Gene Conversion (Recombination) Mediates Expansions of CTG·CAG Repeats

Genetic recombination is a robust mechanism for expanding CTG·CAG triplet repeats involved in the etiology of hereditary neurological diseases (Jakupciak, J. P., and Wells, R. D. (1999) J. Biol. Chem. 274, 23468–23479). This two-plasmid recombination system in Escherichia coli with derivatives of pUC19 and pACYC184 was used to investigate the effect of triplet repeat orientation on recombination and extent of expansions; tracts of 36, 50, 80, and 36, 100, and 175 repeats in length, respectively, in all possible permutations of length and in both orientations (relative to the unidirectional replication origins) revealed little or no effect of orientation of expansions. The extent of expansions was generally severalfold the length of the progenitor tract and frequently exceeded the combined length of the two tracts in the cotransformed plasmids. Expansions were much more frequent than deletions. Repeat tracts bearing two G-to-A interruptions (polymorphisms) within either 171- or 219-base pair tracts substantially reduced the expansions compared with uninterrupted repeat tracts of similar lengths. Gene conversion, rather than crossing over, was the recombination mechanism. Prior studies showed that DNA replication, repair, and tandem duplication also mediated genetic instabilities of the triplet repeat sequence. However, gene conversion (recombinational repair) is by far the most powerful expansion mechanism. Thus, we propose that gene conversion is the likely expansion mechanism for myotonic dystrophy, spinocerebellar ataxia type 8, and fragile X syndrome.

Several human genetic studies on patient materials reported haplotype analyses, especially related to myotonic dystrophy (31)(32)(33)(34) and fragile X syndrome (35)(36)(37), which implicated recombination-mediated TRS instabilities. For these seven cases, it was presumed that the TRS were the sites for the recombination (or gene conversion) events. Substantial linkage analyses were performed utilizing flanking markers, some quite near the repeats, during the mapping of the genes for these neurological diseases (reviewed in Ref. 1). Because the exchange of flanking markers was generally not found (38,39), unequal crossing over as a general mechanism for expansion has not been favored. Alternatively, CAG⅐CTG repeats were reported to be hot spots for recombination (40), and recent experiments in yeast suggested that (CTG⅐CAG) (15) and (CGG⅐CCG) (19) repeats are susceptible to strand breaks. Yeast rad27 strains had augmented instability of the TRS. It is known that the majority of errors that accumulate in rad27 strains are processed via single strand annealing as well as double-stranded break repair (41), which proceeds via gene conversion (42,43).
To evaluate the potential role of genetic recombination (via gene conversion or unequal crossing over) as a mechanism for expansions of CTG⅐CAG repeats, we established (44) an in vivo recombination system in E. coli. The expansion events were dependent on the presence of long CTG⅐CAG sequences in the two-plasmid recombination system and required recombination-proficient cells to give frequent, severalfold expansions. Recombination was proven genetically and biochemically.
Thus, it was concluded that if these reactions occur in humans, recombination may contribute, along with replication and repair, to the expansions responsible for anticipation associated with the hereditary neurological syndromes. Herein, we describe studies on the molecular mechanism of the recombination process (gene conversion or unequal crossing over) and the lack of effect of TRS orientation on the multiple fold expansions.
pRW4436 and pRW4480, which harbor the inserts in orientation II, were constructed by digesting 5 g of pRW3036, which harbors (CTG⅐CAG) 36 , and pRW3080, which harbors (CTG⅐CAG) 80 , respectively, with NheI and XbaI (New England Biolabs Inc.). The digested DNAs were electrophoresed on a 6% polyacrylamide gel. The fragments containing the triplet repeats and the band containing the linearized vector were eluted from the gel and purified by phenol extraction. The inserts and vectors were ligated by treatment for 16 h at 16°C with one unit of T4 DNA ligase (U.S. Biochemical Corp.). The ligation mixture was precipitated and resuspended in 5 l of TE (45) and used to transform E. coli JC10289 via electroporation. Transformants were selected on LB agar plates and grown in 10-ml culture tubes at 37°C in the presence of ampicillin (75 g/ml). Plasmid DNAs were isolated and characterized by restriction mapping. The insert size and orientation were confirmed to be the expected size in orientation II.
pRW4437 and pRW4439 were deletion products of pRW3239 and pRW3238, respectively. They were obtained by repeated recultivation of E. coli AB1157 transformed with either pRW3239 or pRW3238. Initially, E. coli AB1157 was transformed with either pRW3239 or pRW3238, and then transformants were selected on LB plates and subsequently grown in LB medium containing tetracycline (20 g/ml) until the end of log phase (A 600 ϭ 0.9). The cells were harvested, and the DNA was isolated and purified. The DNA was electrophoresed on 1.0% agarose gels, and deletion products were identified. If no deletion products were detected, then 1 l of the culture was used to inoculate fresh medium, and the cells were grown again until late log phase. The cultures were kept growing logarithmically by diluting them 10 6 -fold with fresh, sterile medium. The recultivation was repeated until deletion products were observed. The deletion products were detected as faster migrating supercoiled DNA. Plasmids that contain deletions of TRS are smaller than their respective parent plasmids and thus migrate faster. The TRS in the pACYC184 derivatives were unusually stable and required five or more recultivations to generate a deletion product. The band that contained the deletion was excised from the gel, purified, and used to transform E. coli HB101 (2,3,5,6,11,12). Transformants were subsequently characterized by restriction mapping and DNA sequencing to confirm that the deletion occurred within the TRS.
Bacterial Strains-The genotypes of E. coli AB1157 and the recombination deficient strains JC10289, JC5519, and JC10287 used in the two-plasmid recombination system were described elsewhere (44). All strains were obtained from the E. coli Genetic Stock Center, Yale University, New Haven, CT.
Standard Genetic Techniques-Unless otherwise noted, DNA isolation and agarose gel and polyacrylamide gel electrophoreses were carried out according to standard laboratory protocols (45). Transformations were performed by electroporation (46,47). All plasmids were initially transformed into E. coli JC10289 or HB101 (which are recA Ϫ strains), cultured until mid-log phase, and purified by alkali lysis. The plasmids (see Fig. 1) were then used for cotransformation of the above strains in the two-plasmid recombination system. For the cotransformation experiments, each strain was cotransformed with a variety of plasmid combinations as shown in Table I. The experiments involving the cotransformation of both pUC19 and pACYC184 derivatives containing TRS in orientation I have been published (44) and additional results are listed in Table III. 40 l of washed cells of each strain (5 ϫ 10 7 cells/ml) were transformed with 1 l of the supercoiled DNAs (0.5 g/ml) listed in Fig. 1. For the experiments involving cotransformation, the supercoiled plasmid volume was equally divided between the two test plasmids. A voltage of 1950 V was delivered for 4.1-5.8 ms. The cuvette size was 0.2 mm.
Cotransformants were selected on LB agar plates containing ampicillin and tetracycline. After cotransformation, the cells were allowed to recover in 800 l of SOC medium and kept at 37°C for 1 h or longer. The cells were plated on LB agar that contained ampicillin (75 g/ml) and tetracycline (20 g/ml) and individual colonies were grown for 4 -16 h at 37°C under aerobic conditions. Plasmid purification, gel electrophoresis, and DNA sequence analyses were conducted as described (3,5,6,11,12). Restriction mapping reactions and ligase reactions were conducted as described (45). The uncertainty of the accurate measurement of repeat size varies with repeat length. The error for lengths greater than 130 repeats, between 65 and 130 repeats, and less than 65 repeats was 6, 4, and 2%, respectively. The plasmid inserts and flanking sequences were characterized by dideoxy sequencing on both strands with Sequenase (version 2.0). The pACYC184 primers, purchased from Genosys Inc. were the following: primer 4244 (ACGGTCTTTAAAAAGGCCG), which 3Ј-terminates at map position 95; primer 4245 (CGTCAGTAGCTGAACAGGAGGG), which 3Ј-terminates at map position 522.
Percentage of Gene Conversion versus Unequal Crossing Over-The products of cotransformation of E. coli AB1157 with pRW3036 and pRW3239 were analyzed by DNA sequencing. The pUC19 primer 1211 (New England Biolabs Inc.), which binds 21 bp upstream of the polylinker, was used in the Genosys sequenase 2.0 kit; the triplet repeats were cloned into the HincII site of the polylinker of pUC19. The recombination pathway was determined by analyses of the distal (3Ј of the TRS) flanking sequences. The recombination products were scored as gene conversion events if the TRS flanking sequences were both derived from pUC19. Alternatively, the recombination products were scored as unequal crossing over events if the distal flanking sequence was derived from pACYC184.
Restriction mapping was conducted to determine whether crossing over or gene conversion cause expansion(s). Certain pairs of restriction enzymes were used to analyze products that had expansion(s). Crossover events would generate products that contain two TRS tracts (see Fig. 3B in Ref. 44). One of the tracts would be detected by Bstz17I and NdeI. A second pair of enzymes, BspEI and PstI, would detect the other TRS tract. The products from these digestions were separated on 7% polyacrylamide gels, and the sizes of the fragment(s) were compared with the EcoRI/PstI digest. A third pair of enzymes, NdeI and Hin-dIII, were used to detect products formed by gene conversion. When products contained repeat tracts in orientation II, Sap I and HindIII were used. The fragments were separated on 7% polyacrylamide gels, and the sizes of the fragment(s) were compared with the EcoRI/PstI control fragments.
Sequential Transformation-The pUC19 derivatives pRW3036 and pRW4011 were used to separately transform E. coli AB1157. Transformants were selected on LB agar plates that contained ampicillin (75 g/ml). Individual colonies were selected and grown until mid-log phase, after which they were prepared as competent cells for electroporation (45). Hence these cells harbored a pUC19 derivative containing a TRS, namely, (CTG⅐CAG) 36 or (CAG⅐CTG) 50 . An aliquot of each new strain was analyzed, and the size of the TRS was confirmed by restriction analyses.
Subsequently, the pACYC184 derivatives pRW3238 and pRW3239 were used to transform the new strains (E. coli AB1157 harboring pRW3036 or E. coli AB1157 harboring pRW4011). Cotransformants were selected on LB agar plates that contained ampicillin (75 g/ml) and tetracycline (20 g/ml) and cultured as described above. Conversely, the pACYC184 derivative, pRW3238 (which contains (CAG⅐CTG) 100 ) was used to transform E. coli AB1157. Individual clones that formed on plates that contained tetracycline (20 g/ml) were cultured and prepared as competent cells (45). This new strain (E. coli AB1157 harboring pRW3238) was used for the stepwise transformation of the pUC19 derivatives (see Fig. 1). Cotransformants were selected on LB agar plates that contained ampicillin (75 g/ml) and tetracycline (20 g/ml).

Effect of TRS Orientation on Interplasmid Recombination-
The potential influence of TRS orientation on recombinationmediated expansions was studied biochemically and genetically in a two-plasmid recombination system in E. coli (44). Derivatives of nonhomologous plasmids were used to cotransform isogenic, recombination proficient and deficient strains. Each strain was cotransformed with plasmids containing different lengths of (CTG⅐CAG) tracts cloned in either orientation I or orientation II (defined in Fig. 1). In addition, the number of interruptions (defined under "Experimental Procedures") within the tracts also varied. The plasmids were derivatives of pUC19, which harbors the ampicillin resistance gene, and pA-CYC184 derivatives, which harbors the tetracycline resistance gene (Fig. 1). The two vectors are completely nonhomologous except for short, single copy regions of 5-48 bp (44). Thus, these recombinant plasmids harboring a wide variety of (CTG⅐CAG) tracts enabled our focus on the effects of insert orientation on recombination-mediated instability.
The recombination behavior of these plasmids ( Fig. 1) was analyzed in four E. coli strains that differed in their recombination capacity. Cotransformation of the plasmids is facile (44,48). Each E. coli strain (AB1157, JC5519, JC10287, and JC10289) was transformed and cotransformed with the individual plasmids (Fig. 1). All cotransformants were grown to an A 600 of between 0.5 and 0.9. As a control for the cotransformation of the plasmid pairs, each strain was cotransformed with the vector pACYC184, along with various pUC19 derivatives that contained CTG⅐CAG of different lengths, orientations, and extents of interruptions. Finally, the four strains were cotransformed with all possible pairwise permutations of the plasmids shown in Fig. 1. Thus, the only segments of the nonhomologous plasmids that had identical sequences were the CTG⅐CAG inserts, albeit in different orientations. Cotransformants were cultured in parallel and harvested under identical conditions. The DNA was purified and analyzed on agarose as well as native and denaturing polyacrylamide gels. The number of CTG⅐CAG expansion products and the magnitude of the expansions were compared between the different strains.
Instability of Triplet Repeats-The eight pUC19 derivatives ( Fig. 1) were separately cotransformed with the four pA-CYC184 derivatives. These 32 experiments were conducted in four different isogenic strains. First, the pACYC184 derivatives that contained different lengths of CTG⅐CAG in both orientations were cotransformed with various pUC19 derivatives that contained different lengths of CTG⅐CAG in both orientations. Second, each strain was cotransformed with the vector pA-CYC184 along with various pUC19 derivatives that contained different lengths of CTG⅐CAG in both orientations. Third, the vector pUC19 was cotransformed along with various pA-CYC184 derivatives that contained different lengths of CTG⅐CAG in both orientations. Hence, 128 experiments (32 ϫ 4, for each strain) were conducted.
Restriction analyses of plasmid DNA isolated from individual clones grown for ϳ6 h (15 generations) in the parent E. coli AB1157 showed that multiple fold expansions occurred within the (CTG⅐CAG) or (CAG⅐CTG) tracts. No significant differences in the number of expansions nor the magnitude of the expan-FIG. 1. Plasmids containing (CTG⅐CAG) and (CAG⅐CTG) repeats used in the two-plasmid recombination system. These plasmids are derivatives of the unidirectional replicating pUC19 and pACYC184. Plasmids pRW3036 through pRW4480 are derivatives of pUC19 and pRW4437 through pRW3238 are derivatives of pACYC184 (2,3,5,6,11,12). The types of sequences and their lengths are listed. For example, pRW4011 is a pUC19 derivative that contains a pure insert of (CAG⅐CTG) 50 in orientation II. All pUC19 derivatives contain pure noninterrupted repeating tracts of (CTG⅐CAG) or (CAG⅐CTG) sequences except pRW3753 and pRW3755 (3). The sequence in pRW3239 is not homogenous but contains two G-to-A point mutations at positions 28 and 69 as described previously (3,11). The orientation of the TRS with respect to the direction of replication is defined as follows: plasmids that contain the (CTG) sequence as the leading strand template are designated CTG⅐CAG and are in orientation I; plasmids that contain the (CTG) sequence as the lagging strand template are designated CAG⅐CTG and are in orientation II. For each pair of plasmids with the same length of TRS (i.e. pRW4015 and pRW4011), the relative locations of the origin of replication (Ori), the antibiotic resistance gene, and the TRS are identical except that the orientation of the repeat sequence is switched from orientation I to orientation II. The double-stranded plasmids are shown schematically on the right. The pUC19 derivatives are represented by heavy black circles, and the orientations of the TRS are indicated as either orientation I or orientation II. The double-stranded pACYC184 derivatives are represented by thin circles, and the TRS orientations are indicated as either orientation I or orientation II. Relevant restriction sites are shown. All plasmids have SacI (S) and HindIII (H) sites in common, but the pUC19 derivatives also contain single PstI (P), NdeI (N), and Sap I (SЈ) sites, whereas all pACYC184 plasmids contain single Bstz17 (B) and BspEI (BЈ) sites. The origins of replication (44), which are unique for the two nonhomologous plasmids, are shown as open and striped boxes, respectively. The ampicillin resistance gene (Am r ) is designated as a filled box on pUC19. The tetracycline resistance gene (Tc r ) is designated as a checkered box on pACYC184. pUC19 is a high copy number plasmid with ϳ500 copies/cell (45), whereas pACYC184 has a copy number of ϳ10/cell (44). The plasmids are not drawn to scale. sions were observed between orientations I or II. Examples of multiple fold expansions are shown in Fig. 2. Variations in the number of expansion products, the magnitude of the expansions, and the absence of deletion products and/or the progenitor insert starting length were detected between transformants. The product distribution of CTG⅐CAG lengths varied with each pUC19 derivative used in the different experiments. In the majority of clones, no deletion products were detected, and the progenitor (CTG⅐CAG) length of the original starting material was absent ( Table I).
The products from the cotransformations of plasmids harbor-ing (CTG⅐CAG) sequences in orientations I or II in E. coli AB1157 were distributed over a wide range of (CTG⅐CAG) lengths. The products were digested with SacI and PstI to release the (CTG⅐CAG)-containing fragment from the pUC19 derivatives and separated on a 7% polyacrylamide gel. The patterns of the length distributions were essentially random. Considering the singly transformed cells, lane B shows a 169-bp band, which is the expected size of a fragment containing 36 repeats of (CTG⅐CAG) plus the myotonic dystrophy human flanking sequences. Similarly, lanes C and D show 240and 301-bp bands, which are the expected sizes of fragments containing 50 and 80 repeats, respectively, of (CAG⅐CTG) plus human flanking sequences.
Considering the cotransformed cells, lane E shows the SacI/ PstI fragments from E. coli AB1157 cotransformed with pRW4011 (a pUC19 derivative with (CAG⅐CTG) 50 in orientation II) and pRW3238 (a pACYC184 derivative with a (CAG⅐CTG) 100 tract in orientation II). Surprisingly, instead of recovering the original length insert (lane C), new products that were severalfold longer than the progenitor starting material length were found. Comparison between the product in lane C with the products in lane E clearly demonstrate that the original starting material length (CAG⅐CTG) 50 was converted to several expansion products, some of which were expanded by more than 2-fold. The expansion products, (CAG⅐CTG) 79 and (CAG⅐CTG) 109 , represent the majority of the species. The difference in size between these expansion products suggests an incremental length increase because the progenitor insert length of (CAG⅐CTG) 50 was expanded to (CAG⅐CTG) 79 (an increase of 29 repeats) and/or expanded to (CAG⅐CTG) 109 (an increase of 30 repeats). However, additional sample analyses revealed this to be coincidental because there was no correlation between the size of the insert in the progenitor plasmid and their expansion products.
Lane F shows the analysis of another isolate of the experiment shown in lane E; however, the magnitude of expansion and the number of expansion products are not the same as those contained in lane E. Therefore, a stochastic process that is recombination-dependent must be responsible for generating these expansion products. Again, no deletion products were observed. Another example of expansions that occur regardless of the insert orientation is shown by the cotransformation of pRW4011 and pRW3239 (which contains (CTG⅐CAG) 175 in orientation I) (lane G). Unlike the previously described products (lanes E and F), the sample in lane G contains deletion products as well as expansions. The extent and the magnitude of the deletion products were similar to those recovered from the recombination deficient strains (data not shown). The size of the deletion from the progenitor starting material compared with the sizes of the expansions is not an equal number (deletion of 10 repeats and expansions of 34 and 59 repeats, respectively). Thus, reciprocal exchange between the pUC19 and pA-CYC184 derivatives is unlikely. Lane G shows the presence of expansions as found in lanes E and F where the orientation of  (Fig. 1) were used to transform or cotransform E. coli AB1157. The DNA was isolated, and the (CTG⅐CAG) or (CAG⅐CTG) containing restriction fragments (SacI/PsI I) were subsequently electrophoresed. The expansion of TRS is determined from the change in size of the TRS contained in the original starting plasmid compared with the products. The difference in the electrophoretic mobility of the TRS containing restriction fragments is due to the difference in the number of (CTG⅐CAG) or (CAG⅐CTG) repeats. The left axis shows the migration distance (in numbers of repeats) expected for a SacI/PstI fragment containing 36, 50, 80, 104, or 206 repeats of (CTG⅐CAG) or (CAG⅐CTG) on a 7.0% polyacrylamide gel. Lane A contains the 1-kilobase pair DNA ladder (Life Technologies, Inc.). The sizes of these bands are 517, 502, 398, 344, 298, 220, 201, and 154 bp. All plasmid products (lanes B-Q) were digested with SacI ϩ PstI. Lane B, the SacI/PstI restriction fragment of pRW3036 (which contains (CTG⅐CAG 36 )) in orientation I. Lane C, pRW4011 (which contains (CAG⅐CTG 50 )) in orientation II. Lane D, pRW3080 (which contains (CTG⅐CAG 80 )) in orientation I. Lane E, cotransformation of pRW4011 (which contains (CAG⅐CTG 50 )) in orientation II and pRW3238 (which contains (CAG⅐CTG 100 )) in orientation II. Lane F, cotransformation of pRW4011 (which contains (CAG⅐CTG 50 )) in orientation II and pRW3238 (which contains (CAG⅐CTG 100 )) in orientation II. Lane G, cotransformation of pRW4011 (which contains (CAG⅐CTG 50 )) in orientation II and pRW3239 (which contains (CTG⅐CAG 175 )) in orientation I. Lane H, cotransformation of pRW4011 (which contains (CAG⅐CTG 50 )) in orientation II and pRW3238 (which contains (CAG⅐CTG 100 )) in orientation II. Lane I, cotransformation of pRW4011 (which contains (CAG⅐CTG 50 )) in orientation II and pRW3238 (which contains (CAG⅐CTG 100 )) in orientation II. Lane J, cotransformation of pRW3036 (which contains (CTG⅐CAG 36 )) in orientation I and pRW3238, (which contains (CAG⅐CTG 100 )) in orientation II. Lane K, cotransformation of pRW3036 (which contains (CTG⅐CAG 36 )) in orientation I and pRW3238 (which contains (CAG⅐CTG 100 )) in orientation II. Lane L, cotransformation of pRW4436 (which contains (CAG⅐CTG 36 )) in orientation II and pRW3238 (which contains (CAG⅐CTG 100 )) in orientation II. Lane M, cotransformation of pRW4436 (which contains (CAG⅐CTG 36 )) in orientation II and pRW3238 (which contains (CAG⅐CTG 100 )) in orientation II. Lane N, cotransformation of pRW4480 (which contains (CAG⅐CTG 80 )) in orientation II and pRW3238 (which contains (CAG⅐CTG 100 )) in orientation II. Lane O, cotransformation of pRW4480 (which contains (CAG⅐CTG 80 )) in orientation II and pRW3238 (which contains (CAG⅐CTG 100 )) in orientation II. Lane P, cotransformation of pRW3080 (which contains (CTG⅐CAG 80 )) in orientation I and pRW4439 (which contains (CAG⅐CTG 39 )) in orientation II. Lane Q, cotransformation of pRW3080 (which contains (CTG⅐CAG 80 )) in orientation II and pRW4439 (which contains (CAG⅐CTG 39 )) in orientation II.  176  211  175  208  172  205  169  175  139  120  84  117  81  102  66  91  55  70  34  66  30  39  3  36  0  pRW4011  (CAG⅐CTG) 50  175  125  150  100  149  99  148  98  147  97  102  52  101  51  98  48  80  30  78  28  71  21  69  19  68  18  58  8  52  2  50 0 TABLE I Summary of expansions of (CTG⅐CAG) or (CAG⅐CTG) Plasmids (Fig. 1) were cotransformed into recombination-proficient E. coli, and DNA was isolated from individual colonies grown as described under "Experimental Procedures." The lengths of the repeat expansions in the pUC19 derivatives were determined by DNA sequencing and/or by restriction mapping (usually SacI and PstI were employed). Each line in the table represents a length determination on the recombinant DNA product from an individual clone that was grown on plates containing ampicillin and tetracycline. Section A contains the products of three pUC19 derivatives harboring (CTG⅐CAG) in orientation I cotransformed with a pACYC184 derivative (pRW3238) harboring (CAG⅐CTG) 100 in orientation II. Section B contains the products of a pUC19 derivative (pRW3080) harboring (CTG⅐CAG) 80 in orientation I cotransformed with a pACYC184 derivative (pRW4439) harboring a (CAG⅐CTG) 39 sequence in orientation II. Section C contains the products of five pUC19 derivatives with different lengths of (CAG⅐CTG) in orientation II cotransformed with a pACYC184 (pRW3239) derivatives harboring (CTG⅐CAG) 175 in orientation I. Section D contains the products of three pUC19 derivatives with different lengths of (CAG⅐CTG) in orientation II cotransformed with a pACYC184 (pRW3238) derivatives harboring (CAG⅐CTG) 100 in orientation II. the insert is reversed in the pACYC184 derivative. Nevertheless, in some cases, similar products were formed for certain cotransformations (lane H). Lane H shows the restriction fragment products of pRW4011 and pRW3238 (which contains (CAG⅐CTG) 100 ) cotransformed into E. coli AB1157. Expansions are the only type of products recovered. Second, as in lane G, the progenitor insert length is completely absent. Although, no deletion products were recovered, the expansion product seems to be identical to lane G. Thus, it is possible that different cotransformed DNAs yield the same products, indicating the random nature of the recombination events. Lane I contains another example of the restriction fragments from the cotransformation of pRW4011 and pRW3238 in E. coli AB1157. As for the duplicate experiment in lane F, the progenitor starting length of pRW4011 is recovered, but the expansion product is similar to one of the products shown in lanes E and G. However, unlike lane F, a small amount of a deletion product was detected in lane I. In summary, lanes E-I demonstrate that the recombination mechanism results in the formation of highly variable length alterations in the product (CTG⅐CAG) tracts regardless of repeat orientation.
Considering the cotransformation of other pUC19 derivatives, lane J shows that a 6-fold expansion product was recovered from the cotransformation of pRW3036 (which contains (CTG⅐CAG) 36 ) with pRW3238. This expansion product was longer than the sum of the lengths of the (CTG⅐CAG) tracts from both plasmids; other examples of this behavior were frequent (Table I, sections A, C, and D). Another clone from the cotransformation of pRW3036 with pRW3238 in E. coli AB1157 is shown in lane K. This sample did not contain expansions or deletions and is consistent with previous reports that shorter inserts tend to be more stable than longer repeat inserts in E. coli, yeast, or in humans (1, 3, 16, 44, 49 -51).
Considering the products from the cotransformation of pRW4436 (which contains (CAG⅐CTG) 36 in orientation II) with pRW3238, multiple fold expansions and a two repeat deletion product were recovered. Lane L shows the products of (CAG⅐CTG) 36 that is expanded to (CAG⅐CTG) 51 and (CAG⅐CTG) 207 . The results from a different clone of pRW4436 cotransformed with pRW3238 (lane M) shows expansion products of (CAG⅐CTG) 78 and (CAG⅐CTG) 205 . In summary, the results of lanes J-M illustrate the random formation of multiple fold expansions and show that an expansion bias is associated with the recombination events.
Considering the cotransformation of pUC19 derivatives harboring longer (CTG⅐CAG) tracts than (CTG⅐CAG) 36 , lane N shows the restriction fragment products from the cotransformation of pRW4480 (which contains (CAG⅐CTG) 80 in orientation II) and pRW3238 (which contains (CAG⅐CTG) 100 in orientation II). The (CAG⅐CTG) 80 tract in the pUC19 derivative was expanded by 133 repeats. Another clone from the cotransformation of pRW4480 with pRW3238 (lane O) was expanded by 132 repeats. These similar expansion products represent some of the largest changes in repeat size observed in the twoplasmid recombination system. There seems to be an apparent maximum length to the expansion of the (CTG⅐CAG) repeat tract. Indeed, the largest expansion product detected was a (CAG⅐CTG) 300 tract (Table I, section C). Although the reason for this length limit is uncertain, it was shown that long tracts of TRS have an increased frequency of deletions (1, 7, 8, 10 -12, 16, 49, 51).
Surprisingly, when pACYC184 derivatives harboring shorter lengths of (CAG⅐CTG) than those harbored in the pUC19 derivatives were used in the two-plasmid system, the expansion bias of the pUC19 derivative was dramatically reduced. Lanes P and Q show the products from the cotransformation of pRW3080 (which contains (CTG⅐CAG) 80 ) and pRW4439 (which contains (CAG⅐CTG) 39 ). A variety of deletion products and an expansion of only three repeats were detected in this bay of experiments. In summary, we conclude that the E. coli recombination system significantly promotes the expansion of (CTG⅐CAG) triplet repeats in vivo for sequences cloned in either orientation I or II.
Quantitation of Expansions-More than 110 examples of expansions are listed in Table I, which is divided into four subsections. Section A contains the products from the cotransformation of pRW3238 (which contains (CAG⅐CTG) 100 in orientation II) with various pUC19 derivatives containing (CTG⅐CAG) in orientation I. Section B contains the products from the cotransformation of pRW4439 (a pACYC184 derivative that contains (CAG⅐CTG) 39 in orientation II) with pRW3080. Section C contains the products from the cotransformation of pRW3239 with various pUC19 derivatives containing (CAG⅐CTG) in orientation II. Section D contains the products from the cotransformation of pRW3238 with various pUC19 derivatives containing (CAG⅐CTG) in orientation II. Each subsection is divided into individual entries grouped according to the plasmid pairs used for cotransformation of E. coli AB1157. Each entry represents the analysis of a single colony characterized by restriction mapping and DNA sequencing where appropriate. Table I summarizes the magnitude of the expansion events. Several conclusions may be drawn: 1) expansions are found regardless of the orientation of the inserts in the vectors (Table  I, sections A, C, and D); 2) shorter repeat lengths (36 (data not shown) and 39 units) in the pACYC184 derivatives do not support recombination-mediated expansions (Section B); 3) multiple fold expansions were routinely found (i.e. severalfold longer than the length of the pUC19 progenitor repeat as well as longer than the sum of the two progenitor repeat lengths); 4) the lengths of the expanded repeats were essentially random; 5) the expansion products were of discrete lengths, not a family of lengths as detected by smears of bands; 6) repeat tracts containing interruptions in the pUC19 derivatives (pRW3753 and pRW3755) are much less effective in generating expansion products than similar lengths of pure repeats (pRW4011 and pRW3080), respectively; and 7) the lengths of the interrupted repeat tracts in the low copy number pACYC184 derivatives were unchanged for the cases investigated (ϳ20 isolates).
Effect of Interruptions-Point mutations (polymorphisms) within TRS have an important consequence for hereditary diseases because they increase the stability of the triplet repeat sequences (52)(53)(54). Hence, we made direct comparisons of the extent and magnitude of expansions between interrupted and pure (CAG⅐CTG) repeat tracts used in the recombination system. E. coli AB1157 was cotransformed with pRW3753 (which contains (CAG⅐CTG) 57 with two G-to-A point mutations within the repeat at positions 28 and 43) and pRW3239 and cultured as described above (under "Experimental Procedures"). Parallel control experiments using pRW4011 (which contains a pure repeat tract of similar length as pRW3753) and pRW3239 were also conducted. The products were purified and analyzed by restriction mapping. The products from the progenitor pRW3753 repeats were predominately deletions; less than 10% of the products were expansions of less than 15 repeats (approximately one-fourth of the original length). No multiple fold expansion products were recovered when pUC19 derivatives contained point mutations within the repeat tract. The effect of interruptions was studied further with pRW3755 (which contains (CAG⅐CTG) 73 ), which harbors a longer repeat sequence than pRW3753 and also contains two G-to-A point mutations (at positions 28 and 59); this plasmid was cotransformed with pRW3239 into E. coli AB1157. There was no difference between the repeat instability observed for pRW3753 and pRW3755. In the case of pRW3755, despite the longer length of the repeat, deletions were still the predominant products. In contrast, the products from the control cotransformations of pRW4011 and pRW3239 and those of pRW4480 and pRW3239 were predominantly expansions (Tables I and II). Hence, the in vivo recombination-mediated expansion process is disrupted by the polymorphisms within the repeat. Thus, the presence of two G-to-A interruptions (point mutations) within either 171-or 219-bp tracts profoundly reduced the gene conversion-mediated expansions compared with uninterrupted repeat tracts of similar lengths.
Recombination-deficient Strains-Parallel studies were conducted with all plasmids pairs (Table I) in recombinationdeficient E. coli (strains JC10287, JC10289 (recA Ϫ ) or JC5519 (recBC Ϫ )). Each strain was cotransformed with pairs of plasmids containing different lengths of (CTG⅐CAG) tracts cloned in either orientation I or II. The DNA was isolated and subsequently analyzed by restriction mapping and DNA sequencing. The amount of expansion products obtained from these experiments was below detection limits (data not shown). The digestion pattern of the products from the recombination deficient strains showed that the progenitor length remained the same and that some deletion products were present, as expected. Thus, recombination-deficient strains maintain the progenitor length (when less than 100 repeats) in subsequent generations, as expected. However, in less than 10% of the clones, small increases in insert length did occur. For such cases, the in-crease in size of the repeat tract was less than 10% of the size of the progenitor length. For example, the products from the cotransformation of E. coli JC5519 with pRW3080 (which contains (CTG⅐CAG) 80 ) and pRW3238 ranged in size from (CTG⅐ CAG) 85 , which is an expansion of 5 repeats, to (CTG⅐CAG) 30 which is a deletion of 50 repeats (data not shown). Such changes in length are common occurrences in all E. coli strains (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13). Hence, few or no recombination-mediated expansions were observed between plasmid pairs in recA Ϫ or recBC Ϫ strains.
Expansions are the Dominant Products in the Two-plasmid Recombination System-When the products of recombinationmediated instability were categorized according to the number of expansion events versus the number of deletion events, there was an expansion bias. To specifically address this issue, a cotransformation experiment using pRW4011 (which contains (CAG⅐CTG) 50 ) and pRW3239 (which contains (CTG⅐CAG) 175 ) was conducted in E. coli AB1157 and compared with control experiments. After 20 generations, the progenitor plasmid (pRW4011) was analyzed by restriction mapping and DNA sequencing ( Table II). All of the clones contained expansions of the (CAG⅐CTG) tract. Quantitation of the expansion and deletion products from the cotransformation of E. coli AB1157 with pRW4011 and pRW3239 was performed by scanning polyacrylamide gels containing the end-labeled fragments. The (CAG⅐CTG) tracts were excised from the products with EcoRI and PstI, end-labeled with [␣-32 P]dATP, and electrophoresed on a 7% polyacrylamide gel. The amount of instability was measured by exposing the dried gel to a Molecular Dynamics PhosphorImager screen. The percentage of molecules that contained expansions and/or deletions was measured as the proportion of the total radioactivity in the lane below 1.6 kilobase pairs (the linearized vector backbone).
Ten clones were chosen at random, and five of the clones contained only expansions; no deletions were detected by separation of end-labeled fragments on polyacrylamide gels (Table  II). The other five clones contained both expansions and deletions as well as the original length fragment. The detection limit of radiolabeled materials is ϳ1 ϫ 10 Ϫ12 mol. Thus, when no deletion products were detected and the only products present were expansions, the frequency of formation of expansions compared with that of deletions is as great as 3 ϫ 10 3 . One out of the 10 samples contained a small amount of deletion products (about 6%), whereas the majority of instability was manifested as expansions. The remaining four clones contained an equal number of expansion and deletion products. In summary, the average expansion/deletion ratio of these 10 samples was 1.5 ϫ 10 3 :1. Although these biochemical determinations give a reasonable estimate of the frequency of expansions to deletions, we acknowledge that a genetic assay would be desirable. A genetic assay is under current investigation. 2 Gene Conversion Is the Recombination Mechanism-The mechanism of recombination responsible for generating the observed multiple fold expansions was determined. DNA sequencing analyses were performed on the products formed by the cotransformation of pRW3036, which contains (CTG⅐ CAG) 36 , and pRW3239, which contains (CTG⅐CAG) 175 , in E. coli AB1157. The cotransformations were repeated numerous times on different days, and individual colonies were analyzed. The flanking sequences adjacent to the repeat inserts are unique to the pUC19 and pACYC184 derivatives. Plasmid pRW3036 contains nonhuman flanking sequences adjacent to the repeat, whereas pRW3239 contains human flanking sequences adjacent to the repeat. Therefore, the identification of 2 A. Pluciennik and R. D. Wells, unpublished data.

TABLE II
In vivo expansion bias of (CTG⅐CAG) pRW4011 (which contains (CTG⅐CAG) 50 ) and pRW3239 (which contains (CTG⅐CAG) 175 ) were cotransformed into E. coli AB1157. Individual clones were selected at random, grown to mid-log phase, and harvested, and the DNA was purified. To quantitate the products, and EcoRI/PstI double digestion was performed in EcoRI buffer at 37°C for 4 h to release the fragment containing the TRS. The restriction fragment was subsequently labeled with the Klenow fragment of the E. coli DNA polymerase I and [␣-32 P]dATP. The extent of the instabilities of (CAG⅐CTG) was measured by exposing the dried 7% polyacrylamide gel to a Molecular Dynamics PhosphorImager screen followed by scanning. The amount of radioactivity estimated by the signal intensity in the bands corresponding to the restriction fragments was measured as a proportion of the radioactivity in the lane below the vector band. The disappearance of the restriction fragment corresponding to the original starting length ((CTG⅐CAG) 36 repeats) was compared to the appearance of restriction fragments that were longer than the original starting length (expansions) and restriction fragments that were shorter than the original starting length (deletions). the flanking sequences on both sides of the repeat inserts contained in the expansion products enables differentiation between recombination that occurs via crossing over or gene conversion. Gene conversion is defined as nonreciprocal homologous recombination with retention of the flanking sequences. Dideoxy sequencing of the expansion products was conducted using T7 DNA polymerase (Sequenase) and the M13/pUC forward primer 1211, which is unique to the 5Ј end of the polylinker region of pUC19. Table III shows the analyses of 15 colonies that had varying extents of expansions. The flanking pUC19 sequences were retained for all 15 products; no cases of exchange were found. Although other samples with multiple fold expansions were also sequenced, these samples were of an appropriate size to enable visualization of the entire (CTG⅐CAG) tracts and the 5Ј-and 3Ј-flanking sequences on a single gel. Larger expansion products (as determined by restriction mapping) were also sequenced, and the 3Ј end of the repeat did contain the pUC19 vector flanking sequence, but the exact length of the expansion could not be determined because the 5Ј end of the repeat had migrated off the bottom of the gel.
When the flanking sequence 3Ј to the repeat tract contained CTGCGAAGAGCT as well as the rest of the nonhuman flanking sequence followed by the pUC19 vector sequence, the type of recombination was classified as gene conversion. In summary, DNA sequence analyses confirmed that all 15 products were expanded by gene conversion. However, crossing over does occur in this system because long cointegrated DNAs were reported (44).
Restriction Mapping Strategy to Determine Relative Frequency of Crossing Over versus Gene Conversion-The EcoRI/ PstI double digestion scheme, described above, which analyzed the length polymorphisms of the TRS in the progenitor, differentiated between the original starting material and the expanded TRS in the products. However, it cannot distinguish between all types of recombination that may be responsible for generating the expansion(s). For example, an odd number of crossing over events might generate an expanded TRS tract such that the EcoRI and PstI double digestion would release a TRS-containing fragment. Therefore, products formed by crossing over as well as those formed by gene conversion would be detected using an EcoRI/PstI analysis.
To measure products formed by crossing over (exchange of flanking markers), further restriction mapping analyses were conducted. Two different experiments using unique restriction enzymes were used to identify the relative frequency of crossing over versus gene conversion (retention of flanking sequences). First, products that had expansions of 0.5-3-fold (based on the EcoRI and PstI double digestion, described previously) were subjected to the following double digestions: BspEI and PstI; Bstz17I and NdeI. Because these restriction enzyme pairs are unique to pACYC184 and pUC19 (each vector contains one of the sites (Fig. 1)), any fragments containing TRS generated by the BspEI/PstI or the Bstz17I/NdeI double digestions would indicate that the products were formed via crossing over events between pACYC184 and pUC19. The TRS-containing fragments would appear as lengthened tracts (ϳ341 or 309 bp, respectively) compared with the EcoRI/PstI digest of the progenitor.
Analyses of eight expansion products (individual clones) yielded only the vector backbones and no TRS-containing fragments as detected with end-labeled fragments ("Experimental Procedures") in 7% polyacrylamide gels. Thus, crossing over (exchange of flanking sequences) was not involved in generating these expansions.
Secondly, products formed by gene conversion (retention of flanking markers) were identified via NdeI/HindIII double digestion. This double digestion would only release a fragment of ϳ500 bp from the pUC19 derivative. This fragment is 220 bp longer than the EcoRI/PstI digest. If the expansion products were formed via gene conversion, then a new band (Ն500 bp depending of the size of expansion) would be detected. Analyses of eight expansion products (individual clones) showed fragments containing the TRS as detected by end-labeled fragments in 7% polyacrylamide gels. These products were Ն220 bp longer than the EcoRI/PstI fragments. Thus, all eight products were due to gene conversion events. Therefore, these data plus the DNA sequencing data (described above) agree that gene conversion is the predominant mechanism for these recombination-mediated expansions.
Background TRS Instability-161 individual clones from single plasmid transformations were analyzed to measure the inherent level of tract instability in the progenitor. E. coli AB1157 were transformed with pRW3036 (containing (CTG⅐ CAG) 36 in orientation I), pRW4011 (containing (CAG⅐CTG) 50 in orientation II), and pRW3080 (containing (CTG⅐CAG) 80 in orientation I), respectively. The transformants were plated on agarose containing ampicillin, and clones were subsequently cultured. The DNA was isolated and digested with EcoRI/PstI, and the fragments were separated on 7% polyacrylamide gels. Stepwise Transformation-Experiments were conducted to determine whether the order of introduction of the plasmids contributes to the TRS expansions. The pUC19 derivatives pRW3036 and pRW4011 were used to separately transform E. coli AB1157 cultures. Transformants were cultured in the pres-  36 ) and pRW3239 (which contains (CTG⅐CAG) 175 ) were sequenced using the pUC19 primer 1211 that binds to the human flanking sequence near to the 5Ј end of the (CTG⅐CAG) insert. Both plasmids contain (CTG⅐CAG) inserts in orientation I. The isolated DNA was electrophoresed through a 7% denaturing polyacrylamide gel in Tris-Borate-EDTA buffer at 30 V/cm at 44°C. The gels were washed with 15% acetic acid, 5% methanol, dried, and exposed to film or to a PhosphorImager screen. The numbers show the length of each expansion product and indicate how many repeats were added to the progenitor length of (CTG⅐CAG) 36 . The flanking sequence 3Ј to the (CTG⅐CAG) insert is listed for each isolate. When the 3Јflanking sequences were identical to that of pUC19, the fate of flanking sequences was scored as "retention." When the 3Ј-flanking sequences were identical to that of pACYC184, they were scored as "exchange." ence of ampicillin for 6 h. The DNA was isolated and analyzed by restriction mapping, which confirmed the presence of pRW3036 or pRW4011, as expected. Thus, the progenitor plasmids were transmitted stably. These new strains (E. coli AB1157 harboring pRW3036 and AB1157 harboring pRW4011) were rendered competent. The competent cells were then used for transformation by pACYC184 derivatives, pRW3238 or pRW3239 (Fig. 1) and selected on agar plates containing ampicillin and tetracycline. The DNAs from these clones were digested with SacI/PstI, and the fragments were separated on 7% polyacrylamide gels. The products from these stepwise transformations were expansions and deletions with a bias in favor of expansions. The number and magnitude of the expansions were very similar, if not identical, to those products obtained from numerous experiments described in Fig. 1 and Table I. For a thorough study of stepwise transformation, the order of plasmid introduction was reversed. E. coli AB1157 was transformed initially with the pACYC184 derivatives pRW3238 and pRW3239 and cultured on tetracycline. The cells were rendered competent and used for the transformation by the pUC19 derivatives ( Fig. 1) and selected on plates containing ampicillin and tetracycline. The products were digested with SacI/PstI, and the fragments were separated on 7% polyacrylamide gels. The products of these transformations were expansions and deletions with a bias in favor of expansions. Thus, the order of the plasmid transformation (either pUC19 derivatives transformed first followed by pACYC184 derivatives or pACYC184 derivatives transformed first followed by pUC19 derivatives or both introduced simultaneously) did not influence the number or magnitude of the expansions.

DISCUSSION
This report describes the direct demonstration that gene conversion mediates the expansions of CTG⅐CAG 3 repeats. Gene conversion is the principal recombination mechanism rather than unequal crossing over. Also, the results from a large number of gene conversion-mediated expansion studies with plasmids containing TRS inserts in all possible permutations of the orientations revealed little or no influence of orientation. Hence, these data provide further verification that recombination (not complementary strand slippage at the replication fork) is the responsible mechanism. Gene conversion (recombinational repair) is the nonreciprocal transfer of genetic information from one DNA duplex to another with no exchange of flanking sequences (55)(56)(57). Hence, this would explain the linkage disequilibrium of flanking markers in haplotype analyses (1,(52)(53)(54), if this recombination mechanism is also responsible for expansions in human cells. Whereas our paper (44) was the first to demonstrate the involvement of recombination in TRS expansions, other workers (58) demonstrated that gene conversion plays the major role in controlling the instability of large tandem repeats of ribosomal DNA sequences in yeast. Although the ribosomal tandem repeat sequences are much longer than triplets, the expansion mechanism is probably the same.
The gene conversion (recombinational repair) mechanism responsible for the repeat tract expansions is modeled in Fig. 3. The DNA replication fork stalls when it encounters a CTG⅐CAG sequence (8,13,27), which can result in double-stranded DNA breaks that widen into double-stranded gaps (41,56,59). When breaks occur in the repeat tract (15,19) of the progenitor, the resultant gap may range in size from 40 to 400 repeats (55,60). A search for homology then takes place, and the ends of the progenitor strands invade the TRS in the pACYC184 derivative (Fig. 3, upper right panel). After annealing, the gap is filled in by recombination repair and ligated to finalize the expansion of the TRS tract in the pUC19 derivative. In addition to these standard recombination repair steps, it is possible with the triplet repeat tracts that DNA slippage may enable the formation of hairpin structures (center, right) (2-13, 21, 24 -26, 61-64). Thus, these two processes may act in concert to enhance the formation of expansions. Others have suggested that gene conversion and break-induced replication are cooperative processes (65). Because double-stranded breaks occur frequently under normal growth conditions and recBCD and recA proteins are essential for recombinational repair (55,66,67), this also suggests a coupling of replication and recombination.
Our data ( Fig. 2 and Table I) demonstrate that gene conversion mediates the multiple fold expansions of CTG⅐CAG sequences, irrespective of the orientation of the TRS tracts. These results are in stark contrast to some prior studies in E. coli (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13) and yeast (15,16,68), which demonstrated a marked influence of the orientation of the repeat tracts on deletions and expansions. These results were explained by slippage of the complementary repeat tracts at the replication fork, which was enhanced by the preferential formation of DNA looped conformations (13,63,64,69) that enabled bypass synthesis at the replication fork. When CTG repeats were in the lagging strand template, deletions were more likely than expansions. This behavior could also occur on the leading strand template (49). However, in the studies described herein in recombinationproficient cells with the two-plasmid system, the inversion of the repeat tract does not cause an alteration in the instability,  (15,56,59,66) and is enlarged to a double-stranded gap (55,60). Strand invasion with staggered hybridization initiates the recombinational repair process. Two Holliday-like junctions are formed and DNA repair synthesis (filled circles) takes place on both strands. Simple repeat sequences adopt slipped mispaired structures (3,13,16,69). After the formation of hairpin loop structures, further DNA synthesis may also contribute to elongation of the TRS. Depending on the position of the staggered hybridization, the amount of slippage and the extent of branch migration, different size expansion products will be formed. Resolution of the junctions leaves the flanking sequences unaltered (not exchanged). as expected from recombination mechanisms (55,57,65,70). Recombinational repair (Fig. 3) is dependent on sequence homology but not sequence orientation. Hence, these results further verify that recombination is the mechanism involved in the multiple fold expansion process. Although gene conversion is the simplest mechanism to explain our data, we cannot eliminate the possibility of a double unequal cross-over event. Also, we demonstrated the requirement for recA as well as recBC functions. As shown in Fig. 3, recombination and DNA synthetic processes may act in concert to generate the instabilities.
The lengths of the repeat tracts as well as the absence or the presence of interruptions influences the expansion process. In our two-plasmid recombination system, the pACYC184 derivatives containing only 36 or 39 repeats were unable to effect expansions, whereas 100 or 175 repeats were extremely active, regardless of the length of the TRS in the pUC19 derivatives. Hence, this effect of length is consistent with prior reports (57) on recombination mechanisms and may explain why other workers observed only deletions in their yeast recombination system (30).
A hallmark of triplet repeat diseases is that intermediate size repeats lose a snip (single nucleotide polymorphism) and subsequently expand by severalfold (52)(53)(54). Without the loss of the interruption, the repeat is stably transmitted through a pedigree. We compared the effect of interruptions on gene conversion-mediated expansions with sequences containing no interruptions. However, when the interrupted pUC19 derivatives (pRW3753 and pRW3755) were analyzed, few or no expansions occurred. These plasmids contain two G-to-A point mutations; thus, if the complementary strands slip relative to each other, four mismatches will be generated. Prior work demonstrated that these mismatches effectively destabilize the formation of slipped structures (69). Thus, the polymerase complex is not likely to pause, and the replication fork will not collapse which will limit the number of double-stranded breaks that are inflicted. Whereas four mismatches in 219 bp is a small percentage, this extent of heterology (1.8%) was quite effective in eliminating the formation of multiple fold expansion products in our system. Similarly, other workers have recently demonstrated that as little as 1.2% heterology in a mammalian cell recombination system reduces the effects of recombination (71).
Little or no correlation was found between the length of the CTG⅐CAG tracts in pACYC184 (100 and 175 repeats in length) and pUC19 derivatives with the size of the expansion products. Specifically, any individual clone from a single transformation may contain products of many different sizes. Some clones had only one expansion product, and others had several products. A few clones had both expansions and deletions of the TRS. However, the lengths of these instabilities were essentially random. Interestingly, the repeat tracts were expanded to generate discrete lengths rather than smears representing a large number of related-length molecules. The reason for this behavior is uncertain and is under further investigation. Furthermore, the frequency of expansions to deletions in this gene conversion system is extremely high (approximately 1.5 ϫ 10 3fold). This behavior is in marked contrast to the prior investigations in recombination-deficient E. coli and in yeast where expansions were substantially less frequent than deletions by a ratio of approximately 1:100 (2-13, 15, 16, 68). Thus, we believe that powerful recombination processes may be responsible for the large expansions that are observed in human genetic studies of myotonic dystrophy, spinocerebellar ataxia type 8, Friedreich's ataxia, and fragile X syndrome (reviewed in Ref. 1).
In summary, gene conversion is a robust mechanism for effecting multiple fold expansions of CTG⅐CAG tracts in this genetically and biochemically tractable system. Of course, it is desirable to extend these investigations into human systems to evaluate its role in disease pathogenesis. The development of a suitable human gene conversion system will be required to rigorously evaluate this question. Whereas this will present challenges, it may be noted that a recent investigation on individuals with an Ataxia-Telangiectasia-like disorder showed a link between genetic defects in double-stranded break repair genes and this disease of chromosomal instability (72). Hence, if a link is established between gene conversion mechanisms and the expansions (anticipation) observed in certain human hereditary neurological diseases, new targets for therapeutic interventions may be established.