Nick-dependent and -independent Processing of Large DNA Loops in Human Cells*

DNA loop heterologies are products of normal DNA metabolism and can lead to severe genomic instability if unrepaired. To understand how human cells process DNA loop structures, a set of circular heteroduplexes containing a 30-nucleotide loop were constructed and tested for repair in vitro by human cell nuclear extracts. We demonstrate here that, in addition to the previously identified 5′ nick-directed loop repair pathway (Littman, S. J., Fang, W. H., and Modrich, P. (1999) J. Biol. Chem. 274, 7474–7481), human cells can process large DNA loop heterologies in a loop-directed manner. The loop-directed repair specifically removes the loop structure and occurs only in the looped strand, and appears to require limited DNA synthesis. Like the nick-directed loop repair, the loop-directed repair is independent of many known DNA repair pathways, including DNA mismatch repair and nucleotide excision repair. In addition, our data also suggest that an aphidicolin-sensitive DNA polymerase is involved in the excision step of the nick-directed loop repair pathway.

DNA loop heterologies are unpaired, single-stranded DNA structures that can be generated during DNA metabolism. These structures reflect a form of genetic instability and are considered as an early phenotype in carcinogenesis (1). DNA loops can range in size from a single nucleotide to several thousand nucleotides (nt). 1 Smaller loops (Ͻ20 nt) are generally formed during replication of repetitive DNA sequences (2)(3)(4), and larger loops can arise during recombination events between divergent sequences (reviewed in Ref. 5).
It has been demonstrated that eukaryotic cells are capable of processing DNA loops ranging from 1 to 5,600 nt in length (6 -13). There is accumulating evidence, however, suggesting that eukaryotic cells possess multiple pathways to process these loops, including both the mismatch repair (MMR)-dependent and -independent pathways (6, 11, 12, 14 -20). This redundancy of pathways for loop processing has been recently shown to apply to DNA loops less than 16 nt, but not to those larger than 16 nt. The former can be repaired by both the MMR-dependent and -independent pathway, and the latter can only be processed by the MMR-independent pathway(s) (20). The limit of MMR-dependent loop repair at 16 nt in length serves as a criterion to classify DNA loop heterologies into small loops (Յ16 nt) and large loops (Ն17 nt) (20). For small loops, the MMR-dependent pathway appears to utilize the same general mechanism to process these heterologies, i.e. excision is conducted by exonucleases from the pre-existing strand break to the heterology no matter whether the strand break is located 5Ј or 3Ј to the loop. However, small DNA loops are repaired differently by the MMR-independent pathway(s) (14 -16, 20). Although the processing of small DNA loops with a strand break 5Ј to the heterology occurs by a manner similar to that seen for MMR-dependent processing (i.e. involving exonucleases), the processing of looped heteroduplexes containing a 3Ј strand break seems to involve endonuclease(s) that remove the loop directly, without excision occurring from the nick (20).
Very little is known about the requirements and mechanism of large loop (Ͼ17 nt) processing by the MMR-independent pathway in eukaryotes. Previous studies have indicated that the human MMR-independent pathway(s) can process large loops containing a strand break 5Ј, but not 3Ј to the heterology (14). On the other hand, a very weak 3Ј-directed large loop repair activity has been identified in yeast (16). However, the mechanisms by which these large loop repair pathways operate are largely unknown. To understand large DNA loop repair in human cells, we tested the repair of a series of DNA loop heteroduplexes containing a 30-nt loop and a strand break either 5Ј or 3Ј to the heterology by human nuclear extracts. We demonstrate that there are at least two pathways for large loop repair in human cells, one of which has been previously demonstrated (14). In this work, we identify a human large loop repair activity that specifically removes the loop in a manner independent of a nick, which we designate as the loop-directed pathway. The loop-directed pathway is independent of several major DNA repair pathways, including MMR, nucleotide excision repair (NER), and the Werner syndrome protein WRN. We also show that the two large loop repair pathways have different underlying mechanisms. cultures (HCT116, HCT15, HEC-1A, and AG08802) were harvested from 20 to 30 roller bottles and suspension cells (HeLa, NALM6, GM02345) were harvested from 6-liter cultures in spinner flasks. Nuclear extracts were prepared as previously described (21,22).
Heteroduplex Substrate DNA Construction-Bacterial phage f1MR0 was created by digestion of phage f1MR1 (23) double stranded DNA with XbaI and NheI, gel purification of the large fragment (6,423 bp), and ligation of the compatible ends. The resulting molecule contains a 30-bp deletion relative to phage f1MR24 DNA (22). The sequence of f1MR0 was verified by DNA sequencing. Creation of heteroduplex substrates was performed as previously described (21,23). Briefly, double stranded DNA from one phage (e.g. f1MR0) was linearized by Sau96I, and annealed with single stranded DNA from another phage (e.g. f1MR24). The resulting 5Ј nicked (in the complementary strand) heteroduplex was purified as described (23). To construct a looped substrate with a 3Ј nick, its corresponding 5Ј-nicked substrate was incubated with DNA ligase in the presence of ethidium bromide to form a supercoiled covalently closed circular (ccc) substrate. The later product was then incubated with glycoprotein II protein, an endonuclease that specifically nicks the viral strand at the site of the phage replication origin (24), to yield a nick 3Ј to the loop (see Fig. 1). The nomenclature of substrates follows the format: (nick position) Ϫ (loop size)(loop strand). For example, 5Ј-30V describes a substrate with a 5Ј nick and 30-nt loop in the viral (V) strand (see Fig. 1). 5Ј nicks are located 115 bp away from the loop site on the complementary (C) strand. 3Ј nicks are located 175 bp away from the loop site on the V strand.
Loop Repair Assays-Unless mentioned otherwise, large loop repair was assayed by Southern blot analysis as described (20). Briefly, 100 ng (24 fmol) of DNA substrate was incubated with 75 g of nuclear extracts in 15-l reactions containing 20 mM Tris-HCl (pH 7.6), 110 mM KCl, 5 mM MgCl 2 , 1.5 mM ATP, 1 mM glutathione, and 0.1 mM each of the four dNTPs. Reactions were incubated for 15 min at 37°C and stopped by protease K digestion (30 g/ml) for 15 min at 37°C. DNA was isolated by phenol extraction and ethanol precipitation, and was then digested with SspI and BanII. The resulting products were separated on 6% denaturing polyacrylamide gels (7 M urea, 19:1 acrylamide:bis-acrylamide, 1ϫ Tris-borate buffer), followed by electrotransferring to nylon membrane at 4°C. Membranes were sequentially probed with a 32 Pend-labeled oligonucleotide 20-mer V5216 -5235 (5Ј-ATTGTTCTG-GATATTACCAG-3Ј) and a 32 P-end-labeled 25-mer C5235-5259 (5Ј-GAAGAACTCAAACTATCGGCCTTGC-3Ј) to visualize the repair in the C and V strands, respectively. Probe V5216 -5235 hybridizes to the C strand, whereas probe C5235-5259 hybridizes to the V strand. Autoradiographs were scanned by Kodak Image 2.0.2 software and band intensities were measured. Nonspecific probe binding was corrected for by subtracting "repair" values observed in substrate only or heat-inactivated extract controls. Variation between experiments was generally less than 10%. Where indicated, aphidicolin was added to a final concentration of 0.1 mM and each of the four dideoxynucleotriphosphates (ddNTPs) was added to a final concentration of 0.1 mM each (equimolar with dNTPs in the reaction buffer).
Reaction Intermediate Analysis-Reaction intermediates were trapped during in vitro repair assays by the omission of exogenous dNTPs, the addition of aphidicolin, or the addition of ddNTPs. Purified reaction products were digested with SspI, separated on 6% denaturing polyacrylamide gels, and electrotransferred onto a nylon membrane as described (20). The C strand was probed on the 3Ј end of the SspI fragment (see Fig. 1) using oligonucleotide V5216 -5235 (containing a V strand sequence, but hybridizing to the C strand), and the V strand was probed on the 5Ј end of the fragment using oligonucleotide C5235-5259 (containing a C strand sequence, but hybridizing to the V strand). Size standards were created by digestion of f1MR24 double stranded DNA with SspI (754 bases), SspI and BanII (473 bases), SspI and XcmI (420 bases), or SspI and NheI (408 bases). To probe for single stranded DNA regions between the nick and the loop, repair intermediates trapped by limited DNA synthesis were alternately digested with 2 units each of BseRI and BanII (scoring enzyme). Resulting products were separated by 1% agarose gel electrophoresis and detected by ethidium bromide staining. Band intensities were analyzed by digital photography using Kodak Image 2.0.2 software.

Human Cells Possess Both Nick-dependent and -independent
Large Loop Repair Pathways-To study the repair of large DNA loops in human cells, we constructed a set of DNA substrates containing a 30-nt loop and a strand break, where the loop was placed either on the nicked strand or on the continu-ous strand and the nick was either 5Ј or 3Ј to the heterology (Fig. 1). These substrates were tested for repair by nuclear extracts derived from several human cell lines. The repair was scored using Southern blot analysis by taking advantage of the 30-nt difference in size between looped and non-looped strands. In this assay, a DNA fragment encompassing the loop region was separated by denaturing polyacrylamide gel electrophoresis, transferred to a nylon membrane, and sequentially probed for either the complementary strand or viral strand. A conversion of the strand size from a shorter strand to a longer strand or vice versa reflects the repair of the loop by using looped (30 nt longer) or non-looped (30 nt shorter) strands as the template for synthesis during the repair.
First, we determined the extent of repair to each strand for all four nicked substrates using HeLa nuclear extracts (Fig. 2). As expected, significant repair of the nicked strand was observed for substrates containing a nick (in the C strand) 5Ј to the loop. For the substrate 5Ј-30C, a new species with a size of 443-nt in length was detected in reactions containing active HeLa extract ( Fig. 2A, i, HL) that was not seen for a reaction containing heat-inactivated extract ( Fig. 2A, i, HI). In the case of the substrate 5Ј-30V, a new species 473 nt in size was observed after exposing to an active HeLa extract ( Fig. 2A, ii, HL). These conversions indicate that repair excision occurred in the nicked strand, followed by repair resynthesis using the continuous strand (the V strand) as a template. These observations, which are consistent with a previous study (14), and additional studies that we conducted (described below) suggest that this repair is largely nick-directed (i.e. excision occurs from the nick to the loop via an exonuclease). In contrast to the 5Ј substrates, our results show that only certain looped sub- strates with a 3Ј-nick can undergo HeLa extract-catalyzed repair. A significant level of repair was detected in the nicked strand of the 3Ј-30V substrate as indicated by the fact that there was a conversion of the original longer fragment (473-nt) to a new but shorter fragment (443 nt) in the reaction containing an active HeLa extract ( Fig. 2A, viii, HL). No repair of the nicked strand of the 3Ј-30C substrate was seen, however ( Fig The disparity between the results of the 3Ј-30C and 3Ј-30V substrates suggested to us that the repair observed for the 3Ј-30V substrate may not be nick-directed. Indeed, a previous study has indicated the presence of a non-nick directed, but loop-stimulated pathway for loop repair in yeast cell extracts (15). To test our hypothesis of the presence of a loop-stimulated repair pathway in human extracts that is not nick-directed, we first examined the repair that occurred on the continuous strands of 5Ј-30V and 3Ј-30C substrates. If the loop itself can provoke repair, loop removal should occur on the V strand of substrate 5Ј-30V and the C strand of 3Ј-30C. Results indicate that loop-specific removal indeed occurred in these substrates (Fig. 2, A, iii, for 3Ј-30C, and vi, for 5Ј-30V). Little (if any) repair was observed in the continuous strand containing no loops (see Fig. 2A, iv and v). It is worth mentioning that a residual 473-nt band was detected in substrate 3Ј-30V by the C strand probe ( Fig. 2A, iv, HL), which seemed to be a loop addition (12) in the continuous strand. However, based on our experience in the preparation of 3Ј substrates, which were derived from ligation of their corresponding 5Ј substrates before nicking by glycoprotein II in the viral strand (see "Experimental Procedures" for 3Ј substrate preparations), there is a very small fraction of unli-gated substrates, which still contain a 5Ј nick. Therefore, the residual 473-nt is most likely derived from a background 5Ј nick-directed repair of the unligated substrates.
To confirm the presence of a nick-independent repair pathway in human cells, we tested for loop removal of two ccc substrates that contain a 30-nt loop either in the C (ccc-30C) or V strands (ccc-30V). As shown in Fig. 2B, conversion of the longer fragment to a shorter fragment occurred in the looped strand for both substrates. As expected, little repair was detected in the non-looped strand for both ccc substrates (data not shown). These results, together with the data described above, suggest that a large loop structure in human cells can be removed in a manner independent of a strand break. Therefore, human cells appear to possess at least two large loop repair pathways: one that is directed by a 5Ј strand break, whereas the other is loop-directed loop removal and is nick-independent. In general, nicked-directed repair appears to be more efficient than looped-directed repair as the amount of the former is always higher than the latter (see Fig. 2C).
The Loop-directed Repair Pathway Is Independent of MMR, NER, and WRN Proteins-To determine whether several major DNA repair pathways are involved in large DNA loop repair, we assayed cell lines deficient in various DNA repair pathways for their ability to process the 5Ј-30V substrate, which allows for detection of both nick-dependent and -independent pathways in a single reaction. The cell lines used in this analysis are deficient in MMR (HCT116, HCT15, and NALM6), NER (GM02345 and AG08802), or Werner syndrome protein WRN (WS780). The WRN protein has a 3Ј 3 5Ј helicase activity as well as a 3Ј 3 5Ј nuclease activity, and is believed to participate FIG. 2. Visualization of loop processing reactions by human cell extracts. Loop repair assays for both nick-containing substrates and ccc substrates were performed using 75 g of nuclear extract and 100 ng (24 fmol) of looped substrate, as indicated. DNA was recovered and digested with BanII and SspI prior to electrophoresis through a 6% denaturing polyacrylamide gel. DNA was then transferred onto a nylon membrane and the membrane was hybridized with 32 P-end-labeled oligonucleotide probes as described (20). Repair in the C and V strands was detected by a C strand probe (V5216 -5235) and a V strand probe (C5235-5259), respectively. in recombination repair and double strand break repair (25,26). Repair activities were compared with the levels found in HeLa extract, which is wild type for MMR (21,27), NER (28), and WRN activity (29). Our results indicate that all cell lines tested, regardless of deficiency in MMR, NER, or WRN, showed some degree of activity for both nick-directed and nick-independent pathways (Table I, Figs. 2 and 3). For the nick-directed pathway, nuclear extracts derived from MLH1-deficient HCT116 (Table I), MSH6-deficient HCT15 (Table I), MSH2deficient NALM6 (Table I, Figs. 2 and 3), and WRN-deficient WS780 ( Table I) had levels of repair comparable with those of HeLa cells. Cells with mutations in NER genes XPA (GM02345) or XPG (AG08802) were capable of processing large DNA loops in a nick-directed manner, but the repair level was significantly lower than that in other cells (Fig. 3 and Table I).
To determine whether these cells are partially defective in this loop repair pathway, complementation assays were performed by mixing these two extracts. Although there was a 10% increase in repair, the level was still not as high as in other cells tested (data not shown). However, all extracts displayed a similar level of ability for the nick-independent repair of the substrate (Fig. 3 and Table I). These results indicate that like the nick-directed pathway, loop-directed repair is also independent of the MMR and NER pathways, as well as the WRN protein.
The Loop-directed Repair Pathway Requires Little DNA Synthesis-All known DNA repair pathways include a DNA resyn-thesis step that can be inhibited by a DNA polymerase inhibitor. To assess the impact of DNA synthesis in large loop DNA repair, repair reactions were performed either in the absence of exogenous dNTPs or in the presence of the DNA synthesis inhibitor aphidicolin (an inhibitor for pol ␣, pol ␦, and pol ⑀), or ddNTPs (chain elongation terminators). We first analyzed the requirement for DNA synthesis for the nick-dependent pathway of large DNA loop repair using substrates 5Ј-30V and 5Ј-30C. As shown in Fig. 4A, the nick-directed repair was completely blocked by any inhibitory factor regardless of whether the loop was on the nicked strand (substrate 5Ј-30C) or on the continuous strand (substrate 5Ј-30V), suggesting that like most DNA repair pathways, nick-directed large loop repair requires active DNA synthesis that may involve pol ␣, pol ␦, and/or pol ⑀. This is consistent with a previous report implicating pol ␦ in MMR (30). Similar analysis was performed to determine the impact of DNA synthesis on the nick-independent pathway using substrates 5Ј-30V and 3Ј-30V. Surprisingly, the loop removal for both substrates was not completely inhibited in all three conditions of limited DNA synthesis, although the amount of repair was reduced, with an inhibition of 17% by the omission exogenous dNTPs, 58% by aphidicolin,

FIG. 3. Both nick-directed and loop-directed large loop repair are independent of MMR and NER.
Repair assays were performed as described in reactions using 100 ng of 5Ј-30V substrate and 75 g of nuclear extract proteins, as indicated. Repair was detected by Southern blot analysis using the same probes as described in the legend to Fig. 2. HI, heat-inactivated HeLa extracts; HL, HeLa extracts; N6, MSH2deficient NALM6 extracts; GM, extracts derived from XPA-deficient GM02345 cells; and AG, extracts derived from XPG-deficient AG08802 cells.

FIG. 4.
Loop-directed repair is insensitive to limited DNA resynthesis. Repair assays were performed using conditions that limited DNA synthesis, as indicated. DNA was recovered, digested with SspI and BanII, and separated through a 6% denaturing polyacrylamide gel, followed by Southern blot analysis as described in the legend to Fig. 2. A, nick-directed repair for substrates 5Ј-30V and 5Ј-30C. Repair was detected by probe V5216 -5235 (a probe complementary to the C strand at the 3Ј end of the SspI-BanII fragment). B, loop-directed repair for substrates 3Ј-30V and 5Ј-30V. Repair was detected by probe C5259 -5235 (a probe complementary to the V strand at the 5Ј end of the SspI-BanII fragment). and 67% by ddNTPs (Fig. 4B). These results indicate that nick-independent loop removal requires a greatly reduced amount of DNA synthesis compared with the nick-dependent system.
Because DNA synthesis in DNA repair is preceded by strand excision, the limited inhibition of the loop-directed repair by DNA synthesis inhibitors also suggests a limited repair excision during this reaction. To explore this possibility, repair intermediates were monitored by Southern blot analysis as previously described (20), under the conditions of limited DNA synthesis. Fig. 5, A and B, shows the analysis of nick-directed repair for both 5Ј-30V and 5Ј-30C substrates using a probe that binds to the 3Ј end of the C strand of a fragment produced by digestion with SspI (also see Fig. 1, SspI fragment). As expected, under the normal repair conditions (lanes 1 and 5), only a band with the full size SspI fragment (the top band, 724 nt in length for lane 1, 754 nt in length for lane 5) and a band (524 nt in length for lane 1, and 554 nt in length for lane 5) corresponding to the originally nicked DNA fragment were detected. The top band in these cases represents both complete repair (excision of the loop, resynthesis, and ligation) and direct ligation of the nick prior to repair, whereas the lower band indicates the original substrate that was neither ligated, nor otherwise processed. Under the conditions of limited DNA synthesis (lanes 2-4 and 6 -8), in addition to the whole SspI fragment (724 nt in Fig. 5A and 754 nt in Fig. 5B) and the fragment containing the original nick (524 nt in Fig. 5A and 554 nt in Fig. 5B), smaller bands between 524 and ϳ400 nt were also observed, which represent the excision intermediates. Although a slightly different pattern of intermediate tracts between the nick and the loop was evident between these two substrates, prominent DNA fragments with ends that correspond very close to the loop site were clearly detected in both cases (between the 408-and 420-nt markers), particularly in reactions containing ddNTPs (lanes 4 and 8) and omitting exogenous dNTPs (lanes 2 and 6). Interestingly, at least 50% less repair intermediates were seen in the aphidicolin-containing reactions (lanes 3 and 7, discussed below). These results indicate that an extensive repair excision is associated with nick-directed loop repair and that the bulk of excision occurs between the nick and loop site. Similar experiments were performed to determine excision intermediates for the nick-independent loop repair pathway by using substrates 5Ј-30V and 3Ј-30V. The results from these experiments demonstrated that repair intermediates were difficult to detect. Even after prolonged exposure of the film, we were able to detect only very light bands centered on the loop, which had a similar intensity as many nonspecific repair bands above the loop site (Fig. 5C). Similar results were observed using the 3Ј-30V substrate (not shown). Although less effective repair by the loop-directed pathway may partially contribute to undetectable excision intermediates, the data shown in Fig. 4 suggest to us that loopdirected loop removal is less sensitive to conditions of limited DNA synthesis, which would also explain the less detectable excision intermediates observed in Fig. 5.
An Aphidicolin-sensitive DNA Polymerase Is Involved in Stimulating Repair Excision in the Nick-dependent Pathway-Our data show that an aphidicolin-sensitive DNA polymerase is apparently involved in nick-directed large loop repair, as judged by the fact that no repair products were observed in reactions containing aphidicolin (Fig. 4A). Surprisingly, much less excision intermediates were observed in aphidicolin-containing reactions (Fig. 5, lanes 3 and 7) compared with other reactions that inhibit DNA synthesis. These results suggest that an aphidicolin-sensitive DNA polymerase may be required for loop repair-associated excision. To test this possibility, repair products from reactions with limited DNA synthesis were digested with BseRI and BanII. Whereas BseRI consistently linearizes the DNA substrate (see Fig. 1), digestion of the DNA substrate by BanII depends on the availability of its recognition sequence, because the BanII recognition sequence is located in between the nick and the loop (see SspI fragment in Fig. 1) and is subjected to excision during the repair reaction. Digestion by BanII would occur if there is no excision, or excision followed by resynthesis. If excision occurs without resynthesis in the presence of DNA synthesis inhibitors, a single strand DNA gap that spans the nick and the loop along the shorter distance would be generated, which prevents BanII from being able to cut. If both enzymes are able to digest the DNA under these conditions, fragments of 3.7 and 2.7 kb will be evident, indicating that the region of DNA in between the loop and pre-existing nick is double stranded and that repair excision does not occur in this region.
As an example, a 5Ј G-T heteroduplex was processed by HeLa nuclear extracts under various conditions of limited DNA synthesis. As expected, an increased prominence of the 6.4-kb band was observed regardless of how DNA synthesis was inhibited (Fig. 6, lanes 6 -8), indicative of the presence of a single stranded DNA gap around the BanII sequences as a result of the mismatch-provoked excision and inhibited DNA synthesis. Similar assays were performed using the 5Ј-30C and 5Ј-30V substrates. As shown in Fig. 6, there were significant amounts (26%) of BanII-resistant species (6.4 kb in size) in reactions in the absence of dNTPs for both the 5Ј-30V (lane 11) and 5Ј-30C (lane 16) substrates, indicating production of a single stranded DNA gap within the BanII recognition sequence. However, the amount of the 6.4-kb species was significantly reduced in reactions containing aphidicolin (Fig. 6, lanes 13 and 18). These results suggest that the repair reaction was either completed or inhibited prior to excision. The former possibility is contradictory to the repair data shown in Fig. 4. Therefore, aphidicolin appears to be able to inhibit nick-directed excision as well as synthesis. In other words, an aphidicolin-sensitive DNA polymerase (pol ␣, ␦, or ⑀) is potentially required for the excision step of nick-directed loop repair. The excision mechanism for the nick-directed loop repair pathway is apparently different from that for MMR. For example, regardless of the kinds of DNA synthesis inhibitors, at least a 20-fold increase in the amount of the 6.4-kb band was observed during the repair of a G/T mismatch (see lanes [5][6][7][8]. For looped substrates, there is an ϳ10fold increase when omitting exogenous dNTPs (lanes 11 and 16), but an increase less than 3-fold in reactions containing aphidicolin. Control reactions using homoduplex DNA showed little BanII-resistant 6.4-kb species (Fig. 6, lanes 1-4), implying that the excision is provoked by a loop or a mismatch. DISCUSSION The results presented here demonstrate that there are at least two distinct repair pathways for large DNA loops in human cells: one that is directed by a strand break (nickdirected repair) and the other that is directed by the loop itself (loop-directed repair). The loop-directed repair is a novel loop repair pathway in human cells. Using Southern blot analysis, we show that the repair pathway specifically removes the loop structure from DNA regardless of whether the loop is in the continuous strand or the nicked strand (see Fig. 2, A and B). We further show that the loop-directed pathway requires little DNA synthesis because the repair reaction is marginally sensitive to aphidicolin, ddNTPs, or the absence of exogenous dNTPs (see Fig. 4B), all of which strongly block DNA synthesis. The data in this study demonstrate that the nick-directed repair pathway relies on a nick 5Ј (but not 3Ј) to the loop for repair, and the repair only occurs in the nicked strand, which is in agreement with a previous study in human cell extracts (14). In addition, our data suggest that an aphidicolin-sensitive DNA polymerase is somehow involved in the excision step of the nick-directed pathway (see Fig. 6).
DNA large loop repair has been shown in yeast, and previous data suggest that this pathway requires components from both MMR and NER pathways. Kirkpatrick and Petes (10) have demonstrated that the processing of a 26-nt loop requires the gene products of MSH2 and RAD1, MMR and NER components, respectively. Additionally, the purified RAD1-RAD10 complex is capable of enhancing in vitro repair of a 27-nt loop (15). Our results suggest, however, that loop repair pathways in human cells may occur through a mechanism that is not homologous to that seen in yeast. We demonstrate that both the nick-directed and the loop-directed pathways in human cells seem to be independent of the MMR, NER, and WRN pathways, as judged by the fact that cells defective in MSH2, MLH1, XPA, XPG, or WRN are competent in the repair of large looped heteroduplexes for both nick-and loop-directed mechanisms (see Table I and Fig. 3). Although we did not test the involvement of the XPF gene product (the homolog of yeast RAD1) in loop repair, a recent study has indicated that XPF is not required for nick-directed large loop repair (14). Therefore, it is likely that different mechanisms are used to process large DNA loops in human and yeast cells.
The analysis of excision intermediates and repair products under conditions of limited DNA synthesis indicates that MMR, the nick-directed loop repair, and the loop-directed loop repair pathways utilize distinct mechanisms for repair excision. Like MMR and small loop repair, the excision of the nick-directed large loop repair appears to initiate at the preexisting nick and proceeds toward the loop site (see Refs. 14, 15, and 20; and Fig. 5). However, unlike MMR, where the excision step can occur in the absence of a DNA polymerase, the nick-directed excision appears to depend on an aphidicolinsensitive DNA polymerase (Figs. 5 and 6). For the MMR reaction, none of the three conditions causing limited DNA synthesis had an affect on gap formation between the nick and the mismatch, as a single stranded region that is refractory to BanII digestion is created under all conditions (Fig. 6, lanes  6 -8). In the case of nick-directed large loop repair, aphidicolin decreases excision (Fig. 6, lanes 13 and 18), whereas lack of exogenous dNTPs (Fig. 6, lanes 11 and 16) or the presence of ddNTPs (Fig. 6, lanes 12 and 17) only slightly inhibits repair excision. This finding suggests that an aphidicolin-sensitive DNA polymerase is required for the excision step of nick-directed large loop repair. However, how such a polymerase may play a role in repair excision is a puzzle, as none of the known mammalian DNA polymerases contain a 5Ј 3 3Ј exonuclease activity (31). It is possible that a preformed complex of a po-FIG. 6. An aphidicolin-sensitive DNA polymerase may be involved in nick-directed excision. Repair assays were performed under the conditions of limited DNA synthesis as indicated. DNA products were isolated and digested with BseRI and BanII. Digestion products were separated by a 1% agarose gel electrophoresis and detected by staining with ethidium bromide. Bands of 2.7 and 3.7 kb indicate digestion of DNA substrates by both enzymes, whereas a band of 6.4 kb indicates that the DNA was only digested by BseRI. lymerase and a nuclease is recruited to the nick, with DNA synthesis occurring immediately on the heels of digestion.
Additionally, the nick-independent loop repair pathway appears to operate by a much different mechanism. The most obvious difference is that the loop-directed repair requires little DNA synthesis because all conditions of limited DNA synthesis only slightly reduce the repair (Fig. 4). This is consistent with the fact that only limited excision intermediates and a very small patch of excision can be detected in these reactions (Fig.  5C). Therefore, it seems likely that the loop in the loop-directed pathway is removed by incisions (rather than excision) from activities similar to XPG or XPF/ERCC1, but these activities would need to make incisions either to the immediate 5Ј side of a loop, or on both sides of the loop, creating a strand break or a very small (several nucleotide) gap. The former case would create a flap DNA structure that could be processed by flap endonuclease 1 or a similar activity. In either case, only the looped DNA sequence is removed and there would be no need for extensive DNA synthesis. Clearly, much more work is needed to fully elucidate the mechanisms of the pathways.