Mucosal-associated invariant T cells promote ductular reaction through amphiregulin in biliary atresia

Study design

The aim of this study was to identify the function of MAIT cells on BA. First, we analyzed correlations of liver MAIT cell level and clinical parameters [survival, alanine transaminase (ALT), bilirubin, histological inflammation and fibrosis] in two public cohorts of patients with BA (one cohort was 45 BA patients in US, while the second cohort was 102 BA patients in China). Next, we obtained liver or peripheral blood samples from BA and control patients for bulk RNA sequencing (RNAseq), flow cytometry analysis, immunostaning, and performed wound healing experiment and cell proliferation assay in both chonlangiocyte cell line and biliary organoids. The control patients were pediatric patients with hepatoblastoma (HB), neonatal intrahepatic cholestasis (IHC) and congenital choledochal cysts (CC). For hepatoblastoma, we only obtained adjancent normal liver tissues for study. Finally, we established two in vitro co-culture systems, one is the RRV infected co-culture system to model immune dysfunction of human BA which was validated by single cell RNA sequencing (scRNA-seq) and the other is a multicellular system composed of biliary organoids, LX-2 and MAIT cells to evaluate the role of MAIT cells on ductular reaction.

Ethics statement

All human specimens were collected at The First Affiliated Hospital of Sun Yat-sen University, P.R. China. The protocol was approved by the Institutional Research Ethics Committee at The First Affiliated Hospital of Sun Yat-sen University (approval number 2021-572). The study design conformed to the ethical guidelines of the 1975 Declaration of Helsinki and Istanbul. Written informed consents were obtained from parents of each patient or patients.

The animal protocols were reviewed and approved by the Animal Care and Use Committee of Sun Yat-sen University (number SYSU-IACUC-2022-016).

Patients

Liver samples and peripheral blood were obtained from patients with BA and controls including hepatoblastoma, neonatal intrahepatic cholestasis and congenital choledochal cysts. These types of disease controls have been used in study of BA.29, 30, 31 No matching was done between BA and controls. The diagnosis of BA was defined by an abnormal intraoperative cholangiogram and histological demonstration of obstruction of extrahepatic bile ducts. Liver samples of BA were obtained at the time of hepatoportoenterostomy (HPE). Adjacent normal liver tissues of human hepatoblastoma were collected at radical tumor resection. Clinical paramters including sex assigned at birth, age, ALT, aspartate transaminase (AST), gamma-glutamyl transferase (GGT), total bilirubin, liver pathology (histiological inflammation score and fibrosis score of liver section), failure of jaundice clearance, portal vein hypertension from patients with BA (n = 47) and controls (n = 41) were summarized in Supplemental Table S3. Age has been shown to be a potential confounding factor as it is associated with stage of disease and patient outcome.^4,32 Due to difficulty in obtaining perfect age matched control samples, our study enrolled BA and control patients with age in an appropriate range to minimize the confounding effect of age as previous studies.^31,33 Failure of jaundice clearance were determined by total bilirubin >34.2 μmol/L at 3 months after HPE as previous described.³⁴ Portal vein hypertension were defined by history of a complication including esophageal or gastric variceal bleed, ascites, or hepatopulmonary syndrome or clinical findings consistent with portal vein hypertension [both splenomegaly (spleen palpable >2 cm below the costal margin) and thrombocytopenia (platelet count <150,000/mL)] as previous described.³⁵ Every specimen from one patient was used for one or two different experiments.

Estimation of relative cell abundance of liver MAIT cells in two cohorts of patients with biliary atresia and association with clinical parameters

Level of liver MAIT cells was estimated in two cohorts of BA patients using single sample Gene Set Enrichment Analysis (ssGSEA) method with input of marker genes for MAIT cells and gene expression data as previously described.³⁶ The first cohort (the US cohort) was microarray gene expression data of 47 BA patients in USA, while the second cohort (the China cohort) was bulk liver RNAseq data of 102 patients with biliary atresia and 14 non-BA disease control patients.^37,38 Gene expression data were retrieved from previous deposition in the Gene Expression Omnibus database under accession number GSE15235 for the US cohort and in the Genome Sequence Archive of Beijing Institute of Genomics, Chinese Academy of Sciences under accession number HRA003331 for the China cohort. Clinical parameters including survival, ALT, bilirubin, histological inflammation and fibrosis of the US cohort and China cohort were retrieved from the original pubulications.^37,38 2 patients in the US cohort were removed due to dropped off the study as described in the original publication.³⁰ Then, based on the median level of MAIT cells, patients were dichotomized into MAIT high- and low groups for two cohorts separately. For survival analysis, patients were followed started at the time of HPE surgery and ended at 24 months after HPE as described in the original publication.^37,38 None of these survival data are censored. Kaplan–Meier survival plots were generated and differences in survival were tested by Log-rank test. For other clinical parameters, spearman correlation was used to examine the correlation with level of MAIT cells and data were plotted using corrplot R package. The degree of correlation were defined by the following cutoff: absolute values of spearman correlation coefficient rho, 0–0.19 was regarded as very weak, 0.2–0.39 as weak, 0.40–0.59 as moderate, 0.6–0.79 as strong and 0.8–1 as very strong correlation.³⁹ The marker genes for MAIT cells were obtained from previous publications^40,41 and detailed listed at Supplemental Table S4.

Immunofluorescence staining (IF)

Immunofluorescence on frozen section was performed to examine localization of MAIT cells.

Liver mononuclear cell (LMNC) and PBMC isolation

LMNCs and PBMCs were isolated using density gradient centrifugation and cultured in RPMI medium (Corning, NY, USA) supplemented with 10% fetal bovine serum (FBS; FSP500, ExCell Bio, South America) and 1% Gibco Antibiotic-Antimycotic (15140122, Thermo Fisher Scientific, Waltham, MA, USA). Both cells were maintained in 37 °C incubators with 5% CO₂. LMNCs were isolated from liver tissues of control or BA patients. Liver tissue were dissociated in RPMI medium and filtered on a 70 μm cell strainer. LMNCs were isolated on a 33% Percoll density gradient.

In vitro stimulation assays

For detection of intracellular cytokines, LMNCs or PBMCs were stimulated for 4 h at 37 °C with phorbol 12-myristate 13-acetate (PMA; P1585, Sigma–Aldrich, St. Louis, MO, USA) and ionomycin (I0634, Sigma–Aldrich) at 10 ng/mL and 0.5 μg/mL, respectively, in the presence of brefeldin A (B7651, Sigma–Aldrich) at 3 μg/mL.

For IL-12 and IL-18 stimulation, cells were stimulated in vitro with 50 ng/mL IL-12 (200-12-10, PeproTech, CA, USA) and IL-18 (9124-IL-010, R&D Systems, Minneapolis, MN) for 24 h.

Flow cytometry

Human MAIT cells were identified as CD3⁺CD161⁺Vα7.2⁺ cells. Dead cells were excluded from the analysis using the Ghost Dye™ Red 710 (13-0871, Tonbo Biosciences, San Diego, CA, USA). Cells were labeled with cell surface antibodies before fixation and permeabilization using the Cytofix/Cytoperm slon kit (554,714, BD Biosciences, San Jose, CA, USA) as per the manufacturer's instructions. Cells were subsequently washed and stained for intracellular antibodies for flow cytometric analysis (LSR Fortessa X-20; BD Biosciences).

MAIT cells sorting

MAIT cells from BA or control patients were enriched as CD3 and Vα7.2 double positive cells from LMNCs or PBMCs by fluorescence-activated cell sorting on a LSR Fortessa X-20 (BD Biosciences). RRV-primed or uninfected MAIT were sorted as CD3⁺ Vα7.2⁺ CD161⁺ cells.

RNAseq analysis of MAIT cells

MAIT cells were enriched from LMNCs from adjacent non-tumor liver tissues of HB patients (n = 4) or BA patients (n = 6). Total RNA were used as input material and libraries were prepared by following the SMARTer Stranded Total RNA-Seq kit-Pico Input Mammalian user manual in shanghai biotechnology. The RNA sequencing data has been deposited in GEO database (GSE217466) and can be accessed by token wrufaauodxczhit.

Wound-healing assay

Human cholangiocyte cell line H69 were cultured in 96-well plates. A total of 2 × 10⁴ H69 cells were seeded per well in a 96-well clear flat bottom plate (Corning) and grown to confluency at 37 °C for 2 days and then scratched and washed with PBS before supplementing with different supernatants. Supernatants were collected at 5 days of sorted MAIT cells and diluted with RPMI medium containing 2% FBS in a ratio of 1: 3. Time lapse imaging was recorded every 4 h using microscopy (CKX53, Olympus, Tokyo, Japan) for 24 h. For anti-AREG blocking, 1 μg/mL of the polyclonal anti-AREG antibody was added to the supernatants. The open wound area were analyzed using ImageJ. The percentage of wound healing is defined as the quotient of the initial wound area and the wound area after a fixed time interval.

Cell counting assays

A total of 3 × 10³ H69 cells per well were seeded in 96-well cell culture plates. Next day, 100 ng/mL cytokines were added to the cells or sorted MAIT cells were co-culture with the H69 cells. For anti-AREG blocking, 2 μg/mL of the polyclonal anti-AREG antibody was added. Cell viability was tested by Cell Counting Kit-8 (CCK8) assay (HY-K0301, MCE, New Jersey, United States) after 48 h.

In vitro co-culture experiments of human cholangiocytes and PBMCs

Human cholangiocyte H69 were grown to 80% confluency before infection. RRVs were treated with trypsin at a concentration of 4 μg/mL in DMEM (Corning) for 1 h. Cells were washed with PBS for three times and infected with pretreated RRVs at an MOI of 1 for 1 h. Cells were then washed with PBS and co-cultured with freshly isolated PBMCs derived from healthy donors at the ratio of 1:2 (H69: PBMC). After co-culturing for 48 h, living CD45⁺ cells were sorted using Ghost Dye™ Red 710 and PE/Cyanine7-conjugated anti-CD45 antibody.

For intracellular staining, PBMCs co-culturing for 18 h were stimulated with PMA/ionomycin/brefeldin A for 4 h and the intracellular cytokine expression was analyzed by flow cytometry. For anti-MR1 blocking, 2 μg/mL of the monoclonal anti-MR1 antibody was added during co-culturing.

Single cell RNA sequencing (scRNA-seq) and data analysis

Single cell RNA sequencing libraries were performed according to the manufacturer's instructions (single cell 3′ v2 protocol, 10× Genomics). The libraries were sequenced in Illumina NovaSeq 6000. The demultiplexing, barcoded processing, gene counting and aggregation of sequencing data were made using the Cell Ranger software v2.1.1. The scRNA-seq data has been deposited in GEO database (GSE225177) and can be accessed by token utyngeyohhqpjsb. Sequencing data were processed with Seurat (v3) single-cell analysis pipeline with modifications. All subsequent scRNA-seq analyses were conducted using the Seurat R package (v3.2.2). Detailed information of scRNA-seq analyses is provided in the Supporting materials.

Pairwise overlapping comparisons of differentially expressed genes between RRV infected cell system and human BA liver

Differentially expressed genes for the RRV infected cell system were identified between RRV infected and uninfected groups. Differentially expressed genes for human BA livers were identified between 121 BA livers and 7 normal livers from our previous publication.³³ Pairwise overlapping comparisons were examined by Fisher's exact test in R.

Estimation of cell abundance of immune cell types in 121 patients with biliary atresia and 7 normal controls

Level of immune cell types identified in the RRV infected cell system were measured in 121 patients with BA and 7 normal controls using ssGSEA method with top 20 upregulated genes for each immune cell type and liver expression data as input as previously described.³⁶ Differnece of each immune cell type between BA and NC were tested by wilcoxon rank sum test. Liver expression data of 121 patients with BA and 7 normal controls was obtained from our previous publication.³³ Top 20 upregulated genes for each immune cell type were identified from scRNA-seq of the RRV infected cell system and listed in Supplemental Table S5.

MAIT cells and biliary organoid co-culture

Human biliary organoids were derived from adjacent normal tissues of hepatoblastoma from two patients and cultured as previous described.⁴² For co-culturing, biliary organoids were trypsinized to single cells for 10 min, After dissociation, 3 × 10³ cells was resuspended in growth factor reduced Matrigel (356,231, Corning). Droplets (25 μL) were plated per well into a 24-well plate and allowed to solidify for 10 min at 37 °C before adding 500 μL expansion medium, in which the concentration of recombinant human epidermal growth factor (hEGF; AF-100-15, PeproTech) was adjusted to 2.5 ng/mL. Sorted MAIT cells were added to the media for co-culture. For anti-AREG blocking, 2 μg/mL of the polyclonal anti-AREG antibody was added to the media and supplied every 2 days.

Generation of biliary organoid–LX-2 multicellular system

Biliary organoids were trypsinized to single cells and mixed with LX-2 cells at the ratio of 1:2 (organoid: LX-2). 3 × 10³ mixed cells was resuspended in Matrigel (25 μL) and plated into a 48-well plate for 7 days in expansion medium without TGF-β inhibitor A83-01 and Rho kinase inhibitor Y-27632 dihydrochloride, and then co-cultured with sorted MAIT cells for 3 days.

Analysis of gene expression

The levels of RNA were detected by real-time quantitative RT-PCR (qPCR).

Statistics

Correlations among liver MAIT cells, clinical parameters were performed using spearman correlation and plotted with corrplot R package. For statistical analysis, every patient represented once in any statistical test.

Data are expressed as the mean ± standard deviation (SD) of the indicated number of samples and are from at least three independent experiments or one experiment representative of at least three performed. Differences between two groups were analyzed using either Student's t-test when the variances are equal (normally distributed) or Mann–Whitney test when they are not. All statistical tests were two sided. Differences were considered statistically significant at p < 0.05. Analyses were performed using the GraphPad Prism program V.6 (GraphPad, San Diego, CA).

Role of funders

The funders were not involved in study design, data collection, analysis, interpretation, or writing.

Liver MAIT cells are associated with low survival, advanced fibrosis and ductular profiles in BA

To investigate the clinical implications of liver MAIT cells in human BA, we first performed survival analysis of liver MAIT cells with transplant-free survival after HPE in two cohorts of BA patients. The first cohort (the US cohort) was microarray gene expression data of 47 BA patients in USA, while the second cohort (the China cohort) was bulk liver RNAseq data of 102 BA and 14 non-BA disease control patients^37,38 at the time of diagnosis. 2 patients in the US cohort were removed due to dropped off the study as described in the original publication.³⁰ The abundance of liver MAIT cells were estimated with ssGSEA using marker genes for MAIT cells (Supplemental Table S4). ssGSEA is a gene rank-based method that calculates an enrichment score for each sample and indicates the abundance of MAIT cells. When dichotomizing the patients into MAIT high- and low groups based on median level of liver MAIT cells, we found that patients with high level of liver MAIT cells (MAIT high group) displayed lower survival in both cohorts (Log-rank P = 0.0108 in the US cohort and P = 0.0144 in the China cohort, Fig. 1A and B).

Liver injury, ductular reaction and liver fibrosis has been reported to predictive of low survival in BA.^12,33,43 Thus, we performed correlation analysis among levels of liver MAIT cells, clinical parameters at diagnosis [Serum ALT and conjugated bilirubin (CB) or total bilirubin (TB)], liver pathology (histological inflammation and fibrosis), and expression of KRT19 (a marker for cholangiocytes), COL1A1 (a marker for fibroblasts). Among these parameters, liver MAIT cells were positively correlated with histological fibrosis (rho = 0.31, p = 0.035 in US cohort; rho = 0.23, p = 0.006 in China cohort), expression of KRT19 (rho = 0.76, p < 0.0001 in US cohort, rho = 0.33, p = 0.005 in China cohort) and COL1A1 (rho = 0.4, p = 0.011 in US cohort, rho = 0.21, p = 0.007 in China cohort) in both cohorts (Fig. 1C and D). These human data suggest molecular links among liver MAIT cells, ductular reaction and liver fibrosis.

Liver MAIT cells exhibit activation phenotype and reside around portal tracts in patients with BA

To understand function of MAIT cells in BA, we obtained liver samples and peripheral blood were obtained from patients with BA and controls. Clinical information of these patients were summarized in Supplemental Table S3. To investigate how liver MAIT cells regulate ductular reaction, we first examined the abundance of liver MAIT cells in clinical samples. Using flow cytometry analysis, we found that the frequency of MAIT cells (CD3⁺CD161⁺Vα7.2⁺) among CD3⁺ T cells were significantly reduced in both liver and peripheral blood in BA when compared with the control (Fig. 2A). The frequency of MAIT cells were confirmed by using MR1-5-OP-RU Tetramer staining, which was comparable with staining with Vα7.2 (Supplemental Fig. S1). We thus used Vα7.2 staining for identifying MAIT cells hereafter. MAIT cells from BA exhibited higher activation (marked by HLA-DR and CD38) (Fig. 2B). This is consistent with previous studies showing that liver MAIT cells decreased but were activated in liver diseases.^21,44

The localization of liver MAIT cells was visualized by immunofluorescence staining for CD3 and TCR Va7.2 in human liver samples from BA and adjacent normal liver tissues of hepatoblastoma as control. Most MAIT cells (marked by CD3⁺ TCR Vα7.2⁺) localized around portal tracts in both BA and control (Fig. 2C and Supplemental Fig. S2). Examining the correlation among the abundance of liver MAIT cells measured by flow cytometry, level of clinical parameters and expression of cholangiocyte marker KRT19, fibrosis marker COL1A1 in the same patients (n = 12 for BA, n = 12 for control), we observed that liver MAIT cells were positively correlated with histological inflammation (Spearman correlation rho = 0.46 and p = 0.0485) and fibrosis (rho = 0.47 and p = 0.0102) (Fig. 2D and Supplemental Table S6). Moreover, liver MAIT cells were correlated with expression of KRT19 and COL1A1 only in BA patients (rho = 0.47 and p = 0.0210 for KRT19, rho = 0.50 and p = 0.0296 for COL1A1) but not in control (Fig. 2D and Supplemental Table S6).

Activated MAIT cells highly express AREG and promote cholangiocyte proliferation and migration

To understand molecular characteristic of MAIT cells, we sorted MAIT cells from liver samples of BA and from adjacent normal liver tissue of hepatoblastoma (as control) for bulk RNAseq analysis. We identified 1199 upregulated and 728 downregulated genes in liver MAIT cells from BA (Fig. 3A). Differential expression analysis revealed that several growth factors (such as AREG) were upregulated in BA MAIT cells (Fig. 3A). Among these growth factors, AREG showed the largest upregulation (14.8 fold upregulated, p = 0.0000157) in BA MAIT cells (Supplemental Table S7). Functional enrichment analysis showed that top 10 upregulated biological processes (such as response to wounding, tube morphogenesis, epithelium development) were associated with ductular reaction, while neural related biological processes were downregulated in BA compared to control (Fig. 3B and C). These data was in line with our clinical observations in two BA cohorts that MAIT cells might promote maladaptive wound healing response in BA.

To determine MAIT function on cholangiocyte, we performed wound-healing assay by treating the human cholangiocyte cell line H69 with culture supernatants of MAIT cells derived from BA or control. Supernatants obtained from BA MAIT cell cultures significantly accelerated wound closure (Fig. 3D). Similary, the proliferation-promoting effect of MAIT cells on human cholangiocyte H69 was also validated by CCK8 assay (Fig. 3E). These data suggest that MAIT cells from BA promote cholangiocyte proliferation and migration.

To examine by which growth factor MAIT cells promote cholangiocyte proliferation, we first performed CCK8 assay of cholangiocyte cell line H69 under treatment of AREG, HB-EGF, IGF2 or VEGFA (candidates from Fig. 3A). We found that AREG and HB-EGF had the strongest and similar effect on promoting proliferation of cholangiocyte cell line H69 (Supplemental Fig. S3). As AREG displayed the highest upregulation among these growth factors in sorted BA MAIT cells (from Fig. 3A and Supplemental Table S7), we focused on AREG in the following experiments. To validate AREG expression in protein level, we detected the production of AREG and other effector cytokines in MAIT cells via flow cytometry. Consistent with the result of RNAseq, liver MAIT cells in BA highly expressed growth factor AREG as well as peripheral BA MAIT cells, while other pro-inflammatory or cytotoxic factors remains unchanged (Fig. 3F). Testing the effect of MAIT cells derived AREG on cholangiocytes, we found that accelerated wound closure induced by MAIT cells were blocked by AREG antibody (Fig. 3G). These results together provide initial evidence that MAIT cells in BA exhibited a wound healing signature and promoted cholangiocyte proliferation via AREG.

A RRV infected cell system recapitulates immune dysfunction of human BA

As human liver samples of BA were limited, using clinical liver samples for functional and cellular study was challenged sometimes. Moreover, we observed that MAIT cells in neonatal mice were undetectable in naive or RRV-induced mice (Supplemental Fig. S4), hampering the study of MAIT cells using disease mouse model of BA. To further study human MAIT cells in BA, we developed a new cell system by co-culture of RRV infected human cholangiocyte H69 with human PBMC from healthy donors (Fig. 4A). In this system, human cholangiocyte cell line H69 were infected by RRV, an important cause of BA.^45,46 The RRV infected or uninfected (as control) cholangiocytes were co-cultured with human healthy PBMC to mimic the immune environment in human BA (Fig. 4A). To characterize this cell system, we sorted CD45 positive cells from both groups for single cell RNA sequencing (Fig. 4A). The scRNA-seq analysis identified 10,214 and 10,028 cells in uninfected and RRV infected groups, respectively. This analysis revealed major immune cell types including naïve T cells (CD3D⁺CCR7⁺SELL⁺), memory T cells (CD3D⁺S100A4⁺), MAIT cells (CD3D⁺SLC4A10⁺), cytotoxic T cells (CD3D⁺GZMB⁺PRF1⁺), CD56 dim NK cells (NCAM1⁺KLRF1⁺), CD56 bright NK cells (NCAM1⁺⁺KLRF1⁺), naïve B cells (MS4A1⁺IGHD⁺IGHG1^-), memory B cells (MS4A1⁺IGHD⁻IGHG1⁺), monocytes (CD14⁺LYZ⁺) and dendritic cells (CD14⁻LYZ⁺) (Fig. 4B and C and Supplemental Table S8). Comparing these immune cells between RRV infected and uninfected groups, we found that MAIT cells were decreased while cytotoxic T cells, naïve B cell were increased in RRV infected group compared to uninfected group (Fig. 4D and Supplemental Table S8). These data was in line with several previous observations that innate and adaptive immune cells were increased and infiltrated in BA livers.^47,48

In addition, differential expression and gene set enrichment analysis (GSEA) between RRV infected and uninfected groups showed that enrichment of immune related hallmarks gene sets including HALLMARK_INFLAMMATORY_RESPONSE (NES = 1.25, FDR = 0.019) and HALLMARK_INTERFERON_GAMMA_RESPONSE (NES = 1.29, FDR = 0.011) in the RRV infected group (Fig. 4E). These data suggest that the RRV infected cell system display immune features akin to RRV infected mice model and human BA patients.^49,50 To evaluate clinical implication of this RRV infected cell system, we performed gene overlapping analysis of differentially expressed genes between the RRV infected cell system and human BA livers. Differentially expressed genes for human BA livers were identified between 121 BA livers and 7 normal livers from previous publication.³³ This gene overlapping analysis revealed a transcriptomic similarity between the RRV infected cell system and human BA livers (Fig. 4F). To assess cell similarity, we estimated immune cell types in these 121 BA livers and 7 normal livers with ssGSEA method using top 20 upregulated genes (as gene signature) for each immune type identified in scRNA-seq data of the RRV infected cell system. In line with several previous studies, BA livers contain high levels of cytotoxic T cells, NK cells, B cells, monocytes and dendritic cells (Fig. 4G).^47,51,52 Consisent with flow cytometry data in Fig. 2A, BA livers showed less abundance of MAIT cells compared to control (Fig. 4G). All these data together indicate that the RRV infected cell system exhibit immune dysfunction similar to those in human BA liver.

RRV-primed MAIT cells display wound healing signature and promote cholangiocyte proliferation and migration via AREG

Having established a RRV infected cell system for human BA, we next sought to investigate the function of MAIT cells in this cell system. GSEA analysis revealed that RRV-primed MAIT cells (from RRV infected group) displayed enrichment of wound healing signatures (gene set: GOBP_WOUND_HEALING, GOBP_RESPONSE_TO_WOUNDING, GO_INFLAMMATORY_RESPONSE_TO_WOUNDING), suggesting MAIT cells in this cell system also display wound healing signature as liver MAIT cells from clinical samples (Fig. 5A). Examing wound healing related upregulated genes (from upregulated genes in Fig. 3B) in scRNA-seq data of RRV infected cell system, we found that AREG had the highest expression level among them and that RRV primed MAIT cells exhibited higher expression of AREG when compared to uninfected MAIT cells (Fig. 5B). These data was in keep with the findings of bulk RNAseq data of BA MAIT cells (see Fig. 3A and Supplemental Table S7). Therefore, we employed this cell system for functional studies (Fig. 5C). Flow cytometry analysis showed a decreased frequency of MAIT cells (Fig. 5D), activation of MAIT cells (Fig. 5E) and increased expression of AREG in MAIT cells (Fig. 5F) in the RRV infected group, which was similar with the observation in liver MAIT cells from clinical samples (see Fig. 2A and B and Fig. 3F). Wound-healing assay revealed that supernatants from RRV-primed MAIT cells culture promoted cholangiocyte wound closure, while antagonism of AREG abrogated the effect (Fig. 5G and H). Cell proliferation CCK8 assay also showed that RRV-primed MAIT cells promoted cell growth of H69 via AREG (Fig. 5I).

MAIT cells promote biliary organoid growth via AREG

Biliary organoids are self-organizing three-dimensional structures derived from bile ducts and suitable for the study of bile duct and disease modelling.⁵³ Thus, we co-cultured human biliary organoids derived from normal liver tissues with MAIT cells to further explore its function in promoting ductular reaction. The viability and functionality of biliary organoids were confirmed by a 7-days time-lapse imaging (Supplemental Fig. S5A) and Rhodamine123 assay (Supplemental Fig. S5B) as previously described.⁵⁴ As expected, biliary organoids co-cultured with MAIT cells from human BA liver significantly increased their diameters as compared to control (Fig. 6A). Similarly, diameters of biliary organoids significantly increased when co-cultured with RRV-primed MAIT cells from our RRV infected cell system (Fig. 6B). This effect was abrogated by blocking AREG (Fig. 6B). These results confirmed that MAIT cells promoted biliary organoid growth via AREG.

MAIT cells promote ductular reaction in biliary organoid–LX-2–MAIT cells multicellular system

In addition to cholangiocyte proliferation, inflammation and fibrosis are key features of ductular reaction. We noticed that MAIT cells were positively correlated with liver fibrosis (see Fig. 1C and D). To further explore the effect of MAIT cells in inducing inflammation and fibrosis, we evaluated MAIT cell function in a human biliary organoid–hepatic stellate cells LX-2–MAIT cells multicellular system (Fig. 6C). RRV-primed MAIT cells upregulated the expression of inflammatory genes CCL2, IFNG and profibrogeneic factor TGFB1 (Fig. 6D) in the human biliary organoids–LX-2 multicellular system. Moreover, hepatic stellate cells activation marker TIMP1 were induced by RRV-primed MAIT (Fig. 6E). These data suggested that inflammation and hepatic stellate cells activation were induced by RRV-primed MAIT cells following cholangiocyte proliferation. Interestingly, no difference was found in the activation of LX-2 when it was treated with supernatants from BA patients (Supplemental Fig. S6A) or co-cultured with RRV-primed MAIT cells (Supplemental Fig. S6B). These data together implicate that MAIT cells directly act on cholangiocyte followed by activation of inflammation and fibrosis, leading to ductular reaction eventually.

AREG expression in MAIT cells is TCR-dependent

MAIT cells can be activated by cytokines (such as IL-12 and IL-18) or TCR-dependent pathways.¹⁷ To investigate regulation of AREG expression in MAIT cells, we first treated MAIT cells isolated from clinical samples of BA or control with exogenous IL-12 and IL-18. AREG expression remains unchanged after IL-12 and IL-18 stimulation in both control and BA groups whereas IFN-γ and Granzyme B were increased (Fig. 7A). Next, to test if AREG expression was dependent upon MR1-antigen-TCR interaction, we blocked the TCR mediated activation with a monoclonal anti-MR1 antibody in RRV infected cell system. The AREG production of RRV-primed MAIT cells was significantly reduced upon TCR blockade, while there was no difference in IFN-γ and Granzyme B (Fig. 7B). Taken all together, these data indicate that AREG production is dependent on TCR stimulation.

Liver AREG expression is positively correlated with expression of MR1, ductular reaction markers, chemokines and RAS-MAPK

To further understand potential upstream and downstream factors of AREG, we performed correlation analysis between AREG and some key factors related to BA pathogenesis^4,50 in liver gene expression of two BA cohorts (the US and China cohorts we used in Fig. 1). Confirming the observations that AREG production was dependent on TCR stimulation and that AREG promoted ductular reaction, liver AREG were positively correlated with MR1, ductular reaction marker (KRT19, COL1A1) (Fig. 7C). Interestingly, AREG had strong correlation with CCL2, CXCL2, CXCL8, suggesting that MAIT cells might have interaction with other immune cells such as macrophages and neutrophils. Regarding the downstream effectors, AREG were linked to RAS-MAPK signaling (KRAS, MAPK1, MAP2K1).