Frontline Science: Defects in immune function in patients with sepsis are associated with PD-1 or PD-L1 expression and can be restored by antibodies targeting PD-1 or PD-L1

Defective neutrophil and monocyte function in septic patients correlates with PD-1 and PD-L1 expression, and can be restored by antibodies targeting PD-1 and PD-L1

Abstract

Sepsis is a heterogeneous syndrome comprising a highly diverse and dynamic mixture of hyperinflammatory and compensatory anti-inflammatory immune responses. This immune phenotypic diversity highlights the importance of proper patient selection for treatment with the immunomodulatory drugs that are entering clinical trials. To better understand the serial changes in immunity of critically ill patients and to evaluate the potential efficacy of blocking key inhibitory pathways in sepsis, we undertook a broad phenotypic and functional analysis of innate and acquired immunity in the same aliquot of blood from septic, critically ill nonseptic, and healthy donors. We also tested the ability of blocking the checkpoint inhibitors programmed death receptor-1 (PD-1) and its ligand (PD-L1) to restore the function of innate and acquired immune cells. Neutrophil and monocyte function (phagocytosis, CD163, cytokine expression) were progressively diminished as sepsis persisted. An increasing frequency in PD-L1⁺-suppressor phenotype neutrophils [low-density neutrophils (LDNs)] was also noted. PD-L1⁺ LDNs and defective neutrophil function correlated with disease severity, consistent with the potential importance of suppressive neutrophil populations in sepsis. Reduced neutrophil and monocyte function correlated both with their own PD-L1 expression and with PD-1 expression on CD8⁺ T cells and NK cells. Conversely, reduced CD8⁺ T cell and NK cell functions (IFN-γ production, granzyme B, and CD107a expression) correlated with elevated PD-L1⁺ LDNs. Importantly, addition of antibodies against PD-1 or PD-L1 restored function in neutrophil, monocyte, T cells, and NK cells, underlining the impact of the PD-1:PD-L1 axis in sepsis-immune suppression and the ability to treat multiple deficits with a single immunomodulatory agent.

Introduction

Sepsis is a heterogeneous syndrome that encompasses a gamut of immune responses occurring during the host’s response to a serious, life-threatening infection. The immune phenotype in sepsis ranges from proinflammatory (systemic inflammatory response syndrome) to anti-inflammatory, although these 2 extremes are not mutually exclusive, and a mixture of components of these phenotypes can be detected in most patients [1–7]. The various defects in host immunity that develop with prolonged sepsis have been well-characterized in both clinically relevant animal models [8, 9] and patients with sepsis [10]. Sepsis-induced immunosuppression involves elements of both innate and adaptive immunity and includes increased expression of checkpoint inhibitors, such as PD-1, PD-L1, B and T lymphocyte attenuator (BTLA), T cell membrane protein-3 (TIM-3), LAG3, and others [reviewed in 7, 11–13]. A previous publication from our group described elevated expression of PD-1 and PD-L1 on T cells, NK and NKT cells, and monocytes with a concomitant decrease in T and NK cell function from d 1–21 in patients with sepsis [14, 15]. Furthermore, several groups showed that the functional deficit could be restored by treatment in vitro with mAbs against either PD-1 or PD-L1, suggesting that inhibition of the axis via interaction with the receptor on the defective cell type or the ligand on the complementary cell type can reverse defective immune function [9, 13].

Reports of neutrophil persistence and reduced activity of phagocytes (neutrophils and monocytes) have suggested that there are functional deficits in the innate, first-line response during sepsis both in the mouse cecal ligation puncture model [9, 13, 16, 17] and in patients with sepsis [18–22]. Many mechanisms might contribute to inhibition of phagocytosis, including endotoxin tolerance, soluble mediators such as cytokines, and signaling through cell-associated receptors [9, 19, 23–26]. Importantly, identification of immune suppressive subsets of neutrophils and monocytes, such as MDSCs and G-MDSCs, also termed immature neutrophils, LDNs, or low-density granulocytes, have been characterized in sepsis [27, 28]. In patients with sepsis, both neutrophils and monocytes have been shown to express elevated levels of PD-L1 [14, 29, 30], and higher frequencies of monocytic MDSCs and G-MDSCs have been observed [18, 31–34]. MDSCs have been reported to inhibit T cell function, primarily in cancer models but also in murine sepsis models in a PD-1–dependent manner [13], whereas neutrophils and neutrophil subsets, such as LDN, were very recently reported to mediate inhibition of T cell function in patients with sepsis via direct cell–cell contact through PD-L1 [22, 29].

In this study, we evaluated multiple immune parameters of immunophenotype and dysfunction in T cells, NK cells, neutrophils, and monocytes throughout a prolonged septic ICU time course (up to d 21). In addition, because we had previously determined that T cell function could be restored by targeting either side of the PD-1:PD-L1 axis, we evaluated the ability of anti–PD-1 and anti–PD-L1 mAbs to restore the functions of neutrophils and monocytes. Data presented herein demonstrate 1) an increasing functional deficit in multiple innate and acquired immune responses in the same septic patient sample over time in the ICU, 2) an increase in suppressor phenotype cells associated with functional deficits, 3) correlation of functional deficits with expression of either PD-L1 or PD-1, and 4) restoration of innate and acquired cellular responses ex vivo by treatment with mAbs targeting either arm of the PD-1:PD-L1 axis.

MATERIALS AND METHODS

Study design and setting

This was a prospective observational trial performed at a 1200-bed, university-affiliated hospital in St. Louis, Missouri, between 2013 and 2015. Data collection and analysis were approved by the Human Research Protection Office at Washington University. Informed consent for participation was provided by all patients or their legally authorized representatives.

Patient selection

Septic patients.

Patients admitted to medical or surgical ICUs at Barnes Jewish Hospital, who were older than 18 y of age, and who fulfilled a consensus panel definition of sepsis [35] were included in the study. Sepsis was defined as the presence of systemic inflammatory response syndrome and a known or suspected source of infection. Patients who had undergone bone marrow irradiation or who had received chemotherapy or radiation therapy within the prior 6 mo, patients with HIV infection or viral hepatitis, and patients who were receiving immunosuppressive medications (except corticosteroids at a dose of <10 mg prednisone or equivalent/d) were excluded.

CINS patients.

Control subjects consisted of CINS patients admitted to the medical or surgical ICUs who were not suspected of having infection. Exclusion criteria were identical to that for patients with sepsis.

Healthy volunteers.

Healthy control subjects consisted of 25 healthy, MedImmune or AstraZeneca employees who were anonymously enrolled in the MedImmune, LLC, Research Specimen Collection Program. Donors consisted of 12 women and 13 men (age range, 29–54 y; median age, 41 y). Donors with HIV infection, hepatitis B or C virus, human T lymphotropic virus, or syphilis were excluded. Written consent for blood draws was obtained from each donor. All protocols and informed consent forms were approved by the Chesapeake Institutional Review Board.

Data collection

Patient clinical characteristics.

Baseline demographics included age, gender, APACHE II score, and SOFA score. Data were also collected regarding the requirement for mechanical ventilation, ICU length of stay, secondary infection, and mortality.

Patient hematologic values.

Absolute lymphocyte, absolute granulocyte, and absolute monocyte counts, obtained by the treating physicians as standards of care, were quantitated in the clinical laboratories at Barnes Jewish Hospital and extracted from the patients’ electronic medical records.

Definition of hospital-acquired secondary infections

Data on nosocomial infections occurring while patients were in the hospital were abstracted from medical records using standard Centers for Disease Control and Prevention case definitions (http://www.cdc.gov/hai/). Documentation of secondary infections was performed by an investigator who was blinded to patient stratification.

Blood collection and processing

Patients provided consent for a maximum of 4 blood samples obtained serially on d 1–3 (first blood draw [A]), d 4–7 (second blood draw [B]), d 8–12 (third blood draw [C]), and d 13–21 (fourth blood draw [D]) after sepsis onset. Because the number of C and D samples was limited (because of patient ICU discharge or death), data are presented pooled as “septic C and D.” A single blood draw was taken from nonseptic patients and healthy donors because nonseptic donors were typically discharged from ICU before the second draw. Heparinized blood was collected from septic and CINS patients, and a portion of the blood was processed in the laboratory at Washington University on the same day as the draw. The remainder was transported in polystyrene foam containers with a cold pack overnight and used in the laboratory at MedImmune the following day. Heparinized healthy-donor blood was collected onsite at MedImmune. It was stored overnight under similar conditions as that of the patient blood and was used in the laboratory at MedImmune the following day. Undiluted whole blood was aliquoted into 96-well, U-bottom plates for immunophenotyping or incubation with Abs for functional assays. Fifty to 100 μl/well of undiluted whole blood was incubated in a 37°C, 5% CO₂, incubator overnight with either isotype-control Ab, anti–PD-1 Ab, or anti–PD-L1 Ab (at a final concentration of 10 μg/ml). Isotype control human IgG, anti–PD-1 Abs, and anti–PD-L1 Abs were provided by MedImmune. Excess blood was centrifuged at 1800 rpm for 5 min at room temperature. Plasma was collected and stored at −80°C for subsequent analysis of soluble mediators.

Evaluation of LPS-induced TNF-α production in whole blood

Fifty microliters of whole blood was incubated for 4 h with 0.5 ml RPMI-1640 fortified with 10% FBS, Pen Strep (Thermo Fisher Scientific, Waltham, MA, USA), nonessential amino acids, plus or minus 0.5 ng/ml LPS (Enzo Life Sciences, Farmingdale, NY, USA). At the end of 4 h incubation, the tubes were centrifuged at 1000 relative centrifugal force for 5 min at room temperature. Supernatant (plasma) was harvested, aliquoted into fresh microcentrifuge tubes, and stored at −80°C for quantitation of TNF-α by ELISA.

TNF-α was quantitated using the human TNF-α CytoSet kit (Thermo Fisher Scientific) per manufacturer’s instructions and as previously described [15].

Flow cytometric staining of cells

All flow cytometric analysis and staining was performed on undiluted whole blood. Red cells were lysed using 1× Red Blood Cell Lysis Buffer (BioLegend, San Diego, CA, USA). FACS staining was then performed using standard flow-cytometric staining protocols. Abs for flow cytometric determinations were purchased from BioLegend, BD Biosciences (San Diego, CA, USA), R&D Systems (Minneapolis, MN, USA), or eBioscience (San Diego, CA, USA). Antibody cocktails were prepared by mixing Abs in panels designed to identify certain cell types. Cells blocked with 50 μl/well Human TruStain FcX (BioLegend) were incubated with the 50 μl/well Ab cocktail for 30 min at 4°C. Cells were then fixed with 4% formaldehyde (Electron Microscopy Sciences, Hatfield, PA, USA) for 15 min at 4°C, pelleted, resuspended in FACS buffer, and stored in the dark at 4°C for intracellular staining or acquisition.

For intracellular staining, fixed, surface-stained cells were permeabilized using 1× FACS permeabilization buffer (BioLegend) per manufacturer’s instructions. Ab cocktails were added to the cells in the permeabilization buffer for 20 min. Cells were then washed, resuspended in 150 μl/well 1% fixation buffer (BioLegend), and stored in the dark at 4°C for acquisition. All FACS acquisition was conducted using the LSR-II flow cytometer (Beckon Dickinson, San Jose, CA, USA). Data were then analyzed using FlowJo (version 10; Tree Star, Ashland, OR, USA) and GraphPad Prism 6 software (GraphPad Software, La Jolla, CA, USA).

Flow cytometric cell type identification and immunophenotypic analysis

Flow cytometric staining was performed as above. Cell types were identified, and markers were set for positive populations as shown in Supplemental Fig. 1A–C. Briefly, cellular expression of indicated markers to identify and phenotype cells was performed as follows: monocytes were identified by FSC and SSC properties and by CD14⁺CD11b⁺ immunostaining. Granulocytes were identified by FSC and SSC properties and by CD16⁺CD15⁺ immunostaining. LDNs (G-MDSCs) were further identified by gating on SSC^highCD16⁺CD15⁺CD33⁺CD66b^highCD114⁺CD11b^+/low cells based on previously described staining protocols [27, 34, 36–38] and were distinguished from MDSCs (or monocytic MDSCs) that were defined by a SSC^lowCD14⁺CD11b⁺ CD16⁻CD15⁺ phenotype based on previously described staining protocols [39–41]. Subsets of CD14⁺11b⁺ monocytes were further identified as proinflammatory by CD16⁺CD15⁻ staining [39, 42]. Immunophenotyping of monocytes and granulocytes was performed on unstimulated whole blood that was incubated overnight but without an isotype or anti–PD-1 or anti–PD-L1 Abs. In addition to markers to identify each cell type, cells were stained with Abs recognizing immune suppression and/or activation markers CD163, PD-1, PD-L1, MPO, TNF-α, and IL-10.

Lymphocytes were identified by FSC and SSC properties, as described previously [15]. T cell subsets were further identified by CD3⁺ and CD4⁺ or CD8⁺ immunostaining. NK cells were identified as CD3⁻CD56⁺, whereas NKT cells were identified as CD3⁺CD56⁺. Immunophenotyping was performed on unstimulated cells that were incubated overnight but without isotype or anti–PD-1 or anti–PD-L1 Abs. In addition to markers to identify each cell type, T cells, NK cells, and NKT cells were stained with Abs recognizing immune suppression and/or activation markers PD-1, PD-L1, IFN-γ, granzyme B, and CD107a.

Effects of anti–PD-1 and anti–PD-L1 on phagocytosis of pHrodo Red-labeled Escherichia coli particles

Undiluted whole blood (50 μl/well), which had undergone overnight incubation with either isotype-control Abs, anti–PD-1 Abs, or anti–PD-L1 Abs (5 μl/well for a final concentration of 10 μg/ml mAbs), was used directly in a phagocytosis assay without further stimulation. The phagocytosis assay was performed with 20 μl/well pHrodo Red E. coli Bioparticles kit (Thermo Fisher Scientific) per the manufacturer’s instructions. Cells were then immunostained with cell surface markers for identification of granulocytes and monocytes. The effect of anti–PD-1 and anti–PD-L1 Ab on phagocytosis was quantitated by flow cytometry.

Effects of anti–PD-1 and anti–PD-L1 on stimulated cytokine production and surface activation marker expression by monocytes and neutrophils

Undiluted whole blood, which had undergone overnight incubation with either an isotype-control Abs, anti–PD-1 Abs, or anti–PD-L1 Abs, was stimulated with LPS (L2654, 1 μg/ml in PBS; Sigma-Aldrich, St. Louis, MO, USA) plus 1× brefeldin A (BioLegend)/1× monensin (BioLegend) for 4 h, as previously described [43, 44]. Following stimulation, cells were immunostained with Abs to identify neutrophil, LDN, and monocyte subset cells, as indicated above. Cells were also stained with Ab recognizing surface CD163. Following surface staining, samples were fixed, permeabilized, and stained with fluorescently labeled anti–IL-10, anti-MPO, and anti–TNF-α Abs, as described above.

Effects of anti–PD-1 and anti–PD-L1 on stimulated cytokine production and surface activation marker expression by T and NK cells

Undiluted whole blood, which had undergone overnight incubation with either isotype-control Abs, anti–PD-1 Abs, or anti–PD-L1 Abs, was stimulated with 50 ng/ml PMA (Sigma-Aldrich) and1 μg/ml ionomycin (Sigma-Aldrich) plus 1× brefeldin A/1× monensin for 5 h, as previously described [43, 44]. Following stimulation, cells were immunostained with Abs to identify CD4 Th cells, CD8 cytotoxic T cells, and NK and NKT cells, as indicated above. Cells were also stained with Abs recognizing CD107a. Following surface staining, samples were fixed, permeabilized, and stained with fluorescently labeled anti–IFN-γ and anti–granzyme B Abs.

Statistical analysis

Data were analyzed with the statistical software GraphPad Prism 6. Clinical data are reported as median (IQR). Functional and phenotypic data are reported as means ± sem. For comparison of 2 groups, the independent-samples nonparametric t test (Mann-Whitney U) was employed. One-way ANOVA (Kruskal-Wallis) with Dunn’s multiple comparison tests were used to analyze data in which there were >2 groups. For correlation analyses, Spearman correlation was used along with the linear regression line for best-fit analysis. Significance was reported at P < 0.05.

RESULTS

Clinical and biologic parameters

The relevant clinical and laboratory data for the 17 septic and 9 CINS patients are presented in Table 1. Patients with sepsis had higher APACHE II and SOFA scores as well as longer ICU stays compared with CINS patients (Table 1). One patient with sepsis and one with CINS died by d 28 after ICU admission.

TABLE 1.

Characteristics of patients with sepsis and control CINS patients

Characteristic	Septic	Nonseptic
Patients (n)	17	9
Age (y)
Median	55	66
IQR	42–62	59–71
Gender
Male	10	5
Female	7	4
APACHE II score
Median	15	6
IQR	14–21	5–7
SOFA score
Median	7	2
IQR	5–9	2–3
ALC
Median	1.15	0.90
IQR	0.75–1.60	0.6–1.05
Serum creatinine (mg/dl)
Median	0.84	0.89
IQR	0.67–1.58	0.58–1.06
Stay in ICU (d)
Median	14	4
IQR	9 to 18	3 to 5
28-d mortality, N (%)
Survived	16 (94)	8 (89)
Expired	1 (6)	1 (11)
Hospital mortality, N (%)
Survived	16 (94)	8 (89)
Expired	1 (6)	1 (11)

ALC, absolute lymphocyte count.

Progressive loss of neutrophil and monocyte function in patients with protracted sepsis

We investigated the ability of whole blood SSC^highCD16⁺CD15⁺ neutrophils and SSC^lowCD14⁺CD11b⁺ monocytes to phagocytose labeled E. coli particles ex vivo, and, at the same time, a separate aliquot of whole blood was stained for expression of surface and intracellular activation markers, such as CD163 and TNF-α. Because, in some cases, CINS donors show an immunosuppressive phenotype similar to patients with sepsis [45], and some CINS donors progress to sepsis, both healthy and donors were included as controls for immune function and phenotype, serving as baseline, uninfected controls (healthy), and CINS trauma/injury/surgery, uninfected controls (CINS). As shown in Fig. 1, the ability of neutrophils (Fig. 1A) and monocytes (Fig. 1B) from patients with sepsis to phagocytose E. coli was significantly reduced compared with those from healthy and/or CINS donors, at both the level of percentage of cells that had engulfed the E. coli (percentage positive) and the amount of E. coli taken up quantified by MFI (means ± sem MFI summarized in Fig. 1 legend). Similarly, phagocytosis by monocytes was lower in patients with sepsis than it was in controls, particularly at later time points (MFI data not shown). Furthermore, phagocytic function of both neutrophils and monocytes from patients with sepsis appeared to decline with protracted stays in the ICU (data summarized in Fig. 1 legend, with individual donor progression shown in Fig. 1E and F).

Figure 1. Neutrophil and monocyte function during stays in the ICU.

The function of neutrophils and monocytes are shown during the time in the ICU. Function was determined by flow cytometric methods using whole blood. Phagocytic activity of neutrophils and monocytes in CINS, healthy, or septic donor whole blood was determined by measuring fluorescence emitted by ingested pHrodo Red-labeled E. coli particles (A and B). Surface expression of phagocyte maturation-associated marker CD163 was determined separately in whole blood following LPS stimulation (C and D). Comparison among CINS, healthy, and septic donor samples are shown as the percentage of positive cells. Septic donor sample data are presented as all time points together (Septic All) or separated by the time in the ICU: d 1–3 (Septic A), d 4–7 (Septic B), and d 8–12 and 13–21 (Septic C and D). Panels E to H show progrssion during the time in the ICU for individual septic donors for whom multiple, sequential blood draws were available. Data are presented as means ± sem. P values < 0.05 were considered significant. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, ns = not significant (P > 0.05), ND = not determined. Means ± sem of the MFI phagocytosis by neutrophils was 385.2 ± 55.14 (healthy, P = 0.096 compared with all septic), 367.3 ± 118.6 (CINS, P = 0.708 compared with all septic), 234.9 ± 38.41 (all septic), 330.3 ± 58.7 (Septic A, ICU d 1–3), 126.2 ± 39.6 (Septic B, ICU d 4–7, P = 0.0101 compared with Septic A), and 99.75 ± 33.1 (Septic C and D, ICU d 8–21, P = 0.0011 compared with Septic A). Means ± sem for the percentage of phagocytosis by neutrophils was 32.49 ± 5.52% (Septic A, ICU d 1–3), 14.99 ± 6.28% (Septic B, ICU d 4–7, P = 0.0180 compared with Septic A), and 9.99 ± 5.14% (Septic C and D, ICU d 8–21, P = 0.0151 compared with Septic A). Neutrophil means ± sem for the percentage of CD163 over time in the ICU was 52.29 ± 6.83% (Septic A), 14.66 ± 4.89% (Septic B, P = 0.0003 compared with Septic A), and 16.14 ± 7.60% (Septic C and D, P = 0.0148 compared with Septic A. Phagocytosis and CD163 expression by whole blood neutrophils and monocytes from patients with CINS compared with healthy donors: CINS 32.83 ± 6.77% neutrophil phagocytosis compared with healthy volunteers, 42.28 ± 4.62%, P = 0.253; CINS 49.60 ± 7.57% neutrophil CD163 compared with healthy controls, 51.84 ± 4.85%, P = 0. 985.

The expression of surface activation marker CD163 followed the same pattern as phagocytic activity in both neutrophils and monocytes (Fig. 1C and D), showing significantly reduced levels in patients with sepsis, which declined over time in the ICU (Fig. 1 legend, with individual donor progression shown in Fig. 1G and H). Phagocytosis and CD163 expression by whole blood neutrophils and monocytes from CINS patients were not significantly reduced compared with healthy donors; however, as shown in Fig. 1, a significant difference in phagocytosis and CD163 expression between CINS and septic donors was noted at the later time points (septic groups B and C and D, i.e., between ICU d 4–21).

In addition to the phagocytic function and CD163 expression, we also assessed TNF-α production as an additional marker of monocyte functionality. The present study, in agreement with previous studies [6, 19, 20, 46–48], showed that secreted TNF-α was less in LPS-stimulated whole blood from patients with sepsis compared with control CINS patients, as determined by ELISA, although statistical significance was not reached (Table 2). Healthy donors were not recruited for the ELISA quantitation of TNF-α conducted at Washington University; however, they were included in the flow cytometric analysis for intracellular TNF-α conducted at MedImmune. Interestingly, by the ELISA method, TNF-α secretion did not appear to be progressively reduced over time in the ICU. However, flow cytometric analysis of LPS-induced intracellular TNF-α levels (by MFI and the percentage positive) in individual cell populations from patients with sepsis revealed a trend toward reduced levels in both neutrophils and monocytes compared with healthy (by percentage positive and MFI) and CINS (by MFI) donors (Table 2). TNF-α levels in neutrophils and monocytes from patients with sepsis were lower than those of healthy donors at all sampling time points; however, that difference only reached significance by MFI (amount of TNF-α per cell), not by the percentage of positive neutrophils. Furthermore, monocyte, but not neutrophil, production of intracellular TNF-α did trend toward a steady decline over time in ICU, as determined by both percentage positive and MFI.

TABLE 2.

TNF-α in response to LPS in vitro

Sample source	Neutrophil (%)	Neutrophil (MFI)	Monocyte (%)	Monocyte (MFI)	Whole blood (pg/ml)
CINS (n = 6–17)	11.23 ± 2.75	210.3 ± 31.10	1.12 ± 0.23	24.85 ± 6.54	2248 ± 657
Healthy (n = 18–32)	37.03 ± 6.09 A*	197.1 ± 31.35	4.00 ± 1.36	17.01 ± 4.30 A*	ND
Septic (all, n = 19–36)	22.80 ± 4.71	97.13 ± 16.06 A*** B*	3.02 ± 0.68	10.72 ± 1.28 A*	1395 ± 320.6
Septic A (n = 11–22)	21.73 ± 6.21	84.81 ± 18.63 A** B*	4.08 ± 1.09	16.20 ± 6.18 A*	1372 ± 522.4
Septic B (n = 5–8)	25.91 ± 10.08	124.1 ± 43.24	1.12 ± 0.23	13.62 ± 9.99	1534 ± 487.6
Septic C and D (n = 3–6)	22.57 ± 12.05	116.5 ± 36.10	0.88 ± 0.44	8.48 ± 6.93	1243 ± 184.1

Neutrophil and monocyte expression of TNF-α in response to LPS stimulation of whole blood in vitro was determined by intracellular flow cytometry. Percentage of positive cells and MFI are shown. Comparisons among CINS, healthy, and septic donor samples are indicated with A (compared with CINS sample) and B (compared with healthy donor sample). Septic donor sample data are presented as all time points together (all) and separated by time in ICU: d 1–3, Septic A; d 4–7, Septic B; and d 8–12 and 13–21, Septic C and D. Secreted TNF-α (pg/ml) was determined separately in whole blood after 4 h stimulation with LPS. Data are presented as means ± sem. P values < 0.05 were considered significant. *P < 0.05, **P < 0.01, ***P < 0.001, ND, not determined.

Functional analysis of neutrophils and monocytes showed concordance of multiple functional parameters, such as phagocytic activity with CD163 expression, in neutrophils and TNF-α production in monocytes (Fig. 2). Similar statistically significant correlations were observed between neutrophil phagocytic function and CD163 expression, neutrophil production of TNF-α and MPO (representative data shown in Fig. 2A–C), and monocyte CD163 expression with monocyte phagocytosis, TNF-α, IL-10, and MPO production (representative data shown in Figure 2D–F). This is in agreement with data from a recent report by Santos et al. [48] showing correlation of monocyte activation (CD163 expression) with greater TNF-α production. Moreover, the patients with low neutrophil functions also had low monocyte functions (representative data shown in Figure 2G–I). Neutrophil and monocyte subsets can share similar common markers, such as CD16 and CD14; however, our gating strategy (shown in Supplemental Fig. 1A and B) precluded the possibility that the neutrophils and monocytes analyzed were from the same cell population; thus, our data suggest a multifaceted, innate immunity dysfunction in these patients.

Figure 2. Comparison of neutrophil and monocyte functions.

Function of neutrophils and monocytes were determined by flow cytometric methods using whole blood. Phagocytic activity of neutrophils and monocytes in whole blood was determined by measuring the fluorescence emitted by ingested pHrodo Red-labeled E. coli particles. Surface expression of phagocyte maturation-associated marker CD163 and intracellular expression of TNF-α, MPO, and IL-10 were determined separately in whole blood after LPS stimulation. Graphs presented show comparisons of 2 neutrophil (A–C) or 2 monocytes (D–F) functions and comparisons of neutrophil function with monocyte function (G–I) by the percentage of positive cells (A and D–H) or MFI (B–C and I). Correlation plots are shown with the linear regression line of fit. P values < 0.05 were considered significant. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, ns = not significant (P > 0.05).

Frequency of regular and LDN G-MDSC suppressor-phenotype neutrophils in patients with sepsis over time in the ICU

Different subsets of neutrophils and monocytes have been reported in different disease indications, which show varied immunologic functions and activation states [40, 42]. Because neutrophils and monocyte subset levels are reported to be elevated in patients with sepsis overall [18, 21, 31–34], we quantified levels of neutrophils and monocytes in our healthy donor, CINS, and septic patient samples by flow cytometry. Our results indicated that, in concordance with previously published data [18, 21], absolute granulocyte counts were significantly higher in septic donors compared with CINS donors and remained elevated over the time in the ICU (Fig. 3A); however, absolute monocyte levels from septic donors were not significantly different from CINS donors (Fig. 3B). Similar results were obtained by flow cytometric quantitation of SSC^highCD16⁺CD15⁺ neutrophils and total SSC^lowCD14⁺CD11b⁺ monocytes from septic, CINS, and healthy donors (data not shown). Briefly, flow cytometric analysis of whole blood SSC^highCD16⁺CD15⁺ neutrophils and SSC^lowCD14⁺CD11b⁺ monocytes confirmed the above findings of elevated neutrophil (P = 0.0078) but not monocyte counts (P > 0.05) in septic donors compared with CINS donors. In addition, using flow cytometric analysis, we determined the number of neutrophils in septic patient whole blood was significantly elevated compared with that of healthy donors (P < 0.0001); however, neutrophil numbers in CINS samples were not statistically different from that in healthy donor samples (P > 0.05, data not shown). Numbers of monocytes in both septic and CINS donors were lower than those of healthy donors (P = 0.0157 and P = 0.0106, respectively; data not shown).

Figure 3. Frequency of neutrophils, monocytes, LDNs, and MDSCs during time in the ICU.

Absolute granulocyte and monocyte counts were determined by standard hospital procedures. Populations of low-density, suppressor-phenotype neutrophils LDNs (SSC^highCD16⁺CD15⁺CD33⁺CD66b^highCD114⁺CD11b^+/low) and MDSCs (SSC^lowCD14⁺CD11b⁺CD16⁻CD15⁺) were determined by flow cytometric methods and sequential gating (as shown in Supplemental Fig. 1A and B, respectively) using unstimulated CINS, healthy, and septic donor whole blood. Absolute granulocyte and monocyte numbers (in thousands per cubic millimeter [K/mm³] (A and B) and the percentage of LDNs and MDSCs (C and D) are shown. Comparisons among CINS, healthy, and septic donor samples are shown. Septic donor-sample data are presented as all time points together (Septic All), or separated by time in the ICU: d 1–3 (Septic A), d 4–7 (Septic B), and d 8–12 and 13–21 (Septic C and D). Comparisons between septic donor samples at each time point in the ICU are also shown for LDNs and MDSCs (C and D). Data are presented as means ± sem. P values < 0.05 were considered significant. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, ns = not significant (P > 0.05), ND = not determined.

We further expanded this analysis to determine levels of suppressor phenotype neutrophil and monocyte subsets, LDNs, and MDSCs, via expression of select marker combinations, as described in the “Materials and Methods” section. We identified both LDN and MDSC populations in our donor samples. MDSCs have previously been found in patients with sepsis; however, this is the first report, to our knowledge, describing the presence of LDNs in septic and CINS patients. Interestingly, we found that, although sample numbers were limited at the later time points (septic B and C and D), the frequency of LDNs (percentage of neutrophils) in donors with sepsis appeared to trend toward an increase over time in the ICU (Fig. 3C), whereas MDSCs (CD14⁺11b⁺15⁺16⁻ as percentage of the total CD14⁺11b⁺ monocytes), in septic donors, they were significantly elevated at earlier time points (septic A, ICU d 1–3) compared with healthy donors, and remained at that level to d 21 in the ICU (Fig. 3D) However, additional samples at these late time points (ICU d 13–21, septic C and D) would be needed to further validate this preliminary observation. Frequencies of total monocytes (CD14⁺11b⁺), non-MDSC monocytes (CD14⁺11b⁺15⁻16⁻), and proinflammatory monocytes (CD14⁺11b⁺15⁻16⁺) in septic and CINS patients were not significantly different from those in healthy donors (data not shown).

PD-L1 and PD-1 expression on neutrophil and monocyte subsets over time in ICU

We evaluated classic, inhibitory checkpoint inhibitors and markers of immune suppression, PD-1, and PD-L1 on neutrophil and monocyte subsets. Although levels of suppressor phenotype cells (LDNs and MDSCs) trended toward an increase during the stay in the ICU, levels of checkpoint inhibitors PD-1 and PD-L1 did not. PD-L1 expression on neutrophils and LDNs from patients with sepsis was not significantly different from those of healthy donors at any time during the stay in the ICU by both MFI (Fig. 4A and B) and the percentage of positive cells (Fig. 4C and D).

Figure 4. PD-L1 expression on neutrophils and LDNs during time in the ICU.

Expression of PD-L1 on neutrophils and LDNs was determined by flow cytometric methods using unstimulated septic donor whole blood. The expression of PD-L1 is shown as MFI (A and B) and as the percentage of positive cells (C and D). Comparison among CINS, healthy, and septic donor samples are shown. Septic donor sample data are presented at all time points together (Septic All) or separated by time in the ICU: d 1–3 (Septic A), d 4–7 (Septic B), and d 8–12 and 13–21 (Septic C and D). Data are presented as means ± sem. P values < 0.05 were considered significant. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, ns = not significant (P > 0.05).

PD-L1 expression on monocytes was consistent with our previous data [14], whereby PD-L1 expression on CD14⁺11b⁺ monocytes was greater in septic than in CINS donors (data not shown). In contrast, extension of this analysis to look at monocyte subsets showed that, similar to LDN expression, PD-L1 expression on MDSCs from patients with sepsis was comparable to that of healthy and CINS donors (data not shown). In addition, expression of PD-1 on neutrophil and monocyte subsets from septic donors was not significantly different from those of CINS and healthy donors (data not shown). Further, we observed that expression of both PD-1 (Fig. 5A and B) and PD-L1 (Fig. 5C and D) was significantly greater on suppressor-phenotype subsets of both neutrophils (LDNs) and monocytes (MDSCs) from patients with sepsis (Fig. 5), CINS, and healthy donors (data not shown). This trend was also evident at different times in the ICU when data was separated by blood draw (septic A, B, and C and D); however, statistical significance was not reached because of the low number of samples per group (Supplemental Fig. 2A and B).

Figure 5. PD-1 and PD-L1 expression on neutrophils, monocytes, and suppressor phenotypes LDN and MDSC.

Expression levels of PD-1 (A and B) and PD-L1 (C and D) on neutrophil and monocyte subsets were determined by flow cytometric methods using unstimulated whole blood from septic donors at all time points of their stays in the ICU. Graphs presented show comparison of expression on neutrophils and LDNs and comparisons of the expression on monocytes and MDSCs by MFI. P values < 0.05 were considered significant. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Bowers et al. [37] previously published a report suggesting PD-L1 expression in HIV⁺ patients PBMCs was directly proportional to the levels of LDN [37]. Because both neutrophils and LDNs expressed PD-L1 in our study using whole blood and LDNs are a subset of total CD16⁺15⁺ neutrophils, we examined the correlation between PD-L1 expression in neutrophils and LDNs. A significant correlation was observed between the level of expression of PD-L1 on LDNs and neutrophils, as measured by both MFI of PD-L1 and percentage and number of PD-L1⁺ cells in our septic patient samples (Supplemental Fig. 2C). These data suggest that PD-L1 expression on neutrophils is likely predominantly on the suppressor phenotype LDN subset. A similar analysis comparing PD-L1 expression on monocytes did not show a significant correlation (data not shown).

Reduced neutrophil and monocyte function correlate with PD-L1 and PD-1 expression

Because we detected PD-L1 expression on both suppressor phenotype and total neutrophils and monocytes, we wanted to determine whether there was an association between the cell type expressing the inhibitor molecules and the function of the neutrophil and monocyte cells. As shown in Fig. 6, neutrophil phagocytosis was inversely related to PD-L1 expression on both total neutrophils and monocytes (Fig. 6A and B) but not the suppressor phenotype subsets, LDNs, and MDSCs (P > 0.05, data not shown). However, reduced monocyte phagocytic function only correlated with higher expression levels of PD-L1 on total monocytes (Fig. 6C) but not neutrophils (Fig. 6D), LDNs, or MDSCs (P > 0.05, data not shown). Similar correlations were found when data were analyzed as MFI phagocytosis vs. MFI PD-L1 expression (data not shown). In addition, as expected, given the correlation observed in this study between CD163 expression and phagocytosis (Fig. 2), CD163 expression showed the same inverse correlations with PD-L1 expression as shown for phagocytosis for both neutrophils and monocytes (data not shown).

Figure 6. Reduced neutrophil and monocyte function correlates with PD-L1 expression.

Expression of PD-L1 on neutrophils and monocytes was determined by flow cytometric methods using unstimulated whole blood from all time points of stays in the ICU. The function of neutrophils (A and B) and monocytes (C and D) were determined by flow cytometric methods using whole blood. Phagocytic activity was determined by measuring fluorescence emitted by ingested pHrodo Red-labeled E. coli particles. Correlation plots are shown with the linear regression line of fit. Graphs presented show comparisons of neutrophil and monocyte functions with PD-L1 expression on neutrophils and monocytes as the percentage of positive cells. P values < 0.05 were considered significant. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

As an extension of our findings suggesting the PD-L1 axis may have a role in neutrophil and monocyte function, we also assessed expression of its receptor PD-1 on neutrophils, monocytes, CD8 T cells, and NK cells in the same sample set. Interestingly, we were able to demonstrate a direct negative correlation of PD-1 expression levels on both CD8 T cells and NK cells (Fig. 7A–D) but not for neutrophils or monocytes (data not shown) with the phagocytic function of neutrophils and monocytes. Furthermore, analysis of phagocytic activity by neutrophils and monocytes from septic donors with CD8 T cell PD-1 expression (MFI) above the mean MFI of PD-1 on CD8 T cells from healthy donors also showed significantly reduced phagocytic activity compared with septic donors with mean MFI PD-1 on CD8 T cells below that of healthy donors (Fig. 7E and F). Similar results were obtained when the analysis was conducted using CINS control CD8 T cell PD-1 levels (data not shown). These results are consistent with our previous data showing patients who had greater expression of PD-1 on CD8 T cells than did CINS controls had more suppressed immune systems [14], which is consistent with the idea that PD-1 expression is elevated on CD8 T cells when immune suppression increases [49]. Taken together, these results are consistent with a role for the PD-1:PD-L1 axis in suppression of neutrophil and monocyte function in patients with sepsis.

Figure 7. Reduced neutrophil and monocyte functions correlate with PD-1 expression on CD8 T cells and NK cells.

PD-1 expression on CD3⁺8⁺ T cells and CD3⁻56⁺ NK cells was determined by flow cytometric methods using unstimulated whole blood. The functions of neutrophils and monocytes were determined separately by flow cytometric methods using whole blood. Phagocytic activity of the neutrophils and monocytes in whole blood was determined by measuring the fluorescence emitted by ingested pHrodo Red-labeled E. coli particles. Graphs presented show comparisons of neutrophil and monocyte functions with PD-1 expression on CD8 T cells and NK cells as MFIs. Correlation plots are shown with the linear regression line of fit (A–D). (E and F) Graphs compare phagocytic activity of neutrophils and monocytes in septic donor samples with PD-1 levels (MFI) on CD8⁺ T cells below or above the MFI PD-1 on healthy donor CD8 T cells (healthy donor mean = 17.71). P values < 0.05 were considered significant. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Reduced CD8 T cell and NK cell functions correlate with PD-L1 on LDNs but not on neutrophils or monocyte subsets

As we previously reported, T cell and NK cell functions were deficient in patients with sepsis compared with critically ill donors, and a steady reduction was observed over their stay in the ICU [14]. Because we aimed to correlate innate and acquired immune cell function and phenotype, based on our observation herein that PD-1 expression on CD8 T cells and NK cells was associated with reduced neutrophil and monocyte function, we were interested in determining whether PD-L1 expression on neutrophil or monocyte subsets might be associated with reduced CD8 or NK cell function. Therefore, we performed the reverse analysis and evaluated CD8 T and NK cell function in the context of neutrophil and monocyte phenotype in the same patient samples. The function of CD8 T cells and NK cells was determined by measuring intracellular IFN-γ, granzyme B, and surface expression of degranulation marker CD107a after stimulation with PMA/ionomycin. The function of CD8 T cells and NK cells (IFN-γ production) was negatively correlated with the expression level of PD-L1 on LDNs (Fig. 8) but not total neutrophils, monocytes, or MDSCs (data not shown). In addition, intracellular granzyme B production and surface expression of degranulation marker CD107a on CD8 T cells also showed a significant negative correlation with PD-L1 expression on LDNs (Fig. 8A–C) but not on the other cell types analyzed (data not shown). A similar correlation of higher LDN PD-L1 expression with lower NK cell intracellular IFN-γ and surface CD107a was also noted (Fig. 8D–F). These data were in contrast to our finding that reduced neutrophil and monocyte function correlated with PD-L1 expression on the nonsuppressor-phenotype neutrophils and monocytes (Fig. 6). Furthermore, analysis of CD8 T cell and NK cell activity from septic donors with LDN cell PD-L1 expression (MFI) above the mean MFI of PD-L1 on LDNs from healthy donors also showed significantly reduced activity compared with that of septic donors, with a mean MFI PD-L1 on LDNs below that of healthy donors (Supplemental Fig. 3).

Figure 8:. Reduced CD8⁺ cytotoxic T cell and NK cell functions correlate with PD-L1 expression on LDNs.

Expression of PD-L1 on LDNs was determined by flow cytometric methods using unstimulated whole blood from all time points of stays in the ICU. Markers of CD3⁺8⁺ T cell (A–C) and CD3^-56⁺ NK cell (D–F) functions were determined by flow cytometric methods using whole blood. Intracellular expression of IFN-γ and granzyme B and surface expression of degranulation marker CD107a were determined separately in whole blood after 5 h stimulation with PMA/ionomycin. Correlation plots are shown with the linear regression line of fit. Graphs presented show comparisons of CD8 T cell and NK cell functions as the percentage of positive cells with PD-L1 expression on LDNs as MFIs. P values < 0.05 were considered significant. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Anti–PD-L1 and anti–PD-1 mAbs can restore neutrophil and monocyte function

Given the reciprocal relationship between PD-1 and PD-L1 expression and neutrophil, monocyte, CD8 T cell, and NK cell function, we hypothesized that the functional deficit could be restored by both anti–PD-1– and anti–PD-L1–specific mAbs. Indeed, we observed that, in donor samples with reduced levels of phagocytic function, ex vivo incubation of whole blood with anti–PD-L1 mAb was able to increase phagocytosis of labeled E. coli by both neutrophils (Fig. 9A) and monocytes (Fig. 9B) compared with isotype-control mAb-treated samples. Similarly, anti–PD-1 mAb was also able to restore the deficient phagocytic function in both neutrophils and monocytes compared with isotype-control mAbs. The ability to increase phagocytic function was evident in all cohorts tested (CINS and septic donors at any time point in during their stays in ICU) but with a significant increase in mean group percentage of phagocytosis by neutrophils seen only in the septic patient group (28.62 ± 6.324% with isotype mAbs vs. 53.49 ± 5.817% with anti–PD-L1 mAbs, P = 0.0012; and 50.85 ± 6.163% with anti–PD-1 mAbs, P = 0.0034). As we previously demonstrated with mAb restoration of T cell function [14], there are mAb responders and nonresponders within each group. Similarly, within this study, we observed a mixed population of mAb responders and nonresponders for both neutrophil and monocyte function and with mAbs targeting either PD-1 or PD-L1. We also observed that increases in neutrophil and monocyte functions appeared to be more evident in subjects with the lowest baseline function, although this also occurred in subjects with intermediate function (e.g., as shown by the arrow in Fig. 9A). Consistent with the heterogeneity observed in patients with sepsis, restoration of function by mAbs did not appear to be restricted to any particular time point during stay in ICU in that 1) as in Fig. 1, low phagocytic activity was observed in each septic patient group (A, B, and C and D); and 2) mAb was able to restore function in donors from each of these groups (examples indicated by letters in the Fig. 9A lower-right panel). Additional studies are underway to determine correlation of response to mAb with immune phenotype. This is the first study, to our knowledge, demonstrating the ability of Abs targeting either side of the PD-1:PD-L1 axis to restore both neutrophil and monocyte functions. Additionally, in concordance with our previously published data [14], both anti–PD-L1 and anti–PD-1 mAbs were also able to restore defective CD8 and NK cell functions in this study sample set (data not shown).

Figure 9:. Phagocytic function is restored by anti-PD1 and anti-PDL1 Abs.

The function of neutrophils (A) and monocytes (B) was determined by flow cytometric methods after overnight incubation of whole blood with anti–PD-1, anti–PD-L1, or isotype control mAbs. Phagocytic activity was determined by measuring fluorescence emitted by ingested pHrodo Red-labeled E. coli particles. Graphs presented show comparison of phagocytic function of neutrophils and monocytes as the percentage of positive cells per individual donor with control mAbs, anti–PD-1 mAbs, or anti–PD-L1 mAbs. Arrows indicate an example of a subject with intermediate baseline function who is responding to a mAb. (A) Letters in the lower-right panel show time in the ICU (A, ICU d 1–3; B, ICU d 4–7; C, ICU d 8–12; and D, ICU d 13–21) of specific septic donors who responded to mAbs. The means per group are indicated by the circular symbols, and means ± sem values are shown for each treatment group. P values for pairwise comparisons of isotype vs. antibody treatment are shown. P values < 0.05 were considered significant. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

LDN suppressor neutrophils and neutrophil function correlate with disease severity

Altered functions and frequencies of different cell types were detected in patients with sepsis compared with CINS or healthy donors. In an attempt to elucidate the potential clinical relevance of these findings, we performed analyses comparing the function and phenotype with clinical parameters that were available to us. We found the percentage of LDNs, but not neutrophils (Supplemental Fig. 4A and B), monocytes, or MDSCs (P > 0.05, data not shown), was significantly correlated with higher APACHE II scores. Furthermore, higher PD-L1 expression on neutrophils and LDNs (but not monocyte subsets) trended toward association with higher SOFA scores (Supplemental Fig. 4C and D), corroborating the neutrophil PD-L1 data reported by Wang et al. [29]. As suggested by Bowers et al. [37] and as observed in this study, PD-L1⁺ neutrophils correlated with the PD-L1⁺ LDN subset (MFI and percentage of PD-L1⁺ neutrophils vs. PD-L1⁺ LDNs, Supplemental Fig. 2), thus these data may suggest the association with SOFA score may be related more to the LDN subset of the neutrophils than to all neutrophils. In addition, the patients with better neutrophil function had less-severe disease sequelae, in that, higher levels of phagocytic activity were observed in patients not requiring mechanical ventilation and in patients that didn’t succumb to secondary infection during their stay in the ICU (Supplemental Fig. 4E and F). Furthermore, greater CD163 expression was correlated with lower APACHE II and SOFA scores (P < 0.05, Supplemental Figure 4G and H). In contrast, monocyte function was not significantly correlated with these clinical parameters (data not shown). Moreover, CD8 T cell and NK cell function and expression of PD-1 did not correlate with any clinical parameter tested (data not shown).

DISCUSSION

Recent studies in patients with sepsis have shown a reduced function of neutrophils, including impaired activation and decreased phagocytic capability, on d 3–8 after sepsis onset [22]. However, the function and phenotype of neutrophils and monocytes has not been reported in patients with sepsis over a protracted period, with evaluation of both function and phenotype of the same cells in the same patient samples and in the context of the function and phenotypes of other immune cells. To address these 3 points in a single study, we undertook a broad, longitudinal, phenotypic, and functional analysis of both the innate and acquired immune response using aliquots of the same fresh whole blood sample obtained from CINS and septic donors over time in the ICU and in healthy controls. Similar to previous reports [17, 20, 22, 48], we observed a significant reduction in the ability of both neutrophils and monocytes from septic donors to phagocytose bacteria. Furthermore, we determined that this deficiency extended to other aspects of monocyte and neutrophil function, such that cytokine production and surface expression of CD163 were also reduced over time. In addition, correlations were observed among different functional parameters, such as cytokines, MPO, phagocytosis, and CD163 expression in neutrophils and monocytes, and between neutrophils and monocytes, as described above. These findings were not unexpected, given that CD163 is an innate bacterial sensor that binds gram-negative and gram-positive bacteria and a marker of activation of neutrophils and maturation of monocytes/macrophages into phagocytic cells [23, 24, 50–53]. What was both surprising and interesting was the observation that the dysfunction was correlated in our 2 nonoverlapping cell populations, neutrophils (gated on SSC^highCD16⁺CD15⁺) and monocytes (gated on SSC^lowCD14⁺CD11b⁺), in that lower phagocytic capacity in neutrophils was correlated with lower phagocytic capacity in monocytes and, similarly, reduced expression of CD163, TNF-α, and IL-10 (percentage positive: r² = 0.30, P = 0.015; MFI: r² = 0.34, P = 0.0033; data not shown) were correlated between the 2 cell types. These findings are suggestive of a broad immune dysfunction in patients with protracted sepsis.

Similar to the studies conducted in mice [16, 29], our data suggest immune suppression of phagocytes during sepsis. We showed, in this study, that some functional deficit in neutrophils and monocytes is evident at early and late time points during a stay in the ICU, and in fact, some functions decline with protracted sepsis. This finding, when viewed together with our previously published data [14] and similar current data describing declining T cell function (increased T cell exhaustion), suggests broader, cumulative immune suppression in patients with sepsis and protracted illness. What is not known is whether a hierarchical loss of function exists in neutrophils and monocytes with increasing exhaustion or immune suppression, as has been described for T cell exhaustion [49, 54].

T cell exhaustion has been extensively studied in numerous diseases, including sepsis, and many groups, including our own, have demonstrated a key immunomodulatory role for the checkpoint inhibitors PD-1 and PD-L1 by using Abs or small molecules to block PD-1:PD-L1 interaction [6, 7, 9, 12–14, 30, 37]. Recently, the focus has shifted to looking at the opposite arm of the PD-1:PD-L1 axis, that is, effects on the PD-L1-expressing APCs, neutrophils, monocytes, etc. and the interaction between these 2 sides of the axis. Anti–PD-L1 Abs have been shown to alleviate T cell exhaustion in HIV patient samples by interacting with PD-L1 on neutrophil subsets, such as the newly described suppressor-phenotype cells of granulocyte lineage, called G-MDSC or LDN [20, 26, 33, 34, 37, 38]. Because we were exploring multiple aspects of the PD-1:PD-L1 axis in immune suppression in patients with sepsis, we included analysis of LDNs and MDSCs in our exploratory panel. In contrast to previously published, similar studies [29, 33, 37, 38], in which cells are purified from whole blood before ex vivo analysis, and cryopreserved in some cases, which may result in altered phenotype and/or function with incubation times, we conducted all of our studies (functional assays and phenotyping) in whole blood to preserve the individual patient’s inflammatory milieu. This is important because studies show that removal of the cells from the septic milieu leads to reversal of the immune dysfunction [15].

Herein, we describe for the first time, the increasing presence of LDN during stays in the ICU in patients with sepsis. Our observation of elevated levels of suppressor phenotype neutrophils (LDNs) and monocytes (MDSCs) suggest greater potential for an immunosuppressive phenotype in septic than CINS patients, as previously described [14]. Furthermore, we showed greater PD-L1 expression on LDN and MDSC than their nonsuppressor counterparts, neutrophils, and monocytes, respectively. Importantly, we were able to show a correlation between the presence of LDNs and PD-L1 expression with clinical parameters such as APACHE II and SOFA scores, ventilator use, and secondary infections, suggesting that PD-L1⁺ LDNs may be an important mediator of disease. Similar to our findings above, MDSC phenotype granulocytes (neutrophil-MDSC) have been associated with more-severe disease in sepsis, in that, higher levels of interphase or neutrophil-MDSCs were observed in patients with septic shock vs. nonshock patients with sepsis [28, 33].

Overall, we observed that increased frequency of LDN suppressor neutrophils and reduced neutrophil function correlated with clinical parameters assessed in patients with sepsis. However, monocyte function, frequency of MDSCs, and expression of PD-L1 were not associated with disease severity. Furthermore, CD8⁺ T cell and NK cell function and PD-1 expression did not correlate with any clinical parameter tested in our septic patient cohort. Taken together, these data may suggest a greater impact of neutrophil function and PD-L1⁺ neutrophil subsets than monocytes, T cells, or NK cells on disease severity during sepsis. This further underscores the importance of neutrophils and neutrophil subsets in sepsis, both as cells that are involved in limiting disease and as cells that mediate immune suppression, respectively.

To gain insight into the potential biologic role of LDN and PD-L1 in sepsis, we were able to parse out our analyses to compared LDN and PD-L1 levels with multiple cellular functions. In concordance with previous reports, we found a significant correlation between presence of LDNs and reduced T cell function in our septic patient samples. Our studies are in agreement with those of Darcy et al. [33] describing neutrophil-MDSC in patients with sepsis that were able to suppress T cell proliferation [33]. Similarly, Janols et al. [34] showed patients with sepsis have increased MDSCs and polymorphonuclear cell-MDSCs, and the polymorphonuclear cell-MDSCs suppress T cell proliferation in vitro via a ROS-dependent mechanism. In our study, similar to data from Bowers et al. [37] in their HIV⁺ patient samples, we observed an association of PD-L1 expression on LDNs and reduced T cell and NK cell function. The present study is also consistent with work by Guérin et al. [28] who demonstrated that immature granulocytes were present in blood from patients with sepsis and had potent immunosuppressive properties on T cells [28]. Although a formal demonstration of the suppressive effect of neutrophils and LDNs on CD8 T cell and NK cell function by selective depletion of neutrophils, LDN, CD8, or NK cells from whole blood would have supported our observed correlation of PD-L1/PD-1 expression and diminished function in patients with sepsis, it was not possible to conduct such experiments in this study because of the very limited volume of blood available.

Given the comprehensive nature of our study of septic patient samples, we were also able to extend our analyses to demonstrate correlation of PD-L1 expression and/or LDNs and PD-L1⁺ LDNs with reduced T cell, NK cell, and neutrophil and monocyte activity, and in multiple functional parameters (IFN-γ, granzyme B, and CD107a expression for T cells and NK cells; phagocytosis, CD163, MPO, TNF-α, and IL-10 expression for neutrophils and monocytes). In the reverse analysis, we also noted that greater expression of PD-1 on T and NK cells correlated with reduced neutrophil and monocyte functions, suggesting bilateral inhibition via the PD-1:PD-L1 axis in patients with sepsis. Furthermore, reduced function of T cells, NK cells, neutrophils, and monocytes could all be restored by the addition of not only anti–PD-L1 mAbs but also anti–PD-1 mAbs. Although greater PD-L1 expression on neutrophils and monocytes correlated with reduced phagocytic functions in this study, expression of PD-L1 on these cells during the time in the ICU did not parallel the progressive functional decline observed. However, these data, taken together with our current data, described above, and our previously published data showing that PD-1 expression on CD8 T cells increased with protracted sepsis as PD-L1 decreased [14], highlights the importance of exploring both sides of the checkpoint inhibitor axes for a more-complete understanding of immune suppression in patients with sepsis.

Thus, we show the ability of Abs targeting either side of the PD-1:PD-L1 axis to restore innate immune function through neutrophils and monocytes and to bridge the innate and acquired immunity in patients with sepsis. These data suggest the PD-1:PD-L1 interaction contributes, at least in part, to the immune suppression demonstrated herein by reduced innate and acquired cell function in patients with sepsis, and blockade of this interaction, either through anti–PD-1 or anti–PD-L1, is sufficient to reinstate functionality of neutrophils, monocytes, and CD8 and NK cells.

One caveat remains: because these studies were conducted in whole blood containing mixed cell populations and soluble mediators, it is impossible to say which arm of this axis, or which cell type is more important, in that, whether, for example, anti–PD-L1 might prevent not only the negative signal sent to CD8 and NK cells via interaction of PD-L1 with PD-1 but also the inhibition of neutrophils or monocytes and vice versa to restore function. Purified, specific cell populations would have to be isolated (and/or mixed) to conclusively define this mechanism, and because removing them from their inflammatory or immune-suppressed milieu would inevitably result in the alteration of their native state and phenotype, this type of analysis, although mechanistically informative, might not be relevant for the in vivo situation and development of treatments based around this axis. Alternatively, selective depletion of neutrophils and/or LDNs from whole blood to formally demonstrate their suppressive effect on CD8 T cells and NK cells, and vice versa would be interesting; however, this was not possible in this study because of the limited volume of blood available.

Interestingly, in this study, mAbs against either PD-1 or PD-L1 were able to restore phagocytic function and CD163 expression but did not increase MPO production in neutrophils (data not shown). This is important because phagocytic function is essential for clearance of infection, and although MPO is important for killing of bacteria, other factors, such as neutrophil enzymes, ROS, etc., can also contribute to bacterial killing. However, accumulation of neutrophils in organs, the functional state of neutrophil priming or partial activation, and the extracellular release of neutrophil enzymes, such as MPO, have been implicated in mediating host cell damage and organ failure in mouse models of acute lung injury [55–58]. During severe sepsis, the accumulation and untoward activity of neutrophils may lead to organ dysfunction [59, 60]. In addition, acute lung injury in patients with sepsis is believed to arise from the extracellular release of oxygen radicals and proteolytic enzymes from infiltrating neutrophils [61]; thus, the effect of sepsis treatment with agents such as anti–PD-L1 or anti–PD-1 may be able to restore beneficial immune function without inducing the potentially deleterious responses mediated by neutrophils. Along these lines, Monaghan et al. [58] showed PD-1 knockout mice that were subjected to shock/septic challenge and induced indirect acute lung injury had reduced CD4:CD8 ratios, TNF-α levels, MPO activity, and caspase 3 levels in the lung; improved survival; and reduced lung pathology compared with wild-type control mice.

We hypothesized that suppression of neutrophil and monocyte function could potentially occur indirectly via inhibition of T or NK cell functions, such as cytokine secretion and down-regulation of signals through costimulatory molecules, mediated by neutrophil or monocyte PD-L1 engagement of T cell and NK cell PD-1. Conversely, although controversial, inhibition could potentially occur directly through negative signaling back into the neutrophils or monocytes through cell-associated receptors. To obtain some insight into these potential mechanisms of inhibition, we evaluated the relationship among PD-1 expression on T and NK cells and neutrophil and monocyte functions. Indeed, we were able to demonstrate a negative correlation between PD-1 expression levels on both CD8 T cells and NK cells with phagocytic function of neutrophils and monocytes, but not the function of T and NK cells with the function of neutrophils or monocytes (data not shown). Because reduced neutrophil and monocyte functions were also correlated with their own levels of PD-L1 expression, it may be possible that cross-linking of PD-L1 on these cells may send a negative signal back into the cell to shut down processes. The consensus thus far is that PD-L1 does not possess an intracellular signaling domain; however, Kim et al. [62] suggested PD-L1 (B7-H1) has a potential site for protein kinase phosphorylation in its intracellular domain and shares 16% identity with B7-2 in this region. Because other members of the B7 family—B7-2 (CD86) and another PD-1 ligand B7-DC (PD-L2)—can transmit reverse signals via cross-linking with mAbs [63, 64], it was postulated that PD-L1 might also be able to transmit signals through this region of the intracellular domain. The evidence that PD-L1 can reverse signal back into the cell that expresses it came from a single study by Kim et al. [62] that showed cross-linking of PD-L1 on EBV-transformed B cells using an anti–PD-L1 mAb or PD-1 Ig fusion protein induced ROS generation, mitochondrial disruption, release of apoptotic proteins from the mitochondria, and subsequent apoptosis in those cells [62]. PD-L1 stimulation also induced transcription and translation of the Fas ligand, activated phosphorylation of JNK and c-Jun and down-regulated ERK1/2 and p-Akt. Therefore, given the absence of other cells in this system, Kim et al. [62] provided direct evidence of reverse signaling through PD-L1. Although this concept has not been verified by other investigators, the possibility has been raised that the PD-1:PD-L1 interaction might also induce reverse signaling through PD-L1 to shut down neutrophils or monocyte processes.

To our knowledge, this type of broad, bilateral, multifunctional phenotypic analysis of the PD-1:PD-L1 axis has not previously been reported in sepsis or other indications. This observational study serves to highlight the multifaceted, far-reaching effect of this axis in immune suppression in sepsis and the ability to overcome immune suppression via blockade of either arm of the PD-1:PD-L1 axis, thus, suggesting blocking either PD-1 or PD-L1 could potentially have value in relieving immune suppression in multiple indications. In vitro addition of mAbs at saturating concentrations, as in these studies, does not allow us to distinguish the activity and efficacy between these 2 mAbs. Both were able to restore function in different cell types equally when tested in whole blood. Importantly, the function of the immune effector cell is a dynamic balance that is regulated in real time by competing signals directed by multiple positive and negative costimulatory molecules. Thus, anti–PD-1 and anti–PD-L1 Abs can act rapidly to restore immune effector cell function.

There are several limitations to this study. The median ages of the healthy donors (41 y), patients with sepsis (55 y), and patients with CINS (66 y), were different and may have contributed to some of the differences in immune function that were noted. A second limitation is that this study does not provide a detailed time course for appearance of PD-1 and PD-L1 expression on all the various immune cell phenotypes, such as CD4, CD8, and NK. It would be informative to have more data on the time of appearance of increased levels of PD-1 and PD-L1 on these various leukocyte subsets after sepsis onset.

In conclusion, there are several key findings and one possible, important implication of the current study. One major finding was the recognition of the extent of the defects in neutrophils and monocytes in patients with sepsis; sepsis induced impairments in both phagocytosis and the production of multiple key cytokines. Given the well-documented development of T cell dysfunction and T cell exhaustion occurring in patients with sepsis, these new findings underscore the broad extent of the impairment of host immunity in patients with sepsis. A second major observation in the current study was the increased frequency of PD-L1⁺-suppressor, phenotypic LDNs in patients with sepsis. Importantly, reduced neutrophil and monocyte function correlated with PD-L1 expression and was reversed by incubation with Abs against PD-1 or PD-L1. Thus, the ability of anti–PD-1 and anti–PD-L1 Ab immunotherapy to improve survival in multiple, clinically relevant animal models of sepsis and to restore T cell function in septic patient blood may be due not only to effects on T cells but also to restoration of neutrophil and monocyte functions. Finally, the current work supports the mounting evidence from multiple laboratories suggesting that a key element in the profound immune suppression in sepsis is due to the effects of immature granulocytes and G-MDSCs, which are increased in patients with sepsis [27, 28]. Targeting these newly described cells may provide a novel way to restore host immunity and improve survival in this lethal disease.

AUTHORSHIP

A.C.P. and R.S.H. designed the study and conducted data analysis. A.C.P., K.C., E.R.B., and D.O. performed the experiments. A.M.D. helped in patient recruitment and in data analysis. E.R.B. and D.O. helped in data analysis and in developing study methods. A.C.P., A.M.D., and R.S.H. authored and edited the manuscript.

Glossary

APACHE II

Acute Physiology and Chronic Health Evaluation II

CINS

critically ill nonseptic

FSC

forward light scatter

G-MDSC

granulocytic myeloid-derived suppressor cells

ICU

intensive care unit

IQR

interquartile range

LAG3

lymphocyte activating 3

LDN

low-density neutrophils

MDSC

myeloid-derived suppressor cells

MFI

mean fluorescence intensity

MPO

myeloperoxidase

PD-1

programmed death receptor 1

PD-L1

programmed death ligand 1

ROS

reactive oxygen species

SOFA

sequential organ failure assessment

SSC

side scatter of light

DISCLOSURES

Dr. A. C. Patera is an employee of MedImmune, the global biologics research and development arm of AstraZeneca, and owns stock in AstraZeneca. Dr. R. S. Hotchkiss is a consultant for Medimmune, LLC, and has research funding provided by MedImmune, Bristol-Meyers Squibb, and GlaxoSmithKline. The other authors declare no competing financial interests.