INTRODUCTION

Instability of simple repetitive DNA sequences (microsatellites) has been observed in sporadic tumors of the colon, stomach, pancreas, bladder, hematopoietic system, lung, ovary, breast, and skin and in most tumors from hereditary nonpolyposis colon cancer (for reviews, see Refs. 1 and 2). One possible mechanism for the instability of simple repetitive DNA tracts is due to DNA polymerase slippage (3). In this model, the nascent and template DNA strands dissociate from each other during replication and reanneal in a misaligned configuration, forming a DNA intermediate containing a loop with unpaired repeating units or a hairpin if the looped-out nucleotides are complementary. These secondary DNA structures are referred to as insertion/deletion mismatches (IDM),¹ depending on whether the loop occurs on the nascent or template DNA strand, respectively. If left unrepaired, these IDM structures will be fixed into the genome in subsequent rounds of replication, resulting in the expansion or contraction of repeating DNA units.

Defects in DNA mismatch repair in eukaryotes were shown to result in an increase in the instability of repetitive DNA sequences (4). These findings led to the identification of human MutS and MutL homologues and the genetic basis for hereditary nonpolyposis colon cancer (1, 2); mutations in any of the four known human mismatch repair genes, hMSH2, hMLH1, hPMS1, and hPMS2, are associated with most of these cancers. Similar to the bacterial MutS protein, hMSH2 protein recognizes and binds to DNA containing mismatches (5). In addition, hMSH2 protein can form stable complexes with IDM DNA structures containing up to 14 unpaired nucleotides, suggesting its involvement in the repair of these secondary DNA structures (6). However, crude extracts from certain cancer cell lines defective in mismatch repair still retain the ability to repair heteroduplexes containing 5, 8, or 16 unpaired bases (7). These data suggest that novel mechanisms other than the MutS/MutL-dependent repair system are present for the repair of heteroduplexes containing IDM structures. This is further indicated by the fact that a large proportion of the sporadic tumor cell lines exhibiting microsatellite instability do not appear to have mutations in the mismatch repair genes (8). Recently, a novel activity that binds to DNA insertion/deletion mismatches was identified in Saccharomyces cerevisiae, suggesting the existence of a novel IDM repair pathway other than the MutS/MutL-dependent pathway (9). Furthermore, genetic studies suggest that S. cerevisiae RTH1 5- to 3-exonuclease (10) and exonuclease I from Schizosaccharomyces pombe (11) are also involved in mismatch repair and contribute to the maintenance of microsatellite stability.

Deoxyinosine 3-endonuclease (dI Endo) is a deoxyinosine-specific endonuclease from Escherichia coli. In addition to deoxyinosine, the enzyme also recognizes DNA containing an AP site, urea residue, and mismatched base pairs and cleaves the DNA containing the lesion in a strand-specific manner (12, 13, 14). The gene coding for the enzyme was found to be encoded by the same gene for endonuclease V (nfi),² which was found to be identical to orf 225 coding for a putative protein of unknown function.³ nfi (orf 225) is located at 90 min of E. coli genome (GenBank^TM accession number U00006[GenBank], Ref. 15). nfi was amplified by polymerase chain reaction and cloned into the expression vector pET 22b(+), resulting in plasmid pETI-2.³ Therefore, a large quantity of the homogeneous enzyme can be purified from bacteria BL21(DE3) harboring the plasmid pETI-2 after induction with isopropyl-1-thio--D-galactopyranoside. Its ability to cleave DNA containing base mismatches prompted us to find out whether the enzyme also cleaves DNA containing insertion/deletion mismatches as well as other DNA secondary structures.

MATERIALS AND METHODS

Single-stranded oligodeoxynucleotides (Operon, Alameda, CA) were purified by electrophoresis on a 20% polyacrylamide gel and were eluted from the gels by an IBI electroeluter. Oligodeoxynucleotides were 5-end-labeled with [-³²P]ATP (Amersham International) using T4 polynucleotide kinase or 3-end-labeled with [-³²P]cordycepin 5-triphosphate (DuPont NEN) using deoxynucleotidyl terminal transferase following instructions from the enzyme supplier (U. S. Biochemical Corp.). The labeled oligodeoxynucleotides were annealed to the appropriate complementary strands at 1:1 ratio in 10 mM Tris-HCl, pH 7.5, 0.1 M NaCl, and 2 mM 2-mercaptoethanol by heating the mixture to 90 °C and cooling down gradually to room temperature. The resulting heteroduplexes were used in endonuclease assays.

Purification of Deoxyinosine 3-Endonuclease

To purify deoxyinosine 3-endonuclease, 4 liters of BL21(DE3) haboring pETI-2² were grown in Luria Broth containing 50 µg/ml ampicillin to an absorbance at 600 nm of 0.4. Isopropyl-1-thio--D-galactopyranoside was then added to the culture at a final concentration of 1 mM, followed by further incubation at 37 °C for 3 h. Bacteria were harvested by centrifugation and mixed with 30 ml of buffer B (10 mM Hepes-NaOH, pH 7.4, 1 mM EDTA, 10 mM 2-mercaptoethanol, 1 M NaCl). An equal volume of 0.10-0.11-mm glass beads was added to the cell suspension and homogenized in a Braun MSK homogenizer for 3 min with CO₂ cooling. The glass beads and cell debris were removed by centrifugation at 5,000 rpm for 20 min (Sorvall GS3 rotor) at 4 °C. The glass beads and debris were re-extracted with 30 ml of the same buffer, and the supernatant was collected and pooled. To the pooled supernatant, an equal volume of precipitation buffer (buffer B with 30% PEG-8000) was added and stirred gently with a magnetic stirrer at 4 °C overnight. The mixture was centrifuged at 5,000 rpm for 20 min (Sorvall GS3 rotor) at 4 °C. The supernatant (120 ml) was designated as Fraction I. Fraction I was diluted 10-fold with buffer A (10 mM Hepes-NaOH, pH 7.4, 1 mM EDTA, 10 mM 2-mercaptoethanol) and loaded onto 50 ml of SP-Sepharose Fast Flow (Pharmacia Biotech Inc.) by gravity at a flow rate of 2 ml/min. The column was washed with 200 ml of buffer A and was eluted with a linear gradient (480 ml) from 0.1 M to 0.6 M NaCl in buffer A. Eight ml of each fraction was collected, and the peak fractions containing the 25-kDa protein were pooled as Fraction II. Fraction II was diluted with buffer C (10 mM Tris-HCl, pH 7.4, 10 mM 2-mercaptoethanol) until the concentration of NaCl was about 0.02 M as measured by conductivity and loaded onto a 50-ml Q-Sepharose Fast Flow column (Pharmacia) by gravity at a flow rate of 2 ml/min. The column was then washed with 200 ml of buffer C and eluted with a linear gradient (480 ml) from buffer C to buffer C containing 0.3 M NaCl. The enzyme preparation after Q-Sepharose Fast Flow was apparently homogeneous as judged from silver-stained SDS-PAGE gel. The fractions were pooled and dialyzed against buffer containing 10 mM Tris-HCl, pH 7.5, 0.15 M NaCl, 0.1 mM EDTA, and 50% glycerol and stored at 20 °C for further assay.

Standard reaction contained 20 fmol of DNA substrates with either strand labeled. The reactions were incubated alone or with 1.2 ng of deoxyinosine 3-endonuclease in 10 mM Tris-HCl, pH 7.5, 0.34 mM MnCl₂ at 37 °C for 20 min. The reactions were stopped by adding an equal volume of sequencing loading buffer (90% formamide, 1 mM EDTA, 0.1% xylene cyanol, and 0.1% bromphenol blue) and analyzed by denaturing PAGE as described previously (12, 13, 14).

RESULTS AND DISCUSSION

The ability of dI Endo to cleave IDM-containing oligonucleotides was monitored by an endonuclease assay. The cleavage of oligonucleotide heteroduplex substrates by dI Endo produces shorter oligonucleotides that migrate faster than the parental DNA substrates in a denaturing polyacrylamide gel electrophoresis. Fig. 1 summarized the result with a heteroduplex containing a hairpin (Duplex HP). As a reference, a heteroduplex containing an A/A mismatch (Duplex A/A) with the same sequence context surrounding the lesion was also included in this study. It was previously shown (14) that the strand specificity of dI Endo on mismatch-containing DNA is governed by the 5 terminus; the enzyme cleaves the DNA strand with the mismatch closer to the 5 end. Both DNA strands are cleaved by the enzyme if the mismatch is located in the middle of the oligonucleotide duplex. Since the mismatch in Duplex A/A is at the 19th nucleotide from the 5 terminus of the top strand and the 20th nucleotide from the 5 terminus of the bottom strand, both top and bottom strands were found to be cleaved by dI Endo (Fig. 1B, lanes 1 and 2 and lanes 7 and 8, respectively). The enzyme cleaved both the top and bottom strands at two locations, one immediately 3 and the other one nucleotide 3 to the mismatch. When the top strand of Duplex A/A was 5-end-labeled, the major cleavage site (68% of total top strand cleavage) was observed at the phosphodiester bond immediately 3 to the mismatch, thus generating a 19-mer. A second cleavage site was observed at the phosphodiester bond one nucleotide 3 to the mismatch, generating a 20-mer (Fig. 1B, lane 2). However, when the bottom strand of Duplex A/A was labeled, the major cleavage site (65% of total bottom cleavage) was observed at the phosphodiester bond one nucleotide 3 to the mismatch, generating a 21-mer (Fig. 1B, lane 8). These results confirm our previous observations (14) that while deoxyinosine 3-endonuclease predominantly cleaves deoxyinosine-containing DNA one nucleotide 3 to the mismatch, additional cleavage sites by the enzyme can be observed on mismatch-containing DNA duplex depending on sequence context surrounding the mismatches.

When DNA heteroduplex containing a hairpin (Duplex HP) was used as a substrate for dI Endo, similar cleavage patterns were observed (Fig. 1B). Both DNA strands were cleaved by dI Endo, with a relative cleavage rate of 1:3 for top strand:bottom strand (20% and 60% of total cleavage for top and bottom strand, respectively). When the top strand of Duplex HP was 5-end-labeled, the major cleavage product migrated as a 36-mer (75% of total top strand cleavage, Fig. 1B, lanes 3 and 4), indicating that the major cleavage site was immediately 3 to the hairpin. In addition, the enzyme also cleaved the phosphodiester bond one nucleotide 3 to the hairpin, generating a minor cleavage product migrating as a 37-mer (lanes 3 and 4). When the bottom strand was 5-end-labeled, the major cleavage product observed was a 21-mer (63% of total bottom strand cleavage, lanes 9 and 10), indicating that the cleavage site was one nucleotide 3 to the hairpin. There were two additional minor cleavages induced by the enzyme on the bottom strand, one immediately 3 (20% of total bottom strand cleavage) and the other two nucleotides 3 to the hairpin (17% of total bottom strand cleavage), generating two minor products migrating as 20- and 22-mers, respectively (lanes 9 and 10). The nuclease activity on the heteroduplex containing a hairpin was due to an endonucleolytic activity that is associated with dI Endo and not a contaminating 3- to 5-exonuclease activity that could not bypass the hairpin. This was demonstrated by using 3-end-labeled Duplex HP as a DNA substrate. Using Duplex HP containing 3-end-labeled top strand, two cleavage products were observed, migrating as 18- and 19-mers (Fig. 1B, lanes 5 and 6). When the bottom strand of Duplex HP was 3-end-labeled, three cleavage products were observed, migrating as 18-, 19-, and 20-mers (Fig. 1B, lanes 11 and 12). The cleavage patterns generated using 3-end-labeled substrates are consistent with those observed for 5-end-labeled heteroduplexes.

To show that dI Endo also cleaved DNA containing unpaired nucleotide loops, heteroduplexes containing loops of 2 to 14 nucleotides were prepared (Fig. 2A). Fig. 2B shows that dI Endo cleaved both DNA strands of loop-containing heteroduplexes. Two cleavage sites on the bottom strand were observed for all the heteroduplex DNA substrates containing loops ranging from 2 to 14 unpaired nucleotides. When the bottom strands of Duplex L2, L4, L8, and L14 were 5-end-labeled, the major cleavage product derived from each of these DNA substrates migrated as a 21-mer (by comparing the position of the cleavage products derived from Duplex A/A, Fig. 2B, lanes 11 and 12). Therefore, dI Endo predominantly cleaved the bottom strand of loop-containing heteroduplexes at the phosphodiester bond one nucleotide 3 to the loop. An additional product which migrated one nucleotide faster was also observed, indicating that the additional cleavage site was immediately 3 to the loop. Similarly, two cleavage sites were observed on the DNA strand containing the loop (the top strand, Fig. 2B, lanes 4, 6, 8, and 10). In order to determine the cleavage sites on the top strand, Duplexes L2, L4, L8, and L14 with 3-end-labeled top strand were used as DNA substrates. The major cleavage product derived from each of these DNA substrates (Fig. 2C, lanes 4, 6, 8, and 10) migrated as an 18-mer, which was one nucleotide smaller than the major product derived from Duplex HP (Fig. 1B, lane 12), indicating that the enzyme cleaved one nucleotide 3 to the unpaired loop. In addition, a minor cleavage product which migrated as a 22-mer was observed, especially with heteroduplexes containing 8 and 14 nucleotide loops (Fig. 4B, lanes 7-10), suggesting that the enzyme could also cleave the DNA two nucleotides inside the loop from the 3 duplex juncture.

To confirm that the endonucleolytic activity associated with dI Endo on duplexes containing IDM structures was not due to possible contamination in the enzyme preparation, a substrate competition assay was performed. Duplex PAL (Fig. 3A) containing a hairpin was used as a substrate for dI Endo. It is interesting to note that dI Endo is more active on the top strand of Duplex PAL (Fig. 3B) as compared to Duplex HP (Fig. 1). For Duplex PAL, the relative cleavage of top and bottom strands induced by dI Endo was 3:1, respectively (under the same reaction conditions, the top strand was cleaved 65% while the bottom strand was only cleaved 20%). Using Duplex PAL with 5-end-labeled top strand as DNA substrate, the endonucleolytic activity of dI Endo was assayed in the presence of an unlabeled I/T pair containing DNA duplex (Duplex I/T which has the same DNA sequences as Duplex A/A except that the mismatch A/A pair was replaced with an I/T pair), which is a specific substrate for dI Endo (Fig. 3C). The results were compared with those obtained with a nonspecific competitor, Duplex A/T, which has the same DNA sequences as Duplex I/T except the I/T pair was replaced with an A/T pair. Fig. 3C showed that the presence of a 4-fold excess of the I/T-containing duplex to the enzyme resulted in 50% reduction of the hairpin cleavage activity. However, in the presence of a 20-fold excess of the normal DNA duplex, the hairpin cleavage activity was inhibited only by 20%. The results further substantiated that the endonucleolytic activity on these DNA structures was due to deoxyinosine 3-endonuclease.

In the above experiments, a weak 5- to 3-exonuclease activity was also noticed, releasing 3 to 4 nucleotides from the 5 termini of these oligonucleotides. The exonuclease activity did not seem to be a contaminant since the activity co-purified with the deoxyinosine 3-endonuclease activity over a series of chromatographic columns including Mono S, Mono Q, phenyl-Superose, and Affi-Gel Blue.⁴ The presence of a weak exonuclease activity and its ability to cleave DNA containing different secondary structures led us to think that deoxyinosine 3-endonuclease might be functionally similar to yeast RTH1 class nucleases. RTH1 class nucleases are a group of 5- to 3-exonucleases which is named RTH1 nuclease in S. cerevisiae (16), flap endonuclease (FEN-1) in mouse and human cell (17, 18, 19), and maturation factor 1 and DNase IV from HeLa cells (20, 21), and 5- to 3-exo/endonuclease from calf thymus (22, 23). These nucleases shared sequence homology to the 5-exonuclease domain of DNA polymerase I of E. coli (21) and are functionally similar to the 5- to 3-exonuclease activity of E. coli DNA polymerase I in lagging strand DNA synthesis (16, 24). In addition, yeast RTH1 mutants exhibit high sensitivity to the DNA damaging agent methylmethane sulfonate (16) and increased instability of simple repetitive DNA sequences (10), suggesting that RTH1 nuclease participates in the repair of base damages and mismatch repair. In addition to the 5-3-exonuclease activity, this class of nucleases recognize flap and pseudo-Y DNA, being able to cleave the 5-unpaired tail of these DNA substrates. It is thus interesting to examine whether deoxyinosine 3-endonuclease can also cleave both the flap and pseudo-Y DNA. Fig. 4 shows that dI Endo efficiently cleaved the 5-single-stranded tail of 5-flap and pseudo-Y DNA. The cleavage site was found to be similar to that induced by E. coli DNA polymerase I, being one base inside the double strand juncture (25). Similar to RTH1 class nucleases, dI Endo cannot cleave the 3-single-stranded tail of pseudo-Y and 3-flap DNA.⁴

Our data presented here demonstrated that in addition to deoxyinosine and mismatched base pairs, deoxyinosine 3-endonuclease from E. coli recognized and cleaved DNA containing insertion/deletion mismatches, suggesting that the enzyme might have a role in the repair of these abnormal DNA structures. It was previously shown that methyl-directed mismatch repair system in E. coli recognizes poorly DNA containing loops larger than five nucleotides (26). However, whether E. coli possesses the ability to repair DNA containing larger loops is controversial (27, 28, 29). The discrepancies observed in earlier studies might be due to the assay systems employed. Most of the data on the biological processing of these abnormal DNA structures in E. coli were obtained from bacterial transformation assays using plasmids containing loops of various sizes. These systems are uncoupled from DNA replication. Since these abnormal DNA structures are generated during DNA replication by strand slippage, a recently proposed arm-directed repair model by Leach (30), which suggests that the repair of these secondary structures is directed by the replication fork, thus appeared to be quite attractive.

The cleavage efficiencies of dI Endo on these DNA structures is comparable with those observed previously for mismatches as well as deoxyinosine.⁴ These novel biochemical activities suggest that the enzyme might be involved in the repair of secondary DNA structures, especially those generated on the lagging strand during DNA replication. 1) The endonucleolytic activity of dI Endo on flap and pseudo-Y DNA structures and 5- to 3-exonuclease activity are functionally homologous to those of the RTH1 class nucleases, which are required for lagging strand DNA synthesis. 2) The mismatch and hairpin cleavage activities of the enzyme appear to track from 5 termini of heteroduplexes (14). 3) The endonucleolytic activity of the enzyme on DNA mismatches increases severalfold when a 5 nick is present, suggesting that the enzyme binds to the nick and tracks the DNA in a 5 to 3 direction.² The ability of deoxyinosine 3-endonuclease to recognize both DNA strands of IDM containing duplexes suggests that the strand specificity of the enzyme is directed by mechanisms other than the DNA structure, possibly a 5 terminus or 5 nick, similar to the case with simple base mismatches (14).² This is reasonable since strand slippage can occur on both template and nascent strand. If slippage occurs on the template strand, repair (cleavage) on the strand without the loop (nascent strand) will prevent deletion. On the other hand, if slippage occurs on the newly synthesized strand, repair (cleavage) of the loop in nascent strand will prevent expansion. Therefore, in order to prevent deletion or expansion due to polymerase slippage, a strand-directed mechanism must be able to recognize the nascent strand and correction (cleavage) be made only on the nascent strand. It has been shown previously that a nick can direct a MutS independent mismatch repair in E. coli (31) and is probably the mechanism used in mammalian cells for strand-specific mismatch repair (32, 33). Since deoxyinosine 3-endonuclease tracks DNA in a 5 to 3 direction from a 5 terminus or a 5 nick and induces cleavage 3 to the mismatch or IDM structures, the enzyme thus fulfills the requirements for a factor that can initiate the repair of mismatches or IDM structures, especially those generated on the lagging strand during DNA synthesis, since there is always a 5 nick or a 5 terminus on the lagging strand. It is interesting to note that the enzyme consists of only 223 amino acid residues with a molecular mass of 24.9 kDa, yet it possesses complicated enzyme activities for DNA metabolism. If present, the eukaryotic/human counterpart of dI Endo may play an important role in maintaining genome stability which has been shown to be associated with many different human diseases and cancers.