Abstract
B4galnt2 is a blood group-related glycosyltransferase that displays cis-regulatory
variation for its tissue-specific expression patterns in house mice. The wild type allele, found e.g.
in the C57BL/6J laboratory mouse strain, directs intestinal expression of B4galnt2, which is the
pattern observed among vertebrates, including humans. An alternative allele class found in the
RIIIS/J strain and other mice instead drives expression in blood vessels, which leads to a
phenotype similar to type 1 von Willebrand disease (VWD), a common human bleeding disorder.
Previous studies showed that alternative B4galnt2 alleles are subject to long-term balancing
selection in mice and that variation in B4galnt2 expression influences host-microbe interactions in
the intestine. This suggests that the cost of prolonged bleeding in RIIIS/J allele-bearing mice might
be outweighed by benefits associated with resistance against gastrointestinal pathogens.
However, the conditions under which such trade-offs could lead to the long-term maintenance of
disease-associated variation at B4galnt2 are unclear.
To understand and characterize the potential pathogen-driven selection acting on B4galnt2
in the wild, I first developed a mathematical model based on an evolutionary game framework
with a modified Wright-Fisher process, adjusted to implement diploid individuals. In particular, I
focused on heterozygous mice, which express B4galnt2 in both blood vessels and the
gastrointestinal tract. By comparing simulated to natural populations, I found that the genotype
frequencies observed in nature can be produced by pathogen-driven selection when (i) the fitness
cost of bleeding is roughly half that of infection and (ii) both heterozygotes and RIIIS/J
homozygotes are protected against infection. The resistance of the heterozygote individuals
indicates that a dominant protective function of the RIIIS/J allele is more likely than a protective
loss of intestinal expression. However, the nature of the dominant protective function of the RIIIS/J
allele remains unknown, as the model suggests that the associated vascular expression is not
necessarily linked to the pathogen resistance.
Furthermore, I aimed to identify potential pathogens driving the selection at B4galnt2 by
sampling and phenotyping over 200 newly collected mice from Southern France, where an
intermediate frequency of the RIIIS/J allele is present. Through the multilayer analysis of genetic
patterns, signs of inflammation, and intestinal microbial communities, I could associate several
bacterial genera to patterns consistent with genotype-dependent host-pathogen interaction. One
genus in particular, Morganella, is a likely candidate as it is a well-known opportunistic pathogen
and its abundance, prevalence and activity patterns are associated with increased inflammation in mice with intestinal expression of B4galnt2. Finally, I could identify the relevant species of
Morganella, which represents a new subspecies of the Morganella morganii group, and possesses
virulence-related genes absent from the other Morganella species, which may account for its
potential to drive selection at B4galnt2 via genotype-dependent host-pathogen interactions.
In conclusion, my work provides new insights into the potential evolutionary dynamics taking
place at B4galnt2 in wild populations of house mice, showing that pathogen-driven selection is a
likely cause for the maintenance of both B4galnt2 alleles in the wild. Moreover, my work could be
applied beyond the scope of murine glycosyltransferases, as the methods that I developed can
easily be generalized to other biological models.
