IRF3 contributes to sepsis pathogenesis in the mouse cecal ligation and puncture model

IRF3 is identified as a key mediator of inflammatory markers, bacteremia, and disease pathogenesis, in the cecal ligation and puncture model of sepsis.

Abstract

Much remains to be learned regarding which components of the innate immune response are protective versus detrimental during sepsis. Prior reports demonstrated that TLR9 and MyD88 play key roles in the CLP mouse model of sepsis; however, the role of additional PRRs and their signaling intermediates remains to be explored. In a prior report, we demonstrated that the signal adaptor IRF3 contributes to the systemic inflammatory response to liposome:DNA. We hypothesized that IRF3 might likewise promote sepsis in the CLP model. Here, we present results demonstrating that IRF3-KO mice have reduced disease score, mortality, hypothermia, and bacterial load following CLP versus WT counterparts. This is paired with reduced levels of systemic inflammatory mediators in IRF3-KO mice that undergo CLP. We demonstrate that peritoneal cells from WT CLP mice produce more cytokines than IRF3-KO counterparts on a per-cell basis; however, there are more cells in the peritoneum of IRF3-KO CLP mice. Finally, we show that IRF3 is activated in macrophages cultured with live or sonicated commensal bacteria. These results demonstrate that IRF3 plays a detrimental role in this mouse model of sepsis.

Introduction

Sepsis is the leading cause of death in noncardiac intensive care units and is likely to pose an increasing burden on our healthcare system in the coming years [1, 2]. There are a growing number of individuals >65 years of age in our society who are at greater risk of developing sepsis. This is compounded by an increased number of complex surgical procedures being performed, which can lead to infections that may develop into sepsis. Hence, sepsis is likely to affect a large number of individuals and pose an increasing burden on our healthcare system in the coming years.

Sepsis occurs when an uncontrolled infection induces SIRS. This may progress to septic shock, characterized by a dangerous degree of hypotension, or severe sepsis, culminating in multiple organ dysfunction [3]. Despite carefully coordinated treatments, the mortality rate in sepsis remains high [2]. Novel therapies that modulate the initial inflammatory response could be beneficial to septic patients. However, we still do not fully understand how the various innate immune pathways contribute to sepsis.

An important component of innate immunity is the rapid response to conserved microbial motifs. TLRs are germline-encoded PRRs that recognize conserved microbial motifs [4]. Of relevance to this study, TLR9 resides in endosomes and recognizes CpG motifs in DNA that is taken up via endocytosis, a common entry route for pathogens [5]. TLR activation initiates signal transduction cascades via MyD88 and TRIF [6, 7]. This signaling culminates in the nuclear translocation of NF-κB and IRFs [4] that induce inflammatory mediators, including cytokines and type I IFNs. In addition to the TLRs, PRRs in the cytosol recognize nucleic acids [8–10]. Cytosolic dsRNA activates the retinoic acid inducible gene I-like receptors, initiating a signal transduction cascade via the mitochondrial antiviral signaling adaptor [11] and TBK1, leading to IRF3 and IRF7 activation [9, 12, 13]. Intracellular DNA is recognized in a similar manner [14–16], triggering the signal adaptors STING [17–19] and TBK1, leading to IRF3 activation.

Sepsis is likely to involve multiple pathogen sensors, which synergize to induce the complex pathophysiology of this disease. Mouse models of sepsis have been an important tool in this regard, as mice can be manipulated genetically and experimentally to investigate the importance of different immune pathways in sepsis pathogenesis [20]. In the CLP model, commensal bacteria are allowed to escape from the cecum into the peritoneum, leading to polymicrobial peritonitis that results in sepsis [21]. The related CASP model induces sepsis in a similar manner. As these models induce SIRS in the context of infection, they are arguably the best animal models of clinical sepsis developed to date.

Several reports have investigated how the TLR pathways contribute to sepsis in the CLP and CASP models. Although Gram-negative bacteria expressing LPS are released from the gut, reports indicate that TLR4 does not play a major role in these models [22, 23]. Conversely, TLR9 [24, 25] and MyD88 [19, 22, 26] are key mediators of sepsis in the CLP and CASP models. However, there is evidence than an additional, MyD88-independent pathway contributes to cytokine and chemokine production in these models [19]. Prior reports have generated conflicting data regarding whether the TLR3–TRIF pathway plays a significant role in CLP [26, 27].

We showed previously that IRF3 contributes to the systemic inflammatory response in mice injected with liposome:plasmid DNA, a vector that was developed for use in nonviral gene therapy [28]. In this report, we demonstrate a novel role for IRF3 in the CLP model of sepsis. Mice lacking IRF3 had improved survival, reduced inflammatory mediators, and reduced bacterial load following CLP versus WT congenic mice. We demonstrated that IRF3 is activated in macrophages treated with live and sonicated commensal bacteria in vitro. Hence, IRF3 may represent a novel biomarker or therapeutic target for treating sepsis.

MATERIALS AND METHODS

Mouse strains and the CLP model of sepsis

C57BL/6 mice were purchased from The Jackson Laboratory (Bar Harbor, ME, USA). IRF3-KO mice were a gift from Dr. Ruslan Medzhitov (Department of Immunobiology, Yale University, New Haven, CT, USA). These mice were housed under specific pathogen-free conditions and used at 6–12 weeks of age. The Yale University Institutional Animal Care and Use Committee approved the use of animals in this study. CLP was performed as described previously [29, 30]. Briefly, a midline incision was made in the peritoneum, and the cecum was exteriorized. Seventy-five percent of the cecum was ligated and pierced through and through with a 21G needle, and then a small drop of cecal contents was extruded. The cecum was returned to the peritoneal cavity, and the abdomen was closed in two layers. For sham surgeries, the cecum was exteriorized and then returned to the peritoneum without ligation or puncture. Mice received 1 ml warmed saline s.c. postoperatively and buprenorphine analgesia at 12-h intervals. The mice were monitored for survival and disease score (0=bright, alert, and responsive; 1=slightly lethargic; 2=lethargic and hunched; 3=very lethargic, hunched, and shaky; 4=dead). Surface temperature was determined with an infrared thermometer.

Serum, peritoneal lavage, spleen cell, and IHL preparation and ex vivo cytokine production

Blood was collected by retro-orbital bleed. Serum was prepared in serum separator tubes (Becton Dickinson, Franklin Lakes, NJ, USA), according to the manufacturer's instructions. For peritoneal lavage, mice were killed, and then 3 ml PBS was injected into the peritoneum. The abdomen was gently massaged for 2 min, and then fluid was recovered, and peritoneal cells were spun down. Spleen cells were prepared as described previously [31]. Livers were cut up with scissors; suspended in RPMI, 0.02% (w/v) collagenase IV (Sigma-Aldrich, St. Louis, MO, USA), 0.002% (w/v) DNase I (Sigma-Aldrich), and 1% penicillin/streptomycin; and then incubated at 37°C, 40 min with occasional agitation. The suspension was passed through a 40-μm cell strainer, layered onto an equal volume of Percoll (MP Biomedicals, Santa Ana, CA, USA), and then spun at 400 G, 30 min. IHLs were collected from the interface. To measure ex vivo cytokine production, 1 × 10⁶ cells were cultured in 200 μl RPMI, 10% FCS, and 1% penicillin/streptomycin for 18 h, and cytokine levels were measured in the supernatant by ELISA.

ELISA, serum chemistry, CFU measurements

ELISA kits for IL-6, IL-12/23p40, TNF-α, MCP-1 (eBioscience, San Diego, CA, USA), SAP, and MPO (ALPCO Diagnostics, Salem, NH, USA) were used per the manufacturers' instructions. ALT and AST were measured with Infinity liquid stable reagents (Thermo Electron, Thermo Fisher Scientific, Waltham, MA, USA), and CK was measured with the CK reagent set (Pointe Scientific, Canton, MI, USA). Peritoneal lavage samples were serially diluted with sterile water, plated on BBL agar plates, and incubated for 18 h at 37°C. Bacterial colonies were then counted.

Thioglycollate-elicited macrophages, preparation of commensal bacteria, and Western blot

Mice were injected i.p. with 1.5 ml 4% thioglycollate. Four days later, the animals were killed, and peritoneal lavage was performed as described above. Macrophages were enriched by plating the lavage cells on tissue-culture plates for 90 min at 37°C, 5% CO₂. The plate was washed several times to remove nonadherent cells. To prepare commensal bacteria, mouse cecal contents were inoculated into 10 ml Luria broth and cultured at 37°C on a shaking platform for 5 h. The suspension was centrifuged at 3000 RPM for 15 min, and then the bacterial pellet was washed, resuspended in 2 ml cell culture media, and passed through a 40-μm strainer to remove debris. The bacteria were added to macrophages directly or after 4× 30-s sonication bursts. Western blot was performed as described previously [28] using 1° antibodies against phosphorylated IRF3 serine-386 (Cell Signaling Technology, Danvers, MA, USA), total IRF3, and GAPDH (Santa Cruz Biotechnology, Santa Cruz, CA, USA). Western blot signal was detected with the chemiluminscent Amersham ECL Western blotting analysis system (GE Healthcare Life Sciences, Piscataway, NJ, USA), followed by X-ray film exposure. The film was scanned, and band density was quantified with Image J software.

Statistical analysis

We performed normality tests to determine whether parametric or nonparametric tests were appropriate. For parametric comparisons, we used a two-way t-test for pair-wise comparisons and the ANOVA test followed by a Bonferroni post-test for repeated measures. For nonparametric comparisons, we used Mann-Whitney for pairwise comparisons. Survival curves were compared by the log-rank test. Each experiment was repeated in triplicate. P < 0.05 was considered to be statistically significant.

RESULTS AND DISCUSSION

We first investigated whether IRF3 deficiency affected the sepsis pathophysiology in the CLP model by performing CLP and sham surgeries in IRF3-KO mice and WT C57BL/6 congenic controls (see Materials and Methods). WT mice that underwent CLP exhibited a rapid elevation in sepsis disease score (Fig. 1A), and 100% of the mice died within 51 h of the surgery (Fig. 1B). In contrast, IRF3-KO mice that underwent CLP had a lower sepsis disease score (Fig. 1A) and improved survival: 75% of the mice survived until the end of the experiment (>14 days; Fig. 1B). In the WT mice that underwent CLP, death was preceded by hypothermia; however, this was attenuated in the IRF3-KO mice (Fig. 1C). In WT mice, CLP induced substantial elevations in serum CK at 20 h (indicative of heart or skeletal muscle damage), whereas the level of CK was comparatively lower in IRF3-KO mice that underwent CLP (Fig. 1D). WT mice that underwent CLP exhibited a more subtle elevation in ALT and AST (indicative of liver damage) at 20 h, and this elevation was similar in IRF3-KO mice (not shown). Mice that underwent sham surgeries remained healthy and regained normal body temperature after the surgery, and CK, ALT, and AST levels remained at baseline.

Figure 1. IRF3-KO mice are protected from sepsis in the CLP model.

WT and IRF3-KO mice were subject to CLP (n=8/group) or sham surgeries (n=2/group), and the following measures of sepsis were determined: (A) disease score; we assigned a score based on the degree of lethargy (0=alert, 1=slightly lethargic, 2=lethargic, 3=very lethargic, 4=dead); (B) mouse survival was monitored; (C) mouse surface temperature was measured with an infrared thermometer to determine the degree of hypothermia; (D) mice were eye-bled at 20 h postsurgery, and serum CK levels were determined as a measure of heart/skeletal muscle damage. *P < 0.05; **P < 0.01 by log-rank (survival); two-way ANOVA (temperature, disease score); Mann-Whitney test (CK).

We examined markers of systemic inflammation and organ damage in these mice. IL-6 [32] and TNF-α [33] were previously shown to mediate CLP-induced sepsis, and an array of serum inflammatory mediators become elevated during the SIRS that occurs in this model. Following CLP, WT mice had substantial elevations in serum levels of inflammatory mediators, including cytokines IL-6, TNF-α, and IL-12/23p40, the chemokine MCP-1, and the acute-phase protein serum SAP (Fig. 2A–E). There was a significant reduction in the levels of serum IL-6, TNF-α, MCP-1, and SAP in the IRF3-KO CLP mice at 20 h and a trend toward a reduction in IL-12/23p40 (Fig. 2A–E). Mice that underwent sham surgeries did not exhibit elevations in these inflammatory mediators.

Figure 2. IRF3-KO mice have reduced levels of systemic inflammatory mediators following CLP.

WT and IRF3-KO mice were eye-bled after they underwent CLP (n=8/group) or sham surgeries (n=2/group), and the levels of serum inflammatory mediators were measured, including: cytokines (A) IL-6, (B) TNF-α, (C) IL-12/23p40, (D) chemokine MCP-1, and (E) acute-phase protein SAP. **P < 0.01, two-way ANOVA of the 0- to 21-h time-points, with Bonferroni post-tests.

We quantified bacterial load in the blood and peritoneal lavage 20 h after surgery. WT mice that underwent CLP had a high bacterial load in the blood and lavage; in comparison, IRF3-KO mice that underwent CLP had a lower bacterial load in both compartments (Fig. 3A and B). In mice that underwent sham surgeries, bacteria were not detected in the blood or lavage.

Figure 3. IRF3-KO mice have a lower bacterial load following CLP.

WT and IRF3-KO mice underwent CLP (n=6–8/group) or sham surgeries (n=2/group). (A) Blood and (B) peritoneal lavage were obtained 20 h later and spread on BBL agar plates to quantify bacterial CFUs as a measure of bacterial load in these compartments. *P < 0.05 by Mann-Whitney test.

A prior report indicated that the MyD88 pathway plays a greater role in the liver inflammatory response versus the spleen inflammatory response following CLP [19]. We reasoned that IRF3-mediated cytokine production might also be confined to a specific anatomical compartment. To define the anatomical compartments that produce IL-6, a key inflammatory mediator, we performed CLP on WT and IRF3-KO mice and harvested peritoneal lavage cells, spleen cells, and IHLs 20 h later. We cultured 1 × 10⁶ cells ex vivo and measured IL-6 production by ELISA. We found that peritoneal cells from mice that underwent CLP produced significantly more IL-6 than peritoneal cells from sham controls on a per-cell basis (Fig. 4A). A similar trend was observed for IHL; surprisingly, we did not observe IL-6 production from spleen cells (Fig. 4A). Next, we compared ex vivo cytokine production by peritoneal cells isolated from WT versus IRF3-KO mice after CLP or sham surgeries. We found that 1 × 10⁶ peritoneal cells harvested from WT mice 20 h post-CLP produced significantly more IL-6 and MCP-1 on a per-cell basis than peritoneal cells harvested from IRF3-KO mice post-CLP (Fig. 4B and C). We also observed greater cytokine production by WT IHL versus IRF3-KO IHL (data not shown). However, we found that at 20 h after CLP, the total number of cells in the peritoneum was higher in IRF3-KO mice versus WT mice (Fig. 4D). We directly measured the levels of inflammatory mediators in the peritoneal lavage of WT and IRF3-KO mice 20 h post-CLP surgery and saw no significant differences in IL-6, TNF-α, IL-12/23p40, and MCP-1 (Fig. 5A–D). Together, these data indicate that in WT mice, there were a smaller number of cells producing more cytokines in the peritoneum, whereas in IRF3-KO mice, there were a larger number of cells producing less cytokines, resulting in similar levels of peritoneal cytokines in the two groups at 20 h after CLP.

Figure 4. IRF3-KO mice exhibit less peritoneal cytokine production on a per-cell basis but display a greater number of peritoneal cells following CLP.

(A) Peritoneal cells, spleen cells, and IHL were harvested from WT mice 20 h after CLP (n=6–10/group) or sham surgeries (n=3–4/group). Cells (1×10⁶) were cultured ex vivo, and IL-6 production was measured. (B–D) Peritoneal cells were harvested from WT and IRF3-KO mice 20 h after CLP (n=9/group) or sham surgeries (n=2/group); ex vivo production of (B) IL-6 and (C) MCP-1 was measured as described above. (D) The total number of peritoneal cells was quantified in each group. *P < 0.05; **P < 0.01 by Mann-Whitney test.

Figure 5. IRF3-KO mice have similar levels of local inflammatory mediators following CLP.

WT and IRF3-KO mice underwent CLP (n=15–16/group) or sham surgeries (n=6/group), and the levels of inflammatory mediators were measured in the peritoneal lavage at 20 h, including: cytokines (A) IL-6, (B) TNF-α, (C) IL-12/23p40, and (D) chemokine MCP-1.

Finally, to directly confirm that commensal bacteria activate IRF3, we prepared thioglycollate-elicited macrophages from WT mice and treated them with live or sonicated commensal bacteria derived from the mouse cecum. Live and sonicated bacteria induced phosphorylation of IRF3 serine-396 (indicative of IRF3 activation) within 1 h of treatment, relative to the GAPDH loading control (Fig. 6A and B). In the samples treated with live bacteria, a further increase was observed following 2 h of treatment (Fig. 6A). In contrast, the total amount of IRF3 remained steady, relative to GAPDH, throughout the experiment (Fig. 6A and B).

Figure 6. IRF3 is activated in macrophages treated with commensal bacteria in vitro.

Thioglycollate-elicited macrophages were cultured for 0–2 h with (A) live or (B) sonicated commensal bacteria, derived from the mouse cecum (n=3/time-point, representative of repeated experiments). Serine-396 phosphorylated IRF3 (P-IRF3), total IRF3, and GAPDH were measured via Western blot. P-IRF3/GAPDH relative band density and total IRF3/GAPDH relative band density were quantified via densitometry. **P < 0.01 by one-way ANOVA, followed by Bonferroni post-tests.

In this report, we examined the role of IRF3 in the mouse CLP model of sepsis. We found that IRF3-KO mice were substantially protected from sepsis, with reduced mortality, disease score, hypothermia, and serum CK (Fig. 1A–D), implying that IRF3 plays a detrimental role in this mouse model of sepsis. We observed significantly lower levels of systemic inflammatory mediators in IRF3-KO mice versus WT mice following CLP (Fig. 2A–E), including two key mediators of sepsis: IL-6 [32] and TNF-α [33]. These data demonstrate that in the CLP model, IRF3 contributes to the overwhelming systemic inflammatory response that is a hallmark of sepsis.

We found that the bacterial load was significantly lower in the blood and peritoneum of IRF3-KO mice that underwent CLP versus WT counterparts (Fig. 3). This lower bacterial load is likely a driving factor behind the reduced systemic inflammatory response in IRF3-KO versus WT mice following CLP. However, we cannot exclude the possibility that in WT mice, an overwhelming inflammatory response impairs bacterial clearance (for example, by inducing apoptosis of lymphocytes that would kill bacteria), contributing to the high bacterial load, whereas in IRF3-KO mice, the ability to clear bacteria is preserved, thus lowering the bacterial load. Indeed, it is well known that the compensatory anti-inflammatory response makes an important contribution to the pathophysiology of sepsis (reviewed in ref. [34]).

We found that when cultured ex vivo, peritoneal cells from WT CLP mice produced more IL-6 than peritoneal cells from IRF3-KO CLP mice on a per-cell basis (Fig. 4B and C), and a similar pattern was noted in IHL (data not shown). In contrast, we found that local inflammatory mediators in the peritoneal lavage did not differ greatly between WT and IRF3-KO mice following CLP (Fig. 5A–D). These disparate findings are likely explained by the larger number of cells in the peritoneum of IRF3-KO mice versus WT mice at this time-point (Fig. 4D). The similar levels of peritoneal lavage cytokines in these two groups appear to be the consequence of high levels of cytokine production by fewer cells in the WT mice and low levels of cytokine production by more cells in the IRF3-KO mice. In future studies, it will be interesting to determine whether the increased number of cells in IRF3-KO mice is the consequence of greater infiltration and/or less cell death. Interestingly, a prior report indicates that IRF3 regulates apoptotic signaling [35].

We showed that commensal bacteria derived from the mouse cecum rapidly induced IRF3 phosphorylation in thioglycollate-elicited macrophages during in vitro culture (Fig. 6A and B); this was the case regardless of whether the bacteria were living or killed by sonication. In macrophages treated with live bacteria, there was an increase in IRF3 phosphorylation from 0 h to 1 h, followed by a further increase at 2 h (Fig. 6A); whereas in macrophages treated with sonicated bacteria, the level of IRF3 phosphorylation increased from 0 h to 1 h and then remained steady until 2 h (Fig. 6B). The further increase with live bacteria at 2 h may be a consequence of bacterial replication, leading to additional levels of stimulating ligands. These in vitro data suggest that PAMPs derived from commensal bacteria are sufficient to induce IRF3 activation in peritoneal inflammatory macrophages. Further work is required to elucidate the ligands that activate IRF3 in this context.

Prior reports examined how the TLRs contribute to sepsis pathogenesis in the CLP and CASP models. Notably, TLR9-KO mice have a reduced inflammatory response and improved survival following CLP [24, 25]. Likewise, TLR9 inhibitory oligonucleotides reduce the pathogenesis of sepsis following CLP [24, 25]. Weighardt et al. [19, 22] found a partial reduction in inflammatory mediators in MyD88-KO mice that underwent CASP, and these mice had improved survival versus WT mice. In contrast, Peck-Palmer et al. [26] found that MyD88-KO mice have substantial reduction in the inflammatory response but reduced survival in the CLP model. Importantly, Weighardt et al. [19] found that there is a second MyD88-independent pathway that contributes to cytokine and chemokine production in this context. As MyD88 is not known to activate IRF3, we propose that IRF3 is acting in a separate pathway. We demonstrated previously that TLR9 and IRF3 cooperate to induce SIRS in response to liposome:DNA [28], and we speculate that a similar mechanism is at play in the CLP model. However, the pathway that activates IRF3 in this context remains to be elucidated.

IRF3 can be activated downstream of the TRIF pathway [36] following TLR3 or TLR4 activation. A prior report demonstrated that IRF3-KO mice are protected from LPS-induced endotoxin shock, with reduced mortality and inflammatory mediators [37]. Furthermore, hydroxystilbenes, which inhibit LPS-mediated activation of IRF3, also reduced LPS-induced endotoxin shock mortality and liver inflammation [38]. Although Gram-negative bacteria expressing LPS are released in the CLP and CASP models, TLR4 does not play a major role in the sepsis phenotype in these models [22, 23]. Hence, although the effects of IRF3 that we observed in the CLP model mirror that seen in LPS-induced endotoxemia, this is likely to be the consequence of a different innate immune pathway.

Prior reports have yielded conflicting results regarding the role of the TLR3–TRIF pathway in CLP-induced sepsis. In one report, TLR3-KO mice were protected from CLP-induced mortality and displayed an altered profile of local inflammatory mediators at early time-points, i.e., reduced IFN-β, increased CCR5, CXCL10 at 3 h, and reduced levels of all inflammatory mediators by 24 h [27]. However, in a second report, TRIF-KO mice did not differ significantly from WT mice in inflammatory mediator production following CLP [26]. These apparently disparate results may stem from differences in the experimental model. Regardless, it is plausible that the TLR3-TRIF pathway contributes to IRF3 activation following CLP.

In addition to TLRs, alternative PRRs in the cytosol recognize nucleotides and induce innate immunity. The cytosolic DNA and RNA sensing pathways [9, 14, 15] and the recently identified cyclic diAMP and diGMP sensing pathways [39, 40] all activate IRF3. As DNA activates TLR9 following CLP, we propose that DNA may also stimulate the cytosolic DNA-sensing pathway, activating IRF3 via the STING adaptor. Further experiments are required to determine if this is the case.

In summary, it is likely that multiple pathogen sensors cooperate to induce the complex pathophysiology of sepsis. We have identified IRF3 as one key player in a mouse model of polymicrobial sepsis. Further work will be required to elucidate the ligand and host pathway that drive IRF3 activation and the mechanism whereby IRF3 influences sepsis pathogenesis. Finally, it will be critical to determine the relevance of this pathway in human patients with sepsis.

AUTHORSHIP

W.E.W. designed and performed the experiments, interpreted the data, and wrote the manuscript. A.T.B. performed experiments and assisted with manuscript preparation. D.R.G. helped design the experiments, interpreted the data, and assisted with manuscript preparation.

DISCLOSURES

The authors have no conflict of interest to declare.