Hepatocellular carcinoma (HCC), the fourth leading cause of cancer mortality worldwide, develops almost exclusively in patients with chronic liver disease and advanced fibrosis1,2. Here we

interrogated functions of hepatic stellate cells (HSCs), the main source of liver fibroblasts3, during hepatocarcinogenesis. Genetic depletion, activation or inhibition of HSCs in mouse

models of HCC revealed their overall tumour-promoting role. HSCs were enriched in the preneoplastic environment, where they closely interacted with hepatocytes and modulated

hepatocarcinogenesis by regulating hepatocyte proliferation and death. Analyses of mouse and human HSC subpopulations by single-cell RNA sequencing together with genetic ablation of

TAZ activation in pretumoural hepatocytes and through activation of discoidin domain receptor 1 in established tumours. An increased HSC imbalance between cytokine-producing HSCs and

myofibroblastic HSCs during liver disease progression was associated with increased HCC risk in patients. In summary, the dynamic shift in HSC subpopulations and their mediators during

chronic liver disease is associated with a switch from HCC protection to HCC promotion.

The microarray, RNA-seq, scRNA-seq and snRNA-seq datasets reported in this study have been deposited in the GEO database under the accession numbers GSE174748 and GSE212047. In addition, we

analysed previously published whole liver or isolated HSC scRNA-seq datasets from GSE172492 and GSE158183, normal human liver snRNA-seq data from GSE185477 and the microarray datasets from

GSE15654, GSE49541 and GSE10140. Source data are provided with this paper.

R markdown scripts enabling the main steps of the analysis have been deposited into GitHub (https://github.com/Schwabelab/HSC_in_HCC). Survival of patients with HCC from TCGA dataset was

determined using http://gepia2.cancer-pku.cn/#survival.

This work was supported by grants R01CA190844 and R01CA228483 (to R.F.S.) and R01DK116620 (to R.F.S. and I.T.) and the Columbia University Digestive and Liver Disease Research Center

(1P30DK132710) and its Bioinformatics and Single Cell Analysis Core. J.Z.-R. was supported by the Ligue Nationale contre le Cancer (Equipe Labellisée) and Labex OncoImmunology

(Investissement d’avenir). N.C.H. is supported by a Wellcome Trust Senior Research Fellowship in Clinical Science (ref. 219542/Z/19/Z), the Medical Research Council and a Chan Zuckerberg

Initiative Seed Network Grant. Y.H. was supported by NIH grant CA233794 and Cancer Prevention and Research Institute of Texas grant RR180016. B.I. was supported by NIH grants R37CA258829 and

R21CA263381. These studies used the resources of the Herbert Irving Comprehensive Cancer Center at Columbia University. The Flow Core, Molecular Pathology and Confocal and Specialized

Microscopy shared resources are funded in part through NIH grants P30CA013696 and S10OD020056. A.F. was funded by a Foundation pour la Recherche Medicale postdoctoral fellowship

(SPE20170336778), an American Liver Foundation Postdoctoral Research Award, an International Liver Cancer Association’s Fellowship and the Mandl Connective Tissue Research Fellowship. Y.S.

is supported by the Uehara Memorial Foundation and the Naomi Berrie Diabetes Center Russell Berrie Foundation. D.D. is supported by F31 DK091980. S.B. is funded by Deutsche

Forschungsgemeinschaft grant GZ:BH 155/1-1. S.A. was funded by an American Liver Foundation Postdoctoral Research Fellowship Award, a Cholangiocarcinoma Foundation’s Innovation Award and a

Research Scholar Award from the American Gastroenterological Association. We thank E. Monuki (University of California, Irvine) for the Lhx2-floxed mice; M. Mack (University of Regensburg,

Germany) for the Col1a1-floxed mice; R. Kalluri for the αSMA-TK mice; Y. Yamaguchi (Stamford Burnham Prebys Medical Discovery Institute, La Jolla) for the Has2-floxed mice; and E. Seki

(University of California, Los Angeles), C. Hernandez (The University of Birmingham, UK) and C. Kuntzen (Columbia University) for scientific support and discussions.

Present address: Klinikum Rechts der Isar, Technical University of Munich (TUM), Munich, Germany

Present address: Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Barcelona, Spain

Present address: Department of Health and Nutrition Sciences, Brooklyn College, City University of New York, New York, NY, USA

Present address: Department of Biomedical Engineering, Widener University, Chester, PA, USA

Present address: Department of Gastroenterology, Changzheng Hospital, Shanghai, China

These authors contributed equally: Yoshinobu Saito, Ajay Nair, Dianne Dapito, Le-Xing Yu

Department of Medicine, Columbia University, New York, NY, USA

Aveline Filliol, Yoshinobu Saito, Ajay Nair, Dianne H. Dapito, Le-Xing Yu, Aashreya Ravichandra, Sonakshi Bhattacharjee, Silvia Affo, Qiuyan Sun, Jorge M. Caviglia, Xiaobo Wang, Jin Ku Kang,

 Amit Dipak Amin, Deqi Yin, Oscar M. Rodriguez-Fiallos, Chuan Yin, Adam Mehal, Benjamin Izar, Utpal B. Pajvani, Ira Tabas & Robert F. Schwabe

Department of Systems Biology, Columbia University Irving Medical Center, New York, NY, USA

Liver Tumor Translational Research Program, Harold C. Simmons Comprehensive Cancer Center, Division of Digestive and Liver Diseases, University of Texas Southwestern Medical Center, Dallas,

TX, USA

Department of Pharmacology, School of Medicine, University of California, San Diego, San Diego, CA, USA

Department of Microbiology and Immunology, Columbia University Irving Medical Center, New York, NY, USA

Centre for Inflammation Research, The Queen’s Medical Research Institute, Edinburgh BioQuarter, University of Edinburgh, Edinburgh, UK

John R. Wilson-Kanamori, Sebastian Wallace, Ross Dobie & Neil C. Henderson

Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA

Functional Genomics of Solid Tumors Laboratory, Centre de Recherche des Cordeliers, INSERM, Sorbonne Université, Université de Paris, Paris, France

Institute of Human Nutrition, Columbia University, New York, NY, USA

Jin Ku Kang, Utpal B. Pajvani, Ira Tabas & Robert F. Schwabe

Biomedical Informatics Shared Resource, Herbert Irving Comprehensive Cancer Center, and Department of Biomedical Informatics, Columbia University Irving Medical Center, New York, NY, USA

Department of Pathology, Columbia University Irving Medical Center, New York, NY, USA

MRC Human Genetics Unit, Institute of Genetics and Cancer, University of Edinburgh, Edinburgh, UK

Department of Physiology, Columbia University, New York, NY, USA

A.F. designed experiments, generated, analysed and interpreted data and computational data and drafted the manuscript. Y.S. designed experiments, generated, analysed and interpreted data

related to TAZ. A.N. designed and performed computational analyses of scRNA-seq and snRNA-seq data, including CellPhoneDB and cell trajectories. D.D. designed experiments, generated,

analysed and interpreted data related to Lhx2. L.-X.Y., A.R., S.B., S.A. and Q.S. generated and analysed data. N.F. and Y.H. analysed myHSC/cyHSC imbalances and survival in human cohorts.

H.S. and M.K. provided conceptual input and data on DDR1 activation and degradation. T.M.S. performed and assisted in the flow cytometry analysis (supervised by N.A.). J.M.C. generated

RNA-seq data. D.C. and L.C. performed and analysed the stiffness experiments (supervised by R.G.W.). X.W. and I.T. contributed to studies of TAZ-driven HCC. S.C. and J.Z.-R. analysed mRNA

expression and survival in human cohorts. J.K.K. measured lipid content in the liver (supervised by U.B.P.). A.D.A., S.W. and R.D. performed snRNA-seq. J.R.W.-K. performed computational

analysis of human snRNA-seq. N.C.H. and B.I. oversaw snRNA-seq. D.Y., O.M.R.-F. and A.M. provided technical assistance. C.Y. generated, analysed and interpreted data related to the partial

hepatectomy model. R.A.F. assisted with microarray analysis. H.R. contributed to histopathological tumour evaluation. R.F.S. conceived and oversaw the study, designed experiments, drafted

and edited the manuscript.

B.I. has received honoraria from consulting with Merck, Johnson & Johnson/Janssen Pharmaceuticals, AstraZeneca and Volastra Therapeutics. M.K. is a founder and SAB member of Elgia Pharma and

received research support from Merck and Janssen Gossamer Bio. The other authors declare no competing interests.

Nature thanks Scott Friedman and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

a, qPCR showing Trp53 mRNA in FACS-sorted HSC isolated from p53f/f (n = 2 mice) and p53ΔHSCmice (n = 4 mice). b–d, HCC was induced in p53f/f (n = 11 mice) andp53ΔHSC (n = 14 mice) mice by

injection of DEN (i.p. 25 mg/kg at 2 weeks old) followed by 14 injections of CCl4 (i.p. 0.5 µL/g, 1x/week) starting one month after DEN. HSC activation and fibrogenesis were assessed in by

qPCR for fibrogenic genes Acta2, Col1a1 and Lox in the liver (b). Fibrosis was evaluated by Sirius Red staining (c). HCC is shown by representative liver pictures and the tumour burden

measured by liver/body weight ratio (LBR), tumour number and tumour size (d). e, qPCR showing Rela mRNA in FACS-sorted HSC from Relaf/fand RelaΔHSCmice (n = 3 mice/group). f–h, HCC was

induced in Relaf/f(n = 9 mice) and RelaΔHSC(n = 10 mice) mice by injection of DEN (i.p. 25 mg/kg at 2 weeks old) followed by 17 injections of CCl4 (i.p. 0.5 µL/g, 1x/week). HSC activation

and fibrogenesis was assessed by qPCR for the fibrogenic genes Acta2, Col1a1 and Lox in the liver (f). Fibrosis was evaluated by Sirius Red staining (g). HCC is shown by representative liver

 pictures and tumour burden measured by LBR, tumour number and tumour size (h). i–k, representative images showing senescence in specific cell types by senescence associated

beta-galactosidase (SA-Gal) staining and co-staining for markers or lineage tracers of HSC (Lrat-Cre x TdTom), macrophages (anti-macrophage antibody), endothelial cells (endomucin antibody),

cholangiocytes (CK19 antibody) and hepatocytes (AAV8-TGB-Cre x TdTom) in the CCl4 (n = 3 mice) (i), HF-CDAA diet (n = 1 mouse) (j) and Mdr2KO(n = 1 mouse) (k) mouse models of fibrosis. l–m,

representative images showing senescence in specific cellular compartments by p21 IHC in combination with lineage markers for HSC (Lrat-Cre x TdTom) and hepatocytes (AAV8-TGB-Cre x TdTom)

in the CCl4 (l) and HF-CDAA diet (m) mouse models of fibrosis (from n = 1 mouse per model). Data are shown as mean ± SEM, each data point represents one individual. Scale bars: 400 µm (c,g)

and 100 µm (i–m). LBR: liver/body weight ratio. Statistics: data in b, c, d, e, g, Acta2 mRNA and Col1a1 mRNA in f, and LBR and tumour size in h were analysed by two-tailed Student’s t-test.

Lox mRNA in d and tumour number in h were analysed by two-tailed Mann-Whitney test. Raw data are given in Source Data.

a, Lhx2 mRNA in isolated HSC (n = 5 mice), Kupffer cells (KC), endothelial cells (LSEC) and hepatocytes (n = 3 mice each). b, Lhx2 mRNA by scRNAseq from normal mouse liver (n = 1 mouse). c,

qPCR showing deletion of Lhx2 by Lrat-Cre in whole liver: Lhx2f/f: n = 8 mice, Lhx2ΔHSC: n = 6 mice or FACS-sorted HSC (n = 2 mice/group). d–f, deletion of Lhx2, achieved via Mx1-Cre and

poly I:C injections, increased liver fibrosis, shown by Sirius Red staining: Lhx2f/f: n = 11 mice, Lhx2del: n = 9 mice in non-tumour areas (d), HSC activation measured by qPCR: Lhx2f/f: n = 

11 mice, Lhx2del: n = 8 mice (e); and promoted HCC development Lhx2f/f: n = 11 mice, Lhx2del: n = 9 mice (f) compared to Lhx2f/f littermates. g, Lrat-Cre-mediated Yap1 deletion (YapΔHSC) was

confirmed in FACS-sorted HSC by qPCR: Yapf/f: n = 2 mice, YapΔHSC: n = 3 mice, and western blot (n = 2 mice/group). h, YapΔHSCmice showed reduced fibrosis, evaluated by Sirius Red (n = 15

mice/group) and HSC markers, measured by qPCR (Yapf/f: n = 14 mice, Yap ΔHSC: n = 15 mice), in non-tumour liver tissue from mice treated with DEN+CCl4. i, HSC depletion via Lrat-Cre-induced

DTR significantly reduced Lrat mRNA in the DEN+CCl4 model (DTR neg: n = 15 mice, DTR pos: n = 16 mice). j-k, αSMA staining (n = 13 mice/group) and qPCR for Acta2 and Col1a1 (n = 12

mice/group) showed depletion of αSMA+ cells in non-tumour areas in αSMA-TKposmice compared to αSMA-TKneg littermates after ganciclovir (GCV) injections in DEN+CCl4-induced HCC (j) and

αSMA-TKpos mice developed fewer tumours (n = 13 mice/group) (k). l, Liver fibrosis and deletion of Pdgfrb were determined by Sirius red staining and qPCR for Col1a1 and Pdgfrb in 4 month-old

Mdr2KO PdgfrbΔHSC (n = 13 mice) and Mdr2KO Pdgfrbfl/fl (n = 13 mice) female mice. m, Tumour development was determined in 15 month-old Mdr2KO PdgfrbΔHSC (n = 8 mice) and Mdr2KO Pdgfrbfl/fl

(n = 6 mice) female mice as described above. n, HCC development in mice overexpressing TAZS89A in hepatocytes receiving a NASH-FPC (n = 13 mice) or chow diet (n = 11 mice). o–q, DTRpos mice

displayed efficient HSC depletion in the TAZ+FPC NASH-HCC model compared to DTRneg mice: DTRneg: n = 10 mice, DTRpos: n = 14 mice (o) as well as reduced tumour development: DTRneg: n = 10

mice, DTRpos: n = 14 mice (p), but no reduction of cholesterol and triglycerides measurement in non-tumour liver tissue (untreated: n = 3 mice, TAZ+FPC in DTRneg: n = 6 mice, TAZ+FPC in

DTRpos: n = 7 mice) (q). r–s, Lrat-Crepos DTRpos or DTRneg mice were subjected to DEN+HF-CDAA-induced spontaneous hepatocarcinogenesis, revealing efficient HSC depletion by diphtheria toxin

(DT) injections (n = 4 mice/group) (r) as well as reduced tumour development in DTRpos mice (n = 8 mice) compared to DTRneg mice (n = 6 mice) (s). t–u, αSMA-TKpos or αSMA-TKneg mice were

subjected to NICD+HF-CDAA-induced hepatocarcinogenesis, revealing efficient fibroblast depletion after ganciclovir (GCV) injections: αSMA-TKpos (n = 8 mice) vs αSMA-TKneg mice (

n = 7 mice) (t) as well as reduced tumour development in αSMA-TKpos (n = 8 mice) vs αSMA-TKneg mice (n = 9 mice) (u). Data are shown as mean ± SEM, each data point represents one individual,

all scale bars: 200 µm. Statistics: data in d, all data in e besides Lox mRNA, Sirius Red in h, i, tumour number and tumour size in k, data in l besides Col1a1 mRNA, o, p, r, s, t and data

in u besides tumour number were analysed by two-tailed Student’s t-test. The following data: Lox mRNA in e, f, all data in h besides Sirius Red, j, LBR in k, Col1a1 mRNA in l, m, n, and

tumour number in u were analysed by two-tailed Mann-Whitney test. Data in q were analysed by one-way ANOVA (p