual and (iii)
tially when analogous constraints are relaxed, for example potential mates; (iv) accurate measurement of relatedness;
by allowing purging of inbreeding load, reduced outcross- and (v) appropriate statistical analyses and interpretation.
ing given selfing (‘pollen discounting’), or ecological varia- These requirements might seem obvious, but arguably no
tion in cross-fertilisation failure or inbreeding depression empirical studies have fulfilled them all, especially in wild
[10,26,27]. populations. We highlight key approaches and pitfalls, and
Given that sexual conflict over selfing is not expected, thereby outline a framework for future empirical test.
existing selfing theory cannot be directly applied to predict
the evolution of biparental inbreeding strategies (Box 1). Choosing an appropriate study system
New theory is therefore required. Existing analytical mod- There is limited value in testing for inbreeding preference
els should be extended to quantify the effects of variation in or avoidance in systems where there is little variation in
relatedness on sex-specific selection on inbreeding strate- relatedness between potential mates and, therefore, little
gy. Models predicting the evolutionary resolution of sexual opportunity to express inbreeding strategy. This applies to
conflict should then be formulated. The need to track sex- populations whose sizes, ecologies, or life histories mean
specific selection and feedbacks between inbreeding strat- that relatives rarely interact [17,37,38], and to highly
egy and relatedness, and complex genetic architecture inbred populations comprising uniformly close relatives
underpinning the coevolution of inbreeding strategy and [8]. Rather, the most informative and powerful tests will
inbreeding depression, means that analytical solutions concern populations where there is sufficient variation in
might be restrictive. Individual-based models will then relatedness to create opportunity for inbreeding preference
prove fruitful, especially given the need to consider small and avoidance, and for tolerance to occur as an evolved
populations where variation in relatedness and inbreeding rather than an inevitable constrained outcome. Studies
risk prevail (e.g., [36]). would usefully focus on populations that have experienced
207
Opinion Trends in Ecology and Evolution April 2013, Vol. 28, No. 4

such conditions for numerous generations, and hence ex- which the relatedness of an individual to its actual mate
perienced long-term selection on inbreeding strategy. can be tested. Identifying potential mates is harder than
Small populations where there is currently opportunity identifying actual mates because, by definition, rejected
for inbreeding, but which recently derived from large mates cannot be identified by observing pairings or par-
panmictic populations where inbreeding was previously entage. Moreover, it is often simplistic to assume that all
rare, might not support useful hypothesis tests regarding opposite-sex adults are available as potential mates [44].
predicted inbreeding strategy [11,23,37]. Identifying potential mates therefore requires detailed,
individual-based knowledge of population ecology and life
Identifying actual mates history, ideally including the location of each individual
Directly quantifying inbreeding strategy requires the ac- [6,17,45], timing of settlement [46], age [16,47], and cur-
tual mate of all (or an unbiased sample of) breeding rent or previous reproductive status [17,23]. Truly unbi-
individuals, or the parents of an unbiased sample of result- ased sampling with respect to such variables, which might
ing offspring, to be identified. Mate identities are known co-vary with relatedness, is hard. Field studies are almost
when specific monogamous pairings are experimentally inevitably spatially restricted, meaning that distant or
created [18], but often unknown in observational studies transient potential mates are missed. If relatedness varies
and systems with multiple mating. Whereas mothers are spatially, which is likely when dispersal is restricted and
often identifiable from parental behaviour, paternity fre- inbreeding occurs, unidentified potential mates might be
quently remains unassigned, uncertain, or misassigned. relatively unrelated to focal individuals [45]. Individuals
Accurate molecular genetic paternity assignment is ham- that fail to mate and breed, or whose breeding attempt fails
pered by incomplete or biased sampling of offspring and early, might also go unobserved and hence be excluded as
potential parents, and insufficient or inaccurate genotyp- potential mates [48]. This is problematic if such individuals
ing [39,40]. Furthermore, and critically given the aim of are non-random with respect to relatedness, for example
quantifying inbreeding strategy, the ability to identify due to kin-selected cooperative breeding or reproductive
sires can vary with relatedness. The presence of relatives suppression, or if breeding attempts involving relatives are
within a population means that multiple individuals share more or less successful. Naive sampling can also confound
numerous alleles, resulting in error-prone parentage as- sets of more or less related individuals that are not repro-
signment [41]. Immigrant or transient males might be less ductively active simultaneously and consequently bias
likely to be sampled and also less closely related to resident inferred sets of potential mates [49]. Unbiased tests of
females [42]. These biases are likely to arise in exactly the inbreeding strategy might therefore require populations
populations with high variance in relatedness where adap- where all or almost all adults can be identified with high
tive inbreeding strategies might be expected. Moreover, confidence, and where detailed individual-based data can
early-life mortality due to inbreeding depression can bias be used to define appropriate null models [6,11,23,50].
sampling against inbred offspring, meaning that inbred Given that such models will rarely be demonstrably unbi-
matings go undetected [4]. Mate and parentage assign- ased, the most robust approach might be to formulate sets
ment for the purpose of quantifying inbreeding strategy of increasingly constrained models and verify whether
therefore requires particularly comprehensive or carefully conclusions regarding inbreeding strategy are consistent
structured sampling of offspring and parents, and infor- [16,17,23,30,47,51]. Meeting these conditions might re-
mative molecular markers, to ensure that identified mates quire studies that are explicitly designed to identify unbi-
are unbiased with respect to relatedness. ased sets of potential mates rather than simply relying on
existing data. When potential mates cannot be identified,
Identifying potential mates an alternative approach is to compare relatedness between
A further major requirement is to provide or identify an individual and its first and subsequent mates, or social
appropriate sets of potential mates for focal individuals. and extra-pair mates [6,52,53]. This approach has the
Specific sets of potential mates can be experimentally advantage that both mates can be identified, but the dis-
provided, but should be carefully chosen to ensure biologi- advantages that it measures conditional rather than abso-
cal relevance. Combinations of non-relatives and familiar lute inbreeding strategy and excludes singly mated
or unfamiliar full- or half-sibs are often provided, probably individuals.
because such individuals can be readily produced through
single generations of controlled breeding [18,37,43]. How- Estimating relatedness
ever, dichotomous choice experiments involving sibs and Quantifying inbreeding strategy requires the relatedness
non-relatives will be hard to interpret if sib encounters are between an individual and its actual and potential mates,
rare in nature, and cannot reveal nonlinear inbreeding or the inbreeding coefficients of actual and potential off-
preference or avoidance. Experiments that provide a great- spring, to be estimated. Expected coefficients of related-
er diversity of relatives as potential mates might therefore ness between any two individuals, and corresponding
be more insightful, but will require more complex breeding offspring inbreeding coefficients, can be calculated from
designs to create appropriate individuals of standardised pedigree data irrespective of whether the individuals mat-
age and reproductive history. ed or reproduced (Box 2) [16,17,23]. However, reconstruct-
A major challenge facing observational studies is to ing pedigrees of sufficient depth, completeness, and
identify all the potential mates of an individual, or an accuracy to quantify inbreeding strategy is challenging,
unbiased sample of those mates with respect to related- especially for wild populations. A single pedigree must link
ness, and hence define appropriate null models against not only focal individuals and their actual mates, but also
208
Opinion Trends in Ecology and Evolution April 2013, Vol. 28, No. 4

Box 2. Approaches to estimating relatedness by the same unsampled or unidentified male, the standard
The broad concept of ‘relatedness’ can be refined and measured in
assumption that each offspring had a different unrelated
numerous ways, with different assumptions and utilities [58,68,69]. sire will also cause underestimation of relatedness. Ana-
Understanding inbreeding strategy requires estimation of the lysts should minimise these biases by reconstructing sib-
coefficient of relatedness (r) defined as the proportion of alleles ships [55], implementing ‘gene-dropping’ analyses [56],
that individuals share that are identical by descent due to recent assuming population mean rather than zero relatedness
common ancestry and, hence, relative to a defined baseline
population [58].
among unidentified parents [56], or by applying integrated
Expected coefficients of relatedness (r) and kinship (k) between analyses that allow parentage and effects of relatedness on
pairs of individuals, and coefficients of inbreeding ( f) of actual or parentage to be jointly estimated [39]. The increasing
hypothetical offspring, can be calculated from pedigree data availability of abundant and accurate SNP markers will
assuming Mendelian inheritance, where individual alleles have allow improved parentage inference. However, even so-
parent–offspring transmission probabilities of 0.5. Expected r
between any two individuals x and y is (0.5)n, where n is the
phisticated genomic analysis is unlikely to provide a quick
number of pedigree links connecting x and y. When x and y have or infallible substitute for meticulous collection of the
multiple common ancestors, coefficients stemming from multiple comprehensive multigenerational behavioural, life-histo-
paths are summed. Expected k and f equal r/2 for any pair or ry, and genotypic data that allow complete, accurate pedi-
offspring, respectively. These calculations can incorporate multiple
gree reconstruction [48]. Given that such data are also
common ancestors and hence inbreeding, but values are relative to
defined pedigree founders that are typically assumed to be outbred needed to define appropriate sets of potential mates, stud-
and unrelated. Distant common ancestors contribute little variation ies that fail to collect them might struggle to elucidate
to r and k among contemporary individuals, or f of potential inbreeding strategy.
offspring, because n is large. Consequently, pedigrees covering c. 5– The difficulty of pedigree reconstruction has prompted
10 generations might be sufficient to detect relatively subtle
attempts to estimate pairwise relatedness directly from
inbreeding strategy.
Realised relatedness can be estimated directly from co-dominant genotype data (Box 2). This approach appears beguilingly
marker genotypes (e.g., microsatellites or SNPs) because more easier; it requires similar genotypic data but eliminates
closely related individuals are expected to share more alleles. The the parentage assignment step. Several methods for esti-
simplest estimator is ‘allele sharing’: the number of alleles shared mating relatedness from marker data have been proposed
by two individuals divided by the total number of loci considered.
However, individuals can have identical alleles at a given locus
(Box 2) [57]. Such methods assume that identity by state is
(identity by state) because of their pedigree relation (identity by correlated with identity by descent, and that available
descent) or recurrent mutation. markers accurately measure genome-wide identity by
Identity by state is common if an allele occurs at high frequency in state. In fact, microsatellites frequently mutate to new
the focal population. Therefore, more advanced relatedness esti- alleles that are indistinguishable from existing alleles,
mators incorporate information on population allele frequencies
(estimated from sampled individuals) [70]. These comprise ‘like-
impeding inference of identity by descent from identity
lihood’ estimators [71] and the widely used ‘method-of-moments’ by state, and small sets of markers do not always reliably
estimators. The latter comprise ‘correlation’ estimators that provide capture genome-wide identity by state [58]. Although
a single relatedness estimate for each pair of individuals [72], and marker-based estimators will capture coarse variation in
‘regression’ estimators that compare the genotype of one individual
relatedness in populations where substantial variation
to another and can thus generate different estimates depending on
which individual is compared to which [73,74]. These various exists, they are unlikely to detect subtle variation in
relatedness estimators have different precision and bias, which relatedness that is of interest in the context of inbreeding
depend on characteristics of the study system. Extensive compar- strategy [57]. The approach of estimating parental relat-
isons, reviews, and critiques of these methods are available edness from marker-based offspring heterozygosity (e.g.,
[57,70,75].
[59]) is even less precise than estimating pairwise related-
ness from the genotypes of both parents because it uses
half the information. Indeed, the marker heterozygosity of
all (or an unbiased sample of) their potential mates, and an individual often only weakly predicts its coefficient of
encompass sufficient generations to quantify subtle varia- inbreeding [60,61]. Finally, estimates of parental related-
tion in relatedness (Box 2). These specifications are more ness can be biased if the same markers are used to assign
demanding than those required to estimate inbreeding paternity and estimate relatedness. Such spurious asso-
depression: given that individuals with unknown ancestors ciations can be avoided by using independent markers for
are still the actual or potential mates of other individuals, the two tasks [62,63].
they cannot simply be excluded from analyses (e.g., Although the most informative approach to estimating
[42,54]). relatedness in non-model organisms is currently to use
Pedigree reconstruction involves assembling numerous complete pedigree data, high-density genotyping or se-
parentage assignments. Assignment error can therefore quencing might in future provide more precise estimation.
introduce error and bias into pedigree relatedness esti- Pedigree analysis provides unbiased estimates of expected
mates throughout descendent generations. When an indi- pairwise relatedness but, except for parent–offspring rela-
vidual has no assigned sire (or dam), relatedness is tionships, realised relatedness varies due to recombination
typically calculated assuming that the unknown parent and sampling variance [64]. High-density genotyping
is unrelated to all other pedigree members [42,54]. If allows estimation of realised rather than expected related-
parentage assignment failed precisely because there are ness. This approach might be particularly valuable in
multiple closely related potential parents [41], estimates of evaluating inbreeding strategies in species with long gen-
relatedness among descendants will probably be down- eration times, where sufficient pedigree data cannot real-
wardly biased. Similarly, if multiple offspring were sired istically be collected.
209
Opinion Trends in Ecology and Evolution April 2013, Vol. 28, No. 4

Statistical analyses and interpretation rigorous and informative test and are vulnerable to bias,
Once actual and potential mates are identified and related- meaning that existing conclusions regarding inbreeding
ness estimated, the final hurdle is appropriate statistical strategy should be interpreted cautiously. However, oppor-
analysis. The most obvious and widely implemented analy- tunities to overcome these challenges are available in the
sis is simply to quantify the mean difference in relatedness form of individual-based field studies, carefully designed
between the actual versus potential mates of an individual experiments, and molecular genetic tools that, in combina-
[17,23]. However, several extensions are desirable. It might tion, should allow relatedness between individuals and
prove insightful to quantify other metrics of relatedness actual or potential mates to be adequately estimated. This
distributions, such as the variance. If some individuals avoid objective requires detailed knowledge of the biological
inbreeding whereas others preferentially inbreed, mean system, comprehensive or carefully designed data collec-
observed relatedness could match null expectation but with tion, and sophisticated molecular genetic and statistical
increased variance. Furthermore, unexpected patterns of analysis; empiricists should no longer be content to accept
specific types of inbreeding could occur even when mean current standards with their inherent and typically
relatedness does not differ from expectation [16,17,23]. unquantified biases and uncertainty. Meanwhile, the ulti-
Frequencies of specific categories of inbreeding, including mate aim of discerning general principles will be facilitated
between familiar and unfamiliar relatives, should therefore if authors objectively document the specific strengths and
be quantified. This could facilitate detection of inbreeding limitations of their data, analysis, and inference.
preference mediated by differential imprinting on familiar
versus unfamiliar relatives [17,23,65]. Moreover, because Acknowledgements
overall inbreeding strategy often stems from multiple mate We thank Magdalen College, Oxford and an IEF Marie Curie Fellowship
choice events occurring at different stages throughout an (M.S.), NERC (K.V.S.), the Royal Society (J.M.R.), and the ERC (J.M.P.)
for funding and Deborah Charlesworth, Lukas Keller, Ashleigh Griffin,
individual’s life history (e.g., [17,21,66,67]), analyses should
and two anonymous referees for helpful comments.
ideally consider these stages separately as well as combined
across its lifetime. Inbreeding strategies manifested at dif-
References
ferent life stages should therefore be quantified indepen- 1 Charlesworth, B. and Charlesworth, D. (1999) The genetic basis of
dently. Finally, observed inbreeding strategy should be inbreeding depression. Genet. Res. 74, 329–340
interpreted in the context of current and historical popula- 2 Keller, L.F. and Waller, D.M. (2002) Inbreeding effects in wild
tion ecology because behaviours that previously generated populations. Trends Ecol. Evol. 17, 230–241
3 Blouin, S.F. and Blouin, M. (1988) Inbreeding avoidance behaviours.
inbreeding avoidance or preference might be less effective in Trends Ecol. Evol. 3, 230–233
contemporary ecological conditions [30]. 4 Bretman, A. et al. (2004) Molecular evidence of post-copulatory
When neither inbreeding avoidance nor preference is inbreeding avoidance in the field cricket Gryllus bimaculatus. Proc.
detected, appropriate power analysis should be provided to R. Soc. Lond. B: Biol. Sci. 271, 159–164
quantify the ability of a study to detect departure from 5 Pusey, A. and Wolf, M. (1996) Inbreeding avoidance in animals. Trends
Ecol. Evol. 11, 201–206
random inbreeding. Such power analysis is required to 6 Brouwer, L. et al. (2011) Strategic promiscuity helps avoid inbreeding
support any inference of inbreeding tolerance and, depend- at multiple levels in a cooperative breeder where both sexes are
ing on the null hypothesis, might require simulations that philopatric. Mol. Ecol. 20, 4796–4807
are tailored to specific data sets [8,16,23]. Potential bias 7 Tregenza, T. and Wedell, N. (2002) Polyandrous females avoid costs of
inbreeding. Nature 415, 71–73
and uncertainty stemming from error and uncertainty in
8 Billing, A.M. et al. (2012) Evidence of inbreeding depression but not
mate identities and associated relatedness estimates inbreeding avoidance in a natural house sparrow population. Mol. Ecol.
should also be explicitly considered. Sampling variance 21, 1487–1499
associated with marker-based relatedness estimates might 9 Nelson-Flower, M.J. et al. (2012) Inbreeding avoidance mechanisms:
mean that many studies have little power to detect in- dispersal dynamics in cooperatively breeding southern pied babblers.
J. Anim. Ecol. 81, 876–883
breeding strategy. Integrated analyses that simultaneous-
10 Charlesworth, D. (2006) Evolution of plant breeding systems. Curr.
ly estimate uncertainty in paternity and/or relatedness, Biol. 16, R726–R735
and in resulting relationships between mating or parent- 11 Jamieson, I.G. et al. (2009) Why some species of birds do not avoid
age and relatedness might provide the most powerful and inbreeding: insights from New Zealand robins and saddlebacks. Behav.
robust statistical approach. This will require existing an- Ecol. 20, 575–584
12 Kokko, H. and Ots, I. (2006) When not to avoid inbreeding. Evolution
alytical frameworks to be extended [39]. 60, 467–475
13 Lehmann, L. and Perrin, N. (2003) Inbreeding avoidance through kin
Concluding remarks recognition: choosy females boost male dispersal. Am. Nat. 162, 638–652
Inbreeding shapes genotypic variation and has profound 14 Arct, A. et al. (2010) Kin recognition and adjustment of reproductive
implications for mating system evolution [1,10,27,28,42]. effort in zebra finches. Biol. Lett. 6, 762–764
15 Lemaitre, J.F. et al. (2012) Inbreeding avoidance behaviour of male
Consequently, understanding inbreeding strategy is cen- bank voles in relation to social status. Anim. Behav. 83, 453–457
tral to evolutionary ecology, but substantial advances in 16 Rioux-Paquette, E. et al. (2010) No inbreeding avoidance in an isolated
theory and empirical tests are still required. Explicit the- population of bighorn sheep. Anim. Behav. 80, 865–871
oretical predictions pertaining to inbreeding strategy in 17 Szulkin, M. et al. (2009) Inbreeding avoidance under different null
nonselfing animals are scant. Models that predict the models of random mating in the great tit. J. Anim. Ecol. 78, 778–788
18 Thunken, T. et al. (2007) Active inbreeding in a cichlid fish and its
evolution of inbreeding strategy under sexual conflict, adaptive significance. Curr. Biol. 17, 225–229
including feedbacks with population relatedness structure 19 Lebigre, C. et al. (2010) Female-biased dispersal alone can reduce the
and inbreeding depression, need to be formulated. Existing occurrence of inbreeding in black grouse (Tetrao tetrix). Mol. Ecol. 19,
empirical studies do not fulfil all key requirements for 1929–1939

210
Opinion Trends in Ecology and Evolution April 2013, Vol. 28, No. 4

20 Pärt, T. (1996) Problems with testing inbreeding avoidance: the case of 48 Sardell, R.J. et al. (2010) Comprehensive paternity assignment:
the collared flycatcher. Evolution 50, 1625–1630 genotype, spatial location and social status in song sparrows,
21 Szulkin, M. and Sheldon, B.C. (2008) Dispersal as a means of Melospiza melodia. Mol. Ecol. 19, 4352–4364
inbreeding avoidance in a wild bird population. Proc. R. Soc. B: 49 Robinson, S.P. et al. (2012) Assortative mating for relatedness in a
Biol. Sci. 275, 703–711 large naturally occurring population of Drosophila melanogaster. J.
22 Pusey, A.E. (1987) Sex-biased dispersal and inbreeding avoidance in Evol. Biol. 25, 716–725
birds and mammals. Trends Ecol. Evol. 2, 295–299 50 Reid, J.M. et al. (2008) Individual phenotype, kinship, and the
23 Keller, L.F. and Arcese, P. (1998) No evidence for inbreeding avoidance occurrence of inbreeding in song sparrows. Evolution 62, 887–899
in a natural population of song sparrows (Melospiza melodia). Am. Nat. 51 Stopher, K.V. et al. (2012) Re-mating across years and intralineage
152, 380–392 polygyny are associated with greater than expected levels of inbreeding
24 Hill, W.G. and Weir, B.S. (2007) Prediction of multi-locus inbreeding in wild red deer. J. Evol. Biol. 25, 2457–2469
coefficients and relation to linkage disequilibrium in random mating 52 Varian-Ramos, C.W. and Webster, M.S. (2012) Extrapair copulations
populations. Theor. Popul. Biol. 72, 179–185 reduce inbreeding for female red-backed fairy-wrens, Malurus
25 Fisher, R.A. (1941) Average excess and average effect of a gene melanocephalus. Anim. Behav. 83, 857–864
substitution. Ann. Eugen. 11, 53–63 53 Hansson, B. et al. (2004) Do female great reed warblers seek extra-pair
26 Lande, R. and Schemske, D.W. (1985) The evolution of self-fertilization fertilizations to avoid inbreeding? Proc. R. Soc. Lond. B: Biol. Sci. 271,
and inbreeding depression in plants. 1. Genetic models. Evolution 39, S290–S292
24–40 54 Szulkin, M. et al. (2007) Inbreeding depression along a life-history
27 Goodwillie, C. et al. (2005) The evolutionary enigma of mixed mating continuum in the great tit. J. Evol. Biol. 20, 1531–1543
systems in plants: occurrence, theoretical explanations, and empirical 55 Wang, J. and Santure, A.W. (2009) Parentage and sibship inference
evidence. Annu. Rev. Ecol. Evol. Syst. 36, 47–79 from multilocus genotype data under polygamy. Genetics 181, 1579–
28 Parker, G.A. (1979) Sexual selection and sexual conflict. In 1594
Sexual Selection and Reproductive Competition in Insects 56 Ewing, S.R. et al. (2008) Inbreeding and loss of genetic variation in a
(Blum, M.S. and Blum, N.A., eds), pp. 123–166, Academic reintroduced population of Mauritius kestrel. Conserv. Biol. 22, 395–
Press 404
29 Smith, R.H. (1979) Selection for inbreeding in polygynous animals. 57 Csillery, K. et al. (2006) Performance of marker-based relatedness
Heredity 43, 205–211 estimators in natural populations of outbred vertebrates. Genetics
30 Geffen, E. et al. (2011) Kin encounter rate and inbreeding avoidance in 173, 2091–2101
canids. Mol. Ecol. 20, 5348–5358 58 Blouin, M.S. (2003) DNA-based methods for pedigree reconstruction
31 Bretman, A. et al. (2009) Promiscuous females avoid inbreeding by and kinship analysis in natural populations. Trends Ecol. Evol. 18,
controlling sperm storage. Mol. Ecol. 18, 3340–3345 503–511
32 Tregenza, T. and Wedell, N. (2000) Genetic compatibility, mate choice 59 Frere, C.H. et al. (2010) Inbreeding tolerance and fitness costs in wild
and patterns of parentage: invited review. Mol. Ecol. 9, 1013–1027 bottlenose dolphins. Proc. R. Soc. B: Biol. Sci. 277, 2667–2673
33 Puurtinen, M. et al. (2011) Mate choice for optimal (k)inbreeding. 60 Chapman, J.R. et al. (2009) A quantitative review of heterozygosity-
Evolution 65, 1501–1505 fitness correlations in animal populations. Mol. Ecol. 18, 2746–2765
34 Waser, P.M. et al. (1986) When should animals tolerate inbreeding? 61 Szulkin, M. et al. (2010) Heterozygosity-fitness correlations: a time for
Am. Nat. 128, 529–537 reappraisal. Evolution 64, 1202–1217
35 Parker, G.A. (2006) Sexual conflict over mating and fertilization: an 62 Wang, J.L. (2010) Do marker-based paternity assignments favour
overview. Philos. Trans. R. Soc. Lond. B: Biol. Sci. 361, 235–259 heterozygous and unrelated males? Mol. Ecol. 19, 1898–1913
36 Young, A.G. et al. (2012) Non-additive effects of pollen limitation and 63 Wetzel, D.P. and Westneat, D.F. (2009) Heterozygosity and extra-pair
self-incompatibility reduce plant reproductive success and population paternity: biased tests result from the use of shared markers. Mol. Ecol.
viability. Ann. Bot. 109, 643–653 18, 2010–2021
37 Ala-Honkola, O. et al. (2011) No evidence for postcopulatory inbreeding 64 Hill, W.G. and Weir, B.S. (2011) Variation in actual relationship as
avoidance in Drosophila melanogaster. Evolution 65, 2699–2705 a consequence of Mendelian sampling and linkage. Genet. Res. 93,
38 Alho, J.S. et al. (2012) No evidence for inbreeding avoidance through 47–64
active mate choice in red-billed gulls. Behav. Ecol. 23, 672–675 65 Shutler, D. et al. (2004) Breeding between tree swallows from the same
39 Hadfield, J.D. et al. (2006) Towards unbiased parentage assignment: brood. J. Field Ornithol. 75, 353–358
combining genetic, behavioural and spatial data in a Bayesian 66 Szulkin, M. et al. (2012) Promiscuity, inbreeding and dispersal
framework. Mol. Ecol. 15, 3715–3730 propensity in great tits. Anim. Behav. 84, 1363–1370
40 Wang, J.L. (2010) Effects of genotyping errors on parentage exclusion 67 Szulkin, M. and Sheldon, B.C. (2008) Correlates of the occurrence of
analysis. Mol. Ecol. 19, 5061–5078 inbreeding in a wild bird population. Behav. Ecol. 19, 1200–1207
41 Double, M.C. et al. (1997) Exclusion probabilities for single-locus 68 Rousset, F. (2002) Inbreeding and relatedness coefficients: what do
paternity analysis when related males compete for matings. Mol. they measure? Heredity 88, 371–380
Ecol. 6, 1155–1166 69 Wang, J. (2011) Unbiased relatedness estimation in structured
42 Keller, L.F. (1998) Inbreeding and its fitness effects in an insular populations. Genetics 187, 887–901
population of song sparrows (Melospiza melodia). Evolution 52, 240–250 70 Thomas, S.C. (2005) The estimation of genetic relationships using
43 Gasparini, C. and Pilastro, A. (2011) Cryptic female preference for molecular markers and their efficiency in estimating heritability in
genetically unrelated males is mediated by ovarian fluid in the guppy. natural populations. Philos. Trans. R. Soc. B. 360, 1457–1467
Proc. R. Soc. B: Biol. Sci. 278, 2495–2501 71 Milligan, B.G. (2003) Maximum-likelihood estimation of relatedness.
44 Wang, C. and Lu, X. (2011) Female ground tits prefer relatives as extra- Genetics 163, 1153–1167
pair partners: driven by kin-selection? Mol. Ecol. 20, 2851–2863 72 Wang, J.L. (2002) An estimator for pairwise relatedness using
45 Foerster, K. et al. (2006) A spatial genetic structure and effects of molecular markers. Genetics 160, 1203–1215
relatedness on mate choice in a wild bird population. Mol. Ecol. 15, 73 Lynch, M. and Ritland, K. (1999) Estimation of pairwise relatedness
4555–4567 with molecular markers. Genetics 152, 1753–1766
46 Hansson, B. et al. (2007) No evidence for inbreeding avoidance in a 74 Queller, D.C. and Goodnight, K.F. (1989) Estimating relatedness using
great reed warbler population. Behav. Ecol. 18, 157–164 genetic markers. Evolution 43, 258–275
47 Soulsbury, C.D. et al. (2012) Restrictive mate choice criteria cause age- 75 Bink, M.C.A.M. et al. (2008) Comparison of marker-based pairwise
specific inbreeding in female black grouse, Tetrao tetrix. Anim. Behav. relatedness estimators on a pedigreed plant population. Theor. Appl.
83, 1497–1503 Genet. 117, 843–855

211

You might also like