Sepsis is defined as the host inflammatory response to severe, life-threatening infection with the presence of organ dysfunction1. Sepsis is the most frequent cause of mortality in most intensive care units (ICUs) and is responsible for more than 250,000 deaths in the United States annually2. The incidence of sepsis is increasing owing to the ageing population, who have impaired immunity as a result of immunosenescence2. Despite the litany of failed clinical trials in sepsis, a better understanding of different immunological phases of the disorder and encouraging results from several Phase II clinical trials of immunotherapies in sepsis have brought cautious optimism to the field3,4,5,6,7.

Until recently, most research on sepsis was focused on blocking the initial hyperinflammatory, cytokine-mediated phase of the disorder. Improved treatment protocols have resulted in most patients surviving this initial hyperinflammatory phase and entering a protracted immunosuppressive phase8,9,10,11,12,13. Deaths in this immunosuppressive phase are typically due to failure to control the primary infection or the acquisition of secondary hospital-acquired infections, often with opportunistic pathogens14,15. The recent remarkable success of cytotoxic T lymphocyte antigen 4 (CTLA4)- and programmed cell death 1 (PD1)-specific antibodies as immunotherapies to improve host immunity and to increase survival in cancer patients16,17 is highly encouraging because of the many similarities between the immune defects observed in cancer and those observed in sepsis and because both agents result in improved survival in animal models of sepsis7,10. In this Review, we discuss the panoply of sepsis-induced defects in innate and adaptive immune cells and highlight several promising immunotherapies for the treatment of sepsis.

Controversies over host immunity in sepsis

The current paradigm regarding the host immune response to sepsis is debated2,3,4,5,6,7,18,19. Traditionally, the host immune response to sepsis was considered to be characterized by an initial hyperinflammatory phase that evolved over several days into a more protracted immunosuppressive phase7,8,9 (theory 1). However, recent studies have shown that both pro-inflammatory and anti-inflammatory responses occur early and simultaneously in sepsis18,19,20 (Fig. 1a), although the net initial effect of these competing processes is typically manifested by an early, dominant, hyperinflammatory phase characterized by shock, fever and hypermetabolism. The robustness of the hyperinflammatory phase depends on numerous factors, including pre-existing co-morbidities, nutritional status, microorganism load and virulence factors8,9.

Figure 1: Competing theories of the host immune response in sepsis.
figure 1

a | Theory 1: recent studies show that activation of both pro- and anti-inflammatory immune responses occurs promptly after sepsis onset. Cells of the innate immune system, including monocytes and neutrophils, release high levels of pro-inflammatory cytokines that drive inflammation (blue line; days 1–3). The intensity of the initial inflammatory response varies in individual patients depending on multiple factors, including pathogen load and virulence, patient co-morbidities and host genetic factors. Early deaths in sepsis (top red line; day 3) are typically due to a hyperinflammatory 'cytokine storm' response with fever, refractory shock, acidosis and hypercatabolism. An example of this scenario would be a young patient dying of toxic shock syndrome or meningococcaemia. Most patients have a restoration of innate and adaptive immunity and survive the infection (recovery; blue and green lines; day 6). If sepsis persists, the failure of crucial elements of both the innate and the adaptive immune systems occurs such that patients enter a marked immunosuppressive state (blue and red lines; after day 6). Deaths are due to an inability of the patient to clear primary infections and the development of secondary infections. b | Theory 2: a competing theory of sepsis agrees that there is an early activation of innate immunity and a suppression of adaptive immunity; however, this theory holds that deaths in sepsis are due to the persistent activation of innate immunity that results in intractable inflammation and organ injury. According to this theory, late deaths in sepsis are due to persistent, underlying innate immune-driven inflammation.

PowerPoint slide

Investigators recently reported a new paradigm (Fig. 1b) to describe the host immune response in trauma and sepsis (theory 2). Circulating leukocyte gene expression data from people suffering from trauma and burns showed the rapid and sustained upregulation of genes that regulate the innate immune response and the simultaneous downregulation of genes that regulate the adaptive immune response19. These investigators hypothesized that the best model to describe the host immune response in trauma and sepsis is one of protracted, unabated inflammation driven by the innate immune system with resultant organ dysfunction and failure. Although these investigators agree that the adaptive immune system is impaired, they theorize that patients who die of sepsis have a longer duration and a greater degree of organ injury caused by unabated innate immune-driven inflammation than those who survive. They postulate that this inflammation persists despite the downregulation of the expression of genes that regulate the adaptive immune response and that it is ultimately responsible for patient morbidity and mortality19.

Although we agree with the provocative findings of this group, we believe that this new model — which proposes that morbidity and mortality in sepsis are due to unremitting innate immune-driven inflammation — is unlikely to reflect the actual clinical scenario in most patients. We have shown that patients who die of sepsis have marked immunosuppression12. Immune cells that were harvested from the spleen or lungs of patients with sepsis within 30 to 180 minutes of death showed markedly decreased production of both pro- and anti-inflammatory cytokines, upregulated expression of inhibitory receptors (including PD1), expansion of regulatory T (TReg) cell and myeloid-derived suppressor cell (MDSC) populations, and downregulation of CD28 and HLA-DR-mediated activation pathways12. Collectively, these results show that sepsis induces numerous overlapping mechanisms of immunosuppression involving both the innate and adaptive immune systems.

There are several potential explanations for the different findings in the two studies. First, the results of the gene expression study involved patients with trauma and burns19, and although many of these patients became septic, these patients differ from those who developed sepsis as a primary phenomenon. The mean age of the patients with trauma in the transcriptome study was 33 years, whereas the mean age of patients with sepsis in most developed countries is at least twice this age (65–68 years). This difference in age in the two studies is important because the older patients have underlying defects in immunity due to immunosenescence21. Second, these investigators based their conclusions on gene expression data that, because of post-transcriptional regulation, might not reflect protein levels. Conversely, the post-mortem study of patients with sepsis quantitated actual cytokine levels. Other possible reasons for differences between these two studies are the duration and severity of illness of the patients who were analysed. Many patients with sepsis had protracted disease with a high severity of illness and increased mortality.

We believe that immunosuppression rather than low-grade inflammation is the predominant driving force for morbidity and mortality in sepsis. First, our post-mortem findings are consistent with numerous studies that have examined cytokine production in peripheral-blood mononuclear cells and whole blood from patients with sepsis and have shown severely decreased pro-inflammatory cytokine production22,23,24,25. Second, recent post-mortem studies have reported that a high number of patients who die of sepsis have unresolved opportunistic infections14,15, which is consistent with defective host immunity as a dominant cause of death. Third, the failure of more than 30 clinical trials of different anti-inflammatory agents is inconsistent with the hypothesis that inflammation is a key driving mechanism. Undoubtedly, focal regions of inflammation occur in infected tissue in patients with sepsis. Areas of ischaemic and necrotic tissue are likely to contribute to local inflammation. It is possible that this focal inflammation contributes to organ dysfunction, morbidity and mortality in a subset of patients with sepsis. Nevertheless, we hypothesize that protracted sepsis is predominantly characterized by systemic immunosuppression leading to a failure to eradicate primary infections and to the acquisition of lethal secondary infections. Intriguingly, the gene expression data from children who develop sepsis show similarities to the data from adult patients with trauma-related sepsis: for example, the expression of genes that modulate innate immunity is upregulated, whereas the expression of genes that modulate adaptive immunity is downregulated26. Although these gene expression data would seem to favour theory 2, post-mortem studies of paediatric patients who die of sepsis have shown immunosuppressive features that are characteristic of those observed in adult patients with sepsis, who show a remarkable depletion of immune effector cells (see below). Hopefully, future studies will determine which of these two conflicting hypotheses is correct, as the resolution of this controversy has important therapeutic implications.

Apoptosis and immunosuppression

A fundamental discovery in the field of sepsis occurred when investigators showed that apoptosis causes the marked depletion of immune cells, including CD4+ and CD8+ T cells, B cells, follicular dendritic cells (follicular DCs) and interdigitating DCs, in various organs of patients dying of sepsis, leading to immunosuppression27,28,29,30,31 (Fig. 2). Sepsis-induced immune cell apoptosis has now been confirmed in several post-mortem studies; it affects all age groups (neonatal, paediatric and adult populations)27,30,31 and occurs in response to various microorganisms. Apoptosis of immune cells occurs in lymphoid tissues (spleen, thymus and lymph nodes) and gut-associated lymphoid tissues (GALTs)32. The loss in intestinal intraepithelial and lamina propria lymphocytes might facilitate bacterial translocation into the systemic circulation, thereby perpetuating the systemic inflammatory response and predisposing to secondary infections. Sepsis-induced apoptosis occurs through both death receptor- and mitochondrial-mediated pathways, suggesting that multiple cell death stimuli are activated during sepsis33. Therefore, it is unlikely that blockade of a single apoptotic trigger will prevent lymphocyte cell death during the disorder.

Figure 2: Immunohistochemical staining for CD4+ T cells, CD8+ T cells and HLA-DR in spleens from patients with sepsis or trauma.
figure 2

Spleens from patients with sepsis or patients who required a splenectomy for traumatic injury were obtained and underwent immunostaining for CD4+ T cells, CD8+ T cells or HLA-DR. Sepsis induces the apoptotic death of splenocytes, including CD4+ and CD8+ T cells, leading to marked depletion of these crucial immune effector cells. Note the decrease in numbers of periarteriolar CD4+ and CD8+ T cells (stained brown) in spleens from patients with sepsis compared with patients with trauma (original magnification ×200). The decrease in numbers of CD4+ T cells in patients with sepsis is frequently as severe as occurs in patients with AIDS. The number of HLA-DR+ cells (predominantly B cells) and the levels of HLA-DR expression (as determined by the intensity of the staining) are both decreased in patients with sepsis compared with patients with trauma (HLA-DR+ cells are stained brown). Tissue samples were stained for CD4, CD8 or HLA-DR. 3,3′-diaminobenzidine 4-HCl was used as a chromogen to stain the cells of interest (brown), and a hematoxylin counterstain (blue) was used for background staining. Images courtesy of P. Swanson, University of Washington School of Medicine in Seattle, Washington, USA.

PowerPoint slide

A key question is whether the extensive sepsis-induced apoptosis of immune cells is an epiphenomenon or a major pathophysiological mechanism. Multiple independent laboratories have shown — through the use of various strategies, including transgenic and knockout mice, anti-apoptotic cytokines, caspase inhibitors and death receptor antagonists — that preventing lymphocyte apoptosis improves survival in sepsis34,35,36,37. The detrimental effects of apoptosis are not only related to the severe loss of immune cells but also the impact that apoptotic cell uptake has on the surviving immune cells38. Uptake of apoptotic cells by monocytes, macrophages and DCs results in immune tolerance by inducing anergy or a T helper 2 (TH2) cell-associated immune phenotype with increased interleukin-10 (IL-10) production39. The net result of this effect is that the surviving phagocytic cells cannot combat the remaining pathogenic organisms. Thus, sepsis-induced apoptosis has multiple important effects that impair host defences.

Impact of sepsis on immune cells

Sepsis directly or indirectly impairs the function of almost all types of immune cells. The following section provides an overview of these various immunosuppressive effects on the different cells of the innate and adaptive immune systems.

Neutrophils and MDSCs. Neutrophils are essential for the early control of invading pathogens40. Investigations on neutrophils obtained during the first hours of sepsis uncovered numerous abnormalities (Fig. 3). Most neutrophils normally undergo apoptosis within 24 hours after release from the bone marrow40. Surprisingly, and in contrast to lymphocytes, which undergo accelerated apoptosis, neutrophil apoptosis is delayed during sepsis40. Because of an increase in the release of immature neutrophils and the delayed apoptosis of circulating neutrophils, patients with sepsis typically have markedly increased numbers of circulating neutrophils of various degrees of maturation41. Animal models of sepsis and experimental studies in patients revealed disrupted neutrophil functions, including impaired clearance of bacteria, reduced production of reactive oxygen species (ROS) and decreased recruitment to infected tissues42,43. Loss of chemotactic activity is probably the most frequently documented dysfunction of circulating neutrophils during sepsis43,44. Possible explanations for this include nitric oxide-mediated suppression and reduced expression of CXC-chemokine receptor 2 (CXCR2)44. A current theory is that these alterations in neutrophil function are due to abnormalities in Toll-like receptor (TLR) signalling, which is analogous to the phenomenon that occurs in monocytes during endotoxin tolerance43 (see below).

Figure 3: Impact of sepsis on innate and adaptive immune cells.
figure 3

a | Sepsis has diverse and important effects on all cellular elements that comprise the innate immune system. Sepsis rapidly triggers extensive apoptosis in follicular dendritic cells, dendritic cells, immature macrophages, natural killer (NK) cells and myeloid-derived suppressor cells (MDSCs). Conversely, sepsis delays neutrophil apoptosis — a result that is thought to be secondary to the mechanisms of neutrophil activation. After initial mobilization and activation of neutrophils, subsequent neutrophils that are released from the bone marrow have lower bactericidal functions and decreased cytokine production. Recent data show that a subset of neutrophils release large amounts of the immunosuppressive cytokine interleukin-10 (IL-10). Decreased HLA-DR expression on antigen-presenting cells, including monocytes, macrophages and dendritic cells, is a hallmark of sepsis, which may impair the optimal presentation of microbial antigens to T cells. b | Sepsis causes a marked loss of CD4+ and CD8+ T cells, as well as B cells. Regulatory T (TReg) cells are more resistant to sepsis-induced apoptosis and, consequently, there is an increased percentage of TReg cells in the circulation of patients with sepsis relative to the other lymphocyte subsets. This contributes to the formation of a more immunosuppressive phenotype. Surviving CD4+ and CD8+ T cells either shift from a pro-inflammatory T helper 1 (TH1) cell phenotype to an anti-inflammatory TH2 cell phenotype or develop an 'exhaustive' phenotype that is characterized by increased programmed cell death 1 expression and reduced cytokine secretion. CD4+ T cells have decreased expression of CD28 and reduced T cell receptor (TCR) diversity, which are both likely to contribute to the impaired antimicrobial response to invading pathogens.

PowerPoint slide

Several studies have reported that impaired neutrophil function precedes the development of nosocomial infections. Patients with the most severely reduced neutrophil functions have the highest risk of acquiring nosocomial infections45. Reduced neutrophil function has also been reported in a mouse model of polymicrobial sepsis, in which the mice had an increased susceptibility to secondary infection with Pseudomonas aeruginosa and the subsequent development of pneumonia46. Furthermore, ten patients with septic shock who underwent extracorporeal cell therapy with donor granulocytes showed improvement in various biomarkers of sepsis, as well as decreased sepsis severity47.

Neutrophils show great plasticity in response to a wide range of physiological or pathological conditions43. Sepsis can induce the formation of a subset of neutrophils with suppressive properties. Although neutrophils are often not considered to produce large amounts of cytokines, it has been shown that they produce large amounts of the immunosuppressive cytokine IL-10 during sepsis48. Recently, a subset of mature CD16hiCD62L (also known as L-selectin)low neutrophils that suppressed T cell proliferation was identified in a human model of endotoxaemia in which healthy volunteers were administered with a low dose of endotoxin49. Immunosuppressive neutrophils were also detected in five patients suffering from severe trauma49.

The recently investigated heterogeneity in neutrophils, including immunosuppressive subsets of myeloid cells, offers parallels with MDSCs. In experimental models of sepsis, these cells have been shown to block specific T cell functions, including proliferation and production of interferon-γ (IFNγ) and IL-2 (Refs 50, 51). However, the actual effect of MDSCs might evolve during sepsis, as studies show that MDSCs can either enhance or attenuate the sepsis-associated inflammatory response, depending on the stage of the disorder50,51,52,53. Our group recently showed that there was an increase in the number of cells that were consistent with an MDSC phenotype in the lungs of patients who died of sepsis12. Currently, few clinical studies have investigated MDSCs in patients with sepsis, possibly owing to the complexity of immunophenotyping these cells and the absence of a universally accepted phenotypic definition for these cells in humans54.

DCs. DCs are particularly vulnerable to sepsis-induced apoptosis29. In a post-mortem study of patients with trauma or with sepsis, researchers found a marked reduction in the number of splenic DCs and the percentage area of the spleen that was occupied by DCs in the patients with sepsis compared with patients with trauma29. Similarly, reduced numbers of circulating DCs have been recorded in patients suffering from burns and subsequent sepsis and in patients with severe sepsis55,56,57. Both plasmacytoid and myeloid DCs are effected by sepsis-induced apoptosis55,58 (Fig. 3). Interestingly, DC loss was more apparent in patients with sepsis who died than in survivors55, and it was also more marked in patients who subsequently developed nosocomial infections than in those patients who did not59.

Not only are DC numbers decreased in patients with sepsis, but the surviving DCs also have lower expression of HLA-DR and produce increased levels of IL-10 (Refs 57, 60). In addition, monocyte-derived DCs from patients with sepsis were unable to induce a robust effector T cell response but instead induced either T cell anergy or TReg cell proliferation61. In mouse models of burn injury, DCs possess immunosuppressive properties that impair defences against a subsequent bacterial challenge. Several investigators have shown that preventing sepsis-induced DC apoptosis or improving DC function during the disorder results in enhanced survival62,63,64. Indeed, mice that had selective overexpression of the anti-apoptotic factor B cell lymphoma 2 (BCL-2) in DCs were reported to have improved survival in a lethal model of endotoxic shock62. Importantly, these BCL-2-overexpressing DCs were resistant to sepsis-induced apoptosis, which suggests that DC death is an important determinant of sepsis-induced immunosuppression and mortality.

FMS-related tyrosine kinase 3 ligand (FLT3L) is a DC growth factor that induces a rapid increase in DC numbers. Treatment of burn-injured animals with FLT3L restored DC function and enhanced survival when mice were challenged with P. aeruginosa, which is a pathogen that commonly infects patients with burns63,64,65. The protective effect of FLT3L was also observed following the adoptive transfer of FLT3L-treated DCs. Additional studies documented that treatment with FLT3L increased the secretion of IL-12, IL-15 and IFNγ and had broad effects on the function of CD4+ T cells, natural killer (NK) cells and neutrophils in models of burn infection63,64,65. Intrapulmonary transfer of bone marrow-derived DCs restored DC function and prevented fatal Aspergillus fumigatus infection in mice that had recovered from primary peritonitis66. TLR agonists have also been shown to improve survival from pneumonia occurring in mice after haemorrhagic shock67,68. Possible mechanisms for the beneficial effect of TLR agonists include the increased expression of MHC class II molecules and of the co-stimulatory receptors CD80 and CD86 on DCs67,68. Given these encouraging results, some authors have argued that preserving and/or restoring DC function should be a primary target of investigations into sepsis63,64,65,66,67,68,69.

Monocytes and macrophages. A hallmark of sepsis is the diminished capacity of monocytes to release pro-inflammatory cytokines in response to endotoxin (lipopolysaccharide (LPS)), other TLR agonists and various other bacterial compounds70,71 (Fig. 3). This finding is consistent with the phenomenon of endotoxin tolerance. Monocytes from patients with sepsis typically show a diminished ability to release pro-inflammatory cytokines such as tumour necrosis factor (TNF), IL-1α, IL-6 and IL-12, whereas their ability to release anti-inflammatory mediators, such as IL-1 receptor antagonist (IL-1RA) and IL-10, is either unimpaired or enhanced70,71. These findings show that LPS can still activate monocytes but that their intracellular signalling has shifted towards the production of anti-inflammatory molecules, thereby supporting the concept of monocyte reprogramming72. In clinical studies, the magnitude and the persistent nature of this refractory state is associated with increased mortality and nosocomial infections73.

The mechanisms responsible for endotoxin tolerance are not fully understood70,72. Analysis of mRNA expression levels has shown increased expression of genes encoding inhibitory signalling molecules and inhibitory cytokines and reduced expression of pro-inflammatory and chemokine receptor genes71,72,73,74. Although they have mainly been reported as useful biomarkers for predicting sepsis onset or prognosis, microRNAs are also thought to participate in endotoxin tolerance71. Similarly, recent work highlights a prominent role for epigenetic regulation at different levels, such as histone modification and chromatin remodelling75,76. These epigenetic modifications are restored in vitro after IFNγ stimulation and are associated with the recovery of monocyte cytokine release76,77. Thus, IFNγ might abrogate endotoxin tolerance by facilitating TLR-induced chromatin remodelling78.

Two major consequences of endotoxin tolerance on monocytes and macrophages are an increase in the release of immunosuppressive mediators (mainly IL-10) and a decrease in antigen presentation as a result of reduced expression of HLA-DR; both are associated with a worse outcome in sepsis22,79,80,81. Continued release of IL-10 might contribute to or amplify sepsis-induced immunosuppression and thus might augment susceptibility to secondary microbial infections82,83,84. Indeed, blocking IL-10 reverses endotoxin tolerance ex vivo83,85 and can reverse sepsis-induced immunosuppression and improve survival in a clinically relevant animal model of sepsis86.

Low levels of monocyte HLA-DR expression function as a surrogate marker of monocyte unresponsiveness79,87. Several studies showed an association of reduced monocyte HLA-DR expression with impaired monocyte function (for example, lower levels of TNF and IL-1β release from monocytes in response to bacterial challenges88 and decreased lymphocyte proliferation in response to tetanus toxin, presumably due to impaired antigen presentation89). Most importantly, decreased monocyte HLA-DR expression is associated with increased risk of nosocomial infections and death90,91. It is noteworthy that after adjustment for usual clinical confounders through multivariate analysis, decreased monocyte HLA-DR expression remains an independent predictor of nosocomial infection occurrence and mortality after sepsis73,92. This finding shows that monocyte unresponsiveness and immunosuppression independently contribute to the increased risk of adverse events in sepsis.

NK cells. The preferential location of NK cells in tissues along with their low numbers in peripheral blood render the study of this cell subset difficult, especially in the context of human sepsis. These facts explain why NK cells have not been heavily investigated in sepsis93,94. Studies indicate that sepsis affects both of the main human NK cell subsets (that is the CD56hi and CD56low NK cell subsets); in patients with sepsis the number of circulating NK cells is markedly decreased95,96, often for weeks97, and low numbers of NK cells are associated with increased mortality98. Furthermore, the absolute number of both NK cell subsets is decreased in sepsis99. Reduced NK cell cytotoxic function and cytokine secretion occur during sepsis and following burns and traumatic injuries in animal models and in patients96,100,101,102,103.

Cytokine production by NK cells in response to TLR agonists was impaired after polymicrobial sepsis104, which suggests that NK cells become tolerant to TLR agonists, and have features similar to those of endotoxin tolerance in monocytes105. Impaired IFNγ production in response to TLR agonists such as LPS and CpG has also been reported in NK cells from patients with sepsis99. As NK cells have a central role in antiviral defence, it is possible that impaired NK cell function could lead to the reactivation of latent viruses; this phenomenon has been frequently described in patients in ICUs106,107,108,109. Decreased IFNγ production by NK cells occurs in patients with sepsis and frequently precedes cytomegalovirus reactivation105 in patients with protracted sepsis96. In addition, defective NK cell production of IFNγ could be a factor in the increased incidence of secondary infections and the decreased monocyte HLA-DR expression that occur in septic patients110.

γδ T cells. γδ T cells are a distinct subset of lymphocytes that reside in large numbers in the intestinal mucosa and that possess qualities common to both innate and adaptive immune cells. Although the breadth of antigens to which γδ T cells respond is not completely known, it is clear that they recognize lipid antigens that are present on various invading pathogens111. γδ T cells reside in the intestines and on other mucosal surfaces, where they recognize invading pathogens and mount a prompt, innate-like immune response by releasing IFNγ, IL-17 and various chemokines. As such, γδ T cells can be considered to represent a first line of defence against particular pathogens112. The number of circulating γδ T cells is significantly decreased in patients with sepsis, and more severe depletion occurs in patients who have the greatest severity of illness and mortality112. The loss of γδ T cells in the intestinal mucosa might be particularly detrimental to the host because it might lead to the invasion of intestinal pathogens into the circulation or the peritoneal cavity, thereby causing secondary infections.

CD4+ TH cell subsets. Numerous studies have reported marked effects of sepsis on circulating and tissue T cells, with some of the most considerable effects occurring in CD4+ T cells113,114,115,116,117,118. Mature CD4+ TH cells have been characterized into TH1, TH2 and TH17 cell subsets on the basis of the type of cytokines that they produce on stimulation. Early studies suggested that both TH1 and TH2 cell-associated cytokine production are decreased during the initial immune response to sepsis and trauma119,120,121,122 (Fig. 3). Additional work showed marked reductions in the expression of T-bet and GATA-binding protein 3 (GATA3), which are transcription factors that modulate the TH1 and TH2 cell response, respectively, and this has reinforced the idea that both the TH1 and TH2 cell lineages are suppressed after trauma and sepsis123. These studies also showed that expression of the TReg cell-associated transcription factor forkhead box P3 (FOXP3) is not altered during sepsis, which supports the idea that regulatory functions are maintained or increased during sepsis, whereas effector T cell responses are downregulated (see below)95,122,124. Histone methylation and chromatin remodelling are thought to contribute to the suppression of TH1 and TH2 cell functions by acting at promoter regions of the IFNG and GATA3 genes125.

Although the TH17 cell subset has been less well studied than the TH1 and TH2 cell subsets, there is now general agreement that TH17 cells protect against extracellular bacterial and fungal infections by producing IL-17 and IL-22 (Refs 126, 127, 128, 129). The TH17 cell response is reduced in sepsis, possibly as a result of decreased expression of retinoic acid receptor-related orphan receptor-γt (RORγt), which is the transcription factor that is specific for TH17 cells95,123. This defect in the TH17 cell phenotype in sepsis is likely to be a contributing factor to the increased susceptibility of these patients to secondary fungal infections129. Indeed, mice that survived an initial polymicrobial bacterial infection (but not control mice) become highly susceptible to and rapidly succumb to pulmonary A. fumigatus infection66. Recently, IL-7 therapy has been shown to increase the TH17 cell response and to decrease the number of deaths from secondary infection with Candida albicans130,131.

T cell exhaustion. T cell exhaustion was first described in mice suffering from chronic viral infections and is typified by T cells that have severely impaired effector functions132. Subsequently, T cell exhaustion has been shown in patients with bacterial and parasitic infections, HIV and cancer133. The prolonged duration of sepsis is characterized by high antigen load and increased levels of pro- and anti-inflammatory cytokines, which is an ideal setting for the development of T cell exhaustion. A recent study12 in which spleens were obtained rapidly after the death of patients with sepsis showed evidence that is highly consistent with the occurrence of T cell exhaustion, including: profound suppression of the production of IFNγ and TNF by stimulated T cells; increased expression of PD1 on CD4+ T cells and of programmed cell death 1 ligand 1 (PDL1) on macrophages; and decreased T cell expression of CD127 (the IL-7R α-chain), which is another phenotypic feature of exhausted T cells. Furthermore, this study showed that capillary endothelial cells and bronchial epithelial cells in patients with sepsis had increased PDL1 expression, which in this setting can impair the function of T cells that have migrated to the local area of infection, thereby seriously compromising the ability of the host to eradicate the organisms. A potential causal link between T cell exhaustion and morbidity and mortality in sepsis was provided by studies showing that increased expression of PD1 on circulating T cells from patients with sepsis correlated with decreased T cell proliferative capacity, increased nosocomial infections, and mortality134. Finally, animal studies show that inhibition of the PD1–PDL1 interaction improves survival in several clinically relevant models of sepsis, which is consistent with a key role for T cell exhaustion in the pathogenesis of sepsis (see below)135,136,137.

TReg cells. An increased percentage of circulating TReg cells has been described in patients suffering from septic shock. This increase was observed immediately after the onset of sepsis but persisted only in those patients who subsequently died138. These results were further extended by showing that this relative increase was due to a decrease in effector T cell numbers rather than changes in the absolute numbers of TReg cells119. This finding suggests that TReg cells are more resistant to sepsis-induced apoptosis, possibly because of increased expression of the anti-apoptotic protein BCL-2. Alarmins, including heat shock proteins and histones, are increased in sepsis, and as strong inducers of TReg cells they are also likely to contribute to increased TReg cell numbers in sepsis139. Since this initial work, many groups have confirmed an increase in TReg cell numbers in the blood or the spleen in various animal models of and in patients suffering from trauma and sepsis114,140,141.

Initial studies investigating the role of TReg cells in sepsis-induced immune dysfunction in mice were inconclusive, possibly owing to the use of CD25-specific antibodies, which lack specificity in inhibiting TReg cells142. Recent studies indicate that increased TReg cells are deleterious in sepsis and are associated with decreased effector T cell proliferation and function141. Importantly, these suppressive effects were totally abrogated by using small interfering RNA (siRNA) specific for FOXP3 to block the differentiation of TReg cells141. Other investigators used glucocorticoid-induced TNF-receptor-related protein (GITR; also known as TNFRSF18)-specific antibodies to block TReg cells, which led to improved immune function in sepsis and enhanced microbial killing143. In addition, sepsis-induced immunosuppression was shown to facilitate rapid growth of solid tumours via a TReg cell-mediated effect144.

TReg cells can also suppress innate immune cells. In the context of LPS stimulation, TReg cells inhibit both monocyte and neutrophil function145. Furthermore, TReg cells inhibit IFNγ production by γδ T cells in response to Mycobacterium tuberculosis and precipitate an NK cell-dependent endotoxin tolerance-like phenomenon that is characterized by decreased production of IFNγ and granulocyte–macrophage colony-stimulating factor (GM-CSF)113,146,147. In summary, there is extensive evidence that patients with sepsis and trauma have increased numbers of TReg cells, which, by acting both on innate and adaptive immune cells, impair immunity and contribute to nosocomial infections and mortality.

Immunotherapies in sepsis

Recombinant human IL-7. Considering the severe quantitative and qualitative alterations in T cells that are induced by sepsis, IL-7 has recently emerged as a promising therapeutic agent for septic patients. IL-7 is essential for T cell development and function148. Its effects are mediated via the heterodimeric IL-7R, which is composed of the IL-7R α-chain (CD127) and the common cytokine receptor γ-chain (γc; CD132). This receptor is expressed by most resting human T cells: naive and central memory T cells express the highest levels and TReg cells express the lowest levels. IL-7 upregulates expression of the anti-apoptotic molecule BCL-2, induces the proliferation of peripheral T cells and sustains increased numbers of circulating blood CD4+ and CD8+ T cells. In addition, IL-7 increases the diversity of the T cell receptor (TCR) repertoire, is associated with a reduction in the proportion of TReg cells in the circulation, rejuvenates exhausted T cells by decreasing PD1 expression and increases the expression of cell adhesion molecules, thereby facilitating the trafficking of T cells to sites of infection149,150 (Fig. 4). Recently, a massive loss in diversity of the TCR repertoire was observed in patients with sepsis151. Importantly, this low TCR diversity was also associated with increased risk of death and nosocomial infections. An optimal diversity of the TCR repertoire greatly facilitates an effective immune response against invading pathogens, and this might be one of the most important benefits of IL-7. Thus, beyond supporting the replenishment of lymphocytes, which are markedly depleted during sepsis, IL-7 improves multiple functional aspects of T cells that are considerably altered in sepsis.

Figure 4: Immunotherapy of sepsis.
figure 4

There are several immunotherapeutic agents that have shown promise in reversing the immunosuppressive phase of sepsis, including recombinant interleukin-7 (IL-7), programmed cell death 1 (PD1)- or programmed cell death 1 ligand 1 (PDL1)-specific antibodies, recombinant interferon-γ (IFNγ) and recombinant granulocyte–macrophage colony-stimulating factor (GM-CSF). GM-CSF and IFNγ primarily affect monocytes and macrophages to increase HLA-DR expression and to induce activation. IL-7 and PD1-specific antibodies have the advantage of targeting CD4+ and CD8+ T cells to restore the function of the adaptive immune system. By re-engaging CD4+ T cells, both recombinant IL-7 and PD1-specific antibodies will have effects not only on adaptive immune cells but also indirectly on monocytes and macrophages. PD1- and PDL1-specific antibodies will prevent and/or reverse T cell exhaustion. Thus, the net effects of these antibodies will be to prevent decreased IFNγ production, T cell apoptosis and decreased CD8+ T cell cytotoxicity. IL-7 will block T cell apoptosis, induce T cell proliferation, increase IFNγ production by T cells, increase T cell receptor (TCR) diversity and increase T cell trafficking by increasing the expression of cell integrins such as lymphocyte function-associated antigen 1 (LFA1). The application of immunotherapeutic agents will depend on the use of various biomarkers or tests of immune function to document that patients have entered the immunosuppressive phase of sepsis.

PowerPoint slide

In clinical trials, recombinant IL-7 has been used to treat patients with idiopathic lymphopenia and with lymphopenia-driven diseases, including patients who are infected with HIV and have persistently low lymphocyte numbers despite effective antiretroviral therapy152. Clinical studies in over 250 patients have shown that IL-7 is safe and well tolerated148. The use of IL-7 in clinical trials of sepsis is supported by data from animal models of sepsis131,150,153 (Box 1). In mice with polymicrobial sepsis due to peritonitis, treatment with IL-7 improved T cell viability, trafficking and IFNγ production, and restored the delayed-type hypersensitivity response to recall antigens150. Similar beneficial effects of IL-7 were observed in a murine fungal sepsis model that reproduces the delayed secondary infections that frequently affect patients in ICUs130. Importantly, in these mouse experiments, there was also a more than twofold improvement in survival. The ability of IL-7 to reverse the defects in lymphocyte function that occur in patients with sepsis has been tested in an in vitro model154. Ex vivo IL-7 treatment of T cells from patients with sepsis corrected sepsis-induced defects, including CD4+ and CD8+ T cell proliferation abnormalities, decreased IFNγ production, impaired phosphorylation of signal transducer and activator of transcription 5 (STAT5) and decreased BCL-2 levels154. These data indicate that the IL-7 signalling pathway remains fully operative during sepsis and that IL-7 reverses crucial sepsis-induced defects in T cell function.

PD1- and PDL1-specific antibodies. Another approach that holds great potential in reversing immunosuppression in sepsis involves blockade of the co-inhibitory molecules PD1 and PDL1 (Fig. 4). As discussed previously, PD1 and PDL1 are widely expressed on immune effector cells, endothelial cells and bronchial epithelial cells in patients with sepsis12. Blockade of PD1–PDL1 signalling improves survival in clinically relevant animal models of bacterial sepsis135,136,137 and markedly decreases mortality in primary and secondary fungal sepsis caused by C . albicans155. Thus, PD1 expressed by lymphocytes or PDL1 expressed by monocytes could function as potential biomarkers for the selection of patients with sepsis as candidates for therapy. A recent important study showed that in vitro blockade of PD1 improved IFNγ production and decreased apoptosis of T cells from patients with active M. tuberculosis infections156. A second major finding in this study was that when patients with tuberculosis were effectively treated, the number of PD1-expressing T cells decreased and inversely correlated with the IFNγ T cell response against M. tuberculosis. We believe that this work has major implications for the broader field of sepsis because of the similarities between active tuberculosis and protracted sepsis.

IFNγ. A key immunological defect in sepsis is decreased production of IFNγ, which is a cytokine that is essential for the activation of innate immunity (Fig. 4). Small clinical trials using recombinant IFNγ have been carried out in sepsis. Treatment with recombinant IFNγ reversed monocyte dysfunction in patients with sepsis whose monocytes had decreased HLA-DR expression and produced reduced amounts of TNF in response to LPS. Out of nine patients with severe sepsis who were treated with IFNγ to improve monocyte function, eight survived157. Most recently, IFNγ treatment in a patient with protracted Staphylococcus aureus sepsis caused an increase in monocyte HLA-DR expression and in the numbers of IL-17-expressing CD4+ T cells, and led to the eradication of the bacteria158. Furthermore, IFNγ is effective in treating fungal sepsis in patients with chronic granulomatous disease159. Finally, in a recent trial, patients with HIV who had cryptococcal meningitis and who were treated with IFNγ showed enhanced clearing of fungi from the cerebrospinal fluid compared to control individuals160. Thus, IFNγ might be effective if targeted to patients with sepsis who have entered the immunosuppressive phase, as identified by decreased monocyte HLA-DR expression and/or the development of fungal sepsis. A clinical trial of IFNγ is currently underway in the Netherlands in patients with sepsis who have been determined to have entered the immunosuppressive phase of the disorder ( number: NCT01649921).

Although IFNγ offers real promise as a potential immunotherapy in sepsis because of its ability to restore monocyte function, it will not correct the fundamental defect in T cells that is a major pathological abnormality in sepsis. In the authors' opinion, recombinant IL-7 and PD1-specific antibodies are more likely to be of benefit to patients with sepsis because of their favourable effects on the functions of CD4+ and CD8+ T cells: they both restore IFNγ production in sepsis and have many other beneficial effects on various other immune effector cells.

G-CSF and GM-CSF. Given the many defects in neutrophil function that occur in sepsis, investigators carried out two randomized clinical trials using recombinant granulocyte colony-stimulating factor (G-CSF), which is a drug that increases neutrophil numbers and function, in patients with sepsis resulting from hospital-acquired and community-acquired pneumonia80,81. G-CSF had already been shown to be highly beneficial in decreasing the incidence of sepsis in patients who have abnormally low absolute neutrophil numbers owing to chemotherapy or radiation therapy. Although neutrophil numbers are typically increased in sepsis, investigators postulated that administration of G-CSF to further increase neutrophil numbers might improve pathogen killing. Although white blood cell numbers increased in these patients, there was no effect on overall survival161,162. These two clinical studies indicate that it is unlikely that drugs that enhance neutrophil numbers or function will be of benefit in patients with non-neutropenia-associated sepsis.

Another immunotherapy that is being actively investigated in sepsis is the administration of GM-CSF, which is a cytokine that accelerates the production of neutrophils, monocytes and macrophages11 (Fig. 4). Treatment of ventilator-dependent patients with sepsis who had entered the immunosuppressive phase of the disorder, as identified by persistent decreases in monocyte HLA-DR expression, with recombinant GM-CSF, resulted in the restoration of HLA-DR expression, fewer days on the ventilator and fewer days in the ICU163. In addition, treatment of immunosuppressed paediatric patients with sepsis with recombinant GM-CSF restored TNF production and reduced the acquisition of nosocomial infections11.

Biomarker-guided therapy. A prerequisite for the application of immunotherapy in sepsis is the proper selection of patients. Various biomarkers should help in deciphering whether the patient is in the hyperinflammatory versus the hypoinflammatory phase of the disorder (Fig. 1). Indeed, immunotherapy could worsen the outcome by causing an over-exuberant inflammatory response if applied during the wrong phase of the disorder. Readers are referred to numerous recent reviews that discuss the use of various biomarkers in sepsis91,164,165. Box 2 provides a list of clinical and laboratory findings that could be used to identify immunosuppressed patients who might benefit from immunotherapies. Potential biomarkers include some currently available parameters, such as decreased monocyte HLA-DR expression and increased circulating IL-10 concentrations (both of which assess innate immune function and can be used to stratify patients for GM-CSF or IFNγ treatments), and decreased absolute CD4+ T cell numbers and increased percentage of TReg cells (both of which assess adaptive immunity and can be used to stratify patients for IL-7 therapy). Ideally, ex vivo functional testing of circulating immune cells remains the gold standard for the evaluation of immunity because it directly measures the capacity of cells to respond to pathogens.

Two hallmarks of sepsis-induced immunosuppression have been identified: the decreased capacity of monocytes to produce pro-inflammatory cytokines (mainly TNF) in response to endotoxin challenge; and decreased lymphocyte proliferation. Currently, these two characteristics are thought to provide a reasonable assessment of immune effector cell status but remain barely usable in routine clinical monitoring owing to methodological challenges, such as long incubation time, lengthy cell purification procedures and low standardization. A major challenge is to develop automated or standardized biomarkers that can be used in routine daily practice. More precise cellular monitoring might be possible through the use of innovative therapies, such as monoclonal antibody-based therapies targeting PD1- and CTLA4-expressing cells. Quantifying the mRNA for a panel of candidate genes in whole blood might enable clinicians to identify immunosuppressed patients and to provide the appropriate treatment. In the authors' opinion, the ideal method for identifying immunosuppressed patients would be a combination of cell phenotypic assays (for example, measuring HLA-DR or PD1 expression), functional assays (for example, measuring TNF production in whole blood) and genomic assays166.


The current understanding of the immune response to sepsis is controversial — some investigators suggest that sepsis-induced morbidity and mortality is the result of persistent immune activation with accompanying inflammation, whereas others assert that it is the result of immunosuppression. We support the idea that sepsis-induced immunosuppression with T cell exhaustion is a major abnormality. As discussed in this Review, sepsis induces numerous defects in host innate and adaptive immunity such that, if the invading organisms are not promptly eliminated, the host becomes more susceptible to intractable infection or to new secondary infections. Methods for identifying when patients have entered an immunosuppressive phase of sepsis and for detecting particular defects in immunity will enable the application of potent new immunotherapies that have shown great promise in animal models of sepsis and in early clinical studies of various infectious disorders. In particular, recombinant IL-7 and PD1-specific antibodies reverse fundamental immune defects in sepsis, lead to improved survival in multiple clinically relevant animal models of sepsis, are clinically well tolerated and have an ability to improve immunity in patients with cancer and chronic viral infections. Immunotherapy could therefore represent the next major advance in the treatment of infectious disease and sepsis.