Animal Model of Respiratory Syncytial Virus: CD8⁺ T Cells Cause a Cytokine Storm That Is Chemically Tractable by Sphingosine-1-Phosphate 1 Receptor Agonist Therapy

ABSTRACT

The cytokine storm is an intensified, dysregulated, tissue-injurious inflammatory response driven by cytokine and immune cell components. The cytokine storm during influenza virus infection, whereby the amplified innate immune response is primarily responsible for pulmonary damage, has been well characterized. Now we describe a novel event where virus-specific T cells induce a cytokine storm. The paramyxovirus pneumonia virus of mice (PVM) is a model of human respiratory syncytial virus (hRSV). Unexpectedly, when C57BL/6 mice were infected with PVM, the innate inflammatory response was undetectable until day 5 postinfection, at which time CD8⁺ T cells infiltrated into the lung, initiating a cytokine storm by their production of gamma interferon (IFN-γ) and tumor necrosis factor alpha (TNF-α). Administration of an immunomodulatory sphingosine-1-phosphate (S1P) receptor 1 (S1P1R) agonist significantly inhibited PVM-elicited cytokine storm by blunting the PVM-specific CD8⁺ T cell response, resulting in diminished pulmonary disease and enhanced survival.

IMPORTANCE A dysregulated overly exuberant immune response, termed a “cytokine storm,” accompanies virus-induced acute respiratory diseases (VARV), is primarily responsible for the accompanying high morbidity and mortality, and can be controlled therapeutically in influenza virus infection of mice and ferrets by administration of sphingosine-1-phosphate 1 receptor (S1P1R) agonists. Here, two novel findings are recorded. First, in contrast to influenza infection, where the cytokine storm is initiated early by the innate immune system, for pneumonia virus of mice (PVM), a model of RSV, the cytokine storm is initiated late in infection by the adaptive immune response: specifically, by virus-specific CD8 T cells via their release of IFN-γ and TNF-α. Blockading these cytokines with neutralizing antibodies blunts the cytokine storm and protects the host. Second, PVM infection is controlled by administration of an S1P1R agonist.

INTRODUCTION

Of the 450 million humans with pneumonia each year, approximately four million die (1). A large proportion of respiratory diseases has been attributed to viral infection, and 95% of nasal aspirates from children with respiratory infections are positive for virus (1–4). The human paramyxovirus human respiratory syncytial virus (hRSV) was found in more than 50% of children under the age of 15 afflicted with pneumonia (2). At least 30 million children under the age of 5 become infected with hRSV per year, resulting in nearly 200,000 deaths worldwide (5). In addition, hRSV infection of elderly individuals has become an increasing medical problem (5).

Currently, attempts to treat RSV have been unsatisfactory. Administration of the nucleoside analogue ribavirin has limited efficacy for inhibiting hRSV replication and is often associated with severe side effects. The cytokine storm is a major component of severe respiratory infections, such as those from hRSV; consequently, targeting the hosts' immune response is an alternate strategy (6–8). However, suppression of the hosts' immune response can subvert mechanisms required to control virus replication. For instance, corticosteroids have been used to treat various pulmonary infections, but their broad anti-inflammatory effects can hamper the host's ability to control infection. The outcome can exacerbate virally induced pulmonary injury and may prolong viral shedding that can exaggerate disease (9–11).

“Cytokine storm” defines a combination of cytokines and cellular components that result in an excessive and aberrant inflammatory response that damages host tissues, participating in the enhanced morbidity and mortality. This phenomenon has been documented during infections with influenza virus, hRSV, hantavirus, and severe acute respiratory syndrome coronavirus (SARS-CoV) (8). Mechanistically, virus infection induces the rapid production of type I interferons (IFN), cytokines essential for the production of additional proinflammatory cytokines and stimulation of immune cell activation that consequently amplifies the inflammatory response (8, 12). In addition to cytokines, cells such as dendritic cells (DCs), macrophages, epithelial cells, and endothelial cells play prominent roles in the early antiviral inflammatory response that can damage pulmonary tissues (13–15). Identifying the immune components that are required for the initiation and amplification of a cytokine storm is essential for developing therapeutics at various stop points to alleviate pulmonary injury. Previously, we demonstrated that dampening but not abrogating an influenza virus-induced cytokine storm by utilization of the sphingosine-1-phosphate (S1P) signaling pathway provided significant amelioration of pulmonary inflammation and host survival by limiting immunopathologic injury without compromising the antiviral immune response that controls and eradicates the infection (15–17). S1P is a lysophospholipid ligand for the S1P receptors 1 to 5 (S1P1R to -5R) and plays a role in multiple cellular immunobiological processes, including cytokine secretion, proliferation, adhesion, migration, survival, endocytosis, and endothelial cell barrier function (18–20) (21). Hence, the design and implementation of therapeutic strategies that target the S1P signaling pathway may prove useful for combating a variety of acute respiratory diseases caused by viruses and microbes in which the cytokine storm plays a major pathological role.

PVM is a rodent paramyxovirus used to investigate hRSV pathogenesis. PVM and hRSV are paramyxoviruses; both induce a robust respiratory cytokine storm in their respective hosts, and the intensity of the inflammatory response correlates directly with disease severity (22). Several factors, including CD8⁺ T cells, neutrophils, the chemokine receptor CCR1, and its ligand CCL3, have been shown to exacerbate pulmonary injury following PVM infection (23, 24).

Here we investigated factors that participate in the PVM-induced cytokine storm in C57BL/6 mice. Remarkably, inflammation was undetectable in PVM-infected mice until day 5 postinfection, at which time activated CD8⁺ T cells infiltrated into the lung and secreted tumor necrosis factor alpha (TNF-α) as well as IFN-γ, which initiated the cytokine storm. Oral administration of the immunomodulatory S1P1R agonist RP-002 inhibited the recruitment of activated CD8⁺ T cells, blunted the production of both TNF-α and IFN-γ, and lessened pulmonary injury without compromising control of viral replication. Thus, the paramyxovirus-induced cytokine storm within C57BL/6 mice was dependent on and initiated by CD8⁺ T cells and occurred in the absence of a detectable early innate immune response. Disruption of CD8⁺ T cells, which induced and maintained the cytokine storm, was chemically tractable, with oral S1P1R agonist therapy resulting in significant blunting of disease.

MATERIALS AND METHODS

Mice, virus, and compound.

C57Bl/6J mice, 6 to 8 weeks old, were obtained from The Scripps Research Institute (TSRI) rodent breeding colony and maintained under pathogen-free conditions. Mice were prescreened for PVM infection, and any PVM -positive animals were excluded from the breeding colony. Animal handling conformed to the requirements of the National Institutes of Health and TSRI animal research committee. Recombinant PVM (rPVM) strain 15 (25) was propagated on BHK-21 cells and stored according to published methods (39). Stock virus titers were quantified by plaque assay using BSC-1 cells in the presence of 0.001% of γ-chymotrypsin (40). Mice were infected intranasally (i.n.) with 1.5 × 10³ or 3 × 10³ PFU of rPVM strain 15. RP-002 (Receptos, La Jolla, CA, USA) was dissolved in sterile water at a concentration of 0.75 mg/ml. Six mg/kg of body weight of RP-002 or an equivalent volume of sterile water (vehicle) was administered orally 1, 25, and 49 h following infection. RP-002, (R)-2-(4-(5-(3-cyano-4-isopropoxyphenyl)-1,2,3-oxadiazol-3-yl)-2,3-dihydro-IH-inden-l-ylamino)-N,N-dimethylacetamide hydrochloride, was synthesized according to the published method (E. Martinborough, M. F. Boehm, A. R. Yeager, J. Tamiya, L. Huang, E. Brahmachary, M. Moorjani, G. A. Timony, J. L. Brooks, R. Peach, F. L. Scott, and M. A. Hanson, 19 May 2011, international patent application PCT/US2010/056760). The compound had a 50% effective concentration (EC₅₀) for S1P1 of 0.13 nM, was >100-fold selective versus S1P5, and was 10,000-fold selective versus S1P2, -3, and -4 when assayed as described previously (41). Mice were weighed daily and euthanized if >30% of their initial body weight was lost. Lungs and bronchoalveolar lavage fluids (BALF) were harvested from euthanized mice and frozen at −80°C until used.

Cytokine analysis.

The tracheas of euthanized mice were exposed, transected, and intubated with a blunt 18-gauge needle. One milliliter of phosphate-buffered saline supplemented with Complete Mini, EDTA-free protease inhibitor cocktail (Roche, Indianapolis, IN, USA) was infused three times, and the fluid was then recovered. Harvested BALF was clarified by centrifugation at 3,000 × g for 3 min at 4°C, and the supernatant was stored at −80°C. Cytokine content in the BALF was analyzed by enzyme-linked immunosorbent assays (ELISAs) for cytokine content, i.e., murine CCL2, CCL5, CXCL10, interleukin 1 (IL-1), IL-6, TNF-α, and IFN-γ, with Duoset kits (R&D Systems, Minneapolis, MN, USA). IFN-α was quantified using the VeriKine mouse IFN-α ELISA kit (Pestka Biomedical Laboratories, Inc., Piscataway, NJ, USA).

Antibody administration.

Two hundred fifty micrograms of anti-TNF-α (XT3.11; Bioxcell, West Lebanon, NH, USA) and/or 300 μg of anti-IFN-γ (R4-6A2; Bioxcell) was administered to mice through the intraperitoneal route (i.p.) on days 3 and 4 postinfection for cytokine neutralization. Isotype control antibody, mouse IgG1 (HRPN; Bioxcell), was administered at concentrations equivalent to those of the aforementioned antibodies. For cell depletion experiments, 250 μg of anti-NK1.1 (PK136; Bioxcell) and/or 500 μg of anti-CD8 (53-6.72; Bioxcell) antibodies were administered i.p. to mice 1 day before and 1 day after infection. Rat IgG2a (2A3; Bioxcell) was administered at equivalent concentrations for isotype control-treated mice within the cell depletion experiments.

Cellular analysis by flow cytometry.

Lung and mesenteric lymph nodes (MLN) were harvested from mice perfused with ∼10 ml of sterile phosphate-buffered saline (PBS). Single-cell suspensions from the lung were obtained using the MACs tissue dissociator and lung dissociation protocol (Miltenyi, Gladbach, Germany), washed, resuspended in cell culture medium, and quantified using a hemocytometer and the trypan blue exclusion method. Isolated cells were stained with antibodies raised against murine CD11b (M1/70), CD11c (HL3), Ly6C (HK1.4), Ly6G (1A8), F480 (BM8), NK1.1 (PK136), CD3e (145-2C11), CD4 (L3T4), CD8a (53-6.7), CD90.2 (53-2.1), B220 (RA3-6B2), CD103 (M290), CD69 (HI.2F3), CD44 (IM7), CD40 (MH40-3), CD80 (16-10A1), CD86 (GL1), PD-L1 (MIH5), H-2Kb (AF6-88.5), and I-A/I-E (M5/114.15.2). Antibodies were added to cells at dilutions of 1:100 to 1:400 in fluorescence-activated cell sorting (FACS) buffer (PBS supplemented with 1% fetal bovine serum and 0.05% sodium azide) for 20 to 30 min. Cells were then washed twice with 200 μl FACS buffer, fixed with 2% paraformaldehyde, washed again once with FACS buffer, and then analyzed by flow cytometry.

Intracellular staining.

Cytokine production by NK cells and CD8⁺ T cells within the lung on day 5 after PVM infection was determined by 5 h of incubation of single-cell suspensions in culture medium (RPMI with 10% fetal bovine serum [FBS], 2 mM l-glutamine, and 100 U [each] of penicillin-streptomycin) with 4 μg/ml brefeldin A (Sigma, St. Louis, MO, USA) in the absence of exogenous stimulation. NK cells and CD8⁺ T cells were stained for surface antigen (NK1.1⁺ CD3⁻ and CD8⁺ CD90.2⁺, respectively), fixed, permeabilized with 2% saponin, and then stained for intracellular proteins with fluorescently labeled antibodies specific for murine IFN-γ (XMG1.2) and TNF-α (MP6-XT22). For PVM-specific CD8⁺ T cell analysis, lung single-cell suspensions were suspended in culture medium at a concentration of 1 × 10⁶ to 2 cells per well. Plated cells were incubated for 5 h in the presence of 1 μg/ml PVM immunodominant epitopes L_1025–1059, N_339–347, or an irrelevant peptide (11), as well as 4 μg/ml brefeldin A. CD8⁺ T cells were stained for surface antigen, and intracellular proteins were stained with fluorescently labeled antibodies specific for murine IFN-γ (XMG1.2), TNF-α (MP6-XT22), and IL-2 (JES6-5H4) as described above. Intracellular granzyme B (Life Technologies, Carlsbad, CA, USA) staining was performed directly ex vivo in the absence of stimulation or incubation with brefeldin A. T regulatory (T_reg) cells were determined by Thy1.2, CD4, and CD25 positivity as well as intracellular FoxP3 positivity. Intracellular FoxP3 staining was performed as stated above. Flow cytometric analysis was performed using a digital LSR II instrument (Becton, Dickinson, Franklin Lakes, NJ, USA), and data were analyzed using the software program FlowJo (Tree Star, Inc., Ashland, OR, USA). Absolute numbers of cells were determined by multiplying the frequency of specific cell populations by the total number of viable cells.

Histopathology and determination of lung exudates.

Tissues were harvested and placed in PBS-buffered formalin. Lungs were blocked in paraffin, and 10-μm tissue sections were cut, placed on glass slides, and stained with hematoxylin and eosin. Slides were analyzed by three separate pathologists, who were blinded to the various experimental treatments. The grading score for pulmonary injury and degree of inflammatory infiltration into the lung was recorded in our prior publications (17). Total protein content in the BALF was assessed using the Pierce protein BCA assay kit (Thermo Fisher Scientific, Waltham, MA, USA). IgM content in the BALF was calculated using the mouse IgM quantitation kit (Bethyl Laboratories, Montgomery, TX, USA). Lactate dehydrogenase (LDH) enzymatic activity was determined using the Cytotox 96 nonradioactive cytotoxicity assay (Promega Corporation, Inc., Madison, WI, USA).

Statistics.

For survival studies, a Gehan-Breslow-Wilcoxon test (GraphPad Prism 5; La Jolla, CA, USA) was performed, and a P value of 0.05 (95% confidence level) was deemed significant. An unpaired two-tailed Student t test was calculated using the program Excel (Microsoft, Redmond, WA, USA), when appropriate, to determine significance, which was set at 5%.

Determination of PVM titers.

PVM titers for mouse lung samples were determined by plaque assay on BHK-21 cells under a 0.8% methylcellulose overlay. Plaques were visualized by immunostaining with rabbit antiserum raised against sucrose-gradient-purified PVM followed by a horseradish peroxidase-labeled goat anti-rabbit IgG secondary antibody (KPL, Gaithersburg, MD, USA) (26). Bound antibodies were detected by incubation with a peroxidase substrate (KPL).

Cell isolation from blood and MLN.

Cells were isolated from MLN by mechanical disruption through a 100-μm nylon mesh filter. Two hundred fifty microliters blood was collected from the retro-orbital sinus before perfusion for isolation of peripheral blood mononuclear cells (PBMCs). Red blood cells were lysed by exposing whole blood to 10 ml of a selective red blood cell lysis buffer (ammonium chloride-Tris-HCl, 0.02 M; NH₄Cl, 0.14 M). Cells were then surface stained for flow cytometric analysis or stimulated and then stained intracellularly as described above.

RESULTS

TNF-α and IFN-γ are essential for initiating a cytokine storm during PVM infection.

To characterize the cytokine(s) that exacerbate the PVM-induced cytokine storm, we infected C57BL/6 mice with 1.5 × 10³ PFU of recombinant PVM (rPVM) strain 15 i.n. to induce an inflammatory response and then assessed the types, kinetics, and amounts of cytokines produced within the bronchoalveolar lavage fluid (BALF). The cytokines evaluated were CCL2, CCL5, CXCL10, IL-1α, IL-6, TNF-α, IFN-α, and IFN-γ. All except TNF-α were at background levels until day 5 after infection, at which point all cytokines were elevated above background (Fig. 1A and B). TNF-α was found on day 4 after infection, whereas the cytokine produced at the highest concentration at any time point was IFN-γ (Fig. 1A). This result indicated that the cytokine storm was primarily initiated 5 days following PVM infection, with TNF-α and IFN-γ likely major participants in the process (Fig. 1A and B). Importantly, enhanced levels of PVM over the amount inoculated were detectable within the lung at day 3 postinfection, demonstrating the presence of viral progeny in the absence of robust inflammation at this time point (Fig. 1D). To evaluate the contributions of TNF-α and IFN-γ for the initiation of a PVM-induced cytokine storm, neutralizing antibodies to these cytokines were administered to PVM-infected mice on days 3 and 4 postinfection. Cytokine levels were evaluated in the BALF at day 5 postinfection. The resulting blockade of TNF-α and/or IFN-γ significantly inhibited the secretion of IFN-α (Fig. 1C) and all other cytokines assayed (Fig. 1C and E) when evaluated against results for isotype control-treated mice. In addition, anti-TNF-α antibody administration significantly decreased IFN-γ content, and IFN-γ neutralization limited TNF-α production, demonstrating a dependence of these cytokines on each other's production. Corresponding with these results, innate immune cell recruitment was inhibited by anti-TNF-α and/or anti-IFN-γ antibody administration (Fig. 1F). Numbers of innate immune cells expressing the activation molecules CD69 (Fig. 1G) and major histocompatibility complex (MHC) class II (Fig. 1H) were significantly reduced when TNF-α and/or IFN-γ was neutralized. These results demonstrated that the cytokine storm was not initiated during the early innate immune response but late following PVM infection, at day 5, and secretion of TNF-α as well as IFN-γ was required and essential to commence the cytokine storm.

FIG 1.

TNF-α and IFN-γ are required for initiation of the PVM-induced cytokine storm. (A) IFN-α, TNF-α, and IFN-γ kinetic profiles within the BALF from C57BL/6 mice infected i.n. with 1.5 × 10³ PFU of rPVM strain 15. *, P < 0.05 when judged against day 1 postinfection results for respective cytokines. n = 2 (day 0) or 4 or 5 (all other time points). (B) rPVM strain 15 (1.5 × 10³ PFU) was instilled i.n. in C57BL/6 mice, and kinetic analysis was performed for CCL2, CCL5, CXCL10, and IL-1α (4 mice per group at each time point). (C) C57BL/6 mice were infected i.n. with 3 × 10³ PFU rPVM strain 15 and administered isotype control antibody, anti-TNF-α, and/or anti-IFN-γ neutralizing antibodies. Mice were euthanized on day 5 after infection, and BALF as well as lungs were harvested. (C) IFN-α, TNF-α, and IFN-γ content in the BALF. (D) Kinetic analysis of PVM lung burden by plaque assay. Viral titers were determined by plaque assay in duplicate for each time point. The limit of detection was 1.7 log₁₀ PFU per g of tissue. Circles signify individual mice, and the averages are represented by bars (n = 3 or 4 mice per group per time point). (E) Quantification of cytokines secreted into the BALF from infected mice that received TNF-α and/or IFN-α neutralizing antibody on day 5 postinfection. (F) Enumeration of innate inflammatory cells in the lung by analysis of single-cell suspensions from the lung. (G) Quantification of macrophages/monocytes (F480⁺, CD11b⁺, Ly6G⁻) and NK cells (NK1.1⁺, CD3⁻) positive for the early activation marker CD69. (H) Enumeration of MHC class II-positive macrophages/monocytes. Four or five mice were used per group. Data are representative of 2 independent experiments. Bar graphs indicate the averages ± SEM. *, P < 0.05; **, P < 0.005; ***, P < 0.0005.

PVM-induced cytokine storm requires CD8⁺ T cells.

Next, to identify the cellular sources of TNF-α and IFN-γ, we evaluated the kinetics of immune cell infiltration within the lung following PVM infection. No innate immune cell infiltration was significantly above the background level until day 5 postinfection or later (Fig. 2A), a time that coincided with emergence of cytokine secretion (Fig. 1A and B). Further, innate immune cells expressing activation markers (Fig. 2B) as well as effector CD44⁺ CD8⁺ T cells (Fig. 2C) were found in significantly greater numbers on day 5 after infection. TNF-α is secreted by many cell types, but NK cells and T cells are primary sources of IFN-γ (27). Both activated CD69⁺ NK cells (Fig. 2B) and effector CD44⁺ CD8⁺ T cells (Fig. 2C) were present above background amounts in the lung on day 5 following infection with PVM. Indeed, TNF-α⁺ as well as IFN-γ⁺ NK cells (Fig. 2D) and CD8⁺ T cells (Fig. 2E) occupied the lungs of PVM-infected mice in significantly greater frequencies than were found for uninfected mice. These results suggested that activated NK cells and CD8⁺ T cells produced TNF-α and IFN-γ within the lungs on day 5 after infection, the time when PVM-induced cytokine storm began.

FIG 2.

CD8⁺ T cell depletion abrogates TNF-α and IFN-γ production and blunts the cytokine storm. (A to C) C57BL/6 mice were infected with 1.5 × 10³ PFU of rPVM strain 15 i.n. Kinetic analysis of immune cell recruitment and activation marker expression was performed with single-cell suspensions from the lung. (A) Quantification of macrophage/monocyte, neutrophil and NK cell content. (B) Enumeration of CD69⁺ and MHC class II-positive macrophages/monocytes as well as CD69⁺ NK cells. (C) Determination of activated CD44⁺ CD8⁺ T cell content. Bar graphs represent averages ± SEM. *, P < 0.05 versus values on day 1 after infection. Lung and BALF were harvested from uninfected or rPVM strain 15-infected (3 × 10³ PFU) C57BL/6 mice on day 5 postinfection. Intracellular TNF-α and IFN-γ production by pulmonary NK cells (D) or CD44⁺ CD8⁺ T cells (E) was determined. Values within individual dot plots signify the average frequencies ± SEM for the gated populations. Quantification of NK cells, CD4⁺ cells, and CD8⁺ T cells within the lung (F) or of cytokine production within the BALF (G) was done on day 5 after infection. (H to J) Antibody depletion of NK cells and/or CD8⁺ T cells within infected mice. (H) Quantification of TNF-α and IFN-γ content in the BALF. Enumeration of macrophages/monocytes and neutrophils (I) as well as total numbers of macrophages/monocytes positive for the indicated activation markers (J) is shown. Data are representative of 2 or 3 individual experiments with 4 or 5 mice per group. Bar graphs represent averages ± SEM where applicable. *, P < 0.05; **, P < 0.005; ***, P < 0.0005.

Next, we employed antibodies to deplete NK cells and CD8⁺ T cells to determine their participation in initiating the PVM-induced cytokine storm. The antibodies used depleted >98% of NK cells and of CD8⁺ T cells but did not affect CD4⁺ T cells (Fig. 2F). Antibody depletion of NK cells did not significantly reduce the cytokine content (Fig. 2H and G). In contrast, depletion of CD8⁺ T cells resulted in significant reductions of both TNF-α and IFN-γ (Fig. 2H), as well as IL-1α and IL-6 (Fig. 2G). Antibody depletion of both NK cells and CD8⁺ T cells provided the greatest blunting of all cytokines (Fig. 2H and G). IFN-α production was significantly hampered (Fig. 2G) when NK cell and CD8⁺ T cells were both depleted, demonstrating that the recruitment of both cell populations is likely necessary for IFN-α secretion. Depletion of CD8⁺ T cells significantly decreased the number of innate inflammatory cells (Fig. 2I), as well as expression of activation molecules on such cells (Fig. 2J). NK cell depletion reduced the recruitment of activated immune cells to a lesser extent than CD8⁺ T cell depletion (Fig. 2J). Similar to the neutralization of cytokines, depletion of NK cells and CD8⁺ T cells together was more effective in limiting innate immune cell recruitment (Fig. 2I) and activation (Fig. 2J) than either NK cell or CD8⁺ T cell depletion alone. In contrast, depletion of CD4 T cells had minimal effects on type I and II interferon production or morbidity following PVM infection (data not shown). These results indicate that CD8⁺ T cells are the dominant cell type for initiating a cytokine storm following PVM infection, and both CD8⁺ T cells and NK cells together provided the maximal response with minimal involvement of CD4 T cells.

S1P1R agonist inhibits PVM-induced cytokine storm.

TNF-α, IFN-γ, CD8⁺ T cells, and NK cells contribute to the intensity of the cytokine storm following PVM infection. Previously, we demonstrated that the S1P1R agonist RP-002 inhibited production of TNF-α and IFN-γ as well as the recruitment of activated, CD69⁺ NK cells 1 to 3 days following influenza virus infection (15). We now evaluated the effects of administration of RP-002 or vehicle at 1, 25, and 49 h following infection with 3 × 10³ PFU of PVM. In flow cytometric analysis, lung infiltrates of mice given RP-002 showed significantly fewer CD8⁺ T cells and NK cells than vehicle on day 5 postinfection (Fig. 3A). In addition, fewer activated CD44⁺ CD8⁺ T cells and CD69⁺ NK cells were retrieved from the lungs of RP-002-treated mice (Fig. 3B). Effector CD44⁺ CD8⁺ T cells and NK cells from RP-002 recipients exhibited significantly diminished frequencies of cells positive for granzyme B, a molecule produced following activation and essential for cytolytic activity (Fig. 3C). When single-cell suspensions of lung cells from RP-002 recipients 5 days after PVM infection were stimulated with PVM-specific H-2^b immunodominant CD8⁺ T cell epitopes (28), the frequency of IFN-γ⁺ cells was not enhanced compared to that for similar cells treated with a peptide that was irrelevant to PVM infection (data not shown). Analysis of PVM-specific CD8⁺ T cells within the draining mediastinal lymph nodes (MLN) in RP-002-treated mice on day 5 postinfection revealed significantly fewer PVM-specific CD8⁺ T cells after stimulation with the PVM immunodominant peptide (Fig. 3D) (28) than for mice that received vehicle. Further, after PVM peptide stimulation, IFN-γ⁺ CD8⁺ T cells from RP-002 recipients had fewer singly TNF-α-producing or dually TNF-α/IL-2-producing cells (Fig. 3E), reduced percentages of granzyme B⁺ cells (Fig. 3E), decreased TNF-α and IL-2, and lower granzyme B content within cells positive for their respective effector molecules (Fig. 3F).

FIG 3.

The S1P1R agonist RP-002 inhibits initiation of the PVM-induced cytokine storm. C57Bl/6J mice were infected i.n. with 3 × 10³ PFU of rPVM strain 15 and given orally vehicle or 6 mg/kg of RP-002 1, 25 and 49 h postinfection. On day 5 after infection, the mice were euthanized, and BALF as well as lungs were harvested for analysis. (A) Enumeration of total CD8⁺ T cells (CD8⁺, Thy1.2⁺, CD4⁻) and NK cells (NK1.1⁺, CD3⁻) in single-cell lung suspensions. (B) Quantification of pulmonary CD8⁺ T cells and NK cells that expressed the activation markers CD44 and CD69, respectively. (C) MFI of granzyme B within CD8⁺ T cells and NK cells. (D to F) MLN were harvested, and single cell suspensions were cultured in the presence of the PVM-specific immunodominant peptide L_1052–1060 and brefeldin A for 5 h. (D) Calculated total numbers of IFN-γ⁺ CD8⁺ T cells. (E) Representative dot plots of cells producing TNF-α and IL-2 (above) and granzyme B (below) from the IFN-γ⁺ CD8⁺ T cell population. Values for the gated populations signify the average frequencies ± SEM of cells positive for their respective effector molecules. (F) MFI of IFN-α, TNF-α, IL-2, and granzyme B from CD8⁺ T cells positive for their respective effector molecules. (G) Quantification of cytokine content in the BALF by ELISA. (H) Enumeration of macrophages/monocytes (CD11b⁺, F480⁺, Ly6G⁻) and neutrophils (CD11b⁺, Ly6G⁺, F480⁻) in the lung. (I) Representative dot plots depicting macrophage/monocyte expression of CD69 (left plots) and MHC class II (right plots) within experimental groups. Numbers within individual panels represent the average frequencies ± SEM of gated cells positive for their respective activation molecules. For panels A to F, 4 or 5 mice were used per group. Data are representative of 2 to 4 independent experiments. Bar graphs indicate the averages ± SEM where applicable. *, P < 0.05; **, P < 0.005; ***, P < 0.0005.

RP-002 treatment of PVM-infected mice significantly inhibited the secretion of TNF-α and IFN-γ, as well as the cytokines CCL2, CCL5, CXCL10, IL-1α, and IL-6, within the BALF on day 5 postinfection (Fig. 3G). IFN-α production was not significantly affected by RP-002 treatment (Fig. 3G) and was therefore by itself not sufficient to stimulate a cytokine storm following PVM infection. In correlation with reduced cytokine expression, macrophage/monocyte and neutrophil recruitment diminished significantly in RP-002-treated mice (Fig. 3H). Frequencies of CD69⁺ and MHC class II-positive macrophages/monocytes were decreased within RP-002-treated mice (Fig. 3I). These results indicated that oral administration of the S1P1R agonist RP-002 inhibited the cellular and cytokine factors required for the initiation of a PVM-induced cytokine storm.

S1P1R signaling inhibits PVM-specific CD8⁺ T cell recruitment and proinflammatory cytokine production.

CD8⁺ T cells are required for elimination of PVM, but their antiviral activity elicits severe pulmonary injury that can lead to death (24). To assess this process, RP-002 was administered, and the effect on pulmonary PVM-specific CD8⁺ T cells was quantified on day 8 after infection with 3 × 10³ PFU of PVM. The presence of RP-002 significantly decreased the infiltration of activated effector CD44⁺ CD8⁺ T cells into the lung compared to the values in vehicle recipients (Fig. 4A). The numbers of IFN-γ⁺ CD8⁺ T cells following stimulation with the PVM immunodominant peptides L_1052–1060 (Fig. 4B) and N_339–347 (28) were also significantly lower in mice that received RP-002 than in vehicle recipients. IFN-γ⁺ CD8⁺ T cells from RP-002-treated mice compared to those from vehicle-treated mice displayed increased frequencies of IFN-γ-only producers, reduced percentages of TNF-α and TNF-α/IL-2-producing cells, and fewer cells producing granzyme B (Fig. 4C). Further, the mean fluorescence intensity (MFI) of TNF-α from TNF-α⁺ CD8⁺ T cells from RP-002 recipients was significantly lower than that for vehicle-treated mice (Fig. 4D). Last, reductions in CD8⁺ T cell numbers and antiviral potential correlated with significantly fewer macrophages/monocytes detectable within the lung. Only a minor reduction was observed in the neutrophil population (Fig. 4E).

FIG 4.

RP-002 reduced PVM-specific CD8⁺ T cell responses. C57Bl/6J mice were infected with 3 × 10³ PFU of rPVM strain 15 and treated with vehicle or 6 mg/kg RP-002 orally 1, 25, and 49 h following infection. Lungs were removed on day 8 postinfection for cellular analysis. (A) Enumeration of activated CD44⁺ CD8⁺ T cells. (B to E) Lung single-cell suspensions from experimental mice were cultured with the PVM immunodominant peptide L_1052–1060 (1 mg/ml) and brefeldin A. Cells were incubated for 5 h and assessed for CD8⁺ T cell intracellular cytokine production. (B) Quantification of total IFN-γ⁺ CD8⁺ T cells following peptide stimulation. (C) Representative dot plots of TNF-α- and IL-2-producing (above) and granzyme B-producing (below) cells from the IFN-γ⁺ CD8⁺ T cell population. Values for the gated populations signify the average frequencies ± SEM of cells positive for their respective effector molecules. (D) MFI of TNF-α from TNF-α⁺ CD8⁺ T cells. (E) Calculated total numbers of macrophages/monocytes and neutrophils. (F) Quantification of IFN-γ⁺ CD8⁺ T cells from the MLN. (G) Representative dot plots of IFN-γ⁺ CD8⁺ T cells depicting TNF-α and IL-2 (above) as well as granzyme B (below) production. Numbers within panels signify the average frequencies ± SEM for gated populations. (H) MFI of effector molecules for CD8⁺ T cells positive for their respective antiviral proteins. Four to seven mice were used per group. Data are representative of 3 independent experiments. Bar graphs indicate the averages ± SEM where applicable. *, P < 0.05; **, P < 0.005; ***, P < 0.005.

As observed at day 5 in the MLN and day 8 in the lung, RP-002-treated mice had fewer PVM-specific CD8⁺ T cells (Fig. 4F), and such cells displayed a reduction in TNF-α, IL-2, and granzyme B⁺ expression compared to results for vehicle-treated mice (Fig. 4G). In addition, RP-002 treatment resulted in significantly diminished IFN-γ, TNF-α, IL-2, and granzyme B content within PVM-specific CD8⁺ T cells positive for their respective effector molecules (Fig. 4H). Thus, the S1P1R agonist inhibited PVM-specific CD8⁺ T cell differentiation and effector molecule generation within the MLN that preceded the diminished CD8⁺ T cell-mediated inflammation in the lung.

S1P1R agonist limits PVM-induced pulmonary injury, leading to decreased morbidity and enhanced survival.

The last series of experiments tested whether RP-002 treatment would limit immunopathologic injury and promote host survival. To this end, we performed histopathologic analysis of the lung 8 days following infection for RP-002-treated infected and uninfected mice (Fig. 5A). First, we examined lungs from four uninfected C57BL/6 mice receiving either vehicle or RP-002 orally. In each instance, the alveolar air sac structure was preserved, there was no evidence of hemorrhage, and inflammatory infiltrate was negligible. From these findings, we concluded that first our mouse colony was clean and devoid of any pulmonary injury or infection. Second, oral administration of RP-002 molecules to uninfected C57BL/6 mice did not produce significant pulmonary injury. We then studied five lungs each from PVM-infected mice which received either vehicle or RP-002 therapy. While there were scattered patching locations of pulmonary injury in both groups, lungs from those infected mice receiving just vehicle showed frequent consolidation, hemorrhage, destruction of alveolar air space structures often leading to ballooning of air sacs, and heavy mononuclear lymphoid cell infiltration. On a gradation of 1 to 4 (see Materials and Methods), the average score was 3.6 ± 0.1. This compares to 1.5 ± 0.1 to 1.6 ± 0.1 for lungs from the two groups of uninfected mice studied (the 1.6 ± 0.1 value is for uninfected mice receiving RP-002). Lungs from PVM-infected mice given RP-002 infrequently displayed consolidation, and hemorrhage was not found. The majority of their alveolar air sac retained normal morphology, and the numbers of inflammatory cells present were markedly reduced compared to those for infected mice not receiving RP-002 S1P1R therapy. The score for the PVM-infected mice receiving RP-002 therapy was 2.4 ± 0.1, with a P value of <0.0001 compared to PVM-infected mice receiving only vehicle. Quantitative biochemical analysis paralleled these histologic findings. For RP-002-treated PVM-infected mice, the lesser pulmonary injury and diminished infiltration of inflammatory cells into lungs correlated with significantly less exudate as measured by total protein (Fig. 5B), LDH enzymatic activity (Fig. 5C), and IgM content in their BALF (Fig. 5D) when evaluated against results for vehicle-treated PVM-infected mice at day 8 after infection.

FIG 5.

RP-002 limited immunopathologic injury within PVM-infected mice. resulting in diminished morbidity and improved survival. C57Bl/6J mice were mock infected or infected i.n. with 3 × 10³ PFU PVM and administered vehicle or 6 mg/kg RP-002 orally 1, 25, and 49 h after infection. (A) A representative hematoxylin-and-eosin-stained lung section from mice in each experimental group is shown. Four lungs were analyzed from each uninfected group (two panels on the left) and 6 lungs from each PVM-infected group (two panels on the right). A whole lung lobe from each of the four experimental groups with the tissue magnification is shown. (B to D) Biochemical analysis of lung exudates on BALF from infected and treated mice. Lung exudate markers included the total protein concentration (B), LDH enzymatic activity (C), and IgM content (D). Data are representative of 3 independent experiments with 4 to 7 mice per group. *, P < 0.05. (E) The average percentage of preinfection weight (equivalent to 100%) within experimental groups. For vehicle-treated mice, n = 9; for RP-002-treated mice, n = 10. Mice were removed from the experiment at death or loss of >30% of their initial starting weight: for vehicle-treated mice, 1 mouse at day 9, 1 at day 10, 2 at day 12, 1 at day 14, and 1 at day 16; for RP-002-treated mice, 1 mouse at day 11, 2 at day 12, and 1 at day 14. *, P < 0.05; **, P < 0.005; ***, P < 0.005. (F) Survival within experimental groups. The data represent two combined experiments with a total of 15 mice per group. (G) Lungs were harvested at specified time points and assayed for PVM titer by plaque assay. The limit of detection was 1.7 log₁₀ PFU per g of tissue. Circles represent individual mice, and lines signify the average. P = 0.09, comparing results with vehicle to those with RP-002, on day 5 postinfection, and P = 0.85, comparing results with vehicle to those with RP-002 on day 8 postinfection; n = 5 mice per group per time point.

Vehicle-treated mice infected with PVM exhibited rapid weight loss beginning on day 6 postinfection (Fig. 5E). In contrast, PVM-infected mice receiving RP-002 displayed significantly less weight loss on days 7 to 16 after infection (Fig. 5E). While body weight of mice that received vehicle was static on days 11 to 16 postinfection, RP-002-treated recipients replenished their lost weight to levels equal to or greater than their initial weight over that same time interval. In correlation with their decreased weight loss and diminished cytokine storm and immunopathologic injury, PVM-infected RP-002-treated recipients had significantly longer life spans and higher survival rates than vehicle-treated mice (Fig. 5F). Importantly, although the immune response was dampened, administration of RP-002 did not significantly alter titers of PVM in the lungs on days 5 and 8 after infection (Fig. 5G).

DISCUSSION

The cytokine storm induced by pulmonary viral infections enhances respiratory distress, leading to respiratory failure and frequently death of the host (8). Identifying the underlying factors that elicit such injurious reactions and devising therapeutics that block such factors would provide a viable therapy against virus-induced severe respiratory diseases. During most viral infections, type I IFNs are produced early, and these cytokines stimulate proinflammatory gene expression as well as the secretion of cytokines; therefore, type I IFN, by initiating an early innate immune response, can instigate a cytokine storm, as documented during influenza virus infection (15). In contrast, here we describe a mechanism for the induction of a cytokine storm where the early innate immune response (days 1 to 3) does not apparently play a role, but the secretion of TNF-α and IFN-γ produced by virus-specific CD8 T cells of the adaptive immune response later in infection (day 5) are essential for the cytokine storm. This late onset of the cytokine storm was revealed in the PVM mouse model for hRSV, suggesting the need to reevaluate RSV infection in humans. In the PVM model, either antibody blockade/depletion of TNF-α or IFN-γ and CD8⁺ T cells or oral administration of an immunomodulatory S1P1 receptor-specific agonist, RP-002, blunted the initiation of the cytokine storm and offered protection from this acute respiratory infection.

Neither overt cytokine production nor immune cell infiltration was detected in the BALF and lungs, respectively, until day 5 post-PVM infection. Our findings agree with recent evidence that cytokine mRNA was negligible until day 5 after PVM infection of C57BL/6 mice (25). The timing of the cytokine response suggested two possibilities. First, that earlier than day 5 after infection, PVM was not replicating or, second, PVM suppressed the innate inflammatory response. We found that PVM was detected in the lung at day 3 postinfection at levels greater than the amount of the virus inoculum (Fig. 1D), indicating that the first possibility was unlikely. The second possibility is more probable, since the PVM nonstructural (NS) proteins, NS1 and NS2, have potent anti-IFN activity (29), and innate inflammatory responses may be inhibited by PVM until day 5 after infection within C57BL/6 mice. Further, deletion of NS2 from the PVM genome blunted replication as well as production of specific cytokines, including TNF-α and IFN-γ, which are required for the PVM-induced cytokine storm. The NS2 deletion induced transient, early upregulation of IFN-α, IFN-β, and CXCL10, which likely contributed to host control of PVM replication as well as enhanced survival, probably due to prevention of a fatal cytokine storm (30). Therefore, NS2 suppressed innate antiviral responses, which induced a cytokine repertoire facilitating PVM replication, resulting in enhanced virulence of the virus. It has been reported (31) that IFN-α/β receptor-deficient mice infected with PVM exhibited changes in cytokine profiles compared to their wild-type counterparts but were impaired in their ability to control viral replication and avoid pathological effects. As we report here, IFN-α secretion increased in the BALF beginning on day 5 after PVM infection (Fig. 1A), and antibody neutralization of TNF-α and/or IFN-γ inhibited this IFN-α production (Fig. 1B). Thus, TNF-α and IFN-γ secretion preceded and was required for IFN-α release. Antibody neutralization studies indicated that TNF-α and IFN-γ signaling was the essential mechanistic trigger for the PVM-induced cytokine storm.

The early cellular source of TNF-α and IFN-γ was primarily CD8⁺ T cells and to a lesser extent NK cells (Fig. 2H), since significantly greater reductions in inflammatory markers were noted following CD8⁺ T cell depletion than were seen with NK cell depletion. While the innate immune response is suppressed, recruitment of PVM-specific CD8⁺ T cells that interact with specific PVM H-2-restricted cytotoxic T lymphocyte (CTL) epitopes (28) led to host recognition of PVM-infected cells and initiation of a cytokine storm. TNF-α- and IFN-γ-producing CD44⁺ CD8⁺ T cells appeared within the PVM-infected lung on day 5 after infection (Fig. 2E), indicating that activated antigen-presenting cells (APCs) displaying PVM antigen had migrated to draining lymph nodes early following infection. While it is unclear how APCs became activated and presented PVM antigen without eliciting a detectable early inflammatory response within the lung, accumulation of activated CD44⁺ CD8⁺ T cells in the lung coincided with dramatic upregulation of cytokine secretion and recruitment, as well as activation of inflammatory cells. NK cells also acted as an adjuvant to the PVM-induced cytokine storm, presumably by contributing TNF-α and IFN-γ or by enhancing CD8⁺ T cell activity. This finding is in agreement with a recent finding whereby NK cell-derived IFN-γ contributed to disease severity during hRSV infection of mice (32). Importantly, we have mechanistically decoded a paramyxovirus-induced cytokine storm that required the recruitment of TNF-α- and IFN-γ-producing CD8⁺ T cells, which is responsible for instigating the inflammatory response. These results with the hRSV model stand in sharp contrast to the pathogenesis of the cytokine storm induced by influenza virus infection, where type I IFN production by the innate immune system was essential for the initiation of the cytokine storm (15).

Cytokine secretion and the accumulation of innate (Fig. 1 to 3) (22) and adaptive immune cells (Fig. 1 to 3) within the lung following PVM infection can induce pulmonary injury. CD8⁺ T cells are essential for PVM clearance but achieve this by lysing infected cells, thereby causing pulmonary injury (24). Thus, both the innate and adaptive immune responses contribute to paramyxovirus pathogenesis, although initiation is dependent on the virus-specific CD8⁺ T cell response.

Previously, we reported that oral administration of the S1P1R agonist RP-002 blunted the innate inflammatory response, including TNF-α and IFN-γ secretion during influenza infection (15). Now we show that oral RP-002 treatment significantly protects against the acute respiratory disease caused by PVM by inhibiting recruitment and activation of T cells of the adaptive immune system followed by secretion of multiple cytokines except IFN-α. The lack of inhibition of IFN-α indicates that this cytokine is not necessary for initiation of the PVM-induced inflammatory response. Thus, S1P1R agonist treatment inhibited CD8⁺ T cell accumulation in the lung, which significantly lowered IFN-γ and TNF-α content, thereby blunting the cytokine storm. The detection of fewer PVM-specific CD8⁺ T cells in the MLN of RP-002-treated mice on days 5 (Fig. 3D) and 8 (Fig. 4F) after infection is indicative that CD8⁺ T cell expansion was hampered. It is likely that multiple factors contribute to S1P1R-mediated disruption of CD8⁺ T cell responses during PVM infection. First, Dorsam and Graler (33, 34) reported that retroviral or transgene overexpression of the S1P1R within T cells inhibited their proliferation in response to antigenic challenge, mixed leukocyte reaction, or incubation with monoclonal antibodies against anti-CD3 and anti-CD28. Hence, S1P1R signaling on PVM-specific CD8⁺ T cells can inhibit proliferation. Second, S1P1R signaling on B cells can alter follicular shuttling (35); therefore, administration of RP-002 may have disrupted the S1P-S1P1R-mediated CD8⁺ T cell localization within MLN, thereby hampering proliferation. Third, administration of the promiscuous S1PR agonist FTY720 enhances T regulatory (T_reg) cell activity (36, 37), although the biological consequences of these cells is currently not clear. We noted for our study that RP-002 administration resulted in a significant increase in the frequency of T_reg cells in the MLN on day 5 postinfection (data not shown). Fourth, earlier we demonstrated that administration of a promiscuous S1PR agonist, AAL-R, which signals S1P receptors 1 and 3 to 5, blunted the accumulation of influenza virus-specific CD8⁺ T cells in the lung. The mechanism was by S1P receptor-mediated inhibition of DC activation (16, 38), which likely acted through S1P3R or -4R but was independent of S1P1R signaling (21). Therefore, direct S1P1R signaling on T cells, alterations in T cell localization, and/or perhaps augmented T_reg cell frequencies in the MLN may contribute to reductions in the numbers of PVM-specific CD8⁺ T cells.

The decreased expansion of PVM-specific CD8⁺ T cells in the lung also occurred in the MLN of RP-002-treated mice. The result was a reduced inflammatory profile characterized by diminished antiviral molecule production on both days 5 (Fig. 3E and F) and 8 (Fig. 4G and H) after PVM infection. Further, following egress from lymph nodes and infiltration into the lung on day 8 after infection, PVM-specific CD8⁺ T cells from RP-002-treated mice continued to produce fewer effector inflammatory molecules (Fig. 4C and D). We conclude that S1P1R signaling caused the generation of both fewer and less aggressive CD8⁺ T cells. This led to diminished pulmonary injury by inhibiting CD8⁺ T cell activity, thus limiting the accumulation of exudates within the lung and destruction of alveolar pulmonary cells. The mechanism of RP-002 inhibition of the cytokine storm recorded in the PVM model resulted from the blunting of the later adaptive immune response, thus differing from RP-002's inhibition of the earlier innate immune response observed with influenza infection (15). Nevertheless, both acute respiratory viral infections were successfully treated by blunting the cytokine storm through use of a specific S1P1R agonist. Currently, we are evaluating other acute respiratory distress syndromes, such as hanta virus and SARS infections, whose pathogenesis is also dependent on a cytokine storm, for their chemical tractability by S1P agonist therapy as well as the cell-cell signaling pathways involved. Identification and utilization of an S1P1R agonist and other selective anti-inflammatory therapies that offset damaging immune responses at various stages and sites, regardless of the infectious agent, offer an attractive option for future pharmacologic intervention of acute respiratory diseases and other diseases caused by excessive inflammation and immunopathology.