Methyl-directed Repair of Mismatched Small Heterologous Sequences in Cell Extracts from Escherichia coli *

The methyl-directed DNA repair efficiency of a set of M13mp18 heteroduplexes containing 1–8 or 22 unpaired bases was determined by using an in vitro DNA mismatch repair assay. The unpaired bases of each heteroduplex residing at overlapping recognition sites of two restriction endonucleases allow independent assay of repair on either DNA strand. Our results showed that the repair of small nucleotide heterologies in Escherichia coliextracts was very similar to base-base mismatch repair, being strand-specific and highly biased to the unmethylated strand. Thein vitro activity was also dependent on products ofmutH, mutL, mutS, anduvrD loci and was equally efficient on nucleotide insertions and deletions. The repair levels of small heterologies were affected by base composition of the heterologies. However, the extent of repair of heteroduplexes containing small heterologous sequences was found to decrease with an increase in the number of unpaired bases. Heteroduplexes containing an extra nucleotide of 22 bases provoked very low level of methyl-directed repair.

The methyl-directed DNA repair efficiency of a set of M13mp18 heteroduplexes containing 1-8 or 22 unpaired bases was determined by using an in vitro DNA mismatch repair assay. The unpaired bases of each heteroduplex residing at overlapping recognition sites of two restriction endonucleases allow independent assay of repair on either DNA strand. Our results showed that the repair of small nucleotide heterologies in Escherichia coli extracts was very similar to base-base mismatch repair, being strand-specific and highly biased to the unmethylated strand. The in vitro activity was also dependent on products of mutH, mutL, mutS, and uvrD loci and was equally efficient on nucleotide insertions and deletions. The repair levels of small heterologies were affected by base composition of the heterologies. However, the extent of repair of heteroduplexes containing small heterologous sequences was found to decrease with an increase in the number of unpaired bases. Heteroduplexes containing an extra nucleotide of 22 bases provoked very low level of methyl-directed repair.
The Escherichia coli methyl-directed mismatch repair system monitors the fidelity of DNA replication and recombination in this organism (1). The methyl-directed reaction has been reconstituted in a purified system (2). The reaction is composed of hemimethylated DNA substrate containing a mispair, E. coli MutH, MutL, MutS, and DNA helicase II, along with singlestranded DNA-binding protein, DNA polymerase III holoenzyme, DNA ligase, and any one of the DNA exonucleases Exo I, Exo VII, or RecJ (2,3). The major steps in the excision repair pathway for methyl-directed removal of mismatched DNAs have been well defined. The requisite strand specificity for processing of replication errors is provided by patterns of adenine methylation at d(GATC) sequences (2). Repair is initiated by binding of MutS to the mismatch (4), followed by the addition of MutL (5). Assembly of this complex leads to activation of a latent d(GATC) endonuclease of MutH protein, which incises the unmodified strand at a hemimethylated d(GATC) sequence (6). The resulting strand break, which can occur either 3Ј or 5Ј to the mismatch on the unmethylated strand, suffices to target correction to this strand (3). The ensuing excision reaction, which depends on MutS, MutL, and the cooperative action of DNA helicase II with an appropriate exonuclease, removes that portion of the unmodified strand spanning the d(GATC) site and the mismatch (7). Resynthesis of the excised strand by DNA polymerase III holoenzyme subsequently replaces the misincorporated nucleotide, with ligase restoring covalent integrity to the helix.
In E. coli mismatch correction, adenine methylation of d(GATC) sequences determines the strand on which repair occurs (2,8). With hemimethylated heteroduplex, which is methylated at d(GATC) sequences on only one DNA strand, repair is highly biased to the unmethylated strand, with the methylated strand serving as template for correction. Mismatch repair also occurs on heteroduplex in which neither strand is methylated, but in this case correction shows little strand preference. Regions of DNA in which d(GATC) sequences are fully adenine-methylated are refractory to mismatch repair (2,9). It appears to be the transient undermethylation of newly synthesized d(GATC) sequences in the region immediately following the replication fork that allows mismatch repair to operate only on newly synthesized strands and thereby remove replication errors (9,10).
The E. coli mismatch repair system does not recognize and repair all base-base mismatches with equal efficiency. Generally, the transition G-T mispair is corrected most efficiently, with A-C, C-T, A-A, T-T, G-G, and A-G are repaired at different efficiency. C-C is refractory to the repair (2,11).
The mutator effects observed in E. coli mutH, mutL, mutS, and mutU mismatch repair-deficient mutants, are primarily transition and frameshift mutations (12). The fact that mutants deficient in mismatch repair show increased frequencies of frameshift mutations suggests that the E. coli mismatch repair system can recognize and repair heteroduplexes with one or more unpaired bases. Transfections of E. coli with artificially constructed heteroduplexes and in vitro assays have demonstrated that the different heterologies are subject to correction with different efficiencies. The methyl-directed repair of heteroduplex with one-, two-, and three-base deletions is as efficient as the repair of G-T mismatches (13)(14)(15). Heteroduplexes with a four-base deletion are marginally repaired, and DNA with a five-base deletion is not detectably repaired by the MutHLS system (15,16), but an alternative pathway such as recF-dependent activity may repair large deletions (17). The elements of heterologous structure that are recognized by mismatch binding proteins and features of the repair system that determine repair efficiency are not understood.
Mismatch repair genes are conserved in bacteria and higher organisms. Yeast MSH2 and human hMSH2 are analogous to the E. coli MutS protein, whereas yeast MLH1 and PMS1 and human hMLH1, hPMS1, and hPMS2 correspond to the bacterial MutL protein (reviewed in Ref. 18). A number of studies suggest a functional similarity between the postreplication repair pathway in prokaryotes and eukaryotes (reviewed in Ref. 1). Recently, Umar et al. (19) reported the correction of loops of five or more unpaired bases by human cell extracts. They showed that the repair of 1-4-base loops was strongly dependent on hMLH1 and hMSH2 (19). However, in 5-16-base loops, repair was independent of hMLH1 (19). Human MSH2 had shown specific binding to loops as large as 14 nucleotides (20), although repair of loops larger than five bases has not been tested in the hMSH2-deficient extracts.
The E. coli methyl-directed repair of heteroduplexes containing more than a three-base insertion/deletion has not yet been examined in vitro. Thus, we used the in vitro repair assay and a set of M13mp18 derivatives containing 1-8 or 22 unpaired bases as substrates to determine the size constraints of the repair pathway. We demonstrate here that the methyl-directed mismatch repair pathway efficiently corrects heteroduplexes containing up to five unpaired bases. We also show that loop repair efficiency of heteroduplex is affected by base composition of the heterology.
Construction of M13 Mutants-A Synthetic 22-base pair oligonucleotide linker was inserted into the HindIII cleavage site of M13mp18 (Fig.  1). The product of this construction is dubbed M13LR1, which was further mutagenized with oligonucleotides partially complementary to the inserted linker (22) to create other derivatives. Each of the oligonucleotides used carried 1-8 additional bases, disrupting the recognition sequence of a unique restriction endonuclease site in vector and creating a new, unique restriction endonuclease recognition sequence (see Table I). Mutant M13LR phages were identified by restriction analysis of replication form minipreparation, and mutant sequences were confirmed by dideoxy sequencing (23).
Construction of Heteroduplex DNA-Replicative form DNAs fully methylated at d(GATC) sites were isolated from strain NM522 or JJ119. DNAs devoid of dam methylation at d(GATC) sequences were prepared by using strain RS5033 as host. Heteroduplex DNA substrates were constructed essentially under the conditions of Lu et al. (9). M13LR derivative RF DNA (1 mg) was linearized with BglI and mixed with a 4-fold molar excess of viral DNA, followed by alkaline denaturation and annealing. After isolation of double-stranded DNAs by hydroxylapatite chromatography, linear homoduplex DNA was removed by hydrolyzing with ATP-dependent DNase as described (24). The open circular heteroduplex was purified by Sephadex G-200 (Phamacia Biotech Inc.) chromatography and benzoylated naphthylated DEAE-cellulose (Sigma) chromatography in 10 mM Tris-HCl (pH 7.6), 1 M NaCl, 1 mM EDTA. The strand break was subjected to closure with E. coli DNA ligase in the presence of ethidium bromide (96 mmol of dye/mol of nucleotide). Covalently closed DNA circles were then isolated by equilibrium centrifugation in CsCl/ethidium bromide (9).
Substrates used in this study are summarized in Table II. By pairing different M13LR insertion derivatives, a heteroduplex containing basebase mismatch or a site-specific nucleotide insertion/deletion can be constructed. Unpaired extra nucleotides can be located at either the viral or complementary strand.
Hemimethylated heteroduplexes prepared by using methylated RF and unmethylated viral strands (VϪ/Cϩ) were resistant to cleavage by DpnII, indicating that all d(GATC) sites were in the hemimethylated state. In contrast, substrates prepared from unmethylated RF and methylated viral strands (Vϩ/CϪ) were subject to some cleavage by DpnII. Only 25% of the heteroduplexes were resistant to DpnII, 25% were cleaved once, and 50% were cleaved more than once. Fully unmethylated heteroduplexes (VϪ/CϪ) were formed by annealing unmethylated complementary strands to unmethylated viral strands. Fully methylated heteroduplexes (Vϩ/Cϩ) were prepared by in vitro methylation of hemimethylated heteroduplexes with dam methylase as described (9,15). In Vϩ/Cϩ heteroduplexes, the complementary strands were fully methylated at more than 95% of the d(GATC) sequences, and the viral strands were methylated at about 80%.
Repair Assays-Growth of cells and preparation of cell extracts were as described by Lu et al. (9). The repair reaction in concentrated E. coli lysate was carried out in 10 l containing 0.02 M Tris-HCl (pH 7.6); 5 mM MgCl 2 ; 50 g/ml bovine serum albumin; 1 mM ATP; 0.1 mM concentration each of dATP, dGTP, dTTP, and dCTP; and 0.1 g (20 fmol) of heteroduplex DNA. Cell-free E. coli extracts were included at optimal amounts at 7.5-10 mg of protein/ml. Incubation was at 37°C for 1 h. Then 30 l of 25 mM EDTA (pH 8.0) was added, and the DNA was purified by phenol extraction and ethanol precipitation. The DNA was then analyzed by restriction endonuclease digestion and agarose gel electrophoresis. The ethidium complexes of DNA products were quantitated using a gel documentation CCD camera (UVP Ltd.) (25).

Construction of Heteroduplexes in Which the Small DNA Heterologies Reside in a Similar Sequence Environment-
Starting from phage M13mp18, we have prepared a set of M13LR derivatives that contain extra nucleotides within the polylinker region located between the single EcoRI and HindIII sites of M13mp18 ( Fig. 1 and Table I). This set of M13LR derivatives permits construction of heteroduplexes representing base pair mismatches and 1-8-and 22-base insertion/ deletion heterologies. In each of these heteroduplexes the heterology is located in a similar environment. Moreover, as shown in Table II, each insertion/deletion mispair is located within overlapping restriction endonuclease recognition sites. This approach allows an independent assay of correction on either DNA strand. Digestion of the heteroduplex DNA with AlwNI ( Fig. 1) and the indicator restriction endonuclease, whose recognition site is inactive because of the heterology, will yield a 7.2-kilobase pair fragment only. Similar digestion of DNA in which the recognition sequence has been restored by repair reaction will yield 4.1-and 3.1-kilobase pair fragments. In the case of the heteroduplexes constructed, almost all substrates were refractory to the digestion by the indicator restriction endonucleases (Fig. 2, B and D), except C2cc and C3 were susceptible to the star activity by one of the indicator restriction endonuclease HindIII. In these cases, 20 -40% of untreated C2cc and C3 heteroduplexes were cleaved by HindIII. Thus, the repair on the complementary strand of these two substrates was not determined (Table III).
Methyl-directed Repair of Small DNA Heterologies by MutHLS System in Vitro-To determine whether base inser-tions/deletions were corrected via the methyl-directed mismatch repair pathway in vitro, hemimethylated heteroduplexes containing 1-8-, or 22-base insertion/deletion heterologies were tested for repair in a cell extract from mismatch repair-proficient E. coli strain. A G-T mismatch and an A-C mismatch at position 6303 of M13LR1 were used as positive controls. We choose to compare the efficiency of correction of DNA containing nucleotide heterologies with that of G-T and A-C mismatches, since a methyl-directed repair system has been shown to correct them with high efficiency (11). Table III compares the efficiency of correction of all heteroduplexes as scored by restriction endonuclease digestion, and Fig. 2 illustrates the behavior of the several repair classes in the restric- V 5Ј-AGCTAGCAGCAGCAGCCTCGAG TCGTCGTCGTCGGAGCTCTCGA-5Ј C With M13LR1 as a precursor, a set of additional M13LR phages has been constructed for this study. The underlined nucleotide in the column of mutagenic oligonucleotides indicates the base change for each step. An asterisk indicates phage M13LR6 was mutagenized by the same oligonucleotide as in M13LR7; however, a single C to A* base change was found.
AGCT--CGAGGCTGCTGCTGCT XhoI   TABLE II  M13LR heteroduplex substrates Covalently closed, circular heteroduplexes containing a set of base pair mismatches or small nucleotide insertions/deletions were prepared using the phage DNAs shown. Heteroduplexes with unpaired bases are depicted by C or V, the complementary or viral strand where extra bases are located, followed by the number of unpaired bases. The aa, gg, tt, cc, ac, or gt of C2 or V2 substrates indicates the base composition of the unpaired bases. The n of 4-or 5-base insertion/deletion substrates indicates the marker restriction enzyme NcoI. c, complementary strand; v, viral strand. Relevant restriction endonuclease sequences are underlined (see Table I). The base-base mismatches and unmatched bases are shown in boldface type and hyphens, respectively.
tion assay. As shown in Table III, members of this set were corrected in vitro with different efficiencies. Hemimethylated heteroduplexes of G-T, A-C, and one-, two-, and three-base insertions/deletions were subject to efficient repair in cell-free extracts of E. coli (Table III; Fig. 2A, lanes 1-4). As judged by repair levels, the G-T and A-C base pair mismatches were preferred substrates, with an efficiency comparable with that observed in previous studies (9,11). Heteroduplexes of one-, two-, and three-base insertion/deletion corrected at levels of 50 -100% of that observed for the G-T mispair. Repair of hemimethylated heteroduplexes occurred preferentially on the unmethylated strand. As shown in Fig. 2 and Table III, repair bias exceeded 7:1 (unmethylated:methylated) in all cases of preferred substrates. This finding is similar to previous in vivo (15) and in vitro (14) results. Previous studies examined the repair of heteroduplexes containing more than four extra bases have only been performed by transfection assays with a single set of heteroduplexes (15). Consequently, several heteroduplex substrates containing at least four extra unmatched bases were constructed and tested for in vitro methyl-directed repair. As shown in Table III, substrates containing four-or five-base insertion/deletion were corrected at reduced levels, 30 -70% of the correction observed for G-T mispair, depending on the sequence environment of the loops. As sizes of insertion/deletion increased to six bases and more, the correction level further decreased to 5-50% that of G-T mispair, and showed much less methylation-dependent strand bias (Table III). The differences in repair efficiencies do not reflect the presence of inhibitors in some DNA preparations, since experiments in which each heteroduplex was com-peted with the G-T or A-C substrate resulted in the same hierarchy of correction (not shown).
The heteroduplexes containing loops were subject to kinetic analysis along with well repaired G-T heteroduplex. The methyl-directed reactions of all substrates reached 50% of final repair levels at 15 min of incubation (data not shown), which is consistent with previous in vitro methyl-directed reactions (8).
A comparison of the in vitro repair of C-and V-heteroduplexes (with the former containing nucleotide deletions and the latter containing nucleotide insertions on the unmethylated viral strand) is shown in Table III and Fig. 3. The results demonstrated that repair of both heteroduplex configurations was preferentially directed to the unmethylated strand, indicating that both deletions and insertions were corrected in a methyl-directed fashion.
Methyl-directed mismatch repair is dependent on the product of mutH, mutL, and mutS genes. In addition, MutHLS have been implicated in the repair of one-, two-, and three-base deletion heteroduplexes (14,15). Consequently, cell extracts were prepared from E. coli strains that were defective in each of these gene products and were assayed with small nucleotide insertion/deletion heteroduplexes described above. As shown in Table III and Fig. 3, methyl-directed correction of small nucleotide heterologies was highly dependent on the presence of the mutS gene product in crude extracts. As shown in Table IV, repair of these small nucleotide heterologies also required the presence of mutH, mutL, and uvrD gene products.
Efficiency of in Vitro Correction Depends on the Sequence Composition of Heterology-To determine whether the repair of small nucleotide heterologies depends on the nature of mis-  (Table II) and then subjected to agarose gel electrophoresis to score heteroduplex correction occurring on each DNA strand. A, repair on unmethylated strand. B, untreated heteroduplex substrates hydrolyzed with diagnostic restriction endonucleases as in A. C, repair on methylated complementary strand. D, untreated heteroduplex substrates hydrolyzed with diagnostic restriction endonucleases as in C. The bar pointing to the 7.2-kilobase fragment represents unrepaired substrate, and bars to 4.1-and 3.1-kilobase fragments indicate corrected products. The minor mobility variation of the substrate and product bands from one lane to the next was due to the different optimum ion strength in each restriction endonuclease reaction.  (Table III). The only difference, other than identities of the insertion/deletion is the absence of an A-T base pair adjacent to the bulky heterologies of C2ac and V2gt. Nevertheless, these heteroduplexes were subject to differential repair in E. coli extracts. Analysis of individual heteroduplexes in Table III suggested that V2gg and C2cc are better substrates (with repair from 12.6 to 13.5 fmol/h) than V2gt and C2ca (with repair from 9.0 to 9.4 fmol/h). Given the highly homologous nature of this set of substrates, their differential repair implies that heterology recognition is associated with the correction events as observed in base pair mismatch repair.

Low Level in Vitro Correction of Large DNA Heterologies Is MutHLS-and Methylation
State-dependent-Although functions derived from mutH, mutL, mutS, and uvrD are deficient in heteroduplex correction in the in vitro system, activity is recovered upon mixing of extracts. Results of this type of analysis are shown in Table IV. With well repaired G-T mispair and V2gg dinucleotide insertion heteroduplex, complementation was observed with all possible pairs of extracts and in many cases resulted in a level of repair comparable with that observed with extracts derived from wild-type cells. As in the case of wild type reactions, the repair in mixed extracts was also dependent on the state of DNA methylation, repair being highly biased to unmethylated strands (Table IV). Interestingly, while the poorly repaired C8 deletion and V22 insertion heteroduplex were subject to this type of analysis, a low level of complementation was also observed. Methyl-directed reactions in combined mutant extracts were 2-4-fold higher than reactions in individual extracts, while reaction levels on the methylated strand remained unchanged (Table IV).
With increasing length of heterology, we have occasionally observed low levels of reaction on heteroduplexes that are independent of MutS and methylation state (Fig. 2C, lanes 7 and 9; Table III, V5, V6, V8, C22, and V22). Repair in these cases may be mediated by alternative pathway such as the methyl-independent pathway described by Fishel et al. (26). To ascertain that part of large loop repair was truly methyl-directed, we tested a subset of heteroduplex substrates in all four possible methylation states (VϪ/Cϩ, Vϩ/Cϩ, Vϩ/CϪ, VϪ/CϪ). The combined extracts of mutH471::Tn5 and mutS201:Tn5 were used, since the extracts showed higher repair activity (Table IV). Results of this type of analysis are shown in Table  V.
Repair of either hemimethylated configuration in extracts was directed to the unmethylated strand with different biases. For VϪ/Cϩ heteroduplexes, the biases were Ͼ20:1 for V2gg, 15:1 for V5, and 2.5:1 for C22. In the case of Vϩ/CϪ heteroduplexes, repair was also biased to the unmethylated strand but with reduced values of 2.8:1 for V2gg, 1.3:1 for V5, and 2:1 for C22. This may be due in part to the fact that methylation on viral strands of Vϩ/CϪ heteroduplexes was incomplete (see "Experimental Procedures").
The Dam-dependent repair pathway loses its ability to discriminate the wild-type strand from the deletion strand when DNA is devoid of adenine methylation in d(GATC) sequences. Table V shows the results of reactions with unmethylated heteroduplexes. The results of VϪ/CϪ heteroduplexes repair were consistent with the loss of strand bias, since reaction gave rise to approximately equal levels of repair on both strands.
Since methylated d(GATC) sequences are resistant to nicking by MutH endonuclease, correction of mismatched bases and insertion and deletion heterologies should be greatly reduced. The results of Vϩ/Cϩ heteroduplex in Table V are consistent with this prediction. The repair levels on methylated complementary strands were reduced to background levels. However, the reduced repair levels on methylated viral strands were 2-5-fold higher than background levels. The basis for this difference may be incomplete methylation on viral strands   of Vϩ/Cϩ heteroduplex substrates (see "Experimental Procedures").

DISCUSSION
Genetic studies provided initial evidence supporting the correction of small insertion and deletion heterologies by E. coli mismatch repair. Dohet et al. (13) demonstrated that one-base insertions and deletions are efficiently repaired in vivo (13). Learn and Grafstrom, using an in vitro system, have demonstrated that one-, two-, and three-base deletions can be repaired as efficiently as G-T mismatches in a mutHLS-dependent, methyl-directed process (14). Parker and Marinus (15) and Carraway and Marinus (16) using in vivo assay found that heterologies larger than four bases were not recognized and processed by Dam-dependent mismatch repair. Mismatch repair systems in other cell types have also been implicated in the repair of small insertion/deletion mutations. The Hex system of Streptococcus pneumoniae efficiently corrects one-and twobase heterologies (27), while mismatch repair in Saccharomyces cerevisiae corrects one-base deletion in vivo (28,29) and corrects four-and seven-base heterologies in vitro (30). The human strand break-directed mismatch repair system also repairs one-(31), two-, three-, and four-base insertions/deletions efficiently in vitro (32).
The results present in this report confirm the earlier findings in in vivo and in vitro studies that the methyl-directed DNA repair pathway in repair of 1-3-base insertions is as efficient as base pair mismatches. We further expand the study to determine the maximum size of the insertion or deletion that is repaired by the MutHLS pathway in vitro. Our results showing that insertions/deletions up to four or five bases were subject to methyl-directed repair at levels of 35-70% of that observed for G-T mispair contradict an earlier report showing that heterologies larger than four bases were not repaired (15). To reconcile the differences between their results and ours, we propose that the heteroduplex sequences within overlapping restriction endonuclease recognition sites adopt configurations suitable for repair, whereas the insertions of previous studies do not. Previous studies used only a single set of four-and five-base deletion heteroduplexes (15). We tested several four-or fivebase insertion/deletion heteroduplexes, and the repair levels showed as much as a 2-fold difference (Table III). Thus, the different conclusions of the studies could be explained by sample size differences.
As shown in Fig. 3 and Table III, as the base number of the insertion/deletion increased, generally the repair levels decreased. However, this tendency is not tightly followed by all heterologies; some exceptions do exist. The hemimethylated seven-base insertion heteroduplex was significantly repaired by the methyl-directed mismatch repair pathway (see V7 in Table III). It had been suggested that larger heterology might induce a secondary structure within the loop that recognized by MutS (15). The other possibility is that the realignment of the loop in the duplex may produce transient base pair mismatch. This transient mismatch may be recognized by MutS, and it provoke repair reaction. This is consistent with the finding that low level repair of large loop is MutHLS-dependent and methyl-directed (Tables IV and V). Thus, the MutHLS system may process large heterologies through partial mismatch recognition as well as a co-repair mechanism (16).
With increasing length of heterology, our system detected a low level of Dam-independent reactions on heteroduplexes (Table III). This activity may be mediated by alternative pathway such as methyl-independent RecF pathway (26). The efficiency of correction of these substrates is so near background levels that we have difficulty in determining the exact contributions of Dam-dependent and Dam-independent activities in these large loop reactions.
The MutS protein has been shown to bind DNA at the site of mispaired bases (4,11). Since crystallographic structure of MutS bind mismatched DNAs is not available at present, the elements of heterologous structure that the repair system recognizes and that subsequently determine repair efficiency are not understood. Since base-base mismatches such as T-G, G-G, C-A, A-A, and A-G can assume intrahelical conformations and since several studies suggest that it is this conformation that is recognized (33-35), the enzymatic system responsible for correction must be capable of detecting subtle perturbations in helix structure associated with the presence of the different mispairs. Heteroduplexes with large unpaired nucleotide sequences may form extrahelical loops. Such loop structure may not be readily recognized by MutS. Pursuit of this line of reasoning would imply that a small insertion/deletion accommodated into the double-helix structure can be recognized by MutS. It remains to be seen whether or not this is the case. If it is, then this could be the basis of specificity for the repair of small heterologies. It also can be expected that neighboring sequences would affect the distortion caused by an insertion/ deletion mismatch. Griffith and colleagues (36,37) had shown kinking of heteroduplexes with looped mispair, presumably due to intercalation of extra nucleotides into duplex. Rosen et al. (38,39) had shown duplex containing an extra 1-3 unpaired bases adopt intrahelical conformations, stacking within the duplex. Heteroduplexes can easily accommodate the bulge loop with little structural perturbation beyond the immediate vicinity of the loop itself (38,39). Therefore, it is conceivable that partially stacked bases would still be accessible to bases of other strand to form mismatched base pairs.
Mutations of the E. coli mutS and mutL genes as well as their analogs in yeast, mammalian, and human cell lines result in several hundred-fold enhancement in the frequency of mutations. Our demonstration that the bacterial mismatch repair system can repair insertion or deletion heterologies with different efficiencies suggests that eukaryotic mismatch repair proteins, which are highly homologous to the E. coli enzymes (18), may play an analogous role. Several endometrial and colorectal carcinoma cell lines that are defective in mismatch repair show instability of simple DNA microsatellites (32,40). Since most of the genetic instability found in human diseases is at microsatellite tandem repeats, it is very interesting to know the repair efficiency for small heterologies within tandem repeats. The in vitro assay described here can be modified to test the repair of slippage in simple repeat heterology. The present study provides a basis for further investigation.