Mouse and Human Granzyme B Have Distinct Tetrapeptide Specificities and Abilities to Recruit the Bid Pathway*

Granzyme B is an important mediator of cytotoxic lymphocyte granule-induced death of target cells, accomplishing this through cleavage of Bid and cleavage and activation of caspases as well as direct cleavage of downstream substrates. Significant controversy exists regarding the primary pathways used by granzyme B to induce cell death, perhaps arising from the use of different protease/substrate combinations in different studies. The primary sequence of human, rat, and mouse granzymes B is well conserved, and the substrate specificity and crystal structure of the human and rat proteases are extremely similar. Although little is known about the substrate specificity of mouse granzyme B, recent studies suggest that it may differ significantly from the human protease. In these studies we show that the specificities of human and mouse granzymes B differ significantly. Human and mouse granzyme B cleave species-specific procaspase-3 more efficiently than the unmatched substrates. The distinct specificities of human and mouse granzyme B highlight a previously unappreciated requirement for Asp192 in the acquisition of catalytic activity upon cleavage of procaspase-3 at Asp175. Although human granzyme B efficiently cleaves human or mouse Bid, these substrates are highly resistant to cleavage by the mouse protease, strongly indicating that the Bid pathway is not a major primary mediator of the effects of mouse granzyme B. These studies provide important insights into the substrate specificity and function of the granzyme B pathway in different species and highlight that caution is essential when designing and interpreting experiments with different forms of granzyme B.

is accomplished through a set of shared cytotoxic pathways that recruit the conserved death machinery of target cells. Upon recognition of target cells by cytotoxic lymphocytes, target cell death is induced through pathways initiated by death receptors or the granule exocytosis pathway (3). Granule-mediated target cell killing utilizes several granule components with distinct activities (2). The most abundant components present in cytotoxic lymphocytes are perforin and a family of serine proteases termed granzymes. Granzyme B (GrB) is one of the granzymes that has been demonstrated to function in initiation of cell death by CTL and natural killer cells. This protease utilizes several pathways to induce target cell death, including recruitment of the mitochondrial pathway through the cleavage of Bid (4 -7), activation of caspases (8 -10), and direct cleavage of downstream substrates (11)(12)(13)(14)(15)(16).
The primary substrate specificities of the human and rat proteases have been defined and are extremely similar (17,18). GrB prefers substrates containing P 4 to P 1 amino acids Ile/Val, Glu/ Met/Gln, Pro/Xaa, and aspartic acid N-terminal to the proteolytic cleavage. Non-charged amino acids are preferred at P 1 Ј, and Ser, Ala, or Gly are preferred at P 2 Ј. Although mouse GrB is very similar in sequence to the human and rat proteases and induces apoptotic death of target cells (19), little is known about the substrate specificity of the mouse protease. Several recent studies have suggested possible subtle differences in the specificity of mouse and human GrB (20,21). Because many studies performed to date have utilized mouse and human GrB interchangeably across species, frequently generating controversial results, defining the specificity of the mouse protease is an important priority.
In these studies we have defined the tetrapeptide specificity of mouse GrB and have shown that this differs significantly from human GrB. Such differences are particularly evident for procaspase-3 and Bid, both GrB substrates directly implicated in the apoptotic pathway. Human and mouse granzyme B cleave species-specific procaspase-3 more efficiently than the unmatched substrates. Furthermore, the distinct specificities of the human and mouse protease highlight a previously unappreciated requirement for Asp 192 in the acquisition of catalytic activity upon cleavage of procaspase-3 at Asp 175 . Our studies both in vitro and using intact cells show that human and mouse Bid are not cleaved by mouse GrB (although being efficiently cleaved by human GrB) and indicate that the Bid pathway is not a prominent primary mediator of the effects of mouse GrB. Taken together, the findings described in this manuscript (i) provide new and important insights into the specificity and function of the GrB pathway in different species and (ii) emphasize that caution should be exercised in the design and interpretation of experiments using proteases from different species.

Preparation of Recombinant Enzymes-Recombinant human
GrB was obtained using a baculovirus expression system and purified to homogeneity as described in detail previously (22). The samples were tested for homogeneity on SDS-polyacrylamide gels and for activity using a continuous fluorometric assay with Ac-IETD-7-amino-4-methylcouramin (AMC) as substrate, similar to those described for caspases (23,24). The recombinant human GrB was indistinguishable from native enzyme with regard to K m and K i values for key substrates and inhibitors. Purified recombinant mouse GrB, expressed in Pichia pastoris was purchased from Sigma. Recombinant human caspase 8 was prepared and characterized as described (17).
Determination of Catalytically Active Enzyme Concentration-Concentrations of active enzyme were determined using a potent reversible inhibitor under so called "tight binding" conditions. This relies on the principle that at enzyme concentrations far above K i , most of the inhibitor is enzyme-bound such that titrations initially resemble straight lines nearly independent of K i , with a tangent whose x-axis intercept is the total active enzyme concentration.
where E t is total active enzyme, and I t is total inhibitor (25). Enzyme titrations were conducted using a peptide-mimetic analog of IEPD-aldehyde, previously described as compound 6 (26). It has K i values of 8 and 25 nM for human and mouse GrB, respectively. Observed first-order rate constants were determined using the fluorescent substrate IETD-AMC at 10 M in a buffer composed of 0.1 M HEPES, 0.1% CHAPS, and 10% sucrose, pH 7.5, at room temperature. Nominal enzyme concentrations were 100 -600 nM.
Enzyme concentrations returned from fits to the above equation were in good agreement with nominal values predicted simply from protein concentration and mass 30,000 Da, indicating that essentially all the enzyme in both human and mouse preparations was active. Based on titrations conducted at 3 separate enzyme concentrations, total active human enzyme appeared to be 1.377 Ϯ 0.161 times the nominal prediction and 1.166 Ϯ 0.108 times nominal for mouse. Because the nominal concentrations were quite satisfactory within experimental error, they were used throughout this article. As a corollary, the results indicate that k cat /K m for IETD-AMC is 516 Ϯ 106 M Ϫ1 s Ϫ1 for human and 159 Ϯ 14 M Ϫ1 s Ϫ1 for mouse GrB based on active enzyme concentration.
Determination of Human and Mouse GrB Specificities-A P 1 -7-amino-4-carbamoylmethylcouramin (ACC) diverse tetrapeptide library provided by Charles S. Craik (University of California, San Francisco, CA) was used for the identification of the optimal P 1 amino acid (27). Briefly, this library consists of 20 samples in which only the amino acid in P 1 is held constant (excluding cysteine and including norleucine) and the P 2 , P 3 , and P 4 positions consist of an equimolar mixture of 19 amino acids for a total of 6859 substrate sequences per sample. Sub-strates with ACC as the leaving group show kinetic profiles comparable with those with an AMC leaving group. The substrate samples (final concentration of 50 M substrate mixture/7. 2 nM each compound) dissolved in Me 2 SO were added to 96-well plates and mixed with human and mouse GrB (600 nM final concentration). The appearance of ACC was monitored continuously at room temperature in a Tecan 96-well plate reader, with excitation at 390 nm and emission at 460 nm. Assays were performed in a final volume of 100 l, and the composition of the buffer was 0.1 M HEPES, 10% sucrose, 0.1% CHAPS, and 10 mM dithiothreitol, pH 7.5.
A positional scanning synthetic combinatorial library with the general structure Ac-Xaa-Xaa-Xaa-Asp-AMC was synthesized at Merck Research Laboratories and used to determine GrB specificities in P 2 , P 3 , and P 4 positions. This library has been used in the past for the caspase family and native human GrB and consists of 3 sublibraries of 20 samples each (400 compounds per sample in approximately equimolar concentrations). The screening was carried out as described above but using a 100 M substrate mixture/0.25 M concentrations of each compound and 50 nM human and mouse GrB.
Generation and Cleavage of [ 35 S]Methionine-labeled Substrates-[ 35 S]Methionine-labeled procaspase 3 and Bid (both human and mouse) were generated by coupled in vitro transcription/translation (IVTT) using appropriate full-length cDNAs. For in vitro cleavage reactions, the radiolabeled proteins were diluted in buffer A containing 10 mM Hepes, pH 7.4, 2 mM EDTA, and 1% Nonidet P40 and incubated at 37°C for 1 h in the absence or presence of purified human or mouse GrB as described (12) (at concentrations specified in the legends to Figs. 2-5), or 25 nM purified human caspase-8. 5 mM dithiothreitol was added to the caspase-8 reactions. Unless otherwise specified in the figure legends, no iodoacetamide (IAA) was added to the GrB IVTT cleavage reactions. Reactions were terminated by adding gel application buffer and boiling. The samples were electrophoresed on SDS-polyacrylamide gels, and intact proteins and their cleaved fragments were detected by fluorography.
Site-directed Mutagenesis to Confirm GrB Cleavage Sites-cDNA encoding human procaspase-3 was used as a template to generate clones with (i) a Cys 163 to Ser mutation (in the active site cysteine of procaspase-3), (ii) a Thr 174 to Phe (to "murinize" human procaspase-3), (iii) Asp 175 or Asp 192 to Ala (substitutions in the putative GrB substrate P 1 positions), and (iv) the double mutation Asp 180 to Ala and Asp 181 to Ala. Mutagenesis was performed using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). In each case, incorporation of the desired point mutation without additional changes to the original sequence was confirmed by sequencing. Mutated constructs were used for IVTT reactions, and the radiolabeled protein products were incubated in the absence or presence of GrB and electrophoresed as described above. cDNA encoding mouse procaspase-3 was similarly mutated by changing Asp 175 to Ala to confirm the GrB cleavage site; these data were used for Table 1.  Table 1.

Cleavage of Endogenous Bid in Human and Mouse Cell Lysates-C 2 C 12 mouse muscle cells and Jurkat cells (human)
were maintained in culture using standard procedures before washing in PBS and lysing in buffer A. Cleavage reactions were performed in buffer A with lysate at 1 mg/ml in the presence of the indicated combinations (see Fig. 4B) of 5 mM dithiothreitol (to facilitate activation of the endogenous caspase cascade), 5 mM IAA (to inactivate endogenous caspases), and purified recombinant human GrB, mouse GrB, or caspase-8. Reaction mixtures were incubated at 37°C for 1 h, terminated by the addition of gel application buffer, and electrophoresed on 13% SDS-polyacrylamide gels. After transferring the lysates onto nitrocellulose membranes, Bid was immunoblotted using polyclonal antibodies (R&D Systems, Minneapolis, MN or BD Biosciences) and visualized using enhanced chemiluminescence (Pierce).

Cleavage of Bid and Procaspase-3 in Intact Cultured
Human and Mouse Cells-Human and mouse GrB were delivered to intact Jurkat and C 2 C 12 cells using the Bioporter protein delivery system (Sigma) as previously described for delivery of granzymes and other proteases (28 -30). Briefly, Jurkat cells were washed in Hanks' balanced salt solution (HBSS) and resuspended at 1 ϫ 10 6 cells/ml in medium B (HBSS supplemented with 10 mM Hepes, pH 7.4, and 10 mM MgSO 4 ). C 2 C 12 cells were plated in 6-well plates, and when ϳ80% confluent the cells were washed as above, and the medium was replaced with medium B. Bioporter reagent was preincubated with human or mouse GrB in 40-l volumes according to the manufacturer's directions before diluting with medium B and mixing with an equal volume of cell suspension (Jurkat cells) or culture medium (C 2 C 12 cells). 5 mM IAA was added to each reaction to inactivate endogenous caspases. All cell incubations were performed for 2 h at 37°C in a final volume of 1 ml and contained either Bioporter alone, Bioporter plus human GrB, or Bioporter plus mouse GrB. The amounts of GrB varied from 300 to 3000 ng/ml; the actual amounts used are specified in Fig. 5 legend. Reactions were terminated by scraping the C 2 C 12 cells or lysing the pelleted Jurkat cells at 4°C in buffer A plus a protease inhibitor mixture and adding gel application buffer immediately. Equal numbers of cells (Jurkat cells) or amounts of cell protein (C 2 C 12 lysates) were loaded in each lane on 12 or 14% SDS-polyacrylamide gels (for detection of caspase-3 and Bid, respectively). Caspase-3 and Bid were immunoblotted as described above (the caspase-3 polyclonal antibody was from Cell Signaling, Danvers, MA).

Use of Positional Scanning Synthetic Substrate Combinatorial Libraries to Define the Specificity of Mouse GrB-Rat,
human, and mouse granzymes B are similar in sequence and have been used interchangeably to induce apoptosis of a variety of target cells. Although the tetrapeptide and macromolecular specificity of the human and rat proteases have been well defined, there is currently little known about the tetrapeptide and macromolecular specificity of mouse GrB. We, therefore, initially employed two positional scanning synthetic substrate combinatorial libraries, a P 1 -ACC diverse tetrapeptide library and a library with the general structure Ac-Xaa-Xaa-Xaa-Asp-AMC, previously used to define the tetrapeptide specificities of native human (17) and recombinant rat (18) granzymes B to define the P 1 -P 4 specificity of mouse GrB (Fig. 1). The overall similarity of specificity was clear, with an absolute requirement for Asp in P 1 and a similar but somewhat broader specificity of the mouse protease in P 4 (having a similar preference for isoleucine and valine, an increased preference for leucine, and also tolerating norleucine, Phe, and Pro in this position). P 3 in mouse had a preference for Glu and was otherwise similar for less well tolerated amino acids (Asp, Gln, Gly, Ser). However, P 3 FIGURE 1. Substrate specificities of recombinant human and mouse GrB. Specificities for recombinant human and mouse GrB were determined using two positional scanning libraries, a P 1 -ACC diverse tetrapeptide library to define P 1 and an Ac-Xaa-Xaa-Xaa-Asp-AMC library to define P 2 -P 4 . The y axis represents the rate of ACC or AMC production as a percentage of the maximum rate observed in each experiment. The x axis shows the positionally defined amino acids as represented by the one-letter code. dA, D-alanine; X, norleucine.
in the mouse also accommodated Leu, Val, Phe, Tyr, and Thr, which were not tolerated by the human protease. The most striking differences in the substrate specificity of the human and mouse granzymes B was observed in P 2 . Although the human protease has the unusual and characteristic preference for Pro, Ser, and Thr in P 2 but broadly tolerates many other amino acids, this pattern was completely different in the mouse, where Pro, Ser, and Thr were not preferred. Indeed, Phe, norleucine, Tyr, and Gln were preferred at this site. This finding was somewhat unexpected, as the overall sequence similarity of human, rat, and mouse GrBs is ϳ80% (see "Discussion").
Human and Mouse GrB Both Cleave Human Procaspase-3 but Differ in Their Ability to Generate Active Caspase-3-To determine whether the differences in tetrapeptide cleavage profiles were representative of macromolecular substrates, we examined the cleavage of two proteins that are involved in the apoptotic pathway, Bid and procaspase-3. Both human and mouse GrB cleaved the human caspase-3 precursor, although the former cleaved ϳ6-fold more efficiently than the latter (k cat /K m of 2.3 ϫ 10 3 M Ϫ1 ⅐s Ϫ1 and 0.4 ϫ 10 3 M Ϫ1 ⅐s Ϫ1 , respectively; Table 1). Of note, different fragments were reproducibly generated by the 2 proteases; the signature GrB-induced 21-kDa fragment was generated by both, whereas the 17-kDa fragment (representing autocatalytic removal of the caspase-3 prodomain) was only detected when caspase-3 was cleaved by human GrB (Fig. 2). Because the tetrapeptide specificity of the two proteases differs significantly in P 2 and the Thr at P 2 in the defined IETD 175 cleavage site of human procaspase-3 is preferred by human GrB but poorly tolerated by mouse GrB, we examined whether this failure to achieve autocatalytic removal of the prodomain after mouse GrB cleavage might be caused by cleavage of procaspase-3 at an alternate, closely spaced site. We initially addressed the effect of mutation at IETD 175 on cleavage by mouse and human GrB. This mutation abolished cleavage by both proteases, confirming that both granzymes target this site (Fig. 2, upper  left panel). To confirm that the p17 procaspase-3 fragment observed after cleavage with human GrB in this system required catalytic activity of caspase-3, we mutated the catalytic cysteine at position 163 to serine. Generation of the 17-kDa fragment by human GrB was abolished in this mutant, confirming that acquisition of caspase-3 activity after cleavage by human GrB in these assays is required for generating the 17-kDa fragment (Fig. 2, upper left panel).
Because cleavage of human procaspase-3 by mouse GrB did not result in caspase-3 activation, we addressed whether mouse GrB interacts with this substrate in a way that prevents acquisition of enzymatic activity. Of particular importance in this regard is the proximity of the IETD 175 cleavage site to the active site cysteine 163 . We, therefore, addressed (i) whether the kinetics of acquisition of caspase-3 activity differed for mouse and human GrB and (ii) whether there were any stable complexes formed between mouse GrB and procaspase-3 that were not observed with human GrB. Even prolonged incubation of caspase-3 cleaved by mouse GrB failed to generate the 17-kDa   [ 35 S]Methionine-radiolabeled human and mouse Bid were generated by in vitro transcription translation reactions, cleaved with human and mouse GrB (in the absence of IAA), and electrophoresed as described under "Experimental Procedures." k cat /K m values were calculated from the scanned densitometric data as described. Unless indicated "not cleaved" or marked with a reference number, we confirmed the GrB cleavage site by mutagenesis as described under "Experimental Procedures." fragment (data not shown). Furthermore, no complex of mouse GrB with either intact or cleaved caspase-3 was observed (data not shown). Recent studies have highlighted an inhibitory loop at the N terminus of the small subunit of caspase-3 that appears to influence caspase-3 autoactivation (31). Just downstream of this "safety catch" loop is another site that is potentially cleavable by GrB (IEAD 192 ). When the Asp at position 192 was mutated to Ala, no effect on cleavage by mouse GrB was noted.

Antigen
In contrast, this mutation completely abolished generation of the 17-kDa large subunit after initial cleavage by human GrB (Fig. 2, right panel), strongly implicating a role for Asp 192 in the activation of caspase-3 by human GrB. Interestingly, this mutation at Asp 192 also prevents the activation of caspase-3 by caspase-8. Using several gel systems we were unable to visualize a difference in the sizes of the small subunit consistent with a difference of 17 amino acids, suggesting that the role of Asp 192 may not involve a cleavage event. Importantly, murinizing the GrB cleavage site in human procaspase-3 so that it contained a more preferred tetrapeptide (mutating Thr 174 to Phe) enhanced the efficiency of cleavage of procaspase-3 by mouse GrB, with improved production of the 17-kDa fragment (Fig. 2, right panel). In contrast, cleavage of the murinized substrate by human GrB was inhibited, with minimal generation of the 17-kDa cleavage fragment. Interestingly, this mutation completely abolished the cleavage of procaspase-3 by caspase-8. Taken together, these data demonstrate that human and mouse granzymes B differ in the efficiency with which they cleave human procaspase-3 and, most importantly, in their ability to generate the active caspase. Mouse GrB Efficiently Cleaves and Activates Mouse Procaspase-3-Although human and mouse procaspase-3 are strikingly conserved, there are changes in the sequence of the two proteases around the safety catch region. Thus, the VDDD sequence in human procaspase-3 is replaced by TDEE in the mouse. We, therefore, addressed whether mouse GrB induced the cleavage and activation of mouse procaspase-3. In striking contrast to its relatively inefficient cleavage of human procaspase-3 and its failure to generate the autocatalytically active form of this molecule, mouse GrB efficiently cleaved mouse procaspase-3, producing predominantly the 17-kDa fragment (Fig. 3, lanes 1-6). Interestingly, mouse GrB was as efficient in cleaving and activating mouse procaspase-3 as was human GrB on its human counterpart (k cat /K m of 5.0 ϫ 10 3 M Ϫ1 ⅐s Ϫ1 compared with 2.3 ϫ 10 3 M Ϫ1 ⅐s Ϫ1 ). To define whether Asp 192 was also required for acquisition of proteolytic activity by mouse caspase-3, we mutated this residue to alanine and addressed the effects on cleavage by mouse GrB. Interestingly, generation of the 17-kDa fragment was also abolished in this setting (data not shown), demonstrating that Asp 192 has a critical role in acquisition of caspase-3 activity across species. Of note, the addition of 5 mM IAA to the reaction mixes (Fig. 3, lanes 7-12) abolished the autocatalytic removal of the caspase-3 prodomain; hence, only the signature 21-kDa GrB cleavage fragment was detected under these conditions.
Human GrB Directly Cleaves Bid, Whereas Mouse GrB Requires Intervening Activation of the Caspase Cascade-We next addressed cleavage of mouse and human Bid by GrB. Human GrB cleaved human Bid efficiently (k cat /K m value of 5.3 ϫ 10 3 M Ϫ1 ⅐s Ϫ1 ; Table 1); this substrate was completely resistant to cleavage by mouse GrB. We also performed similar experiments on mouse Bid, which has a high overall sequence similarity to its human counterpart, with identical caspase-8 cleavage sites (LQTD 60 ) and very similar GrB cleavage sites (IEAD 75 in human versus IEPD 75 in mouse). As seen with the human protein, mouse Bid was efficiently cleaved by human GrB (k cat /K m value of 1.4 ϫ 10 4 M Ϫ1 ⅐s Ϫ1 ) but was not cleaved with the mouse protease ( Fig. 4A and Table 1). Interestingly, when human or mouse GrB was added to extracts of human or mouse cells and endogenous Bid cleavage was assayed by immunoblotting, human GrB could directly cause processing of Bid even in extracts in which caspase activity was prevented by prior treatment with IAA. The prominent fragment detected migrated more rapidly on SDS-PAGE than the fragment generated by purified caspase-8, consistent with the GrB cleavage at IEAD 75 defined previously and demonstrating that GrB efficiently and directly cleaves Bid in these extracts Fig. 4B). In contrast, mouse GrB was unable to cleave Bid in cell lysates in which caspases were inactivated but induced Bid cleavage (albeit inefficiently) in lysates supporting caspase activity. It is noteworthy that the fragment generated under these circumstances comigrated with that observed in lysates to which purified caspase-8 was added (Fig. 4B).
Because post-translational modifications may influence susceptibility to proteolysis, we established whether the in vitro lysate findings with Bid are also true in intact cells. Human and mouse GrB were delivered into intact Jurkat (human) or C 2 C 12 (mouse) cells using the Bioporter protein delivery system after first performing dose-response experiments to optimize conditions and to validate the system (Fig. 5, B and C). Consistent with the in vitro findings depicted in Fig. 4, data obtained in the intact cell system showed that although human and mouse Bid were both cleaved by human GrB, neither was cleaved by mouse GrB (Fig. 5A, upper panel). Of note, IAA was added to the reactions to inactivate endogenous caspases. To assess whether human and mouse GrB were indeed active in this system (especially since no cleavage of Bid was noted with mouse GrB), the  1-3 and 7-9) and mouse (lanes 4 -6 and 10 -12) procaspase-3 were generated by IVTT. The radiolabeled proteins were incubated with 100 nM concentrations of purified human (hu) or mouse (m) GrB in the absence (lanes 1-6) or presence (lanes 7-12) of 5 mM IAA. After 1 h at 37°C, the reactions were terminated, and gel samples were electrophoresed on 12% SDS-PAGE. Intact proteins and cleavage fragments were detected by fluorography.
same gel samples were also immunoblotted with an antibody recognizing procaspase-3 and its cleavage fragments (Fig. 5A, lower  panel). Both human and mouse GrB cleaved procaspase-3 to its 21-kDa form, confirming catalytic activity of both GrB species in the intact target cells. The 17-kDa fragment generated by autocatalytic removal of the caspase-3 prodomain was not detected because IAA was included in the cell incubations; these data are consistent with the cleavage of IVTT procaspase-3 in vitro, performed in the absence or presence of IAA (Fig. 3). Thus, although mouse GrB effectively activates mouse caspase-3, it is incapable of directly recruiting the Bid pathway. This may explain in part some of the differences previously observed regarding the requirement for Bid in the granzyme pathway when using human and mouse granzymes B.

DISCUSSION
These studies demonstrate that although human and mouse granzymes B share significant sequence similarity, the proteases are quite different in their overall tetrapeptide and macromolecular specificities. Using positional scanning synthetic substrate combinatorial libraries, several striking observations about the mouse and human proteases were made. (i) Both proteases have an absolute requirement for cleavage after aspartic acid; (ii) although the human protease prefers Pro, Ser, Thr, Ala, Asn, and Gln in the P 2 position, mouse GrB is poorly tolerant of Pro, Ser, Thr, Ala, and Asn and favors large hydrophobic residues (Phe and Tyr) in this position; (iii) there is broader tolerance for additional amino acid residues in the P 4 position in the mouse protease, which shows extended specificity to include Leu, Phe, and Pro over the human protease.
Cleavage of human procaspase-3 by mouse GrB at IETD 175 is ϳ6-fold less efficient than by the human protease. This is apparently due at least in part to primary differences in tetrapeptide specificities, as murinizing the GrB cleavage site in human procascpase-3 to a sequence preferred by mouse GrB (IETD to IEFD) greatly enhances the efficiency of human procaspase-3 cleavage by mouse GrB. It should be noted, however, that murinized human procaspase-3 is not the physiological version of procaspase-3 in mouse, which is identical to human procaspase-3 in almost 30 positions before the cleavage site (and has an identical IETD 175 to human). The observation that mouse GrB cleaves mouse procaspase-3 more efficiently than it does human procaspase-3 suggests that additional determinants of macromolecular specificity residing outside the region upstream of the scissile bond are distinct in the mouse and human granzymes and account for these changes in specificity. Some of these determinants may relate to prime side residues, whereas others may be more distant (see a discussion of inhibitory loop below).  The Bioporter protein delivery system was used to deliver human (hu) and mouse (m) GrB (both at 1500 ng/ml in panel A or at the indicated concentrations in panels B and C) to intact C 2 C 12 (panel A) or Jurkat (panels A-C) cells. Those cell mixtures without GrB were incubated in the presence of the Bioporter reagent alone. All incubations were performed for 2 h at 37°C before terminating the reactions, electrophoresing the gel samples, and immunoblotting Bid and caspase-3 as described under "Experimental Procedures." In panel A, for both Bid and caspase-3, data from a single autorad was used, but intervening lanes between the Jurkat and C 2 C 12 sets were spliced out. Similarly, the blots shown in panels B and C were also from a single autorad in each case that was spliced between the 300 and 1500 ng/ml hu GrB lanes. DTT, dithiothreitol.
The crystal structures of both human and rat GrB are available, and residues that determine substrate specificity in the different subsites have been defined (22,32,33). When residues that contact the substrate in human and rat proteins are compared against residues at equivalent positions in the mouse protease, the similarities and differences are very striking ( Table 2). All three proteases have Arg in position 226, consistent with the absolute requirement for Asp in the P 1 position in all 3 proteases. Interestingly, despite the significant differences in specificity of human and mouse GrB noted above, most of the residues that primarily contact the substrate in the S 2 -S 4 subsites are identical in the human and mouse proteases. The following differences were notable. (i) Asn 218 in the human protease is replaced by Lys in the mouse. (ii) S 4 subsite residues that contact substrate were significantly different in the human and mouse proteases. Thus, Tyr 174 in the human is an Arg residue in the mouse protease, whereas Arg 217 in human is Tyr in the mouse, and (iii) there were also significant differences on the prime side. Thus, mouse differs from human by having an Ala and Ile in positions 40 and 41, in comparison to dibasic Lys and Arg in human GrB. The accumulation of positively charged residues around this subsite in the human protease has been proposed to account for the aversion of GrB to arginine and lysine in the prime position. It is of interest that several of the sites in macromolecules that are efficiently cleaved by mouse GrB do contain positively charged amino acids on the prime side. Taken together, these data indicate that human and mouse GrB have some critical differences in structure that likely play a role in determining novel specificity against tetrapeptide and macromolecular substrates.
The change in Asn 218 is of particular interest. Mutations in this residue in combination with other mutations have been shown to alter P 2 -P 4 extended specificity. Such mutations were noted to change extended specificity preferentially over catalytic power. Asn 218 plays a critical role in substrate specificity, not limited to P 3 specificity. In the Craik study (34) the Asn 218 mutation introduced cooperative effects in substrate recognition by decoupling the independence between P 2 and P 4 positions. It is of particular interest that when Craik and co-workers (34) mutated Ile 99 and Asn 218 in rat GrB, they generated a protease with specificity changes that appeared similar to those that we have observed here for mouse GrB (34). Thus, in those mutants, Pro, Ser, and Thr are not preferred in P 2 , whereas norleucine, Phe, and Tyr are. Furthermore, the P 4 specificities are also very similar to those defined here for the mouse protease. We propose that the changes in the mouse P 2 and prime side residues play important roles in determining the unique specificity of mouse GrB.
To address the cleavage of macromolecular substrates by human and mouse GrB, we studied procaspase-3 and Bid, which are substrates for both upstream caspases and GrB and whose cleavage plays a large role in apoptosis. Procaspase-3 is cleaved by both mouse and human GrB at IETD 175 , generating a 21-kDa fragment. Surprisingly, the second signature fragment at 17 kDa, generated by autocatalytic cleavage of the N-terminal prodomain, is formed only when procaspase-3 is cleaved by human GrB and not by mouse. Mutational analysis demonstrated that both proteases cleave procaspase-3 at IETD 175 and confirmed that generation of the 17-kDa fragment requires the catalytic site cysteine. The distinct activities of the human and mouse granzymes B, with both able to cleave at Asp 175 but only human able to generate significant autocatalytic activity of human procaspase-3, points out a previously unrecognized intermediate step in caspase-3 activation. Several potential explanations for this observation exist, including the existence of secondary cleavage sites exclusively cleaved by human GrB. Interestingly, consensus sites for cleavage by human GrB and caspase-8 exist just downstream of the IETD cleavage site at VDDD 181 and at VEAD 192 . Of note, VDDD 181 comprises the inhibitory loop in caspase-3, which functions to limit autoactivation of the protease and has been termed the safety catch (31). When VDDD 181 was mutated to VDAA, activation by human and mouse GrB was augmented, and the 20-kDa fragment was now further processed by mouse GrB (Fig. 2, lower panel). We propose that human GrB effectively releases the inhibition of VDDD, whereas mouse GrB does not. It is possible that a secondary cleavage event at Asp 181 and/or Asp 192 removes the inhibitory residues.
Interestingly, these studies have also shown an unappreciated role for Asp 192 in caspase-3 activation. When VEAD 192 was mutated to VEAA, human procaspase-3 was still efficiently cleaved by human GrB and caspase-8, but generation of the 17-kDa fragment synonymous with autocatalysis was prevented for both proteases. The same observations were made for mouse GrB and mouse procaspase-3. Defining the mechanism by which the Asp 192 mutation prevents caspase-3 activation will require additional studies. The failure of mouse GrB to directly and efficiently activate human caspase-3 is, however, not predictive of its effects against mouse procaspase-3, which is both efficiently cleaved and activated by its speciesmatched protease. Mouse procaspase-3 has a different sequence in the safety catch region, which may account for these differences.
It is also noteworthy that although human GrB efficiently cleaves human and mouse Bid, mouse GrB cleaves neither. Similar observations have been suggested in other studies. For example, a recent study demonstrated that Bid was an ineffi-  (20). Also, Bredemeyer et al. (35) failed to observe Bid cleavage using a proteomic approach to define mouse GrB substrates. These data may explain in part the dissimilar conclusions reached when the role of Bid in GrBinduced death has been addressed previously. Studies examining the role of human GrB have concluded that Bid plays an important role downstream of GrB (4,5,7,36). In contrast, some studies using mouse GrB have concluded that GrB can cause death in the absence of Bid (37). It is noteworthy that Bid is cleaved downstream of mouse GrB but that such cleavage requires caspase activity and is marked by generation of the caspase-8 fragment rather than that generated by GrB. These data suggest that the proapoptotic function of mouse GrB is largely dependent on caspase activation. These data also demonstrate that despite significant homology between apoptotic proteases and their substrates in different species, caution should be exercised before assuming that protease-substrate pairs function similarly.