A requirement for CD45 distinguishes Ly49D-mediated cytokine and chemokine production from killing in primary natural killer cells

Mice with a targeted disruption of exon-6 in the CD45 gene have been previously shown to have enhanced number of splenic NK cells (32), however, it is unclear how this arises. We investigated the development and phenotype of NK cells from mice with a targeted disruption of CD45 exon-12 (33). These mice (hereafter referred to as CD45^−/−) had a four- and fivefold increase in the percentage of NK1.1⁺CD49b⁺TCR-β⁻ NK cells in the spleen and liver, respectively (Fig. 1 A). This increased percentage equates to significantly enhanced numbers of NK1.1⁺CD49b⁺TCR-β⁻ cells in CD45^−/− mice with 19 ± 2 × 10⁶ and 1.3 ± 0.2 × 10⁶ NK cells in the spleen and liver, respectively, compared with 2.4 ± 0.3 × 10⁶ and 2 ± 0.4 × 10⁵ (P < 0.05; n = 8) in control mice. This enhanced NK cellularity is unlikely to result from increased production in BM as the percentage of NK precursors (CD122⁺NK1.1⁻DX5⁻CD3⁻CD19⁻) (34) was unchanged (Fig. 1 A) and no difference was observed in bromodeoxyuridine (BrdU) uptake by BM NK1.1⁺CD49b⁺ cells (Fig. 1 B). We did observe however, greater turnover of NK cells in the spleen (P = 0.008), but normal turnover in the liver (Fig. 1 B). These findings indicate that the increased number of NK cells in the spleen of CD45^−/− mice results from enhanced peripheral expansion. Given that enhanced NK cell turnover in CD45^−/− mice appears to occur only in the spleen and may reflect a consequence of the CD45-deficient environment, we prepared mixed hematopoietic chimeras. The NK cells in the spleen of the chimeras were mostly of CD45^−/− origin (80% Ly5⁻) indicating the increased in vivo production is intrinsic to CD45^−/− NK cells (Fig. 1 C), consistent with a previous report (32). We next examined whether CD45 was required for normal maturation by examining cells for the acquisition of activating and inhibitory receptors. We observed a significantly reduced proportion of NK1.1⁺DX5⁺ cells expressing the activating receptor Ly49D (P = 0.043) in both the spleen and liver of CD45^−/− mice (Table I; and unpublished data). The acquisition of other Ly49 receptors and CD94-NKG2A/C/E and NKG2D, thought to occur before the expansion stage in NK development (35), was not significantly affected by the absence of CD45. All receptors expressed on CD45^−/− NK cells were at equivalent levels to controls except for a significant down-regulation of the activating receptor CD244 (P < 0.001), a phenotype that is stable in mixed chimeras (unpublished data). Interestingly, in vitro expansion of CD45^−/− NK cells using IL-15 returned CD244 expression and the percentage of Ly49D⁺ NK cells to levels equivalent to controls (Fig. 1 D and unpublished data).

CD45 and the SFK have been implicated as positive and negative regulators of cytokine responses in multiple cell types (27), prompting us to investigate the cytokine responsiveness of CD45^−/− NK cells. We observed CD45^−/− NK cells to proliferate normally across a range of IL-15 concentrations and different time points in tissue culture (Fig. 2, A and B). We recently showed that IL-21 functionally differentiates NK cells both in vivo and in vitro (36). IL-21 was able to slow IL-15–mediated division in both control and CD45^−/− NK cells (Fig. 2 B) and further differentiate these cells as characterized by a down-regulation of NK1.1 (Fig. 2 C) and the acquisition of a larger, more granular phenotype and reduced viability (unpublished data). Taken together, these data show CD45 not to be required for IL-15 and IL-21 responsiveness, and that the cells derived from these cultures are indistinguishable by phenotype from controls.

An important cellular response triggered by activating receptors on NK cells is the production of cytokines. To determine if CD45 is required for cytokine production, IL-15–expanded NK cells were cultured overnight in the presence of plate-bound mAbs. We found CD45 to be critical for the production of IFN-γ after ligation of the ITAM-containing receptors Ly49D, NKG2D, and CD16 (P < 0.039) but not NK1.1 or the IL-12 and IL-18 receptors (Fig. 3 A). Interestingly, inhibitors of SFK (PP2 and SU6656) mimicked these results (P < 0.04; Fig. 3 B and unpublished data) suggesting a role for CD45 in the activation of these kinases in NK cells, whereas inhibitors of the PI3K pathway (Wortmannin and Ly294002) had little effect on IFN-γ production (unpublished data). Stimulation of mouse NK cells by coculture with CHO cells is predominantly through Ly49D recognition (37), and this also failed to elicit IFN-γ from CD45^−/− NK cells at ratios of 1:10 and 1:1 for 24 h (Fig. 3 C). CD16 stimulation of control NK cells induced GM-CSF, which was significantly impaired in the absence of CD45 (P = 0.0004; Fig. 3 D) or after treatment of control cells with PP2 (P = 0.0003; Fig. 3 E). We also measured these parameters using freshly isolated NK1.1⁺CD49b⁺TCR-β⁻ cells and found that while cytokine production was reduced across all stimuli compared with in vitro expanded cells, CD45^−/− NK cells still secreted vastly less cytokine compared with control cells (Fig. 3 F). Collectively these data highlight the critical requirement for SFK activation and CD45 in cytokine production after ITAM receptor stimulation.

NK cells are a potent source of inflammatory chemokines such as RANTES, MIP-1α, and MIP-1β (6, 38). These are critical for recruiting effector cells to the site of antigen challenge thus aiding in the clearance of pathogens. Ly49D ligation on NK cells has been shown to result in the robust production of MIP-1α and MIP-1β (38) Ly49D-stimulated CD45^−/− NK cells produced between 100- and 1,000-fold less MIP-1α (P < 0.0001) and MIP-1β (P < 0.0002) than control NK cells (Fig. 4 A). Furthermore, the T cell and monocyte attractant RANTES was also abundantly produced by NK cells after Ly49D ligation and this also displayed an essential requirement for CD45 (P < 0.0002; Fig. 4 B). The role of CD45 in this process appears to be the activation of the SFK because inhibition of SFK activity by PP2 induced the same phenotype, blocking the production of MIP-1α (P < 0.0005), MIP-1β (P = 0.0025), and RANTES (P < 0.01; Fig. 4 C). Taken together, these findings identify CD45 and the SFK as critical signaling molecules in the production of chemokines from NK cells.

CD45-exon 6–deficient mice were previously shown to have normal cytotoxicity against YAC-1 cells (32), which are recognized via NKG2D. To address whether CD45^−/− NK cells could also kill non-NKG2D target cells, we tested their ability to lyse a range of targets. As expected CD45^−/− NK cells efficiently lysed NKG2D ligand⁺ target cells including YAC-1 and RMA-S Rae1b. Interestingly, control and CD45^−/− NK cells were equally efficient at killing CHO cells, a process predominantly mediated through Ly49D (37; Fig. 5, A and B). These data suggest that unlike cytokine production, cytotoxicity triggered via Ly49D and NKG2D is CD45-independent. The MHC class I–deficient cell line RMA-S was killed more efficiently than MHC class I expressing RMA cells by NK cells from both knockout and control mice, indicating that NK inhibitory receptor signaling is intact in the absence of CD45. IL-21 enhanced control and CD45^−/− NK cell killing against all targets to a similar degree, supporting the independence of IL-21 signaling and CD45 (Fig. 2).

As cytokine production after ITAM-containing receptor ligation was dependent on CD45, we investigated whether these same receptors could participate in redirected cytolysis of FcR⁺ P815 cells, a process known to be Syk/Zap70 dependent (20). IL-15–activated control and CD45^−/− NK cells displayed equal basal killing and both responded to ITAM-dependent Ly49D and CD16 engagement with increased killing, most notably at an E/T ratio of 27:1 (Fig. 5 C). It is possible that the initial in vitro expansion in IL-15 altered the activation status of the cells and may therefore not accurately represent the requirement for these ITAM signaling molecules for in vivo lysis. We addressed this issue in a nonmanipulated setting by addressing in vivo killing using BM allograft rejection, whereby BALB/c BM (H-2d) was transferred into irradiated control or CD45^−/− B6 (H-2b) mice. Antibody depletion studies have previously shown that H-2d rejection is dependent on Ly49D⁺ NK cells (39). We observed that both C57BL/6 and CD45^−/− mice possessed significantly fewer splenic colonies when H-2d BM grafts are transferred compared with isogenic H-2b BM grafts (P = 0.01; Fig. 5 D) indicating that rejection of H-2d BM by Ly49D⁺ NK cells occurs independently of CD45. Similarly CHO killing is only slightly impaired when NK cells are treated with PP2 (Fig. 5 E), whereas PI3K inhibitors have a much greater effect on lysis supporting the notion that Ly49D-mediated lysis is independent of CD45 and SFKs. Taken together, these findings demonstrate that both in vivo and in vitro natural cytotoxicity of a variety of targets by NK cells occurs independently of CD45. The increased number of NK cells in CD45^−/− mice may alter the E/T ratio in vivo, masking potential differences in killing efficiency. However, since CD45-deficient mice have substantially greater spleen mass (unpublished data), target cell numbers may also be elevated, making accurate in vivo E/T calculations difficult.

As some NK cell functions were impaired after Ly49D and CD16 activation in the absence of CD45, we investigated signal propagation in these cells. The total tyrosine phosphorylation (Fig. 6 A) differed considerably between CD45^−/− and control IL-15–expanded NK cells. We note the virtual absence of phosphoproteins of around 95 and 105 kD in size and a dramatic reduction of a phosphorylated band at ∼50-kD protein in the absence of CD45 (Fig. 6 A). Furthermore, in contrast to controls, Ly49D stimulation failed to increase phosphorylation of 97-, 68-, and 58-kD proteins in CD45^−/− NK cells. As protein tyrosine phosphorylation was not enhanced in CD45^−/− NK cells we assessed the activation status of Syk, it being the dominant Ly49D/DAP12-induced protein tyrosine kinase (PTK) in NK cells (21, 40). Consistent with earlier studies (40), Ly49D stimulation of control NK cells induced a large increase in phosphorylated Syk after 10 min (Fig. 6 B). However, very little phosphorylated Syk was detected in lysates from equivalently treated CD45^−/− NK cells (Fig. 6 B). CD16 stimulation similarly failed to activate Syk in CD45^−/− NK cells (unpublished data) suggesting a general role for CD45 in activating Syk after ITAM activation in NK cells. Zap70 is also expressed in NK cells however its role in Ly49D signaling is largely unexplored. We found that unlike Syk, phosphorylation of Zap70 was not enhanced by Ly49D ligation in both control and CD45^−/− NK cells (Fig. 6 C). This was not due to inefficient stimulation, as tyrosine phosphorylation of a number of other proteins was induced when whole cell lysates were examined. Ly49D stimulation is known to activate PLCγ1 resulting in an increase of intracellular calcium (40). We observed such calcium flux in control NK cells after Ly49D stimulation but saw no change in the intracellular calcium level in identically treated CD45^−/− NK cells (Fig. 6 D). CD45^−/− NK cells were capable of releasing calcium after ionomycin treatment, indicating that the defect in CD45^−/− NK cells was specific for the Ly49D pathway (Fig. 6 E). These finding show CD45 is essential for full Syk phosphorylation and calcium flux after activation of ITAM-containing receptors.

CD16 stimulation of human NK cells results in SFK- and Syk-dependent p38 phosphorylation and enhances cytotoxicity and IFN-γ mRNA accumulation, processes that are sensitive to inhibition of MAPK p38 (41, 42). Furthermore, c-Jun NH₂-terminal kinase (JNK) is phosphorylated after NK cell interaction with immune complexes (42) and is also strongly implicated in coupling surface receptor induced signals to IFN-γ production in T cells (43, 44). To investigate whether the impaired ITAM-mediated IFN-γ production in CD45^−/− NK cells was associated with reduced activity of the MAPKs, we assessed the phosphorylation state of p38 and JNK after Ly49D ligation. Indeed, CD45^−/− NK cells showed reduced phosphorylation of p38 after a 2- and 10-min stimulation with impaired JNK phosphorylation observed at 10 min, confirming that Syk is important in coupling surface receptor signals to the activation of JNK and p38 (Fig. 7 A). Interestingly, despite inducing IFN-γ and augmenting lysis of P815 cells, NK1.1 stimulation did not activate p38, raising the possibility that p38 is specifically downstream of ITAM-containing receptors in primary NK cells. Analysis of MAPK-Erk1/2 phosphorylation was also performed and whereas total Erk1/2 was readily observed in primary NK cells, we were unable to detect phosphorylation of these proteins after Ly49D stimulation (unpublished data).

The ability of Ly49D on CD45^−/− NK cells to recognize and lyse CHO cells implies that Ly49D/DAP12 stimulation activates signaling proteins required for cytolysis. Pharmacological studies have highlighted the importance of the PI3K pathway in natural killing (14, 20, 45) and that this pathway can be activated in an ITAM-independent manner (10, 18). To assess whether the PI3K pathway is activated after Ly49D/DAP12 stimulation, we measured the level of phospho-Akt in lysates of Ly49D-stimulated NK cells. We observed a striking increase in phospho-Akt after 10 min of Ly49D stimulation in both control and CD45^−/− NK cells (Fig. 7 B). This finding suggests that, like the NKG2D pathway, Ly49D may also activate the PI3K pathway and that Akt is less reliant than MAPK JNK/p38 on CD45-mediated SFK and Syk activity for phosphorylation after Ly49D stimulation. The guanine nucleotide exchange factor Vav1 is also implicated in natural cytotoxicity, as it is phosphorylated after both CD16 and NKG2D-DAP10 stimulation (21) and Vav1^−/− NK cells have reduced natural cytotoxicity against NKG2D-ligand⁺ target cells (46, 47). We found levels of phospho-Vav1 to be increased after Ly49D stimulation of control NK cells but almost undetectable in CD45^−/− NK cells irrespective of stimulation (Fig. 7 C). Consistent with CD45 mediating its effects through the SFK, control NK cells treated with PP2 not only failed to increase Vav1 phosphorylation after stimulation, but actually showed decreased levels of phospho-Vav1 compared with untreated NK cells (Fig. 7 C). Taken together, CD45 appears to be differentially required for the optimal activity of signaling molecules downstream of Ly49D, with normal Akt phosphorylation being observed in the presence of impaired Syk, Vav1, and p38 and JNK phosphorylation in CD45-deficient NK cells.

This study has identified CD45 as pivotal in signal transduction from a prototypic ITAM-containing receptor on NK cells, determining distinct functional outcomes from similar ligand interactions. We show for the first time in NK cells that modulation of a cell surface molecule can determine the nature of the response. CD45^−/− NK cells fail to produce IFN-γ, GM-CSF, MIP-1α/β, and RANTES when ITAM-containing surface receptors are stimulated, a process that normally results in robust production of these soluble factors. Another cellular consequence from engagement of Ly49D on NK cells is the release of lytic granules capable of killing engaged cells. In contrast to cytokine production, NK cells devoid of CD45 efficiently killed cells expressing ligands for Ly49D. The ability of SFK inhibitors to block cytokine/chemokine production from CD16-, NKG2D-, or Ly49D-stimulated NK cells but only minimally effect cytotoxicity mimics the CD45-null phenotype and strongly suggests that CD45 lies upstream of the SFK and that its activity is differentially required for Ly49D responses. Furthermore, the impaired Ly49D- and CD16-mediated phosphorylation of Syk in CD45^−/− NK cells suggests that CD45-mediated SFK activity is important for the recruitment of Syk to phosphorylated ITAMs and/or the phosphorylation of recruited Syk. An important question from our findings is how does engagement of the same receptor lead to a bifurcation of signaling pathways that differ in the requirement for CD45?

The failure of CD45^−/− NK cells to secrete IFN-γ after CD16, NKG2D, or Ly49D stimulation is consistent with the phenotype of Syk^−/−/Zap70^−/− mice (14, 20) and with the pharmacological inhibitors of Syk being able to impair Ly49D-mediated IFN-γ production (38). Impaired Syk activity in Ly49D-stimulated CD45^−/− NK cells coincides with hypophosphorylation of Vav1 and JNK and a failure to release intracellular calcium. The involvement of Vav1 in this process, however, is unclear as CD16-mediated IFN-γ production has previously been shown to occur independently of Vav1 (46). In contrast, calcium and JNK are likely to be involved in this process. Augmented IFN-γ production by T cells after CD158j ligation strongly correlated with increased phospho-JNK (44), while exogenous HIV-1 Tat, which blocks calcium influx can inhibit the production of IFN-γ after interactions with dendritic cells (48). These findings define a signal transduction pathway showing absolute dependence on CD45-SFK-Syk activity for cytokine production from ITAM-containing receptors (Fig. 8 A), raising the question of how cytotoxicity is differentially regulated.

Signal transduction pathways leading to cytotoxicity have been defined in NK cells to some extent. It is clear for example that killing through the Ly49D receptor requires activation of Syk/Zap70 as mice deficient in both fail to lyse CHO cells (20). Similarly, DAP12-deficient NK cells fail to induce cytotoxicity after Ly49H-mediated recognition of m157-expressing cells (3) and analysis of NK cell lines has shown DAP12 to activate the PI3K pathway in a Syk-dependent manner (49). Although there is some evidence that Zap70 can compensate for an absent Syk (50), the presence of a nonfunctional Syk is dominant, suggesting that Zap70 normally has little role in the induction of cytotoxicity (21, 40). We noted an equal increase in phospho-Akt, a direct product of PI3K activation, in Ly49D-stimulated CD45^−/− NK cell lysates compared with control lysates. This finding identifies for the first time that Ly49D activates the PI3K pathway in primary NK cells and that this signaling pathway can function independently of CD45 and in the absence of normal levels of Syk activity (Fig. 8 A). Interestingly, Ly49D-mediated IFN-γ production has been shown to be sensitive to pharmacological inhibitors of PI3K, Syk, and SFK (38) suggesting that multiple pathways contribute to IFN-γ production after Ly49D stimulation. Even though our findings support roles for Syk and SFK activity in IFN-γ production, we failed to find any evidence for PI3K involvement in this process. The basis of this difference is currently unclear but may reflect differences in the derivation of the NK cells or the assay conditions.

Whereas it is apparent that cytotoxicity after NKG2D ligation requires PI3K recruitment to the YxxM motif of DAP10 (21) and activation, processes which occur independently of Syk and Zap70 (21) and calcium release (15), DAP10 is not recruited to Ly49D after its ligation (10, 18, 51) and is therefore an unlikely mediator of the resultant cytotoxicity. It is, however, interesting to note that both NKG2D/DAP10 (18) and Ly49D/DAP12 activate PI3K, placing this lipid kinase centrally to both cytotoxic signaling pathways.

Recently, DAP12-mediated cytotoxicity in contrast to NKG2D/DAP10-mediated killing was shown to be largely Vav independent (47). Our data support the notion that Vav1 is not required for DAP12-mediated CHO lysis, and suggest that Syk lies upstream of Vav1. We have not addressed the phosphorylation state of Vav1 after NKG2D ligation, although recent findings (47) would predict it to be normal since CD45^−/− NK cells efficiently kill YAC-1 and RMA-S Rae1β cells.

It is puzzling to note that although both DAP12 and DAP10 appear to require tyrosine phosphorylation for the recruitment and activation of kinases (21, 40, 52), only DAP12-mediated pathways appear affected in CD45^−/− NK cells with DAP10-mediated killing intact, a phenotype also seen in the Syk/Zap70^−/− mice. Furthermore, the intact NKG2D-mediated killing by primary CD45^−/− and Syk/Zap70^−/− NK cells contrasts with their failure to secrete IFN-γ via NKG2D stimulation and suggests that the requirement of SFKs and Syk in DAP10 phosphorylation is less stringent than that of DAP12. The finding that PP2 treatment only mildly affects YAC-1 killing (20) supports this. Furthermore, it appears that cytotoxicity is only completely blocked if PI3K, SFK, and Syk are inactivated in combination (20), highlighting the importance of natural killing and thus the numerous mechanisms leading to its triggering. Collectively this evidence suggests a signaling cascade after Ly49D ligation that involves PI3K but is distinct from the DAP10-mediated pathway, which can function independently of Syk/Zap70 (20).

Our observation that CD45^−/− NK cells differentially activate cytotoxicity and cytokine production after stimulation through prototypical ITAM-containing receptors adds significantly to the understanding of these phenomenon by associating this dichotomy with discrete changes in the activity of signaling molecules. These results suggest two possible models for the induction of effector functions after ITAM receptor stimulation. In one there is a bifurcation of signaling with one pathway running through PI3K to cytotoxicity and the second through Vav1/Ca²⁺/JNK to cytokine production. The alternative proposal is that cytotoxicity and cytokine production are triggered at different signaling thresholds.

Ly49D can stimulate the phosphorylation of Akt, a critical substrate in cytotoxicity (16, 20, 45), with this pathway showing an absolute requirement for Syk/Zap70 (20). Reduced Ly49D-mediated phospho-Syk and basal phospho-Zap70 in CD45^−/− NK cells does not affect the efficiency of CHO killing, supporting the notion that the small amount of PTK activity in CD45^−/− NK cells, as apposed to none in Syk^−/−/Zap70^−/− cells, is above the threshold necessary to activate cytolysis but not cytokine production (Fig. 8 B). It is also possible that the numerous receptor interactions between an NK cell and its target may activate Syk in a CD45-independent manner. These more extensive interactions may also induce Ca²⁺ flux in CD45^−/− NK cells, thereby fulfilling the apparent requirement for Ca²⁺ in killing (53). The failure of Ly49D mAb to induce Erk1/2 phosphorylation, whereas NK cells incubated with CHO cells displayed an increase in Erk1/2 phosphorylation (47), may support this notion. We have shown that Ly49D ligation activates Syk in a manner that is largely CD45 dependent, but have yet to examine the activation of Syk in the context of cell–cell interactions.

CD45 is essential for B and T cell antigen receptor signaling, development and maturation (26). However this is not the case for NK cells. Peripheral expansion of CD45^−/− NK cells was enhanced in vivo, a phenomenon specific to the spleen, cell intrinsic and independent of IL-15 responsiveness. Furthermore, CD45^−/− NK cells express all the surface receptors characteristic of full maturation (35). The reduced percentage of Ly49D⁺ NK cells in vivo was particularly interesting since this occurs in Syk^−/−/Zap70^−/− mice (20) and suggests that CD45 activation of SFK and Syk may be involved in positive signaling through Ly49D influencing the survival or selection of Ly49D⁺ cells in vivo. CD45 has also been identified as a Jak phosphatase implicated in the negative regulation of cytokine signaling (54). The normal response of CD45^−/− NK cells to IL-15, -12, -18, and -21 in vitro excludes cytokine hyperresponsiveness as an contributing factor in the enhanced NK cellularity in CD45^−/− mice.

Our findings suggest a model of the biochemical pathways activated in NK cells after interactions between ligands such as H-2d, m157, Rae1β, and Ig-Fc fragments and ITAM-containing receptors such as Ly49D, Ly49H, NKG2D, and CD16 (Fig. 8 A). Upon ligand binding, receptors are drawn together recruiting ITAM-containing adaptor molecules such as DAP12. CD45 then activates the SFKs allowing phosphorylation of the ITAM and the recruitment of PTKs (Syk/Zap-70) via their SH2 domain binding to phosphorylated ITAM, where they may be phosphorylated by the SFKs. Syk can now activate Vav1 and signaling pathways required for the production of cytokines/chemokines including the release of calcium and the activation of the MAPKs, JNK, and p38. In contrast, PI3K activation after ITAM phosphorylation requires only a small enhancement of Syk/Zap70 activity and is responsible for NK cell cytotoxicity. This model postulates segregated signaling pathways dictated by their dependence on CD45-SFK–mediated activation of Syk with the production of cytokines/chemokines absolutely dependent on CD45. In contrast cytotoxicity is functional in the absence of CD45. This finding has important implication for NK cell effector functions during development and differentiation, as CD45 expression on NK cells appears to change depending on maturation and activation state of the cell (32). Furthermore, it may be possible to regulate these functions specifically by targeting CD45 on the surface of NK cells.

C57BL/6 and CD45^−/− (gift from Dr. V. Tybulewicz (National Institute for Medical Research, Mill Hill, UK; reference 33) mice were bred and maintained at The Walter and Eliza Hall Institute of Medical Research. CD45 mutation was backcrossed at least eight times onto the C57BL/6 background before being made homozygous and maintained by brother–sister mating. All mice were used between 8 and 20 wk of age. Competitive BM chimeras were generated by i.v. transfer of 2 × 10⁶ CD45^−/− and C57BL/6 bone marrow cells to C57BL/6 recipients after two doses of irradiation at 5.5 Gy. Animal experimentation was in strict accordance with protocols approved by the Royal Melbourne Hospital Animal Ethics Committee.

Antibodies specific for NK1.1 (PK136), Ly49A (YE132), Ly49C/I (SW5E6), Ly49D (4E5/E1), Ly49G2 (4D11), CD16 (24G.2), B220 (RA3-6B2), Ly5.2 (ALI 4A2), Ly5.1 (A20.1), CD3 (KT3-1.1), CD11b (Mac-1; M1/70), Gr-1 (RB6-8C5), TCR-β (H57-5921), IFN-γ (HB170 and XMG1.2), GM-CSF (MP1-22E9 and MP1-31G6), and IL-10 (JES5-2A5.1) were prepared as described previously (36). Antibodies against CD94 (18d3), DX5, NKG2A/C/E (20d5), NKG2D (CX5), CD244 (SW2B4), Ly49F (HBF-719), CD8 (56–7.3), BrdU (B44), CD122 (TM-1), IL-10 (SXC-1), and CD49b (HM2) were purchased from (BD Biosciences). Anti-Ly49H (3D10) (55) was a gift from Dr. A. Scalzo (University of Western Australia, Nedlands, Australia). Antibodies used in immunoblot analysis and immunoprecipitation were; anti-phosphotyrosine and anti-Akt (Upstate Biotechnology); anti-Vav1 (blotting), anti-actin, and anti-Syk (Santa Cruz Biotechnology, Inc.); anti–phospho-P44/42 MAPK (Thr202/Tyr204), anti–phospho-Akt (Ser473), anti-Vav1 (IP), anti–phospho-P38 (Thr180/Try182), anti-P38, anti–Zap-70 (blotting), anti–phospho-Jnk (Thr183/Tyr185), and anti-JNK (Cell Signaling Technology); anti–phospho-Src (Tyr418; Biosource); anti–Zap-70 (IP; Caltag Laboratories); anti–rabbit-HRP, anti–goat HRP (Silenus); anti–mouse IgG1, -2b, and -κ (Southern Biotechnology Associates, Inc.). Single cell suspensions were prepared by forcing organs through metal sieves. Liver lymphocytes were isolated from a 40–80% Percoll gradient (Amersham Biosciences) centrifuged for 20 min at 2,500 rpm. For flow cytometry, single-cell suspensions were stained with the appropriate mAb in PBS containing 2% FCS. Biotinylated mAb were revealed by Cy5- or PE-streptavidin (Southern Biotechnology Associates, Inc.). Cells were analyzed on an LSR (BD Biosciences) with dead cells being excluded by propidium iodide staining. Cell sorting was performed using FacsDIVA (BD Biosciences) or MOFLO (Cytomation) to >98% purity.

Sorted NK cells were cultured in Nunclon^TM six-well flat bottom tissue culture plates (Nunc) at 10⁶ cells/ml in IMDM supplemented with 10% FCS and 50 ng/ml Gentamycin (Sigma-Aldrich). Cells were typically expanded for 5–7 d in 50 ng/ml recombinant human IL-15 (R&D Systems). For cytokine/chemokine production, MULTIWELL^TM 48-well tissue culture plates (BD Labware) were coated with 20 μg/ml of the indicated mAb overnight in PBS at 4°C. NK cells were washed extensively and cultured for 24 h at 10⁵ cells/well in 0.5 ml of IMDM with 10% FCS. For NK–CHO cocultures, 5 × 10⁴ purified NK cells were incubated overnight with either 5 × 10⁴ or 5 × 10⁵ CHO cells. Supernatants from 5 × 10⁵ CHO cells culture alone were used as negative controls. Cytokines used included: IL-12 (2 ng/ml), IL-18 (10 ng/ml), and IL-21 (100 ng/ml; all from R&D Systems). IFN-γ and IL-10 production was determined by ELISA as described previously (36). GM-CSF was detected by sandwich ELISA using MP1-22E9 as a capture Ab and MP1-31G6-biotin as a detection Ab. RANTES, MIP-1α, and -β Quantikine kits (R&D Systems) were used according to the manufacturer's instructions. Cell proliferation was performed by culturing 5 × 10⁴ cells/well in 96-well, flat-bottom plates with a titration of IL-15 (0.5–50 ng/ml) for 48 h with or without IL-21. Plates were then pulsed with 1 μCi of [³H]thymidine (NEN Life Science Products) and harvested after 8 h onto glass fiber filters (Packard instrument Co.) and incorporation was determined by scintillation counting. The cytotoxicity of freshly isolated and in vitro expanded NK cells were measured using a 4-h ⁵¹Cr-release assay described previously (56). The YAC-1, CHO, P185, RMA, RMA-S, and RMA-S-Rae1β tumor cell lines were loaded with 200 μCi of ⁵¹Cr (ICN Pharmaceutical). Reverse ADCC assays were performed exactly as described previously (20). The pharmacological inhibitors PP2, Ly294002, and SU6656 (Calbiochem) were solubilized in DMSO and incubated with NK cells for 30 min at 37°C before culturing. PP3 and DMSO were used as controls.

Rejection of mismatched BM were performed by i.v injection of 2 × 10⁶ BALB/c (H-2^d) or 10⁵ C57Bl/6 (H-2^b) BM cells in PBS into recipients after two doses of 5.5 Gy. After 8 d, spleens were harvested and colonies counted using a dissecting microscope. NK cell turnover was measured by addition of 0.5 mg/ml of BrdU (Sigma-Aldrich) and 2% glucose to drinking water. After 2 wk, organs were harvested, single cell suspensions prepared, stained with mAb directed against cell surface proteins, and fixed as described previously (57). Cells were then resuspended in 50 μg/ml Dnase1 (Sigma-Aldrich) in PBS, 1 mM CaCl₂, and 1 mM MgSO₄ for 30 min at 37°C before being stained with anti-BrdU.

IL-15–cultured NK cells (2–5 × 10⁷) were washed resuspended in PBS/1% FCS. Cells were stained with biotinylated mAbs for 15 min on ice, washed, stimulated with 20 μg/ml avidin, and lysed using RIPA buffer. Cell lysates or precipitated proteins were processed as described previously (58). Samples were separated using 4–20% and 8% gels (Gradipore) and transferred to nitrocellulose membranes (Amersham Pharmacia Biotech). Filters were probe with primary antibody overnight at 4°C, extensively washed, and revealed by incubation with secondary Abs for 30 min. Blots were developed using SuperSignal ECL reagent (Pierce Chemical Co.). Ca²⁺ release was assessed by loading IL-15–expanded NK cells with indo-1 acetoxymethyl (Molecular Probes) as described previously (59). Loaded cells were stained with biotinylated anti-Ly49D in the presence 2.4G2 to prevent FcR binding. After washing, 2 × 10⁶ cells were sampled using a LSR (BD Biosciences) for 30 s to establish baseline fluorescence then avidin was added to 20 μg/ml and the analysis continued for 5 min.

Data were analyzed with Microsoft Excel software applying the two-tailed Student's t test. The null hypothesis was rejected and differences deemed significant when P < 0.05.

We thank V. Tybulewicz, M. Smyth, Y. Hayakawa, F. Colucci, and A. Scalzo for the gift of reagents and helpful discussions. We thank R. Gerl, J. Brady, A. Light, and The Walter and Eliza Hall Institute of Medical Research support staff for technical assistance.

This work was supported by The National Health and Medical Research Council of Australia and The Walter and Eliza Hall of Medical Research Metcalf Fellowship (to S.L. Nutt).

The authors have no conflicting financial interests.