FBP1 controls liver cancer evolution from senescent MASH hepatocytes

Abstract

Hepatocellular carcinoma (HCC) originates from differentiated hepatocytes undergoing compensatory proliferation in livers damaged by viruses or metabolic-dysfunction-associated steatohepatitis (MASH)¹. While increasing HCC risk², MASH triggers p53-dependent hepatocyte senescence³, which we found to parallel hypernutrition-induced DNA breaks. How this tumour-suppressive response is bypassed to license oncogenic mutagenesis and enable HCC evolution was previously unclear. Here we identified the gluconeogenic enzyme fructose-1,6-bisphosphatase 1 (FBP1) as a p53 target that is elevated in senescent-like MASH hepatocytes but suppressed through promoter hypermethylation and proteasomal degradation in most human HCCs. FBP1 first declines in metabolically stressed premalignant disease-associated hepatocytes and HCC progenitor cells^4,5, paralleling the protumorigenic activation of AKT and NRF2. By accelerating FBP1 and p53 degradation, AKT and NRF2 enhance the proliferation and metabolic activity of previously senescent HCC progenitors. The senescence-reversing and proliferation-supportive NRF2-FBP1-AKT-p53 metabolic switch, operative in mice and humans, also enhances the accumulation of DNA-damage-induced somatic mutations needed for MASH-to-HCC progression.

Extended Data Fig. 1 |. FBP1, ALDOB and TP53 are downregulated in human HCC.

a, Relative staining intensities of the indicated proteins from Fig. 1a. b, Relative ALDOB amounts in NT and T tissues from the CPTAC-LIHC database. c, Relative FBP1 and ALDOB amounts in NT and T tissues from the PXD006512-LIHC database. d, Percent survival of PXD006512-LIHC patients stratified according to ALDOB expression by “best expression cut-off”. Significance determined by log-rank test. e, FBP1 expression in TCGA-LIHC with WT (n = 308) and mutant (n = 107) TP53. f, Relative FBP1 promoter methylation levels in the TCGA-LIHC database. g, Relative FBP1 promoter methylation levels in HCC patients from the TCGA database. Low and high were defined as in Methods. h-j, Relative promoter methylation levels of G6PC (h), PCK1 (i) and TP53 ( j) in the TCGA-LIHC database. Data in a, b, c, e, f, h, i and j are mean ± SEM. Statistical significance determined by two-sided unpaired t-test or Mann–Whitney U test (a, b, c, e, f, h, i and j) based on data normality distribution. **P < 0.01, ***P < 0.001, ns, not significant. Box plots show center line (median), box limits (first and third quartiles) and whiskers (outer data points).

Extended Data Fig. 2 |. FBP1 is transcriptionally activated by TP53.

a, Schematic of the FBP1 gene 5′ and 3′ regions, showing putative TP53 binding sites and amplicons used for ChIP-qPCR. b, Relative amounts of Fbp1, p21 and Tp53 mRNAs in primary hepatocytes from Tp53^F/F and Tp53^ΔHep mice. c, IB demonstrating shRNA mediated TP53 knockdown and FBP1 downregulation in human hepatocytes. d, Relative FBP1, p21^CIP1 and TP53 mRNA amounts in human primary hepatocytes stably transfected with shCtrl, shTP53#1 and shTP53#2. e, f, Relative FBP1, p21^CIP1 and TP53 mRNAs in HepG2 (e) and SK-HEP-1 (f) cells stably transfected with shCtrl, shTP53#1 and shTP53#2. IB showing TP53 knockdown in SK-HEP-1 cells is on the right. g-j, ChIP-qPCR probing TP53 recruitment to the Fbp1 gene in Tp53^F/F and Tp53^ΔHep livers (g), SK-HEP-1 (h), HepG2 (i) and Huh7 ( j) cells (n = 3 BR each). k, Relative FBP1, p21^CIP1and TP53 mRNAs (left, middle) and proteins (right) in NCD and HFD and CSD and HFrD fed Tp53^F/F and Tp53^ΔHep mice (16 weeks; n = 5 biological replicates/BR). Quantification of relative normalized protein amounts is shown below each strip. Data in b, d, e, f, g, h, i, j and k are mean ± SEM. Statistical significance determined by two-sided unpaired t-test or Mann–Whitney U test (b, g, h, i and j) and one-way ANOVA with Tukey post-hoc tests or Kruskal–Wallis test with Dunn post-hoc tests (d, e, f and k) based on data normality distribution. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, ns, not significant.

Extended Data Fig. 3 |. FBP1 and TP53 are induced in MASH and blunt diet-induced DNA damage.

a, IB analysis of the indicated proteins in livers of HFrD-fed MUP-uPA mice collected at the indicated time points. Quantification of relative normalized protein amounts is shown below each strip. b, Frozen liver sections from the indicated mice stained for 53BP1 and P-γH2AX (n = 4–5 BR). Scale bars, 10 μm. Quantification of 53BP1- and P-γH2AX-positive cells per HMF is shown underneath. c, IB analysis of the indicated liver proteins after 22 weeks of CSD or HFrD feeding. MW markers, densitometric quantification of protein ratios (HFrD/CSD) and P values are on the right. d, Representative IHC of MUP-uPA livers after 32 weeks of HFrD feeding (n = 4–5 BR). Scale bars, 50 μm. Quantification of staining intensity/HMF of indicated proteins is shown underneath. e, MUP-uPA/Fbp1^F/F and MUP-uPA/Fbp1^ΔHep livers at 22 weeks of CSD or HFrD. f, g Representative IHC (f) and Image J (g) quantification of MUP-uPA/Fbp1^F/F and MUP-uPA/Fbp1^ΔHep livers in 22 weeks after HFrD (n = 7–9 BR). Scale bars, 50 μm. Data in b, c, d and g are mean ± SEM. Statistical significance determined by one-way ANOVA with Tukey post-hoc tests or Kruskal–Wallis test with Dunn post-hoc tests (b, g) and two-sided unpaired t-test or Mann–Whitney U test (d) based on data normality distribution. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Extended Data Fig. 4 |. FBP1 and TP53 are induced in human MASH and downregulated in daHep.

a, Relative FBP1 and ALDOB expression from the GSE162694 human normal and MASH liver dataset (normal, n = 31; MASH, 0, n = 35; 1, n = 30; 2, n = 27; 3, n = 8; 4, n = 12). b, Representative IHC of human normal and MASH livers (n = 7–9 BR). Image J quantifications are to the right. c, SR staining and 53BP1 and P-γ-H2AX IHC of human normal and MASH liver tissues (n = 7–9 BR). Image J quantification is shown underneath. d, UMAP visualizations and unsupervised clustering of 78250 hepatocyte nuclei from integrated GSE185477, GSE174748, GSE192742 and GSE212046 datasets. Five hepatocyte subsets were annotated based on gene expression and liver pathology metadata (top left). FBP1 and TP53 expression shown by UMAP visualization (top right and bottom left). NRF2 pathway signatures containing n = 141 target genes in daHep and HCC are shown in the bottom right. The location of daHep nuclei is highlighted by blue ellipses. e, UMAP visualizations and unsupervised clustering of 12,540 mouse hepatocyte nuclei from the GSE200366 dataset. Four subsets previously identified, three representing normal hepatocyte zonation (Zone_1_Hep, Zone_2_Hep, Zone_3_Hep) and one daHep cluster. Density maps of Fbp1 and Dnmt1 mRNA expression in the UMAP space are shown with the daHep cluster highlighted by blue ellipses. Data in a, b and c are mean ± SEM. Statistical significance determined by one-way ANOVA with Tukey post-hoc tests or Kruskal–Wallis test with Dunn post-hoc tests (a) and two-sided unpaired t-test or Mann–Whitney U test (b, c) based on data normality distribution. *P < 0.05, **P < 0.01, ****P < 0.0001. Box plots show center line (median), box limits (first and third quartiles) and whiskers (outer data points).

Extended Data Fig. 5 |. Fbp1 ablation promotes MASH to HCC progression.

a, Normal hepatocytes and HCC progenitor cells (HcPC) were isolated from 22 weeks HFrD fed MUP-uPA/Fbp1^F/F and MUP-uPA/Fbp1^ΔHep livers (n = 5). Numbers/liver and diameters of aggregates are shown underneath. Scale bars, 100 μm. b, c, IB analysis of the indicated proteins in normal hepatocytes and HcPC from a. MW markers, densitometric quantification of protein ratios (HcPC/Normal hepatocytes) and P values are depicted on the right. d, Tumour numbers (left) and volumes (right) of MUP-uPA/Fbp1^F/F and MUP-uPA/Fbp1^ΔHep livers after 40 weeks of HFrD (n = 5–7 BR each). e, Representative CD44, MYC, and AFP staining of liver sections from above mice. Scale bars, 50 μm. Relative staining intensities per HMF are below. f, IB analysis of liver lysates from above mice prepared after 40 weeks of HFrD. MW markers, densitometric quantification of protein ratios (MUP-uPA/Fbp1^ΔHep/MUP-uPA/Fbp1^F/F) and P values are shown on the right. g, Tumour numbers (left) and volumes (right) of MUP-uPA/Fbp1^F/F and MUP-uPA/Fbp1^ΔHep livers after 40 weeks of HFrD +/− Nutlin-3a treatment (25 mg/kg) 2x/week for 8 weeks (n = 4–5 BR). h, MUP-uPA/Fbp1^F/F and MUP-uPA/Fbp1^ΔHep livers were transduced with AAV8-Ctrl, AAV8-FBP1 and AAV8-FBP1^E98A 6–8 weeks after HFrD initiation (n = 4–5 BR). Mice were examined at 40 weeks. Tumour numbers (left) and volumes (right) are shown. Data in a, d, e, g and h are mean ± SEM. Statistical significance determined by two-sided unpaired t-test or Mann–Whitney U test (a, d and e) and one-way ANOVA with Tukey post-hoc tests or Kruskal–Wallis test with Dunn post-hoc tests (g, h) based on data normality distribution. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Extended Data Fig. 6 |. FBP1 loss and AKT activation license NRAS^G12V-induced tumorigenesis.

a, Overall survival of Fbp1^F/F and Fbp1^ΔHep mice receiving NRAS^G12V HTVI (n = 6–8 each). Significance was determined by log-rank test. b, IHC analysis of Fbp1^F/F and Fbp1^ΔHep livers at the indicated times post-NRAS^G12V HTVI (n = 4–5 BR per time point). Scale bars, 50 μm. c, Relative staining intensities per HMF from b. d, IB of liver lysates 1, 4, and 12 weeks after NRAS^G12V HTVI of Fbp1^F/F and Fbp1^ΔHep mice. e, f, IHC of FFPE liver sections from indicated mice livers(e). Scale bars, 50 μm. Relative staining intensities (f). g, IB of NT and T lysates of Fbp1^ΔHep mice 16 weeks after NRAS^G12V HTVI. h, IB of Tp53^F/F and Tp53^ΔHep liver lysates 1 and 4 weeks post NRAS^G12V HTVI. Data in c and f are mean ± SEM. Statistical significance determined by two-sided unpaired t-test or Mann–Whitney U test (c, f) based on data normality distribution. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Extended Data Fig. 7 |. FBP1 inhibits NRAS^G12V-induced tumorigenesis regardless of its catalytic activity.

a, Schematic of NRAS^G12V HTVI followed by AAV8 viral transduction. b, H&E and SR staining of liver sections from Fig. 3e (n = 4–5 BR). Scale bars, 50 μm. SR staining intensities per HMF determined by Image J are shown to the right. c, IB analysis of Fbp1^F/F and Fbp1^ΔHep liver lysates from Fig. 3e. d, IHC of FFPE liver sections from 3e. Scale bars, 50 μm. e, Relative staining intensities per HMF determined by Image J of d. f, ITT of mice from Fig. 3e. AUC quantifications are shown below. Data in b, e and f are mean ± SEM. Statistical significance determined by one-way ANOVA with Tukey post-hoc tests or Kruskal–Wallis test with Dunn post-hoc tests (b, e and f) based on data normality distribution. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, ns, not significant.

Extended Data Fig. 8 |. SA-β-Gal expression in primary hepatocytes and its ATM/ATR dependence.

a, b, Primary Fbp1^F/F and Fbp1^ΔHep hepatocytes were transduced with NRAS^G12V (a) or treated with etoposide (20 μM, 12 h) (b) and stained for SA-β-Gal (n = 4 BR). Scale bars, 100 μm. Image J quantification is shown on the right of a and in b. c, Image J quantification of primary hepatocytes prepared from WT mice 8 weeks post-HTVI with Ctrl or NRAS^G12V that were treated with ATMi (KU60019, 5 μM) or ATRi (AZD6738, 2 μM) for 6 h before SA-β-Gal staining (n = 4 BR). d, Image J quantification of primary hepatocytes prepared from MUP-uPA mice fed for 22 weeks with CSD or HFrD and treated with ATMi (KU60019, 5 μM) or ATRi (AZD6738, 2 μM) for 6 h before staining for SA-β-Gal (n = 3 BR). e, f, IB analysis of WT livers after 12 weeks HTVI of Ctrl or NRAS^G12V (e) and MUP-uPA mice fed with CSD and HFrD for 22 weeks (f). All mice were injected with ATMi (KU60019, 10 mg/kg) or ATRi (AZD6738, 10 mg/kg) 2x/week for 4 weeks (n = 4 BR). g, Primary hepatocytes prepared from 8 weeks old Fbp1^F/F mice, were transfected with NRAS^G12V and after 2 days were infected with AAV-TBG-GFP and AAV-TBG-Cre (10⁷ virus copies per well) for 2 days before staining for SA-β-Gal (n = 3 BR). Image J quantification is shown. h, Primary hepatocytes from WT mice were transfected with NRAS^G12V and after 2 days were treated with BpV (1 μM) and NK252 (2 μM) for 20 h before staining for SA-β-Gal (n = 3 BR). Image J quantification is shown. i, Experimental scheme. MUP-uPA/Fbp1^F/F mice were fed HFrD for 22 weeks, infected with AAV-TBG-Cre and AAV-TBG-GFP and analysed 4 weeks later (n = 5 BR). j, IB analysis of liver lysates from above mice. Quantification of relative normalized protein amounts is shown below each strip. k, IHC of FFPE liver sections from above mice. Scale bars, 50 μm. Relative staining intensities per HMF determined by Image J are shown to the right. Data in a, b, c, d, g, h and k are mean ± SEM. Statistical significance determined by one-way ANOVA with Tukey post-hoc tests or Kruskal–Wallis test with Dunn post-hoc tests (a, b, c, d, g, h and k) based on data normality distribution. *P < 0.05, ***P < 0.001, ****P < 0.0001, ns, not significant.

Extended Data Fig. 9 |. NRF2 activation licences NRAS^G12V-induced HCC development.

a-d, IHC of FFPE liver sections from Nrf2^Tg/Tg and Nrf2^Act-Hep livers 1, 4, 12 and 16 weeks after NRAS^G12 HTVI (n = 4–5 BR) (a, c). Scale bars, 50 μm. Relative staining intensities per HMF determined by Image J (b, d). e, Tumour numbers and volumes in Nrf2^Tg/Tg and Nrf2^Act-Hep mice 12 and 16 weeks after HTVI of NRAS^G12V (n = 4–5 BR). Data in b, d and e are mean ± SEM. Statistical significance determined by two-sided unpaired t-test or Mann–Whitney U test (b, d) and Kruskal–Wallis test with Dunn post-hoc tests (e) based on data normality distribution. **P < 0.01, ***P < 0.001, ****P < 0.0001.

Extended Data Fig. 10 |. NRF2 and FBP1 inversely control cell senescence and proliferation genes.

a, Venn diagram showing overlap in upregulated genes in Nrf2^Act-Hep vs Nrf2^Tg/Tg and Fbp1^ΔHep vs Fbp1^F/F livers (n = 3 BR). b, Venn diagram showing overlap in downregulated genes in Nrf2^Act-Hep vs Nrf2^Tg/Tg and Fbp1^ΔHep vs Fbp1^F/F livers (n = 3 BR). c, GO BP enrichment analysis of the overlapping upregulated genes in Nrf2^Act-Hep vs. Nrf2^Tg/Tg and Fbp1^ΔHep vs. Fbp1^F/F livers 1 week after NRAS^G12V HTVI (n = 3 BR). Dot size corresponds to the number of genes associated with each GO BP term, and dot colour corresponds to statistical significance. d, GO BP enrichment analysis of overlapping downregulated genes in Nrf2^Act-Hep vs. Nrf2^Tg/Tg and Fbp1^ΔHep vs. Fbp1^F/F livers 1 week after NRAS^G12V HTVI (n = 3 BR). e, f, Venn diagrams depicting the overlap in genes that are upregulated (e) or downregulated (f) in Fbp1^ΔHep vs. Fbp1^F/F, Nrf2^Act-Hep vs. Nrf2^Tg/Tg and Tp53^ΔHep vs. Tp53^F/F NCD livers. g, h, Venn diagrams depicting the overlap in genes that are upregulated (g) or downregulated (h) in Fbp1^ΔHep vs. Fbp1^F/F, Nrf2^Act-Hep vs. Nrf2^Tg/Tg 1 week after NRAS^G12V HTVI and Tp53^ΔHep vs. Tp53^F/F HFD-fed livers. i, j, Heatmaps representing cell-senescence- (i) and cell-cycle- (j) associated genes in Nrf2^Act-Hep livers 3 weeks after NRAS^G12V HTVI followed by AAV8-Ctrl and AAV8-FBP1 transduction (n = 3 BR).

Extended Data Fig. 11 |. NRF2 activation and FBP1 loss enable propagation of NRAS induced mutations.

a, INDEL types in HCCs from MUP-uPA/Fbp1^ΔHep mice fed HFrD or HFD for 40 weeks. b, Mutational signatures displayed based on trinucleotide frequency of HFD-induced MUP-uPA/Fbp1^ΔHep HCCs (1 tumour per run). c, d Duplex sequencing showing the number of mutations (c) and INDELs (d) in each signature in the indicated mice. NT- non-tumour, T-tumour. e, Number of mutations in each INDEL signature expressed as somatic mutations per megabase in HCCs from NRAS^G12V transduced Nrf2^Act-Hep livers. f, g, Number of mutations and INDEL types in HCCs of NRAS^G12V transduced Nrf2^Act-Hep livers for 16 weeks.

Extended Data Fig. 12 |. HCC emerges from senescent progenitors.

a, Schematic representation of the p21-Cre construct. The diagram was created using BioRender. b, Relative Cre and p21^CIP1 mRNA amounts in primary hepatocytes transfected with the indicated vectors and treated +/− etoposide (20 μM, 48 h) (n = 3 BR). c, Schematic of experimental protocol. Mice were transduced with the indicated AAV8 vectors 2 weeks (day −14) before NRAS^G12V HTVI. Two weeks later, the mice were i.p. injected with PTENi (BpV) (0.5 mg/kg) or Ctrl for 8 weeks and were sacked at 16 weeks post HTVI (n = 3 BR) for HCC analysis. d, IB analysis of the indicated proteins in liver lysates from above mice. Quantification of relative normalized protein amounts is shown underneath each strip. e, f H&E and IHC of FFPE liver sections from above mice. Scale bars, 20 μm. Relative staining intensities per HMF were determined by Image J are shown to the right. g, Tumour numbers (top) and volumes (bottom) in mT/mG livers transduced with AAV8-Ctrl, AAV8-TBG-Cre or AAV8-p21-Cre 2 wks prior to NRAS^G12 HTVI, followed by BpV (HOpic, 0.5 mg/kg) or Ctrl i.p. injections 2x/wk for 8 wks and analysis at 16 wks post-HTVI (n = 3 each). Data in b, e, f and g are mean ± SEM. Statistical significance determined by two-sided unpaired t-test or Mann–Whitney U test (e, f and g) and Kruskal–Wallis test with Dunn post-hoc tests (b) based on data normality distribution. **P < 0.01, ***P < 0.001, ****P < 0.0001.

Fig. 1 |. p53 dependence, FBP1 downregulation and promoter hypermethylation in human HCC.

a, Representative IHC analysis of 83 human HCC tumour (T) and adjacent NT tissues. Scale bars, 50 μm. The relative staining intensities (Methods) are shown on the right. b, The relative FBP1 amounts in NT and tumour tissues in the CPTAC-LIHC database. c, PXD006512-LIHC patient survival stratified according to FBP1 expression by best expression cut-off. Significance was determined using a log-rank test. d, FBP1 promoter and 5′ region (chromosome (chr.) 9:94633352–94643777) methylation frequency at individual CpG sites (coloured dots; blue, NT hepatocytes; red, HCC). e, Relative Fbp1 mRNA in WT and Trp53^ΔHep tumour and NT tissues 9 months after DEN injection. n = 3 biological replicates. For a,b,e, data are mean ± s.e.m. Statistical significance was determined using two-sided unpaired t-tests or Mann–Whitney U-tests (a,b) and Kruskal–Wallis tests with Dunn post hoc test (e) based on data normality distribution; *P < 0.05, **P < 0.01, ***P < 0.001. The box plots show the median (centre line), first and third quartiles (box limits) and outlier datapoints (whiskers).

Fig. 2 |. DNA-damage-induced FBP1 and p53 upregulation in MASH.

a, Immunoblot (IB) analysis of liver lysates from MUP-uPA mice fed CSD or HFrD for 22 weeks. Molecular masses are shown on the on the right. b, Representative comet assays of isolated hepatocytes. n = 3 biological replicates. Original magnification, ×10. Scale bars, 100 μm. Tail DNA and comet tail length quantifications are shown on the right. HMF, high magnification field. c, Representative IHC analysis of MUP-uPA livers at 22 weeks of HFrD. Scale bars, 50 μm. The relative staining intensities of normal versus early neoplastic hepatocytes are shown on the right. d, Relative FBP1 (left) and ALDOB (right) amounts in low- and advanced-fibrosis human MASLD proteomes. e, Representative and quantified comet assays of hepatocytes (hepa) and HcPCs. n = 3 biological replicates. Original magnification, ×10. Scale bars, 100 μm. f, Mouse hepatocyte subtype distribution, as determined using snRNA-seq. n = 3. g, Immunoblot analysis of liver lysates collected at 22 weeks of HFrD. For b–f, data are mean ± s.e.m. Statistical significance was determined using one-way analysis of variance (ANOVA) with Tukey post hoc tests or Kruskal–Wallis test with Dunn post hoc tests (b,e) and two-sided unpaired t-test or Mann–Whitney U-test (c,d,f) based on data normality distribution; NS, not significant; ****P < 0.0001. The box plots show the median (centre line), first and third quartiles (box limits) and outlier datapoints (whiskers).

Fig. 3 |. FBP1 or p53 loss and AKT activation license NRAS^G12V tumorigenesis.

a, Tumour numbers (left) and volumes (right) in Fbp1^F/F and Fbp1^ΔHep livers 12 and 16 weeks after NRAS^G12V HTVI. n = 4–5 biological replicates. b, Sirius Red staining and AFP, CD44 and MYC IHC at 16 weeks after NRAS^G12V HTVI. Scale bars, 50 μm. Quantified Sirius-Red-positive areas and relative staining intensities are shown on the right. c, Immunoblot analysis of the indicated liver lysates at 16 weeks after NRAS^G12V HTVI. d, Tumour numbers (top) and volumes (bottom) of Trp53^F/F and Trp53^ΔHep livers 12 weeks after NRAS^G12V HTVI. n = 4–5 biological replicates. e, The tumour numbers (top) and volumes (bottom) of Fbp1^F/F and Fbp1^ΔHep livers transduced with AAV8-Ctrl, AAV8-FBP1 and AAV8-FBP1(E98A) were examined 16 weeks after NRAS^G12V HTVI, respectively. n = 4–5 biological replicates. f, The tumour numbers (left) and volumes (right) of the indicated livers 16 weeks after NRAS^G12V HTVI with or without MK2206 (100 mg per kg) treatment for the last 6 weeks. n = 4 biological replicates. g, Immunoblot analysis of liver lysates from f. h, The tumour numbers (top) and volumes (bottom) of the indicated livers at 16 weeks after NRAS^G12V HTVI with or without nutlin-3a (25 mg per kg) treatment for the last 6 weeks. n = 4 biological replicates. For a,b,d–f,h, data are mean ± s.e.m. Statistical significance was determined using one-way ANOVA with Tukey post hoc tests or Kruskal–Wallis test with Dunn post-hoc tests (a,e,f,h) and two-sided unpaired t-tests or Mann–Whitney U-tests (b,d) based on data normality distribution.

Fig. 4 |. NRF2-induced FBP1 degradation relieves senescence to license tumorigenesis.

a, NRF2 distribution in HCC cells transfected with control (Ctrl) or FBP1 short hairpin RNA (shFBP1). n = 3 biological replicates. Scale bars, 10 μm. b, Immunoblot analysis of liver lysates from 8-week-old Nrf2^Tg/Tg and Nrf2^Act-Hep mice. c, Immunoblot analysis of liver lysates from 9-month-old Atg7^F/F and Atg7^ΔHep mice. d, Immunoblot analysis of MYC–NRF2(E79Q)-expressing HCC cells treated with ERKi (SCH772984), MEKi (PD98059) or AKTi (MK2206) at 5 μM for 6 h. e, Co-immunoprecipitation (co-IP) analysis of V5–FBP1 with Flag–TRIM28 from lysates of HEK293T cells that were transfected with or without HA–ERK2(GOF) and proteasome inhibitor (MG132; 10 μM, 6 h) treatment. Gel-separated co-IPs were immunoblotted using anti-Flag antibodies. f, TRIM28-induced ubiquitylation of transfected V5–FBP1 variants in MG132-treated (10 μM, 6 h) HEK293T cells. Gel-separated V5 immunoprecipitates were immunoblotted with anti-MYC antibodies. Blots represent at least three biological replicates. WCL, whole-cell lysates. g, Representative NQO1 and P-ERK1/2 IHC in tumour and NT tissues of 83 human HCCs. Scale bars, 50 μm.

Fig. 5 |. NRF2 and FBP1 inversely affect the hepatocyte transcriptome.

a, Gene Ontology Biological Process (GO BP) enrichment analysis of overlapping upregulated genes in Nrf2^Act-Hep versus Nrf2^Tg/Tg and Fbp1^ΔHep versus Fbp1^F/F livers. n = 3 biological replicates. The dot size and colour represent gene number and statistical significance, respectively. b, Gene Ontology Biological Process enrichment analysis of overlapping downregulated genes in Nrf2^Act-Hep versus Nrf2^Tg/Tg and Fbp1^ΔHep versus Fbp1^F/F liver. n = 3 biological replicates. c, Heat maps of cell senescence mRNAs in Nrf2^Tg/Tg and Nrf2^Act-Hep livers 1 week after control or NRAS^G12V HTVI. d, Heat maps of cell-cycle mRNAs in the above livers. e, Cell senescence (CS) (top) and cell cycle (bottom) scores of FBP1 protein amounts in CPTAC-LIHC data. n = 330. For e, data are mean ± s.e.m.

Fig. 6 |. Mutational signatures and HCC emergence from DNA-damaged and senescent progenitors.

a, Mutational signatures based on the trinucleotide frequency of HFrD-induced MUP-uPA Fbp1^ΔHep HCCs (1 tumour per run). b, Mutational signatures displayed as described above of NRAS^G12V-induced Nrf2^Act-Hep HCCs. c, DAPI-stained frozen mT/mG liver sections (n = 3) with the indicated treatment were scanned using the Olympus SLIDEVIEW VS200 slide scanner. Scale bars, 20 μm. Subs, substitutions.