Hot Zones for Liver Cancer: Metabolic Zonation, Ferroptosis, and the Origins of HCC

Abstract

Hepatocellular carcinoma (HCC) is a major cause of cancer-related death and is often resistant to treatment, partly because it develops after decades of chronic injury in a metabolically heterogeneous organ. Anatomically, the liver is organized into lobules with three concentric zones, each with distinct gene expression and metabolic programs; however, it remains unclear whether cancer favors any particular zone. In a new study, Guo and colleagues use sophisticated mouse genetics to introduce cancer-driving mutations in specific liver zones. By combining this approach with spatial transcriptomics, they track premalignant hepatocytes within the tissue microenvironment. They discover that periportal zone 1 clones expand and persist more than the rare, relatively less fit clones in pericentral zone 3. Paradoxically, however, HCC mainly arises from zone 3 hepatocytes across multiple oncogenic models, revealing a striking disconnect between clonal expansion and tumorigenic potential. A functional screen of transcripts enriched in zone 3 identifies the glutathione S-transferases Gstm2 and Gstm3 as key factors in transformation. These enzymes maintain redox balance and suppress ferroptosis in early mutant cells. Genetic deletion, hepatocyte-specific knockdown, and irreversible chemical inhibition of GSTMs all increase oxidative stress and ferroptosis, preventing tumor formation. Ectopic expression of Gstm3 in zone 1 is sufficient to reprogram these cells into a cell-of-origin compartment. These findings define the “tumorigenic zonation” of the liver and suggest that targeting ferroptosis vulnerability could be a promising therapeutic strategy for HCC.

A Heterogeneous Organ, a Heterogeneous Cancer

Hepatocellular carcinoma (HCC) remains a leading cause of cancer-related death worldwide (1). It develops after years of chronic injury, regeneration, and fibrosis, and yet when it emerges, it is aggressive, molecularly complex, and often resistant to drugs. Many hepatocytes in cirrhotic livers quietly accumulate driver mutations, but only a small fraction ever become cancerous. Why those cells, and not their neighbors?

The healthy liver provides an important clue. Hepatocytes are organized along the portal–central axis into metabolic zones (zone 1, periportal; zone 2, midlobular; zone 3, pericentral) with distinct oxygen levels, gene expression programs, and metabolic functions (Fig. 1A; refs. 2, 3). Several studies have suggested that metabolic zonation influences where damage occurs and where cancer begins (4-7). However, until recently, zonation was largely invisible to scientists. Liver zones are not separated by membranes and cannot be isolated using traditional bulk methods (8).

Figure 1.

Metabolic and tumorigenic zonation. A, Normal zonation. Hepatocytes are arranged along the portal–central axis (zones 1–3). Zone-specific metabolic functions are indicated. BD, bile duct; CV, central vein; HA, hepatic artery; PC, pericentral; PP, periportal; PV, portal vein. B, Divergent fates of premalignant clones. Ctnnb1/Arid2 mutations were introduced into zone 1 or zone 3 hepatocytes using inducible CreERT2 mouse lines. Ectopic GS expression was used to lineage-trace premalignant cells. Large, fast-expanding zone 1 clones rarely developed into HCC. In contrast, small, diminishing zone 3 clones had a higher potential to develop into HCC. C, GSTM2/3-driven ferroptosis protects zone 3 cells from ferroptosis. GSTM2/3 in zone 3 hepatocytes shields cells from lipid peroxidation and ferroptosis, thus promoting tumor initiation. Inhibition or deletion of GSTM2/3 reactivates ferroptosis and suppresses tumor initiation. Created in BioRender. Porat-Shliom, N. (2026) https://BioRender.com/phcc1ag.

In a recent Science article, Guo and colleagues (9) turn this limitation into an opportunity. Using zonal Cre/Lox drivers, single-nucleus genotyping, and spatial transcriptomics, they visualize premalignant hepatocytes in their native lobular neighborhoods and address a seemingly simple question: Do all hepatocytes have the same tumorigenic potential, or are certain metabolic zones intrinsically “hot spots” for HCC?

Mapping Premalignant Clones in Space

To mimic human premalignancy, Guo and colleagues introduced common HCC mutations into specific zones, rather than transforming the entire liver. They focused on CTNNB1 (β-catenin) and ARID2, a co-mutated gene pair in human HCC. Using conditional alleles for Ctnnb1 (exon 3 deletion stabilizes β-catenin, leading to continuous activation) and Arid2 (exon 4 deletion leading to loss of function), combined with tamoxifen-inducible CreERT2 drivers, they generated mosaic, zone-specific Ctnnb1/Arid2 double-mutant clones in an otherwise normal liver (Fig. 1B).

In wild-type (WT) livers, glutamine synthetase (GS) is restricted to a narrow ring of zone 3 hepatocytes surrounding the central vein. Because β-catenin activation induces ectopic GS expression outside this region, the authors used GS as a lineage marker for premalignant cells, enabling long-term fate mapping of mutant clones.

Next, the authors induced the mutations in a zonated manner. In zone 1, they used Gls2-CreERT2, whereas in zone 3 they employed Cyp1a2-CreERT2 with a Rosa26tdTomato reporter to mark all Cre-recombined cells (Fig. 1B). Recombination at endogenous loci is intentionally inefficient, producing somatic mosaicism that more closely mirrors human cirrhotic livers, where only a fraction of hepatocytes in each nodule carry driver mutations.

This experimental design, which combined zonal Cre drivers, endogenous alleles, and a reliable GS readout, enabled Guo and colleagues to trace premalignant cells in situ for months and study how their location influences clonal fitness and malignant potential.

Clones in Zone 3 Give Rise to HCC

After inducing mutations at 1 month of age, the team tracked GS⁺ clones over time. In zone 1 mutant livers, GS⁺ Ctnnb1/Arid2-mutant clones persisted and frequently expanded around portal tracts for up to 6 months. Single-nucleus DNA sequencing showed that double-mutant cells accounted for roughly 7% of hepatocyte nuclei at 5 months. The cell proliferation markers 5-ethynyl-2’-deoxyuridine (EdU) and Ki-67 revealed that mutant cells were more proliferative than their WT neighbors, indicating a clear fitness advantage.

In zone 3 mutant livers, the opposite occurred. The proportion of GS⁺ cells decreases over time. By 6 months, ectopic GS⁺ clones were rare, and most remaining GS staining was limited to the normal pericentral ring. Single-nucleus genotyping reveals that double mutants account for less than 1% of hepatocytes, and EdU labeling shows that proliferation predominantly occurs in WT cells, rather than mutant cells. Mutant hepatocytes are less competitive in zone 3 than in zone 1.

Spatial transcriptomics provided insights into the possible underlying mechanisms. Regardless of their origin, Ctnnb1/Arid2-mutant hepatocytes develop a zone 3–like gene expression profile, characterized by high expression of Glul and Cyp1a2. However, zone 1 mutants express higher levels of cell-cycle genes (Ccnd1, Cdk1, and Anln) than zone 3 mutants or WT hepatocytes, consistent with their increased expansion rate. Importantly, there were no clear zonal differences in expression of genes related to canonical apoptosis, tumor suppressor mechanisms, or senescence at this premalignant stage. If clone size and fitness were predictive of cancer, one would expect more tumors in zone 1. Unexpectedly, the results were the opposite.

Across different allelic combinations, HCCs primarily originate from zone 3 hepatocytes. In mice with Ctnnb1/Arid2 double mutations, about 90% of zone 3 mice, compared with around 60% of zone 1 mice, developed tumors. Blinded histology confirmed malignant lesions in 60% of zone 3 versus 12.5% of zone 1 livers. Activation of Ctnnb1, alone or with Ctnnb1/Arid2 heterozygosity, again favored zone 3. Additional HCC-inducing models, including diethylnitrosamine-induced HCC and NRAS^G12V/shp53 tumors, also demonstrated a zone 3 bias in cell of origin, indicating that this is a broad phenomenon rather than an anomaly of β-catenin–driven disease.

This work indicates that liver zonation influences both the initiation and progression of HCC. Although premalignant clones thrive in zone 1, zone 3 is the main “hot zone” for HCC initiation. Once tumors form, however, bulk RNA sequencing shows that cancers derived from zone 1 and zone 3 are transcriptionally similar, sharing WNT signatures and exhibiting only modest differences in cell-cycle and metabolic pathways. The key insight is that clonal expansion is a poor indicator of cancer risk. Instead, the local metabolic environment and stress response circuitry determine which mutated cells survive long enough to become malignant.

GSTM2/3 and a Zonated Ferroptosis Shield

What makes zone 3 so permissive to the development of HCC?

Because established tumors from different zones are transcriptionally similar, Guo and colleagues reasoned that the key factors must be embedded in the normal zonation program. Drawing on prior spatial transcriptomics data, they compiled a list of genes enriched in zone 3 hepatocytes and tested their function in a rapid CTNNB1/MYC hydrodynamic transfection model. Among several candidates, the glutathione S-transferase family members mu 2 and 3 (Gstm2 and Gstm3) emerged as important protumor factors.

GSTM2/3 proteins are highly expressed in zone 3 hepatocytes. They are induced by β-catenin and NRF2 signaling and are associated with poor survival when highly expressed in human HCC (for GSTM3). Deleting Gstm2 or Gstm3 in mice, via CRISPR/Cas9 knockout (KO) in the CTNNB1/MYC hydrodynamic overexpression model, reduced microtumor size, lowered liver-to-body weight ratios, and improved survival. Accordingly, whole-body Gstm3 KO mice develop fewer tumors than WT mice. Conversely, hepatocyte-restricted Gstm3 overexpression increases tumor burden, even though the nonintegrating adeno-associated virus (AAV) vector was largely absent from rapidly dividing tumor nodules, suggesting a role in initiation rather than progression.

Elegant gain-of-function experiments further support the hypothesis of spatial regulation in HCC initiation and progression. Expressing Gstm3 in zone 1 hepatocytes using Rosa-LSL-Gstm3-V5 allele shifts the HCC cell of origin. Lineage tracing with a Tomato reporter shows that under CTNNB1/MYC hydrodynamic transfection, most HCCs in control mice arise from unlabeled, zone 3 hepatocytes, whereas in Gstm3-overexpressing mice, the majority of tumors are Tomato⁺ and thus zone 1–derived tumors.

Mechanistically, the authors link GSTM2/3 to ferroptosis, a form of iron-dependent cell death caused by lipid peroxidation (10). Gstm3 deletion or ethacrynic acid–mediated GSTM inhibition in vivo increases reactive oxygen species, the highly reactive aldehyde 4-hydroxynonenal (4-HNE), and TUNEL⁺ cell death in hydrodynamic CTNNB1/MYC-transfected livers and significantly reduces tumor burden. In HCC cell lines and primary hepatocytes, Gstm3 KO increases sensitivity to the GPX4 inhibitor RSL3, whereas Gstm3 overexpression was protective. The ferroptosis inhibitor liproxstatin rescued this effect, further confirming that ferroptosis is the underlying mechanism for cell death. Together, zone 3–enriched Gstm2 and Gstm3 help mutated hepatocytes avoid ferroptosis, allowing them to progress to HCC, and their inhibition prevents liver cancer formation in several models (Fig. 1C).

From Metabolic to “Tumorigenic” Zonation: Implications for Classification and Therapy

Conceptually, this work reframes the relationship among chronic liver disease, clonal evolution, and HCC:

• Spatial origin matters. HCCs with similar driver genotypes can develop in different zones, but tumors in zone 3 grow in a microenvironment characterized by high oxidative stress and robust ferroptosis defenses. From a diagnostic standpoint, including zonation signatures and GSTM2/3 expression in HCC classification could help identify tumors with shared vulnerabilities even when their driver mutations differ.

• Premalignant clone burden is an unreliable indicator of risk. Large, rapidly expanding Ctnnb1/Arid2-mutant clones in zone 1 rarely become cancerous, whereas small, stressed clones in zone 3 disproportionately lead to HCC. Risk models that rely only on mutation burden or clone size fail to account for this spatial aspect.

• Zonal ferroptosis defenses are promising targets. The study demonstrates that short-term genetic or pharmacological inhibition of GSTM2/3 selectively eliminates premalignant clones and delays or prevents tumor development, indicating that harnessing ferroptosis could be an effective strategy for preventing HCC in high-risk livers.

More broadly, the work provides a blueprint for the study and characterization of other solid tumors. Combining spatially resolved genetics, lineage tracing, and functional screens can provide therapeutic insights based on where cancer begins, not just which mutations it has.