Data collection and network analysis

The genetic changes of Hep3B cells after resistance to sorafenib in the GSE248769 dataset (n = 3), and expression profiling of paired sorafenib-resistant and -sensitive tumor-derived organoids from 4 HCC patients (3 replicates for each sample) from the GSE182593 dataset were obtained from the publicly available GEO database. Differential expression analysis of genes was performed in GEO2R (https://www.ncbi.nlm.nih.gov/geo/geo2r/), and p-values were corrected by Benjamini & Hochberg (false discovery rate). Genes that were DEGs in HCC-SR were screened using adjusted p < 0.01, |Log2FoldChange| > 1 as the cutoff threshold. The latest gene annotations of the KEGG pathway were obtained using KEGG rest API (https://www.kegg.jp/kegg/rest/keggapi.html) as a background, and DEGs were mapped to the background collection. The enrichment analysis was performed using the R package clusterProfiler (version 3.14.3). HumanTFDB (http://bioinfo.life.hust.edu.cn/HumanTFDB#!/) was used to obtain the human transcription factor (TF) list, and the Jaspar database (https://jaspar.elixir.no/) was searched to predict the binding sites of transcription factors on DNA fragments.

Data obtained from the TCGA-LIHC program were used for analyzing the prognostic significance of gene expression on overall survival (OS) and relapse-free survival (RFS) in HCC patients in Kaplan-Meier Plotter (https://kmplot.com/analysis/index.php? p=service&cancer=liver_rnaseq), for analyzing the correlation between gene expression in SangerBox (https://kmplot.com/analysis/index.php? p=service&cancer=liver_rnaseq), for correlation of immune cell infiltration in the SangerBox (http://vip.sangerbox.com), and analysis of the correlation between two gene expressions in HCC tumor tissues in the GEPIA (http://gepia.cancer-pku.cn/index.html).

Cell culture

Human HCC cells HCCLM3 (Serial: SCSP-5093) and HuH-7 (Serial: SCSP-526) were procured from the National Collection of Authenticated Cell Cultures (Shanghai, China). The mouse HCC cell line Hepa1-6 stably expressing firefly luciferase (Hepa1-6-Luc, Serial: ml-CC2168) was purchased from MLBio (Shanghai, China). All HCC cells were grown in DMEM (Gibco, Carlsbad, CA, USA) plus 10% FBS (Gibco). Transformed human liver epithelial-2 (THLE-2, CL-0833) was purchased from Procell (Wuhan, Hubei, China) and cultured using bronchial epithelial growth medium (CC-3170, Cellverse Bioscience Technology Co., Ltd., Shanghai, China) supplemented with 5 ng/mL EGF, 70 ng/mL Phosphoethanolamine, and 10% FBS. The culture environment for all cell lines was a 37 °C incubator in 5% CO2/95% air. All cells were identified by STR and were free of mycoplasma, bacterial, or fungal contamination.

Cell infection and treatment

Overexpression lentiviral vectors LV-AQP7, LV-MGRN1, LV-TBX19, shRNA lentiviral vectors LV-shAQP7 1# (GGAGGAAGTTTCCGGTCTATGCTCGAGCATAGACCGGAAACTTCCTCC), 2# (TGGAGGATTCTGTGGCGTATGCTCGAGCATACGCCACAGAATCCTCCA), LV-shMGRN1 1# (CCTGAGAGTTTCATAACAGAACTCGAGTTCTGTTATGAAACTCTCAGG), 2# (TGGAGCATGATGAGCACTCTTCTCGAGAAGAGTGCTCATCATGCTCCA), and the control vectors were designed and synthesized by VectorBuilder (Guangzhou, Guangdong, China) at a viral titer of 109 TU/mL. HCC cells were plated in 24-well plates at 5 × 104 cells/well, cultured overnight, and infected (MOI = 10). After 48 h of lentiviral infection, the cells were cultured in a puromycin-containing culture medium to screen for stable cells. Sorafenib (IS0220, Solarbio, Beijing, China), dissolved in DMSO, was treated at different concentrations for 24 h, whereas Fatostatin (HY-14452, MedChemExpress, Monmouth Junction, NJ, USA) was treated at 1 µM for 24 h to inhibit lipogenesis [11].

RT-qPCR

Total RNA from cells extracted with TRIzol (15596026CN, Invitrogen Inc., Carlsbad, CA, USA) was reverse transcribed by BeyoRT II cDNA First Strand Synthesis Kit (D7168L, Beyotime, Shanghai, China). The cDNA was subjected to qPCR on a CFX96 Real-Time Fluorescence Quantitative PCR Instrument (Bio-Rad Laboratories, Hercules, CA, USA) using BeyoFast SYBR Green qPCR Mix (D7260, Beyotime). Relative gene expression analysis was conducted using the 2−ΔΔCT, with GAPDH used as an internal control. See Table 1 for primer sequences.

Table 1 The primer sequences for RT-qPCRWestern blot analysis

Cells were lysed on ice using RIPA lysis buffer (C0045, TargetMol, Boston, MA, USA, centrifuged at 10,000 g for 5 min, and the supernatant was quantified by Pierce BCA Protein Assay Kits (A55864, Thermo Fisher Scientific Inc., Waltham, MA, USA). An equal amount of protein was transferred to the PVDF membrane by SDS-PAGE gel electrophoresis, and the membrane was incubated with 5% BSA for 1 h at room temperature (RT). After incubating with appropriate primary and secondary antibodies, the immunoreactions were visualized with ECL. The primary antibodies used were anti-AQP7 (1:3000, PA5-102891, Invitrogen), anti-TBX19 (1:1000, GTX77878, GeneTex, Inc., Alton Pkwy Irvine, CA, USA), anti-MGRN1 (1:2000, 11285-1-AP, ProteinTech Group, Chicago, IL, USA), anti-ubiquitin (1:2000, 10201-2-AP, ProteinTech Group), anti-β-Tubulin (1:2000, 10068-1-AP, ProteinTech Group), anti-GAPDH (1:5000, 81640-5-RR, ProteinTech Group), and goat anti-rabbit IgG H&L (HRP) (1:1000, ab6721, Abcam, Cambridge, UK).

Establishment of HCC-SR cells

The establishment of HCC-SR was conducted as previously described [12, 13]. HCC cells were cultured in a medium containing 4 µM Sorafenib for 2 weeks, followed by increasing the concentration of Sorafenib by 10% at an interval of two weeks on the cells. Surviving cells were screened each time for subsequent Sorafenib induction cultures until the maximum dose (10 µM) was reached. HCC-SR cells were cultured in the presence of 1 µM Sorafenib to maintain their resistance. Sorafenib was discontinued for three days before conducting the following experiments.

CCK8

To assess the resistance of the cells to sorafenib, the cells were incubated with a medium containing different concentrations of sorafenib for 48 h. Cell viability was assessed using CCK8 (HY-K0301, MedChemExpress) by detecting the OD value at 450 nm, and the data were normalized according to the viability of cells cultured in a medium without sorafenib (0 µM).

Nile red staining

Lipid accumulation in HCC cells was detected by Nile Red staining as described previously [14]. Briefly, the cells were seeded onto coverslips of 6-well plates at a density of 1 × 105 cells/well and fixed with 4% paraformaldehyde at RT, followed by incubation with 1 ng/mL Nile red (HY-D0718, MedChemExpress) for 10 min at RT. After sealing, the cells were observed by fluorescence microscopy, and the fluorescence intensity was quantitatively analyzed by ImageJ software.

Intracellular lipid analysis

For the detection of intracellular lipids, free fatty acid (FFA) and TG contents in cell lysate were measured by FFA content test kit (ml092765, MLBio) and TG assay kit (ab65336, Abcam), and all operations were performed.

TUNEL assay

Sorafenib-induced apoptosis was evaluated using the One-Step TUNEL apoptosis assay kit (C1088, Beyotime). Briefly, HCC-SR cells treated with sorafenib (4 µM) for 24 h were fixed with 4% paraformaldehyde and permeabilized with 0.3% Triton X-100. Apoptotic cells were subsequently labeled by TUNEL assay solution at 37 °C for 60 min, followed by incubation with DAPI for 10 min to label nuclei (both in the dark). The TUNEL-positive cell rate was viewed under a fluorescence microscope and calculated.

Colony formation assay

HCC-SR cells were seeded into 6-well plates at 1000 cells/well and treated with sorafenib (4 µM) for 24 h. Cells without sorafenib treatment were used as controls. Subsequently, the medium was replaced with fresh medium, and the cells were conventionally cultured for 14 d. The colonies were fixed with 4% paraformaldehyde at RT for 15 min, stained with crystal violet for 10 min, observed, and counted. The survival fraction of sorafenib-treated cells was calculated by normalizing to the number of colonies in the control group.

Isolation and induction of peripheral blood mononuclear cells (PBMCs)

Peripheral blood samples from healthy donors were obtained from the Affiliated Chuzhou Hospital of Anhui Medical University, following the Ethics Committee of the Affiliated Chuzhou Hospital of Anhui Medical University, and informed consent was obtained from all participants. Written informed consent was obtained from all subjects in this study. Peripheral blood samples from five healthy volunteers were collected, and Ficoll-Paque PLUS (17144002, Cytiva, Shanghai, China) was added and centrifuged at 400 g for 40 min at 20 °C to collect PBMCs. Subsequently, NK cells or monocytes were extracted from PBMCs by Dynabeads Untouched Human NK Cell Kit (11349D, Invitrogen) and Dynabeads Untouched Human Monocyte Kit (11350D, Invitrogen), respectively. Isolated monocytes were differentiated into mature macrophages by treatment with 100 ng/mL M-CSF for 5 days [15].

NK cell-mediated killing assay

NK cells (1 × 105 cells) were added to 96-well plates inoculated with target cells (HCC cells) at a 2:1 effector/target cell (E/T) ratio and incubated for 24 h [16]. NK cell-mediated tumor cell killing was detected by lactate dehydrogenase (LDH) activity assay kits (E-BC-K046-M, Elabscience, Wuhan, Hubei, China). LDH content was assessed by measuring the OD at 450 nm. The percentage of cytotoxicity was calculated by the following formula [17]: % of cytotoxicity = (experimental LDH release from co-cultured effector and target cells - LDH spontaneously released from effector cells alone - LDH spontaneously released from target cells alone)/(maximum amount of LDH from target cells - LDH spontaneously released from target cells alone) × 100.

In addition, the levels of cytotoxic cytokines released from NK cells in the co-culture supernatants were also detected by human interferon-gamma (IFN-γ) ELISA Kit (E-EL-H0108, Elabscience) and human granzyme B (GzmB) ELISA Kit (E-EL-H1617, Elabscience).

Transwell co-culture assays

HCC cells were co-cultured with macrophages by the Transwell co-culture system [18] to investigate the link between HCC cells and the polarization of TAMs. In brief, 1 × 105 HCC cells were seeded in the apical chamber of Transwell with 0.4 μm pore polyester membrane insert (CLS3460, Sigma-Aldrich Chemical Company, St Louis, MO, USA). The basolateral chamber contained 2 × 105 macrophages, which were collected after 48 h of co-culture. Expression of macrophage polarization-associated proteins was assessed by flow cytometry using FITC anti-human CD86 antibody (374204, BioLegend, San Diego, CA, USA) and FITC anti-human CD206 antibody (321104, BioLegend). The expression of M1/M2 polarization factors in macrophages was measured using RT-qPCR.

Animal experiment

Six-week-old male C57BL/6J mice (19 ± 2 g) were purchased from SLAC Laboratory Animal Co., Ltd. (Shanghai, China) and were housed under a 12 h:12 h light/dark cycle with controlled temperature (22 ± 1 °C) and humidity (40 ± 5%). All experiment protocols were approved by the Ethics Committee Board for Animal Experiments at The Affiliated Chuzhou Hospital of Anhui Medical University.

Referring to a previous report for orthotopic implantation [19], after anesthetizing the mice by tribromoethanol (240 mg/kg, ip), 4 × 106 Hepa1-6-Luc cells with different genetic interventions mixed with Matrigel were administered into the left lobe of the liver (n = 5/group). After five days, the mice were initiated as needed to undergo sorafenib treatment. Based on the human dose of sorafenib (800 mg/day), the equivalent dose in mice, converted by body surface area = 800 mg × 0.0026/0.02 kg = 104 mg/kg, was administered by gavage (oral gavage; i.g) daily. Tumor load in the mice was monitored by a small-animal live-imaging system by injecting 90 µL luciferin (10 µg/µL; Pierce, Waltham, MA, USA) into the peritoneum of mice on the fifth day of Hepa1-6-Luc cell injection, and every 7 days thereafter. After 2 weeks of sorafenib treatment, all mice were euthanized by intraperitoneal injection of an overdose of sodium pentobarbital (150 mg/kg), and liver tumors were harvested.

Immunohistochemistry

For paraffin-embedded mouse orthotopic tumor sections (5 µM), heat-mediated antigen retrieval was performed in citrate buffer after routine dewaxing and hydration. To suppress endogenous peroxidase activity, 3% H2O2 was applied for 30 min at RM. Following blocking with normal goat serum for 1 h at RM, the tissue sections were subjected to overnight incubation with the diluted (all at 1:100) primary antibodies to AQP7 (PA5-102891, Invitrogen), TBX19 (GTX77878, GeneTex), MGRN1 (11285-1-AP, ProteinTech Group), and CD31 (80530-1-RR, ProteinTech Group) at a temperature of 4◦C, succeeded by the application of the secondary antibody IgG H&L (HRP) (1:1000, ab6721, Abcam) at ambient temperature for 1 h. Subsequent steps entailed the utilization of the DAB in tandem with hematoxylin to detect immunoreactive cells. Finally, the percentage of positively stained area was calculated using Image-J.

Oil red O staining

Frozen sections of mouse orthotopic tumors (10 µM) were rewarmed for 10 min at RM. The sections were rinsed with ddH2O and stained with freshly prepared Oil Red O Staining Working Solution (C0157S, Beyotime) for 10 min at RT. The nuclei were subsequently counterstained with hematoxylin. Images were taken using an inverted microscope, and the percentage of oil red O-positive area was calculated using Image-J.

Tumor single-cell suspension Preparation and flow analysis

Mouse orthotopic tumors were minced and detached with digestive solutions (1 mg/mL collagenase II, 1 mg/mL collagenase IV, 300 µg/mL DNase I, and 0.01% HEPES dissolved in DMEM) for 1 h at 37 °C. The cell suspension was filtered through a 40 μm cell filter. Single-cell suspensions were resuspended in PBS. Red blood cells (RBC) in the suspensions were lysed with RBC Lysis Buffer (00–4333-57, Invitrogen), and the Dead Cell Removal Kit (130-090-101, Miltenyi Biotec, Bergisch Gladbach, Germany) was used for the removal of dead cells. The remaining cells were incubated with FITC anti-mouse NK-1.1 antibody (#156507, BioLegend), PE anti-mouse IFN-γ antibody (505807, BioLegend), FITC anti-mouse F4/80 antibody (157309, BioLegend), and PE anti-mouse CD206 antibody (141705, BioLegend) and loaded to examine the proportion of positive immune cells.

ChIP-qPCR

The binding ability of TBX19 to the AQP7 promoter fragment (chr9: 33401332–33403674) was assessed using the SimpleChIP Enzymatic ChIP Kit (#9003, Cell Signaling Technologies, Beverly, MA, USA). HCC cells (4 × 106 cells) were incubated with 1% formaldehyde for 10 min at RT to cross-link the proteins to the DNA, followed by the addition of glycine solution and mixing for 5 min at RT to terminate the cross-linking. Chromatin crosslinks were prepared, and 10 µL of diluted chromatin samples were taken as a 2% sample input control. Immunoprecipitation was carried out by incubation with anti-TBX19 (GTX77878, GeneTex) overnight at 4 °C, and Normal Rabbit IgG was used as a control. DNA-protein crosslinks were captured by adding Protein G Magnetic Beads and incubating at 4 °C for 2 h with rotation. After elution of chromatin and decrosslinking, enrichment of the AQP7 promoter (forward primer: 5’-GGAGGTGGAAGTGTTGGAGA-3’, reverse primer: 5’-GGCTGGTGTTTCTCTGCAAA-3’) in the pull-downs was quantified by qPCR. Signals obtained from each immunoprecipitation are expressed as a percent of the total input chromatin.

Luciferase reporter assay

The AQP7 promoter fragment (chr9: 33401332–33403674) was subcloned into the pMCS-Firefly Luc vector (16155, Thermo Fisher Scientific Inc., Waltham, MA, USA) to construct the AQP7 promoter luciferase reporter vector. The above vectors were co-transfected with the control vector pGMLR-TK Renilla Luciferase Reporter Gene Plasmid (11557ES03, Yeasen Biotechnology Co., Ltd., Shanghai, China) into cells stably overexpressing TBX19 by Lipofectamin2000 (Thermo Fisher). After 48 h, the transcriptional regulation of the AQP7 promoter was examined using the Dual-Luciferase Reporter Gene Assay Kit (11402ES60, Yeasen), with Renilla Luciferase as an internal reference.

Co-immunoprecipitation (co-IP) assay

HCC cells overexpressing MGRN1 or not were treated with 50 µM MG132 (S2619, Selleck, Shanghai, China) for 6 h. Cells were then harvested to prepare whole cell lysates. Protein A/G MagBeads (36417ES03, Yeasen) and rabbit IgG were added to the remaining lysate and incubated for 2 h at 4 °C to remove non-specific binding proteins. After overnight incubation with anti-TBX19 (GTX77878, GeneTex) at 4 °C, the protein complexes were harvested by Protein A/G MagBeads, and IP samples were obtained by boiling with a loading buffer. Proteins enriched in IP samples were assessed by western blot analysis.

Protein stability assay

HCC cells overexpressing MGRN1 or not were treated with 100 µg/mL of the eukaryotic protein synthesis inhibitor cycloheximide (CHX, S7418, Selleck) for the indicated periods (0 h,3 h, 6 h). The cells were assayed for TBX19 protein expression by western blot assay, and the rate of TBX19 protein degradation was assessed by normalization to protein expression at 0 h.

Statistical analyses

The data is presented as a bar chart plus scatter dots. Statistical significance was assessed using GraphPad 8.0.2 Software (GraphPad, San Diego, CA, USA). Unpaired Student’s t-test and one-way/two-way ANOVA with Tukey’s or Sidak’s post-test for multiple comparisons were performed when the variance among groups. IC50 values of sorafenib on cells were determined using Nonlin fit. p values of less than 0.05 were considered significant.