In the clinical trial, all participants were recruited from the Fifth Medical Center of the People’s Liberation Army (PLA) General Hospital for screening from June 2014 to September 2016. Eligible patients met the diagnostic criteria for ACLF, suggested by the Asian Pacific Association for the Study of the Liver (APASL) (1, 24), which are: (i) pre-existing diagnosed or undiagnosed chronic liver disease; (ii) acute deterioration with exacerbating jaundice (serum total bilirubin ≥ 5 mg/dL); (iii) international standard ratio (INR) ≥ 1.5 or plasma prothrombin activity (PTA; <40%); and (iv) comply with ascite and/or encephalopathy within 4 weeks. The inclusion criteria were as follows: (i) 18–70-year-old male or female; and (ii) serum hepatitis B surface antigen or HBV DNA was detected for at least 6 months. The exclusion criteria were as follows: (i) super-infection or co-infection with other hepatotropic and non-hepatotropic virus; (ii) a previous application of any immune modulator or cytotoxic/immunosuppressive therapy within the previous 12 months; (iii) hepatocellular carcinoma or extrahepatic malignancy; (iv) co-existence of severe renal, lung, brain,or heart disease or other liver disease such as alcoholic liver disease, Wilson disease, drug-induced liver injury, or autoimmune hepatitis; (v) presence of sepsis; (vi) malignant jaundice leaded by obstructive or hemolytic jaundice; and (vii) prolonged prothrombin time due to hematologic system disease.

Ascites was identified by clinical manifestations, diagnostic paracentesis, and abdominal imaging examination; Acute kidney injury was diagnosed according to the International Club of Ascites (ICA) corresponding standard (25); A West Haven classification was applied in the diagnosis of hepatic encephalopathy (26); Corresponding formulas were applied for calculation of the model for end-stage liver disease (MELD), MELD-sodium (MELD-Na), and chronic liver failure-sequential organ failure assessment (CLIF-SOFA) scores (3, 27, 28).

Diagnosis of infection was made based on clinical presentation, laboratory values and imaging examination. (i) spontaneous bacterial peritonitis: polymorphonuclear cell count in ascitic fluid ≥250/mm³. (ii) pneumonia: pulmonary imaging changes (infiltration, consolidation, or cavitation) plus at least 1 of the following presentation (fever ≥38°C, leucocyte count >12,000/mm³ or <4,000/mm³) plus at least 1 of the following symptoms or signs (new cough, sputum production, dyspnea, rales or bronchial breath sounds, etc.) and/or etiological evidence. (iii) urinary tract infection: Patient had at least 1of the following clinical presentations (fever, urinary tract irritation or suprapubic tenderness, etc.) and positive urine culture; or at least 2 of the above presentations and more than 10 leukocytes/μL in urine. (iv)bacteremia: positive blood culture (29, 30).

PASS 11.0 software (NCSS, Kaysville, UT, USA) was used to calculate the sample size. Based on our previous study, we expected that G-CSF therapy could improve the survival rate at 180 days by approximately 20% in the G-CSF group compared to that in the control group. With a statistical power of 80%, we required 52 patients in each group to detect this meaningful difference at a significance level of 5%. Considering the possible loss of patients to follow-up (10%), 114 patients were included in this study. The computer-generated randomization number code was prepared for each patient. The patients were randomly allocated to the G-CSF group (receive G-CSF therapy plus standard medical therapy) or control group (only receive standard medical therapy).

G-CSF (SL Pharm, Beijing, China) was injected subcutaneously at a dose of 5 μg/kg every day for 6 days and then every other day until day 18 (total 12 doses) in G-CSF group patients; thereafter, physical examination and laboratory tests were conducted. All patients received standard medical therapy, including intensive care monitoring, antiviral therapy, supportive therapy, and prevention and treatment for complications. All patients with infection were treated with antibiotics. Patients received albumin, terlipressin, and so on if required.

A total of 114 patients meeting the inclusion and exclusion criteria were enrolled in the study (Figure 1). Of these, 56 patients received both standard medical therapy and G-CSF treatment (G-CSF group), and 58 patients received standard medical therapy only (control group). All patients were followed up for at least 180 days after treatment commencement. During the follow-up period, one patient in the G-CSF group discontinued treatment, and one patient each in the G-CSF group and control group underwent liver transplantation. Thus, 111 patients were included in the final analysis.

Peripheral blood samples were obtained from the enrolled patients and loaded into 10 mL heparin anticoagulant tubes for the following experiments.

Peripheral blood mononuclear cells (PBMCs) were obtained from patients with HBV-ACLF (n = 12), before and after G-CSF treatment, to evaluate the effect of G-CSF on monocytes. Surface markers and intracellular cytokine expression in monocytes, expressed as frequency or mean fluorescence intensity (MFI), were measured using flow cytometry on the BD FACSymphony A5 analyzer (BD Biosciences, UK). Data were analyzed using the FlowJo software (Tristar, San Carlos, CA, USA).

For surface marker staining, APC-Cy7-conjugated anti-CD45 (BD Biosciences), PerCP-Cy5.5-conjugated anti-CD14 (BD Biosciences), APC-conjugated anti-CD16 (BD Biosciences), BV711-conjugated anti-HLA-DR (BD Biosciences), PE-conjugated anti-MerTK (R&D Systems), PE-Cy7-conjugated anti-CD163 (eBioscience), BV421-conjugated anti-CCR2 (BioLegend), FITC-conjugated anti-CX3CR1 (BioLegend), and BUV-737-conjugated anti-CD86 (BD Biosciences) antibodies were used. For intracellular cytokine staining (ICCS), BV-421-conjugated anti-IL-6 (BD Biosciences), PE-CF594-conjugated anti-IL-10 (BD Biosciences), and BV-510-conjugated anti-TNF-α (BioLegend) were used. Dead cells were excluded using Fixable Viability Stain 440UV (BD Bioscience).

For surface marker labeling, PBMCs were incubated with surface fluorescent-labeled monoclonal antibodies. For intracellular staining, PBMCs were incubated with or without 100 ng/mL lipopolysaccharide (LPS) and protein transport inhibitor (1μL/mL, BD GolgiPlug) for 6 h, followed by staining with surface markers, and fixation, permeabilization, and staining with the corresponding fluorescent intracellular antibodies.

PBMCs were obtained from patients with HBV-ACLF who did not receive G-CSF therapy. CD14⁺ monocytes were further isolated using magnetic bead separation (Miltenyi, Bergisch Gladbach, Germany). The purity of monocyte separation was examined using flow cytometry, and CD14⁺ monocytes with a purity >95% were used in vitro experiments. Isolated monocytes were incubated in complete media with 20 ng/mL G-CSF or an equivalent volume of PBS (control) for 24 h for surface staining. For intracellular staining, the isolated CD14⁺ monocytes were stimulated with G-CSF (20 ng/ml) for 18 h. After 18 h incubation, LPS (100 ng/ml) and protein transport inhibitor (1μL/mL, BD GolgiPlug) was added and continue to incubation for 6 h, followed by staining with surface markers, and fixation, permeabilization, and staining with the corresponding fluorescent intracellular antibodies. After all the above steps, the monocytes were harvested and analyzed for their surface phenotype and intracellular cytokine levels using flow cytometry. Staining was performed as previously described.

The effect of G-CSF therapy on monocyte phagocytosis and the oxidative burst capacity of patients with HBV-ACLF was evaluated. Monocyte phagocytosis and oxidative burst capacity were tested using the PHAGOTEST kit (CELONIC, Germany) and PHAGOBURST kit (CELONIC, Germany) following the manufacturer’s protocol, respectively, and then assessed by flow cytometry.

The study protocol and informed consent form were approved by the Human Ethics Committee of the Fifth Medical Center of the PLA General Hospital. All the procedures were in accordance with the ethical guidelines of Declaration of Helsinki. Written informed consent was obtained from patients or their guardians before enrollment. This trial was registered at ClinicalTrials.gov (NCT02331745).

In the clinical trial, continuous variables with normal distribution and skewed distribution were expressed as mean ± standard deviation (SD) and median [interquartile range (IQR)], respectively. Categorical variables were expressed as numbers and percentages. The chi-square test (for categorical variables) and Student’s t-test (for continuous variables, normal distribution) or Mann–Whitney U test (for continuous variables, skewed distribution) were used to compare differences between the two groups. Patients lost to follow-up contributed to the censored data. Survival rates were calculated using the Kaplan–Meier method and compared using the Log-Rank test. In the crude model and three multivariate adjusted models, Cox proportional hazards model was used for multivariate regression analysis to estimate the hazard ratios (Hazard Ratio, HRs) and 95% confidence intervals (Confidence Interval, CIs) for the risk of 180-day mortality. Statistical analyses and graphing were conducted using the statistical packages R version 3.4.3 (The R Foundation, Vienna, Austria), EmpowerStats (X&Y Solutions, Inc., Boston, MA, USA) and the MedCalc 15.2.2 (Ostend, Belgium). Statistical significance was set at P < 0.05.

In the Experimental section, statistical analysis was performed using GraphPad Prism 7.00 (GraphPad Software Inc., La Jolla, CA, USA). Data are presented as median (IQR). Non-parametric analyses, such as Kruskal–Wallis and Mann–Whitney U tests, were applied to comparisons across the different groups. Comparisons within the same individual were performed using Wilcoxon’s matched-pair test. Statistical significance was set at P < 0.05.

As shown in Table 1, the median age was 43.9 years [n = 91 men (82.0%) and 20 women (18.0%)]. Seventy patients (63.1%) patients had liver cirrhosis. MELD, MELD-Na and CLIF-SOFA scores were 23.7 (21.0–26.4), 22.9 (17.6–29.2), and 7.2 ± 1.0, respectively. The most common complication was ascites (88.3%), followed by infection (36.9%). Serum bilirubin and CRP in the G-CSF group were higher than those in the control group, but the differences were not statistically significant (P = 0.066 and 0.051, respectively). Meanwhile, the serum creatinine level in the control group was slightly higher than that in the G-CSF group, but both were within the normal range, and no statistical difference was detected (P = 0.090). Table 1 shows that there were no significant differences in the patients’ demographic and clinical characteristics between the control and G-CSF groups.

The patients tolerated the treatment well, and no severe side effects were observed. One of the patients developed a mild rash during the follow-up and discontinued all drugs, including G-CSF. However, no evidence was shown that the rash was induced by G-CSF. The results showed that the G-CSF treatment protocol was safe.

During the first month, the main new-onset complications were infection, hepatic encephalopathy, and acute kidney injury (Supplementary Table 1). The frequency of complications was not significantly different (P > 0.05) between the control and G-CSF groups.

Of the 111 patients, 66 survived, 40 died, and 5 lost to 180 days of follow-up; thus, the survival probability at day 180 was 62.6%. In the G-CSF group, 36 patients survived, 14 patients died, and 4 patients lost to follow-up; hence, the survival probability at day 180 was 72.2%. In the control group, 30 patients survived, 26 patients died, and 1 patient lost to follow-up, with a survival probability of 53.8%. The differences between the two groups were statistically significant (P = 0.0242; Figure 2).

We further explored the association between baseline and day 7 monocyte count and 180-day mortality in HBV-ACLF patients treated with or without G-CSF. As shown in Table 2, in both the crude and adjusted models (Models 1, 2, and 3), the baseline monocyte count was positively correlated with the 180-day mortality risk of HBV-ACLF [HR: 2.90 (1.41, 5.93), P = 0.0036 in Model 3], and the stratification analyses showed that this positive correlation was mainly contributed by patients in the G-CSF group [HR: 15.48 (3.60, 66.66), P = 0.0002 in Model 3]. However, after six consecutive days of treatment, the control group did not demonstrate significant changes, whereas the positive association between monocyte count on day 7, and the risk of death was significantly weakened in the G-CSF-treated group [HR: 1.10 (0.50, 2.43), P = 0.8080 in Model 3]. Similar relationship was observed between monocyte count and 90-day mortality (Supplementary Table 2). Therefore, we speculate that G-CSF may benefit patients with HBV-ACLF by altering the number or function of monocytes.

According to traditional gating, monocyte subtypes can be identified by the expression of CD14 and CD16 (Figure 3A). We evaluated the frequencies of the three monocyte subsets (described above). As shown in Figure 3B, although there was no significant difference in the proportion of classic, intermediate, and non-classical monocytes before (day 0) and after G-CSF treatment (P = 0.8981, 0.7113 and 0.9953, respectively), the intermediate and non-classical monocytes demonstrated a decreasing trend, whereas classical monocytes demonstrated an increasing trend. Thus, a shift in monocyte subpopulations was detected after G-CSF treatment.

To fully evaluate the effect of G-CSF on monocytes in HBV-ACLF, we designed detailed immunophenotypic analyses to detect the levels of activation/inflammation markers (HLA-DR, CD86, MerTK, and CD163) and tissue-homing markers (CCR2 and CX3CR1) before (day 0) and after the administration of G-CSF (n = 12). As shown in Figure 4, after treatment with G-CSF, the expression of M1-like markers (HLA-DR and CD86) in monocytes decreased (HLA-DR: day 0 vs. day 3 vs. day 7 vs. day 14 vs. day 28 = 94.9% vs. 78.5% vs. 80.1% vs. 82.1% vs. 92.6%, respectively; P = 0.0148; CD86: day 0 vs. day 3 vs. day 7 vs. day 14 vs. day 28 = 57.1% vs. 30.5% vs. 28.6% vs. 43% vs. 40.3%, respectively; P = 0.0764), whereas the expression of M2-like marker (MerTK) increased (day 0 vs. day 3 vs. day 7 vs. day 14 vs. day 28 = 39.3% vs. 71.5% vs. 56.3% vs. 55.4% vs. 37.5%, respectively; P = 0.0002). There was no significant change in the expression of CCR2 and CX3CR1 in monocytes according to MFI before and after G-CSF treatment (P = 0.1074 and 0.8889, respectively). The above results indicate that circulating monocytes after G-CSF therapy showed an anti-inflammatory/pro-restorative phenotype in patients with HBV-ACLF.

We evaluated whether G-CSF treatment in patients with HBV-ACLF modulated the ability of monocytes to produce IL-6, TNF-α, and IL-10 following LPS challenge or not. As shown in Figure 5, the secretion of TNF-α, IL-6, and IL-10 by monocytes decreased after G-CSF therapy without LPS stimulation (TNF-α: day 0 vs. day 3 vs. day 7 vs. day 14 vs. day 28 = 6.6% vs. 1.6% vs. 2.0% vs. 1.6% vs. 12.5%, respectively; P < 0.0001; IL-6: day 0 vs. day 3 vs. day 7 vs. day 14 vs. day 28 = 24.0% vs. 7.5% vs. 10.9% vs. 8.4% vs. 44.6%, respectively; P = 0.0025; IL-10: day 0 vs. day 3 vs. day 7 vs. day 14 vs. day 28 = 0.9% vs. 0.5% vs. 0.2% vs. 0.5% vs. 1.6%, respectively; P = 0.0004). In contrast to pre-treatment (day 0), cytokine secretion in monocytes showed a decreased response to LPS stimulation after G-CSF treatment (TNF-α: day 0 vs. day 3 vs. day 7 vs. day 14 vs. day 28 = 30.8% vs. 18.8% vs. 21.0% vs. 23.6% vs. 26.3%, respectively; P = 0.0439; IL-6: day 0 vs. day 3 vs. day 7 vs. day 14 vs. day 28 = 69.4% vs. 64.5% vs. 61.8% vs. 72.3% vs. 74.7%, respectively; P = 0.0611; IL-10: day 0 vs. day 3 vs. day 7 vs. day 14 vs. day 28 = 4.2% vs. 2.7% vs. 3.0% vs. 4.5% vs. 4.6%, respectively; P = 0.0099). Hence, G-CSF reduces the inflammatory response of monocytes in HBV-ACLF patients.

We performed in vitro experiments to further clarify the effect of G-CSF on the phenotype and function of monocytes in patients with HBV-ACLF. As shown in Figure 6, after exogenous addition of G-CSF to monocytes from HBV-ACLF patients, the expression of M1-type markers (HLA-DR and CD86) decreased, whereas the expression of M2-type markers (CD163 and MerTK) increased (P < 0.01). There was no significant difference in the expression of homing receptors CCR2 and CX3CR1 between the two groups (P > 0.05; Figure 6C). Concurrently, although no statistically significant difference was detected, G-CSF decreased the secretion of pro-inflammatory factors (IL-6 and TNF-α) after LPS stimulation, whereas IL-10 secretion was slightly increased (Figure 6D). Except for the secretion of IL-10, these results were consistent with those observed in vivo. G-CSF tend to promote the transition of monocytes to M2 type in HBV-ACLF patient.

We tested and compared phagocytosis (Figure 7A) and oxidative burst (Figure 7B) of monocytes freshly isolated from HBV-ACLF patients before (day 0) and after G-CSF treatment. It was observed that phagocytosis of monocytes showed an upward trend, whereas oxidative burst showed a downward trend after the administration of G-CSF; however, the discrepancies were not statistically significant in view of the small numbers (P all <0.05).