Functional Dichotomy in Natural Killer Cell Signaling: Vav1-Dependent and -Independent Mechanisms

The protooncogene Vav1 is one of three members of the Vav family of guanine nucleotide exchange factors (GEFs) for the small GTPases of the Rho family 1. Its product, p95^vav1 (or Vav1) contains an array of structural motifs including a Dbl homology domain, a single Src homology (SH)2 and two SH3 domains, a pleckstrin homology domain, and a calponin homology domain, enabling it to interact with different proteins involved in multiple signal transduction pathways 1. Vav1 is expressed preferentially in hematopoietic cells 2 and is rapidly phosphorylated after stimulation of growth factor receptors and Ag receptors on B and T lymphocytes 3,4,5. Upon tyrosine phosphorylation, Vav1 promotes Rac1 and other Rho family proteins to their active GTP-bound state 6,7,8,9,10, thereby mediating vital functions such as survival, differentiation, motility, and cell division 1,11. In lymphocytes, activation of the small Rho family proteins by Vav1 promotes Ag–receptor capping, actin polymerization, and Ag-induced proliferation of B and T cells in vitro and effective T cell selection in vivo 5,12,13,14,15,16,17. Therefore, it is clear that Vav1 is essential for T cell development and for Ag–receptor-induced responses of B and T lymphocytes. In contrast, signaling requirements for NK cell development are not fully appreciated, although soluble growth factors (including Flt3L/Flk2L, stem cell factor [SCF], IL-7, and IL-15, which appear to be the dominant cytokines) and stromal elements in the bone marrow are essential 18,19,20,21,22. Because Vav1 has been reported to be part of the signaling pathways activated by the receptors for Flt3L/Flk2L and SCF 23,24, it may be crucial in early NK cell development.

Vav1 may participate in signaling pathways during the activation of NK cells. Proximal events, including phosphorylation of phosphatidylinositol 3 kinase (PI3K) and phospholipase C (PLC), participate to sustain the rise in intracellular calcium, a necessary second messenger during NK cell effector functions 25. In T cells, Vav1 is a critical transducer of TCR signals to the calcium pathway, and in its absence, T cells fail to initiate IL-2 gene transcription and proliferate 26. Vav1 has the ability to bind and be regulated by the lipid substrates and products of PI3K 27. Vav1 has also been proposed to enhance the production of substrates for PLCγ2 through activation of phosphatidylinositol 4-phosphate 5 kinase 28. In line with this, the PI3K-specific inhibitor wortmannin abolishes antibody-dependent cell cytotoxicity (ADCC; reference 29), and mice deficient for PLCγ1 have decreased natural cytotoxicity 30.

Vav1 is phosphorylated upon contact with tumor targets and upon cross-linking of the sole FcR (FcγRII/III, CD16) expressed by NK cells 31,32 and therefore may be required for NK cell functions. Virally infected cells produce IFN-α/β, which are potent activators of NK cells and have been shown to induce phosphorylation of Vav1 in several hematopoietic lineages, including lymphocytes 33.

Binding of NK cells to tumor cells and stimulation of NK cells via CD16 results in activation of the mitogen-activated protein kinase (MAPK) cascades. One of this cascades, the extracellular signal-regulated kinase (ERK) is part of the intracellular signaling leading to IFN-γ production, granule exocytosis, and cytotoxicity by NK cells 34,35,36. In T cells, Vav1 is required to transduce TCR signals to the ERK pathway 26, thereby mediating gene transcription. By analogy, Vav1 may be involved in the process of IFN-γ production in NK cells. On the other hand, p38 MAPK is also activated upon binding of NK cells to tumor targets and leads to cytokine production and cytotoxicity 37. However, Vav1 does not seem to be an essential regulator either of the p38 MAPKs or of the c-Jun NH₂-terminal kinase (JNK, or stress-activated kinase, SAPK) in T cells 15,16,26.

Inactivation of the Vav1/Rac1 pathway in human NK cell lines has helped address the role of Vav1 in NK cell functions. As such, Vav1 has been shown to be critical for signaling through CD16 and through the receptor(s) mediating natural cytotoxicity 32,38. In these studies, it was reported that natural cytotoxicity and ADCC were both decreased in human NK cell lines in the presence of a dominant-negative form of Rac1 32, or antisense oligonucleotides for Vav1 38. However, whether deficiency in Vav1 will affect NK cell ontogeny or cytokine production has not yet been investigated. Here, we characterize NK cell differentiation in Vav1^−/− mice.

Vav1^−/− mice on the B10.BR background were derived as described 14. Vav1^−/−, B10.BR (wt), Rag2/γc^−/− 39, and Rag2^−/− mice 40 were housed at the animal facility of the Pasteur Institute.

Cell suspensions were prepared from spleen, thymus, liver, and bone marrow. They were depleted of red cells and stained for flow cytometry as described 39. mAbs directly conjugated to FITC, PE, Tricolor (TRI), allophycocyanin, or biotin included mAbs specific for CD3, CD4, CD8, CD11a (LFA-1), CD11b (Mac-1), CD19, CD45R (B220), CD90 (Thy-1.2), CD117 (c-kit), CD122 (IL-2Rβ), CD161 (NK1.1), DX5, 2B4, Ly49A, Ly49C/I, Ly49D, Ly49G2, TCR-α/β, IgM (all from PharMingen), and H-2K^k (Caltag Laboratories). Biotin-conjugated mAbs were revealed by streptavidin–TRI (Caltag). Cells (10⁶) were first incubated with anti-FcγRII/III (hybridoma 2.4G2) for 20 min on ice to avoid unspecific binding to low-affinity FcRs. To purify cells, splenocyte suspensions were stained with mAbs specific for NK1.1–PE and CD3–FITC: CD3⁻NK1.1⁺ NK cells and CD3⁺NK1.1⁻ T cells were sorted using a FACStar™ (Becton Dickinson). The purity of the sorted populations was reproducibly >95%. Alternatively, splenocytes were passed over nylon wool columns to deplete myeloid and B cells and thereafter depleted of T cells by panning on anti–TCR-β (clone H57) coated Optilux^® Petri dishes (Falcon). Cells were then expanded in IL-2 (1,000 U/ml) for 7–10 d.

A standard lung clearance assay was used to test the ability of mice to reject tumors in vivo as described 41. Briefly, RMA-S cells (10⁶) were labeled in 100 μCi of ⁵¹Cr for 1 h at 37°C, and after several washings, 10⁵ cells per mouse were injected intravenously. 4 h later, mice were killed, and the residual radioactivity was measured in the explanted lungs. In a typical experiment, total radioactivity of the injected inoculum (10⁵ cells/200 μl) would be ∼25–50,000 cpm. In a second tumor model, modified from that described in reference 43, RMA-S cells (2.5 × 10⁵) were injected intraperitoneally, and 48 h later the peritoneal exudate cells recovered, counted, and stained with mAbs specific for CD3, NK1.1, and H-2K^k. Residual tumor cells could readily be distinguished by their larger size, CD3 staining, and absence of H-2K^k staining. In the lymphoma metastases model, mice were injected intravenously with H-2–compatible (H-2^k) but class I^low RDM4 cells (10⁴), and survival was scored for up to 14 d. Those mice that survived were killed at day 14 and their livers fixed in Bouin's medium for 48 h. Macroscopic metastases were readily detected by visual inspection. For the infection model, 10⁴ Listeria monocytogenes (L.m.) bacteria (strain LO28) were injected intravenously, and 48 h later the serum was collected from each mouse and the IFN-γ content was quantified by ELISA (Genzyme). Bacterial burden in livers and spleens was determined 48 h after injection as described previously 42.

A standard ⁵¹Cr-release assay was used to measure NK activity in vitro as described 43. Target cells (H-2^a murine YAC-1 thymoma, H-2^b murine macrophage tumor line IC-21, H-2^b murine EL-4 thymoma, H-2^d murine P815 mastocytoma, Con A–activated B10.BR and β2m^−/− blasts, and H-2^b murine lymphomas RMA and transporter associated with antigen processing [TAP]-deficient RMA-S) were labeled with 100 μCi ⁵¹Cr (ICN Pharmaceuticals), and 2.5–5 × 10³ targets were incubated with graded numbers of effector cells in 200 μl of medium for 4 h. For natural cytotoxicity, effector cells were NK-enriched splenocytes (by passage on nylon wool columns) from untreated mice or from mice primed 40 h earlier with 0.2 mg of poly(I:C) (Sigma-Aldrich). NK-enriched splenocytes and FACS^®-purified NK cells were expanded in IL-2 (1,000 U/ml) for 7–10 d and thereafter used as effectors for (a) ADCC using EL-4 cells coated with anti–Thy-1.1 mAb (lysis of EL4 cells without the Ab never exceeded 10% at the highest E/T ratios); (b) “reverse” ADCC (R-ADCC) using FcR⁺ P815 cells in presence or absence of anti-NK1.1, anti-2B4, or anti-CD16 mAbs; (c) MHC inhibition and recognition of “missing self” using Con A–activated B10.BR and β2m^−/− blasts or RMA and RMA-S cells. MHC class I⁺ Con A blasts (wt) or MHC class I⁺ RMA cells were both resistant to lysis (<5%), even at the highest E/T ratios. Radioactivity released into the cell-free supernatant was measured, and the percent specific lysis was calculated as following: 100 × (experimental release − spontaneous release)/(maximum release − spontaneous release).

Sorted NK cells (2 × 10⁴ cells/200 μl) were cultured in flat-bottomed microtiter plates in human IL-2 (1,000 U/ml) and stimulated with murine IL-12 (2 ng/ml; PeproTech) or mAbs specific for NK1.1 (clone PK136, 20 μg/ml), 2B4 (5 μg/ml), CD16 (75 μg/ml), or control anti–Gr-1 mAb (10 μg/ml). Wells were precoated with mAbs (50 μl/well) for 4 h before adding NK cells. After 24 h, the cell-free supernatants were collected and the amount of IFN-γ was quantitated by ELISA (Genzyme). In parallel experiments, FACS^®-purified NK and T cells (2 × 10⁵) were incubated in the presence of brefeldin A (10 μg/ml) for 4 h at 37°C on flat-bottomed microtiter plates precoated with anti-NK1.1 or anti-CD3. Cells were then washed, fixed in 2% paraformaldehyde for 30 min on ice, thereafter stained with anti-IFN-γ–FITC (PharMingen) in 0.5% saponin. As controls, cells were incubated in medium alone or with PMA (50 ng/ml) and ionomycin (1 μg/ml).

IL-2–activated NK cells (2 × 10⁵) were labeled with 1 μg/ml of 2′-7′-biscarboxyethyl-5(6)-carboxyfluorescein (BCECF; Calbiochem) for 30 min at room temperature, washed, and mixed with a 1:1 ratio of YAC-1 targets that had been labeled with 40 μg/ml of hydroethydine (HE; Polysciences, Inc.) for 10 min at room temperature. Cell mix was spun at 1,000 rpm for 5 min and incubated for 1, 2, 5, 10, or 20 min at 37°C. After gentle resuspension, cells were analyzed by flow cytometry. BCECF emits in FL1 and HE emits in FL3, therefore conjugates can be readily detected as double positive (FL1/FL3). Controls were sham conjugates obtained by centrifuging and incubating under the same conditions differentially labeled (HE and BCECF) YAC-1 cells. Background double-positive staining was reproducibly <1%.

Purified NK and T cells were labeled with 5 μg/ml of Indo-1 (Molecular Probes) for 45 min at 37°C. Washed cells were resuspended in 10% RPMI and divided into aliquots (5 × 10⁵ cells each) that were incubated for 30 min at 4°C with 20 μg/ml of one of the following mAbs: anti-2B4, anti-NK1.1, anti-CD16, anti-Ly49D, or anti-CD3. Cells were then washed and kept at room temperature until analysis, which was done using an LSR flow cytometer (Becton Dickinson). Immediately before acquisition, cells were resuspended in 500 μl of prewarmed medium. Acquisition was performed for 45 s to set the baseline and then briefly interrupted to add 25 μg/ml of goat F(ab′)₂ anti–mouse IgG to cross-link anti-2B4 and anti-NK1.1 mAbs or goat F(ab′)₂ anti–rat IgG to cross-link anti-CD16, anti-Ly49D, or anti-CD3 mAbs, and finally interrupted after 4–5 min to stimulate cells with 1 μg/ml of ionomycin. Cross-linking Abs failed to generate calcium flux in Indo-1–labeled cells not preincubated with primary mAbs.

NK and T sorted cells were first expanded in IL-2. Before analysis, cells were starved in IL-2–free medium for 4 h at 37°C and thereafter incubated on ice for 30 min with 20 μg/ml of anti-NK1.1 or anti-CD3 mAbs, followed by cross-linking with the relevant F(ab′)₂ Ab for 3 min at 37°C. As a positive control, cells were stimulated with 50 ng/ml PMA for 15 min at 37°C. Cells were lysed in 1% NP-40–containing buffer as described 44, and soluble proteins were separated by SDS-PAGE. ERK activation was assessed by immunoblotting with the anti–phospho-ERK mAb E4 (Santa Cruz Biotechnology). The amount of ERK proteins was assessed by reprobing the same membrane with a polyclonal anti-ERK Ab (Santa Cruz Biotechnology). Autoradiographies were quantified by densitometry.

IL-2–activated NK cells were mixed with YAC-1 cells at a 2:1 ratio. After 4 h, the specific release of esterase was measured in 25 μl of cell-free supernatants after addition of 175 μl of PBS containing 0.22 mM of DNP (Sigma-Aldrich) and 0.2 mM of N-α-benzyloxycarbonyl-l-lysine thiobenzyl ester (BLT; Sigma-Aldrich) as described 44. Maximal release was obtained by measuring the esterase content in the supernatant of NK cells after three freeze–thaw cycles.

Statistical significance of data was evaluated by applying Student's t test (Microsoft Excel^® software).

We enumerated lymphocyte subsets in the spleen, bone marrow, liver, and thymus of Vav1^−/− mice and littermate controls (Vav1^+/− or Vav1^+/+, hereafter referred to as wt). While T and NK-T cells were reduced in Vav1^−/− mice, NK cell percentages and absolute numbers in the spleen and bone marrow of Vav1^−/− mice were found to be greater than in controls (Fig. 1 A and Table). Mature NK cells express a constellation of surface markers, including members of the Ly49 family of MHC receptors, whose upregulation starts at 10 d of age and reaches levels comparable to those of mature NK cells at 2–3 wk of age 45. We found that the expression pattern of 13 different markers on splenic and marrow NK cells (including Ly49A, Ly49C/I, Ly49D, Ly49G2, 2B4, DX5, CD2, CD11b, CD16, CD69, CD90, CD117, and CD122; Fig. 1 B and data not shown) was comparable between Vav1^−/− and wt NK cells. Therefore, Vav1 is not required for NK cell development, unlike T and NK-T cell development.

NK cells recognize cells with disparate class I MHC expression and rapidly kill tumor cells that have lost expression of MHC class I 46. This function can be evaluated in vivo using a short-term assay, where radiolabeled RMA-S (class I–negative) tumor cells are injected intravenously. NK cell antitumor activity is inversely proportional to the radioactivity measured in the lungs. This function is T cell independent, because NK cell–depleted mice cannot reject the injected tumor cells 41. Thus, NK cell–deficient Rag2/γc^−/− mice retained ∼20% radioactivity, while only ∼1% was found in wt mice (Fig. 2 A) or in Rag2^−/− mice (data not shown). In contrast, Vav1^−/− mice displayed reduced NK cell activity, as about three- to fourfold more radioactivity was detected (3.4 ± 1% in Vav1^−/− versus 0.9 ± 0.2% in wt; P = 0.002).

In a second model of NK cell–dependent tumor rejection 47, we tested the capacity of Vav1^−/− mice to eliminate intraperitoneally injected RMA-S cells. As shown in Fig. 2 B, 60% of the injected tumor cells were retained in NK cell–deficient Rag2/γc^−/− mice. In contrast, only ∼0.5% of the inoculated cells were recovered in wt mice, while Vav1^−/− mice retained ∼16-fold more cells (Fig. 2 B), showing a considerable disadvantage in controlling tumor growth. Importantly, the absolute numbers of NK cells in the peritoneal cavity of wt and Vav1^−/− mice were similar, and the injection of tumor cells induced a similar local increase in NK cell numbers (data not shown). Finally, wt and Vav1^−/− mice were injected with H-2–compatible (H-2^K) RDM4 lymphoma cells, which express low levels of class I and therefore should be tolerated by T cells yet sensitive to NK cell lysis. All wt mice (four of four) survived for 14 d after the inoculum, and no signs of liver metastases were found at sacrifice (Table). In contrast, three of five Rag2/γc^−/− mice succumbed during the 14-d period, and the two surviving mice presented numerous liver metastases. Vav1^−/− mice showed signs of NK cell deficiency, as one of six mice died, and two of the remaining five mice showed liver metastases at sacrifice. Collectively, these results show that Vav1^−/− mice have reduced NK cell–mediated antitumor activity in vivo.

NK cells can secrete cytokines (notably IFN-γ, TNF-α, and GM-CSF) during viral or bacterial infections 48. Previous studies suggested that NK cell–derived IFN-γ is critical for early control of infection by L.m. 49. Therefore, we infected Vav1^−/− mice and wt mice intravenously with 10⁴ L.m., and bacterial burden was measured 48 h later in the liver and spleen (Fig. 2 C). Controls included Rag2/γc^−/− and Rag2^−/− mice. NK cell–deficient animals displayed an average of ∼10-fold more colonies in liver and spleen (data not shown) as compared with wt mice (P = 0.002; Fig. 2 C) or Rag2^−/− mice (data not shown). In contrast to the differences found between wt and Vav1^−/− mice in tumor clearance, no significant difference was found between wt and Vav1^−/− mice in terms of bacterial burden. Thus, Vav1^−/− mice can control the early phases of L.m. infection. As IFN-γ appears an essential cytokine for early control of L.m. infection, we measured the IFN-γ levels in the sera of infected mice. Fig. 2 D shows that IFN-γ production by Vav1^−/− infected mice was similar to that of wt infected mice. Taken together, these results suggest a functional dichotomy for NK cell functions: normal tumoricidal activity is Vav1 dependent, while IFN-γ production is Vav1 independent.

To dissect the cellular mechanisms responsible for defects in Vav1^−/− NK cells in vivo, we compared the lytic activity of freshly isolated and IL-2–activated NK cells from Vav1^−/− and wt mice. As shown in Fig. 3A,Fig. c,Fig. d,Fig. e, target cell lysis by Vav1^−/− NK cells was defective, ranging from 20 to 50% of control values. The assays included: (a) natural cytotoxicity versus murine thymoma cells (YAC-1), mastocytoma (P815), or macrophage cell line (IC-21); (b) ADCC using EL4 cells precoated with α-CD90; (c) killing of class I–deficient cells (β2m^−/− Con A–activated blasts and TAP-deficient RMA-S thymomas); and (d) “reverse” ADCC (R-ADCC) versus FcR⁺ P815 cells in the presence of α-CD16, α-2B4, or α-NK1.1 mAbs. These results strongly suggest that all forms of NK cell–mediated cytolysis depend on Vav1, irrespective of which cellular receptor is activated. Importantly, prior activation of NK cells with poly(I:C) in vivo or with IL-2 in vitro did not rescue the lytic defect in Vav1^−/− NK cells (Fig. 3B and Fig. d), and no significant difference was observed in the proliferation induced by IL-2 (data not shown).

In contrast, Vav1^−/− NK cells responded with normal IFN-γ production (Fig. 4 A), after stimulation of 2B4, CD16, and NK1.1 activating receptors, and with IL-12. It should be noted that stimulation of Vav1^−/− NK cells via the same receptors (2B4, CD16, and NK1.1) did not result in normal cytotoxicity of FcR⁺ P815 cells (Fig. 3 E). To better evaluate the dichotomy between cytokine production and cytotoxicity observed in Vav1^−/− NK cells, we measured the levels of intracellular IFN-γ produced by wt and Vav1^−/− NK cells 4 h after activation. This can be compared with the 4-h Cr-release assay used to measure cytotoxicity. No significant defect in IFN-γ production was detected, as similar percentages of wt and mutant NK cells stained positively for intracellular IFN-γ upon NK1.1 stimulation (Fig. 4 B). In contrast, CD3 stimulation did not result in IFN-γ production by Vav1^−/− T cells (Fig. 4 B).

It has been shown that conjugate formation is decreased in human NK cell lines expressing a dominant-negative form of the Vav1 substrate Rac1 32. We therefore tested the ability of Vav1^−/− NK cells to form stable conjugates using a flow cytometric approach. As shown in Fig. 5 A, conjugate formation by Vav1^−/− NK cells was readily detectable, ruling out an essential role for Vav1 in this process. A detailed kinetic study revealed that Vav1^−/− NK cells formed somewhat less conjugates, although the observed decrease never exceeded 30%, and was not statistically significant in four of five time points analyzed (Fig. 5 B). Therefore, although Vav1 may be required for normal conjugate formation under these conditions, the extent of the lytic defect (up to 80% reduction), strongly suggests that postbinding mechanisms account for defective killer activity of Vav1^−/− NK cells.

Vav1 is required for normal calcium flux in response to Ag–receptor signaling in T cells 26, and by analogy, reduced cytolysis by Vav1^−/− NK cells could result from a similar defect in proximal signaling. To test this, we compared the rise in intracellular calcium in NK cells upon cross-linking membrane receptors. Normal calcium flux was induced after stimulation of the NK1.1 and CD16 receptors (Fig. 6) as well as the 2B4 and Ly49D receptors (data not shown) in both wt and Vav1^−/− NK cells. In contrast, and as expected from previous reports 26, CD3 cross-linking did not induce calcium flux in Vav1^−/− T cells (Fig. 6). Thus, Vav1 is not essential to transduce signals to the calcium pathway in NK cells and acts downstream of the rise in intracellular calcium to control the NK cell cytotoxicity machinery.

Vav1 appears essential in transducing TCR signals to the ERK pathway in T cells 26, although studies using T cells from two other Vav1 mutant mice did not support an essential role for Vav1 in ERK activation 15,16. ERKs have been implicated in the control of both cytotoxicity and IFN-γ production by NK cells 34,35,36. We therefore analyzed ERK1/2 phosphorylation in NK cells after stimulation of NK1.1 and compared it to CD3-initiated ERK1/2 phosphorylation in T cells. We found a reduced activation of ERK1 (eightfold) and ERK2 (2.5-fold) in Vav1^−/− T cells (Fig. 7 A), confirming the essential role of Vav1 in TCR-mediated ERK activation 26. NK1.1-mediated ERK1 phosphorylation was 4.4-fold reduced in Vav1^−/− NK cells, while no significant reduction in activation of ERK2 was detected (Fig. 7 A). However, stimulation of Vav1^−/− NK cells via Ly49D resulted in reduced activation of both ERK1 (threefold) and ERK2 (twofold) kinases (data not shown). Collectively, these results suggest that Vav1 controls the ERK pathway in NK cells.

Intracellular polarization of NK cell cytotoxic granules toward sensitive target cells precedes granule exocytosis. The Vav1/Rac1 pathway has been suggested to be a critical regulator of this process 32. To directly measure granule exocytosis upon contact with targets, we quantitated the granzyme A esterase content in the supernatants of NK cell cultures stimulated with targets. Although total esterase content was similar in wt and mutant NK cells (data not shown), Vav1^−/− NK cells showed a significant reduction in esterase exocytosis compared with wt NK cells (7.2 ± 4.5 versus 17.3 ± 5%, P = 0.03; Fig. 7 B). Thus, the absence of Vav1 results in defective granule exocytosis and reduced ability to lyse tumor targets.

The data presented here show that Vav1 is an essential regulator of NK cell–mediated cytolysis of tumor cells. On the other hand, Vav1 is dispensable for NK cell development and cytokine production. Thus, Vav1^−/− mice generate normal numbers of phenotypically mature NK cells but show a dichotomy in NK cell–mediated effector functions. They are competent for IFN-γ–dependent control of early bacterial infections yet have a reduced capacity to reject tumors. We could define the cellular mechanisms that are Vav1 dependent. The absence of Vav1 had only minor effects on conjugate formation, and calcium flux was normal. In contrast, the absence of Vav1 reduced ERK activation and exocytosis of cytotoxic granules that allow NK cells to lyse their targets.

The phenotype of NK cell functions in Vav1^−/− mice is reminiscent of that of Beige mice, the murine model for Chediak-Higashi syndrome, a monogenic disease caused by mutations in the lysosomal trafficking regulator (Lyst) gene 50. Although the function of the Lyst protein has not yet been fully understood, it has been suggested to interact with microtubules to regulate protein sorting and late endosomal organization 51. Whatever the mechanisms underlying the Chediak-Higashi syndrome, the end result is a severely impaired granule exocytosis, much like what we have observed in Vav1^−/− NK cells. Interactions between Vav1 and Lyst have never been documented, but it is likely that they both regulate components of the cytoskeleton (actin filaments and microtubules, respectively) required for successful exocytosis. However, NK cells of Beige mice have a complete defect in cellular cytotoxity, whereas NK cells of Vav1^−/− mice have reduced capacity to kill tumor targets. How can we explain the quantitative defect in the lytic activity of Vav1^−/− NK cells?

Cell-mediated cytotoxicity can be induced by calcium-dependent, perforin-mediated mechanisms and calcium-independent, FasL-mediated pathways 52. The residual cytotoxicity of Vav1^−/− NK cells could be accounted for FasL-dependent mechanisms, which would not depend on exocytosis. In fact, we found that target cell lysis was totally abrogated in calcium-free medium in both wt and Vav1^−/− NK cells (data not shown), ruling out calcium-independent mechanisms as responsible for the residual cytotoxic activity seen in Vav1^−/− NK cells.

Activation of small GTPases can be induced by several GEFs, and functional redundancy among members of this family may account for some degree of compensation in the absence of Vav1. Human NK cells have been recently shown to express Vav2, which regulates the development of cell-mediated cytotoxicity 53. We have also detected Vav2 proteins in murine NK cells by intracellular staining (our unpublished observation), suggesting that suboptimal granule exocytosis and residual cytolysis by Vav1^−/− NK cells may be accounted for by compensatory mechanisms (activation of Rac1 or Rho?) initiated by Vav2. The analysis of NK cell functions in mice deficient for Vav1 and Vav2 will help answer this question.

Inactivation of the Vav1 substrate Rac1 reduces the capacity of human NK cells to form stable conjugates with target cells 32. However, our data indicate only a minor role for Vav1 at this stage, suggesting essential roles for Vav1 in postbinding target cell lysis. Moreover, GEFs other than Vav1 may control Rac1-dependent conjugate formation, explaining the partial discrepancy between our results and those reported in the literature 32.

Our data underscore fundamentally divergent roles for Vav1 in the biology of T and NK lymphocytes. First, T and NK-T cell development and T cell functions are impaired in Vav1^−/− mice, a consequence of the fact that T cells must signal through the same (Ag) receptor to complete differentiation and to exert their effector functions. Instead, requirement for Vav1 in NK cell functions (at least for cell-mediated cytotoxicity) dissociates from requirement for development. Several cytokine receptors are thought to drive NK cell differentiation (i.e., Flt3, c-kit, IL-7R, IL-15R), and we can suggest that Vav1 is not essential for transducing signals through any of these receptors, although it is phosphorylated upon ligation of some of them 23,24. Moreover, it is likely that NK cell cytotoxicity is triggered by ligand–receptor systems different from those required for development, otherwise developing NK cells might be constitutively stimulated in the absence of inhibitory signals (which develop only later in life) and potentially dangerous.

Second, proximal signaling (measured by rise in intracellular calcium) in Vav1^−/− NK cells is not impaired as opposed to defective capping and calcium flux upon cross-linking of TCRs in Vav1^−/− T cells 14,15,16. Therefore, the abnormal biological responses of Vav1^−/− T cells seem to be due to defective cellular activation. Along these lines, cytotoxic CD8⁺ T cells of Vav1^−/− mice fail to generate a primary response to viral infections, while they are able to mount a normal CTL secondary response, suggesting that Vav1 is required for T cell–mediated cytotoxicity only when the activation threshold is high, during a primary response 54. But this is unlikely to occur in Vav1^−/− NK cells, as we show here that IL-2 activation in vitro and stimulation with poly(I:C) in vivo does not rescue the NK cell lytic defect. Although several receptors have been reported to be able to activate NK cell effector functions 55, the ones that activate natural cytotoxicity remain unknown 56. This incomplete knowledge has obviously been an obstacle for advancing our understanding of the molecular mechanisms transducing NK cell activating signals into effector functions 57. The NK cell receptors we have used to elicit calcium flux (NK1.1, CD16, 2B4) are known to enhance cytotoxicity, but it would be interesting to measure rise in intracellular calcium upon target cell binding.

Vav1 has been shown to control the ERK pathway in T cells 26, using the same Vav1^−/− mice analyzed in this study. However, two independent groups, using mice carrying different Vav1 mutations, did not find an essential role for Vav1 in ERK activation 15,16. We confirm here the results by Costello et al. 26 showing that Vav1^−/− T cells have a reduced ERK activation. Moreover, we show that Vav1 controls the ERK pathway in NK cells. Although both cytotoxicity and IFN-γ production have been shown to depend on ERK 34,35,36, we found that only cytotoxicity was affected in Vav1^−/− NK cells. It is possible that different thresholds of ERK phosphorylation control cytokine productions and granule exocytosis, explaining the divergent effect seen in Vav1^−/− NK cells.

Finally, receptor ligation in Vav1^−/− NK cells resulted in divergent effects on tumor cell killing and on cytokine production, showing that Vav1 differentially controls NK cell effector functions. This situation is not paralleled in T cells, where TCR-initiated events tested so far (calcium flux, proliferation, apoptosis, IL-2 gene transcription, cytotoxicity) are all defective in the absence of Vav1 14,15,16,26. Our results are consistent with a functional dichotomy in NK cell signaling with respect to Vav1, where Vav1-dependent cytotoxicity contrasts with Vav1-independent cytokine production.

We would like to thank Fabienne Mazerolles (Necker Hospital, Paris) for helping with the detection of conjugate formation, Petter Hoglund (Karolinska Institute, Stockholm) for suggestions on the lung clearance assay, Claude Leclerc (Pasteur Institute, Paris) for providing the RDM4 cell line, Olivier Gaillot (Necker Hospital, Paris) and Thomas Ranson (Pasteur Institute, Paris) for help with the L.m. infections, and Emma Colucci-Guyon (Pasteur Institute, Paris) for discussions and for critical reading of the manuscript. The skillful technical help of Erwan Corcuff and Odile Richard is much appreciated.

This work was supported by grants from the Pasteur Institute (Paris, France), Fondation pour la Recherche Medicale (France), Ligue Nationale contre le Cancer (France), Association contre le Cancer (France), Medical Research Council (UK), and Biotechnology and Biological Sciences Research Council (BBSRC). F. Colucci is a recipient of a Poste Vert position sponsored by the Institute National pour la Santé et la Recherche Médicale (France). E. Rosmaraki is a recipient of a Marie Curie Training and Mobility of Researchers grant from the European Commission. S. Bregenholt was supported by a postdoctoral fellowship from the Danish Research Agency.