Identification of a Second MutL DNA Mismatch Repair Complex (hPMS1 and hMLH1) in Human Epithelial Cells*

Deficiencies of MutL DNA mismatch repair-complex proteins (hMLH1, hPMS2, and hPMS1) typically result in microsatellite instability in human cancers. We examined the association patterns of MutL proteins in human epithelial cancer cell lines with (HCT-116, N87, SNU-1, and SNU-638) and without microsatellite instability (HeLa, AGS, KATO-III, and SNU-16). The analysis of hMLH1, hPMS2, and hPMS1 was performed using Northern blot, Western blot, and co-immunoprecipitation studies. Our data provide evidence that MutL proteins form two different complexes, MutL-α (hPMS2 and hMLH1) and MutL-β (hPMS1 and hMLH1). Gastric and colorectal cancer cells lines with microsatellite instability lacked detectable hMLH1. Decreased levels of hMLH1 protein were associated with markedly reduced levels of hPMS2 and hPMS1 proteins, but the RNA levels of hPMS1 and hPMS2 were normal. In this study, we describe the association of hPMS1 with hMLH1 as a heterodimer, in human cells. Furthermore, normal levels of hMLH1 protein appear to be important in maintaining normal levels of hPMS1 and hPMS2 proteins.

Deficiencies of MutL DNA mismatch repair-complex proteins (hMLH1, hPMS2, and hPMS1) typically result in microsatellite instability in human cancers. We examined the association patterns of MutL proteins in human epithelial cancer cell lines with (HCT-116, N87, SNU-1, and SNU-638) and without microsatellite instability (HeLa, AGS, KATO-III, and SNU-16). The analysis of hMLH1, hPMS2, and hPMS1 was performed using Northern blot, Western blot, and co-immunoprecipitation studies. Our data provide evidence that MutL proteins form two different complexes, MutL-␣ (hPMS2 and hMLH1) and MutL-␤ (hPMS1 and hMLH1). Gastric and colorectal cancer cells lines with microsatellite instability lacked detectable hMLH1. Decreased levels of hMLH1 protein were associated with markedly reduced levels of hPMS2 and hPMS1 proteins, but the RNA levels of hPMS1 and hPMS2 were normal. In this study, we describe the association of hPMS1 with hMLH1 as a heterodimer, in human cells. Furthermore, normal levels of hMLH1 protein appear to be important in maintaining normal levels of hPMS1 and hPMS2 proteins.
The DNA mismatch repair system (MMR) 1 maintains the sequence integrity of the genome by recognizing and repairing mispaired nucleotides that result from misincorporation during DNA replication. The genes encoding for DNA MMR proteins are highly conserved throughout evolution. In humans, there are two sets of MMR proteins, corresponding to homologues of the bacterial MutHLS system. The human MutS proteins consist of hMSH2, hMSH3, and hMSH6 and the human MutL proteins include hMLH1, hPMS1, hPMS2, and hMLH3 (1)(2)(3).
The role of hPMS1 in human tissues has not been clarified. In yeast, the two MutL-␣ homologues (MLH1 and PMS1) are known to form heterodimers and are essential for MMR. In humans, the two MutL-␣ homologues (hMLH1 and hPMS2) also function as a heterodimer (1). Furthermore, hMSH2, hMLH1, and hPMS2 can be co-immunoprecipitated in the presence of DNA and ATP (11). Another MutL protein, the Saccharomyces cerevisiae MLH3 was shown to interact with MLH1 in a two-hybrid system (2). Similarly, the association of hMLH1 and hMLH3 was described in human cells (12). Recently, using a yeast two-hybrid assay and coexpression of baculoviruses carrying cDNAs encoding hMLH1, hPMS1, and hPMS2 in Sf9 cells, hPMS1 was shown to form a heterodimer with hMLH1 (3). Although a function of hPMS1 in DNA mismatch repair has not been demonstrated, a role for this protein in DNA repair is supported by studies in hPMS1 knockout mice, which showed that, although not developing tumors, these mice display poly(A) tract mutation frequencies above normal levels (13).
Defects in human mismatch repair mechanisms have been implicated in the susceptibility to some inherited and sporadic colorectal, gastric, and endometrial cancers. Characteristically, these cancers display instability of microsatellite repeats, which is an important trait in the identification of human tumors with MMR gene defects (14 -17). Germline mutations of DNA MMR genes occur in more than 90% of hereditary nonpolyposis colorectal cancer (HNPCC) cases, and approximately 70% of the cases are due to germline mutations in hMLH1 and hMSH2 (14). Mutations in other MMR genes (hPMS1, hPMS2, hMSH3, and hMSH6) are infrequent (18). Recently, various missense mutations of hMLH1 in HNPCC were found to have a decreased ability to physically interact with hPMS2 (19). Other functional alterations of the human MutL homologues may be critical in tumorigenesis.
Alterations of the MutL complex are also important in subsets of sporadic colorectal and gastric cancers displaying microsatellite instability (MSI) associated with reduced expression of hMLH1 in the absence of germline mutations and derived colorectal and gastric cancers were shown to have MSI associated with decreased expression of hMLH1 (15,20,21). Epigenetic silencing of the hMLH1 gene caused by hypermethylation of a CpG island in the promoter region was recently found to be an important cause of MMR deficiency in sporadic gastric cancer (21)(22)(23). In this study, we used established epithelial cancer cell lines to identify the pattern of association of the human MutL proteins (hMLH1, hPMS2, and hPMS1) in cells with and without MSI. We show that epithelial cells from colorectal, cervical, and gastric cancers express the hPMS1 gene and that this protein associates with hMLH1 but not with hPMS2. Furthermore, we examined the RNA expression of MutL proteins to assess if loss of MutL proteins in cells with MSI was associated with absence of mutL transcripts.
MSI Assay Using Polymerase Chain Reaction-To characterize the MSI status of gastric cancer cell lines, DNA was extracted from eight or nine different cell clones per marker examined as described previously (27). The microsatellite markers included the BAT26 mononucleotide repeat markers and three dinucleotide markers, as described previously (17). Briefly, one of the primers in each set was end-labeled with [␥-32 P]ATP using T4 polynucleotide kinase (Promega Biotec, Madison, WI). Polymerase chain reactions (PCRs) were performed for 35 cycles, consisting of 1 min at 94°C, 1 min at 50°C, and 1 min at 72°C, with the exception of BAT-40, which required annealing temperatures of 40°C. The PCR products were separated through 7% polyacrylamide gels containing 5.6 M urea and 32% formamide.
Western Blot-Whole cell protein extracts from cultured cancer cell lines were prepared in Laemmli sample buffer (28). The same amounts of protein (15 g) were resolved through SDS-polyacrylamide gel electrophoresis and transferred to an Immobilon-P membrane (Millipore, Bedford, MA). The membranes were incubated overnight at 4°C with antibodies against hMLH1 (clone G168-15; Pharmingen, San Diego, CA), hMSH2 (clone FE11; Oncogene Research, Cambridge, MA), hPMS2 (C-20; Santa Cruz Biotechnology; Santa Cruz, CA), and hPMS1 (C-20 and K-20; Santa Cruz Biotechnology). Antibody concentrations were 1:1000 for all antibodies used. After washing, the filters were incubated with peroxidase-labeled anti-mouse antibodies (hMLH1, hMSH2) or peroxidase-labeled anti-rabbit antibodies (hPMS1, hPMS2) (Amersham Pharmacia Biotech) for 1 h at room temperature. The proteins were then detected using an enhanced chemiluminescence detection system (ECL; Amersham Pharmacia Biotech) and exposed to x-ray film. The membranes were then stripped and reprobed with antihuman actin antibody (Roche Molecular Biochemicals, Indianapolis, IN).
Co-immunoprecipitation-Cultured cell lines were lysed with buffer containing 50 mM Tris-HCl (pH5), 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, and proteinase inhibitor. The supernatants were incubated with 50 l of protein G-agarose (Roche Molecular Biochemicals) for preclearing of nonspecific protein adsorption. The protein Gagarose pellets were removed by centrifugation, and 2 g of anti-hMLH1 monoclonal antibody (clone G168-15), anti-hPMS2 polyclonal antibody (C-20), anti-hPMS1 polyclonal antibodies (C-20 and K-20), or goat serum used as negative control (Santa Cruz Biotechnology) were added to the supernatant and incubated at 4°C overnight. Immunoprecipitation was performed with 50 l of protein G-agarose for 3 h. The protein-agarose complexes were then collected by repeated centrifugation and washing as described by the manufacturer and were resuspended in SDS-gel loading buffer (50 mM Tris-HCl, pH 6.8, 100 mM dithiothreitol, 2% SDS, 0.1% bromphenol blue, and 10% glycerol). The proteins were resolved through SDS-polyacrylamide gel electrophoresis, followed by Western blot analysis using anti-hMLH1, anti-hMSH2, or anti-hPMS1 antibodies as described in the previous section.
RNA Isolation and Preparation of cDNA Probes-Total RNA was extracted from gastric cancer cell lines (AGS, KATO-III, SNU-1, and SNU-16), HCT-116 cells, and HeLa cells by using a total RNA isolation kit (Ambion, Austin, TX). The cDNA probes used for Northern blot analyses were prepared by cloning the reverse transcription (RT)-PCR products into plasmid vectors and were released by restriction enzyme digestion. Primers spanning the coding region of the three DNA mismatch repair genes (hMLH1, hPMS1, and hPMS2) were selected and synthesized by Life Technologies Inc. Primers for hMSH2 were adopted from Liu et al. (29). The oligonucleotide primer sequences are listed in Table I. Total RNA (1 g) was directly amplified by one-step RT-PCR (Titan One-Tube RT-PCR system; Roche Molecular Biochemicals). PCR products of hPMS1 and hPMS2 were cloned into the TA cloning vector (PCR2.1; Invitrogen, Carlsbad, CA), and the EcoRI-digested fragments were labeled and used as probes for subsequent Northern blotting. For hMLH1 and hMSH2, PCR products were cloned into the cloning vector pT7Blue (Novagen, Madison, WI), and the restriction fragments obtained after digestion with BamHI and HincII were used as probes. The cDNA probe sequences were verified by sequence analysis.
Northern Blot Analysis-Total RNA (20 g) was separated by electrophoresis through 1% formaldehyde-agarose gels and transferred to positively charged nylon membranes (Zeta-Probe; Bio-Rad, Hercules, CA). The cDNA probe was labeled with [␣-32 P]dCTP using random oligonucleotide priming (Roche Molecular Biochemicals). Hybridization of membranes was carried out with radiolabeled probes (10 6 cpm/ml) in hybridization buffer (ULTRAhyb; Ambion) at 45°C overnight. Autoradiography was performed by exposing the blots at Ϫ70°C. The blots were stripped and rehybridized with ␣-32 P-labeled glyceraldehyde-3phosphate dehydrogenase (G3PDH) cDNA probe as control.

Microsatellite Instability in Established Gastrointestinal
Cancer-derived Cell Lines-Established cancer cell lines have been productive models to study DNA mismatch repair deficiencies underlying microsatellite instability in gastrointestinal cancers. We therefore used a number of gastric cancer cell lines and a colorectal cancer cell line to investigate alterations of the MutL complex. The MSI-positive cell lines studied included the colorectal cancer line HCT-116 (30) and the gastric cancer cell line SNU-1 (31). Control mismatch repair-competent, MSI-negative HeLa (30) and KATO-III cells were also tested (Table II). In addition, we tested additional gastric cancer cell lines (AGS, N-87, and SNU-16) and determined their patterns of microsatellite instability with four microsatellite markers that included three dinucleotide markers and one mononucleotide marker (Table II). Eight or nine different cell clones were examined with each microsatellite marker. SNU-1 cells displayed marked instability in both the dinucleotide and in the BAT marker (Fig. 1), whereas N-87 cells only showed one clone with microsatellite instability at the D13S170 dinucleotide repeat (Table II).
Expression of DNA Mismatch Repair Proteins in Gastrointestinal Cancer Cell Lines-The expression of the MutL protein homologues hMLH1, hPMS2, and hPMS1 and of the MutS homologue hMSH2 were examined by Western blot analyses (Fig. 2). Western blot analyses of MSI-positive SNU-1 gastric cancer cells showed that this cell line lacked detectable hMLH1 and hPMS2 and showed markedly reduced levels of hPMS1. HCT-116 cells (Fig. 2) and another gastric cancer line (SNU-638, data not shown) also showed decreased levels of the three MutL proteins. Expression of hMSH2 was similar in all lines tested (Fig. 2). Next, we examined whether the RNA levels of the three MutL proteins were similarly decreased. Northern blot analyses of the MutL genes revealed normal levels of hPMS1 and hPMS2 mRNA in both SNU-1 (Fig. 3) and HCT-116 cells (data not shown), whereas hMLH1 mRNA was almost undetectable in these two cell lines (Fig. 3).  Co-immunoprecipitation of MutL Complex Proteins-The decreased protein levels of hPMS1 and hPMS2 with normal levels of the corresponding transcripts in HCT-116 and SNU-1 cell lines and associated decreased levels of hMLH1 protein suggest the possibility that hPMS1 and hPMS2 are unstable in the absence of hMLH1. A human MutL-␣ complex consisting of hPMS2 and hMLH1 has been described in human cells (11,32).
It is likely that hPMS1 acts like the yeast homologue MLH3 and binds to hMLH1. To test this hypothesis we performed immunoprecipitation of protein extracts from the mismatch repair competent cell lines (HeLa, AGS, KATO-III, and SNU-16) and from the MSI-positive gastric cancer cell line (SNU-1) using an antibody that specifically recognizes hMLH1. After blotting to polyvinylidene difluoride membranes, the transferred proteins were identified with antibodies against hMLH1, hPMS1, and hPMS2 (Fig. 4A). To further examine the physical association pattern of MutL proteins, we investigated whether hPMS1 binds hPMS2 by performing protein co-immunoprecipitation. Whole cell protein extracts from MSI-negative AGS cells, which express the three MutL proteins, were immunoprecipitated with hPMS2 antibody, blotted to polyvinylidene difluoride membranes, and reacted with either hMLH1, hPMS1, or hPMS2 antibodies (Fig. 4B). The hMLH1 and hPMS2 antibodies recognized their target proteins after immunoprecipitation by the hPMS2 antibody, but hPMS1 antibodies did not (Fig. 4B). In addition, proteins immunoprecipitated with anti-hPMS1 reacted with anti-hMLH1 and anti-hPMS1 antibodies but not with anti-hPMS2 antibody, further supporting the notion that hPMS1 binds hMLH1 but not hPMS2. This finding demonstrates that hPMS2 and hPMS1 are present in the cell bound to hMLH1, but that hPMS1 and hPMS2 do not associate with each other. Interestingly, the amount of hPMS1 immunoprecipitated by anti-hPMS1 was similar to the amount of this protein immunoprecipitated by hMLH1 (Fig. 4B). Similarly, the amount of hPMS2 immunoprecipitated by anti-hPMS2 was similar to the amount of this protein immunoprecipitated by hMLH1 (Fig. 4B). These findings suggest that hPMS1 and hPMS2 proteins exist in the cell predominantly bound to hMLH1, with a minimal free-protein pool (Fig. 4B). Several theoretical possibilities for the association of hMLH1, hPMS1, and hPMS2 are depicted in Fig. 4C. Our data suggest that the most likely model of MutL complex formation in human cells is depicted as F and G in Fig. 4C, i.e. mutually exclusive heterodimers of hPMS2-hMLH1 (MutL-␣) and hPMS1-hMLH1 (MutL-␤). AGS 0/9 0/9 0/9 0/9 N-87 1/9 0/9 0/9 0/9 SNU-16 0/9 0/9 0/9 0/9 KATO-III 0/9 0/9 0/9 0/9 SNU-1 4/8 4/8 4/8 3/7 HCT-116 4/9 4/9 7/9 3/8   3. RNA levels of hMLH1, hPMS1, hPMS2, and hMSH2 in gastric cancer cell lines. Northern blot analysis of total RNA from AGS, KATOIII, SNU-1, and SNU-16 gastric cancer cell lines, using cDNA probes specific for the DNA mismatch repair genes hMLH1, hPMS1, hPMS2, and hMSH2 (see Materials and Methods) and the control glyceraldehyde-3-phosphate dehydrogenase (G3PDH) gene. SNU-1 cells show almost undetectable levels of hMLH1, whereas hPMS1, hPMS2, and hMSH2 do not appear to be reduced.

DISCUSSION
In this study we examined the expression of the three human MutL homologue proteins (hMLH1, hPMS1, and hPMS2) in established gastrointestinal cancer cell lines and determined their association patterns. The colorectal cancer cell line HCT-116 has been reported to carry a nonsense mutation in the hMLH1 gene that results in undetectable protein in Western blots and is associated with microsatellite instability (26). In addition, we examined the expression of human MutL proteins in gastric cancer cell lines, including the SNU-1 line, which also displays absent hMLH1 protein as a result of a nonsense mutation in the coding region of the gene (25). We confirmed that HCT-116 and SNU-1 cells have instability at both mononucleotide and dinucleotide repeat microsatellite loci. However, in addition to undetectable hMLH1 protein, a marked reduction in the levels of hPMS1 and hPMS2 proteins was also demonstrated. In this study, we also showed that the level of hMLH1 RNA was markedly reduced in both cell lines but that hPMS1 and hPMS2 RNA levels were normal. The decreased hPMS1 and hPMS2 protein levels with normal RNA levels in cells lacking hMLH1 protein suggest that hMLH1 protein is required to maintain normal levels of hPMS1 and hPMS2, which indicate that protein stability may be an important mechanism to maintain functional levels of DNA mismatch repair proteins. In support of this possibility, Buermeyer et al. (33) reported that mPMS2 protein is decreased in fibroblasts from MLH1 knockout mice and that expression of hMLH1 restored normal levels of mPMS2.
In addition to decreased hMLH1 protein levels, hMLH1 RNA levels were also low in SNU-1 and HCT-116 cells. This finding might be explained by data from recent studies indicating that hMLH1 promoter methylation was associated with low levels of hMLH1 RNA and protein in gastric cancers with a high level of MSI (21)(22)(23). Similar findings have been described in MSI-H colorectal cancers (15,20,34). The requirement of hMLH1 for normal cellular levels of hPMS1 and hPMS2 also indicates that these proteins probably interact with each other. A complex containing hMLH1 and hPMS1 was only recently described by co-expression of baculoviruses carrying cDNAs encoding hMLH1, hPMS1, and hPMS2 in Sf9 cells, and in a yeast two-hybrid assay (3). To determine the patterns of interaction of the three MutL proteins in vivo, we performed co-immunoprecipitation combined with Western blot analysis using cell lines with competent DNA mismatch repair function. Our results showed that a specific antibody against hMLH1 not only precipitated hMLH1, but also hPMS1 and hPMS2 (Fig. 4). Therefore, co-immunoprecipitation confirmed that the MutL proteins form a complex consisting of hMLH1, hPMS1, and/or hPMS2. Several models for the association of the three MutL proteins are possible, as represented in Fig. 4C. The results of co-immunoprecipitation studies indicate that hPMS2 and hPMS1 are present in the cell bound to hMLH1, but that hPMS1 and hPMS2 do not associate in the cell, resulting in two types of MutL heterodimers. The model of MutL protein association can thus be depicted as F (hMLH1-hPMS2 or MutL-␣, because it was the first MutL complex described) and G (hMLH1-hPMS1 or MutL-␤) (Fig.  4C). The identification of two different forms of the MutL complex in human cells potentially increases the complexity of DNA MMR deficiencies associated with human disease, namely, in HNPCC and in sporadic cancers. In view of the apparent redundancy of the protein partners of hMHS2 (hMSH3 and hMSH6) and of hMLH1 (hPMS2, hMLH3, and hPMS1), it is logical that the most critical genes in DNA MMR are hMHS2 and hMLH1, matching the most frequent MMR gene deficiencies in the hereditary colorectal cancer syndromes and MSI-positive sporadic cancers (14,21). In other words, the redundancy of hPMS1 and hPMS2 may explain why these genes are rarely found mutated in cancers with the mutator phenotype.
In conclusion, our data provide evidence for the association of hPMS1 with hMLH1 as a heterodimer in human epithelial cancer cells. In addition, normal levels of hMLH1 protein appear to be important in maintaining normal levels of hPMS1 and hPMS2 proteins, suggesting that PMS proteins are unstable in the absence of hMLH1. These findings raise the possibility that the assembly and stability of MutL complex proteins might be regulated by post-translational mechanisms. FIG. 4. Immunoprecipitation of MutL proteins from cancer cell lines. A, immunoprecipitation was performed with the anti-hMLH1 antibody, using protein extracts from the different cell lines. Whole protein extract from HeLa cells before immunoprecipitation was run in the same gel to assess the integrity of protein in the extract and to identify the correct hMLH1 protein size (H1); the next lanes represent protein immunoprecipitated with hMLH1 antibody from HeLa (H) AGS (A), KATO-III (K), and SNU-1 (S1) cells, and the last lane (G) shows the results of protein extract from HeLa cells immunoprecipitated with control goat serum (G). Western blots were performed with specific antibodies against hMLH1, hPMS1 C-20(C) or K-20(K), and hPMS2, as indicated on the left. B, immunoprecipitation of cell extracts from AGS gastric cancer cells with antibodies against hPMS2 or hPMS1 or with goat serum followed by Western blot and immunodetection with antibodies against hMLH1, hPMS1, and hPMS2 as indicated on the left. C, possible patterns of association of the MutL proteins. The results of co-immunoprecipitation indicate that the most likely complexes are the heterodimers schematically represented by F and G, corresponding to MutL-␣ and MutL-␤, respectively.