CD14+ CD15− HLA-DR− myeloid-derived suppressor cells impair antimicrobial responses in patients with acute-on-chronic liver failure

Objective Immune paresis in patients with acute-on-chronic liver failure (ACLF) accounts for infection susceptibility and increased mortality. Immunosuppressive mononuclear CD14+HLA-DR− myeloid-derived suppressor cells (M-MDSCs) have recently been identified to quell antimicrobial responses in immune-mediated diseases. We sought to delineate the function and derivation of M-MDSC in patients with ACLF, and explore potential targets to augment antimicrobial responses. Design Patients with ACLF (n=41) were compared with healthy subjects (n=25) and patients with cirrhosis (n=22) or acute liver failure (n=30). CD14+CD15−CD11b+HLA-DR− cells were identified as per definition of M-MDSC and detailed immunophenotypic analyses were performed. Suppression of T cell activation was assessed by mixed lymphocyte reaction. Assessment of innate immune function included cytokine expression in response to Toll-like receptor (TLR-2, TLR-4 and TLR-9) stimulation and phagocytosis assays using flow cytometry and live cell imaging-based techniques. Results Circulating CD14+CD15−CD11b+HLA-DR− M-MDSCs were markedly expanded in patients with ACLF (55% of CD14+ cells). M-MDSC displayed immunosuppressive properties, significantly decreasing T cell proliferation (p=0.01), producing less tumour necrosis factor-alpha/interleukin-6 in response to TLR stimulation (all p<0.01), and reduced bacterial uptake of Escherichia coli (p<0.001). Persistently low expression of HLA-DR during disease evolution was linked to secondary infection and 28-day mortality. Recurrent TLR-2 and TLR-4 stimulation expanded M-MDSC in vitro. By contrast, TLR-3 agonism reconstituted HLA-DR expression and innate immune function ex vivo. Conclusion Immunosuppressive CD14+HLA-DR− M-MDSCs are expanded in patients with ACLF. They were depicted by suppressing T cell function, attenuated antimicrobial innate immune responses, linked to secondary infection, disease severity and prognosis. TLR-3 agonism reversed M-MDSC expansion and innate immune function and merits further evaluation as potential immunotherapeutic agent.


Hepatology InTrODuCTIOn
Acute-on-chronic liver failure (ACLF) is characterised as high morbidity and mortality due to profound activation of systemic inflammatory response syndrome (SIRS) and development of multiple-organ dysfunction. 1 Recent reports indicate that activation of SIRS responses in ACLF results in immune dysregulation and is postulated to be responsible for defective immune responses to microbial cues, termed immune paresis. 2 3 Immune paresis is associated with increased frequency of infectious events in cirrhosis 4 and frequently leads to acute decompensation, extrahepatic organ failure and increased mortality. 5 Development of secondary bacterial infections in patients with ACLF is associated with a 30-day mortality of 49% and is highly predictive of adverse outcome. 4 Over the previous decade, it has emerged that myeloid-derived suppressor cells (MDSCs) are elicited under various pathological conditions. MDSCs were initially defined in murine cells expressing CD11b and Gr-1 antigen, defined by their ability to suppress T cell proliferative and antitumour responses. However, human MDSCs are a heterogeneous population including a CD14 -CD11b + CD33 + CD15 + polymorphonuclear fraction (PMN-MDSC) and a CD11b+CD14 + HLA-DR low/mononuclear fraction (M-MDSC). 6-8 M-MDSCs have recently been identified in a wide number of hepatic (eg, chronic viral infection, hepatocellular cancer), [9][10][11] non-hepatic systemic (eg, sepsis), 12 13 organ-specific inflammatory diseases 14 15 and malignancies. 16 17 The increased frequency of this suppressive immune cell population has been associated with impaired T cell responses and serves as an immunological biomarker of disease severity and outcome.
Although the ability of MDSC to supress T cell responses has been extensively documented, their innate immune and antimicrobial responses remain relatively unexplored. Furthermore, while the existence of a HLA-DR low/− monocyte population has been described in acute hepatic inflammatory disorders, [18][19][20][21] little is known about their lineage, innate and adaptive immune function, and their candidacy as a potential immunotherapeutic target to restore antimicrobial responses in patients with ACLF.
We hypothesised that persistent exposure to pathogen-associated molecular patterns (PAMPs), activation of SIRS and the development of ACLF trigger the expansion of immunosuppressive M-MDSCs that serve to impair both innate and adaptive responses to microbial agents thereby contributing to the increased frequency of infections encountered in patients with ACLF. We therefore sought to delineate the presence and function of M-MDSC in patients with ACLF, and explore potential targets to modulate their function in order to augment antimicrobial responses.

MATerIAls AnD MeTHODs Patients and sampling
From January 2013 to June 2015, we consecutively recruited patients admitted to King's College Hospital for stable cirrhosis (n=22), ACLF (n=41) or acute liver failure (ALF; n=30). Patients with ACLF fulfilled the established diagnostic criteria developed by the European Association for the Study of the Liver-CLIF (Consortium on Chronic Liver Failure). 1 Cirrhosis was diagnosed by a previous liver biopsy or clinical presentation with typical ultrasound or CT imaging. Exclusion criteria were age younger than 18 years, malignancy and immunosuppressive therapy other than corticosteroids. Within 24 hours of admission to hospital, blood was sampled for ex vivo analysis of monocyte differentiation and function. Results were compared with healthy subjects (n=25). For patients with ALF and ACLF, sequential tests were done on days 3, 5, 7 and 14 after admission when feasible. Power calculations indicate that in order to detect significant differences in immune function and phenotype (eg, tumour necrosis factor-alpha (TNF-α) secretion) with 80% statistical power at the 5% significance level, a minimum of 28 patients would need to be recruited into the ACLF and ALF study groups.
Routinely assessed full blood count, international normalised ratio, liver and renal function tests, lactate and clinical variables were entered prospectively into a database. The following disease severity scores were calculated: Child-Pugh, Model for End-Stage Liver Disease (MELD), CLIF-SOFA (Sequential Organ Failure Assessment), 1 North American Consortium for Study of End-stage Liver Disease (NACSELD), 22 Acute Physiology and Chronic Health Evaluation II, Simplified Acute Physiology Score II, SOFA scores; infections and 28-day survival were documented. The study had been approved by the King's College Hospital Ethics Committee (12/LO/0167). Assent was obtained by the patients' nominated next of kin if they were unable to provide informed consent themselves.

CD14+ cell isolation
CD14+ monocytic cells were isolated from peripheral blood mononuclear cells (PBMCs) using the CD14 Microbeads or Pan-Monocyte Isolation Kit (Miltenyi Biotec, Bergisch Gladbach, Germany) as previously described. 23 Purity of monocytes was assessed by flow cytometry.

Mixed lymphocyte reaction
CD14+ monocytic cells from a healthy donor were isolated using the Pan-Monocyte Isolation Kit (Miltenyi Biotec) as previously described. 23 Cells were cultured in 25% plasma from patients with ACLF (n=8) or healthy controls (n=2) for 16 hours and cocultured with allogeneic CD3+ T cells from a different healthy donor, isolated using the Pan-T-cell Isolation Kit (Miltenyi Biotec) in a 1:1 ratio. T cell stimulation was induced with anti-CD2/CD3/CD28 beads (T cell Activation/ Expansion Kit; Miltenyi Biotec) according to the manufacturer's protocol (2.5×10 5 cells in 250 µL of Roswell Park Memorial Institute medium, 5% AB serum (Sigma-Aldrich, Gillingham, Dorset, UK), 1% penicillin/streptomycin). Cells were stained with carboxyfluorescein succinimidyl ester at day 0. Proliferation was assessed at day 3 of coculture. Experiments were performed in triplicate.

ex vivo phagocytosis assay in PbMC
For the assessment of phagocytosis of PBMCs in vitro, 500 000 healthy PBMCs per well were cultured (24 hours) in complete medium containing 20% healthy controls or ACLF plasma. Cells were stimulated with various agents (polyI:C 10 µg/mL, LPS 1 and 10 µg/mL, CpG-ODN2006 1 and 10 µg/mL) for 4 hours. Harvested cells were supplemented by 10% autologous plasma and pHrodo E. coli Red BioParticles (Invitrogen) were added for 60 min (see online supplementary methods).

Phagocytosis of CD14+ cells: Cell-IQ
CD14+ cells were isolated and cultured (24 hours) in complete media with 25% plasma from two healthy subjects and two patients with ACLF. Cells were supplemented with 10% human AB serum and pHrodo E. coli Red BioParticles (Invitrogen).

Quantification of bacterial 16s rDnA levels in whole blood using TaqMan qrT-PCr
The procedure was carried out under strict aseptic conditions as previously reported 24 (see online supplementary methods).

In vitro models for the generation of M-MDsC-like cells
The models were adapted from Pena et al. 25 9×10 6 PBMCs per well were cultured on a 12-well plate in 2000 µL X-Vivo medium (Lonza, Basel, Switzerland) in a 37°C, 5% CO 2 environment. Cells were stimulated with or without LPS 10 ng/mL, Pam3CSK4 5 µg/mL (Invivogen) or a combination of LPS and Pam3CSK4, respectively, for 24 hours. After this time, cells were rechallenged with LPS or Pam3CSK4 for a further 4 hours, and then subjected to diverse experiments such as intracellular staining of cytokine production, immunophenotyping, phagocytosis and viability assays as described above.
For the plasma experiments, CD14+ cells were cultured (16 hours) in plasma from healthy subjects (n=3) and patients with cirrhosis (n=3) and ACLF (n=6), respectively. Subsequently, the M-MDSC subset was identified and phenotyped as described above. Parts of the cells were transferred to fresh medium, stimulated with LPS 100 µg/mL (5 hours) and supernatants were used for S100A8/A9 ELISA (Systems).

Phagocytosis PCr array
The RT 2 Profiler phagocytosis PCR array (Cat No 330231 PAHS-173ZA) was purchased from SABiosciences, Quiagen, Manchester, UK. The process from RNA extraction to cDNA synthesis and quantitative RT-PCR was performed exactly as per manufacturer's protocol (see online supplementary methods).

statistical analyses
Data are expressed as the median/IQR unless otherwise specified. For data that did not follow a normal distribution, the significance of differences was tested using the Mann-Whitney or Wilcoxon tests; Spearman correlation coefficients and area under the receiver operating characteristic curve (AUROC) were calculated. Graphs were drawn using Prism 7.0a (GraphPad, La Jolla, California).

Patient characteristics
In comparison to patients with ACLF (n=41), patients with stable cirrhosis (n=22) and ALF (n=30) were included. Clinical characteristics of the different groups including disease stratification scores, the evidence of infection and prognosis are summarised in table 1. A 28-day transplant-free survival was 39% in the ACLF group and 56.6% in the ALF group, while 17.1% and 23.3% were transplanted, 43.9% and 20% died, respectively.

Hepatology Circulating M-MDsCs in AClF have impaired innate responses to microbial challenge and refer to poor prognosis and infection susceptibility
Compared with healthy subjects, CD14+ cells from patients with ACLF exhibit attenuated proinflammatory responses (TNF-α/IL-6) to TLR-2 (Pam3CSK4), TLR-4 (LPS) and TLR-9 (CpG) challenge indicating reduced innate responses to several pathogen-induced TLR signalling pathways (see online supplementary figure 1). Compared with classical CD14+HLA-DR+ monocytes, M-MDSCs in ACLF were characterised by defective proinflammatory cytokine production in response to all different TLR ligands (figure 2A). Furthermore, S100A8/A9 protein levels, an inflammatory mediator synthesised and secreted preferentially by M-MDSC, 26 were significantly higher in the plasma of patients with ACLF compared with healthy subjects and patients with cirrhosis ( figure 2B).
Median HLA-DR expression over the first 14 days following admission predicted 28-day survival with a sensitivity of 74% and a specificity of 77% for the criterion MFI >1617 and the onset of infection within 28 days of hospitalisation with a sensitivity of 75% and a specificity of 67% for the criterion MFI <1500. Furthermore, it predicted both the presence of primary infection (AUROC 0.82/p=0.04, sensitivity 75%, specificity 71%, criterion MFI <1346; data not shown) and the development of secondary infection (AUROC 0.73/p=0.03, sensitivity 67%, specificity 67%, criterion MFI <1488) (figure 2E). These observations corroborate previous reports linking HLA-DR expression on monocytes with adverse prognosis in patients with ACLF 20 21 and newly suggest an association between the persistence of the M-MDSC population during the course of disease and infection susceptibility.

M-MDsCs display impaired bacterial uptake and clearance in AClF
Phagocytosis of bacteria is an important function of myeloid cells with regard to effective clearance of invading microorganisms and during sepsis. The proportion of CD14+ cells able to phagocytose E. coli particles was significantly reduced in patients with ACLF (median 97.4%) and ALF (median 82.4%; both p<0.0001, figure 3A,B) when compared with patients with cirrhosis and healthy controls (median 99%). Also, capacity to phagocytose S. aureus particles was reduced in both patients with ACLF and ALF and interestingly, the capacity to phagocytose gram-negative (E. coli) particles correlated with capacity to phagocytose gram-positive (S. aureus) particles (see online supplementary figure 2A). The production of reactive oxidative metabolites following uptake of E. coli was preserved in both patients with ACLF and ALF compared with healthy subjects (see online supplementary figure 2B).
A phagocytosis assay was designed to differentially assess phagocytosis capabilities of CD14+HLA-DR+ monocytes compared with CD14 + HLA-DR −

Transcriptional analysis of circulating monocytes links impaired phagocytosis to downregulation of Tlr pathways
In order dissect which intracellular pathways may be involved in defective bacterial clearance and innate responses in ACLF, a phagocytosis-specific RT-PCR array was performed comparing mRNA extracts from CD14+ cells of patients with ACLF bearing an expanded M-MDSC population compared with healthy subjects (figure 4A). Differential expression more than twofold revealed higher expression of three genes (FcγR1/CD64, MERTK and MFGE8) and reduced expression of 16 genes interestingly including TLR-3 and TLR-9, CD14 and interferon-γ. The findings indicate multiple defective pathways of innate immune responses in CD14+ cells from patients with ACLF with marked reductions in   responses and therefore represent a potential therapeutic target.

Generation of the immunosuppressive M-MDsC population in patients with AClF may result from circulating bacterial products stimulating Tlr pathways and circulating cytokines
In addition to impaired pathogen responses and clearance mechanisms, we also detect elevated levels of bacterial DNA in whole blood from patients with ACLF compared with cirrhosis and in cirrhosis compared with healthy subjects. Furthermore, bacterial DNA titres negatively correlated with the expansion of CD14 + HLA-DR − M-MDSCs (figure 5A) and with a number of indices of disease severity (figure 5B). We therefore hypothesised that continual exposure to PAMPs gives rise to the immunosuppressive M-MDSC with impaired antimicrobial responses as described in vivo. In an in vitro model administering recurrent LPS stimulation 25  In addition to PAMPs, soluble factors such as cytokines have been reported to induce M-MDSCs. We and others had previously shown that numerous cytokines (TNF-α, IL-6, IL-10, IL-8) are upregulated in ACLF in comparison to cirrhosis and healthy subjects, respectively. 23 27 We therefore studied the effect of ACLF plasma on healthy CD14+ monocytes in vitro and indeed observed the expansion of the M-MDSC population revealing a CD16 low MERTK low phenotype. Also, the secretion of S100A8/A9 protein levels significantly increased in the cell culture supernatants of CD14+ cells after incubation with ACLF plasma. Similar to data reported above, the viability of generated M-MDSC was not reduced compared with CD14+HLA-DR+ monocytes. The data implicate the likely involvement of other soluble mediators such as proinflammatory cytokines causing the generation of M-MDSC in patients with ACLF ( figure 5E).

In vitro Tlr-3 stimulation reduces the frequency of M-MDsCs and augments antimicrobial responses in AClF
The downregulated expression of TLRs raised the question whether regulation of innate immune function by TLRs was attenuated, and whether targeted TLR agonists may have the potential to reduce the frequency of M-MDSC. Previous reports and studies in man have identified the therapeutic use of TLR-3 agonism therapy as an adjuvant for vaccinations as well as for induction of antitumour immune responses by decreasing the frequency of MDSC. [28][29][30][31] Moreover, studies demonstrated that TLR-3 agonist polyI:C promoted phagocytosis of bacteria in vitro. 32 To explore the effect of different TLR agonists on innate immune responses, we incubated healthy CD14+ cells in plasma from patients with ACLF in the presence or absence of TLR agonists (TLR-3, TLR-4, TLR-9). Interestingly, TLR-3 agonist polyI:C significantly reduced the proportions of M-MDSC population while simultaneously enhancing their phagocytosis capacity in ACLF ( figure 6A,B). Conversely, treatment with TLR-4 and TLR-9 agonists or NFκ-B inhibitor N-acetylcysteine did not reduce or in fact further increased proportions of M-MDSC and further impaired bacterial phagocytosis ( figure 6C).

DIsCussIOn
In this study, we have shown a marked expansion of M-MDSCs 6 8 that impair antimicrobial responses in patients with ACLF. Their biological relevance is highlighted through their strong association with indices of disease severity and incidence of infectious complications in these patients. In addition to impairing T cell responses, M-MDSCs are typified by impaired (1) secretion of proinflammatory cytokines in response to stimulation with a wide range of TLR ligands and (2) phagocytosis of bacteria. Analyses of these cells reveal a distinct immunophenotype characterised by reduced expression of tissue scavenger, costimulatory and phagocytosis receptors (eg, MERTK, CD64, CD86, CD163).
M-MDSCs are not only elevated in patients where primary infection has been the trigger for the development of ACLF but their persistence is also linked with the development of secondary infections. These data echo data from other inflammatory pathologies where the expansion of these cells 6 8 correlates with poor prognosis. 9 10 13-19 33-35 These observations urge to prospectively evaluate the role of M-MDSCs as biomarkers of infection and survival in large cohorts of patients with cirrhosis and ACLF.
Our data also reveal a less pronounced but significant elevation in M-MDSC in patients with cirrhosis. Here, we suggest that they are likely to evolve as a consequence of continual exposure to elevated concentrations of gut-derived PAMPs and proinflammatory mediators that are probably implicated in disease progression. Further studies are required to detail how these cells evolve and their role and migratory patterns in the liver and other tissue sites.

Hepatology
The generation of M-MDSC involves a complex interplay of numerous soluble circulating factors. 36 37 Using in vitro models, we were able to recapitulate the M-MDSC phenotype with comparable defects of innate immune function, suggesting that the generation and persistence of M-MDSC in ACLF could be related to repetitive stimulation by elevated circulating titres of PAMPs. This hypothesis is further strengthened by our observation of increased bacterial DNA levels in patients with ACLF, which correlated with the frequency of M-MDSCs. Among TLRs, TLR-2/6 38 and TLR-4 39-41 ligands have been reported to promote MDSC population. Consistent with these findings, our data reveal that recurrent stimulation of TLR-2 and/or TLR-4 induces a similar suppressive M-MDSC-like population as described ex vivo.
Given that both ALF and ACLF are conditions characterised by a profound activation of SIRS responses, in addition to continual PAMP exposure, we also propose that increases in circulating inflammatory mediators also contribute to the expansion and differentiation of the M-MDSCs. Our in vitro data support this concept where we demonstrate that incubation of CD14+ The transcriptional profile of CD14+ myeloid cells in ACLF suggests these cells are functionally reprogrammed to promote uptake of apoptotic cells while concomitantly suppressing T cell activation, TLR-triggered proinflammatory responses and phagocytosis of bacteria. A similar transcriptional profile has been recently reported in patients with septic shock where the leucocyte transcriptome is reprogrammed towards resolution of tissue injury and with concomitant downregulation of genes that prime innate and adaptive antimicrobial responses. 42 Further evidence to support this theory is provided in experimental lung injury models where, following clearance of necrotic/apoptotic cells, macrophages have an attenuated uptake of bacteria and augmented secretion of anti-inflammatory mediators. 43 A similar mechanism may explain how these immunological defects evolve in ACLF, where the host response to severe tissue injury and organ dysfunction is skewed towards tissue repair responses to the detriment of antimicrobial programmes. We therefore hypothesised that selective activation of TLR pathways may represent an immunotherapeutic strategy in order to reverse M-MDSC expansion and improve antimicrobial responses in ACLF.
Deng et al had observed that TLR-3 agonist polyI:C promoted bacterial uptake in murine peritoneal macrophages via a TRIF-IRF-3-mediated mechanism. 32 TLR-3 is classically activated by double-stranded RNA from viruses, but may also be released from degraded bacteria and necrotic cells. 44 45 In experimental models, TLR-3 activation was associated with an increased innate immune cell infiltrate and enhanced infection clearance. 46 47 In tumour models, TLR-3 activation reverses the expansion and immunosuppressive function of MDSCs 29 30 and has been used as an immunotherapeutic agent to promote tumouricidal responses. 29 48 49 Consistent with these observations, TLR-3 reduced the frequency of M-MDSC and augments innate immune responses and pathogen uptake in patients with ACLF. These findings may relate to the fact that TLR-3 differentially regulates phagocytosis and may 'rebalance' the phagocytosis processes which are skewed predominantly towards apoptotic cell clearance in ACLF to the detriment of antimicrobial responses. Given that TLR-3 activation has been shown to be protective in experimental models of chronic liver injury, a TLR-3 agonist may improve antibacterial responses without having a detrimental effect on tissue repair processes. 32 50 Further work using in vitro and in vivo models of disease is required to further understand how activation of the TLR-3 pathway modulates inflammatory and antimicrobial responses before considering its use in human studies.
Although M-MDSCs correlate with indices of disease severity, incidence of infectious complications, the data presented in this study strongly suggest but cannot conclusively prove that the increased frequencies of M-MDSC are responsible for reduced pathogen clearance in vivo. Moreover, the immune read-outs developed here do not represent point-of care testing that could be introduced for clinical use in multicentre cohort studies. Future prospective work is therefore required to further evaluate Hepatology the utility of M-MDSC as a predictive biomarker of infection susceptibility and an immunotherapeutic target.
In conclusion, we describe a marked expansion of M-MDSC in ACLF. These cells are characterised by attenuated immune responses to pathogens, defective pathogen clearance mechanisms and a higher incidence of overt infectious complications. Perpetual exposure to elevated concentrations of PAMPs and inflammatory cytokines involved in SIRS response play a pathogenic role in the expansion of this immunosuppressive population. In vitro data indicate that TLR-3 agonism may reduce the frequency of M-MDSC and therefore merit further evaluation as potential novel immunotherapeutic strategy to restore antimicrobial responses and reduce infectious complications in patients with ACLF.