Increased Instability of Human CTG Repeat Tracts on Yeast Artificial Chromosomes during Gametogenesis

Abstract

Expansion of trinucleotide repeat tracts has been shown to be associated with numerous human diseases. The mechanism and timing of the expansion events are poorly understood, however. We show that CTG repeats, associated with the human DMPK gene and implanted in two homologous yeast artificial chromosomes (YACs), are very unstable. The instability is 6 to 10 times more pronounced in meiosis than during mitotic division. The influence of meiosis on instability is 4.4 times greater when the second YAC with a repeat tract is not present. Most of the changes we observed in trinucleotide repeat tracts are large contractions of 21 to 50 repeats. The orientation of the insert with the repeats has no effect on the frequency and distribution of the contractions. In our experiments, expansions were found almost exclusively during gametogenesis. Genetic analysis of segregating markers among meiotic progeny excluded unequal crossover as the mechanism for instability. These unique patterns have novel implications for possible mechanisms of repeat instability.

More than 14 sites in the human genome have been found to include unstable trinucleotide repeats (1). This instability manifests itself in changes in the number of repeats between successive generations and in changes during the life span of a single person (25). In more than 10 of these sites, the increase in the number of repeats causes severe diseases, such as fragile-X syndrome, myotonic dystrophy, and Huntington disease, etc. As the number of repeats increases, disease onset becomes earlier and the severity of disease increases. This phenomenon is also known as “anticipation.” Two basic models have been employed to explain trinucleotide repeat instability (11). The first associates the instability with recombination, and the second associates it with the replication process. Recombination models explain the increase in the number of repeats as unequal crossover or gene conversion between homologous chromosomes or sister chromatids. These models are supported by the following findings. In fragile-X disease, the transition from premutation (54 to 200 repeats) to mutation (more than 200 repeats) always occurs while the X chromosome is transmitted through the mother. As the female has two X chromosomes, whereas the male has only one, this suggests that the mechanism could be unequal crossover between the X chromosomes. Furthermore, in male meiosis, the X chromosome is inactivated (16), hence even recombination between the sister chromatids may be inhibited on this chromosome (but may occur in female meiosis). Moreover, recombination seems to be the source of instability of the CEB1 minisatellites in humans (2) and the ribosomal DNA genes (15) and CUP1 repeats (23) in the yeast Saccharomyces cerevisiae. The alternative, slippage-during-replication model (11) suggests that during replication the template and the new strand dissociate from each other. One of the DNA strands creates a new structure, for example, a hairpin, which results in contraction or expansion in the next generation, depending on which strand created the hairpin. This model predicts that the same tract in opposite orientations at the same site should have different probabilities of changing. The prediction has been verified to some extent (4, 10). Both the recombination and slippage models are compatible with recent findings (5, 19) that in rad27-deleted strains of S. cerevisiae there are high levels of trinucleotide repeat tract instabilities.

Another important issue is whether expansion in the number of trinucleotide repeats occurs during meiosis or mitotic divisions. Although it is known that both somatic and germ line tissues show instability (14), it is not known at what stage the changes occur, whether in gametogenesis or in the first mitotic divisions in the embryo (9).

To illuminate the mechanism and timing of trinucleotide instability, we inserted a 1.4-kb fragment from the DMPK gene, associated with myotonic dystrophy, into the same position in two differentially marked copies of the experimental yeast artificial chromosome (YAC) YAC12 (20). The instability of these repeat tracts was examined both in meiosis and in mitotic cell divisions in strains with one or two YACs.

MATERIALS AND METHODS

Media and genetic methods.

Standard yeast media were used (17). Sporulation was performed as previously described (20). Tetrads were dissected on yeast extract-peptone-dextrose plates, and the plates were incubated at 30°C to allow spores to germinate and form spore colonies. The spore colonies were replicated onto a series of plates lacking uracil, adenine, histidine, lysine, tryptophan, or leucine and incubated at 30°C to determine the segregation of genetic markers. Spore colonies that contained the markers of a YAC were analyzed by PCR to determine the lengths of CTG tracts.

YAC constructions.

The YACs we used were versions of YAC12 (20), with either URA3 near the centromere and HIS3 at the end of the long arm or LYS2 near the centromere and TRP1 at the end of the long arm. The YACs had an insertion of the ADE2 gene in a position 225 kb from their centromere (20), which was used to direct integration of the sequences containing the CTG repeat tracts. A 1.4-kb SmaI-HindIII fragment (kindly provided by the Norman Arnheim laboratory) of the DMPK gene containing (CTG)₆₀ (60 repeats) was integrated into the HpaI site of the implantation vector pGS534 (20) in both orientations. The resulting two plasmids, pDS38 and pDS39, were digested with restriction enzymes NotI and ClaI, and the 2.8-kb fragments containing the inserts as well as their bracketing sequences were isolated and used for directed integration into YAC12 by lithium acetate transformation (17). Transformed cells were plated on a medium that selected for the YAC (e.g., was devoid of uracil and histidine) and contained a limited amount of adenine (5 μg/ml), to allow colonies which lost ADE2 to develop a red pigment (20). Red colonies were isolated and were checked by pulsed-field gel electrophoresis to establish that the YACs had maintained their original lengths and by Southern analysis to confirm that integration of the CTG repeat tract had occurred at the desired site on the YAC. We thus obtained the following three versions of YAC12: YAC12A, which is marked with LYS2-TRP1 at its ends and contains a 1.4-kb SmaI-HindIII DMPK fragment with (CTG)₅₄ repeats at a site 225 kb from the centromere; YAC12B, marked with URA3-HIS3 and containing the inserted 1.4-kb SmaI-HindIII DMPK fragment with (CTG)₆₁ repeats at the same location and in the same orientation as the insert in YAC12A; and YAC12C, marked with URA3-HIS3 and containing the 1.4-kb SmaI-HindIII DMPK fragment in the orientation opposite that of the inserts in YAC12A and YAC12B, with a (CTG)₅₇ repeat tract.

Yeast strains.

The S. cerevisiae diploid strains used in this study are isogenic except for the repeats and the markers on the ends of the YACs; they were derived from strains described previously (20) and are of the following genotype: MATα/MATa ura3-52/ura3-52 ade2-101/ade2-101 trp1Δ1/trp1Δ1 lys2-801/lys2-801 his3Δ200/his3Δ200 leu2Δ1/leu2Δ1 CEN6/ΔCEN6::LEU2-CEN11. Strain yCH380 was generated by mating the haploid strain yDS366 (MATα, YAC12A) with the haploid strain yDS358 (MATa, YAC12B). yCH278 was generated by mating the haploid strain yDS358 (MATa, YAC12B) with the haploid strain yPH858 (MATα, no YAC). yCH386 was generated by mating the haploid strain yDS371 (MATa, YAC12C) with the haploid strain yPH857 (MATα, no YAC).

PCR analysis of single colonies.

Isolated colonies were removed from the agar plate and suspended in 100 μl of sterile water, heated to 96°C for 5 min, and chilled on ice for 5 min. Ten microliters was used as a template. The sizes of the repeats were determined by radioactive PCR analysis with the following oligonucleotides as primers: MDK409 (GAAGGGTCCTTGTAGCCGGGAA) and MDK410 (AGAAAGAAATGGTCTGTGATCCC). The PCR mixture included 0.2 μl of KlenTaq enzyme and 0.2 μl of [α-³²P]dCTP. The reaction mixtures were cycled 30 times at 94°C for 1 min, 52°C for 1 min, and 72°C for 1 min. The amplified products were analyzed on a 6% polyacrylamide gel. PCR product sizes were determined by comparison with an M13 sequence.

RESULTS

Higher instability in meiosis than in mitotic divisions.

Normally, trinucleotide repeat tracts in the human genome are present at corresponding, parallel sites on a given pair of homologs. Nevertheless, recent studies of trinucleotide repeat instability in model systems, whether in Escherichia coli (18) or in budding yeast (4, 5, 7, 10, 19), have employed a single copy of an inserted repeat tract somewhere in the host genome. In mice, single copies (3, 8, 13) or several copies in tandem (3, 13) were inserted. To model the natural situation more closely, we constructed diploid yeast strains with two homologous human DNA YACs, both copies of YAC12 (20) with short inserts containing trinucleotide repeats. The two YACs are identical except for the markers at their ends, and both YACs have autonomously replicating sequences (ARS) near both telomeres. PCR was used to monitor changes in the number of repeats. The tests employed primers which bracketed the CTG repeat tract in the DMPK gene and gave PCR products of 66 to 234 bp, depending on the sizes of the repeat tracts (Fig. 1).

FIG. 1.

PCR analysis of numbers of human CTG repeats on YACs. PCR was performed on colonies that germinated from single spores. [α-³²P]dCTP was included in the amplification reaction. (N) Normal tetrad. Two spore colonies have tracts of 54 repeats (a and d), and two have tracts of 61 repeats (b and c). (1) Type A tetrad. Two spore colonies have the original tract of 61 repeats (a and d), and two have a new tract length of 24 repeats (b and c). (2) Type B tetrad. All four spore colonies show the original tract size (54 or 61 repeats), and one of the colonies (a) also has an additional, new tract size (61 + 11 repeats). (3) Type C tetrad. Three spore colonies show the original tract sizes of 54 repeats (a) and 61 repeats (b and c), and the fourth shows a new size of repeat tract, 46 repeats (d). (M-13) Sequence of M13 phage.

PCR tests were performed on DNA of 224 mitotic diploid colonies obtained from streaking six different colonies from strain yCH380, which contained two copies of YAC12, one with 54 CTG repeats and the markers URA3 and HIS3 (YAC12A) and the other with 61 repeats and the markers LYS2 and TRP1 (YAC12B). The strain was shown to carry only one copy of each of these two YACs by segregation of their markers among the progeny of 270 dissected tetrads. PCR tests revealed contractions in the sizes of the tracts in 41 mitotic colonies (18.3%). These contractions could be assigned to 39 different types, based on repeat size, indicating that they represent independent events (and not a few jackpots of early events that were propagated in the cell population from which the mitotic colonies were derived). There were no expansions of repeat tracts in this sample.

The contractions were assigned to two groups, according to their distribution among the cell population of a given colony (Table 1). Type 1 colonies contained two YACs, one with the original-size trinucleotide repeat tract and one with a contracted size, for example, YACs with 61 and 32 repeats. Contraction events leading to type 1 colonies, where all cells of the colony carry the contracted repeat tract on one YAC and an original tract on the other YAC, have probably occurred in a cell division cycle before the colony was established. Six colonies contained two YACs with the same contracted repeat sizes. We consider these colonies to be type 1, as this pattern could result from a type 1 event followed by mitotic gene conversion. Type 2 colonies gave three CTG tract sizes in PCR tests, the two original tract sizes and a new, contracted size, for instance 61, 54, and 32 repeats. Based on the intensity of the bands representing the three tract sizes in the PCR tests and streaking showing individual colonies with two tract sizes only (e.g., either 61 and 54 or 61 and 32), one can determine which of the two original YACs has undergone contraction. Type 2 contraction events have probably occurred during one of the cell division cycles following establishment of the colonies on which the PCR tests were performed.

TABLE 1.

Types of instability

Cell sample	Type	No. of repeats
Original diploid		61; 54
Mitotic colonies (diploid)	Normal	61; 54
	Type 1	61; 32
	Type 2	61; 54, 32
Spore colonies from dissected tetrads (haploid)	Normal tetrad	61; 61; 54; 54
	Type A tetrad	61; 61; 24; 24
	Type B tetrad	61; 61; 54; 54, 11
	Type C tetrad	61; 61; 54; 46

The sizes of contracted repeat tracts were determined by comparing the sizes of the PCR products on sequencing gels with an M13 sequence (Fig. 1). The sizes of CTG tracts after contraction ranged between 5 and 48 repeats; 68% of the contractions involved deletions of 21 to 50 repeats (Table 2).

TABLE 2.

Distribution of contraction sizes of CTG repeat tracts

Size of deletion (no. of repeats)	No. of contractions
	yCH380 [(CTG)54/(CTG)61]a		yCH278 [(CTG)61]b		yCH345 [(CTG)52/(CTG)0]c		yCH386 [(CTG)57]d
	Mitotic colonies	Tetrads	Mitotic colonies	Tetrads	Mitotic colonies	Tetrads	Mitotic colonies
1–10	2	2	1	3	2	2	1
11–20	9	7	1	1	5	16	0
21–30	9	14	7	6	7	15	3
31–40	13	10	4	11	8	6	7
41–50	6	17	5	9	1	0	7
51–61	2	3	1	0	0	0	0
Total	41	53	19	30	23	39	18

^aSample size: 224 colonies, 187 tetrads. 

^bSample size: 103 colonies, 73 tetrads. 

^cSample size: 80 colonies, 76 tetrads. 

^dSample size: 116 colonies. 

To check trinucleotide repeat instability in meiosis, we determined repeat lengths in all spore colonies from 187 dissected tetrads. The spore colonies were haploid and normally contained a single YAC per cell. Fifty-three tetrads showed contractions in the sizes of repeats in one or two of the spore colonies, and five showed expansions. As in the mitotic cells, the sizes of new repeat tracts resulting from contraction ranged from 5 to 55 repeats. Based on their repeat lengths, the contractions could be assigned to 47 different groups, suggesting that almost all were the result of independent events. Seventy-seven percent of the contractions were deletions of between 21 and 50 repeats (Table 2). The five expansion events increased by 18, 15, 10, 10, and 5 repeats (to tract sizes of 72, 76, 71, 64, and 66 repeats, respectively).

Among the dissected tetrads, we classified the contraction or expansion events into three types (Fig. 1 and Table 1). Type A tetrads had two spores with one of the original sizes and two spores with the same contracted size or four spores with the same new contracted size, for example, 61, 24, 24, and 61 repeats. Type A tetrads resulted from events that occurred before meiosis (premeiotic events) or in G₁ of meiosis. Type B tetrads had four spores with the original sizes, but one also included a new, different size, for example, 61 + 11, 61, 54, and 54 repeats (the first spore colony was a mixture of cells with either 61 or 11 CTG repeats). Type B tetrads represented changes that had occurred after meiosis (postmeiotic events). Type C tetrads had three spores with the original sizes and a fourth with a new size, for example, 54, 61, 61, and 46 repeats. Type C tetrads represented changes that had occurred during meiosis at the four-strand stage. We believe that the three types of changes, especially types A and C, occur at different stages of sporogenesis which are parallel to the major stages of gametogenesis. Table 3 shows the distributions of instability types arising from meiosis as well as those found among the “mitotic” colonies arising from plating of single cells. In Tables 1 and 2, types A and B among the meiotic tetrads correspond to types 1 and 2 of the mitotic colonies. Individual cells undergoing meiosis correspond to the founder cells that started the mitotic colonies.

TABLE 3.

Changes in repeat tracts among mitotic colonies and tetrads

Strain (cell division process)	Repeat tracta	Sample size	No. (%) of changesb			Total no. (%) of changes
			Type 1 or A	Type 2 or B	Type C
yCH380 (mitosis)	(CTG)61/(CTG)54	224 colonies	30 (13.4)	11 (4.9)		41 (18.3)
yCH380 (meiosis)	(CTG)61/(CTG)54	187 tetrads	30 (16)	20 (10.7)	8 (4.3)	58 (31)
yCH278 (mitosis)	(CTG)61	103 colonies	10 (9.7)	9 (8.7)		19 (18.4)
yCH278 (meiosis)	(CTG)61	73 tetrads	8 (10.9)	17 (23.3)	6 (8.2)	31 (42.4)
yCH345 (mitosis)	(CTG)52/(CTG)0	80 colonies	12 (15)	12 (15)		24 (30)
yCH345 (meiosis)	(CTG)52/(CTG)0	76 tetrads	11 (14.5)	20 (26.3)	12 (15.8)	43 (56.6)
yCH386 (mitosis)	(CTG)57	116 colonies	7 (6)	11 (9.5)		18 (15.5)

^aTwo repeat tracts represent two YACs, each with one tract of repeats, whereas one repeat tract represents a single YAC. 

^bTypes of changes are described in the text and in Table 1. 

From type C tetrads we can calculate directly the meiotic rate of change in the number of repeats, T_c, as we know the number of cells in our sample that have undergone meiosis (187 cells, leading to 187 tetrads from strain yCH380). Thus, for yCH380, T_c is 4.3% (8 of 187 tetrads). This is similar to the rate of instability of (CAG)_28–31 repeats in the human androgen receptor gene among single sperm (26). Both cases represent repeat instability during meiosis.

For the other types of tetrads we need to calculate indirectly the rate of repeat tract instability. The rate of repeat changes of type A can be estimated from the frequencies in Table 3. Sixteen percent of tetrads of strain yCH380 were of type A. These events occurred before the cells entered meiosis (or before premeiotic DNA replication). The cells that were exposed to meiosis-inducing conditions (nitrogen and glucose starvation on sporulation plates) were taken from a single colony, of approximately 10⁷ cells, originating from a single cell that contained the two homologous YACs with the two original trinucleotide repeat tracts, (CTG)₅₄ and (CTG)₆₁. Thus, at least 23 mitotic generations led to the meiotic cell population of 10⁷ cells. A type A event could have occurred with probability T_a at any of these 23 preceding generations, and the expectation of type 1 not having occurred is (1 − T_a)²³. Since 84% of tetrads did not have type 1 events, we calculate that (1 − T_a)²³ is 0.84 and T_a is 0.0076, or 0.76% per cell generation. Thus the rate of type A events, T_a, is approximately one-sixth of the rate of meiotic type C events, T_c (4.3%). The frequency of type 1 events in the mitotic cell population of yCH380 was 13.4%, compared with 16% in the meiotic type A population. The frequency of type A contractions in the meiotic population, excluding four cases of expansion, was 13.9%. This suggests that T_a may be slightly higher during divisions on sporulation plates than in “normal” mitotic cells growing on vegetative medium.

Type B events are represented by spore colonies that contain cells with original repeat tracts as well as cells with altered (contracted) tracts. These alterations must have occurred during growth of the colonies. Among the 187 dissected tetrads of yCH380 there were 20 cases of type B instability; therefore, the frequency of these events, calculated per tetrad, is 10.7% for yCH380 (Table 3). But if these were mitotic events that occurred following germination, during postmeiotic growth of the spore colonies, they should have been calculated per spore colony to give a frequency of 20 events out of 748, or 2.7%. For the mitotic colonies (Table 3), 4.9% of events were of type 2, which corresponds to type B among the tetrads. As the mitotic type 2 events could have occurred for either of the two YACs carried by diploid yCH380, it is not surprising that the frequency of type 2 events was twice as high as that found for the haploid spore colonies that carried only one YAC each. Postmeiotic segregation (PMS) of events that actually occurred during meiosis and were not repaired could be expected to give rise to spore colonies with a mixture of two sizes of repeat tract, the original and a new tract size. Such a spore colony would be regarded as belonging to type B tetrads, although it resulted from a meiotic (and not a postmeiotic) event. However, as the corrected frequency of type B events (per spore colony, per YAC) is only slightly higher for spore colonies than for mitotic colonies of yCH380 (2.7% versus 2.45%), we see here no compelling evidence of PMS.

Expansions were found only among the dissected tetrads of yCH380. Four cases of expansion were found in type A tetrads, and one was found in a type C tetrad. No expansions were found in the “mitotic” colonies or among the 20 type B tetrads. Type C expansion must have occurred during meiosis. The four type A events may have occurred during late premeiotic divisions or in G₁ of meiosis. However, if the type A expansion events had occurred in the earlier “mitotic” divisions preceding meiosis, we might have expected to see expansion events in our mitotic experiments. We therefore suggest that type A expansions are likely to have occurred during the last cell division cycle before meiosis or in premeiotic G₁, possibly already on the sporulation plate.

A CTG repeat tract on a single YAC is more unstable than repeat tracts on two homologous YACs.

To check whether the higher instability in meiosis of trinucleotide repeats is due to recombination between the two homologous YACs, we also examined this instability in an isogenic strain, yCH278, containing a single YAC with 61 repeats. In this strain, recombination could occur only between repeats on sister chromatids or within the same chromatid. Of 103 mitotic colonies that were checked by PCR, 19 contained contracted repeat tracts. Seventy-three tetrads were also dissected from this strain. For each tetrad, the two spores with YACs were checked by PCR. Thirty tetrads showed a contraction, and only one tetrad showed an expansion, which occurred during meiosis (i.e., type C), to 72 repeats. Table 3 contains the results. As for mitotic and meiotic cells of strain yCH380, more than 80% of the contracted sizes were between 21 and 50 repeats (Table 2).

The rate of repeat instability during meiosis for strain yCH278 was 8.2%. If we compare the estimated rate of instability before meiosis (calculated from type A, as shown above for yCH380) to the value obtained during meiosis (type C), we find again that the former is 10 times smaller than the latter. The most surprising result is the high frequency of “postmeiotic” instability events (type B) among tetrads of strain yCH278 (23.3%). This value is 2.2 times higher than the comparable frequency among tetrads of yCH380. This actually represents a 4.4-fold-higher value, since in strain yCH380 such events could occur on either of the two YACs. This difference between yCH380 and yCH278 in the frequency of type B events is statistically significant (χ² = 5.8 [1 degree of freedom]). One explanation for the higher frequency of type B events in tetrads of yCH278 is the absence of a second YAC carrying the repeats. A second copy of repeats could have provided a template for repair of an instability event. Indirect evidence for such a mechanism are two tetrads from yCH380 with 61, 54, 12, and 12 and 61, 54, 5, and 5 repeats. In these cases, which were counted as type C, there appear to have been interactions between the two homolog during the instability event, because the two original-size repeats changed to the same contracted size. Such repair is expected to be more efficient during meiosis, and the high level of type B events could be accounted for by meiotic events related to PMS.

To find out whether the higher frequency of contraction events in the strain with a single YAC is due to the absence of a second YAC or whether it results from the absence of repeat sequences on the homologous chromosome (YAC), we constructed strain yCH345. This strain has two homologous YACs (both YAC12s but with different genetic markers), one of which contains an insert of the DMPK gene with a (CTG)₅₂ tract and the other without an insert. Basically, the frequencies of contractions were the same as in strain yCH278 (Table 3), which contains only a single YAC. Of the 43 tetrads that contained changes in repeat tract length (out of 76 tetrads), four were expansions. In two cases the expansions were in type A tetrads, to 53 and 54 repeat tracts, and two expansions were in type C tetrads, both to 59 repeats. The 80 mitotic colonies of yCH345 contained 23 colonies that underwent contraction of the repeat tract, and one showed a large expansion, to 78 repeats (type 1). This was the only expansion we observed among mitotically growing colonies.

The overall similarities between the results obtained for strains yCH278 and yCH345 suggest that the lower frequency of type B contractions in tetrads of strain yCH380 is probably due to the presence of the CTG repeat tract on the homologous YAC.

No effect of orientation on repeat tract instability.

Previous studies have shown that trinucleotide repeat instability depends on the orientation of repeats in relation to the replication fork (4, 10, 12). We therefore examined instability in the isogenic diploid strain yCH386, which carries a single YAC, into which the same fragment as in yCH278, but with 57 CTG repeats, was inserted at the same site but in the opposite orientation relative to the flanking YAC sequences. This was verified by Southern analysis. We examined by PCR the DNA of 116 vegetatively growing colonies derived from strain yCH386 and found 18 contractions. Surprisingly, both the frequencies (Table 3) and distributions (Table 2) of the contractions were not different from those of yCH278. Thus, the orientation of the repeat tract did not alter its stability in mitosis.

DISCUSSION

To answer questions of the mechanism and timing of trinucleotide repeat expansion, we used implanted copies of repeats on YACs and compared the rates of repeat instability during meiosis and mitosis. We also examined the influence on instability of the presence of a second copy of the repeat tract, on a homologous chromosome (a YAC). CTG repeats were inserted into the YACs and showed a high level of trinucleotide repeat instability. Most of the changes in tract length were contractions, and the very few expansions that were observed were all in tetrads. Other investigators (4, 10, 12) have also shown that most of the changes in CTG repeat tracts in mitotically growing cells of S. cerevisiae are contractions. A few expansions were found in two of these studies (4, 10). Our results differ from those of previous studies in two respects. First, when we inverted the orientation of the insert with the CTG repeats relative to the flanking YAC sequences, the instability of the repeat tract did not change. In contrast, Freudenreich et al. (4) and Maurer et al. (10) did find differences in tract instability between the two orientations, a so-called “orientation effect,” in samples considerably smaller than ours. Secondly, unlike previous studies, we have also examined spore colonies derived from dissected tetrads, found among them a much higher instability than in mitotically derived colonies, and were able to analyze the origin of instability by genetic analysis of the tetrads.

Our findings about the instability of CTG repeat tracts may be explained by one of two mechanisms, namely, modified replication slippage (Fig. 2a) or intramolecular recombination (Fig. 2b). As mentioned above, most of the changes in tract length were contractions, and the very few expansions that we observed were all in tetrads. A recombinational mechanism would be compatible with these data if the recombination events usually occurred between repeats on the same chromatid. Unequal crossover between chromatids can be ruled out, because in no case did we obtain the expected two reciprocal products together (contraction and a corresponding expansion in the same tetrad or colony) or evidence of mitotic reciprocal recombination of outside markers (homozygosis of the telomeric marker) in type 1 colonies or among type A tetrads.

FIG. 2.

Proposed models for contraction or expansion of trinucleotide repeat tracts. (a) During replication, the strands dissociate and a hairpin is created on one of the strands. If the hairpin is on the template strand, the new strand contracts to a new tract size that is equal to the original size minus the size of the hairpin. If the hairpin is on the new strand, the repeat tract expands to a new size that is equal to the original size plus the size of the hairpin. (b) Recombination, e.g., between two loops on a nucleosome, creates a new, contracted tract size that equals the original size minus the number of triplets between the loops.

More than 75% of the contraction events deleted 21 to 50 triplets, with a conspicuous dearth of contractions of small sizes. This dearth of small contractions could not be explained by PCR selectively amplifying small repeat tracts (large contractions) rather than large tracts (small contractions), since there was no problem with amplification of the original-size tracts. In contrast, small contractions were abundant in the case of dinucleotide repeats (21). If deletions of trinucleotide repeats resulted from intramolecular recombination (Fig. 2b) and there were restrictions on the minimal size of the loop during the process, for instance due to DNA rigidity, one could expect a spectrum of large contractions, as we have found here. Furthermore, it is possible that the association of DNA with the nucleosome during intramolecular recombination (Fig. 2b) and the nucleosome size determine the size of deletions. There is evidence that CTG repeats are the strongest natural nucleosome positioning elements (22) and therefore that intramolecular recombination could indeed happen on nucleosomes wrapped by the repeat tracts. The range of contractions between 21 and 50 repeats corresponds to one to two loops around the nucleosome core. Expansions might also occur on the nucleosomes during recombination between sister or homologous chromatids, but here a minimum-size limit is not expected. Reciprocal recombination should lead to a corresponding contraction in another spore colony in the tetrad that contains an expanded repeat tract. This was not found in the four type C tetrads that contained expansions. This means that reciprocal, unequal crossover, the source of instability of minisatellites (15, 23), is not the mechanism for trinucleotide repeat expansion. Therefore, a nonreciprocal, gene conversion-like mechanism must be postulated for the expansions.

To explain the data with a replication slippage model requires certain modifications. One assumes a hairpin to be an intermediate in the slippage process, and due to the higher stability of longer hairpins, large contractions should be more abundant (6). The probability of creating and maintaining a given-size hairpin equals the stability (−ΔG) of this hairpin multiplied by the number of places where this hairpin can be formed in tracts of a given number of repeats. Thus, very long hairpins, although very stable, can be formed only rarely, whereas a hairpin of, say, 30 repeats can start at 31 different positions within the 61-repeat tract. Calculations based on these arguments may explain the distribution of contraction sizes.

In other studies (4, 10), the orientation of inserts containing CTG repeats had an effect on the frequency of instability events. It was suggested that a (CTG)_n hairpin is more stable than a (CAG)_n hairpin and that the lagging strand is more prone to slippage events than the leading one, due to differences in the flexibility of the strands (4, 10). Thus, a hairpin as an intermediate in replication slippage would suggest that when the repeat tract is inserted in the opposite orientation it alters contraction and expansion frequencies and size distributions. This we did not see (strain yCH386 [Tables 2 and 3]). The effect of the orientation of the repeats depends on the location of the nearest ARS (10). The absence of an orientation effect in our experiments could be reconciled with the slippage mechanism if we take into account that the CTG tract was not inserted in the YAC by itself but with the flanking 1.2-kb sequence, which could contain an ARS. However, we carefully examined the 1.2-kb sequence of the insert and did not find that it contained an ARS consensus sequence. From another study, we know that transcription through a repetitive dinucleotide tract destabilizes the tract (24). Here, we do not know whether the insert on the YAC in either orientation is transcribed.

Expansions of trinucleotide repeat tracts in humans have been suggested to occur in the parental germ line, or postzygotically, in early divisions of the embryo (9). In our system, most expansions occurred during sporogenesis, in meiosis, or in the preceding cell divisions, thus sharing many fundamental features with gametogenesis in higher eukaryotes. Moreover, it is clear that meiosis and its surrounding divisions show high rates of repeat instability (χ² values for the comparison of instability frequencies in tetrads versus mitotic colonies in strains yCH380, yCH278, and yCH345 were highly significant, at 10.9, 12.7, and 12.1, respectively). High rates of instability of trinucleotide repeats during gametogenesis would thus explain why the transition from premutation to mutation in humans requires passing between generations. A comparison of the distributions of contraction sizes of meiotic and mitotic cells in yCH380 (Table 2) suggests that the mechanism of contraction is probably the same in the two types of cell division, but the rates are different. We propose that two mechanisms could lead to the instability of trinucleotide repeat tracts in our experiments: replication slippage, with a secondary structure (hairpin) on one of the DNA strands as an intermediate, or intramolecular recombination between repeats along the tract length, which leads mostly to deletions of large sizes. Involvement of the repeat tracts in recombination could lead also to expansion events, although in these cases two chromatids need to be involved. Meiotic timing and a recombinational mechanism for trinucleotide repeat instability in yeast fit well together and suggest a similar association as a basis for trinucleotide repeat-related diseases in humans.