Generation of Catalytically Active Granzyme K fromEscherichia coli Inclusion Bodies and Identification of Efficient Granzyme K Inhibitors in Human Plasma*

Granzymes are granule-stored lymphocyte serine proteases that are implicated in T- and natural killer cell-mediated cytotoxic defense reactions after target cell recognition. A fifth human granzyme (granzyme 3, lymphocyte tryptase-2), renamed as granzyme K (gene name GZMK), has recently been cloned from lymphocyte tissue. For its further characterization we successfully generated catalytically active enzyme in milligram quantities per liter of Escherichia coli culture. The natural proform of granzyme K with the amino-terminal propeptide Met-Glu was expressed as inclusion bodies and converted to its active enzyme by cathepsin C after refolding of precursor molecules. Recombinant granzyme K cleaves synthetic thiobenzyl ester substrates after Lys and Arg withk cat/K m values of 3.7 × 104 and 4.4 × 104 m −1 s−1, respectively. Granzyme K activity was shown to be inhibited by the synthetic compounds Phe-Pro-Arg-chloromethyl ketone, phenylmethylsulfonyl fluoride, PefablocSC, and benzamidine, by the Kunitz-type inhibitor aprotinin and by human blood plasma. The plasma-derived inter-α-trypsin inhibitor complex, its bikunin subunit, and the second carboxyl-terminal Kunitz-type domain of bikunin were identified as genuine physiologic inhibitors with K i values of 64, 50, and 22 nm, respectively. Inter-α-trypsin inhibitor and free bikunin have the potential to neutralize extracellular granzyme K activity after T cell degranulation and may thus control unspecific damage of bystander cells at sites of inflammatory reactions.

Perforin and granzymes are major components of cytosolic granules that play important roles in the secretory pathway of T and natural killer (NK) 1 cell-mediated cytotoxicity against virally infected host cells, tumor cells, and antigenetically al-tered untransformed host cells. Upon specific target cell recognition granules are exocytosed toward the target cell membrane, and perforin and granzymes are released in concert (1)(2)(3)(4). The two most abundant granzymes of cytolytic T cells are the granzymes A and B that were shown to contribute to the induction of apoptosis and DNA fragmentation in target cells (2, 4 -6). So far, five distinct human granzymes A, B, H, M, and K have been described at the molecular genetic level (7), but very little is known about the enzymatic specificities, biological functions, and regulation of activity of the human granzyme H (8,9), granzyme M (10 -12), and granzyme K (GzmK) (13,14). All granzymes are synthesized as pre-pro-enzymes in the rough endoplasmic reticulum and converted into active enzymes in a two-step process by cleavage of the signal peptide and subsequent removal of the propeptide by a similar, presumably identical dipeptidyl aminopeptidase of cytosolic granules, called cathepsin C (15,16).
Although granzymes are released from storage granules in a tightly controlled manner after recognition of target cells, low constitutive levels of granzyme A and B can be detected in human blood plasma (17) suggesting a continuous secretion by these cells in healthy individuals. At sites of active inflammation and during chronic inflammatory diseases, human blood plasma levels of granzymes A and B are elevated, and additional extracellular functions for granzymes have been inferred recently (17)(18)(19). Soluble extracellular granzyme A has been shown to stimulate interleukin-6, interleukin-8, and tumor necrosis factor-␣ production of human peripheral blood mononuclear cells and purified monocytes, presumably via a membrane receptor that is different from the thrombin receptor (18,19). These extracellular actions of granzyme A appear to be controlled by a highly abundant plasma inhibitor, ATIII, which can inactivate circulating granzyme A levels rapidly (20). GzmK (granzyme 3, lymphocyte tryptase-2) was first discovered and purified as a second Z-Lys-SBzl-cleaving esterase in lymphokine-activated killer cells (21). It differed from human granzyme A with respect to its amino acid sequence and monomeric structure under nonreducing conditions (21). Later, a similar granzyme was isolated and cloned from the rat NK tumor cell line RNK-16 (22,23) which was capable of inducing apoptosis in vitro (22). The human GzmK has recently been cloned (13,14) and shown to possess hydrolytic activities for the thioester substrates Z-Lys-SBzl and Z-Arg-SBzl (21) and for a highly basic peptide representing the SV40 nuclear localization signal (24). High mRNA levels of GzmK are detected in NK cells and activated T cells but are absent in normal tissues that do not contain high numbers of these cells (13).
Recombinant granzymes have been produced in various eu-karyotic and prokaryotic host systems including yeast (25), baculovirus-infected insect cell lines (26), mammalian cell lines (16,27), Escherichia coli (28), and Bacillus subtilis (24) by secretion and transport of precursor forms carrying signal sequences. To overcome various problems with these systems, we have developed a procedure that utilizes inclusion bodies (IB) from E. coli as the starting material for in vitro refolding. Amino-terminally extended and catalytically inactive precursors of GzmK were refolded into soluble proteins and then subjected to limited proteolysis by reaction with a dipeptidylaminopeptidase, cathepsin C, which appeared to be suitable to remove selectively extra amino acid residues at the amino terminus that are not present in catalytically active GzmK. We hypothesized that human GzmK, as the closest structural and functional relative of human granzyme A, could also be controlled by extracellular fluid phase inhibitors, and thus we investigated the inhibitory capacity of candidate inhibitors from human plasma. Our search for efficient inhibitors was guided by previous observations that mouse granzyme A, a major Z-Lys-SBzl esterase of lymphocytes, is inhibited by ATIII (20) and that aprotinin, a Kunitz-type inhibitor of bovine pancreas, is inhibitory toward human GzmK (24). The major Kunitz-type inhibitor of human plasma, known as inter-␣-trypsin inhibitor (I␣I) according to its electrophoretic mobility between ␣ 1 and ␣ 2 plasma proteins, however, is a mixture of various complexes between bikunin (which consists of two Kunitz-type inhibitor domains) and H2 or H3 or H1 and H2 of the three structurally related polypeptide chains H1 to H3 (29). I␣I occurs in concentrations between 0.4 and 0.7 mg/ml plasma, whereas free bikunin represents only about 2% of total plasma bikunin (30,31). Here, we report that I␣I is the most important physiological GzmK inhibitor of human plasma and that its inhibitory activity is mediated by the second carboxyl-terminal domain of bikunin.
Expression-Overnight cultures of E. coli B834 (DE3) harboring expression plasmids were grown in Luria-Bertani broth (LB) containing 50 g/ml kanamycin, 2% glucose and inoculated in LB/kanamycin without glucose. Expression was induced by addition of isopropyl-1-thio-␤-D-galactopyranoside to a final concentration of 1 mM at an A 600 nm of 0.5 to 0.8. Cultures were grown for 3-4 h after induction at 37°C with moderate shaking at 150 rotations/min and harvested by centrifugation.
Refolding and Purification-Bacteria were lysed in 50 mM Tris-HCl, 2 mM MgCl 2 containing 10 g/ml DNase I, and 0.25 mg/ml lysozyme, pH 7.2, by French press or sonification. IB were harvested by centrifugation and washed twice in 50 mM Tris-HCl, 60 mM EDTA, 1.5 M NaCl, 6% Triton X-100, pH 7.2, followed by two washing steps in 50 mM Tris-HCl, 60 mM EDTA, pH 7.2. Purified IB were solubilized in 6 M guanidinium chloride, 100 mM Tris-HCl, 20 mM EDTA, 15 mM GSH, 150 mM GSSG, pH 8.0, overnight at room temperature (RT) in an end-over-end rotator followed by dialysis against 6 M guanidinium chloride, pH 5.0, at 4°C. Refolding was initiated by diluting solubilized proteins (usually 10 -15 mg/ml) in 100 volumes of 50 mM Tris-HCl, 0.5 M L-arginine, 20 mM CaCl 2 , 1 mM EDTA, 0.1 M NaCl, 0.5 mM L-cysteine, pH 8.5, at RT, followed by an additional incubation period of 2 days at RT. The refolding solution was dialyzed against 100 volumes of PBS, pH 7.0, at 4°C until conductivity of the dialyzed solution was equal to the dilution buffer. After centrifugation at 30,000 ϫ g refolded proteins were filtrated and loaded onto a Mono S-Sepharose column (Amersham Pharmacia Biotech) in PBS and further purified by fast protein liquid chromatography using 20 column volumes of a linear salt gradient from 0.15 to 2 M NaCl in the same buffer at RT. GzmK concentrations were determined with the Bradford (Coomassie dye binding) assay.
Processing by Cathepsin C and Active Site Titration-Fast protein liquid chromatography peak fractions of various refolded precursor forms were desalted by dialysis against PBS, pH 6.5, and concentrated to approximately 10 mg/ml. Bovine cathepsin C (Sigma) was first activated at 37°C in PBS, 10 mM 2-mercaptoethanolamine, 75 mM sodium acetate, pH 5.0, for 30 min and then dialyzed against PBS, 75 mM sodium acetate, pH 5.5, before incubation with GzmK precursors at RT. Five units of cathepsin C and 10 mg of GzmK precursor in equal volumes of activation buffer were mixed and incubated at RT. The increase of enzymatic activity was monitored using N ␣ -benzyloxycarbonyl-L-lysine-thiobenzyl ester (Z-Lys-SBzl) as described below, and the incubation was stopped after activity reached a plateau. The activated granzyme was separated from cathepsin C by cation exchange chromatography as described above. The concentration of active GzmK was determined by burst titration using 4-nitrophenyl-4Ј-guanidinobenzoate (Sigma). The amount of bikunin, Kunitz domain 2 of bikunin, and I␣I were determined by titration with active site titrated trypsin.
Activity Assays and Kinetic Studies-Activity assays were performed at RT in 50 mM Tris-HCl, 0.15 M NaCl, 0.01% Triton X-100, pH 7.6, containing 0.3 mM 5,5Ј-dithiobis(2-nitrobenzoic acid) (Ellman's reagent) (Sigma) in 96-well microtiter plates with a reaction volume of 150 l. Substrate concentrations were 0.3 mM for activity assays and between 0.05 and 1.5 mM for kinetic studies. Z-Lys-SBzl (Sigma) was dissolved in water, Z-Arg-SBzl, Boc-Ala-Ala-Met-SBzl, and Boc-Ala-Ala-Asp-SBzl (Enzyme Systems Products, Livermore, CA) in Me 2 SO. Concentrations of active site titrated GzmK were 3 nM for activity assays and 0.5-7.5 nM for kinetic studies. Activity was determined as the rate of hydrolysis by measuring increase of absorbance at 405 nm wavelength over time with a Dynatech MR4000 microplate reader. Equilibrium dissociation constants K i for reversible GzmK-inhibitor interactions and association rate constants (k a ) for GzmK-antithrombin III (ATIII) interactions were determined as described (32).
Inhibition and Plasma Assays-Various inhibitors were tested using 0.3 mM Z-Lys-SBzl in the presence of 3 nM active site titrated GzmK. I␣I was a generous gift of Dade Behring (Marburg, Germany); bikunin and bikunin D2 were produced in recombinant form in Pichia pastoris. 2 ␣ 2 -Macroglobulin was purchased from Sigma; ATIII and heparin cofactor II were from Hematologic Technologies Inc.; p-aminobenzamidine was from Fluka; aprotinin (bovine pancreas trypsin inhibitor) was kindly provided by Bayer, Leverkusen, Germany. GzmK was preincubated with various concentrations of each inhibitor for 15 min at RT before adding Z-Lys-SBzl substrate. The decrease in the rate of hydrolysis was monitored, and kinetic values were calculated by nonlinear regression. For plasma assays, human EDTA-plasma of a young and healthy donor was filtered (0.22 m) and preincubated with 3 nM GzmK for 15 min at RT before adding substrate. ␣ 2 -Macroglobulin was inactivated for 4 h at 23°C at pH 8.3 by addition of methylamine (Sigma) to freshly isolated plasma at a final concentration of 0.2 M. Anti-human ATIII antibodies and antisera against human I␣I, bikunin, and Hchains of I␣I were kindly provided by Dade Behring (Marburg, Germany). Whole IgG fractions were purified over a ProsepG column according to the manufacturers' instructions (Bioprocessing, Princeton, NJ).
Antibodies were preincubated with plasma or purified inhibitors for 15 min at RT before addition of GzmK, followed by a second incubation for 15 min before addition of the substrate. Controls were run with the respective component (plasma and inhibitor) in the absence of GzmK to measure background Z-Lys-SBzl activities. These values were subtracted from test wells.
Measurement of I␣I in Human Plasma-The concentration of I␣I in the applied plasma probes was determined by radial immunodiffusion. Holes of 2 mm in diameter were punched in horizontal agarose gels containing 3% PEG 4000, 1% SeaKem agarose (FMC Bioproducts, Rockland, ME), 0.02% sodium azide, 16.8 mg/liter rabbit anti-human I␣I (Dako, Glostrup, Denmark) in 25 mM barbital buffer (3.7 mM 5,5-2 R. Friebel and H. Tschesche, manuscript in preparation. diethylbarbituric acid, 21.3 mM 5,5-diethylbarbituric acid sodium salt, 1.3 mM calcium lactate, 0.7% sodium azide, pH 8.6) and were filled with 5 l of plasma probes or various dilutions of titrated I␣I as a standard. The gels were incubated at 4°C until ring formation was complete, dried, and stained with 0.1% Amido Black in 45% methanol, 2% acetic acid for 3-5 min, and destained in 90% methanol, 2% acetic acid. I␣I concentrations of plasma probes were determined from calibration curves for the areas of circles obtained by immunodiffusion of different I␣I standard concentrations.

Recombinant Expression and
Renaturation-Cathepsin C only removes an even number of amino acid residues (dipeptides) from the amino terminus of proteins until it is blocked by certain residues such as lysine and arginine in the first position or proline in the first or second position (33). We therefore studied the processing of different precursor forms of human GzmK (Fig. 1) by the endogenous E. coli methionine aminopeptidase (MAP) and the efficiency of in vitro processing of renatured GzmK precursors without specific stop sequences by bovine cathepsin C.
The presequences of GzmK precursors that were expressed in E. coli consisted of either Met followed by a modified amino terminus Val-Ile-Gly-Gly or started with Met-Gly-Glu or Met-Glu ( Fig. 1) followed by the natural amino acid sequence of mature human GzmK, Ile-Ile-Gly-Gly. High levels of expression were achieved for all constructs in the E. coli host B834 (DE3) with yields of 50 -75 mg of IB protein per liter of culture comprising up to 50% total bacterial protein. Due to the high yield and insolubility of the recombinant proteins, IBs could be efficiently separated from total bacterial proteins and subjected to the protein refolding procedure without further purification steps. Fig. 2 shows total bacterial extracts before and after induction of protein expression and a preparation of inclusion bodies. After replacement of the refolding buffer by PBS, soluble material was analyzed by SDS-PAGE and shown to have the expected size of human GzmK.
The degree of methionine removal in E. coli by the endogenous MAP was examined by amino acid sequencing of renatured protein precursors after purification by cation exchange chromatography and estimated from the ratio of the two phenylthiohydantoin-derivatives recovered at each successive Edman degradation cycle. Methionine residues were poorly cleaved from the precursor Met-Gly-Glu with two-thirds of these proforms starting with the desired dipeptide Gly-Glu. Moreover, the methionine residue before Glu in the precursor Met-Glu-GzmK remained quantitatively attached in renaturated precursor molecules. Since amino-terminal methionine residues followed by Val are often quantitatively removed from endogenous bacterial proteins (34), we constructed a maturelike amino terminus by replacing the first Ile residue by the smaller Val residue. Amino-terminal sequencing of the refolded protein, however, revealed that the methionine residue was still present in the majority of molecules (more than 90% of the refolded material). Furthermore, very little Z-Lys-SBzl activity was observed in such protein preparations indicating that the majority of this protein is catalytically inactive and that a free amino-terminal Val or Ile residue is required to adopt the catalytically active conformation.
The recovered GzmK precursors were then subjected to the exopeptidolytic activity of bovine cathepsin C. Amino-terminal amino acid sequencing confirmed successful conversion of both proforms, Gly-Glu-GzmK and Met-Glu-GzmK. Amino acid residues from shorter products were not detected by Edman degradation despite the fact that the first linear cathepsin C stop sequences occur at residue positions 5 (Lys) and 9 (Pro) in GzmK. As expected, conversion of precursor forms from the Met-Gly-Glu-construct was incomplete and heterogeneous, and the yield of the desired mature GzmK equaled the fraction that had lost the amino-terminal methionine already during biosynthesis in E. coli. The precursor molecules from the Met-Glu-GzmK construct, however, were trimmed quantitatively to the length of the mature protein as documented by amino-terminal sequencing. In this way the precursor construct with the dipeptide Met-Glu sequence at the amino terminus of mature GzmK was shown to be the most favorable design of a precursor molecule that can be processed in vitro to the authentic enzyme by means of the dipeptidyl-aminopeptidase cathepsin C. Glutamic acid at the second position prevents any removal of methionine residues in E. coli and, together with methionine, is a well suited substrate residue at P1 position for bovine cathepsin C in vitro. By using Met-Glu-precursor molecules for GzmK, 20% of the IB material was refolded and subjected to conversion. Half of it was finally recovered after ion exchange chromatography in highly concentrated and homogenous form.
Functional Characterization of Recombinant GzmK-Enzymatic activity of GzmK was assayed using thiobenzyl esters and Ellman's reagent. None of the precursor forms exhibited enzymatic activity using Z-Lys-SBzl as a chromogenic substrate (see Fig. 3, black bars, M-E precursor). When compared with total protein concentrations, little Z-Lys-SBzl activity was recovered from the Met-Val-Ile-Gly-Gly precursor variant after chromatographic purification. This finding is fully consistent with our observation that the majority of molecules still carried the methionine residue at the amino terminus. The highest yield of active GzmK was obtained with the precursor construct Met-Glu-GzmK because the proform could be quantitatively converted into the active mature enzyme. In contrast, the Met-Gly-Glu-precursor construct resulted only in the recovery of 50 -65% catalytic activity because of the large fraction of methionine-carrying precursor molecules that were not converted to mature GzmK.
Mature GzmK cleaved the synthetic substrates Z-Lys-SBzl and Z-Arg-SBzl containing Lys and Arg at position P1, respectively, but no other thiobenzyl ester substrates with Met or Asp at P1 (Fig. 3). Kinetic (Michaelis-Menten) parameters were determined for Z-Lys-SBzl and Z-Arg-SBzl using active site titrated GzmK. K m values of 1.7 and 0.14 mM and k cat values of 63.1 and 6.2 s Ϫ1 were found for Z-Lys-SBzl and Z-Arg-SBzl, respectively. k cat /K m values amounted to 3.7 ϫ 10 4 M Ϫ1 s Ϫ1 for Z-Lys-SBzl and 4.4 ϫ 10 4 M Ϫ1 s Ϫ1 for Z-Arg-SBzl. These kinetic parameters are very similar to those obtained recently for biosynthetically folded GzmK that was produced by extracellular secretion from B. subtilis (24).
To characterize further the functional and biochemical properties of refolded GzmK, we tested its sensitivity toward various synthetic and peptidic protease inhibitors. Inhibition assays were carried out as previously reported (24) with an excess of inhibitors and active site titrated GzmK. Compared with published results, no differences in enzymatic inhibition were found with 10 synthetic inhibitors tested (Table I), except for N-tosyl lysine chloromethyl ketone which inhibited Bacillusderived GzmK completely (24). Phe-Pro-Arg-chloromethyl ketone was identified as a new synthetic peptide inhibitor for GzmK showing complete inhibition at an inhibitor to enzyme ratio of 3 ϫ 10 4 . Aprotinin was inhibitory to the same extent as in a previous study (24). With the use of Z-Lys-SBzl, a pH optimum of 8.2 to 8.6 was determined for GzmK which agrees with that of natural GzmK purified from lymphokine-activated killer cells (21). Our enzyme preparations kept catalytic activity over weeks at RT and 4°C, even in diluted solutions and could be stored over long periods at Ϫ20°C. Active site titration of our several independent GzmK preparations confirmed that 80 -90% of all molecules were enzymatically functional after conversion. Whereas 100 nM solutions of Bacillus-derived GzmK were previously used in Z-Lys-SBzl assays (24), we found that concentrations of less than 3 nM recombinant refolded GzmK are sufficient to analyze Z-Lys-SBzl esterase activity under comparable assay conditions. In summary, the described procedure yields high amounts of recombinant GzmK with high specific activity, good stability, and enzymatic properties that are characteristic for naturally occurring GzmK.
I␣I Is a Natural Plasma Inhibitor of GzmK-Potent plasma inhibitors exist for most granule-associated serine proteases like leukocyte elastase, cathepsin G, and proteinase 3 (35). To search for GzmK-specific inhibitors, we incubated a constant amount of recombinant GzmK with increasing amounts of total plasma and measured Z-Lys-SBzl esterase activity. GzmK was already inhibited by plasma proteins at low concentrations in a dose-dependent manner (Fig. 4, upper panel). Higher amounts of EDTA plasma led to a further decrease of GzmK activity (data not shown). Since background levels of Z-Lys-SBzl esterase activity increased with increasing volumes of EDTA plasma (hatched bars in Fig. 4), subsequent experiments were done in buffers containing only 2.5% EDTA plasma. Addition of heparin to 2.5% EDTA plasma at a final concentration of 0.5 units/ml led to a further decrease of GzmK activity after 15 min of incubation (Fig. 4, lower panel), indicating that the major heparin-accelerated serine protease inhibitors of human plasma, ATIII or heparin cofactor II, could mediate inhibition. To identify the most relevant plasma inhibitor(s), several known inhibitors were tested individually. The most prominent plasma inhibitor for trypsin-like enzymes, ␣ 1 -proteinase inhibitor, and heparin cofactor II had no effect on GzmK (data not shown). In contrast, ATIII inhibited GzmK with an apparent association rate constant of 1.7 ϫ 10 3 M Ϫ1 s Ϫ1 . In the presence of 0.5 units/ml heparin the association rate constant was about 5-fold higher (8.2 ϫ 10 3 M Ϫ1 s -1 ). ATIII concentrations equivalent to those in the plasma assays, however, showed no inhibitory potential (Fig. 5, 3rd column).
Since inhibition of GzmK by ␣ 2 -macroglobulin (data not shown) and ATIII in the absence of heparin did not appear to

TABLE I Inhibition of human GzmK by various inhibitors
Remaining activities of 3 nM human GzmK after incubation with various inhibitors in percent of initial activity at the indicated concentrations. After 60 min of preincubation at RT, residual activity was determined in triplicate. Initial activities were determined in buffers containing the same proportion of organic solvents. Inhibitor to enzyme ratios (I:E) were calculated on the basis of active site titrated GzmK. Molar concentrations of active aprotinin were determined using active site titrated bovine trypsin. CMK, chloromethyl ketone; TPCK, N-tosyl phenylalanine chloromethyl ketone; TLCK, N-tosyl lysine chloromethyl ketone; PMSF, phenylmethylsulfonyl fluoride. account for the major inhibitory capacity of human plasma, we considered I␣I, the predominant form of bikunin in human plasma, as another possible inhibitor for human GzmK. The carboxyl-terminal Kunitz-type domain of bikunin (bikunin D2) is known to inhibit trypsin-like enzymes with Lys and Arg specificity at the S1 subsite, but an endogenously produced cognate enzyme with high affinity for this domain has not yet been identified. First, we tested bikunin D2 which had been produced in P. pastoris in recombinant form and determined its equilibrium dissociation constant (K i ). Domain 2 efficiently blocked GzmK activity with a K i of 22 nM. Since protease binding may be affected by the presence of the first domain of bikunin or the presence of heavy chains in I␣I, we also determined the K i values for both the complete recombinant bikunin inhibitor and the purified plasma-derived I␣I. The K i value for bikunin (domains 1 ϩ 2) was 50 nM, whereas the natural plasma inhibitor had a K i value of 64 nM for GzmK. These values are 2-and 3-fold higher, respectively, than that (22 nM) observed with the individual bikunin domain 2. Nevertheless the K i value of I␣I for human GzmK is lower than that determined previously for human plasmin (130 nM) and neutrophil elastase (150 nM) (36) and lower than that determined for the inhibition of factor Xa (450 nM) and plasma kallikrein (410 nM) by bikunin D2 (37). When we mixed purified I␣I with bikunin D2 at the same molar ratio as present in 2.5% normal plasma (30,38) and at a total concentration of 26 nM, GzmK activity was inhibited to the same degree (Fig. 5, 5th column). The addition of ATIII to bikunin D2 or to I␣I plus bikunin D2 showed no additional effect (Fig. 5, 6th and 7th columns). By using a polyclonal rabbit serum against I␣I and the technique of radial immunodiffusion, we found 2 M I␣I in the undiluted plasma sample (50 nM in 40-fold diluted plasma). This value is twice as much as required to achieve the same degree of GzmK inhibition with purified I␣I (see Fig. 5, 4th and 5th columns) indicating that the total inhibitory capacity of human plasma does not exceed that of the isolated bikunin-containing inhibitors. These findings suggested that I␣I and the small amount of free bikunin of human plasma accounted for most of the GzmK inhibitory activity (Fig. 5).
To prove this observation, we determined IgG antibody concentrations for bikunin, I␣I, and heparin-activated antithrombin III that were sufficient to neutralize the inhibitory capacity of these inhibitors toward GzmK in assays with purified components (Fig. 6, upper panel). The same amount of IgG antibodies was then added to 2.5% human plasma, and GzmK activity was measured after 15 min incubation. The 2nd and 3rd columns in Fig. 6, lower panel, show that anti-bikunin-and anti-I␣I-specific IgGs antagonize the inhibitory properties of these inhibitors and almost completely prevent the inactivation of GzmK in plasma dilutions. In control experiments we added antibodies specific for the H-chains of I␣I and for ATIII to human plasma and observed no antagonizing effect on GzmK inhibition (Fig. 6, lower panel, 4th and 5th columns). DISCUSSION Purification of catalytically active granzyme A was previously reported following periplasmic expression in E. coli (28). This approach has certain limitations and drawbacks. Not only are some granzymes bactericidal but translocation of the serine protease to the periplasmic space results in exposure to periplasmic bacterial proteases. Moreover, the export and signal peptide cleavage capacity is often exhausted in periplasmic expression systems yielding a mixture of folded and unfolded recombinant protein in the bacterial lysate. In contrast, expression yields are often very high with cytosolic expression vec-tors. The formation of inclusion bodies not only protects the bacteria from the toxic effects of the biologically active enzyme but also protects the recombinant protein from degradation by cytosolic bacterial proteases.
A correctly engineered amino terminus starting with I(I/ V)GG is essential for a granzyme to adopt its active conformation. The amino-terminal methionine, however, is only removed by the endogenous MAP when small-sized amino acid residues like glycine, alanine, serine, threonine, valine, or proline follow at the second position (34). Our initial attempts to overcome this problem for GzmK were unsuccessful. Valine naturally occurs at the amino terminus of some serine proteases and was thus considered as a suitable replacement of the first Ile residue. However, when we replaced Ile to facilitate removal of amino-terminal Met residues (34), over 90% of recombinant molecules still carried the Met residue as shown by amino-terminal sequencing. After refolding of this protein preparation, we observed little enzymatic activity suggesting that the vast majority of the recombinant enzyme was catalytically inactive. Recovery of some enzymatic activity, corresponding with the minor fraction of biosynthetically converted GzmK, however, indicates that a Val at the amino terminus could functionally replace the natural Ile in GzmK.
To generate highly homogeneous starting material for GzmK, we attached the prosequences Met-Gly-Glu and Met-Glu to its amino terminus. When the methionine was followed by a Gly, the amino-terminal methionine was removed in about 70% of refolded protein by MAP during biosynthesis of inclusion bodies. The refolded precursors for GzmK thus started with the propeptide sequence Gly-Glu as present in granzyme B or to a lesser extent with Met-Gly-Glu. In contrast, when the methionine was followed by Glu, the starting methionine remained associated with the polypeptide chain. Whereas the dipeptide Gly-Glu was shown to be efficiently removed from the amino terminus of granzyme B by cathepsin C in vitro (27), removal of the natural propeptide of GzmK, Met-Glu, by cathepsin C had not yet been investigated. Our observations now indicate that both dipeptides attached to GzmK are efficiently cleaved off in vitro by cathepsin C. Thus we conclude that GzmK and granzyme B are most likely activated by the same processing enzyme in vivo. Since about 30% of all precursor molecules of the Met-Gly-Glu-construct still carried Met at the first position, the extension sequence Met-Glu was the most appropriate and the perfect solution to generate structurally homogeneous starting material.
In this way, folding and conversion of the Met-Glu-precursor in vitro closely mimic the natural generation of granzymes in the endoplasmic reticulum after cleavage of the signal peptide. This particular feature of our protocol may explain the high folding yields and the efficient processing of the recombinant protein in vitro. The procedure described here is the first that enables the preparation of active human GzmK and its natural precursor from aggregated E. coli inclusion bodies by refolding and exopeptidolytic processing. The enzyme exhibits the same biochemical properties and specificities as biosynthetically folded GzmK purified from lymphokine-activated killer cells or supernatants of GzmK secreting B. subtilis (21,24). Our procedure yields pure and active enzyme in the milligram range and, thus, overcomes problems of cost-intensive low level expression in mammalian cell lines.
In view of the close structural and functional similarities between granzymes A and K and the previous observation that granzyme A is inhibited by ATIII (20), we first considered ATIII as the major inhibitor of GzmK in human plasma. In the absence of heparin, inhibition of GzmK by purified ATIII was slow, and the apparent association rate constant of 1.7 ϫ 10 3 FIG. 6. Effect of polyclonal IgG antibodies to I␣I, bikunin, and ATIII on the inhibition of GzmK. Upper panel, inhibition of GzmK in the presence of monospecific antibodies to purified inhibitors (black bars) was compared with that without neutralizing antibody (hatched bars). Bikunin D2 (30 nM), I␣I (76 nM), and ATIII (125 nM) were preincubated with monospecific antibodies for 15 min before the addition of 3 nM GzmK. Inhibitor to enzyme ratios (I:E) are shown below the upper panel. Lower panel, the same amount of polyclonal antibodies specific for bikunin (2nd column), I␣I (3rd column), H-chains of I␣I (4th column), or ATIII (5th column) was mixed with a 2.5% plasma dilution and preincubated for 15 min before the addition of 3 nM GzmK. The remaining GzmK activity was determined 15 min later. Percentages of Z-Lys-SBzl activity with standard errors are given. Each antibody was affinity purified on protein G-Sepharose and used at final concentrations of 120 g/ml. Only bikunin and I␣I antibodies neutralize the inhibitory effect of total human plasma toward GzmK. M Ϫ1 s Ϫ1 was in the same order of magnitude as for granzyme A. However, inhibition of GzmK occurred with diluted plasma samples at ineffective ATIII concentrations and was independent of ATIII. Moreover, antibodies to ATIII did not change the kinetics of inhibition in human plasma indicating that ATIII is not the major inhibitor of GzmK in plasma.
The results obtained in this study clearly demonstrated that I␣I is the most important inhibitor for GzmK in human plasma. By using recombinant bikunin and Kunitz domain 2 of bikunin, we showed that the second Kunitz domain accounts for the inhibitory effect of I␣I complexes. The equilibrium dissociation constant for GzmK and bikunin is in the same order of magnitude as has been found for other endogenous protease/inhibitor pairs including C1s/C1r and C1 inhibitor (35). Thus inhibition by I␣I is expected to be relevant at physiological concentrations.
So far, endogenous physiological targets for human I␣I have not been identified. Bikunin in I␣I and free bikunin inhibits trypsin of the exocrine pancreas (36), acrosin from human sperm cells (39) and plasmin (36), but not factor Xa and plasma kallikrein. Blood coagulation factor Xa and plasma kallikrein are only recognized by the second Kunitz domain of bikunin but not by intact bikunin and I␣I for steric reasons. Protease binding to the recognition loop of D2 is restrained by amino-terminal Kunitz-type domain D1 which most likely interacts with D2-bound small trypsin-like proteases and blocks binding of larger proteases (40). Equilibrium dissociation constants for the interaction of the latter two enzymes with the isolated Kunitz domain 2 are nevertheless 20-fold higher (37) than that for GzmK. So far, plasmin was the only enzyme of the interstitial and intravascular space that binds reasonably well to intact I␣I complexes with a K i value of 130 nM (36). Plasmin, however, is much faster and better controlled by its cognate inhibitor ␣ 2 -antiplasmin than by I␣I (35), indicating that I␣I is not a natural regulator of plasmin activity. As a result of these considerations and our new findings, we strongly suggest that lymphocyte-derived GzmK is a physiological target for I␣I and free bikunin and is controlled by this plasma protein family.
Inhibitors of human plasma and interstitial fluids restrict the extracellular actions of enzymes after their release or activation and thus prevent inappropriate effects on bystander cells and distant tissues. Granzymes that are controlled by fluid phase inhibitors are expected to act on substrates that are present on the surface of host cells or in the extracellular compartment. The existence of a potent GzmK-specific Kunitztype inhibitor in human plasma supports the idea that GzmK fulfills additional extracellular functions besides its intracellular role in DNA fragmentation of target cells (22). In analogy to human granzyme A (41,42) it may activate a G-protein-coupled thrombin-like receptor on the membrane of target cells or may process as yet unidentified extracellular mediators of inflammation in the local microenvironment of activated T cells.