Microsatellite Evolution at Two Hypervariable Loci Revealed by Extensive
Avian Pedigrees
Nadeena R. Beck, Michael C. Double, and Andrew Cockburn
Evolutionary Ecology Group, School of Botany and Zoology, Australian National University, Canberra, Australia
Genealogies generated through a long-term study of superb fairy-wrens (Malurus cyaneus) were used to investigate
mutation within two hypervariable microsatellite loci. Of 3,230 meioses examined at the tetranucleotide locus (Mcyl8),
45 mutations were identified, giving a mutation rate of 1.4%. At the dinucleotide locus (Mcyl4) 30 mutations were
recorded from 2,750 meioses giving a mutation rate of 1.1%. Mutations at both loci primarily (80%; 60/75) involved the
loss or gain of a single repeat unit. Unlike previous studies, there was no significant bias toward additions over deletions.
The mutation rate at Mcyl8 increased with allele size, and very long alleles (.70 repeats) mutated at a rate of almost
20%. The length of the mutating allele and allele span, however, were strongly correlated so it was not possible to isolate
the causative factor. Allele size did not appear to affect mutation rate at Mcyl4, but the repeat number was considerably
lower at this locus. The gender of the mutating parent was significant only at Mcyl8, where mutations occurred more
frequently in maternal alleles. However, at both loci we found that alleles inherited from the mother were on average
larger than those from the father, and this in part drove the higher mutation rate among maternally inherited alleles at
Mcyl8.
Introduction
The most widely accepted mechanism of microsatellite mutation is the slipped-strand mispairing model,
in which the misaligned reassociation of replicating DNA
strands after DNA polymerase slippage results in the
insertion or deletion of one or more repeat units (Levinson
and Gutman 1987; reviewed by Eisen 1999). Microsatellite mutations are estimated to occur between 1023
and 1025 times per locus per generation (Crozier et al.
1999; Hancock 2000), a rate several orders of magnitude
higher than at other loci. However, previous studies on
microsatellite evolution have revealed a large variation in
mutation rates, not only between loci, but also between
alleles at the same locus (Primmer et al. 1996; Wierdl,
Dominska, and Petes 1997; Chakraborty et al. 1997;
Primmer and Ellegren 1998; Primmer et al. 1998; Neff and
Gross 2001).
Mutation rates have been found to be influenced
primarily by the sex of the individual, the nature of the
repeat motif, and the length of the allele. Mutation rates
have been reported to be up to five times as high in
paternally transmitted alleles, probably because of the
greater number of cell divisions involved in spermatogenesis relative to oogenesis (Ellegren and Fridolfsson 1997;
Primmer et al. 1998; Ellegren 2000). The effect of repeat
motif on mutation rate is unclear. Weber and Wong (1993)
found that the mutation rate at tetranucleotide loci in
humans was four times higher than at dinucleotide loci,
a conclusion that has some support (Hastbacka et al. 1992;
Zahn and Kwiatkowski 1995). Subsequent studies,
however, suggest that mutation rate is inversely related
to motif size (Chakraborty et al. 1997; Anderson et al.
2000). Although larger repeat units might be less prone to
slippage, there is some evidence that mismatch repair
mechanisms are less able to recognize misalignments
involving four or more base pairs (Eisen 1999). Finally,
many studies in different organisms have found a mutaKey words: microsatellites, Malurus cyaneus, mutation.
E-mail: [email protected].
Mol. Biol. Evol. 20(1):54–61. 2003
DOI: 10.1093/molbev/msg005
Ó 2003 by the Society for Molecular Biology and Evolution. ISSN: 0737-4038
54
tional bias in favor of long alleles over short alleles within
the same locus (Primmer et al. 1996; Wierdl, Dominska,
and Petes 1997; Primmer et al. 1998; Anderson et al.
2000). In an analysis of 592 AC microsatellite loci from
five vertebrate classes, Neff and Gross (2001) also found
that mutation rate, inferred from microsatellite variability,
increased with average allele length. Amos et al. (1996)
suggested that heterozygotes with a large size difference
between alleles (large allele span) have an elevated rate of
mutation rate. However, neither Primmer et al. (1998) nor
Ellegren (2000) found any evidence to support this
conclusion, suggesting instead that a large allele span is
likely to involve a long allele which is more prone to
replication errors due to slippage.
Previous studies have found that most mutations
involve single repeat changes consistent with a step-wise
model, and that there is significant bias toward additions
over deletions (Amos et al. 1996; Primmer et al. 1996,
1998; Neff and Gross 2001). Longer alleles are, however,
more likely to display deletions than shorter alleles, and
these deletions are likely to involve repeat numbers larger
than two; perhaps providing a ‘‘length ceiling’’ preventing
infinite growth (Weirdl et al. 1997; Primmer et al. 1998;
Ellegren 2000). Primmer et al. (1998) also identified
mutational differences between male and female barn
swallows (Hirundo rustica), finding that additions are
more common in males, but the magnitude of size
alterations are greater in females.
Clearly, the mechanisms of microsatellite mutations
are still not fully understood and need to be further
investigated to devise more accurate models of mutation
for use in population and paternity analysis (Primmer et al.
1998). This study takes advantage of extensive pedigrees
resulting from a long-term study of superb fairy-wrens
(Malurus cyaneus) to investigate the mutation characteristics of two hypervariable microsatellite loci. Hypervariable loci such as these represent only a very small
proportion of the total microsatellite loci present in the
genome and are therefore likely to be subject to different
mutational processes, resulting in a higher mutation rate
(Webster, Smith, and Ellegren 2002). However, it is
Avian Microsatellite Evolution 55
Table 1
Characterization of the Two Microsatellite Loci Used in This Analysis of Mutation Events
Locus
Repeat Motifa
No. of
Individuals
Allele Size
Range (bp)
No. of
Alleles
Observed
Heterozygosity
Expected
Heterozygosity
Mcyl4
Mcyl8
(GT)26AT(GT)3
(AAAG).50
392
392
140–188
169–401
22
57
0.92
0.98
0.92
0.96
NOTE.—Statistics are based on the genotypes of 392 randomly selected individuals.
Double et al. 1997b.
a
important to have an understanding of the processes
affecting the evolution of these unusual loci, as highly
polymorphic microsatellites are preferentially selected for
use in population studies and genome mapping.
Methods
Study Species and Population
This study was based on the population of superb
fairy-wrens (Malurus cyaneus) resident in the Australian
National Botanic Gardens (ANBG) and immediate surrounds. Superb fairy-wrens are a cooperatively breeding
species in which males display extreme natal philopatry
(Mulder 1995). In contrast, females are obligate dispersers.
Whereas 43% of females disperse within 14 weeks of
fledging, the remainder are forced off their natal territory at
the onset of the following breeding season when the
dominant female no longer tolerates them (Mulder 1995).
Females disperse an average of 11.8 territory widths and
are not known to travel or settle with close relatives.
Dispersing females that obtain a breeding vacancy breed in
the first year, producing up to two clutches of three to four
eggs. Although socially monogamous, this species has the
highest known rate of extra-pair fertilization of any bird,
with 76% of offspring sired by males outside the social
group (Mulder et al. 1994).
The ANBG population of superb fairy-wrens has
been the subject of an extensive behavioral and genetic
study for over 12 years, resulting in a demographic data set
including social and reproductive information for over
4,000 individuals (Langmore and Mulder 1992; Mulder
et al. 1994; Mulder 1995; Dunn and Cockburn 1999;
Double and Cockburn 2000). The study area encompasses
approximately 75 territories but is embedded in a large,
panmictic population that extends from the study area in
all directions. All individuals within the study area are
color banded, and year-round censuses have determined
group histories and social relationships. To assign
paternity to offspring within this population, more than
2,500 individuals have been genotyped at five or more
microsatellite loci (Double et al. 1997b). The extreme natal
philopatry observed in males and the breeding philopatry
of females of this species have allowed the construction of
large pedigrees for use in the verification of genotypes and
the identification of microsatellite mutation events.
Microsatellite Genotyping
Throughout the 12-year study, blood was collected
from all nestlings on territories within the study area and
from adults dispersing into the ANBG. DNA was isolated
from blood samples by ammonium acetate extraction
(Richardson et al. 2001) after digestion with proteinase K
(Progen). Microsatellite loci were amplified in polymerase
chain reaction (PCR) mixes consisting of approximately
50 lg of template DNA, 1.2 ll of Opti-prime 103 PCR
buffer, 1.2 ll of 2 mM dNTPs, 1.2 ll of 2 lM forward and
fluorescently labeled reverse primers, 1.0 ll of 25 mM
MgCl, 0.1 ll AmpliTaq DNA polymerase (PerkinElmer)
and ddH2O to a total reaction volume of 12 ll. PCR was
performed on an FTS-960 Thermal Sequencer (Corbett
Research) using the following profile: initial template
DNA denaturation at 948C for 3 min, followed by 35
cycles of denaturation at 948C for 30 s, 558C for 30 s,
polymerization at 728C for 30 s or 45 s, and a final extension at 728C for 3 min.
PCR products were visualized on a 5.3% polyacrylamide gel using an ABI Prism 377 automated
sequencer (PerkinElmer). All samples were run with
TAMRA 500 internal size standard (PerkinElmer), and
allele sizes were determined using GeneScan 3.1 and
Genotyper 2.0 software (PerkinElmer/ABI).
Genetic fathers were identified by exclusion (Double
and Cockburn 2000). Every potential father within the
study area has been sampled and genotyped. Up to seven
microsatellite loci were used for paternity assignment in
the study population; six dinucleotide repeat loci and one
tetranucleotide repeat locus (GenBank accession numbers
U82385–U82382; Double et al. 1997b). With allele
frequencies from the five most commonly used loci, the
probability of false assignment to a male unrelated to the
true sire was 9.8 3 1025. Using a simulation-based
approach (Double et al. 1997a), we estimated the
probability of false assignment to a single first-order
relative of the true sire was 0.08. This probability fell to
0.04 if all seven loci were used. Because all males within
the study area are genotyped, a false assignment could
occur only if the true sire was outside our study area and
had never been sampled or if the true sire produced
a mutant allele and another male matched all the
offspring’s paternal alleles.
Mutation Detection
The present study sought mutations at two of the
microsatellite loci used to assign paternity; the dinucleotide Mcyl4 and the tetranucleotide Mcyl8 (table 1).
Sequences for these two microsatellite loci are available
from GenBank (accession numbers U82388 and U82392).
The two loci were chosen because of the clarity of their
GeneScan profiles. Mutations were detected by comparing
the genotype of an offspring with that of both its parents.
56 Beck et al.
Only individuals for whom both maternal and paternal
genotypes were available were included in the analysis.
PCR conditions for Mcyl4 were changed during the study
and improved the clarity of the genotyping profiles.
Previous ambiguous profiles were discarded, and therefore
the number of meioses examined at Mcyl8 was greater
than at Mcyl4 (3230 versus 2750).
The reliable identification of mutations depends on
being confident of correctly matching offspring to parents
for genotype comparison. Observations of hundreds of
laying sequences over the last 12 breeding seasons have
revealed no evidence for intraspecific brood parasitism
(Mulder et al. 1994). Therefore the maternal alleles of the
offspring could always be confidently assigned.
Only those individuals whose paternity had been
assigned based on matching genotypes at five or more loci
were included in the analysis of mutation events. If
a putative mutation was identified with five loci, then the
remaining two loci were examined to further lower the
probability of incorrect paternity assignment. Individuals
with ambiguous genotypes or in whom the putative father
was mismatched at more than one locus were excluded
from the analysis. Where possible, individual genotypes
and mutations were confirmed by tracking the inheritance
of each allele through genealogies. For example, if an
offspring inherited a mutated allele from its mother, we
first checked the genotypes of other offspring produced by
the mother and, if possible, the mother’s parents to confirm
the size of the progenitor allele. In an attempt to verify the
existence of a mutant allele, we also checked if the
offspring had successfully reproduced and if so whether
the mutated allele had been inherited.
Initially we used three different protocols to identify
the progenitor allele: (1) assume that the smallest
mutational change in allele size was most likely (Primmer
et al. 1998; Crozier et al. 1999; Ellegren 2000); (2) assume
that the largest mutational change allele size was most
likely; and (3) random assignment of the progenitor allele.
The distribution of mutational changes produced by these
three protocols was then compared to the distributions
expected under single-step and multi-phase models of
microsatellite evolution.
Statistical analyses were performed with the software
package JMP (SAS Institute). We used logistic regression
to analyze a discrete response (e.g., mutation yes/no) to
a continuously distributed explanatory variable (e.g., allele
length). Terms were assessed by the change in deviance
which approximates a v2 distribution. We used two-tailed
tests and rejected the null hypothesis when P , 0.05.
Results
Mutation Rate
At the dinucleotide Mcyl4 locus, 1,375 individuals
were examined, representing 2,750 meioses. Of these, 30
mutations were recorded, yielding a mutation rate of 1.1%.
These meioses were derived from 236 females and 258
males. The maximum number of offspring produced by
any one individual was 51, with the average being 5.3 6
5.9 SD for males and 5.8 6 5.0 SD for females. The
mutation rate at the tetranucleotide Mcyl8 locus was
FIG. 1.—Mutational changes in repeat number presented as
a percentage of the total number of mutation events detected (Mcyl4:
N 5 30; Mcyl8: N 5 45). We identified the change in size between
progenitor and mutant alleles by assuming (1) that the largest size change
was most likely, (2) that the smallest size change was most likely, or (3)
by random assignment of the progenitor allele.
found to be 1.4% (45 mutations from 1,615 individuals:
3,230 meioses). At this locus, the meioses were derived
from 254 females and 288 males. Again, the maximum
number of offspring produced by any one individual was
51, with the average being 5.6 6 6.1 SD for males and
6.4 6 5.4 SD for females. In 65 of the 75 mutation events,
the size of the suspected progenitor allele could be
confirmed by examination of the genotypes from the
progenitor’s own parents and/or non-mutant offspring. In
eight cases the size of the mutant allele could be verified
because the mutant allele was passed on to the following
generation. In all other cases the individual with the
mutant allele failed to breed successfully within our study
area or exclusively transmitted the nonmutant allele.
Magnitude and Directionality of Mutations
Initially, we identified progenitor using three different
assumptions (fig. 1). Under the first assumption, that the
smallest change in allele size was most likely, the average
size of mutations was 1.7 6 1.4 SD repeat units at Mcyl4,
and 1.5 6 2.6 SD repeat units at Mcyl8. The majority of
mutant alleles (Mcyl4: 20/30; Mcyl8: 40/45) differed
from the parental allele by an increase or decrease of just
Avian Microsatellite Evolution 57
four and six repeats, whereas at Mcyl8, a mutation added
four repeats and another removed 18 repeats.
Allele Length
At Mcyl4, allele length did not influence mutation
rate (logistic regression: v2 5 0.25, df 5 1, P 5 0.6; fig.
2a). In contrast, the mutation rate at Mcyl8 was greatly
influenced by allele length, with the longest alleles
mutating at a rate of almost 20% of meioses (logistic
, 0.001; fig. 2b). At
regression: v2 5 31.4, df 5 1, P ,
both loci, allele length did not influence the frequency of
deletions (logistic regression: Mcyl4: v2 5 1.4, df 5 1,
P 5 0.2; Mcyl8: v2 5 0.04, df 5 1, P 5 0.8).
Interpretation of the effects of allele length may be
biased if the more common alleles often occur in heterozygotes with adjacent alleles. Mutations are masked if an
allele mutates to the other allele carried by the individual. As
longer alleles are generally rare, heterozygotes are less likely
to have adjacent alleles. Therefore, mutations might be
recognized more easily in rare, longer alleles. At Mcyl8 we
found that longer alleles were less likely to occur in
a heterozygote with an allele of adjacent length (logistic
regression: v2 5 8.9, df 5 1, P . 0.005). However, even the
most common alleles had only a 20% chance of occurring
with an adjacent allele. Given a mutation rate of 1.4%, fewer
than 1 in 1,000 mutations is likely to be masked in this way.
Therefore, we conclude that the effect of allele size on
mutation rate at this locus is real.
FIG. 2.—Allele frequency distribution (shaded) and mutation
frequency distribution (black) at Mcyl4 and Mcyl8. At Mcyl8 alleles
were pooled into groups of two repeat units. The allele frequency
represents the proportion of meioses observed in a particular allele class.
The mutation rate was calculated from the number of mutations in an
allele size class, divided by the total number of meioses within that class.
one repeat unit. Assuming the largest change in allele size
was most likely, the average mutation size was 5.2 6 2.7
SD repeat units at Mcyl4 and 20.4 6 13.1 SD repeat units
at Mcyl8. Finally, random assignment of the progenitor
allele resulted in an average mutation size of 3.9 6 3.0 SD
at Mcyl4 and 7.6 6 9.6 SD at Mcyl8. For both loci, only
the assumption of the smallest change in allele size
produced a distribution of mutational changes that is easily
reconciled with the expectation of a predominance of
single-step mutations as predicted by single-step and
multi-phase models of microsatellite evolution (e.g., Di
Rienzo et al. 1994). We followed this assumption in all
subsequent analyses.
The difference between the number of additions and
deletions at Mcyl4 (15 versus 13, respectively; v2 5 0.14,
df 5 1, P 5 0.71) and Mcyl8 (23 versus 21, respectively;
v2 5 0.09, df 5 1, P 5 0.76) was not significant. For two
mutations at Mcyl4 and one at Mcyl8, the mutant allele
lay equidistant between two potential progenitor alleles
and so it was uncertain if an addition or deletion had
occurred. For each locus only two mutations involved
more than two repeat units. At Mcyl4, mutations deleted
Allele Span
Allele span refers to the difference in length between
the alleles of a heterozygote (Amos et al. 1996). At
Mcyl4, parents with large allele spans were no more likely
to pass on mutant alleles than those with small allele spans
(logistic regression: v2 5 0.3, df 5 1, P 5 0.6). Seventeen
mutations originated from the parent with the largest
difference in allele size, whereas 11 occurred in the parent
with the smallest allele span (contingency table analysis:
v2 5 1.3, df 5 1, P 5 0.26). One mutation occurred in an
individual whose parents had identical allele spans, and the
parental origin of another could not be determined because
the parents had identical alleles.
At Mcyl8, parents with a large allele span were more
likely to pass a mutant allele on to their offspring (logistic
regression: v2 5 11.1, df 5 1, P , 0.001), and 68%
(28/41) of mutations occurred in the parent with the largest
allele span (v2 5 5.49, df 5 1, P 5 0.02). However, allele
span and the length of the progenitor allele were highly
correlated (r2 5 0.19, F 5 8.99, df 5 1, P 5 0.005), so it
was not possible to determine whether allele span or allele
length had the greater influence on mutation rate.
Similarly, the effect of allele span and length could not
be dissected by analyzing the frequency data; 17 of the 41
mutations did not involve the largest parental allele, of
which 10 (58%) were inherited from the parent with the
largest allele span (v2 5 1.8, df 5 1, P 5 0.17). At Mcyl8
the parental origin of three mutations could not be
58 Beck et al.
Table 2
Effect of Gender on Allele Sizes at Mcym4 and Mcym8
N
Average
Allele Size (SE)
Mcyl4
Mothers
Fathers
Female offspring
Male offspring
Maternally inherited alleles
Paternally inherited alleles
434
388
920
1,298
1,291
1,291
24.9
25.0
24.8
24.9
25.3
24.5
(0.20)
(0.21)
(0.14)
(0.12)
(0.13)
(0.11)
Mcyl8
Mothers
Fathers
Female offspring
Male offspring
Maternally inherited alleles
Paternally inherited alleles
486
476
1,074
1,536
1,557
1,557
59.2
58.0
59.2
58.7
59.7
57.5
(0.57)
(0.52)
(0.36)
(0.30)
(0.30)
(0.29)
identified, and one mutation occurred in an individual
whose parents had identical allele spans.
Influence of Gender on Mutation
At Mcyl4 more mutations originated from a paternal
allele (66%; 19/29), although the difference was not
significant (v2 5 2.8, df 5 1, P 5 0.09). The parental
origin of one mutation could not be identified at this locus.
By contrast, mutant alleles at Mcyl8 were more likely to
be maternally derived (74%; 31/42; v2 5 9.5, df 5 1, P 5
0.002). This result could be driven by a tendency for
females to pass on larger alleles (see below). However, in
a logistic regression model which initially included, and
therefore controlled for, allele length, we still found that
the mutation rate was significantly influenced by gender of
the parent donating the allele (v2 5 7.9, df 5 1, P 5 0.02).
The parental origin of three mutations at Mcyl8 could not
be identified because the parents had identical alleles at
this locus.
Gender bias in mutation rates could be explained by
a difference in the average allele length. We found no
difference in allele length at either Mcyl4 or Mcyl8 either
between mothers and fathers or between male and female
offspring (table 2). At both loci, however, alleles inherited
from the mother were on average longer that those from
the father (table 2). Although the effect is slight (about half
of one repeat at each locus), it is highly significant. The
longest allele was maternally derived in 53.5% (691/1291)
of offspring at Mcyl4 and 52.9% (824/1557) of offspring
at Mcyl8. This result is most likely due to differential
reproductive success of males and females within the study
area. When only a single offspring was randomly selected
for each mother and father, there was no significant
difference between alleles inherited from males or females
for either locus (Mcyl4: t 5 0.8, df 5 498, P 5 0.4;
Mcyl8: t 5 0.8, df 5 541, P 5 0.4). There was no
evidence to suggest that either sex was preferentially
passing on their larger or smaller alleles at either locus (all
P . 0.1).
At both loci the gender of the parent with the
progenitor allele did not influence whether the mutation
was an addition or a deletion (table 3; Mcyl4: contingency
Statistical
Comparison
F 5 0.002, P 5 0.96
F 5 0.34, P 5 0.56
F 5 22.7, P 0.001
F 5 2.40, P 5 0.12
F 5 0.99, P 5 0.32
F 5 27.0, P 0.001
table analysis: v2 5 0.02, df 5 1, P 5 0.9; Mcyl8: v2 5
1.1, df 5 1, P 5 0.29). There was also no evidence at
either locus that mutations of more than a single repeat
unit were more commonly inherited from fathers or
mothers (Mcyl4: Fisher Exact test, P 5 0.2; Mcyl8:
Fisher Exact test, P 5 0.1).
Discussion
Correct Paternity Assignment
Identifying mutations based on comparisons with
parental genotypes poses some potential problems, particularly in a species that exhibits high levels of extra-pair
paternity and strong male philopatry. In these circumstances
the possibility of incorrectly assigning a close relative as the
sire becomes more likely (Double et al. 1997a). For
example, it is possible for the true sire to pass on a mutated
allele, causing paternity to be assigned to his brother. This
phenomenon may occur infrequently, but it could result in
either an underestimate of mutation rate, where a mutation
results in an allele of a high frequency in the population, or
an overestimate of mutation rate where the correct sire has
not been sampled. Although errors are possible, the nature of
the mutations themselves argues for their reliable identification. First, we found that almost all mutated alleles
differed by a single repeat from the progenitor allele, an
unlikely result if false paternity assignment were the cause of
a mismatch. Second, the size distribution of mutating alleles
at Mcyl8 is very different from the allele frequency
distribution, and it is consistent with the pattern reported
in other studies (e.g., Primmer et al. 1996, 1998). Third, the
majority of mutations detected in this study occurred in the
maternal line. As there is no evidence of intraspecific brood
parasitism in superb fairy-wrens (Mulder 1994), mutations
occurring in the maternal line can be identified with some
confidence. Finally, offspring with mutant alleles were
evenly distributed throughout our study area. If incorrect
paternity assignments had led to incorrect attribution of
mutation, we would expect the majority of mutations to be
clustered on the periphery of the study area, where potential
fathers are less likely to have been sampled.
Other potential sources of bias in the identification of
mutations are genotyping errors and null alleles. We are
Avian Microsatellite Evolution 59
Table 3
Summary of Mutations at Mcym4 and Mcym8
Mcyl4
Mcyl8
Source of
Mutant Allele
Addition
Deletion
Addition
Deletion
Maternal allele
Paternal allele
5 (2)
9 (8)
5 (3)
8 (7)
14 (14)
7 (5)
17 (15)
4 (3)
NOTE.—At Mcyl4 the mutating parent could not be identified for one mutation, and two mutations could not be classified as an addition or deletion. At
Mcyl8 the mutating parent could not be identified for three mutations. Numbers
in parentheses indicate mutations involving a single repeat unit.
able to exclude genotyping errors as a potential source of
bias, because all offspring genotypes are cross-checked
with the maternal genotype, ruling out loading errors. In
addition, where paternity is ambiguous, individuals are
regenotyped for clarification.
The presence of null alleles may artificially inflate
mutation estimates, because an individual possessing a null
allele will be scored as a homozygote. However, the
proportion of mutant individuals that were homozygous
was not significantly different from the proportion of
homozygotes in the general population (v2 5 1.26, df 5 1,
P . 0.05). Also there was no heterozygote deficiency
(table 1), which would be expected if null alleles were
frequent.
The Influence of Motif on Mutation Rates
Mutation rates were similar at both Mcyl4 and
Mcyl8 (1.1% and 1.4%, respectively). Weber and Wong
(1993) found that tetranucleotide loci had higher mutation
rates than dinucleotide loci, whereas more recent studies
have found that mutation rate is inversely proportional
to motif size (Chakraborty et al. 1997; Anderson et al.
2000). Although it makes intuitive sense that a longer
motif would be more stable, many microsatellite systems
(including that used in this study) include highly variable
tetranucleotide loci (Li, Huang, and Brown 1997; Primmer
et al. 1998), and such hypervariability itself implies
a higher mutation rate. When selecting microsatellites
for population analyses, however, there is always
a bias toward more variable, and hence more informative,
loci.
maternal mutations, the significant difference in average
length of alleles inherited from each sex contributes to an
elevated mutation rate in females, particularly at this locus,
where allele length has such a significant effect on
mutation rate.
Directionality and Magnitude of Mutations
Our study found no significant difference between the
number of additions and deletions at either locus. This
result is contrary to the findings of Amos et al. (1996) and
Primmer et al. (1996, 1998), who found marked directionality in microsatellite mutations, favoring additions.
This result led Primmer et al. (1998) to hypothesize that
there must be a microsatellite length ceiling preventing
uncontrolled growth. It has been suggested that a mechanism counteracting gradual expansion would involve very
large deletions of 50 to 100 repeat units (Weber and Wong
1993). Primmer et al. (1998) found no evidence for
deletions of such magnitude and suggested that such
events are probably rare, and so would not be detected in
typical samples sizes. However, the practice of assigning
mutations to the allele requiring the smallest number of
repeat changes could, in some cases, mask large changes
in repeat number.
However, our findings are consistent with observations by Crozier et al. (1999), who found no evidence of
directionality in an ant microsatellite. They suggested that
this locus was approaching an equilibrium frequency
distribution of allele sizes. If microsatellites eventually
reach an equilibrium allele frequency distribution where
there is no pronounced directionality in mutations, then
a mechanism preventing infinite microsatellite growth
through large deletions becomes unnecessary. If this is the
case, however, it might be expected that deletions would
be concentrated among the largest alleles at a locus. This
pattern has been found in other studies (e.g., Wierdl,
Dominska, and Petes 1997; Ellegren 2000) but appears not
to be the case in fairy-wrens, where deletions are not more
common in longer alleles.
Most of the mutations in this study involved length
changes of only one repeat unit, consistent with the twophase model of microsatellite mutation (Di Rienzo et al.
1994). Our findings are concordant with previous studies
and support the applicability of the two-phase model to
population and genetic analyses.
The Influence of Sex on Mutation Rates
The effect of the sex of the mutating parent was only
significant at Mcyl8, where almost three times as many
mutations occurred in the maternal line as in the paternal
line. This is in contrast to previous studies where
microsatellite mutation rates have been found to be up to
five times as high in males as in females, a difference that
is generally attributed to the higher number of mitoses
involved in spermatogenesis (Ellegren 2000). There is
some evidence that sex affects different loci differently,
with other studies finding more uniform mutation rates
with respect to sex (e.g., Jeffreys et al. 1988; Talbot et al.
1995). Although the conservative identification of paternal
mismatches as mutations might contribute to an excess of
Allele Length
Investigation of the distribution of mutations at the
two loci reveals significant differences. At Mcyl8, the size
distribution of mutating alleles is extremely skewed
toward higher repeat numbers, with the longest alleles
mutating at rates of almost 20%. However, allele size does
not appear to affect mutation rate at Mcyl8 until repeat
numbers extend beyond 70. Below this threshold,
mutations rates are similar to those at Mcyl4, where
mutating alleles are more evenly distributed with respect to
allele size but allele length does not exceed 40 repeat units.
A similar pattern was found by Primmer et al. (1996), who
60 Beck et al.
found that mutation rates were significantly higher in
alleles greater than 80 repeat units.
The significant difference in allele lengths inherited
from mothers and fathers was unexpected but may reflect
the biology of this species. There is extreme skew in the
reproductive success of superb fairy-wrens. Within our
study area 5% of males sire 50% of offspring, and a single
male is related to over 500 of the 2,500 individuals
genotyped in this study (unpublished data). This skew
appears to have generated the significant difference in
allele lengths inherited from mothers and fathers for the
two loci examined here.
The similarity between the mutation rates at these two
loci may reflect the influence of a combination of length
and motif. We suggest that while the tetranucleotide motif
may be more stable, the mutation rate at Mcyl8 is driven
primarily by allele length. In contrast, at Mcyl4, the
stability of comparatively short alleles counters a less
stable dinucleotide motif. Thus similar mutation rates at
different loci may obscure the operation of different
mutational mechanisms.
Allele Span
Amos et al. (1996) suggested that microsatellite
mutations were more likely to occur in heterozygous
individuals with a larger difference between the sizes of
their two alleles. This effect of allele span differed between
Mcyl4 and Mcyl8. Significantly more mutations at
Mcyl8 occurred in the parent with the largest span;
however, this effect was not seen at Mcyl4. Primmer et al.
(1998) suggested that large allele spans are more likely to
involve large alleles whose length would be the predominant factor affecting mutation rate. In our study allele
span and allele length were strongly correlated, so we
could not decipher which factor had the greater influence
on mutation rate.
Conclusions
The results of this study provide further support for
the influence of allele length (or allele span) on the
mutation rates of microsatellite loci and that stepwise
mutations predominate. However, we did not find a predominance of additions resulting in an increase in allele
size—a phenomenon that has been consistently reported in
other studies. We also found, in contrast to previous
studies, that a disproportionate number of mutations at
Mcyl8 originated from maternal meiotic events. This
seems to be due in part to the surprising finding that in our
study population maternally derived alleles tended to be
longer than those inherited from the father. Although the
mutation rate at these two microsatellite loci appears
similar, the mutational processes involved appear to differ.
At the tetranucleotide locus allele length or span seems to
have the greatest influence whereas for Mcyl4 perhaps the
dinucleotide repeat motif raises the mutation rate across
the entire allele range. It seems that different underlying
mechanisms influence different loci, and further studies are
needed to more fully understand the complexities of
microsatellite evolution.
Acknowledgments
This study was funded by the Australian Research
Council and the Australian National University. We thank
Rod Peakall and two anonymous referees for their
insightful comments and guidance.
Literature Cited
Amos, W., S. J. Sawcer, R. W. Feakes, and D. C. Rubinsztein.
1996. Microsatellites show mutational bias and heterozygote
instability. Nat. Genet. 13:390–391.
Anderson, T. J. C., X. Z. Su, A. Roddam, and K. P. Day. 2000.
Complex mutations in a high proportion of microsatellite loci
from the protozoan parasite Plasmodium falciparum. Mol.
Ecol. 9:1599–1608.
Chakraborty, R., M. Kimmel, D. N. Stivers, L. J. Davison, and
R. Deka. 1997. Relative mutation rates at di-, tri-, and
tetranucleotide microsatellite loci. Proc. Natl. Acad. Sci. USA
94:1041–1046.
Crozier, R. H., M. E. Kaufmann, M. E. Carew, and Y. C. Crozier.
1999. Mutability of microsatellites developed for the ant
Camponotus consobrinus. Mol. Ecol. 8:271–276.
Di Rienzo, A., A. C. Peterson, J. C. Garza, A. M. Valdes, M.
Slatkin, and N. B. Freimer. 1994. Mutational processes of
simple-sequence repeat loci in human populations. Proc. Natl.
Acad. Sci. USA 91:3166–3170.
Double, M. C., A. Cockburn, S. C. Barry, and P. E. Smouse.
1997a. Exclusion probabilities for single-locus paternity
analysis when related males compete for matings. Mol. Ecol.
6:1155–1166.
Double, M. C., and A. Cockburn. 2000. Pre-dawn infidelity:
females control extra-pair mating in superb fairy-wrens. Proc.
R. Soc. Lond. Ser. B. 267:465–470.
Double, M. C., D. Dawson, T. Burke, and A. Cockburn. 1997b.
Finding the fathers in the least faithful bird: a microsatellitebased genotyping system for the superb fairy-wren Malurus
cyaneus. Mol. Ecol. 6:691–693.
Dunn, P. O., and A. Cockburn. 1999. Extrapair mate choice and
honest signaling in cooperatively breeding superb fairy-wrens.
Evolution 53:938–946.
Eisen, J. A. 1999. Mechanistic basis for microsatellite instability.
in D. B. Goldstein and C. Schlötterer, eds. Microsatellites:
evolution and applications. Oxford University Press, Oxford.
Ellegren, H. 2000. Heterogeneous mutation processes in human
microsatellite DNA sequences. Nat. Genet. 24:400–402.
Ellegren, H., and A. K. Fridolfsson. 1997. Male-driven evolution
of DNA sequences in birds. Nat. Genet. 17:182–184.
Hancock, J. M. 2000. Microsatellites and other simple sequences:
genomic context and mutational mechanisms. in D. B.
Goldstein and C. Schlötterer, eds. Microsatellites: evolution
and applications. Oxford University Press, Oxford.
Hastbacka, J., A. de la Chapelle, A. Kaitila, I. Sistonen, A.
Weaver, and E. Lander. 1992. Linkage disequilibrium
mapping in isolated founder populations: dystrophic dysplasia
in Finland. Nat. Genet. 2:204–211.
Jeffreys, A. J., N. J. Royle, V. Wilson, and Z. Wong. 1988.
Spontaneous mutation rates to new length alleles at tandem
repetitive hypervariable loci in human DNA. Nature 332:278–
281.
Langmore, N. E., and R. A. Mulder. 1992. A novel context for
bird song: predator calls prompt male singing in the
kleptogamous superb fairy-wren, Malurus cyaneus. Ethology
90:143–153.
Levinson, G., and G. A. Gutman. 1987. Slipped-strand mispairing: a major mechanism for DNA sequence evolution. Mol.
Biol. Evol. 4:203–221.
Avian Microsatellite Evolution 61
Li, S. H., Y. J. Huang, and J. L. Brown. 1997. Isolation of
tetranucleotide microsatellites from the Mexican jay Aphelocoma ultramarina. Mol. Ecol. 6:499–501.
Mulder, R. A. 1995. Natal and breeding dispersal in a cooperative,
extra-group-mating bird. J. Avian Biol. 26:234–240.
Mulder, R. A., P. O. Dunn, A. Cockburn, K. A. Lazenby-Cohen,
and M. J. Howell. 1994. Helpers liberate female fairy-wrens
from constraints on extra-pair mate choice. Proc. R. Soc.
Lond. Ser. B. 255:223–229.
Neff, B. D., and M. R. Gross. 2001. Microsatellite evolution in
vertebrates: inference from AC dinucleotide repeats. Evolution 55:1717–1733.
Primmer, C. R., H. Ellegren, N. Saino, and A. P. Møller. 1996.
Directional evolution in germline microsatellite mutations.
Nat. Genet. 13:391–393.
Primmer, C. R., and H. Ellegren. 1998. Patterns of molecular
evolution in avian microsatellites. Mol. Biol. Evol. 15:997–
1008.
Primmer, C. R., N. Saino, A. P. Møller, and H. Ellegren. 1998.
Unraveling the processes of microsatellite evolution through
analysis of germ line mutations in barn swallows (Hirunda
rustica). Mol. Biol. Evol. 15:1047–1054.
Richardson, D. S., F. L. Jury, K. Blaakmeer, J. Komdeur, and T.
Burke. 2001. Parentage assignment and extra-group paternity
in a cooperative breeder: the Seychelles warbler (Acrocephalus sechellensis). Mol. Ecol. 10:2263–2273.
Talbot, C. C., D. Avramopoulos, S. Gerken, A. Chakravarti, J. A.
Armour, N. Matsunami, R. White, and S. E. Antonarakis.
1995. The tetranucleotide repeat polymorphism d21s1245
demonstrates hypermutability in germline and somatic cells.
Hum. Mol. Genet. 4:1193–1199.
Weber, J. L., and C. Wong. 1993. Mutation of human short
tandem repeats. Hum. Mol. Genet. 2:1123–1128.
Webster, M. T., N. G. C. Smith, and H. Ellegren. 2002.
Microsatellite evolution inferred from human-chimpanzee
genomic sequence alignments. Proc. Natl. Acad. Sci. USA
99:8748–8753.
Wierdl, M., M. Dominska, and T. D. Petes. 1997. Microsatellite
instability in yeast—dependence on the length of the microsatellite. Genetics 146:769–779.
Zahn, L. M., and D. J. Kwiatkowski. 1995. A 37-marker PCRbased genetic linkage map of human chromosome 9—
observations on mutations and positive interference. Genomics 28:140–146.
Edward Holmes, Associate Editor
Accepted September 6, 2002