The Interaction of DNA Mismatch Repair Proteins with Human Exonuclease I

Exonucleolytic degradation of DNA is an essential part of many DNA metabolic processes including DNA mismatch repair (MMR) 1 and recombination. Human exonuclease I (hExoI) is a member of a family of conserved 5’ →→→→ 3’ exonucleases which are implicated in these processes by genetic studies. Here, we demonstrate that hExoI binds strongly to hMLH1, and we describe interaction regions between hExoI and the MMR proteins hMSH2, hMSH3, and hMLH1. In addition, hExoI forms an immunoprecipitable complex with hMLH1/hPMS2 in vivo . The study of interaction regions suggests a biochemical mechanism of the involvement of hExoI as a downstream effector in MMR and/or DNA recombination.


INTRODUCTION
DNA is susceptible to exogenous and endogenous damaging agents as well as spontaneous hydrolysis resulting in nucleotide damage, abasic (AP) sites, and DNA strand breaks (1). In addition, DNA sequence alterations may be caused by nucleotide misincorporation errors during DNA replication (2) and recombination between divergent DNA parents (3). Highly conserved DNA repair mechanisms have been identified, and inactivation of these pathways has been linked to a variety of diseases including cancer (4). DNA mismatch repair (MMR) is one of the best studied repair pathways (5)(6)(7), and inactivating mutations in the human MMR genes are causally involved in the development of sporadic and hereditary cancers such as the common cancer susceptibility syndrome Hereditary Non-Polyposis Colon Cancer (HNPCC) (8). Schmutte et al.;page 4 EXO1 (32), S.pombe EXO1 (33), the D.melanogaster TOSCA protein (34), human and yeast FEN1 (35)(36)(37), and the human hExoI (32,38,39). S.pombe EXO1 degrades DNA starting at both double-strand ends and DNA nicks, and an exo1strain exhibited a 4-to 13-fold increased mutation rate in a G→T transversion assay (33). While the mutation rate of S.cerevisiae msh2 mutants was several-fold higher than exoI mutants, this rate was not significantly elevated in exo1 msh2 double mutants (40,41). These results suggested that msh2 and exo1 were likely to be epistatic (although additivity was not entirely ruled out). Yeast strains harboring a replicative DNA polymerase δ defective in its 3'→5' proofreading exonuclease (pol3-01) in combination with a deletion of exo1 or msh2 display similarly elevated mutation rates (41), and haploid yeast strains with a DNA polymerase δ proofreading defect combined with a deletion of exo1 or msh2 are lethal (41). Taken as a whole, these results suggest yeast EXO1 plays a role in mutation avoidance and supports a model in which the 5'→3' activity of EXO1 and the 3'→5' exonuclease activity of Pol δ participate in a bidirectional MMR (41). Several lines of evidence also indicate that EXO1 functions in additional pathways. EXOI expression is UV inducible, and EXO1-mutant yeast strains display a mild UV sensitivity (42). Biochemical and genetic evidence in yeast (43) and the observation of high hExoI/HEX1 expression in testis (44) suggests the involvement of these proteins in meiotic and mitotic recombination. hExoI also shows RNAse H activity, and it may function as backup for FEN1 in RNA primer removal during lagging strand DNA synthesis (45).
We have recently shown that hExoI interacts directly with hMSH2 in vitro (38). hExoI has been shown to functionally complement a yeast exo1 mutant strain, and the truncated version of hExoI, termed HEX1 (or hExoIa), has both single-stranded and double-stranded 5'→3' exonuclease activity in vitro (45). In this study, we examined the interactions of hExoI with other human MutS and MutL homologs. Characterization of the interaction regions suggests a mechanism for recognition, signaling and excision of the damage. Our results support a role for by guest on March 23, 2020 http://www.jbc.org/ Downloaded from hExoI in down-stream events associated with MMR and/or other metabolic processes involving MMR enzymes.

MATERIALS AND METHODS
Protein-Protein Interaction -The GST/IVTT protein-protein interaction assay was performed essentially as described (38). Briefly, hExoI, hMSH2, hMSH3, hMSH6, hMLH1, hPMS2 and relevant fragments of these constructs were subcloned into pGEX-4T-2 (Pharmacia Biotech), which allows high expression of a glutathione S-transferase (GST) fusion protein, and were transformed into E.coli XL1-blue. A 10ml overnight culture was diluted with LB medium to 800ml and grown at 37˚C until the OD 600 of the culture was 0.5. Adding IPTG to a final concentration of 1 mM and incubation for 2 hr at 30˚C induced protein expression. The culture was harvested by centrifugation at 2000 xg for 15 min at 4˚C. The bacterial pellet was resuspended in PBS containing 1mM phenylmethylsulfonylfluoride (PMSF), 2µg/ml leupeptin, and 2µg/ml pepstatin. The cells were lysed by 2 freeze-thaw cycles in the presence of 100µg/ml lysozyme and Triton X-100 (0.2%) and 5 mM dithiothreitol (DTT). DNAse I was then added to a final concentration of 20µg/ml. Lysates were incubated on ice for 30 min and cellular debris removed by centrifugation. The cleared supernatant was incubated with glutathione agarose beads (SIGMA) for 1 h at 4˚C under gentle continuous agitation. Under these conditions, approximately 20-50ng of protein was bound to 25µl beads. Samples were pelleted for 1 min at 300 xg and each pellet was washed 3 times with 500µl of binding buffer (20mM Tris pH7.5, 10% glycerol, 150mM NaCl, 5mM EDTA, 1mM DTT, 0.1% Tween 10, 0.75mg/ml BSA, 0.5mM PMSF, 1µg/ml leupeptin, and 1µg/ml pepstatin). Lysates containing the unmodified pGEX vector were treated similarly and used as a negative control. GST and GST-fusion protein binding to the beads was verified by denaturing gel electrophoresis (SDS-PAGE) and quantitated using BSA as a standard as previously described (46). To produce 35 S-labeled proteins, hExoI, hMSH2, hMSH3, hMSH6, hMLH1, hPMS2 and fragments thereof where subcloned into pET24d (Novagen), and 1µg of these constructs was used in in vitro transcription/translation (IVTT) (TNT Coupled Reticulocyte Lysate System; Promega). The labeled proteins were quantitated as described (46). Equivalent amounts of protein from each IVTT mix was added to the suspension of protein-bound glutathione beads in 500µl of binding buffer. The mixture was incubated at 4˚C for 1 h under gentle agitation. Subsequently, the beads were washed 3 times with binding buffer and then resuspended in SDS-PAGE loading buffer. Proteins were resolved on a 10% SDS-PAGE gel, visualized and quantified using a Molecular Dynamics PhosphorImager System.

Quantitation of Protein
Interactions -The GST-IVTT interaction assay system is not absolutely quantitative and is likely to depend on the relative association constant (k assoc ) of the individual interacting proteins. Thus, subtle changes in the relative concentration of interacting peptides may influence the ultimate measure of interaction. In order to provide modest control between experiments for such concentration-dependent processes, we determined the approximate molar concentrations of the GST-fusion protein and the IVTT protein as previously described (46). Interaction experiments were designed to contain an equivalent ratio of the test peptides.
Relative interaction (Int rel ) was determined as the fraction of the peptide interaction-ratio (IR p ) divided by the wild type interaction-ratio (IR wt

RESULTS
We have previously demonstrated that hExoI interacts with hMSH2 (38), suggesting a role in human MMR and/or recombination repair processes that involve the MMR lesion-recognition processes. To further explore the role of hExoI, we examined its interaction with several other human MMR proteins using the glutathione-S-transferase-fusion (GST)/in vitro transcription translation (IVTT) bait-prey interaction system as previously described (38,46 following incubation and direct examination (data not shown). Any background of smaller-thanpredicted peptide fragments present in the IVTT reaction and/or interaction experiments were found to represent internal translation start sites contained in the IVTT subclone which are generated in the IVTT reaction (data not shown). In addition, treatment of the GST/IVTT extracts with DNAse did not alter the results suggesting that contaminating DNA does not mediate any of the observed interactions (data not shown). While it is possible that our peptide-deletion constructs do not adopt an entirely wild type conformation, it is important to note that the interaction regions identified for hMSH2, hMSH3 and hMSH6 by this methodology (46) have been largely confirmed by comparison to the crystal structure of bacterial MutS (12,13).

The interaction regions between hExoI and hMSH2
We determined the interaction region between GST-hMSH2 with IVTT-hExoI ( Figure 2A). were deleted. These results suggest that the carboxy-terminal amino acids 604-846 of hExoI were likely involved in the specific interaction with hMSH2 (see shaded area of Figure 2A). This [hExoI(del 10)] continue to display binding to GST-hMSH2.
We similarly determined the reciprocal interaction region of the IVTT-hMSH2 protein with GST-hExoI ( Figure 2B). interaction with hExoI, and that hMSH2-(aa 261-600) may further stabilize this interaction and/or allow enhanced peptide-folding of the interaction region (see shaded area of Figure 2B).
The interaction region between IVTT-hMSH3 and GST-hExoI was determined to be also N-terminal ( Figure 3B). Of the subdivided hMSH3 overlapping peptides, only the peptide containing amino acids 1-617 [hMSH3(del 1)] was found to significantly interact with GST-hExoI (note: the interaction was actually greater than the full-length protein). Subdivision of amino acids 1-297 of hMSH3 [hMSH3(del 4-6)] resulted in further definition of the interaction region with hExoI. Since all three peptides displayed a significant Int rel-hMSH3 , we conclude that the minimal hExoI interaction region of hMSH3 spans amino acids 75-297 (see shaded area of Figure 3B).
The lack of an additional interaction region outside hMSH3-(aa 75-297) was confirmed with the hMSH3(del 7) IVTT-peptide. It appears that del7 has a slightly increased mobility (compare del4, 297 amino acids to del 7, 334 amino acids). This may be due to region of high polarity in this protein fragment or due to a slightly longer electrophoresis of the gel containing del6 and del7.

The interaction regions between hExoI and hMLH1
Our initial studies suggest that hExoI displayed a more robust interaction with hMLH1 compared to hMSH2 (Figure 1). We further defined the region(s) of IVTT-hExoI interaction with GST-hMLH1 ( Figure 4A).  Figure 4A).

Interaction between hExoI and hMLH1 is affected by mutations found in HNPCC
MLH1 is essential for MMR in eukaryotic cells (47,48). Mutational inactivation or abolished expression by promoter hypermethylation in mammalian cells leads to a mutator phenotype (49)(50)(51)(52), and up to 50% of all HNPCC cases show mutations in hMLH1 (8). We have previously demonstrated that missense mutations of hMLH1 found in well-defined HNPCC kindreds affect the interaction of hMLH1 with both hPMS1 and hPMS2 suggesting a pathogenic consequence of these mutations ( (53); unpublished results). We tested the effect of several of these missense mutations of hMLH1 on its interaction with hExoI ( Figure 4C). We found that four of these missense mutations (hMLH1-L574P; hMLH1-K616∆; hMLH1-R659P; hMLH1-A681T) reduced the interaction with hExoI >80%. These missense mutations also affected the interaction of hMLH1 with hPMS1 and hPMS2. However, the interaction between hMLH1 and hPMS1 or hPMS2 appeared to be more reduced by these missense mutations than the interaction between hMLH1 and hExoI [compare ref. (53) and Figure 4C]. A missense mutation (hMLH1-L582V) that did not affect the interaction of hMLH1 with either hPMS1 or hPMS2, also did not affect the interaction between hMLH1 and hExoI. These results suggest that hMLH1 mutations found in HNPCC kindreds, which affect its interaction with its heterodimeric partners hPMS1 or hPMS2, may also affect its interaction with hExoI. These results have significant implications when considering the functional effects of HNPCC mutations.

Interaction between hExoI and hMLH1 does not compete with binding to hPMS2
A summary of the interaction regions between the human MMR proteins and hExoI are shown in Figure 5. Because the interaction between hExoI and hMLH1 overlaps the C-terminal interaction region for hPMS2 (53), it was formally possible that hExoI might interfere with the MutL homolog heterodimer interactions. To test this possibility and to determine the physiological by guest on March 23, 2020 http://www.jbc.org/ Downloaded from relevance of the hExoI and hMLH1 interaction we performed co-immunoprecipitation studies using HeLa total cell extracts ( Figure 6). Antibodies for hExoI, hMLH1, and hPMS2, each reciprocally co-precipitated all three proteins ( Figure 6, Lanes 1-3). In contrast, these proteins were not precipitated when no antibody or pre-immune serum was present in the incubation mixture ( Figure   6, Lane 4; data not shown). The hExoI antibody was found to recognize the hExoI protein but not hMLH1 or hPMS2 (Lane 5). Similar specific recognition was demonstrated with the hMLH1 and hPMS2 antibodies (data not shown). While we cannot exclude that hExoI binds individually to hMLH1 and hPMS2 in the cell, the weak interaction between hExoI and hPMS2 in vitro supports the concept that some fraction of the hExoI, hMLH1 and hPMS2 proteins exist as an immunoprecipitable complex in the cell. In addition, our results support the conclusion that hExoI can interact with the hMLH1-hPMS2 heterodimer. We were unable to reciprocally coimmunoprecipitate hMSH2 or hMSH6 with the hExoI antibody under the same conditions (data not shown). Because hMSH2 may exist in at least three different conformational states that depend on its binding to adenosine nucleotides (16,54), it is possible that the cellular interaction of hMSH2 with hExoI requires a specific form of hMSH2 and/or its heterodimeric partners, or that other factors alter this interaction.

DISCUSSION
The excision of mismatched or damaged nucleotides during MMR as well as other DNA metabolic processes such as DNA recombination requires an exonucleolytic enzyme activity.
Genetic and biochemical evidence suggests that the 5' 3' double-strand DNA exonuclease, hExoI, is involved in these processes in eukaryotic cells (45). hExoI interacts strongly with the MMR protein hMSH2 in the GST/IVTT assay (38), and the S.cerevisiae orthologs have also been shown to interact by yeast two-hybrid analysis (40). Here we demonstrate that hExoI interacts with the MMR protein hMLH1 in vitro, and that hExoI may exist in a complex with hMLH1-hPMS2 in vivo. It is important to reiterate that the interaction regions described here could be substantially influenced by inappropriate or altered peptide folding associated with the deletion constructs. However, our studies clearly provide a framework for further analysis.
The global interaction of hExoI with MMR proteins suggests both a specific and mechanistic association. We have previously proposed a model for MutS homolog function in which mismatched nucleotides, lesions, or DNA structures provoke ADP→ATP exchange and the formation of a hydrolysis-independent sliding clamp capable of diffusion along the DNA backbone (16). The sliding clamp model proposes that a C-terminal "hinge" region maintains a stable contact between hMSH2 and hMSH3 while an N-terminal "clasp" domain changes conformation in response to mismatch provoked ADP→ATP exchange and clamp formation (16,46,55). The crystal structure of the E.coli MutS and Taq MutS homodimers bound to a G/T mismatch has recently been solved (12,13). Comparison of these structures with corresponding regions in the human homologs have largely confirmed the stable contact between the C-terminus "hinge domain" of hMSH2 with hMSH3 or hMSH6. The overall structure resembles a "pair of praying hands" (13) or a clamp (16) surrounding the mismatched DNA. While the "clasp" region identified for hMSH2 fits well with the structural data, the similar interaction regions identified on hMSH3 and hMSH6 appear to encompass other functional domains (46,55). It is possible that the structural transitions associated with ATP-binding by the MutS homologs will identify additional interaction regions (46).
We have observed that the interaction regions for hExoI with hMSH2 and hMSH3 appear to cover, but are not identical to, the N-terminal "clasp" regions identified between hMSH2 with hMSH3 ( Figure 5). These interactions may reflect the limitations of the GST/IVTT interaction system, as well as the two-hybrid interaction system, where partial peptides may fold and/or interact inappropriately. With this substantial caveat in mind, the core domain for interaction with hExoI that we have identified here (hMSH2 aa 600-672) covers the lower portion of the Walker We have previously demonstrated equivalent C-terminal interaction regions between hMLH1 and hPMS2 that are affected by missense mutations found in HNPCC (53). Interestingly, HNPCC mutations, which affect the interaction between hMLH1 and hPMS2, also affect the interaction with hExoI ( Figure 4C).
There are a number of caveats associated with peptide interactions. However, the identification of HNPCC mutations, which affect the interaction between full-length hMLH1 and hExoI in addition to the demonstration of immunoprecipitable complexes of these proteins from human cells, appears to provide a biological significance to these studies. A comprehensive biochemical study of the functions associated with the purified MMR proteins and hExoI is in progress.

ACKNOWLEDGMENTS
We thank Hans-Jürg Alder and the Kimmel Nucleic Acids Facility for oligonucleotide synthesis and sequencing. This work was supported by NIH grant CA56542 (R.F.)  the IVTT proteins were also incubated with GST alone. After washing, the proteins were eluted from the beads, separated by SDS-PAGE and visualized by autoradiography. Quantitation was performed as previously described (46). Double lines in the grid above the gels indicate separate gels. Lower panels illustrate the constructs and summarize the interaction results (Int rel ). The shaded area marks the interaction region of hExoI with hMSH2. Panel B) 35 S-labeled hMSH2 full length and deletion peptides were precipitated with GST-hExoI, and the interaction was analyzed as in Panel A. The dark shaded area marks the region essential for the interaction of hMSH2 with hExoI while the lighter shaded area illustrates the region stabilizing the interaction. peptides that displayed greater that 100% Int rel-hExoI (relative to the wild type hExo1 protein) are marked with a star (*). Panel B) 35 S-labeled hMSH3 full length and deletion peptides were similarly precipitated with GST-hExoI, and the interaction was analyzed as above. The shaded areas mark the consensus interaction regions of hMSH3 with hExoI. Interaction between truncated peptides that displayed greater that 100% Int rel-hMSH3 (relative to the wild type hMSH3 protein) are marked with a star (*). Materials and Methods). As a control for nonspecific binding to the GST or glutathione beads, the IVTT proteins were also incubated with GST+glutathione beads alone. Double lines in the grid above the gels indicate separate gels. Lower panels illustrate the constructs used, and the shaded area marks the consensus interaction region of hExoI with hMLH1. Interaction between truncated peptides that displayed greater that 100% Int rel-hExoI (relative to the wild type hExoI protein) are marked with a star (*). Panel B) 35 S-labeled hMLH1 full length and deletion peptides were precipitated with GST-hExoI (see legend to Figure 2 and Materials and Methods). The shaded areas mark the consensus interaction regions of hExoI with hMLH1. Panel C) hMLH1 point mutants were generated as described (53), precipitated with full length GST-hExoI, and interaction analyzed as above. The lower panel illustrates the quantification of interaction relative to wild type hMLH1 (Int rel-hMLH1 ). Interaction between truncated peptides that displayed greater that 100% Int rel-hMLH1 (relative to the wild type hMLH1 protein) are marked with a star (*). Antibodies for hExoI, hMLH1, and hPMS2 were able to co-precipitate all three proteins suggesting that they exist as a cellular complex. As a control, no signal was generated in the absence of a precipitating antibody or with pre-immune serum (Lane 4; data not shown). The specificity of the hExoI, hMLH1, or hPMS2 antibodies was tested by Western analysis using purified recombinant hExoI (Lane 5).