veys of wild populations, suggesting that mating prefer-
ences could indeed be the elusive source of selection main-
taining MHC polymorphisms (Potts et al. 1991; Hedrick
1992). In this study, nine populations of wild-derived mice
carrying four possible MHC haplotypes were maintained in
seminatural enclosures until approximately 150 pups were
born in each enclosure. MHC genotypes of pups born to
the original founders revealed a significant 27% reduction
of homozygotes from random mating expectations. Com-
parisons with controlled laboratory breedings suggested
that abortional selection was not operating to alter geno-
type ratios (Alberts and Ober 1993). Moreover, genotypes
of unborn litters from female founders suggested the ho-
mozygote deficiency was not due to differential mortality of
neonates. Only MHC-based disassortative mating prefer-
ences could explain the pup genotype patterns. Behavioral
observations also strongly suggested that these disassor-
tative matings were primarily due to female choice, as fe-
males, not males, were observed leaving their territories to
engage in extra-pair matings; these matings explained three-
quarters of the heterozygote excess (Potts et al. 1992). Fur-
thermore, communal nesting and nursing tended to occur
among females with shared MHC haplotypes, providing the
first experimental support for a general role of the MHC
in kin recognition (Manning, Wakeland, and Potts 1992;
Manning et al. 1995).
In another seminatural enclosure study, Penn and Potts
(1998b) corroborated data from previous laboratory as-
says, showing that MHC-based mating preferences of fe-
males are established via phenotype matching. In other
words, females use odor information from their rearing as-
sociates (nestmates) rather than their own MHC odors to
inform mate selection (Beauchamp et al. 1988; Eklund
1997). Mating preferences of females that had been cross-
fostered onto MHC-dissimilar litters as pups were com-
pared to the preferences of in-fostered females (from MHC-
similar foster litters). In direct contrast to preferences of the
in-fostered control females, cross-fostered females pre-
ferred males matching their own MHC genotype over males
carrying the MHC genotypes of their foster families. This
study additionally established females as the choosier sex,
as a male-specific preference would have yielded similar
pup genotype frequencies across all populations, regardless
of the female fostering treatment.
Taken together, these seminatural population experi-
ments provide strong evidence that females learn their own
MHC identity from families, that this information is used
for mate choice and kin recognition (which could addition-
ally prevent inbreeding), and that mating behavior is ca-
pable of maintaining MHC polymorphisms, even in the ab-
sence of pathogens. Yet, all this work was done using mice
that differed by entire MHC haplotypes, with variation
among all classical antigen-presenting genes, as well as dif-
ferences at numerous linked genes associated with the MHC
region. Consequently, unambiguous evidence connecting
MHC-based mating preferences to the highly polymorphic
antigen-presenting genes is still lacking.
Recently, we used seminatural enclosures to study mice
that differed by naturally occurring MHC mutations at a
single antigen-presenting locus. On a homogeneous ge-
netic background, odors mediated by these allelic differ-
ences are distinguishable by untrained mice (Carroll et al.
2002). However, in the context of a randomly segregating
wild background, mice did not differentiate among MHC
types, mating randomly rather than disassortatively (Car-
roll 2001). This result underscores the need for more stud-
ies to disprove or validate MHC antigen-presenting alleles
as the ultimate source of MHC-based sexual selection. As
our study lacked power to detect less than a 10% deficiency
of homozygous pups, the result does not preclude the oper-
ation of low levels of sexual selection (even a 1% preference
favoring MHC-dissimilar mates can have substantial evo-
lutionary impact, propelling MHC-diversifying selection).
However, it does suggest that other genetic loci are ex-
tremely important in the expression of odor-based cues and
odor-based behavior. For example, major urinary proteins
(MUPs), the most abundant proteins in mouse urine, also
exhibit a high level of polymorphism, and are found at
levels 100,000-fold higher than MHC protein fragments
(Brennan 2001). Although unlinked and genetically un-
related to MHC, the family of genes giving rise to MUP
chemical signals is also used by mice for individual recog-
nition (Hurst et al. 2001) and is therefore one of the many
possible signals competing with MHC-based odors. The
rich source of odor from MUPs might easily overwhelm
new odors arising from novel MHC mutations. The absence
of MHC-based mating preferences for single gene differ-
ences suggests that MHC-based mating preferences may be
more important in maintaining already existing MHC di-
versity (balancing selection), than in driving the incorpora-
tion of new alleles (diversifying selection).Social structure and MHC-based selection
in other rodents
Although genetic diversity at various MHC loci has been
measured in a number of rodent species, and several of
these studies have found evidence for balancing selection at
MHC loci (Seddon and Baverstock 1999; Richman et al.
2001; Hambuch and Lacey 2002; Sommer 2003), very little
data beyond that obtained from house mice have revealed
MHC-based sexual selection. At best, there are limited data
in rodents suggesting a connection between MHC diversity
and social structure. For example, the endangered MalagasySexual Selection: Using Social Ecology to Determine Fitness Differences 61