|Home | About | Journals | Submit | Contact Us | Français|
The hMSH2(M688R) mismatch repair (MMR) gene mutation has been found in five large families from Tenerife, Spain, suggesting it is a Lynch syndrome or hereditary non-polyposis colorectal cancer (LS/HNPCC) founder mutation. In addition to classical LS/HNPCC tumors, these families present with a high incidence of central nervous system (CNS) tumors normally associated with Turcot or constitutional mismatch repair deficiency (CMMR-D) syndromes. Turcot and CMMR-D mutations may be biallelic, knocking out both copies of the MMR gene. The hMSH2(M688R) mutation is located in the ATP hydrolysis (ATPase) domain. We show that the hMSH2(M688R)–hMSH6 heterodimer binds to mismatched nucleotides but lacks normal ATP functions and inhibits MMR in vitro when mixed with the wild-type (WT) heterodimer. Another alteration that has been associated with LS/HNPCC, hMSH2(M688I)–hMSH6, displays no identifiable differences with the WT heterodimer. Interestingly, some extracolonic tumors from hMSH2(M688R) carriers may express hMSH2–hMSH6, yet display microsatellite instability (MSI). The functional analysis along with variability in tumor expression and the high incidence of CNS tumors suggests that hMSH2(M688R) may act as a dominant negative in some tissues, while the hMSH2(M688I) is most likely a benign polymorphism.
LS/HNPCC is an autosomal dominant inherited disease with a penetrance of 80–85% (1,2). The majority of LS/HNPCC is caused by mutation or epigenetic silencing in the DNA MMR genes hMLH1, hMSH2, hMSH6 and hPMS2 although vast majority of the cases are linked to the hMLH1 or hMSH2 genes (2,3). The hMSH2 and hMSH6 proteins form a heterodimer that recognizes base/base and small insertion/deletion mispairs of one or two nucleotides that often arise during DNA synthesis (4). The MMR genes are also involved in the DNA damage response pathways (5–8).
Both hMSH2 and hMSH6 possess ATP-binding domains. The Walker A motif (residues 669–676, GxxxxGKS/T), which coordinates the phosphoryl moiety of the ATP, and the Walker B motif (residues 748–751, DExx), which coordinates the magnesium cofactor, are the most conserved regions of MutS homologs (MSH) (9). These ATP processing domains are essential for the function of MSH proteins in MMR (10,11). The Salmonella typhimurium MutS (K622A) substitution mutation in the Walker A motif was the first identified dominant negative phenotype in MMR (12). Subsequent genetic studies have found that the vast majority of MSH-dominant negative mutations were located in the Walker A domain (13). A comparable K→A substitution mutation of the yeast Msh2 and Msh6 Walker A motif suggested that the dominant negative phenotype was characterized by an inability to perform downstream processes following mismatch binding (14). The Walker A motif mutation in either the hMSH2 or hMSH6 subunit did not affect mismatch binding but did contribute differently to the dominant mutator effects (15–17).
We have previously described the hMSH2(M688R) founder mutation in five large families diagnosed with Lynch syndrome type II from the Island of Tenerife (Canary Islands, Spain). The tumors of index patients showed high microsatellite instability (MSI-H), and the majority of the colorectal tumors examined showed loss of heterozygosity of the WT allele (18). The hMSH2(M688R) mutation is located between the Walker A and B motifs of the hMSH2 protein. The M688 residue resides near the base of a conserved α-helix, which ends in the Walker A site, and interacts with β-sheets that form the base of the Walker B site (11). The Met residue is partially conserved and is found in mouse, rat and yeast among others but is substituted for a Leu in worms (Caenorhabditis elegans) and Ile in Rickettsia rickettsia that causes Rocky Mountain spotted fever (Figure 1A). Considering the structure of the hMSH2–hMSH6 (11), we postulated that the hMSH2(M688R) mutation could disrupt the conserved α-helix and affect the allosteric communication between the mismatched DNA and the ATP-binding sites. The hMSH2(M688I) variant has also been considered as a cause for Lynch syndrome in several families (19–21). Here we have examined the functional consequences of the hMSH2(M688I) and the hMSH2(M688R) alterations. We have determined that the hMSH2(M688I) is most likely a polymorphism, while the hMSH2(M688R) may function as a tissue-dependent dominant mutator that could explain the increased number of CNS tumors in the Tenerife families, which are normally associated with Turcot or CMMR-D syndromes (22–26).
hMSH2(M688R) and hMSH2(M688I) were generated by site-directed mutagenesis. hMSH2 WT and variants (pFastBac™ 1; Invitrogen) and His-tagged hMSH6 (pFastBac™ HT A) were coexpressed in SF9 insect cells and purified by FPLC using Ni-column (Ni-NTA agarose; Qiagen) and PBE-94 (PolyBuffer Exchanger-94; Amersham) as described previously (27). Peak fractions were dialysed against 25mM HEPES pH 7.8, 150–200mM NaCl, 20% glycerol, 1mM DTT and 0.1mM EDTA. Protein concentration was determined by Bradford method and aliquots were frozen in liquid nitrogen and stored at −80°C.
The oligonucleotides were prepared as previously described (28). The concentration was determined by spectrophotometry. For gel shift assays, DNA was labeled using [γ-32P]-ATP and T4 PNK (New England Biolabs). The reaction was purified with Illustra™ ProbeQuant G-50 kit (GE Healthcare). Labeled DNA was tested by native PAGE. For the preparation of labeled 3′ biotin–streptavidin blocked-end DNA, 3′ biotin blocked-end DNA (500ng) purified by HPLC was labeled using T4 polynucleotide kinase as previously described (10). To block the ends with streptavidin, a reaction containing 0.1mM EDTA, duplex DNA (18.4–19.4 pmol) and 250 DNA molar-fold excess Streptavidine (Prozyme) was performed at room temperature for 2h. Samples were electrophoresed on 4% polyacrylamide (29:1) gel at 4°C. Streptavidin–biotin blocked-end DNA was purified by gel extraction as previously described (29). For surface plasmon resonance (SPR) experiments, annealed duplex DNA contained only a 3′ biotin blocked-end.
SPR experiments were performed as previously described (28). Biotinylated DNA was immobilized on the streptavidin-coated chip surface by non-covalent capture. Experiments were carried out at 25°C. Binding experiments consisted of seven protein curves: 0, 10, 20, 50, 100 or 200nM of hMSH2–hMSH6 in 25mM HEPES pH 7.8, 130mM NaCl, 1mM DTT, 0.1mM EDTA, 10mM MgCl2, 6% glycerol, 150 µg/ml acetylated BSA (Promega) and 0.005% P-20 surfactant (BIAcore). The binding is reflected in the increase on the response units with time. To test the ATP-induced dissociation, reaction buffer containing 1mM ATP was added following binding of hMSH2–hMSH6 [100 nM].
Steady-state ATPase analysis was performed as previously described (27,30,31). Experiments were performed with 5× excess homoduplex or heteroduplex DNA or in absence. The data were fit to the Michaelis–Menten equation to generate values for K m and V max (kcat). Standard of deviation was calculated from at least three independent experiments.
ADP→ATP exchange was performed as previously described (27,30,31). Briefly, hMSH2–hMSH6 [50 nM] was incubated in exchange buffer (25mM HEPES pH7.8, 130mM NaCl, 1mM DTT, 2mM MgCl2, 15% glycerol) for 10min at room temperature in the presence of 2.3 µM [3H]-ADP. Exchange buffer containing heteroduplex G/T DNA (200nM) and cold ATP [25 µM] was added to start the ADP→ATP exchange. Reactions were stopped by dilution with 4ml of ice-cold Stop Buffer (25mM HEPES pH 7.8, 130mM NaCl, 10mM MgCl2) and immediately filtered onto Millipore HAWP02500 filters (25mm, 0.45 µm). Filters were dried and counted for radiolabel. The data were fit to a single exponential decay to determine the t1/2. Standard of deviation was calculated from at least three independent experiments.
EndoLys-C digestion reactions were performed as previously described (27) at 37°C in 15 µl reaction mixtures containing 1× proteolysis buffer (25mM HEPES pH 7.8, 130mM NaCl, 2mM MgCl2, 0.1mM EDTA, 1mM DDT, 20% glycerol), 1 µg hMSH2–hMSH6 and endoLys-C (sequencing grade; Roche). When specified, ADP [25 µM] or ATP-γ-S [50 µM] were added prior to protease addition and incubated on ice for 10min. To start the proteolysis, 0, 2.5, 5, 20 or 80ng of EndoLys-C was added. Samples were incubated at 37°C for 45min, stopped by Laemmli buffer addition and boiled. The proteolytic products were separated on 8% SDS–PAGE gels and silver-stained as described (32).
The adenosine nucleotide-binding efficiency of each ATPase protomer was examined by UV crosslinking. No DNA was added to the reactions. Adenosine nucleotides [32P-γ]-ATP (3000 Ci/mmol), [32P-α]-ATP (3000 Ci/mmol), [32P-α]-ADP (3000 Ci/mmol) and [35S-γ]-ATP (1250 Ci/mmol) (Perkin Elmer) were used. [32P-α]-ADP was prepared by [32P-α]-ATP hydrolysis using the hexoquinase enzymatic activity. The purity of [32P-α]-ADP was determined by thin layer chromatography. UV crosslinking experiments were performed on ice in 10 µl reactions containing hMSH2–hMSH6 [200 nM], 25mM HEPES (pH 7.8), 110mM NaCl, 1mM DTT, 10% glycerol in the presence/absence of MgCl2 [5 mM]. Increasing concentrations of adenosine nucleotides (0, 1, 5, 10, 50, 100 and 500nM) were added to the reactions. If the reactions were performed under hydrolysis conditions, 5mM MgCl2 was added to the reaction mixture. Samples were transferred to 96-well plates, placed on ice and UV-crosslinked on a Stratalinker-UV-Crosslinker 1800UV (Stratagene). Samples were electrophoresed on 8% polyacrylamide (29:1) gel and dried at 80°C (Biorad Laboratories). Results were visualized and quantified using Typhoon 9410 phosphorimager (GE Healthcare). The amount of crosslinking was fit to the Hill equation assuming independent subunit binding (Hill coefficient: n = 1) and half saturation (S0.5) was determined for each subunit. If the adenosine nucleotide binding is rate-limiting, the S0.5 is equivalent to equilibrium dissociation constant (K D) (33).
Reactions were performed in 20 µl containing 10fmol of homoduplex or heteroduplex [32P]-labeled DNA in 25mM HEPES pH 7.8, 130mM NaCl, 2mM MgCl2, 1mM DTT, 100–200 µg/ml acetylated BSA (Promega), 25 µM ADP (Sigma) and 15% glycerol. Poly(dI-dC) (5ng/µl) (Amersham) was added as non-specific competitor. For gel shift assays using blocked-end DNA, streptavidin–biotin blocked-ends 32P-DNA (1fmol), HPLC-purified cold DNA (9fmol) and streptavidin [850 nM] were incubated on ice for 10min prior to the binding reaction. PCR-purified homoduplex DNA (5ng/µl) was used as non-specific competitor. Samples were evaluated following electrophoresis at 4°C, on 4% polyacrylamide (29:1 bis), 4% glycerol gel in TBE buffer. For EMSA assays using streptavidin–biotin blocked-end DNA, the electrophoresis was developed on 4% polyacrylamide (37.5:1 bis) and 5% glycerol gel in TAE buffer. Gels were dried, visualized using Thyphoon 9410 Phosphorimager and quantified using Quantity One® software (Bio-Rad).
MMR assays were performed with either purified proteins or nuclear extracts using a circular DNA containing a G/T mismatch (24fmol) and a strand scission 5′ to the mismatch (34). In the purified system, mismatch-provoked excision was determined in reactions containing RPA (800fmol), hMLH1–hPMS2 (400fmol), EXO1 (5fmol) and indicated amount of hMSH2–hMSH6 as previously described (35). Reactions (20 µl) were assembled on ice, incubated at 37°C for 10min and terminated by Proteinase K digestion. DNA samples were recovered by ethanol precipitation after two phenol extractions to remove proteins. The mismatch-provoked excision (generating a single-stranded gap) was scored by restriction enzyme digestion as described previously (34,35). In the nuclear extract system, the complete MMR reaction (including mismatch-provoked excision, gap-filling and ligation) was assayed using 50 µg of nuclear extract as previously described (36).
Immunohistochemical (IHC) staining was performed on 4 µm thick, formalin-fixed, paraffin-embedded tumor sections using Dako TechMate staining robot (DAKO, Glostrup). Samples were deparaffinized and rehydrated through graded alcohol to water. Antigen retrieval was achieved by microwave treatment in 10mM citrate buffer (pH 6). Subsequently, samples were cooled with citrate buffer at room temperature for 1h and washed with distilled water. EnVision™ + kits (K4007, DAKO) were used to prepare the samples for immunostaining. Slides were incubated overnight at 4°C with primary antibody against hMLH1 (clone 14, Oncogene, 1:10; Calbiochem), hMSH2 (clone GB121, Oncogene, 1:20; Calbiochem) and hMSH6/GTBP (mAB clone 44, 1:400; Becton Dickinson Transduction Lab). After overnight incubation, samples were placed in Dako TechMate immunostaining robot following the manufacturer recommendations. This system uses a goat antibody against mouse conjugated with horseradish peroxidase. The final samples are immuno- and hematoxylin-stained. Slides were dehydrated in graded alcohols, cleared in and coverslipped. A tonsil tissue corresponding to reactive lymphoid hyperplasia was used as negative control, omitting the primary antibody in the process described above. Only samples showing positive immunostaining, adjacent to the tumoral tissue (normal mucosa, inflammatory elements in the submucous or dermal chorion), were considered. If these elements were not present, the procedure was repeated at least three times. If these nontumoral elements were still negative for inmunostaining, these samples were deemed not useful for IHC.
The hMSH2(M688R) mutation has been associated with 104 tumors in five Tenerife families (Table 1, updated from ref. 37,38). All tumors analysed contained the c.2063T→G in exon 13 of the hMSH2 gene that results in the missense alteration hMSH2(M688R) and displayed MSI-H. Many tumors in these families displayed loss of heterozygosity of the WT allele (37,38). Colorectal cancer was the most frequent tumor (51.9%) followed by endometrial cancer (9.6%), skin and soft tissue tumors (6.7%) and CNS tumors (5.8%) (six patients: glioblastomas, three patients; other CNS tumors, three patients). The first leiomyosarcoma related to LS/HNPCC was described in these families (38). A high proportion of breast cancer has been found (6%), although the numbers are likely to be an overestimation since vast majority of the cases appeared in a single family. The average age of appearance of the first tumor in these families was 49.1±0.13 years. Six family members (6.7%) developed their first tumor before age 30 years. These included a glioblastoma in one carrier (6 years of age), a leiomyosarcoma and brain tumor in one carrier (18 and 21 years) and colon cancer in two other carriers (25 and 28 years). Moreover, a patient unavailable for mutation testing but whose father carried the hMSH2(M688R) mutation developed a lymphoma at 5 years.
The glioblastoma presented in three carriers at age 6, 39 and 55 years. Two CNS tumors were largely uncharacterized since they occurred prior to comprehensive recordkeeping and presented at 39 and 58 years of age. One carrier presented with a CNS tumor at age 21 years following a leiomyosarcoma at age 18 years (see below). Importantly, Turcot and CMMR-D syndromes include a high incidence of CNS tumors (22–26). Both Turcot and CMMR-D syndromes may present with biallelic MMR mutations and develop childhood tumors including lymphoma that is sometimes combined with characteristics of neurofibromatosis type 1 (26). Useful glioblastoma tissues including the childhood onset tumor were unavailable. However, a blood sample from the 6-year-old glioblastoma carrier was examined for other MMR mutations and only found to contain the hMSH2(M688R) mutation. A missense alteration of the identical residue, hMSH2(M688I), has been linked to LS/HNPCC (19–21). Functional analysis of hMSH2(M688I) has suggested either a slight mutator activity in a yeast system or no functional effect in a mouse system (39,40).
The hMSH2(M688R)–hMSH6 and hMSH2(M688I)–hMSH6 proteins were isolated to better than 95% purity free of any nuclease or other contaminants (Supplementary Figure 1A, available at Carcinogenesis Online). Mismatch binding was examined using simple gel shift analysis (Supplementary Figure 1B, available at Carcinogenesis Online) as well as a more quantitative SPR technology (Figure 1B, Table 2). We noted an increase in the K D for the hMSH2(M688R)–hMSH6 compared to the WT protein (Table 2). Mismatch binding by hMSH2(M688I)–hMSH6 appeared largely identical to the WT hMSH2–hMSH6 (Table 2). The difference in the hMSH2(M688R)–hMSH6 binding activity was attributable to an increased koff, suggesting a modest decrease in the binding stability of the mutant protein with a mismatch (Table 2). Gel shift binding analysis, which is more sensitive to alterations in koff, confirmed this conclusion (Supplementary Figure 1B, available at Carcinogenesis Online). We also note a difference in hMSH2(M688R)–hMSH6 binding saturation. These results suggest that the hMSH2(M688R)–hMSH6 preparations likely contained an inactive fraction that reduced the overall binding-specific activity. Studies of the thermal stability of the hMSH2(M688R)–hMSH6 protein showed little if any difference with the WT hMSH2–hMSH6, suggesting that once purified the protein is stable (data not shown) (41). Taken as a whole, we conclude that the hMSH2(M688R)–hMSH6 heterodimer displays reduced but significant mismatch DNA-binding activity.
The hMSH2 subunit of WT hMSH2–hMSH6 is normally bound with ADP that inhibits the uncontrolled ATP hydrolysis by the hMSH6 subunit and primes the heterodimer for high-affinity mismatch binding (10,15,27). Binding to a mismatch provokes ADP release that is followed by ATP binding to both the hMSH2–hMSH6 subunits, which results in dissociation from the mismatch and the formation of a hydrolysis-independent ATP-bound sliding clamp (10,42). Iterative mismatch binding followed by sliding clamp formation results in multiple ATP-bound sliding clamps surrounding the mismatch (10,42). It is these ATP-bound sliding clamps that recruit downstream MMR machinery and ultimately transmits the mismatch binding information to a distant DNA strand scission where the mismatch excision reaction is initiated (4,43).
The ability of hMSH2–hMSH6 to form an ATP-bound sliding clamp that dissociates from a mismatch may be easily examined using SPR (Figure 1B). We found that both the WT and hMSH2(M688I)–hMSH6 proteins rapidly dissociate from a mismatch when ATP is introduced (koff•ATP = 0.18/s and 0.17/s, respectively; Table 2, Figure 1B). In contrast, the hMSH2(M688R)–hMSH6 remains stably bound to the mismatch (koff•ATP = 0.007/s; Table 2, Figure 1B). These results indicate that the hMSH2(M688R)–hMSH6 heterodimer may remain associated with the mismatch for at least 2–3min. We expanded these results using double blocked-end oligonucleotides as previously described (10). In these studies, ATP-bound hMSH2–hMSH6 sliding clamps may be trapped on an oligonucleotide containing a mismatch that is blocked at both DNA ends by biotin–streptavidin. Specific mismatch binding activity by the WT, hMSH2(M688I)–hMSH6 and hMSH2(M688R)–hMSH6 proteins was first confirmed with these substrates (Supplementary Figure 2A, available at Carcinogenesis Online). The addition of ATP may then trap sliding clamps (compare Figure 1B with Supplementary Figure 2B, available at Carcinogenesis Online). Both the WT and hMSH2(M688I)–hMSH6 heterodimers dissociate from unblocked DNA but are retained by the blocked-end mismatched DNA substrate (Supplementary Figure 2B, available at Carcinogenesis Online). In contrast, the hMSH2(M688R)–hMSH6 remains bound to both the unblocked and blocked-end substrates. A similar effect has been observed when Walker A/B residues were mutated (14,17). These results are consistent with the conclusion that the WT and hMSH2(M688I)–hMSH6 form stable ATP-bound sliding clamps while hMSH2(M688R)–hMSH6 remains largely bound to the mismatch.
A mismatch-dependent ATP hydrolysis (ATPase) cycle exhibited by MSH proteins may be examined using short oligonucleotides containing unblocked DNA ends (27). Once an ATP-bound sliding clamp dissociates from the mismatch, it may slide off the open end where it will complete the ATP hydrolytic cycle leaving it in the ADP-bound form. This ADP-bound form may initiate another round of mismatch binding, sliding clamp formation, oligonucleotide dissociation and hydrolysis in a steady-state cycle. Previous studies have determined that the ATPase kcat most accurately reflects the efficiency of the steady-state ATPase activity since the K m is significantly below the cellular ATP concentration (28,33). The steady-state mismatch-dependent ATPase activity of the WT and hMSH2(M688I)–hMSH6 appeared nearly identical (kcat•WT = 19.2±1.3/min; kcat•hMSH2(M688I) = 14.9±0.6/min; Table 2). This is especially true when one corrects for the background hydrolysis in the absence of mismatched DNA (see Table 2). In contrast, the hMSH2(M688R)–hMSH6 heterodimer appeared completely defective in the mismatch-provoked steady-state ATPase activity and displayed the near-background activity (kcat•hMSH2(M688R) = 4.4±0.5/min) exhibited by non-mismatched DNA (DNA G/C; kcat•hMSH2(M688R) = 2.9±0.1/min) or in the absence of DNA (kcat•hMSH2(M688R) = 1.8±0.2/min; Table 2).
The defect in the hMSH2(M688R)–hMSH6 steady-state ATPase appeared to be associated with both the ability of a mismatch to provoke ADP release as well as adenosine nucleotide binding (Figure 1C and and1D).1D). We found that both the WT and hMSH2(M688I)–hMSH6 proteins efficiently released ADP (ADP→ATP exchange) in the presence of a mismatched DNA (t1/2•WT = 7 s; t1/2•hMSH2(M688I) = 7 s; Figure 1C, Table 2). However, the hMSH2(M688R)–hMSH6 heterodimer appeared to bind less ADP and the mismatch-dependent rate of release displayed considerable longer kinetics (t1/2•hMSH2(M688R) = 220 s; Figure 1C, Table 2). Moreover, we were unable to detect significant adenosine nucleotide crosslinking by the hMSH2(M688R)–hMSH6 heterodimer under a variety of conditions where the WT and hMSH2(M688I)–hMSH6 proteins crosslinked to adenosine nucleotide with near equivalent S0.5 for the individual subunits (Figure 1D). The inability to properly process adenosine nucleotide was reflected in a failure to undergo appropriate protein conformational transitions as determined by partial proteolytic digests (Supplementary Figure 3, available at Carcinogenesis Online). Together, these results strongly suggest that the hMSH2(M688R)–hMSH6 heterodimer is fundamentally defective in the ATP processing activities required to form the sliding clamps required for MMR (10,42). While we observe modest dissimilarity between the WT and hMSH2(M688I)–hMSH6 proteins, these differences generally fall within the margins of error, suggesting that the hMSH2(M688I) substitution has little if any affect on ATP processing functions.
We examined the effect of hMSH2(M688R) mutation on mismatch-provoked excision reaction as well as the complete MMR reaction (Figure 2A). Mismatch-provoked excision contained purified RPA, EXO1, hMLH1–hPMS2 and hMSH2–hMSH6 and is significantly more sensitive than the complete MMR reaction that uses MMR protein-deficient cellular extracts in a complementation analysis (35). Mismatch-provoked excision was scored as the conversion of a circular mismatched DNA to a circular DNA containing a single-stranded DNA gap that is resistant to a specific restriction endonuclease (Figure 2A) (34). The addition of WT hMSH2–hMSH6 resulted in an excision gap-containing product (excision = 40%; Figure 2B, lane 2; Supplementary Figure 4A, lanes 2–4, available at Carcinogenesis Online). In contrast, the addition of hMSH2(M688R)–hMSH6 did not produce a measurable excision gap-containing product (Figure 2B, lane 3; Supplementary Figure 4A, lanes 5–7, available at Carcinogenesis Online). The hMSH2(M688I)–hMSH6 displayed approximately 35% of the WT activity in this excision assay. The inclusion of equivalent amounts of hMSH2–hMSH6 with hMSH2(M688R)–hMSH6 reduced the excision product by 3-fold (excision = 14%; Figure 2B, lane 4). The reduction in excision product generally increased with increasing hMSH2(M688R)–hMSH6 and was completely absent with an 8-fold excess of hMSH2(M688R)–hMSH6 over the WT protein (Figure 2B, lane 7). These results suggest that the hMSH2(M688R)–hMSH6 protein can functionally inhibit the WT hMSH2–hMSH6 during the mismatch excision process. Similar mixing studies with the hMSH2(M688I)–hMSH6 protein suggest a modest inhibition of excision (data not shown)
The complete MMR reaction was examined using a cellular extract deficient for hMSH2 (N6 cells) (44). In these studies, nick-directed MMR results in the reconstitution of a restriction endonuclease site (35). The WT hMSH2–hMSH6 protein complemented the hMSH2-deficient extract (Figure 2C, lane 2; Supplementary Figure 4B, lanes 1–4, S0.5 ≈ 90ng, available at Carcinogenesis Online). In contrast, the hMSH2(M688R)–hMSH6 protein was unable to complement the hMSH2-deficient extract (Figure 2C, lane 3; Supplementary Figure 4B, lanes 5–8, available at Carcinogenesis Online). Importantly, mixing experiments (Figure 2C, lanes 4–7) demonstrated that the hMSH2(M688R)–hMSH6 protein may inhibit the WT hMSH2–hMSH6 reaction (S0.5 ≈ 500ng). These results suggest that hMSH2(M688R)–hMSH6 is capable of interfering with the normal MMR reaction. We regard it likely that a component of the crude cellular extract (N6) used in these complementation studies reduced the effectiveness of hMSH2(M688R)–hMSH6 inhibition compared with the excision assay system. Interestingly, the hMSH2(M688I)–hMSH6 protein displayed approximately 50% of the WT activity in the complete MMR reaction (Supplementary Figure 4B, lanes 9–12, available at Carcinogenesis Online). Taken together, we conclude that hMSH2(M688R)–hMSH6 may function as a dominant negative in MMR. Since a reduction in MMR can affect drug resistance and mutation rates (45,46), it appears likely that the expression of hMSH2(M688R)–hMSH6 protein could ultimately drive tumorigenesis even in the presence of the WT protein. In contrast, the hMSH2(M688I)–hMSH6 protein appears largely similar to the WT protein and is unlikely to affect MMR.
Eighteen tumor blocks that were amenable to IHC analysis were collected from Tenerife hospitals. Tumor samples were MSI-H, and 16 of these tumors showed loss of both hMSH2 and hMSH6 expression compared with the normal adjacent tissues [Figure 3, see representative tumors marked for tumor tissue (T) and adjacent normal tissue (N)]. Two samples (11%), a bladder tumor and the brain tumor (see above), displayed a positive pattern of staining for the hMSH2 and hMSH6 proteins in both the tumor and adjacent normal tissues (Figure 3).
The bladder tumor was the first of two that arose in a patient within a period of 7 years. The first tumor presented a strong expression pattern for the MMR proteins (Figure 3). The second showed no expression of hMSH2 or hMSH6 (data not shown). The brain tumor arose in the patient who initially presented with leiomyosarcoma and later developed a lung metastasis. IHC analysis of the highly necrotic undifferentiated brain tumor showed focally positive hMSH2 and hMSH6 protein expression (Figure 3). It is important to note that the majority of the available tumor tissues examined (89%) showed loss of hMSH2 expression. However, the identification of several MSI-H tumors expressing hMSH2 as well as the high incidence of CNS tumors suggested that the hMSH2(M688R) protein might be expressed in some tissues and may function differently during tumorigenesis than classic loss-of-expression LS/HNPCC mutations.
Members of the Tenerife families carrying the hMSH2(M688R) mutation may develop a single cancer during their lifetime, while other members developed multiple cancers and often in a short period of time. In addition, some family members developed multiple tumors prior to the age of 30 years, including unusual childhood tumors such as glioblastoma and lymphoma. In recent years, several case reports have described children with Turcot or CMMR-D syndromes with compound heterozygous or homozygous MMR gene defects (22–26). These patients developed tumors at early ages and sometimes combined with characteristics of neurofibromatosis type 1. While LS/HNPCC-related tumors are clearly observed, lymphoma, hematological and CNS tumors are most common in these patients (22–25). Turcot and CMMR-D were ruled out with our 6-year-old glioblastoma carrier where we identified a heterozygous hMSH2(M688R) mutation and no other mutations in the hMSH2, hMSH6 or hMLH1 genes. However, the age of onset and the likelihood that another 5-year-old patient with lymphoma also was a hMSH2(M688R) carrier suggested a potentially more aggressive genetic phenotype than simple LS/HNPCC.
The collection of tumor pathology samples in conditions adequate for IHC has only recently been structured on the Canary Islands that include Tenerife. Despite this technical difficulty, we were able to collect 18 tumor samples from hMSH2(M688R) carriers that were amenable to IHC analysis. All of these tumors were MSI (data not shown). Sixteen of the 18 samples, including all of the classic LS/HNPCC type tumors, showed loss of hMSH2 and hMSH6 expression with normal expression of hMLH1. Two samples (11%) displayed expression of hMSH2 and hMSH6 proteins within the tumor. These was a metachronous bladder tumor from a patient who developed a second bladder tumor, which paradoxically showed loss of hMSH2 and hMSH6 expression. The other was a brain tumor from a patient who originally developed a leiomyosarcoma at 18 years that showed loss of hMSH2 and hMSH6 expression. The leiomyosarcoma patient also developed a lung metastasis that displayed loss of hMSH2 and hMSH6 expression while the brain tumor showed expression of both hMSH2 and hMSH6. These results suggest that the brain tumor was a metachronous MSI tumor that expressed the hMSH2–hMSH6 complex although an unusual metastasis cannot be entirely ruled out. Nevertheless, the expression of hMSH2 in this subset of MSI tumors suggested a molecular mechanism of tumor promotion where an MMR defect was present irrespective of MMR protein expression.
Because of the location of the hMSH2(M688) residue, we considered a dominant negative function as one possible explanation for tumor promotion activity in the presence of protein expression. Since a number of studies have implicated the hMSH2(M688I) alteration as causative for LS/HNPCC (19–21), we purified both the hMSH2(M688I)–hMSH6 and hMSH2(M688R)–hMSH6 proteins. Biochemical analysis of the hMSH2(M688I)–hMSH6 heterodimer suggested only modest differences compared with the WT. We conclude that the hMSH2(M688I) alteration is likely to be a non-causative polymorphism. In contrast, the hMSH2(M688R)–hMSH6 heterodimer binds mismatched nucleotides but is completely incapable of processing ATP normally. Moreover, the hMSH2(M688R)–hMSH6 protein significantly interferes with mismatch-dependent excision gap formation at a 1:1 ratio with the WT hMSH2–hMSH6 and to a lesser extent the complete MMR reaction catalyzed by an hMSH2-deficient extract complemented by the WT hMSH2–hMSH6. These studies are consistent with the conclusion that hMSH2(M688R)–hMSH6 may function as a dominant negative for MMR in vitro.
Can the hMSH2(M688R) function as a dominant negative in vivo? The ability of the hMSH2(M688R)–hMSH6 heterodimer to function as a dominant negative will depend on its cellular expression. Moreover, a complete inhibition of MMR function would require an excess of mutant protein expression compared to the WT allele. However, it is clear that even a partial inhibition may result in modestly elevated mutation rates and increased drug resistance, which could ultimately lead to the selection of a complete MMR defect and enhanced drug resistance (45,46).
The IHC results suggest that the mutant protein may be expressed in some tissues but not in others. These observations are consistent with the hypothesis that the hMSH2(M688R) mutation may function as a tissue-dependent dominant negative. In such a case, its effectiveness as a dominant negative would depend on its tissue expression. Based on the IHC results, one might predict that the hMSH2(M688R)–hMSH6 heterodimer may be expressed in brain and lymphoid tissues. Such a tissue-dependent dominant negative could explain the relatively high incidence of CNS and early-onset tumors that appear similar to Turcot and CMMR-D syndromes. It is interesting to note that reports of dominant negative MMR mutations suggest it is likely to be extremely rare and often controversial (47–50). Continued surveillance of the Tenerife families and acquisition of new early-onset tumors amenable to IHC will ultimately confirm whether the hypothesis that the hMSH2(M688R) mutation functions as a tissue-dependent dominant negatve.
This work was supported by 28/03 FUNCIS (to E.S. and Y.B.) and National Institutes of Health grants CA104333 (to L.G.) and CA67007 (to R.F.).
The authors wish to thank Sarah Javaid and Nidhi Punja for technical assistance; Robert Forties for help with statistical analysis and curve fitting; and members of the Fishel and Salido laboratories for helpful discussions.
Conflict of Interest Statement: None declared.