Harnessing the versatility of iNKT cells in a step-wise approach to sepsis immunotherapy

Abstract

Sepsis results from a heavy-handed response to infection that may culminate in organ failure and death. Many patients who survive acute sepsis become immunosuppressed and succumb to opportunistic infections. Therefore, to be successful, sepsis immunotherapies must target both the initial and the protracted phase of the syndrome to relieve early immunopathology and late immunosuppression, respectively. Invariant natural killer T (iNKT) cells are attractive therapeutic targets in sepsis. However, repeated treatments with α-galactosylceramide (α-GalCer), the prototypic glycolipid ligand of iNKT cells, result in anergy. We designed a novel ‘double-hit’ treatment that allows iNKT cells to escape anergy and exert beneficial effects in biphasic sepsis. We tested the efficacy of this approach in the sublethal cecal ligation and puncture (CLP) mouse model, which mirrors polymicrobial sepsis with progression to an immunosuppressed state. Septic mice were treated with OCH {[(2S, 3S, 4R)-1-O-(α-D-galactopyranosyl)-N-tetracosanoyl-2-amino-1,3,4-nonanetriol]}, a T_H2-polarizing iNKT cell agonist, before they received α-GalCer. This regimen reduced the morbidity and mortality of CLP, induced a transient but robust IFN-γ burst within a pro-inflammatory cytokine/chemokine landscape, transactivated NK cells, increased MHC class II expression on macrophages, and restored delayed-type hypersensitivity to a model hapten, consistent with recovery of immunocompetence in protracted sepsis. Structurally distinct T_H2-polarizing agonists varied in their ability to replace OCH as the initial hit, with their lipid chain length being a determinant of efficacy. The proposed approach effectively exploits iNKT cells’ versatility in biphasic sepsis and may have translational potentials in the development of new therapies.

Introduction

Sepsis is a catastrophic syndrome triggered by a dysregulated host response to infection leading to hyperinflammation and an early cytokine storm, which could lead to vital organ failure. Although sepsis takes many lives in its early phase, improved critical care has resulted in a shift in the mortality pattern of sepsis. Accordingly, most deaths from sepsis now occur due to immunosuppression in the protracted phase of the syndrome, in which patients succumb to secondary or opportunistic infections (1–3). In addition, sepsis may elevate the long-term risk of certain malignancies (4). However, many studies to date have focused on early sepsis, and numerous clinical trials targeting APCs, conventional T cells, or their products (e.g., inflammatory cytokines) have failed (5, 6). There is increasing appreciation that optimal immunotherapy of sepsis requires a two-pronged approach, namely to prevent exaggerated inflammation early on while boosting immunity in the protracted phase (2, 3, 6).

Invariant natural killer T (iNKT) cells are innate-like T lymphocytes with tremendous immunomodulatory properties and therapeutic potentials in a variety of disease models and settings. They secrete a wide array of inflammatory cytokines copiously and rapidly after they detect microbe-derived or synthetic glycolipid Ags, typified by α-galactosylceramide (α-GalCer), presented by CD1d (7–10). In doing so, iNKT cells activate or regulate the functions of APCs (11, 12) and multiple downstream effector cell types belonging to both the innate and adaptive arms of immunity (13–17). Importantly, iNKT cells are highly versatile entities and can be polarized to produce predominantly T_H1- or T_H2-type cytokines (18), a characteristic that has been explored in preclinical studies and exploited in clinical trials for cancer and infectious diseases.

Several studies have suggested a pathogenic role for iNKT cells in murine sepsis (6, 19–21). We reported increased frequencies of iNKT cells among circulating T cells of septic patients and also demonstrated the benefit of skewing iNKT cell responses towards a T_H2-like phenotype in the feces-induced peritonitis (FIP) mouse model of acute intraabdominal sepsis (21). Whether iNKT cell polarization can be achieved in a septic phase-tailored fashion has not been addressed before. This is perhaps because iNKT cells undergo long-lasting anergy following exposure to α-GalCer, which renders them unresponsive to subsequent treatments with this glycolipid (22). This represents a major impediment to the success of iNKT cell-based immunotherapies for various conditions, including but not limited to sepsis. To address this limitation, we devised a two-step treatment regimen that prevented iNKT cell anergy through sequential administration of different glycolipid agonists. This approach was efficacious in a clinically relevant mouse model of biphasic sepsis in which hyperinflammation is followed by immunosuppression.

Materials and Methods

Mice

C57BL/6 (B6) and BALB/c mice were purchased from Charles River Canada Inc. (St. Constant, Quebec). We also maintained a breeding colony of B6 mice in our institutional barrier facility. β2M^−/− mice on a B6 background were provided by Dr. Anthony Jevnikar (Western University). Age-matched cohorts of adult male mice were used in this study. Our animal experiments were performed in compliance with the Canadian Council on Animal Care guidelines and following animal use protocols that were reviewed and approved by Animal Care and Veterinary Services at Western University.

Glycolipids

KRN7000 (α-GalCer) from Funakoshi Co. Ltd was prepared in a vehicle containing 0.5% Tween 20, 56 mg/mL sucrose and 7.5 mg/mL histidine, heated at 80°C, aliquoted and stored at −80°C. Shortly before use, α-GalCer aliquots were thawed, reheated at 80°C for 10 minutes, and diluted in sterile PBS for injection. Lyophilized OCH {[(2S, 3S, 4R)-1-O-(α-D-galactopyranosyl)-N-tetracosanoyl-2-amino-1,3,4-nonanetriol]} was supplied by the NIH Tetramer Core Facility (Atlanta, GA), reconstituted with sterile water and stored at 4°C until use. Dry, solvent-free C20:2 was solubilized in a PBS solution containing 1% DMSO and 0.5% Tween 20 and stored at −20°C. Aliquots were thawed, sonicated at 37°C for 5 minutes, heated at 80°C for 1 minute, and then diluted in PBS for injections. Lyophilized PBS-25 (Paul B Savage-25) and PBS-128 were formulated for direct dissolution in water and subsequent injections.

Polymicrobial sepsis models

In the vast majority of experiments, we employed a sublethal version of cecal ligation and puncture (CLP) with slight modifications (23). Ten-to-twelve-week-old male B6 mice were placed in a plastic chamber containing 2% vaporized isofluorane. Once stable, mice were transported into a makeshift surgical station where anesthesia was maintained by 1% isofluorane applied through a nose cone. The abdominal skin was disinfected using sterile gauze pads presoaked in a 2% chlorhexidine solution, wiped with 70% ethanol, and subjected to additional disinfection with 0.5% chlorhexidine. Following a midline laparotomy, the distal end of the cecum was externalized and ligated at a 0.5-cm distance from the apex. Cecum was then perforated twice using a 27-gauge needle. The peritoneal cavity and abdominal skin were closed with sutures, and 1 mL of normal saline was administered s.c. behind the ear. Mice were left to recover in separate cages under a heat lamp for 30 minutes. For sham mice, the procedure was identical except that the cecum was neither ligated nor punctured. All animals were injected s.c. with 0.5 mg/kg of buprenorphine twice, once 20 minutes before and again 24 hours after the surgery.

To induce severe CLP in separate cohorts of mice, one-third of the cecum was ligated followed by 3 perforations made with a 25-gauge needle. In a limited number of experiments, we also employed the FIP model of polymicrobial sepsis. Briefly, a slurry containing 200 mg/mL of fecal material was prepared in PBS after pooling feces from 20 age- and sex-matched donors residing in the same barrier environment. To induce FIP, mice were injected i.p. with 50 or 100 μL of the above slurry.

Animals were closely monitored, and their morbidity was scored in a blinded manner using a murine sepsis scoring (MSS) system we previously described (24). Four, 18, 24 and 48 hours after CLP, each mouse was assigned a score of 0 to 4 for each of several criteria, including coat and eye appearance, respiration rate, consciousness level, motility, and response to various stimuli. A weight loss of ≥20% and/or unresponsiveness to physical provocation were considered experimental endpoints.

In vivo treatments

Four μg of each glycolipid were administered i.p. in a final volume of 200 μL to naïve or septic mice as indicated. Primary and secondary glycolipid injections were separated by 4 days. Control animals received an equal volume of a corresponding vehicle solution.

Phase-tailored glycolipid immunotherapy of septic mice involved two injections. Mice received 4 μg of either OCH or PBS-25 i.p. 4 hours after CLP, which was followed by a 4-μg i.p. injection of α-GalCer on day 4 post-CLP. Control mice were given the appropriate vehicles following the above schedule, and sham controls were left untreated. Additional control cohorts received α-GalCer followed by α-GalCer, α-GalCer followed by OCH, or OCH followed by OCH, as indicated.

Serum cytokine measurements

Saphenous blood was collected from non-septic mice 2, 8 and 24 hours after a sole glycolipid injection or following a secondary challenge as indicated. Septic animals that had received successive injections of OCH and α-GalCer (or corresponding vehicles) were bled 12 hours after the secondary α-GalCer (or vehicle) treatment. Sera were isolated via centrifugation at 17 × 1000 g for 15 minutes at 4°C, and then stored at −80°C. Serum cytokine levels were quantified using ELISA kits from ThermoFisher Scientific (Waltham, MA) or by Luminex xMAP technology employed by Eve Technologies (Calgary, Alberta).

Customized gene expression analysis

Five days after CLP, glycolipid- and vehicle-treated survivors were sacrificed for their liver. Hepatic CD3^-NK1.1⁺ cells were stained and sorted, after dead cell and doublet exclusion, by a BD FACSAria III cytometer achieving a purity of >99%. RNA was extracted using a Purelink RNA Mini Kit (ThermoFisher) and converted to cDNA using the Invitrogen SuperScript VILO MasterMix. Quantitative PCR was performed using Custom TaqMan Array 96-Well Fast Plates (ThermoFisher) in a StepOnePlus Real-Time PCR System. Changes in gene expression were assessed by the ΔΔCT method.

Cytofluorimetric analyses

Mice were sacrificed for their spleen and/or liver, which were mechanically homogenized in sterile PBS. Liver tissue homogenates were subsequently placed in a 33.75% Percoll PLUS solution (GE Healthcare, Chicago, IL) and spun at 700 × g, without brake, for 12 minutes at room temperature to remove parenchymal cells. Brief treatment with ACK lysis buffer and filtration through 70-μm pores of a cell strainer rid the resulting splenic and non-parenchymal hepatic mononuclear cell (HMNC) preparations of erythrocytes and debris, respectively. To block Fcγ receptors before staining, cells were briefly incubated on ice with 20 μL of culture supernatant from a hybridoma producing an anti-CD16/CD32 mAb (clone 2.4G2). Staining then followed using fluorophore-conjugated mAbs to surface CD3ε (clone 145–2C11), CD107a (1D4B), F4/80 (BM8), I-A/E^b (AF6–120.1), NK1.1 (PK136), intracellular granzyme (GZM) A (GzA-3G8.5) and/or intracellular IFN-γ (XMG1.2) as indicated. Splenic and hepatic iNKT cells were identified through co-staining with an anti-TCRβ mAb (H57–597) and PBS-57-loaded mouse CD1d tetramers. Labeled mAbs and isotype controls were all from ThermoFisher, and CD1d tetramer reagents were provided by the NIH Tetramer Core Facility.

Cell surface staining was carried out at 4°C for 30 minutes. For intracellular staining, an eBioscience Intracellular Fixation & Permeabilization Buffer Set was utilized. To detect cytolytic molecules present in hepatic NK cells, bulk HMNCs were co-cultured at a 30:1 ratio with YAC-1 thymoma cells in the presence of 10 μg/mL brefeldin A (BFA) (Sigma). After 4 hours at 37°C and 6% CO₂, cells were washed, co-stained with anti-CD3 and anti-NK1.1 mAbs, fixed and permeabilized, and stained for indicated effector molecules. For CD107a staining, anti-CD107a mAb was present at 1 μg/mL in 4-hour cultures containing HMNCs, YAC-1 cells, 2 μM monensin (BioLegend, San Diego, CA) and BFA. Stained cells were washed and interrogated using a FACSCanto II flow cytometer, and data were analyzed using FlowJo software version 10.

⁵¹Cr release assays

YAC-1 cells were grown in RPMI-1640 medium supplemented with 10% heat-inactivated FBS, 0.1 mM MEM nonessential amino acids, GlutaMAX-I, 1 mM sodium pyruvate, 10 mM HEPES, 100 U/mL penicillin and 100 μg/mL streptomycin, which we simply refer to as medium. To prepare target cells for cytotoxicity assays, YAC-1 cells were labeled with 100 μCi of Na₂⁵¹CrO₄, with occasional shaking, inside an incubator set at 37°C and 6% CO₂. After 90 minutes, labeled cells were washed, resuspended in medium, and seeded at 10,000 cells/well in a U-bottom microtiter plate.

HMNCs from glycolipid- and vehicle-treated mice were prepared and employed as a source of effector cells against ⁵¹Cr-labeled YAC-1 cells at indicated ratios. After 4 hours at 37°C and 6% CO₂, microplates were spun, and a 100-μL co-culture supernatant sample was harvested from each well for reading in a γ-counter. Experimental release (ER) values were obtained from wells in which effector and ⁵¹Cr-labeled target cells were co-present. Spontaneous release (SR) and total release (TR) were determined in supernatant samples collected from wells in which target cells were suspended in medium or in 1% Triton X-100, respectively. Specific cytotoxicity against YAC-1 target cells was calculated using the following equation: % specific lysis = [(ER - SR) ÷ (TR - SR)] × 100.

In vivo killing assays

The in vivo lytic function of α-GalCer-transactivated NK cells was assayed using a method we described elsewhere (16). Erythrocyte-depleted, naïve target splenocytes from WT and β2M^−/− B6 mice were labeled with two different concentrations of CFSE, typically 0.2 μM and 2 μM respectively. Target cells were extensively washed, mixed in equal numbers, and co-injected at 1 × 10⁷ total cells in 200 μL PBS into the tail vein of glycolipid- and vehicle-treated CLP survivors. Three hours later, the recipients were euthanized for their spleen and liver in which CFSE-labeled target cells were traced by flow cytometry. Percent specific lysis of β2M^−/− target cells was calculated using the following formula: % specific killing = {1 - [(CFSE^high event number in organ ÷ CFSE^low event number in organ) ÷ (CFSE^high event number within mixed target cells before injection ÷ CFSE^low event number within mixed target cells before injection)]}×100.

Evaluation of delayed-type hypersensitivity (DTH)

On day 4 post-CLP, indicated cohorts of sepsis survivors were injected s.c., behind the ear, with 100 μL of a 10-mM 2,4,6-trinitrobenzene sulfonic acid (TNBS) solution (Sigma-Aldrich, St. Louis, MO). Four days later, mice were challenged via a 50-μL injection of the same solution in the left footpad and also received 50 μL of PBS in the right (control) foot. After 24 hours, footpad swelling was measured using a digital caliper after deducting the baseline thickness recorded prior to the TNBS challenge. In experiments in which septic mice received sequential glycolipid treatments, the sensitizing dose of TNBS was given on day 4 post-CLP twenty minutes after the second glycolipid injection.

Statistical analyses

Graphpad Prism 6 software was employed to compare various datasets. We used unpaired Student t-tests and ANOVA as appropriate, and conducted log rank tests for survival analyses. Differences with p values equal to or less than 0.05 were considered statistically significant.

Results

OCH-primed iNKT cells retain in vivo responsiveness to α-GalCer

Sepsis is a complex syndrome consisting of hyper- and anti-inflammatory phases. We previously proposed the possibility of using iNKT cell glycolipid agonists therapeutically for sepsis (6). iNKT cell ligands that either suppress or promote inflammatory responses do exist. However, administration of even a single dose of α-GalCer, the prototypic example of such ligands, abrogates or attenuates iNKT cell responses to a recall challenge with the same molecule (22). Whether other glycolipid combinations follow the same pattern has remained largely unclear. We sought to determine if in vivo priming with OCH, a truncated T_H2-polarizing analog of α-GalCer (25), alters subsequent iNKT cell responses to α-GalCer. The reason we chose this sequential treatment (OCH → α-GalCer) was two-fold. First, OCH treatment alone is protective in the context of FIP, a well-established model of acute sepsis (21). Second, treatment with α-GalCer relieves immunosuppression in certain other models or conditions (26, 27). Therefore, we posited that if iNKT cells retain their responsiveness to α-GalCer following an initial treatment with OCH, the OCH → α-GalCer regimen should alleviate both sepsis-induced hyperinflammation and immunosuppression.

We first tested the above hypothesis in the absence of sepsis. Naïve mice were treated with α-GalCer, OCH or vehicle followed, 4 days later, by a second injection of α-GalCer and quantitation of serum IL-4 and IFN-γ levels (Fig. 1A). As anticipated, pretreating B6 mice with vehicle did not prevent their subsequent IL-4 and IFN-γ responses to α-GalCer, which reached their peak levels at 2 hours and between 8–24 hours, respectively (Fig. 1B–C). In vivo exposure to α-GalCer before a second treatment with this glycolipid (α-GalCer → α-GalCer) dramatically reduced serum IL-4 levels (Fig. 1B) and abolished IFN-γ production (Fig. 1C). This was consistent with previous reports in which iNKT cell anergy was judged by a hyporesponsive state after an ex vivo challenge with α-GalCer (22). In contrast, priming with OCH did not alter the magnitude of the recall IL-4 response to α-GalCer (Fig. 1B). In addition, the OCH → α-GalCer treatment resulted in a sharp rise in serum IFN-γ levels, which were even higher than the levels detected in the vehicle → α-GalCer cohort at the 2-hour time point (Fig. 1C). Serum IFN-γ levels in the OCH → α-GalCer cohort were maximal at 8 hours and returned to the baseline at 24 hours. It needs to be noted that vehicles used to dissolve the above glycolipids are known not to generate cytokine responses on their own. Therefore, we did not include a vehicle → vehicle group in these experiments.

Fig. 1. OCH-primed iNKT cells remain responsive to α-GalCer.

Indicated strains of mice were injected i.p. with 4 μg of OCH or α-GalCer, or with vehicle. Four days later, mice were challenged with 4 μg of α-GalCer before they were either sacrificed for their spleen and liver or bled at indicated time points (A). Serum IL-4 (B,E) and IFN-γ (C,F) levels were quantified by ELISA. Using separate cohorts of B6 mice, the intracellular IFN-γ content of iNKT cells was determined by flow cytometry two hours after the α-GalCer challenge. Representative plots for hepatic iNKT cells and summary data for both hepatic and splenic iNKT cells are shown (D). Error bars represent SEM. Statistical comparisons were carried out using two-way ANOVA with a Tukey’s post-hoc test. *, ** and *** denote differences between mice treated with OCH → α-GalCer and those receiving α-GalCer → α-GalCer, with p<0.05, p<0.01 and p<0.001, respectively. † and ††† indicate differences between OCH → α-GalCer- and vehicle → α-GalCer-treated cohorts with p<0.05 and p<0.001, respectively. N.S. indicates a non-significant difference.

In separate experiments, unlike in α-GalCer → α-GalCer-treated mice, hepatic and splenic iNKT cells from animals receiving OCH → α-GalCer contained large quantities of IFN-γ two hours after the α-GalCer challenge (Fig. 1D). Therefore, the latter regimen allows iNKT cells to maintain their vigorous IFN-γ production capacity.

Next, we extended our findings to another standard mouse strain, namely BALB/c. These animals are reportedly more vulnerable than B6 mice to CLP-induced lethality (28). Similar to B6 mice, BALB/c mice that had received two doses of α-GalCer had diminished IL-4 and IFN-γ concentrations in their circulation (Fig. 1E–F), indicative of iNKT cell anergy. However, they were able to launch an augmented response to α-GalCer if they had been primed with OCH first.

Treatment with OCH → α-GalCer reduces the severity and lethality of sepsis

Since OCH did not anergize iNKT cells in our two-step stimulation regimen, we tested its efficacy as the initial component of a ‘double-hit’ immunotherapeutic protocol for biphasic sepsis. This required an in vivo model with low early mortality. Although very informative, the FIP model simulates only the acute phase of the syndrome (6). In a pilot experiment, i.p. injection of B6 mice (n=3) with 100 μL of a fecal slurry resulted in 100% mortality within 24 hours. Therefore, we resorted to the CLP model, the gold standard of sepsis models in rodents.

CLP involves laparotomy and ligation of the cecum, which is then perforated to allow fecal content to leak into the peritoneal cavity (29). A mild or sublethal form of CLP can be conducive to survival in a fraction of septic animals and their progression to an immunosuppressed state (6, 23). In our hands, ligating one-third of the cecum, which was then punctured thrice with a 25-guage needle, led to severe sepsis with a mortality rate of 80% by day 2 post-surgery. By comparison, ligating a smaller portion of the cecum followed by two perforations inflicted by a 27-gauge needle resulted in relatively mild sepsis, typically with a ~40–60% survival rate. Therefore, we proceeded to test the efficacy of OCH → α-GalCer treatment in the latter CLP model.

Following a protocol that is schematically illustrated in Fig. 2A, B6 mice were subjected to mild CLP before they received OCH (or vehicle), monitored for signs of morbidity using a scoring system we previously developed (24), re-injected with α-GalCer (or vehicle), and continued to be observed. As expected, unlike sham-operated controls, animals that underwent CLP showed signs of overt morbidity, which intensified over time before a maximum MSS was reached 48 hours after the surgery (Fig. 2B). We found OCH to significantly minimize the severity of sepsis in this timeframe (Fig. 2B). Furthermore, OCH → α-GalCer reproducibly reduced the mortality of septic mice (Fig. 2C). The observed survival advantage was evident early after OCH administration (Fig. 2C), consistent with low MSS scores recorded for animals that had been primed with OCH but not yet challenged with α-GalCer (Fig. 2B). Of note, in the absence of the initial anti-inflammatory stimulation with OCH, α-GalCer treatment did not alter the lethality of CLP (Fig. 2D).

Fig. 2. Treatment with OCH → α-GalCer reduces CLP-induced morbidity and mortality.

B6 mice were subjected to sublethal CLP four hours before they received 4 μg of OCH or vehicle i.p. according to a regimen schematically illustrated in A. Animals were monitored, and their 48-hour morbidity was recorded using a murine sepsis score (MSS) described in Materials and Methods. Error bars represent SEM, and * denotes p<0.05 when comparing OCH- and vehicle-treated septic mice by unpaired Student’s t-tests (B). Survivors were treated with 4 μg of α-GalCer or vehicle i.p. (A,C), and continued to be monitored. Kaplan-Meier survival curves were generated with a weight loss of ≥ 20% and/or unresponsiveness to provocation defining the endpoints. * indicates p<0.05 based on a log-rank test (C). Separate cohorts of B6 mice that had survived CLP surgery were injected with α-GalCer or vehicle in the absence of OCH pretreatment and were monitored for survival (D).

Taken together, the above results demonstrate that OCH → α-GalCer treatment reduces both the morbidity and the mortality of sublethal CLP, which is afforded by the early OCH hit and not reversed by subsequent α-GalCer treatment.

Sequential treatment with OCH and α-GalCer yields a systemic pro-inflammatory cytokine picture

Given the clear benefit of OCH → α-GalCer in our CLP model, it was pertinent to investigate the impact of this regimen on blood cytokine and chemokine levels in the face of an ongoing septic challenge. To this end, we used an extensive panel of cytokines/chemokines to quantify pro- and anti-inflammatory mediators in the serum of CLP survivors 12 hours after they were given the second hit with α-GalCer. This time point was chosen because of our interest in relieving sepsis-induced immunosuppression and also to closely mimic the timeline used in our iNKT cell anergy experiments (Fig. 1). Similar to non-septic mice (Fig. 1C), CLP survivors that had undergone treatment with OCH → α-GalCer had significantly more IFN-γ in their blood (Fig. 3). Moreover, they had elevated levels of TNF-α, IL-2, IL-5, eotaxin, CCL2, CXCL9 and CXCL10. In contrast, no statistically significant differences were found between vehicle → vehicle and OCH → α-GalCer cohorts in terms of the circulating levels of several T_H2-polarizing and/or anti-inflammatory mediators, such as IL-4, IL-10, IL-13 and TGF-β (Fig. 3). These findings are consistent with an overall pro-inflammatory signature induced by the OCH → α-GalCer treatment. In addition, they reinforce our conclusion that iNKT cells from OCH-primed animals maintain their ability to respond to α-GalCer even in a septic milieu.

Fig. 3. Treating septic mice with OCH → α-GalCer gives rise to a pro-inflammatory blood cytokine pattern.

B6 mice underwent sublethal CLP surgery before they were treated with either OCH → α-GalCer or vehicle → vehicle. Saphenous blood was collected 12 hours after the second hit, and serum levels of indicated cytokines and chemokines were determined using a multiplexing platform. Error bars represent SEM. Statistical analyses were performed by unpaired Student’s t-tests, and * and ** denote statistically significant differences with p<0.05 and p<0.01, respectively. N.S. indicates a non-significant difference.

Treatment with OCH → α-GalCer permits NK cell transactivation

Glycolipid-stimulated iNKT cells are known to transactivate a multitude of downstream effector cell types, including NK cells that play critical roles in antipathogen immunity. This is largely owed to iNKT cells’ capacity to secrete IFN-γ amply and swiftly (30). We found OCH → α-GalCer to induce IFN-γ production by iNKT cells (Fig. 1D) as well as a robust but transient rise in serum IFN-γ levels (Fig. 1C and Fig. 1F). Furthermore, iNKT cells harbored by septic mice were not anergized by this treatment and remained capable of triggering IFN-γ production (Fig. 3).

To begin to explore the functional significance of the above phenomenon, we examined gene expression by NK cells in CLP survivors (Fig. 4A–B). Transcriptomic analyses of hepatic NK cells from glycolipid-treated animals revealed elevated mRNA levels of Gzmb, Gzma and Perforin, suggesting enhanced cytolytic potentials (Fig. 4B). Consistent with this signature, HMNCs prepared from OCH → α-GalCer-treated survivors were much more potent than those harvested from vehicle-treated or sham-operated mice in destroying YAC-1 cells (Fig. 4C), the prototypic mouse NK cell targets. At the protein level, intracellular GZM A was highly abundant in NK cells, albeit at comparable levels between glycolipid- and vehicle-treated mice (93.8 ± 1.9% and 90.8 ± 1.9%, respectively; n=3/group). We found increased cell surface expression of CD107a (LAMP-1) in NK cells from OCH → α-GalCer-treated animals, suggesting that more efficient degranulation was responsible, at least partially, for their augmented cytolytic effector function (Fig. 4D).

Fig. 4. Sequential treatments of septic mice with OCH and α-GalCer augments the expression of cytotoxic effector molecules, degranulation and lytic function by NK cells.

B6 mice were subjected to sublethal CLP before they were injected i.p. with OCH (or vehicle) and α-GalCer (or vehicle) (A). Twenty-four hours after the second hit, hepatic CD3^-NK1.1⁺ NK cells were stained, FACS-purified and pooled for RNA extraction and RT-PCR. Fold increases/decreases in the expression of indicated genes, relative to the vehicle → vehicle condition, were used to generate a heat map (B). Bulk non-parenchymal hepatic mononuclear cells (HMNCs) were co-incubated for 4 hours with ⁵¹Cr-labeled YAC-1 target cells. The ⁵¹Cr activity of culture supernatant samples was then determined by a γ counter, and the specific lysis of target cells was calculated using a formula that is detailed in Materials and Methods (C). In parallel, bulk HMNCs were co-cultured with YAC-1 cells in the presence of monensin and BFA. After 4 hours, the surface expression of CD107a on NK cells was analyzed by flow cytometry (D). In separate experiments, 24 hours after the second hit, wild-type and β2M^−/− splenocytes, which were respectively labeled with a low and a high dose of CFSE, were mixed at a 1:1 ratio and injected via tail vein into glycolipid-treated and control CLP survivors. Three hours later, target cells in the spleens (E) and in the livers (F) were tracked and distinguished based on their differential CFSE labeling intensities, and their relative abundance was used to calculate % specific killing of β2M^−/− target cells as described in Materials and Methods. Error bars represent SEM. Statistical analyses were performed using two-way ANOVA with a Tukey’s post-hoc test (C) or unpaired Student’s t-tests (D-F). * and ** denote differences with p<0.05 and p<0.01, respectively.

To validate our results in an in vivo setting, we took advantage of a CFSE-based killing assay, which we recently optimized for transactivated NK cell-mediated cytotoxicity (16). In this assay, naïve splenocytes from β2 microglobulin (β2M)-deficient B6 mice that are devoid of cell surface MHC class I molecules serve as NK targets. These targets were more efficiently removed from both the spleen and the liver of OCH → α-GalCer-treated mice (Fig. 4E–F), further supporting the conclusion that this therapeutic regimen results in greater NK cell-mediated cytotoxicity.

The α-GalCer hit in OCH → α-GalCer-treated mice alleviates CLP-induced immunosuppression

One of the primary goals of our step-wise treatment approach was to avert late immunosuppression. As predicted, CLP survivors in our model were immunosuppressed. This was judged by the meager swelling of their footpad that had been injected with a recall dose of TNBS following s.c. priming with this hapten (Fig. S1A–B). This was indicative of a suboptimal delayed-type hypersensitivity (DTH) response. We also found the splenic F4/80⁺ macrophages of CLP survivors to express low levels of MHC class II molecules (Fig. S1C). The above readouts were used due to their clinical relevance. Many patients who survive sepsis’s early hyperinflammatory phase do not exhibit normal DTH skin reactions to standard Ags (31). In addition, HLA-DR expression on CD14⁺ monocytes is routinely assessed in the clinic not only to identify septic patients in an immunosuppressed state but also to monitor the efficacy of the immunotherapies they receive (32).

To evaluate the efficacy of OCH → α-GalCer in reversing immunosuppression, mice were given OCH (or vehicle) shortly after they underwent CLP surgery, followed 4 days later by TNBS sensitization and α-GalCer (or vehicle) injection. Four days later, they were challenged via an intrafootpad (i.f.p.) injection of TNBS and subsequently assessed for footpad swelling and MHC II expression (Fig. 5A). As with mice treated with OCH → α-GalCer without TNBS priming/challenge (Fig. 2C), a clear survival advantage was manifest in this cohort (Fig. 5B). Importantly, OCH → α-GalCer could significantly increase footpad swelling as an indication of a partially restored DTH response (Fig. 5C). We also detected increased expression of I-A/I-E on a per cell basis, as judged by the geometric mean fluorescence intensity (gMFI) of staining for these MHC class II molecules, on splenic macrophages (Fig. 5D–E).

Fig. 5. Administering α-GalCer to OCH-primed CLP survivors restores DTH reaction to TNBS and MHC II levels on macrophages.

B6 mice underwent sublethal CLP surgery before they were treated i.p. with indicated glycolipids (or vehicles), sensitized s.c. with TNBS and then challenged i.f.p. with TNBS following the timeline depicted in A. Kaplan- Meier survival curves were generated with experimental endpoints defined as a weight loss of ≥20% and/or unresponsiveness to physical provocation (B). Twenty-four hours after the TNBS recall, the thickness of the injected footpad was measured for each mouse using a caliper, from which the baseline pre-challenge thickness was deducted. Error bars represent SEM (C). On day 9 post-CLP, survivors were sacrificed, and splenic F4/80⁺ macrophages were assessed by flow cytometry for their expression level of I-A/E^b. Representative histograms (E) and summary geometric mean fluorescence intensity (gMFI) data (F) are shown. One-way ANOVA with Tukey’s post-hoc tests was employed to determine statistically significant differences with p<0.05, p<0.01 and p<0.001, which are denoted by *, ** and ***, respectively. Sham controls underwent a surgical procedure in which their cecum was neither ligated nor perforated. They also did not receive any treatments with glycolipids (or vehicles).

Of note, in a separate experiment in which α-GalCer was administered in the absence of a prior OCH treatment, neither the DTH response to TNBS nor the expression of I-A/I-E could be recovered (Fig. S2A–B).

In additional control experiments, we tested several combinations in parallel with OCH → α-GalCer treatment, including α-GalCer → α-GalCer, α-GalCer → OCH, and OCH → OCH. In the absence of sepsis, none of the above treatments resulted in appreciably high blood cytokine levels (Fig. S3A). In cohorts that were subjected to CLP, whenever OCH was administered as the initial hit, sepsis-induced mortality was ameliorated (Fig. S3B). In contrast, using α-GalCer in this capacity was futile (Fig. S3B). Importantly, with the sole exception of OCH → α-GalCer treatment, every combination that we tested failed to rescue the animals’ DTH response (Fig. S3C).

Collectively, the above results demonstrate the dual benefit of OCH → α-GalCer treatment in CLP-induced sepsis. While the initial OCH hit serves to prevent death, the subsequent α-GalCer challenge boosts at least some aspects of innate and adaptive immune responses.

PBS-25, but not C20:2, is efficacious as the initial hit in the treatment of CLP-induced sepsis

A number of glycolipids with activities similar to OCH have been previously reported. We set out to ascertain whether the beneficial bioactivity of OCH as the initial hit was unique to this molecule or could be mimicked by other T_H2-skewing iNKT cell agonists. One such agonist is C20:2 (33), which we previously used as a solitary treatment for FIP-induced acute sepsis (21). As depicted in Fig. S4A and consistent with our previous findings (21), administering a single dose of C20:2 to B6 mice resulted in an early IL-4 burst and a subsequent IFN-γ response. However, priming with C20:2 resulted in a dwarfed recall response to α-GalCer (Fig. S4B), suggesting that iNKT cells had been anergized. Unlike OCH, C20:2 has a relatively long acyl chain. Therefore, we tested two other glycolipids, namely PBS-25 and PBS-128 (34, 35), for their capacity to replace OCH since they are more similar to OCH in terms of their lipid chain length.

In our head-to-head comparisons, a single i.p. injection of PBS-25 yielded a higher IL-4:IFN-γ ratio than did PBS-128 (Fig. 6A). Moreover, priming naïve B6 mice with either PBS-25 or PBS-128 did not induce iNKT cell anergy since IL-4 and IFN-γ responses to second stimulation with α-GalCer were intact in both conditions (Fig. 6B). However, compared to PBS-128, an initial PBS-25 treatment potentiated less IL-4 and more IFN-γ production in response to α-GalCer (Fig. 6B). Therefore, we pursued the usage of PBS-25 in our sepsis model. We found treatment with PBS-25 → α-GalCer to significantly reduce the mortality of CLP (Fig. 6C), which was reminiscent of the OCH → α-GalCer treatment (Fig. 2C and Fig. 5B). These results suggest that the structure of iNKT cell agonists contributes, at least partially, to their anergy-inducing property, or lack thereof.

Fig. 6. In vivo priming with PBS-25 does not compromise subsequent iNKT cell responses to α-GalCer, and can serve as the initial hit in the treatment of biphasic sepsis.

Four μg of α-GalCer, PBS-128 or PBS-25, glycolipid agonists of iNKT cells with varying acyl chain lengths, were injected i.p. into naïve B6 mice, which were bled 2, 8 and 24 hours later for serum cytokine analyses. Peak levels of IL-4 and IFN-γ were used to calculate the IL-4:IFN-γ ratio for each agonist (A). Separate cohorts of naïve B6 mice were given 4 μg of α-GalCer, PBS-128, PBS-25 or vehicle i.p. Four days later, all animals were challenged with a 4-μg i.p. injection of α-GalCer as the second hit, and serum IL-4 and IFN-γ concentrations were quantitated by ELISA at indicated time points. Error bars represent SEM. Statistical comparisons were performed using unpaired Student’s t-tests. * and *** denote differences between cohorts receiving PBS-128 → α-GalCer and α-GalCer → α-GalCer, with p<0.05 and p<0.001, respectively. †† and ††† indicate differences between PBS-25/PBS-128 → α-GalCer- and vehicle → α-GalCer-treated mice with p<0.01 and p<0.001, respectively. Significant differences between PBS-25 → α-GalCer- and α-GalCer → α-GalCer-treated mice with p<0.01 and p<0.001 are denoted by ## and ###, respectively (B). Kaplan-Meier survival curves were generated using data from B6 mice that were subjected to sublethal CLP before they were treated with PBS-25 → α-GalCer or with vehicle → vehicle. A log-rank test was employed to perform statistical analysis, and * indicates p<0.05 (C).

Discussion

Accumulating evidence suggests that iNKT cells play important roles in acute polymicrobial sepsis. Hu et al. demonstrated a substantial drop in hepatic iNKT cell frequencies of B6 and BALB/c mice after severe CLP (20). This was accompanied by elevated CD25 and CD69 levels on the surface of residually detectable iNKT cells indicating their activation. Moreover, iNKT cell deficiency in Jα18^−/− mice was protective against CLP. Heffernan and co-workers subsequently reported that following CLP, iNKT cells migrate out of the liver to accumulate in the peritoneal cavity, the site of polymicrobial infection, where they influence the phagocytic activity of macrophages (36). In addition, these investigators demonstrated a lower bacterial burden in the peritoneal cavity of septic Jα18^−/− mice compared with their wild-type (WT) counterparts. In a prospective clinical study, we found the peripheral blood iNKT:T cell ratio to be higher in patients with sepsis than in non-septic trauma patients (21). We also reported that the severity of FIP-induced sepsis was low in Jα18^−/− mice but worsened when these animals were reconstituted with WT iNKT cells. Finally, we found a single injection of OCH within 20 minutes of the fecal challenge to reduce the MSS scores and to prolong the survival of septic mice.

The above studies have painted a generally pathogenic picture of iNKT cells during acute sepsis but have fallen short of addressing the role(s) and/or the therapeutic potentials of iNKT cells in sepsis-induced immunosuppression, a problem that is common in the clinic and potentially fatal. The consequent opportunistic infections that arise can be stubbornly difficult to resolve even with broad-spectrum antibiotics and infection source control (1). A retrospective review of macroscopic postmortem findings indicated that about 77% of surgical ICU patients who had died from sepsis had continuous septic foci (37), suggesting a failure to clear infection with the inciting pathogen, nosocomial microbes, or both. Immunosuppression in patients with protracted sepsis is evidenced by weak DTH skin reactions (31) and frequent reactivation of one or more latent viruses, such as cytomegalovirus, Epstein-Barr virus, herpes simplex virus and human herpesvirus-6 (38).

To date, iNKT cells have not been studied in biphasic sepsis models. In the current investigation, we optimized and used a mild, sublethal form of CLP with hyperinflammatory and immunosuppressive phases that mimic real-life scenarios (6, 23). There has been a general consensus that sequential targeting of iNKT cells induces anergy, which can be obviously counterintuitive in therapeutic settings. Here, we have provided strong evidence to the contrary when glycolipid combinations are carefully chosen. We designed and employed a step-wise treatment regimen in which OCH (or PBS-25) and α-GalCer were administered to septic mice sequentially and in a phase-tailored fashion. This novel approach reduced the mortality of CLP and remedied the problem of iNKT cell anergy. As a result, OCH-exposed iNKT cells retained responsiveness to α-GalCer, which was administered in the protracted phase of sepsis to mitigate immunosuppression.

Since OCH was synthesized and found to polarize iNKT cells towards a T_H2-type phenotype (25), it has been used to ward off T_H1-mediated immunopathology in a wide range of conditions. For instance, we employed OCH to reduce morbidity and/or pathology encountered in cardiac allotransplantation (39) and in the HLA-DR4-transgenic mouse models of rheumatoid arthritis (40) and toxic shock syndrome (41). We also demonstrated the therapeutic benefit of OCH in the FIP model of acute sepsis (21).

In the vast majority of our experiments, we used OCH as the ‘initial hit’ enabling iNKT cells to counter sepsis-induced hyperinflammation without undergoing anergy. Importantly, the impact of OCH in this capacity could be readily phenocopied through treatment with PBS-25, but not with C20:2. Although a potent T_H2-polarizing compound, C20:2 is a diene with a relatively long and hydrophobic acyl chain and as such structurally dissimilar to OCH (33). By contrast, PBS-25 is more similar to OCH in terms of solubility in aqueous environments, lipid chain length and affinity for CD1d (34, 35). Therefore, we propose that the structure of iNKT cell agonists, especially the length of their lipid chains, is a determinant of their ability or inability to escape anergy. This should in turn inform whether an agonist can be used in multi-dose treatment modalities for a variety of conditions, including sepsis.

In our system, the second hit with α-GalCer rescued innate and adaptive responses represented by NK cell-mediated cytotoxicity and DTH, respectively. These responses largely depend on IFN-γ, a prominent pro-inflammatory cytokine that rose sharply but transiently following the second hit. It was therefore not surprising that the IFN-γ-inducible chemokines CXCL9 and CXCL10 were also elevated. These chemokines bind and signal through CXCR3 to regulate NK cell and T cell trafficking to the site of infection during sepsis (42). We also detected increased levels of the potent pleiotropic cytokines IL-2 and TNF-α along with CCL2, IL-5 and eotaxin. CCL2 is a monocyte chemoattractant with important roles in bacterial clearance in septic animals (43). IL-5 and eotaxin, another IFN-γ-inducible chemokine (44), are key to eosinophil growth and recruitment. Of note, higher IL-5 levels and eosinophil counts may be associated with survival in clinical sepsis (45, 46) although the reportedly protective property of IL-5 may not be always linked to eosinophils (47). The interplay between IL-5, eotaxin and eosinophils in the contexts of protracted sepsis and sepsis-induced immunosuppression will need to be further clarified.

Serum levels of several T_H2-type and anti-inflammatory cytokines, namely IL-4, IL-10, IL-13 and TGF-β, remained statistically unaltered after sequential treatments with OCH and α-GalCer. Therefore, this protocol changed the overall cytokine landscape of sepsis in favor of a pro-inflammatory profile, which should serve the host well during an immunosuppressive phase.

IFN-γ is a major culprit of early sepsis immunopathology (6). However, its controlled release during protracted sepsis likely benefits the host by restoring immunocompetence. In our model, systemic administration of α-GalCer in the absence of OCH pretreatment was neither deleterious nor capable of reversing immunosuppression. Intriguingly, however, priming with OCH before α-GalCer treatment culminated in upregulated MHC II on macrophages, augmented NK cell-mediated cytotoxicity and restored DTH reactions, all of which depend, at least partially, on the activity of IFN-γ. We speculate that the OCH → α-GalCer treatment achieves this feat by inducing a transient IFN-γ spike, which should be adequate to reverse immunosuppression. By the same token, sustained IFN-γ, which is avoided in our regimen, would unleash excessive inflammatory responses leading to tissue injury and/or immunological exhaustion. This needs to be taken into serious consideration when designing other iNKT cell-based therapies as the second hit. α-C-GalCer, a C-glycoside analog of α-GalCer with strong IFN-γ and IL-12 production capacities, has been described (48, 49). In addition, α-GalCer-pulsed dendritic cells (DCs) may be tested as the second hit in lieu of the free-floating glycolipid (50), especially if more than one treatment will be necessary. This approach would not only circumvent iNKT cell anergy but may also optimize DC functions in orchestrating adaptive responses (11, 51) to counter sepsis-induced immunosuppression (52). Regardless of the nature of the second hit, a desirable agent or modality will need to be one that will be inflammatory enough to boost antimicrobial immunity during protracted sepsis but not too inflammatory causing tissue damage and organ failure.

The next important question will be whether α-GalCer can be combined with other treatments to maximize the host’s ability to fight secondary infections during protracted sepsis. For instance, PD-1-based immune checkpoint inhibitors may be beneficial. Interfering with PD-1 signaling is known to prevent α-GalCer-induced anergy (53). This may be particularly important when the second hit may need to be delivered repeatedly. Secondly, a link has been established between PD-1 and the immunological shortcomings of septic patients, including their T cell exhaustion. Boomer et al. found high frequencies of PD-1⁺ CD4⁺ T cells and PD-1 ligand 1 (PD-L1)⁺ APCs among the splenocytes of septic patients harvested rapidly after their death (54). Moreover, within the postmortem lungs of septic patients, the expression of PD-1 by CD4⁺ cells and that of PD-L1 by plasmacytoid pre-dendritic cells were enhanced in comparison with control tissues from lung cancer resections. Brahmamdam et al. reported that PD-1 blockade after CLP rescues the DTH response of septic mice (55). In a separate study, blocking PD-L1 either before or after CLP reduced the mortality of sepsis (56). Also importantly, this approach lowered the bacterial burden in the peripheral blood and within the peritoneal cavity of septic mice. We are currently investigating how PD-1-based immune checkpoint inhibitors perform as part of our second hit (i.e., OCH → α-GalCer plus anti-PD-1/PD-L1) in alleviating sepsis-induced immunosuppression.

Another attractive pathway that can be targeted is coupled to the IL-7 receptor system, which promotes cellular viability and growth. Exposing septic patients’ T cells to rIL-7 ex vivo enhances their proliferation, IFN-γ production, STAT5 phosphorylation, and Bcl-2 upregulation in response to TCR triggering (57). Recently, in a prospective, randomized, placebo-controlled trial, the safety of rIL-7 (CYT107) administration to septic patients and its ability to reverse CD4⁺ and CD8⁺ lymphopenia were verified (58). We noted a significant decrease in serum IL-7 levels of septic mice that had been sequentially treated with OCH and α-GalCer (Fig. 3). Therefore, adding IL-7 to our second hit may present a potential therapeutic opportunity by preventing immune cell apoptosis and improving T cell functions to bolster innate and adaptive antipathogen immunity.

In summary, in this investigation, we have taken advantage of the tremendous versatility of iNKT cells to design a novel, phase-tailored protocol for the treatment of sepsis-induced hyperinflammation and immunosuppression. iNKT cells constitute attractive therapeutic targets in sepsis for several reasons. First, the prognosis of sepsis is partially determined by the speed with which its treatment gets underway. iNKT cells are among the few T cell types that launch swift responses to antigenic stimulation and may, as such, make a difference quickly (59). Second, the availability of multiple T_H2- and T_H1-glycolipid agonists of iNKT cells should allow for testing additional and carefully timed protocols similar to what we have described herein. Third, iNKT cell stimulation results in secondary activation of downstream effector cell types. As such, targeting iNKT cells will have wide-ranging impacts on ensuing host responses. We assessed the cytolytic effector function of transactivated NK cells in this work. However, there are numerous other effectors that can be influenced, ideally to the septic host’s benefit. Fourth, glycolipid agonists of iNKT cells work beyond the species barrier (60). In fact, some of the same glycolipids employed in mouse studies have shown promise in clinical trials for malignancies and viral diseases (18). Therefore, we anticipate the findings of this study to be translatable. Fifth, iNKT cells are restricted by CD1d, which is monomorphic (61). Therefore, glycolipid ligands of iNKT cells, such as those used in this work, should be useful in genetically diverse human populations. This is a tempting possibility that warrants further investigation.

Key Points.

• In vivo priming with OCH does not curb subsequent iNKT cell responses to α-GalCer.

• Sequential treatment with OCH and α-GalCer reduces CLP-induced mortality.

• α-GalCer treatment of OCH-primed septic mice restores their immunocompetence.

Non-standard Abbreviations used in this article:

α-GalCer

α-galactosylceramide

β2M

β2 microglobulin

BFA

brefeldin A

CLP

cecal ligation and puncture

DC(s)

dendritic cell(s)

ER

experimental release

FIP

feces-induced peritonitis

gMFI

geometric mean fluorescence intensity

GZM

granzyme

HMNC(s)

[non-parenchymal] hepatic mononuclear cell(s)

iNKT

invariant natural killer T (cell)

MSS

murine sepsis score

OCH

[(2S, 3S, 4R)-1-O-(α-D-galactopyranosyl)-N-tetracosanoyl-2-amino-1,3,4-nonanetriol]

PBS-25

Paul B Savage-25

PBS-128

Paul B Savage-128

PD-1

programmed cell death-1 (PD-1)

PD-L1

PD-1 ligand 1

SR

spontaneous release

TNBS

[2,4,6-]trinitrobenzene sulfonic acid

TR

total release