Sepsis is defined as life-threatening organ dysfunction due to a dysregulated host response to infection (1). Immunosuppression in part through T cell dysregulation is a critical component of continued pathology during sepsis and eventual mortality (2). Mechanisms by which patients become immunosuppressed during sepsis include upregulation of inhibitory molecules, expansion of the regulatory T cell population, and downregulation of CD28 and HLA-DR-mediated activation pathways (2, 3). CD28 is a T cell costimulatory receptor which, upon ligation with CD80/CD86, transmits a costimulatory signal important for T cell activation, proliferation, and pro-inflammatory responses (4), making it a potential target for sepsis immunomodulatory therapy. Furthermore, there is evidence of significantly decreased expression of CD28 and its stimulatory ligands, CD80/CD86, in human septic patients (2).

Given this possible role of CD28 in sepsis immunosuppression, multiple studies have found that modulation of CD28 and its ligands impact the emergence of immune dysregulation during sepsis, and thus impact mortality. One study found that deletion of CD80 or CD86 decreased inflammatory cytokine production during murine cecal ligation and puncture (CLP) and improved mortality (5). Similar findings were shown with combined blockade or deletion of CD80/86 in the same murine model of sepsis (6). Further evidence of interruption of the CD28/B7 complex through use of a B7-2 and CD28 dimer interface mimetic peptides prevented superantigen binding to the CD28/B7 complex and attenuated inflammatory cytokines (7). A peptide antagonist directed just to CD28 showed similar results and improved survival in a murine model of necrotizing soft-tissue infection (8). In contrast, another study showed antibody-mediated CD28 stimulation could be protective for intestinal immune responses during sepsis due to its role in T cell transendothelial migration (9).

Given these somewhat conflicting results, our group recently investigated the role of CD28 costimulatory signaling during sepsis through the use of an agonistic αCD28 Ab (10). Additionally, given that CD28 expression is altered on memory T cells vs. naïve T cells in both mice and humans, and given that laboratory mice possess significantly fewer memory T cells than adult humans (11), we utilized our novel “memory mouse” model in which mice receive specific pathogen inoculation to generate memory T cell populations similar in frequency to that of adult humans (12). This allowed us to also investigate the role of CD28 in a more physiologically relevant murine sepsis model. We found that in naïve septic mice, αCD28 Ab treatment resulted in worsened mortality, but in the immunologically experienced septic mice, a significant survival advantage was observed. This was associated with reduced T cell apoptosis in septic immunologically experienced mice but not in the naïve septic mice (10).

Here, we investigated the mechanisms underlying the survival advantage observed in immunologically experienced mice treated with agonistic anti-CD28. We found that CD8⁺ T cells, and not regulatory T cells or CD4⁺ conventional T cells (Tconv), are necessary for the effect. Depletion of CD8⁺ T cells increased both systemic inflammation and mortality during sepsis in the immunologically experienced host.

This study addressed the role of CD28 signaling during sepsis-induced immune dysregulation in immunologically experienced hosts. Results demonstrate a functional role of CD28 agonism during sepsis in immunologically experienced mice, which are likely more physiologically similar to immunologically experienced adult humans, as compared to naïve laboratory mice housed under specific pathogen-free conditions. More specifically, data indicate that memory CD8⁺ T cells following CD28 agonism during sepsis may function in an immunosuppressive manner, insofar as depletion of these cells leads to increased systemic inflammatory cytokines and a trend toward lower bacterial counts in both the blood and peritoneal fluid. This conclusion is also consistent with the observed increase in Foxp3⁺ CD8⁺ T cells (putative suppressor cells) under these conditions. The observation of both increased immunity (as measured by increased inflammatory cytokine levels and a trend toward decreased bacterial load) and increased mortality leads us to speculate that following the deletion of CD8⁺ T cells in immunologically experienced septic mice treated with agonistic anti-CD28, the mice are dying from an overexuberant inflammatory response, and not immune incompetence. It is also in line with our previously published data showing that the beneficial effect of CD28 agonism on sepsis survival in immunologically experienced hosts is dependent on IL-10 (10). Jointly with our previous findings, this current study indicates that CD28 agonism in immunologically experienced mice promotes the generation of an immunosuppressive environment that quells systemic inflammation and is beneficial for survival during sepsis.

As discussed above, reports regarding the role of CD28 signaling in sepsis have been somewhat contradictory. However, the preponderance of studies analyzing the role of CD28 and/or its ligands CD80/86 in models in which the recipients used are immunologically naïve have demonstrated that deleting or antagonizing CD28 signaling results in improved sepsis mortality. Importantly, in keeping in line with these studies, we recently showed that CD28 agonism was moderately deleterious in naïve septic hosts (10). In contrast, CD28 agonism conferred a significant survival benefit when administered to immunologically experienced septic hosts. Of note, this increased survival was associated with increased IL-10 production from CD4⁺ Foxp3⁺ Treg, and systemic blockade of IL-10 abrogated the protective effect of the CD28 agonist in septic immunologically experienced mice (10). Thus, in the previously published study we surmised that CD4⁺ Foxp3⁺ IL-10-producing cells could be mechanistically responsible for the observed increase in sepsis survival in CD28 agonist-treated animals (10). However, here we show that depletion of either CD4⁺ T cells or CD25⁺ T cells, widely used as a strategy to deplete CD25⁺ CD4⁺ Foxp3⁺ Treg, did not reverse the survival benefit of CD28 agonist in immunologically experienced septic mice. Instead, we observed that depletion of CD8⁺ T cells removed the survival benefit conferred by the CD28 agonist in immunologically experienced septic mice. Owing to the observation that CD8⁺ T cells in CD28 agonist-treated immunologically experienced septic mice exhibited increased IL-10 secretion and our published finding that IL-10 neutralization erases the benefit of CD28 agonism in immunologically experienced septic mice (10), we conclude that the immunoregulatory effect of IL-10-producing CD8⁺ Foxp3⁺ T cells contributes to the salutary effect of CD28 agonism during sepsis in immunologically experienced hosts.

Immunosuppressive Foxp3⁺ CD8⁺ T cells have been described in other studies. Conditions that result in the generation of suppressive CD8⁺ T cells (in vivo or in vitro) include continuous antigen stimulation with staphylococcal enterotoxin B (15), administration of anti-CD3 mAb (18), stimulation using autologous LPS-activated dendritic cells (19), and priming naïve CD8⁺ T cells with CD40 ligand-activated plasmacytoid dendritic cells (20). Most relevant to our study, suppressive CD8⁺ T cells have also been shown to be induced through increasing CD28 costimulatory signaling in the presence of TGF-β (16). Because TGF-β is significantly upregulated during sepsis (21), our results are consistent with the hypothesis that CD28 agonism during sepsis directly increases a population of immunosuppressive CD8⁺ Foxp3⁺ T cells (15–17). This hypothesis of a direct effect of CD28 signaling is consistent with our data showing that a subset of CD8⁺ Foxp3⁺ Treg are CD28⁺ under these conditions. However, many previous studies report suppressive CD8⁺ T cells to be CD28 negative (22, 23). It is also possible that the precursor to Foxp3⁺ CD8⁺ T cells may begin as CD28⁺ cells which are modified by CD28 signaling, but subsequently downregulate CD28 and upregulate Foxp3. Anti-CD28 antibody could also be impacting CD8⁺ Foxp3⁺ T cells indirectly by freeing up more CD80/CD86 for CTLA-4 ligation and signaling. This hypothesis is supported by our findings of a significant increase in CTLA-4 expression on CD8⁺ Foxp3⁺ following administration of the anti-CD28 mAb in septic immunologically experienced mice. These possibilities warrant further investigation. Of note, CTLA-4 expression was not different on bulk CD4⁺ or CD8⁺ T cells, nor on CD44^hi CD8⁺ or CD44^lo CD8⁺ T cells when comparing cells isolated from vehicle control-treated and anti-CD28-treated septic immunologically experienced mice (data not shown).

CD28 agonism during sepsis also led to increased IL-2 production from bulk CD8⁺ T cells, driven mainly by CD44^hiCD8⁺ T cells ( Figure 4 ). IL-2 is essential for generation, survival, and functional activity of Foxp3⁺ regulatory T cells (24), further supportive evidence that CD28 agonism could promote an immunosuppressive environment. The significance of the observed increase in IL-10 secretion from Foxp3⁺ CD8⁺ T cells during sepsis is based on studies showing that IL-10 can be both beneficial (25–27) and detrimental (28, 29) to survival during sepsis. In the setting of CD28 agonism, we previously observed that neutralization of IL-10 significantly worsened mortality in septic immunologically experienced mice (10), evidence to support the beneficial effect of immunosuppression observed in immunologically experienced mice with CD28 agonism.

While here we show that CD28 agonism promoted IL-10 production in CD8⁺ Foxp3⁺ T cells, we have previously shown that CD28 agonism promoted the proliferation and IL-10 production of traditional CD4⁺ Foxp3⁺ T cells, and functionally increased their suppressive capacity, in the context of immunologically experienced but not naive recipients (10). The mechanisms underlying these related findings are not clear, but could be related to differential effects of CD28 agonism on the recently-described memory Treg subset (30). Following exposure to antigen in the periphery, thymus-derived Treg become activated, proliferate and differentiate into more potent suppressors (30). Following the resolution of inflammation, a subset of these activated Treg can persist and function to suppress subsequent immune responses in a more accelerated, recall fashion when antigen is re-encountered. Extensive phenotypic and functional analysis of memory Treg has been hampered by a lack of specific phenotypic markers to define these cells in both human and mouse. Thus, further investigation is needed to dissect the role of CD28 signaling on memory Treg cells during sepsis. However, based on the results presented here as well as our previously published work, it is tempting to speculate that CD28 signaling may differentially impact naive and memory Treg cells, in a manner analogous to the differential requirements for CD28 costimulation on naive vs. memory conventional T cells (40, 41). One limitation of our study is that differences in the expression and regulation of CD28 on distinct T cell subsets in mice vs. humans may exist. Indeed, the results of the TGN141 study (31), in which agonistic anti-CD28 was administered to five healthy human volunteers who experienced catastrophic cytokine storm and which was not predicted from studies in animal (canine) models (32), may highlight differences in CD28 regulation or signaling under these conditions. Additional studies are therefore required to understand the role of CD28 on memory and regulatory T cell subsets in humans, both in health and in the context of immune dysregulation.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was approved by Emory University Institutional Animal Care and Use Committee. The study was conducted in accordance with the local legislation and institutional requirements.

MF: Writing – review & editing, Writing – original draft, Visualization, Supervision, Project administration, Funding acquisition, Formal analysis, Data curation, Conceptualization. JA: Writing – review & editing, Writing – original draft, Visualization, Investigation, Formal analysis, Data curation. YS: Writing – review & editing, Visualization, Investigation, Formal analysis, Data curation. CD: Writing – review & editing, Visualization, Investigation, Formal analysis, Data curation. MW: Writing – review & editing, Investigation. ZL: Writing – review & editing, Investigation. EB: Writing – review & editing, Investigation, Formal analysis. CC: Writing – review & editing, Supervision, Project administration, Funding acquisition, Conceptualization.