The number of patients with non-viral, nonalcoholic steatohepatitis-associated liver cancer has been increasing with the prevalence of obesity. Increasing evidence shows that the tumor microenvironment (TME) is crucial for tumor progression. We previously showed that hepatic stellate cells (HSCs) play key roles in tumorigenic TME of obesity-associated hepatocellular carcinoma (HCC) by secreting senescence-associated secretory factors that suppress anti-tumor immunity.1Yoshimoto S. et al.Nature. 2013; 499: 97-101Crossref PubMed Scopus (1263) Google Scholar, 2Loo T.M. et al.Cancer Discov. 2017; 7: 522-538Crossref PubMed Scopus (201) Google Scholar, 3Schwabe R.F. et al.J Hepatol. 2020; 72: 230-238Abstract Full Text Full Text PDF PubMed Scopus (91) Google Scholar Recent advances in single-cell RNA-sequencing (scRNA-seq) techniques have enabled the characterization of different cell types in various tissues. Therefore, development of a refined method to characterize HSCs in TME under various conditions is necessary to understand their role in HCC progression. Isolation of HSCs from normal liver via perfusion from the inferior vena cava (IVC) using a retrograde approach has been reported.4Mederacke I. et al.Nat Protoc. 2015; 10: 305-315Crossref PubMed Scopus (261) Google Scholar However, this method is not applicable to highly steatotic liver or steatohepatitic tumor tissues in obese mice as the accumulation of adipose tissue in obese mice hinders IVC visualization, thereby making cannulation via IVC difficult. Therefore, we attempted HSC isolation through the portal vein (PV), which is clearly visible under all conditions. We aimed to develop a method to isolate HSCs at a low temperature (6 °C) to minimize over digestion by enzymes and heat-associated stress after dissociation (at 37 °C), which reportedly induce genes such as Fos and Jun.5O'Flanagan C.H. et al.Genome Biol. 2019; 20: 210Crossref PubMed Scopus (70) Google Scholar Furthermore, we identified a cell surface marker abundantly expressed in HSCs, which is useful for HSC cell sorting from murine and human liver tumor tissues. In this proposed protocol, cannulation via PV and digestion using ice-cold enzymatic solution (Figure 1, A‒B) are important steps for successful HSC isolation. We confirmed that the ice-cold enzyme solutions (pronase E, collagenase, and DNase I) retained enzymatic activity at 6 °C. The HSCs were concentrated using Nycodenz gradient medium (Supplementary Figure 1, A‒D). The expression of Lrat, a marker for HSCs, revealed maximum HSC fraction in 13% Nycodenz gradient medium (Supplementary Figure 1, A). Pronase E was used to destruct hepatocytes,6Werner M. et al.PLoS One. 2015; 10e0138655Google Scholar and almost no live hepatocyte contamination was confirmed. In contrast, increased contamination with macrophages (Clec4f expressing cells) and liver sinusoidal endothelial cells (Stab2 expressing cells) was observed at higher concentrations of Nycodenz (Supplementary Figure 1, B‒C). Therefore, we decided to use 13% Nycodenz gradient medium for the subsequent experiments including flow cytometry and cell sorting. To screen HSC-specific cell surface markers, we sorted CD31-/CD45- cells from the HSC-rich 13% Nycodenz fraction and performed scRNA-seq using the CD31-/CD45- fraction, excluding liver sinusoidal endothelial cells and immune cells (Figure 1, C). In the t-distributed stochastic neighbor embedding (t-SNE) analysis (Figure 1, D), clusters 0, 1, 2, and 6 predominantly expressed Lrat, indicating that they were HSC fractions (Figure 1, E). However, the other clusters consisted mainly of hepatocytes and HCC cells (Hnf4a-expressing cells), but almost no cholangiocytes (Krt19-expressing cells) were observed (Supplementary Figure 2, A). Among the top 50 genes expressed in the HSC clusters, we focused on 11 genes encoding cell surface molecules expressed in HSCs obtained from the livers of normal diet (ND)-fed mice (clusters 0 and 6 in Figure 1, E) and in those obtained from non-tumor (NT) (cluster 1) and tumor (T) (cluster 2) segments of high-fat diet (HFD)-fed mice (Supplementary Figure 2, B). All 3 segments comprised Itga1 (encoding integrin alpha1, CD49a in CD classification), indicating that Itga1 was the most commonly expressed surface molecule gene. Moreover, Itga1 distribution highly overlapped with Lrat distribution in the HSC clusters (Figure 1, F), indicating that Itga1 (CD49a) could be an excellent HSC cell surface marker in the livers of both ND-fed and HFD-induced HCC-bearing mice. Therefore, we sorted the CD49a-high cell population among CD31-/ CD45- cells (Supplementary Figure 1, E; Figure 1, G) and conducted scRNA-seq analysis. Itga1 (CD49a) was confirmed as a promising HSC marker comparable to Lrat (Figure 1H), and there was almost no contamination of other liver cell types. Furthermore, the content of retinoid, a reported HSC marker in normal liver,4Mederacke I. et al.Nat Protoc. 2015; 10: 305-315Crossref PubMed Scopus (261) Google Scholar was significantly reduced by HFD-induced tumor progression, whereas high CD49a expression was maintained (Supplementary Figure 2, C), indicating that CD49a could be used as an HSC marker under various conditions. The heat-stress response and related modulation in gene expression occur following enzymatic digestion at 37 °C.5O'Flanagan C.H. et al.Genome Biol. 2019; 20: 210Crossref PubMed Scopus (70) Google Scholar,7Adam M. et al.Development. 2017; 144: 3625-3632Crossref PubMed Scopus (145) Google Scholar To evaluate the side effects of high-temperature enzyme incubation on transcriptome, bulk RNA-sequencing of 24346 genes of CD49a-high HSCs was performed at 37 °C or 6 °C, and the results were compared. A total of 3071 genes were overexpressed at 37 °C compared with those at 6 °C (Figure 1, I). Subsequently, pathway enrichment analysis was performed on 1144 commonly overexpressed genes. Notably, expression of several ribosomal genes, which reflects stress-mediated translation,8Vind A.C. et al.Nucleic Acids Res. 2020; 48: 10648-10661Crossref PubMed Scopus (25) Google Scholar was upregulated (Figure 1, J). We identified the top 20 co-expressed genes. The common stress responsive transcription factor Atf3 and prototypical immediate-early response genes (Fos and Jun gene families) were overexpressed at 37 °C (Figure 1, K). Furthermore, the enrichment analysis revealed upregulation of other stress response pathways such as ER stress signaling and inflammatory pathways under heat-stressed conditions. Nevertheless, the expression of cold shock protein-encoding genes including Cirbp, Csds1, and Csds2 remained unchanged (Figure 1, K), suggesting that HSC isolation at 6 °C could minimize the generation of enzyme and heat-stressed transcriptional artefacts. We applied this protocol to collect HSCs with high CD49a expression from the livers of db/db mice (a well-established obese model) (Supplementary Figure 3, A) and confirmed the high yield and high purity of HSCs based on Lrat expression. We also applied this protocol to isolate HSCs from human HCC tissue (T-hHSCs) (Supplementary Figure 3, B‒C). Consistent with murine HSCs, human HSCs from HCC tissue were confirmed as a CD49a-high population among CD31-/ CD45- cells. The mRNA expression of ACTA2 suggested that T-hHSCs predominantly consisted of activated HSCs as we observed previously1Yoshimoto S. et al.Nature. 2013; 499: 97-101Crossref PubMed Scopus (1263) Google Scholar,2Loo T.M. et al.Cancer Discov. 2017; 7: 522-538Crossref PubMed Scopus (201) Google Scholar (Supplementary Figure 3, D). In conclusion, we developed a method to isolate HSCs from highly steatotic liver and steatohepatic HCC tissues utilizing PV-mediated enzymatic cold perfusion method to avoid heat-induced artefact gene expression. Furthermore, we identified CD49a as a reliable HSC marker under various conditions in this procedure. These results lay a foundation for investigating the role of HSCs in liver under both normal and steatotic HCC conditions. The authors thank Dr. Yoshimi Yukawa-Muto and the Department of Surgery for their help in collecting human samples used in this study. We thank Dr. Tomonori Kamiya and the laboratory members of Pathophysiology for their technical assistance. This manuscript has been edited by English editing company, Editage, which was supported by the Takeda Science Foundation . Yi Cheng, PhD (Conceptualization: Supporting; Investigation: Lead; Writing – original draft: Supporting) Ryota Yamagishi, PhD (Data curation: Supporting; Formal analysis: Supporting; Funding acquisition: Supporting; Investigation: Equal; Supervision: Supporting; Writing – review & editing: Supporting) Yoshiki Nonaka, MS (Data curation: Lead; Formal analysis: Lead; Investigation: Supporting) Misako Sato-Matsubara, PhD (Funding acquisition: Supporting; Investigation: Supporting) Norifumi Kawada, MD, PhD (Funding acquisition: Supporting; Resources: Lead) Naoko Ohtani, MD, PhD (Conceptualization: Lead; Funding acquisition: Lead; Investigation: Supporting; Project administration: Lead; Supervision: Lead; Writing – original draft: Lead; Writing – review & editing: Lead) C57BL/6 mice (CLEA Japan) and the 2 diets (normal diet, CE2; CLEA Japan and a high-fat diet [HFD], D12492; Research Diet) were used.1Yoshimoto S. et al.Nature. 2013; 499: 97-101Crossref PubMed Scopus (1268) Google Scholar Obese mice (>45 g; 30–40 weeks) were used in accordance with the protocols approved by the Animal Care and Use Committee of Osaka City University (Approval number: 17206). Chemically induced liver carcinogenesis was performed as described previously.1Yoshimoto S. et al.Nature. 2013; 499: 97-101Crossref PubMed Scopus (1268) Google Scholar The mice were anesthetized, and the portal vein was cannulated. After confirming the blood backflow, the catheter was connected to the perfusion line with EGTA perfusion solution (all solutions shown in Supplementary Table 1 should be ice-cold). As circulation started, the inferior vena cava was cut to discard the perfused solution. The perfusion was sequentially performed using 35, 60, and 75 mL of EGTA, Pronase E (Sigma), and collagenase (Wako) solutions until the entire liver swelled up and softened (Figure 1, A). The liver was then explanted into a 10-cm dish on ice and soaked with enzyme buffer. Next, the non-tumor (NT) and tumor (T) sections were separated carefully. The NT sections turned fluid-like with pipetting, whereas the T sections remained solid and had to be minced (Figure 1, B). They were then transferred into 2 beakers with 45 mL of in vitro digestion solution and incubated for 30 minutes (6° C). The incubated tissues were filtered through a 70-μm cell-strainer into two 50-mL tubes (NT and T), centrifuged at 600 × g for 10 minutes (4 °C), and the pellet was re-suspended in GBSS/B. Nycodenz (AXS) solution (0.286 g/mL) was prepared by vortexing before use. The Nycodenz gradient medium (6%–17%) was prepared, and 8 mL of cell-Nycodenz gradient medium (with the ratio shown in Supplementary Figure 1, D) was transferred into moistened 15-mL tubes and overlaid with 3 mL of GBSS/B. The solutions were then centrifuged at 1700 × g for 20 minutes (accel & brake 0; 4 °C). A white intermedia layer of cells containing HSCs was obtained. A fat layer on the top was aspirated to avoid contamination. The white layer was collected and added into 40 mL of GBSS/B and subjected to low-speed centrifugation (50 × g for 5 minutes) to exclude hepatocytes. The supernatant was centrifuged at 1000 × g for 10 minutes (4 °C); thereafter, 3 mL of RBC lysis buffer (1×, Biolegend) was added to the pellet and incubated for 4 minutes to remove the erythrocytes, and then 2% fetal bovine serum (FBS)-phosphate buffered saline (PBS) was added to make up the volume to 14 mL for ceasing the reaction. Another step of high-speed centrifugation was performed at 1000 × g for 10 minutes (4 °C), and the pellet containing primary HSCs was re-suspended in 2% FBS-PBS solution and used for quantitative polymerase chain reaction (PCR), flowcytometry, and RNA-sequencing (RNA-seq). Although the yield of HSCs slightly varied in different mice, we could always isolate more than 106 cells from one ND mouse liver regardless of mouse age (Supplementary Figure 1, F) and 10 × 106 cells from the NT region and 106 cells from the T region of one HFD-fed mouse liver. Total RNA was extracted from primary HSCs using RNAiso Plus (Takara) and reverse-transcription and quantitative PCR were performed as previously described.1Yoshimoto S. et al.Nature. 2013; 499: 97-101Crossref PubMed Scopus (1268) Google Scholar The primer sequences are listed in Supplementary Table 1. The isolated HSCs were pre-incubated with unlabelled anti-CD16/32 mAb (BioXcell) to avoid non-specific binding to FcγR. The cells were stained with antibodies against CD31, CD45.2, and CD49a (Biolegend), and analyzed using the Attune NxT Cytometer (Thermo-Fisher Scientific); and sorted using the SONY-SH800 cell sorter (SONY). Data were processed using FlowJo-Version-10 software. Dead cells were excluded using propidium iodide (Biolegend) gating. Antibodies used are listed in Supplementary Table 1. CD31-/CD45.2- cells from ND, NT, and T of mouse livers were sorted for single-cell RNA-seq. Single-cell dispensing and library preparation were performed according to the protocol of the ICELL8 cx 3′ DE kit.2Takara Bio. ICELL8 cx 3ˈ DE Kit User Manual. 32.Google Scholar Libraries were sequenced on Illumina Novaseq 6000 with 25–150-bp paired-end reads to produce an average read of 934,102/cell. RNA-seq data of 1613 cells were obtained. The expression level of target genes in each cell was calculated using the mappa analysis pipeline (Demuxer and analyser version 1.0, ICELL8). From the gene expression matrix, quality control, data clustering, visualization, and differential expression analysis were performed using Seurat version 2.3.4 in R.3Butler A. et al.Nat Biotechnol. 2018; 36: 411-420Crossref PubMed Scopus (3537) Google Scholar The HSC clusters were confirmed using t-SNE of marker gene expression, including DES (desmin), CYGB (cytoglobin), and Lrat. CD49a-high cells among CD31-/CD45.2- cells from ND, NT, and T tissues of mouse livers were sorted for single-cell RNA-seq (ICELL 8, Takara). The libraries were sequenced to produce an average read of 892086/cell and RNA-seq data of 1390 cells were obtained. The expression level of target genes in each cell was calculated using the Cogent NGS Analysis Pipeline. The total RNA was extracted from sorted-CD49a-high cells among CD31-/CD45.2- cells using TRIzol (Invitrogen). Bulk RNA-seq libraries were prepared using the SMART-Seq v4 protocol (Takara). The libraries were sequenced on Illumina Hiseq4000. For each sample, depths of at least 27.6 million paired-end reads were generated. Mapping was performed using tophat. The expression level was calculated using cuffdiff and feature counts. From the gene expression matrix, differential expression analysis was performed with R version 3.6.3. The liver tumor tissue (approximately 5 × 5 × 5 mm3 tissue), obtained from a patient with HCC who underwent partial hepatectomy, was minced and stirred with enzymes for 40 minutes (6 °C), and separated following the same procedure as the murine cell isolation method. We used 20% Nycodenz gradient medium to harvest the maximum number of cells from the limited tumor tissues. Thereafter, CD49a-high cells among CD31-/CD45.2- cells were isolated. We isolated 7.3 × 105 HSCs labelled as T-hHSCs. The study was conducted following the Helsinki Declaration II; written informed consent was obtained from each patient according to the protocol approved by the ethics committee of Osaka City University (Approved number: 3722). The RNA-seq data are available in the GEO databases (accession numbers: HSCs at 37°C or 6°C: GSE192598, CD31-/CD45.2-/CD49a+ HSCs: GSE192582, and CD31-/CD45.2-HSCs: GSE192637)Supplementary Figure 2CD49a can be a marker for HSC isolation. (A) Unsupervised t-SNE clustering of the sorted CD31-/CD45.2- cells revealed the expression of a cholangiocyte marker (Krt19) and hepatocyte marker (Hnf4a). (B) The gene expression in the HSC clusters (ND cluster 0 and 6; NT cluster 1; T cluster 2) was compared and analyzed using log2-Fold change (logFC). The top 11 genes encoding cell surface markers are illustrated in an ascending order of fluctuation. (C) Flow cytometry plots of HSCs from ND and HFD mice. The average body weight of mice: ND, 32.6 g (range, 31‒34 g; n = 3); HFD, 58.5 g (range, 55‒61 g; n = 3).View Large Image Figure ViewerDownload Hi-res image Download (PPT)Supplementary Figure 3CD49a-high HSCs can be isolated from the db/db mouse liver and the human HCC tissue. (A) Application of the cold isolation protocol for obtaining HSCs from db/db mice (n = 3). The average body weight (BW) of mice was 51.8 g (range, 48‒57 g). The sorted CD49a-high and CD49a-low cells among CD31-/CD45- cells were subjected to qPCR analysis for Lrat expression. Data were obtained in triplicate using the samples of 3 different mice and are presented as fold-change difference in comparison with pre-sorted HSCs. Statistical significance was determined using Bonferroni’s multiple comparisons test (see the raw data table by the bar graph). (B) The histology of post-sustained virologic responce (hepatitis C virus) human HCC tissue (hematoxylin and eosin staining, scale bar, 200 μm) and the patient’s information. (C) Method for isolating HSCs from T-hHSCs. (D) Flow cytometry analysis and qPCR analysis of T-hHSCs. The rectangle shows the double-positive fraction of CD49a expression and retinoid content. T-hHSCs, LI90 (human HSC line), or HepG2 (human HCC cell line) cells were subjected to qPCR analysis for evaluating ACTA2 expression. Data were obtained in triplicate using these samples. Statistical significance was determined using Bonferroni’s multiple comparisons test (see the raw data table by the bar graph).View Large Image Figure ViewerDownload Hi-res image Download (PPT)Supplementary Table 1ASolutionsSolutionsReagentsFinal concentrationEGTA perfusion solutionNaCl136.89 mMKCl5.37 mMNaH2PO40.64 mMNa2HPO40.85 mMHEPES9.99 mMNaHCO34.17 mMEGTA0.50 mMGlucose5.00 mMEnzyme bufferNaCl136.89 mMKCl5.37 mMNaH2PO40.64 mMNa2HPO40.85 mMHEPES9.99 mMNaHCO34.17 mMCaCl2·2H2O3.81 mMGey’s balanced salt solution B, GBSS/BNaCl136.89 mMKCl4.96 mMMgCl2·6H2O1.03 mMMgSO4·7H2O0.28 mMNa2HPO40.42 mMKH2PO40.22 mMGlucose5.50 mMNaHCO32.70 mMCaCl2·2H2O1.53 mMGey’s balanced salt solution A, GBSS/AKCl4.96 mMMgCl2·6H2O1.03 mMMgSO4·7H2O0.28 mMNa2HPO40.42 mMKH2PO40.22 mMGlucose5.50 mMNaHCO32.70 mMCaCl2·2H2O1.53 mMPronase E perfusion solutionPronase E in enzyme buffer1 mg/mLCollagenase perfusion solutionCollagenase in Enzyme buffer1.33 mg/mLIn vitro digestion solutionPronase E0.556 mg/mLCollagenase0.556 mg/mL DNase I in Enzyme buffer0.02 mg/mL Open table in a new tab Supplementary Table 1BChemicalsReagentsCompany, catalog numberPronase ESigma-Aldrich, cat. no. 1.07433CollagenaseWako, cat. no. 032-22364DNaseIRoche, cat. no. 11284932001Nycodenz AGAXS, cat. no. 1002424EGTANACALAI, cat. no. 15214-92Sodium chloride (NaCl)Wako, cat. no. 191-01665Potassim chloride (KCl)NACALAI, cat. no. 28514-75SodiumNACALAI, cat. no. 31720-65di-SodiumNACALAI, cat. no. 31726-05HEPESDOJINDO, cat. no. 340-01371Sodium bicarbonateNACALAI, cat. no. 31212-25d-(+)-GlucoseNACALAI, cat. no. 16806-25Calcium chlorideNACALAI, cat. no. 06731-05Magnesium chlorideNACALAI, cat. no. 20909-55Magnesium sulfateNACALAI, cat. no. 21003-75Potassium phosphateNACALAI, cat. no. 28721-55RBC lysis buffer (10×)Biolegend, cat. no. 420302D-PBS (−)FUJIFILM, cat. no. 045-29795 Open table in a new tab Supplementary Table 1CEquipmentEquipmentCompany, catalog numberPerfusion pumpCole-Parmer, 07516-10Perfusion lineCole-Parmer, 06424-1424-gauge catheterTERUMO SURFLO, SR-OT2419CP70-μm Cell strainerAS ONE, 3-6649-02Water bathTHERMAL ROBO, TR-1ARBottle-top filter, 0.45 μmThermo scientific, cat. no. 291-3345Refrigerated benchtopBeckman Coulter, Allegra X-15RCell sorterSONY-SH800, cat. no. SH800SAttune NxT AcousticThermo Fisher, cat no. A24860 Open table in a new tab Supplementary Table 1DPrimersGenePrimer sequencesMouse GAPDH5'-CAACTACATGGTCTACATGTTC-3' (forward)5'-CACCAGTAGACTCCACGAC-3' (reverse)Mouse Lrat5'-TACTGCAGATATGGCTCTCG-3' (forward)5'-GATGCTAATCCCAAGACAGC-3' (reverse)Mouse Clec4f5'-TGCAGGAAGCTGTGGCTGCA-3' (forward)5'-TCCCGCCACGGCTTCTTGTC-3' (reverse)Mouse Stab25'-ATTGCCTCTAACGGGGTTCT-3' (forward)5'-ATCGCACCAGTAACCAGGAC-3' (reverse)Mouse albumin5'-CCCGAAGCTTGATGGTGTGA-3' (forward)5'-GTCTGGCTCAGACGAGCTAC-3' (reverse)Human Acta25'-ACCTCATGAAGATCCTGACT-3' (forward)5'-TTCAAAGTCCAGAGCTACAT-3' (reverse)Human GAPDH5'-GTGGTCTCCTCTGACTTCAAC-3' (forward) 5'-TGAGCTTGACAAAGTGGTCG-3' (reverse) Open table in a new tab Supplementary Table 1EAntibodiesAntibodiesCompany, catalog numberCD16/32 mAb (2.4G2)BioXcell, #BE0307CD31Biolegend, #102514CD45.2Biolegend, #109813CD49aBiolegend, #142603Propidium iodideBiolegend, #421301 Open table in a new tab