A Role for IFN-αβ in Virus Infection-Induced Sensitization to Endotoxin¹

The outcome of bacterial septic shock in humans is determined by the magnitude of the inflammatory response (1). Thus, characterizing factors impacting on this magnitude are critical to understanding the disease. Underlying viral infection can modulate secondary inflammatory challenges (2, 3, 4). The impact of immune activation by virus on the outcome of bacterial infection is not well understood. In humans there are many examples of links between severe bacterial disease and underlying, concurrent viral disease. For instance, influenza viral infection is well known to predispose humans to severe morbidity and mortality from common respiratory bacterial pathogens (5, 6, 7). Varicella also predisposes for severe streptococcal and staphylococcal toxic shock-like syndrome (8, 9). The mechanism for this coupling is not well understood. Animal models demonstrate that many viruses such as coxsackie, influenza, lymphocytic choriomeningitis virus (LCMV),³ and murine CMV heighten the sensitivity to second inflammatory challenge (2, 3, 4, 10, 11, 12, 13, 14, 15). Pathways for this sensitization have not been clearly defined.

Key proinflammatory mediators required for a lethal response to bacteria or endotoxin include TNF-α, IFN-γ, and IL-12 (16, 17, 18, 19, 20). TNF-α is a primary mediator of much of the pathology leading to lethality from endotoxin (21, 22, 23, 24, 25, 26). IFN-γ participates in this response by enhancing the sensitivity to TNF-α (27, 28, 29). IL-12 is important because it induces IFN-γ (17, 18). Many reports have established that prior exposure to IFN-γ can enhance the inflammatory and lethal effects of endotoxin (19, 20, 30, 31). This has been demonstrated as IFN-γ-dependent priming to endotoxin in the Schwartzman reaction and in dual challenge models using viral infection followed by superantigen or endotoxin challenge (4, 11, 19). Viral infections are potent stimuli for type 1 and/or type 2 IFNs (IFN-αβ and/or IFN-γ) (32, 33). Some viruses, such as LCMV, induce primarily an early IFN-αβ response, which can actively inhibit the induction of IL-12 and, subsequently, IFN-γ (33, 34, 35). Later during LCMV infection IFN-αβ levels fall, and T cells become prominent producers of IFN-γ. Exquisite sensitivity to a second inflammatory challenge such as superantigen, IL-12, or endotoxin has been demonstrated late during viral infections when T cells are activated to become IFN-γ-producing cells (3, 4, 11, 12). IFN-γ signaling results in the formation of activated STAT1 homodimers, but IFN-αβ signaling activates STAT1 and STAT2, leading to the formation of STAT1 and STAT2 heterodimers and fewer STAT1 homodimers. Thus, the IFNs can have some common functions (32). The ability to potentiate the response to a second inflammatory challenge may be such a function. However, little is known about the impact of IFN-αβ on LPS priming. IFN-αβ are produced early during many viral infections and are administered therapeutically for chronic viral hepatitis (36). They are also being used to augment the immune response in other infections and several malignancies (37, 38, 39). Because of this, any similarity to IFN-γ in terms of priming for second inflammatory challenges may be important in complications during viral infections or IFN-αβ therapies.

LCMV infection provides a system in which the effects of IFN-αβ or IFN-γ on endotoxin sensitivity can be assessed because LCMV induces an early IFN-αβ response that actively inhibits IL-12 and IFN-γ. Alternatively, in mice genetically deficient in IFN-αβ signaling, LCMV induces an early IFN-γ response (34, 35). In this report we demonstrate that sensitization for endotoxin occurs early during the innate response to LCMV. T, B, NK cells, IFN-γ, and IL-12 are not required. Early sensitization to endotoxin is mediated by IFN-αβ. Because IFN-γ can be induced in the absence of LCMV-induced IFN-αβ, both type 1 and 2 IFN functions must be blocked to reduce this sensitization. This can be demonstrated in both IFN-αβγ receptor-deficient or STAT1-deficient mice where signaling for both type 1 and 2 IFNs is interrupted. These data demonstrate for the first time that viral infection-induced IFN-αβ is capable of potentiating the response to endotoxin and, like IFN-γ, may be a critical determinant of the outcome of septic shock. They also suggest possible mechanisms for detrimental conditions resulting from IFN-αβ therapies.

All mice were housed in specific pathogen-free conditions until initiation of experiments. The strains of mice used included C57BL/6, 129 SvEv (Taconic Laboratory Animals and Services, Germantown, NY). Mice genetically deficient in STAT1 on C57BL/6 background (STAT1^−/−) and mice deficient in IFN-αβ and/or γ receptors (IFN-αβ R^−/−, IFN-γ R^−/−, IFN-αβγ R^−/−) on a 129SvEv background were obtained from B & K Universal Limited (North Humberside, U.K.) (IFN-αβ R^−/−) or from Dr. Joan Durbin at Ohio State University (STAT1^−/−, IFN-γ R^−/−, IFN-αβγ R^−/−) and bred here. E26 mice, deficient in NK and T cells, were established with CBA × C57BL/6 backgrounds as described (40). They were bred in strict isolation by brother-to-sister mating in the animal care facility at Brown University. T and B cell-deficient C57BL/6-recombination activation gene 1 (RAG-1) mutants (RAG-1^−/−) mice were also bred at Brown University (41). IFN-γ-deficient and IL-12p40-deficient mice on C57BL/6 background as well as their controls were obtained from The Jackson Laboratory (Bar Harbor, ME). STAT4-deficient mice were products of C57BL/6 and 129 breeding and were originally obtained from Dr. J. Ihle, St. Jude’s Children’s Research Hospital (Memphis, TN) and bred in our facility. Their C57BL/6 × 129 F₂ controls were obtained from The Jackson Laboratory.

Experiments were initiated on day 0 with mice either not infected or infected i.p. with 2 × 10⁴ PFU of LCMV Armstrong strain clone E350 (34).

Animals were injected daily i.p. with 1 × 10⁵ U of recombinant human IFN-α A/D (IFN-α), active on all mouse cells, donated by Dr. Michael Brunda (Hoffman-LaRoche, Nutley, NJ) or purchased from PBL Biomedical Laboratories (New Brunswick, NJ), for 1 or 2 days before LPS challenge. Vehicle treatment was 0.25 M ammonium acetate, 0.2 M NaCl at pH 2.5 diluted 1:140 in PBS (Brunda preparation) or PBS with 0.1% BSA (PBL preparation).

NK depletion was achieved by pretreatment with anti-NK1.1 mAb (0.165 mg of PK136 mAb) on day −1 relative to LCMV infection. This protocol is >90% effective at removing NK cells from the spleen cell population (42, 43). Control animals were treated with similar preparations made with P3NS1 cells. Both reagents were purified in our laboratory from ascites fluid.

Mice were injected i.p. with endotoxin (LPS) Escherichia coli strain 0111:B4 (Difco, Detroit, MI) day 1 or 2 after virus or IFN-α treatments described above. The dose administered was determined by the strain of mice used. C57BL/6 mice were much more sensitive to LPS compared with 129 SvEv and mice >8 wk of age were more sensitive compared with younger mice. For this reason, all experiments were done using age-matched controls. A careful LPS dose titration was performed to determine the optimal dose for age and strain for experiments. For survival experiments using the C57BL/6 strain, all mice were 5–6 wk of age and received an LPS dose ranging from 50, 75, and 100 μg. For serum TNF-α experiments using C56BL/6 mice, 100 μg of LPS was injected; and for 129 SvEv mice, 200 μg of LPS was injected.

Mice infected with virus or injected with IFN-α on day 0 and controls were injected with LPS i.p. on day 1 or 2. Blood was collected 1.5 and 6 h after LPS injection as well as from controls receiving no LPS. Whole blood was centrifuged, and serum was decanted and stored at −20°C.

Mice infected with virus or injected with rIFN-α on day 0 and controls were sacrificed on day 1 or 2. Peritoneal leukocytes were obtained by peritoneal lavage with 10 ml of cold sterile PBS. Spleens were harvested and manually disrupted to obtain a single cell suspension of splenic leukocytes. Peritoneal and splenic leukocytes were briefly treated with NH₄Cl to lyse erythrocytes. Viability was confirmed by trypan blue exclusion. Splenic and peritoneal leukocytes were cultured at 1 × 10⁶ and 5 × 10⁵ cells/well, respectively, in 96-well plates with a titration of LPS. Ten-fold dilutions of LPS from 100 to 0.1 mg/ml were added to cells at a 1:1 ratio. Therefore, the final concentrations of the culture solution were half the concentration appearing on the figure legends. Cells were cultured at 37°C for 24 h in RPMI 1640 plus 10% FBS; supernatant was decanted and stored at −20°C. Due to low cell yields per mouse, peritoneal cells were pooled from each group.

TNF-α and IFN-γ concentrations in serum and leukocyte culture supernatant were determined by standard sandwich ELISA. The capture Ab for TNF-α was clone TN3-19.2 hamster anti-mouse/rat TNF-α (BD PharMingen, San Diego, CA). The detection Ab was a polyclonal rabbit anti-mouse TNF-α Ab (Endogen, Woburn, MA). Limit of detection for diluted serum and culture supernatant TNF-α was 0.08 ng/ml. IFN-γ capture Ab was obtained from ascites from clone XMG1.2. The detection Ab was rabbit polyclonal anti-mouse IFN-γ supplied by Dr. Phillip Scott (University of Pennsylvania, Philadelphia, PA). The limit of detection for diluted samples was 0.08 ng/ml. Colorimetric changes of enzyme substrates were detected at 405 nm wavelength using a Spectramax 250 reader (Molecular Devices, Sunnyvale, CA).

Results are given as mean ± SEM. Two-tailed Student’s t tests were run on SigmaStat (SPSS, Chicago, IL). For survival studies, p values were obtained using the nonparametric Mantel-Cox test (Statview; Abacus Concepts, Berkeley, CA).

To examine sensitivity to LPS during the innate response to LCMV, LPS-induced lethality was assessed at day 2 of LCMV infection. Day 2 was chosen because at that time point, LCMV induces high levels of IFN-αβ with low to undetectable levels of IFN-γ protein. C57BL/6 mice were injected with 50, 75, or 100 μg of LPS i.p. and observed for 14 days. Increased lethality was seen in LCMV-infected mice compared with uninfected mice (Fig. 1, A–D). Thus, underlying early infection with LCMV resulted in increased sensitivity to LPS.

TNF-α is induced by LPS and is required for LPS-induced lethality. For this reason, it is presented here as an end-point to represent the magnitude of the inflammatory response. LCMV-infected mice were challenged at day 2 with i.p. LPS to examine inflammatory cytokine production during this dual challenge. Increased LPS-induced TNF-α was observed in the serum of LCMV-infected compared with uninfected mice (Fig. 2,A). Splenic and pooled peritoneal leukocytes from day 2 LCMV-infected mice also produced more TNF-α in response to ex vivo LPS stimulation compared with those from uninfected mice (Fig. 2, B and C). Therefore, LCMV sensitizes for both LPS-induced lethality and TNF-α production.

Because IFN-γ is known to contribute to lethality from LPS, we used mice deficient in NK, T, and/or B cells to demonstrate that early LCMV-induced sensitization for LPS can occur in the absence of IFN-γ-producing cells. Increased TNF-α induction was seen in LCMV-infected immunocompetent, E26 mice deficient in NK and T cells, and RAG-1^−/− mice deficient in T and B cells (Table I). Ab-mediated NK depletion also did not abolish virus-induced sensitization for LPS (Table I). In addition, under conditions where all uninfected mice survived, LCMV-infected E26 mice displayed similar LPS-induced lethality compared with C57BL/6 controls (83 vs 100%). Thus, LCMV-induced sensitization for LPS does not require the presence of T or NK cells, which are known to be producers of IFN-γ.

The innate production of IFN-γ is generally IL-12 dependent via signaling through STAT4 (44). By using mice genetically deficient in either IL-12p40 or STAT4, unable to produce innate IFN-γ in response to certain viral or LPS challenges, and by using IFN-γ-deficient mice, it was evident that virus-induced sensitization for LPS can occur in the absence of IFN-γ (Table II). Taken together, these data indicate that virus-induced sensitization for LPS does not require cells known to produce IFN-γ or IFN-γ itself.

Because IFN-αβ is detectable at high levels in early LCMV infection, we hypothesized that type 1 IFNs were responsible for LCMV-induced priming for LPS response. To assess this, we injected C57BL/6 mice with IFN-α followed by LPS at day 2. LPS-induced lethality was modestly higher in IFN-α-pretreated mice compared with vehicle-treated mice (100 vs 70%). Serum TNF-α levels were higher in IFN-α-treated mice compared with those treated with controls (Fig. 3,A). Splenic leukocytes and pooled peritoneal leukocytes obtained from mice 24 h after IFN-α treatment were similarly sensitized to LPS (Fig. 3, B and C). These data indicate that IFN-α can sensitize for LPS.

To determine whether IFN-αβ induced early during LCMV infection was responsible for sensitization for LPS, IFN-αβ R^+/+ and IFN-αβ R^−/− mice were injected with LPS at day 2 of LCMV infection. LPS priming for TNF-α was intact in LCMV-infected IFN-αβ R^−/− mice (Fig. 4 A).

Previous work from this group reported an early LCMV-induced IL-12 and IFN-γ response in the absence of IFN-αβ signaling (35). This is evident here in LCMV-infected IFN-αβ R^−/− mice who had higher serum IFN-γ compared with IFN-αβ R^+/+ controls (510 ± 102 vs 84 ± 15 pg/ml) and IL-12p40 (2033 ± 178 vs 338 ± 47 pg/ml) (p < 0.05). To prove that virus-induced IFN-γ revealed in this setting contributes to the sensitization for LPS seen in the absence of IFN-αβ signaling, the same experiment was performed examining IFN-αβγ R^+/+ (wild type), IFN-γ R^−/−, IFN-αβ R^−/−, and IFN-αβγ R^−/− mice. Only when both IFN-αβ and IFN-γ signaling at the receptor level is interrupted is LPS induction of TNF-α reduced (Fig. 4,A). Importantly, priming in the IFN-γ R^−/− mice is intact, demonstrating again that IFN-γ is not necessary for sensitization early in LCMV. A similar pattern of priming was seen in splenic and pooled peritoneal leukocytes isolated from LCMV-infected mice (Fig. 4, B and C). These data demonstrate that either LCMV-induced IFN-αβ or IFN-γ revealed in the absence of IFN-αβ signaling can mediate sensitization for LPS. Only by blocking both IFNs can this sensitization, as indicated by TNF-α production, be reduced.

STAT1 is a transcription factor important to signaling for both IFN-αβ and IFN-γ. Because the above experiments indicate that both IFN-αβ and IFN-γ can contribute to virus-induced sensitization for LPS, the role of signaling through their common transcription factor, STAT1, was examined. Mice genetically deficient in STAT1 provide a tool for assessing the importance of a pathway common to both IFNs. Reduced sensitization for LPS in vivo and ex vivo was observed during LCMV infection in STAT1^−/− compared with STAT1^+/+ mice (Fig. 5, A–C). These data, like that presented in IFN-αβγ R^−/− mice, also demonstrate that by disrupting IFN-αβ and IFN-γ signaling, LCMV-induced priming for LPS is significantly reduced.

This work demonstrated increased sensitivity to LPS during the innate response to LCMV. An ∼2-fold increase in LPS-induced lethality and a 2- to 6-fold increase in TNF-α production were seen in LCMV-infected compared with uninfected mice. Sensitization to LPS at day 2 of LCMV, when high systemic levels of IFN-αβ are present, did not require the presence of T, B, NK cells or IFN-γ. IFN-αβ R^−/− mice showed more sensitivity to LPS during LCMV infection; however, in the absence of IFN-αβ signaling, LCMV induces IL-12 and IFN-γ, which are known to enhance the sensitivity to LPS. For this reason, it was necessary to compare sensitization in IFN-αβγ R^+/+, IFN-γ R^−/−, and IFN-αβγ R^−/− mice to assess the role of IFN-αβ. In both the IFN-αβγ R^+/+ and IFN-γ R^−/− mice sensitization was intact. Only in the IFN-αβγ R^−/− mice was priming reduced. Mice deficient in STAT1, a transcription factor important in both IFN-αβ and IFN-γ signaling, also showed reduced priming for LPS during LCMV infection compared with STAT1^+/+ controls. Pretreatment of mice with IFN-α sensitized to the effects of LPS. These data demonstrate for the first time that IFN-αβ can play an underappreciated role in sensitization for LPS.

These data demonstrate several important points concerning the immune cross talk between viral infection and bacterial challenge. There are many examples clinically of coupling of viral infection with serious bacterial infection such as influenza with severe bacterial pneumonia and varicella with severe secondary skin infections. The link between bacterial and viral infection traditionally has been explained as opportunistic. Host defenses are compromised by injury to skin and epithelial barriers during viral infection providing a portal of entry for secondary bacterial pathogens. Our data extend the understanding of the impact of immune activation by virus on the magnitude of the inflammatory response and ultimate outcome to bacterial challenge. It is possible that immune activation from common viral infections routinely shapes the magnitude of the immune response to more dangerous bacterial challenges. Because the magnitude of the inflammatory response early in septic shock has correlated well in humans to outcome, it is plausible that the host IFN response to underlying viral infection may be a critical determinant of outcome in septic shock as well as other severe inflammatory states.

Many studies have shown that IFN-γ can augment LPS-induced TNF-α production and lethality (19, 29, 30, 31). In fact, previous work from this group demonstrated a 5- to 10-fold increase in TNF-α production and lethality from LPS when administered during the adaptive immune response to LCMV (4). At this time, the immune response to LCMV is characterized by the activation and expansion of CD8 T cells, which are potent IFN-γ producers. LPS-induced lethality during the adaptive response to LCMV infection required IFN-γ, NK, and T cells (4). In this report we demonstrate that IFN-αβ induced early during the innate response to LCMV can also augment LPS-induced lethality and TNF-α production by 4- to 5-fold. This pathway for sensitization to LPS is clearly different in that it does not require IFN-γ, NK, or T cells. Although virus-induced IFN-αβ is less effective than IFN-γ at sensitization for LPS, this work shows that viral infections can modulate the response to secondary inflammatory challenges at times when either IFN is present. The results raise important questions concerning potential detrimental conditions resulting from high dose IFN-α and IFN-β therapy.

Importantly, in this in vivo model of virus-induced IFN-αβ, our group previously reported that blocking of IFN-αβ signaling revealed the production of an earlier alternative pathway for virus-induced IL-12 and IFN-γ (34). Using this system, these observations have been extended to the sensitization for LPS at day 2 through either virus-induced IFN-αβ or through virus-induced IFN-γ revealed in the absence of IFN-αβ signaling. This observation is important for two reasons. First, it indicates redundancy in the IFN response to LCMV revealing two different pathways for enhancing sensitivity to endotoxin, which can only be examined by in vivo analysis. Second, it provides important insight into the mechanism of virus-induced sensitization for the response to LPS. Because IFN-αβ and IFN-γ can lead to priming, the pathways involved must be common to both. Signaling through STAT1, used by both IFN-αβ and IFN-γ, must occur to maximize LCMV-induced priming. These data point to molecular cross talk between IFN signaling and LPS signaling in an in vivo system. Although the signaling pathways for IFN-αβ, IFN-γ, and LPS are well described, overlap between these pathways has not been well defined (32). At what level this overlap occurs is unclear; however, IFN signaling to the point of activation of STAT1 is necessary for LPS priming. More importantly, these data implicate IFN-activated STAT1 as a key regulator of the magnitude of the response to LPS. This sequence of events may be critical in determining the magnitude of the response to second inflammatory challenges in viral infections as well as following the therapeutic use of IFN-αβ.

Although potentiation of the LPS response by IFN-γ has been extensively documented, the mechanism for this is poorly understood (19, 29, 30, 31, 45). Recently, Held et al. showed this and observed that pretreatment of a macrophage cell line with IFN-γ augmented LPS-induced DNA binding of NF-κB resulting in increased expression of inducible NO synthase mRNA and nitrite production (46). They did not determine a role for STAT1 in this priming but their work does begin to elucidate a molecular pathway for IFN-γ modulation of LPS signaling. Our work is the first to demonstrate sensitization for LPS through virus-induced IFN-αβ or IFN-γ and that this sensitization requires STAT1.

Our results can be contrasted to those of Tzung et al., who reported that IFN-α protected from endotoxin-induced mouse mortality (47). Those investigators chose different doses and types of LPS. Moreover, they used either 1 h pretreatment, simultaneous, or post-LPS treatment with recombinant IFN-α to diminish mortality as well as TNF-α production by Kupffer cells. Their results may be related to the studies from our group demonstrating that IFN-αβ can act to inhibit IL-12 induction (34) as well as responsiveness to IL-12 for IFN-γ induction (35). However, here we demonstrate that LCMV-induced IFN-αβ signaling through STAT1 enhances rather than inhibits the response to LPS. The differences between the reported IFN-αβ effects may be related to dose, type, or timing of LPS. Although phosphorylation of STAT1 and its binding to DNA occur quickly, the necessary events for endotoxin sensitization may require more time than examined by Tzung et al. Teleologically, the IFN-αβ-enhancing rather than -inhibitory effects seen in LPS sensitization may exist to allow the host to respond to overwhelming bacterial infections during virus-immune activation.

We attempted to demonstrate decreased LPS-induced lethality during LCMV infection in the absence of IFN signaling. The results of these experiments were somewhat puzzling. LCMV infection alone was nonlethal for IFN-αβγ R^+/+, IFN-αβ R^−/−, IFN-γ R^−/−, and IFN-αβγ R^−/− mice. Although LCMV-infected, LPS-treated IFN-αβγ R^−/− mice showed increased average survival times compared with IFN-αβγ R^+/+, IFN-αβ R^−/−, and IFN-γ R^−/− mice, 100% lethality was eventually seen in all groups. Conversely, pretreating with IFN-α to increase LPS-induced lethality was at best marginally successful (see Results). Thus, although IFNs play an important role, these data indicate that other factors are also promoting sensitization to LPS during viral infection.

In summary, these studies extend previous work from this group describing immune modulation by underlying viral infection leading to sensitization to a subsequent bacterial product challenge. Here, we show that virus-induced sensitization to endotoxin can occur through either IFN-αβ or IFN-γ due to overlapping signaling through STAT1. Thus, pathways for critical immune cross talk capable of determining the magnitude and outcome of the response to sequential inflammatory challenges are defined. In humans, the outcome of bacterial septic shock correlates to the magnitude of the inflammatory response. Thus, these studies suggest that underlying viral infection leading to the induction of IFN-γ or IFN-αβ or their therapeutic administration resulting in STAT1 activation may be critical to outcome by modulating the sensitivity to subsequent inflammatory stimuli.

We thank Drs. Thais Salazar-Mather, Melanie Ruzek, and Leslie Cousens along with Gary Pien and Stacey Carlton for help with experiments and stimulating discussions, and Dr. Phillip Scott for his gift of rabbit anti-mouse IFN-γ sera.