Background
Sepsis is a dysregulated host response to infection [
1], with concomitant immune activation and suppression. Sepsis-related immunosuppression contributes to poor outcomes by increasing the risk of nosocomial infection and death [
2‐
4]. A common feature of sepsis-related immunosuppression is impaired lymphocyte function, with increased expression of inhibitory checkpoint molecules, such as programmed-death 1 protein (PD-1) [
2]. PD-1 serves to limit excessive immune responses by negatively regulating lymphocyte activation and function, and promoting immune cell apoptosis. It has two known ligands: programmed death ligand-1 (PD-L1), which is widely expressed by a variety of immune and non-immune cell types; and programmed death ligand-2 (PD-L2), which is expressed by antigen-presenting cells [
2]. Increased expression of PD-1 and PD-L1 by T cells, monocytes and neutrophils has been demonstrated in sepsis, while upregulation of the PD-1 pathway is associated with higher mortality [
2,
5‐
8]. As this dysfunction is potentially reversible with anti-PD-1 or anti-PD-L1 monoclonal antibody treatment [
2,
5‐
8], manipulating the PD-1 pathway represents a potential target for sepsis trials.
Against this background, we hypothesized that lymphocyte surface PD-1, PD-L1 and PD-L2 expression by B and T cell subsets will vary by CD27 expression status. CD27 is a marker of lymphocyte activation; CD27 positive (CD27+) B cells correspond to memory B cells [
9], while CD27 negative (CD27-) T cells represent a high antigen-recall subset of memory T cells [
10]. Another rationale for assessing CD27-based memory lymphocyte subsets is the selective depletion of memory B cells in sepsis [
8]; it is not known whether PD-1 expression varies by lymphocyte memory status. We therefore measured PD-1, PD-L1 and PD-L2 surface expression on CD27+ and CD27- subsets of CD4+ T and B lymphocytes using flow cytometry in adult patients with sepsis on the intensive care unit (ICU). We compared expression between patients with sepsis and healthy controls, and between subgroups of patients with sepsis by nosocomial infection and survival status. PD-1 and PD-L1 also exist in a soluble form in serum; however, the relevance of these soluble forms to sepsis pathogenesis is unclear. We hypothesized that cell-surface PD-1, PD-L1 and PD-L2 expression would correlate with serum concentrations, and so we measured corresponding serum PD-1 and PD-L1 levels in the same samples.
Discussion
The novel findings from this pilot study include the first report of higher B cell expression of PD-1, PD-L1 and PD-L2 in sepsis, and differential expression of PD-1 by CD27 status in both B and CD4+ T cells. We also report results that are consistent with the published literature such as higher PD-1 and PD-L1 expression in CD4+ T cells in sepsis compared with controls [
7,
17], which gives external validity to our report. The overexpression of these checkpoint inhibitors in most patients with sepsis is consistent with the published literature suggesting this is a feature of sepsis-related immunosuppression.
PD-L2 has been less well-studied in sepsis than PD-L1, as PD-L1 is the more important binding partner for PD-1. The contribution of PD-L2 to the pathophysiology of sepsis remains unknown although increased PD-L2 expression by monocytes was reported in one observational study of patients with septic shock [
7]. A key finding of our present study in critically ill adult patients with sepsis was higher expression of PD-1/PD-L in the lymphocyte subsets associated with memory status, i.e. CD27+ B cells and CD27- CD4+ T cells. Memory lymphocytes are formed after encountering a specific pathogen, and are vital for generating rapid and effective immune responses upon future encounters with the same pathogen [
18]. We chose CD27 as a marker of memory status. CD27 expression is associated with activation in circulating B cells [
9]. CD27+ B cells are larger and exhibit greater and more rapid proliferation and immunoglobulin production in response to antigenic stimulation [
9]. In CD4+ T cells, loss of CD27 expression is seen in memory cells at a late stage in differentiation, and is associated with an increased capacity for IL-4 production [
19]. Functional studies of CD4+ T cells report that CD27 expression distinguishes two distinct subpopulations, of which the CD27- subset shows a stronger antigen-recall response and increased cytokine secretion [
10]. Relatively higher PD-1 expression on CD27- T cells may therefore have a greater negative effect on antigen-specific responses in both B cells and T cells, as there is T cell-dependent B cell development within the germinal centers of secondary lymphoid organs [
20].
We measured soluble PD-1 (sPD-1) and sPD-L1 levels, as high levels of sPD-1 or sPD-L1 in sepsis could reduce the efficacy of anti-PD-1 or anti-PD-L1 antibody therapy by neutralisation. Furthermore, should serum levels correlate with cell surface expression this may offer a potential point-of-care biomarker to identify patients who could benefit from early PD-1 pathway blockade. We did not find any significant differences in sPD-1 or sPD-L1 levels between patients with sepsis and controls; of note, levels were towards the lower limit of detection in the majority of subjects. Previous studies measuring serum sPD-1/L levels in sepsis have yielded inconsistent results [
21‐
23] (Additional file
17: Table S6). Importantly, concurrent cell surface expression was not measured in any of these studies. Timing of measurement may contribute to the differences; we measured sPD-1/PD-L1 within 12 h of ICU admission whereas the others varied from time of presentation to the emergency department [
21] to within 24 h of ICU admission [
22,
23]. Our pilot study results suggest that sPD-1 or sPD-L1 levels within 12 h of ICU admission do not identify patients with high cell surface PD-1/L expression [
11]. This needs confirmation in larger cohorts, ideally using the same inclusion criteria as those planned for interventional trials.
In contrast to one previous study [
7], we did not observe significant differences in PD-1/PD-L1 expression by either survival or nosocomial infection status. Aside from our small sample size, there are several alternative explanations. The kinetics of these checkpoint inhibitors is unknown in critically ill patients with sepsis. There is also a variable degree of immunosuppression even at the time of ICU admission though we specifically excluded any patients with previously documented immunosuppression. However, our timing of sampling within 12 h of ICU admission may be too early for differentiating survival status. This inference is supported by a recent study examining PD-1 expression by CD4+ T cells in a sepsis cohort using serial measurements on days 1, 3 and 7 of ICU admission, which found that while all septic patients had raised PD-1 at days 1 and 3, only survivors had normalised PD-1 expression by day 7 [
17]. This highlights the need for further work to characterize how PD-1/L expression changes over the course of sepsis, how this relates to outcome, and the optimal recruitment window for any future trial of anti-PD-1 therapy. With regard to nosocomial infection, we found that lymphocyte PD-L1 expression was significantly higher in those who subsequently developed an infection, but only when the ICU length of stay exceeded 7 days. This may be explained by the competing risk of nosocomial infection with early ICU discharge or death. The additional risk provided by over-expression of these checkpoint inhibitors may be overwhelmed by stronger risk factors for mortality such as age, comorbidity and illness severity [
14].
When interpreting our results, key limitations to consider include the small sample size, the use of healthy controls instead of non-sepsis critical illness controls, and that this was a post-hoc sub-study designed to test a hypothesis to inform future trials. We chose healthy controls as critically ill patients exhibit a range of immune deficits similar to those seen in patients with sepsis [
24]; the use of non-infected critically ill controls could confound the association between PD-1 expression and outcomes. The key strengths of the study include the hypothesis-driven set of experiments that highlight the need for further research to define PD-1, PD-L1 and PD-L2 expression in sepsis, and its relationship to two competing events - nosocomial infection and death [
11].