The origin of hepatocellular carcinoma depends on metabolic zonation

Previous studies have shown that cells in zone 3 can transform into HCC.

“Previous studies have shown that cells in zone 3 can transform into HCC (22–24), but there has not been a head-to-head comparison to assess cancer formation from different zones.”

• cites [22] Adebayo Michael AO Inhibiting glutamine-dependent mTORC1 activation ameliorates liver cancers driven by β-catenin mutations. Cell Metab. 29, 1135– 1150.e6 (2019). doi: 10.1016/j.cmet.2019.01.002 30713111 PMC6506359 sourceCould not verify

  The Introduction describes β-catenin activity in zone-3 hepatocytes and β-catenin mutations in some HCCs. The Results localize GS and p-mTOR-S2448 to zone-3 hepatocytes in normal liver, but neither passage establishes that those cells transformed into HCC. Searches for pericentral HCC origin and zone-3 tumor initiation found no passage demonstrating that transition. (LLM, conf=0.72)

  Experimental — accuracy not yet measured
• cites [23] Kurosaki S Cell fate analysis of zone 3 hepatocytes in liver injury and tumorigenesis. JHEP Rep. 3, 100315 (2021). doi: 10.1016/j.jhepr.2021.100315 34345813 PMC8319533 sourceSupported
• cites [24] Ang CH Lgr5+ pericentral hepatocytes are self-maintained in normal liver regeneration and susceptible to hepatocarcinogenesis. Proc. Natl. Acad. Sci. U.S.A 116, 19530– 19540 (2019). doi: 10.1073/pnas.1908099116 31488716 PMC6765306 sourceSupported

The HCC driver genes TP53 and CTNNB1 are not the most frequently observed mutant genes in the cirrhotic liver.

“The HCC driver genes TP53 (tumor protein p53) and CTNNB1 (catenin beta 1) are not the most frequently observed mutant genes in the cirrhotic liver (7–9), suggesting that these mutations do not drive extensive clonal expansion events before tumor initiation.”

• cites [7] Zhu M Somatic mutations increase hepatic clonal fitness and regeneration in chronic liver disease. Cell 177, 608– 621.e12 (2019). doi: 10.1016/j.cell.2019.03.026 30955891 PMC6519461 sourceSupported
• cites [8] Brunner SF Somatic mutations and clonal dynamics in healthy and cirrhotic human liver. Nature 574, 538– 542 (2019). doi: 10.1038/s41586-019-1670-9 31645727 PMC6837891 sourcePartly supported

  The driver-mutation analysis (section labelled Abstract; page unavailable) identifies ALB and ACVR2A, rather than TP53 or CTNNB1, as significant among the 30 HCC genes tested. It does not give a gene-frequency ranking specifically for cirrhotic liver; its reported variants combine normal and cirrhotic samples. (LLM, conf=0.78)

  Experimental — accuracy not yet measured
• cites [9] Ng SWK Convergent somatic mutations in metabolism genes in chronic liver disease. Nature 598, 473– 478 (2021). doi: 10.1038/s41586-021-03974-6; 34646017 sourceCould not verify

  No passage from our copy addresses whether TP53 and CTNNB1 are among the most frequently mutated genes in cirrhotic liver. Searches for both gene names and for mutations in cirrhosis returned only bibliographic material, with no findings or page-level evidence to assess the claim. (LLM, conf=0.70)

  Experimental — accuracy not yet measured

TP53 and CTNNB1 mutations do not drive extensive clonal expansion events before tumor initiation.

“The HCC driver genes TP53 (tumor protein p53) and CTNNB1 (catenin beta 1) are not the most frequently observed mutant genes in the cirrhotic liver (7–9), suggesting that these mutations do not drive extensive clonal expansion events before tumor initiation.”

• cites [7] Zhu M Somatic mutations increase hepatic clonal fitness and regeneration in chronic liver disease. Cell 177, 608– 621.e12 (2019). doi: 10.1016/j.cell.2019.03.026 30955891 PMC6519461 sourcePartly supported

  The Results (Ultra-deep targeted sequencing identified recurrent somatic mutations; page unavailable) report TRP53 mutations in 6% of patients and say CTNNB1 was not detected in more than two patients, suggesting different selection pressures in HCC and background liver. These findings support the citing sentence’s observation about mutation frequency, but the paper does not establish that TP53 and CTNNB1 mutations do not drive extensive clonal expansion before tumor initiation. (LLM, conf=0.82)

  Experimental — accuracy not yet measured
• cites [8] Brunner SF Somatic mutations and clonal dynamics in healthy and cirrhotic human liver. Nature 574, 538– 542 (2019). doi: 10.1038/s41586-019-1670-9 31645727 PMC6837891 sourcePartly supported

  The concluding passage (labeled Abstract in our copy; page unavailable) says driver mutations in non-malignant liver did not drive large clonal expansions. The driver-screening passage in the same section identifies other genes but does not establish this finding specifically for TP53 or CTNNB1, or show how frequently those two genes were mutated in cirrhotic liver. (LLM, conf=0.83)

  Experimental — accuracy not yet measured
• cites [9] Ng SWK Convergent somatic mutations in metabolism genes in chronic liver disease. Nature 598, 473– 478 (2021). doi: 10.1038/s41586-021-03974-6; 34646017 sourceCould not verify

  No passage from our copy addresses whether TP53 or CTNNB1 mutations drive extensive clonal expansion before tumor initiation. Searches for both genes and for clonal expansion in cirrhotic liver returned no relevant passage, so the claim cannot be assessed from the available evidence. (LLM, conf=0.70)

  Experimental — accuracy not yet measured

Mice with Cre recombinase activity in distinct zonal populations permit induction of HCC mutations in a variety of normal cell types.

“The recent generation of mice with Cre recombinase activity in distinct zonal populations permitted us to induce HCC mutations in a variety of normal cell types (18).”

• cites [18] Wei Y Liver homeostasis is maintained by midlobular zone 2 hepatocytes. Science 371, eabb1625 (2021). doi: 10.1126/science.abb1625 33632817 PMC8496420 sourceSupported

Activating CTNNB1 mutations on their own may not be sufficient to drive cancer.

“Given that activating CTNNB1 mutations on their own may not be sufficient to drive cancer (26), we sought to co-mutate another gene that would accelerate cancer development.”

• cites [26] Dong B Activating CAR and β-catenin induces uncontrolled liver growth and tumorigenesis. Nat. Commun 6, 5944 (2015). doi: 10.1038/ncomms6944 25661872 PMC4324535 sourcePartly supported

  The Introduction and Discussion state that β-catenin activation alone does not induce spontaneous liver tumors in mouse models, indicating that additional hits are needed. The Results likewise report that β-catenin activation alone did not induce liver tumors in the tested mice. This supports the cautious claim for mouse liver cancer, but the paper does not establish it for cancer generally. (LLM, conf=0.78)

  Experimental — accuracy not yet measured

The inefficient CreER recombination faithfully modeled the somatic mosaicism seen in human livers before cancer development.

“While this showed that CreER was not 100% efficient, it also faithfully modeled the somatic mosaicism seen in human livers before cancer development (7–9, 33).”

• cites [7] Zhu M Somatic mutations increase hepatic clonal fitness and regeneration in chronic liver disease. Cell 177, 608– 621.e12 (2019). doi: 10.1016/j.cell.2019.03.026 30955891 PMC6519461 sourcePartly supported

  The Introduction reports somatic mutations and evidence of clonal expansion in non-dysplastic human liver tissue, supporting the claim that mosaicism exists before cancer. The paper does not assess inefficient CreER recombination or establish that it faithfully models that mosaicism; its mouse methods describe other approaches. (LLM, conf=0.86)

  Experimental — accuracy not yet measured
• cites [8] Brunner SF Somatic mutations and clonal dynamics in healthy and cirrhotic human liver. Nature 574, 538– 542 (2019). doi: 10.1038/s41586-019-1670-9 31645727 PMC6837891 sourcePartly supported

  The paper describes non-malignant human liver as a “clonal patchwork” and healthy liver as a community of small clones (Abstract; page unavailable), supporting the claim that mosaicism exists before liver cancer. It does not examine CreER recombination or establish that inefficient recombination faithfully models that mosaicism. (LLM, conf=0.84)

  Experimental — accuracy not yet measured
• cites [9] Ng SWK Convergent somatic mutations in metabolism genes in chronic liver disease. Nature 598, 473– 478 (2021). doi: 10.1038/s41586-021-03974-6; 34646017 sourceCould not verify

  The supplied passage contains only the paper’s title and author information (no section or page). Searches for somatic mosaicism, CreER recombination, and hepatocyte clones found no eligible passage in our copy establishing that inefficient CreER recombination faithfully models mosaicism in human livers before cancer development. (LLM, conf=0.72)

  Experimental — accuracy not yet measured
• cites [33] Wang Z Positive selection of somatically mutated clones identifies adaptive pathways in metabolic liver disease. Cell 186, 1968– 1984.e20 (2023). doi: 10.1016/j.cell.2023.03.014 37040760 PMC10321862 sourcePartly supported

  The Discussion (page unavailable) says the authors established MOSAICS to model widespread somatic mosaicism, supporting the general modeling point. The Results describe an AAV-based system using a Cre fusion protein, not inefficient CreER recombination; the paper does not establish that CreER faithfully reproduces mosaicism in human livers before cancer develops. (LLM, conf=0.82)

  Experimental — accuracy not yet measured

Ectopic Ctnnb1 activation by itself induces glutamine synthetase (Glul/GS) expression outside the endogenous GS expression normally observed in a two- to three-cell ring of hepatocytes around the central vein.

“In the liver, ectopic Ctnnb1 activation by itself induces glutamine synthetase (gene: Glul; protein: GS) expression outside of the endogenous GS expression normally observed in a two- to three-cell ring of hepatocytes around the central vein (CV) (28–32).”

• cites [28] Gebhardt R, Mecke D, Heterogeneous distribution of glutamine synthetase among rat liver parenchymal cells in situ and in primary culture. EMBO J. 2, 567– 570 (1983). doi: 10.1002/j.1460-2075.1983.tb01464.x 6138251 PMC555062 sourcePartly supported

  The Results (page unavailable) report that GS-positive rat liver cells form a narrow zone, usually 1–3 cell layers, directly around terminal hepatic venules. The paper does not address ectopic Ctnnb1 activation or show that it induces GS expression outside that zone. (LLM, conf=0.90)

  Experimental — accuracy not yet measured
• cites [29] Hale G Correlation of exon 3 β-catenin mutations with glutamine synthetase staining patterns in hepatocellular adenoma and hepatocellular carcinoma. Mod. Pathol 29, 1370– 1380 (2016). doi: 10.1038/modpathol.2016.122 27469330 PMC5149418 sourcePartly supported

  The Introduction (page not available) says that β-catenin activates GLUL transcription, that normal glutamine synthetase expression is confined to hepatocytes around terminal hepatic venules, and that expression extends through much of β-catenin-activated tumors. It does not specify a two- to three-cell ring or establish that ectopic Ctnnb1 activation by itself induces expression outside that ring; the Discussion (page not available) cautions that the relationship between β-catenin activation and glutamine synthetase staining is complex. (LLM, conf=0.85)

  Experimental — accuracy not yet measured
• cites [30] Huang W-J, Tsai J-H, Jeng Y-M, Complementary roles of β-catenin and glutamine synthetase immunostaining in diagnosis of chemotherapy-treated and untreated hepatoblastoma. J. Formos. Med. Assoc 116, 549– 553 (2017). doi: 10.1016/j.jfma.2016.09.013 28277286 sourcePartly supported

  The Introduction identifies GS as a downstream target of Wnt/β-catenin signaling, and the Results report diffuse GS staining in hepatoblastoma versus a perivenular pattern in nontumorous liver. But this immunostaining study does not establish that ectopic Ctnnb1 activation by itself induces GS; the Discussion allows for other mechanisms. Nor does it specify a two- to three-cell ring around the central vein. Page numbers were not available in our copy. (LLM, conf=0.86)

  Experimental — accuracy not yet measured
• cites [31] Loeppen S Overexpression of glutamine synthetase is associated with β-catenin-mutations in mouse liver tumors during promotion of hepatocarcinogenesis by phenobarbital. Cancer Res. 62, 5685– 5688 (2002). 12384525Not checked — no full text of the source
• cites [32] Sekine S, Lan BY-A, Bedolli M, Feng S, Hebrok M, Liver-specific loss of β-catenin blocks glutamine synthesis pathway activity and cytochrome p450 expression in mice. Hepatology 43, 817– 825 (2006). doi: 10.1002/hep.21131 16557553 sourceNot checked — no full text of the source

Among 14 frequently mutated or copy number-altered genes in HCC, CTNNB1 mutations were most likely to co-occur with ARID2 mutations.

“Among 14 frequently mutated or copy number–altered genes in HCC (25), CTNNB1 mutations were most likely to co-occur with ARID2 mutations (Fig. 1A).”

• cites [25] Cancer Genome Atlas Research Network, Comprehensive and integrative genomic characterization of hepatocellular carcinoma. Cell 169, 1327– 1341.e23 (2017). doi: 10.1016/j.cell.2017.05.046 28622513 PMC5680778 sourceSupported

Many GST protein family members act as tumor promoters by influencing reactive oxygen species (ROS).

“Many GST protein family members act as tumor promoters by influencing reactive oxygen species (ROS) (51–56).”

• cites [51] Collino A, Sustained activation of detoxification pathways promotes liver carcinogenesis in response to chronic bile acid-mediated damage. PLOS Genet. 14, e1007380 (2018). doi: 10.1371/journal.pgen.1007380 29734330 PMC5957449 sourcePartly supported

  The Results sections on hepatocyte gene expression and detoxification enzymes report that several glutathione S-transferases are upregulated in Mdr2-/- livers and place them within detoxification pathways active during tumor development. They also report an oxidative-stress response, but do not show that GST family members themselves act as tumor promoters by influencing ROS. No page number is available in our copy. (LLM, conf=0.83)

  Experimental — accuracy not yet measured
• cites [52] Townsend DM, Tew KD, The role of glutathione-S-transferase in anti-cancer drug resistance. Oncogene 22, 7369– 7375 (2003). doi: 10.1038/sj.onc.1206940 14576844 PMC6361125 sourcePartly supported

  Table 1 (page unavailable) lists products generated by reactive oxygen species as substrates of α and μ GSTs, and the discussion of GSTs as regulators of the MAP kinase pathway links GSTπ and GSTμ to cell-survival signaling. The paper does not say that many GST family members promote tumors by influencing ROS; it describes elevated GST expression in tumors and says GSTπ may be either a cause or an effect of transformation. (LLM, conf=0.84)

  Experimental — accuracy not yet measured
• cites [53] Chen T, GSTM3 deficiency impedes DNA mismatch repair to promote gastric tumorigenesis via CAND1/NRF2-KEAP1 signaling. Cancer Lett. 538, 215692 (2022). doi: 10.1016/j.canlet.2022.215692 35487311 sourceNot checked — no full text of the source
• cites [54] Huang J, Tan PH, Tan BKH, Bay BH, GST-pi expression correlates with oxidative stress and apoptosis in breast cancer. Oncol. Rep 12, 921– 925 (2004). doi: 10.3892/or.12.4.921 15375523 sourceNot checked — no full text of the source
• cites [55] Fiander H, Schneider H, Dietary ortho phenols that induce glutathione S-transferase and increase the resistance of cells to hydrogen peroxide are potential cancer chemopreventives that act by two mechanisms: The alleviation of oxidative stress and the detoxification of mutagenic xenobiotics. Cancer Lett. 156, 117– 124 (2000). doi: 10.1016/S0304-3835(00)00368-2 10880760 sourceNot checked — no full text of the source
• cites [56] Cederbaum AI, Nrf2 and antioxidant defense against CYP2E1 toxicity. Subcell. Biochem 67, 105– 130 (2013). doi: 10.1007/978-94-007-5881-0_2 23400918 sourceNot checked — no full text of the source

In human HCCs, higher expression of GSTM3 was associated with shorter survival.

“In human HCCs, higher expression of GSTM3 was associated with shorter survival (fig. S7E) (44, 45).”

• cites [44] Chandrashekar DS, UALCAN: A portal for facilitating tumor subgroup gene expression and survival analyses. Neoplasia 19, 649– 658 (2017). doi: 10.1016/j.neo.2017.05.002 28732212 PMC5516091 sourceCould not verify

  The Methods (Data Analyses; page unavailable) describe survival analyses by gene expression, and the Results (Usage of UALCAN; page unavailable) discuss liver hepatocellular carcinoma. Neither reports a GSTM3 survival result or says whether higher GSTM3 expression is associated with shorter survival. Searches for GSTM3 and its full gene name found no such passage in our copy. (LLM, conf=0.76)

  Experimental — accuracy not yet measured
• cites [45] Chandrashekar DS, UALCAN: An update to the integrated cancer data analysis platform. Neoplasia 25, 18– 27 (2022). doi: 10.1016/j.neo.2022.01.001 35078134 PMC8788199 sourceCould not verify

  The Methods section (Survival analyses; page not provided) describes how UALCAN compares survival by expression level. Searches for GSTM3 and HCC found no passage in our copy reporting whether higher GSTM3 expression is associated with shorter survival in human HCCs. (LLM, conf=0.75)

  Experimental — accuracy not yet measured

Human HCCs with CTNNB1 and ARID2 mutations from two independent studies also showed up-regulation of many of the same WNT pathway genes (NOTUM, AQP9, TBX3, ABCB11, GLUL) as observed in mouse models.

“Notably, human HCCs with CTNNB1 and ARID2 mutations from two independent studies (25, 35) also showed up-regulation of many of the same WNT pathway genes (NOTUM, AQP9, TBX3, ABCB11, GLUL) as observed in our mouse models (fig. S6B and table S4).”

• cites [25] Cancer Genome Atlas Research Network, Comprehensive and integrative genomic characterization of hepatocellular carcinoma. Cell 169, 1327– 1341.e23 (2017). doi: 10.1016/j.cell.2017.05.046 28622513 PMC5680778 sourcePartly supported

  The Results sections on Somatic Mutations and Other Signaling Pathways (pages unavailable) report CTNNB1 and ARID2 mutations in human HCC and frequent alterations in WNT signaling. They do not report that HCCs with those mutations up-regulate NOTUM, AQP9, TBX3, ABCB11, or GLUL. Searches for the named genes and their expression did not establish that part of the claim. (LLM, conf=0.83)

  Experimental — accuracy not yet measured
• cites [35] Chen L, Deep whole-genome analysis of 494 hepatocellular carcinomas. Nature 627, 586– 593 (2024). doi: 10.1038/s41586-024-07054-3 38355797 sourceNot checked — no full text of the source

CTNNB1 or MYC alone was insufficient to drive cancer within 4 weeks.

“CTNNB1 or MYC alone was insufficient to drive cancer within 4 weeks (38), but delivering both led to carcinogenesis within 1 to 2 weeks on the FVB strain background.”

• cites [38] Molina-Sánchez P, Cooperation between distinct cancer driver genes underlies intertumor heterogeneity in hepatocellular carcinoma. Gastroenterology 159, 2203– 2220. e14 (2020). doi: 10.1053/j.gastro.2020.08.015 32814112 PMC7726023 sourcePartly supported

  The Results section (‘In vivo screen identifies cooperating driver genes in HCC’; page unavailable) reports that individual genetic alterations were not associated with mortality within eight months and that two drivers were required for efficient liver tumorigenesis in C57BL/6 mice. It does not establish that CTNNB1 or MYC alone failed to produce cancer specifically within four weeks; mortality is not the same measure as tumor formation. (LLM, conf=0.82)

  Experimental — accuracy not yet measured

Gstm2 has been shown to increase liver tumorigenesis through inactivation of ASK1, a kinase that promotes apoptosis.

“Gstm2 has been shown to increase liver tumorigenesis through the inactivation of ASK1, a kinase that promotes apoptosis (49, 50).”

• cites [49] Lan T, Hepatocyte glutathione S-transferase mu 2 prevents non-alcoholic steatohepatitis by suppressing ASK1 signaling. J. Hepatol 76, 407– 419 (2022). doi: 10.1016/j.jhep.2021.09.040 34656650 sourceNot checked — no full text of the source
• cites [50] Nakagawa H, Apoptosis signal-regulating kinase 1 inhibits hepatocarcinogenesis by controlling the tumor-suppressing function of stress-activated mitogen-activated protein kinase. Hepatology 54, 185– 195 (2011). doi: 10.1002/hep.24357; 21488081 sourcePartly supported

  The Discussion (page unavailable) concludes that ASK1 acts as a tumor suppressor in hepatocarcinogenesis, and the apoptosis results report fewer apoptotic tumor cells in ASK1-deficient mice. The paper supports the ASK1 portion of the claim, but searches for Gstm2 and its proposed inactivation of ASK1 found no passage supporting that mechanism. (LLM, conf=0.87)

  Experimental — accuracy not yet measured

At 5 months after tamoxifen, the zone 1 mutant mouse had 63% nonmutant, 8.3% Ctnnb1 single-mutant, 21% Arid2 single-mutant, and 7.3% double-mutant cells.

“At 5 months after tamoxifen, which preceded any cancer development, the zone 1 mutant mouse had 63% nonmutant, 8.3% Ctnnb1 single-mutant, 21% Arid2 single-mutant, and 7.3% double-mutant cells (Fig. 2, E and F).”

No mice from any of the mutant models developed macroscopic tumors before the 6-month time point.

“No mice from any of the mutant models developed macroscopic tumors before the 6-month time point, so we performed cancer analysis 6 months after tamoxifen (Fig. 3, A and B, and table S1).”

Tumors from both zones did not segregate according to zone of origin in RNA-seq expression.

“We transcriptionally profiled tumors from both zones using RNA sequencing (RNA-seq) and found that the tumor expression did not segregate according to zone of origin (Fig. 3P and fig. S6A).”

Mutant clones in zones 1 and 3 had high expression of zone 3 genes such as Glul and Cyp1a2.

“Mutant clones in zones 1 and 3 had high expression of zone 3 genes such as Glul and Cyp1a2.”

In the zone 1 mutant liver, GS+ hepatocytes were ectopically located in zone 1.

“In the zone 1 mutant liver, GS+ hepatocytes were ectopically located in zone 1.”

In zone 3 mutant mice, there was a similar absolute number of proliferating hepatocytes in comparison to WT mice.

“In zone 3 mutant mice, there was a similar absolute number of proliferating hepatocytes in comparison to WT mice (fig. S4, A and B).”

When considering only 8- to 10-month post-tamoxifen mice, there were still more surface tumors in zone 3 mice (mean 10.3 versus 3.67; P = 0.0253).

“When considering only 8- to 10-month post-tamoxifen mice, there were still more surface tumors in zone 3 mice (mean: 10.3 versus 3.67; P = 0.0253) (Fig. 3D).”

The fractional GS+ area showed a reduction of mutant clone abundance in zone 3 mice compared with zone 1 mice.

“The fractional GS+ area (GS+ area/total tissue area) also showed a reduction of mutant clone abundance in zone 3 mice compared with zone 1 mice (Fig. 2D).”

At 5 months after tamoxifen, the zone 3 mutant mouse had 88% nonmutant, 0.62% Ctnnb1 single-mutant, 11% Arid2 single-mutant, and 0.69% double-mutant cells.

“In the zone 3 mutant mouse, there were 88% nonmutant, 0.62% Ctnnb1 singlemutant, 11% Arid2 single-mutant, and 0.69% double-mutant cells (Fig. 2, E and G).”

After 6 months, the GS+ mutant population decreased in size in zone 3 mutant mice.

“After 6 months, the Tomato+ population remained stable (fig. S5, A and B), whereas the GS+ mutant population decreased in size (Fig. 2C and fig. S5C), suggesting that WT cells from within zone 3 expanded.”

Blinded histologic analyses showed that 15 of 25 zone 3 mice (60%) and 2 of 16 zone 1 mice (12.5%) harbored malignant lesions, defined as dysplastic nodules, early HCC, or HCC.

“Blinded histologic analyses showed that 15 of 25 zone 3 mice (60%) and 2 of 16 zone 1 mice (12.5%) harbored malignant lesions, defined as dysplastic nodules, early HCC, or HCC (Fig. 3, E to H, and table S2; subcategories defined by the most common lesion type in the section).”

In either zone, only 2 of all 58 (3.4%) Arid2 single mutant mice developed lesions by 14 months.

“In either zone, only 2 of all 58 (3.4%) Arid2 single mutant mice developed lesions by 14 months, indicating that Arid2 mutations alone were insufficient for carcinogenesis (Fig. 3C).”

Zone 3 livers harbored a much higher number of surface tumors than zone 1 livers (mean 9.62 versus 3.68; P = 0.0002).

“Also, zone 3 livers harbored a much higher number of surface tumors (mean: 9.62 versus 3.68; P = 0.0002) (Fig. 3C).”

In livers containing mutant hepatocytes, Ctnnb1/Arid2 mutant cells took on a zone 3 identity regardless of their location.

“In livers containing mutant hepatocytes, Ctnnb1/Arid2 mutant cells took on a zone 3 identity regardless of their location.”

After a 10-day EdU pulse chase, there was an increased absolute number of proliferating hepatocytes in zone 1 mutant versus WT livers.

“After a 10-day 5-ethynyl-2′-deoxyuridine (EdU) pulse chase, we observed an increased absolute number of proliferating hepatocytes in zone 1 mutant versus WT livers (fig. S4, A and B).”

After 6 months, the Tomato+ population remained stable in zone 3 mutant mice.

“After 6 months, the Tomato+ population remained stable (fig. S5, A and B), whereas the GS+ mutant population decreased in size (Fig. 2C and fig. S5C), suggesting that WT cells from within zone 3 expanded.”

51/64 (80%) double-mutant mice with mutations in either zone developed cancers, usually in a multifocal pattern.

“In contrast, 51/64 (80%) double-mutant mice with mutations in either zone developed cancers, usually in a multifocal pattern (Fig. 3, A to C, and table S1).”

Mutant clones in zone 3 appeared to have a hybrid pattern of expression that included zone 2 genes in addition to zone 3 genes.

“Mutant clones in zone 3 appeared to have a hybrid pattern of expression that included zone 2 genes in addition to zone 3 genes.”

The relative GS+ area in zone 3 mutant livers decreased at 2 and 6 months.

“In contrast, the relative GS+ area in zone 3 mutant livers decreased at 2 and 6 months (Fig. 2C).”

Mutant clones in zone 1 appeared to have a hybrid pattern of expression that included zone 2 and 3 genes in addition to zone 1 genes.

“Mutant clones in zone 1 appeared to have a hybrid pattern of expression that included zone 2 and 3 genes in addition to zone 1 genes (fig. S3D).”

More Ki-67+ nuclei were observed in zone 1 mutant livers at the 6 month time point.

“In agreement with EdU tracing, we also observed more Ki-67+ nuclei in the zone 1 mutant livers at the 6 month time point (compare Fig. 2J and fig. S4, C and D).”

The frequency of GS+ mutant hepatocytes in zone 1 mutant mice was stable over time.

“The frequency of GS+ mutant hepatocytes in zone 1 mutant mice was stable over time (Fig. 2A).”

In the Ctnnb1 heterozygous;Arid2 heterozygous model, zone 3 livers harbored a much higher number of surface tumors than zone 1 livers (mean 18.64 versus 0.94; P < 0.0001).

“In addition, zone 3 livers harbored a much higher number of surface tumors (mean: 18.64 versus 0.94; P < 0.0001) (Fig. 3K).”

The frequency of mutant hepatocytes in zone 3 mutant mice decreased over time.

“In contrast, the frequency of mutant hepatocytes in zone 3 mutant mice decreased over time (Fig. 2B).”

Expression of Glul and Cyp1a2 in mutant clones was higher than in zone 1 hepatocytes.

“In all cases, the expression of these genes was higher than in zone 1 hepatocytes.”

Among EdU+ hepatocytes in zone 1 mutant mice, a majority (65%) were also GS+ Ctnnb1 mutants.

“Among the EdU+ hepatocytes in zone 1 mutant mice, a majority (65%) were also GS+ Ctnnb1 mutants.”

Zone 1 mutant clones had higher expression of Ccnd1, Ccnb2, Cdk1, and Anln than did surrounding WT hepatocytes and zone 3 mutant clones.

“As for proliferation, zone 1 mutant clones had higher expression of Ccnd1, Ccnb2, Cdk1, and Anln than did surrounding WT hepatocytes and zone 3 mutant clones.”

In all examined mutant cells from ApoC4-CreERT2;Arid2fl/fl;Ctnnb1(ex3)fl/+ mice with nuclear and/or cytoplasmic CTNNB1, high GS staining was also observed.

“In all examined mutant cells from ApoC4-CreERT2;Arid2fl/fl; Ctnnb1(ex3)fl/+ mice, which had nuclear and/or cytoplasmic CTNNB1, high GS staining was also observed.”

The relative GS+ cell area normalized to the 1-month time point in zone 1 mutant livers was largely unchanged over time, with the exception of an increase at 3 months.

“The relative GS+ cell area, normalized to the 1-month time point (set at 100% for each zonal mouse model), in the zone 1 mutant livers was largely unchanged over time, with the exception of an increase at 3 months.”

In the zone 3 mutant liver, GS+ hepatocytes were ectopically located in zones 2 and 3 but were farther from the CV than would be expected for a normal zone 3 hepatocyte.

“In the zone 3 mutant liver, GS+ hepatocytes were ectopically located in zones 2 and 3 but were farther from the CV than would be expected for a normal zone 3 hepatocyte (Fig. 2I).”

No additive reductions in gross or microtumor burden were observed in the combination group of ethacrynic acid plus CRISPR-mediated Gstm3 deletion.

“No additive reductions in gross or microtumor burden in the combination group were observed, suggesting that ethacrynic acid was acting through GSTM3 (Fig. 5H).”

Compared with vehicle, ethacrynic acid resulted in a 75% reduction in microtumor area.

“Compared with vehicle, ethacrynic acid did not affect HDT efficiency (fig. S9A) but did result in a lower LW/BW and a 75% reduction in microtumor area (Fig. 5, B to D).”

Ethacrynic acid led to increased peroxide levels compared with vehicle treatment.

“Ethacrynic acid led to increased peroxide levels compared with vehicle treatment, suggesting successful inhibition of GSTM activity (Fig. 5E).”

Two doses of Gstm3 siRNA before CTNNB1/MYC HDT reduced LW/BW ratios and microtumor burden compared with nontargeting control siRNA.

“We gave two doses of Gstm3 siRNA before CTNNB1/MYC HDT, which reduced LW/BW ratios and microtumor burden compared with the nontargeting control siRNA (Fig. 4, M to O).”

One week after introduction of oncogenes, sgGstm3 and sgGstm2 KO livers showed a decrease in the number and area of microscopic tumors.

“One week after the introduction of oncogenes, sgGstm3 and sgGstm2 knockout (KO) livers showed a clear decrease in the number and area of microscopic tumors, indicating a block in tumor initiation (Fig. 4, E and F).”

There was no significant difference in the total number of tumors between control and Gstm3 overexpression groups (6.2 versus 14.2 tumors per WT or overexpression liver, P = 0.1518).

“There was no significant difference in the total number of tumors between groups (6.2 versus 14.2 tumors per WT or overexpression liver, P = 0.1518) (fig. S8D).”

There was a significant increase in the mean number of Tomato+ HCCs originating from zone 1 in GSTM3 overexpression versus control (1.3 versus 8.7 HCC per mouse, P = 0.0453).

“There was a significant increase in the mean number of Tomato+ HCCs that originated from zone 1 (1.3 versus 8.7 HCC per mouse in control versus overexpression, P = 0.0453; Fig. 4S).”

Cre-driven Ctnnb1/Arid2 tumors were phenotypically and transcriptionally similar regardless of zone of origin.

“The fact that Cre-driven Ctnnb1/Arid2 tumors were phenotypically and transcriptionally similar regardless of zone of origin (Fig. 3P and fig. S6A) supported the idea that Gstm2/3 in hepatocytes, rather than in established tumors, was responsible for differences in tumor initiation.”

Compared with vehicle, ethacrynic acid did not affect HDT efficiency.

“Compared with vehicle, ethacrynic acid did not affect HDT efficiency (fig. S9A) but did result in a lower LW/BW and a 75% reduction in microtumor area (Fig. 5, B to D).”

Ethacrynic acid-treated WT FVB mice (2 days before and 2 days after HDT) showed decreased LW/BW five days after the last injection.

“Five days after the last injection, the ethacrynic acid–treated mice showed a decreased LW/BW, further supporting the observation that baseline levels of Gstm3 are important in initiation (Fig. 5F).”

Ethacrynic acid prevented Ctnnb1-driven HCC initiation by increasing ferroptosis in premalignant cells.

“Genetic deletion of Gstm2 or Gstm3 impaired HCC initiation, and ethacrynic acid, a loop diuretic known to inhibit GSTM2 and GSTM3, also prevented Ctnnb1-driven HCC initiation by increasing ferroptosis in premalignant cells.”

• AssumptionRests on an assumption: Ethacrynic acid inhibits GSTM2 and GSTM3.

Five of seven murine Gstm family members had increased expression in zone 3 compared with zone 1.

“Examination of all seven murine Gstm family members showed that five had increased expression in zone 3 compared with zone 1, suggesting that Gstm2 and Gstm3 may be representatives of a family-wide effect (fig. S7F).”

Whole-body Gstm3 KO mice developed less CTNNB1/MYC tumors than WT mice.

“KO mice developed less CTNNB1/MYC tumors than did WT mice (Fig. 4, G and H, and fig. S7H), supporting the concept that Gstm3 within zone 3 hepatocytes was required for tumor initiation.”

There was no difference in the mean number of Tomato− HCCs originating from zone 3 between control and GSTM3 overexpression (4.8 versus 5.5 HCC per mouse, P = 0.8765).

“As expected, there was no difference in the mean number of Tomato− HCCs that originate from zone 3 (4.8 versus 5.5 HCC per mouse in control versus overexpression, P = 0.8765; Fig. 4R).”

Delivering both CTNNB1 and MYC led to carcinogenesis within 1 to 2 weeks on the FVB strain background.

“CTNNB1 or MYC alone was insufficient to drive cancer within 4 weeks (38), but delivering both led to carcinogenesis within 1 to 2 weeks on the FVB strain background.”

Alternative liver cancer models, driven by either mutagens or other oncogenes, also arose in zone 3.

“Alternative liver cancer models, driven by either mutagens or other oncogenes, also arose in zone 3.”

In HCC53N, liproxstatin-1 abrogated RSL3-induced cell death.

“In HCC53N, ferroptosis induction with a small-molecule inhibitor of GPX4 (RSL3) killed cells, and the ferroptosis inhibitor liproxstatin-1 abrogated RSL3-induced cell death (Fig. 5I, left panel).”

• AssumptionRests on an assumption: RSL3 induces ferroptosis by inhibiting GPX4, and liproxstatin-1 inhibits ferroptosis.

Tumors driven by NRAS/shp53 came primarily from zone 3.

“Similarly, tumors driven by NRAS/shp53 came primarily from zone 3, demonstrated by a higher proportion of Tomato+ tumors in the zone 3–labeled mice and Tomato− tumors in the zone 1–labeled mice (fig. S15, B and C).”

• AssumptionRests on an assumption: The Gls2-CreERT2 and Cyp1a2-CreERT2 lines specifically label zone 1 and zone 3 hepatocytes, respectively.
• AssumptionRests on an assumption: Tomato expression is a reliable lineage-tracing marker that persists in labeled hepatocytes and their progeny.

Gstm3 deletion increased 4HNE in NRAS/shp53 context.

“Gstm3 deletion also increased 4HNE in this context (fig. S16B).”

Fsp1 overexpression did not affect the CTNNB1/MYC model in the control setting.

“Fsp1 overexpression did not affect the CTNNB1/MYC model in the control setting but did eliminate the LW/BW suppressing effects of Gstm3 deletion (Fig. 5R).”

• AssumptionRests on an assumption: Fsp1 overexpression does not have off-target effects that affect tumorigenesis.

Other zone 3 genes promote tumorigenesis.

“We also identified other zone 3 genes that promote tumorigenesis, suggesting that GSTMs are not alone in their protumor functions.”

Liproxstatin-1 reversed ethacrynic acid-induced sensitization to RSL3-mediated cell death.

“This effect was reversed by liproxstatin-1, demonstrating that ethacrynic acid–induced sensitization to RSL3-mediated cell death acts through ferroptosis (Fig. 5I, right panel).”

• AssumptionRests on an assumption: RSL3 induces ferroptosis by inhibiting GPX4, and liproxstatin-1 inhibits ferroptosis.

Transient knockdown with a single dose of siRNA before HDT reduced LW/BW ratios.

“Gstm3 deletion increased survival in NRAS/shp53 mice (fig. S15D), and transient knockdown with a single dose of siRNA before HDT reduced LW/BW ratios (fig. S15E), indicating that basal Gstm3 expression in zone 3 hepatocytes is necessary for NRAS/shp53-driven HCC.”

NRAS/shp53 livers showed decreased GSH/GSSG ratios in mice receiving sgGstm3 or sgNrf2.

“Additionally, NRAS/shp53 livers showed decreased GSH/GSSG ratios in mice receiving sgGstm3 or sgNrf2, indicating that loss of Gstm3 increases oxidative stress (fig. S16A).”

Gstm2 KO HCC53N and HCC53M cells did not show significantly increased sensitivity to ferroptosis.

“Gstm2 KO HCC53N and HCC53M cells did not show significantly increased sensitivity to ferroptosis (fig. S13, A and B, and Fig. 5N).”

Cell death in Gstm3 KO HCC53N cells was reversed by liproxstatin-1.

“As expected, RSL3 caused more cell death in KO cells than in WT cells (Fig. 5J), and cell death was reversed by liproxstatin-1 (fig. S10B and Fig. 5K).”

• AssumptionRests on an assumption: RSL3 induces ferroptosis by inhibiting GPX4, and liproxstatin-1 inhibits ferroptosis.

GSH was reduced after ethacrynic acid treatment.

“GSH was increased after Gstm3 overexpression (fig. S12A); GSH was reduced after ethacrynic acid treatment and after Gstm3 deletion (fig. S12B), collectively indicating that this is one mechanism by which ferroptosis occurs.”

• AssumptionRests on an assumption: GSH is a critical cofactor for GPX4-mediated protection against ferroptosis.

Gstm3 deletion led to an increase in hepatocyte cell death measured by TUNEL.

“Concurrently, Gstm3 deletion led to an increase in hepatocyte cell death measured by TUNEL (fig. S9D).”

GSH was increased after Gstm3 overexpression.

“GSH was increased after Gstm3 overexpression (fig. S12A); GSH was reduced after ethacrynic acid treatment and after Gstm3 deletion (fig. S12B), collectively indicating that this is one mechanism by which ferroptosis occurs.”

• AssumptionRests on an assumption: GSH is a critical cofactor for GPX4-mediated protection against ferroptosis.

Tumors caused by DEN came primarily from zone 3.

“Indeed, we found that tumors caused by DEN came primarily from zone 3 (fig. S15A).”

• AssumptionRests on an assumption: The Gls2-CreERT2 and Cyp1a2-CreERT2 lines specifically label zone 1 and zone 3 hepatocytes, respectively.
• AssumptionRests on an assumption: Tomato expression is a reliable lineage-tracing marker that persists in labeled hepatocytes and their progeny.

Ethacrynic acid had a stronger suppressive effect on GSH than Gstm3 deletion.

“However, ethacrynic acid had a stronger suppressive effect on GSH than Gstm3 deletion (fig. S12B), suggesting that other GSTs or GSH regulators might be inhibited.”

Ethacrynic acid sensitized HCC53N cells to ferroptosis.

“Consistent with an antiferroptotic role for GSTM3, ethacrynic acid sensitized cells to ferroptosis (Fig. 5I, middle panel).”

• AssumptionRests on an assumption: RSL3 induces ferroptosis by inhibiting GPX4, and liproxstatin-1 inhibits ferroptosis.

In all cell lines, ethacrynic acid promoted RSL3-induced cell death to a greater degree in WT cells than in Gstm3 KO cells.

“In all cell lines, ethacrynic acid promoted RSL3-induced cell death to a greater degree in WT cells than in Gstm3 KO cells (Fig. 5, J and N, and fig. S11, A to D), and similarly, ethacrynic acid abrogated the protective effects of overexpression (fig. S11E).”

In sgGstm3-treated livers, superoxide and peroxide levels were increased.

“In sgGstm3-treated livers, superoxide and peroxide levels were increased, suggesting increased oxidative stress (fig. S9, B and C).”

Fsp1 overexpression eliminated the LW/BW suppressing effects of Gstm3 deletion.

“Fsp1 overexpression did not affect the CTNNB1/MYC model in the control setting but did eliminate the LW/BW suppressing effects of Gstm3 deletion (Fig. 5R).”

• AssumptionRests on an assumption: Fsp1 overexpression does not have off-target effects that affect tumorigenesis.

Mutant clones in different locations in the liver lobule have divergent fates.

“Perhaps the most unexpected result was that mutant clones in different locations in the liver lobule have divergent fates.”

Gstm3 was sufficient to protect against RSL-induced ferroptosis.

“Gain-of-function experiments showed that Gstm3 was sufficient to protect against RSL-induced ferroptosis (fig. S10, A and B, and Fig. 5, L and M).”

• AssumptionRests on an assumption: RSL3 induces ferroptosis by inhibiting GPX4, and liproxstatin-1 inhibits ferroptosis.

In livers injected with CTNNB1/MYC, Gstm3 deletion resulted in increased 4HNE.

“In livers injected with CTNNB1/MYC, Gstm3 deletion resulted in increased 4HNE (fig. S14A), supporting a role for GSTM3 in regulating lipid peroxidation in vivo.”

GSH was reduced after Gstm3 deletion.

“GSH was increased after Gstm3 overexpression (fig. S12A); GSH was reduced after ethacrynic acid treatment and after Gstm3 deletion (fig. S12B), collectively indicating that this is one mechanism by which ferroptosis occurs.”

• AssumptionRests on an assumption: GSH is a critical cofactor for GPX4-mediated protection against ferroptosis.

Mice with zone 3 mutations developed HCC more frequently than those with zone 1 mutations.

“Notably, mice with zone 3 mutations developed HCC more frequently than those with zone 1 mutations.”

Ethacrynic acid increased lipid peroxidation in cells treated with RSL3.

“Furthermore, ethacrynic acid increased lipid peroxidation in cells treated with RSL3 (fig. S9E).”

Gstm3 deletion increased survival in NRAS/shp53 mice.

“Gstm3 deletion increased survival in NRAS/shp53 mice (fig. S15D), and transient knockdown with a single dose of siRNA before HDT reduced LW/BW ratios (fig. S15E), indicating that basal Gstm3 expression in zone 3 hepatocytes is necessary for NRAS/shp53-driven HCC.”

DKO induced a stronger ferroptotic response than either single KO.

“DKO induced a stronger ferroptotic response than either single KO, indicating that Gstm2 and Gstm3 both inhibit ferroptosis (Fig. 5N).”

Fsp1 overexpression rescued tumor initiation when Gstm3 was deleted.

“As indicated by microtumor quantification, Fsp1 overexpression rescued tumor initiation when Gstm3 was deleted (Fig. 5, S and T).”

• AssumptionRests on an assumption: Fsp1 overexpression does not have off-target effects that affect tumorigenesis.

RSL3 caused more cell death in Gstm3 KO HCC53N cells than in WT cells.

“As expected, RSL3 caused more cell death in KO cells than in WT cells (Fig. 5J), and cell death was reversed by liproxstatin-1 (fig. S10B and Fig. 5K).”

• AssumptionRests on an assumption: RSL3 induces ferroptosis by inhibiting GPX4, and liproxstatin-1 inhibits ferroptosis.

Ethacrynic acid abrogated the protective effects of Gstm3 overexpression.

“In all cell lines, ethacrynic acid promoted RSL3-induced cell death to a greater degree in WT cells than in Gstm3 KO cells (Fig. 5, J and N, and fig. S11, A to D), and similarly, ethacrynic acid abrogated the protective effects of overexpression (fig. S11E).”

In HCC53N, ferroptosis induction with RSL3 killed cells.

“In HCC53N, ferroptosis induction with a small-molecule inhibitor of GPX4 (RSL3) killed cells, and the ferroptosis inhibitor liproxstatin-1 abrogated RSL3-induced cell death (Fig. 5I, left panel).”

• AssumptionRests on an assumption: RSL3 induces ferroptosis by inhibiting GPX4, and liproxstatin-1 inhibits ferroptosis.

Arid2 mutations alone were insufficient for carcinogenesis.

“In either zone, only 2 of all 58 (3.4%) Arid2 single mutant mice developed lesions by 14 months, indicating that Arid2 mutations alone were insufficient for carcinogenesis (Fig. 3C).”

In zone 1 mutant mice, Tomato reporter expression confirmed that CreER recombination was restricted to hepatocytes around portal triads.

“In zone 1 mutant mice, Tomato reporter expression confirmed that CreER recombination was restricted to hepatocytes around portal triads (PTs) (Fig. 1G).”

Co-mutating Ctnnb1 and Arid2 resulted in frequent liver tumor formation.

“As predicted by The Cancer Genome Atlas (TCGA) (Fig. 1A), co-mutating Ctnnb1 and Arid2 resulted in frequent liver tumor formation.”

In zone 1, mutant GS+ hepatocytes were more competitive than WT hepatocytes owing to increased proliferation.

“These data suggested that in zone 1, mutant GS+ hepatocytes were more competitive than WT hepatocytes owing to increased proliferation.”

The relative fitness of mutant hepatocytes before cancer development depended on the zone of origin.

“Taken together, the relative fitness of mutant hepatocytes before cancer development depended on the zone of origin.”

In zone 3 mutant mice, Tomato expression centered around central veins and extended into zone 2.

“In zone 3 mutant mice, Tomato expression centered around CVs and extended into zone 2 (Fig. 1H).”

The degree of premalignant clone expansion in different zones did not predict tumorigenic potential.

“The above results showed that (i) the degree of premalignant clone expansion in different zones did not predict tumorigenic potential and (ii) the lack of profound gene expression differences in tumors from distinct zones suggested that, once established, the tumors were phenotypically similar and may not have revealed the differences that originally influenced tumor emergence.”

Ectopic GS expression allowed lineage tracing of Ctnnb1 mutant hepatocytes.

“Thus, ectopic GS expression allowed us to lineage trace Ctnnb1 mutant hepatocytes.”

Tumors from both zone 1 and zone 3 groups stained strongly for GS, showing that these tumors were driven by oncogenic CTNNB1.

“Tumors from both groups stained strongly for GS, showing that these tumors were driven by oncogenic CTNNB1 (Fig. 3, I and J).”

WT cells from within zone 3 expanded.

“After 6 months, the Tomato+ population remained stable (fig. S5, A and B), whereas the GS+ mutant population decreased in size (Fig. 2C and fig. S5C), suggesting that WT cells from within zone 3 expanded.”

Serial staining for CTNNB1 and GS supported a 1:1 relationship between Ctnnb1 activation and ectopic GS expression.

“Serial staining for CTNNB1 and GS supported a 1:1 relationship between Ctnnb1 activation and ectopic GS expression (fig. S1, B and C).”

Gstm3 overexpression was associated with a shift in tumor origin from zone 3 to zone 1 (chi-square P = 0.000059).

“This indicated that Gstm3 overexpression was associated with a shift in tumor origin from zone 3 to zone 1 (chi-square analysis, P = 0.000059).”

Baseline levels of Gstm3 are important in initiation.

“Five days after the last injection, the ethacrynic acid–treated mice showed a decreased LW/BW, further supporting the observation that baseline levels of Gstm3 are important in initiation (Fig. 5F).”

Gstm2/3 in hepatocytes, rather than in established tumors, was responsible for differences in tumor initiation.

“The fact that Cre-driven Ctnnb1/Arid2 tumors were phenotypically and transcriptionally similar regardless of zone of origin (Fig. 3P and fig. S6A) supported the idea that Gstm2/3 in hepatocytes, rather than in established tumors, was responsible for differences in tumor initiation.”

Expression of GSTM3 in zone 1 was sufficient to shift the zone of origin from zone 3 to zone 1.

“These data indicated that the expression of GSTM3 in zone 1 was sufficient to shift the zone of origin from zone 3 to zone 1.”

• AssumptionRests on an assumption: The Gls2-CreERT2 and Cyp1a2-CreERT2 lines specifically label zone 1 and zone 3 hepatocytes, respectively.
• AssumptionRests on an assumption: Tomato expression is a reliable lineage-tracing marker that persists in labeled hepatocytes and their progeny.

Ethacrynic acid was acting through GSTM3.

“No additive reductions in gross or microtumor burden in the combination group were observed, suggesting that ethacrynic acid was acting through GSTM3 (Fig. 5H).”

Gstm2 and Gstm3 may be representatives of a family-wide effect.

“Examination of all seven murine Gstm family members showed that five had increased expression in zone 3 compared with zone 1, suggesting that Gstm2 and Gstm3 may be representatives of a family-wide effect (fig. S7F).”

These findings reveal how spatial context shapes tumor development.

“These findings reveal how spatial context shapes tumor development and highlight ferroptosis regulation as a potential target for liver cancer prevention.”

Reductions of the targeted mRNAs and proteins were observed in the normal liver as well as in isolated liver tumors.

“Reductions of the targeted mRNAs and proteins were observed in the normal liver as well as in isolated liver tumors (fig. S7, A to D).”

Zonation influences the clonal dynamics and transformation potential of premalignant cells in the liver.

“Zonation influences the clonal dynamics and transformation potential of premalignant cells in the liver.”

Lineage tracing with Tomato allowed determination of the zone of origin.

“Lineage tracing with Tomato allowed us to determine the zone of origin (Fig. 4, P and Q).”

Different levels of Gstm3 in hepatocytes before transformation could influence the initiation of HCCs.

“These results showed that different levels of Gstm3 in hepatocytes before transformation could influence the initiation of HCCs.”

In Gls2-CreER;RosaLSL-tdTomato control and Gls2-CreER;RosaLSL-tdTomato;RosaLSL-Gstm3-V5 mice, tamoxifen-induced Tomato and/or GSTM3-V5 expression was restricted to zone 1.

“In Gls2-CreER;RosaLSL-tdTomato (control) and Gls2-CreER; RosaLSL-tdTomato;RosaLSL-Gstm3-V5 mice, tamoxifen-induced Tomato and/ or GSTM3-V5 expression was restricted to zone 1 (fig. S8B).”

Increased peroxide levels suggest successful inhibition of GSTM activity.

“Ethacrynic acid led to increased peroxide levels compared with vehicle treatment, suggesting successful inhibition of GSTM activity (Fig. 5E).”

Gstm3 and Gstm2 deletion blocks tumor initiation.

“One week after the introduction of oncogenes, sgGstm3 and sgGstm2 knockout (KO) livers showed a clear decrease in the number and area of microscopic tumors, indicating a block in tumor initiation (Fig. 4, E and F).”

Gstm3 within zone 3 hepatocytes was required for tumor initiation.

“KO mice developed less CTNNB1/MYC tumors than did WT mice (Fig. 4, G and H, and fig. S7H), supporting the concept that Gstm3 within zone 3 hepatocytes was required for tumor initiation.”

Chemical inhibition of GSTMs recapitulated the HCC inhibitory effects of Gstm3 deletion.

“Therefore, chemical inhibition of GSTMs recapitulated the HCC inhibitory effects of Gstm3 deletion.”

GSTM2/3, antioxidant phase II detoxification enzymes that perform glutathione conjugation of xenobiotics, are contributors to increased zone 3 transformation in part by inhibiting ferroptosis.

“GSTM2/3, antioxidant phase II detoxification enzymes that perform glutathione conjugation of xenobiotics, are contributors to increased zone 3 transformation in part by inhibiting ferroptosis.”

As a consequence of oxidative stress, mutations may accumulate more frequently in cells within zone 3.

“As a consequence of oxidative stress, mutations may accumulate more frequently in cells within zone 3.”

The existence of antioxidant and antiferroptotic mechanisms in zone 3, including GST-related enzymes, allows these enzymes to be co-opted for tumor cell survival.

“However, the existence of antioxidant and antiferroptotic mechanisms in zone 3, including GST-related enzymes, allows these enzymes to be co-opted for tumor cell survival.”

Basal Gstm3 expression in zone 3 hepatocytes is necessary for NRAS/shp53-driven HCC.

“Gstm3 deletion increased survival in NRAS/shp53 mice (fig. S15D), and transient knockdown with a single dose of siRNA before HDT reduced LW/BW ratios (fig. S15E), indicating that basal Gstm3 expression in zone 3 hepatocytes is necessary for NRAS/shp53-driven HCC.”

• AssumptionRests on an assumption: Gstm3 is more highly expressed in zone 3 than in zone 1.

Fsp1 overexpression was confirmed on the mRNA and protein levels.

“Fsp1 overexpression was confirmed on the mRNA and protein levels (fig. S14, B and C).”

Loss of Gstm3 increases oxidative stress.

“Additionally, NRAS/shp53 livers showed decreased GSH/GSSG ratios in mice receiving sgGstm3 or sgNrf2, indicating that loss of Gstm3 increases oxidative stress (fig. S16A).”

Ethacrynic acid-induced sensitization to RSL3-mediated cell death acts through ferroptosis.

“This effect was reversed by liproxstatin-1, demonstrating that ethacrynic acid–induced sensitization to RSL3-mediated cell death acts through ferroptosis (Fig. 5I, right panel).”

• AssumptionRests on an assumption: RSL3 induces ferroptosis by inhibiting GPX4, and liproxstatin-1 inhibits ferroptosis.

Clone expansion immediately after establishment of truncal Ctnnb1 mutations is unlikely to be a feature of hepatic premalignancy.

“Clone expansion immediately after establishment of truncal Ctnnb1 mutations is unlikely to be a feature of hepatic premalignancy.”

GSTMs are not alone in their protumor functions.

“We also identified other zone 3 genes that promote tumorigenesis, suggesting that GSTMs are not alone in their protumor functions.”

This is one mechanism by which ferroptosis occurs.

“GSH was increased after Gstm3 overexpression (fig. S12A); GSH was reduced after ethacrynic acid treatment and after Gstm3 deletion (fig. S12B), collectively indicating that this is one mechanism by which ferroptosis occurs.”

• AssumptionRests on an assumption: GSH is a critical cofactor for GPX4-mediated protection against ferroptosis.

GSTM3 promotes tumorigenesis in zone 3-biased models of HCC.

“Consistent with our prediction, these data suggest that GSTM3 promotes tumorigenesis in zone 3–biased models of HCC.”

• AssumptionRests on an assumption: Gstm3 is more highly expressed in zone 3 than in zone 1.

Increased superoxide and peroxide levels in sgGstm3-treated livers suggest increased oxidative stress.

“In sgGstm3-treated livers, superoxide and peroxide levels were increased, suggesting increased oxidative stress (fig. S9, B and C).”

One mechanism by which GSTM3 promotes tumorigenesis in zone 3 is through ferroptosis inhibition in premalignant hepatocytes.

“Thus, one mechanism by which GSTM3 promotes tumorigenesis in zone 3 is through ferroptosis inhibition in premalignant hepatocytes.”

• AssumptionRests on an assumption: RSL3 induces ferroptosis by inhibiting GPX4, and liproxstatin-1 inhibits ferroptosis.
• AssumptionRests on an assumption: GSH is a critical cofactor for GPX4-mediated protection against ferroptosis.
• AssumptionRests on an assumption: Fsp1 overexpression does not have off-target effects that affect tumorigenesis.

The existence of a zonal bias was not specific to HCCs driven by Ctnnb1/Arid2 but may be a more general phenomenon.

“These data suggested that the existence of a zonal bias was not specific to HCCs driven by Ctnnb1/Arid2 but rather may be a more general phenomenon.”

• AssumptionRests on an assumption: The Gls2-CreERT2 and Cyp1a2-CreERT2 lines specifically label zone 1 and zone 3 hepatocytes, respectively.
• AssumptionRests on an assumption: Tomato expression is a reliable lineage-tracing marker that persists in labeled hepatocytes and their progeny.

GSTM3 regulates lipid peroxidation in vivo.

“In livers injected with CTNNB1/MYC, Gstm3 deletion resulted in increased 4HNE (fig. S14A), supporting a role for GSTM3 in regulating lipid peroxidation in vivo.”

Other GSTs or GSH regulators might be inhibited by ethacrynic acid.

“However, ethacrynic acid had a stronger suppressive effect on GSH than Gstm3 deletion (fig. S12B), suggesting that other GSTs or GSH regulators might be inhibited.”

The degree of clonal expansion of mutant hepatocytes did not necessarily predict their tumorigenic potential.

“Overall, our study demonstrated that the degree of clonal expansion of mutant hepatocytes did not necessarily predict their tumorigenic potential.”