A Fatal Cytokine-Induced Systemic Inflammatory Response Reveals a Critical Role for NK Cells¹

Activation of monocytes/macrophages by bacteria, fungi, viruses, or their products results in the rapid production of monokines such as TNF-α, IL-1, IL-12, IL-15, and IL-18, which in turn induce NK cell production of IFN-γ and TNF-α (1, 2, 3, 4). IL-12 appears to be pivotal to the NK cell response, as there is only modest NK cell cytokine production in the absence of this factor (5, 6). We have demonstrated previously that the combination of IL-15 and IL-12 exerts a profound synergy upon resting NK cell production of IFN-γ, TNF-α, and MIP-1α³ (6, 7, 8). These proinflammatory cytokines and chemokines play a critical role in the clearance of obligate intracellular pathogens and, in some cases, the promotion of sepsis, shock, and death (9). An identical profile of NK cell cytokine and chemokine production can be induced by the combination of IL-2 and IL-12 (5, 6). This reflects the fact that the heterotrimeric IL-15R and IL-2R share the IL-2Rβ and γ signaling subunits and differ only in the specificity of their high affinity α-chains (10). The ability to obtain synergistic immunomodulatory effects via activation of the IL-2/15R and the IL-12R expressed on NK cells has led to investigations of this cytokine combination for the immunotherapy of neoplastic disease (11, 12). In the current study, we examined the effects of administering IL-2 or IL-15 in combination with IL-12 in a murine toxicity model. While the dose of the individual cytokines was well tolerated, the administration of IL-2 or IL-15 in combination with IL-12 induced a lethal systemic inflammatory response that did not require any of the major proinflammatory factors or signaling pathways felt to be active in the induction of septic shock. However, the lethal toxicity of this cytokine combination therapy was critically dependent upon NK cells, but not B or T cells.

Purified, yeast-derived rhuIL-2 (Chiron, Emeryville, CA) or rhuIL-15 (Immunex, Seattle, WA) was administered at a dose of 3 × 10⁵ U/day via the i.p. route. rIL-12 of murine (mu) origin (Genetics Institute, Cambridge, MA) was administered i.p. at a dose of 1 μg/day. rhuIL-2 and rmuIL-12 were administered daily until the death of the animal. rmuIL-10 was supplied by Schering-Plough (Kenilworth, NJ). Platelet-derived huTGF-β1 (which has activity in murine systems) was purchased from R & D Systems (Minneapolis, MN) and reconstituted according to the manufacturer’s recommendations in 4 mM HCl supplemented with 0.1% BSA (13). Rat IgG1 anti-muIFN-γ mAb (Endogen, Cambridge, MA), and hamster IgG anti-muIFN-γ mAb (Genzyme Diagnostics, Cambridge, MA) were administered at a dose of 100 μg per mouse via the i.p. route 12 h before cytokine therapy and then daily thereafter. Control Abs were purchased from Sigma (St. Louis, MO). A dimeric rhuTNF receptor p80/IgG1 Fc fusion protein (TNFR-Fc; Immunex) was used in TNF-α neutralization experiments (100 μg/mouse/day i.p. beginning 24 h prior cytokine treatment) (14). Human IgG was used as a control for the fusion protein (Baxter, Glendale, CA). Dexamethasone was purchased from Moore Medical (New Britain, CT). Depletion of NK cells was accomplished via i.p. administration of an anti-asialo GM1 Ab (Wako BioProducts, Richmond, VA) every 3 days beginning 2 wk before the injection of cytokines (0.2 mg/mouse) (15). NK cell numbers postdepletion were evaluated via flow-cytometric analysis of murine splenocytes using a phycoerythrin (PE)-labeled pan-NK mAb (clone DX5; PharMingen, San Diego, CA). Mice were depleted of monocytes/macrophages via i.v. and i.p. injection of the F4/80 mAb (an IgG2b mAb) 48 and 24 h before cytokine therapy (16). Macrophage numbers postdepletion were evaluated via the enumeration of plastic-adherent cells following a 2-h culture of PBMCs, splenocytes, bone marrow cells, or peritoneal cells in 24-well plastic culture dishes (1 × 10⁶ cells/well in RPMI 1640 supplemented with 10% FBS) (7). All cytokine reagents contained less than 0.015 EU/ml endotoxin, as measured by the E-Toxate system (Sigma).

Female mice age 4–6 wk were utilized in all experiments. C.B-17 scid/scid (SCID) mice (BALB/c background), splenectomized C.B-17 SCID mice, sham-operated C.B-17 SCID mice, and inbred BALB/c mice were purchased from Taconic Farms (Germantown, NY). Perforin-deficient (^−/−) mice, CD3ε transgenic mice, IFN-γ^−/− mice, and Fas^lpr/lpr mice were purchased from The Jackson Laboratory (Bar Harbor, ME) (17, 18, 19, 20). TNFR p55^−/− mice, TNFR p75^−/− mice, and TNFR p55^−/−/TNFR p75^−/− mice were provided by Immunex (21). Type I IL-1R^−/− mice, type I IL-1R^−/−/TNFR p55^−/− mice, and type I IL-1R^−/−/TNFR p75^−/− mice were also provided by Immunex (22). TNFR p55^−/− and Fas^lpr/lpr/TNFR p55^−/− mice were produced by Dr. Klaus Pfeffer (Munich, Germany) (20, 23). MIP-1α^−/− mice were the gift of Dr. O. Smithies (Chapel Hill, NC) (24). Mice deficient in the enzyme-inducible nitric oxide synthase (iNOS^−/−) were the gift of Dr. Ricardo Gazzinelli (Bethesda, MD) (25). Mice deficient in the IL-1β-converting enzyme (ICE^−/−) and STAT1^−/− mice were produced as described (26, 27). All mice were housed in a specific pathogen-free environment and given food and water ad libitum.

Serum levels of IFN-γ and TNF-α were measured using ELISAs obtained from Endogen (Woburn, MA). IL-1β and IL-6 levels were measured using ELISAs obtained from Biosource International (Camarillo, CA). KC and MIP-2 levels were measured using ELISAs from R & D Systems. Serum chemistries were performed on mouse serum using a Vitros 500 analyzer (Johnson and Johnson, Raritan, NJ). Histopathologic examination of cytokine-treated SCID mice and determination of mean wet-to-dry lung ratios were performed as described (28). NK cells were isolated from the spleens of cytokine-treated C.B-17 SCID mice and analyzed for endonucleolytic cleavage of cellular DNA via a flow-cytometric assay using propidium iodide and DNA gel electrophoresis, as described (12, 29).

Serially diluted serum was analyzed for haptoglobin and α1-acid glycoprotein by immunoelectrophoresis (30). The area under the precipitation peak was quantitated in arbitrary units using the National Institute of Health Image program 1.61. The data for each peak were then converted into mg/ml values by comparison with the values obtained with calibrated mouse acute phase plasma.

An E1A-deleted recombinant adenovirus (rAd) expressing a dominant-negative form of the I-κBα protein was constructed (31). A rAd engineered to express the lacZ gene served as a control. A total of 10⁹ PFUs of rAd/I-κB or rAd/lacZ was administered to mice via tail vein injection 48 h before treatment with IL-2 plus IL-12 (32). In vivo protein expression derived from these adenoviral vectors was confirmed via immunohistochemistry, as described (7).

Statistical significance was analyzed by the Student’s t test.

Intraperitoneal administration of rhuIL-2 (3 × 10⁵ U/day) plus rmuIL-12 (1 μg/day) was lethal to C57BL/6 mice within 4 to 6 days (Fig. 1). The combination of IL-15 plus IL-12 elicited identical results, which was expected given that IL-15 signals through components of the IL-2R (6, 33). No deaths were observed in control mice receiving daily injections of IL-2, IL-15, or IL-12 alone. Similar results were obtained with IL-2 or IL-15 plus IL-12 in several different species of mice, including BALB/c, 129, B6 × 129, CD-1, and C.B-17 mice bearing the scid/scid (SCID) mutation (data not shown). Indeed, SCID mice, which lack B and T cells (34), exhibited 100% mortality within 3 to 5 days of the initiation of treatment and were utilized extensively in the majority of experiments.

SCID mice receiving IL-2, IL-12, or IL-2 plus IL-12 were subjected to histopathologic evaluations. IL-2-associated findings included mononuclear cell infiltrates in the portal areas of the liver, splenic extramedullary hemopoiesis, and pulmonary interstitial mononuclear cell infiltrates. IL-12-associated changes included mild hyperplastic and degenerative changes of the gastrointestinal mucosa, and scattered foci of apoptotic lymphoid cells within the lymphoid organs. These IL-2- and IL-12-induced lesions were mild, of late onset (72–96 h), and consistent with those previously described for these cytokines (35, 36). Changes in SCID mice treated with IL-2 plus IL-12 included both significant exacerbation of IL-2- and IL-12-associated lesions as well as novel changes such as fibrinoid necrosis of lymphoid tissue in the spleen and lymph nodes, confluent foci of necrosis in pancreatic exocrine tissue, macrophage activation and proliferation, and fibrinoid necrosis of pulmonary arterioles. Apoptosis of SCID lymphocytes (i.e., NK cells) within the lymphoid tissues (spleen and lymph nodes) was markedly enhanced (Fig. 2,A). Analysis of nonadherent splenic NK cells from cytokine-treated SCID mice by DNA gel electrophoresis and propidium iodide staining (12) confirmed this observation (Fig. 2,B and data not shown). Lesions of the gastrointestinal tract associated with IL-12 administration were also markedly exacerbated by the combination of IL-2 and IL-12 (Fig. 2,C). Pulmonary pathology was prominent and included perivascular and septal mononuclear cell infiltrates associated with multifocal hemorrhage and alveolar edema. The formation of pulmonary edema at the 72-h time point was significantly greater in mice receiving the combination of IL-2 plus IL-12 (p < 0.05, Fig. 2,D) (28). Analysis of serum chemistries revealed significant increases in the liver enzymes ALT, AST, and LDH (5.3-, 9.9-, and 2.4-fold increases over baseline, respectively) as well as acute phase proteins beginning 24–48 h after treatment with IL-2 plus IL-12, but not following the administration of IL-2 or IL-12 alone (Fig. 2 E) (37).

Serum levels of IFN-γ and TNF-α rose rapidly in SCID mice treated with IL-2 plus IL-12, peaked at approximately 24 h, and remained elevated until death (Fig. 3, A and B). Anti-asialo GM1 Ab-treated SCID mice did not exhibit elevated serum levels of IFN-γ or TNF-α after receiving IL-2 plus IL-12, suggesting that this cytokine combination acted directly on NK cells to induce production of IFN-γ and TNF-α (not shown). Serum levels of IL-1β, IL-6, KC, and MIP-2 were also found to be elevated during combined administration of IL-2 with IL-12 (summarized in Fig. 3,C). The elevated levels of proinflammatory cytokines seen with the combination of IL-2 plus IL-12 were not the result of a simple additive effect, since administration of IL-2 or IL-12 alone did not elicit significant cytokine production in SCID mice (with the exception of IFN-γ production in mice receiving IL-12). Of note, elevations in TNF-α and IL-1β occurred early in the course of treatment (i.e., within 1–3 h), whereas peak levels of IL-6, KC, and MIP-2 appeared later in the disease course, as is observed in animal models of septic shock (Fig. 3, B and C, and data not shown) (9, 38).

We investigated the mechanism of this fatal cytokine-induced inflammatory response treatment using cytokine neutralization strategies and genetically altered mouse strains. Results are summarized in Table I. Neutralization of IFN-γ or TNF-α did not afford protection to SCID mice treated with IL-2 plus IL-12, nor did genetic deficiencies in IFN-γ or the TNFR complex (p55 and/or p75). We noted somewhat prolonged survival within both the control and experimental groups in our initial experiments with TNFR p55^−/− mice. However, in subsequent experiments with the identical strain of TNFR p55^−/− mouse and an independently generated strain, we found that all background mice and TNFR p55^−/− mice died within 5 to 7 days (Table I and data not shown). These data in conjunction with the results of our TNF-α neutralization experiments led us to conclude that TNF-α was not solely responsible for the toxicity seen in this model. STAT1^−/− mice were also susceptible to the toxicity of IL-2 plus IL-12, which is significant because STAT1 is critical for IFN-γ-induced gene regulation and also for the induction of apoptosis by TNF-α (39, 40). IFN-γ and TNF-α were simultaneously neutralized using both an anti-IFN-γ mAb and a TNFR-Fc soluble receptor construct (14, 41). Mortality rates of the experimental and control groups were identical. TNF-α was neutralized in IFN-γ^−/− mice, and IFN-γ was neutralized in TNFR p55^−/−/TNFR p75^−/− mice, without effect. Type I IL-1R^−/− mice, IL-1R^−/−/TNFR p55^−/− mice, and IL-1R^−/−/TNFR p75^−/− mice all succumbed to the lethal effects of IL-2 plus IL-12, whereas mice receiving IL-2 alone or IL-12 alone exhibited minimal toxicity (not shown). In addition, IL-1R^−/−/TNFR p55^−/− mice were treated with a neutralizing Ab to muIFN-γ during treatment with IL-2 plus IL-12, yet this intervention did not ameliorate toxicity or prolong survival (not shown). Several alternative effector molecules were considered as possible mediators of toxicity; however, mice with targeted genetic deficiencies in MIP-1α, ICE, perforin, Fas, and iNOS remained completely susceptible to the toxic effects of IL-2 plus IL-12.

Dexamethasone, TGF-β1, and ibuprofen have been used effectively to prevent morbidity and death in animal models of septic shock and other inflammatory processes (42, 43, 44). These agents were administered in pharmacologically relevant doses before the start of cytokine treatment and then daily thereafter; however, none was capable of ameliorating the toxicity of IL-2 plus IL-12 (Table II). Inhibition of NF-κB signaling in vivo via overexpression of I-κB in the liver was also ineffective in preventing mortality in this model, in contrast to its ability to protect mice from the lethal effects of endotoxin (45). Indeed, mice expressing the I-κB protein were actually more susceptible to the toxic effects of IL-2 plus IL-12 than were mice treated with the control vector (p < 0.05, Table II). This experiment was repeated in TNFR p55^−/− mice because of the role I-κB has in potentiating TNF-mediated apoptosis (31, 46) (M. Karin, unpublished observation). The toxicity of IL-2 plus IL-12 in TNFR p55^−/− mice overexpressing I-κB was essentially identical in the experimental and control groups (Table II).

SCID mice that lack T and B lymphocytes undergo massive NK cell apoptosis following administration of IL-2 plus IL-12 (Fig. 2, A and B). To determine whether the toxicity of this model was mediated by NK cells, we administered IL-2 plus IL-12 to SCID mice depleted of NK cells by pretreatment with an anti-asialo GM1 Ab (Fig. 4,A). IL-2 (or IL-15) plus IL-12 elicited minimal toxicity when administered to SCID mice depleted of NK cells, and 100% of mice in this group survived, whereas control mice receiving IL-2 plus IL-12 all died within 5 days of the initiation of treatment (Fig. 5,A). To confirm this observation, we administered IL-2 and IL-12 to SCID mice that had been depleted of NK cells by splenectomy and observed 100% survival (47). Sham-operated SCID mice treated with IL-2 plus IL-12 exhibited 100% mortality at 5 days (Fig. 5,B). Furthermore, CD3ε transgenic mice that completely lack mature NK cells (and T cells) due to a developmental block (18) showed absolutely no toxicity when treated with IL-2 plus IL-12. Control mice of the appropriate background all died between 4 and 8 days (Fig. 5 C).

Given the ability of NK cell-derived cytokines to potentiate macrophage effector functions (1, 2, 3, 48), the role of macrophages in the toxicity of this model was investigated. SCID mice were depleted of monocytes and macrophages by approximately 50% in the peripheral blood, spleen, and bone marrow, and by >95% in the peritoneal cavity by injecting the F4/80 mAb (16) via the i.v. and i.p. routes 48 and 24 h before the administration of IL-2 plus IL-12 (Fig. 4,B). Mice receiving the F4/80 mAb tolerated the administration of IL-2 plus IL-12 significantly better than mice receiving the control Ab and exhibited a 50% survival rate (p < 0.05). Control mice all died within 3 to 4 days (Fig. 5,D). IL-10 is a potent macrophage deactivator (49). Pretreatment of SCID mice with rmuIL-10 afforded significant protection from the toxicity of IL-2 plus IL-12 (66% survival overall, p < 0.02, Table II). Taken together, these data suggest that monocytes/macrophages, in addition to NK cells, have a role in mediating the lethal toxicity of IL-2 plus IL-12.

Administration of IL-12 in combination with IL-2 or IL-15 induced a systemic inflammatory response that rapidly progressed to a fatal shocklike state. The NK cell compartment was the only lymphocyte population responsible for mediating this toxicity because mice that underwent depletion of NK cells or had congenital absence of NK cells were completely protected from the lethal effects of this cytokine combination. Treatment of SCID mice with IL-2 plus IL-12 resulted in high serum levels of IFN-γ and TNF-α, and production of these cytokines was dependent upon the presence of NK cells. However, neither IFN-γ nor TNF-α was required for the fatal inflammatory reaction induced by IL-2 plus IL-12. Other effector molecules of the NK cell and macrophage compartments, namely MIP-1α, IL-1, ICE, Fas, perforin, iNOS, and the STAT1 pathway of signal transduction, were eliminated as critical mediators of toxicity in this model.

Our histopathologic findings and analysis of serum cytokine levels indicated that the administration of IL-2 plus IL-12 had induced a severe systemic inflammatory response. High circulating levels of IL-1β, IL-6, and the IL-8 homologues (KC and MIP-2) appeared only in the serum of mice receiving the combination of IL-2 plus IL-12, which is significant in that the production of these cytokines is limited to periods of inflammation, tissue injury, and immunologic challenge (38). Indeed, the sequence of cytokine induction was highly reminiscent of that observed in experimental models of endotoxemia and in humans diagnosed with septic shock (9, 38). The presence of high serum levels of IL-6 at the time of death is particularly significant in that they represent the net effect of biologically active IL-1β and TNF-α, and have been found to correlate inversely with survival in patients with septic shock (50). Thus, treatment of mice with IL-2 plus IL-12 induces many of the same proinflammatory mediators that are active in sepsis. The presence of increased pulmonary edema, multiple organ system toxicities, and an acute phase response in mice receiving the combination of IL-2 and IL-12 lends additional support to this observation (9, 51). Induction of proinflammatory mediators has also been implicated in the pathogenesis of the shocklike states associated with high dose cytokine therapy in humans, and we initially hypothesized that the lethal reaction to IL-2 plus IL-12 might be the result of additive toxicities induced by the overlapping actions of known proinflammatory factors (52, 53). However, no protection was afforded by simultaneous neutralization of TNF-α and IFN-γ or utilization of IL-1R^−/−/TNFR p55^−/− and IL-1R^−/−/TNFR p75^−/− mice. The data also suggest that the fatal toxicity of IL-2 plus IL-12 was not mediated via NF-κB signaling within the liver; however, it is possible that NF-κB was only partially inhibited and therefore still able to activate the transcription of NF-κB-responsive genes during this intense inflammatory response (45). We cannot exclude the possibility that other cytokines, cytokine receptors, or effector molecules might combine with IL-1, TNF-α, or IFN-γ following administration of IL-2 plus IL-12 to induce a lethal inflammatory response. However, the data suggest that stimulation of the NK cell compartment with IL-2 plus IL-12 might result in the production of novel factors or factors that have yet to be characterized as proinflammatory. Such a hypothesis is supported by our characterization of several unique pathologic lesions that have not been observed in other models of shock (9, 51).

The appearance of IL-1β, IL-6, and other macrophage-derived cytokines in the circulation, and the presence of activated and proliferating macrophages in the splenic bed following the administration of IL-2 plus IL-12 implied that this treatment had directly or indirectly activated the macrophage compartment (54). The importance of macrophages in the toxicity of this model was confirmed by studies in which partial depletion of the macrophage compartment resulted in a 50% survival rate for mice receiving IL-2 plus IL-12. The protection afforded by IL-10 pretreatments also suggested a role for the macrophage compartment in the toxicity of this model; however, we cannot rule out the possibility that IL-10 may exert its protective effects via some other pathway (49). The prevention of death by the congenital absence or depletion of NK cells and the improved survival by the partial depletion of macrophages suggested that the toxicity of this model is the result of interactions between these two cell populations. NK cells constitutively express the IL-2/15R, and activation of this receptor complex results in further up-regulation of NK cell IL-12R expression (55). Thus, we suspect that the toxicity of this model first involves stimulation of NK cells by IL-2/15 plus IL-12, followed by an NK cell-dependent activation of macrophages. The inability of IL-2 plus IL-12 administration to induce TNF-α in NK-depleted SCID mice also supports our proposed sequence of events, as NK cells and macrophages are both sources of this cytokine. Thus, mobilization of macrophage effector function may occur as a result of continuous and uncontrolled production of NK cell factors in response to IL-2 plus IL-12 (2), although the precise cellular source of these proinflammatory cytokines cannot be determined based upon measurements of serum cytokines. Our postulate that chronic stimulation of the NK cell compartment via daily injection of IL-2 plus IL-12 results in macrophage activation is not an unprecedented one (1, 2, 56, 57). However, the inability of IFN-γ and TNF-α neutralization strategies to prevent death was unexpected and suggests that NK cells may be able to activate macrophage effector functions via alternate pathways.

To the best of our knowledge, the present model represents the first example of a cytokine-induced shock syndrome that is mediated by components of the innate immune system. Our current understanding of the role of the NK cell in other models of shock is quite limited, but there is evidence that NK cells can be activated during septic events and may contribute to the pathogenesis of this condition via the secretion of IFN-γ, which acts primarily to augment macrophage function. One well-studied example of this phenomena is the Shwartzman reaction in which an intradermal priming dose of LPS is followed 24 h later by an i.v. LPS challenge (58). The first LPS dose induces the production of IL-12, which stimulates the release of IFN-γ (presumably by NK cells) and permits the priming of macrophages. Upon subsequent LPS challenge, sensitized macrophages release massive amounts of TNF-α and IL-1β, which mediate the lethal effects of this treatment (58). Heremans et al. have demonstrated that depletion of NK cells before the induction of the generalized Shwartzman reaction leads to a 70% reduction in mortality and significantly lower levels of IFN-γ and TNF-α following the systemic injection of LPS (59). In contrast, mice depleted of CD4⁺ or CD8⁺ T cells were still highly susceptible to the lethal effects of the Shwartzman reaction. Ozmen et al. analyzed the Shwartzman reaction in detail and found that simple coinjection of TNF-α plus IL-1β or TNF-α plus IFN-γ was sufficient to induce lethality following priming of mice with IL-12 or IFN-γ alone (60). However, regardless of the priming event, no mortality was observed if IL-1β and IFN-γ were administered in the absence of TNF-α. Thus, in distinct contrast to our model, TNF-α is crucial for the lethality of the Shwartzman reaction. Furthermore, our experiments with IFN-γ^−/− and STAT1^−/− mice would suggest that IFN-γ priming (following IL-12 administration) is not a critical event in our model.

The induction of NK cell apoptosis in the present model was dependent upon the coadministration of IL-2 with IL-12. Nonadherent splenocytes from cytokine-treated SCID mice did not exhibit endonucleosomal DNA degradation or the morphologic features of apoptosis unless mice had received both cytokines, results that were confirmed in vitro (not shown). Apoptosis of NK cells has also been observed following coactivation with IL-2 and engagement of the low affinity FcRγ (or cross-linking of CD94), and after cell-mediated lysis of leukemic targets cells in the presence of IL-2 (61, 62, 63, 64). We have demonstrated previously that in vitro stimulation of resting human NK cells with IL-2/15 and IL-12 leads to the production of large amounts of IFN-γ and TNF-α, followed by the induction of programmed cell death at approximately 48–72 h (12). Activation-induced apoptosis appears to be a common strategy for the removal of activated effector cells and attenuation of the immune response to specific pathogens (65). The ability of IL-2 plus IL-12 to induce NK cell apoptosis in vivo is further evidence for the central role of the NK cell compartment in the toxicity of this treatment. The existence of this suicide pathway following NK cell activation further suggests that unbridled amplification of NK cell effector functions may be deleterious to the organism. This apoptotic regulatory mechanism was apparently insufficient to limit toxicity in the present model, possibly due to the intensity and rapidity of the NK cell response to administration of IL-2/15 plus IL-12. Activation-induced apoptosis associated with the induction of a shocklike state has also been described following experimental activation of T cells with staphylococcal enterotoxins (superantigens) or anti-CD3 Ab (66, 67). However, in contrast to our model, the toxicity of these T cell-mediated shock syndromes is mediated primarily by TNF-α.

In summary, we have demonstrated that the administration of IL-2 or IL-15 in combination with IL-12 results in a fatal systemic inflammatory response that is critically dependent upon the NK cell compartment. This lethal shocklike reaction does not appear to be mediated by any of the known cytokine products of the NK cell compartment, nor any of the effector molecules associated with NK cell cytotoxic activity. This is the first time that the NK cell (or any cellular compartment) has been identified as an important component of cytokine-induced systemic inflammation. Elucidation of the factors involved in this novel inflammatory pathway may therefore have relevance for understanding the complications of high dose cytokine therapy as well as other forms of shock.

We thank Dr. Ricardo Gazzinelli for his assistance with experiments involving the inducible nitric oxide synthase-deficient murine strain, and Dr. Peter Kantor for analysis of serum chemistries.