Patterns of dendritic cell and monocyte subsets are associated with disease severity and mortality in liver cirrhosis patients

Liver cirrhosis is often complicated by an immunological imbalance known as cirrhosis-associated immune dysfunction. This study aimed to investigate disturbances in circulating monocytes and dendritic cells in patients with acute decompensation (AD) of cirrhosis. The sample included 39 adult cirrhotic patients hospitalized for AD, 29 patients with stable cirrhosis (SC), and 30 healthy controls (CTR). Flow cytometry was used to analyze monocyte and dendritic cell subsets in whole blood and quantify cytokines in plasma samples. Cirrhotic groups showed higher frequencies of intermediate monocytes (iMo) than CTR. AD patients had lower percentages of nonclassical monocytes than CTR and SC. Cirrhotic patients had a profound reduction in absolute and relative dendritic cell numbers compared with CTR and showed higher plasmacytoid/classical dendritic cell ratios. Increased plasma levels of IL-6, IL-10, and IL-17A, elevated percentages of CD62L+ monocytes, and reduced HLA-DR expression on classical monocytes (cMo) were also observed in cirrhotic patients. Patients with more advanced liver disease showed increased cMo and reduced tissue macrophages (TiMas) frequencies. It was found that cMo percentages greater than 90.0% within the monocyte compartment and iMo and TiMas percentages lower than 5.7% and 8.6%, respectively, were associated with increased 90-day mortality. Monocytes and dendritic cells are deeply altered in cirrhotic patients, and subset profiles differ between stable and advanced liver disease. High cMo and low TiMas frequencies may be useful biomarkers of disease severity and mortality in liver cirrhosis.


Liver cirrhosis is often complicated by an immunological imbalance known as cirrhosis-associated immune dysfunction. This study aimed to investigate disturbances in circulating monocytes and dendritic cells in patients with acute decompensation (AD) of cirrhosis. The sample included 39 adult cirrhotic patients hospitalized for AD, 29 patients with stable cirrhosis (SC), and 30 healthy controls (CTR).
Flow cytometry was used to analyze monocyte and dendritic cell subsets in whole blood and quantify cytokines in plasma samples. Cirrhotic groups showed higher frequencies of intermediate monocytes (iMo) than CTR. AD patients had lower percentages of nonclassical monocytes than CTR and SC. Cirrhotic patients had a profound reduction in absolute and relative dendritic cell numbers compared with CTR and showed higher plasmacytoid/classical dendritic cell ratios. Increased plasma levels of IL-6, IL-10, and IL-17A, elevated percentages of CD62L + monocytes, and reduced HLA-DR expression on classical monocytes (cMo) were also observed in cirrhotic patients. Patients with more advanced liver disease showed increased cMo and reduced tissue macrophages (TiMas) frequencies. It was found that cMo percentages greater than 90.0% within the monocyte compartment and iMo and TiMas percentages lower than 5.7% and 8.6%, respectively, were associated with increased 90-day mortality. Monocytes and dendritic cells are deeply altered in cirrhotic patients, and subset profiles differ between stable and advanced liver disease. High cMo and low TiMas frequencies may be useful biomarkers of disease severity and mortality in liver cirrhosis. Monocyte and DC subsets are altered in liver cirrhosis patients compared with healthy controls. Cirrhotic patients had a higher percentage of monocytes than CTR, but this difference was not apparent in absolute numbers (Fig. 1a). Significant differences were observed in monocyte subset (cMo, iMo, and ncMo) frequencies. As expected, cMo was the predominant subset in CTR patients, followed by ncMo and iMo. Both cirrhotic groups had higher frequencies of iMo (CD14 + CD16 + ) than CTR. The AD group had a lower frequency of ncMo than CTR and SC (p ≤ 0.01). No significant differences in cMo percentages within the monocyte compartment were observed between groups. However, when evaluating the proportion of cMo in relation to leukocytes, we found that AD patients had higher cMo percentages than CTR and SC individuals (p ≤ 0.05). Cirrhotic patients, particularly those with AD, showed a substantial reduction in absolute and relative numbers of circulating DCs compared with CTR. Significant differences were found between SC and AD groups, as depicted in Fig. 1b. Interestingly, when considering only the DC compartment, we observed that AD patients showed reduced cDC and increased pDC percentages, resulting in an increased pDC/cDC ratio. Table S2 (Supplementary material) shows monocyte and DC subsets frequencies of the three groups evaluated in this study. Monocyte surface expression of HLA-DR was investigated. For this, we measured the MFI of HLA-DR on cMo, iMo, ncMo, and TiMas and calculated the TiMas/cMo HLA-DR ratio. No differences in iMo, ncMo, or TiMas HLA-DR expression were observed between groups. However, AD patients showed a lower MFI of cMo HLA-DR than CTR individuals (p ≤ 0.01) and a higher TiMas/cMo HLA-DR ratio than the other groups ( Fig. 2  a,b). Monocyte expression of the adhesion molecule CD62L was also assessed. SC and AD showed higher percentages of CD62L + monocytes than CTR (p ≤ 0.05 and p ≤ 0.01, respectively) ( Fig. 2 a,b).

Cytokine measurements in liver cirrhosis patients.
Cytokine measurements demonstrated that IL-6 was very elevated in cirrhotic patients, particularly in AD individuals (Fig. 3a. Similar results were observed for IL-10 but without differences between cirrhotic groups (Fig. 3b). IL-17A was found to be elevated in SC patients (Fig. 3 c,d). Noteworthy, this cytokine was high in SC patients with previous clinical complications or hospitalization for liver disease (SCD) in comparison with SC patients with compensated cirrhosis (CC) (p ≤ 0.05). In fact, this was the only parameter that differed significantly when the SC group was compared according to these features.

Monocyte and DC alterations are associated with cirrhosis progression and clinical complications.
Spearman's correlation coefficient was used to identify correlations between numerical variables of interest. DC and monocyte frequencies were compared according to MELD and Child-Pugh scores (Table S3, The results demonstrated that patients with more advanced disease have higher cMo frequencies, lower TiMas frequencies, and lower DC frequencies. A moderate positive correlation was found between TiMas/cMo HLA-DR ratio and MELD score. Regarding the DC compartment, high pDC frequency and pDC/cDC ratio correlated moderately and positively with Child-Pugh score. IL-6 correlated positively with MELD and Child-Pugh scores, whereas IL-17A correlated negatively with the scores. MELD scores were categorized into three groups (≤ 9, 10-19, and ≥ 20) and Child-Pugh scores into three classes (A, B, and C). Patients with advanced liver disease showed altered distribution of monocyte subsets, lower HLA-DR expression on cMo, lower DC frequencies, and higher pDC proportion within the DC compartment  We also compared cirrhotic patients according to the presence of ascites and HE (Table 2). Ascitic patients showed higher cMo and lower TiMas frequencies within the monocyte compartment as well as lower DC frequencies. Similar results were observed for patients with HE.
Factors associated with ACLF are shown in Table S4 (Supplementary material). ACLF was associated with serum creatinine, C-reactive protein, alanine aminotransferase, IL-10, HE, MELD score, and iMo and DC percentages.
Association between monocyte percentages and mortality. Six patients (15%) died within the first 90 days, all of them belonging to the AD group. Univariate Cox regression analysis was performed to investigate factors associated with 90-day mortality (Table S5, Supplementary material). Bacterial infection, Child-Pugh C classification, serum C-reactive protein, IL-6, percentages of cMo, iMo, and TiMas within the monocyte compartment, percentage of CD62L + monocytes, and pDC/cDC ratio were associated with 90-day mortality. Multivariate Cox regression was not carried out because of the low number of events (six deaths).
The AUROC of cMo percentage (within the monocyte compartment) for predicting 90-day mortality was 0.869 ± 0.078 (p = 0.004, 95% CI 0.716-1.000), and the best cutoff was 90%. At this cutoff, cMo frequency showed a sensitivity of 83.3%, specificity of 75.8%, positive predictive value of 38.5%, and negative predictive value of 96.2% for 90 day-mortality. Figure 5a exhibits the Kaplan-Meier curve for 90-day mortality according to cMo frequency at a cutoff of 90%. The Kaplan-Meier survival probability was 96.2% in subjects with cMo frequencies < 90.0% and 61.5% in those with values > 90.0% (p = 0.003).
The AUROC of iMo percentage (within the monocyte compartment) for estimating 90-day mortality was 0.944 ± 0.037 (p = 0.001, 95% CI 0.872-1.000), and the best cutoff was 5.7%. At this cutoff, iMo frequency showed a sensitivity of 100%, specificity of 87.9%, positive predictive value of 60.0%, and negative predictive value of 100%. The Kaplan-Meier curve for 90-day mortality based on iMo frequency at a cutoff of 5.7% is depicted in Fig. 5b. The Kaplan-Meier survival probability was 100% in subjects with iMo frequencies > 5.7% and 60% in those with values ≤ 5.7% (p < 0.001).
The AUROC of CD62L + monocytes and pDC/cDC ratio did not show significant results.

Discussion
In patients with inflammatory liver disease, monocytes are recruited from the blood and then differentiate into distinct functional subsets of macrophages and DCs that regulate inflammation, fibrogenesis, and resolution 33 .
In these cases, the phenotype of circulating monocytes may be a valuable indicator of hepatic recruitment and systemic inflammatory state 31 . Previous studies observed an increase in circulating and hepatic CD16 + monocytes in cirrhotic patients; these cells are believed to play a role in the perpetuation of inflammatory and profibrogenic signals during liver disease progression 18,31,34 . In this study, we observed an increase of circulating iMo (CD14 + CD16 + ) in cirrhotic patients compared with healthy CTR individuals. In addition, we detected alterations in the proportions of circulating monocytes, particularly in patients with more advanced liver disease. These findings suggest intensification of monocyte recruitment from bone marrow to peripheral blood, which results in increased circulating cMo frequency. Depending on the stimuli received, cMo differentiate into iMo, although it is not yet known whether differentiation occurs in circulation or in the inflamed tissue. iMo cells differentiate into ncMo, and, after acting on the tissue, return to peripheral blood 3 . We observed that circulating ncMo levels were reduced in AD patients in comparison with CTR. CD62L, an adhesion molecule belonging to the selectin family, is expressed by cells for endothelial adherence and tissue migration 35 . Cirrhotic patients had higher percentages of monocytes expressing CD62L than CTR, as also observed in a previous study 31 . This finding suggests intensification of recruitment and migration of monocytes from peripheral blood to inflammatory sites.
HLA-DR is a class II human leukocyte antigen typically expressed in antigen-presenting cells. In this study, HLA-DR expression on cMo was reduced. Such a reduction has been observed in cirrhotic patients 31,36,37 , and www.nature.com/scientificreports/ is a well-known characteristic of ACLF and septic shock 38 . Furthermore, reduced HLA-DR expression, combined with enhanced IL-6 and IL-10 and decreased IL-1, TNF-a, and nitric oxide synthase, is typical of immune paralysis 16 . The results of the present study indicate that a high TiMas/cMo HLA-DR ratio occurs because of a reduction in HLA-DR expression on cMo and not because of increased expression on iMo or ncMo, which did not differ significantly between groups. As previously mentioned, when compared with CTR, cirrhotic patients showed high iMo percentages. However, in comparing between cirrhotic patients, we found that those with more advanced disease showed higher cMo and lower iMo and ncMo percentages. Moreover, we found that lower frequencies of iMo and TiMas and higher frequencies of cMo within the monocyte compartment (cutoffs of ≤ 5.7%, < 8.6%, and ≥ 90%, respectively) were associated with mortality. Monocytes are crucial mediators of host defense, and infections are responsible for much of the morbidity and mortality in patients with AD of cirrhosis. This could indicate that the occurrence of CAID in the last stages of liver disease, when the immune system collapses, is marked by the disappearance of TiMas in peripheral blood. These unprecedented findings show that circulating monocyte subsets could represent useful biomarkers in clinical practice.
The current paradigm suggests that pDCs and cDCs have a common myeloid progenitor in the bone marrow, known as monocyte-DC progenitor (MDP), which can differentiate into DC-committed progenitors, giving rise to cDCs and pDCs 11,39 . DCs play a key role in the induction of innate and adaptive immune responses against specific antigens; however, its role in the initiation and progression of liver diseases is poorly elucidated. DCs are thought to promote resolution rather than progression of fibrosis 40 .
CAID is characterized by impairment of phagocytosis and antigen presentation, which might be associated with reduced frequencies or impaired function of circulating DCs 14 . Alterations in DCs, such as numerical reduction, impaired production of inflammatory cytokines, and increased production of immunosuppressive IL-10, have been described in HCV infection 21,22 . Impaired maturation and impaired expression of specific immune checkpoints and TLR molecules were detected in circulating DCs of patients with chronic HBV infection 23,41 . Ouaguia and colleagues found that circulating DCs are reduced in frequency and absolute numbers in patients chronically infected with HBV. The authors suggested the occurrence of a specific mechanism for active recruitment of circulating pDCs to HBV-infected liver 23 .
In the current study, we observed a profound reduction in both absolute and relative numbers of circulating DCs in cirrhotic patients, especially in those with signs of advanced liver disease. The proportion of pDCs within the DC compartment was higher in cirrhotic patients, mainly in the AD group, resulting in an augmented pDC/cDC ratio. Despite the proinflammatory role of pDCs (exerted by secretion of type I IFN), these cells were associated with induction of immune tolerance and immunosuppression. In transplant patients, high pDC/cDC ratios were associated with lower rejection rates [42][43][44] . www.nature.com/scientificreports/ The cytokine profile observed in this study-elevated plasma levels of IL-6 and IL-10, particularly in AD patients, and elevated plasma levels of IL-17A, particularly in SC patients-corroborates the results of a previous study carried out by our research group, which analyzed a cohort of 118 SC patients, 130 patients hospitalized for AD cirrhosis, and 30 healthy controls 45 . It has been reported that high levels of Th2-related cytokines, such as IL-10, are associated with chronic HCV infections 46 . In vitro tests showed that cMo isolated from peripheral blood differentiate into iMo after stimulation with TGF-β and IL-10 47 , indicating that IL-10 could be related to the increased numbers of circulating iMo in cirrhotic patients compared with healthy individuals. In accordance with a previous study 45 , here, IL-6 levels were found to be elevated in cirrhotic patients, mainly in those with advanced liver disease. IL-6 and IL-4 are associated with Th cell differentiation into Th2 and deeply involved in monocyte activation and chemotaxis toward the inflamed region 46 . IL-17A is a highly versatile proinflammatory cytokine produced by Th17 cells. The significance of circulating IL-17A has not been completely elucidated, particularly in the case of patients with hepatic diseases. A possible explanation for the low IL-17A levels in AD patients as compared with SC patients is the inhibitory effect of high IL-10 levels on Th17 cells 48 .
Overall, our findings indicate the presence of systemic effects induced by liver cirrhosis that influence the immune-hematopoietic system. CAID profile varied according to disease severity, as evidenced by phenotypic differences between cirrhotic patients at different stages of liver disease. The results revealed alterations in circulating monocyte levels and their differentiation to TiMas, as well as an important reduction of circulating DCs and increased pDC/cDC ratio in cirrhotic patients. A greater understanding of the immunological aspects of liver cirrhosis may contribute to the development of improved therapeutic approaches and facilitate detection of patients that could benefit most from immunotherapy.

Methods
Patients and controls. This prospective cohort study included consecutive adult cirrhotic patients hospitalized for acute decompensation (AD) at the University Hospital of the Federal University of Santa Catarina, Brazil, between March 2018 and December 2018. In addition to hospitalized patients, 29 adult cirrhotic patients attending the outpatient clinic at the same hospital were included as the stable cirrhosis (SC) group. Cirrhosis diagnosis was established histologically or by a combination of clinical, imaging, and laboratory findings in patients with evidence of portal hypertension.
Patients with hepatocellular carcinoma, receiving immunosuppressive therapy, seropositive for human immunodeficiency virus (HIV), with active Epstein-Barr virus or cytomegalovirus infection, with a history of neoplasm (except skin cancer) or cirrhosis of autoimmune etiology were excluded from the study. A questionnaire was used to collect information not available from medical records. Laboratory results were obtained for SC and AD cohorts.
Thirty age-and gender-matched healthy subjects undergoing routine health examination were selected as the control (CTR) group, using AD group as reference. Healthy individuals were interviewed using a questionnaire. Exclusion criteria included history of hepatic disease, history of neoplasm (except skin cancer), infections in the previous seven days, heavy alcohol consumption, diabetes mellitus, autoimmune disease, chronic inflammatory disease, and use of immunosuppressive therapy.
Written informed consent was obtained from all participants. The study was approved by the Ethics Committee of the Federal University of Santa Catarina (CEPSH UFSC protocol no. 1822/2011).
Correlations between laboratory and sociodemographic variables were investigated. Child-Pugh 26 and Model for End-Stage Liver Disease (MELD) scores 27,28 were used to assess the severity of hepatic disease. Acute-onchronic liver failure (ACLF) was defined as proposed by the EASL-CLIF Consortium 29 .
Peripheral blood was drawn into 4 mL vacutainers (Vacuette, Greiner Bio-One, Austria) containing ethylenediaminetetraacetic acid (EDTA) for anticoagulation. Samples were obtained at routine office visits for SC patients and on the first day of hospitalization for the AD cohort. Blood from all individuals was processed within 24 h of collection for flow cytometric analysis and plasma separation. Plasma was stored at − 80 °C until cytokine analysis.  Table S1 (Supplementary material).

Immunophenotyping of monocytes and
Peripheral blood-EDTA samples (300 µL) were stained with an eight-color panel previously designed and tested by our group to identify monocyte and DC subsets. CD62L expression was analyzed by using a different panel, which also contained antibodies for monocyte identification (HLA-DR, CD4, and CD45). Detailed information on staining procedures and acquisition parameters can be found in our previous study 30 .
Data acquisition was performed using a three-laser FACSCanto II flow cytometer equipped with FACSDiva software version 6 (BD Biosciences, USA). For homogeneous cell analysis with good sensitivity, 500,000 to 1,000,000 gated CD45 + events were recorded for each individual. Data were analyzed using Infinicyt version 1.7.0 (Cytognos, Spain) and the web-based application Cytobank 7.3.0 (Beckman Coulter, USA) (Fig. 6). The gating strategy was adapted from Cardoso and Santos-Silva 30 and previous studies 20, 31,32 . Cutoff limits were defined based on fluorescence-minus-one (FMO), negative, and positive controls. Automatic gates, applied using the analysis strategy tool, were supervised by two operators during analysis. Absolute numbers were calculated by a dual-platform flow cytometry method using a XE 2100 blood cell analyzer (Sysmex, Japan). performed by determining the mean fluorescence intensity (MFI) of this marker on cMo, iMo, ncMo, and TiMas. MFI values were then compared between groups. The MFI ratio of TiMas HLA-DR to cMo HLA-DR (hereafter referred to as TiMas/cMo HLA-DR ratio) was calculated to reduce the variability associated with antibody batches and equipment conditions. The ratio is expected to be positive, as ncMo and iMo have greater HLA-DR expression than cMo. Because CD62L expression is bimodal, we chose to report the values of monocytes expressing this molecule as a percentage. HLA-DR and CD62L expression was also analyzed by the tool heatmap of the web-based application Cytobank 7.3.0 (Beckman Coulter, USA).
Statistical analysis. The normality of data distribution was assessed using Shapiro-Wilk test. Continuous variables were compared using Student's t-test or analysis of variance (ANOVA) with Post-hoc Bonferroni in the case of normal distribution and Mann-Whitney U or Kruskal-Wallis tests in the remaining cases. Categorical variables were evaluated with the chi-square test. Correlations between numerical variables were evaluated using Spearman's correlation coefficient. Univariate and multivariate Cox regression analyses were used to evaluate the association between variables of interest and survival. The best cutoffs of monocytes for predicting mortality were chosen on the basis of the area under receiver operating characteristic (AUROC) curves. Survival curves were calculated by the Kaplan-Meier method, and survival differences between groups were compared using a log-rank test. P-values of equal or less than 0.05 were considered statistically significant. Statistical analysis were performed using SPSS software version 17.0 (SPSS, Chicago, IL, USA).