Altered PGE2-EP2 is associated with an excessive immune response in HBV-related acute-on-chronic liver failure

In total, 323 subjects were recruited from outpatients or inpatients of the First Affiliated Hospital of Zhejiang University between October 2016 and December 2017. These subjects were divided into four groups: (i) HB-ACLF: 135 CHB patients with ACLF diagnosed by the criteria of the Asian-Pacific Association for the Study of the Liver (APASL)—i.e., “acute hepatic insult manifesting as jaundice (≥ 5 mg/dL) and coagulopathy, complicated within 4 weeks by ascites and/or encephalopathy in a patient with previously diagnosed or undiagnosed chronic liver disease.” [11]; (ii) acute decompensated cirrhosis (AD): 30 AD patients were characterized by cirrhosis complicated with ascites, hepatic encephalopathy and upper gastrointestinal bleeding. Cirrhosis was defined by previous endoscopy, liver biopsy, radiological evidence, or clinical manifestation of liver decompensation. (iii) chronic hepatitis B (CHB): 128 age-/sex-matched patients with stable chronic hepatitis B were included. Chronic hepatitis B was diagnosed by serum HBsAg positive for more than six months together with histology or imaging or laboratorial or clinical evidence of cirrhosis or liver fibrosis or long-term liver inflammation; (iv) healthy controls (HC): 180 age-/sex-matched healthy subjects were used as controls. Multi-organ failure was diagnosed with two or more extrahepatic organ failures happen according to the chronic liver failure sequential organ failure assessment (SOFA) score. Exclusion criterions were as follows: age younger than 18 years, human immunodeficiency virus infection, pregnancy, immunotherapy, cancer, and a history of autoimmune diseases. Written consent was obtained from each subject or their nominated next of kin if the participants could not provide informed consent. This study was performed according to the principles of the Helsinki Declaration and was approved by the Ethic Committee of the First Affiliated Hospital of Zhejiang University. The baseline characteristics of the patients are shown in (Additional file 1: Table S1). The model for end-stage liver disease (MELD) score and CLIF-consortium organ failure (CLIF-C) score were calculated to assess the severity of the disease.

CD3-percp-cy5.5 (Catalogue #45-0036-42), CD8-APC (Catalogue #17-0086-42), CD8-FITC (Catalogue #11-0086-42), CD14-APC (Catalogue #17-0149-42), CD56-FITC (Catalogue #4278380), CD16-PerCP-eFluor™710 (Catalogue #46-0168-42), CD11b-PerCP-eFluor™710 (Catalogue #46-0110-80), IFN-γ-PE (Catalogue #12-7319-42), TLR2-FITC (Catalogue #11-9922-41) and Mouse IgG1-PE (Catalogue #12-4714-81) were all bought from eBioscience. HLA-DR-PE (Catalogue #555812), IL-10-APC (Catalogue #554707), CXCR3-APC (Catalogue #550967) and IFN-γ-FITC (Catalogue #561053), Rat IgG2a-APC (Catalogue #554690) and mouse IgG-APC (Catalogue #5065947) were purchased from BD Biosciences. CD16-FITC (Catalogue #302006), HLA-DR-APC (Catalogue #307609), CD3-FITC (Catalogue #317306) and TLR4-APC (Catalogue #312815) were purchased from Biolegend. EP2-PE (Catalogue #10477) and Rabbit IgG-PE (Catalogue D5-1610) were purchased from Cayman Chemical.

Whole-blood samples were collected within 24 h after admission. After centrifugation, the plasma was collected and stored immediately at − 80 °C. Peripheral blood mononuclear cells (PBMCs) were isolated using Ficoll-Hypaque density gradient centrifugation (GE healthcare, Pittsburgh, USA).

The plasma cytokine levels of ACLF patients were measured using a multiplex panel (Bio-Rad, Hercules, CA), according to the manufacturer’s instructions. PGE2 was measured using the Prostaglandin E Metabolites ELISA kit (Cayman Chemical, Michigan, USA). Both measurements were performed according to the manufacturer’s instructions. The detection limits were as follows:IL-1β, 17.3 pg/ml; IL-1rα, 15 pg/ml; IL-2, 16.5 pg/ml; IL-4, 11 pg/ml; IL-5, 27 pg/ml; IL-6, 24.8 pg/ml; IL-7, 11 pg/ml; IL-8, 14.5 pg/ml; IL-9: 21 pg/ml; IL-10: 33.3 pg/ml; IL-12: 15.5 pg/ml; IL-13: 13.5 pg/ml; IL-5; 57.3 pg/ml; IL-17: 24.8 pg/ml; eotaxin, 17 pg/ml; FGF basic, 14 pg/ml; G-CSF, 36.3 pg/ml; GM-CSF, 16 pg/ml; IFN-γ, 15.5 pg/ml; IP-10, 14.5 pg/ml; MCP-1, 18.5 pg/ml; MIP-1α, 8.8 pg/ml; PDGF-bb, 32.5 pg/ml; MIP-1β, 31.5 pg/ml; RANTES, 15.8 pg/ml; TNF-α, 12.5 pg/ml; VEGF, 109.3 pg/ml; PGE2, 2 pg/ml.

Isolated PBMCs or 100 μl of heparin-treated peripheral whole blood was incubated with antibodies for 15 min at room temperature. After incubation, whole-blood samples were lysed with FACS lysing solution and were washed with phosphate-buffered saline before analysis using an Accuri C6 cytometer (Accuri, BD). The isotype IgG was used as the control.

To explore factors driving EP2 expression, PBMC from healthy controls were incubated for 7 days with a range of IP-10 and MIP-1β and assessed for EP2 expression.

Gating strategies of flow cytometry for immune subsets are introduced in Figure S1 (Additional file 2): CD8⁺T cells (CD3⁺/CD8⁺), CD4⁺T cells (CD3⁺/CD8⁻), NK (CD3⁻/CD56⁺), NKT (CD3⁺/CD56⁺), monocytes (CD14⁺), neutrophils (CD16⁺).

In vitro, PBMCs (2 × 10⁵/well) were stimulated with lipopolysaccharide (LPS) (100 ng/ml; Sigma-Aldrich, St. Louis, MO) for 72 h in the presence or absence of AH6809 (150 μM; Cayman Chemical) or DMSO (0.1%) in 1640 RPMI medium (Invitrogen, Oslo, Norway) supplemented with 10% fetal bovine serum (FBS, Invitrogen), 100 U/ml of penicillin, and 100 U/ml of streptomycin (Invitrogen) at 37 °C in 5% CO₂. For the last 6 h, monensin (1.7 μg/ml; Biolegend) was added to the medium to inhibit cytokine secretion, and the cells were re-stimulated with LPS (100 ng/ml). Next, the cells were incubated with fixation/permeabilization solution (BD biosciences) at 4 °C for 20 min. After incubation, the cells were washed, stained with IFN-γ-PE and IL-10-APC for 15 min at room temperature and analyzed by flow cytometry.

In vitro, PBMCs (2 × 10⁵/well) were stimulated with LPS (100 ng/ml; Sigma-Aldrich, St. Louis, MO) for 72 h in the presence or absence of AH6809 (150 μM) or butaprost (50 μM; Sigma) or DMSO in 10% FBS RPMI medium. Next, the supernatants were collected and stored at − 80 °C immediately. Cytokines in the supernatants were measured using a multiplex panel (Bio-Rad, Hercules, CA) according to the manufacturer’s instructions. The limitations of detection have been mentioned above.

One hundred microliters of whole blood were incubated with AF488-conjugated Escherichia coli (K-12 strain) bio-particles (Molecular probe, Eugene, OR, USA) in 96-well plates and was analyzed by FACS according to previously described protocols [6]. For blocking experiments, the blood was preincubated with AH6809 or DMSO for 1 h at 37 °C in 5% CO₂.

One hundred microliters of whole blood samples were incubated with or without heat-inactivated E. coli (8.04*10⁷cfu/ml) or PMA (50 ng/ml) in 96-well plates for 30 min. Next, the cells were assessed for oxidative burst using an ROS assay kit (Genecopoeia, MD, USA) and were analyzed by FACS according to the manufacturer’s protocol. For blocking experiments, blood was pre-incubated with AH6809 or DMSO for 1 h at 37 °C in 5% CO₂.

The data are shown as the mean ± standard deviation (SD), mean ± standard error of the mean (SEM), median (range) or number (percentage). For comparisons between two independent groups, the Mann–Whitney U test was used. Comparisons between paired groups were performed by the Wilcoxon signed-rank test. Correlations between variables were calculated by the Spearman’s rank correlation test. P < 0.05 at two sides was considered statistically significant. All statistical analyses were performed using GraphPad Prism 6 (GraphPad Software).

We confirmed an elevation of various cytokines levels in HB-ACLF (Additional file 2: Figure S2a). Additionally, spontaneous ROS production in monocytes and neutrophils was significantly enhanced in HB-ACLF than in HC and CHB (Additional file 2: Figure S2bc). No differences were found between CHB and healthy controls. These results were partly consistent with those of a previous report [12].

It was reported that endotoxins are significantly elevated in patients with ACLF who can easily develop Gram-negative bacterial infection. These infections are severe and are associated with intense systemic inflammation and poor clinical outcomes [13]. To determine whether endotoxins or Gram-negative bacterial infection is involved in the elevated levels of plasma cytokines, we investigated the immune response of white blood cells to Gram-negative bacteria or LPS. Patients with ACLF showed a remarkable increase in the monocytic and neutrophil ROS production upon PMA and E. coli stimulation compared with those of the HC and CHB groups (Fig. 1a). There was a significant increase in the frequency of LPS-stimulated PBMCs producing IL-10 or IFN-γ in ACLF, but not in CHB, compared with healthy controls (Fig. 1b). As expected, in neutrophils and monocytes, TLR4 was significantly upregulated in ACLF only. However, CD16, TLR2 and CD11b were unexpectedly down regulated in patients with ACLF and AD than in patients with CHB and HC (Additional file 2: Figure S3). There was no difference in TLR4 expression on CD8⁺ T cells among the three groups (Additional file 2: Figure S3). Regarding antigen-presenting function, monocytes in ACLF exhibited lower expression of HLA-DR than both HC and CHB (Additional file 2: Figure S3).

To further demonstrate the hypersensitivity of blood immune cells in ACLF, we isolated peripheral monocytes from HC and differentiated them into macrophages by M-CSF. These differentiated macrophages incubated in ACLF plasma exhibited enhanced M1 polarization after LPS stimulation while their ability for M2 polarization was significantly suppressed after IL-4 stimulation compared with macrophages incubated in HC plasma (Additional file 2: Figure S4ab). Such a phenomenon was also seen in THP-1 cell lines (Additional file 2: Figure S4 cd).

Next, we assumed that immune inhibitory factors might be involved in the hyper-reaction of blood immune cells to endotoxins or E. coli. Thus, we investigated the expression of EP2, an immune modulatory molecule, on the blood innate and adaptive immune cells. As shown in Fig. 2b and Figure S5a, EP2 expression on CD8⁺ T cells were significantly lower in ACLF than in HC and CHB by frequency (p < 0.0001; p = 0.001), MFI (p < 0.0001, p = 0.045) and count (p < 0.0001; p = 0.002). In contrast, no difference of EP2 expression on CD8⁺ T cells was found between CHB and HC. There was a decrease of the count of EP2 positive CD4⁺ T cell and NK cells in ACLF than HC (p = 0.012) and CHB (p = 0.023) respectively (Additional file 2: Figure S5bc). However, EP2 expression on CD4⁺ T cell and NK cells by frequency and MFI showed no significant difference among the three groups (Additional file 2: Figure S5bc). On monocyte, the MFI of EP2 and the count of EP⁺ cells were higher in ACLF than in CHB (p = 0.005; p = 0.007) and HC (p = 0.013; p = 0.038), but no significant differences of the frequency of EP2⁺ monocyte were found between ACLF than HC (p = 0.16) (Additional file 2: Figure S5d). On neutrophils, only MFI of EP2 was increased in ACLF than HC (p = 0.0006) (Additional file 2: Figure S5e). EP2 expression on NKT cells did not differ among these groups by either frequency or MFI or count (Additional file 2: Figure S5f). We also found that plasma PGE2, the ligand of EP2, showed higher levels in ACLF than in HC and CHB (p = 0.0002; p = 0.033, Fig. 2c).

We further investigated the association of PGE2-EP2 expression with disease severity. Notably, patients with MOF showed decreased frequency of EP2-positive CD8⁺ T cells compared with patients without MOF (Fig. 3a), and ACLF patients who died or received transplantation during the 30-day follow-up had a lower PGE2 level (Fig. 3b). Moreover, a strong negative association was found between the frequency of EP2-positive CD8⁺ T cells and both MELD (r = − 0.41, p = 0.017, Fig. 3c) and CLIF-C scores (r = − 0.48, p = 0.004, Fig. 3c). In addition, the frequency of EP2-positive CD8⁺ T cells was significantly negatively associated with MIP-1β (r = − 0.66, p = 0.024, Fig. 3d) and IP-10 (r =  − 0.67, p = 0.028, Fig. 3d). CXCR3 is the receptor of IP-10. Interestingly, the frequency of CXCR3⁺ CD8⁺ T cells was significantly lower in ACLF than in HC, while CXCR3⁺ monocytes were both higher in CHB and ACLF than in HC, indicating that increased IP-10 in ACLF exerted its function through monocytes but not in CD8⁺ T cells (Additional file 2: Figure S6). Nevertheless, no significant correlation was found between PGE2-EP2 expression and the 90-day mortality (Additional file 2: Figure S7ab). Additionally, the plasma PGE2 level in ACLF showed no correlation with the development of MOF or prognostic score (Figure S7 cd).

All these data indicated an altered PGE2-EP2 axis might be involved in the progression of ACLF.

Then, we investigated whether MIP-1β or IP-10 decreased the frequency of EP2⁺CD8⁺T cells. After incubated with the combination of MIP-1β and IP-10 for 7 days, PBMC isolated from HC showed decreased frequency of EP2⁺CD8⁺T cells (Fig. 4). However, there was no change in the frequency of EP2⁺CD8⁺T cells when incubated with MIP-1β or IP-10 alone (Fig. 4).

To determine the function of EP2-positive immune cells, we compared the phenotypes of EP2-positive cells with those of EP2-negative immune cells. Among all three groups, EP2-positive neutrophils and monocytes showed higher TLR2 and TLR4 expression than EP2-negative ones (Fig. 5a, b), indicating a higher response of EP2-positive cells to TLR4 and TLR2 ligands, such as LPS. There was also a marked increase in HLA-DR and CD11b expression on EP2⁺ monocytes compared with that on EP2- (Fig. 5c, d). Additionally, IFN-γ and IL-10 production in response to LPS was stronger in EP2⁺ PBMCs than in EP2- (Fig. 5e). After whole blood was stimulated with E. coli in vitro, ROS production was enhanced in EP2⁺ monocytes and neutrophils than in EP2⁻ ones (Fig. 5f). Intriguingly, the difference in these phenotypes between EP2⁺ immune cells and EP2⁻ immune cells seemed always larger in the ACLF group than in the HC and CHB groups, suggesting an immune imbalance in ACLF. These phenomena suggested an immune-activation role of EP2-positive immune cells.

To explore the exact role of EP2 in immune cells in ACLF patients, we blocked EP2 and observed the change in cytokines, ROS production and phagocytosis. Blockade of EP2 on PBMCs from ACLF but not from CHB or HC could increase the secretion of IFN-γ and IL-10 (Fig. 6a). Moreover, this blockade also increased the secretion of many other cytokines, such as IL-6, TNF-α, and MCP-1 in ACLF (Fig. 6b). Additionally, ROS production in monocytes and neutrophils from ACLF patients was markedly enhanced after EP2 blockade in vitro (Fig. 6c). However, such blockade decreased monocytic phagocytosis in the ACLF group but not in the HC or CHB group (Fig. 6d). No significant change was seen in the phagocytosis of neutrophils (Fig. 6d).

To explore potential drugs for the treatment of ACLF, we employed butaprost, an EP2-selective agonist, to PBMCs isolated from patients with ACLF. There was a significant decrease in IL-2, IL-4, TNF-α, MIP-1β and FGF basic in the supernatants after EP2 was agonized under stimulation of LPS, indicating an anti-inflammatory role of butaprost in ACLF (Fig. 7). In the meantime, there was also a marked increase of G-CSF (Fig. 7), a cytokine that have been proved to promote liver regeneration [14].