It is widely accepted that the large trinucleotide repeat expansions observed in many neurological diseases occur during replication. However, genetic recombination has emerged as a major source of instability for tandem repeats, including minisatellites, and recent studies raise the possibility that it may also be responsible for trinucleotide repeat expansions. We will review data connecting tandem repeat rearrangements and recombination in humans and in eukaryotic model organisms, and discuss the possible role of recombination in trinucleotide repeat expansions in human neurological disorders.
Two classes of tandem repeat sequences, minisatellites and microsatellites, have gained increasing attention from the scientific community over the past decade. Microsatellites are tandem arrays of short (usually <10 bp) units, while minisatellites are tandem arrays of longer units (>10 and <100 bp). These two kinds of sequence are widespread in all eukaryotic genomes (reviewed in Charlesworth et al., 1994) from yeast to mammals. A growing number of neurological disorders have been found to result from the expansion of a particular class of microsatellites, called trinucleotide repeats (reviewed in Cummings and Zoghbi, 2000).
In contrast to other forms of microsatellite instability such as that found in mismatch repair (MMR) mutants, trinucleotide repeat expansions observed in neurological diseases are often quite large (several times the initial number of repeats). The occurrence of these expansions during the lifetime is not precisely known; they might be meiotic, pre‐ or post‐zygotic. Trinucleotide repeats are not the only class of tandem repeats prone to expansions. Three minisatellites are known to share this property. Expansion of one of them is associated with a human disease (Lalioti et al., 1997). The others, two AT‐rich repeats, are responsible for fragile sites (Yu et al., 1997; Hewett et al., 1998).
Two types of mechanism have been proposed to be involved in tandem repeat instability: replication and recombination. The genome‐wide microsatellite instability observed in MMR mutant cells (Simpkins et al., 1999 and references therein) strongly supports a replicative mechanism, and the somatic instability of trinucleotide repeats is generally viewed as a consequence of replication (reviewed in McMurray, 1999; Richard et al., 1999a). However, several recent results argue that recombination might also play a role in microsatellite expansions. The aim of this paper is to review the data connecting tandem repeat rearrangements with recombination, and to discuss the possible involvement of recombination in trinucleotide repeat expansions.
Gene conversion as a major source of tandem repeat instability
Initially, when people were thinking of tandem repeat rearrangements in terms of recombination, they were thinking of crossovers (Figure 1), i.e. a reciprocal transfer of genetic information. However, gene conversion (non‐reciprocal transfer of information) has recently emerged as a major cause of tandem repeat instability. Gene conversion and crossover are often but not always associated, and a series of models, culminating in the model proposed by Szostak et al. (1983) (see Figure 2A), summarizes this relationship. However, it is mostly gene conversion without crossover that destabilizes tandem repeats.
Welch et al. (1990) were the first to establish a correlation between non‐crossover gene conversion and tandem repeat rearrangements, when they observed meiotic expansions and contractions of the yeast CUP1 locus. Tandem repeat rearrangements that were likely to have resulted from gene conversion were also observed in Drosophila (Pâques and Wegnez, 1993; Thompson‐Stewart et al., 1994; Delattre et al., 1995; Pâques et al., 1996) and provided the basis for a new model for tandem repeat rearrangements. These events were thought to result from the excision of a Drosophila P transposon. Such excisions are repaired by gene conversion, which is not usually associated with crossovers. A second series of models, termed SDSA, for synthesis‐dependent strand annealing (reviewed in Pâques and Haber, 1999), accounts for non‐crossover gene conversions. In SDSA models, the two newly synthesized strands unwind from their template and anneal with each other (Figure 2B). Pâques and Wegnez (1993) postulated that out‐of‐frame re‐annealing of tandem repeats would explain the observed rearrangements detected after excision of a P transposon (Figure 2C).
Meiotic recombination in human minisatellites
Since the discovery of the first minisatellites in the human genome (Jeffreys et al., 1985), mechanisms leading to high spontaneous mutation rates of these hypervariable DNA sequences have been thoroughly studied. Jeffreys et al. (1988) proposed that small mutations involving gain or loss of 4–10 repeats could occur by a mitotic replication slippage mechanism, while large mutation events involving gain or loss of up to 200 repeats were more compatible with a meiotic recombinational process.
In further analyses, minisatellite mutations found in the germline were precisely mapped (Buard and Vergnaud, 1994; Jeffreys et al., 1994), and typical features of the mutation process were observed. First, there was a polarity in the mutational process, i.e. repeats were gained or lost preferentially at one end of the repeat array. Secondly, complex rearrangement events suggested that the new mutant allele gained genetic information from both the sister chromatid (intra‐allelic event) and the homologous chromosome (inter‐allelic event). Thirdly, exchange of flanking markers was rare, indicating that recombination occurred with few associated crossovers.
Again, these results were explained in terms of SDSA (Buard and Jeffreys, 1997). This mechanism provided a simple explanation for the complex events observed: both templates were used by the two ends of a double‐strand break (DSB) to copy information, and the two newly synthesized strands were assembled in the recipient molecule by annealing. SDSA could also explain the low crossover rate (Figure 2B). Related SDSA models, where DNA synthesis occurred at only one end of the DSB, may also account for other features of minisatellite rearrangements observed in yeast or humans (Pâques et al., 1998; Debrauwère et al., 1999).
By mapping these rare crossovers in the vicinity of the MS32 minisatellite, Jeffreys et al. (1998) found a meiotic hot spot (a chromosomal region where meiotic recombination is frequently observed) centered ∼200 bp upstream of MS32 and extending into the beginning of the repeat array. Their work strongly suggests that hypervariability arises in proximity to a meiotic hot spot. Their finding also explains mutational polarity, since meiotic DSBs are initiated on one side of the minisatellite and recombination propagates within the repeat array from the initiation point. The gene conversion gradient decreases from the initiation point, as observed for meiotic hot spots in yeast (Nicolas et al., 1989).
Analysis of somatic mutations at MS32 revealed simple deletions or duplications, with no detectable bias toward one extremity of the repeat array (Jeffreys and Neumann, 1997). The authors thus concluded that somatic mutational processes involved in MS32 expansions and contractions were different from those found in the germline.
The introduction of two minisatellites, MS32 and CEB1, into yeast confirmed that they are more unstable during meiosis than during mitosis. Gains and losses of repeats were again complex events involving both inter‐ and intra‐allelic recombination (Appelgren et al., 1997; Debrauwère et al., 1999). In addition, it was shown that CEB1 meiotic instability depended on DSB formation at two well characterized meiotic hot spots, one of which lies upstream and the other downstream of the repeat array. Interestingly, instability disappeared when CEB1 was moved to a chromosomal region where no meiotic DSB could be detected.
Gene conversion‐induced rearrangements of tandem repeats in yeast
Recently, Pâques et al. (1998) designed an experimental system to study the effect of mitotic gene conversion on tandem repeat instability in Saccharomyces cerevisiae. A DSB was made on a yeast chromosome, which could be repaired using a homologous sequence containing a tandem repeat as donor (Figure 3). Both donor and recipient were recovered and analyzed for tandem repeat rearrangements.
Different kinds of tandem repeat were studied including 5S rRNA genes, trinucleotide repeats and minisatellites (Pâques et al., 1998; Richard et al., 1999b, 2000; F. Pâques, G.‐F. Richard and J.E. Haber, unpublished results), all of which rearranged with a high frequency during gene conversion (Table 1). Furthermore, rearrangements were observed mainly in the recipient molecule, whereas the donor was extremely stable, as predicted by SDSA models (Figure 2C). Interestingly, the CAG39 repeat (i.e. an array of 39 CAG triplets) yielded only contractions, whereas both contractions and expansions were observed with CAG98 (Table 1).
These results demonstrate the destabilizing effect of gene conversion on different kinds of tandem repeat. The model proposed in Figure 2C postulates that, during gene conversion‐associated DNA synthesis, slippage of the newly synthesized strands can occur, leading to addition or subtraction of a whole number of repeats. The frequency of such ‘repair‐slippage’ was estimated to be 800‐fold higher than that for S‐phase replication‐slippage (Richard et al., 1999b).
Is DSB‐repair involved in trinucleotide repeat expansions?
Trinucleotide repeats and some minisatellites are fragile sites in human (reviewed in Sutherland et al., 1998) and in yeast (Freudenreich et al., 1998). Fragile sites are loci at which chromosome breakage occurs in the presence of specific chemicals. We suggest that a mitotic or meiotic DSB could initiate the expansion process. Meiotic instability of CAG repeats in yeast has been correlated with formation of a nearby DSB in one study using long CAG tracts (Jankowski et al., 2000), but not in another using shorter CAG tracts (Moore et al., 1999). White et al. (1999) showed that CCG repeats did not exhibit a higher rate of instability during meiosis than during vegetative growth in yeast, whereas Cohen et al. (1999) found that instability of CTG repeats carried by YACs was higher during meiosis than during mitotic divisions. A very recent study showed that CA dinucleotide repeats were also destabilized during meiosis, in yeast, and that progression of strand exchange during gene conversion was inhibited by the microsatellite (Gendrel et al., 2000). In transgenic mice, CAG repeats exhibited a higher instability in the germline than in somatic cells, with a propensity towards expansion rather than contraction (Seznec et al., 2000).
In human diseases, it is common to observe expansions of more than twice the original size. Such large expansions in yeast and humans could occur by successive rounds of unwinding/re‐invasion of the donor sequence by the newly synthesized strand, allowing DNA synthesis to proceed more than once within the repeats (Pâques et al., 1998; Figure 2C′).
An interesting case of large contraction of a CAG repeat was described during transmission of a myotonic dystrophy allele. O'Hoy et al. (1993) reported a large reduction of the number of CAG triplets associated with what they called a ‘discontinuous gene conversion event’. The resulting allele was a patchwork of both maternal and paternal alleles. Buard and Vergnaud (1994), Debrauwère et al. (1999) and Tamaki et al. (1999) also found complex recombination events in minisatellites. In those cases, gene conversion was inter‐allelic (between homologous chromosomes), intra‐allelic (between sister chromatids) or a patchwork of both. All these observations are compatible with the unwinding/re‐invasion model (Figure 2C′).
Even though more conclusive evidence will be required to account for dramatic trinucleotide repeat expansion by a gene conversion mechanism in humans, recent data suggest that recombination can lead to large scale contractions and expansions within tandem repeat sequences at a high frequency. Experimental systems can be designed to study the effect of mitotic and meiotic recombination on tandem repeat instability. It will be particularly interesting to determine whether any kind of sequence can be expanded during gene conversion, or whether there is a sequence (or structure) specificity to this process. Moreover, even though some trans‐acting factors have been shown to be involved in expansions during gene conversion (Richard et al., 2000), their precise function needs to be determined. Hopefully, further experiments in model organisms will allow us to unravel the molecular basis of tandem repeat complex rearrangements during gene conversion, and perhaps help us to understand how dramatic trinucleotide repeat expansions occur in humans.
We would like to thank Allyson Holmes and members of the Unité de Génétique Moléculaire des Levures for insightful suggestions. G.‐F.R. is the recipient of a post‐doctoral fellowship from the Institut Pasteur.
- Copyright © 2000 European Molecular Biology Organization