Acquisition of Murine NK Cell Cytotoxicity Requires the Translation of a Pre-existing Pool of Granzyme B and Perforin mRNAs

Immunity

Article Acquisition of Murine NK Cell Cytotoxicity Requires the Translation of a Pre-existing Pool of Granzyme B and Perforin mRNAs Todd A. Fehniger,1 Sheng F. Cai,1 Xuefang Cao,1 Andrew J. Bredemeyer,1 Rachel M. Presti,2 Anthony R. French,3,* and Timothy J. Ley1,* 1

Division of Oncology, Department of Internal Medicine Division of Infectious Disease, Department of Internal Medicine 3 Division of Rheumatology, Department of Pediatrics Siteman Cancer Center, Washington University School of Medicine, St. Louis, MO 63110, USA *Correspondence: [email protected] (A.R.F.), [email protected] (T.J.L.) DOI 10.1016/j.immuni.2007.04.010 2

SUMMARY

Although activated murine NK cells can use the granule exocytosis pathway to kill target cells immediately upon recognition, resting murine NK cells are minimally cytotoxic for unknown reasons. Here, we showed that resting NK cells contained abundant granzyme A, but little granzyme B or perforin; in contrast, the mRNAs for all three genes were abundant. Cytokineinduced in vitro activation of NK cells resulted in potent cytotoxicity associated with a dramatic increase in granzyme B and perforin, but only minimal changes in mRNA abundance for these genes. The same pattern of regulation was found in vivo with murine cytomegalovirus infection as a physiologic model of NK cell activation. These data suggest that resting murine NK cells are minimally cytotoxic because of a block in perforin and granzyme B mRNA translation that is released by NK cell activation. INTRODUCTION Natural-killer (NK) cells are innate immune lymphocytes that are important for early host defense against infectious pathogens and tumors through two primary mechanisms: contact-dependent cytotoxicity and cytokine production (Biron and Brossay, 2001; Yokoyama et al., 2004). NK cells were first identified by their ability to kill tumor target cells without prior sensitization via a specific antigen receptor, a trait that distinguishes NK cells from cytotoxic T cells (Herberman et al., 1975; Kiessling et al., 1975). NK cell responses to potential target cells are determined by the integration of multiple triggering signals from families of cell-surface-activating and -inhibitory NK receptors (Lanier, 2005; Yokoyama and Kim, 2006). Although NK cells can induce target-cell death through several mechanisms, the granule-exocytosis pathway is the primary one used by NK cells to kill tumor and virus-infected cells 798 Immunity 26, 798–811, June 2007 ª2007 Elsevier Inc.

(Grossman et al., 2003; Henkart, 1985; Lieberman, 2003; Russell and Ley, 2002). Therefore, to kill a target cell, an NK cell must be triggered by engagement of NK cell receptors, and it must contain functional cytotoxic granules. Cytotoxic lymphocyte granules contain many component proteins, including perforin (Prf1) and granzymes (Gzms); Prf1 facilitates the entry and/or trafficking of Gzms into target cells, where these proteases induce cell death (Russell and Ley, 2002). Of the Gzms, A and B are the best characterized, and they have been shown to initiate target-cell death by cleaving nonoverlapping sets of substrates (Lieberman, 2003; Russell and Ley, 2002). Much of our understanding of murine NK cell Prf1 and Gzm expression is derived from experiments with human NK cells, cell lines, cytotoxic T cells, and lymphokine-activated-killer (LAK) cells (Babichuk et al., 1996; Ebnet et al., 1995b; Garcia-Sanz et al., 1990; Grossman et al., 2003; Russell and Ley, 2002; Smyth et al., 1995). Studies utilizing CD8+ cytotoxic T lymphocytes (CTL) have shown that the expression of Prf1 and Gzmb are controlled primarily at the level of transcription in effector T cells—a finding that has been generalized to all cytotoxic lymphocytes (Babichuk and Bleackley, 1997; Babichuk et al., 1996; GarciaSanz et al., 1990; Hanson and Ley, 1990; Heusel et al., 1994; Shresta et al., 1999). Studies that used murine LAK cells (generated by long-term culture of splenocytes in high doses of IL-2) as ‘‘NK cell surrogates’’ have shown that these cells contain abundant mRNA and protein for Gzma, Gzmb, and Prf1 (Pham et al., 1996; Revell et al., 2005; Shresta et al., 1995, 1997, 1999). The current view of murine NK cells as being ‘‘armed’’ with Prf1 and Gzm protein was derived from these prior studies; however, it is not clear that human NK cells, LAK cells, cell lines, or T cells reflect the true biology of murine NK cells. Paradoxically, it has long been recognized that resting murine NK cells are minimally cytotoxic against tumor targets in vitro, and it is unclear why activation is required to induce cytotoxicity. Here we show that most resting murine NK cells contain Gzma, but little or no Prf1 or Gzmb protein, yet they express abundant mRNAs for all three genes. Whereas resting NK cells are minimally cytotoxic against tumor target cells in vitro, cytokine-induced in vitro activation

Immunity Perforin and Granzyme A/B in Murine NK Cells

Figure 1. Resting Murine NK Cells Express Abundant Gzma but Little Gzmb or Prf1 Protein Resting splenocytes from B6 mice (A), freshly isolated mononuclear cells from various B6 mouse tissues (B), or resting splenocytes from the indicated strains of mice (C) were gated on NK1.1+CD3 or DX5+CD3 NK cells and analyzed for expression of intracellular Gzma, Gzmb, or Prf1. Individual flow-cytometry plots are shown and are representative of at least 30 (A), 5 (B), or 3 (C) independent experiments (see Results for summary statistics). Numbers indicate the percentage of cells in each quadrant.

of NK cells results in potent killing associated with a dramatic increase in Gzmb and Prf1 protein, with minimal changes in mRNA abundance. Furthermore, the expression of Gzmb and Prf1 protein in NK cells is induced after murine-cytomegalovirus (MCMV) infection in vivo, and both Gzmb/ and Prf1/ B6 mice are more susceptible to MCMV than are wild-type (WT) mice. These data suggest that resting murine NK cells are minimally cytotoxic because of a block in Prf1 and Gzmb mRNA translation that is released by NK cell activation.

RESULTS Resting Murine NK Cells Contain Abundant Gzma but Minimal Gzmb or Prf1 Protein Although it is widely thought that all NK cells constitutively express Prf1 and Gzms in cytotoxic granules, murine NK cells are minimally cytotoxic against tumor targets in vitro for unclear reasons. We therefore analyzed resting murine NK cells for their Gzma, Gzmb, and Prf1 protein expression by using intracellular flow cytometry (Figure 1). For Immunity 26, 798–811, June 2007 ª2007 Elsevier Inc. 799

Immunity Perforin and Granzyme A/B in Murine NK Cells

Figure 2. Cytokine-Activated Murine NK Cells Express Abundant Gzmb and Prf1 Protein (A) Cytokine-activated NK cells express Gzmb and Prf1 protein by flow cytometry. Splenocytes were cultured with no cytokines (media) or the following cytokines for 48 hr: rhIL-2 (1000 IU/mL), rmIL-12 (100 ng/mL), rmIL-15 (100 ng/mL), rmIL-18 (100 ng/mL), or rmIFN-a (100 U/mL), and flow-cytometry plots of NK1.1+CD3 NK cell expression of Gzma, Gzmb, and Prf1 are shown. Results from an individual experiment are shown and are representative of at least three independent experiments. Numbers indicate the percentage of cells in each quadrant.

800 Immunity 26, 798–811, June 2007 ª2007 Elsevier Inc.

Immunity Perforin and Granzyme A/B in Murine NK Cells

all flow-cytometry experiments, bulk splenocytes were first gated on NK1.1+CD3 or DX5+CD3 NK cells, and the expression of Gzma, Gzmb, and Prf1 was examined with monoclonal antibodies (mAbs) that have proven specificity in the appropriate mutant mice (Figure S1 in the Supplemental Data available online; Grossman et al., 2004b). Surprisingly, the majority of resting splenic NK cells from C57Bl/6 (B6) mice expressed abundant Gzma (86.4% ± 2.1% positive, n = 22); however, there was little or no expression of Gzmb (3.3% ± 4.6% positive, n = 22) or Prf1 (1.9% ± 2.2% positive, n = 14) (Figures 1A, 1B, and 2). NK cells in the blood, liver, and bone marrow demonstrated a similar Gzma+GzmbPrf1 phenotype (Figure 1B). Fewer bone-marrow NK1.1+CD3 NK cells expressed Gzma, compared to cells derived from the spleen, liver, or blood (Figure 1B). A similar resting phenotype was observed in 129/SvJ, BALB/c, CBA, and A/J mouse strains (Figure 1C). A/J mice had fewer Gzma+ NK cells at rest, with a smaller amount of Gzma per positive NK cell. Murine CD8+ and CD4+ T cells in resting spleens do not express Gzma, Gzmb, or Prf1, as expected (data not shown). Prf1 and Gzm protein expression in resting murine NK cells differs from that of human peripheral blood CD56+CD3 NK cells, which constitutively express Gzmb and Prf1 protein (Figure S2; Bratke et al., 2005; Grossman et al., 2004b). Thus, murine NK cells unexpectedly contain abundant Gzma but minimal Gzmb or Prf1 protein. Activated Murine NK Cells Express Gzmb and Prf1 Protein and Contain Azurophilic Granules Cytokines are well-characterized stimulators of NK cell cytotoxicity in vitro. We therefore examined the ability of IL-2, IL-12, IL-15, IL-18, and IFN-a to alter Gzma, Gzmb, and Prf1 protein expression in murine NK cells. Splenocytes were cultured in the indicated cytokines for 48 hr, and NK1.1+CD3 NK cells were analyzed (Figure 2A). Whereas IL-2, IL-12, IL-18, and IFN-a all increased the percentage of NK cells expressing Gzmb and Prf1 after 48 hr, IL-15 consistently induced the highest amount. Stimulation of flow-sorted (R99% NK1.1+CD3) NK cells with these cytokines for 48 hr revealed that IL-2 and IL-15 induced similar amounts of Gzmb and Prf1, whereas IL12, IL-18, and IFN-a induced less expression of these proteins (Figure S3). We therefore performed an in-depth analysis of Gzma, Gzmb, and Prf1 protein expression by using IL-15 to activate NK cells (Figures 2B–2D). The percentage of Gzma+ NK cells did not change substantially with 48 hr of IL-15 activation. IL-15 stimulated a timeand dose-dependent increase in the percentage of cells expressing Gzmb and Prf1 protein (Figures 2B–2D). The

percentage of Gzmb+ NK cells increased after 6 hr and continued to increase at 24 hr (86.4% ± 9.1%, p < 0.0001) and 48 hr (86% ± 11.8%, p < 0.0001). Prf1 protein was increased in a small percentage of NK cells at 6–12 hr, with further increases at 24 hr (30.4% ± 13.3%, p < 0.01) and 48 hr (45.2% ± 19.6%, p < 0.001; Figures 2B and 2C). IL-15 increased Gzmb and Prf1 abundance at concentrations of 5 ng/mL; however, maximal induction was observed with R50 ng/mL (Figure 2D). Increased NK cell Gzmb and Prf1 was observed after IL-15 stimulation in all strains tested (129/SvJ, BALB/c, CBA, and A/J; Figure S4). To extend these findings, we used flow sorting to isolate highly purified NK cell populations (R99% NK1.1+CD3; data not shown) from the spleens of B6 Rag1/ mice. We compared purified NK cells and bulk WT B6 splenocytes for Prf1 and Gzmb protein expression at rest, and after stimulation with IL-15 for 24 hr or 48 hr, by using immunoblot analysis (Figure 3A). These experiments utilized an independent Prf1 mAb and a different polyclonal Gzmb antibody, both of which have proven specificities in their respective mutant mice (Revell et al., 2005; Shresta et al., 1999). Purified resting NK cells contained minimal Gzmb and Prf1 protein according to this assay, with increased abundance after 24 and 48 hr of stimulation with IL-15. Consistent with the immunoblot data, the percentage of Prf1-containing cells (and the Prf1 MFI) measured by flow cytometry increased only modestly at 24 hr, with a more dramatic increase apparent in both assays at 48 hr. No signal was detected for Gzmb and Prf1 by immunoblot in resting bulk splenocytes, and minimal signals were detected after 48 hr of IL-15 stimulation. Because NK cells comprise only 3% of total splenocytes, the low signal intensities detected suggest that these proteins are induced only in the NK population, consistent with the flow-cytometry data. Microscopic examination of flow-sorted resting NK cells revealed small lymphocytes with little cytoplasm and few cytoplasmic granules (Figure 3B). In contrast, IL-15-activated NK cells were large and had open nuclear chromatin, abundant cytoplasm, a prominent Golgi, and abundant azurophilic (cytotoxic) granules. Thus, resting murine NK cells require short-term activation (e.g., by IL15 or IL-2) to become ‘‘armed’’ with Gzmb and Pfr1 protein and to develop large granular lymphocyte morphology. Gzma, Gzmb, and Prf1 mRNA Change Minimally with NK Cell Activation We next examined the expression of Gzma, Gzmb, and Prf1 mRNAs in flow-sorted (R99% NK1.1+CD3) resting

(B and C) IL-15-treated NK cells express Gzmb and Prf1 protein in a time-dependent fashion. Splenocytes were stimulated with 100 ng/mL IL-15 for 0 (resting), 6, 12, 18, 24, or 48 hr, and NK1.1+CD3 NK cells were analyzed for expression of intracellular Gzma, Gzmb, and Prf1. Representative flowcytometry plots from individual splenocyte cultures (B), and summary scatter plots of the percent positive NK1.1+CD3 NK cells for Gzma, Gzmb, and Prf1 from at least six independent experiments (C) are shown, with each data point representing an individual mouse. Horizontal bars represent mean percent positive NK1.1+CD3 cells. *p < 0.01; **p < 0.001; ***p < 0.0001. (D) IL-15-activated NK cells express Gzmb and Prf1 in a dose-dependent fashion. Splenocytes were stimulated with the indicated concentrations of rmIL-15 for 48 hr, and NK1.1+CD3 NK cells were analyzed by flow cytometry for intracellular Gzma, Gzmb, and Prf1. Results shown are representative of at least three independent experiments.

Immunity 26, 798–811, June 2007 ª2007 Elsevier Inc. 801

Immunity Perforin and Granzyme A/B in Murine NK Cells

Figure 3. IL-15 Induces Gzmb and Prf1 Protein Expression and Granule Formation in Murine NK Cells (A) Immunoblots of resting and activated NK cells or bulk splenocytes for Prf1 (66 kDa) and Gzmb (30 kDa). Protein lysates were made from bulk splenocytes or flow-sorted NK cells at rest (fresh) or after stimulation with IL-15 for 24 hr or 48 hr. The murine NK cell line 36.2 (which constitutively expresses Prf1 and Gzmb) is shown as a positive control (+ control), and actin is used as a loading control. NK, NK cells; SP, spleen; NS, nonspecific band. Results shown are representative of two independent experiments. (B) IL-15 induces the formation of azurophilic granules in purified murine NK cells. Lightmicroscopic evaluation (10003) of flow-sorted murine NK cells at rest or after 24 hr stimulation with IL-15 (100 ng/mL). Note the large size, open nuclear chromatin, prominent Golgi, and abundant azurophilic granules in the IL15-stimulated NK cells, compared with resting NK cells. Results shown are representative of three independent experiments.

and IL-15-activated murine NK cells by using Affymetrix 430v2.0 gene-expression microarrays (Figure 4). Gzma was one of the most abundant mRNAs in resting NK cells, and IL-15 stimulation did not markedly alter its mRNA abundance (0.69-fold change). Both Gzmb and Prf1 mRNAs were abundantly expressed in resting NK cells; whereas IL-15 modestly increased Gzmb mRNA abundance (2.5-fold, p < 0.05) at 24 hr, Prf1 mRNA expression was unchanged (0.78-fold change). Gzma, Gzmb, and Prf1 all ranked (mean rank order from three microarrays) in the top 500 expressed genes in resting (Gzma, #3; Gzmb, #411; Prf1, #141) and IL-15-activated (Gzma, #4; Gzmb, #30; Prf1, #182) NK cells (of the 45,037 probe sets represented on the array) (Figure 4C). To confirm these findings, quantitative real-time RT-PCR was performed via RNA from flow-sorted NK cells at rest and after 24 hr IL-15 stimulation (Figure 4D). Consistent with the microarray findings, Gzma and Prf1 mRNA expression was not significantly different between resting and IL-15-activated NK cells. Gzmb mRNA abundance increased modestly after 24 hr of IL-15 stimulation (4.2fold, p < 0.05). Thus, resting murine NK cells are prearmed with abundant Gzmb and Prf1 mRNA, but minimal Gzmb and Prf1 protein. Cytokine Activation Arms Murine NK Cells to Become Potent Cytotoxic Effectors We next evaluated the cytotoxic potential of resting and IL-15-activated splenocytes in a flow-based killing assay (FloKA). Resting splenocytes were used as effectors against the NK-sensitive target-cell lines RMAS and YAC1, with minimal killing observed in a 4 hr FloKA (Figure 5A). At an effector:target (E:T) ratio of 20:1, resting 802 Immunity 26, 798–811, June 2007 ª2007 Elsevier Inc.

splenocytes induced only 5.6% ± 2.8% (mean ± SD, n = 7) 7-AAD+ RMAS target cells, and 7.3% ± 4.1% 7AAD+ YAC1 (n = 3) target cells. In contrast, IL-15-activated splenocytes used at an identical E:T ratio yielded 54.8% ± 11.1% 7-AAD+ RMAS (n = 14, p < 0.0001) and 53.2% ± 13.6% 7-AAD+ YAC1 (n = 7, p < 0.001) cells. The NK cell-insensitive EL4 cell line was minimally killed by resting or IL-15-activated splenic NK cells, as expected. We observed the same killing phenotype for all strains of mice evaluated (Figure 5B). Identical B6 splenocyte effectors and RMAS targets were also compared in parallel Cr51-release assays, which showed similar results (Figures 5C and 5D). The percentages of NK1.1+CD3 or DX5+CD3 NK cells changed only slightly between fresh (2.6% ± 0.4%), 24 hr IL-15 (2.5% ± 0.3%), or 48 hr IL-15 (3.6% ± 0.6%) spleen cultures, and thus had a minimal impact on the E:T ratios used. Flow-sorted NK cells from Rag1/ mice demonstrated minimal cytotoxicity against RMAS targets at rest, which was dramatically increased after stimulation with IL-15 (Figure S5). Because NKG2D is also induced after IL-15 stimulation, and could potentially mediate increased NK cytotoxicity, we performed experiments comparing IL-15-activated splenocytes preincubated with a blocking NKG2D (C7) mAb or a hamster IgG control mAb (Ho et al., 2002). There was no effect of NKG2D blockade on killing, suggesting that activation of NKG2D did not contribute to increased cytotoxicity in this setting (data not shown). Thus, murine NK cells are minimally cytotoxic at rest and require activation (e.g., by IL-15) to acquire the potential for potent cytotoxicity against NK-sensitive target cells in vitro.

Immunity Perforin and Granzyme A/B in Murine NK Cells

Figure 4. Resting Murine NK Cells Express Abundant Gzma, Gzmb, and Prf1 mRNA, but Only Gzmb mRNA Increases with IL-15 Stimulation (A and B) Gzma, Gzmb, and Prf1 mRNA expression detected on microarrays from resting and IL-15-activated NK cells. (A) Flow-sorted NK cells were stimulated with 100 ng/mL of IL-15 for 0 hr (rest) or 24 hr, and total mRNA was isolated, labeled, and hybridized with Affymetrix 430v2.0 expression microarrays. Signal intensities of probe sets for Gzma, Gzmb, Prf1, and b-actin (control for [D]) from three independent NK cell experiments are shown with the mean indicated by horizontal bar. *p < 0.05. (B) Representative scatter plot of probe set values from three resting (x axis) and IL-15-activated (y axis) NK cell pairs, with Gzma, Gzmb, and Prf1 probe set values designated with arrows. (C) Microarray signal intensity (mean ± SD) from three independent resting and IL-15-activated (24 hr) NK cell pairs for Prf1 and murine Gzm probe sets on the Affymetrix 430v2.0 microarray. Also shown is the signal intensity of resting, unmanipulated WT spleen cells (n = 2 spleens, with arrays for both performed in duplicate) for comparison. Values in italics indicate that these signals were Affymetrix ‘‘absent’’ calls (no specific detectable signal present) on the microarrays. (D) Quantitative real-time RT-PCR comparing resting and 24 hr IL-15-activated NK cell expression of Gzma, Gzmb, and Prf1 mRNAs. The mean ± SD expression from at least four independent experiments is shown as fold-change of IL-15-activated (24 hr) compared to resting NK cells, utilizing the comparative Ct method, with b-actin used as the control. *p < 0.05.

IL-15-Activated NK Cells Require Gzmb and Prf1 for Potent Cytotoxicity In Vitro We next examined whether Gzma, Gzmb, or Prf1 were required for NK cell cytotoxicity in vitro (Figures 5E and 5F). Splenocytes from WT mice, Prf1/ mice, and

Gzmb/ mice (all in the B6 background) were used at rest or after 48 hr in IL-15 as effectors in FloKAs against RMAS tumor targets (Figure 5E). In this report, ‘‘Gzmb/’’ refers to mice with a retained Pgk-Neo cassette in the Gzmb gene. These mice are deficient for Gzmb and also Immunity 26, 798–811, June 2007 ª2007 Elsevier Inc. 803

Immunity Perforin and Granzyme A/B in Murine NK Cells

Figure 5. Resting Murine Splenocytes Are Minimally Cytotoxic, and IL-15-Activated Splenocytes Require Gzmb and Prf1 to Efficiently Kill Tumor Targets In Vitro (A) Resting murine splenocytes exhibit minimal cytotoxicity against the NK-sensitive RMAS and YAC1 tumor cell lines that is increased by IL-15 activation. FloKAs (4 hr) were performed with resting or activated (48 hr IL-15) B6 splenocytes as effectors and the indicated tumor cell line as targets, at E:T ratios of 5:1, 10:1, and 20:1. The NK-insensitive EL4 cell line is included as a negative control. For (A), (B), (C), (E), and (F), FloKA results shown are the percent target cells that are 7-AAD positive (mean ± SD of replicate wells) from one experiment. The data (mean ± SD) shown are representative of at least three independent experiments. (B) Multiple strains of mice exhibit minimal resting NK cell cytotoxicity, which is increased by IL-15 activation. Splenocytes from the indicated mouse strains isolated fresh (resting) or after 48 hr of IL-15 activation were used as effector cells in a 4 hr FloKA against RMAS targets at the indicated E:T ratios. Results (mean ± SD) shown are representative of three independent experiments. (C and D) FloKA and Cr51-release assays yield similar results, showing a time-dependent increase in NK cell cytotoxicity after IL-15 stimulation. WT B6 splenocytes were isolated and used fresh (0 hr) or at 12 hr, 24 hr, or 48 hr of IL-15 stimulation as effector cells in both FloKA (C) and Cr51-release (D) assays against identical RMAS targets. Results (mean ± SD) shown are representative of two independent experiments. (E) Splenocytes from WT, Gzmb/, and Prf1/ (B6) were used as effector cells in a 4 hr FloKA at rest or after 48 hr of IL-15 stimulation against RMAS tumor targets at 5:1, 10:1, and 20:1 E:T ratios. Results (mean ± SD) shown are from one experiment representative of at least three independent experiments. (F) Splenocytes from WT, Gzma/, Gzmb/, Gzma/Gzmb/, and Prf1/ (129/SvJ) were used as effector cells in a 4 hr FloKA after 48 hr of IL-15 stimulation against RMAS tumor targets. Results (mean ± SD) are representative of at least three independent experiments. For all experiments, spontaneous target cell death (FloKAs, percent targets positive for 7-AAD in target-only wells; Cr51-release, CPM in target-only wells) was routinely <10% and subtracted to provide specific cytotoxicity data. Summary and comparative statistics are detailed in the Results section.

804 Immunity 26, 798–811, June 2007 ª2007 Elsevier Inc.

Immunity Perforin and Granzyme A/B in Murine NK Cells

display reduced Gzmc and Gzmf expression in LAK cells and are therefore often referred to as Gzmb clusterdeficient mice (Pham et al., 1996; Revell et al., 2005). As previously noted (Figures 5A–5D), resting splenocytes from all genotypes exhibited minimal cytotoxicity. IL-15activated Gzmb/ splenocytes (20:1 E:T, 21.7% ± 5.8% 7-AAD+, n = 7) had a significant defect in their cytotoxic activity compared with WT NK cells (20:1 E:T, 54.2% ± 13.2% 7-AAD+, n = 7, p < 0.001). Similarly, IL-15-activated Prf1/ splenocytes (20:1 E:T, 10.9% ± 5.2% 7AAD+, n = 6) also had a severe cytotoxic defect compared with WT splenocytes (20:1 E:T, 55.8% ± 8.8% 7-AAD+, n = 6, p < 0.001). We repeated similar experiments with IL-15-activated splenocytes by using mice derived in the 129/SvJ background, so that we could test strainmatched Gzma/ and Gzma/Gzmb/ mice (Figure 5F). Results with WT, Gzmb/, and Prf1/ 129/SvJ mice were similar to those observed with B6 mice carrying the same mutations. Splenocytes from Gzma/ mice (20:1 E:T, 51.2% ± 5.9% 7-AAD+, n = 6) had no detectable defect in killing (20:1 E:T, 54.5% ± 13.8% 7-AAD+, n = 6). Splenocytes from Gzma/Gzmb/ mice (20:1 E:T, 14.2% ± 3.1% 7-AAD+, n = 5) demonstrated significantly less killing than those from Gzmb/ mice (20:1 E:T, 19.4% ± 3.1% 7-AAD+, n = 5, p < 0.01). Splenocytes from Prf1/ mice consistently displayed the lowest cytotoxic activity of all genotypes analyzed (20:1 E:T, 5.4% ± 3.3% 7-AAD+, n = 4; p < 0.01 compared to Gzma/Gzmb/). These data show that activated murine NK cells require Gzmb and Prf1 to acquire potent cytotoxic potential against tumor targets in vitro. Poly I:C Induces NK Cell Expression of Gzmb and Prf1 Protein In Vivo Mice were injected i.p. with poly I:C or PBS, and splenic NK cells were analyzed after 24 hr for Gzma, Gzmb, and Prf1 expression by flow cytometry and also were tested for cytotoxic potential (Figure 6). The spleens of poly I:Cinjected mice had a markedly higher percentage of NK cells positive for Gzmb and Prf1 than control mice (Figures 6A and 6B). As expected (Shresta et al., 1995), the splenocytes from poly I:C-injected Gzmb/ and Prf1/ mice displayed reduced killing (Figure 6C). NK Cells Express Gzmb and Prf1 Protein In Vivo after MCMV Infection To further investigate patterns of Gzma, Gzmb, and Prf1 expression in NK cells activated in vivo, we analyzed the well-characterized MCMV mouse model during the NK cell-dependent phase of viral clearance. WT B6 mice were infected with sublethal doses of Smith strain MCMV and then analyzed on days 1, 2, 4, 6, and 8 after MCMV infection for Gzma, Gzmb, and Prf1 expression in NK cells (Figures 7A and 7B). The percentages of NK cells expressing Gzmb and Prf1 increased concurrently, peaked at day 2–4 after infection, and returned to baseline by day 8. In contrast, the percentage of NK cells expressing Gzma was essentially constant throughout the infection. Interestingly, both Ly49H+ and Ly49H NK cells

Figure 6. In Vivo Treatment with Poly I:C Induces NK Cell Gzmb and Prf1 Protein and Cytotoxicity (A and B) Mice were injected i.p. with 200 mg poly I:C or PBS (vehicle control), and after 24 hr, splenic NK1.1+CD3 NK cell expression of intracellular Gzma, Gzmb, and Prf1 was analyzed by flow cytometry. Flow-cytometry data representative of at least three independent experiments are shown (A), with a significant increase in the percent NK cells positive (mean ± SD) for Gzmb and Prf1 after poly I:C injection (B). *p < 0.001. (C) Splenocytes from poly I:C- or PBS-injected mice were used as effectors in a FloKA against RMAS target cells at a 100:1 E:T ratio. Results (mean ± SD) are representative of two independent experiments.

acquired Gzmb and Prf1 during MCMV infection, suggesting that this upregulation was not specific for NK cells stimulated by Ly49H-m157 interactions (data not Immunity 26, 798–811, June 2007 ª2007 Elsevier Inc. 805

Immunity Perforin and Granzyme A/B in Murine NK Cells

Figure 7. MCMV Infection Induces NK Cell Gzmb and Prf1 Protein, which Are Required for Host Defense and Survival (A and B) NK cell expression of Gzma, Gzmb, and Prf1 protein by flow cytometry after MCMV infection. Mice were infected with 5 3 104 pfu MCMV/ mouse, and splenocytes were harvested from naive mice or at days 1, 2, 4, 6, and 8 after infection (n = 3 mice/time point) and analyzed for NK cell expression of intracellular Gzma, Gzmb, and Prf1. Representative plots from individual mice are shown (A), and the percent NK cells positive for Gzma, Gzmb, and Prf1 are summarized for one experiment (B), with each dot indicative of one mouse and the mean shown as a horizontal bar with SD. These data are representative of 2–5 independent experiments. (C) MCMV titers in wild-type (WT), Gzmb/, and Prf1/ mice (B6 background). Mice were infected with 1 3 105 pfu MCMV/mouse. Spleens were harvested on day 3 after infection and analyzed by plaque assay for MCMV titers. Combined results from two independent experiments demonstrate that Gzmb/ (n = 9, p < 0.001) and Prf1/ (n = 5, p < 0.01) had significantly higher MCMV titers in the spleen compared to WT mice (n = 9). (D) Survival after MCMV infection in WT, Gzmb/, and Prf1/ mice. Groups of WT, Gzmb/, and Prf1/ (B6 background) mice were infected with MCMV (day 0) and followed for survival. Kaplan-Meier analysis is shown, with significantly inferior survival compared to WT in both Gzmb/ and Prf1/ mice (p < 0.001) at two different MCMV doses: left, 1 3 105 pfu/mouse (females); right, 2.5 3 105 pfu/mouse (males). All p values shown on the survival curves indicate significance compared to WT mice. There is also a significant difference between the Gzmb/ and Prf1/ curves (p = 0.003) at the 1 3 105 pfu dose in females (left).

shown). Similarly, IL-15-activated splenocytes showed similar induction of Gzmb and Prf1 protein at early time points in both Ly49H+ and Ly49H subsets (Figure S6). Therefore, murine NK cells acquire Gzmb and Prf1 protein during the innate immune response to MCMV in vivo. 806 Immunity 26, 798–811, June 2007 ª2007 Elsevier Inc.

Gzmb/ and Prf1/ Mice Exhibit Higher Mortality than WT Mice after MCMV Infection Although previous reports have identified defects in hostdefense against MCMV infection in Prf1/ mice (Loh et al., 2005; Tay and Welsh, 1997), the role of Gzms for

Immunity Perforin and Granzyme A/B in Murine NK Cells

innate NK cell responses after MCMV infection have not yet been well defined. Spleens from MCMV-infected WT, Gzmb/, and Prf1/ B6 mice were assayed for MCMV titers at day 3 after infection (Figure 7C). Both Gzmb/ (p < 0.001) and Prf1/ (p < 0.01) mice had significantly elevated viral titers in the spleen at day 3 (compared with WT mice), demonstrating that control of MCMV infection was compromised in these mice. We next monitored the survival of age- and sexmatched WT B6, Gzmb/, and Prf1/ mice infected with 1 3 105 or 2.5 3 105 pfu/mouse Smith strain MCMV (Figure 7D). Kaplan-Meier analysis demonstrated a significant increase in mortality in both Gzmb/ and Prf1/ mice (Figure 7D, p < 0.001), with the observed deaths occurring between days 5 and 8 after infection. At the 2.5 3 105 pfu dose of MCMV, Gzmb deficiency was nearly as detrimental as the absence of Prf1; however, at the lower dose of 1 3 105 pfu, greater mortality was observed in Prf1/ mice (p = 0.003). Additional experiments at varying doses of MCMV demonstrated that the LD50 for Gzmb/ mice was 25% that of WT mice, whereas the LD50 of Prf1/ mice was <10% that of WT B6 mice (data not shown). Thus, Gzmb/ and Prf1/ mice clear MCMV inefficiently and exhibit increased mortality after MCMV infection. DISCUSSION In this report, we demonstrate that murine NK cells derived from common laboratory strains of mice housed in SPF conditions contain little or no Gzmb and Prf1 protein, which provides a mechanism to explain why they are minimally cytotoxic at rest. NK cells isolated from the blood, spleen, liver, and bone marrow of mice all had similar, low amounts of Gzmb and Prf1, but they contained abundant Gzma protein. Although several cytokines (IL-12, IL18, IFN-a) were able to induce fractions of NK cells to translate Gzmb and Prf1 protein in bulk splenocyte cultures, only IL-2 and IL-15 stimulated substantial amounts in both bulk splenocytes and flow-sorted NK cells, with IL15 being the most efficient. The striking increase in Gzmb and Prf1 protein abundance was accompanied by only small changes in mRNA levels, suggesting that activation is due to the release of a block in mRNA translation. The requirement for Gzmb and Prf1 for potent NK cytotoxicity in vitro was confirmed with Gzmb/ and Prf1/ mice. The minimal residual killing detected in Prf1/ mice suggests that other contact-dependent killing pathways (e.g., FasL, TRAIL) are minimally utilized in this experimental system, consistent with the small amounts of these proteins detected on the surface of IL-15-activated NK cells at 24–48 hr (data not shown). Most studies examining murine NK cell cytotoxicity over the past 15 years have utilized either poly I:C-injected mice or IL-2-activated adherent NK cells (i.e., LAK cells) as ‘‘surrogates’’ for NK cells. Our data strongly suggest that the minimal in vitro cytotoxicity exhibited by resting mouse NK cells is due to the low amounts of Gzmb and Prf1 in these cells, and that poly I:C or long-term culture

in high-dose IL-2 was necessary to induce expression of these proteins in these past studies, allowing the cells to become cytotoxic. In contrast, a few reports have suggested that freshly purified, resting murine NK cells exhibit substantial cytotoxicity against tumor target cells in vitro (Kim et al., 2002; Vosshenrich et al., 2005). Although the reason for the discrepancy is not immediately clear, it is possible that undetected infections of the Rag-deficient mice used in these studies may have caused NK cell activation or that technical approaches to cytotoxicity measurements are now more sensitive and reproducible. Several groups have reported rapid (6–24 hr) clearance of intravenously injected tumor cells and have therefore suggested that resting NK cells may be fully ‘‘armed.’’ However, our studies have clearly shown that IL-15 can begin to induce Gzmb and Prf1 production from a preexisting pool of mRNA within 6 hr of treatment in vitro, and therefore these data are not inconsistent with our observations. Regardless, the complex variables that determine the clearance rates of many tumor cell lines in vivo are not yet well understood (e.g., tumor cell trafficking, immune activation, effector compartments, etc.), and all could play roles in clearance kinetics. Finally, in vivo tumor clearance may not be an accurate measure of Prf1dependent NK cytotoxicity (Grundy et al., 2006). We have demonstrated that resting murine NK cells are ‘‘pre-armed’’ with high amounts of Gzmb and Prf1 mRNA, which NK cells can rapidly translate into protein upon activation in vitro and in vivo. The use of such a mechanism by murine NK cells to control Gzmb and Prf1 would allow for the rapid production of effector proteins without the need for new gene transcription. Interestingly, resting murine NK cells have previously been shown to constitutively express IFN-g mRNA—but not protein—which is rapidly translated upon activation (Stetson et al., 2003). Previous reports have also demonstrated that murine memory CTLs express abundant Gzmb mRNA, but do not contain detectable amounts of Gzmb protein (Kaech et al., 2002). Similarly, we have shown a more rapid increase in Gzmb protein production after the in vitro activation of human memory CD4+ T cells compared to naive CD4+ T cells (Grossman et al., 2004b). Thus, the pre-arming of cytotoxic lymphocytes with effector mRNAs may be a general mechanism that many kinds of cytotoxic lymphocytes use to allow these cells to rapidly respond to dangerous stimuli. The presence of abundant Gzma protein in resting murine NK cells contrasts sharply with that of Prf1 and Gzmb. Gzma is one of the most highly expressed genes in both resting and activated NK cells, suggesting a key role in murine NK cell physiology. Because Gzma mRNA and protein expression remain relatively constant in resting and activated NK cells, this gene exhibits a fundamentally different pattern of regulation than does Gzmb and Prf1. Despite the abundance of Gzma, Gzmadeficient NK cells kill tumor targets normally in vitro, consistent with previous reports that used other cytotoxic effectors (Ebnet et al., 1995a; Mullbacher et al., 1996; Pereira et al., 2000; Shresta et al., 1999). NK cells from Immunity 26, 798–811, June 2007 ª2007 Elsevier Inc. 807

Immunity Perforin and Granzyme A/B in Murine NK Cells

Gzma/Gzmb/ mice kill less efficiently than NK cells derived from Gzmb/ mice, suggesting that Gzma may serve as a ‘‘fail-safe’’ mechanism for killing target cells that contain inhibitors of Gzmb (Shresta et al., 1999; Russell and Ley, 2002). While several previous reports have examined late, nonNK-dependent MCMV responses in Prf1/, Gzma/, and/or Gzmb/ mice (Cheeran et al., 2005; Dix et al., 2003; Riera et al., 2000), the role of these proteins for innate, NK-dependent responses (Biron et al., 1999; Bukowski et al., 1983) had not previously been fully examined. Specific NK control of MCMV in B6 mice results from NK cell recognition of the virally encoded m157 protein (Arase et al., 2002; Smith et al., 2002) on the surface of infected cells by the activating NK receptor Ly49H (Brown et al., 2001; Daniels et al., 2001; Lee et al., 2001). The importance of Prf1 for early NK-dependent responses was demonstrated by elevated viral titers and decreased survival during MCMV infection in Prf1/ mice (Loh et al., 2005; Tay and Welsh, 1997). Here, we show that MCMV infection induces NK cell Prf1 and Gzmb protein production that peaks 2–4 days after infection and disappears by day 8. MCMV induces a number of candidate cytokines, which are produced by plasmacytoid dendritic cells and other innate immune cell types (Andoniou et al., 2005; Andrews et al., 2003; Dalod et al., 2003; Krug et al., 2004; Nguyen et al., 2002). Our in vitro studies suggest that IL-15 is a leading candidate for the observed induction; however, it is likely that multiple pathways synergize to activate NK Gzmb and Prf1 protein in vivo. To begin to address whether NK receptors play a role in this synergistic activation pathway, we evaluated the ability of activating NK receptor ligation (plate-bound NK1.1 or NKG2D or coculture with Baf/3 cells that express m157 [Ho et al., 2002; Kim et al., 2005]) to induce Gzmb or Prf1 protein in splenic NK cells after 12–48 hr. Ligation of individual NK receptors (NK1.1, NKG2D, or Ly49H) in these settings did not result in increased NK cell Gzmb or Prf1 protein translation, according to experimental conditions that result in IFN-g production (data not shown). Thus, cytokines (and possibly other signals) may well represent the arming pathway that stimulates the translation of Gzmb and Prf1, which provides the potential for potent cytotoxicity. NK-receptor ligation probably represents the ‘‘trigger’’ that leads to directed granule release at the NK-target-cell synapse. Experiments examining viral titers and survival demonstrate an important role for Gzmb cluster enzymes in MCMV clearance and confirm the essential role of Prf1. We were unable to evaluate the impact of Gzma/ NK cells during MCMV infection because the mutation created in our laboratory was derived in 129/SvJ mice, which lack Ly49H and specific NK-mediated MCMV resistance. Definitive evaluation of the contribution of Gzma to innate host defense against MCMV awaits the backcross of our Gzma/ mice to the B6 background, which is in progress. A recent study that used a lower MCMV dose (1 3 104 pfu) reported no significant differences in viral titers in the spleens of Gzmb/ and Gzma/ mice compared with 808 Immunity 26, 798–811, June 2007 ª2007 Elsevier Inc.

WT mice, although higher titers were observed in Gzma/Gzmb/ and Prf1/ mice (van Dommelen et al., 2006). In addition, no mortality was observed in Gzmb/ or Gzma/Gzmb/ mice. However, because LD50 values for the Gzm-deficient mice were not determined, it is possible that the apparent discrepancy between the studies reflects differences in the MCMV doses used. Regardless, both studies suggest that Gzma may play a role in NK-dependent MCMV responses in vivo. Our findings in murine NK cells contrast sharply with observations made with the majority of resting human NK cells from the peripheral blood of healthy individuals, which express Gzma, Gzmb, and Prf1 protein and are cytotoxic (Bratke et al., 2005; Cooper et al., 2001; Grossman et al., 2004b). One potential explanation for the observed differences may derive from the chronic exposure of humans to infectious pathogens; SPF laboratory mice have few exposures to pathogens that activate the innate immune system. Alternatively, there may be fundamental differences between murine and human NK cell cytotoxic granule gene regulation, with different mechanisms of prearming at the levels of mRNA versus protein. However, it is notable that the CD56bright human NK cell subset is minimally cytotoxic at rest, expresses little Gzmb or Prf1 but considerable Gzma protein, and becomes potently cytotoxic only after cytokine activation (Bratke et al., 2005; Cooper et al., 2001); this mimics our findings with mouse NK cells. A similar profile of Gzm and Prf1 protein expression has been noted in CD8+ T cells in patients who fail to clear chronic viral infections (Andersson et al., 1999; Chen et al., 2001); a posttranscriptional mechanism could also help to explain these findings. The differences between mouse and human NK cell biology are important, because mice serve as an essential model system to understand innate immune responses to infection and malignancy. The identification of differences between mouse and human NK cell Prf1 and Gzm expression patterns highlight the importance of assessing both species to obtain a complete understanding of NK cell biology.

EXPERIMENTAL PROCEDURES Mice and Cell Lines WT C57Bl/6, 129/SvJ, BALB/c, CBA, A/J, and Rag1/ (B6) mice were obtained from the Jackson Laboratory. The following mouse strains were previously described: Gzmb/ mice (Heusel et al., 1994), which were backcrossed to B6J for 12+ generations (Revell et al., 2005); and Gzma/ (Shresta et al., 1999), Gzmb/ (Revell et al., 2005), and Gzma/Gzmb/ mice, which were all derived in the 129/SvJ background (Revell et al., 2005). Prf1/ mice (B6) were obtained from Jackson Laboratory (Kagi et al., 1994) and were backcrossed to 129/SvJ with a speed congenic approach for six generations; founders used to generate the colony had >98% 129/SvJ markers. EL4 (ATCC), YAC1, and RMAS (a kind gift of W. Yokoyama) cell lines were grown in K10 media (RPMI 1640, 10% FCS, 10 mM HEPES, 1% NEAA, 1% sodium pyruvate, 1% L-glutamine, 13 penicillin/streptomycin, 0.57 mM bME). The murine 36.2 NK line was maintained in K10 + 1000 U/mL recombinant (r) human IL-2 (MacIvor et al., 1999). All mice were maintained in SPF housing, and all experiments were conducted in accordance with institutional animal care and use guidelines.

Immunity Perforin and Granzyme A/B in Murine NK Cells

Reagents and Antibodies mAbs used were anti-mouse NK1.1 (PK136), CD3 (145-2C11), CD49b (DX5), CD16/32 (2.4G2) (BD Biosciences), Prf1 (eBioscience, OMAKD), and Gzmb (Caltag, GB12). The anti-mouse Gzma mAb was produced in our laboratory and is available from Santa Cruz (3G8.5), with Gzma specificity shown in Figure S1. For immunoblots, rat antiPrf1 (P1-8, Kamiya, 1:5000), polyclonal rabbit anti-Gzmb (526B, 1:1000) (Shresta et al., 1999), and goat anti-actin (C-11, Santa Cruz, 1:1000) were used and developed with HRP secondary Abs (Amersham, 1:5000). Anti-NKG2D mAbs C7 and A10 were kind gifts of W. Yokoyama. Cytokines were obtained from R&D Systems (rmIL-15, rmIL-12, rmIL-18, and rmIFN-a) or Chiron (rhIL-2). All cytokines were endotoxin free and stored at 80 C after reconstitution in PBS + 0.1% BSA. Poly I:C was obtained from Sigma. Cell Isolation and Stimulation Spleens were harvested and single-cell suspensions were generated by pressing through sterile 70 mM nylon mesh. RBCs were lysed and cells washed in complete K10 medium. For in vitro cytokine stimulation, splenocytes were plated at 10 3 106/mL in 24-well plates in 100 ng/mL IL-15 (or at the indicated concentration), 1000 U/mL IL-2, 100 ng/mL IL-12, or 100 ng/mL IL-18. Cells were harvested by pipetting with cold PBS after 48 hr (or as indicated), and analyzed by flow cytometry or used as effector cells in FloKAs. For cell sorting, splenocytes were surface stained with anti-NK1.1 or DX5 and anti-CD3 mAbs, and NK1.1+CD3 or DX5+CD3 cells were isolated on a MoFlo cell sorter (routinely R99% pure). Flow-Based Killing and Cr51 Release Assays Flow-based killing assays (FloKAs) (4 hr) were performed as described (Grossman et al., 2004a, 2004b; Revell et al., 2005), utilizing resting or activated splenocytes or purified NK cells as effectors, and the NKsensitive RMAS, YAC1, or NK-resistant EL4 tumor cell lines as targets at E:T ratios of 20:1, 10:1, and 5:1 (or as indicated) in 2–3 replicate wells. Controls included target-only wells for background (spontaneous) cell death and effector-only wells to ensure that target cells only were in the CFSE gate. Results are shown as mean percent 7-AAD+ target cells ± SD, with background death subtracted. Cr51-release assays were performed as previously described (Shresta et al., 1995). Intracellular Staining and Flow Cytometry Splenocytes or tissue mononuclear cells were stained for surface markers and intracellular Gzma, Gzmb, or Prf1 as described (Grossman et al., 2004a, 2004b; Revell et al., 2005) with directly conjugated Gzma (1:400 dilution), Gzmb (1:400 dilution), or Prf1 (2.5 mg/mL) mAbs. Sample data were acquired on a Cytek-modified FACScan (BD) or FC500 (Coulter) flow cytometer and analyzed with CXP (Coulter) or FloJo (TreeStar) software. Isotype controls were used to set quadrant gates. Immunoblots Protein lysates were generated from resting or IL-15-activated bulk spleen or flow-sorted NK cells. Reducing SDS-PAGE, transfer to PVDF membranes, and immunoblots for Prf1, Gzmb, or actin were performed as described (Revell et al., 2005; Shresta et al., 1999). RNA Isolation and Microarrays RNA isolation and microarray processing and analysis were performed as described (Yuan et al., 2007), with flow-sorted NK cell samples or spleens at rest or after 24 hr of IL-15 activation, with linear amplification, on Affymetrix 430v2.0 expression arrays. Analysis was performed with DecisionSite for Functional Genomics software (Spotfire, Somerville, MA) or Microsoft Excel (Microsoft, Redmond, WA). Scatter plots were generated after removing probesets that were Affymetrix ‘‘absent’’ calls on all microarrays. Microarray data are available at the Gene Expression Omnibus (GEO) microarray database (http:// www.ncbi.nlm.nih.gov/geo/), accession number GSE7764.

Quantitative Real-Time RT-PCR Total RNA was isolated from paired (resting and 24 hr IL-15-activated) flow-sorted NK cell samples (2.5 3 105) with the RNeasy Micro Kit (Qiagen). qRT-PCR was performed as described for Gzma and Gzmb (Revell et al., 2005), and for Prf1 the following primers were used: forward, 50 -GATGTGAACCCTAGGCCAGA-30 ; reverse, 50 -AAA GAGGTGGCCATTTTGTG-30 . Comparisons were made by the comparative Ct method, with b-actin serving as the comparator, because this mRNA did not significantly change between resting and IL-15activated NK cells (Figure 4). Data are presented as fold-change relative to resting NK cell samples, which are set to a value of 1. MCMV Infection, Plaque Assays, and Poly I:C Injections Smith strain MCMV was kindly provided by W. Yokoyama (Washington University, St. Louis, MO). A salivary gland stock of MCMV was prepared from BALB/c mice that had been i.p. injected with 1 3 106 pfu tissue-culture-propagated MCMV, and the titer of the stock was determined via standard plaque assay (Brown et al., 2001) with NIH 3T12 fibroblasts (ATCC; Manassas, VA). For experiments, mice were injected i.p. with 5 3 104 to 2.5 3 105 pfu/mouse salivary gland Smith MCMV stock. For survival curves, mice were observed daily after infection for 21 days, and moribund mice were euthanized per institutional guidelines. For viral titer plaque assays, spleens from day 3 after infection were frozen in D10 media and homogenized by bead beating (BioSpec), and titers were read in triplicate as described (Heise and Virgin, 1995) and are presented as log pfu/spleen. Poly I:C (200 mg) or vehicle (PBS) was injected i.p., and 24 hr later, spleens were harvested to assess Gzm/Prf1 expression and cytotoxicity. Light Microscopy Light microscopy on flow-sorted murine NK cells stained with Wright/ Giemsa were performed as described (Grossman et al., 2004a). Statistics Two-tailed t tests were used to determine significant differences (p < 0.05); Kaplan-Meier analysis was performed with Prism software (Graph Pad, San Diego, CA). Supplemental Data Six figures are available at http://www.immunity.com/cgi/content/full/ 26/6/798/DC1/. ACKNOWLEDGMENTS This work was supported by grants by the National Institutes of Health (DK49786 to T.J.L., AI059083 to A.R.F.). We thank the Hi Speed Flow Core, the Multiplexed Gene Analysis Core, and the Bioinformatics Core of Siteman Cancer Center for their invaluable support. M. Hoock provided excellent animal husbandry, and M. Hapak provided outstanding assistance with Cr-release assays. We thank M. Cooper, B. Carreno, J. Payton, D. Link, and T. Graubert for insightful discussions. The Gzma mAb was developed in part with the WUSM Hybridoma Center (supported by the Department of Pathology and Immunology and NIH grant P30 AR048335). T.J.L. holds the license for the Gzma monoclonal antibody clone 3G8.5 (Santa Cruz). The authors have no other special/competing interests to disclose. Received: November 17, 2006 Revised: March 28, 2007 Accepted: April 4, 2007 Published online: May 31, 2007 REFERENCES Andersson, J., Behbahani, H., Lieberman, J., Connick, E., Landay, A., Patterson, B., Sonnerborg, A., Lore, K., Uccini, S., and Fehniger, T.E. (1999). Perforin is not co-expressed with granzyme A within cytotoxic

Immunity 26, 798–811, June 2007 ª2007 Elsevier Inc. 809

Immunity Perforin and Granzyme A/B in Murine NK Cells

granules in CD8 T lymphocytes present in lymphoid tissue during chronic HIV infection. AIDS 13, 1295–1303. Andoniou, C.E., van Dommelen, S.L., Voigt, V., Andrews, D.M., Brizard, G., Asselin-Paturel, C., Delale, T., Stacey, K.J., Trinchieri, G., and Degli-Esposti, M.A. (2005). Interaction between conventional dendritic cells and natural killer cells is integral to the activation of effective antiviral immunity. Nat. Immunol. 6, 1011–1019. Andrews, D.M., Scalzo, A.A., Yokoyama, W.M., Smyth, M.J., and Degli-Esposti, M.A. (2003). Functional interactions between dendritic cells and NK cells during viral infection. Nat. Immunol. 4, 175–181. Arase, H., Mocarski, E.S., Campbell, A.E., Hill, A.B., and Lanier, L.L. (2002). Direct recognition of cytomegalovirus by activating and inhibitory NK cell receptors. Science 296, 1323–1326. Babichuk, C.K., and Bleackley, R.C. (1997). Mutational analysis of the murine granzyme B gene promoter in primary T cells and a T cell clone. J. Biol. Chem. 272, 18564–18571. Babichuk, C.K., Duggan, B.L., and Bleackley, R.C. (1996). In vivo regulation of murine granzyme B gene transcription in activated primary T cells. J. Biol. Chem. 271, 16485–16493. Biron, C.A., and Brossay, L. (2001). NK cells and NKT cells in innate defense against viral infections. Curr. Opin. Immunol. 13, 458–464. Biron, C.A., Nguyen, K.B., Pien, G.C., Cousens, L.P., and SalazarMather, T.P. (1999). Natural killer cells in antiviral defense: function and regulation by innate cytokines. Annu. Rev. Immunol. 17, 189–220. Bratke, K., Kuepper, M., Bade, B., Virchow, J.C., Jr., and Luttmann, W. (2005). Differential expression of human granzymes A, B, and K in natural killer cells and during CD8+ T cell differentiation in peripheral blood. Eur. J. Immunol. 35, 2608–2616. Brown, M.G., Dokun, A.O., Heusel, J.W., Smith, H.R., Beckman, D.L., Blattenberger, E.A., Dubbelde, C.E., Stone, L.R., Scalzo, A.A., and Yokoyama, W.M. (2001). Vital involvement of a natural killer cell activation receptor in resistance to viral infection. Science 292, 934–937. Bukowski, J.F., Woda, B.A., Habu, S., Okumura, K., and Welsh, R.M. (1983). Natural killer cell depletion enhances virus synthesis and virus-induced hepatitis in vivo. J. Immunol. 131, 1531–1538. Cheeran, M.C., Gekker, G., Hu, S., Palmquist, J.M., and Lokensgard, J.R. (2005). T cell-mediated restriction of intracerebral murine cytomegalovirus infection displays dependence upon perforin but not interferon-gamma. J. Neurovirol. 11, 274–280. Chen, G., Shankar, P., Lange, C., Valdez, H., Skolnik, P.R., Wu, L., Manjunath, N., and Lieberman, J. (2001). CD8 T cells specific for human immunodeficiency virus, Epstein-Barr virus, and cytomegalovirus lack molecules for homing to lymphoid sites of infection. Blood 98, 156–164. Cooper, M.A., Fehniger, T.A., and Caligiuri, M.A. (2001). The biology of human natural killer-cell subsets. Trends Immunol. 22, 633–640. Dalod, M., Hamilton, T., Salomon, R., Salazar-Mather, T.P., Henry, S.C., Hamilton, J.D., and Biron, C.A. (2003). Dendritic cell responses to early murine cytomegalovirus infection: subset functional specialization and differential regulation by interferon alpha/beta. J. Exp. Med. 197, 885–898. Daniels, K.A., Devora, G., Lai, W.C., O’Donnell, C.L., Bennett, M., and Welsh, R.M. (2001). Murine cytomegalovirus is regulated by a discrete subset of natural killer cells reactive with monoclonal antibody to Ly49H. J. Exp. Med. 194, 29–44. Dix, R.D., Podack, E.R., and Cousins, S.W. (2003). Loss of the perforin cytotoxic pathway predisposes mice to experimental cytomegalovirus retinitis. J. Virol. 77, 3402–3408. Ebnet, K., Hausmann, M., Lehmann-Grube, F., Mullbacher, A., Kopf, M., Lamers, M., and Simon, M.M. (1995a). Granzyme A-deficient mice retain potent cell-mediated cytotoxicity. EMBO J. 14, 4230– 4239.

810 Immunity 26, 798–811, June 2007 ª2007 Elsevier Inc.

Ebnet, K., Levelt, C.N., Tran, T.T., Eichmann, K., and Simon, M.M. (1995b). Transcription of granzyme A and B genes is differentially regulated during lymphoid ontogeny. J. Exp. Med. 181, 755–763. Garcia-Sanz, J.A., MacDonald, H.R., Jenne, D.E., Tschopp, J., and Nabholz, M. (1990). Cell specificity of granzyme gene expression. J. Immunol. 145, 3111–3118. Grossman, W.J., Revell, P.A., Lu, Z.H., Johnson, H., Bredemeyer, A.J., and Ley, T.J. (2003). The orphan granzymes of humans and mice. Curr. Opin. Immunol. 15, 544–552. Grossman, W.J., Verbsky, J.W., Barchet, W., Colonna, M., Atkinson, J.P., and Ley, T.J. (2004a). Human T regulatory cells can use the perforin pathway to cause autologous target cell death. Immunity 21, 589– 601. Grossman, W.J., Verbsky, J.W., Tollefsen, B.L., Kemper, C., Atkinson, J.P., and Ley, T.J. (2004b). Differential expression of granzymes A and B in human cytotoxic lymphocyte subsets and T regulatory cells. Blood 104, 2840–2848. Grundy, M.A., Zhang, T., and Sentman, C.L. (2006). NK cells rapidly remove B16F10 tumor cells in a perforin and interferon-gamma independent manner in vivo. Cancer Immunol. Immunother., in press. Published online December 8, 2006. 10.1007/s00262-006-0264-1. Hanson, R.D., and Ley, T.J. (1990). Transcriptional activation of the human cytotoxic serine protease gene CSP-B in T lymphocytes. Mol. Cell. Biol. 10, 5655–5662. Heise, M.T., and Virgin, H.W., IV. (1995). The T-cell-independent role of gamma interferon and tumor necrosis factor alpha in macrophage activation during murine cytomegalovirus and herpes simplex virus infections. J. Virol. 69, 904–909. Henkart, P.A. (1985). Mechanism of lymphocyte-mediated cytotoxicity. Annu. Rev. Immunol. 3, 31–58. Herberman, R.B., Nunn, M.E., Holden, H.T., and Lavrin, D.H. (1975). Natural cytotoxic reactivity of mouse lymphoid cells against syngeneic and allogeneic tumors. II. Characterization of effector cells. Int. J. Cancer 16, 230–239. Heusel, J.W., Wesselschmidt, R.L., Shresta, S., Russell, J.H., and Ley, T.J. (1994). Cytotoxic lymphocytes require granzyme B for the rapid induction of DNA fragmentation and apoptosis in allogeneic target cells. Cell 76, 977–987. Ho, E.L., Carayannopoulos, L.N., Poursine-Laurent, J., Kinder, J., Plougastel, B., Smith, H.R., and Yokoyama, W.M. (2002). Costimulation of multiple NK cell activation receptors by NKG2D. J. Immunol. 169, 3667–3675. Kaech, S.M., Hemby, S., Kersh, E., and Ahmed, R. (2002). Molecular and functional profiling of memory CD8 T cell differentiation. Cell 111, 837–851. Kagi, D., Ledermann, B., Burki, K., Seiler, P., Odermatt, B., Olsen, K.J., Podack, E.R., Zinkernagel, R.M., and Hengartner, H. (1994). Cytotoxicity mediated by T cells and natural killer cells is greatly impaired in perforin-deficient mice. Nature 369, 31–37. Kiessling, R., Klein, E., Pross, H., and Wigzell, H. (1975). ‘‘Natural’’ killer cells in the mouse. II. Cytotoxic cells with specificity for mouse Moloney leukemia cells. Characteristics of the killer cell. Eur. J. Immunol. 5, 117–121. Kim, S., Iizuka, K., Kang, H.S., Dokun, A., French, A.R., Greco, S., and Yokoyama, W.M. (2002). In vivo developmental stages in murine natural killer cell maturation. Nat. Immunol. 3, 523–528. Kim, S., Poursine-Laurent, J., Truscott, S.M., Lybarger, L., Song, Y.J., Yang, L., French, A.R., Sunwoo, J.B., Lemieux, S., Hansen, T.H., and Yokoyama, W.M. (2005). Licensing of natural killer cells by host major histocompatibility complex class I molecules. Nature 436, 709–713. Krug, A., French, A.R., Barchet, W., Fischer, J.A., Dzionek, A., Pingel, J.T., Orihuela, M.M., Akira, S., Yokoyama, W.M., and Colonna, M. (2004). TLR9-dependent recognition of MCMV by IPC and DC

Immunity Perforin and Granzyme A/B in Murine NK Cells

generates coordinated cytokine responses that activate antiviral NK cell function. Immunity 21, 107–119. Lanier, L.L. (2005). NK cell recognition. Annu. Rev. Immunol. 23, 225– 274. Lee, S.H., Girard, S., Macina, D., Busa, M., Zafer, A., Belouchi, A., Gros, P., and Vidal, S.M. (2001). Susceptibility to mouse cytomegalovirus is associated with deletion of an activating natural killer cell receptor of the C-type lectin superfamily. Nat. Genet. 28, 42–45. Lieberman, J. (2003). The ABCs of granule-mediated cytotoxicity: new weapons in the arsenal. Nat. Rev. Immunol. 3, 361–370. Loh, J., Chu, D.T., O’Guin, A.K., Yokoyama, W.M., and Virgin, H.W., IV. (2005). Natural killer cells utilize both perforin and gamma interferon to regulate murine cytomegalovirus infection in the spleen and liver. J. Virol. 79, 661–667. MacIvor, D.M., Pham, C.T., and Ley, T.J. (1999). The 50 flanking region of the human granzyme H gene directs expression to T/natural killer cell progenitors and lymphokine-activated killer cells in transgenic mice. Blood 93, 963–973. Mullbacher, A., Ebnet, K., Blanden, R.V., Hla, R.T., Stehle, T., Museteanu, C., and Simon, M.M. (1996). Granzyme A is critical for recovery of mice from infection with the natural cytopathic viral pathogen, ectromelia. Proc. Natl. Acad. Sci. USA 93, 5783–5787. Nguyen, K.B., Salazar-Mather, T.P., Dalod, M.Y., Van Deusen, J.B., Wei, X.Q., Liew, F.Y., Caligiuri, M.A., Durbin, J.E., and Biron, C.A. (2002). Coordinated and distinct roles for IFN-alpha beta, IL-12, and IL-15 regulation of NK cell responses to viral infection. J. Immunol. 169, 4279–4287. Pereira, R.A., Simon, M.M., and Simmons, A. (2000). Granzyme A, a noncytolytic component of CD8(+) cell granules, restricts the spread of herpes simplex virus in the peripheral nervous systems of experimentally infected mice. J. Virol. 74, 1029–1032. Pham, C.T., MacIvor, D.M., Hug, B.A., Heusel, J.W., and Ley, T.J. (1996). Long-range disruption of gene expression by a selectable marker cassette. Proc. Natl. Acad. Sci. USA 93, 13090–13095. Revell, P.A., Grossman, W.J., Thomas, D.A., Cao, X., Behl, R., Ratner, J.A., Lu, Z.H., and Ley, T.J. (2005). Granzyme B and the downstream granzymes C and/or F are important for cytotoxic lymphocyte functions. J. Immunol. 174, 2124–2131. Riera, L., Gariglio, M., Valente, G., Mullbacher, A., Museteanu, C., Landolfo, S., and Simon, M.M. (2000). Murine cytomegalovirus replication in salivary glands is controlled by both perforin and granzymes during acute infection. Eur. J. Immunol. 30, 1350–1355. Russell, J.H., and Ley, T.J. (2002). Lymphocyte-mediated cytotoxicity. Annu. Rev. Immunol. 20, 323–370. Shresta, S., MacIvor, D.M., Heusel, J.W., Russell, J.H., and Ley, T.J. (1995). Natural killer and lymphokine-activated killer cells require granzyme B for the rapid induction of apoptosis in susceptible target cells. Proc. Natl. Acad. Sci. USA 92, 5679–5683. Shresta, S., Goda, P., Wesselschmidt, R., and Ley, T.J. (1997). Residual cytotoxicity and granzyme K expression in granzyme A-deficient cytotoxic lymphocytes. J. Biol. Chem. 272, 20236–20244.

Shresta, S., Graubert, T.A., Thomas, D.A., Raptis, S.Z., and Ley, T.J. (1999). Granzyme A initiates an alternative pathway for granule-mediated apoptosis. Immunity 10, 595–605. Smith, H.R., Heusel, J.W., Mehta, I.K., Kim, S., Dorner, B.G., Naidenko, O.V., Iizuka, K., Furukawa, H., Beckman, D.L., Pingel, J.T., et al. (2002). Recognition of a virus-encoded ligand by a natural killer cell activation receptor. Proc. Natl. Acad. Sci. USA 99, 8826– 8831. Smyth, M.J., Browne, K.A., Kinnear, B.F., Trapani, J.A., and Warren, H.S. (1995). Distinct granzyme expression in human CD3- CD56+ large granular- and CD3- CD56+ small high density-lymphocytes displaying non-MHC-restricted cytolytic activity. J. Leukoc. Biol. 57, 88–93. Stetson, D.B., Mohrs, M., Reinhardt, R.L., Baron, J.L., Wang, Z.E., Gapin, L., Kronenberg, M., and Locksley, R.M. (2003). Constitutive cytokine mRNAs mark natural killer (NK) and NK T cells poised for rapid effector function. J. Exp. Med. 198, 1069–1076. Tay, C.H., and Welsh, R.M. (1997). Distinct organ-dependent mechanisms for the control of murine cytomegalovirus infection by natural killer cells. J. Virol. 71, 267–275. van Dommelen, S.L., Sumaria, N., Schreiber, R.D., Scalzo, A.A., Smyth, M.J., and Degli-Esposti, M.A. (2006). Perforin and granzymes have distinct roles in defensive immunity and immunopathology. Immunity 25, 835–848. Vosshenrich, C.A., Ranson, T., Samson, S.I., Corcuff, E., Colucci, F., Rosmaraki, E.E., and Di Santo, J.P. (2005). Roles for common cytokine receptor gamma-chain-dependent cytokines in the generation, differentiation, and maturation of NK cell precursors and peripheral NK cells in vivo. J. Immunol. 174, 1213–1221. Yokoyama, W.M., and Kim, S. (2006). How do natural killer cells find self to achieve tolerance? Immunity 24, 249–257. Yokoyama, W.M., Kim, S., and French, A.R. (2004). The dynamic life of natural killer cells. Annu. Rev. Immunol. 22, 405–429. Yuan, W., Payton, J.E., Holt, M.S., Link, D.C., Watson, M.A., Dipersio, J.F., and Ley, T.J. (2007). Commonly dysregulated genes in murine APL cells. Blood 109, 961–970.

Accession Numbers The microarray expression data have been deposited in the Gene Expression Omnibus (GEO) microarray database under the accession number GSE7764.

Note Added in Proof After this paper was submitted, Lucas et al. have shown that IL-15 presented ‘‘in trans’’ on the surface of dendritic cells during the innate immune response to infection arms resting murine NK cells by increasing Gzmb expression and cytotoxicity against tumor targets in vitro (Lucas, M., Schachterle, W., Oberle, K., Aichele, P., and Diefenbach, A. (2007). Dendritic cells prime natural killer cells by trans-presenting interleukin 15. Immunity 26, 1–15).

Immunity 26, 798–811, June 2007 ª2007 Elsevier Inc. 811