Hydrocortisone (water-soluble) was from Sigma-Aldrich, cat# H0396. Other materials are listed in Major Resources Tables in Supplemental Materials .

SR-BI^fl/fl mice were made as described (32). SF1cre mice were from the Jackson Laboratory (#012462). The SF1cre mice were backcrossed with C57BL/6J and bred with SR-BI^fl/fl mice to yield SF1creSR-BI^fl/fl and SR-BI^fl/fl littermates (please refer to the Major Resources Tables for background information). The animals were bred at the University of Kentucky’s animal facility. Littermates at 10-14-week old were randomly allocated into the groups for all animal experiments. Sex and cage effect affect survival in CLP-induced sepsis. Cage-mates and sex-match were considered whenever possible. Blinding was performed for survival and biochemical analyses. No data were excluded. Animal care and experiments followed Reporting of In Vivo Experiments (ARRIVE) Guidelines, and were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Kentucky with approval # 2020-3511 (Role of SR-BI in inflammatory diseases) on June 6 of 2022. Mouse tail DNA was used for PCR genotyping. Animals were fed a standard laboratory diet and kept with a 10/14 h light/dark cycle.

CLP was performed following the protocol described previously (25). Briefly, mice were anesthetized by inhalation of 2–5% isoflurane in 100% oxygen. A midline incision (~1.0 cm) was made on the abdomen. The cecum was ligated and punctured twice with a 25-gauge needle, and gently compressed to extrude a small amount of cecal material. The cecum was returned to the abdomen, and the incision was closed with 5–0 Ethilon suture material. The mice were subsequently resuscitated with 1 ml phosphate-buffered saline via i.p. Survival was monitored for a 7-day period.

Plasma corticosterone was quantified with a kit from ENZO Life Sciences. Cytokines (IL-6 and TNF-α) were quantified with corresponding ELISA kits from eBioscience. Plasma cytokines were also analyzed by Eve Technologies using Mouse Cytokine Array/Chemokine Array 31-Plex (MD31). The serum alanine aminotransferase (ALT) levels were quantified with a kit from POINTE SCIENTIFIC, Inc to determine the degree of liver damage. The blood urea nitrogen (BUN) concentrations were measured with a kit from the QuantiChrom to determine the degree of kidney injury.

BODIPY ™ FL-conjugated Escherichia coli BioParticles (E. coli, K-12 strain, Life technologies) were suspended in sterile PBS to a final concentration of 1×10⁹ CFU/ml. The BODIPY-E. coli were intraperitoneally injected at 1×10⁸/mouse at 17 h post CLP. One hour later, the peritoneal fluid, blood and spleen were harvested for fluorescence-activated cell sorting (FCAS) analysis as we described previously (33).

Water-soluble hydrocortisone was dissolved in PBS and i.p. injected at 3mg/kg body weight following CLP. Survival was monitored for 7 days.

Bone marrow-derived macrophages (BMDM) were cultured as described previously (34). BMDM cells (5x10⁵ cell/well; 6 well plate) were cultured for 7 days and then stimulated with 2 ng/ml LPS or co-stimulated with 2 ng/ml LPS plus 0.4 ng/ml IFN-γ in the presence of 0, 31.25 (physiological level), 100 and 500 nM (stress levels) hydrocortisone for 18 h. The supernatants were harvested for cytokine analysis and cells were harvested for qRT-PCR.

The survival assay was analyzed by Log-Rank test. Significance in experiments comparing two groups was determined by 2-tailed Student’s t test. Significance in experiments comparing more than two groups was evaluated by One Way ANOVA, followed by post hoc analysis using Tukey’s test. Means were considered different at p< 0.05. The statistical analysis was done using GraphPad Prism 9 and consulted with Dr Bing Huang, a biostatistician.

SR-BI is highly expressed in adrenal gland and mediates the uptake of cholesterol ester into adrenal cortex cells, which is responsible for iGC production in stress conditions (28). We recently generated new SR-BI^fl/fl mice by flanking exon 2 of SR-BI (32) ( Figure 1A ). The SR-BI^fl/fl mice displaced normal SR-BI expression in the liver and adrenal gland compared to C56BL/6J mice ( Figure 1B ). We generated SF1creSR-BI^fl/fl mice by breeding SF1Cre mice with SR-BI^fl/fl mice. The SF1creSR-BI^fl/fl mice had complete knockout of adrenal SR-BI with normal SR-BI expression in the liver ( Figure 1C ). At the basal physiologic condition, plasma corticosterone concentration did not show significant difference between SR-BI^fl/fl mice and SF1creSR-BI^fl/fl mice. One-hour post ACTH stimulation, SR-BI^fl/fl mice had a delta of GC at ~ 200 ng/ml (> 90 ng/ml or 9 µg/dL), but SF1creSR-BI^fl/fl mice failed to generate iGC ( Figure 1D ). Thus, SF1creSR-BI^fl/fl mouse is a model of RAI.

Similar to data from ACTH stimulation, SR-BI^fl/fl mice showed a significant increase in iGC prodection at 4 and 20 h after CLP treatment. On the contrary, SF1creSR-BI^fl/fl mice failed to show an increase in iGC production ( Figure 1E ). SF1creSR-BI^fl/fl mice were susceptible to CLP-induced septic death as shown by 6.7% survival in SF1CreSR-BI^fl/fl mice versus 86.4% in SR-BI^fl/fl mice; p=0.0001 ( Figure 1F ), Blood urea nitrogen (BUN) concentration was about 2 times higher in SF1creSR-BI^fl/fl mice than SR-BI^fl/fl mice 20 h after CLP treatment, indicating kidney injury in SF1creSR-BI^fl/fl mice ( Figure 1G ), and there was no significantly difference in serum alanine aminotransferase level between SF1creSR-BI^fl/fl mice and SR-BI^fl/fl mice ( Figure 1H ).

Glucocorticoid has potent anti-inflammatory activity (35). To investigate why RAI mice are susceptible to sepsis, we measured 31 plasma cytokines and the phagocytic capability of phagocytes in CLP-treated mice. Four hours post CLP, cytokine levels were markedly increased in both SR-BI^fl/fl and SF1creSR-BI^fl/fl mice and there was no significant difference in cytokine levels between SR-BI^fl/fl and SF1creSR-BI^fl/fl mice, indicating similar inflammatory response in both SR-BI^fl/fl and SF1creSR-BI^fl/fl mice at the early stage of sepsis ( Figure 2 and Supplemental Table S1 ). Twenty hours post CLP, most of the cytokines were decreased in SR-BI^fl/fl mice, indicating well-controlled inflammatory response in SR-BI^fl/fl mice; however, for SF1creSR-BI^fl/fl mice, most of cytokines, including Eotaxin, GM-SCF, IL-1B, IL-2, IL-5, IL-6, IL-12(p40), IL-13, KC, MIP-1α, TNF-α and VEGF, were at high levels or even higher than that of 4 h post CLP, indicating uncontrolled inflammatory response in SF1creSR-BI^fl/fl mice ( Figure 2 and Supplemental Table S1 ). It is worth noting that the SF1creSR-BI^fl/fl mice had physiological levels of GC ( Figure 1D ). Thus, the data indicate that iGC is responsible for controlling inflammatory response and mice with RAI have systemic hyper-inflammatory responses due to a lack of iGC in sepsis.

We then utilized BODIPY ™ FL-conjugated Escherichia coli BioParticles to test the phagocytosis in sepsis. In peritoneal fluids, compared to SR-BI^fl/fl mice, SF1creSR-BI^fl/fl mice had a 60% lower percentage of MFI of E. coli in CD11b⁺ cells, a 73% lower percentage of MFI of E. coli in macrophages (CD45⁺CD11b⁺F4/80⁺), and a 56% lower percentage of MFI of E. coli in neutrophils (CD45⁺CD11b⁺Ly6g⁺) ( Figure 3 ). These data suggest that iGC is required for phagocytic activity of immune effector cells and mice with RAI have impaired phagocytic activity due a lack of iGC.

We used SF1creSR-BI^fl/fl mice as a RAI animal model to investigate the efficacy of GC treatment. Hydrocortisone (3 mg/kg body weight) treatment significantly decreased the survival rate of SR-BI^fl/fl mice ( Figure 4A ); on the contrary, hydrocortisone treatment significantly increased the survival of SF1creSR-BI^fl/fl mice ( Figure 4B ). SF1creSR-BI^fl/fl mice with GC treatment had a significantly decrease in plasma IL-6 concentrations compared to SF1creSR-BI^fl/fl mice without GC treatment at 4 and 20h post CLP; on the contrary, SR-BI^fl/fl mice with GC treatment displayed moderately higher plasma level of IL-6 compared to SR-BI^fl/fl mice without GC treatment ( Figure 4C ). SF1creSR-BI^fl/fl mice with GC treatment also had a 24% decrease in BUN concentrations at 20h post CLP ( Figure 4D ). Thus, these data show that GC supplementation benefits septic mice with RAI but harms septic mice without RAI. Our earlier study showed that GC treatment with 8 mg corticosterone/kg body weigh decreased survival from 80% to 40% in CLP-wild type mice. Our earlier study further showed that treatment with a stronger GC cocktail decreased survival to 8% in CLP-wild type mice (31). In this study, we used GC at 3 mg/kg. These data suggest that GC treatment to mice without RAI is harmful, the stronger GC treatment, the more harmful. Extra exogenous GC may cause immunosuppression in mice without RAI, given the potent immuno-suppressive activity of GC. Our earlier study in wild type mice showed that exogenous GC suppresses corticosterone production, decreases plasma IL-6 levels and increases blood and peritoneal bacterial load (31).

We utilized bone marrow-derived macrophages (BMDMs) to test our hypothesis that stress levels of GC (iGC) are required for effective suppression of inflammatory response. As shown in Figure 5A , hydrocortisone at physiological concentration (31.5 nM) significantly reduced IL-6 and TNF-α production in LPS-stimulated BMDMs. In sepsis, multiple inflammatory cytokines are generated and cross-talk between cytokines further augments inflammatory response. For example, IFN-γ, secreted by Th1 T lymphocytes, functions as a co-stimulator of LPS to stimulate macrophage to generate higher levels of cytokines (36). We tested our hypothesis in LPS and IFN-γ co-stimulated BMDMs. Compared to LPS stimulation alone, LPS/IFN-γ co-stimulation significantly enhanced IL-6 and TNF-α production ( Figure 5B ). Physiological concentrations of hydrocortisone hardly controlled IL-6 and TNF-α production and stress levels (100 or 500 nM) of hydrocortisone were required to significantly control IL-6 and TNF-α production ( Figure 5B ). These results show that stress levels of GC are required for the effective suppression of pro-inflammatory cytokine production in LPS/IFN-γ co-stimulated macrophages. Glucocorticoids induce several anti-inflammatory genes such as MAP kinase phosphatase-1 (MKP-1), which plays an important role in the feedback control of MAP kinase signaling (37, 38). We next investigated the effect of GC on the expression of MKP-1. As shown in Figure 5C , stress levels of hydrocortisone (500 nM) significantly increased the mRNA expression of MKP-1 in LPS or LPS/IFN-γ stimulated cells.

RAI is common in septic patients (8–10). The concept of RAI was introduced in 2004 Surviving Sepsis Campaign guidelines (39) and GC therapy was recommended for patients with RAI, which is supported by the French trail that shows a beneficial effect of corticosteroid treatment in septic shock patients with RAI (11). However, a later CORTICUS trial failed to validate these findings in a less severe group of septic patients with RAI (10). In 2008 Surviving Sepsis Campaign guidelines (40), the diagnosis of RAI was not recommended. In 2008, the term RAI was replaced by critical illness-related corticosteroid insufficiency (CIRCI) by a task force of experts in critical care medicine (41), “The task force coined the term CIRCI to describe the dysfunction of the hypothalamic-pituitary-adrenal axis that occurs during critical illness.” CIRCI includes absolute, relative adrenal insufficiency, and GC resistant. The task force pointed out some potential technical limitations associated with RAI, paradoxically, the same ACTH test is used for diagnosis of CIRCI. As pointed out by Loriaux et. al (42), “When the (ACTH) test is applied to stressed patients, for example, septic patients, the baseline cortisol value is an ACTH-stimulated number. The ACTH is endogenous, secreted in response to the stress. In this situation, the ‘Δ’ criterion is no longer applicable. This criterion, however, has been used to interpret the test in many of the published studies on relative adrenal insufficiency.” The same flaws apply to the diagnosis of CIRCI in critically ill patients. Nevertheless, the flaw associated with ACTH test is a technical issue and the same flaw applies to both RAI and CIRCI. It should not be a reason to abandon RAI. As outlined by the task force, “The mechanisms leading to dysfunction of the HPA axis during critical illness are complex and poorly understood and likely include decreased production of ACTH, and cortisol and the dysfunction of their receptors.” (41), a lack of ACTH will lead to impaired adrenal stress response (iGC), resulting in RAI. In another word, RAI is likely present in critically ill patients. Thus, rather than abandoning RAI, further efforts should be made to explore the contribution of RAI to sepsis and to develop new diagnosis for RAI. In this study, using SF1creSR-BI^fl/fl mice as a RAI model, we demonstrated that RAI mice failed to produce iGC and were susceptible to CLP-induced sepsis death. Supplementation of GC rescued RAI mice. Thus, RAI is a risk factor for death in this mouse model of sepsis.

Inflammatory response is an essential host response. It is required for fighting against infections but hyperinflammatory response causes organ injury and death in sepsis (3–7). For septic patients, some have controlled inflammatory response along with recovery; some develop immunosuppression who are susceptible to secondary infections; and some displace uncontrolled hyperinflammatory response. Thus, identification of inflammatory endotype is critical for an effective therapy. Upon sepsis challenge, both SR-BI^fl/fl and SF1creSR-BI^fl/fl mice developed systematic inflammatory response at the initial stage of sepsis. Twenty hours post CLP, the inflammatory response was well-under control in SR-BI^fl/fl mice, but maintained at hyperinflammatory status in SF1creSR-BI^fl/fl mice. Supplementation of GC controlled the inflammatory response in SF1creSR-BI^fl/fl mice. Thus, RAI is an endotype of sepsis which features persistent hyperinflammatory response.

So far, extensive clinical trials have had little impact on patient survival (43). A limitation is that these therapies were applied to all septic patients nonselectively. However, septic patients have heterogeneous clinical endotypes. While multiple studies have examined septic patients from the perspective of isolated biomarkers, few studies characterize septic patients according to endotype, and understanding of septic endotypes has not yet translated to patient care (44). Given the complexity of sepsis, emerging voices (44, 45), including ours (46, 47), have called for a precision medicine approach for sepsis therapy. Targeting a subgroup of patients with an endotype is a key component for a successful precision medicine approach (48, 49). In this study, we identified RAI as a risk factor and an endotype of sepsis, and showed that GC treatment benefits mice with RAI but harms mice without RAI. It is worth noting the dose of GC treatment. Three milligram hydrocortisone/kg body weight in mouse is equivalent to 15 mg hydrocortisone/62.5 kg body weight in human (50). Based on an early study, it is a low stress dose (51), 13-fold less than the current recommended GC dose (200 mg/day). Our study reveals the adrenal stress response (endogenous iGC) as a mechanism of controlling hyperinflammatory response in RAI and provides a mechanistic support for a low stress dose of GC for sepsis treatment, which may reduce the side effect of GC. This proof-of-concept study supports a precision medicine approach for sepsis therapy – selectively applying GC therapy for a subgroup of patients with RAI.

There are concerns that the rodent CLP model may not fully mimic human sepsis due to potential differences in host response between human and rodents. GC is a potent immune regulator at both physiological and stressed conditions. GC is present at 20 - 200ng/ml in circulation at physiological conditions in both human and mouse; upon septic stress, human mounts a robust adrenal stress (iGC production), and the GC levels are upregulated by 5 - 10 folds (52). Upon CLP challenge, mouse mounts a similar adrenal stress to septic patients, and the GC levels are upregulated by 5 - 10 folds (28, 29, 31). Thus, CLP induces similar adrenal stress response to septic patients, and in term of elucidating the role of iGC in sepsis, CLP is an appropriate and useful model. It is worth noting that the current Survival Sepsis Guidelines do not recommend to identify septic patients with RAI because a lack of clinical evidence (16). As discussed in the Introduction, the ACTH stimulation test may not appropriately identify patients with RAI due to technical limitations, which may prevent us from truly evaluating GC therapy in septic patients. Our study encourages effort to explore new diagnosis for RAI and then to re-evaluate the efficacy of GC therapy in patients with RAI.