Main

Direct lysis of tumor cells is a key function of natural killer (NK) cells but, until now, the signal events that drive NK lytic function were not well defined. Syk70, Pyk-2, Vav, Rac1 and the mitogen-activated protein kinase–extracellular signal–regulated kinase signaling element (MAPK-ERK) have all been implicated in lysis by NK cells1,2,3,4,5,6. Each of these events is triggered by target ligation in NK cells but the mechanism by which these molecules interact to mediate NK lysis was unknown. We recently found that MAPK-ERK is a critical signal element in driving NK cytotoxicity, through its control of perforin and granzyme B movement toward target cells6.

In T cells, MAPK-ERK activation is triggered by T cell receptor binding to antigen, and is coupled to the pathway that involves GRB-2, Sos, Ras, Raf1 and MAPK kinase (MEK)7,8,9. Association of the GRB-2-Sos complex to the receptor recruits Ras to the plasma membrane, resulting in Raf-1 binding to GTP-bound Ras. Subsequent activation of Raf-1 leads it to phosphorylate MEK-1, which in turn becomes activated as a dual kinase with the ability to activate ERK. It was expected that the same pathway would be involved in NK direct lysis. However, target ligation in NK cells activates a Ras-independent ERK pathway that leads to NK lytic function10. Ras inactivation by pharmacological disruption with the farnesyl transferase inhibitor FTI-277 or by vaccinia virus–mediated delivery of dominant-negative N17Ras had no adverse effect on NK lysis or ERK activation. Such inhibition did not disrupt the mobilization and redistribution of perforin and granzyme B within NK cells towards the contacted target cell. Instead, Ras inactivation effectively blocked interleukin 2 (IL-2)–induced ERK activation suggesting that growth-related function, but not tumoricidal function, may require Ras10. We set out to define the Ras-independent pathway that is activated by target ligation and controls NK cell cytotoxicity.

Results

ERK1, ERK2 and NK lytic function

To investigate the signaling pathways involved in the regulation of NK direct tumor lysis, we tested whether ERK1 and ERK2 (ERK1/2) were activated by conjugation of NK92 cells with Raji target cells. ERK1/2 activation peaked at 2 min, dropping back to basal levels by 10 min. PD98059, a specific MEK inhibitor, effectively suppressed this activation as well as NK lytic function against tumor cells (data not shown). More significantly, PD98059 blocked perforin and granzyme B movement in NK92 cells towards target cells (Fig. 1a,b). These results confirm that ERK1/2 are responsible for the mobilization and redistribution of perforin and granzyme B to effect lysis6.

Figure 1: Suppression of NK92 lytic function by PI3K inhibitors.
figure 1

Blockade of (a) perforin and (b) granzyme B movement by inhibition of PI3K or MEK. NK92 cells pretreated with DMSO, PD98059 (PD, 50 μM), LY294002 (LY, 50 μM) or wortmannin (WM, 50 nM) were mixed with equal numbers of Raji tumor cells for 5 min at 37 °C. The cytospun cells were then stained with (a) FITC–anti-perforin and TRITC–anti-IgM (b) or with FITC–anti-granzyme B and TRITC–anti-IgM. (c) Akt activation triggered by target ligation. NK92 cells rested in IL-2–free medium for 4 h at 37 °C were mixed with fixed Raji tumor cells for 0–30 min at 37 °C. Cell lysates were analyzed by immunoblotting (WB) with antibodies to active Akt (anti-phosph-AKT) and reprobed with antibodies to pan-Akt (anti-panAKT) to check for equal loading. (d,e) Suppression of NK cytotoxicity and Akt activation. IL-2–starved NK92 cells were treated with LY294002 (LY), wortmannin (WM) or DMSO for 30 min at 37 °C. The cells were then divided, with one group tested for lysis of 51Cr-labeled Raji tumor cells and the rest incubated with fixed Raji cells for 5 min at 37 °C, before immunoblotting with antibodies to active Akt. The same membranes were reprobed with antibodies to pan-Akt to check for equal loading. (f) Cell viability. Each group of NK92 cells was examined by the Annexin-V apoptosis assay. (Results represent one of at least five independent experiments.)

Full size image

PI3K and NK lytic function

We investigated whether phosphoinositide-3 kinase (PI3K) was an upstream regulator of ERK in the NK lytic process. We found that target engagement induced expression of PI3K in NK92 cells within 2 min, as determined by Ser473 phosphorylation of Akt (Fig. 1c). Control samples of either media or dimethylsulphoxide (DMSO, which was used as a solvent for the pharmacological inhibitors) had the largest lytic capacity, whereas inhibition of PI3K by LY294002 or wortmannin suppressed NK lysis of tumor cells in a dose-dependent pattern (Fig. 1d,e). All doses effectively suppressed Ser473 phosphorylation of Akt (Fig. 1d,e). Inhibition of PI3K also blocked the movement of perforin and granzyme B towards target cells (Fig. 1a,b). NK cell viability was intact in all groups, excluding the possibility that the nonspecific toxicity of these pharmacological reagents might result in loss of NK function (Fig. 1f).

PI3K controls MEK1/2 and ERK1/2

Because the inhibitory effects of LY294002 and wortmannin on NK lysis and perforin and granzyme B movement were similar to those of PD98059, we determined whether PI3K inhibition could interfere with ERK and the upstream regulator of ERK, MEK. Immunoblotting (western blotting) of whole NK92 cell lysates after target engagement demonstrated that both LY294002 and wortmannin treatment significantly inhibited ERK1/2 and MEK1 and MEK2 (MEK1/2) phosphorylation in a dose-dependent manner (Fig. 2a,b). This inhibition correlated with impaired lysis seen in these cells. ERK1/2 and MEK1/2 phosphorylation and kinase activity were blocked by treatment with 50 μM of LY294002 or 50 nM of wortmannin (Fig. 2c,d). These results revealed that PI3K regulates ERK 1/2 through MEK1/2 and that this pathway controls NK lysis.

Figure 2: Regulation of ERK1/2 and MEK1/2 activation by PI3K after target engagement.
figure 2

(a,b) Inhibition of ERK1/2 and MEK1/2 activation by LY294002 or wortmannin. Aliquots of NK92 cell lysates, as used in Fig. 1d,e, were immunoblotted with antibodies to active MAPK (anti-AMAPK) or antibodies to active MEK (anti-AMEK) and then reprobed with antibodies to pan-ERK to check for equal loading. (c) Comparison of the effects of LY294002 and wortmannin versus PD98059 on ERK1/2 and MEK1/2 phosphorylation. NK92 cells pretreated with LY294002, wortmannin, PD098050 or DMSO for 30 min at 37 °C were incubated with fixed Raji tumor cells for 5 min at 37 °C, before sequential immunoblotting with antibodies to active MAPK, active MEK and pan-ERK. (d) Comparison of the effects of LY294002 and wortmannin versus PD98059 on ERK1/2 and MEK1/2 kinase activities. ERK1/2 and MEK1/2 immunoprecipitated (IP) from the whole cell lysates used in c were analyzed using in vitro kinase assays with myelin basic protein (MBP) as a substrate. After autoradiography, the same membrane was probed with antibodies to pan-ERK or pan-MEK to check for equal loading. (Results represent one of at least five independent experiments.)

Full size image

Rac1 regulates NK cytotoxicity

In light of the reported association of Rac1 with NK lysis4, we examined Rac1 function in NK92 cells, as it has also been linked to exocytosis of secretory granules in mast cells11. We found that vaccinia virus vector–mediated introduction of dominant-negative Rac1 (N17Rac1), but not an irrelevant cDNA (CD56), into NK92 cells inhibited NK92 lysis of tumor cells markedly (Fig. 3a). However, expression of dominant-negative Rho or Cdc42 had minimal effects on NK function (data not shown). The viability of NK92 cells was similar for all groups (Fig. 3b).

Figure 3: Rac1 regulation of NK cytotoxicity.
figure 3

(a) Effect of N17Rac1 and V12Rac1 on NK cytotoxicity. Dominant-negative Rac1 (N17Rac1), constitutively active Rac1 (V12Rac1) and a CD56 control were introduced into NK92 cells by recombinant vaccinia virus–mediated gene transduction, followed by testing for 51Cr-labeled Raji tumor cells. (Results represent one of at least four independent experiments.) ( b) Cell viability. Each group of NK92 cells was examined by the Annexin-V apoptosis assay.)

Full size image

Rac1 and MEK and ERK

Because of the importance of ERK in NK lysis5,6, we inspected target cell–triggered ERK and MEK activation before and after N17Rac1 or V12Rac1 expression in NK92 cells. After target engagement, immunoblotting of whole NK92 cell lysates revealed that recombinant Rac1 was expressed in the appropriate samples (Fig. 4a). N17Rac1 blocked both ERK1/2 and MEK1/2 activation triggered by target cell ligation. In contrast, dominant-negative Rho and Cdc42 did not interfere with target cell–induced ERK1/2 or MEK1/2 activation (data not shown). Correspondingly, in vitro kinase assays showed that N17Rac1 inhibited the kinase activities of MEK1/2 and ERK1/2 stimulated by target signals (Fig. 4b). In contrast, constitutively active V12Rac1 stimulated both phosphorylation and kinase activities of MEK1/2 and ERK1/2 even in the absence of target cells (Fig. 4a,b). These data indicate that Rac1, like PI3K, primarily regulates NK cytotoxicity through MEK and ERK.

Figure 4: Rac1 regulation of ERK1/2 and MEK1/2 activation triggered by tumor cell ligation.
figure 4

(a) Regulation of ERK1/2 and MEK1/2 phosphorylation by Rac1. The NK92 cells used in Fig. 3a were incubated with fixed Raji tumor cells for 5 min at 37 °C before sequential immunoblotting with antibodies to Rac1, active MAPK, active MEK and pan-ERK. (b) Regulation of ERK1/2 and MEK1/2 kinase activities by Rac1. ERK1/2 and MEK1/2 immunoprecipitated from the cell lysates used in a were analyzed by in vitro kinase assays using MBP as a substrate. After autoradiography the same membrane was probed with antibodies to pan-ERK or pan-MEK. (Results represent one of at least four independent experiments.)

Full size image

Rac1 acts downstream of PI3K in MEK pathway

We speculated that PI3K and Rac1 lie within the same pathway in NK92 cells and that PI3K regulates NK cytotoxicity through Rac1. Thus we examined whether constitutively active V12Rac1 could rescue NK cytotoxicity blocked by PI3K inhibition. NK92 cells pretreated with LY294002, wortmannin, PD98059 or DMSO were infected with recombinant V12Rac1 vaccinia virus and reincubated with LY294002, wortmannin, PD98059 or DMSO, respectively, before analysis of tumoricidal function. The control groups (mock, DMSO, CD56 and DMSO + CD56) exhibited similarly high lytic capability, whereas PD98059, LY294002 and wortmannin all suppressed NK lysis (Fig. 5a). Infection with V12Rac1 vaccinia virus recovered the suppressed lytic function in LY294002- or wortmannin-treated cells, but did not restore the impaired lysis in PD98059-treated cells (Fig. 5a). Analysis of cell viability demonstrated that LY294002, wortmannin or PD98059 treatment + viral infection did not cause any loss of viability in NK92 cells (Fig. 5b). Because V12Rac could restore the cytotoxicity in LY294002- or wortmannin-treated, but not PD98059-treated NK92 cells, our results indicate that Rac1 is downstream of PI3K but upstream of MEK.

Figure 5: Constitutively active Rac1 in the rescue of NK lytic function and MEK-ERK activities lost by PI3K inhibition.
figure 5

(a) Effect of V12Rac1 on NK lytic function. NK92 cells pretreated for 30 min with LY294002 (50 μM), wortmannin (50 nM), PD98059 (50 μM) or DMSO were infected with recombinant V12Rac1 or CD56 vaccinia virus. Cells were then reincubated with LY294002, wortmannin, PD98059 or DMSO for 2.5 h and tested separately for lytic function against 51Cr-labeled Raji tumor cells. (Results represent one of at least four independent experiments.) (b) Cell viability. Each group of NK92 cells was examined by Annexin-V apoptosis assay. (c,d) Effect of V12Rac1 on MEK1/2 and ERK1/2 phosphorylation and kinase function. NK92 cells from a were incubated with fixed Raji tumor cells for 5 min at 37 °C and lysed. One-sixth of the lysates were sequentially immunoblotted with antibodies to active MAPK, active MEK and pan-ERK. The rest were immunoprecipitated with anti-MEK or anti-ERK before analysis using in vitro kinase assays with MBP as a substrate. After autoradiography the same membrane was probed with antibodies to pan-MEK or pan-ERK. (Results represent one of at least four independent experiments.)

Full size image

If Rac is upstream of MEK, V12Rac1 should restore the phosphorylation and kinase activities of both MEK1/2 and ERK1/2, which are lost in LY294002- or wortmannin-treated NK92 cells. Immunoblotting of whole NK92 cell lysates after target engagement demonstrated that V12Rac1 significantly re-elevated the inhibited phosphorylation and kinase activities of MEK1/2 and ERK1/2 in LY294002- or wortmannin-treated, but not PD98059-treated, NK92 cells ( Fig. 5c,d). These results provided additional support for the hypothesis that Rac1 is downstream of PI3K and upstream of MEK and MAPK in signaling pathway that controls NK lytic function.

PI3K, Rac1 and MEK involvement in lysis

To ensure that the newly defined pathway did not only occur in the NK92 cell line and its target Raji, as is used in this study, we validated its existence in freshly isolated NK cells and used two other tumor targets, K562 and 721.2216. Using large granular lymphocytes (LGLs) prepared from the whole blood of a normal donor, we confirmed that wortmannin and LY294002 could inhibit fresh NK cell lysis of K562 tumor cells and, at the same time, block MAPK- and MEK-activation triggered by K562 ligation (Fig. 6a,b). Expression of V12Rac1 in these LGLs restored NK lysis and MAPK-MEK activation, but not in PD98059-treated LGLs. Thus the results from NK92 cells were recapitulated in fresh NK cells. The same experiments done on LGLs using 721.221 tumor target cells yielded similar results. Thus, the PI3K-Rac1-MEK-MAPK pathway is biologically relevant in mediating NK function.

Figure 6: PI3K, Rac1, MEK and ERK in target lysis of freshly isolated NK cells.
figure 6

(a,b) Freshly isolated IL-2–cultured human LGLs pretreated with LY294002 (50 μM), wortmannin (50 nM), PD98059 (50 μM) or DMSO for 30 min were infected with recombinant V12Rac1 or CD56 vaccinia virus. Cells were then reincubated with LY294002, wortmannin, PD98059 or DMSO for 2.5 h before testing for lysis of 51Cr-labeled K562 tumor cells or for MEK and ERK activation by immunoblotting with antibodies to active MEK or active MAPK. (Results represent one of two independent experiments.)

Full size image

p21-activated kinase 1 and NK cytotoxicity

Because a well characterized effector of Rac1, p21-activated kinase 1 (PAK1), can also activate MAPK in epithelial cells and T cells12,13, we explored the role of PAK1 in NK lysis and MAPK activation. Expression of kinase-deficient PAK1 suppressed NK lysis of tumor cells, whereas NK92 cells infected with the control CD56 viral vector remained high in their lytic capacity, equal to that in mock-infected control NK92 cells (Fig. 7a). Examination of whole cell lysates from the same NK92 cells showed that kinase-deficient (KD) PAK1, unlike control CD56, also effectively suppressed MEK and ERK activation by target ligation (Fig. 7b). Thus, PAK1 is upstream of both MEK and ERK in the NK lytic process.

Figure 7: Regulation of MEK1/2, ERK1/2 and NK cytotoxicity by PAK1, under the control of PI3K and Rac1.
figure 7

(a) Regulation of NK cytotoxicity by PAK1. Dominant-negative PAK1 and a CD56 control were introduced into NK92 cells by recombinant vaccinia virus–mediated gene transduction and tested for cytotoxicity using 51Cr-labeled Raji tumor cells. (b) Regulation of MEK1/2 and ERK1/2 activities by PAK1. NK92 cells were incubated with fixed Raji tumor cells for 5 min at 37 °C, and sequentially immunoblotted with antibodies to PAK1, active MAPK, active MEK and pan-ERK. (c) Effect of PI3K inhibition on PAK1 activation. Whole NK92 cell lysates, treated as in Fig. 2c, were immunoprecipitated with anti-PAK1 for analysis of PAK1 activity using an in vitro kinase assay with MBP as a substrate. After autoradiography, the same membrane was reprobed with anti-PAK1. (d) Effects of N17Rac1 and V12Rac1 on PAK1 activity triggered by target cell ligation. NK92 cells, treated as in Fig. 3a, were analyzed for PAK1 activation using an in vitro kinase assay. (e) Ability of constitutively active Rac1 to rescue PAK1 activity lost by PI3K inhibition. NK92 cells, treated as in Fig. 5a, were analyzed for PAK1 activation using an in vitro kinase assay. (Results represent one of at least four independent experiments.)

Full size image

Analysis of whole NK92 cell lysates showed that target cell engagement sharply elevated PAK1 kinase activity in mock-, DMSO- and PD98059-treated cells, but not in LY294002- or wortmannin-treated cells (Fig. 7c). This indicated that PAK1 is under the control of PI3K during this process. To address whether PI3K controls PAK1 via Rac1, we examined PAK1 activity in N17Rac1 or V12 Rac1 recombinant vaccinia virus–infected NK92 cells before and after conjugation with target cells. Introduction of N17Rac1 into NK92 cells totally blocked PAK1 activation triggered by target ligation, whereas introduction of V12Rac1 strongly stimulated PAK1 activity even in the absence of target cells (Fig. 7d). Thus, PAK1 is also subject to Rac1 regulation in NK cells.

PAK1 functions upstream of MEK1/2

To address where PAK1 occurs in the PI3K-Rac pathway, PAK1 protein was immunoprecipitated from the whole NK92 cell lysates after target engagement and analyzed for kinase activity. The results showed that V12Rac1 restored PAK1 activation in LY294002- or wortmannin-treated NK92 cells (Fig. 7e) to levels that were similar to those in the DMSO control. PD98059 could not inhibit PAK1 function in NK92 cells and V12Rac had no added effect. Thus, PAK1 is downstream of PI3K and Rac1 but upstream of MEK. Therefore, we have demonstrated that Rac1 and PAK1 sequentially function upstream of MEK, but downstream of PI3K, in the signaling cascade that drives NK cytotoxicity.

Discussion

We have previously shown that MAPK is a Ras-independent critical regulator in the NK cell lysis of tumor cells6,10. We show here, using biological, biochemical and gene-manipulation methods, that PI3K is the key upstream regulator of MAPK-ERK in the NK lytic process. Target ligation in NK cells caused a short but marked burst of PI3K activation. Wortmannin- or LY294002-inhibition of PI3K in NK cells blocked ERK activation and perforin–granzyme B movement towards target cells, as well as tumor lysis.

We determined that Rac1, but not Rho or Cdc42, acted downstream of PI3K to regulate PAK1, MEK1/2 and ERK1/2 activation, which critically controls NK lytic function. Dominant-negative Rac1 inhibited PAK1, MEK1/2 and ERK1/2 activation as well as NK lytic capacity whereas constitutively active Rac1 restored all these activities. Constitutively active Rac1 could reverse the inhibitory effects of wortmannin and LY294002 but not PD98059 in NK cells. These results were reproducible using freshly isolated human NK cells against two known tumor targets. Dominant-negative PAK1 suppressed NK lytic function as well as MEK1/2 and ERK1/2 activation, pinpointing its essential role upstream of MEK. These results document a pivotal role for PI3K in the directional activation of Rac1→PAK1→MEK1/2→ERK1/2 function, which is coupled to lytic granule mobilization and redistribution, resulting in target lysis.

PI3K regulates a variety of cellular processes including reorganization of the actin cytoskeleton14, induction of cellular survival and protection from apoptosis15,16, antigen and IL-2 growth factor–mediated activation in T cells17,18, membrane ruffling in T cells19 and chemotaxis in neutrophils20. Given this diversity, it is not surprising that PI3K is linked to many signal pathways. In actin reorganization PI3K is closely associated with the Rho family GTPases14, whereas for cell survival it is well documented that the serine and threonine kinase Akt functions downstream of PI3K15,16. In integrin signaling, activation of Raf1, MEK and ERK together with Akt is dependent on PI3K21. In T cells, CD5 costimulation requires PI3K to activate guanine nucleotide exchange factor (GEF), Vav and Rac117, whereas IL-2 receptor triggering can activate PI3K-mediated protein kinase B (PKB) and p70 S6 kinase18. In several systems—including integrin cross-linking in fibroblasts21, CD3–IL-2 activation in T cells22 and chemokine activation in neutrophils20—PI3K is associated with MAPK activation. However, in many cases these studies mainly report signal events without linking them to function. Among the available pathways described for PI3K, the results presented here illustrate that the Rac1→PAK1→MEK1/2→ERK1/2 pathway is the one that PI3K utilizes to drive lytic function in NK92 cells through granule mobilization.

Rac1 has been primarily demonstrated to mediate PI3K regulation of cytoskeleton rearrangement and/or membrane ruffling induced by growth factors and cytokines19,23. Our data indicate that Rac1, acting downstream of PI3K, critically controls yet another function, tumoricidal capacity in NK cells. However, although Rac1 has so far been closely associated with c-Jun N-terminal kinase (JNK) signal transduction14,24,25, our results show that Rac1 utilizes ERK for target-induced signaling. There is further supporting evidence that Rac can activate ERK, as shown in fibroblasts and epithelial cells26,27. In addition it has been reported that Vav, the GEF for Rac1, binds directly to and becomes activated by the products of PI3K, which then directly activate Rac28. Thus it is easily understandable why Vav, together with Rac, has also been implicated in NK cytotoxicity4.

We have attempted to define the effects of N17Rac1 and V12Rac1 on perforin–granzyme B movement in NK92 cells by using immunostaining. So far, pharmacological inhibitors such as LY294002, wortmannin and PD98059 have shown the most consistent ability to disrupt granule polarization. Viral infection itself tended to cause changes in cytoskeletal organization, so that it was difficult to interpret the data. The use of overexpressed proteins can also produce undesired effects, for example, unwanted signal events triggered by cell apoptosis, particularly if the multiplicity of infection is higher than that which can be sustained by a cell while it is still deemed to be healthy. For this reason, every batch of virus vector is carefully titered and monitored for optimum expression without cell toxicity. Mast cells, like NK cells, have secretory granules and a recent report indicated that Rac controls granule release in these cells11. Whether the same pathway described here for NK cells operates in mast cells for granule mobilization needs to be determined. In addition to Rac, Syk70 has been associated with NK lysis1. It is possible that Syk70 interacts with the PI3K pathway that we have described here, based on recent reports that Syk70 and PI3K can crosstalk29,30.

PAK1, the well defined Rac1 effector, has been implicated in mediating cytoskeletal reorganization, such as the formation of pseudopodia, membrane ruffles and phagocytic cups in human neutrophils31 and cellular motility of fibroblasts32. Using kinase-deficient PAK1 expression, we have demonstrated here that PAK1 acts directly upstream of MEK in NK cells and have documented a new role for PAK1 in controlling lytic function against tumor cells.

Overall, these results establish a pivotal role for PI3K in the regulation of NK cytotoxicity, and identify the pathway of PI3K→Rac1→PAK1→MEK1/2→ERK1/2 as controlling this process. Although links between some of these molecules have been recognized in other systems, none have probed the exact sequence of events that lead to lytic function. We have now identified signal transducers and linked them together sequentially in a signaling cascade in NK cells leading to the lysis of tumor cells. This lytic process is driven by a MAPK-ERK pathway that is Ras-independent but Rac1-dependent.

Methods

Cells.

NK92 cells, provided by H. G. Klingeman, Terry Fox Laboratory, Vancouver, Canada, were maintained in α-minimum essential medium supplemented with 20% fetal calf serum, 100 U/ml of IL-2 and 5×10−5 M 2-mercaptoehanol. LGLs possessing high NK activity were freshly isolated from normal donor blood as previously described and cultured in IL-2–containing medium for 3 days before testing6. The NK target cells, Raji, K562 and 721.221, were cultured in RPMI1640 media containing 10% fetal calf serum.

Cytotoxicity assay.

A 51Cr-release assay was done as previously described, using Raji, K562 or 721.221 tumor cells as targets for NK92 or primary NK-LGL effector cells6. Briefly, 51Cr-labeled target tumor cells were added for 4 h at 37 °C to effector cells (5×103 cells/well in 96-well round-bottom microplates), resulting in effector:target ratios of 50:1 to 2.5:1. The percentage of specific 51Cr release was determined by the equation: (experimental cpm – spontaneous cpm)/total cpm incorporated × 100. All determinations were in triplicate, and the s.e.m. of all assays were calculated and were typically around 5% of the mean or less.

Detection of ERK, MEK, PAK1 and AKT phosphorylation and kinase activities.

Antibodies to phosphorylated ERK1/2 (Thr202/Tyr204), MEK1/2 (Ser217/221) and PKB (Ser473) were from New England Biolabs Inc. (Beverly, MA). Pan-ERK and pan-MEK were from Transduction Laboratories (Lexington, KY). Polyclonal antibodies to ERK1, ERK2, MEK1, MEK2, Rac1 and PAK1 were from Santa Cruz Inc. (Santa Cruz, CA). The MEK inhibitor PD98059 and the PI3K inhibitors LY294002 and wortmannin were purchased from Calbiochem (La Jolla, CA). The Annexin apoptosis kit was purchased from R&D Systems (Minneapolis, MN). Cell lysates prepared from IL-2–rested NK cells, which had been incubated with the pharmacological reagents or infected with the relevant recombinant vaccinia virus vectors, were exposed to fixed target tumor cells for 0–30 min at 37 °C. After gel electrophoresis, the active forms of MAPK, MEK and Akt were detected with phosphospecific antibodies. Equal loading was assessed by reblotting with antibodies to pan-ERK or pan-Akt (Cell Signaling Technology Inc., Beverly, MA). The immunoprecipitation and in vitro kinase assays were carried out as previously described6.

Immunostaining.

NK92 cells, either untreated or treated with PD098059 (50 μM), LY294002 (50 μM), of wortmannin (50 nM) or DMSO for 30 minutes at 37 °C, were added to Raji cells at a 1:1 ratio. The cells were spun rapidly at 1000 rpm for 1 min in a cold microcentrifuge, then incubated for 0–5 min at 37 °C. The cells were then centrifuged onto a microscope slide and fixed at −20 °C with methanol:acetone (3:1) for 20 min6. Polyclonal rabbit antibodies to human IgM (Sigma, St. Louis, MO) were used to differentiate Raji B lymphocytic cells from NK92 cells. Monoclonal antibodies to human perforin (Endogen/T cell Sciences, Woburn, MA) and anti–granzyme B were used to detect these lytic components from NK92 cells10. Anti-IgM together with anti-perforin (or anti–granzyme B), each diluted to 1:200, were applied to the slide for 1 h. The slides were then washed in PBS and incubated for 25 min with tetramethylrhodomine isothiocyanate (TRITC)-labeled goat antibody to rabbit IgG (Sigma) or fluorescein isothiocyanate-labeled (FITC)–goat antibody mouse Ig (Sigma). After washing in PBS, the slides were covered with coverslips in mounting media of antifade and 4,6-diamidino-2-phenylindole. Immunofluorescence was observed with a Leitz Orthoplan 2 microscope and images were captured by a charge coupled device (CCD) camera with the Smart Capture Program (Vysis, Downers Grove, IL). On each slide, 100 NK92 and Raji conjugates were evaluated for perforin or granzyme B mobilization.

Several controls, NK92 cells alone or Raji tumor cells alone stained with FITC-labeled goat antibodies to mouse immunoglobulin or TRITC-labeled goat antibodies to rabbit IgG, were used to check for nonspecific binding of the secondary antibodies. Nonspecific binding was not detected and the results were omitted from the figures for clarity.

Vaccinia viral delivery of N17Rac1, V12Rac1 and kinase-deficient PAK1.

NK92 cells were incubated with recombinant vaccinia viruses encoding N17Rac1, V12Rac1 or kinase-deficient PAK1, respectively, for 1.5 h at 37 °C in serum-free medium at a multiplicity of infection of five. CD56-expressing vaccinia viral vector was used as a control. The cells were then washed three times and starved in serum-free medium containing 0.5% bovine serum albumin for 2.5 h before the tumoricidal, immunoblotting and in vitro kinase assays. For experiments combining inhibitors and virus vector infection, NK92 cells were pretreated for 30 min at 37 °C with LY294002 (50 μM), wortmannin (50 nM) or PD98059 (50 μM) before virus infection as described above except that the inhibitors were reintroduced for the last 2.5 h of infection.