Proteolysis of the mismatch repair protein MLH1 by caspase-3 promotes DNA damage-induced apoptosis.

Caspases are critical proapoptotic proteases that execute cell death signals by selectively cleaving proteins at Asp residues to alter their function. Caspases trigger apoptotic chromatin degradation by activating caspase-activated DNase and by inactivating a number of enzymes that sense or repair DNA damage. We have identified the mismatch repair protein MLH1 as a novel caspase-3 substrate by screening small pools of a human prostate adenocarcinoma cDNA library for cDNAs encoding caspase substrates. In this report, we demonstrate that human MLH1 is specifically cleaved by caspase-3 at Asp(418) in vitro. Furthermore, MLH1 is rapidly proteolyzed by caspase-3 in cancer cells induced to undergo apoptosis by treatment with tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) or the topoisomerase II inhibitor etoposide, which damages DNA. Importantly, proteolysis of MLH1 by caspase-3 triggers its partial redistribution from the nucleus to the cytoplasm and generates a proapoptotic carboxyl-terminal product. In addition, we demonstrate that a caspase-3 cleavage-resistant D418E MLH1 mutant inhibits etoposide-induced apoptosis but has little effect on TRAIL-induced apoptosis. These results indicate that the proteolysis of MLH1 by caspase-3 plays a functionally important and previously unrecognized role in the execution of DNA damage-induced apoptosis.

The caspase family of cysteine proteases is an essential effector of the apoptotic cell death program that catalyzes many of the biochemical and morphological events of apoptosis by concerted proteolytic actions on a subset of intracellular proteins (1,2). Caspases are organized in a proteolytic cascade in which initiator procaspases are activated by oligomerization via recruitment to distinct caspase-activating complexes; these apoptotic signals are then amplified by the mitochondria. In the extrinsic pathway, ligands of the tumor necrosis factor (TNF)-␣ 1 family (e.g. TNF-␣, TNF-related apoptosis-inducing ligand (TRAIL), and Fas ligand) bind to their death domaincontaining receptors, an event that leads to the recruitment of the death domain-containing protein FADD and subsequently to the recruitment and activation of procaspases-8 and -10 (1)(2)(3)(4)(5). In the intrinsic pathway, caspase-2 rather than caspase-9 has recently emerged as the likely initiator caspase activated by genotoxic stress such as DNA damage and chemotherapeutic drugs (6 -8). In response to these stimuli, procaspase-2 is recruited to a large cytosolic complex, the components of which have yet to be identified (but are apparently different from the apoptosome discussed below), resulting in its oligomerization and activation (8). Importantly, the apoptotic signals initiated by death ligands or genotoxic stress are amplified by the mitochondria; caspase-8 (extrinsic pathway) and caspase-2 (intrinsic pathway) trigger the mitochondrial release of cytochrome c and other proapoptotic mediators such as Smac/DIABLO (6, 7, 9 -12). Procaspase-9 is then activated by oligomerization in the apoptosome, a large cytosolic complex composed of cytochrome c, Apaf-1, ATP, and procaspase-9 (13,14). Caspases-8, -9, and -10 proteolyze and activate downstream caspases, including caspases-3, -6, and -7, which execute the apoptotic cell death signal by cleaving a number of intracellular protein targets (5,15,16).
The oligonucleosomal degradation of chromosomal DNA is one of the defining irreversible features of apoptotic cells that facilitates packaging the fragmented genome into apoptotic bodies (1,17). Caspases promote DNA fragmentation by a variety of mechanisms. First, at least under certain circumstances, caspases induce the mitochondrial release of apoptosis-inducing factor and endonuclease G, which interact with each other to induce chromosomal DNA fragmentation (18 -20). Second, caspases proteolyze and inactivate ICAD (also known as DFF-45), an inhibitor of the caspase-activated DNase (CAD) (21)(22)(23). ICAD normally binds to and suppresses the DNase activity of CAD. Caspase cleavage of ICAD releases it from CAD, thereby activating CAD, which degrades chromosomal DNA between nucleosomes. Third, caspases cleave several enzymes that sense or repair damaged DNA, including poly(ADPribose) polymerase (PARP), RAD21, RAD51, ATM, the catalytic subunit of the DNA-dependent protein kinase, the Bloom syndrome protein (BLM), and BRCA1 (24 -31). In the case of RAD21, caspase cleavage generates a proapoptotic carboxylterminal product that is sufficient to induce apoptosis (25,26). In this way, caspases promote DNA fragmentation by activating apoptotic DNases and by systematically subverting the DNA repair machinery.
Here we report that MLH1, a component of the conserved DNA mismatch repair (MMR) complex, is a novel and specific caspase-3 substrate. Indeed, our results are the first to indicate a linkage between caspases and MMR proteins. The MMR system recognizes and repairs mispaired or unpaired nucleotides that result from errors in DNA replication (32,33). The mammalian homologues of the Escherichia coli MutS protein form heterodimers (MSH2-MSH6 and MSH2-MSH3) that bind to nucleotide mismatches and recruit heterodimers of the MutL homologues (MLH1-PMS2, MLH1-PMS1, or MLH1-MLH3) to base mismatches (32)(33)(34). These MLH1-containing heterodimers in turn function as adaptor proteins that link the MutS proteins to the DNA repair/replication machinery, resulting in the excision and repair of the mismatch-containing, newly synthesized DNA strand. Inactivating mutations of MLH1 and MSH2 occur commonly in hereditary nonpolyposis colon cancer and less commonly in other carcinomas, and they result in a "mutator" phenotype characterized by instability of repetitive microsatellite DNA sequences (33,(35)(36)(37)(38). MMR proteins have also been implicated in homologous recombination repair of DNA double-strand breaks, in cell cycle checkpoint activation, and in the execution of the apoptotic response to DNA damage induced by alkylating agents and other drugs that modify nucleotides or inhibit topoisomerase II (39 -47).
In the present work, we show that MLH1 is rapidly and specifically cleaved by caspase-3 at Asp 418 in cells induced to undergo apoptosis by treatment with TRAIL (extrinsic apoptotic pathway) or etoposide (intrinsic apoptotic pathway), a chemotherapeutic drug that inhibits topoisomerase II and induces DNA double-strand breaks (48). Furthermore, we demonstrate that caspase proteolysis of MLH1 induces its partial relocalization from the nucleus to the cytoplasm and produces a proapoptotic carboxyl-terminal product. We also show that a caspase cleavage-resistant mutant of MLH1 inhibits apoptosis induced by etoposide but has little effect on TRAIL-induced apoptosis, thereby indicating a novel role for caspase proteolysis of MLH1 in the execution of DNA damage-induced apoptosis.
Small Pool Expression Cloning-Small pool expression cloning to identify cDNAs encoding caspase substrates was performed as described previously (25, 49 -51) with the following exceptions. Small pools (48 cDNAs/pool) of a human prostate adenocarcinoma cDNA library (Invitrogen, catalogue product 11597010) were used to make 35 S-labeled protein pools with the TNT SP6-coupled transcription/translation system (Promega) as described previously (49,50). 35 S-Labeled protein pools were incubated with buffer control or 25 ng of caspase-1, -2, -3, or -8 for 1 h at 37°C; the cleavage reactions were then separated by SDS-PAGE and visualized by autoradiography as described previously (49,50). Single cDNAs encoding putative caspase substrates were isolated by systematically subdividing small cDNA pools and retesting the corresponding 35 S-labeled protein pools as described previously (49,50).
Cell Culture, Apoptosis Induction, and Immunoblotting-Human PC-3 prostate carcinoma cells, TSU-Pr1 bladder carcinoma cells, MCF-7 breast carcinoma cells, or RKO colorectal carcinoma cells stably transfected with empty pcDNA3 vector (designated RKO.3 cells, kindly provided by D. Boothman, Case Western University School of Medicine) were all grown in Dulbecco's modified Eagle's medium (Mediatech) with 10% heat-inactivated fetal calf serum (Invitrogen). Three apoptotic inducers were used in these studies: soluble TRAIL (purified as below) and TNF-␣ (R&D Systems), which activate the death receptor apoptotic pathway (1), and etoposide (Sigma), a topoisomerase II inhibitor (48). Recombinant soluble TRAIL (amino acids 95-281) was produced in E. coli from pET15b plasmid (Novagen) containing a truncated TRAIL cDNA (53), and the His-tagged protein was purified under native conditions with the QIAexpress System (Qiagen) as described previously (54). PC-3 cells were treated with 2 g/ml TRAIL for 0 -16 h, while TSU-Pr1 cells were treated with 50 M etoposide for 0 -36 h. Immunoblotting of whole cell lysates was performed as described previously (52) using the following antibodies: an MLH1 monoclonal antibody (BD Biosciences), a protein kinase C␦ polyclonal antibody (Santa Cruz Biotechnology), a procaspase-3 monoclonal antibody (BD Biosciences), or a RAD21 polyclonal antibody (25). To determine whether caspases were responsible for the apoptotic proteolysis of MLH1, PC-3 or TSU-Pr1 cells were preincubated for 1 h with vehicle or 50 M Z-VAD-fmk, a broad spectrum caspase inhibitor. Cells were then treated with 2 g/ml TRAIL or 50 M etoposide for 0 -24 h.
Transfections, Indirect Immunofluorescence, and Apoptosis Assays-PC-3 or RKO.3 cells were transiently transfected with 1.2 g of pEGFP-N1 plasmids using LipofectAMINE PLUS Reagent (Invitrogen). After 24 h, cells were fixed in 4% paraformaldehyde for 10 min at room temperature, washed, and incubated with the DNA fluorochrome Hoescht 33258 (10 g/ml, Sigma) for 30 min at room temperature. GFP-tagged proteins and nuclei were visualized by fluorescence microscopy using a Nikon Eclipse E400 microscope as described previously (25,51). For apoptosis assays, Ն200 GFP-positive cells were scored for apoptotic nuclei (i.e. condensed or fragmented); all experiments were performed at least in triplicate.

Identification of MLH1 as a Caspase-3 Substrate by Small
Pool Expression Cloning-We have described recently a small pool expression cloning strategy to screen cDNA libraries for cDNAs encoding caspase substrates (25, 49 -51). In this report, small pools (48 cDNAs/pool) of a human prostate adenocarcinoma cDNA library (Invitrogen) were transcribed and translated in vitro in the presence of [ 35 S]methionine, and the corresponding 35 S-labeled protein pools were incubated with recombinant caspases. As shown in Fig. 1A, a ϳ66-kDa protein (indicated by an asterisk) present in 35 S-labeled protein pool 10 was specifically cleaved by caspase-3 (C3), but not by the other caspases tested (caspase-1, -2, or -8), into two products, ϳ45 and 24 kDa in size (indicated by arrows). The enzymatic activity of each protease was verified with a known substrate (data not shown). To isolate the putative caspase-3 substrate present in protein pool 10, cDNA pool 10 was further subdivided into smaller pools, and the corresponding 35 S-labeled protein pools were reincubated with caspase-3. This process was repeated until a single cDNA encoding a ϳ66-kDa protein cleaved by caspase-3 into the appropriately sized fragments was identified (Fig. 1B). This cDNA was sequenced and found to be a partial MLH1 cDNA (36,37). MLH1 Is Specifically Proteolyzed by Caspase-3 at Asp 418 in Vitro-To determine whether the full-length MLH1 protein is cleaved by caspases in vitro, we incubated 35 S-labeled fulllength human MLH1 with recombinant caspases. As shown in Fig. 2A, 35 S-labeled MLH1 was selectively proteolyzed by caspase-3 into two major cleavage products, which were ϳ45 and 40 kDa in size (indicated by arrows). In contrast, none of the other caspases (caspase-1, -2, -6, -7, or -8) examined cleaved MLH1. The activity of each caspase was confirmed using a known substrate (data not presented). To identify the caspase cleavage site in human MLH1, we substituted the Asp residue (Asp 418 ) at a potential caspase-3 cleavage site (Asp-Lys-Thr-Asp 418 -2-Ile 419 ) with a Glu residue. Unlike WT MLH1 (Fig.  2B, left panel), the 35 S-labeled D418E MLH1 protein was not cleaved by caspase-3 (right panel), indicating that Asp 418 is the caspase-3 cleavage site in vitro.
MLH1 Is Cleaved by Caspase-3 in Cancer Cells Undergoing Apoptosis-To determine whether MLH1 is cleaved in cancer cells during the induction of apoptosis, we treated human PC-3 prostate carcinoma cells with 2 g/ml TRAIL for 0 -16 h or human TSU-Pr1 bladder carcinoma cells with 50 M etoposide for 0 -36 h. As demonstrated in Fig. 3A, MLH1 was rapidly proteolyzed into a ϳ45-kDa product (indicated by an arrow) in cells treated with TRAIL (within 4 h) or with etoposide (within 12 h). The caspase substrate protein kinase C␦ (55) was used as a positive control in these studies and was cleaved into its characteristic fragment (Fig. 3A, indicated by an arrow) in apoptotic cells. Importantly, the size of the apoptotic MLH1 cleavage product was similar to the larger of the two products generated by caspase-3 in vitro. Furthermore, MLH1 proteolysis occurred at a similar time after exposure to apoptotic stimuli as did caspase-3 activation (seen as a reduction in procaspase-3 levels because of proteolytic processing). Taken together, these findings suggest that MLH1 might be cleaved by caspase-3 in apoptotic cells. Consistent with this notion, the broad spectrum caspase inhibitor Z-VAD-fmk potently suppressed MLH1 cleavage in PC-3 cells treated with TRAIL or etoposide (Fig. 3B, upper panels) or in TSU-Pr1 cells treated with these same apoptotic stimuli (lower panels). To evaluate specifically the role of caspase-3 in the apoptotic proteolysis of MLH1, we treated caspase-3-deficient MCF-7 breast carcinoma cells (56) with 10 ng/ml TNF-␣ and 1 g/ml cycloheximide for 0 -12 h. As shown in Fig. 3C, MLH1 was not cleaved in caspase-3-deficient MCF-7 cells treated with TNF-␣, while RAD21, a known substrate of multiple caspases (25), was proteolyzed in these cells. In addition, MLH1 was not cleaved in MCF-7 cells treated with etoposide (data not shown). These results indicate that MLH1 is a specific proteolytic target of caspase-3 that is cleaved in cancer cells in response to diverse apoptotic stimuli.
Caspase Proteolysis of MLH1 Triggers Its Partial Relocalization from the Nucleus to the Cytoplasm and Generates a Proapoptotic Carboxyl-terminal Product-To begin to assess the functional consequences of MLH1 proteolysis by caspase-3, we transiently transfected PC-3 prostate carcinoma cells with GFP-tagged constructs encoding WT MLH1 or the amino-terminal (amino acids 1-418, N-MLH1) or carboxyl-terminal (amino acids 419 -756, C-MLH1) caspase cleavage products. As shown in Fig. 4A (upper panels), WT MLH1 and C-MLH1 were expressed in the nuclei of transfected cells, while N-MLH1 was found in both the nucleus and the cytoplasm. The nuclei (Fig.  4A, lower panels) of cells transfected with WT MLH1 or N-MLH1 construct were intact (i.e. non-apoptotic). Although most cells transfected with C-MLH1 also had intact nuclei, a subset of these cells had fragmented, apoptotic nuclei (Fig. 4B, (57)) with each of the MLH1 constructs revealed that C-MLH1 was proapoptotic (Fig. 4C). In contrast, WT MLH1 and N-MLH1 did not induce apoptosis above background levels observed in vector-transfected cells. Overall, these results indicate that caspase cleavage of MLH1 alters its subcellular localization and produces a proapoptotic carboxyl-terminal product.
A Caspase Cleavage-resistant MLH1 Mutant Protects against Apoptosis Induced by Etoposide-To examine whether the pro-

FIG. 2. Human MLH1 is selectively proteolyzed by caspase-3 at Asp 418 in vitro.
A, 35 S-labeled full-length human MLH1 is specifically cleaved into two products of ϳ45 and 40 kDa (indicated by arrows) by caspase-3 in vitro. 35 S-Labeled MLH1 was incubated with buffer control (C) or 2.5 or 25 ng of caspase-1, -2, -3, -6, -7, or -8 (C1-C8) for 1 h at 37°C. B, substitution of Asp 418 in MLH1 with a Glu residue produces a mutant MLH1 protein (D418E) that is resistant to cleavage by caspase-3 in vitro. 35 S-Labeled WT or mutant D418E MLH1 was incubated with buffer control or caspase-3 (2.5 or 25 ng) for 1 h at 37°C. FIG. 1. Identification of MLH1 as a putative caspase-3 substrate by small pool expression cloning. A, presence of a ϳ66-kDa protein (indicated by an asterisk) in 35 S-labeled protein pool 10 that is selectively cleaved by caspase-3 (C3), but not by buffer control (C) or caspase-1, -2, or -8 (C1, C2, or C8), into two proteolytic fragments, which are ϳ45 and 24 kDa in size (indicated by arrows). 35 S-Labeled protein pools were made from small pools of a human prostate adenocarcinoma cDNA library by coupled transcription/translation in vitro, and these 35 S-labeled protein pools were screened for caspase substrates as described under "Experimental Procedures." B, a single cDNA encoding the putative caspase-3 substrate was isolated from cDNA pool 10 by subdividing the pool and incubating the corresponding 35 S-labeled protein pools with caspase-3. Clone 10 4G encodes a ϳ66-kDa protein that is proteolyzed by caspase-3 into cleavage products (indicated by arrows) of similar size to those observed in A. Sequence analysis indicated that this clone is a partial MLH1 cDNA. apoptosis induction by TRAIL or etoposide. In contrast, mutant D418E MLH1-expressing cells were partly protected against etoposide-induced apoptosis, while their apoptotic response to TRAIL was largely unaffected. These results indicate that MLH1 proteolysis is a functionally important event in the execution of DNA damage-induced apoptosis. DISCUSSION We have identified the mismatch repair protein MLH1 as a novel, functionally relevant substrate of caspase-3 that is rapidly proteolyzed in cells induced to undergo apoptosis by stimuli that engage the extrinsic (TRAIL) or the intrinsic (etopo-side) apoptotic pathways. MLH1 is selectively cleaved by caspase-3 (but not by other caspases tested) in vitro and is not cleaved in apoptotic caspase-3-deficient MCF-7 cells. These results indicate that MLH1 is a specific substrate of caspase-3, one of the major downstream executioner caspases (1,2). Intriguingly, experiments performed with cells derived from caspase-3 knockout mice or with caspase-3-deficient MCF-7 cells have revealed that caspase-3 is required for apoptotic chromatin condensation and DNA fragmentation, perhaps because ICAD proteolysis, and therefore CAD activation, is impaired in the absence of caspase-3 (16,56,58). MLH1, then, can be added to a short list of caspase-3-specific substrates, which also includes other proteins involved in DNA repair such as RAD51, BLM, and topoisomerase I (27,30,59). Human MLH1 is cleaved at a DXXD consensus caspase-3 cleavage motif (60, 61) (Asp-Lys-Thr-Asp 418 -2-Ile 419 ) near the middle region of the protein. Although the caspase-3 cleavage site (Asp 418 ) in human MLH1 is not conserved in non-human species (replaced by a glutamic acid residue in rodents), murine MLH1 is cleaved during apoptosis (data not shown), thereby suggesting that proteolysis of MLH1 may be a general feature of apoptosis.
Furthermore, our observation that the caspase cleavage-resistant D418E MLH1 mutant inhibited apoptosis induced by the chemotherapeutic drug etoposide suggests that MLH1 proteolysis is a functionally important event in DNA damageinduced apoptosis. Although MLH1 is cleaved in response to diverse apoptotic stimuli, the cleavage-resistant MLH1 mutant had little effect on TRAIL-induced apoptosis, thereby underscoring the specificity of our observations. How, then, might MLH1 proteolysis contribute to the execution of apoptosis initiated by DNA damage? In general, caspase cleavage of a protein can promote apoptosis by activating a latent proapoptotic function or by disrupting an antiapoptotic or prosurvival function (1,2). We observed that full-length MLH1 was unable to induce apoptosis in transiently transfected cancer cells, regardless of their endogenous MLH1 expression. These results are in agreement with those of Brieger and co-workers (62), whereas a second group reported a modest proapoptotic effect of fulllength MLH1 in transient transfection experiments (63). In contrast, we found that the carboxyl-terminal MLH1 caspase cleavage product (C-MLH1), but not the amino-terminal cleavage product (N-MLH1), induced apoptosis in transfected cells. These findings suggest that caspase proteolysis of MLH1 unmasks its cell death function by generating a proapoptotic carboxyl-terminal product.
Consistent with our results, MLH1 has been demonstrated previously to mediate the apoptotic response to DNA damage, including damage induced by alkylating agents, base analogues, adduct-forming drugs such as cisplatin, and etoposide, which induces DNA double-strand breaks (43)(44)(45)(46)(47)(48). Cells deficient in MLH1 or its interacting protein PMS2 are resistant to apoptosis induced by cisplatin and have a defective p73 or p53 response to this drug (46,64,65). Hence, the tolerance of MMR-deficient cells to these agents reflects, at least in part, defective activation of the p53/p73 apoptotic response. MMR proteins, then, are key components of the apoptotic response to DNA damage. Indeed, PMS2 directly binds/stabilizes p73 and enhances its proapoptotic activity, suggesting that MMR proteins may target components of the cell death machinery to damaged DNA and trigger their activation (64). Additional support for this notion comes from the observation that MLH1 binds to the proapoptotic oncoprotein c-MYC (66). Furthermore, MLH1 is a component of the BRCA1-associated genome surveillance complex (which also includes MSH2, MSH6, ATM, BLM, RAD50, and other components) that has been implicated in the detection and repair of damaged DNA (67). Interestingly, several of the components of this complex (BRCA1, BLM, ATM, and MLH1) are cleaved by caspases (28,30,31). Like MLH1, BRCA1 cleavage by caspases promotes DNA damage-induced apoptosis (31). Finally, caspase proteolysis of RAD21, a cohesin component that repairs DNA double-strand breaks, produces a carboxyl-terminal cleavage product that induces apoptosis (25,26). The results reported here indicate for the first time that proteolysis of MLH1 plays a similarly important role in activating its proapoptotic function in response to DNA damage, perhaps by deregulating its interactions with other MMR proteins, components of the BRCA1 complex, or apoptotic proteins. Taken together, these observations suggest that components of the DNA repair machinery are converted to apoptotic executioners by caspase cleavage, an elegant and efficient strategy to signal DNA damage-induced apoptosis.