Spatial Transcriptomics Reveals Distinct Architectures but Shared Vulnerabilities in Primary and Metastatic Liver Tumors

Abstract

Background: Primary hepatocellular carcinoma (HCC) and liver metastases differ in origin, progression, and therapeutic response, yet a direct high-resolution spatial comparison of their tumor microenvironments (TMEs) within the liver has not previously been performed. Methods: We applied high-definition spatial transcriptomics to fresh-frozen specimens of one HCC and one liver metastasis (>16,000 genes per sample, >97% mapping rates) as a proof-of-principle two-specimen study, cross-validated in human proteomics and patients' survival datasets. Transcriptional clustering revealed spatially distinct compartments, rare cell states, and pathway alterations, which were further compared against an independent systemic dataset. Results: HCC displayed an ordered lineage architecture, with transformed hepatocyte-like tumor cells broadly dispersed across the tissue and more differentiated hepatocyte-derived cells restricted to localized zones. By contrast, liver metastases showed two sharply compartmentalized domains: an invasion zone, where proliferative stem-like tumor cells occupied TAM-rich boundaries adjacent to hypoxia-adapted tumor-core cells, and a plasticity zone, which formed a heterogeneous niche of cancer-testis antigen-positive germline-like cells. Across both tumor types, we detected a conserved metabolic program of "porphyrin overdrive," defined by reduced cytochrome P450 expression, enhanced oxidative phosphorylation gene expression, and upregulation of FLVCR1 and ALOX5, reflecting coordinated rewiring of heme and lipid metabolism. Conclusions: In this pilot study, HCC and liver metastases demonstrated fundamentally different spatial architectures, with metastases uniquely harboring a germline/neural-like plasticity hub. Despite these organizational contrasts, both tumor types converged on a shared program of metabolic rewiring, highlighting potential therapeutic targets that link local tumor niches to systemic host-tumor interactions.

Keywords: cancer metabolism; heme metabolism; plasticity; porphyrin metabolism; prostaglandin; tumor microenvironment.

Figure 1

HCC Visium HD data with tissue overlay at single-cell resolution. (A) Cell types are color-coded according to the legend and shown with matching percentage distributions in the accompanying pie chart. The largest group, comprising approximately one-third of all cells, consists of tumor hepatocytes; (B) Heatmap of the top 50 marker genes for each cell group, with normalized expression values used as the scale; (C) Tumor hepatocytes display a dispersed, clump-like spatial pattern, whereas more differentiated hepatocellular carcinoma cells, likely derived from pericentral hepatocytes, are localized to a single anatomical region; (D) A cluster of epithelial cells of non-hepatocyte origin is positioned adjacent to tumor hepatocytes and exhibits elevated expression of stress-response genes, including multiple heat shock proteins. (E) Schematic summarizing the core spatial patterns observed in HCC versus liver metastasis. Primary HCC displays more differentiated tumor cells that cluster together and remain localized within defined regions, while transformed populations are more dispersed. The liver metastasis exhibits an epithelial-derived invasion zone, reflecting foreign lineage infiltration.

Figure 2

Liver metastasis Visium HD spatial transcriptomic data overlaid with tissue morphology. (A) Two distinct spatial zones are identified within the tumor. Zone Invasion represents the liver metastasis invasive front, where colorectal cancer cells are bordered by tumor-associated macrophages (TAMs) along the tumor edge. Zone Plasticity in metastasis contains germline-like, dedifferentiated cells composed of multiple highly plastic cell types, each shown in a different color; (B) The tumor compartment comprises rapidly proliferating, stem-like cells at the tumor periphery and a more differentiated tumor core with necrotic features; (C) Residual hepatocytes within the tumor invasion zone are interspersed among invading tumor cells, whereas more differentiated hepatocytes form small clumps resembling mature hepatocytes; (D) Two TAM subtypes are distinguished: one at the Zone Invasion boundary, forming a fringe around invading tumor cells and marking the interface between invading colonic epithelial cells and surrounding tissue; and an M2-like TAM population in Zone Plasticity, intermingled with germline-like and highly plastic cells; (E) Prostaglandin signaling in the liver metastasis. Expression of the inflammatory lipid-producing enzyme encoding gene ALOX5 originates primarily from stromal rather than tumor epithelial cells. Both M2-like TAMs and plastic TAMs in Zone Plasticity, as well as pEMT cells, exhibit elevated ALOX5 expression.

Figure 3

Heme and porphyrin metabolism gene expression in HCC and liver metastasis. Log₂ fold changes are shown for comparisons between more aggressive and more differentiated tumor cell populations. * indicate significant differential expression with adjusted p value < 0.05. (A) HCC: Comparison of more transformed versus more differentiated hepatocyte-derived tumor cells. ALAS1, ALAD, HMBS, UROS, UROD, CPOX, PPOX, and FECH are heme biosynthesis genes. Cytochrome P450 enzymes are indicated by CYP genes, and electron transport chain (ETC) genes are denoted by COX genes; (B) Liver metastasis: Comparison of tumor stem-like outer zone cells versus tumor core cells. HMOX1 and HMOX2 are heme degradation genes; FLVCR1 and FLVCR2 encode heme/porphyrin exporters. ETC-related genes are marked, including Complex I (NDUF genes), Complex III (UQCR genes), and Complex IV (COX genes); (C) Schematic: Model of aberrant heme metabolism in liver tumors. Heme biosynthesis gene expression is reduced, while protoporphyrin IX (PPIX) accumulation is enhanced, as inferred from an imbalance in heme production (ALAS1 downregulated, HMBS upregulated). Increased FLVCR1 expression suggests elevated heme/porphyrin export. Cytochrome genes are decreased, in contrast to electron transport chain (ETC)-related genes, which are upregulated in more aggressive cancer cell populations.

Figure 4

Immune cell aggregates in HCC and liver metastasis. (A) In HCC, immune cell populations appear as dispersed clusters along boundaries between different cell type regions, particularly at interfaces with biliary-like cell groups. This cell population harbors B-lineage cells expressing IGHM, IGHA1, and CD79A. Adjacent biliary-like cells express markers such as SOX9 and HAMP; (B) In liver metastasis, immune cells form scattered clusters throughout the tissue and are frequently adjacent to myofibroblastic cancer-associated fibroblast (myCAF) populations. These immune clusters show elevated expression of IL6 and CXCL12; (C) Two distinct CAF populations are identified in liver metastasis. These CAF types form separate clusters in UMAP space, independent of their physical spatial location. CAFs are predominantly located in the metastatic invasion zone, whereas myCAFs are enriched in the metastatic germline-like dedifferentiation zone. myCAFs express MYL9 in addition to classic fibroblast genes.

Figure 5

Increased cellular complexity in liver metastasis compared to HCC, driven by highly plastic cells in Zone Plasticity (germline-like dedifferentiation zone). (A) UMAP of HCC cell types, generated from gene expression profiles without incorporating spatial coordinates, shows that the majority of cells are differentiated; (B) UMAP of liver metastasis reveals markedly more complex cellular lineages; (C) UMAP of Zone Plasticity in liver metastasis (germline-like dedifferentiation zone) shows cell types lacking differentiation features and displaying high plasticity. These include germline-like cells, plastic ALOX5-positive cells, M2-like TAMs, Neuro GABA cells, myCAFs, and pEMT cells. Germline-like cells express genes such as HOXC9 and ZNF genes; (D) UMAP of individual cell types within Zone Plasticity of metastasis, including germline-like cells, Neuro GABA cells, M2-like TAMs, and plastic ALOX5-positive cells, show overlapping or adjacent positions, indicating intrinsic relatedness in their lineage or state transitions.

Figure 6

Protein expression and prognostic significance of FLVCR1 and ALOX5 in human liver tumors. (A) Protein expression data from 99 normal and 100 tumor liver samples (HMP dataset). (B) Protein expression data from 165 normal and 165 tumor liver samples (HMP dataset). FLVCR1 protein levels are significantly different between tumors and normal tissues (nonparametric Kruskal–Wallis test, p < 0.01). (C) Kaplan–Meier analysis of FLVCR1 overall survival, demonstrating its value as a validated prognostic marker. Results were confirmed in an independent validation set. (D) Kaplan–Meier analysis of ALOX5 overall survival, showing that ALOX5 expression is a potential prognostic marker. KM curves were generated from TCGA liver cancer datasets.

Figure 7

Tumor-bearing mice exhibit distant liver metabolic reprogramming in heme and prostaglandin pathways. UMAP of liver cell populations from two normal mice (upper cluster) and two tumor-bearing mice (lower cluster), reanalyzed from Vandenbon et al. In both groups, Cyp2a1 expression was maintained (fold change < 2), reflecting conserved hepatic metabolic activity. In contrast, livers from tumor-bearing mice exhibited significant upregulation of Mki67 (adjusted p < 0.05), indicating enhanced cell cycle activity. Moreover, the heme/porphyrin export gene Flvcr1 (adjusted p < 0.05) and the prostaglandin pathway enzyme Alox5 (adjusted p < 0.05) were elevated in tumor-bearing livers, consistent with metabolic reprogramming. These changes suggest that tumors can trigger early (“pioneer”) signals of heme and lipid pathway activation in distantly affected organs.