Expression of Human Recombinant Granzyme A Zymogen and Its Activation by the Cysteine Proteinase Cathepsin C*

Human granzyme A is one of the serine proteinases present in the granules of cytotoxic T lymphocytes and natural killer cells. Granzymes are synthesized as inac- tive proenzymes with an amino-terminal prodipeptide, which is processed during transport of granzymes to the cytotoxic granules, where they are stored as active proteinases. In this study, we explored the possibility of producing recombinant granzymes. Recombinant human granzyme A zymogen was expressed in several eu- karyotic cell lines (HepG2, Jurkat, and COS-1) after infection with a recombinant vaccinia virus containing full-length granzyme A cDNA. Immunoblot analysis of cell lysates showed that all infected cells produced a disulfide-linked homodimer of identical molecular weight as natural granzyme A. Infected HepG2 cells pro- duced the largest amount of this protease (approximate-ly 160 times more than lymphokine activated killer (LAK) cells). The recombinant protein only had high mannose type oligosaccharides as did the natural pro- tein. Although infected HepG2 and COS cells contained high granzyme A antigen levels, lysates from these cells did not show any granzyme A proteolytic activity. How-ever, the inactive proenzyme could be converted into active granzyme A by incubation with the thiol proteinase cathepsin C (dipeptidyl peptidase I). This study is the first to demonstrate expression of an active recombinant human cytotoxic lymphocyte pro- teinase and conversion of inactive progranzyme A into an active enzyme by cathepsin C. We suggest that a similar approach can be used for the production of other granzymes and related proteinases.

Activated cytotoxic T lymphocytes and natural killer cells contain specialized cytoplasmic granules, which are able to lyse susceptible targets (1,2). These so called "cytotoxic granules" among others contain the pore-forming protein perforin (3) and a family of highly homologous serine proteinases, termed granzymes (4). Seven different granzymes have been identified in the mouse (granzymes A-G) (4,5) and four in humans: granzyme A, granzyme B, granzyme H, and granzyme 3 (6 -13). Experiments with purified proteins as well as with knockout mice have indicated a direct involvement of granzymes A and B in target cell DNA fragmentation and apoptosis (14 -17).
Human and mouse granzymes all contain the catalytic triad consisting of histidine, aspartic acid, and serine, typical for serine proteinases (4,5). Furthermore, six cysteine residues involved in intrachain disulfide bond formation are conserved among the granzymes. The structure of granzyme A is unique among the granzymes in that it contains an additional disulfide bond and forms disulfide-linked homodimers via a free cysteine residue at position 76 (6,9). The nucleotide sequence of human granzyme A predicts a proteinase of 262 amino acids, a 26amino acid hydrophobic prepeptide and a short propeptide consisting of a glutamic acid and a lysine residue. Aminoterminal sequence analysis of granzyme A purified from cytotoxic granules revealed that it is stored as a fully processed, active, disulfide-linked, 50-kDa homodimer, i.e. without preand propeptide (4,6,9). Both catalytic centers of the homodimer are active (18) and preferably cleave synthetic substrates after lysine or arginine (6,19). Granzymes are structurally related to granular proteinases from myeloid leucocytes such as cathepsin G, elastase, and mast cell proteinases. Also, these proteinases are synthesized as inactive precursor molecules and stored in the granules as active proteinases (4,6). Furthermore, they also have a propeptide consisting of two charged amino acid residues (20 -22). In addition, there is indirect evidence that the thiol proteinase cathepsin C (also known as dipeptidyl peptidase I) is able to process these propeptides, as inhibition of this enzyme impairs the generation of active cathepsin G, elastase, or granzyme A (21).
We have expressed granzymes A and B in prokaryotic systems (23), but recombinant proteins obtained in this way were inactive because of incomplete folding and aggregation. Here we describe the expression of recombinant human granzyme A (rGA) 1 as a zymogen by HepG2 cells infected with a granzyme A recombinant vaccinia virus. Granzyme A zymogen could be converted into an active enzyme by cathepsin C, providing further evidence for the involvement of cathepsin C in the processing of granzyme A.
Cell Culture-Human 143 tk Ϫ fibroblasts and RK-13 rabbit kidney cells were maintained in Eagle's medium supplemented with 10% (v/v) fetal calf serum, streptamycin, penicillin, and nonessential amino acids. SV 40-transformed COS-1 monkey cells, the human T-helper cell line Jurkat, and the human liver cell line HepG2 were grown in Iscove's modified Dulbecco's medium supplemented with 10% (v/v) heat-inactivated fetal calf serum, streptamycin, penicillin, and ␤-mercaptoethanol (medium for Jurkat cells also contained 20 units/ml interleukin-2; Chiron, Emeryville, CA). Lymphokine-activated killer (LAK) cells were prepared by culturing human peripheral blood mononuclear cells (obtained from healthy donors by Percoll density gradient centrifugation) at a concentration of 0.5 ϫ 10 6 cells/ml for 7 days with 1,000 units/ml interleukin-2 in the same medium as for Jurkat cells. LAK cell lysate (20 ϫ 10 6 cells/ml) was prepared as described below for vacciniainfected cells.
Construction of Recombinant Vaccinia Virus for the Expression of Human Granzyme A-To obtain full-length granzyme A cDNA, including the nucleotides coding for the pre-and propeptide, specific primers containing the appropriate restriction sites were prepared based on the published granzyme A cDNA sequence (9). First strand cDNA was prepared from mRNA of LAK cells as described (23) and amplified with the polymerase chain reaction using the granzyme A primers. The amplified cDNA fragment was isolated, digested with the appropriate restriction enzymes, and ligated into the SacI ϫ SphI sites of the vaccinia virus recombinant vector p11k-ATA-18 (24) (Fig. 1). The authenticity of the cloned cDNA was confirmed by nucleotide sequence analysis (Sequenase kit; U.S. Biochemical Corp.). Isolation of plasmid DNA, conditions for digestion by restriction enzymes, and agarose gel electrophoresis were as described (25).
Granzyme A recombinant virus was prepared by homologous recombination of the granzyme A/p11k-ATA-18 plasmid with the temperature-sensitive vaccinia virus mutant ts7. Plaques were selected and purified as described (26). Briefly, subconfluent plates of human 143 tk Ϫ fibroblasts were infected for 1 h with vaccinia virus ts7 (0.1 plaqueforming unit/cell). Subsequently, cells were incubated with fresh medium for 2 h at the permissive temperature of 33°C, transfected with a calcium phosphate co-precipitate of wild type vaccinia virus DNA (100 ng/10 5 cells) and an equivalent amount of recombinant plasmid DNA, and incubated for 2 h at the nonpermissive temperature, 39.5°C. Medium was then removed, and cells were rinsed and incubated for 48 h at 39.5°C. Infected cells and culture medium were collected and freezethawed once. Dilutions of this material, containing recombinant virus, were used to select for tk Ϫ viral plaques by infecting human 143 tk Ϫ fibroblast cultures incubated in the presence of bromodeoxyuridine (100 g/ml). Single plaques were purified and amplified in RK-13 cells and stored at Ϫ20°C. This viral stock was titrated on RK-13 cells to determine the number of plaque-forming units/ml of virus stock.
Granzyme A Expression in Different Cell Lines-Cell lines were infected with granzyme A recombinant vaccinia virus by incubating subconfluent monolayers of HepG2 or COS-1 cells in 10-cm 2 dishes with 10 plaque-forming units of recombinant virus/cell. Jurkat cells, suspended at 10 ϫ 10 6 cells/ml, were incubated with 500 plaque-forming units/cell for 60 min at 37°C, diluted 10-fold with fresh medium, and then further incubated for 24 or 48 h at 37°C. Thereafter, cells and supernatant were harvested. The supernatant was centrifuged at 3,000 rpm to remove cellular debris, incubated with 1% (w/v, final concentration) Nonidet P-40 to inactivate viral particles, and stored at Ϫ70°C for further analysis. Cells were washed with phosphate-buffered saline (PBS), resuspended at a concentration of 2 ϫ 10 6 cells/ml in ice-cold lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1% (w/v) Nonidet P-40), gently mixed, and left for 30 min in melting ice. The mixture was then centrifuged for 10 min at 200 ϫ g (4°C) to remove the nuclei and the supernatant was stored at Ϫ70°C until used.
As a control, a recombinant vaccinia virus containing an insert coding for the serine proteinase domain of human factor-XII (rFXII.lpc) was used (27). Alternatively, the virus was omitted during the incubation.
Enzyme Assays-Granzyme A proteolytic activity was measured essentially as described (28). Briefly, 20 l of cell lysate was added to 100 l of 0.1 M Tris-HCl, pH 8.0, 0.5% (v/v) Nonidet P-40, 0.3 mM BLT, and 0.3 mM dithiobis(2-nitrobenzoic acid) (Aldrich-Chemie, Steinheim, Germany). The absorbance at 414 nm was read over 1 h at 37°C on a microplate reader. The amount of granzyme A activity was expressed as the increase of absorption at 414 nm/min.
Glycosidase Treatment-Total cell lysates (with a final concentration of 0.25% (v/v) Nonidet P-40) were first denatured in 0.5% (w/v) SDS at 95°C for 5 min, after which the SDS was quenched by addition of a 2-fold excess of Nonidet P-40. Then PNGase F at a final concentration of 3 milliunits/ml was added for 16 h at 37°C in the presence of 2 mM PMSF. Similarly, samples were treated with endo H by incubation for 16 h at 37°C in 9 mM NaAc, pH 5.5, in the presence of 2 mM PMSF and 3 milliunits of the enzyme/ml. All samples were centrifuged at 13,000 rpm for 2 min, prior to analysis by immunoblotting.
Activation of rGA Zymogen by Cathepsin C-LAK cell lysate or supernatant from HepG2 cells infected (for 48 h) with granzyme A or FXII.lpc recombinant viruses was dialyzed against 50 mM NaAc, pH 5.0, 30 mM NaCl, 1 mM EDTA, 10 mM cysteine (Pierce) at 4°C. Samples were then incubated for varying time intervals at 37°C with or without 0.02 units of purified bovine cathepsin C. Samples were neutralized and stored on ice until analyzed for granzyme A activity. This activity was analyzed either by incubating samples with benzamidin-Sepharose, whereafter bound fraction (i.e. active granzyme A) was analyzed on immunoblot (see below), or by determining the conversion of BLT substrate. To prevent interference of cysteine in the chromogenic assay, samples were dialyzed against PBS at 4°C prior to analysis.
Affinity Purification of Granzyme A Species Using Benzamidin-Sepharose-Cell lysates or supernatant was diluted in lysis buffer to a final volume of 250 -500 l and incubated with 10 l of benzamidin-Sepharose for 4 h at room temperature on a head-over-head rotator. The Sepharose beads were then washed four times with 1 ml of lysis buffer. The supernatant of the last washing step was carefully removed, after which SDS-sample buffer was added to the Sepharose beads. The mixtures were incubated for 5 min at 100°C and centrifuged for 3 min at 13,000 ϫ g. The supernatant was electrophoresed on SDS-polyacrylamide (12.5%, w/v) gels and subsequently analyzed by immunoblot (see below).
Immunoblotting-Cell lysates or benzamidin-Sepharose precipitates were separated on 12.5% (w/v) polyacrylamide gels. Proteins were then transferred onto nitrocellulose sheets (Schleicher and Schuell), which then were incubated for 30 min with blocking buffer, i.e. PBS containing 5% (w/v) nonfat dry milk (Protifar; Nutricia, Zoetermeer, the Netherlands) and 0.1% (w/v) Tween 20. The sheets were then incubated with monoclonal antibody GrA-8 (at 2.5 g/ml in the same buffer), for 14 h at room temperature. Sheets were washed by repeated (3 times) incubation for 10 min with PBS, 0.1% (w/v) Tween 20, and then probed with horseradish peroxidase-conjugated goat anti-mouse immunoglobulins diluted in blocking buffer for 2 h. After a wash with PBS, 0.1% (w/v) Tween, and one with PBS alone, sheets were developed for 2 min in chemiluminescent detection reagent (Amersham International) and exposed for 15-120 s to Kodak XS1 films (Kodak, Rochester, NY).

Construction of a Granzyme A Recombinant Vaccinia
Virus-A full-length granzyme A cDNA was cloned from activated T lymphocytes as described under "Materials and Methods." The nucleotide sequence of the clone obtained was identical to that published by Gershenfeld et al. (9). The cDNA was inserted into the vaccinia virus recombination vector p11k-ATA-18, under the control of the 11k late promotor (24) (Fig. 1). It encoded the complete protein including the signal peptide and the prodipeptide (Fig. 1). The cDNA was intergrated into the viral DNA of the wild type virus by recombination as described elsewhere (24). The promotor used enables the synthesis of a large amount of foreign polypeptides in the late phase of the viral infection, initiating translation at their authentic start codon.
Characterization of rGA Produced by Different Cell Lines-Jurkat, HepG2, and COS-1 cells were infected with rGA vaccinia virus. Twenty-four and 48 h after infection all cell lines produced granzyme A protein. Cell lysates as well as culture supernatants from Jurkat or HepG2 cells obtained 48 h after infection appeared to contain rGA only when infected with granzyme A recombinant virus (Fig. 2, lanes 3-6 and lanes  9 -12, respectively), but not after infection with a control virus (Fig. 2, lanes 7-8 and lanes 13-14, respectively). RecGran A consisted of a disulfide-linked homodimer with a relative mobility (M r ) identical to the natural protein (LAK cells; Fig. 2,  lanes 1 and 2).
Each granzyme A monomer has one glycosylation site to which an N-linked high mannose oligosaccharide is bound (9,29). No difference in Endo H susceptibility between LAK cellderived granzyme A (Fig. 3, lanes 1 and 2) and rGA dimer from Jurkat or HepG2 was found ( lanes 5 and 8, respectively). Additionally, digestion of rGA with PNGase F, which cleaves high mannose as well as complex-type sugar chains, yielded a similar shift in M r as treatment with Endo H, indicating that glycosylation of rGA was identical with that of the natural protein, i.e. of the high mannose type. Deglycosylated rGA protein from Jurkat and HepG2 cells migrated with the same M r compared with deglycosylated granzyme A from LAK cells, indicating that the molecular weight of the protein backbone of the recombinant species was similar to that of natural granzyme A.
The Amount of rGA Produced by HepG2 Is Considerably Higher than That Produced by LAK Cells-In general, infected cells produced a maximal amount of rGA antigen 48 h postinfection. During infection an increase of cell lysis and a concomitant release of rGA into the supernatant was observed (Fig. 2,  lanes 11 and 12). The extent of cell lysis and subsequent granzyme A release into the supernatant varied during different infection experiments.
Compared with the number of LAK cells (Fig. 2, lanes 1 and  2), 53 times fewer HepG2 cells (lanes 9 and 10) and 10 times fewer Jurkat cells (lanes 3 and 4) were analyzed, whereas the amount of HepG2 cell supernatant tested (lanes 11 and 12) corresponded to 10 times fewer cells. However, in spite of testing a lower number of infected cells, the intensity of protein bands rGA (compare, for example, lanes 2 and 10, Fig. 2), was equal to or higher than that observed with LAK cells, suggesting the latter produced less granzyme A than the infected cells. This was further assessed semiquantitatively using an immunoblot; per given amount of cells, infected Jurkat cells produced approximately 20 times, COS-1 40 times, and HepG2 cells even 160 times more granzyme A antigen (48 h postinfection) than did LAK cells.
Recombinant Granzyme A Shows No Proteolytic Activity-Enzymatic activity of the recombinant protein was determined in two ways; first, cell lysates were tested for hydrolysis of the chromogenic substrate BLT, and, second, the affinity of recombinant proteins for benzamidin-Sepharose was determined, as benzamidin only binds to proteolytically active granzyme A. In these experiments LAK, HepG2, Jurkat, and COS-1 cell lysates were adjusted to contain approximately equal amount of granzyme A antigen as assessed by immunoblotting. Lysates from COS-1 or HepG2 cells showed hardly any BLT hydrolysis (the activity corresponded to less than 0.5% of that of LAK cells; Table I). In agreement herewith, rGA from HepG2 cells did not bind to benzamidin-Sepharose (Fig. 4, lane 5), whereas natural granzyme A from LAK cells did (Fig. 4, lane 2). In contrast, lysates of infected Jurkat cells had significant BLT activity, corresponding to about 20% of that present in LAK cell lysates (Table I). No BLT activity was observed in lysates from Jurkat cells infected with FXII.lpc recombinant virus, consistent with the observed lack of granzyme A antigen (Fig. 2, lanes 7 and 8).
The BLT activity observed with infected Jurkat cells was due to  3-6 and lanes 9 -12) or with a control  virus (FXII, lanes 7-8 and lanes 13-14). As a positive control, natural granzyme A-containing LAK cells are analyzed in lanes 1 and 2. The even lanes contain reduced samples, and the odd lanes contain nonreduced samples. The following number of cells or cell equivalents (in case of medium) per lane were analyzed: lanes 1 and 2, 1 ϫ 10 6 ; lanes  3-8 and lanes 11 and 12, 0.16 ϫ 10 6 ; lanes 9 and 10 and lanes 13 and  14, 0.03 ϫ 10 6 . 1 and 2), infected HepG2 (lanes 3-5) and Jurkat (lanes 6 -8) cells. Control samples (Ϫ) were treated identically except that no enzyme was added. the presence of proteolytically active rGA, since part of the latter bound to benzamidin-Sepharose (Fig. 4, lanes 7 and 8). Preincubation of granzyme A from LAK cells or infected Jurkat cells with PMSF, an inhibitor of serine proteinases, prevented binding to benzamidin-Sepharose, demonstrating the specificity of this binding (Fig. 4, lanes 3 and 9, respectively).

FIG. 3. Immunoblot showing PNGase F (PF) and Endo H (EH) digestion of granzyme A in LAK cells (lanes
Recombinant Granzyme A Is Converted into a Proteolytic Active Species by Cathepsin C-Although rGA from HepG2 displayed similar biochemical features (i.e. M r , dimerization, and high-mannose type glycosylation) as natural granzyme A from cytotoxic cells, hardly any proteolytic activity toward BLT substrate was observed. Presumably, the inability of HepG2 cells to produce proteolytically active granzyme A was due to inappropriate processing of the activation dipeptide. Therefore, rGA-containing HepG2 cell lysates were incubated with the lysosomal cysteine proteinase cathepsin C and assessed for granzyme activity (Fig. 5). Upon incubation with cathepsin C, rGA in HepG2 cell lysates appeared to bind to benzamidin-Sepharose (Fig. 5A, lane 6), whereas this was not observed upon incubation of rGA with buffer alone (lane 7). No difference in M r was observed prior to and after cathepsin C treatment (lanes 4 and 6, respectively), indicating only a minor modification of rGA occurred after cathepsin C treatment. As expected, natural granzyme A from LAK cell lysates was resistant to cathepsin C treatment, as incubation with cathepsin C did not alter reactivity toward benzamidin-Sepharose (lanes 2 and 3).
Apparently, after processing of the prodipeptide, cathepsin C was not able to further modify active granzyme A.
The effect of cathepsin C on the proteolytic activity of rGA in HepG2 cell lysates was also assessed (Fig. 5B). This proteolytic activity increased during incubation with cathepsin C, whereas no increase of BLT activity by rGA was detected when cathepsin C was omitted. It is to be noted that cathepsin C itself, at the concentrations used, did not convert the BLT substrate (not shown). Thus, together these results demonstrated that cathepsin C was able to convert recombinant granzyme A in HepG2 cell lysates into a proteolytically active enzyme. DISCUSSION Here we report the expression of recombinant human granzyme A in mammalian cells by a granzyme A recombinant vaccinia virus and the ability of cathepsin C to convert rGA zymogen into an active proteinase. Using immunoblotting, we estimated that infected HepG2 cells produced about 160 times more granzyme A than LAK cells. These levels exceeded that produced by transient expression in COS cells using a mammalian expression vector (not shown). This is in concordance with previous studies comparing conventional transient expression systems with expression by vaccinia virus (30).
Different cell lines were infected by recombinant vaccinia virus harboring cDNA coding for full-length granzyme A including the signal and propeptide. Infected cell lines produced a disulfide-linked homodimer with the same molecular weight as natural granzyme A. In vitro translation of granzyme A mRNA, using rabbit reticulocytes and dog microsomes in the presence of oxidized glutathione, produced a granzyme A dimer unable to bind to benzamidin-Sepharose. 2 Apparently, dimerization can occur before formation of active granzyme A. Furthermore, these in vitro translation experiments showed that dimerization of granzyme A likely takes place in the rough endoplasmic reticulum, i.e. before the proteinase is activated. Experiments with PNGase F and Endo H showed that rGA only contained high mannose-type oligosaccharides, similar to natural granzyme A. Binding of the mannose 6-phosphorylated sugar to the mannose 6-phosphate receptor leads to selective transport of granzyme A to the cytotoxic granules (31,32). The majority of the rGA produced by HepG2 or COS cells was retained in the infected cells. Immunofluorescence showed a granular staining pattern (not shown) comparable with that in LAK cells (23), suggesting recombinant protein was targeted to lysosomes. Alternatively, overproduction of recombinant protein, combined with the virus infection, may have destroyed the cellular architecture, leading to an accumulation of the protein

Relative proteolytic activity of recGran A compared with an equal amount of natural granzyme A antigen from LAK cells
Different cell lines were infected with granzyme A or with the FXII.lpc (control) recombinant vaccinia virus. 20 ϫ 10 3 infected and noninfected cells of each cell line were tested for granzyme A proteolytic activity by determining the hydrolysis of the chromogenic substrate BLT (increase in absorption at 414 nm after 60 min). Additionally the amount of recGran A antigen produced by each cell line was compared with the granzyme A antigen content of an equal amount of LAK cells. Finally, the relative granzyme A activity was calculated by comparing the proteolytic activity and the amount of granzyme antigen to natural granzyme A from LAK cells. ND, not determined.  3) and rGA of infected HepG2 (lanes 5 and 6) and Jurkat cells (lanes 8 and 9) to benzamidin-Sepharose in the presence (؉) or absence (؊) of PMSF. 80 ϫ 10 4 LAK, 0.5 ϫ 10 4 HepG2, and 8 ϫ 10 4 Jurkat cells were incubated for 4 h with bezamidin-Sepharose, and the granzyme A bound was analyzed on immunoblot as described under "Materials and Methods." In addition, for each cell type a complete lysate (cl) of equal cell numbers was tested (lanes 1, 4, and 7). in the endoplasmic reticulum or the Golgi apparatus.
Granzymes are synthesized as inactive precursor molecules with a short prodipeptide consisting of Glu-Lys in the case of human granzyme A and of Gly-Glu in human granzymes B and H and mouse granzyme B. Mast cell and neutrophil serine proteinases contain similar, short acidic propeptides. Formation of proteolytically active human elastase and cathepsin G involves a dual proteolytic processing pathway. First, the signal peptide is cleaved off, generating an inactive zymogen; second, the amino-terminal dipeptide and a carboxyl-terminal extension are removed, thereby generating active enzyme (20). Although rGA produced by vaccinia-infected cells consisted of a homodimer with similar molecular weight as natural granzyme A, it had no proteolytic activity, except for rGA produced by Jurkat cells. Similar observations have been made for COS cells transfected with full-length cDNA coding for human leukocyte elastase or murine granzyme B cDNA (22,33), whereas COS cells transfected with a mutant granzyme B cDNA lacking the prodipeptide did produce active granzyme B (22). Together these data suggest that HepG2 and monkey COS cells, in contrast to Jurkat cells, lack the ability to process the prodipeptide of granzymes and related proteinases correctly.
It has been suggested that the lysosomal cysteine proteinase dipeptidyl peptidase I, previously termed cathepsin C, is the putative enzyme involved in the processing of the amino-terminal propeptides of granzymes and myeloid associated serine proteinases (21). In agreement with that, high levels of cathepsin C occur in the spleen and other lymphoid or myeloid cells (34,35). Thus, the inability of COS and HepG2 cells infected with recombinant vaccinia virus, coding for the full-length granzyme A cDNA, to generate active rGA may have been due to low levels of cathepsin C in these cells. In contrast, Jurkat cells, a human T-helper leukemia cell line, constitutively express granzyme A (9) and thus presumably contain cathepsin C. Part of the rGA produced by infected Jurkat cells indeed appeared to be active, although this activity was only 20% compared with natural granzyme A. Overproduction of recombinant protein, together with suppression of host protein synthesis by the viral infection, may have disturbed complete processing of recombinant progranzyme A by Jurkat cells.
Activation of serine proteases results from the ability of the ␣-amino group of the first isoleucin of the mature enzyme ( Fig.  1), generated after processing of the propeptide, to form an ion pair with the aspartic acid of the catalytic pocket. This interaction enables the formation of a functional catalytic center (20). Apparently, after removal of propeptide by cathepsin C, rGA had proteolytic activity. No decrease in esterolytic activity or loss in affinity for benzamidin-Sepharose was observed after prolonged cathepsin C treatment of natural or activated rGA. This indicates that active granzyme A is refractory to further cathepsin C treatment. If cathepsin C-mediated N-terminal dipeptide cleavage would proceed, the protease would lose its stable conformation and, therefore, its proteolytic activity.
This study is the first to demonstrate expression of an active recombinant human cytotoxic lymphocyte proteinase. By expressing rGA as an inactive zymogen, proteolytic damage to the expression system was minimized, thereby allowing high expression levels. Furthermore, the cysteine proteinase cathepsin C was shown to be able to convert rGA zymogen into an active enzyme, implying that cathepsin C may be involved in the processing of natural granzymes. A similar strategy may be feasible for the expression of other granzymes or related proteinases. FIG. 5. A, activation of rGA from infected HepG2 cells by cathepsin C. Lysates of LAK cells or HepG2 cells were incubated with cathepsin C in the presence of cysteine, after which binding of granzyme A to benzamidin-Sepharose was analyzed on immunoblot (lanes 3 and 6, respectively). HepG2 cell lysate incubated with buffer alone and absorbed to benzamidin-Sepharose is shown as control (lane 7). In addition, cell lysates of LAK cells (lane 1) and HepG2 cells (lane 4), not incubated with cathepsin C, as well as their benzamidin-bound fraction (lanes 2 and 5, respectively) were analyzed. All lanes contain lysates or absorbed fractions equivalent to 200 ϫ 10 3 LAK or 5 ϫ 10 3 HepG2 cells. B, time course of activation of rGA from infected HepG2 cells by cathepsin C. The amount of BLT hydrolysis is expressed as LAK cell equivalents. HepG2 cell lysate is incubated with cathepsin C and cysteine (ࡗ), with cysteine alone (F), or without cathepsin C and cysteine (Ç), as described under "Materials and Methods."