Article Text

Download PDFPDF

Original research
Targeting monocyte-intrinsic enhancer reprogramming improves immunotherapy efficacy in hepatocellular carcinoma
  1. Man Liu1,2,
  2. Jingying Zhou1,
  3. Xiaoyu Liu1,
  4. Yu Feng1,
  5. Weiqin Yang1,
  6. Feng Wu3,
  7. Otto Ka-Wing Cheung1,
  8. Hanyong Sun4,5,
  9. Xuezhen Zeng1,
  10. Wenshu Tang1,
  11. Myth T S Mok1,
  12. John Wong6,
  13. Philip Chun Yeung6,
  14. Paul Bo San Lai6,
  15. Zhiwei Chen7,
  16. Hongchuan Jin8,
  17. Jie Chen2,
  18. Stephen Lam Chan9,
  19. Anthony W H Chan3,
  20. Ka Fai To3,10,
  21. Joseph J Y Sung4,11,
  22. Minhu Chen2,
  23. Alfred Sze-Lok Cheng1,11
  1. 1 School of Biomedical Sciences, The Chinese University of Hong Kong, Hong Kong, China
  2. 2 Department of Gastroenterology, The First Affiliated Hospital, Sun Yat-Sen University, Guangzhou, China
  3. 3 Department of Anatomical and Cellular Pathology, The Chinese University of Hong Kong, Hong Kong, China
  4. 4 Department of Medicine and Therapeutics, The Chinese University of Hong Kong, Hong Kong, China
  5. 5 Ren Ji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai, China, Shanghai, China
  6. 6 Department of Surgery, The Chinese University of Hong Kong, Hong Kong, China
  7. 7 AIDS Institute, Department of Microbiology and Research Center for Infection and Immunity, The University of Hong Kong, Hong Kong, China
  8. 8 Labortaory of Cancer Biology, Key Laboratory of Biotherapy in Zhejiang, Sir Runrun Shaw hospital, Medical School of Zhejiang University, Hang Zhou, China
  9. 9 Department of Clinical Oncology, The Chinese University of Hong Kong, Hong Kong, China
  10. 10 State Key Laboratory of Translational Oncology, The Chinese University of Hong Kong, Hong Kong, China
  11. 11 State Key Laboratory of Digestive Disease, The Chinese University of Hong Kong, Hong Kong, China
  1. Correspondence to Professor Alfred Sze-Lok Cheng, The Chinese University of Hong Kong, Hong Kong, China; alfredcheng{at}cuhk.edu.hk

Abstract

Objective Hepatocellular carcinoma (HCC), mostly developed in fibrotic/cirrhotic liver, exhibits relatively low responsiveness to immune checkpoint blockade (ICB) therapy. As myeloid-derived suppressor cell (MDSC) is pivotal for immunosuppression, we investigated its role and regulation in the fibrotic microenvironment with an aim of developing mechanism-based combination immunotherapy.

Design Functional significance of MDSCs was evaluated by flow cytometry using two orthotopic HCC models in fibrotic liver setting via carbon tetrachloride or high-fat high-carbohydrate diet and verified by clinical specimens. Mechanistic studies were conducted in human hepatic stellate cell (HSC)-peripheral blood mononuclear cell culture systems and fibrotic-HCC patient-derived MDSCs. The efficacy of single or combined therapy with anti-programmed death-1-ligand-1 (anti-PD-L1) and a clinically trialled BET bromodomain inhibitor i-BET762 was determined.

Results Accumulation of monocytic MDSCs (M-MDSCs), but not polymorphonuclear MDSCs, in fibrotic livers significantly correlated with reduced tumour-infiltrating lymphocytes (TILs) and increased tumorigenicity in both mouse models. In human HCCs, the tumour-surrounding fibrotic livers were markedly enriched with M-MDSC, with its surrogate marker CD33 significantly associated with aggressive tumour phenotypes and poor survival rates. Mechanistically, activated HSCs induced monocyte-intrinsic p38 MAPK signalling to trigger enhancer reprogramming for M-MDSC development and immunosuppression. Treatment with p38 MAPK inhibitor abrogated HSC-M-MDSC crosstalk to prevent HCC growth. Concomitant with patient-derived M-MDSC suppression by i-BET762, combined treatment with anti-PD-L1 synergistically enhanced TILs, resulting in tumour eradication and prolonged survival in the fibrotic-HCC mouse model.

Conclusion Our results signify how non-tumour-intrinsic properties in the desmoplastic microenvironment can be exploited to reinstate immunosurveillance, providing readily translatable combination strategies to empower HCC immunotherapy.

  • cancer immunobiology
  • fibrosis
  • hepatocellular carcinoma
  • immunotherapy
  • liver immunology

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Significance of this study

What is already known on this subject?

  • Recent clinical trials of immune checkpoint blockade (ICB) therapy in advanced hepatocellular carcinoma (HCC) showed promising but not yet fully satisfactory response rates.

  • Emerging evidence suggest that the de novo resistance of HCC may be primarily attributable to myeloid-derived suppressor cells (MDSCs) that counteract the infiltration of cytotoxic T lymphocytes into the tumour.

  • Most HCCs develop from liver fibrosis or cirrhosis, but whether and how the fibrotic microenvironment impacts ICB efficacy remain unknown.

What are the new findings?

  • Increased monocytic MDSC (M-MDSC) in fibrotic microenvironment contributes to aggressive tumour growth in orthotopic HCC mouse models and patients with HCC.

  • Hepatic stellate cell (HSC) induces M-MDSC accumulation and immunosuppression through p38 MAPK-mediated enhancer reprogramming.

  • Pharmacological inhibition of p38 MAPK signalling ablates HSC-M-MDSC crosstalk to inhibit HCC growth.

  • Enhancer disruption by BET bromodomain inhibitor suppresses HCC patient-derived M-MDSCs and enhances tumour-infiltrating lymphocytes and ICB efficacy in fibrosis-associated HCC model.

Significance of this study

How might it impact on clinical practice in the foreseeable future?

  • Our preclinical data support a crucial role for fibrotic microenvironment in driving aggressive HCC growth via a profound immunosuppressive mechanism.

  • These findings demonstrate that pharmacological inhibition of M-MDSC enhancer reprogramming can improve antitumour responses elicited by ICB therapy, which provides a strong rationale for clinical trial of BET bromodomain and programmed death-1-ligand-1 coblockade.

Introduction

Hepatocellular carcinoma (HCC) is the second leading cause of cancer-related deaths globally, with an annual incidence rate of approximately 8 50 000 cases.1 Because of its heterogeneity, lack of early diagnostic biomarkers, frequent relapse on surgery and marginal effect of the available molecular therapy, treatment of HCC remains challenging.1 2 Major risk factors for HCC include chronic hepatitis B and C infections, alcohol abuse and non-alcoholic fatty liver disease associated with obesity and diabetes. Fibrosis is the common pathological niche linking the majority of these chronic liver diseases to HCC, but its contribution to hepatocarcinogenesis is only partially understood.3 Recent HCC integrative genomic studies have not only uncovered key genes involved in signalling deregulation and metabolic reprogramming but also highlighted a transformation of the immune microenvironment from activating effector cells to suppressive cells.2 4 While a few druggable driver mutations have been revealed for a fraction of patients,2 4 understanding the non-cancer cell-autonomous cues and the altered immunological landscape might provide more effective therapeutic strategies for HCC.

Therapeutic blockade of T cell coinhibitory molecules including cytotoxic T lymphocyte-associated protein 4 (CTLA-4), programmed cell death receptor 1 (PD-1) or its ligand (PD-L1) has demonstrated durable antitumour responses and long-term remissions in a subset of patients with a range of solid or haematological cancers.5 6 Two recent phase I/II trials of PD-1 checkpoint inhibitors in patients with advanced HCC have produced promising results,7 8 yet the objective response rates were relatively low (17%–20%) compared with other solid tumours such as melanoma and lung cancer (~40%–60%).5 While multiple tumour intrinsic and extrinsic factors might explain the therapeutic outcomes on immune checkpoint blockade (ICB),9 10 the de novo resistance of HCC may be primarily attributable to the strong immunosuppressive forces that counteract the infiltration of lymphocytes into the tumour.11 12

Myeloid-derived suppressor cells (MDSCs) have emerged as universal negative regulators of adaptive immunity in many pathologic conditions including malignancy.13 Two major subsets of MDSC have been identified: monocytic (M-MDSC) and polymorphonuclear (PMN-MDSC), which share phenotypic and morphologic features with monocyte and neutrophil, respectively.14 In some circumstances, M-MDSCs have been reported to possess higher immunosuppressive activity than PMN-MDSCs within the tumour microenvironment.13 15 The plasticity of MDSCs further emphasises the influences of environmental factors and cellular crosstalk in specific tumour contexts.16 Activation of hepatic stellate cells (HSCs), residing between the hepatocytes and the liver vasculature, has been established as a central driver of fibrosis.3 Interestingly, emerging data suggest that activated HSCs also contribute to liver immune tolerance through interaction with monocytes.17–19

Precisely, how HSCs modulate monocytic differentiation at molecular level and the role of this crosstalk in tumour immune evasion of fibrosis-associated HCC are poorly understood. However, strong evidence suggests that enhancers, which are fundamental determinants of gene expression, may play a key role in this context.20–22 In addition, it is possible that tissue microenvironmental signals influence the selection of lineage-specific enhancers in immune cell fate determination.23 24 Here, we sought to determine the cellular and molecular basis of the establishment of HCC-prone fibrotic microenvironment that enables rational design of cotargeted combinatorial immunotherapy.

Methods

Human specimens

Paired tumour and non-tumour liver tissues from 197 patients with HCC were collected for tissue microarray immunohistochemistry. An additional cohort of 122 liver tissues from patients with HCC and eight histologically normal livers from patients with benign focal nodular hyperplasia were collected for reverse transcription-quantitative polymerase chain reaction (RT-qPCR) analysis. Coimmunofluorescence and immunohistochemistry were also performed in a subset of patients with HCC. Peripheral blood, fresh tumour and non-tumour tissues from 54 patients with HCC were obtained from Prince of Wales Hospital (Hong Kong) for flow cytometry analysis and drug treatment. Peripheral blood mononuclear cells (PBMCs) isolated from 50 healthy donors collected from Hong Kong Red Cross served as controls or used for MDSC in vitro induction.

Cell culture, in vitro MDSC generation and immunosuppressive function analysis

The Hepa1-6 cell line with stable luciferase expression was derived from parental Hepa1-6 cells (ATCC CRL-1830) by transduction with the retroviral vector pBABE-luc-puro. The LX2 cells were purchased from Merck Millipore. Cell lines were grown in dulbecco’s modified eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS) at 37°C in a 5% CO2 humidified atmosphere. Conditional medium (CM) of LX2 was  harvested at 48 hours after culture. PBMCs isolated from anonymous healthy donors were cultured in LX2-CM for 7 days and then analysed for MDSC expansion and T cell suppression (see online supplementary methods).

Supplemental material

Molecular analyses 

RT-qPCR and chromatin immunoprecipitation-qPCR (ChIP-qPCR) were performed as previously described.25 Primer sequences for PCR and antibodies for Western blot analysis are listed in the online supplementary methods.

Establishment of murine HCC models

To model tumour growth in fibrotic liver, 6- to 8-week-old male C57BL/6 mice were treated with CCl4 (20% in 150 µl olive oil/mouse, v/v) by oral gavage for 4 or 7 weeks. Age-matched mice administered with olive oil (150 µl/mouse) served as control. Mice were injected by Hepa1-6 at a dose of 5×106 cells via intrahepatic route and then sacrificed at 3 weeks post-tumour cell implantation or humane endpoint. The tumour growth was monitored by In Vivo Xtreme device (Bruker) indicated by luciferase intensity. Tumour and matched non-tumour liver tissues were collected for primary cell isolation and subsequent immune profiling analysis.25 To further delineate tumour growth and immune profiling, another fibrotic liver model was utilised by feeding male 6- to 8-week C57BL/6 mice with high-fat high-carbohydrate (HFHC) diet or control diet (CD) for 24 weeks, followed by intrahepatic injection of Hepa1-6 cells. An additional spontaneous fibrosis-HCC model was performed using CD or HFHC diet-fed male C67BL/6 mice with neonatal diethylnitrosamine treatment as described previously.26

In vivo treatment studies

For GW856553 study, liver fibrosis was initiated in C57BL/6 mice by 2 weeks of CCl4 exposure followed by CCl4 with vehicle or GW856553 (5 mg/kg/day) cotreatment by intraperitoneal injection for additional 2 weeks. Mice were intrahepatically injected with Hepa1-6 and sacrificed 3 weeks after cell inoculation. In another group of mice, GW856553 was administered after tumour cell implantation. For i-BET762 combined immunotherapy study, liver fibrosis was initiated in C57BL/6 mice by 4 weeks of CCl4 exposure followed by Hepa1-6 intrahepatic injection. On day 7 following tumour cell inoculation, mice were divided into four groups: vehicle plus LTF-2 (IgG2b control, 10 mg/kg); i-BET762 (20 mg/kg) plus LTF-2; vehicle plus 10F.9G2 (10 mg/kg); i-BET762 plus 10F.9G2. I-BET762 or vehicle were orally administrated for 2 weeks. The antibodies were intraperitoneally injected on days 7, 12 and 17 as described previously.25 Mice were closely monitored for tumour progression by in vivo imaging. Mice were sacrificed 3 weeks after tumour cell inoculation.

Statistical analysis

Data are presented as mean ± SEM. The difference between groups was evaluated by Student’s t-test using GraphPad Prism V.7 Software (GraphPad Software). P<0.05 was regarded as statistically different. The clinicopathological features in patients with HCC with high and low CD33 immunohistochemical score were compared using Fisher’s exact test for categorical variables and the Mann-Whitney U test for continuous data. The Kaplan-Meier survival analysis was performed, where the overall survival time was calculated from the date of curative surgery to death or last follow-up of patients. The Cox proportional hazards model was used for multivariable analysis of patient survival.

Results

Liver fibrosis-associated M-MDSCs contribute to aggressive HCC development

To obtain insights into the role of fibrotic microenvironment in HCC development, we have established an orthotopic HCC model in immunocompetent C57BL/6 mice treated with CCl4 for 4 weeks (figure 1A),27 28 which developed perivenular and pericellular fibrosis with focal bridging of collagen fibres as shown by Sirius Red staining and expression of alpha smooth muscle actin, a marker of activated HSCs (online supplementary figure 1A). Notably, the mouse hepatoma exhibited much accelerated growth in the CCl4-induced fibrotic livers than the olive oil-administered controls, as shown by the increased tumorous luciferase intensity (figure 1B) and tumour weight (p<0.01; figure 1C).

Figure 1

Liver fibrosis-associated M-MDSCs contribute to aggressive HCC development. (A) Schematic diagram of liver fibrosis-associated HCC mouse model. (B) Tumour growth monitored by Bruker in vivo imaging system (n≥7/group). (C) Representative liver tumour morphology and statistical analysis of tumour weight at the endpoint. (D) Representative flow cytometry dot plots of CD11b+Ly6G-Ly6C+M-MDSCs, CD11b+Ly6G+Ly6CintPMN-MDSCs; and percentages as well as absolute numbers of M-MDSCs in liver- and tumour-infiltrating CD45immune cells. (E) Percentage of M-MDSCs in liver-infiltrating CD45immune cells was positively correlated with tumour weight. (F) The representative flow cytometry dot plots and percentages of CD8+cytotoxic T cells in tumour-infiltrating CD45immune cells. (G, H) Percentage of cytotoxic T cells in tumour-infiltrating CD45immune cells was negatively correlated with tumour weight and M-MDSC proportion in liver-infiltrating CD45immune cells. (I) The proportions and absolute numbers of CD33+CD11b+CD14+HLA-DR−/lowM-MDSC in liver- and tumour-infiltrating CD45immune cells from HCC patients with (n=24) or without liver fibrosis (n=7). (J) Representative pictures of H&E,   alpha smooth muscle actin  immunohistochemistry as well as CD33 and CD14 coimmunofluorescence in human HCC specimen. Colocalisation of CD33 and CD14 is shown in the merged images. Hoechst served as positive control for cell nuclei staining. Scale bar=100 µm. (K) Kaplan-Meier overall survival and disease-free survival curves of HCC patients with high (n=49) or low (n=148) expressions (stratified top 25% as high expression group) of CD33 in non-tumour liver tissues as determined by immunohistochemistry (see insert). Error bars represent mean ± SEM, *p<0.05; **p<0.01; ***p<0.001. HCC, hepatocellular carcinoma; M-MDSCs, monocytic myeloid-derived suppressor cells.

To investigate the immunological landscape responsible for accelerated HCC growth, we measured the levels of CD11b+Ly6G-Ly6C+M-MDSCs and CD11b+Ly6G+Ly6CintPMN-MDSCs (figure 1D), which are instrumental for immunosuppressive tumour microenvironment.25 29 We observed a significant increase in M-MDSCs (p<0.05; figure 1D), but not PMN-MDSCs (online supplementary figure 1B), in the tumour-surrounding fibrotic livers compared with control livers, while both subtypes of MDSCs showed no difference within the tumour (figure 1D and online supplementary figure 1B). We found that the level of M-MDSCs may be correlated with the degree of liver fibrosis (online supplementary figure 2). Moreover, the liver-infiltrating M-MDSC proportion was significantly correlated with tumour mass (r2=0.6635, p<0.0001; figure 1E), whereas tumor-infiltrating M-MDSC proportion had no significant correlation (data not shown). On the  contrary, the proportion of tumour-infiltrating CD8T cells was significantly decreased in CCl4-treated compared with control mice (p<0.05; figure 1F) and negatively correlated with tumour mass (r2=0.6035, p<0.0001; figure 1G). Importantly, the liver-infiltrating M-MDSC proportion exhibited inverse correlation with tumour-infiltrating CD8T cells (r2=0.3363, p<0.005; figure 1H).

To consolidate the role of liver fibrosis-associated M-MDSCs, we have established an additional orthotopic HCC model in C57BL/6 mice fed with HFHC diet for 24 weeks,30 which developed both obesity and non-alcoholic steatohepatitis with significant fibrosis (online supplementary figure 3A,B). We also observed a significant increase in M-MDSCs (p<0.05), but not PMN-MDSCs, in the tumour-surrounding fibrotic livers compared with the control livers of mice fed with CD (online supplementary figure 3C,D). Moreover, the liver-infiltrating M-MDSC proportion was significantly correlated with tumour mass (r2=0.5389, p<0.005) and negatively correlated with tumour-infiltrating CD8T cells (r2=0.6460, p=0.0005; online supplementary figure 3E,F). These findings were further partially recapitulated in a spontaneous fibrosis-associated HCC model using CD or HFHC-fed male mice with neonatal diethylnitrosamine treatment,26 31 which exhibited positive correlations among the degree of liver fibrosis, M-MDSCs and tumour multiplicity (online supplementary figure 4). Thus, our findings from multiple HCC models consistently demonstrate that M-MDSCs in fibrotic liver may contribute to aggressive HCC development by inhibiting T cell infiltration.

Next, we investigated the clinical relevance of M-MDSC by measuring the liver- and tumour-infiltrating levels in HCC patients with (n=24) or without (n=7) underlying fibrosis/cirrhosis. In human, the equivalent to M-MDSC is defined as CD33+CD11b+CD14+HLA-DR−/low.14 Consistent with the mouse model, the proportion of M-MDSCs was significantly higher in the tumour-surrounding fibrotic livers than the non-fibrotic livers of patients with HCC (p<0.001; figure 1I). Moreover, the level of M-MDSCs was significantly higher in the livers than tumours (p<0.05; figure 1I), which was supported by colocalisation (figure 1J and online supplementary figure 5A) and correlated expressions of CD33 and CD14 in the tumour-surrounding livers (r2=0.3446; p<0.0001; online supplementary figure 5B). To further determine whether the proportion of M-MDSCs impacts the prognosis of patients with HCC, we performed CD33 immunohistochemistry in an HCC tissue microarray, which includes 197 cases with recorded clinical follow-up data. The Kaplan-Meier analysis revealed that patients with HCC with high CD33 expression in liver (figure 1K), but not tumour (online supplementary figure 6), were significantly correlated with shorter overall (p<0.01) and disease-free survival rates (p<0.01). Consistently, clinicopathological correlation analysis showed that high CD33 expression in liver (online supplementary table 1), but not tumour (online supplementary table 2), was significantly associated with larger tumour size, increased vascular invasion and advanced tumour stage. Multivariate analysis further showed positive association between high liver CD33 expression and poorer patient survival (online supplementary tables 3 and 4), thus highlighting the clinical significance of M-MDSCs in the tumour-surrounding liver microenvironment.

HSC induces M-MDSC accumulation and function through p38 MAPK signalling

To investigate how the HCC-prone fibrotic microenvironment influences monocyte-intrinsic signalling, we first recapitulated MDSC expansion from human PBMCs using the CM of an activated HSC cell line LX2 (figure 2A). In this culture system, M-MDSCs but not PMN-MDSCs were specifically induced (p<0.001; figure 2B and online supplementary figure 7), which phenotypically resembled those isolated from fibrosis-associated HCC patients (figure 2B and online supplementary figure 8). Similar to the patient-derived M-MDSCs, the HSC-induced M-MDSCs also significantly suppressed T cell proliferation (p<0.001; figure 2C). RNA sequencing (RNA-seq) analysis of the sorted HSC-induced M-MDSCs and the untreated CD14monocytes uncovered 869 differentially expressed genes (p<0.05 and fold-change >2) that were significantly enriched in key signalling pathways (p<0.05; figure 2D). Most of them (tumour necrosis factor (TNF), nucleotide-binding oligomerization domain (NOD)-like receptor, extracellular-signal-regulated kinase (ERK), receptor for advanced glycation end products (RAGE), nuclear factor-κB (NF-κB)) have been reported to drive monocyte polarisation to MDSC,32–34 but the role of MAPK signalling is poorly defined (figure 2D). Multiple genes in the MAPK cascade were distinctively upregulated in HSC-induced M-MDSCs when compared with CD14monocytes (figure 2E), which were validated by RT-qPCR (p<0.05; supplementary figure 9A). Western blot analysis of phosphorylated p38 at Thr180/Tyr182 (p-p38Thr180/Tyr182) further verified strong p38 MAPK activation in HSC-induced M-MDSCs (figure 2F).

Figure 2

HSCs activate p38 MAPK signalling pathway during M-MDSC differentiation. (A) Schematic diagram of M-MDSC generation and phenotypic/functional analysis in vitro. (B) The representative flow cytometry dot plots and percentages of CD14+HLA-DR−/lowM-MDSCs generated in vitro or from patients with HCC ex vivo (n≥22/group). (C) CFSElow proliferated CD3+ T cells were shown in histogram and dot plot graph in indicated groups (n=5/group). (D) Signalling pathway enrichment analysis of 869 differentially expressed genes in M-MDSC identified by RNA-seq (n=3, cut-off of p<0.05, fold-change >2). (E) Differentially upregulated genes of MAPK signalling pathway are shown in hierarchical clustering heatmap. (F) Phosphorylation of p38Thr180/Tyr182 and total p38 expressions were determined by western blot. β-actin served as loading control. Error bars represent mean ± SEM. ***p<0.001. HCC, hepatocellular carcinoma; HSCs, hepatic stellate cells; M-MDSC, monocytic  myeloid-derived suppressor cell; PBMCs, peripheral blood mononuclear cells.

Cell identity, differentiation and response to extrinsic cues depend on the transduction of cellular signals, which converge on chromatin to elicit gene activation or suppression.35 We next investigated the role of p38 MAPK signalling on M-MDSC gene expressions by a specific inhibitor, SB203580. Compared with the human CD14monocytes, HSC-induced M-MDSCs highly expressed a myeloid lineage-determining transcription factor CCAAT-Enhancer-Binding Protein Beta (C/EBPβ),14 36 which could be abrogated by SB203580 treatment (figure 3A,B). Similarly, the S100 family members, especially the highly expressed S100A8, S100A9 and S100A12 (online supplementary figure 9B) that exert M-MDSC-mediated immunosuppression,14 37 were also simultaneously regulated by p38 MAPK signalling (figure 3A).

Figure 3

p38 MAPK signalling controls M-MDSC accumulation and function via C/EBPβ-mediated chromatin regulation. (A) SB203580 (10 µM) suppressed mRNA expression of C/EBPβ as well as S100A8, S100A9 and S100A12 in HSC-induced M-MDSCs (n≥3/group). (B) C/EBPβ protein level was determined by western blot in monocytes or HSC-induced M-MDSCs with or without SB203580 (10 µM) treatment. (C–D) Relative mRNA level of C/EBPβ, S100A8, S100A9 and S100A12 and mean fluorescence intensity (MFI) of S100A8/9 and S100A12 was determined by RT-qPCR and flow cytometry, respectively, in M-MDSC transfected with small-interfering RNAs against control sequence (si-Ctrl) or C/EBPβ (si-C/EBPβ) (n=6). (E) Proportion of CD14+HLA-DR−/lowcells in M-MDSCs with or without C/EBPβ knockdown was  measured by flow cytometry (n=6). (F–H) Chromatin immunoprecipitation-quantitative PCR (ChIP-qPCR) of monocytes or HSC-induced M-MDSCs with or without SB203580 (10 µM) treatment. C/EBPβ, H3S28ph and p300 enrichment are shown at indicated gene regions (n≥4). (I) Schematic diagram of SB203580 treatment regimen in vitro. (J) Representative flow cytometry dot plots and proportions of CD14+HLA-DR−/lowM-MDSCs in PBMCs cultured in DMEM or LX-CM with or without SB203580 treatment (n=8). (K) Representative histogram of CFSElow proliferated CD3T cells cocultured with monocytes (DMEM control) or with LX2-CM-induced M-MDSCs with or without SB203580 (10 µM) treatment. Cumulative data are shown in dot plot graph (n=4). Error bars represent mean ± SEM. *p<0.05; **p<0.01; ***p<0.001. HSC, hepatic stellate cell; M-MDSCs, monocytic  myeloid-derived suppressor cells; PBMCs,  peripheral blood mononuclear cells.

We speculated that HSC-stimulated p38 MAPK signalling is pivotal for M-MDSC accumulation and function via induction of C/EBPβ-mediated chromatin regulation. We first found that knockdown of C/EBPβ reduced S100A8/9/12 expressions (p<0.05; figure 3C,D) and M-MDSC proportion (p<0.05; figure 3E). We next showed that p38 MAPK signalling regulated C/EBPβ expression (figure 3A) and binding to its own and S100A8/9/12 promoters in HSC-induced M-MDSCs (figure 3F). As a bona fide coactivator for C/EBPβ, p300-dependent transcription is augmented by p38 MAPK-mediated phosphorylation of histone 3 serine 28 (H3S28ph).38 39 Indeed, the C/EBPβ and S100A8/9/12 promoter occupancies of H3S28ph and p300 were also simultaneously regulated by p38 MAPK signalling (figure 3G,H). To further determine its functional significance, we treated the human PBMC culture system with SB203580 (figure 3I) and found that p38 MAPK inhibition significantly abolished HSC-induced M-MDSC expansion (p<0.001; figure 3J) and impaired the T cell suppressive activity (p<0.05; figure 3K).

Blockade of p38 MAPK signalling abrogates HSC-M-MDSC crosstalk and inhibits HCC growth

To investigate the effect of p38 MAPK signalling in vivo, we treated C57BL/6 mice with GW856553, a clinically trialled p38 MAPK inhibitor,40 or vehicle control during fibrosis induction by CCl4 and followed by tumour cell implantation (figure 4A). Administration of GW856553 significantly suppressed fibrosis-induced M-MDSC expansion (p<0.01; figure 4B) and tumour growth (figure 4C), leading to a significant reduction in HCC tumorigenicity (p<0.05; figure 4D). Notably, GW856553 also significantly reduced the liver-infiltrating M-MDSC proportion (p<0.05; figure 4E,F), which was positively and negatively correlated with tumour mass (r2=0.7026, p<0.0001; figure 4G) and tumour-infiltrating CD8T cell proportion (r2=0.5683, p<0.0001; figure 4H), respectively. In contrast, administration of GW856553 after tumour cell implantation did not reduce HCC tumorigenicity (online supplementary figure 10A,B). Some tumours in GW856553-treated mice were even larger than those of the untreated mice, which concurs with the tumour-suppressive effect of p38 MAPK.41 Indeed, knockdown of tumour cell-intrinsic p38 MAPK by RNA interference significantly increased tumorigenicity (p<0.001; online supplementary figure 10A,B). Collectively, these data suggest that myeloid-intrinsic p38 MAPK signalling is crucial for HCC growth.

Figure 4

p38 MAPK inhibition suppresses M-MDSCs and fibrosis-associated HCC development. (A) Schematic diagram of GW856553 (5 mg/kg) treatment regimen in the fibrotic HCC mouse model (n≥10/group). (B) M-MDSC proportion in PBMCs was detected weekly in mice with or without GW856553 treatment. (C) Representative luciferase images of tumours in vehicle or GW856553-treated groups taken on a weekly basis. (D) Representative liver tumour morphology and statistical analysis of tumour weight of the vehicle or GW856553-treated groups at the endpoint. (E) Representative flow cytometry dot plots as well as (F) percentages and absolute numbers of CD11b+Ly6G-Ly6C+M-MDSCs in liver-infiltrating CD45immune cells. (G–H) Liver-infiltrating M-MDSCs was positively correlated with tumour weight and negatively correlated with tumour-infiltrating cytotoxic T cell proportion. (I) Intracellular p-p38Thr180/Tyr182 level in peripheral CD14cells from healthy donor and fibrotic HCC patients are shown in flow cytometry histogram and MFI dot plot graph (n≥6/group). (J) Schematic diagram of GW856553 (10 µM) treatment regimen in the fibrotic HCC patient-derived PBMCs. (K) Representative flow cytometry dot plots and proportion of CD14+HLA-DR–/lowM-MDSCs in fibrotic HCC patient-derived PBMCs with or without GW856553 treatment (n=7). (L) Histogram graph and MFI of HLA-DR, CD40, CD80, CD86 on CD14cells from fibrotic HCC patient-derived PBMCs with or without GW856553 treatment (n=6). Error bars represent mean ± SEM. *p<0.05; **p<0.01; ***p<0.001. HCC, hepatocellular carcinoma; M-MDSCs, monocytic  myeloid-derived suppressor cells; MFI, mean fluorescence intensity; PBMCs, peripheral blood mononuclear cells.

Next, we detected a significantly higher level of p-p38Thr180/Tyr182 in the circulating CD14monocytes of the fibrotic HCC patients when compared with the healthy donors (p<0.05; figure 4I), supporting the clinical relevance of our findings. Furthermore, treatment of HCC patient-derived PBMCs with GW856553 significantly decreased CD14+HLA-DR−/lowM-MDSC proportion (p<0.001; figure 4J, K) and increased the expressions of HLA-DR and costimulatory molecules (CD40, CD80 and CD86) (p<0.05; figure 4L), indicative of monocyte maturation. In corroboration with the animal study, our findings demonstrate that activation of p38 MAPK functionally links HSC-induced M-MDSC accumulation with suppression of antitumour immunosurveillance in fibrotic liver environment. p38 MAPK-mediated enhancer reprogramming determines M-MDSC identity and immunosuppression.

Given the importance of p300 in the activation of transcriptional enhancers39 and the emerging roles of enhancer regulation in immune cells,23 24 we next investigated the p300-dependent enhancer activity in M-MDSC. First, we found that the histone mark for enhancer activity, histone H3 lysine 27 acetylation (H3K27ac), was globally increased during HSC-mediated induction of M-MDSCs (figure 5A). Moreover, the H3K27ac level was dose-dependently reduced by a specific p300 inhibitor, C646 (figure 5B). Notably, the M-MDSC induction by HSC could be abolished by C646 (p<0.01; online supplementary figure 11), highlighting the importance of p300-mediated H3K27ac in M-MDSC accumulation.

Figure 5

Crucial role of p38 MAPK-mediated enhancer reprogramming in M-MDSC identity and immunosuppression. (A) Representative H3K27ac immunoblot in monocytes from DMEM or LX2-CM cultured PBMCs at indicated time points. (B) Representative H3K27ac immunoblot in DMEM or LX2-CM-cultured monocytes treated with DMSO or C646 (1, 5, 10 µM). (C) RNA-seq profiles of monocytes and M-MDSCs at indicated FANTOM 5 eRNA regions of C/EBPβ and S100A8/9/12 (n=3). (D) C/EBPβ and S100A8/9/12 eRNA expressions in monocytes or M-MDSCs by RT-qPCR (n=5). (E) ChIP-qPCR heatmap of C/EBPβ, P300, H3K27ac and BRD4 occupancies at enhancer regions of C/EBPβ and S100A8/9/12 in monocytes, M-MDSCs with DMSO, SB203580 (10 µM) or C646 (5 µM) treatment (n≥4). (F) Schematic diagram of JQ1 (100 nM) treatment regimen in the healthy donor PBMCs cultured in DMEM or LX2-CM. (G) C/EBPβ and BRD4 enrichment at C/EBPβ and S100A8/9/12 enhancer regions in each group were determined by ChIP-qPCR (n≥4). (H) C/EBPβ and S100A8, S100A9, S100A12 mRNA levels in each group were determined by RT-qPCR (n≥3). (I) Representative flow cytometry dot plots and proportions of CD14+HLA-DR−/lowM-MDSCs in PBMCs cultured in DMEM or LX2-CM treated with or without JQ1 (100 nM) (n=9). (J) Representative histogram of CFSElow proliferated CD3T cells cocultured with monocytes (DMEM control) or LX2-CM-induced M-MDSCs with or without JQ1 (100 nM) treatment. Cumulative data are shown in dot plot graph (n=4). (K) HLA-DR, CD40, CD80, CD86 expression levels on CD14cells cultured in DMEM control or LX2-CM with or without JQ1 (100 nM) treatment are shown by histogram and MFI dot plot graph (n=6). Error bars represent mean ± SEM; *p<0.05; **p<0.01; ***p<0.001. M-MDSCs, monocytic  myeloid-derived suppressor cells; MFI, mean fluorescence intensity; PBMCs, peripheral blood mononuclear cells .

Transcriptional enhancers are key regulatory elements driven by combinatorial assembly of lineage-determining transcription factors, coactivators and bromodomain and extraterminal domain (BET) family proteins including BRD4, which recruits transcriptional complexes to H3K27ac for enhancer RNA (eRNA) synthesis.42 Using the RNA-seq data, we uncovered non-coding RNAs that overlap with the FANTOM5 enhancers,20 including those targeting C/EBPβ and S100A8/9/12 (figure 5C). By RT-qPCR, we verified the overexpression of these eRNAs in HSC-induced M-MDSCs compared with CD14monocytes (figure 5D). As predicted by our recently proposed method,22 the C/EBPβ and S100A8/9/12 enhancers were enriched with C/EBPβ, p300, H3K27ac and BRD4 (figure 5E). Moreover, treatment with either SB203580 or C646 abrogated their enhancer binding in HSC-induced M-MDSCs (figure 5E), which was consistent with the crucial role of p38 MAPK signalling and its interaction with p300.39

To further investigate the significance of enhancer regulation in M-MDSC, we treated the human PBMC culture system with a BET inhibitor JQ1 (figure 5F). Concordant with the enhancer coregulatory function of C/EBPβ and BRD4,43 JQ1 simultaneously blocked their recruitment to the C/EBPβ and S100A8/9/12 enhancers (figure 5G), leading to dramatic inhibition of target gene expressions (figure 5H). Importantly, JQ1 treatment significantly reduced HSC-induced M-MDSC expansion (p<0.001; figure 5I) and impaired the T cell suppressive activity (p<0.01; figure 5J), which were accompanied by upregulation of the maturation markers (p<0.05; figure 5K). Overall, these results suggest that p38 MAPK transduces the extrinsic signals to enhancer reprogramming for M-MDSC accumulation and function.

BET bromodomain inhibition suppresses fibrotic HCC patient-derived M-MDSCs

We next investigated the clinical potential of BET inhibition by treating fibrotic HCC patient-derived PBMCs with a small-molecule BET inhibitor, i-BET762 (also known as GSK525762),44 which is currently under investigation in phase I/II clinical trials (NCT02964507/NCT03150056) (figure 6A). Compared with vehicle control, i-BET762 treatment significantly reduced the CD14+HLA-DR−/low M-MDSC proportion (p<0.001; figure 6B). The expressions of the immunosuppressive S100A8/9/12 proteins in M-MDSCs were also significantly reduced (p<0.05; figure 6C), while the maturation markers were concordantly upregulated (p<0.05; figure 6D). These findings demonstrate the sensitivity of HCC patient-derived M-MDSCs to BET inhibition.

Figure 6

BET bromodomain inhibitor suppresses HCC patient-derived M-MDSCs. (A) Schematic diagram of i-BET762 (500 nM) treatment regimen in fibrotic HCC patient-derived PBMCs. (B) Representative flow cytometry dot plots and proportion of CD14+HLA-DR−/low M-MDSCs in fibrotic HCC patient-derived PBMCs with or without i-BET762 treatment (n=10). (C–D) S100A8/9 and S100A12 as well as HLA-DR, CD40, CD80, CD86 expressions on CD14cells from fibrotic HCC patient-derived PBMCs with or without i-BET762 treatment are shown by histogram and MFI dot plot graph (n≥9). Error bars represent mean ± SEM. *p<0.05; **p<0.01; ***p<0.001. HCC, hepatocellular carcinoma; M-MDSCs, monocytic  myeloid-derived suppressor cells; MFI, mean fluorescence intensity; PBMCs, peripheral blood mononuclear cells.

BET bromodomain inhibition enhances PD-L1 blockade efficacy to eradicate fibrotic HCC and prolong survival

Given the crucial role of MDSCs as the Achilles’ heel of cancer immunotherapy,45–47 we next evaluated the therapeutic efficacy of i-BET762 using the aforementioned CCl4-induced fibrotic HCC model. i-BET762 (20 mg/kg) or vehicle control was administered44 in combination with 10 mg/kg PD-L1 blockade antibody (10F.9G2) or the control rat IgG2b (LTF-2) (figure 7A),25 followed by weekly in vivo imaging (figure 7B), immune cell profiling and tumour measurement at the time of sacrifice (figure 7C,D). As expected, i-BET762 treatment significantly reduced the level of M-MDSCs in the fibrotic liver (p<0.05; figure 7E,F) and increased the level of tumour-infiltrating CD8T cells (p<0.05; figure 7G), leading to a significant reduction in tumour mass (p<0.05; figure 7D). Although PD-L1 blockade antibody also significantly reduced tumorigenicity to similar extent with i-BET762 (p<0.01; figure 7D), single treatment alone was unable to eradicate the large hepatoma (figure 7B-D).

Figure 7

Coblockade of BET bromodomain and PD-L1 exhibits synergistic efficacy in HCC eradication and prolonging host survival. (A) Schematic diagram of single or combined i-BET762 (20 mg/kg) and anti-PD-L1 antibody 10F.9G2 (10 mg/kg) treatment in fibrotic HCC mouse model (n≥6/group). (B) Tumour growth was monitored by Bruker in vivo imaging system. (C) Representative liver tumour morphology and (D) statistical analysis of tumour weight at the endpoint (below). (E) Representative flow cytometry dot plots as well as (F) percentages and absolute numbers of CD11b+Ly6G-Ly6CM-MDSCs in liver-infiltrating CD45cells of mice in each group. (G) Representative flow cytometry dot plots and percentages of CD8cytotoxic T cells in tumour-infiltrating CD45cells of mice in each group. (H–I) Liver-infiltrating M-MDSCs was negatively correlated with tumour-infiltrating cytotoxic T cells and positively correlated with tumour weight. (J) Kaplan-Meier survival analysis of fibrotic HCC mice that received single or combined i-BET762 and anti-PD-L1 antibody treatment (n≥8/group). Error bars represent mean ± SEM. *p<0.05; **p<0.01; ***p<0.001. HCC, hepatocellular carcinoma; M-MDSCs, monocytic myeloid-derived suppressor cells; PD-L1, programmed death-1-ligand-1.

Notably, BET and PD-L1 coblockade markedly improved the antitumour efficacy when compared with either treatment (p<0.05; figure 7C,D), which was accompanied by significant reduction in the level of liver-infiltrating M-MDSCs (p<0.05; figure 7E,F) and marked increase in tumour-infiltrating CD8T cells (p<0.001; figure 7G). These intratumoral T cells expressed a myriad of cytolytic cytokines, especially IL-2 that may play a crucial role in the suppression of tumorigenicity via promotion of T cell cytotoxicity activity (online supplementary figure 12). We further observed significant negative and positive correlations between liver-infiltrating M-MDSCs and tumour-infiltrating CD8T cells (r2=0.1832, p<0.05; figure 7H) and tumour mass (r2=0.4586, p<0.001; figure 7I), respectively, thus supporting the importance of M-MDSC in promoting HCC growth in fibrotic liver via its T cell suppressive activity.

Since ICB therapy provided durable objective responses in only a fraction of advanced HCC patients,7 8 we further investigated whether our combined immunotherapy could prolong the host survival. While all mice died within 30 days in our aggressive fibrosis-associated HCC model (figure 7J), either i-BET762 or PD-L1 antibody treatment showed significantly higher survival rate (p<0.05 and 0.01, respectively; figure 7J). Notably, combined BET and PD-L1 inhibition elicited marked long-term survival benefits (p<0.001, 0.01 and 0.05 compared with untreated, i-BET762 or PD-L1 antibody, respectively; figure 7J). In fact, 75% (6/8) of mice receiving the coblockade therapy have survived for more than 6 months, in contrast to only 11.1% and 12.5% in i-BET762 and PD-L1 antibody groups, respectively (figure 7J). In addition, the combination immunotherapy was well-tolerated, and neither sign of tissue toxicity nor body weight loss was observed (online supplementary figure 13). Overall, these findings demonstrate that targeted inhibition of BET bromodomain enhances the efficacy of anti-PD-L1 to eradicate aggressive HCC and achieve long-term survival.

Discussion

Accumulating evidence suggests that cancer cell-intrinsic properties can orchestrate the immune landscape of tumour-bearing hosts.10 48 However, whether the surrounding tissue microenvironment modulates tumour immunity and sensitivity to immunotherapy remain obscure.6 Using multiple fibrosis-associated HCC models, our preclinical data support a crucial role for peritumoral fibrotic microenvironment in driving aggressive tumour growth via a profound immunosuppressive mechanism (figure 8). We found that accumulation of M-MDSCs induced by the profibrogenic HSCs form a barricade to restrain cytotoxic T lymphocyte infiltration to the tumour. Strikingly, targeting the HSC-M-MDSC crosstalk in fibrotic liver enhanced the efficacy of anti-PD-L1 therapy to unleash effective cytotoxic T cell responses for hepatoma eradication. In human HCCs, the tumour-surrounding fibrotic/cirrhotic liver was also markedly enriched with M-MDSCs, which was significantly associated with more aggressive tumour phenotypes and poor survival rates. These observations are consistent with the clinical impact of high levels of circulating M-MDSC on poor prognosis of patients with HCC49 50 and limited anti-CTLA-4 responsiveness in patients with melanoma.51 52 Our data illuminate a novel mechanism-based combination strategy to overcome the overwhelming de novo resistance of HCC to ICB.7 8

Figure 8

Schematic representation of HCC immune evasion via monocyte-intrinsic enhancer reprogramming in the fibrotic microenvironment. Therapeutic coblockade of BET bromodomain and PD-L1 can simultaneously inhibit liver-infiltrating M-MDSCs and enhance tumour-infiltrating CD8+ T cells, resulting in eradication of aggressive HCC with underlying fibrotic liver.  HCC, hepatocellular carcinoma; HSC, hepatic stellate cell; M-MDSCs, monocytic myeloid-derived suppressor cells; PD-L1, programmed death-1-ligand-1.

This work has advanced the understanding of deregulation of the liver immunological network in HCC development. Using our human HSC-PBMC culture system, which reflects the situation in a fibrotic liver with preferential enrichment of M-MDSCs that phenotypically resemble the HCC patient-derived M-MDSCs, we have uncovered a fibrosis-elicited p38 MAPK signalling that triggers monocyte to M-MDSC differentiation via C/EBPβ-mediated chromatin regulation. In addition to eliciting the proinflammatory responses of macrophage in the proliferation and transformation of premalignant cells,53 the activated HSCs also induce M-MDSCs to establish an immunosuppressive liver environment for uncontrolled HCC growth. This notion is supported by the marked reduction in intrahepatic M-MDSCs and the correlated T cell influx in fibrosis-bearing mice on GW856553 treatment, leading to marked retardation in tumorigenicity. Our findings demonstrating the impact of the clinically trialled p38 MAPK inhibitor on fibrotic HCC patient-derived M-MDSCs further suggests that other cancer types with p-p38-high MDSCs, such as melanoma and bladder cancer,54 55 may also be amenable to pharmacological restoration of immune surveillance.

We and others have interrogated the crosstalk between tumour cells and MDSCs in HCC microenvironment.25 26 56 57 While hepatoma-intrinsic pathways such as nuclear factor-kappa B and mammalian target of rapamycin signalling have been shown to induce PMN-MDSC expansion and immunosuppression,25 26 here we show that the environment-driven chromatin deregulation of the myeloid lineage-determining factor C/EBPβ and S100A8/9/12 effectors fuels the accumulation of immunosuppressive M-MDSCs in the tumour-surrounding liver. Our data suggest that mechanistically, activated HSCs release factors to activate monocyte-intrinsic p38 MAPK signalling, which in turn recruits C/EBPβ and p300 for H3K27 acetylation and BRD4 binding at enhancers, upregulating expressions of eRNAs and M-MDSC-specific genes such as S100A9.37 Pharmacological inhibition of p38 MAPK, p300 or BRD4 resulted in codepletion of C/EBPβ and BRD4 and thus abrogation of active enhancer identity at M-MDSC-associated enhancers. This is consistent with recent studies reporting that tissue environmental signals control the transcriptional programmes of macrophage and regulatory T cell via selective activation of lineage-specific enhancers,23 24 thus reinforcing the importance of context-dependent enhancer reprogramming for immune cell fate determination.

One notable finding from this epigenetic study is the therapeutic efficacy of BET inhibition in the suppression of HSC-induced and fibrotic HCC patient-derived M-MDSCs. It remains to be elucidated in future studies the transcriptomes and enhancer landscapes contributing to the tumorigenic role of MDSCs. Nevertheless, recent studies have indicated the therapeutic potential of depleting or inactivating MDSC to empower cancer immunotherapy.45–47 Besides the abrogation of expansion and immunosuppressive activity, BET inhibitors also facilitated M-MDSC maturation with higher costimulatory molecule expressions, which may acquire antigen-presenting cell-like properties such as T cell priming and activation. While this advantageous effect warrants validation, the facts that BET inhibition also transcriptionally suppresses tumour cell-intrinsic oncogenic pathways58 and fibrotic responses of HSC59 underscore the necessity of further testing this attractive therapeutic strategy in spontaneous models of fibrosis and HCC development.31

By using a preclinical model of fibrosis-associated HCC, we have demonstrated that BET inhibition therapy-induced enhancer remodelling of M-MDSC confers sensitivity of HCC to ICB. The translational potential of this study has implications beyond HCC. The desmoplastic response of other cancers such as pancreatic ductal adenocarcinoma and castration-resistant prostate cancer also drives the generation of immunosuppressive microenvironment, which is often resistant to immune checkpoint therapies.5 60 61 We show that the stromal interactions in the desmoplastic tumours may be transformed by iBET-induced attraction of CD8T cells to the tumour microenvironment, thus reinstating immune surveillance and responsiveness to ICB. In this regard, a phase I/IIa trial of a BET inhibitor with or without anti-PD-1 in advanced solid or hematologic malignancies (NCT02419417) would substantiate the hypotheses of our work. In summary, we have devised a combination epigenetic immunotherapy that promotes MDSC maturation and cytotoxic T lymphocyte infiltration in immune-refractory tumours, thereby eliciting superior synergistic efficacy against intrinsic resistance for durable survival benefits.

References

Footnotes

  • ML and JZ contributed equally.

  • Contributors Study concept and design: ML, JZ and ASLC; Data acquisition and analysis: ML, JZ, XL, YF, WY, FW, HS, XZ, WT, MTSM, AWHC and ASLC; Bioinformatics analysis: FW and OKWC; Clinical resources: JW, PCY, PBSL, ZC, AWHC and KFT; Writing of manuscript: ML, JZ and ASLC; Critical review of manuscript: ML, JZ, PBSL, HJ, JC, SLC, KFT, ZC, JJYS, MC and ASLC; Supervision: ASLC; Funding acquisition: ML, JZ, ZC, JJYS, MC and ASLC.

  • Funding This project is supported by the University Grants Committee through the Collaborative Research Fund (C4045-18W) and the Theme-based Research Scheme (T11-706/18-N), the Health and Medical Research Fund (16170451), the Terry Fox Foundation, the Focused Innovations Scheme-Scheme B (1907309) from the Chinese University of Hong Kong and the Li Ka Shing Foundation (Canada). Man Liu is supported by the China Post-doctoral International Exchange Program and China Postdoctoral Science Foundation.

  • Competing interests None declared.

  • Ethics approval All animal experiments were approved by the Animal Experimentation Ethics Committee of the Chinese University of Hong Kong (AEEC-CUHK). Studies using human specimen were approved by the Joint CUHK-NTEC Clinical Research Ethics Committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Patient consent for publication Obtained.