Residual Cytotoxicity and Granzyme K Expression in Granzyme A-deficient Cytotoxic Lymphocytes*

Cytotoxic lymphocytes contain granules that have the ability to induce apoptosis in susceptible target cells. The granule contents include perforin, a pore-forming molecule, and several granzymes, including A and B, which are the most abundant serine proteases in these granules. Granzyme B-deficient cytotoxic T lymphocytes (CTL) have a severe defect in their ability to rapidly induce apoptosis in their targets, but have an intact late cytotoxicity pathway that is in part perforin-dependent. In this report, we have created mice that are deficient for granzyme A and characterized their phenotype. These mice have normal growth and development and normal lymphocyte development, activation, and proliferation. Granzyme A-deficient CTL have a small but reproducible defect in their ability to induce 51Cr and125I-UdR release from susceptible allogeneic target cells. Since other granzyme A-like tryptases could potentially account for the residual cytotoxicity in granzyme A-deficient CTL, we cloned the murine granzyme K gene, which is linked to granzyme A in humans, and proved that it is also tightly linked with murine granzyme A. The murine granzyme K gene (which encodes a tryptase similar to granzyme A) is expressed at much lower levels than granzyme A in CTL and LAK cells, but its expression is unaltered in granzyme A−/− mice. The minimal cytotoxic defect in granzyme A−/− CTL could be due to the existence of an intact, functional early killing pathway (granzyme B dependent), or to the persistent expression of additional granzyme tryptases like granzyme K.

The cytolytic granules of cytotoxic T lymphocytes (CTL) 1 and natural killer (NK) cells contain several different neutral serine proteases called granzymes, the pore-forming protein perforin, and other proteins of unknown function. The granule exocytosis model of cellular cytotoxicity postulates that granzymes, in concert with perforin, induce target cell apoptosis after cytotoxic lymphocytes specifically recognize the target cell, and release their granule contents into the intercellular space between the effector and target cells (1). Perforin is thought to produce channels in the target cell membrane through which the granzymes pass; in the target cell, granzymes are thought to cleave and activate critical substrates, thereby initiating the apoptotic program of the cell. Several distinct granzymes from mice, rats, and humans have been cloned and characterized. To date, granzymes A through G and metase-1 have been identified in mice (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13), granzymes A, B, K, and metase-1 in rats (14 -19), and granzymes A, B, H, K, and metase-1 in humans (20 -29). Except for granzyme B, which is required for the rapid induction of target cell apoptosis during CTL attack (30 -32), the precise functions of the other granzymes are not yet known.
Granzymes A and B are the most abundantly expressed granzymes found in CTL stimulated with a wide variety of activation protocols. Granzyme A is a tryptase (which cleaves after lysine or arginine at the P1 position) and granzyme B is an Asp'ase (which cleaves after aspartic acid at the P1 position). The genes encoding these proteases have been mapped to two different chromosomal loci: 1) murine and human granzyme A are found on chromosome 13 and a syntenic region of human chromosome 5, respectively (23,33,34), and 2) both murine and human granzyme B are found on syntenic regions of chromosome 14 (35)(36)(37). The granzyme B gene is part of a cluster of tightly linked granzymes and other serine protease genes that are specifically regulated and expressed in distinct hematopoietic cell types. The murine granzyme B locus contains granzymes B-G, cathepsin G, and MMCP-2 (and probably other mast cell chymases); the human granzyme B locus is known to contain granzymes B and H, cathepsin G, and mast cell chymase. Our laboratory has demonstrated that the insertion of PGK-Neo cassette in the murine granzyme B gene disrupts the expression of granzymes C, D, F, and G in LAK/NK cells, potentially confounding interpretations of some of the results obtained from these mice (38). Specifically, the defect observed in the ability of granzyme BϪ/Ϫ LAK cells to induce apoptosis of susceptible targets could be due in part to the loss of granzyme C, D, F, or G. However, CTL derived from MLR predominantly express granzymes A and B. Thus, in our studies of granzyme BϪ/Ϫ CTL, the defect in the rapid induction of apoptosis is most likely due to the loss of granzyme B itself.
Studies of granzyme BϪ/Ϫ mice have revealed that CTL are equipped with a late, granzyme B-independent mechanism(s) of cytotoxicity (30,31). Several lines of evidence implicate granzyme A as a potential candidate for this pathway. Specifically, 1) purified granzyme A from mouse CTL or rat NK granules induces apoptosis of detergent-permeabilized or perforin-permeabilized target cells (16,39); 2) noncytotoxic rat basophilic leukemia cells transfected with perforin and granzyme vectors * This work was supported by National Institutes of Health Grants DK49786 and CA49712. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The induce DNA fragmentation in target cells only when they express both perforin and granzyme A or B (40,41); and 3) granzyme A is a "late" acting protease, since NK granulepurified granzyme A induces apoptosis only after 14 h of incubation time, compared with 2 h of incubation time needed for the production of granzyme B-induced apoptosis (16).
Another tryptase called granzyme K is also a candidate for granzyme B-independent cytotoxicity. The murine orthologue of human granzyme K (28,29) has not yet been identified, but rat granzyme K has been purified and its cDNA has been cloned (16,18). Similar to granzyme A, granzyme K purified from a rat large granular lymphoma cell line (RNK-16) induces apoptosis of target cells in a late 14-h cytotoxicity assay (16). Moreover, human granzymes A and K map to the same location on chromosome 5q11-q12 (42). These data suggested that murine granzyme K is likely to exist, and that it may be located near the granzyme A gene on mouse chromosome 13.
In this paper, we report the cloning of murine granzyme K and establish the tight linkage between murine granzymes A and K. We have created a null mutation of granzyme A by replacing exon 2 and a significant portion of intron 2 with a standard PGK-Hygro cassette. We have found that our granzyme AϪ/Ϫ mice are viable and have normal development, fertility, and hematopoiesis, similar to the mice reported by Ebnet et al. (43), which bear a different mutation in the granzyme A gene. Expression of the granzyme K gene is not altered by the insertion of the PGK-Hygro cassette in the granzyme A locus. Granzyme AϪ/Ϫ CTL have only a slightly reduced ability to attack and kill allogeneic target cells, perhaps because they have an intact granzyme B cluster, and/or perhaps because they exhibit persistent expression of granzyme K (and/or D), which is also a potent tryptase.

EXPERIMENTAL PROCEDURES
Molecular Cloning of Murine Granzyme K cDNA-Total RNAs derived from mouse splenocytes activated with concanavalin A/IL-2 for 2 days, or from an NK-like cell line (a CD3Ϫ, NK1.1ϩ splenic tumor line harvested from transgenic mice expressing simian virus 40 (SV40) T antigen driven by the human granzyme H promoter) 2 were isolated using a previously described method (44). The first strand of cDNA was synthesized with reverse transcriptase (Life Technologies, Inc.) and random hexamer primers. Using KlenTaqLA (Wayne Barnes, Washington University, St. Louis), double-stranded cDNA templates were then PCR amplified with the following pair of primers that were derived from rat sequences and corresponded to the most conserved regions between the human and rat granzyme K cDNAs: 5Ј-CATTCCAGGC-CTTTTATGGC-3Ј and 5Ј-GGATCAAAAGCAAGCTTGCC-3Ј. A 650-bp PCR product was subcloned into the pCR 2.0 vector (Invitrogen, San Diego). The inserts were sequenced with an ABI373 automated sequencer.
Molecular Cloning of Murine Genomic Granzyme K-A 500-bp murine granzyme A probe spanning the two alternatively spliced leader exons of granzyme A was generated using PCR amplification of mouse genomic DNA. The PCR amplified probe was sequenced and tested in Southern analysis to confirm its ability to detect unique mouse granzyme A bands. This probe was then used to screen a murine BAC library (Genome Systems, St. Louis, MO). Four BAC clones (BAC 1254 -1257) that hybridized strongly with the granzyme A fragment were obtained and subsequently tested for the presence of granzyme K gene in Southern analysis using the 650-bp murine granzyme K cDNA as the probe. One BAC clone, BAC 1255, hybridized with the granzyme K cDNA probe. DNA obtained from BAC 1255 was then digested with HindIII or PstI, and two HindIII and one PstI fragments were subcloned into pUC19 based vectors and sequenced.
Mapping of the Distance between Murine Granzymes A and K Loci-DNA isolated from BAC 1255 was digested with rare cutting restriction endonucleases, electrophoresed on 1% agarose gels using a CHEF DRII apparatus (Bio-Rad), transferred onto Hybond (Amersham) membranes, and probed with either the 500-bp 5Ј murine granzyme A probe or a 2.0-kb murine granzyme K HindIII genomic fragment containing exons 1 and 2. The hybridizing and washing conditions of these Southern blots were performed as described previously (37).
Production of Activated Lymphocytes-Murine splenocytes were activated either with concanavalin A/IL-2, or with one-way mixed lymphocyte reactions (MLR) as described previously (30). Mouse spleen cells were also cultured with high dose (1000 units/ml) recombinant human IL-2 for 10 days to generate LAK cells, as described previously (31).
Analysis of mRNA-Total cellular RNA was isolated from activated lymphocytes and analyzed using an S1 nuclease protection assay, as described (45). Probes specific for murine granzyme A, murine granzyme B, and murine ␤ 2 -microglobulin genes were previously reported (30). A probe that specifically detected correctly spliced exon 4 of murine granzyme K was created using a 1.8-kb BglII genomic fragment isolated from the plasmid containing the 4.7-kb HindIII genomic fragment of granzyme K. Correctly spliced granzyme K mRNA protects a probe fragment of 119 nucleotides from S1 digestion. All probes were labeled to approximately the same specific activity.
Construction of the Murine Granzyme A Targeting Vector-The 500-bp murine granzyme A DNA probe that spans the alternatively spliced leader exons was used to screen a 129/Sv mouse genomic library (Stratagene). A clone containing an 8-kb BamHI fragment containing most of the murine granzyme A gene was obtained. The 8-kb genomic fragment was subcloned into pUC9, and Southern blot and partial sequencing analyses confirmed that this DNA contained the 5Ј-flanking region and exons 1 through 4 of mouse granzyme A. This 8-kb BamHI fragment was used to generate the granzyme A gene targeting vector. A 2.7-kb EcoRI-HindIII region containing exon 2 of the granzyme A gene was replaced with a 1.6-kb fragment containing the hygromycin phosphotransferase gene driven by the phosphoglycerate kinase I gene promoter (PGK-hygro (46)), resulting in an exon 2 deletion mutation in the granzyme A gene. The PGK-hygro cassette was inserted in the opposite transcriptional orientation from granzyme A. The completed targeting vector consisted of a 3.5-kb BamHI-EcoRI genomic fragment containing 5Ј-flanking sequence and leader exons upstream from PGKhygro, and a 1.9-kb HindIII-BamHI genomic fragment including exons 3 and 4 downstream from PGK-Hygro.
Production of Granzyme A Null Mutant Mice-RW4 ES cells (derived from 129/SvJ mice) were transfected with the granzyme A targeting vector, and hygromycin-resistant clones were isolated essentially as described (30). The resistant ES clones were screened by Southern blotting using a 327-bp 3Ј external probe that was PCR amplified with primers based on published mouse granzyme A genomic sequences (47,48). The external probe detected a 9-kb wild-type and a 5-kb mutant allele of PstI-digested mouse genomic DNA. Two out of 110 ES clones (ES 64 and ES 144) were correctly targeted, and these two clones were further confirmed in a Southern analysis with a hygromycin phosphotransferase specific probe. C57BL/6 blastocysts were then microinjected with ES cells from each of the two targeted clones and implanted into pseudopregnant Swiss Webster foster females, as described previously (30). Chimeric males with a high percentage of agouti color were mated with C57BL/6 females, and their offspring were examined for germline transmission of the disrupted granzyme A allele by Southern analysis using the 3Ј external probe described above. ES 64 derived heterozygotes were intercrossed to produce homozygous mutant mice. 2  Flow Cytometric Analysis-Thymocytes, splenocytes, and activated splenocytes from day 5 MLRs were stained with fluorescein isothiocyanate-or phycoerythrin-conjugated monoclonal antibodies against various mouse cell surface markers and analyzed using a FACScan flow cytometer and CellQuest version 1.2.2 software (Becton-Dickinson), as described previously (30).
Cytotoxicity Assays-Standard 51 Cr or 125 I-UdR release assays were performed exactly as described (50). BamHI, EcoRI, HindIII, and PstI sites are indicated, but these restriction maps are not complete. Note that only the granzyme A gene contains the alternatively spliced leader exons.

RESULTS
Isolation of Murine Granzyme K cDNA-To identify the mouse homologue of human and rat granzyme K, we performed reverse transcriptase-PCR using primers based on rat cDNA sequences that corresponded to the most conserved regions between the human and rat (77% identity between human and rat granzyme K cDNAs). We amplified a 650-bp product from Con/IL-2 activated mouse splenocytes and a mouse NK-like cell line. Sequence analysis confirmed that the 650-bp fragment from each cDNA source contained a highly homologous cDNA encompassing exons 2, 3, 4, and most of 5. Fig. 1 shows the deduced amino acid sequence of this cDNA. The 650-bp cDNA sequence isolated was not full-length, but the 5Ј region, all of exons 1 through 5, and a putative polyadenylation signal (AATAAA) were identified in the genomic sequence (Fig. 2).
The mouse granzyme K cDNA, which is predicted to encode 264 amino acids, is 89.0 versus 77.6% identical to the rat and human cDNAs, respectively. Based on previous work, the NH 2 terminus of the mature, active protease is designated ϩ1 at the isoleucine position shown in Fig. 1. The deduced amino acid sequence of murine granzyme K shares 86.0 versus 70.9% identity with the rat and human homologues, respectively. In the aligned amino acid sequences of the mouse, rat, and human granzyme K, the catalytic triad residues (His 41 , Asp 90 , and Ser 188 ) and six of eight cysteine residues that are expected to form three internal disulfide bonds (26 -42, 123-155, and 173-194) are conserved among all three species (Fig. 1). The two remaining cysteine residues (184 and 208) may form a fourth disulfide bond. The granzyme K sequence does not contain any consensus sites for N-linked glycosylation.
Among the murine granzymes, granzyme K cDNA is most similar to granzyme A (53.5% identity). Granzymes A, K, and other tryptases have an aspartic acid residue at position 182 (six residues amino-terminal to the active site serine) that is thought to determine specificity for lysine or arginine in the P1 position. However, in contrast to granzyme A (which forms a 60-kDa homodimer due to an interchain bond made by the ninth cysteine) granzyme K probably acts as a monomer because it does not contain the additional cysteine residue. Therefore, the active, mature murine granzyme K protein is predicted to be a 26-kDa monomer, in keeping with the size of purified rat (16,18) and human granzyme K (24).
Genomic Organization and Linkage of the Granzyme K and A Genes in Mice-We screened a mouse BAC library derived from 129/SvJ mice using a 500-bp genomic fragment containing the alternatively spliced leader exons of mouse granzyme A, and obtained four clones (BAC 1254 -1257) that hybridized with this probe. Southern analysis of DNA from the BAC clones (using probes encompassing various regions of murine granzyme A locus) confirmed the presence of the entire granzyme A genomic sequence in all four BAC clones. These clones were further examined by Southern blotting using a 650-bp murine granzyme K cDNA probe spanning exons 2 through 5; one clone, BAC 1255, hybridized strongly with this probe. A Hin-dIII or PstI digest of BAC 1255 DNA probed with the 650-bp granzyme K cDNA probe revealed two HindIII (2.0 and 4.7 kb) and one PstI (2.2 kb) fragment that were subcloned into pUC 19 vectors. All three plasmids were completely sequenced using oligonucleotides based on the cDNA and the newly derived genomic sequences. The 2.0-kb HindIII fragment contained the 5Ј region and exons 1 and 2; the 4.7-kb HindIII piece contained exons 3, 4, and most of 5; and the 2.2-kb PstI fragment included exons 4 and 5 and 3Ј-flanking sequence. The complete genomic structure of mouse granzyme K is shown in Fig. 2. Consensus intron donor and acceptor sites are present in the expected position ( Fig. 2A). The entire gene consists of 5 exons and 4 introns (Fig. 2B), a common feature for this family of genes. Granzyme A spans over 10 kb and contains two alternatively spliced leader exons (47,48); the granzyme K gene encompasses ϳ6 kb of genomic DNA, and appears to have only one leader-encoding exon.
Since the localization of both granzymes A and K to BAC 1255 indicated that the two genes are closely linked, we next attempted to map the precise distance between them. DNA from BAC 1255 was treated with rare-cutting restriction enzymes and analyzed using Southern blotting with mouse granzyme A or K-specific probes. The granzyme A-specific probe consisted of a 500-bp DNA fragment that spans the alternatively spliced exons (described above). The 2.0-kb HindIII granzyme K probe contained the 5Ј region and exons 1 and 2, and detected unique granzyme K bands in Southern blots of mouse genomic DNA. SunI, AscI, SgfI, and PacI all cleaved the insert once; NotI released the insert of 145 kb, which contained both genes. However, a 65-kb PmeI fragment of BAC 1255 DNA hybridized to both the granzyme A and K probes, revealing that the two genes are at most 65 kb apart (data not shown). None of the co-digests with enzymes that cut once or twice within the insert localized the two genes on a smaller fragment, so the precise distance between the two genes is not yet resolved.
Expression of Murine Granzyme K mRNA in CTL and LAK Cells-We investigated the expression pattern of murine granzyme K mRNA by S1 nuclease protection analysis, using a granzyme K-specific probe end-labeled in exon 4 (Fig. 3). Correctly spliced granzyme K mRNA protected an exon 4-derived fragment of 119 nucleotides from S1 digestion. A tissue survey revealed no detectable granzyme K mRNA in murine bone

FIG. 3. Expression of the murine granzyme K mRNA in MLRderived CTL and LAK cells.
An S1 protection analysis was performed using the previously described probes specific for mouse ␤ 2microglobulin, granzyme A, and granzyme B, and a newly generated mouse granzyme K probe. As depicted at the bottom of the figure, correctly spliced granzyme K mRNA protects an exon 4-derived probe fragment of 119 nucleotides from S1 digestion. RNAs isolated from day 5 MLR-derived CTL (lanes 2 and 3) or high dose IL-2 activated LAK cells (lanes 4 and 5) were simultaneously hybridized with ␤ 2 -microglobulin and granzyme B probes (lanes 2 and 4) or with granzymes A and K probes (lanes 3 and 5). Lane 6 contains a no RNA control (hybridized with all four probes). marrow, heart, kidney, liver, lung, small intestine, large intestine, resting spleen, or thymus (data not shown). We next examined granzyme K mRNA levels in day 5 MLR-derived CTL, and in LAK cells generated in splenocyte cultures activated with 1000 units/ml IL-2 for 10 days. In addition to the exon 4 granzyme K probe, probes for granzyme A and ␤ 2microglobulin were used as controls for lymphocyte activation, and for RNA quality and content, respectively. The results in Fig. 3 show the presence of granzyme K mRNA, along with the expected expression of granzyme A and B mRNAs, in both MLR-derived CTL and LAK cells. Densitometric analysis of these autoradiographs revealed that granzyme K mRNA levels are ϳ1.5% that of granzyme A in day 5 MLR cells, and ϳ15% that of granzyme A in LAK cells.
Creation of Mice with a Targeted Disruption of the Granzyme A Gene-We generated granzyme AϪ/Ϫ mice using the targeting vector shown in Fig. 4A (top panel). The mutant, homologous recombinant granzyme A allele contains a deletion of exon 2 (which contain the histidine residue of the catalytic triad). Using a 3Ј probe that lies just outside of the sequence used for targeting (Fig. 4A, middle panel), two correctly targeted 129/ SvJ-derived RW4 ES clones (out of 110 screened) were identified. Both ES clones were injected into C57BL/6J blastocysts. One clone gave rise to highly chimeric males that transmitted the mutant granzyme A locus in the germline. In Fig. 4B, which shows a Southern analysis of PstI-digested genomic tail DNA of offspring generated by mating the mutant granzyme A heterozygotes, the expected 9.0-kb wild type and 5.0-kb targeted granzyme A alleles were detected by the 3Ј probe described above. Heterozygote matings yielded mice with approximately Mendelian ratios of 40:59:34 (ϩ/ϩ:ϩ/Ϫ:Ϫ/Ϫ) mice. Granzyme Aϩ/Ϫ and Ϫ/Ϫ mice have normal development and fertility, and are grossly indistinguishable from their wild type counterparts.
Lack of Granzyme A Expression in Granzyme AϪ/Ϫ Cytotoxic Lymphocytes-S1 protection and BLT esterase activity assays were performed with day 5 MLR-derived CTL and/or LAK cells generated from granzyme Aϩ/ϩ or Ϫ/Ϫ mouse spleens. S1 protection analysis, using a previously described exon 4-specific granzyme A probe (which lies outside of the mutant region), revealed a complete absence of granzyme A mRNA in granzyme AϪ/Ϫ MLR-derived CTL (Fig. 5A). Granzyme AϪ/Ϫ cytotoxic effector cells revealed a normal ability to activate in response to allogeneic stimuli (MLR) or to IL-2 in LAK cell cultures, since equivalent levels of granzyme B mRNA were detected in granzyme Aϩ/ϩ and Ϫ/Ϫ CTL. Unaltered levels of granzyme K mRNA were detected in the AϪ/Ϫ mice; the presence of the PGK-hygro cassette in the granzyme A gene therefore does not reduce the expression of the closely linked granzyme K gene. We next measured the ability of MLR-derived granzyme Aϩ/ϩ or Ϫ/Ϫ CTL to cleave the synthetic substrate BLT. Fig.  5B shows that a markedly reduced amount of BLT is hydrolyzed by lysates from granzyme A-deficient CTL; however, a small amount of residual BLT esterase activity remains in these mutant CTL.
Normal Development and Activation of Peripheral Lymphocytes in Granzyme AϪ/Ϫ Mice-We measured the subsets of lymphocytes in the thymus and spleen of granzyme AϪ/Ϫ mice. Fig. 6 is a flow-cytometric analysis of granzyme Aϩ/ϩ or Ϫ/Ϫ lymphocyte populations stained with mAbs against murine CD3, CD4, CD8, B220, and NK1.1 cell surface molecules. Granzyme Aϩ/ϩ or Ϫ/Ϫ thymocytes included the typically predominant population of immature CD4ϩCD8ϩ (double-positive) cells and small percentages of mature CD4ϩ or CD8ϩ (single positive) cells (Fig. 6A). Similarly, both wild type and mutant mice had virtually identical proportions of all expected subsets of immune cells in their spleens (Fig. 6B). Day 5 MLR-derived granzyme Aϩ/ϩ and Ϫ/Ϫ CTL ((H-2 b ) anti-Balb/c (H-2 d )) had equivalent percentages of CD3ϩ, CD4ϩ, and CD8ϩ cells, and the majority of these CTL were CD8ϩ, as expected (Fig. 6C). Moreover, cell counts revealed similar numbers of wild type and mutant CTL generated in MLR (data not shown), suggesting that granzyme AϪ/Ϫ splenocytes proliferate normally. Together, the cell counts and FACS profiles imply that granzyme A-deficient T lymphocytes develop normally and undergo proper activation upon allogeneic stimulus in vitro.
Slightly Reduced Cytotoxic Activity of Granzyme A-deficient CTL against Allogeneic Targets-The normal proliferation and activation of granzyme AϪ/Ϫ splenocytes in response to allostimulus in vitro allowed us to specifically address the ability of the mutant, day 5 MLR-derived CTL to attack allogeneic targets using standard lytic assays. Fig. 7 shows results of 51 Cr release and 125 I-UdR release assays at varying effector to target (E:T) ratios at an early (2 h) or a late (8 h) time after the incubation of effector and target cells together. In control experiments, both the wild type and mutant CTL lacked cytotoxic activity against the syngeneic, H-2 b expressing EL4 targets at all time points and E:T ratios tested (data not shown). Granzyme AϪ/Ϫ CTL (H-2 b anti-H-2 d ) were able to induce nearly normal levels of 51 Cr or 125 I-UdR release from the H-2 d ex-pressing allogeneic TA3 (Fig. 7A), P815 (Fig. 7B), or YAC-1 (Fig. 7C) target cells at all E:T ratios tested (at both early and late time points). The slightly reduced levels of 51 Cr and 125 I-UdR release mediated by the mutant CTL were consistently observed in multiple experiments against a wide variety of target cells, including L1210 cells transfected with a sense or antisense Fas cDNA construct (data not shown). Although the differences shown are small, they are consistent and statistically significant.

DISCUSSION
In this report, we have described the phenotype of granzyme AϪ/Ϫ mice obtained by gene targeting. Similar to granzyme AϪ/Ϫ mice created by Ebnet and colleagues (43), our granzyme A-deficient animals have normal development and fertility. Cytotoxic lymphocytes derived from these mice have normal programs of activation and proliferation, and express normal levels of the tightly linked granzyme K gene. In addition, standard lytic assays performed with allo-specific CTL derived from these mice reveal a small defect in their ability to induce 51 Cr and 125 I-UdR release from allogeneic target cells at all time points and E:T ratios tested. The initial analysis of our granzyme AϪ/Ϫ mice therefore corroborates the phenotype observed in cytotoxic effector cells derived from granzyme Adeficient mice produced by Ebnet and colleagues (43).
The small defect in cytotoxicity present in granzyme A-deficient mice is puzzling, since granzyme A is known to be an enzyme that is capable of inducing apoptosis in target cells. However, the similar phenotypes observed in the mice described by Ebnet et al. (43) and in this report strongly suggest that the minimal phenotype is, in fact, correct. The mice de- scribed by Ebnet et al. (43) contain an insertional mutation of PGK-Neo in exon 4; this mutant gene could potentially be alternatively spliced to produce mutant granzyme A molecules that contain exons 1, 2, 3, and 5. However, no granzyme A protein could be detected by Western analysis in that study, and minimal residual tryptase activity was noted on BLT esterase studies. No defect in the cytotoxicity of granzyme AϪ/Ϫ CTL was observed, but these mice failed to clear the cytopathic pox virus ectromelia in normal fashion (51). In contrast, our mutation removes one of the exons that contains part of the catalytic triad, and eliminates nearly the entire amino terminus of the protein. A small amount of residual tryptase activity was detected in the activated CTL from our mice; a definite small reduction in cytotoxicity was observed, in contrast to the normal cytotoxicity detected in the Ebnet et al. (43) mice. The mice made by Ebnet et al. (43) are in an inbred C57BL/6 background, and our mice are in a mixed BL6 ϫ 129 background. Therefore, two different mutations in this gene (created in two different genetic backgrounds) have yielded similar phenotypes, making it extremely likely that these results represent an accurate view of the granzyme A null mutant state.
There are several potential reasons why granzyme AϪ/Ϫ CTL may have minimally reduced cytotoxicity in vitro. First, granzyme AϪ/Ϫ mice have a functioning granzyme B locus. In previous work from our laboratory, we have shown that enzymes in the granzyme B cluster are required for the rapid induction of apoptosis in susceptible target cells (30,31,38). Since the rapid killing pathway is intact in granzyme A-deficient animals, the granzyme A defect may be nearly impossible to detect. By creating animals that are doubly deficient for granzymes A and B, we may be able to precisely define the role of granzyme A for the late, granzyme B-independent cytotoxicity that we have previously described (30,31,52). Careful analysis of the cytotoxic phenotypes of these animals will also allow us to determine whether any other molecules in cytolytic granules contribute significantly to "perforin-dependent" cytotoxicity. If granzyme A ϫ B cluster-deficient animals have the same phenotype as those that are perforin-deficient, then it would seem unlikely that additional granule proteins contribute significantly to granule-mediated cytotoxicity. Experiments designed to test these hypotheses are in progress.
An alternative explanation for the minimal cytotoxic defect of granzyme A deficient mice may be the "rescue" of granzyme A function by other granzymes with a similar specificity. As noted above, granzyme K is a potent tryptase. Despite its close linkage with granzyme A, the PGK-hygro cassette does not alter granzyme K expression in granzyme AϪ/Ϫ CTL. In addition to granzyme K, granzyme D is another known CTL tryptase (8), and its expression is unaltered in granzyme A-deficient mice as well (data not shown), since it lies in the granzyme B locus on chromosome 14. Regardless, granzymes D and K must contribute only minimally to the total tryptase activity present in activated CTL, since BLT esterase activity was reduced by more than 80% in our granzyme A-deficient mice. However, the formal possibility exists that the low level of residual tryptase activity provided by granzymes D, K, and perhaps other granzymes is sufficient to rescue the mice from the consequences of granzyme A deficiency.
As an initial step toward understanding the contribution of these other granzymes for CTL function, we have cloned the murine granzyme K gene which encodes a tryptase that is similar to granzyme A. Granzyme K has three structural features that are common to all serine proteases (Fig. 1): 1) the mature, active protease contains the characteristic NH 2 -terminal sequence (IIGG); 2) the catalytic triad residues (H, D, and S) are present at homologous positions flanked by conserved peptides; and 3) six cysteine residues (out of a total of eight) are conserved in homologous positions and are expected to form internal disulfide bonds. Of all the known granzymes with these common features, granzyme K is most similar to granzyme A. Both granzymes A and K have the same specificitydetermining aspartic acid residue located six residues aminoterminal to the active site serine. Other substrate specificity defining amino acids, including glycine at position 205 and 209, are also conserved in identical positions in granzyme K. Besides these common tryptase determining residues, the two extra cysteines (which are not characteristic of all serine proteases) are located in homologous positions in granzymes A and K. These common structural features suggest that these two proteases may cleave similar physiological substrates and have similar functions. However, granzyme K does not contain the ninth cysteine residue that allows granzyme A to form homodimers. Furthermore, there are several stretches of amino acid residues that are conserved in granzyme K but not in granzyme A, suggesting that granzyme K could have slightly different substrate specificities and/or functions from granzyme A.
Highly related serine protease genes tend to occur in clusters (e.g. the "granzyme B" cluster and the "neutrophil elastase" cluster). Extensive studies with both the mouse and human granzyme B gene clusters have demonstrated that some genes within this cluster are expressed specifically in the activated lymphocyte compartment (35)(36)(37). Our present study, together with previously published work from the human system, establishes that granzymes A and K genes are closely linked as well. In the human system, Fink and colleagues (34) have previously demonstrated that the granzyme A gene maps to 5q11-12, a region that is syntenic between human chromosome 5 and mouse chromosome 13, band D, where the mouse granzyme A gene is localized. Moreover, using fluorescent in situ hybridization, Baker et al. (42) have shown that the human granzyme K also maps to 5q11-12, thus identifying a new locus of the tryptase family of serine proteases. In our study, we extend these findings by identifying both murine granzymes A and K genes in a 65-kb mouse genomic fragment derived from a BAC clone. The significance of the clustering of these genes is not yet clear, but it suggests that these genes share common regulatory elements, such as locus control regions.
Previous work has suggested that the expression of granzymes A and K is restricted to activated T cells and NK cells.
Our results indicate that day 5 MLR-derived CTL and LAK cells express much lower levels of granzyme K than granzyme A mRNA. This observation corroborates data obtained by Hameed and colleagues (24), who have purified both granzymes A and K from IL-2 activated human LAK cells; they found that granzyme K protein is much less abundant than granzyme A. However, in that study, granzyme K had twice the specific activity of granzyme A. Rat granzyme K has been demonstrated to be a highly expressed granzyme that can be easily purified from RNK-16, the rat large granular lymphocyte leukemia cell line (16,18); Shi and colleagues (16) have demonstrated that granzyme K is more abundant than granzyme A in RNK-16 cells. Further analysis of various activated cytotoxic lymphocyte populations is needed to more accurately determine the relative abundance of granzyme A versus K in cyto-toxic lymphocytes, and to more precisely define the roles of these granzymes in cell mediated cytotoxicity.