Immunogenetics (2015) 67:323–335

DOI 10.1007/s00251-015-0835-4

ORIGINAL PAPER

Positive selection drives the evolution of a major

histocompatibility complex gene in an endangered Mexican
salamander species complex
Karen E. Tracy 1 & Karen M. Kiemnec-Tyburczy 1 & J. Andrew DeWoody 2 &
Gabriela Parra-Olea 3 & Kelly R. Zamudio 1

Received: 23 October 2014 / Accepted: 12 March 2015 / Published online: 7 April 2015
# Springer-Verlag Berlin Heidelberg 2015

Abstract Immune gene evolution can be critical to spe-
cies survival in the face of infectious disease. In partic-
ular, polymorphism in the genes of the major histocom-
patibility complex (MHC) helps vertebrates combat nov-
el and diverse pathogens by increasing the number of
pathogen-derived proteins that can initiate the host’s ac-
quired immune response. In this study, we used a com-
bination of presumably adaptive and neutral markers to
investigate MHC evolution in populations of five sala-
mander species within the Ambystoma velasci complex,
a group consisting of 15 recently diverged species, sev-
eral of which are endangered. We isolated 31 unique
MHC class II β alleles from 75 total individuals from
five species in this complex. MHC heterozygosity was
significantly lower than expected for all five species,
and we found no clear relationship between number of
MHC alleles and species range, life history, or level of
heterozygosity. We inferred a phylogeny representing
the evolutionary history of Ambystoma MHC, with
which we found signatures of positive selection on the
overall gene, putative peptide-binding residues, and al-
lelic lineages. We identified several instances of trans-
species polymorphism, a hallmark of balancing selection
observed in other groups of closely related species. In
contrast, we did not detect comparable allelic diversity
or signatures of selection on neutral loci. Additionally,
we identified 17 supertypes among the 44 unique
Ambystoma alleles, indicating that these sequences may
encode functionally distinct MHC variants. We therefore
have strong evidence that positive selection is a major
evolutionary force driving patterns of MHC polymor-
phism in this recently radiated species complex.
Keywords Ambystoma . Balancing selection . Disease .
Immunogenetics . MHC

Electronic supplementary material The online version of this article

(doi:10.1007/s00251-015-0835-4) contains supplementary material,
Introduction
which is available to authorized users.
Major histocompatibility complex (MHC) genes are common-
* Karen E. Tracy ly used to study immunogenetic evolution because they en-
ket45@cornell.edu code key proteins in the acquired immune response of verte-
brates (Hughes and Yeager 1998). Within vertebrates, mam-
1
Department of Ecology and Evolutionary Biology, Cornell
malian immune systems have received the most attention,
University, Ithaca, NY, USA while studies of amphibian immunity have primarily concen-
2
Department of Forestry and Natural Resources, Purdue University,
trated on two model species: the African clawed frog
West Lafayette, IN, USA (Xenopus laevis) and the Western clawed frog (Silurana
3
Departamento de Zoología, Instituto de Biología, Universidad
tropicalis) (Robert and Cohen 2011). Immunogenetic studies
Nacional Autonoma de México Ciudad Universitaria, Ciudad de of wild amphibians, particularly salamanders, are even more
México, Mexico limited. Our current knowledge of salamander cellular immu-
4
Present address: Graduate Group in Immunology, University of nity comes primarily from two closely related species: the
California Davis, Davis, CA, USA tiger salamander (Ambystoma tigrinum) and the axolotl
324
(Ambystoma mexicanum) (Chen and Robert 2011), with little
comparative data from other salamander species (but see
Cohen 1971). A. mexicanum laboratory strains are common,
useful models because of their regenerative capacity and rel-
ative immunodeficiency (Shaffer 1993; Tournefier et al. 1998;
Voss et al. 2009). However, A. mexicanum is critically endan-
gered in the wild and lab-bred individuals cannot be used to
accurately assess immunocompetence and adaptive potential
in natural populations.
Pathogen-driven declines and extinctions of species can
correlate with a lack of variation in and selection on immune
genes like MHC; studies of immune gene variation and selec-
tion in wild populations therefore help to interpret otherwise
poorly understood declines (Frank 2002). For example, de-
clines in endangered Tasmanian devil (Sarcophilus harrish)
populations devastated by transmissible facial tumors are as-
sociated with high sequence similarity (Siddle et al. 2007) and
low allelic diversity (Cheng et al. 2012) in the MHC class I
and II genes, respectively. Amphibians are one of the most
threatened vertebrate groups, and among other pressures,
emerging infectious diseases contribute considerably to the
global decline of amphibian populations (Daszak et al.
2003). Two pathogens with worldwide distributions, a chytrid
fungus (Batrachochytrium dendrobatidis, or Bd) and multiple
species of ranavirus (genus Ranavirus), have recently caused
serious amphibian population declines and extinctions
(Skerratt et al. 2007; Fisher et al. 2009; Gray et al. 2009).
Ranaviruses and Bd infect both salamander (urodele) and frog
(anuran) species (Fisher et al. 2009; Gray et al. 2009).
Recently, researchers discovered another species of chytrid
fungus (Batrachochytrium salamandrivorans) that may be
an obligate salamander pathogen (Martel et al. 2013).
Studies of evolution of amphibian immune genes and immu-
nodeficiency are therefore an important step in understanding
host response to pathogens.
In this study, we focus on five congeneric species in the
Ambystoma velasci complex: Ambystoma altamirani,
Ambystoma andersoni, Ambystoma dumerilii,
A. mexicanum, and A. velasci. This complex includes 15 spe-
cies distributed in central Mexico (Irschick and Shaffer 1997;
Frost 2014). Many of the species in the complex are endan-
gered or critically endangered (IUCN 2014), including
A. mexicanum, the original wild source for laboratory strains.
Our five focal species vary in life history traits, range size,
level of heterozygosity, and conservation status (Table 1,
Fig. 1). Although recent speciation within the complex has
made exact phylogenetic relationships between species diffi-
cult to determine (O’Neill et al. 2013), high divergence be-
tween all five populations at microsatellite markers suggests
that these Ambystoma species are genetically isolated (Parra-
Olea et al. 2012). Isolation and the repeated evolution of pae-
domorphosis in this complex may have contributed to speci-
ation events, recent bottlenecks, and low heterozygosity
Immunogenetics (2015) 67:323–335
values; previous microsatellite studies showed generally
higher allelic diversity in transforming than in paedomorphic
populations (Parra-Olea et al. 2012). This study illustrates the
utility of studying genetic variation, including variation in
MHC, in a system where recently diverged lineages have dif-
ferent evolutionary trajectories.
Immune genes of the MHC produce proteins that are crit-
ical for specific parasite detection in the host, meaning that
MHC allelic diversity directly influences vertebrate host abil-
ity to combat pathogens (ex. Wegner et al. 2003; Kurtz et al.
2004). As a result, selection often maintains unusually high
polymorphism at particular MHC loci (Hughes and Yeager
1998; Hedrick and Poulin 2002). Laboratory challenge exper-
iments, assays of natural populations, and simulations all
show that balancing selection and host-pathogen coevolution
contribute to MHC polymorphism (Hedrick and Poulin 2002;
Borghans et al. 2004; Li et al. 2014), which may be main-
tained by mate choice (Bos et al. 2009; Baratti et al. 2012).
Previous studies show that anuran MHC genotypes are asso-
ciated with resistance to Bd (Savage and Zamudio 2011) and
the bacterial pathogen Aeromonas hydrophila (Barribeau et al.
2008). Likewise, ranavirus infection status is associated with
particular MHC supertypes in wild populations of common
frogs (Rana temporaria) (Teacher et al. 2009), underscoring
the link between wildlife disease susceptibility and variation
of MHC immune genes.
Data on MHC for natural populations of salamanders are
still scarce. To date, MHC class II has only been characterized
(with a focus primarily on β loci) in seven species of the
approximately 684 known species of salamanders (Frost
2014). Four newt species have multiple MHC class II β loci
and potential pseudogenes (Babik et al. 2008; Babik et al.
2009; Nadachowska-Brzyska et al. 2012); the Chinese giant
salamander also has at least two MHC II β loci (Zhu et al.
2014). In contrast, the two Ambystoma salamanders studied
thus far only express a single class II β gene, the DAB locus
(Sammut et al. 1999; Laurens et al. 2001; Bos and DeWoody
2005). Studies of A. mexicanum class II MHC conflict in
estimates of allelic diversity and efficacy of antigen binding
(Tournefier et al. 1998; Laurens et al. 2001; Richman et al.
2007). Work on both A. mexicanum and A. tigrinum has re-
vealed alternative splicing of DAB mRNA transcripts, which
may or may not influence immunocompetence (Laurens et al.
2001; Bulut et al. 2008). More data from wild populations are
therefore needed to assess the general immunogenetic vari-
ability within Ambystoma.
In this study, we characterized genetic variation at the DAB
locus in populations of five Ambystoma salamander species.
We expected to find higher allelic diversity at and selection on
the DAB locus in our focal species than was found in previous
studies on lab-bred A. mexicanum because this would be con-
sistent with the results of other studies on wild salamander
populations (Richman et al. 2007). We investigated whether
Immunogenetics (2015) 67:323–335 325

Critically endangered

Critically endangered

Critically endangered
selection has occurred over the evolutionary history of these

Conservation status
diverging lineages by comparing DAB allelic diversity to that

Least concern
of presumably neutrally evolving markers and testing for sig-

Endangered
natures of positive selection on particular alleles and sites in
the DAB locus. We interpreted our results in the context of
each species’ unique ecology and life history. We predicted
that both allelic diversity and selection would be strongly in-
Obligate paedomorph

Obligate paedomorph

Obligate paedomorph
fluenced by the unique ecology and evolutionary history of
each species, with the least allelic diversity and weakest selec-
Transformer

Transformer
tion found in critically endangered, obligate paedomorphic
Life history

species with restricted ranges and low population

heterozygosity.

Wide distribution in Mexico

Multiple small populations

Materials and methods

Focal loci

The primary function of classical MHC genes is to initiate an

One lake

One lake

One lake

acquired immune cascade, which recognizes and eliminates

Range

pathogens that have invaded a host (Klein 1986). A mature

MHC protein binds to a pathogen-derived peptide and pre-
sents this epitope on the cell surface; this complex then inter-
Lat: 19.509, Long:
GPS coordinates

acts with T cells to trigger an immune response. We focused

Long: 101.64
Lat: 19.288
Lat: 19.053

Lat: 19.813

on one class of MHC, class II. Only specific immune cells

Lat: 19.63
−101.797

−99.103

−98.304
−99.311

express class II genes, which recognize extracellular patho-

Long:

Long:

Long:

gens and activate helper T cells to initiate an antibody re-

sponse (Hughes and Yeager 1998; Parham and Janeway
2005). At the DAB locus, the only known class II β chain
Lake Patzcuaro, State of Michoacan
Range, life history, and level of heterozygosity of five Ambystoma study species

Lake Zacapu, State of Michoacan

gene in Ambystoma, we investigated the second exon, which

Lake Zempoala and surrounding

Parque El Chamizal, San Martin

Texmelucan, State of Puebla

encodes the β1 domain. The β1 domain is essential for pep-

creeks, State of Morelos

tide binding (and therefore parasite recognition) in the host

Xochimilco, Mexico City

(Bulut et al. 2008).

Population sampled

We also amplified two reference loci, presumed to be neu-

trally evolving, from the focal species for comparison with
DAB. The nuclear reference locus is a portion of the 3′ un-
translated region (UTR) of G1D6, a gene that exhibits high
similarity to Homo sapiens karyopherin alpha 6. Previous
phylogenetic work on Ambystoma has utilized this marker
(Weisrock et al. 2006). Cytochrome b (cytb) is a protein-
Mean population heterozygositya

coding mitochondrial gene commonly used in amphibian phy-

logeographic studies (e.g., Mattoccia et al. 2011; Kieswetter
and Schneider 2013).
0.425

0.487
0.228

0.745

DNA extraction and PCR

Unknown

Parra-Olea et al. (2012)

We utilized Ambystoma liver, gill, or tail clippings from wild-

caught individuals at a single population of each species
(Parra-Olea et al. 2012). We also used five A. tigrinum sam-
A. mexicanum

ples collected from various wild populations in the USA and

A. altamirani

A. andersoni

A. dumerilii

A. velasci

Mexico (MVZ 144955, 173474, 187202, 230026, and

Table 1

Species

230144; deposited in the Museum of Vertebrate Biology,

UC Berkeley). We extracted genomic DNA using the
a
326
Fig. 1 Locality figure with range
for each focal species (IUCN
2014). The inset of central
Mexico shows the four small
species ranges filled in black
A. velasci
Km
Qiagen DNeasy Kit following manufacturer’s protocols
(Qiagen, Germantown, Maryland). Primers MHCβ1_F and
MHCβ1_R (Bos and DeWoody 2005) were used to amplify
the β1-containing region of the DAB locus in all species using
polymerase chain reaction (PCR). Amplification reactions
(15 μL of total volume) included 0.5 U GoTaq Flexi DNA
polymerase in 1× Flexi Buffer (Promega, Madison,
Wisconsin), 0.2 mM each deoxynucleoside triphosphates
(dNTP), 1.0 μM each primer, and 1.25 μL of DNA template.
The thermal profile consisted of an initial denaturation step at
95 °C for 3 min, 29 cycles of 95 °C for 30 s, 60 °C for 30 s,
and 72 °C for 30 s, and a final extension step of 72 °C for
7 min.
The G1D6 locus was amplified using primers G1D6 5.1
and G1D6 3.1 (Weisrock et al. 2006). Amplification reactions
(15 μL of total volume) included 0.6 U Taq DNA polymerase
in 1× PCR Reaction Buffer (Roche, Indianapolis, Indiana),
0.17 mM additional MgCl2, 0.25 mM each dNTP, 0.66 μM
each primer, and 1.25 μL of DNA template. The thermal pro-
file consisted of an initial denaturation step at 94 °C for 5 min,
34 cycles of 94 °C for 60 s, 60.2 °C for 60 s, and 72 °C for
60 s, and a final extension step of 72 °C for 5 min. The cytb
locus was amplified using primers MVZ15 and MVZ16
(Moritz et al. 1992). Amplification reactions (15 μL of total
volume) included 0.6 U Taq DNA polymerase in 1× PCR
Reaction Buffer, 0.5 mM additional MgCl2, 0.25 mM each
dNTP, 0.66 μM each primer, and 2 μL of DNA template.
The thermal profile consisted of an initial 94 °C for 5 min,
34 cycles of 94 °C for 60 s, 54.5 °C for 30 s, and 72 °C for
60 s, and a final step of 75 °C for 5 min. Negative controls
Immunogenetics (2015) 67:323–335
A. andersoni
A. dumerilii
A. mexicanum
A. altamirani
Km
were used for all PCRs to ensure that there was no
contamination.
Cloning and sequencing
MHC markers are highly polymorphic and often contain mul-
tiple variable sites that are difficult to correctly phase when
both alleles are sequenced simultaneously from PCR prod-
ucts. To avoid this potential issue, we cloned MHC amplicons
using the pGEM-T vector system (Promega Madison,
Wisconsin) following manufacturer instructions. We grew
transformed cells for 15 h at 37 °C on Luria agar plates with
ampicillin. We then amplified 8–12 colonies per individual
using PCR with M13 universal primers. We verified
amplicons on an agarose gel and prepared products in the
appropriate size range for sequencing by cleaning with an
exonuclease and alkaline phosphatase reaction: 1.5 U Exo I
and 0.15 U SAP (Affymetrix: USB, Santa Clara, California).
We sequenced the products using Big Dye v.3.1 chemistry on
an ABI 3730xl capillary DNA sequencer (Life Technologies,
Grand Island, New York). We sequenced the G1D6 and cytb
amplicons directly without cloning because cytb is haploid
and we did not expect G1D6 to be highly variable.
Assessment of genetic diversity and population genetics
We estimated allelic diversity and richness in FSTAT v.2.9.3.2
(Goudet 1995) for each species at all three loci: DAB, G1D6,
and cytb. Additionally, we searched for instances of trans-
species polymorphism (TSP), a phenomenon in which alleles
Immunogenetics (2015) 67:323–335
or allelic lineages are maintained across speciation events,
because TSP provides evidence of balancing selection
(Klein et al. 2007). We identified TSP by comparing each
unique allele in a given species to the alleles from all other
species in BioEdit 7.1.3 (Hall 1999).
For DAB and G1D6, we used HW-Quickcheck
(Kalinowski 2006) to quantify observed and expected hetero-
zygosities, and we used GenePop v.4.2 (Rousset 2008) to
determine pairwise F ST values. We used DnaSP v.5.10
(Librado and Rozas 2009) to estimate nucleotide diversity
per site (π), average number of nucleotide differences (k),
and number of segregating (variable) sites (S) for each species
at all three loci. To obtain estimates of how much differentia-
tion existed between the five species, we performed Fisher
exacts tests on pairwise FST values of DAB and G1D6 in
GenePop v.4.2 (Rousset 2008). The significance values were
calculated with Bonferroni corrections for multiple tests (ad-
justed α = 0.005) following the methods described by
Raymond and Rousset (1995). We created haplotype net-
works to visually compare allelic diversity at DAB, G1D6,
and cytb in Splitstree v.4.12.8 (Huson and Bryant 2006).
Assessment of potential functional diversity
To investigate how much total allelic diversity corresponds to
functional diversity in Ambystoma, we grouped MHC alleles
by Bsupertypes^ based on amino acid properties and putative
functional differences (Doytchinova and Flower 2005; Ellison
et al. 2012). Using this method, we separate alleles into
supertypes on the basis of similarities in structural and func-
tional features inferred from amino acid properties. After we
aligned translated amino acid sequences, we constructed a
matrix based on five amino acid principal property scales
(Sandberg et al. 1998). This matrix was based only on the
eight amino acid residues in our sequence that were both pre-
sumed to be peptide-binding residues and found to be under
positive selection (highlighted in gray in Fig. 4). In PAST v.
2.04 (Hammer et al. 2001), we ran cluster analyses on this
matrix to separate alleles based on the similarity of their amino
acid properties at our eight sites of interest. These analyses
used four different algorithms: Ward’s, paired correlation,
paired cosine, and paired Euclidian; we used bootstrapping
to assess support for each tree generated from the clustering,
and we defined Bsupertype^ as a group of alleles that clustered
together in all four trees.
Tests for recombination and phylogenetic reconstruction
Recombination can impact analyses of selection because for
recombining sequences, a single phylogenetic tree does not
accurately represent the history of the loci under study
(Schierup et al. 2001; Anisimova et al. 2003). To assess
whether the DAB, G1D6, and cytb alleles we recovered
327
exhibited signatures of recombination, we evaluated our
datasets using tests for significant correlation between physi-
cal distance along the gene and three measures of linkage
disequilibrium implemented in PERMUTE (Wilson and
McVean 2006). This program calculates r 2 (Hill and
Robertson 1968), D′ (Lewontin 1964), and G4 (McVean
et al. 2002). Significance was determined by permuting the
sites 999 times and recalculating the correlation coefficient for
each permuted dataset. We considered recombination breaks
in our sequences if the majority of the tests (e.g., two of three)
were statistically significant.
We used MrBayes v.3.2.1 (Ronquist et al. 2012) to recon-
struct the evolutionary relationships among Ambystoma DAB
alleles, including alleles previously sequenced by other inves-
tigators. The phylogeny includes all of our newly isolated
sequences, plus additional four alleles from A. mexicanum,
seven alleles from A. tigrinum, and an outgroup MHC class
II sequence from a Northern crested newt, Triturus cristatus
(Table S1). Although the outgroup sequence is part of a clade
named BDAB^ by Babik et al. (2009), phylogenetic analyses
including all Ambystoma DAB sequences and all T. cristatus
sequences indicated that, because of the high amount of se-
quence divergence between the sequences of the two salaman-
der groups, it is not possible to determine whether the
T. cristatus DAB or DBB sequences are orthologous to
Ambsytoma DAB (data not shown). The outgroup sequence
was therefore chosen because it was one of the sequences least
diverged from Ambystoma. We first used MrModelTest v.2.3
in PAUP v. 4.0b10 (Swofford 2003) to infer the best-fit model
on the basis of hierarchical likelihood ratio tests (hLRT). We
then ran an MCMC analysis to approximate the posterior
probabilities of trees on MrBayes, using the best-fit model
selected, for ten million generations, with the first million
discarded as burn-in. MrBayes produced summary statistics
from the resulting trees by assessing the proportion of output
trees with each taxon bipartition. A consensus, majority-rule
tree was also generated based on these summary statistics.
Tests of selection
To determine if the β1 domain in Ambystoma has any signa-
tures of positive selection, we implemented dN/dS nucleotide
substitution ratio tests using the phylogeny produced by
MrBayes. For tests of selection, all focal taxa were treated
together. Our dataset for the tests of selection was the same
as that used to reconstruct the evolutionary history of the DAB
locus. For tests of selection, we used the best fit model of
sequence evolution according to MrModelTest v.2.3. First,
we tested for a significant signal of selection over the entire
alignment of DAB alleles with PARRIS (Scheffler et al. 2006).
We also used this test to determine whether cytb was evolving
neutrally, using neighbor-joining trees generated by the pro-
gram. The 3′ untranslated region of G1D6 does not contain an
328 Immunogenetics (2015) 67:323–335

open reading frame, and thus we did not conduct substitution The number of unique DAB alleles found per species ranged
ratio tests on this gene. Second, we tested for individual resi- from three in A. altamirani to 11 in A. dumerilii (Table 2). Of
dues under selection using the mixed effects model of evolu- the 17 A. tigrinum and A. mexicanum alleles we retrieved from
tion (MEME), which uses codon-based maximum likelihood GenBank (Table S1), seven of these were identical to alleles we
to detect evidence of diversifying and episodic selection identified. The average nucleotide diversity ranged from 0.008
(Murrell et al. 2012). In addition to MEME, we also used to 0.072 across all species. We identified several clear cases of
single likelihood ancestor counting (SLAC), fixed effects like- trans-species polymorphism (TSP) (Fig. 2). Because of low
lihood (FEL), random effects likelihood (REL) (Pond and support for some branches in our phylogeny, we only consid-
Frost 2005a), and fast unbiased Bayesian approximation ered identical alleles as evidence of TSP, although a more robust
(FUBAR) (Murrell et al. 2013) methods to test for individual phylogeny might yield additional evidence of TSP between
sites under selection. These four additional methods, like similar alleles. We amplified two alleles shared between
MEME, detect diversifying and purifying selection, but they A. mexicanum and A. velasci (Amme/Amve_DAB*24 and
are not designed to detect episodic selection. To ensure that we Amme/Amve_DAB*30) and one allele shared between
took into account the important effects of episodic selection, A. andersoni and A. velasci (Aman/Amve_DAB*10). We also
we chose MEME as our primary method of analysis of selec- found one A. mexicanum allele (Amme/Amti_DAB*27) shared
tion on individual codons. However, we also included SLAC, with a US A. tigrinum allele from the literature (GenBank ac-
FEL, REL, and FUBAR for comparison. Third, we performed cession no. DQ071905). In contrast, A. dumerilii and
a test of branch-site random effects likelihood (BranchREL) A. altamirani shared no alleles with any other species.
(Pond et al. 2011), which we used to determine whether par- We calculated population genetic summary statistics for the
ticular MHC allelic lineages were undergoing episodic diver- DAB locus (Table 2) and the two neutral markers (Table 3).
sifying selection. This last test employs likelihood ratio tests to For the DAB locus, the observed heterozygosity was much
identify all lineages where a given fraction of sites are posi- lower than expected heterozygosity for all five species.
tively selected. All software that we used for tests of positive Observed heterozygosity was lowest in A. altamirani at 0.08
selection is hosted on the Datamonkey server, www. and highest (though still less than half of expected heterozy-
datamonkey.org (Pond et al. 2005; Pond and Frost 2005b; gosity) in A. dumerilii at 0.42. Generally, A. altamirani was
Delport et al. 2010). the species with the lowest measures of diversity, both in
number of alleles and level of divergence between alleles.
For example, allelic richness was lowest in A. altamirani at
2.857 (and highest in A. dumerilii at 11.000). Likewise, the
Results average number of nucleotide differences between alleles
within a species ranged from 1.855 in A. altamirani to
Genetic diversity and population genetics 16.089 in A. velasci. Of the other four species, which were
generally similar in these measures of DAB diversity,
We isolated a total of 31 unique DAB alleles in the five focal A. mexicanum was slightly less genetically variable than
Ambystoma species (GenBank accession numbers in A. velasci, A. andersoni, and A. dumerilii. When we compared
Table S1, genotypes in Table S2), plus two unique alleles from allelic richness, nucleotide diversity per site, average number
the five MVZ A. tigrinum samples sequenced for comparison of nucleotide differences, and number of segregating sites
(Table S1). The DAB alleles were 222 bp in length and
contained 67 variable sites. None of the sequences had inser- Table 2 Population genetic summary statistics for Ambystoma MHC
tions or deletions. Although we did not perform independent DAB sequences
amplifications to account for misincorporation, detecting the N n Ra HO HE π k S
same allele in multiple individuals greatly increases the like-
lihood that it is a real allele, rather than an artifactual point A. altamirani 19 3 2.857 0.00* 0.28 0.008 1.855 8
mutant. Only four of the 31 novel alleles we detected were A. andersoni 13 9 8.834 0.08* 0.86 0.062 13.820 43
both found in only a single individual and different from an- A. dumerilii 12 11 11.000 0.42* 0.88 0.057 12.634 34
other allele by a single nucleotide (Table S3). Furthermore, the A. mexicanum 18 6 5.352 0.06* 0.57 0.040 8.933 35
apparent existence of only one MHC class II β chain gene in A. velasci 13 5 4.994 0.38* 0.73 0.072 16.089 39
Ambystoma (Laurens et al. 2001; Bos and DeWoody 2005;
Bulut et al. 2008) minimizes some of the difficulties with N number of individuals sequenced for locus, n number of alleles, Ra
allelic richness, HO and HE observed and expected heterozygosities, π
amplifying multigene families. The segment of G1D6 se- nucleotide diversity per site, k average no. of nucleotide differences, S
quenced was 281 bp long and had 5 variable sites. The ampli- number of segregating sites
fied portion of the cytb locus included 696 bp with 21 variable *P<0.01, probability of departure from Hardy-Weinberg equilibrium ex-
sites. pectations is significant
Immunogenetics (2015) 67:323–335 329

Fig. 2 Bayesian tree of TrCr_DAB

genealogical relationships among Amti_DAB*37
Amme_DAB*28
DAB alleles from six Ambystoma * Amme_DAB*41
species. Branches are named in * Aman_DAB*09
the format: Species_Locus*Allele Amve_DAB*31
Amti_DAB*35
#, where species is abbreviated as Amti_DAB*36
the first two letters of the species Amal_DAB*01
and genus names. The outgroup * Amal_DAB*02
Amal_DAB*03
branch (not to scale) is a single Aman_DAB*04
* Aman/Amve_DAB*10
MHC class II β sequence from
the northern crested newt, * Aman_DAB*05
Aman_DAB*06
T. cristatus (GenBank accession Aman_DAB*08
no. FJ448027). A different- Amti_DAB*40
colored dot is used to indicate Aman_DAB*11
* Aman_DAB*12
each species (see key in figure). Amve_DAB*29
Branch support values are given Amdu_DAB*23
Amme/Amve_DAB*24
on each branch, and B*^ is used Amme_DAB*44
*
where support is greater than 0.9. Amme_DAB*42
Instances of trans-species poly- Amme_DAB*25
Amme/Amve_DAB*30
morphism (alleles amplified from
more than one species) are given * Amti_DAB*33
Amti_DAB*38
with both species names and dot Amti_DAB*34
Amme_DAB*43
colors. Lineages found to be un- Amdu_DAB*18
der positive selection by Branch- Amdu_DAB*15
Amme/Amti_DAB*27
REL are indicated in bold. The
Amti_DAB*32 = A. tigrinum
scale bar represents the number Aman_DAB*07 = A. mexicanum
of substitutions per site. Alleles Amdu_DAB*21
32–38 and 41–44 were sequenced Amdu_DAB*19 = A. andersoni
by previous investigators and
Amdu_DAB*16 = A. velasci
Amdu_DAB*22
were retrieved from GenBank Amti_DAB*39 = A. altamirani
(accession numbers in Table S1) Amdu_DAB*13
Amme_DAB*26
= A. dumerilii
*
Amdu_DAB*20
Amdu_DAB*14
Amdu_DAB*17

among the species, we found that DAB was more diverse
than G1D6 and cytb in all five focal species (Table 3,
Fig. 3).
Pairwise FST values for DAB ranged from 0.0981 between
A. dumerilii and A. andersoni to 0.5638 between
A. mexicanum and A. altamirani (Table 4). All pairwise con-
trasts were statistically significant. In contrast, FST values var-
ied more for G1D6 with some very high values (>0.9) and
others very low (∼0.1). Notably, the pairwise FST values were
a mixture of both similar and contrasting patterns of pairwise
genetic differentiation between species. For example,
A. velasci and A. altamirani showed no significant FST at the
neutral marker G1D6 but had one of the highest pairwise FST
values at DAB. Pairwise comparisons of A. altamirani and
A. andersoni, on the other hand, revealed similarly high levels
of FST at both loci. Finally, the pairwise FST of G1D6 between
A. dumerilii and A. andersoni was high while the FST value for
DAB was much lower. These patterns are expected when parts

Table 3 Population genetic

summary statistics for G1D6 and N n Ra HO HE π k S
cytb in five Ambystoma species
(presented as G1D6/cytb) A. altamirani 17/12 1/2 1.000/2.000 0.00/NA 0.00/NA 0.000/0.001 0.000/0.530 0/1
A. andersoni 18/10 1/3 1.000/2.842 0.00/NA 0.00/NA 0.000/0.001 0.000/0.800 0/4
A. dumerilii 13/10 4/3 4.000/2.918 0.38/NA 0.40/NA 0.003/0.001 0.902/0.556 4/2
A. mexicanum 16/15 2/3 1.970/2.658 0.00/NA 0.12/NA 0.000/0.001 0.121/0.381 1/2
A. velasci 14/7 2/1 1.997/1.000 0.14/NA 0.14/NA 0.001/0.000 0.275/0.000 2/0

NA not applicable, N number of individuals sampled, n number of alleles, Ra allelic richness, HO and HE observed
and expected heterozygosities, π nucleotide diversity per site, k average no. of nucleotide differences, S number
of segregating sites
*P<0.01, probability of departure from Hardy-Weinberg equilibrium expectations is significant
330 Immunogenetics (2015) 67:323–335

DAB G1D6

0.01

Cytb
Tcri_DAB

0.01
0.01

Fig. 3 Similarity clustering of DAB, G1D6, and cytb. Dots represent (T. cristatus) is not to scale. Dots with multiple colors represent trans-
unique alleles, not separate individuals, and dot color corresponds to species polymorphism. A separate scale bar is provided for each cluster.
species identity following the key in Fig. 2. The outgroup branch The scale bar represents the number of substitutions per site

of the genome evolve at different rates (Holsinger and Weir
2009).
MHC supertypes and potential functional diversity
We found 17 supertypes among the 44 unique Ambystoma
alleles (Fig. S1), suggesting than many nucleotide substitu-
tions between alleles may correspond to functional differences
in binding affinities (Doytchinova and Flower 2005). Because
some branch support values were low, we did not make de-
tailed comparisons of functional diversity between
Ambystoma species. However, we observed that
A. andersoni and A. dumerilii, the two species with the highest
allelic richness at DAB, tended to have multiple alleles per
supertype, while A. altamirani had only three alleles, but each
allele clustered into a separate supertype. Therefore, even
Ambystoma species like A. altamirani that have low total
MHC polymorphism may have relatively high functional
MHC polymorphism.
Recombination and phylogenetic reconstruction
No recombination was detected at the neutral loci cytb or
G1D6, as determined by the three tests in PERMUTE.
Before phylogenetic inference, we first ran tests of recombi-
nation on the DAB alleles. Of the three tests of recombination
used, only one test (r2) showed significance (P=0.003), while
the other two tests were not significant (D′: P=0.77, G4: P=
0.81). These results, combined with the short length of our
fragment, and simulation results that show that D′ outperforms
r2 for fine mapping (Devlin and Risch 1995), indicated that
recombination would not likely influence our analyses. Thus,
we proceeded with phylogenetic analyses and downstream
tree-based analyses of selection. The best-fit model, as chosen
by hierarchical likelihood ratio tests (hLRT), was F81 + I + G.
Our phylogenetic reconstruction showed a complex rela-
tionship among 44 unique DAB alleles from the six
Ambystoma species (Fig. 2). The topology shows no clear
grouping by species, with the exception of a few small groups

Table 4 Pairwise estimates of

MHC (upper diagonal) and G1D6 A.altamirani A.andersoni A.dumerilii A.mexicanum A.velasci
(lower diagonal) FST among
Ambystoma species A. altamirani – 0.435 0.436 0.564 0.506
A. andersoni 0.826 – 0.098 0.269 0.175
A. dumerilii 1 0.821 – 0.265 0.177
A. mexicanum 0.052 0.733 0.938 – 0.227
A. velasci 0.008 0.747 0.939 −0.015 –

Values in italics are significant (P<0.005)

Immunogenetics (2015) 67:323–335
of very highly similar alleles. For example, all three
A. altamirani alleles formed a highly supported clade. Five
of the A. andersoni alleles had high sequence similarity and
clustered together with high support. One of these alleles
(Aman/Amve_DAB*10) was identical to an allele recovered
from A. velasci. Four other more divergent A. andersoni DAB
alleles were scattered throughout the tree, and two of these
were most closely related to two different A. velasci alleles.
A. dumerilii alleles were distributed throughout the tree.
Although no A. andersoni alleles were shared with other spe-
cies, one (Aman_DAB*07) was nested within a cluster of four
closely related A. dumerilii alleles. Alleles from A. mexicanum
and A. tigrinum were also scattered throughout the tree, with
no species clustering.
Positive selection
We did not detect a significant signal of selection acting on
cytb (P=1.000), suggesting it was evolving under predomi-
nately neutral and/or purifying forces. G1D6 was not part of
an open reading frame and thus was not tested for selection
using codon-based methods.
Our first assessment of positive selection in PARRIS de-
tected a significant overall signal of positive selection in the
DAB alignment (P<0.001). We then used multiple methods to
infer which amino acid residues had experienced positive se-
lection. We focus here on the results from MEME (Fig. 4)
because this method detects both diversifying and episodic
selection and it is the method recommended by the
Datamonkey server, which hosts all of the selection test soft-
ware we used (Delport et al. 2010). The sites found to be under
selection by SLAC, FEL, REL, and FUBAR analyses were
generally congruent with those found by MEME (Table S4).
Sites where selection was detected by MEME but none of the
other four methods were presumed to be under episodic selec-
tion (Table S4). We used putative peptide-binding residues of
β1 previously described for humans (Brown et al. 1993) and
applied to X. laevis (Kobari et al. 1995) as putative peptide-
binding residues in the Ambystoma β1. We then compared
these residues to the sites MEME determined to be under
selection (Fig. 4). Half of the 16 putative peptide-binding res-
idues were positively selected (3, 5, 20, 22, 49, 62, 63, and
66). Seven residues underwent no selection (1, 29, 30, 52, 53,
59, 70), and FEL, FUBAR, REL, and SLAC detected that
only one residue underwent negative selection (74). Out
of the 58 remaining sites that are not likely to bind
foreign peptides, only two showed evidence of positive
selection (18 and 23).
We then used BranchREL to detect significant evidence of
diversifying episodic selection in the DAB phylogeny (Fig. 2).
Historical positive selection was detected on the branches
leading to Amdu_DAB*26 and Amdu_DAB*17 and on the
ancestor of Aman_DAB*09 and Amve_DAB*31 DAB alleles.
331
Discussion
Early studies reported a single fixed allele at the DAB locus in
A. mexicanum (Laurens et al. 2001) compared to much higher
DAB allelic diversity in wild populations of tiger salamanders
(A. tigrinum) (Bos and DeWoody 2005), suggesting that axo-
lotls have lower potential for adaptive immune responses.
However, these reports of low DAB allelic diversity in
A. mexicanum were likely biased by small sample sizes and
sampling from single populations (or potentially inbred labo-
ratory populations). Richman et al. (2007) first documented
higher allelic diversity in A. mexicanum by recovering five
unique alleles (two of which we also amplified in our study)
in nine wild A. mexicanum individuals, which motivated us to
undertake this study of additional Ambystoma populations.
Our study greatly increases estimates of MHC class II allelic
richness and selection in ambystomatid salamanders. By char-
acterizing the MHC allelic diversity in wild populations of
A. mexicanum and four additional, previously uncharacterized
Ambystoma species, we revealed 31 novel Ambystoma DAB
alleles. Although further studies are needed to establish the
long-term effect of MHC polymorphism on these populations,
we hypothesize that the relatively high MHC allelic diversity
in the three endangered paedomorphic species (A. andersoni,
A. dumerilii, and A. mexicanum) may be beneficial to the
long-term viability of these species.
The positive selection we detected provides further evi-
dence that wild Ambystoma populations have a functional
and variable MHC class II β locus. We detected positive se-
lection at the levels of individual codons, particular lineages,
and the DAB alignment overall. We did not detect positive
selection in our neutral markers, making it unlikely that
genome-wide selective process is causing global signals of
positive selection or false signals of selection on the DAB
locus. DAB is also much more diverse than the neutral loci
we sequenced. Although allelic diversity cannot be quantita-
tively compared between the three loci (due to differences in
mutation rates), a qualitative comparison alone shows that
DAB has far higher allelic diversity than the neutral loci
(Fig. 3). This result is consistent with balancing selection,
which favors high allelic diversity.
Additionally, because of recent speciation within the com-
plex, we were able to easily detect trans-species polymor-
phism (TSP), an important phenomenon in MHC evolution
that is most common in closely related species (Klein et al.
2007). In some instances, TSP provides evidence of balancing
selection; long-term selective pressure drives high MHC poly-
morphism in an ancestral population and allows advantageous
allelic lineages to persist across speciation events (Klein et al.
2007). Because we also found shared alleles for G1D6 and
cytb (Fig. 3), it is difficult to distinguish the effects of TSP due
to selection from signals of incomplete lineage sorting, herein
defined as when the phylogeny of a gene differs from the
332 Immunogenetics (2015) 67:323–335

+ + + + +++ ++ + + +
++
P P P P P PP P PP P PP P
PP P P
. |
....|....| ....|....|| ....|....|
. . ....|....| ....|....|
....|....| ....|....| .. . ....
10 2
20 30 40 50 60 70
Amal_DAB*01 M Y
TQMKYECLYL V RWSYNQQQFL
NGSERVRYVV W SA ESWNSQKEVL EQRRAEVDTF
HFDSDTGVFK ADDLLGVPSA QR
RR EV CRHN
Amal_DAB*02 . G
....G..... . ........V.
.......... . T .......... ..K..A....
.......... ........T. .K
K A ....
Amal_DAB*03 . G
....G..... . ........V.
.......... . T .......... .RE.......
.......... ........T. RE
E . ....
Aman_DAB*04 L F
..L.F..RI. . .......P.V
..T....... . D .Y........ ..K......L
..Y....... ....F...D. .K
K . ..Y.
Aman_DAB*05 L F
..L.F..HI. . .......P.V
..T....... . D .Y........ ..K......L
.......... ....F...D. .K
K . ..Y.
Aman_DAB*06 L F
..L.F..HI. . .......P.V
..T....... . D .Y........ ..K......L
.......... ....F...D. .K
K . ..Y.
Aman_DAB*07 R S
..R.SG.HF. E .Y.......V
.........E Y T ........I. .........Y
........YE ..TPF...T. .. . ..F.
Aman_DAB*08 L F
..L.F..HI. . .......P.V
..T....... . D .Y.......P ..K......L
.......... ....F...D. .K
K . ..Y.
Aman_DAB*09 ..G.S..HF.
G S ..T....F.D
D .Y.......V
Y .........E ..TPF...D.
D .Y........ ..A..A....
.A
A A ..Y.
Aman_DAB*10 ..L.F..RI.
L F ..T.......
. .......P.V
. .......... ....F...D.
D .Y........ ..K......L
.K
K . ..Y.
Aman_DAB*11 ....G..HF.
. G ..T....F.L
L ........V.
. .......... ........T.
T ........I. .........Y
.. . ..Y.
Aman_DAB*12 W S
..W.S..HF. A ..........
..T......A . . ..L....... ..E..K...Y
........YE .......... .E
E K ..Y.
Amdu_DAB*13 . G
....G..HF. L .........V
..T....F.L . T ........I. .........Y
........YE ..TPF...T. .. . ..F.
Amdu_DAB*14 ....G..HF.
. G ..T......E
E .Y.......V
Y ........YE ..TPF...D.
D .......... ..E..A...Y
.E
E A ..Y.
Amdu_DAB*15 ....G..HF.
. G ..T......E
E .Y.......V
Y .........E ........T.
T ........I. .....A...Y
.. A ..Y.
Amdu_DAB*16 R S
..R.S..HF. L .Y.......V
.........L Y T .......... ..E..A...Y
........YE ..TPF...T. .E
E A ..Y.
Amdu_DAB*17 V H
..V.H..HF. L .Y.......V
..T......L Y D .......... .....A...Y
........YE ..TPF...D. .. A ..Y.
Amdu_DAB*18 V H
..V.H..HF. E .Y......V.
..T......E Y . ..L....... ..K......Y
........YE .......... .K
K . ..Y.
Amdu_DAB*19 R S
..R.S..HF. E .Y.......V
.........E Y D ........I. .....A...Y
........YE ..TPF...D. .. A ..Y.
Amdu_DAB*20 W S
..W.S..HF. A ......RP.V
..T....F.A . D .......... ..E..A...Y
........YE ..TPF...D. .E
E A ..Y.
Amdu_DAB*21 R S
..R.S..HF. E .Y.......V
.........E Y T ........I. .........Y
........YE ..TPF...T. .. . ..F.
Amdu_DAB*22 W S
..W.S..HF. E .Y.......V
.........E Y T .......... ..E..A...Y
........YE ..TPF...T. .E
E A ..Y.
Amdu_DAB*23 W S
..W.S..HF. E .Y......V.
..T......E Y . ..L....... ..E..A....
.......... .......... .E
E A ....
Amme_DAB*24 W S
..W.S..HF. A ..........
..T......A . . ..L....... ..E..A....
.......... .......... .E
E A ....
Amme_DAB*25 . G
....G..HF. . ........V.
..T....F.. . . ..L....... ..E..A....
.......... .......... .E
E A ....
Amme_DAB*26 . G
....G..HF. L .........V
..T....F.L . T .......... ..E..A...Y
........YE ..TPF...T. .E
E A ..Y.
Amme_DAB*27 V H
.EV.H..HF. E .Y.......V
..T....F.E Y D .......... .....A...Y
.......... ........D. .. A ..Y.
Amme_DAB*28 . G
....G..HF. . .Y....L..V
..T....... Y T .Y........ ..K......V
........YE ..TPF...T. .K
K . ....
Amve_DAB*29 W S
..W.S..HF. A ........V.
..T......A . . ..L....... ..E..K...Y
........YE .......... .E
E K ..Y.
Amve_DAB*30 . G
....G..HF. . ........V.
..T....F.. . . ..L.....I. ..E.....R.
.......... .......... .E
E . ....
Amve_DAB*31 G S
..G.S..HF. D .Y.......V
..T....L.D Y D .Y......F. .....A....
.........E ..TPF...D. .. A ..F.
Amti_DAB*32 . G
....G..HF. E .Y.......V
..T....F.E Y D .......... .....A...Y
.......... ........D. .. A ..Y.
Amti_DAB*33 . G
....G..HF. . ........V.
..T....... . . ........I. ........R.
.........Q .......... .. . ..F.
Amti_DAB*34 ..L.C..HF.
L C ..T....F.E
E .Y....R.I.
Y .....A...E ...M....T.
T K.L....... ..........
.. . ....
Amti_DAB*35 V H
..V.HK.HF. E SY...E..VV
.........E Y D .YL....... .DV..A....
.........Q ..SPF...D. DV
V A ..Y.
Amti_DAB*36 V H
..V.H..HF. . .F.....P.V
.......... F D KY........ ..A.......
.........Q ..TPF...D. .A
A . ..F.
Amti_DAB*37 V H
.EV.H..HI. E .Y.......V
..T....F.E Y T .Y......F. ..K.......
.........Q ..SPF...T. .K
K . ..F.
Amti_DAB*38 . G
....G..HF. . ........VV
..T....... . . ........I. ........R.
.........Q .......... .. . ..F.
Amti_DAB*39 R S
.ER.S..HF. E .Y.......V
..T....F.E Y T ........I. ..Q..K...Y
.........Q ..TPF...T. .Q
Q K ..F.
Amti_DAB*40 V G
..V.G..HF. A .......P.V
..T....L.A . . ........I. ..A..Q....
.........E .......... .A
A Q ..F.
Amme_DAB*41 . G
....G..HF. . .Y....L..V
..T....... Y T .Y......F. ..E..A...V
........YE ..TPF...T. .E
E A ....
Amme_DAB*42 W S
..W.S..HF. . ..........
..T....... . . ..L....... ..E..A....
.......... .......... .E
E A ...X
Amme_DAB*43 ....G..HF.
. G ..T....F..
. .C......V.
C .......... ........T.
T .......... ..A..A...V
.A
A A ..F.
Amme_DAB*44 W S
..W.S..HF. A ..........
..T......A . N ..L....... ..E..A....
.......... ........N. .E
E A ....
Tcri_DAB - -
---------- . ..V..R..VV
----...FL. .V . KQ...D.AF. ..A..D..R.
.....V.F.E ..MEA.E... .A
A D ....
Xela_DAB G A
F.G.AQ.YFR A .YIN..EEYA
..TDN..FLA Y Q DY......I. D.H......L
Y....V.S.I .KNDF.RVQ. .H
H . ..Y.

Fig. 4 Alignment of a subset of representative DAB amino acid from Kobari et al. (1995) and noted by a P. The sites found under positive
translations from five Ambystoma species, Triturus cristatus and selection by MEME are labeled with a + symbol. Putative peptide-
Xenopus laevis. Amino acid identity is represented by dots (.) and binding residues that are also under positive selection are highlighted in
amino acids predicted to be involved with peptide binding are taken gray

phylogeny of the species because groups of alleles in diverg-
ing species have not yet reached reciprocal monophyly.
However, in the context of our other findings, the potential
signal of TSP lends support to our balancing selection
hypothesis.
Our results reveal additional complexity when we interpret
population genetic statistics for the DAB sequences (Table 2)
in the broader context of life history, range, population hetero-
zygosity, and conservation status of the individual Ambystoma
species (Table 1). Because they are obligately paedomorphic
endangered species existing in single populations restricted to
single lakes, we expected A. andersoni, A. dumerilii, and
A. mexicanum to have lower genetic diversity at both neutral
and functional genes. A. altamirani and A. velasci, in contrast,
are transforming species (presumably with greater potential
for between-population admixture); thus, we expected them
to have higher genetic diversity at both DAB and neutral loci.
We also expected species with lower heterozygosity (based on
microsatellite data) to have lower genetic diversity at all loci.
However, our predictions regarding ecological traits and
MHC genetic diversity were not supported. For example,
A. andersoni, despite the ecology described above and the
low level of population heterozygosity (Table 1), has the
highest number of segregating sites of any of the five species
and the second highest allelic richness. At the other extreme,
A. altamirani, a transforming species with intermediate popu-
lation heterozygosity, has dramatically lower values for all
measures of MHC genetic diversity than the other four spe-
cies. Our analysis of MHC supertypes may partially explain
this discordance because the species with higher overall
Immunogenetics (2015) 67:323–335
numbers of alleles tend to have more Bfunctionally similar^
alleles than species with lower overall numbers of alleles.
We noted that all five Ambystoma species in our study have
significantly lower observed than expected heterozygosities
for MHC, but not for G1D6 or most microsatellite markers
(Parra-Olea et al. 2012). This pattern could reflect amplifica-
tion bias, selective amplification, and null alleles at the DAB
locus, which would suggest even higher variability than we
detected in our focal species. However, several factors suggest
that null alleles are neither present in large numbers nor
strongly influencing our results. Most importantly, the mating
system of Ambystoma is expected to result in overall low
levels of heterozygosity at MHC: A. tigrinum, a close relative
of our focal species, mates assortatively relative to MHC (Bos
et al. 2009). If this mating preference also occurs in other
Ambystoma species, we would expect the signatures of non-
random mating at the DAB locus to manifest as departure from
Hardy-Weinberg equilibrium, in contrast to other unlinked
neutral markers. Furthermore, the positive selection we detect-
ed at the DAB locus violates the expectations of Hardy-
Weinberg equilibrium used to determine expected heterozy-
gosity, and selection can impact heterozygosity at selected loci
differently than neutral (non-selected) loci. Although uncom-
mon, heterozygote deficiency at MHC loci has been docu-
mented in some species, including the African buffalo and
the mouse lemur (Wenink et al. 1998; Schad et al. 2004).
Thus, further research on the mating systems and pathogens
of these species may shed light on this unusual pattern of
MHC genotype frequencies.
We hypothesize that DAB polymorphism in Mexican
Ambystoma has historically been maintained over evolution-
ary time, likely in response to environmental pressures, such
as pathogen exposure, that favor high polymorphism. We de-
tected positive selection at the levels of individual codons,
particular lineages, and the DAB alignment as a whole. The
31 unique alleles we isolated across species provide a large
sample size for tests of positive selection that were previously
unavailable. We found that approximately half of the putative
peptide-binding residues have undergone positive selection,
supporting the interpretation that the β1 domain in
Ambystoma species evolved to better recognize and present
antigens from pathogens. In addition, pairwise FST compari-
sons support the inference that a process other than demogra-
phy is shaping MHC evolution in some populations. Under a
neutral model, pairwise FST (a measure of genetic differentia-
tion) is expected to be similar for most loci in the genome.
Significantly lower genetic differentiation of MHC compared
to neutral loci may indicate balancing selection on MHC while
higher MHC differentiation may be a signature of selection
favoring different MHC alleles in different populations.
Our results suggest that both balancing and differing
directional selection may be occurring in Ambystoma
populations.
333
Complicating the picture, we also found much lower MHC
heterozygosity than expected and no clear relationship be-
tween allelic diversity and life history, range, or mean popu-
lation heterozygosity. We hypothesize that current selection
and mating preferences are driving these trends, but more
information is needed to elucidate current, as opposed to his-
torical, selective forces in wild Ambystoma populations. Our
study therefore provides valuable insight into historical selec-
tion in Ambystoma and the complexities of immunogenetic
studies in wild populations. We also contribute raw material
for future studies on Ambystoma salamanders and other uro-
deles. As we try to understand the potential impact of amphib-
ian declines around the world, a thorough understanding of
amphibian host immune gene evolution is critical.
Acknowledgments We thank D. Weisrock for providing A. dumerilii
tissue samples for this study, A. E. Savage and D. Rodriguez for the
assistance with molecular protocols, R. C. Bell for the assistance with
phylogenetic analyses, A. Ellison for supertyping of MHC, and M. Yuan
for the assistance with figures. Members of the Zamudio lab group pro-
vided invaluable advice and feedback throughout this study. We also
thank our two anonymous reviewers for their feedback, which greatly
improved the final version of the manuscript. Funding for this study
was provided by National Science Foundation Grants (DEB-0815315
and DEB-1120249 to K. R. Zamudio) and an award from the Dextra
Undergraduate Research Endowment Fund (to K. E. Tracy).
Conflict of interest The authors declare that they have no conflict of
interest.
Ethical approval The tissues used for this study were collected by
Gabriela Parra-Olea (Parra-Olea et al. 2012). Tissues were collected under
the permit FAUT-0106, issued by Secretaria del Medio Ambiente y
Recursos Naturales and protocol #1999-0010 issued by the Cornell
University Institutional Animal Care and Use Committee.
References
Anisimova M, Nielsen R, Yang Z (2003) Effect of recombination on the
accuracy of the likelihood method for detecting positive selection at
amino acid sites. Genetics 164:1229–1236
Babik W, Pabijan M, Radwan J (2008) Contrasting patterns of variation
in MHC loci in the Alpine newt. Mol Ecol 17:2339–2355. doi:10.
1111/j.1365-294X.2008.03757.x
Babik W, Pabijan M, Arntzen JW et al (2009) Long-term survival of a
urodele amphibian despite depleted major histocompatibility com-
plex variation. Mol Ecol 18:769–781. doi:10.1111/j.1365-294X.
2008.04057.x
Baratti M, Dessi-Fulgheri F, Ambrosini R et al (2012) MHC genotype
predicts mate choice in the ring-necked pheasant Phasianus
colchicus. J Evol Biol 25:1531–1542. doi:10.1111/j.1420-9101.
2012.02534.x
Barribeau SM, Villinger J, Waldman B (2008) Major histocompatibility
complex based resistance to a common bacterial pathogen of am-
phibians. PLoS One 3:e2692. doi:10.1371/journal.pone.0002692
Borghans JAM, Beltman JB, De Boer RJ (2004) MHC polymorphism
under host-pathogen coevolution. Immunogenetics 55:732–739.
doi:10.1007/s00251-003-0630-5
334
Bos D, DeWoody J (2005) Molecular characterization of major histocom-
patibility complex class II alleles in wild tiger salamanders
(Ambystoma tigrinum). Immunogenetics 57:775–781. doi:10.1007/
s00251-005-0038-5
Bos DH, Williams RN, Gopurenko D et al (2009) Condition-dependent
mate choice and a reproductive disadvantage for MHC-divergent
male tiger salamanders. Mol Ecol 18:3307–3315. doi:10.1111/j.
1365-294X.2009.04242.x
Brown J, Jardetzky T, Gorga J et al (1993) 3-Dimensional structure of the
human class II histocompatibility antigen HLA-DR1. Nature 364:
33–39. doi:10.1038/364033a0
Bulut Z, McCormick C, Bos D, DeWoody J (2008) Polymorphism of
alternative splicing of major histocompatibility complex transcripts
in wild tiger salamanders. J Mol Evol 67:68–75. doi:10.1007/
s00239-008-9125-1
Chen G, Robert J (2011) Antiviral immunity in amphibians. Viruses 3:
2065–2086. doi:10.3390/v3112065
Cheng Y, Sanderson C, Jones M, Belov K (2012) Low MHC class II
diversity in the Tasmanian devil (Sarcophilus harrisii).
Immunogenetics 64:525–533. doi:10.1007/s00251-012-0614-4
Cohen N (1971) Amphibian transplantation reactions: a review. Am Zool
11:193–205. doi:10.1093/icb/11.2.193
Daszak P, Cunningham AA, Hyatt AD (2003) Infectious disease and
amphibian population declines. Divers Distrib 9:141–150. doi:10.
1046/j.1472-4642.2003.00016.x
Delport W, Poon AFY, Frost SDW, Pond SLK (2010) Datamonkey 2010:
a suite of phylogenetic analysis tools for evolutionary biology.
Bioinformatics 26:2455–2457. doi:10.1093/bioinformatics/btq429
Devlin B, Risch N (1995) A comparison of linkage disequilibrium mea-
sures for fine-scale mapping. Genomics 29:311–322
Doytchinova IA, Flower DR (2005) In silico identification of supertypes
for class II MHCs. J Immunol 174:7085–7095
Ellison A, Allainguillaume J, Girdwood S et al (2012) Maintaining func-
tional major histocompatibility complex diversity under inbreeding:
the case of a selfing vertebrate. Proc R Soc B-Biol Sci 279:5004–
5013. doi:10.1098/rspb.2012.1929
Fisher MC, Garner TWJ, Walker SF (2009) Global emergence of
Batrachochytrium dendrobatidis and amphibian chytridiomycosis
in space, time, and host. Annu Rev Microbiol 63:291–310. doi:10.
1146/annurev.micro.091208.073435
Frank SA (2002) Genetic variability of hosts. http://www.ncbi.nlm.nih.
gov/books/NBK2401/. Accessed 3 May 2014
Frost, DR (2014) Amphibian species of the world: an online reference.
Version 6.0. American
Goudet J (1995) FSTAT (Version 1.2): a computer program to calculate F-
statistics. J Hered 86:485–486
Gray M, Miller D, Hoverman J (2009) Ecology and pathology of am-
phibian ranaviruses. Dis Aquat Organ 87:243–266. doi:10.3354/
dao02138
Hall TA (1999) BioEdit: a user-friendly biological sequence alignment
editor and analysis program for Windows 95/98/NT. Nucleic Acids
Symp Ser 41:95–98
Hammer O, Harper DAT, Ryan PD (2001) PAST: paleontological statis-
tics software package for education and data analysis. Palaeontol
Electron 4:Unpaginated
Hedrick PW, Poulin R (2002) Pathogen resistance and genetic variation at
MHC loci. Evolution 56:1902–1908. doi:10.1554/0014-3820(2002)
056[1902:PRAGVA]2.0.CO;2
Hill WG, Robertson A (1968) Linkage disequilibrium in finite popula-
tions. Theor Appl Genet 38:226–231. doi:10.1007/BF01245622
Holsinger KE, Weir BS (2009) Genetics in geographically structured
populations: defining, estimating and interpreting F-ST. Nat Rev
Genet 10:639–650. doi:10.1038/nrg2611
Hughes AL, Yeager M (1998) Natural selection at major histocompati-
bility complex loci of vertebrates. Annu Rev Genet 32:415–435.
doi:10.1146/annurev.genet.32.1.415
Immunogenetics (2015) 67:323–335
Huson DH, Bryant D (2006) Application of phylogenetic networks in
evolutionary studies. Mol Biol Evol 23:254–267. doi:10.1093/
molbev/msj030
Irschick DJ, Shaffer HB (1997) The polytypic species revisited: morpho-
logical differentiation among tiger salamanders (Ambystoma
tigrinum) (Amphibia: Caudata). Herpetologica 53:30–49
IUCN (2014) The IUCN red list of threatened species. Version 2014.2.
http://www.iucnredlist.org. Accessed 21 Oct 2014
Kalinowski ST (2006) HW-QUICKCHECK: an easy-to-use computer
program for checking genotypes for agreement with Hardy–
Weinberg expectations. Mol Ecol Notes 6:974–979. doi:10.1111/j.
1471-8286.2006.01456.x
Kieswetter CM, Schneider CJ (2013) Phylogeography in the northern
Andes: complex history and cryptic diversity in a cloud forest frog,
Pristimantis w-nigrum (Craugastoridae). Mol Phylogenet Evol 69:
417–429. doi:10.1016/j.ympev.2013.08.007
Klein J (1986) Natural history of the major histocompatibility complex.
New York
Klein J, Sato A, Nikolaidis N (2007) MHC, TSP, and the origin of spe-
cies: from immunogenetics to evolutionary genetics. In: Campbell A
(ed) Annual review of genetics book series. California Annual
Reviews, Palo Alto, pp 281–304
Kobari F, Sato K, Shum B et al (1995) Exon-intron organization of
Xenopus MHC class II β chain genes. Immunogenetics 42:376–385
Kurtz J, Kalbe M, Aeschlimann PB et al (2004) Major histocompatibility
complex diversity influences parasite resistance and innate immuni-
ty in sticklebacks. Proc R Soc B-Biol Sci 271:197–204. doi:10.
1098/rspb.2003.2567
Laurens V, Chapusot C, del Rosario OM et al (2001) Axolotl MHC class
II β chain: predominance of one allele and alternative splicing of the
β1 domain. Eur J Immunol 31:506–515. doi:10.1002/1521-
4141(200102)31:2<506::AID-IMMU506>3.0.CO;2-P
Lewontin RC (1964) The interaction of selection and linkage. I. General
considerations; heterotic models. Genetics 49:49
Li L, Wang BB, Ge YF, Wan QH (2014) Major histocompatibility com-
plex class II polymorphisms in forest musk deer (Moschus
berezovskii) and their probable association with purulent disease.
Int J Immunogenet 41:401–412. doi:10.1111/iji.12135
Librado P, Rozas J (2009) DnaSP v5: a software for comprehensive
analysis of DNA polymorphism data. Bioinformatics 25:1451–
1452. doi:10.1093/bioinformatics/btp187
Martel A, Spitzen-van der Sluijs A, Blooi M et al (2013)
Batrachochytrium salamandrivorans sp nov causes lethal
chytridiomycosis in amphibians. Proc Natl Acad Sci U S A 110:
15325–15329. doi:10.1073/pnas.1307356110
Mattoccia M, Marta S, Romano A, Sbordoni V (2011) Phylogeography
of an Italian endemic salamander (genus Salamandrina): glacial
refugia, postglacial expansions, and secondary contact. Biol J Linn
Soc 104:903–922. doi:10.1111/j.1095-8312.2011.01747.x
McVean G, Awadalla P, Fearnhead P (2002) A coalescent-based method
for detecting and estimating recombination from gene sequences.
Genetics 160:1231–1241
Moritz C, Schneider C, Wake D (1992) Evolutionary relationships within
the Ensatina eschscholtzii complex confirm the ring species inter-
pretation. Syst Biol 41:273–291. doi:10.2307/2992567
Murrell B, Wertheim JO, Moola S et al (2012) Detecting individual sites
subject to episodic diversifying selection. Plos Genet. doi:10.1371/
journal.pgen.1002764
Murrell B, Moola S, Mabona A et al (2013) FUBAR: a fast, uncon-
strained Bayesian approximation for inferring selection. Mol Biol
Evol. doi:10.1093/molbev/mst030
Nadachowska-Brzyska K, Zieliński P, Radwan J, Babik W (2012)
Interspecific hybridization increases MHC class II diversity in two
sister species of newts. Mol Ecol 21:887–906. doi:10.1111/j.1365-
294X.2011.05347.x
Immunogenetics (2015) 67:323–335
O’Neill EM, Schwartz R, Bullock CT et al (2013) Parallel tagged
amplicon sequencing reveals major lineages and phylogenetic struc-
ture in the North American tiger salamander (Ambystoma tigrinum)
species complex. Mol Ecol 22:111–129. doi:10.1111/mec.12049
Parham P, Janeway CA (2005) The immune system. Garland Science,
New York
Parra-Olea G, Zamudio KR, Recuero E et al (2012) Conservation genet-
ics of threatened Mexican axolotls (Ambystoma). Anim Conserv 15:
61–72. doi:10.1111/j.1469-1795.2011.00488.x
Pond SLK, Frost SDW (2005a) Not so different after all: a comparison of
methods for detecting amino acid sites under selection. Mol Biol
Evol 22:1208–1222. doi:10.1093/molbev/msi105
Pond SLK, Frost SDW (2005b) Datamonkey: rapid detection of selective
pressure on individual sites of codon alignments. Bioinformatics 21:
2531–2533. doi:10.1093/bioinformatics/bti320
Pond SLK, Frost SDW, Muse SV (2005) HyPhy: hypothesis testing using
phylogenies. Bioinformatics 21:676–679. doi:10.1093/
bioinformatics/bti079
Pond SLK, Murrell B, Fourment M et al (2011) A random effects branch-
site model for detecting episodic diversifying selection. Mol Biol
Evol 28:3033–3043. doi:10.1093/molbev/msr125
Raymond M, Rousset F (1995) GENEPOP (Version 1.2): population
genetics software for exact tests and ecumenicism. J Hered 86:
248–249
Richman AD, Herrera G, Reynoso VH et al (2007) Evidence for
balancing selection at the DAB locus in the axolotl, Ambystoma
mexicanum. Int J Immunogenet 34:475–478. doi:10.1111/j.1744-
313X.2007.00721.x
Robert J, Cohen N (2011) The genus Xenopus as a multispecies model for
evolutionary and comparative immunobiology of the 21st century.
Dev Comp Immunol 35:916–923. doi:10.1016/j.dci.2011.01.014
Ronquist F, Teslenko M, van der Mark P et al (2012) MrBayes 3.2:
efficient Bayesian phylogenetic inference and model choice across
a large model space. Syst Biol 61:539–542. doi:10.1093/sysbio/
sys029
Rousset F (2008) GENEPOP’007: a complete re-implementation of the
GENEPOP software for Windows and Linux. Mol Ecol Resour 8:
103–106. doi:10.1111/j.1471-8286.2007.01931.x
Sammut B, Du Pasquier L, Ducoroy P et al (1999) Axolotl MHC archi-
tecture and polymorphism. Eur J Immunol 29:2897–2907. doi:10.
1002/(SICI)1521-4141(199909)29:09<2897::AID-IMMU2897>3.
0.CO;2-2
Sandberg M, Eriksson L, Jonsson J et al (1998) New chemical descriptors
relevant for the design of biologically active peptides. A multivariate
characterization of 87 amino acids. J Med Chem 41:2481–2491. doi:
10.1021/jm9700575
Savage AE, Zamudio KR (2011) MHC genotypes associate with resis-
tance to a frog-killing fungus. Proc Natl Acad Sci 108:16705–
16710. doi:10.1073/pnas.1106893108
Schad J, Sommer S, Ganzhorn JU (2004) MHC variability of a small
lemur in the littoral forest fragments of southeastern Madagascar.
335
Conserv Genet 5:299–309. doi:10.1023/B:COGE.0000031137.
50239.d3
Scheffler K, Martin DP, Seoighe C (2006) Robust inference of positive
selection from recombining coding sequences. Bioinformatics 22:
2493–2499. doi:10.1093/bioinformatics/btl427
Schierup MH, Mikkelsen AM, Hein J (2001) Recombination, balancing
selection and phylogenies in MHC and self-incompatibility genes.
Genetics 159:1833–1844
Shaffer HB (1993) Phylogenetics of model organisms: the laboratory
axolotl, Ambystoma mexicanum. Syst Biol 42:508–522. doi:10.
1093/sysbio/42.4.508
Siddle HV, Kreiss A, Eldridge MDB et al (2007) Transmission of a fatal
clonal tumor by biting occurs due to depleted MHC diversity in a
threatened carnivorous marsupial. Proc Natl Acad Sci U S A 104:
16221–16226. doi:10.1073/pnas.0704580104
Skerratt LF, Berger L, Speare R et al (2007) Spread of chytridiomycosis
has caused the rapid global decline and extinction of frogs.
EcoHealth 4:125–134. doi:10.1007/s10393-007-0093-5
Swofford D (2003) PAUP*. Phylogenetic analysis using parsimony (*and
other methods). Version 4. Sinauer, Sunderland, Massachusetts
Teacher AGF, Garner TWJ, Nichols RA (2009) Evidence for directional
selection at a novel major histocompatibility class I marker in wild
common frogs (Rana temporaria) exposed to a viral pathogen
(Ranavirus). PLoS One 4:e4616. doi:10.1371/journal.pone.
0004616
Tournefier A, Laurens V, Chapusot C et al (1998) Structure of MHC class
I and class II cDNAs and possible immunodeficiency linked to class
II expression in the Mexican axolotl. Immunol Rev 166:259–277
Voss SR, Epperlein HH, Tanaka EM (2009) Ambystoma mexicanum, the
axolotl: a versatile amphibian model for regeneration, development,
and evolution studies. Cold Spring Harb Protoc 2009:pdb.emo128.
doi: 10.1101/pdb.emo128
Wegner KM, Kalbe M, Kurtz J et al (2003) Parasite selection for immu-
nogenetic optimality. Science 301:1343. doi:10.1126/science.
1088293
Weisrock DW, Shaffer HB, Storz BL et al (2006) Multiple nuclear gene
sequences identify phylogenetic species boundaries in the rapidly
radiating clade of Mexican ambystomatid salamanders. Mol Ecol
15:2489–2503. doi:10.1111/j.1365-294X.2006.02961.x
Wenink PW, Groen AF, Roelke-Parker ME, Prins HHT (1998) African
buffalo maintain high genetic diversity in the major histocompati-
bility complex in spite of historically known population bottlenecks.
Mol Ecol 7:1315–1322. doi:10.1046/j.1365-294x.1998.00463.x
Wilson DJ, McVean G (2006) Estimating diversifying selection and func-
tional constraint in the presence of recombination. Genetics 172:
1411–1425. doi:10.1534/genetics.105.044917
Zhu R, Chen Z, Wang J et al (2014) Extensive diversification of MHC in
Chinese giant salamanders Andrias davidianus (Anda-MHC) re-
veals novel splice variants. Dev Comp Immunol 42:311–322. doi:
10.1016/j.dci.2013.10.001