Immunosuppression in Patients Who Die of Sepsis and Multiple Organ Failure FREE

Sepsis is responsible for more than 225 000 deaths annually in the United States.¹ Developing new therapies for sepsis has been particularly challenging, with more than 25 unsuccessful drug trials.^2- 6 Characterized by an initial intense inflammatory response or “cytokine storm,” patients with sepsis may present with fever, shock, altered mental status, and organ dysfunction.^5- 8 Numerous investigative agents have been directed at down-modulating this initial phase. Improved clinical management algorithms have led to survival of the majority of patients in this early period.^9- 11Quiz Ref IDHowever, those who survive early sepsis often develop nosocomial infections with organisms not typically pathogenic in immunocompetent hosts and have reactivation of latent viruses.^9,12- 13 These observations have led to the controversial hypothesis that the early hyperinflammatory state evolves to a subsequent hypoinflammatory state with significant immunosuppression.^14- 19 Although animal studies demonstrate progression to an immunosuppressive phase, epidemiologic studies in clinical sepsis are lacking.^14- 19 The purpose of this investigation was to assess evidence of immunosuppression in sepsis and to determine mechanisms that might be responsible for the presumed impaired immunity. Cells from spleen and lung were studied to compare and contrast the functional status and phenotype of cells from a lymphoid organ and a peripheral organ that is a frequent site of nosocomial infection.

An overview of study design including the purpose of the various immunologic tests is shown in Table 1. Methods are described in further detail in the eAppendix).

Patients with sepsis who died while undergoing treatment in surgical or medical intensive care units (ICUs) were included. Sepsis was defined using a consensus panel definition: presence of microbiologically proven, clinically proven, or suspected infection and presence of systemic inflammatory response syndrome (SIRS). The diagnosis of SIRS required at least 2 of the following: hypothermia (≤36°C) or hyperthermia (≥38°C); tachycardia (≥90/min); tachypnea (≥20 breaths/min) and/or arterial PCO₂ 32 mm Hg or lower and/or mechanical ventilation; and leukocytosis (≥12 000/μL) or leukopenia (≤4000/μL) and/or left-shifted white blood cell differential count of 10% or higher.²⁰

To try to limit potential confounding effects of other conditions affecting immunity, patients with cancer, chronic viral infections (human immunodeficiency virus, hepatitis B or C), or autoimmune diseases and patients taking high-dose corticosteroids (hydrocortisone, >200 mg/d) or immunosuppressive medications were excluded.

Control spleens were obtained from critically ill patients with no sepsis including those declared dead by neurological criteria and trauma patients who had emergent splenectomy because of splenic injury (see eAppendix for inclusion and exclusion criteria). Patients who met brain death criteria typically had been mechanically ventilated for 48 to 72 hours, required vasopressors to maintain adequate organ perfusion, and had associated organ injuries. Patients requiring mechanical ventilation for more than 4 days were excluded because of the high incidence of ventilator-associated pneumonia. Control lung tissue was obtained from the excess tissue of transplant donor lungs and from the non–tumor-involved tissue of lobectomies performed for cancer (eAppendix).

All studies were approved by the Washington University or St John's Hospital human research protection offices. For tissue obtained by rapid autopsy of patients who died in the ICU, the studies were determined by the Washington University and St John's Hospital human research protection offices not to constitute human subjects research because there was no interaction with the patient prior to death. After the patient had died, written permission to collect postmortem tissue for research purposes was obtained from the patient's next of kin. For lung resection specimens, all participants provided written informed consent prior to surgery. For lung tissue obtained from transplant donors, the tissue was collected after death and provided to the laboratory completely anonymized; therefore, it was also determined not to constitute human subjects research. However, the transplant donation consent signed by next of kin included a provision that tissue may be used for research purposes. Similarly, spleens were obtained from Mid-America Transplant, St Louis, Missouri, from brain-dead organ donors whose next of kin gave written consent for use of tissues for research purposes.

Spleen samples were obtained from 40 patients with sepsis and 29 patients without sepsis within 30 to 180 minutes of death. Splenocytes were dissociated and resuspended in sterile media.²¹ Splenocytes either were studied acutely or stored at 4°C for subsequent analysis within 12 to 72 hours. Cell counting and viability were determined as described previously.²¹

Lungs samples from patients with sepsis (n = 34) or without sepsis (n = 20) were immediately processed. Lungs were fixed overnight for immunohistochemistry or single-cell suspensions were prepared, counted, viability determined, and stained for flow cytometry.

Mononuclear cells were prepared by density gradient centrifugation.²¹ Cell viability was greater than 85% to 90%. Cells were stimulated with lipopolysaccharide, anti-CD3/anti-CD28, or phorbol 12-myristate 13-acetate (PMA)/ionomycin. Supernatants were harvested at 5 and 22 hours and cytokines quantitated by enzyme-linked immunosorbent assay.

Cell suspensions were incubated with isotype control or type-specific antibodies. To determine the percentage of cells positive for each marker, the percentage positive isotype control was subtracted from that within the positive gate of the type-specific marker. The mean fluorescence index (MFI) was determined by subtracting the geometric MFI of the isotype control from the geometric MFI of the type-specific marker.

For spleen tissue, formalin-fixed sections underwent antigen retrieval followed by incubation with primary antibodies or isotype-matched controls. Slides were sequentially incubated with biotinylated antibody and peroxidase-labeled streptavidin and immunoreactive cells were visualized with diaminobenzidine-chromogen. Lung tissue was fixed in paraformaldehyde. Sections underwent endogenous peroxidase quenching and antigen retrieval followed by incubation with primary antibodies or isotype-matched controls.²² Slides were incubated in horseradish peroxidase (HRP)–conjugated antibodies followed by visualization with HRP-AB-C substrate.

Slides were evaluated in blinded fashion and scored as described previously.²³

Differences in sepsis vs nonsepsis cytokine production and phenotypic expression were analyzed by 2-tailed nonparametric t test (Mann-Whitney U test) using the statistical program GraphPad Prism, version 5.0 (GraphPad Software). Statistical significance was set at P ≤ .05.

The most common etiologies of sepsis were ventilator-associated pneumonia and peritonitis (Table 2 and eTable 1). Other causes included necrotizing fasciitis, retroperitoneal abscess, infected intravascular catheters, urinary tract infections, intrapelvic abscess, and osteomyelitis. The nonsepsis patient population for control spleen tissues consisted of 20 patients who met brain death criteria and 9 patients who had splenectomies because of traumatic injury (Table 2 and eTable 2). Control populations for lung consisted of specimens from lung transplant donors (n = 10) or lobectomy for lung cancer (n = 10).

The mean ages of patients with sepsis and controls were 71.7 (SD, 15.9) and 52.7 (SD, 15.0) years, respectively. The median number of ICU days for sepsis patients was 8 (range, 1-195), while the median duration of sepsis was 4 days, with a range of 1 to more than 40 days (Table 2). Control patients were in ICUs for 4 or fewer days. Patients with sepsis had numerous comorbidities; in contrast, comorbidities were much less frequent in organ donor controls. The mean serum albumin level for sepsis patients was 2.4 (SD, 0.62) g/dL, with a range of 1.0 to 3.9 g/dL (normal hospital laboratory range for albumin is 3.6-5.0 g/dL) (eAppendix and eFigure 4).

Compared with controls, splenocytes from sepsis patients produced significantly fewer cytokines (tumor necrosis factor [TNF], interferon [IFN-γ], and interleukins [IL] 6 and 10) at 5 and 22 hours (Figure 1, eFigure 1, eTable 3, and eTable 4). Decreased cytokine production was profound at 5 hours, when selected sepsis tissues produced minimal amounts of cytokines regardless of stimulus. For example, cytokine production stimulated with anti-CD3/anti-CD28 at 5 hours from controls vs sepsis patients, respectively, resulted in TNF, 5361 (95% CI, 3327-7485) pg/mL vs 418 (95% CI, 98-738) pg/mL; IFN-γ, 1374 (95% CI, 550-2197) pg/mL vs 37.5 (95% CI, −5 to 80) pg/mL; IL-6, 3691 (95% CI, 2313-5070) pg/mL vs 365 (95% CI, 87-642) pg/mL; and IL-10, 633 (95% CI, −269 to 1534) pg/mL vs 58 (95% CI, −39 to 156) pg/mL (P < .001 for all). There were similar reductions in 5-hour lipopolysaccharide- and PMA/ionomycin–stimulated cytokine secretion (Figure 1, eFigure 1, eTable 3, and eTable 4).

By 22 hours, some splenocytes from sepsis patients exhibited partial recovery of cytokine production. Collectively, secretion of cytokines by stimulated sepsis splenocytes at 5 hours was less than 10% of that of controls. However, by 22 hours, secretion had increased to approximately one-third that of controls and, for selected sepsis patients, cytokine production was similar to that of controls (Figure 1). Cell viability at 5 and 22 hours was 93% and 86%, respectively, for controls vs 81% and 60%, respectively, for sepsis patients.

To determine if the age differential between controls and sepsis patients was associated with changes in stimulated cytokine production, comparison of data from sepsis patients aged 52 years or younger (n = 5) vs older than 52 years (n = 21) was performed and revealed no statistical differences between these 2 groups (eFigure 2). Furthermore, for virtually all cytokines, sepsis patients in both age groups were statistically different from controls. To determine if duration of sepsis affected cytokine secretion, data from patients who had sepsis for 4 or fewer days were compared with that of patients who had sepsis for more than 4 days. No significant difference was seen between these 2 groups (eFigure 3). Likewise, because of potential confounding effects of low-dose corticosteroids on cytokine production, comparisons were also made between sepsis patients receiving corticosteroids (n = 9) and not receiving corticosteroids (n = 15). Data from both of these groups were statistically different from that of controls but were not different from that of each other (eTable 4). To determine if the patient nutritional state (as reflected by serum albumin) had an effect on cytokine secretion in sepsis patients, TNF and IL-6 secretion were plotted against patient serum albumin (eFigure 4). No correlation was observed.

To assess possible etiologies for the markedly depressed cytokine secretion, we performed flow cytometric analysis and examined expression of cell surface receptors important in cellular activation (Figure 2, Figure 3, and eTable 5). The splenic T cells of sepsis patients had increased expression of activation marker CD69 on both CD4 and CD8 subsets and of IL-2Rα (CD25) on CD4 T cells, when analyzed both as percentage of positive cells and as cellular expression (as determined by geo-mean fluorescence intensity [MFI]). Expression of the percentage positive cells and MFI of the potent positive costimulatory receptor CD28 was decreased on CD4 T cells in sepsis; MFI for CD28 was also decreased on CD8 T cells. Inhibitory members of the CD28 family programmed cell death 1 (PD-1) (CD279) and cytotoxic T-lymphocyte antigen 4 (CTLA-4) (CD152) were also examined. Sepsis patients had significantly increased percentages of CD4 T cells expressing PD-1 and increased CTLA-4 positive CD8 T cells. Both MFI and the percentage of cells expressing the IL-7 receptor α (IL-7Rα, CD127), which is critical for cell survival and is decreased in cell exhaustion, were reduced on CD4 and CD8 T cells in sepsis spleen.

CD28 and PD-1 are engaged by their ligands (CD80/CD86 for CD28, PD-L1/PD-L2 for PD-1), which are expressed primarily on antigen-presenting cells (APCs), including dendritic cells and macrophages. Similar to findings on T cells, the costimulatory ligand CD86 was decreased and the inhibitory ligand, PD-L1, was increased on both macrophages and other APCs (Figure 2 and Figure 3). HLA-DR expression was highly significantly decreased on these APCs.

Cells from lung tissue were extensively characterized for activating and inhibitory receptors (eAppendix), and data are presented for markers in which we detected meaningful differences (Figure 4 and eTable 6). The majority of T cells from lungs of all groups expressed PD-1. However, the MFI of PD-1 was higher on CD4 cells from sepsis patients compared with both transplant and lung cancer controls. Analysis of B- and T-lymphocyte attenuator (BTLA) (CD272), another T-cell inhibitory ligand, revealed that BTLA was also expressed on the majority of T cells but was increased primarily on cells from lung cancer patients. Interestingly, both PD-L1 and PD-L2 expression were markedly increased on plasmacytoid dendritic cells of sepsis patients, presenting opportunity for potent inhibitory receptor-ligand interactions. In contrast to plasmacytoid dendritic cells, PD-L2 was not detected on myeloid dendritic cells, and PD-L1 was expressed at low levels that were not statistically different between sepsis patients and controls.

Expansion of suppressive cells, including regulatory T cells and myeloid-derived suppressor cells (MDSCs), has been reported in sepsis and provides another plausible mechanism for immunosuppression.¹⁰ In spleen, regulatory T cells were increased approximately 2-fold in sepsis vs control patients (mean, 8.5% [SD, 0.9%] vs 4.6% [SD, 0.6%], respectively; P < .05). In contrast, in lung, no increase in regulatory T cells was detected, but there were increased cells consistent with an MDSC phenotype, defined by Lin1-/HLA-DR low/CD33/CD11b high staining, isolated from sepsis vs control tissue (47.9% vs 15.7%; P ≤ .01).

Evaluation of spleen tissue demonstrated cellular loss in periarteriolar lymphoid sheath (PALS) and diminished number and size of splenic follicles in sepsis patients, as previously reported.²⁴ Significantly reduced numbers of CD4, CD8, and HLA-DR cells (dendritic cells, macrophages, and B cells) typified sepsis splenic tissue (Figure 5), though a small subset of sepsis tissue retained a “normal” complement of these cell populations (eTable 7). Nineteen of 22 sepsis patients (but no controls [n = 12]) had either decreased HLA-DR positivity, depletion of HLA-DR cells, or both. Conversely, in red pulp, increased expression of HLA-DR on sinusoidal endothelial lining cells was noted in 16 of 26 sepsis patients (Figure 5) but not in any controls. PD-1 was noted in follicular dendritic cells but generally not in T-cell zone dendritic elements. In contrast, PD-L1 was more typically seen in T-cell zone dendritic cells. In sepsis spleen tissue with severe cellular depletion, the dominant PD-L1–positive PALS population was capillary endothelium, and this was increased relative to controls (eFigure 5). Macrophages were consistently PD-L1 positive in all cell zones and were increased in PALS and red pulp. PD-L1 was uniformly present in arteriolar and sinusoidal endothelium in both sepsis and control spleen. PD-1 reactivity was unevenly expressed in arteriolar endothelium and absent on sinusoidal linings. Both B and T lymphocytes were variably positive for PD-1 and PD-L1 in both sepsis and control patients.

Signaling through inhibitory receptors on T cells requires engagement by their cognate ligand. While typically expressed by APCs, these may be induced on nonimmune cells by inflammatory cytokines, perhaps serving to locally inhibit T cells and dampen inflammation. Immunohistochemistry showed intense airway epithelial staining for herpes virus entry mediator (HVEM), the receptor for BTLA, on lung isolated from the majority of sepsis patients but not controls (Figure 6 and eTable 8). This was true for patients both with and without pneumonia. Herpes virus entry mediator was expressed on macrophages from both groups. We also stained for PD-L1 and PD-L2, the ligands for PD-1 (Figure 7). PD-L1 and PD-L2 appeared to be expressed in sepsis lung and lung resections to a greater degree compared with transplant donor lung.

Early sepsis is characterized by excessive inflammation in what is often termed a “cytokine storm.”^2- 8 As sepsis persists, patients often have reactivation of endogenous viruses and develop nosocomial infections with opportunistic pathogens.^9,12- 13 Investigators have argued that these findings suggest that patients with sepsis enter an immunosuppressive state, but this is highly controversial and most potential therapy for sepsis remains focused on blocking immune activation.^14,25- 28 While several potential abnormalities have been identified, a comprehensive analysis of the immune status of patients who die of sepsis has not been conducted and mechanistic explanations remain speculative.^25- 29Quiz Ref IDThe present study shows that splenocytes from sepsis patients had highly significant functional impairments as evidenced by major reductions in cytokine secretion. Multiple inhibitory mechanisms were identified, including dominance of inhibitory over activating receptors, expansion of suppressive cell types, and induction of inhibitory ligands on both APCs and tissue parenchymal cells. These findings are present in the setting of apoptosis-induced depletion of immune cells (Figure 5).^23- 24 Both proinflammatory and anti-inflammatory cytokines were impaired. These spleen data are consistent with multiple previous studies examining blood that showed similar degrees of cytokine suppression in sepsis.^25,30- 32 Thus, the severely reduced cytokine production observed in spleen is consistent with a systemic abnormality. In a subset of sepsis patients, cytokine production at 22 hours was comparable with controls without sepsis, suggesting that in some patients, defective cytokine secretion may be reversible if cells are removed from the sepsis milieu.

Of particular interest is the expression of ligands for T-cell inhibitory receptors on tissue parenchymal cells. Compared with controls, splenic capillary endothelial cells of sepsis patients had increased expression of the inhibitory ligand PD-L1 (eFigure 5). Similarly, lung cells from sepsis patients, both with and without pneumonia, demonstrated intense airway epithelial staining for HVEM, a key regulator of host inflammatory response in autoimmunity and infection (Figure 6).³³Quiz Ref IDSignificantly, the majority of lung T cells expressed BTLA, one of several ligands for HVEM and another regulator of host inflammation.³³ Thus, the required elements for activation of this immunosuppressive pathway are present in sepsis. Collectively, these data demonstrate that parenchymal cells of spleen and lung express important immunoregulatory proteins and that HVEM appears to be specifically induced in lungs of sepsis patients.

Quiz Ref IDThe PD-1 pathway has emerged as an important mechanism that inhibits T-cell function. Expression of PD-L1 and PD-L2 by dendritic cells promotes a tolerogenic phenotype, resulting in T-cell suppression.^34- 36 In lung from sepsis patients, both PD-L1 and PD-L2 were detected on resident dendritic cells and airway epithelial cells. Thus, these pathways might be engaged on T cells trafficking through lung, resulting in localized inhibition of T cells and thereby predisposing to infection at this site. In spleen, PD-1 expression was detected on CD4 and CD8 T cells, and both PD-L1 and PD-L2 are known to be expressed on endothelial cells (eFigure 5) and impair CD8 T-cell cytolytic properties.³⁷

A long-standing question in sepsis is whether immunologic changes are organ specific.³⁸ Although differences existed, both spleen and lung shared common immunosuppressive mechanisms. CD86 and HLA-DR expression were decreased in cells isolated from both spleen and lung (Figure 1). In contrast, regulatory T cells were increased in spleen but not in lung. T cells isolated from both organs of sepsis patients had increased expression of the major inhibitory receptor PD-1. Relative to spleen, there were twice as many PD-1–expressing cells in lung, suggesting the possibility of selective recruitment of PD-1–expressing cells to lung. In both organs, expression of ligands for major inhibitory receptors was detected on APCs and parenchymal cells.

Functional unresponsiveness of T cells during chronic viral infections has long been recognized and recently mechanistically described as a state of cell “exhaustion.”^39- 41 Driven by persistent antigen exposure, exhaustion is typified by progressive loss of cell function.^39- 41 Phenotypically, exhaustion is characterized by persistent expression of multiple inhibitory receptors, including PD-1, TIM-3, and LAG-3, along with decreased expression of CD127 (IL-7Rα) and CD62L. The functional and phenotypic characteristics described for cell exhaustion are similar to the present findings in sepsis, a state that is highly likely to result in protracted antigen exposure and cell stimulation.³⁹ As noted in the present study, splenic T cells demonstrated profound reductions in cytokine production as well as increased expression of PD-1and CD69 and decreased CD127. Collectively, these findings suggest that T-cell exhaustion may be an important immunosuppressive mechanism in sepsis.

Quiz Ref IDThe present study has a number of important therapeutic implications. Most investigative agents in sepsis have been directed at blocking inflammation and immune activation. Although such therapies may be successful if applied early, they may be harmful if applied later in the immunosuppressive phase. As supportive therapies of sepsis have improved, early deaths have decreased and most patients enter a more protracted phase, with evidence of impaired immunity made manifest by infections with relatively avirulent organisms.^9,14- 15 An important part of implementing more targeted therapies will be to accurately determine the immune status of individual patients during their disease. By using a combination of functional assays and flow cytometry to characterize immune cell phenotypic changes, it may be possible to identify patients with more severe immune compromise, for which targeted immune-enhancing therapies may be beneficial. A recent study that treated sepsis patients with granulocyte-macrophage colony-stimulating factor based on monocyte HLA-DR expression suggests that such therapies might be successful.⁴² Our findings suggest additional targets. Interleukin 7 has an excellent clinical safety record and has been used to enhance immunity in patients with cancer, human immunodeficiency virus type 1, and hepatitis C. Interleukin 7 restored immunity in patients with persistent viral infections and has improved viral clearance and survival in animal models of chronic viral disease and sepsis.^21,43- 45 The decreased IL-7Rα expression documented in the present study could be causative in T-cell loss in sepsis, and IL-7 administration might ameliorate this problem Another potential therapeutic approach is to interfere with inhibitory receptor signaling on T cells. We have demonstrated up-regulation of multiple inhibitory receptors and their ligands, including PD-1 and BTLA. Blockade of PD-1/PD-L1 interactions in chronic viral infection has been shown to reverse the functional unresponsiveness of T cells and to enhance viral clearance.⁴⁰ Blockade of PD-1 improves survival in animal models of fungal infection and sepsis.^46- 48

This study has several important limitations of the ability to generalize the findings to sepsis at large. These limitations include the relatively small sample size and the heterogeneous nature of both sepsis and control patients. In this regard, the control groups were markedly different from the sepsis patients in many aspects; for example, nutritional status, degree of comorbidities, and length of illness. Although the hypoalbuminemia noted in the sepsis patients could be due to a number of causes such as increased vascular permeability, poor nutritional status was undoubtedly a major contributing factor, and malnutrition has numerous effects on host immunity that could be responsible for some of the observed immunologic findings.⁴⁹ Also, the study was confined to sepsis patients who died in the ICU, some after a considerable duration of sepsis. Thus, the process may be different for sepsis patients who die sooner of sepsis, who do not die in the ICU, or who make a good recovery, as well as for a broader case mix in general. Finally, it is possible that findings in sepsis patients were prodromal events and not reflective of patients with sepsis in general but of patients with sepsis who had unsuccessful responses to supportive measures. It should be emphasized that deaths within the first 72 hours of sepsis in previously healthy patients with infections of highly virulent organisms are associated with extremely elevated proinflammatory cytokines, and these deaths are likely secondary to an overexuberant immunoinflammatory response.^7- 8,14 Thus, the present study serves as a bridge between preclinical and early clinical findings and must be viewed cautiously.

In conclusion, these data provide a unique insight into the status of the immune system during sepsis, not only in a lymphoid organ but in peripheral tissue. Identification of potential receptor-ligand interactions and signaling pathways leading to immunosuppression may allow for targeted therapeutic interventions to restore host immunity.