REVIEWS

MICROSATELLITES: SIMPLE
SEQUENCES WITH COMPLEX
EVOLUTION
Hans Ellegren
Few genetic markers, if any, have found such widespread use as microsatellites, or simple/short
tandem repeats. Features such as hypervariability and ubiquitous occurrence explain their
usefulness, but these features also pose several questions. For example, why are microsatellites
so abundant, why are they so polymorphic and by what mechanism do they mutate? Most
importantly, what governs the intricate balance between the frequent genesis and expansion of
simple repetitive arrays, and the fact that microsatellite repeats rarely reach appreciable lengths?
In other words, how do microsatellites evolve?

HETEROZYGOSITY Assuming chance association of nucleotides, the proba- multiple alleles, which is in sharp contrast to unique
The proportion of individuals in bility of finding the sequence CACACACACACACA- DNA. With the advent of PCR in the late 1980s, the
a population that carry two CACACA more than once in the human genome is analysis and genotyping of microsatellite polymor-
different alleles at a locus. negligible. However, perfect or near-perfect tandem phisms became straightforward (see TIMELINE). Micro-
GENE FLOW
iterations of short sequence motifs of this kind are satellites quickly became the marker of choice in genome
The transfer of alleles within and extremely common in eukaryotic genomes and, in the mapping, and subsequently also in population genetics
between populations that arises case of the human genome, they are found at hundreds studies and related areas.
from migration and dispersal. of thousands of places along chromosomes1. This par- For a neutral marker, the degree of polymorphism
ticular genomic feature is not restricted to (CA)n repeats is proportional to the underlying rate of mutation.
— every possible motif of mono-, di, tri- and tetranu- Given the extensive polymorphism of microsatellites,
cleotide repeats is vastly overrepresented in the genome. it follows that mutations must occur frequently — an
Ever since their discovery in the early 1980s, the ubiqui- assumption that is supported by direct observations5.
tous occurrence of microsatellites — also referred to as The rate and direction of mutations constitute two
short tandem repeats (STRs) or simple sequence repeats basic factors in the estimation of genetic distance on
(SSRs) — has puzzled geneticists. Why are they so com- the basis of microsatellite data. By applying theoretical
mon? Do they fulfill some function or are they simply models of microsatellite evolution to empirical data,
junk DNA sequences that should perhaps be viewed as population geneticists attempt to, for example, deter-
‘selfish DNA’2,3? Addressing these questions is important mine how long ago two populations diverged, or mea-
if we wish to understand how genomes are organized sure the amount of GENE FLOW between populations.
and why most genomes are filled with sequences other However, despite the extensive use of microsatellite
Department of Evolutionary than genes. markers over the past 15 years, it is clear that many
Biology, Evolutionary Biology Microsatellites are among the most variable types theoretical models fail to accurately explain allele fre-
Centre, Uppsala University, of DNA sequence in the genome4. In contrast to unique quency distributions in natural populations. Importantly,
Norbyvägen 18D, DNA, microsatellite polymorphisms derive mainly from it seems that microsatellite evolution is a far more com-
SE-752 36 Uppsala, Sweden.
e-mail: variability in length rather than in the primary sequence. plex process than was previously thought. A deeper
[email protected] Moreover, genetic variation at many microsatellite loci is understanding of the evolutionary and mutational
doi:10.1038/nrg1348 characterized by high HETEROZYGOSITY and the presence of properties of microsatellites is therefore needed, not

NATURE REVIEWS | GENETICS VOLUME 5 | JUNE 2004 | 4 3 5

REVIEWS

Timeline | The early history of microsatellites

Sequence analysis of Demonstration of extensive

alleles at the globin locus length variability of tandem Microsatellites used to Fine-scale analysis of the
found a varying number of repetitive DNA as revealed by Development of PCR- derive the first detailed genetic relationships among
short sequence DNA fingerprinting of based microsatellite map of the human human populations made
motifs122,123 *. minisatellites 125. genotyping127–129. genome131. possible by microsatellites132.

1981 1982 1985 1986 1989 1991 1992 1993 1994

Identification of a novel repeated Regions of ‘cryptic simplicity’ Microsatellites introduced Large numbers of microsatellite
element — alternating pyrimidine–purine identified as an important for studies of natural mutations identified from
polymers — in eukaryotic genomes, source of genetic variation126. populations 84,130. pedigree analysis in humans 5.
with Z-DNA-forming potential 124.

*The importance of these findings seemed to be insignificant, as laborious cloning and sequencing procedures were required to analyse
the polymorphisms122,123.

only to understand how the genome is organized, but It is more complicated to define the minimum num-
also to correctly interpret and use microsatellite data in ber of iterations needed for a repetitive sequence to be
population genetics studies. referred to as a microsatellite. For instance, the sequence
Recent new information provides clues to the mys- CACA occurs frequently in the human genome: should
tery of microsatellite repeats. First, whole-genome it be seen as (CA)2 microsatellites or just as unique
sequence data provide an unbiased picture of the occur- sequence? In practice, the threshold that is used when
rence and genomic distribution of repetitive elements. describing the occurrence of a microsatellite in a genomic
Second, large-scale pedigree analysis in different organ- sample data set must be specified. Unfortunately, no real
isms gives direct insight into the characteristics of consensus has been reached on this matter; whereas
de novo mutation events. Third, molecular studies of the some use a minimum number of base pairs, others use a
DNA replication machinery show what might go wrong minimum number of repeat units, and in both cases,
during microsatellite replication. Here, I review these the numbers have varied. The issue is further compli-
new findings and summarize our current knowledge cated by the lack of agreement on how much degener-
about microsatellite evolution. Emerging from the new acy should be accepted for characterizing a slightly
data is the picture of a heterogeneous mutation process, imperfect tandem repetitive sequence as a microsatellite.
showing distinct differences in rates and patterns of Mismatch considerations are particularly important
mutation among loci and species. when using algorithms (such as RepeatMasker, Sputnik
and Tandem Repeats Finder; see online links box and
The genome biology of microsatellites BOX 1) to search large genomic sequences for repeats.
What is a microsatellite? Genomes are scattered with It is appropriate to further classify microsatellites
simple repeats. Tandem repeats occur in the form of according to their association with coding sequence as
iterations of repeat units of almost anything from a sin- this is related to the mutational and selective forces that
gle base pair to thousands of base pairs. Mono-, di-, tri- operate on different types of repeat. The bulk of simple
and tetranucleotide repeats are the main types of repeats are embedded in non-coding DNA, either in the
microsatellite, but repeats of five (penta-) or six (hexa-) intergenic sequence or in the introns. Microsatellites
nucleotides are usually classified as microsatellites as that are used as genetic markers are usually of this type
well. Repeats of longer units form minisatellites or, in and are generally assumed to evolve neutrally. Their fre-
the extreme case, satellite DNA. The term satellite DNA quency and distribution should therefore reflect the
originates from the observation in the 1960s of a frac- underlying mutation process. In coding DNA, selection
tion of sheared DNA that showed a distinct buoyant against frameshift mutations effectively hinders the
density, detectable as a ‘satellite peak’ in DENSITY GRADIENT expansion of everything other than trinucleotide
CENTRIFUGATION, and that was subsequently identified repeats6, for which there might be further length con-
as large centromeric tandem repeats. When shorter straints related to protein function7. Trinucleotide
(10–30-bp) tandem repeats were later identified, they repeats associated with human disease comprise a spe-
came to be known as minisatellites. Finally, with the dis- cial class of microsatellites in coding DNA. These loci
covery of tandem iterations of simple sequence motifs, undergo extensive repeat expansions, the mutational
the term microsatellites was coined. The difference mechanism of which is thought to differ from that of
between the terms micro- and minisatellites might not most microsatellites in the genome. For instance, the
DENSITY GRADIENT
CENTRIFUGATION
be obvious per se, but it is motivated by the difference in establishment of hairpin structures with a relatively high
Separation of biomolecules on the mutational mechanisms of repeats of just a few amount of base-pair complementarities might stabilize
the basis of their density. nucleotides and of ten or more (see below). loops that are generated during replication slippage.

436 | JUNE 2004 | VOLUME 5 www.nature.com/reviews/genetics

 REVIEWS

Details of the evolution of expanded trinucleotide negatively correlated with genome size16. This has been
repeats have been described elsewhere and will not be attributed to the fact that microsatellites are underrepre-
considered further here. sented in the repetitive parts of the plant genome that
are involved in genome expansion, such as the long ter-
Microsatellite distribution. The initial analysis of the minal repeats of RETROTRANSPOSONS16. Another peculiar
draft sequence of the human genome concluded that feature of most plant genomes is that (AT)n is the most
microsatellites account for 3% of the genome1. There common motif among dinucleotides17. Assuming that,
are more than one million microsatellite loci in the on a genomic scale, microsatellite sequences are at equi-
human genome, although the exact number greatly librium, the contrasting distributions of microsatellite
depends on the parameters of the search algorithm (for motifs in different genomes strongly indicate that there
example, gap and mismatch penalties). This number is interspecific variation in the mechanisms of mutation
also includes an appreciable proportion of interrupted or repair of specific motifs. Alternatively, there might be
microsatellites and many that are probably monomor- variation in the selective constraints that are associated
phic. Dinucleotide repeats dominate, followed by with different microsatellite motifs.
mono- and tetranucleotide repeats, and trinucleotide Are microsatellites equally common everywhere in
repeats are least dominant. Again, however, it is a matter the genome? There seem to be no distinct differences in
of how microsatellites are defined. Among repeats that density between intergenic regions and introns14. Base
are at least 12 bp long, mononucleotide repeats out- composition influences microsatellite density, which is
number dinucleotide repeats; the reverse situation is not consistent with their neutral origin and random genera-
valid until a higher threshold is used. Among dinu- tion by mutation18. There is, however, evidence for
cleotides, (CA)n repeats are most frequent, followed by regional variation in microsatellite frequency that can-
(AT)n, (GA)n and (GC)n, the last type of repeat being not be explained by base composition18, and, in the
rare. Note that there are only four possible types of din- human and mouse genomes, microsatellite density is
ucleotide repeat, because CA = AC = GT = TG, GA = nearly twofold higher near the ends of chromosome
AG = CT = TC, AT = TA, and GC = CG. arms8. What accounts for this heterogeneity remains to
Data from the mouse genome have confirmed the be explained. In several species, the density and/or the
abundance of microsatellites but have also revealed length distribution of microsatellites on the X chromo-
impressive differences8. If identical search criteria are some differs from that on the autosomes8,18. It might be
used, the mouse genome proves to be repeat-rich with a result of factors such as sex differences in the mutation
two–threefold more microsatellites than humans. rate, differences in EFFECTIVE POPULATION SIZE between the
Moreover, microsatellites are longer in mice than in X chromosome and autosomes, and the efficiency of
humans, and the same holds true for the rat–human selection on hemizygous chromosomes.
comparison9. Preliminary data from other mammalian Microsatellites are also frequently found in the prox-
genomes indicate that rodent genomes have particularly imity of interspersed repetitive elements such as short
high microsatellite numbers. This might be a general interspersed repeats (SINEs) and long interspersed ele-
phenomenon — that microsatellite occurrence differs ments (LINEs). For example, human Alu repeats often
between related species. In fact, differences might even have a microsatellite-like structure at their 3′ ends19 that
occur between such closely related species as humans might arise from the introduction of poly(A) tails of
and chimpanzees10, and within the genus Drosophila11,12. reversed transcribed messages when element insertion
Microsatellite density tends to positively correlate takes place. This is consistent with the observation that
with genome size13–15. Among fully sequenced eukary- mononucleotide arrays (A)n and other types of A-rich
otic genomes, microsatellite density is highest in mam- microsatellite dominate at these sites. Other examples
mals. However, in plants, microsatellite frequency is include the intimate association of microsatellites with

Box 1 | Informatics approaches to finding microsatellites in a genomic data set

Sputnik
Sputnik uses a recursive algorithm to search for repeats of two–five nucleotides in sequence files in FASTA format.
Insertions, mismatches and deletions are tolerated but affect the overall score. If the score falls below a cutoff
threshold, the search is abandoned and begun again at the next nucleotide. Sputnik does not compute an entire
identity matrix first and then pick the best of the hits; instead, it starts at the beginning and compares the patterns
RETROTRANSPOSONS until the score falls below a cutoff threshold.
Mobile elements that spread in RepeatMasker
the genome through an RNA
RepeatMasker does not use a recursive algorithm. It scans for di- to pentameric repeats and simple repeats that are
intermediate.
shorter than 20 bp, and those with >10% divergence from a perfect repeat are ignored.
EFFECTIVE POPULATION SIZE Tandem Repeats Finder
The theoretical size of an
This program works without the need to specify either the pattern or the pattern size. It models tandem repeats
idealized population that has
the same magnitude of random according to the percentage identity and frequency of indels (insertions/deletions) between adjacent pattern copies
genetic drift as the actual and uses statistically-based recognition criteria. The program can return a copy of the original sequence with the
population. tandem repeats masked out.

NATURE REVIEWS | GENETICS VOLUME 5 | JUNE 2004 | 4 3 7

REVIEWS

retrotransposon-like elements in barley20, and (AT)n removal of one repeat unit at a fixed rate (a symmetric
microsatellites that are frequently found to be juxta- forward–backward random walk that is independent
posed with miniature inverted repeat-transposable of repeat length)27–30. However, it soon became appar-
elements (Micropon-4) in rice21. In some of these cases, ent that a simple SMM does not lead to stationary
microsatellites have evolved from internal A-rich struc- microsatellite-length distributions31. For example, the
tures, and it is also possible that insertion of an inter- fact that microsatellites seem to show an upper size
spersed element might in itself be favoured at sites with limit is incompatible with the SMM. Extensions of the
a pre-existing microsatellite22. SMM have therefore introduced an upper limit on
Microsatellites are present in low numbers in pro- allele sizes32–35, or a mutational bias such that large alle-
karyotes. This is particularly true for longer repeats, for les mutate preferentially to alleles of smaller sizes36–38.
which the numbers are lower than would be expected on Other approaches have involved more complex stepwise
the basis of nucleotide composition23, in sharp contrast models39–42. The parameters of the mathematical models
to the situation in eukaryotic genomes. Even short pro- are tested against measures of variability (heterozygosity,
karyotic microsatellites might still vary in length24. variance of repeat counts, SKEWNESS) that are observed
Unusually long microsatellites are sometimes associated within populations, and, more recently, against micro-
with virulence factors, in which case, they act as trans- satellite distributions in genomic data sets.
lation and transcriptional ‘switches’; therefore, their An attractive model of microsatellite evolution holds
presence is maintained by positive selection25. that a genome-wide distribution of microsatellite repeat
length that is at equilibrium results from a balance
The mutation process between length and point mutations43,44. According to
Mutation models. A mutation model of microsatellite this model, two opposing mutational forces operate on
evolution is needed if allele frequency data from two microsatellite sequences. Length mutations, the rate of
groups of individuals (for example, populations or which increases with increasing repeat count, favour
species) are to be used for estimating the genetic dis- loci to attain arbitrarily high values, whereas point
tance between them. A wide range of models of the mutations break long repeat arrays into smaller units.
evolutionary dynamics of microsatellites has been pre- At equilibrium, there will be a steady-state distribution
sented, most of which derive from the stepwise mutation of repeat lengths governed by the rate of length muta-
model (SMM)26 (FIG. 1). Adopted for microsatellites, tion and the rate of point mutation. This model, or
the original SMM postulates that a mutation alters derivatives thereof, has been well received in recent
the length of a repetitive array through the addition or years because it can explain differences in microsatellite

a b c
Frequency

Frequency

Frequency

–3 –2 –1 1 2 3 –3 –2 –1 1 2 3 –3 –2 –1 1 2 3
Repeat length Repeat length Repeat length

d Short alleles Long alleles

Frequency

Frequency

–3 –2 –1 1 2 3 –3 –2 –1 1 2 3
Repeat length Repeat length

Figure 1 | Microsatellite mutation models. The magnitude and direction of mutation events according to different forms of the
stepwise mutation model. Mutations are indicated by the change in the number of repeat units; for example, +1 is an expansion of
SKEWNESS one repeat unit. a | A simple model that only involves one-step changes. b | A model that involves multi-step changes. c | A model
Deviation from the normal with directionality in favour of repeat expansions. d | A length-dependent model in which short alleles tend to increase in size,
distribution. whereas long alleles show a bias towards contraction.

438 | JUNE 2004 | VOLUME 5 www.nature.com/reviews/genetics

 REVIEWS

a b
Increase in repeat length Decrease in repeat length
1 2 3 4 1 2 3 4 400
Initiation 200
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 0
Fragment length
3 4 3 4
1 2 1 2 400
Dissociation
200
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
0
Fragment length
3
1 2 4 Rehybridization 1 2 3 4
and 400
1 2 3 4 5 6 7 8 9 10 misalignment 1 2 4 5 6 7 8 9 10 200
3 0
Fragment length
3
1 2 4 5 6 7 8 9 10 11 The new strand 1 2 3 4 5 6 7 8 9 10
is a different 400
1 2 3 4 5 6 7 8 9 10 length to the 1 2 4 5 6 7 8 9 10 200
template 3 0
Fragment length

Figure 2 | Replication slippage. a | After the replication of a repeat tract has been initiated, the two strands might dissociate. If the
nascent strand then realigns out of register, continued replication will lead to a different length from the template strand. If misalignment
introduced a loop on the nascent strand, the end result would be an increase in repeat length. A loop that is formed in the template
strand leads to a decrease in repeat length. b | Replication slippage also occurs during in vitro amplification of microsatellites, in this case
mainly in the form of repeat contractions. These events can be recognized as minor peaks — known as stutter bands — that differ from
each main product by multiples of the repeat unit length.

distribution among species and provides an elegant as products that are shorter than the size of the allele
solution to the problem of why microsatellites do not being amplified. The in vitro rate of contraction muta-
expand into enormous arrays. However, even with this tion caused by Taq polymerase must therefore be much
model, evolutionary dating of divergence times is not higher than the rate of expansions52. For this reason,
necessarily trivial45. PCR slippage probably cannot be used to gain insight
into the mutational dynamics of microsatellites in vivo.
Mutation mechanism. Length changes in microsatellite Recombination-like processes that involve unequal
DNA are generally thought to arise from replication crossover or GENE CONVERSION introduce mutations in the
slippage — that is, transient dissociation of the replicat- larger minisatellite sequences53. There is little evidence
ing DNA strands followed by misaligned reassociation46 that recombination would also contribute to microsatel-
(FIG. 2a). When the nascent strand realigns out of register, lite mutations. Genomic microsatellite distributions are
renewed replication will lead to the insertion or deletion associated with sites of recombination54, most probably
of repeat units relative to the template strand. Most of as a consequence of repetitive sequences being involved
these primary mutations are corrected by the MISMATCH- in recombination rather than being a consequence of
REPAIR SYSTEM, and only the small fraction that was not it55. Moreover, most tests for a correlation between
repaired ends up as microsatellite mutation events47. recombination rate and microsatellite density or muta-
In vitro experiments that use purified eukaryotic or bility have failed to demonstrate such an effect18,56.
prokaryotic enzymes confirm that DNA polymerase is Furthermore, there is no evidence of any systematic dif-
the only enzymatic activity needed for slippage48. ferences in the rates and patterns of microsatellite muta-
Slippage involves DNA polymerase pausing, during tions between autosomal and Y-linked markers; the
which the polymerase dissociates from the DNA. On fact that Y-chromosome sequences are not involved in
dissociation, only the terminal portion of the newly meiotic recombination therefore does not influence
synthesized strand separates from the template and the mutation process57,58. Neither do the observations
subsequently anneals to another repeat unit49. of similar patterns of microsatellite mutations in
MISMATCH-REPAIR SYSTEM Replication slippage also occurs during PCR amplifi- somatic cells59 and in germ cells support a role for
An enzymatic system for the
correction of errors that are
cation of microsatellite sequences in vitro (FIG. 2b). A recombination.
introduced during DNA characteristic feature of such amplifications is the pres-
replication or recombination ence of ‘stutter bands’ — that is, minor products that Character of observed mutations. Although improved
when an incorrect base is differ in size from the main product by multiples of the mutation models are now available, it can be difficult to
incorporated into the daughter
length of the repeat unit50,51. Quantitative experiments assess to what extent they reflect true evolutionary
strand, or when small
insertion–deletion loops are show that the Taq polymerase slippage rate increases processes. Fortunately, the high rate of mutation at
being formed. with the number of repeat units and is inversely corre- microsatellite loci makes it possible to observe mutation
lated with repeat unit length52. PCR-induced stutter events directly. Specifically, pedigree analysis offers a
GENE CONVERSION bands have been observed by many microsatellite users; means for mutation detection (BOX 2), and data on
A meiotic process of directed
change in which one allele
tetranucleotide repeat markers typically give fewer stut- de novo mutations have now been reported for a range
directs the conversion of a ter bands than dinucleotide and, in particular, of loci and organisms5,60–64. The general pattern that
partner allele to its own form. mononucleotide repeats. Stutter bands generally appear emerges is compatible with replication slippage in which

NATURE REVIEWS | GENETICS VOLUME 5 | JUNE 2004 | 4 3 9

REVIEWS

new variants differ from their progenitor alleles by inte- Heterogeneity is also seen in the propensity of muta-
gral numbers of repeats. The fact that mutations some- tion events to lead to different forms of alterations in
times involve more than one repeat unit means that the microsatellite size. Directionality in the mutation process
single-step mutation model is not valid in most cases. in favour of gains over losses has been observed for many
One important conclusion from observations of human markers39,58,65,66,68,69 and for bird microsatel-
spontaneous mutations is that the mutation process lites71,72. However, Xu et al.70 found no such bias, and
seems to be heterogeneous with respect to loci, repeat Huang et al.56 found only a modest excess of contrac-
types and organisms. For instance, most human studies tions in their studies of human microsatellites. Whatever
find that <15% of mutation events are multi-step the cause of this heterogeneity, it will be interesting to see
changes5,58,65–68. However, the three largest human stud- whether directionality is related to microsatellite length
ies of this kind that have so far been presented reveal at the level of individual loci. Everything else being equal,
contradictory results. In agreement with other studies, we should expect loci that have a tendency to expand by
Ellegren69 and Xu et al.70 found 11–14% multi-step mutation to grow more often than those that tend to
mutations among 102 and 236 mutation events, respec- contract. To add to the complexity, several studies have
tively. By contrast, Huang et al.56, in an analysis of 97 found evidence of a negative correlation between direc-
mutation events, reported 63% multi-step changes. tion/magnitude of mutation and allele size56,63,64,69,70,72,74
What accounts for this discrepancy remains unclear, but — that is, long alleles being biased towards contraction.
it might indicate that some loci are more prone to large If generally true, this would offer a mechanistic explana-
changes than others. Analyses of individual loci in other tion for the stationary genomic length distributions
organisms have revealed highly variable proportions of seen at microsatellite loci. Interestingly, mutations from
multi-step changes, in the range of 5–75% (REFS three bacterial species show a downward bias, which can
62,63,71–73). A more extensive screening of zebrafish perhaps account for the rarity of microsatellites in
markers found 68% multi-step changes61. prokaryotic genomes75.

Box 2 | Using pedigrees to detect microsatellite mutations

The most straightforward approach for the study of Male
microsatellite mutations is direct observations of allele
transmissions in parent–child pairs (see figure).A mutant
allele (asterisk) is identified as being incompatible with
Mendelian inheritance because it is different in size from
194 221 Fragment length
parents’ alleles. Data from other markers are needed to
confirm that non-congruence between parental and Female
offspring genotypes is not a result of incorrect parentage133.
One limitation of this approach is that detectable mutations
are restricted to cases in which child genotypes cannot be
generated by transmission from parents’ genotypes.A
230 Fragment length 280
mutation to a character state that is already present in one
of the parents can therefore remain undetected. However, Child 1
likelihood-based estimates of mutation rates that take such
events into account have been developed134.
A related concern derives from the fact that even when
the mutant allele is different from parents’ alleles, it is
not evident from which progenitor allele it originates. Fragment length
The standard assumption made is that the smallest Child 2
mutational change is the most probable. However, this
introduces circularity when such data are used to
support the stepwise mutation model. Fortunately,
simulations show that this assumption is valid in most
cases, at least when compared with the alternative of Fragment length
random assignment to the progenitor allele73. Child 3
An alternative approach for the identification of germline
mutations is offered by the analysis of single sperm, or small
pools of sperm67,97. This gives virtually unlimited access to *
gametes from specific males to screen, allowing individual-
specific estimates of the mutation rate to be made. However, Fragment length 280

a disadvantage with this approach is that it is technically Child 4

demanding to amplify such small amounts of DNA, and
that it therefore requires most stringent laboratory routines
to avoid contamination. On the other hand, as it does not
require access to pedigrees, it might be useful for
applications such as genetic toxicology screenings. Fragment length

440 | JUNE 2004 | VOLUME 5 www.nature.com/reviews/genetics

 REVIEWS

0.1 Not only the birth but also the death of micro-
satellites can be captured by sequence comparisons83.
Such studies have supported the idea that, in the long
run, point mutations break up perfect repeats and
0.01
reduce the mutation rates of microsatellite loci.
Clearly, long microsatellite alleles do not persist
indefinitely. However, microsatellite evolution is a
Mutation rate

dynamic process; therefore, repeats might shrink as

well as expand over evolutionary timescales. In fact,
0.001 the removal of microsatellite interruptions by repli-
cation slippage 81 means that point mutations in
microsatellite arrays do not necessarily lead to decay
but might represent only a transition state during the
evolution of microsatellites. This feature should be
0.0001
incorporated into the model of Kruglyak et al.44 and
others43, which indicates that point mutations usually
0 5 10 15 20 25 30
destroy perfect repeats.
Marker
Intraspecific comparisons that reveal the sequence
Figure 3 | Microsatellite mutation rates in the human genome. Observed sex-average structure of individual alleles provide further evidence
mutation rates (log scale, with 95% confidence intervals) for human microsatellites obtained
for the complexity of the mutation process36,77,79,84,85.
from pedigree analysis. Markers are given in the following order: 1, D10S1214; 2, D12S1090;
3, ACTBP2; 4, D19S253; 5, D9S302; 6, D3S1744; 7, FGA; 8, HUMvWA31; 9, D22S683; There are numerous loci in a range of organisms at
10, D18S51; 11, D8S1179; 12, D21S11; 13, CYP19; 14, D3S1358; 15, HUMCSF1P0; which alleles differ not only in repeat length but also
16, D18S849; 17, D13S317; 18, D17S5; 19, D5S818; 20, D7S820; 21, Penta E; 22, D16S539; in repeat structure. Perhaps an extreme example of
23, HUMFESFPS; 24, HUMF13A01; 25, HUMLIPOL; 26, D1S80; 27, HUMF13B; 28, D2S1338; allele structure comes from 3 sequenced alleles at a
29, HUMTH01; and 30, HUMTPOX. Data from The Annual Report Summary for 2000 from the bird tetranucleotide repeat locus: allele 1, (AAAG)12;
US Parentage Testing Standards Committee.
allele 2, (AAAG)22A(AAAG)12; and allele 3, (AAAGA-
GAG)6(A)4(AG)3(AAAG)3(AG)9AA(AG)3(AAAG)2(A
G)2(AAAG)2(AGAGAAAG)15(AAAG)24 (REF. 80).
Sequence data. Another empirical approach to study-
ing microsatellite evolution involves characterizing the Mutation-rate variation
sequence structure of alleles within species, or compar- There is no uniform microsatellite mutation rate; the
ing the sequence of orthologous loci in different rates differ among loci and among alleles, and, perhaps
species. Using this approach, the effect of mutations as a consequence, also among species86. The single most
accumulated over evolutionary timescales can be read- important factor to affect mutation rate that has so far
ily studied, although it might be difficult to determine been discovered is microsatellite length — mutation
the precise order and character of individual mutation rate increases with an increasing number of repeat
events if they have occurred at high rates during the units. Intuitively, this seems understandable — more
time period being surveyed. Note that in contrast to repeat units give more opportunities for replication slip-
base substitutions, for which an infinite allele model is page. A length-dependent mutation rate explains part of
applicable, microsatellite alleles are often identical in the mutation-rate variation at several scales. The low
state (structure) but not by descent76. For example, two mutation rate in Drosophila melanogaster is compatible
chromosomes that are drawn from a population with with microsatellites being much shorter in flies than in
the sequence (GT)17 at a particular microsatellite locus vertebrates87. Within species, measures of repeat lengths
might have reached this state through mutations from correlate with mutability88. Furthermore, a positive cor-
(GT)16 or (GT)18 alleles. relation between allele size and mutation rate has been
Nevertheless, sequence analysis has shed light on the seen in many organisms60,66,72–74. The precise character of
genesis of new microsatellite loci, particularly in cases the relationship between repeat count and mutation
in which changes in orthologous microsatellite sequ- rate is less clear. Some studies have found a linear rela-
ences can be mapped on a phylogenetic tree77–81. Such tionship89, whereas some recent data indicate a power or
studies confirm that short repetitive sequences — with exponential relationship between size and rate10,88.
as few as two or three repeat units — are the starting But repeat length is not the sole cause of microsatel-
point for subsequent microsatellite expansion. These lite mutation-rate variation. As can be seen in FIG. 3, the
primary repeats can arise from normal base substitu- mutation rate of individual loci in a selected set of
tions, such as an A–G transition in GTATGT to human markers varies within two orders of magnitude,
GTGTGT. In addition, a significant proportion of new which cannot be attributed to repeat length. Similar
two-repeat loci arise from insertion mutations that are observations are made in other organisms62,63,72,90. One
duplications of adjacent sequence82. Interestingly, this important consequence of this variation is that the mean
would be compatible with a recent model proposed by mutation rate of a set of markers will vary a lot depending
Dieringer and Schlotterer42, which indicates that a on which particular markers are used. Moreover, as
length-independent mutation process operates on most markers that are used in genetic studies are
short microsatellites. selected on the basis of being (highly) polymorphic and,

NATURE REVIEWS | GENETICS VOLUME 5 | JUNE 2004 | 4 4 1

REVIEWS

given the expected relationship between polymor- frame of that gene can easily be monitored. Mutation
phism and mutability, the observed rates might not profiles have in this way been particularly well charac-
provide a representative picture for the genome as a terized in the yeast Saccharomyces cerevisiae, in which
whole. repeats that are integrated into chromosomes have also
What else matters? One possibility is that sequences been studied.
that flank the microsatellite affect the mutation rate91,92. These studies confirm several observations from
That inherent characteristics of individual loci are germline transmissions. Mismatch repair is identified
involved is indicated by covariation in levels of variabil- as crucial to microsatellite stability as mutation rates
ity at orthologous loci in related species93. However, in prokaryotic and eukaryotic cells that are deficient in
in addition to a flanking-sequence effect, this observa- mismatch repair are increased by several orders of
tion could also be compatible with an effect of, for magnitude compared with wild-type cells46,47,98,99.
example, TRANSCRIPTION-COUPLED REPAIR18,94, chromatin Mutations in genes that encode proof-reading exonu-
structure, regional sequence context and local point- cleases and some DNA polymerases have also been
mutation-rate variation. As for the last possibility, an implicated in repeat instability, although they have a
extension of the balance model of microsatellite evolu- more modest effect compared with mismatch-repair
tion states that not only will the equilibrium length deficiency100–102. In all systems, mutation rate increases
distribution of simple repeats be dependent on the with repeat length46,103,104, but interruptions stabilize
species-specific rate of point mutation, but the length repeat tracts105,106. The orientation of repeats — that is,
of individual repeat loci will also depend on the local whether a particular motif is on the coding or the com-
point-mutation rate, for which there is evidence for plementary strand — does not seem to affect the muta-
significant heterogeneity within genomes95. In a study tion rate98, with the exception of long trinucleotide
of orthologous microsatellite loci in the mouse and repeat arrays107. Observations of the destabilization of
rat, a negative correlation between microsatellite microsatellites by elevated levels of transcription would
length and substitution rate in nearby flanking sequ- support a role for transcription-coupled repair103.
ence was found96. This would indicate that when point The effect of sequence composition on the relative
mutations occur at high frequency, they seem to hin- instability of repeats is less clear. A study in Escherichia
der further microsatellite expansion (mutability) by coli found no significant difference in the mutability
introducing interruptions or imperfections in the of CA- and GA-repeats of similar length108, in contrast
repeat array. to observations in human cells59 (see also REF. 92).
Assuming a replication origin of microsatellite Conflicting observations are also made with respect to
mutations, we should expect the mutation rate to cor- the effect of the length of the repeat unit, as seen, for
relate with the number of germline cell divisions. By example, in di- versus tetranucleotide repeats99,109.
extension, mutations should be more frequent in males However, in E. coli110 and in human111 and yeast112 cells,
than in females, and in older males than in younger the mutability of G-mononucleotide repeats is higher
males. Although seemingly reasonable, these predic- than that of A-repeats of the same length, potentially
tions are only partly supported by empirical data. Two owing to stronger stacking interactions among Gs or Cs
of the large studies of human mutations find three–four than among As and Ts.
times more mutations in men69,70, which is close to Insertions generally outnumber deletions in eukary-
recent data on the male-to-female mutation-rate ratio otic cells103,113, whereas the opposite is true in E. coli46. In
for point mutations. However, Huang et al. saw no sex both cases, large deletions are frequently seen in long
bias in microsatellite mutation rate in their study. repeat tracts103. In general, at least two explanations
Moreover, significant variation in the mutational sex- might account for observations of a directional bias in
bias has been documented for swallow microsatellites, microsatellite mutation. The primary rate of slippage
with at least one locus showing a male-biased rate, mutation might be higher for insertions than for dele-
whereas others have female-biased rates60,72. Female- tions. Displaced loops might be more easily introduced
biased rates for individual loci have been reported for in the newly synthesized strand (which results in an
other organisms as well73,90. Attempts to correlate insertion) than in the template strand. Alternatively,
human microsatellite mutation rates with the father’s mismatch repair might more easily recognize or more
age have either failed to find such an effect58,68,97 or only efficiently repair displaced loops on the template strand
found a small effect66. than on the nascent strand99. That mismatch repair is
involved in a directional bias is indicated by the fact
Experimental approaches that mutations in some mismatch-repair genes, such as
The empirical data on microsatellite mutations de- yeast MSH3, differentially affect the rate of insertions
scribed above all refer to spontaneous events observed and deletions114. In D. melanogaster, mismatch repair
after germline transmissions. An experimental approach preferentially recognizes and/or corrects primary
to the study of microsatellite evolution is offered by the expansion mutations to leave an excess of contractions
TRANSCRIPTION-COUPLED analysis of instability of artificial plasmid-borne and, generally, (AT)n mutations are repaired more effi-
REPAIR microsatellite sequences introduced into bacterial or ciently than (GT)n changes115. If the character of mis-
Preferential repair of the
transcribed strand of an active
eukaryotic cells. By constructing plasmids with repeats match repair differs between groups of organisms, we
gene that is performed by that are associated with a resistance or a reporter gene, might expect consequent differences in microsatellite
excision-repair pathways. length mutations that disrupt or restore the reading frequency92.

442 | JUNE 2004 | VOLUME 5 www.nature.com/reviews/genetics

 REVIEWS

The future of microsatellites realistically, the use of many markers might compensate
The evolutionary process of simple repeats is far from for heterogeneity in mutational properties among loci.
simple. One important implication of the complexity of Microsatellites continue to find their application in
microsatellite evolution is, therefore, that care needs to areas such as linkage mapping, paternity testing, forensics
be taken when using microsatellite data in population and for the inference of demographic processes. More
genetics studies. For instance, significant mutation-rate recently, they have found most use in linkage-disequilib-
heterogeneity among loci means that it might be difficult rium mapping studies, in which associations between
to translate estimates of genetic distance into absolute markers and trait loci are searched for in population sam-
timescales. Similarly, directional biases in the mutation ples117, and in hitchhiking mapping, in which genome-
process have important consequences for the interpreta- wide screens for regions that show signs of selection are
tion of differences in allele size distributions among made118. But there are also prospects for new applications.
species, particularly if the character of the bias differs Given their high mutation rate, microsatellites offer a
among species 116. Future mathematical models of realistic means to study how the overall genomic muta-
microsatellite evolution should therefore aim to incor- tion rate is affected by environmental factors (genetic tox-
porate as many of the different forms of mutational het- icology). Elevated rates of microsatellite mutations in the
erogeneity as possible. Those who use microsatellites in germline have been seen in animals and plants that are
population genetics studies should select only the exposed to ionizing radiation119,120, and similar observa-
markers that are well characterized in terms of muta- tions have been made for minisatellites in humans121.
tional properties (mutation rates, directionality, Estimating microsatellite mutation rates in samples that
whether all alleles of equal length are identical in sequ- are exposed to different forms of radiation or toxic com-
ence), and, preferably, use markers that show uniform pounds could, when properly set in relation to data from
rates and patterns. Alternatively, but in many species less control groups, help to make risk assessments.

1. International Human Genome Sequencing Consortium. 18. Bachtrog, D., Weiss, S., Zangerl, B., Brem, G. & 33. Feldman, M. W., Bergman, A., Pollock, D. D. &
Initial sequencing and analysis of the human genome. Schlotterer, C. Distribution of dinucleotide microsatellites in Goldstein, D. B. Microsatellite genetic distances with range
Nature 409, 860–921 (2001). the Drosophila melanogaster genome. Mol. Biol. Evol. 16, constraints: analytic description and problems of estimation.
The first description and analysis of a publicly 602–610 (1999). Genetics 145, 207–216 (1997).
released assembly of the human genome. 19. Arcot, S. S., Wang, Z., Weber, J. L., Deininger, P. L. & 34. Pollock, D. D., Bergman, A., Feldman, M. W. &
2. Doolittle, W. F. & Sapienza, C. Selfish genes, the phenotype Batzer, M. A. Alu repeats: a source for the genesis of Goldstein, D. B. Microsatellite behavior with range
paradigm and genome evolution. Nature 284, 601–603 primate microsatellites. Genomics 29, 136–144 (1995). constraints: parameter estimation and improved distances
(1980). 20. Ramsay, L. et al. Intimate association of microsatellite for use in phylogenetic reconstruction. Theor. Popul. Biol.
3. Orgel, L. E. & Crick, F. H. Selfish DNA: the ultimate parasite. repeats with retrotransposons and other dispersed repetitive 53, 256–271 (1998).
Nature 284, 604–607 (1980). elements in barley. Plant J. 17, 415–425 (1999). 35. Stefanini, F. M. & Feldman, M. W. Bayesian estimation of
4. Weber, J. L. Informativeness of human (dC-dA)n. (dG-dT)n 21. Temnykh, S. et al. Computational and experimental analysis range for microsatellite loci. Genet. Res. 75, 167–177
polymorphisms. Genomics 7, 524–530 (1990). of microsatellites in rice (Oryza sativa L.): frequency, length (2000).
5. Weber, J. L. & Wong, C. Mutation of human short tandem variation, transposon associations, and genetic marker 36. Garza, J. C., Slatkin, M. & Freimer, N. B. Microsatellite allele
repeats. Hum. Mol. Genet. 2, 1123–1128 (1993). potential. Genome Res. 11, 1441–1452 (2001). frequencies in humans and chimpanzees, with implications
6. Metzgar, D. & Wills, C. Evidence for the adaptive evolution 22. Wilder, J. & Hollocher, H. Mobile elements and the genesis for constraints on allele size. Mol. Biol. Evol. 12, 594–603
of mutation rates. Cell 101, 581–584 (2000). of microsatellites in dipterans. Mol. Biol. Evol. 18, 384–392 (1995).
7. Albà, M. M. & Guigó, R. Comparative analysis of amino acid (2001). 37. Zhivotovsky, L. A. A new genetic distance with application to
repeats in rodents and humans. Genome Res. 14, 23. Field, D. & Wills, C. Abundant microsatellite polymorphism in constrained variation at microsatellite loci. Mol. Biol. Evol.
549–545 (2004). Saccharomyces cerevisiae, and the different distributions of 16, 467–471 (1999).
8. Mouse Genome Sequencing Consortium. Initial microsatellites in eight prokaryotes and S. cerevisiae, result 38. Calabrese, P. & Durrett, R. Dinucleotide repeats in the
sequencing and comparative analysis of the mouse from strong mutation pressures and a variety of selective Drosophila and human genomes have complex, length-
genome. Nature 420, 520–562 (2002). forces. Proc. Natl Acad. Sci. USA 95, 1647–1652 (1998). dependent mutation processes. Mol. Biol. Evol. 20,
9. Rat Genome Sequencing Project Consortium. Genome 24. Metzgar, D., Thomas, E., Davis, C., Field, D. & Wills, C. The 715–725 (2003).
sequence of the Brown Norway rat yields insights into microsatellites of Escherichia coli: rapidly evolving repetitive 39. Cooper, G., Burroughs, N. J., Rand, D. A., Rubinsztein, D. C.
mammalian evolution. Nature 428, 493–521 (2004). DNAs in a non-pathogenic prokaryote. Mol. Microbiol. 39, & Amos, W. Markov chain Monte Carlo analysis of human
10. Webster, M. T., Smith, N. G. & Ellegren, H. Microsatellite 183–190 (2001). Y-chromosome microsatellites provides evidence of biased
evolution inferred from human–chimpanzee genomic 25. Himmelreich, R. et al. Complete sequence analysis of the mutation. Proc. Natl Acad. Sci. USA 96, 11916–11921 (1999).
sequence alignments. Proc. Natl Acad. Sci. USA 99, genome of the bacterium Mycoplasma pneumoniae. 40. Nielsen, R. & Palsboll, P. J. Single-locus tests of
8748–8753 (2002). Nucleic Acids Res. 24, 4420–4449 (1996). microsatellite evolution: multi-step mutations and
Provides an unbiased comparison of microsatellite 26. Ohta, T. & Kimura, M. A model of mutation appropriate to constraints on allele size. Mol. Phylogenet. Evol. 11,
length in humans and chimpanzees on the basis of estimate the number of electrophoretically detectable alleles 477–484 (1999).
orthologous genome sequence data. in a finite population. Genet. Res. 22, 201–204 (1973). 41. Renwick, A., Davison, L., Spratt, H., King, J. P. & Kimmel, M.
11. Pascual, M., Schug, M. D. & Aquadro, C. F. High density of Classic description of the stepwise mutation model. DNA dinucleotide evolution in humans: fitting theory to facts.
long dinucleotide microsatellites in Drosophila subobscura. 27. Shriver, M. D., Jin, L., Chakraborty, R. & Boerwinkle, E. Genetics 159, 737–747 (2001).
Mol. Biol. Evol. 17, 1259–1267 (2000). VNTR allele frequency distributions under the stepwise 42. Dieringer, D. & Schlotterer, C. Two distinct modes of
12. Schlotterer, C. & Harr, B. Drosophila virilis has long and mutation model: a computer simulation approach. Genetics microsatellite mutation processes: evidence from the
highly polymorphic microsatellites. Mol. Biol. Evol. 17, 134, 983–993 (1993). complete genomic sequences of nine species. Genome
1641–1646 (2000). 28. Valdes, A. M., Slatkin, M. & Freimer, N. B. Allele frequencies Res. 13, 2242–2251 (2003).
13. Hancock, J. M. Simple sequences in a ‘minimal’ genome. at microsatellite loci: the stepwise mutation model revisited. 43. Bell, G. I. & Jurka, J. The length distribution of perfect dimer
Nature Genet. 14, 14–15 (1996). Genetics 133, 737–749 (1993). repetitive DNA is consistent with its evolution by an unbiased
14. Toth, G., Gaspari, Z. & Jurka, J. Microsatellites in different 29. Kimmel, M. & Chakraborty, R. Measures of variation at DNA single-step mutation process. J. Mol. Evol. 44, 414–421
eukaryotic genomes: survey and analysis. Genome Res. repeat loci under a general stepwise mutation model. Theor. (1997).
10, 967–981 (2000). Popul. Biol. 50, 345–367 (1996). 44. Kruglyak, S., Durrett, R. T., Schug, M. D. & Aquadro, C. F.
15. Katti, M. V., Ranjekar, P. K. & Gupta, V. S. Differential 30. Kimmel, M., Chakraborty, R., Stivers, D. N. & Deka, R. Equilibrium distributions of microsatellite repeat length resulting
distribution of simple sequence repeats in eukaryotic Dynamics of repeat polymorphisms under a forward- from a balance between slippage events and point mutations.
genome sequences. Mol. Biol. Evol. 18, 1161–1167 (2001). backward mutation model: within- and between-population Proc. Natl Acad. Sci. USA 95, 10774–10778 (1998).
16. Morgante, M., Hanafey, M. & Powell, W. Microsatellites are variability at microsatellite loci. Genetics 143, 549–555 (1996). Provides an integrated model of microsatellite
preferentially associated with nonrepetitive DNA in plant 31. Di Rienzo, A. et al. Mutational processes of simple- evoluton that takes length mutations as well as base
genomes. Nature Genet. 30, 194–200 (2002). sequence repeat loci in human populations. Proc. Natl substitutions into account.
17. Lagercrantz, U., Ellegren, H. & Andersson, L. The Acad. Sci. USA 91, 3166–3170 (1994). 45. Calabrese, P. P., Durrett, R. T. & Aquadro, C. F. Dynamics of
abundance of various polymorphic microsatellite motifs 32. Nauta, M. J. & Weissing, F. J. Constraints on allele size at microsatellite divergence under stepwise mutation and
differs between plants and vertebrates. Nucleic Acids Res. microsatellite loci: implications for genetic differentiation. proportional slippage/point mutation models. Genetics 159,
21, 1111–1115 (1993). Genetics 143, 1021–1032 (1996). 839–852 (2001).

NATURE REVIEWS | GENETICS VOLUME 5 | JUNE 2004 | 4 4 3

REVIEWS

46. Levinson, G. & Gutman, G. A. High frequencies of short 70. Xu, X., Peng, M. & Fang, Z. The direction of microsatellite 96. Santibanez-Koref, M. F., Gangeswaran, R. & Hancock, J. M.
frameshifts in poly-CA/TG tandem repeats borne by mutations is dependent upon allele length. Nature Genet. A relationship between lengths of microsatellites and nearby
bacteriophage M13 in Escherichia coli K-12. Nucleic Acids 24, 396–399 (2000). substitution rates in mammalian genomes. Mol. Biol. Evol.
Res. 15, 5323–5338 (1987). 71. Primmer, C. R., Saino, N., Moller, A. P. & Ellegren, H. 18, 2119–2123 (2001).
Firm demonstration of replication slippage as a main Directional evolution in germline microsatellite mutations. Offers a suggestion for how variation in microsatellite
mechanism for microsatellite instability. Nature Genet. 13, 391–393 (1996). length might relate to point-mutation-rate
47. Strand, M., Prolla, T. A., Liskay, R. M. & Petes, T. D. 72. Primmer, C. R., Saino, N., Moller, A. P. and Ellegren, H. heterogeneity.
Destabilization of tracts of simple repetitive DNA in yeast by Unraveling the process of microsatellite evolution through 97. Brohede, J., Arnheim, N. & Ellegren, H. Single molecule
mutations affecting DNA mismatch repair. Nature 365, analysis of germ line mutations in barn swallows Hirundo analysis of the hypermutable tetranucleotide repeat locus
274–276 (1993). rustica. Mol. Biol. Evol. 15, 1047–1054 (1998). D21S1245 through sperm genotyping: a heterogeneous
48. Schlotterer, C. & Tautz, D. Slippage synthesis of simple 73. Beck, N. R., Double, M. C. & Cockburn, A. Microsatellite pattern of mutation but no clear male age effect. Mol. Biol.
sequence DNA. Nucleic Acids Res. 20, 211–215 (1992). evolution at two hypervariable loci revealed by extensive Evol. 21, 58–64 (2004).
Experimental evidence that DNA polymerase is the avian pedigrees. Mol. Biol. Evol. 20, 54–61 (2003). 98. Henderson, S. T. & Petes, T. D. Instability of simple
only enzymatic activity needed for replication 74. Harr, B. & Schlotterer, C. Long microsatellite alleles in sequence DNA in Saccharomyces cerevisiae. Mol. Cell.
slippage. Drosophila melanogaster have a downward mutation bias Biol. 12, 2749–2757 (1992).
49. Hile, S. E. & Eckert, K. A. Positive correlation between DNA and short persistence times, which cause their genome- Introduces an experimental approach to the study of
polymerase α-primase pausing and mutagenesis within wide underrepresentation. Genetics 155, 1213–1220 microsatellite mutations, using artificial plasmid-
polypyrimidine/polypurine microsatellite sequences. J. Mol. (2000). borne repeat sequences associated with a resistance
Biol. 335, 745–759 (2004). 75. Metzgar, D., Liu, L., Hansen, C., Dybvig, K. & Wills, C. or reporter gene.
50. Hauge, X. Y. & Litt, M. A study of the origin of ‘shadow Domain-level differences in microsatellite distribution and 99. Sia, E. A., Kokoska, R. J., Dominska, M., Greenwell, P. &
bands’ seen when typing dinucleotide repeat content result from different relative rates of insertion and Petes, T. D. Microsatellite instability in yeast: dependence on
polymorphisms by the PCR. Hum. Mol. Genet. 2, 411–415 deletion mutations. Genome Res. 12, 408–413 (2002). repeat unit size and DNA mismatch repair genes. Mol. Cell.
(1993). 76. Estoup, A., Jarne, P. & Cornuet, J. M. Homoplasy and Biol. 17, 2851–2858 (1997).
51. Murray, V., Monchawin, C. & England, P. R. The mutation model at microsatellite loci and their 100. Strauss, B. S., Sagher, D. & Acharya, S. Role of
determination of the sequences present in the shadow consequences for population genetics analysis. Mol. Ecol. proofreading and mismatch repair in maintaining the stability
bands of a dinucleotide repeat PCR. Nucleic Acids Res. 21, 11, 1591–1604 (2002). of nucleotide repeats in DNA. Nucleic Acids Res. 25,
2395–2398 (1993). A useful overview of the implications of microsatellite 806–813 (1997).
52. Shinde, D., Lai, Y., Sun, F. & Arnheim, N. Taq DNA homoplasy in evolutionary studies. 101. Tran, H. T., Keen, J. D., Kricker, M., Resnick, M. A. &
polymerase slippage mutation rates measured by PCR and 77. Messier, W., Li, S. H. & Stewart, C. B. The birth of Gordenin, D. A. Hypermutability of homonucleotide runs in
quasi-likelihood analysis: (CA/GT)n and (A/T)n microsatellites. microsatellites. Nature 381, 483 (1996). mismatch repair and DNA polymerase proofreading yeast
Nucleic Acids Res. 31, 974–980 (2003). 78. Orti, G., Pearse, D. E. & Avise, J. C. Phylogenetic mutants. Mol. Cell. Biol. 17, 2859–2865 (1997).
53. Berg, I., Neumann, R., Cederberg, H., Rannug, U. & assessment of length variation at a microsatellite locus. 102. Gutierrez, P. J. & Wang, T. S. Genomic instability induced by
Jeffreys, A. J. Two modes of germline instability at human Proc. Natl Acad. Sci. USA 94, 10745–10749 (1997). mutations in Saccharomyces cerevisiae POL1. Genetics
minisatellite MS1 (locus D1S7): complex rearrangements 79. Angers, B. & Bernatchez, L. Complex evolution of a 165, 65–81 (2003).
and paradoxical hyperdeletion. Am. J. Hum. Genet. 72, salmonid microsatellite locus and its consequences in 103. Wierdl, M., Dominska, M. & Petes, T. D. Microsatellite
1436–1447 (2003). inferring allelic divergence from size information. Mol. Biol. instability in yeast: dependence on the length of the
54. Majewski, J. & Ott, J. GT repeats are associated with Evol. 14, 230–238 (1997). microsatellite. Genetics 146, 769–779 (1997).
recombination on human chromosome 22. Genome Res. 80. Primmer, C. R. & Ellegren, H. Patterns of molecular evolution 104. Yamada, N. A. et al. Relative rates of insertion and
10, 1108–1114 (2000). in avian microsatellites. Mol. Biol. Evol. 15, 997–1008 (1998). deletion mutations in dinucleotide repeats of various
55. Treco, D. & Arnheim, N. The evolutionarily conserved 81. Harr, B., Zangerl, B. & Schlotterer, C. Removal of lengths in mismatch repair proficient mouse and
repetitive sequence d(TG. AC)n promotes reciprocal microsatellite interruptions by DNA replication slippage: mismatch repair deficient human cells. Mutat. Res. 499,
exchange and generates unusual recombinant tetrads phylogenetic evidence from Drosophila. Mol. Biol. Evol. 17, 213–225 (2002).
during yeast meiosis. Mol. Cell. Biol. 6, 3934–3947 (1986). 1001–1009 (2000). 105. Petes, T. D., Greenwell, P. W. & Dominska, M. Stabilization
56. Huang, Q. Y. et al. Mutation patterns at dinucleotide 82. Zhu, Y., Strassmann, J. E. & Queller, D. C. Insertions, of microsatellite sequences by variant repeats in the yeast
microsatellite loci in humans. Am. J. Hum. Genet. 70, substitutions, and the origin of microsatellites. Genet. Res. Saccharomyces cerevisiae. Genetics 146, 491–498
625–634 (2002). 76, 227–236 (2000). (1997).
57. Heyer, E., Puymirat, J., Dieltjes, P., Bakker, E. & de Knijff, P. 83. Taylor, J. S., Durkin, J. M. & Breden, F. The death of a 106. Bacon, A. L., Farrington, S. M. & Dunlop, M. G. Sequence
Estimating Y chromosome specific microsatellite mutation microsatellite: a phylogenetic perspective on microsatellite interruptions confer differential stability at microsatellite
frequencies using deep rooting pedigrees. Hum. Mol. interruptions. Mol. Biol. Evol. 16, 567–572 (1999). alleles in mismatch repair-deficient cells. Hum. Mol. Genet.
Genet. 6, 799–803 (1997). 84. Schlotterer, C., Amos, B. & Tautz, D. Conservation of 9, 2707–2713 (2000).
58. Kayser, M. et al. Characteristics and frequency of polymorphic simple sequence loci in cetacean species. 107. Maurer, D. J., O’Callaghan, B. L. & Livingston, D. M.
germline mutations at microsatellite loci from the human Nature 354, 63–65 (1991). Orientation dependence of trinucleotide CAG repeat
Y chromosome, as revealed by direct observation in 85. Colson, I. & Goldstein, D. B. Evidence for complex instability in Saccharomyces cerevisiae. Mol. Cell. Biol. 16,
father/son pairs. Am. J. Hum. Genet. 66, 1580–1588 (2000). mutations at microsatellite loci in Drosophila. Genetics 152, 6617–6622 (1996).
59. Hile, S. E., Yan, G. & Eckert, K. A. Somatic mutation rates 617–627 (1999). 108. Eckert, K. A. & Yan, G. Mutational analyses of dinucleotide
and specificities at TC/AG and GT/CA microsatellite 86. Ellegren, H. Microsatellite mutations in the germline: and tetranucleotide microsatellites in Escherichia coli:
sequences in nontumorigenic human lymphoblastoid cells. implications for evolutionary inference. Trends Genet. 16, influence of sequence on expansion mutagenesis. Nucleic
Cancer Res. 60, 1698–1703 (2000). 551–558 (2000). Acids Res. 28, 2831–2838 (2000).
60. Brohede, J., Primmer, C. R., Moller, A. & Ellegren, H. 87. Schug, M. D., Mackay, T. F. & Aquadro, C. F. Low mutation 109. Lee, J. S., Hanford, M. G., Genova, J. L. & Farber, R. A.
Heterogeneity in the rate and pattern of germline mutation at rates of microsatellite loci in Drosophila melanogaster. Relative stabilities of dinucleotide and tetranucleotide
individual microsatellite loci. Nucleic Acids Res. 30, Nature Genet. 15, 99–102 (1997). repeats in cultured mammalian cells. Hum. Mol. Genet. 8,
1997–2003 (2002). Characterizaton of mutation rate and repeat lengths 2567–2572 (1999).
61. Shimoda, N. et al. Zebrafish genetic map with 2000 of microsatellites in D. melanogaster, revealing 110. Sagher, D., Hsu, A. & Strauss, B. Stabilization of the
microsatellite markers. Genomics 58, 219–232 (1999). significant differences from vertebrate genomes. intermediate in frameshift mutation. Mutat. Res. 423, 73–77
62. Fitzsimmons, N. N. Single paternity of clutches and sperm 88. Leopoldino, A. M. & Pena, S. D. The mutational spectrum of (1999).
storage in the promiscuous green turtle (Chelonia mydas). human autosomal tetranucleotide microsatellites. Hum. 111. Boyer, J. C. et al. Sequence dependent instability of
Mol. Ecol. 7, 575–584 (1998). Mutat. 21, 71–79 (2003). mononucleotide microsatellites in cultured mismatch repair
63. Gardner, M. G., Bull, C. M., Cooper, S. J. & Duffield, G. A. 89. Sibly, R. M. et al. The structure of interrupted human AC proficient and deficient mammalian cells. Hum. Mol. Genet.
Microsatellite mutations in litters of the Australian lizard microsatellites. Mol. Biol. Evol. 20, 453–459 (2003). 11, 707–713 (2002).
Egernia stokesii. J. Evol. Biol. 13, 551–560 (2000). 90. Crozier, R. H., Kaufmann, B., Carew, M. E. & Crozier, Y. C. 112. Harfe, B. D. & Jinks-Robertson, S. Sequence composition
64. Jones, A. G., Rosenqvist, G., Berglund, A. & Avise, J. C. Mutability of microsatellites developed for the ant and context effects on the generation and repair of
Clustered microsatellite mutations in the pipefish Camponotus consobrinus. Mol. Ecol. 8, 271–276 (1999). frameshift intermediates in mononucleotide runs in
Syngnathus typhle. Genetics 152, 1057–1063 (1999). 91. Glenn, T. C., Stephan, W., Dessauer, H. C. & Braun, M. J. Saccharomyces cerevisiae. Genetics 156, 571–578
65. Amos, W., Sawcer, S. J., Feakes, R. W. & Rubinsztein, D. C. Allelic diversity in alligator microsatellite loci is negatively (2000).
Microsatellites show mutational bias and heterozygote correlated with GC content of flanking sequences and 113. Twerdi, C. D., Boyer, J. C. & Farber, R. A. Relative rates of
instability. Nature Genet. 13, 390–391 (1996). evolutionary conservation of PCR amplifiability. Mol. Biol. insertion and deletion mutations in a microsatellite sequence
66. Brinkmann, B., Klintschar, M., Neuhuber, F., Huhne, J. & Evol. 13, 1151–1154 (1996). in cultured cells. Proc. Natl Acad. Sci. USA 96, 2875–2879
Rolf, B. Mutation rate in human microsatellites: influence of 92. Bachtrog, D., Agis, M., Imhof, M. & Schlotterer, C. (1999).
the structure and length of the tandem repeat. Am. J. Hum. Microsatellite variability differs between dinucleotide repeat 114. Strand, M., Earley, M. C., Crouse, G. F. & Petes, T. D.
Genet. 62, 1408–1415 (1998). motifs-evidence from Drosophila melanogaster. Mol. Biol. Mutations in the MSH3 gene preferentially lead to deletions
67. Holtkemper, U., Rolf, B., Hohoff, C., Forster, P. & Evol. 17, 1277–1285 (2000). within tracts of simple repetitive DNA in Saccharomyces
Brinkmann, B. Mutation rates at two human Y-chromosomal 93. Harr, B., Zangerl, B., Brem, G. & Schlotterer, C. cerevisiae. Proc. Natl Acad. Sci. USA 92, 10418–10421
microsatellite loci using small pool PCR techniques. Hum. Conservation of locus-specific microsatellite variability (1995).
Mol. Genet. 10, 629–633 (2001). across species: a comparison of two Drosophila sibling 115. Harr, B., Todorova, J. & Schlotterer, C. Mismatch repair-
Introduction of sperm typing as an alternative means species, D. melanogaster and D. simulans. Mol. Biol. Evol. driven mutational bias in D. melanogaster. Mol. Cell 10,
for microsatellite mutation detection. 15, 176–184 (1998). 199–205 (2002).
68. Myhre Dupuy, B., Stenersen, M., Egeland, T. & Olaisen, B. 94. Mellon, I., Rajpal, D. K., Koi, M., Boland, C. R. & 116. Amos, W., Hutter, C. M., Schug, M. D. & Aquadro, C. F.
Y-chromosomal microsatellite mutation rates: differences in Champe, G. N. Transcription-coupled repair deficiency Directional evolution of size coupled with ascertainment bias
mutation rate between and within loci. Hum. Mutat. 23, and mutations in human mismatch repair genes. Science for variation in Drosophila microsatellites. Mol. Biol. Evol. 20,
117–124 (2004). 272, 557–560 (1996). 660–662 (2003).
69. Ellegren, H. Heterogeneous mutation processes in human 95. Ellegren, H., Smith, N. G. & Webster, M. T. Mutation rate 117. Ohashi, J. & Tokunaga, K. Power of genome-wide linkage
microsatellite DNA sequences. Nature Genet. 24, 400–402 variation in the mammalian genome. Curr. Opin. Genet. Dev. disequilibrium testing by using microsatellite markers.
(2000). 13, 562–568 (2003). J. Hum. Genet. 48, 487–491 (2003).

444 | JUNE 2004 | VOLUME 5 www.nature.com/reviews/genetics

 REVIEWS

118. Schlotterer, C. Hitchhiking mapping — functional genomics 126. Tautz, D., Trick, M. & Dover, G. A. Cryptic simplicity in DNA is 134. Whittaker, J. C. et al. Likelihood-based estimation of
from the population genetics perspective. Trends Genet. 19, a major source of genetic variation. Nature 322, 652–656 microsatellite mutation rates. Genetics 164, 781–787
32–38 (2003). (1986). (2003).
119. Ellegren, H., Lindgren, G., Primmer, C. R. & Moller, A. P. 127. Litt, M. & Luty, J. A. A hypervariable microsatellite revealed
Fitness loss and germline mutations in barn swallows by in vitro amplification of a dinucleotide repeat within the Acknowledgements
breeding in Chernobyl. Nature 389, 593–596 (1997). cardiac muscle actin gene. Am. J. Hum. Genet. 44, The author would like to acknowledge two particularly helpful
120. Kovalchuk, O., Kovalchuk, I., Arkhipov, A., Hohn, B. & 397–401 (1989). reviewers who provided useful comments on the manuscript. The
Dubrova, Y. E. Extremely complex pattern of microsatellite This paper, and references 128 and 129, introduce the author’s work was supported in part by the Swedish Research
mutation in the germline of wheat exposed to the post- use of PCR for genotyping microsatellites. Council for Environment, Agricultural Sciences and Spatial
Chernobyl radioactive contamination. Mutat. Res. 525, 128. Weber, J. L. & May, P. E. Abundant class of human DNA Planning.
93–101 (2003). polymorphisms which can be typed using the polymerase
121. Dubrova, Y. E. et al. Human minisatellite mutation rate after Competing interests statement
chain reaction. Am. J. Hum. Genet. 44, 388–396 (1989).
the Chernobyl accident. Nature 380, 683–686 (1996). The author declares that he has no competing financial interests.
129. Tautz, D. Hypervariability of simple sequences as a general
122. Spritz, R. A. Duplication/deletion polymorphism 5′- to the
source for polymorphic DNA markers. Nucleic Acids Res.
human β-globin gene. Nucleic Acids Res. 9, 5037–5047 Online links
17, 6463–6471 (1989).
(1981).
123. Miesfeld, R., Krystal, M. & Arnheim, N. A member of a new 130. Ellegren, H. DNA typing of museum birds. Nature 354, 113 DATABASES
repeated sequence family which is conserved throughout (1991). The following terms in this article are linked online to:
eucaryotic evolution is found between the human δ- and 131. Weissenbach, J. et al. A second-generation linkage map of Entrez: http://www.ncbi.nih.gov/Entrez
β-globin genes. Nucleic Acids Res. 9, 5931–5947 (1981). the human genome. Nature 359, 794–801 (1992). MSH3
124. Hamada, H. & Kakunaga, T. Potential Z-DNA forming 132. Bowcock, A. M. et al. High resolution of human evolutionary
sequences are highly dispersed in the human genome. trees with polymorphic microsatellites. Nature 368, 455–457 FURTHER INFORMATION
Nature 298, 396–398 (1982). (1994). RepeatMasker: http://www.repeatmasker.org
125. Jeffreys, A. J., Wilson, V. & Thein, S. L. Hypervariable 133. Dawid, A. P., Mortera, J. & Pascali, V. L. Non-fatherhood or Sputnik: http://espressosoftware.com/pages/sputnik.jsp
‘minisatellite’ regions in human DNA. Nature 314, 67–73 mutation? A probabilistic approach to parental exclusion in Tandem Repeats Finder: http://c3.biomath.mssm.edu/trf.html
(1985). paternity testing. Forensic Sci. Int. 124, 55–61 (2001). Access to this links box is available online.

NATURE REVIEWS | GENETICS VOLUME 5 | JUNE 2004 | 4 4 5