Orientation-dependent and sequence-specific expansions of CTG/CAG trinucleotide repeats in <em>Saccharomyces cerevisiae</em>

Abstract

Expansions of endogenous trinucleotide repeats (TNRs) underlie more than 10 human genetic disorders, including Huntington’s disease and myotonic dystrophy (reviewed in refs. 1 and 2). Several lines of evidence indicate that TNR alterations occur differently than for other microsatellites. For example, large increases in TNR number range from 10 to 2,000 repeats among affected kindreds (reviewed in refs. 1 and 2). Also, the propensity to undergo large TNR expansions is unaffected by the mismatch repair status of the cell (3, 4). These and other features suggest that TNR instability results from an unusual mechanism.

Although each triplet repeat disease exhibits unique genetic features (reviewed in ref. 1), there are several common aspects that provide clues to the mechanism of TNR instability. First is sequence specificity; all known TNR diseases result from expansions of the sequences CNG (where N is any nucleotide) or GAA. Second, a single genetic event is sufficient to incorporate many additional TNRs. Since the unaffected allele often is unaltered (5, 6), this suggests that repeats are added de novo by a replicational mechanism. Third, instability often is governed by a threshold number. TNR lengths below the threshold are stable whereas alleles above the threshold are prone to expansions (reviewed in ref. 1).

Examination of TNR structures suggests that DNA hairpins may play an important role in expansions. When single-stranded, a subset of TNRs forms secondary structures in vitro (7, 8). For example, 15-repeat molecules of CTG and CAG form hairpins that melt at 47° and 38°, respectively (summarized in ref. 8). Most other TNRs show no predisposition to expand and are unlikely to form stable secondary structures (7, 8). In a second corollary, duplex DNA containing CTG/CAG and CGG/CCG repeats form slipped-strand structures upon denaturation and reannealing (9), indicating that slipped structures can form in competition with duplexes. A third parallel is that hairpin formation in vitro is dictated by a threshold number of repeats. Based on modeling studies, a minimum number of repeats is important to stabilize the hairpin (7).

The noteworthy correlations between TNR genetics and formation of secondary structures such as hairpins led a number of authors to suggest that expansions result from aberrant lagging-strand DNA replication (2, 5, 10–14). As diagrammed in Fig. ​Fig.11A, lagging-strand DNA replication involves synthesis of ≈150- to 300-nt Okazaki fragments that subsequently are processed and ligated together. If TNR sequences such as CTG or CAG are present, hairpins could form on the lagging daughter strand. Extension of these hairpins by DNA synthesis would lead to the presence of extra triplet repeats. If unrepaired, the anomalous strand would template for an expanded second strand in the next round of DNA synthesis. The hairpin model for TNR expansions therefore predicts not only sequence specificity, with expansions arising most often from sequences that readily form hairpins, but also suggests possible orientation effects for complementary pairs such as CTG and CAG. In other words, the presence of the stronger, CTG hairpin-forming sequence on the lagging daughter strand should yield more expansions than the reciprocal case for a CAG tract.

Figure 1

Hairpin model for TNR expansions. The figure shows the possible behavior of CAG and CTG tracts in our experiments, assuming that (CTG)₂₅ repeats form stable hairpins in vivo more readily than (CAG)₂₅. In A–D, chromosomal DNA replication proceeds from left to right and the lagging strand synthesis is on top. The direction of the URA3 reporter (sense strand, 5′ → 3′) is indicated by the open arrow. In A, CTG sequences on the lagging daughter strand are predicted to form hairpins occasionally. Incorporation of the hairpin into the replicated product and failure to repair this structure ultimately would lead to an expanded allele. B shows an orientation effect, in which CAG sequences occupy the lagging daughter strand. If CAG sequences form less-stable hairpins than CTG, the CAG configuration will exhibit fewer expansions. In C and D, the entire reporter has been inverted to the opposite direction (depicted by the open arrow and the notations 3ARU GTC and 3ARU GAC, respectively). Inversion of the reporter is predicted to affect expansions because the sequences present on the lagging daughter strand will be altered. Inversion of the sequences in A (genetically unstable) results in the situation in D, which should yield fewer expansions. Similarly, inversion of the reporter from B (low rate of expansions) will yield the C scenario and should increase the rate of expansions.

Direct experimental proof for the hairpin model has been elusive. In transgenic mice, triplet repeats are quite stable (summarized in ref. 15). Among the alterations reported, all have been either losses or small gains. Thus, these animals have not revealed much about the mechanism of TNR expansions. In bacteria and yeast, contractions occur much more frequently than expansions (13, 16, 17). Assays that monitor TNRs by nonselective physical techniques (PCR or Southern blotting) tend to lack the sensitivity for facile detection of expansions. Sensitivity can be increased in yeast by the presence of a rad27 mutation that increases the frequency of expansions and contractions (18, 19). Under these conditions, expansions become abundant enough to detect by physical means. However, interpretation of results is complicated by the pleiotropic nature of rad27 mutations (20–22). In other experiments (19), CTG repeats were shown to act as fragile sites in yeast. Unfortunately, chromosome breakage also results in recombinational deletion of the TNR sequences, making it difficult to characterize the nature of expansions.

In this paper we describe a new genetic assay for TNR expansions in yeast that is selective and quantitative. Using this assay, we observe a number of characteristics for CTG and CAG expansions that are consistent with predictions of the hairpin model. To our knowledge, this is the most conclusive in vivo evidence supporting the hairpin-based, replicational model for CTG and CAG expansions.

(C,A,G)₂₅ refers to a random sequence of 25 C, A, and G residues with no repeat units, as described in Methods and Materials.

Chromosomal alignment refers to the direction of URA3 relative to the chromosomal sequences (Fig. ​(Fig.1).1). “Forward” indicates that the URA3 reporter and the endogenous LYS2 gene are aligned in the same direction (as in Fig. ​Fig.11A and B), and “reverse” indicates opposite directions (Fig. ​(Fig.11C and D).

Acknowledgments

ABBREVIATIONS