How ALKBH5 changes TIRAP RNA to increase radiation-related liver scarring and reduce liver cancer sensitivity to radiation

ALKBH5 was up‐regulated in irradiated LX2 cells, especially at 8 Gy dose (Figure S1A). Notably, ALKBH5 promoted irradiated LX2 activation and reduced cell apoptosis (Figures 1A and S1B). We also confirmed the conclusion in isolated primary mouse HSC (Figure 1H–J). To identify the genes regulated by ALKBH5, we performed an integrated approach combining methylated RNA immunoprecipitation‐sequencing (MeRIP‐seq) and RNA‐sequencing (RNA‐seq). The common motif of m⁶A in LX2 cells was GGACUU (Figure S1D). Compared with the control group, there were less unique m⁶A peaks (10 523 vs. 11 547) and less genes with unique modification peaks (9581 vs. 10 261) in ALKBH5 knockdown group (Figure S1E). There were 3549 differential m⁶A peaks (fold change ≥ 2, p < .05) and 485 differentially expressed genes (fold change ≥ 2, p < .05) in ALKBH5 knockdown group, and 56 genes that showed differences in m⁶A peaks and gene expression were obtained by integrated analysis. Then the co‐expression between 56 genes and ALKBH5 was verified in normal liver tissues of the TCGA database and Genotype‐Tissue Expression database, and 17 genes co‐expressed with ALKBH5 were obtained by intersection. ALKBH5 promoted the activation of NF‐κB (Figure S1C), whereas NF‐κB pathway activation is related to the proliferation and activation of HSC that promotes the development of RILI.7, 8 After intersection with the above 17 genes and NF‐κB pathway gene set from the KEGG database (https://www.kegg.jp/↗), only toll‐interleukin 1 receptor domain‐containing adaptor protein (TIRAP), the upstream regulatory molecule of NF‐κB, was identified to be the target gene for ALKBH5 to regulate the activation of NF‐κB pathway. Further analysis showed that the m⁶A site of TIRAP was distributed in 3' untranslated region (3′UTR) (Figure 1C). ALKBH5 promoted m⁶A modification of TIRAP and bound to the wild‐type binding sites in TIRAP 3′UTR, suggesting that ALKBH5 mediates the m⁶A modification of TIRAP mRNA (Figure 1D,E). In addition, actinomycin D assay showed that ALKBH5 improved the stability of TIRAP mRNA (Figure 1F). Previous studies have reported that m⁶A reader YTH N6 methyladenosine RNA binding protein 2 (YTHDF2) enables to recognise m⁶A and promotes the degradation of hypermethylated mRNA.24 Therefore, we further explored the effect of YTHDF2 on TIRAP expression. As shown in Figure 1G, YTHDF2 knockdown restored the down‐regulated expression of TIRAP caused by ALKBH5 knockdown, indicating that TIRAP mRNA was hypermethylated and then recognised and degraded by YTHDF2. These results suggest that ALKBH5 mediates the m⁶A modification of TIRAP mRNA and regulates TIRAP expression in a YTHDF2‐dependent manner.

Down‐regulation of TIRAP could simulate the effect of ALKBH5 knockdown, that is, inhibiting the activation and proliferation of LX2 cells and promoting apoptosis (Figure 2A,B), while TIRAP overexpression reversed the effects of ALKBH5 inhibition (Figure 2C,D). ALKBH5 upregulated the expression of TIRAP and its downstream genes interleukin 1 receptor associated kinase 1 (IRAK1) and TNF receptor associated factor 6 (TRAF6), promoted the activation of NF‐κB and JNK/Smad2 and led to up‐regulation of anti‐apoptosis factor Bcl‐2 and activation markers (α‐SMA and collagen 1), while TIRAP knockdown partially alleviated the promoting effects of ALKBH5 overexpression (Figure 2E). On the contrary, silencing ALKBH5 down‐regulated the expression of TIRAP and downstream effectors in LX2 (Figure 2E) and primary HSCs (Figure 1K), while TIRAP overexpression could reverse the inhibitory effects of ALKBH5 (Figure 2E). These results suggest that TIRAP is the target gene that mediates the regulatory effects of ALKBH5.

Interestingly, we also found that TIRAP can up‐regulate ALKBH5 expression (Figure 2F), suggesting that there is a positive feedback loop between ALKBH5 and TIRAP. It was predicted by the JASPAR database (https://jaspar.genereg.net/↗) that there were several binding sites in ALKBH5 promoter region for NF‐κB p65, the downstream effector of TIRAP (Figure 2H). Furthermore, we found that NF‐κB p65 inhibitor JSH‐23 suppressed ALKBH5 expression and inhibited the binding of NF‐κB p65 and binding site 2 in ALKBH5 promoter (Figure 2G–I). These results indicate that the activation of TIRAP/NF‐κB pathway promotes ALKBH5 expression, and there is a positive feedback regulation between ALKBH5 and TIRAP.

In the mouse model of RILI, the ALKBH5/TIRAP axis was significantly activated in liver tissue and HSC (Figure S2A–C). HSC activation markers α‐SMA and collagen 1 deposition were mainly distributed around the vessels (Figure S2D,E), suggesting that activated HSC is mainly distributed around the hepatic vessels after irradiation, which provides favourable conditions for HSC to participate in the recruitment of peripheral blood cells into the liver. Compared with the non‐irradiation group, more F4/80⁺ macrophages infiltrated the liver of the irradiation group (Figure S2F). CLEC4F is used to identify Kupffer cells, while IBA1 is a pan‐macrophage marker of Kupffer cells and infiltrating monocytes/macrophages. After irradiation, IBA1⁺CLEC4F⁻ monocytes were recruited around the injured hepatic sinusoids, while IBA1⁺CLEC4F⁺Kupffer cells were mainly distributed in liver tissue (Figure S2G), suggesting that irradiation promoted peripheral blood monocyte infiltration.

To test whether IR‐HSC promoted monocyte migration and polarisation, the human monocyte THP‐1 cells were co‐cultured with the culture medium of irradiated LX2 cells (IR‐LX2 CM). We found that the migration of THP‐1 was increased, the expression of M2‐related markers (CD163, CD206, IL10) was increased, while the expression of M1‐related markers (interleukin 1 beta (IL1B), tumor necrosis factor (TNF), nitric oxide synthase 2 (NOS2)) was reduced (Figure). Similar results were observed in the co‐culture experiment of mouse IR‐HSC CM and bone marrow‐derived monocytes (BMDMs) (Figure). These results suggested that IR‐HSC promoted monocyte migration and polarisation to M2. However, the ALKBH5‐silenced IR‐LX2 CM significantly reduced THP‐1 migration and polarisation to M2 (Figure). S3A–C S5A S3E,F

To identify the role of ALKBH5 in monocyte infiltration and polarisation, MeRIP‐seq and RNA‐seq revealed that ALKBH5 regulates the m⁶A modification and expression of chemokines CCL5 and CXCL12 (Figure 3A). CCL5 expression was up‐regulated in IR‐LX2 cells, irradiated liver tissue and HSC of mice, but there was no significant change in CXCL12 (Figure 3A,B). Notably, CCL5 expression in isolated irradiated mouse HSC was down‐regulated by silencing ALKBH5 (Figure 3B). ALKBH5 mediated m⁶A modification of CCL5 and improved the stability of CCL5 mRNA (Figure 3C,D). Silencing YTHDF2 restored the down‐regulation of CCL5 caused by ALKBH5 knockdown (Figure 3E), indicating that ALKBH5 mediates the m⁶A modification of CCL5 mRNA and regulates CCL5 expression in a YTHDF2‐dependent manner. In addition, we also found that TIRAP can reverse the regulation of ALKBH5 on CCL5 expression (Figure 3F). Inhibition of NF‐κB activation by JSH‐23 also down‐regulated the expression of CCL5 (Figure 3G). Collectively, these results suggested that ALKBH5 also promotes CCL5 expression through the activation of TIRAP/NF‐κB pathway.

ALKBH5 promoted the secretion of CCL5 by HSC (Figure S3D). Adding CCL5 neutralising antibody to the co‐culture of IR‐LX2 CM and THP‐1 could inhibit the migration of THP‐1 (Figure 3H). Since the common receptors of CCL5 are CCR1, CCR3 and CCR5, THP‐1 cells were transfected with CCR1, CCR3 and CCR5 siRNA or pretreated by maraviroc (a selective inhibitor of CCR5) and then co‐cultured with IR‐LX2 CM. We found that THP‐1 migration was inhibited by CCR5 siRNA or using maraviroc but not CCR1 or CCR3 siRNA (Figure 3I). CCL5 neutralising antibody or maraviroc pretreatment reduced the morphological change and M2 polarisation of THP‐1 cells (Figure 3I,J). Similar results were observed in the co‐culture of mouse IR‐HSC CM and BMDMs (Figure S4B,C). It has been reported that the activation of AKT, signal transducer and activator of transcription (STAT), extracellular regulated protein kinase (ERK), JNK and NF‐κB pathways could be induced by CCL5/CCR5 axis.25 In this study, IR‐LX2 CM promoted AKT phosphorylation and inhibited STAT3 phosphorylation in THP‐1 cells but had no effects on ERK, NF‐κB and JNK phosphorylation (Figure 3K1). In addition, CCL5 promoted AKT phosphorylation and inhibited STAT3 phosphorylation in THP‐1 cells in a dose‐dependent manner, which could be reversed by maraviroc blockade (Figure 3K2). Similar results could also be observed by adding CCL5 neutralising antibody or pretreatment with maraviroc in the co‐culture of IR‐LX2 CM and THP‐1 (Figure 3K3). In order to determine whether the CCL5 secretion of HSC mediated by ALKBH5 promotes the migration and polarisation of monocytes, LX2 cells were transfected with ALKBH5 siRNA and then stimulated with CCL5. Compared with the control group, ALKBH5‐silenced IR‐LX2 CM reduced THP‐1 migration, morphological change and M2 polarisation, promoted AKT phosphorylation and inhibited STAT3 phosphorylation, which could be reversed by CCL5 stimulation (Figures S3E,F and 3K4,M). After the same treatment as above, similar results were observed in the co‐culture of mouse IR‐HSC CM and BMDMs (Figure S5D,G). On the contrary, co‐culture of ALKBH5‐overexpressing IR‐LX2 CM and THP‐1 promoted THP‐1 migration, morphological change and M2 polarisation, inhibited AKT phosphorylation but promoted STAT3 phosphorylation, which could be reversed by CCL5 neutralisation antibody or maraviroc (Figure S4A–D). In addition, when THP‐1 cells were pretreated with AKT inhibitor MK2206 and co‐cultured with IR‐LX2 CM, the migration, morphological change and M2 polarisation of THP‐1 cells decreased (Figure S4E–G). Similar results were observed in the co‐culture of mouse IR‐HSC CM and BMDMs (Figure S5E,F). Moreover, maraviroc administration in the mouse model of RILI significantly decreased the recruitment of IBA1⁺CLEC4F⁻ monocytes in perivascular tissue (Figure S2G). Maraviroc treatment also significantly reduced the level of Ly6C⁺ monocyte and decreased the level of M2 marker CD163 more significantly than that of M1 marker CD86 (reduced 36.83% vs. 7.62%; Figure S2H), suggesting that blocking CCR5 predominantly inhibits the monocyte recruitment and M2 polarisation.

Next, we explored the role of ALKBH5 in HSC in the RILF model. We used HBAAV‐GFAP‐shALKBH5 to knock down ALKBH5 in HSC. As expected, HBAAV‐GFAP‐shALKBH5 could effectively down‐regulate the expression of ALKBH5/TIRAP axis in HSC (Figure 4A,B). HBAAV‐GFAP‐shALKBH5 inhibited α‐SMA and collagen 1 expression in liver tissue and HSC, decreased α‐SMA expression around the vessel and reduced perivascular collagen deposition (Figure 4C,D). In addition, HBAAV‐GFAP‐shALKBH5 treatment down‐regulated CCL5 expression in liver tissue and HSC (Figure 4E). As expected, the silence of ALKBH5 suppressed the infiltration of IBA1⁺CLEC4F⁻ monocytes in perivascular tissue and Ly6C⁺ monocytes in liver (Figure 4F,H). Notably, M2 marker CD163 decreased significantly, while M1 marker CD86 decreased slightly after ALKBH5 knockdown (reduced 51.23% vs. 17.95%; Figure 4G,H). In summary, ALKBH5 inhibition in HSC alleviates RILF and suppresses monocyte infiltration and M2 polarisation.

Next, we explored the effect of IR‐HSC‐educated monocytes on HSC. We found that the CM from IR‐LX2 CM‐stimulated THP‐1 cells could reduce the cell apoptosis, promote fibrotic markers expression and significantly up‐regulate the expression of ALKBH5, TIRAP, BCL2 and CCL5 of IR‐LX2 cells (Figure S6A–C). ALKBH5 knockdown in IR‐LX2 cells could reverse these effects (Figure S6D–F), indicating that these effects were correlated with the further upregulation of ALKBH5. Subsequently, the medium from IR‐LX2, blank THP‐1 and IR‐LX2 CM‐stimulated THP‐1 (IR‐LX2 CM‐THP‐1) was collected for cytokine microarray detection. Compared with the medium of IR‐LX2 and blank THP‐1, CCL20 was significantly increased in the medium of IR‐LX2 CM‐THP‐1 (Figure 5A). Consistently, CCL20 expression was significantly up‐regulated in THP‐1 cells after co‐culturing with IR‐LX2 CM (Figure S6G1). Additionally, CCL5 increased CCL20 expression in a dose‐dependent manner, while maraviroc blockade could reverse it (Figure S6G2). Co‐culturing with ALKBH5‐silenced IR‐LX2 CM decreased CCL20 expression in THP‐1 cells (Figure S6G3). Moreover, CCL20 expression was decreased in MK2206 pretreated THP‐1 cells co‐cultured with IR‐LX2 CM (Figure S6G3). Similar results were observed in the co‐culture of mouse IR‐HSC CM and BMDMs (Figure S8A–D). These results suggested that CCL5 secreted by IR‐HSC promotes monocyte to produce CCL20 through the AKT pathway.

Moreover, we found that CCL20 promoted NF‐κB activation and up‐regulated ALKBH5 expression in a dose‐dependent manner in IR‐LX2 cells (Figure S7A). CCL20 stimulation reduced the cell apoptosis and promoted CCL5 secretion as well as cell activation of IR‐LX2 (Figure S7B–D). Notably, CCR6 inhibitor down‐regulated the expression of CCL5 and ALKBH5/TIRAP axis and inhibited cell activation and NF‐κB pathway in CCL20‐treated IR‐LX2 cells (Figure S7E,F). In order to determine whether CCL20 regulates ALKBH5 expression through NF‐κB, LX2 cells were treated with CCL20 combined with JSH‐23. The results showed that JSH‐23 inhibited NF‐κB activation and ALKBH5/TIRAP axis in CCL20‐treated IR‐LX2 cells (Figure S7F). In addition, ALKBH5 knockdown inhibited NF‐κB activation and TIRAP downstream effectors expression, which could be partially reversed by CCL20 (Figure S7G). Furthermore, we collected IR‐LX2 CM‐stimulated THP‐1 CM to co‐culture with IR‐LX2 for validation (Figure 5B). The results showed that CCL20 neutralisation antibody, CCR6 inhibitor or JSH‐23 treatment increased cell apoptosis but reduced the expression of CCL5, cell activation markers and ALKBH5/TIRAP axis (Figure 5C–E). Similar results were observed in the mouse IR‐HSC stimulated by CCL20 or co‐cultured with IR‐HSC CM‐stimulated BMDM CM (Figure S8E–I). In summary, blocking the CCL20/CCR6 axis inhibits HSC activation and CCL5 production and down‐regulates the expression of ALKBH5/TIRAP axis mediated by NF‐κB activation in IR‐HSC.

Next, we explored the effect of IR‐LX2 CM‐stimulated THP‐1 CM on the radiosensitivity of HCC cells. We found that IR‐LX2 CM‐stimulated THP‐1 CM reduced the apoptosis of irradiated HCC (IR‐HCC) cells and up‐regulated Bcl‐2 expression (Figure S9A,B). Interestingly, ALKBH5 expression increased in IR‐HCC cells after co‐culture (Figure S9B). In IR‐HCC cells, ALKBH5 knockdown promoted cell apoptosis and inhibited cell proliferation (Figure S9C,D), suggesting that ALKBH5 inhibition increases the radiosensitivity of HCC cells. Notably, CCL20 stimulation reduced the apoptosis of IR‐HCC cells (Figure S9E). Similar to the effect of CCL20 on HSC, CCL20 also promoted NF‐κB activation and up‐regulated ALKBH5 expression in a dose‐dependent manner in IR‐HCC cells (Figure S9F). This finding inspired us to speculate whether ALKBH5‐mediated TIRAP m⁶A modification and expression also exist in HCC cells. MeRIP confirmed that ALKBH5 knockdown promoted m⁶A modification of TIRAP in HCC cells (Figure 5F). CCR6 inhibitor or JSH‐23 inhibited NF‐κB activation and ALKBH5/TIRAP axis in CCL20‐stimulated HCC cells (Figure S9G,H). CCL20 stimulation could partially reverse the effect that ALKBH5 knockdown inhibited the expression of TIRAP and downstream effectors in IR‐HCC cells (Figure S9G,H). Furthermore, in the co‐culture of IR‐LX2 CM‐stimulated THP‐1 CM and IR‐HCC cells, CCL20 neutralising antibody, CCR6 inhibitor or JSH‐23 treatment increased cell apoptosis and inhibited ALKBH5/TIRAP axis (Figure 5G,H). These results suggested that blocking the CCL20/CCR6 axis increases HCC radiosensitivity by inhibiting NF‐κB /ALKBH5/TIRAP axis.

Moreover, we also eliminated the effect of CCL5 expression of HCC cells on monocyte migration and polarisation. Irradiation did not affect CCL5 expression in HCC cells (Figure). IR‐HCC cells secreted a small amount of CCL5, and CCL20 stimulation could not induce HCC cells to produce CCL5 (Figure). In the co‐culture of IR‐LX2 CM‐stimulated THP‐1 CM and IR‐HCC cells, CCL20 neutralising antibody, CCR6 inhibitor or JSH‐23 treatment had no significant effect on CCL5 production (Figure). In addition, the silence of ALKBH5 in IR‐HCC cells did not affect CCL5 expression (Figure). Furthermore, the CM from CCL5‐silenced IR‐HCC cells had no effect on the migration and polarisation of THP‐1 cells (Figure). These results suggested that CCL5 expression of HCC cells has no effect on monocyte migration and polarisation. S10A S10B S10C S10D S10E,F

HBAAV‐TBG‐shALKBH5 or vector was injected through the tail vein 1 week before irradiation, CCR6 inhibitor was injected intraperitoneally twice a week after irradiation, and the experiment was completed 3 weeks after the last irradiation (Figure 6A). Treatment of HBAAV‐TBG‐shALKBH5 or/and CCR6 inhibitor significantly reduced the serum level of CCL5 and CCL20, suppressed the infiltration of Ly6C⁺ monocytes and M2 macrophage but not M1 macrophage in liver tissue (Figure 6B,C). Additionally, treatment of HBAAV‐TBG‐shALKBH5 or/and CCR6 inhibitor markedly decreased tumour volume and liver/body weight ratio, promoted tumour necrosis and inhibited ALKBH5/TIRAP axis in tumour (Figure 6D,E,G). Moreover, the serum level of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) increased markedly in irradiated mice but was significantly attenuated by HBAAV‐TBG‐shALKBH5 or/and CCR6 inhibitor treatment (Figure 6F). After combined treatment, α‐SMA and collagen deposition around the vessel and tumour decreased (Figure 6H). In summary, ALKBH5 inhibition can reduce monocyte infiltration and M2 polarisation, increase HCC radiosensitivity and alleviate RILF.

In order to better understand the relationship between ALKBH5/TIRAP axis and RILI and HCC radiosensitivity, we collected samples from patients responding to radiotherapy without concomitant RILI and patients showing nonresponse to radiotherapy but with concomitant RILI. High expression of ALKBH5/TIRAP axis was detected in tumour and paratumour tissues before or post radiotherapy from HCC patients who showed non‐response to radiotherapy but with concomitant RILI, compared to those patients responding to radiotherapy without RILI (Figure 7A–C), suggesting that these indicators may be used as potential markers for predicting radiotherapy response and complications of HCC. Moreover, we found that high expression of ALKBH5 and/or TIRAP was significantly associated with the poor overall survival of glioblastoma and thyroid cancer patients receiving radiotherapy from TCGA database(Figure 7D–E), suggesting ALKBH5 and TIRAP have prognostic value for certain patients receiving radiotherapy.

Male C57BL/6 mice (6‐7 weeks old and 16–20 g) were maintained under pathogen‐free conditions in the Experimental Animal Center of Southern Medical University. All animal experiments were specifically approved by Southern Medical University Experimental Animal Ethics Committee. ALKBH5‐short hairpin RNA (shRNA) encapsulated by adeno‐associated virus containing HSC‐specific promoter of glial fibrillary acidic protein (HBAAV‐GFAP‐shALKBH5, 1×10¹² vg/mL) or liver‐specific promoter (HBAAV‐TBG‐shALKBH5, 1×10¹² vg/mL) and vector (1×10¹² vg/mL) were designed and synthesised by Hanbio Biotechnology. For mice irradiation, the right liver lobe of mice was irradiated with X‐rays (8 Gy × 5 times, once a day). For the evaluation of maraviroc in modulating monocyte infiltration, intraperitoneal injection of maraviroc (10 mg/kg) was performed once a day after the initiation of irradiation. For the evaluation of ALKBH5 expression of HSC in modulating RILF and monocyte infiltration, HBAAV‐GFAP‐shALKBH5 or vector (100 uL per mouse) was injected once a week before irradiation. Three weeks after the last irradiation, the mice were sacrificed, and liver tissues were collected for histological, immunohistochemistry and immunofluorescence analysis and flow cytometry. The remaining fresh liver tissues were used for qRT‐PCR and western blot analysis.

Orthotopic HCC models were established according to the method described previously.37 Hepa 1–6 cells (5×10⁵) were suspended in 20 µL PBS and injected into the right liver lobe of C57BL/6 mice. After 2 weeks, the mice were randomly grouped and the right liver lobe was irradiated with X‐rays (8 Gy×5 fractions, once a day), and a non‐irradiated group was set up as a control. The irradiation groups were treated as follows: (1) single irradiation; (2) tail vein injection of HBAAV‐TBG‐shALKBH5 (100 uL per mouse) once a week before irradiation; (3) tail vein injection of vector (100 uL per mouse) once a week before irradiation; (4) intraperitoneal injection of CCR6 inhibitor (10 mg/kg, MedChemExpress) twice a week after initiation of irradiation; (5) tail vein injection of HBAAV‐TBG‐shALKBH5 combined with an intraperitoneal injection of CCR6 inhibitor. Tumour volume and liver/body weight ratio were evaluated at 3 weeks after the last irradiation. Tumour volume was calculated using the formula: tumourvolume (mm³) = (smallest diameter² × largest diameter)/2.38 The venous blood of mice was drawn from the eye orbit of mice, and then the mice were sacrificed at 3 weeks after the last irradiation. The levels of AST and ALT in serum were detected by an automatic biochemistry analyzer. The right liver lobes were collected for histological, immunohistochemistry and immunofluorescence analysis and for monocyte–macrophage detection by flow cytometry.

Seventeen HCC samples were collected from patients who received radiotherapy for intrahepatic tumour during 2018–2021 (for detailed patient information, see Table). Among them, 15 samples were collected before radiotherapy, and two samples were collected post radiotherapy. The radiotherapy doses ranged from 40 to 60 Gy with 5–20 fractions. All the patients conducted the computed tomography (CT) or magnetic resonance imaging (MRI) image examination before and after radiotherapy. The treatment effects were evaluated within 6 months after the last irradiation. Informed consent in writing was obtained from each patient. The study protocol was approved by the review committee of the Nanfang Hospital of Southern Medical University. S1

The treatment response was assessed according to the Modified Response Evaluation Criteria in Solid Tumour. After radiotherapy, it was defined as ‘response’ that the tumour volume was significantly reduced, the signs of ‘fast in and fast out’ were lost or the tumour had no obvious enhancement in the arterial phase, while the increased tumour volume, local recurrence or new lesions were defined as ‘nonresponse’. For image evaluation of RILI, in the area of focal oedema or congestion caused by veno‐occlusive disease and hepatic fibrosis after irradiation, the irradiated area usually displays hypoattenuation on unenhanced CT. On contrast‐enhanced CT, it could show hyperdensity in all enhanced phases, hypodensity in the arterial and portal venous phases or isodensity in all enhanced phases. For MRI, the irradiated area usually exhibits hypointensity on T1‐weighted images, hyperintensity on T2‐weighted images, and a defined hypointense area on the hepato‐biliary phases of gadolinium‐ethoxybenzyl‐diethy‐lenetriamine pentaacetic acid (Gd‐EOB‐DTPA)‐enhanced MRI.10

The rest of the materials and methods are included in the Supplementary Information.