a b s t r a c t

Objective: To evaluate the efficacy of soluble programmed death-1 (sPD-1) for risk stratification and prediction of 28-day mortality in patients with sepsis, we compared serum sPD-1 with procalcitonin , C-reactive protein (CRP), and the Mortality in Emergency Department Sepsis (MEDS) score.

Methods: A total of 60 healthy volunteers and 595 emergency department (ED) patients were recruited for this prospective cohort study. According to the severity of their condition on ED arrival, the patients were allocated to the systemic inflammatory response syndrome group (130 cases), sepsis group (276 cases), severe sepsis group (121 cases), and septic shock group (68 cases). In addition, all patients with sepsis were also divided into the survivor group (349 cases) and nonsurvivor group (116 cases) according to the 28-day outcomes.

Results: When the Severity of sepsis increased, the levels of sPD-1 gradually increased. The levels of sPD-1, PCT, CRP and the MEDS score were also higher in the nonsurvivor group compared to the survivor group. Logistic re- gression suggested that sPD-1, PCT, and the MEDS score were independent risk factors for 28-day mortality of patients with sepsis. Area under the curve (AUC) of sPD-1, PCT and the MEDS score for 28-day mortality was 0.725, 0.693, and 0.767, respectively, and the AUC was improved when all 3 factors were combined (0.843).

Conclusion: Serum sPD-1 is positively correlated with the severity of sepsis, and it is valuable for risk stratification of patients and prediction of 28-day mortality. Combining sPD-1 with PCT and the MEDS score improves the prognostic evaluation.

(C) 2017

Introduction

Sepsis is a common life-threatening disease, and severe immuno- suppression plays an important role in the deterioration of patients with sepsis [1]. Negative co-stimulatory molecules can negatively regu- late cell proliferation, differentiation, and apoptosis, and play an impor- tant role in the immune function of patients with sepsis. As a member of the CD28 superfamily, programmed death-1 (PD-1) mediates a nega- tive co-stimulatory signal [2]. In addition, PD-1 and programmed death ligand-1 (PD-L1) can effectively inhibit the function of T and B cells and suppress proliferation of T cells, thus playing an important part in immune regulation [3-5]. It has been reported that membrane- bound PD-1/PD-L1 plays an important part in immune suppression dur- ing sepsis [6]. Many co-stimulatory factors such as B7-H3 and cytotoxic T lymphocyte-associated antigen-4 (CTLA-4) exist in both membrane- bound and soluble forms [7,8]. Similar to the membrane-bound mole- cules, the soluble molecules also have their corresponding biological

* Corresponding author at: Emergency Department of Beijing Chaoyang Hospital, 8# Worker’s Stadium South Road, Chaoyang District, Beijing 100020, China.

E-mail address: lcscyyy@163.com (C. Li).

function-they circulate in the blood stream like cytokines and play their part in the immune response. Two forms of PD-1 have been found: membrane-bound and soluble forms. Compared to membrane- bound PD-1, soluble PD-1 (sPD-1) lacks the third exon of PD-1 that en- codes the transmembrane region and has its own biological function [9]. In recent years, sPD-1 has become the focus of increasing attention.

It is currently believed that sPD-1 can promote T-cell responses through blocking PD-1/PD-L1 pathway. Excessive sPD-1 can specifically block the PD-1/PD-L1 signaling pathway, leading to immune imbalance, so that auto-reactive T cells cannot be effectively inhibited or cleared, which results in pathological immune damage [9]. The level of sPD-1 is low in healthy individuals but is highly expressed in the serum of pa- tients with rheumatoid arthritis (RA) [10], chronic hepatitis [11], and aplastic anemia [12]. Studies have shown that sPD-1 is also involved in anti-tumor [13] and anti-virus immunity [14]; however, the correla- tion between sPD-1 and sepsis has rarely been reported. Our research group has previously demonstrated the correlation between the expres- sion of membrane-bound PD-1/PD-L1 and the prognosis of sepsis [15]. Thus, we speculate that sPD-1 in serum may also be associated with risk stratification and prognosis of patients with sepsis. In this study, we measured sPD-1 levels in the peripheral blood of patients in the

http://dx.doi.org/10.1016/j.ajem.2017.07.002

0735-6757/(C) 2017

emergency department (ED) with different degrees of sepsis severity, and we evaluated the usefulness of sPD-1 for risk stratification and pre- diction of 28-day mortality in patients with sepsis. We hoped that PD-1 could be used as an immunological marker for early assessment of the severity of sepsis and for monitoring Disease progression.

Materials and methods

Study subjects

This prospective cohort study was conducted in the ED of an urban, tertiary care, teaching hospital with an average number of 260,000 ED visits per year. Consecutive patients who visited the ED with suspected infection and had 2 or more criteria of Systemic Inflammatory Response Syndrome were enrolled from January 2015 to December 2015. Exclusion criteria were as follows: being b 18 years of age; presenting with surgical trauma, blood diseases, autoimmune diseases, HIV infec- tion, liver disease (for example, hepatitis, cirrhosis), end-stage renal dis- ease (requiring dialysis), tumors, and pregnancy; and receiving hormone therapy. The definitions of SIRS, sepsis, severe sepsis, and sep- tic shock were consistent with the diagnostic criteria established by the 2001 International Sepsis Definitions Conference [16]. For patients who had been admitted to the ED twice or more during this study, only data obtained from the first admission were used. Patients included may have one or more previous comorbidities, including chronic obstructive pulmonary disease (COPD), cardiovascular disease (coronary heart dis- ease and/or heart failure), chronic kidney disease (excluding those re- quiring dialysis), diabetes mellitus, and cerebrovascular disease. First, according to the patients’ condition on ED arrival, all patients were allo- cated to a SIRS group (suspected infection on ED arrival, but eventually confirmed no infection), a sepsis group, a severe sepsis group, or a septic shock group. Second, all patients with sepsis (including sepsis, severe sepsis, and septic shock) were followed up for 28 days, and the survivor group and nonsurvivor group were identified according to the 28-day outcomes. Those discharged earlier than 28 days were followed up by telephone to define outcomes. This study was in line with medical ethics standards approved by the Ethics Committee of Beijing Chaoyang Hos- pital, and it was carried out with informed consent from the patients or their families.

Data collection

On arrival at the ED, the following patient information was recorded: age; gender; address; telephone number; vital signs (heart rate, blood pressure, respiratory rate, oxygen saturation, and temperature); altered conscious state; nursing-home resident status; comorbidities; medical history; routine blood data; white blood Cell differential count; blood gas analysis; Liver and kidney function test; coagulation function test; procalcitonin levels; C-reactive protein levels; chest X- ray; and bacteriological tests of sputum, blood, or urine. In addition, the Mortality in Emergency Department Sepsis (MEDS) score was calcu- lated by summing the points of 9 variables: terminal illness, age N 65 years, bands N 5%, tachypnea or hypoxia, septic shock, platelet count, nursing-home resident, lower respiratory tract infection, and al- tered mental state [17]. Peripheral blood serum was collected immedi- ately after the patient’s arrival in the ED, and stored at -80 ?C for the determination of sPD-1 level.

Measurements of sPD-1, PCT, and CRP levels

The level of sPD-1 was measured using an enzyme-linked immuno- sorbent assay kit (DuoSet Human PD-1, R&D systems, Minneapolis, MN, USA) on an automatic microplate reader (Multiskan MK3; Thermo, West Palm Beach, FL). The PCT level was measured using an enzyme- linked fluorescence immunoassay with a miniVIDAS immunoassay ana- lyzer (BioMerieux, Durham, NC, USA). The CRP level was measured

using the QuickRead CRP kit (Orion Diagnostica Oy, Espoo, Finland) by following the rapid immune turbidity method.

Statistical analysis

Data were analyzed by SPSS 17.0 (SPSS Inc., Chicago, IL, USA). Age, score, and test results were all non-normally distributed. They were presented as medians with interquartile ranges [M (Q_L,Q_U)]. The Kruskal-Wallis test was applied for multigroup comparisons, and the Mann-Whitney U test was applied for comparison between 2 groups. Categorical clinical variables were tested using the chi-square test. Lo- gistic regression was used to determine independent risk factors of 28-day mortality in patients with sepsis. New combinations of the MEDS score and relevant biomarker(s) were constructed by logistic re- gression model for conversion with the prediction probability (P) as the analysis indicator; the degree of fitness of the model was tested by the Hosmer-Lemeshow goodness-of-Fit test. Receiver operating character- istic (ROC) curves were used to compare the Predictive capacity of rel- evant variables, and the cutoff value of each variable was determined by the Youden index. The sensitivity, specificity, positive predictive values, and the negative predictive value were then calculated. The area under the ROC curve (AUC) was compared using MedCalc 11.6 sta- tistical software (MedCalc Software, Ostend, Belgium). P b 0.05 indicates a statistically significant difference.

Results

Characteristics of the patients

A total of 595 patients and 60 healthy volunteers were enrolled in this study. There were 5 groups, 60 cases in the healthy control group, 130 cases in the SIRS group, 276 cases in the sepsis group, 121 cases in the severe sepsis group, and 68 cases in the septic shock group. white blood cell count and platelet count were significantly different in the multigroup comparison among the 5 groups (P b 0.001), but WBC count showed no difference between the severe sepsis group and the septic shock group (P = 0.848), and platelet count showed no significant difference between the SIRS group and the healthy control group (P = 0.051). There were no significant differences in age, gender, and comorbidities among the 5 groups (P N 0.05). Among the 4 groups of SIRS, sepsis, severe sepsis, and septic shock, the CRP and PCT levels, MEDS score, and 28-day mortality were significantly different and the increase was higher according to the severity of sepsis (P b 0.05) (Table 1).

Levels of sPD-1 among the 5 groups

Levels of sPD-1 showed a significant difference among the 5 groups (P b 0.001). However, there was no significant difference in levels of sPD-1 between the SIRS group and the healthy control group (P = 0.462). The level of sPD-1 was higher in the sepsis group than in the SIRS and healthy control groups (P b 0.05), and sPD-1 levels of the sep- sis, severe sepsis, and septic shock groups showed a significant differ- ence in pairwise comparison (P b 0.001). As the severity of sepsis increased, the levels of sPD-1 gradually increased (Fig. 1).

Comparison between the survivor group and the nonsurvivor group

According to the 28-day outcomes, all the patients with sepsis (in- cluding sepsis, severe sepsis, and septic shock) were separated into the survivor group (349 cases) and nonsurvivor group (116 cases). Age, CRP, PCT, MEDS score, and sPD-1 level were higher in the nonsurvivor group than in the survivor group (P b 0.001), while gender, WBC count, platelet count, and site of infection showed no significant difference between the 2 groups (P b 0.05) (Table 2). Logistic regression analysis demonstrated that sPD-1, PCT, and the MEDS score were

survival curves of sPD-1″>Table 1

Characteristics of the five groups.

Characteristics	Healthy control	SIRS	Sepsis	Severe sepsis	Septic shock	P-value
	(n = 60)	(n = 130)	(n = 276)	(n = 121)	(n = 68)
Male, n (%)	33 (55.0)	75 (57.7)	158 (57.2)	67 (55.4)	37 (54.4)	0.984
Age, years	66 (56-75)	67 (55-78)	68 (56-77)	70 (58-77)	71 (58-79)	0.371
WBC (x109/L)	6.51 (5.89-7.66)	10.30 (7.22-13.28)	12.53 (9.88-15.59)	14.07 (11.34-16.91)	14.91 (10.19-18.41)	b0.001
Platelets, (x109/L)	238 (211-268)	222 (191-255)	199 (163-250)	176 (137-238)	149 (94-214)	b0.001
CRP (mg/L)	/	30 (12-62)	50 (17-105)	73 (36.5-237.5)	99 (49-158)	b0.001
PCT (ng/ml)	/	0.21 (0.05-0.55)	2.29 (0.77-5.00)	4.66 (2.19-7.45)	6.18 (4.16-10.06)	b0.001
MEDS score	/	3 (3-6)	8 (5-11)	11 (8-14)	16 (13-19)	b0.001
sPD-1 (ng/ml)	0.76 (0.31-1.29)	0.80 (0.43-1.30)	1.01 (0.55-1.58)	1.72 (1.07-2.94)	2.70 (1.31-4.83)	b0.001
Comorbidities, n (%)
COPD	/	47 (36.2)	75 (27.2)	41 (33.9)	24 (35.3)	0.215
Cardiovascular disease	/	49 (37.7)	93 (33.7)	50 (41.3)	31 (45.6)	0.225
Chronic renal disease	/	35 (26.9)	52 (18.8)	22 (18.2)	13 (19.1)	0.237
Diabetes mellitus	/	31 (23.8)	67 (24.3)	33 (27.3)	20 (29.4)	0.765
Cerebrovascular disease	/	22 (16.9)	28 (10.1)	16 (13.2)	12 (17.6)	0.170
Mortality, n (%)	/	10 (7.7)	37 (13.4)	42 (34.7)	37 (54.4)	b0.001

SIRS: systemic inflammatory response syndrome; WBC: white blood cells; CRP: C-reactive protein; PCT: procalcitonin; MEDS: Mortality in Emergency Department Sepsis; sPD-1: soluble programmed death-1; COPD: chronic obstructive pulmonary disease.

WBC, platelets, and sPD-1 were significantly different in multi-group comparison among the five groups (P b 0.001); levels of CRP, PCT, MEDS score, and 28-day mortality were significantly different among the four groups of SIRS, sepsis, severe sepsis, and septic shock (P b0.001).

independent risk factors for 28-day mortality of patients with sepsis (P b 0.05), while age and CRP were not (P N 0.05) (Table 3).

Comparison of sPD-1, PCT, and the MEDS score for predicting 28-day mortality

The ability of sPD-1, PCT, and the MEDS score to predict 28-day mor- tality of patients with sepsis was compared using the ROC curve (Fig. 2), and their AUC, cutoff value, sensitivity, specificity, positive predictive value, and negative predicted values were determined (Table 4). The AUC values of sPD-1, PCT and the MEDS score were 0.725, 0.693, and 0.767, respectively. The AUC of the MEDS score was significantly higher than that of PCT (P = 0.024), while the AUC of sPD-1 showed no

significant difference compared to that of PCT and the MEDS score (P = 0.417 and 0.225). The MEDS score showed high sensitivity (87.9%) but low specificity (54.3%), while sPD-1 showed the highest specificity (91.8%).

Survival curves of sPD-1

Using the cutoff value determined by the ROC curve, Kaplan-Meier survival curves of sPD-1 were established. Kaplan-Meier survival analy- sis indicated that patients with sepsis with sPD-1 levels higher than

2.38 ng/ml had a lower Probability of survival at 28 days than patients with lower sPD-1 levels (Log-rank = 90.40; P b 0.001) (Fig. 3).

Fig. 1. Levels of sPD-1 in the 5 groups. There is no significant difference in the levels of sPD-1 between the SIRS group and the healthy control group (P = 0.462). Levels of sPD-1 are higher in the sepsis group than in the SIRS and healthy control groups (P b 0.05). In cases of sepsis, severe sepsis, and septic shock, as the severity of sepsis increases, the levels of sPD-1 gradually increase (P b 0.001). sPD-1: soluble programmed death-1; SIRS: systemic inflammatory response syndrome.

Table 2

Comparison between the survivor group and nonsurvivor group.

Characteristics	Survivors	Nonsurvivors	P-value
	(n = 349)	(n = 116)
Male, n (%)	197 (56.4)	65 (56.0)	0.938
Age, years	67 (55-77)	71 (58-79)	b0.001
WBC (x109/L)	13.23	14.91	0.077
Platelets, (x109/L)	(10.10-15.83) 193 (148-244)	(10.19-18.41) 149 (94-214)	0.076
CRP (mg/L)	58 (23-105)	99 (41-160)	b0.001
PCT, (ng/ml)	2.80 (0.94-5.43)	6.12 (2.89-9.14)	b0.001
MEDS score	8 (3 -13)	14 (11-18)	b0.001
sPD-1 (ng/ml)	1.17 (0.66-1.75)	2.16 (1.05-4.28)	b0.001
Primary site of infection, n
(%)
Respiratory	186 (53.3)	63 (54.3%)	0.849
Abdominal	71 (20.3)	22 (19.0)	0.748
Urinary	46 (13.2)	14 (12.1)	0.757
Cerebral	23 (6.6)	10 (8.6)	0.461
Others	22 (6.3)	7 (6.0)	0.917

WBC: white blood cells; CRP: C-reactive protein; PCT: procalcitonin; MEDS: Mortality in Emergency Department Sepsis; sPD-1: soluble programmed death-1.

Combinations of sPD-1, PCT, and MEDS score

The efficacy of the combinations of sPD-1, PCT, and the MEDS score to predict 28-day mortality of patients with sepsis was also tested. The AUC values of the three combinations (MEDS + sPD-1, MEDS + PCT, and MEDS + PCT + sPD-1) were 0.829, 0.792, and 0.843, respectively, and all of them higher than the AUC of the three factors when used alone (P b 0.05). The Hosmer-Lemeshow test indicated that the three combinations showed a high degree of fitness (P = 0.892, 0.631, and 0.824, respectively). The combination of MEDS + PCT + sPD-1 showed higher AUC than the MEDS + PCT combination (P = 0.002), and the sensitivity, specificity, positive predictive value, and negative predictive value of the MEDS + PCT + sPD-1 combination were 81.6%, 83.4%, 71.7%, and 89.8%, respectively (Fig. 2, Table 4). The MEDS + sPD-1 com- bination had a larger AUC than the MEDS + PCT combination, but the difference was not statistically significant (P = 0.052). There was no sig- nificant difference in the AUC between the combinations of MEDS

+ sPD-1 and MEDS + PCT + sPD-1 (P = 0.077).

Discussion

In this study, we found that peripheral blood sPD-1 levels of patients with sepsis were higher than those of healthy control and SIRS patients, and the sPD-1 levels were positively correlated with the severity of sep- sis. In addition, the sPD-1 level was also higher in patients of the nonsurvivor group than in those of the survivor group. Our results sug- gest that sPD-1 levels can be used for risk stratification of patients with sepsis; the patients with sPD-1 levels higher than 2.38 ng/ml tend to

Table 3

Logistic regression analysis of independent risk factors for 28-day mortality in patients with sepsis.

Fig. 2. Comparison of the ROC curves for predicting 28-day mortality. The AUC of sPD-1, PCT and the MEDS score are 0.725, 0.693, and 0.767, respectively. The AUC of the MEDS score is higher than that of PCT (P = 0.024), while the ACU of sPD-1 shows no significant difference compared to those of PCT and the MEDS score (P = 0.417 and 0.225). There are 3 combinations of sPD-1, PCT, and the MEDS score: MEDS + sPD-1, MEDS + PCT, and MEDS + PCT + sPD-1. The AUC of the 3 combinations are 0.829, 0.792, and 0.843, respectively, and all of them are higher than the AUC of the 3 factors when used alone (P b 0.05). The combination of MEDS + PCT + sPD-1 shows higher AUC than the MEDS + PCT combination (P = 0.002). The MEDS + sPD-1 combination shows a larger AUC than the MEDS + PCT combination, but the difference is not statistically significant (P = 0.052). There is no significant difference in the AUC between the combinations of MEDS + sPD-1 and MEDS + PCT + sPD-1 (P = 0.077). ROC: receiver operating characteristic; AUC: area under the curve; sPD-1: soluble programmed death-1; PCT: procalcitonin; MEDS: Mortality in Emergency Department Sepsis.

have more severe symptoms and to be at higher risk of death within 28 days of admission. Our study demonstrated that sPD-1 was an inde- pendent risk factor for 28-day mortality of patients with sepsis and was useful in prediction of 28-day mortality. Therefore, sPD-1 may be used as an immunological biomarker for early assessment of the severity of sepsis and prognosis.

immune dysfunction is thought to be one of the main causes of death in patients with sepsis [1,18], and negative co-stimulatory molecules such as PD-1 may play an important role [19]. It has been reported that the expression of PD-1 on the surface of peripheral blood macro- phages and monocytes was increased in a mouse model of sepsis [20], and administration of a PD-l antagonist significantly improved survival of mice with sepsis [21]. A clinical study in recent years also confirmed that PD-1 expression on peripheral blood T-cells was significantly in- creased in patients with sepsis [22]. The present study found that sPD- 1 levels were increased in the patients with sepsis, and as the severity

Variable	?	Standard error	Wald	P-value	Odds ratio	95% CI Lower Upper		of sepsis increased, the levels of sPD-1 increased. Therefore, sPD-1 may also play a role in the immune regulation of sepsis.
Age	0.016	0.010	2.515	0.113	1.016	0.996	1.036	It has been demonstrated that sPD-1 can promote T-cell responses
sPD-1	0.617	0.105	34.569	b 0.001	1.854	1.509	2.277	through blocking membrane-bound PD-1/PD-L1 pathway in vitro [23].
MEDS score	0.149	0.030	25.029	b 0.001	1.161	1.095	1.231	sPD-1 can bind PD-L1 and block PD-1/PD-L1 interactions [24]. Many clinical studies have shown that patients with severe sepsis or septic

PCT	0.051	0.023	5.017	0.025	1.052	1.006	1.100
CRP	0.004	0.003	2.839	0.092	1.004	0.999	1.009
Constant	-5.696	0.755	56.886	0.000	0.003

?: regression coefficient; CI: confidence interval; sPD-1: soluble programmed death-1; MEDS: Mortality in Emergency Department Sepsis; PCT: procalcitonin; CRP: C-reactive protein.

shock had significantly higher PD-1/PD-L1 expression on the surface of cells [25-27]. PD-1/PD-L1 pathway blockade in vitro with anti-PD-1 and anti-PD-L1 antibodies markedly decreased sepsis-induce lympho- cyte apoptosis and restored the ability of immune effector cells to pro- duce Proinflammatory cytokines, such as tumor necrosis factor (TNF)

Table 4

Comparison of sPD-1, PCT, MEDS score, and combinations of these parameters for predicting 28-day mortality in patients with sepsis.

Variable	AUC	Standard error	95% CI	Cutoff value	Sensitivity %	Specificity %	PPV %	NPV %
sPD-1	0.725	0.030	0.667-0.784	2.38 ng/ml	50.5	91.8	83.3	69.6
PCT	0.693	0.029	0.636-0.749	4.68 ng/ml	64.7	68.7	58.2	74.6
MEDS score	0.767	0.024	0.721-0.813	11	87.9	54.3	70.0	78.6
MEDS score + sPD-1	0.829	0.022	0.786-0.873	/	85.1	73.4	71.1	86.5
MEDS score + PCT	0.792	0.023	0.748-0.837	/	90.6	62.7	79.8	80.4
MEDS score + PCT + sPD-1	0.843	0.021	0.802-0.884	/	81.6	83.4	71.7	89.8

AUC: area under the curve; CI: confidence interval; PPV: positive prediction value; NPV: negative prediction value; sPD-1: soluble programmed death-1; PCT: procalcitonin; MEDS: Mor- tality in Emergency Department Sepsis.

? and interleukin (IL)-6 [22,25]. A recent study reported that TNF? and IL-6 could stimulate the increase of sPD-1 levels in cell culture superna- tants [28]. The overexpression of PD-1 on CD4⁺ and CD8⁺ T cells and el- evated serum levels of sPD-1 have been found in patients with aplastic anemia or RA [12,29]. Some researchers proposed that the function of overexpressed PD-1 on T cells to restrict over-self-reaction was counteracted by the excessive production of sPD-1 [12]. Other studies also suggested that when such immune regulation was uncontrolled, the excessive sPD-1 blocked the PD-1/PD-L1 pathway and led to aber- rant activation and proliferation of T cells [9,10]. We found that sPD-1 levels were significantly increased in patients with severe sepsis and septic shock, and a marked increase in sPD-1 levels may indicate im- mune dysfunction in patients with sepsis.

Our previous study has demonstrated that PD-L1 expression in monocytes is an independent predictor of 28-day mortality in patients with septic shock [15]. This study also suggested that sPD-1 performed as an independent predictor of 28-day mortality of patients with sepsis, and could be used for risk stratification and prognostic evaluation for patients with sepsis. Moreover, the detection of serum sPD-1 is more easily and rapidly performed than the detection of membrane-bound PD-1/PD-L1 by flow cytometry, and it is more suitable for clinical applications.

A recent study found that sPD-1 was elevated in Acute respiratory distress syndrome and might represent a potential biomarker

Fig. 3. Survival curves of sPD-1. Kaplan-Meier survival analysis indicates that patients with sepsis with sPD-1 levels higher than 2.38 ng/ml have a lower probability of survival at 28 days than patients with lower sPD-1 levels (Log-rank = 90.40; P b 0.001). sPD-1: soluble programmed death-1.

for ARDS [30]. However, we also noticed that another recent study by Lange et al. reported that sPD-1 concentrations were lower in patients with sepsis and septic shock than in healthy Control subjects, and were not associated with disease severity and mortality [31]. The dis- crepancies between our results and those of Lange et al. may be due to the fact that their sPD-1 levels of healthy controls (median 2.9, inter- quartile ranges 0.9-9.1 ng/ml) were relatively high. Conversely, the levels of sPD-1 were low in healthy controls from many other studies and were in agreement with ours [10,12,23]. There is a hypothesis that the basal expression of sPD-1 may be different between various lin- eages. This observation deserves to be further examined. The discrepan- cies between our study and those of Lange et al. may be due to the timing of the sample acquisition. The samples from Lange study were collected at different times after ICU admission and the patients had al- ready been subjected to therapy in the ICU. The samples of our patients were collected on ED arrival. Therefore, the patients in the Lange study may have been in the immunosuppression stage of sepsis, whereas the patients in our study may be in the early phase of sepsis. Additionally, whether the levels of sPD-1 change over time remains unknown. There- fore, it is possible that the level of sPD-1 varies and further studies are needed to investigate this.

The biomarkers PCT and CRP are commonly used to detect inflam-

mation [32-34], and the MEDS score is used to predict the mortality of patients with suspected infection in the ED [17]. Our study found that the levels of sPD-1, CRP, and PCT and the MEDS score gradually in- creased as the disease progressed, and sPD-1 and PCT levels as well as the MEDS score were independent predictors for 28-day mortality in patients with sepsis. The Predictive ability of sPD-1 was similar to that of PCT and the MEDS score, sPD-1 was more specific for patients with sepsis, but the sensitivity of sPD-1 was low. We speculate that if the samples of analysis were to restrict to the most critical patients (such as only septic shock patients), the sensitivity of sPD-1 probably would be improved. We also found the sensitivity of the MEDS score was the highest. So we combined sPD-1 with the MEDS score. The prognostic evaluation was improved when sPD-1 was combined with the MEDS score, and the combination of MEDS + PCT + sPD-1 showed signifi- cantly superior predictive ability compared with the combination of MEDS + PCT. These results suggest that sPD-1 helps to improve the pre- dictive capacity of current clinical indicators for the prognosis of pa- tients with sepsis. Because sPD-1 reflects the status of the immune system, combining sPD-1 with the use of biomarkers of inflammation and a scoring system may help clinicians to assess the prognosis of pa- tients with sepsis.

This study has several limitations. First, we examined only sPD-1 in

the peripheral blood of the patients, and the expression of membrane- bound PD-1/PD-L1 was not measured. The correlation between mem- brane-bound PD-1/PD-L1 and sPD-1 during sepsis remains to be ex- plored. Second, all the patients in this study were selected from the ED, who were tested and sampled only on ED arrival, without continu- ous monitoring. In addition, this is only a single-center study, and mul- ticenter studies with a larger sample size are required to further validate these results.

Conclusions

This study demonstrated that serum sPD-1 levels significantly in- creased in patients with sepsis, especially in those with severe sepsis or septic shock. Serum sPD-1 is an independent risk factor for 28-day mortality in patients with sepsis, and is a valuable indicator for risk stratification and prediction of 28-day mortality. Combining sPD-1 with PCT and MEDS score significantly improves the prediction for 28- day mortality from sepsis. Of note, serum sPD-1 detection is more easily and rapidly performed than the detection of membrane-bound PD-1/ PD-L1 by flow cytometry. However, the clinical application of serum sPD-1 measurements for patients with sepsis requires further investigation.

Funding

This work was supported by the 2015 Annual Special Cultivation and Development Project for Technology Innovation Base of Beijing Key Laboratory of Cardiopulmonary Cerebral Resuscitation (No. Z151100001615056).

Acknowledgements

We wish to express our gratitude for the generosity of the patients who participated in this study. We sincerely thank the colleagues of our emergency department for their excellent clinical assistance.

References

1. Boomer JS, To K, Chang KC, Takasu O, Osborne DF, Walton AH, et al. Immunosup- pression in patients who die of sepsis and multiple organ failure. JAMA 2011;306: 2594-605.
2. Ishida Y, Agata Y, Shibahara K, Honjo T. Induced expression of PD-1,a novel member of the immunoglobulin gene superfamily, upon programmed cell death. EMBO J 1992;11:3887-95.
3. Freeman GJ, Long AJ, Iwai Y, Bourque K, Chernova T, Nishimura H, et al. Engagement of the PD-1 immunoinhibitory receptor by a novel B7 family member leads to neg- ative regulation of lymphocyte activation. J Exp Med 2000;192:1027-34.
4. Okazaki T, Honjo T. PD-1 and PD-1 ligands: from discovery to clinical application. Int

   Immunol 2007;19:813-24.

   Gianchecchi E, Delfino DV, Fierabracci A. Recent insights into the role of the PD-1/ PD-L1 pathway in immunological tolerance and autoimmunity. Autoimmun Rev 2013;12:1091-100.

5. Sharpe AH, Wherry EJ, Ahmed R, Freeman GJ. The function of programmed cell death 1 and its ligands in regulating autoimmunity and infection. Nat Immunol 2007;8:239-45.
6. Zhang G, Hou J, Shi J, Yu G, Lu B, Zhang X. Soluble CD276 (B7-H3) is released from monocytes, dendritic cells and activated T cells and is detectable in normal human serum. Immunology 2008;123:538-46.
7. Ward FJ, Dahal LN, Wijesekera SK, Abdul-Jawad SK, Kaewarpai T, Xu H, et al. The sol- uble isoform of CTLA-4 as a regulator of T-cell responses. Eur J Immunol 2013;43: 1274-85.
8. Dai S, Jia R, Zhang X, Fang Q, Huang L. The PD-1/PD-Ls pathway and autoimmune diseases. Cell Immunol 2014;290:72-9.
9. Greisen SR, Rasmussen TK, Stengaard-Pedersen K, Hetland ML, Horslev-Petersen K, Hvid M, et al. Increased soluble programmed death-1 (sPD-1) is associated with dis- ease activity and radiographic progression in early rheumatoid arthritis. Scand J Rheumatol 2014;43:101-8.
10. Cheng HY, Kang PJ, Chuang YH, Wang YH, Jan MC, Wu CF, et al. Circulating pro- grammed death-1 as a marker for sustained high hepatitis B viral load and risk of Hepatocellular carcinoma. PLoS One 2014;9:e95870.
11. Wu H, Miao M, Zhang G, Hu Y, Ming Z, Zhang X. Soluble PD-1 is associated with ab- errant regulation of T cells activation in aplastic anemia. Immunol Invest 2009;38: 408-21.
12. Xiao H, Huang B, Yuan Y, Li D, Han LF, Liu Y, et al. Soluble PD-l facilitates 4-1BBL-trig- gered antitumor immunity against murine H22 hepatocarcinoma in vivo. Clin Can- cer Res 2007;13:1823-30.
13. Onlamoon N, Rogers K, Mayne AE, Pattanapanyasat K, Mori K, Villinger F, et al. Sol- uble PD-1 rescues the proliferative response of simian immunodeficiency virus-spe- cific CD4 and CD8 T cells during chronic infection. Immunology 2008;124:277-93.
14. Shao R, Fang Y, Yu H, Zhao L, Jiang Z, Li CS. Monocyte programmed death ligand-1 expression after 3-4 days of sepsis is associated with risk stratification and mortality in septic patients: a prospective cohort study. Crit Care 2016;20:124.
15. Levy MM, Fink MP, Marshall JC, Abraham E, Angus D, Cook D, et al. 2001 SCCM/ ESICM/ACCP/ATS/SIS international sepsis definitions conference. Crit Care Med 2003;31:1250-6.
16. Shapiro NI, Wolfe RE, Moore RB, Smith E, Burdick E, Bates DW. Mortality in the Emergency Department Sepsis (MEDS) score: a prospectively derived and validated clinical prediction rule. Crit Care Med 2003;31:670-5.
17. Torgersen C, Moser P, Luckner G, Mayr V, Jochberger S, Hasibeder WR, et al. Macro- scopic postmortem findings in 235 surgical intensive care patients with sepsis. Anesth Analg 2009;108:1841-7.
18. Liu Q, Li C. Programmed Cell Death-1/Programmed Death-ligand 1 Pathway: A New Target for Sepsis. Chin Med J (Engl) 2017;130:986-92.
19. Huang X, Venet F, Wang YL, Lepape A, Yuan ZL, Chen YPS, et al. PD-1 expression by macrophages plays a pathologic role in altering microbial clearance and the innate inflammatory response to sepsis. Proc Natl Acad Sci U S A 2009;106:6303-8.
20. Brahmamdam P, Inoue S, Unsinger J, Chang KC, McDunn JE, Hotchkiss RS. Delayed administration of anti-PD-1 antibody reverses immune dysfunction and improves survival during sepsis. J Leukoc Biol 2010;88:233-40.
21. Chang K, Svabek C, Vazquez-Guillamet C, Sato B, Rasche D, Wilson S, et al. Targeting the programmed cell death 1: programmed cell death ligand 1 pathway reverses T cell exhaustion in patients with sepsis. Crit Care 2014;18:R3.
22. Liu C, Jiang J, Gao L, Wang X, Hu X, Wu M, et al. Soluble PD-1 aggravates progression of collagen-induced arthritis through Th1 and Th17 pathways. Arthritis Res Ther 2015;17:340.
23. He YF, Zhang GM, Wang XH, Zhang H, Yuan Y, Li D, et al. Blocking programmed death-1 ligand-PD-1 interactions by local gene therapy results in enhancement of antitumor effect of secondary lymphoid tissue chemokine. J Immunol 2004;173: 4919-28.
24. Zhang Y, Li J, Lou J, Zhou Y, Bo L, Zhu J, et al. Upregulation of programmed death-1 on T cells and programmed death ligand-1 on monocytes in septic shock patients. Crit Care 2011;15:R70.
25. Guignant C, Lepape A, Huang X, Kherouf H, Denis L, Poitevin F, et al. Programmed death-1 levels correlate with increased mortality, Nosocomial infection and immune dysfunctions in septic shock patients. Crit Care 2011;15:R99.
26. Jia Y, Zhao Y, Li C, Shao R. The expression of programmed death-1 on CD4+ and CD8+ T lymphocytes in patients with type 2 diabetes and severe sepsis. PLoS One 2016;11:e0159383.
27. Bommarito D, Hall C, Taams LS, Corrigall VM. Inflammatory cytokines compromise programmed cell death-1 (PD-1)-mediated T cell suppression in inflammatory ar- thritis through up-regulation of soluble PD-1. Clin Exp Immunol 2017;188:455-66.
28. Wan B, Nie H, Liu A, Feng G, He D, Xu R, et al. Aberrant regulation of synovial T cell activation by soluble costimulatory molecules in rheumatoid arthritis. J Immunol 2006;177:8844-50.
29. Monaghan SF, Chung CS, Chen Y, Lomas-Neira J, Fairbrother WG, Heffernan DS, et al. Soluble programmed cell death receptor-1 (sPD-1): a potential biomarker with anti- inflammatory properties in human and experimental acute respiratory distress syn- drome (ARDS). J Transl Med 2016;14:312-20.
30. Lange A, Sunden-Cullberg J, Magnuson A, Hultgren O. Soluble B and T lymphocyte attenuator correlates to disease severity in sepsis and high levels are associated with an increased risk of mortality. PLoS One 2017;12:e0169176.
31. Nobre V, Harbarth S, Graf JD, Rohner P, Pugin J. Use of procalcitonin to shorten anti- biotic treatment duration in septic patients: a randomized trial. Am J Respir Crit Care Med 2008;177:498-505.
32. Becker KL, Snider R, Nylen ES. Procalcitonin assay in systemic inflammation, infec- tion, and sepsis: clinical utility and limitations. Crit Care Med 2008;36:941-52.
33. Tschaikowsky K, Hedwig-Geissing M, Braun GG, Radespiel-Troeger M. Predictive value of procalcitonin, interleukin-6, and C-reactive protein for survival in postoper- ative patients with severe sepsis. J Crit Care 2011;26:54-64.