This study was based on samples previously collected for the purpose of paternity testing in Canadian populations of barn swallows and tree swallows. Both species are socially monogamous passerine birds, but with high levels of extrapair paternity 3334. The barn swallow samples consisted of those already reported by Kleven et al. 33 and additional samples collected during 2003 and 2004. Barn swallows were genotyped with six to nine polymorphic microsatellite markers, including HrU10, and a detailed description of the markers, their variability and the paternity determination are presented elsewhere 33.

Tree swallows were genotyped with three polymorphic microsatellite markers, including HrU10. In cases where one of the three markers showed an allelic mismatch between offspring and one of the putative parents, an additional triplet of microsatellite markers were analyzed to distinguish mutation events from extrapair paternity. Details about the microsatellite markers, their polymorphism and the parentage determination of tree swallows are provided as additional file [see Additional file 1].

Mutations were detected by comparing the genotype of the offspring with that of its biological parents. We only included individuals for whom the genotypes of both biological parents were available. Furthermore, to avoid the problem of non-amplifying alleles, we only included mutations in parents that were heterozygous at the HrU10 locus. We assumed the smallest mutational change in allele size to be most likely in cases where more than one parental allele could be the progenitor allele. To verify observed mutation events, we amplified microsatellite fragments twice for the parents and offspring involved in these cases.

The HrU10 locus was amplified according to the protocol described in 33. In addition, the reverse primer HrU10-EXT-R2 (5'-GCTGCTGTTCGAGGAAATAA-3') was designed to improve sequencing of tree swallow alleles. To reduce time consumption and lab costs we optimized a simple isolation strategy for HrU10 alleles that did not require cloning. Alleles with size differences > 10 base pairs (bp) were separated on MetaPhor^® agarose gels (Cambrex, East Rutherford, NJ) or on standard SeaKem^® LE Agarose (Cambrex) if the alleles differed > 50 bp in size. The allele of interest was subsequently purified with the Nucleospin^® Extract II gel extraction Kit (Macherey-Nagel, Düren, Germany), and sequenced directly in both directions using the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA) according to the manufacturer's recommendations on an ABI 3100 Genetic Analyser (Applied Biosystems). As a further control, the sequence lengths were compared to previous fragment length analysis of the microsatellite 33.

Pedigree analysis based on 2076 meiotic events in barn swallows and 496 meiotic events in tree swallows revealed 41 mutations for the microsatellite locus HrU10 in barn swallows and 15 in tree swallows. According to fragment length analyses all mutations involved a gain or loss of five or ten bp, i.e. equivalent to one or two repeat units, which is consistent with the assumption of replication slippage. This represents a slippage rate of 1.97 × 10^-2 in barn swallows and 3.02 × 10^-2 in tree swallows. No other types of indels were observed. Sixty-six HrU10 alleles involved in the observed germ line mutation were sequenced, that is, 33 sets of the offspring and the donor parent (GenBank accession numbers EU295565–EU295630). There were no site (four colonies) or year (two years) effects on mutation rates in barn swallows (both P >0.1). The tree swallows were only sampled for one year at one location.

Woodruff et al. 35 approached the issue of "clustered mutations", which implies that related individuals may inherit identical genetic changes, contrasting an assumption that mutations are independent events. In this respect six of the parental individuals gave rise to two mutations in the same family in the barn swallow population. In these cases, identical lengths where observed only twice. Only one of these two incidents of identical mutant lengths gave adequate sequences of both mutants. Nevertheless, in the particular case (Mut12 and Mut18 [see Additional file 2]) where sequences were obtained from both mutants, the outcome of the two mutations was dissimilar, verifying that the mutations represent independent events.

The nucleotide sequences confirmed that there was a gain or a contraction of one or two pentanucleotide units in all the 33 germ lines (26 in barn swallows and seven in tree swallows [see Additional file 2]). No nucleotide substitutions or other indels were detected. The sequence data revealed several different IMs for the HrU10 alleles in both species, all of which consisted of distinct rearrangements of 1–30 pyrimidine bases (T and C). No purine bases were detected in the pyrimidine rich strand of the HrU10 microsatellite. The subsequent statistical analyses were performed on the barn swallow parental alleles only, as the number of alleles that gave adequate sequences in tree swallows was considered too low (n = 7) for statistical testing.

First, we tested the hypothesis that longer microsatellites are more unstable and will consequently contain more IMs. If such a correlation occurred, then one would expect an upper size limit for the length of perfect repeats, and, accordingly there should be no correlation between the length of the longest stretch of perfect repeats of a given HrU10 allele and the number of IMs. Correlation tests between total microsatellite length and (1) number of IMs (Spearman: r_s = 0.55, P = 0.003), and (2) total number of nucleotides contributing to the IMs (r_s = 0.54, P = 0.004) were significant. No significant correlation was found between the longest stretch of perfect tandemly repeated units and (1) number of IMs (r_s = 0.09, P = 0.66) or (2) total number of nucleotides in the IMs (r_s = 0.14, P = 0.5) (Figure 1).

We further determined roughly in which part of the HrU10 allele a slippage event had taken place. The 16 shortest alleles showed one long (17–29 units) and one or two short (1–2 units) stretches of perfectly repeated units that are separated by IMs. The 10 longest alleles showed two long (14–55 units) and up to three short (0–5 units) stretches of perfectly repeated units. All replication induced mutations could be attributed to the long (≥14 units) stretches of perfect repeats. In 21 of 26 alleles (81%), the slippage mutation affected the longest stretch of perfectly repeated units. It is noteworthy that the other 5 mutations that did not occur in the longest stretch of perfect repeats, all affected the alleles that were among the 10 longest.

In the barn swallow, 15 of the 26 (58%) observed mutations resulted in mutant alleles that were homoplasic to another sequenced allele in the data set. Nine (35%) of these mutations resulted in homoplasic alleles with respect to length, and six (23%), mutant alleles were also identical in sequence to another allele. No incidents of homoplasy due to mutations were detected in the tree swallows, but the number of sequences was significantly lower.

The genotyped adult population consisted of 376 and 144 individuals in the barn and tree swallow populations, respectively. The allele size frequencies for the HrU10 locus in the two species are illustrated in Figure 2 (note: 95 bp in the flanking regions of the microsatellite are not included in the presented sequences but in the fragment analyses). Due to seven non-amplifying alleles, only 745 barn swallow alleles were included. The median allele size of the HrU10 microsatellite in the adult barn swallow population was 231 bp (± 2.0 SE, range = 175–581 bp, n = 704 alleles). Median size of the microsatellite of the mutant barn swallow alleles was 241 bp (± 13.8, range = 193–586, n = 41), i.e. significantly longer than for the entire population (Mann-Whitney U test: Z = -2.4, P = 0.02). The median allele size for the HrU10 locus in the tree swallow population was 280 bp (± 4.6, range = 186–605, n = 272 alleles), and the median microsatellite length of the mutant tree swallow alleles was 323 bp (± 24.5, range = 222–518, n = 15). The mutation rates showed a tendency to be positively correlated with allele sizes for both barn swallows (GLM with binominal error distribution and logit link: χ²₁ = 9.26, P = 0.002) and tree swallows (GLM with binominal error distribution and logit link: χ²₁ = 3.35, P = 0.067). Estimations of mutation rates in relation to allele sizes are illustrated in Figure 3.

There were 26 (65%) expanding and 14 (35%) contracting mutations in the barn swallow population, which was not significantly different from equity (binominal test (two-tailed): P = 0.08), but may indicate a bias toward expansion. Directionality of the mutation was impossible to determine in one case because of a 5 bp difference to both possible parental alleles. No indication of directional bias for HrU10 mutations was found in tree swallows with 7 (47%) expansions and 8 (53%) contractions (binominal test: P = 1.0). Directionality of mutations were not significantly affected by allele size neither in barn swallows (GLZ with binomial error distribution; χ² = 0.2, P = 0.7) nor in tree swallows (χ² = 0.1, P = 0.78). Furthermore, there was no positive correlation between length of the longest perfectly repeated core microsatellite unit and the directionality of the mutations in barn swallows (GLZ with binomial error distribution; χ² = 0.1, P = 0.81).

In both species, there was a tendency for mutations to be maternally transmitted, as 66% (27/41) of the barn swallow mutations were observed in the female germ line (binominal test: P = 0.06) and 80% (12/15) in tree swallows (binominal test: P = 0.035).

All IMs consisted of different arrangements of Cs and Ts. No introduction of purines on the pyrimidine-rich strand was observed. The repeated motif of the HrU10 locus is a pentanucleotide, and its spatial conformation during a slippage event may involve a larger loop compared to a smaller repeat unit. This may indicate that all the HrU10 IMs are results of uncompleted replication slippage. If so, one needs to assume that the loop on the nascent or template strand will not include an entire unit before it realigns. This hypothesis is in line with all the IMs detected for HrU10, except for the (TTCCC)₆-repeat in Mut5 and Mut6 [see Additional file 2], which happened to be the longest HrU10 alleles sequenced in the barn swallows. These two mutants have originated from the same parental allele that is 3.5 times longer than the median allele length.

An interesting feature related to the IMs, is the relatively constant maximal length of stretches of perfectly repeated core units, indicating a threshold length for stable arrays of perfectly repeated microsatellite units (Figure 1). One explanation for such a threshold level may be a selection pressure to retain stabile conformations. Such a selection pressure has been suggested to be important for preserving folding potential in repeated di-nucleotides 2136 and tri-nucleotides 37. In this respect, nothing is currently known for pentanucleotides.

Because the longest stretch of perfectly repeated units contributes most to the total microsatellite length in the HrU10 alleles with only one long stretch of perfect tandem repeats, it is not surprising that the slippage events were restricted to this region of the microsatellite. Furthermore, theoretical and experimental approaches have suggested that slippage rates are approximately zero in very short repeated regions 38394041, supporting a low probability for slippages to occur in the short stretches in the HrU10 alleles. Nevertheless, if length of perfectly repeated tandem units is more crucial for microsatellite instability than total microsatellite length, one would predict a consistent bias for slippage to affect the longest stretch of alleles comprising at least two long (>14 units) stretches of perfect core units. However, only 50% (5/10) of the mutations were introduced in the longer of the two stretches. This result is in agreement with a theory of total length of entire microsatellite being a more important factor for microsatellite stability than longest motif of perfect tandemly units. The observed threshold level for length of perfectly repeated units before introductions of IMs might then be explained by incomplete slippage (described above) being an increasingly more important mutational mechanism as the allele is destabilized due to growth. However, the statistical power for such a conclusion is relatively poor. It is noteworthy that an opposite pattern was detected in an in vitro system of mono- and dinucleotides in a mutation study of human cell lineages 38. Selection pressure for the maintenance of IMs has also been proposed important for tri-nucleotide repeat stability in genes related to various types of spinocerebellar ataxia 163742.

The results from the fragment length analyses also support the prediction of long HrU10 alleles to be more unstable, as the probability of a slippage event was positively related to allele sizes. The mutation rates were estimated to be more than three-fold higher in the longest compared with the smallest allele classes in the two swallow populations (Figure 3). Higher mutation rates for longer repeated regions have been reported in a number of previous studies 51011303236383943444546. The higher mutability in longer alleles can be explained by stabilization patterns concerning the mismatch-repair system, which may be less effective and as a result generate a relatively large probability for insertion of slippage events if the repeated region is sufficiently long, as suggested by Wierdl et al. 32. The logic in this theory is that the number of possible conformations increases proportionally with the increase of repeated microsatellite units. However, if this hypothesis was correct, the microsatellite could, in theory, obtain uncontrolled growth and finally strike a large region on the chromosome. Nevertheless, microsatellites seems to have an upper size limit that rarely exceed 50 repeated units 31. Still, some of the HrU10 alleles uncovered in our study include almost 100 pentanucleotide units.

It has been suggested that short microsatellites tend to gain additional units whereas long microsatellites are more likely to lose units during a mutation event 3947. However, the support for this hypothesis is ambiguous. Primmer et al. 11, Eckert et al. 13 and Vigouroux et al. 48 have reported an overall directionality bias of slippage leading to expansion. Our data showed a similar tendency for barn swallow, though not statistically significant. Xu et al. 49 showed a constant slippage rate for expansion and an increasing slippage rate for contraction with increasing allele size, which is consistent with a general observation of an upper size limit for repeated regions. However, our data on the HrU10 locus do not support that slippage directionality is length dependent, neither for total length of perfectly repeated core units nor for the total length of the entire microsatellite region. This result is in agreement with earlier studies for the HrU10 locus 30.

Weber and Wong 50; Garza et al. 51; and Primmer et al. 52 have put forward hypotheses for the upper size limit for microsatellites. They postulate that the contraction of microsatellites is caused by large deletions that occur when a microsatellite reaches its maximal length potential. An example of such large deletion has been reported by Colson and Goldstein 53 who reported one incidence of an absent microsatellite in one allele of the U1951 locus in Drosophila melanogaster. Other examples include a 27 units contraction of a dinucleotide microsatellite in Ranunculus carpaticola 46, and a 18 units contraction in a tetranucleotid repeat in superb fairy-wrens (Malurus cyaneus) 43. It has also been suggested that if the balance between slippage and point mutations favours point mutations within the repeated region, the mutations may interrupt the feature of the microsatellite without enhancing large contractions 54, and eventually give rise to new diversity. Kruglyak et al. 55 developed a Markov chain model which confirmed that infinite microsatellite growth can be disabled by introductions of point mutations. In the sequence data presented here, all sequences contain IMs. These must have been introduced by other mutation mechanisms than regular slippage of entire units.

Because of the higher number of mitotic cell divisions in male than in female germ lines, it is plausible to expect that evolution of microsatellites, to some extent, is male-driven 1415. However, we observed a bias for mutations at the HrU10 locus to be maternally transmitted in both barn swallows and tree swallows. Similar results have been reported by Brohede et al. 30 who observed a 2.5–5 fold increase in slippage rates in several hypermutable markers in females barn swallows compared to males, including the HrU10 microsatellite. Beck et al. 43 also uncovered a bias favouring maternally transmitted slippages for one locus in the superb fairy-wren (Malurus cyaneus).

More than 50% of the HrU10 mutations sequenced in this study resulted in a mutant allele size homoplasic to another sequenced allele in the dataset, and 23% of the mutations resulted in alleles homoplasic in both allele length and sequence (identical alleles). Such homoplasy is only detectable through sequencing of observed mutations in known pedigrees and has to our knowledge not earlier been confirmed by empirical data. Size homoplasy may be problematic according to Estoup et al. 22 in instances with "(1) high mutation rates and (2) large population sizes together with (3) strong allele size constraints". The mutation rates observed in this study (1.97 × 10^-2 per meiosis in barn swallow, 3.02 × 10^-2 per meiosis in tree swallow) are among the highest ever reported for microsatellites, and our estimate is also concordant with that provided by Brohede et al. 30 for the same locus. A total number of 2070 meiotic events for the barn swallow population presented here is the largest data set ever reported for an avian pentanucleotide microsatellite. Although it seems unlikely that there is a strong allele size constraint for the HrU10 locus, our empirical results confirm that most of the observed mutations resulted in an electromorph already present among the 66 sequenced alleles. In consequence, homozygous individuals in terms of fragment length analyses are indeed not necessarily homozygous in terms of nucleotide sequences. Discrepancies between allelic variation detectable through fragment length analyses and sequence analyses have also been reported in other microsatellite studies 35356. Estoup et al. 22 approached the issue of size homoplasy theoretically based on frequently applied mutations models such as e.g. the SMM and the KAM. The proposed index of size homoplasy can be explained as the probability of two electromorphic alleles not being of common descent. However, these indexes have certain limitations when applied to the data on the HrU10 locus provided here. First, the proposed homoplasy estimates relates to length and not sequence. Accordingly, there is no parameter that distinguishes between the two types of homoplasy. Second, the homoplasy index for the SMM does not include a parameter for the number of allelic states in a population, which is certainly crucial for homoplasy estimates. Third, the KAM accounts for allelic states, but assumes equal probability to mutate towards any of the other K-1 alleles. This is certainly not the case for the HrU10 locus in barn swallows.

Many studies have focused on the occurrence of microsatellite size homoplasy within different taxa (e.g: humans (Homo sapiens) and chimpanzees (Pan troglodytes) 57, mammalian carnivores 5859, birds 6061, salmonids 3, pipefish (Syngnathus typhle) 62, crabs (Limulus polyphemus) 63, two bee species and the fresh water snail Bulinus truncates 64 and fruit flies (Drosophila) 65). These examples support the notion that caution must be taken when microsatellite data are collected by electromorphic genotyping only. The high rate of mutations leading to size homoplasy in the present study provides support for alleles of identical size being attributed to common descent and hence causing bias in population genetic estimates.

Estoup et al66 and Estoup and Cournet 67 suggested that the amount of size homoplasy is more important in interspecific than in intraspecific comparisons. Nevertheless, our results for HrU10 provide evidence that homoplasy may play an important role also within populations. This conclusion is in agreement with the results published by van Oppen et al. 68, who found equally high amounts of homoplasy when comparing individuals among closely related taxa and among more distantly related species. HrU10 is a frequently used marker for different genetic analyses of bird populations, especially parentage studies 3369707172 because of the high allelic diversity enabling a powerful marker for parentage testing. However, our study indicates that HrU10 should be used with caution whenever homoplasy may cause biased estimates of relatedness and genetic diversity.