Although the functions of granzyme A (GzmA) and GzmB are well-defined, a number of orphan granzymes of unknown function are also expressed in cytotoxic lymphocytes. Previously, we showed that a targeted loss-of-function mutation for GzmB was associated with reduced expression of several downstream orphan granzyme genes in the lymphokine-activated killer cell compartment. To determine whether this was caused by the retained phosphoglycerate kinase I gene promoter (PGK-neo) cassette in the GzmB gene, we retargeted the GzmB gene with a LoxP-flanked PGK-neo cassette, then removed the cassette in embryonic stem cells by transiently expressing Cre recombinase. Mice homozygous for the GzmB null mutation containing the PGK-neo cassette (GzmB−/−/+PGK-neo) displayed reduced expression of the closely linked GzmC and F genes in their MLR-derived CTLs and lymphokine-activated killer cells; removal of the PGK-neo cassette (GzmB−/−/ΔPGK-neo) restored the expression of both genes. Cytotoxic lymphocytes derived from mice with the retained PGK-neo cassette (GzmB−/−/+PGK-neo) had a more severe cytotoxic defect than those deficient for GzmB only (GzmB−/−/ΔPGK-neo). Similarly, GzmB−/−/+PGK-neo mice displayed a defect in the allogeneic clearance of P815 tumor cells, whereas GzmB−/−/ΔPGK-neo mice did not. These results suggest that the retained PGK-neo cassette in the GzmB gene causes a knockdown of GzmC and F expression, and also suggest that these granzymes are relevant for the function of cytotoxic lymphocytes in vitro and in vivo.
Cytotoxic lymphocytes can use multiple pathways to induce target cell death, including secretory mechanisms (e.g., TNF-α and IFN-γ) and contact-dependent mechanisms, including the granule exocytosis pathway and the Fas pathway (1, 2, 3). In several studies the killing of tumor cells has been shown to be dependent on the granule exocytosis pathway, but variable results have been obtained with different tumor targets in different laboratories, suggesting that more than one mechanism is relevant, and that different mechanisms may be important for different tumors (1, 4, 5, 6, 7, 8, 9, 10, 11).
The secretory granules of cytotoxic lymphocytes (CTL, NK, and LAK cells) migrate along microtubules to the site of contact between the effector and target cells, where a tight immunologic synapse is formed (1, 2, 3). There, the contents of the cytotoxic granules are secreted into the small intercellular space, where perforin can polymerize in the presence of calcium to form membrane perturbations that may be relevant for the entry of other granule proteins, including granzymes. Perforin also appears to play a critical role in the trafficking of granzymes once they enter the target cell. Granzymes, a class of neutral serine proteases, are thought to induce target cell death by cleaving critical intracellular substrates (1, 2). Granzyme A (GzmA)7 and B have very different substrate specificities, and appear to cleave nonoverlapping sets of target cell proteins (1, 2, 12, 13). Although GzmB induces classical apoptosis associated with mitochondrial depolarization, nuclear collapse and fragmentation, and activation of caspase-activated DNase, GzmA induces cell death that has many of the features of apoptosis, but is clearly different from death induced by GzmB (1, 2, 14, 15, 16). The apparent nonredundancy of these two granzymes might be explained in part by the fact that viruses and tumor cells sometimes express inhibitors of granzymes, presumably to thwart the ability of cytotoxic lymphocytes to kill these cells (1). Under these circumstances, alternative granzymes with different specificities would be required to induce cell death.
In both humans and mice, a variety of additional granzymes of unknown function (orphans) are expressed in cytotoxic lymphocytes (17). In the mouse a large number of functional granzyme genes are found downstream from GzmB in a locus commonly known as the GzmB gene cluster. GzmC lies ∼24 kb downstream from GzmB and is most closely related to human GzmH, which is the only orphan granzyme gene found downstream from human GzmB. In the mouse, however, several additional granzyme genes lie downstream from GzmC, including (5′->3′) GzmF, G, L, N, D, and E, followed by cathepsin G (a promyelocyte-specific serine proteinase of similar structure) and several mast cell chymases (17).
In our initial GzmB knockout mouse, we removed the coding sequences from exon 1 and part of intron 1 and replaced these sequences with a functional phosphoglycerate kinase I gene promoter (PGK)-neo cassette (18). We later learned that mixed strain mice containing this mutation display significantly reduced the expression of several downstream granzyme genes (C, F, and D) in the lymphokine-activated killer (LAK) cell compartment (19). To determine whether this was a neighborhood knockdown effect caused by the retained PGK-neo cassette, we retargeted the GzmB gene with a LoxP-flanked PGK-neo cassette, then removed it from the targeted ES cells with Cre recombinase. Removal of the PGK-neo cassette restored the expression of the downstream granzymes, which allowed us to compare the functions of cytotoxic lymphocytes deficient for GzmB only (GzmB−/−/ΔPGK-neo) or GzmB plus reduced levels of GzmC and F (GzmB−/−/+PGK-neo). To our surprise, cytotoxic lymphocytes deficient for GzmB, C, and F displayed a more striking defect in killing than cytotoxic lymphocytes deficient for GzmB only; this was true not only for LAK cells, but also for CTL derived from MLR. To validate these results in vivo, we assessed the ability of these mice to clear an allogeneic tumor (P815, H-2Kd) and again found that mice deficient for GzmB, C, and F have a more significant clearance defect than those deficient for GzmB only. In sum, these results have shown that the retained PGK-neo cassette in the GzmB gene does cause a neighborhood effect, and the reduced expression of GzmC and F is relevant for the functions of CTLs and LAK cells in vitro and in vivo.
Materials and Methods
Construction of the GzmB targeting vector
A 4.16-kb EcoRI fragment containing the entire GzmB locus was subcloned into pUC19 to generate the backbone of the GzmB targeting vector, similar to that previously described (18). A 350-bp AvrII fragment containing the initiation and start sites and the entire first exon (and part of the first intron) of the GzmB gene was replaced with a 1.6-kb NotI fragment containing a LoxP-flanked neomycin phosphotransferase gene driven by PGK-neo (20). The resulting targeting vector contained 1.03 kb of genomic DNA from the GzmB 5′ flank upstream from the PGK-neo insertion and 2.75 kb of genomic DNA containing exons 2–5 downstream from PGK-neo. The targeting vector was linearized with SalI and gel-purified before electroporation into Rw4 embryonic stem cells (ES; 129/SvJ).
Generation of 129/SvJ mice containing the GzmB−/−/+PGK-neo and GzmB−/−/ΔPGK-neo mutations
GzmB+/−/+PGK-neo ES were generated by electroporation of the GzmB targeting vector in Rw4 ES cells, neomycin-resistant clones were expanded, and homologous recombination events were identified by Southern blot analysis using an external probe (data not shown) (18). Correct targeting was confirmed by Southern blot analysis using an internal probe after EcoRI digestion (Fig. 1). ES clones with specific homologous recombination were injected into C57BL/6 blastocysts that were implanted into pseudopregnant Swiss-Webster mice. Male chimeras were bred with 129/SvJ female mice, and germline transmission of the mutant alleles was confirmed by Southern blot analysis. Heterozygous animals were intercrossed to generate homozygous GzmB−/−/+PGK-neo deficient mice. GzmB+/−/ΔPGK-neo ES clones were generated by transiently transfecting GzmB+/−/+PGK-neo ES clones with a Turbo-Cre-expressing plasmid as previously described (20, 21), resulting in the recombination of LoxP sites and the deletion of the internal PGK-neo cassette. Resultant GzmB+/−/ΔPGK-neo clones were screened for neomycin sensitivity, and deletion of the PGK-neo cassette was confirmed by Southern blot analysis (Fig. 1). Mutant mice were generated exactly as described above. Heterozygous animals were intercrossed to generate homozygous GzmB−/−/ΔPGK-neo deficient mice. Both mutant strains were maintained as homozygous breeding colonies in the 129/SvJ background.
MLR and LAK cell preparations for in vitro cytotoxicity and flow cytometry assays
MLR and LAK cell preparations were performed as previously described (18, 22). For MLR preparations, stimulator (BALB/c wild-type (WT); H-2Kd) and effector (129/SvJ WT, GzmB−/−/+PGK-neo, and GzmB−/−/ΔPGK-neo; all H-2Kb) spleens were harvested into PBS, and single cell suspensions were made by gently crushing spleens between frosted microscope slides. Red cells were lysed in ACK buffer (154 mM NH4Cl, 10 mM KHCO3, and 0.1 mM EDTA; sterile-filtered) and washed twice in PBS. The cell suspension from each BALB/c spleen was resuspended in 5 ml of complete K10 medium (RPMI 1640, 10% heat-inactivated FCS, 10 mM HEPES, 1% nonessential amino acids, 1% sodium pyruvate, 1% l-glutamine, 1× penicillin/streptomycin, 0.57 μM β-ME) and 50 U/ml recombinant human IL-2 (rhIL-2) and irradiated (2000 cGy). The cell suspensions from WT, GzmB−/−/ΔPGK-neo, and GzmB−/−/+PGK-neo spleens were each resuspended in 40 ml of complete K10 medium and 50 U/ml rhIL-2. Irradiated BALB/c stimulators and 129/SvJ, GzmB−/−/ΔPGK-neo, or GzmB−/−/+PGK-neo responders were combined into 45 ml of complete K10 medium and 50 U/ml rhIL-2. Samples were mixed, plated in three 100-mm non-tissue culture-treated plates, and incubated for 4 days. Before FloKA, the mononuclear cells were purified over Ficoll (Sigma-Aldrich), washed with PBS, and resuspended in complete K10 medium and 50 U/ml rhIL-2.
Quantitative real-time RT-PCR (qRT-PCR)
Total RNA was isolated from 10 × 106 LAK cells or Ficoll-purified cells from MLRs using the RNeasy Mini Kit (Qiagen). One microgram of total RNA from each sample was reverse-transcribed into cDNA using the TaqMan reverse transcriptase (ABI; Applied Biosystems) and analyzed by PCR. The relative mRNA abundance of each granzyme mRNA was measured by the 7300 real-time PCR system from Applied Biosystems, using primer sets that were proven to amplify only the self granzyme cDNA using the conditions described below. The primer sets used in these studies were as follows: GzmA-specific primer set: forward, 5′-AGA CCG TAT ATG GCT CTA CT-3′; and reverse, 5′-CCC TCA CGT GTA TAT TCA TC-3′; GzmB-specific primer set: forward, 5′-ATC AAG GAT CAG CAG CCT GA-3′; and reverse, 5′-TGA TGT CAT TGG AGA ATG TCT-3′; GzmC-specific primer set: forward, 5′-ATG AGT TTC TGA AAG TTG GTG-3′; and reverse, 5′-TCA TTA GAA CGG TCA TCA GG-3′; GzmF-specific primer set: forward, 5′-TGA GGT TTG TGA AAG ATA ATG-3′; and reverse, 5′-TCA CTG GTG TTG TCC TTA TC-3′; GzmG-specific primer set: forward, 5′-GTT CAT TAA GTC TGT GGA TAT CG-3′; and reverse, 5′-GTC TTG GAA TAG GTG TAC CAG-3′; GzmD-specific primer set: forward, 5′-AAA CAG CTC TGT CCA AAG CTC-3′; and reverse, 5′-CAA ATC TCT GTG GTC TCA GTG-3′; and GzmE-specific primer set: forward, 5′-CTG CTC ACT GCA GGA ACA GGA-3′; and reverse, 5′-CAA ATC TCT GTG GTC TCA GTG-3′. The qRT-PCRs were conducted in the presence of the DNA intercalating dye SYBR Green. PCR conditions used for all primer sets were as follows: 95°C hot start for 10 min, followed by 40 amplification cycles of 95°C for 30 s (denaturing), 59°C for 40 s (annealing), and 72°C for 40 s (extension). A PCR amplification profile was derived by recording the SYBR Green fluorescence intensity, which was in linear relation to the amount of formed PCR product (as a function of PCR cycle number). Standard curves were generated by plotting the PCR cycle number at which a reaction entered exponential amplification vs the amount of input DNA of plasmids containing the cDNA of each granzyme gene. The standard curves were then used for a determination of sample template concentration.
Western blot analysis
Ficoll-purified MLR and LAK cell lysates were prepared and analyzed using standard Western blotting techniques as previously described (16, 22). Primary Abs included rabbit anti-mouse GzmA antiserum (MA2A) (16), rabbit anti-mouse GzmB antiserum (526B) (22), rabbit anti-mouse GzmC antiserum (428A) (23), and goat anti-mouse β-actin-HRP antiserum (Santa Cruz Biotechnology). Primary Abs were detected using secondary goat anti-rabbit HRP or horse anti-goat HRP with standard chemiluminescence procedures (Amersham Biosciences). Recombinant murine GzmA, B, and C were purified as previously described (16, 23, 24). One hundred nanograms of each recombinant granzyme was used to demonstrate antiserum specificity. Twenty-five to 50 μg of MLR or LAK cell protein was used in each lane.
WT C57BL/6J, BALB/c, and 129/SvJ mice were obtained from The Jackson Laboratory. GzmB−/−/+PGK-neo- and GzmB−/−/ΔPGK-neo mice in the 129/SvJ background were generated as described above. The gld/gld and perforin−/− mice in C57BL/6 backgrounds were previously described (25). GzmB−/−/+PGK-neo mice in the C57BL/6J background were generated by backcrossing our previously described mutant mice (18) to C57BL/6J mice for 10 generations. All mice were bred and kept in pathogen-free housing in accordance with Washington University School of Medicine animal care guidelines, using protocols approved by the animal studies committee.
Intracellular granzyme staining and flow cytometry
Unstimulated splenocytes or MLR cells were stained for surface markers and intracellular GzmB expression using a GzmB-specific Ab as previously described (26). Briefly, 1 × 106 cells were first labeled with fluorochrome-conjugated Abs against cell surface markers (anti-mouse CD4, CD8, and DX5; BD Pharmingen). Samples were fixed and permeabilized (Cytofix/Cytoperm; BD Pharmingen), then stained with primary conjugated anti-GzmB Ab (GB12; Caltag Laboratories) diluted at 1/400 in staining buffer. During all steps of staining, permeabilization, and washing, anti-murine FcγRII/III-blocking Ab (0.1 μg/μl, final concentration; BD Pharmingen) and human albumin (1%; Aventis Behring) were used to block nonspecific FcR Ab binding. Samples were analyzed on a FACScan (BD Biosciences). All FACScans depicted are representative of four or more independent experiments.
Flow-based killing assay (FloKA) and [125I]5-iodo-2[prime]-deoxyuridine ([125I]UdR) release assay
CTL and LAK cells were assessed by FloKA as previously described (26). Briefly, allogeneic (P815 and TA3-H-2Kd; American Type Culture Collection) or MHC class I-deficient (RMAS and YAC-1; American Type Culture Collection) murine target cells were washed with PBS and labeled with 125 nM (final concentration) CFSE (Molecular Probes). Labeling reactions were stopped with complete K10 medium. Labeled cells were added to 96-well, V-bottom, tissue culture-treated plates (Corning Glass) along with the indicated effector cells (CTLs from MLRs, or LAK cells; see above) in complete medium containing 50 U/ml rhIL-2. E:T cell ratios varied from 5:1 to 20:1 in different experiments. Immediately before analysis, 1 μg/ml (final concentration) of 7-aminoactinomycin D (7-AAD; Calbiochem) was added to each sample. Control samples (targets only) had effectors added immediately before analysis at the indicated E:T cell ratio. [125I]UdR release assays were performed as previously described (18, 27). All experiments examining WT vs Gzm-deficient mice were performed in parallel. All samples were analyzed in duplicate. Data presented in graphic form are the combination of three or more individual experiments. The data shown in individual FACS scans are representative of three or more individual experiments. Statistical analyses were performed by one-way-ANOVA with Bonferroni post-test analysis or Student’s t test, using PRISM version 3.0a for Macintosh (GraphPad).
Allogeneic tumor cell clearance in vivo
All mice were sublethally irradiated (400 cGy) on day −1 and subsequently injected i.v. in the lateral tail vein with 1 × 107 P815 mastocytoma cells on day 0. WT, GzmB−/−/+PGK-neo, perforin−/−, and gld/gld mice (all on the C57BL/6J background) were monitored over a 60-day period. WT, GzmB−/−/ΔPGK-neo, and GzmB−/−/+PGK-neo mice (on the 129/SvJ background) were similarly injected and monitored.
Creation of mice with targeted mutations in the GzmB gene
We generated GzmB−/−/+PGK-neo mice using the targeting vector shown in Fig. 1,A (top panel). The targeted GzmB allele contained a 350-bp deletion that starts in the 5′ untranslated region of the gene, and removes all the coding sequences of exon 1, and part of intron 1 (18). Targeted ES cells containing the mutant GzmB+/−/+PGK-neo allele contained the PGK-neo selection cassette flanked by two LoxP sites. We generated GzmB+/−/ΔPGK-neo ES cells by transiently transfecting the GzmB+/−/+PGK-neo ES clones with pTurbo-Cre, which resulted in recombination of the two LoxP sites and deletion of the internal PGK-neo selection cassette (Fig. 1,A, top panel). Using Southern blot analysis with internal and external probes along with new restriction sites introduced by the targeting vector, we demonstrated correct targeting of the vector in two independent Rw4 (129/SvJ) ES clones; precise deletion of the PGK-neo cassette in the targeted clones was likewise demonstrated with Southern blot analysis. The correctly targeted ES clones were injected into C57BL/6 blastocysts, which were then implanted into pseudopregnant female recipients. All ES clones gave rise to chimeric male mice that were then directly mated to 129/SvJ mice, and the mutations were transmitted through the germline. The expected EcoRI fragments were detected in the tail DNA of mice heterozygous for the mutations (4.4 kb GzmB+/−/+PGK-neo, 2.8 kb GzmB+/−/ΔPGK-neo, and 4.2 kb WT) with Southern blot analysis using an internal probe (Fig. 1, A and B) and also an external probe (data not shown). Heterozygous matings yielded offspring with the expected Mendelian frequencies. Homozygous GzmB−/−/+PGK-neo and GzmB−/−/ΔPGK-neo mice were found to have normal development and fertility.
Neighborhood effect on downstream granzyme gene expression caused by the retained PGK-neo cassette
We measured the expression of multiple granzyme genes in cells derived from MLRs and LAK preparations from both GzmB−/−/+PGK-neo and GzmB−/−/ΔPGK-neo mice using qRT-PCR with primer pairs specific for each granzyme gene (Fig. 2). All primer pairs were shown to amplify only the cloned self target cDNA in qRT-PCRs (data not shown). However, annealing efficiencies varied among the primer pairs, so comparisons of the absolute abundances of different granzyme mRNAs are not valid. We therefore normalized the abundance of each granzyme mRNA to the level detected in WT LAK cells. In all experiments activations were performed concurrently with all three genotypes. Neither GzmB mutant strain had detectable GzmB mRNA expression in LAK cells or CTLs (Fig. 2). There was no significant difference in the expression levels of GzmA in any of the mutant mice, because GzmA is located on a different chromosome than GzmB (17). The qRT-PCR analysis of GzmC and F mRNAs in LAK cells derived from GzmB−/−/+PGK-neo mice demonstrated that expression was reduced 5- to 6-fold compared with that in WT LAK cells, but was normal in LAK cells derived from GzmB−/−/ΔPGK-neo mice (Fig. 2). GzmC and F expression in MLR-derived cells from GzmB−/−/+PGK-neo mice was reduced compared with that in WT cells, but expression levels were very low, and the difference was not statistically significant. GzmC and F expression in GzmB−/−/ΔPGK-neo CTLs was reproducibly increased compared with that in WT animals (Fig. 2). mRNAs for GzmD, E, and G were not detectably expressed in MLR-derived CTLs, but all were detected in LAK cells and were not altered in GzmB−/−/+PGK-neo derived LAKs.
Examination of granzyme protein levels from MLR and LAK cell lysates yielded a similar expression profile (Fig. 3). MLR and LAK cell lysates from both GzmB−/−/+PGK-neo and GzmB−/−/ΔPGK-neo mice lacked a detectable GzmB signal using Western blot analysis (Fig. 3, left and middle panels). LAK cells from GzmB−/−/+PGK-neo mice had an ∼80% reduction in GzmC protein expression compared with WT LAKs (Fig. 3, middle panels). In contrast, LAK cell lysates from GzmB−/−/ΔPGK-neo mice had restoration of GzmC protein expression to WT levels. Analysis of MLR-derived protein lysates from GzmB−/−/ΔPGK-neo mice demonstrated GzmC protein expression levels that were elevated over the GzmC protein levels found in wild-type CTLs (Fig. 3, left panels), similar to those observed with mRNA analysis. GzmA protein expression was equivalent in LAK cell lysates from WT and the two GzmB knockout strains (Fig. 3, middle panels); GzmA protein could not be detected in day 4 MLR-derived lysates, probably because this gene is activated later than GzmB and C (28). Recombinant murine granzymes were used for each Western blot as a control for antiserum specificity (Fig. 3, right panels), and each blot was analyzed for β-actin protein levels as an internal control for protein loading. Specific Abs for GzmD, E, F, and G are not yet available.
Normal lymphocyte subsets in GzmB-deficient mice
To ensure that the observed differences in granzyme mRNA and protein levels between GzmB−/−/+PGK-neo and GzmB−/−/ΔPGK-neo mice were not due to skewed granzyme-expressing populations, we examined these mice for the main lymphocyte populations known to express granzymes, namely CD4+ and CD8+ T cells and NK cells (DX5-positive cells). As shown in Fig. 4,A, no differences were found in the percentages of these lymphocyte subpopulations between WT mice and the two GzmB-deficient mouse strains. Resting WT splenocytes do not contain detectable amounts of GzmB (Fig. 4 B), in contrast to resting human PBMCs, where 15–25% of resting CD8+ T cells contain detectable amounts of this enzyme (26).
We next examined the intracellular GzmB expression profile from WT MLR-derived cells using flow cytometry and an intracellular staining assay for GzmB (26). These studies demonstrated that the majority of GzmB-expressing cells in MLRs are CD8+ T cells (Fig. 4,C); however, a small percentage of CD4+ T cells also express GzmB, a finding consistent with that of human PBMC (26). As expected, no GzmB intracellular expression was detected in either CD4+ or CD8+ T cells in MLRs from GzmB−/−/+PGK-neo and GzmB−/−/ΔPGK-neo mice (Fig. 4,C). The polyclonal Ab against GzmC used in the Western blots in Fig. 3 was not useful in flow-based studies; mAbs specific for murine GzmA or C are not yet available. A small amount of nonspecific staining (presumably representing activated macrophages) was observed in MLR-derived cells from all mouse strains (Fig. 4 C), which was not prevented by FcγRII-III-blocking Abs.
Defects in cytotoxicity in vitro
To determine whether the two GzmB knockout strains have defects in cytotoxicity, we first examined CTLs generated from MLR for their ability to kill allogeneic P815 or TA3 tumor cells (both from an H-2Kd background) using a flow-based killing assay (FloKA). We found that GzmB−/−/+PGK-neo CTLs were significantly less efficient at inducing target cell death than WT CTLs at all time points tested (representative data are shown in Fig. 5,A, R1 and R2 gates). The same cytotoxic defects were detected with both P815 and TA3 targets, and the findings were consistent and highly reproducible. The most striking defect was the inability of these effectors to generate the small, 7-AADlow+high cells that are located in the R1 gate, which probably represent cells in the late stages of apoptosis. CTLs derived from both Gzm-deficient strains induced an increased number of target cells in the R2 gate (large cells that are 7-AADhigh, probably representing cells with early apoptotic events), suggesting that the induction of target cell death is delayed with GzmB-deficient effectors, a finding that corroborates earlier studies of CTL function with these mice (18). GzmB−/−/ΔPGK-neo generated CTLs demonstrated a less severe cytotoxic defect than GzmB−/−/+PGK-neo CTLs (Figs. 5,A and 6,A, R1 gates). A similar defect in the production of late stage apoptotic events was seen with LAK cells generated from both Gzm-deficient strains against two different NK-sensitive target cell lines (RMAS and YAC-1 cells; Fig. 6,B). The most severe cytotoxic defect was observed with GzmB−/−/+PGK-neo LAK cells, with GzmB−/−/ΔPGK-neo LAKs again demonstrating an intermediate defect (Fig. 6 B). Although the severity of the cytotoxic phenotype was partly dependent upon the target cells used, the trends were the same for all target cells tested.
When we examined GzmB−/−/+PGK-neo and GzmB−/−/ΔPGK-neo LAKs for their ability to release dsDNA fragments using classic [125I]UdR release assays, we detected no difference between the two Gzm-deficient strains regardless of time or E:T cell ratio (Fig. 5,B). CTLs from both Gzm-deficient strains displayed a severe defect in [125I]UdR clearance at 2 h, but this defect was not apparent after 8 h (Fig. 5 B), similar to results previously described (18). No abnormalities were detected in 51Cr release with GzmB-deficient LAK cells, as expected (data not shown).
Defects in tumor clearance in vivo
To determine whether the cytotoxic defects detected in vitro were relevant in vivo, we examined the clearance of P815 cells in sublethally irradiated (400 cGy) C57BL/6 mice deficient for perforin, GzmB−/−/+PGK-neo, or FasL (gld). The GzmB−/−/+PGK-neo and perforin-deficient mice were equivalently impaired in their ability to survive a P815 challenge, compared with WT mice (Fig. 7,A). In contrast, gld/gld mice were able to clear P815 cells as efficiently as WT mice (Fig. 7 A).
To examine the role of orphan granzymes in allogeneic tumor clearance, we compared the clearance of P815 cells in sublethally irradiated (400 cGy) 129/SvJ mice (WT vs GzmB−/−/+PGK-neo vs GzmB−/−/ΔPGK-neo). GzmB−/−/+PGK-neo mice were significantly impaired in their ability to clear P815 cells (Fig. 7 B). In contrast, GzmB−/−/ΔPGK-neo mice cleared P815 cells as efficiently as WT 129/SvJ mice.
In this report we describe two strains of mice that are deficient for GzmB. In one, a retained PGK-neo cassette in the GzmB gene causes knockdown of the most closely linked genes, GzmC and F. Removal of the PGK-neo cassette from the GzmB gene by Cre-mediated recombination restored the expression of GzmC and F to normal levels in LAK cells; however, in MLR-derived CTLs, GzmC and F were expressed at higher than normal levels. The cytotoxic potential of MLR-generated CTLs from both GzmB-deficient strains was reduced, but the defect was more severe in mice with the retained PGK-neo cassette, implying that GzmC and/or F may be relevant for in vitro cytotoxicity. Finally, these results were corroborated in vivo, where the allogeneic clearance of P815 tumor cells was significantly impaired in GzmB−/−/+PGK-neo mice, but was normal in mice deficient for GzmB only. These data suggest that the orphan GzmC and/or F are relevant for cytotoxic lymphocyte functions in vitro and in vivo.
Retained PGK-neo cassettes in a targeted locus can cause the expression of tightly linked genes to be reduced (19, 20, 29, 30, 31). Although many examples of this neighborhood effect have been reported, the mechanism and rules that govern the effect are not yet clear. The interposition of an active selectable marker cassette between a locus control region (LCR) and its functional targets can significantly reduce the expression of those linked genes over tens of kilobases. This effect can often be eliminated by removing the selectable marker cassette, using LoxP/Cre (or Frt/Flp) recombination technology (20, 29, 30, 31). Indeed, removal of the PGK-neo cassette from the GzmB gene restored expression of GzmC and F, which are located 24.7 and 44.8 kb downstream from GzmB, respectively. The expression of the granzyme genes located further downstream (G, D, and E) was not affected by the retained PGK-neo cassette, which suggests that the effect is limited to the 5′ end of this multigene cluster. We previously examined the effect of a retained PGK-neo cassette in the murine cathepsin G gene, which lies just downstream from the GzmE gene (32); this retained cassette knocked down the expression of the mast cell chymase gene just downstream from it, but did not affect the expression of the granzymes found upstream, suggesting a unidirectional component of the neighborhood effect.
In our previously described mouse model of GzmB deficiency (18), the GzmB gene was targeted with a standard PGK-neo cassette, and expression studies were performed in mice on a mixed strain background (129/SvOla×C57BL/6). In that mouse model, the expressions of GzmC, F, and D were all significantly reduced in LAK cells (19). In the current study both targeted mutations were made in a 129/SvJ ES line (Rw4). Chimeric males made from these ES cells were bred to 129/SvJ females, so that the mutations were immediately anchored in a pure 129/SvJ background. In the current model we detected a striking reduction in GzmC and F expression in LAK cells, but detected no reduction in GzmD expression. This may reflect strain-specific differences in the mouse models, or it could explain differences in the measurement technology. In the previous study experiments were performed using S1 nuclease protection assays (19, 32); in the current study, quantitative RT-PCR studies were performed, using primers absolutely specific for each granzyme mRNA. Regardless, it is clear that the retained PGK-neo cassette caused a knockdown of GzmC and F expression in both mutant mouse strains, and that this expression was restored upon removal of the cassette.
We do not yet understand why expression of the GzmC and F genes is higher than normal in MLR-derived CTLs from GzmB−/−/ΔPGK-neo mice. The targeted mutation of the GzmB gene creates a deletion between two naturally occurring AvrII sites. The 5′ end of the deletion is in the very short 5′-untranslated region of GzmB. The deletion extends 350 bp downstream and includes all coding sequences within the first exon and part of the first intron of the gene. The transcription start site of the gene is therefore removed. It is possible that removal of the transcriptional start site disrupts a functional interaction between a putative LCR and the 5′ end of the GzmB gene, allowing the LCR to scan downstream for the next available promoter (i.e., GzmC). Alternatively, it is possible that high level transcription of the GzmB gene normally causes interference with the transcription of GzmC and F; this transcriptional interference would be relieved by the start site deletion. However, the expression levels of GzmC and F are normal in LAK cells from GzmB−/−/ΔPGK-neo mice, making it less likely that these mechanisms are involved. Alternatively, it is possible that a specific subpopulation of cells within an MLR preferentially expresses GzmC and F in this setting; we cannot address this issue experimentally as yet, because mAbs that permit flow-based detection of these granzymes are not yet available.
The generation of these two GzmB-deficient mouse strains has allowed us to assess the importance of granzyme genes downstream from GzmB in cytotoxic killing assays in vitro and in vivo. Traditional [125I]UdR and 51Cr release assays did not reveal significant differences between the cytotoxic profiles of LAK cells with and without the PGK-neo cassette. However, FloKA revealed that the CTLs derived from GzmB−/−/+PGK-neo mice had a more significant killing defect than those from GzmB−/−/ΔPGK-neo mice. Previous work by Lecoeur et al. (33, 34) revealed that target cells with the FSClow/7-AADhigh phenotype represent late apoptotic target cells (as defined by increased activation of caspases-3 and -8, as well as increased mitochondrial depolarization, DNA fragmentation, and annexin V staining). In contrast, FSChigh/7-AADhigh target cells (R2 gate) were shown to represent cells in the early stages of apoptosis (defined by lower levels of mitochondrial depolarization, DNA fragmentation, and annexin V staining) (33, 34). In this study we noted a significant decrease in the percentage of late apoptotic target cells (e.g., FSClow/7-AADlow+high; R1 gate) induced by MLR-generated CTLs and LAK cells from both GzmB-deficient strains, but the defect was more severe with CTLs from the GzmB−/−/+PGK-neo mice. CTLs from both strains caused excess target cells to appear in the early apoptotic gate (R2), suggesting that the full induction of target cell death requires additional time if GzmB is missing, consistent with previous results (18). Alternatively, the differences in the target cell FLoKA patterns between the two strains may represent specific, biochemically unique apoptotic events induced by GzmC and/or F. Indeed, data from our laboratory have demonstrated that GzmC can directly induce mitochondrial depolarization and nuclear collapse in a noncaspase-dependent manner (23). Regardless, these findings suggest that the FloKA assay can detect subtle changes in target cells undergoing cell death that are not detected by traditional cytotoxicity assays (that rely on the release of radiolabeled proteins and/or DNA from target cells). The strong correlation between the cytotoxic defects detected by this method and the tumor clearance defect detected in vivo suggests that these changes may indeed be physiologically relevant.
We detected a strong defect in the clearance of P815 tumor cells in perforin-deficient mice and in mice with our original GzmB−/−/+PGK-neo mutation on the C57BL/6 background. These data strongly suggest that GzmB and/or the orphan granzymes affected by this mutation (i.e., GzmC, F, and D) are relevant for the perforin-dependent clearance of this tumor cell line in vivo. These data are similar to those reported by Pardo et al. (4), who showed that GzmA and B cluster enzymes are important for perforin-dependent clearance of RMAS cells in vivo. However, our data are different from those reported by Smyth, Trapani, and colleagues (5, 6), who detected perforin-dependent, but granzyme-independent, clearance of several tumor cell lines in vivo. The reasons for these discrepancies are not yet clear. Differences in laboratory stocks of tumor cell lines could be relevant, and subtle differences in the strain backgrounds of the mutant mice could also be important, because GzmB−/−/+PGK-neo mice on the C57BL/6J background are clearly more susceptible to death than the same mice on the 129/SvJ background (see Fig. 7). Clearly, additional studies will be required to understand the causes of these laboratory-to-laboratory differences.
In addition to the defect in allogeneic tumor cell clearance in GzmB−/−/+PGK-neo mice, we have detected a defect in the control of gammaherpesvirus (γHV68) latency and reactivation in these mice (35); this defect is more severe than that in mice deficient for GzmB only. These data support the observations reported in this study and implicate GzmC and/or F in the control of gammaherpesviruses in vivo.
The neighborhood effect in LAK cells derived from GzmB−/−/+PGK-neo mice predominantly affects GzmC and F. Recent studies in our laboratory have demonstrated that GzmC can induce target cell death that has many of the features of apoptosis, including nuclear collapse, annexin V positivity, and mitochondrial depolarization. However, GzmC does not cause nuclear fragmentation or double-stranded intranucleosomal DNA cleavage, and it does not cleave any of the recognized substrates of GzmB (23). The ability of GzmC to cause cell death is related to its protease activity, because a mutation in its active site serine significantly reduces its ability to kill target cells in vitro; however, the specificity of this protease and its relevant substrates are currently unknown. Because GzmC can induce target cell death, its reduced expression in GzmB−/−/+PGK-neo-derived CTLs may well explain the reduced cytotoxicity of these cells, at least in part. The ability of GzmF to induce cell death has not yet been examined; likewise, its substrates are not yet known. Regardless, it is important to note that the overexpression of GzmC and F in MLR-derived CTLs does not rescue the cytotoxicity defect caused by the loss of GzmB, suggesting that the enzymes are not functionally redundant. Additional studies and knockout mice will be required to dissect the relative contributions of GzmC and F to the phenotypes observed in this report.
In summary, these studies have shown that a retained PGK-neo cassette in the GzmB gene causes a knockdown of the expression levels of GzmC and F just downstream from the GzmB gene. CTLs derived from these mice have a more severe cytotoxic defect than those from mice deficient for GzmB only. These data clearly implicate GzmC and F in CTL functions and suggest the need for additional studies of their substrate specificities and killing activities. Furthermore, because GzmC is the probable orthologue of human GzmH (17), it will be important to learn whether human GzmH is important for the ability of human cytotoxic lymphocytes to kill their targets.
We thank the Siteman Cancer Center High Speed Cell Sorter Core and the Embryonic Stem Cell Core for technical support, and Kelly Schrimpf and Mieke Hoock for excellent mouse husbandry. Nancy Reidelberger provided expert editorial assistance.
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.
This work was supported by grants from the National Institutes of Health (5T32AI07163 (to P.A.R.) and DK49786 (to T.J.L.)) and the Hope Street Kids Foundation (to W.J.G.).
Abbreviations used in this paper: Gzm, granzyme; 7-AAD, 7-aminoactinomycin; ES, embryonic stem cell; FloKA, flow-based killing assay; LAK, lymphokine-activated killer; PGK-neo, phosphoglycerate kinase I gene promoter; qRT-PCR, quantitative RT-PCR; rh, recombinant human; [125I]UdR, [125I]5-iodo-2[prime]-deoxyuridine; WT, wild type; LCR, locus control region.