Baseline characteristics

A total of 18 patients with HCC were included in this study. Baseline characteristics showed no significant differences in age, tumor size, or other clinicopathological variables between the two groups based on the presence or absence of the VETC pattern (Supplementary Table 1).

Digital spatial profilingGlobal transcriptomic differences

Principal component analysis (PCA) demonstrated clear separation between tumor and non-neoplastic liver samples, with partial distinction between VETC-positive and VETC-negative HCC (Fig. 2a).

Fig. 2

Digital spatial profiling. a Principal component analysis (PCA) plot based on gene expression profiles. The non-neoplastic liver tissue (purple) clusters in the upper right region, clearly distinguishing itself from the tumor samples. Vessels encapsulating tumor clusters (VETC)-positive hepatocellular carcinoma (HCC) (teal) and VETC-negative HCC (yellow) partially overlap but also form distinct clusters in some areas. b Volcano plot comparing differentially expressed genes (DEGs) between VETC-positive and VETC-negative HCC. Genes significantly upregulated in VETC-positive HCC include LAMTOR2, ZNHIT1, DPP4, and MLXIPL (red dots). Angiogenesis-related genes (orange dots), epithelial–mesenchymal transition (EMT)-related genes (blue dots) and lipid metabolism-related genes (green dots) are upregulated in VETC-positive HCC. Immune-related genes (grey dots) and cholesterol metabolism-related genes (purple dots) are downregulated in VETC-positive HCC. c Heatmap of hallmark pathway enrichment scores in HCC and non-neoplastic liver tissues. In VETC-positive tumors, pathways related to mTOR, lipid metabolism, WNT/β-catenin, Hedgehog, EMT, and angiogenesis are enriched, while pathways related to inflammation are suppressed. Rows represent hallmark gene sets, and columns correspond to individual samples. The samples are categorized into VETC-positive HCC (red labels), VETC-negative HCC (blue labels), and non-neoplastic liver tissue (black labels). The color scale represents pathway enrichment scores, with red indicating upregulation and blue indicating downregulation of the respective pathways. Hierarchical clustering is applied to both samples and pathways. d Quantification of gene set variation analysis using VETC-positive HCC-specific DEGs. Enrichment of angiogenesis, WNT/β-catenin, Hedgehog, EMT, p53 signaling, and lipid metabolism pathways is observed in VETC-positive tumors, with concomitant suppression of interferon- and inflammatory signaling pathways. Red bars indicate pathways enriched in VETC-positive HCC, while blue bars represent pathways enriched in VETC-negative HCC. e Venn diagram showing consistent upregulation of four genes (LAMTOR2, ZNHIT1, DPP4, and MLXIPL) in VETC-positive HCC compared to VETC-negative HCC and non-neoplastic liver tissues. Among the upregulated genes, four genes—LAMTOR2, DPP4, ZNHIT1 and MLXIPL—were consistently increased in VETC-positive regions compared with both VETC-negative tumors and non-neoplastic liver tissue. f–i Boxplots of expression levels of key DEGs (LAMTOR2, ZNHIT1, DPP4, and MLXIPL) across VETC-positive HCC, VETC-negative HCC, and non-neoplastic liver tissues. f LAMTOR2 expression was significantly higher in VETC-positive HCC compared to both VETC-negative HCC and non-neoplastic tissue. LAMTOR2 expression was also elevated in VETC-negative HCC compared to non-neoplastic liver tissue. g ZNHIT1 expression was significantly increased in VETC-positive HCC compared to both VETC-negative HCC and non-neoplastic tissue, with no significant difference between VETC-negative HCC and non-neoplastic liver tissue. h DPP4 expression was significantly higher in VETC-positive HCC compared to VETC-negative HCC and non-neoplastic tissue, but no significant difference was observed between VETC-negative HCC and non-neoplastic liver tissue. i MLXIPL expression was increased in VETC-positive HCC compared to VETC-negative HCC, but no significant difference was found between VETC-positive and VETC-negative HCC and non-neoplastic liver tissue. Statistical significance is indicated by *p < 0.05 and **p < 0.01. "ns" denotes non-significant differences. (VETC−HCC vessels encapsulating tumor clusters-negative hepatocellular carcinoma, VETC + HCC vessels encapsulating tumor clusters-positive hepatocellular carcinoma, EMT epithelial mesenchymal transition)

Differential gene expression

Spatial transcriptomic analysis identified differentially expressed genes (DEGs) between VETC-positive and VETC-negative regions, and DEGs between VETC-positive regions and normal liver tissue (Fig. 2b and Supplementary Fig. 3a–c). In VETC-positive HCC, 39 upregulated and 235 downregulated DEGs were identified compared to VETC-negative HCC. The upregulated genes were involved in angiogenesis, epithelial–mesenchymal transition (EMT), and lipid metabolism. Conversely, the downregulated genes were associated with immune response and cholesterol metabolism.

Pathway enrichment analysis

Gene set variation analysis (GSVA) revealed consistent enrichment of angiogenesis, WNT/β-catenin, Hedgehog, EMT, p53 signaling, and lipid metabolism pathways in VETC-positive tumors, with concomitant suppression of interferon and inflammatory signaling pathways (Fig. 2c, d, Supplementary Fig. 6c).

Identification of candidate genes

Among the upregulated genes, four genes—LAMTOR2, DPP4, ZNHIT1 and MLXIPL—were consistently increased in VETC-positive regions compared with both VETC-negative tumors and non-neoplastic liver tissue (Fig. 2e–i). All four genes exceeded the limit of quantitation (LOQ) value in almost all areas of interest (Supplementary Fig. 3d). These findings suggest that these genes may serve as key molecular markers of VETC-positive HCC, warranting further investigation.

Immunohistochemical validation

Based on The Cancer Genome Atlas (TCGA) data, we observed that LAMTOR2 RNA expression was significantly elevated in tumor tissues compared to normal tissues (Supplementary Results and Supplementary Fig. 4). Consequently, we decided to evaluate LAMTOR2 protein expression through IHC staining performed on TMA slides for further validation. IHC revealed distinct expression patterns of LAMTOR2 across normal liver tissue, VETC-negative HCC, and VETC-positive HCC. In normal liver tissue, LAMTOR2 showed either negative staining or faint to weak cytoplasmic expression in hepatocytes, with no peripheral accentuation (Fig. 3a). In HCC, LAMTOR2 was expressed in the cytoplasm of tumor cells, and a subset of cases demonstrated peripheral accentuation, a pattern exclusive to HCC (Fig. 3b, c). This peripheral accentuation pattern was significantly more frequent in VETC-positive HCC (n = 5/7, 71%) compared to VETC-negative HCC (n = 3/11, 27%) or non-neoplastic liver tissue (n = 0/24, p < 0.001). In addition, moderate intensity of peripheral accentuation was observed at a significantly higher rate in VETC-positive HCC (n = 3/7, 43%) compared to VETC-negative HCC (n = 2/11, 18%) or non-neoplastic liver tissue (n = 0/24, p < 0.001, Fig. 3d).

Fig. 3

Immunohistochemical analysis of LAMTOR2 expression. a–c Representative immunohistochemistry (IHC) images of LAMTOR2 expression (200x, scale bars: 100 µm). a A non-neoplastic liver tissue showing focal faint or no cytoplasmic expression in hepatocytes. b Vessels encapsulating tumor clusters (VETC)-negative hepatocellular carcinoma (HCC) displaying moderate cytoplasmic expression in tumor cells without cell peripheral accentuation. c VETC-positive HCC showing a peripheral accentuation pattern of expression with moderate intensity in tumor cells. d, e Analysis of LAMTOR2 IHC expression and correlation with RNA levels in the study cohort (n = 42). d Distribution of peripheral accentuation intensity levels in non-neoplastic liver, VETC-negative HCC, and VETC-positive HCC. The frequency of peripheral accentuation pattern was higher in VETC-positive HCC compared to VETC-negative HCC, and moderate intensity of peripheral accentuation was more common in VETC-positive HCC as well. No peripheral accentuation pattern was observed in non-neoplastic liver. Chi-square test showed a significant p value between the three groups. e A linear correlation between LAMTOR2 peripheral accentuation intensity and RNA expression levels, indicating that higher peripheral staining is associated with increased RNA expression, with a significant p value observed in the Kruskal–Wallis test. f–h Validation of LAMTOR2 IHC findings in the independent tumor microarray (TMA)-based HCC cohort (n = 276). f Comparison of LAMTOR2 cytoplasmic expression and peripheral accentuation intensity between tumor cores (n = 579) and normal cores (n = 289). Both cytoplasmic expression and peripheral accentuation pattern were observed at higher frequencies in tumor tissue compared to normal tissue. The peripheral accentuation pattern was exclusive to tumor tissue. g Comparison of LAMTOR2 peripheral accentuation intensity between VETC-positive HCC (n = 66) and VETC-negative HCC (n = 210). The frequency of peripheral accentuation pattern was higher in VETC-positive HCC compared to VETC-negative HCC, and moderate intensity was also more common in VETC-positive HCC. h Clinicopathological findings, including Edmondson–Steiner (E–S) grade, vascular invasion, VETC status and tumor relapse, according to the presence of LAMTOR2 peripheral accentuation pattern. Statistical significance is indicated as *p < 0.05 and **p < 0.01. (VETC−HCC vessels encapsulating tumor clusters-negative hepatocellular carcinoma, VETC + HCC, vessels encapsulating tumor clusters-positive hepatocellular carcinoma, E–S Edmondson–Steiner grade, MVI microvascular invasion, VETC vessels encapsulating tumor clusters)

A strong statistical correlation was observed between peripheral accentuation pattern of LAMTOR2 and its RNA expression levels (median (IQR) RNA expression with and without peripheral accentuation; 10.4 (9.6–12.0) vs. 5.0 (4.1–6.1); p < 0.001). Furthermore, a linear relationship was identified between the intensity of peripheral accentuation and RNA expression levels (negative: 5.0 (4.1–6.1), weak: 9.9 (7.8–9.9), moderate: 11.3 (10.9–13.9), p = 0.002, Fig. 3e). These findings demonstrate that LAMTOR2 protein localization and intensity, particularly peripheral accentuation, are closely linked to its RNA expression.

LAMTOR2 protein expression in a validation cohort

LAMTOR2 IHC staining was performed on the IHC validation cohort of 276 HCC cases. For each case, 1–4 tumor cores and 1–2 normal cores were included, resulting in 579 tumor cores and 289 normal cores analyzed. There was significant difference in LAMTOR2 expression between tumor and normal tissues. Among tumor cores, 16% (90/579) exhibited moderate cytoplasmic expression, 74% (429/579) weak expression, and 10% (60/579) negative expression, compared to 3% (10/289), 83% (240/289), and 13% (39/289) in normal cores, respectively (p < 0.001, Fig. 3f). Notably, peripheral accentuation of LAMTOR2, a feature exclusive to tumor tissue, was observed in 24% (136/579) of cases. Specifically, 2% (11/579) of tumor cores showed moderate expression, and 22% (125/579) weak expression, whereas no peripheral accentuation was detected in normal tissue cores (p < 0.001).

Of the 276 cases analyzed, 65 cases (24%) exhibited LAMTOR2 peripheral accentuation. HCCs with peripheral accentuation demonstrated a significant association with VETC positivity, compared to cases without peripheral accentuation (24/65, 37% vs. 42/211, 20%; p = 0.008, Fig. 3g, h and Supplementary Table 2). Furthermore, these cases exhibited a marginally higher proportion of Edmondson–Steiner grade III/IV tumors (55/65, 83% vs. 152/211, 72%; p = 0.060) and a significantly higher recurrence rate (49/65, 75% vs. 120/211, 57%; p = 0.011). No significant differences were observed in the frequency of vascular invasion or the macrotrabecular massive (MTM) subtype between cases with and without peripheral accentuation. Although microvascular invasion (MVI) and MTM subtype have been previously reported to associate with the VETC pattern, and were also more frequent in VETC-positive tumors in our cohort (MVI: 39/66, 59% vs. 77/210, 37%; p = 0.002; MTM: 12/66, 20% vs. 11/210, 10%; p = 0.092), their distribution did not significantly differ according to LAMTOR2 expression. These findings highlight the unique expression patterns of LAMTOR2 in HCC, particularly its peripheral localization, which is strongly associated with VETC positivity and aggressive tumor features.

Single-cell transcriptomic analysis

To further substantiate these observations, scRNA-seq was performed on the scRNA-seq validation cohort of five biopsied HCC cases, including one VETC-positive case. Basic information is presented in Supplementary Table 3, and detailed results are provided in supplementary results section. Unsupervised clustering identified seven major cell clusters (Fig. 4a, Supplementary Figs. 5a, d, 6g), among which LAMTOR2 expression was markedly elevated in HCC cells compared to other cell types (p < 0.0001, Fig. 4b, Supplementary Fig. 5b, c).

Fig. 4

Single-cell RNA sequencing. a Uniform Manifold Approximation and Projection (UMAP) visualization of single-cell RNA-sequencing data from five hepatocellular carcinoma (HCC) patients, showing clustering of seven major cell types. b LAMTOR2 expression is significantly higher in copyKAT-defined aneuploid HCC cells compared to diploid non-neoplastic hepatocytes (**** indicates p < 0.0001). c Violin plot of LAMTOR2 expression in tumor cells stratified by VETC score-based grouping (VETC-high = top 33% vs VETC-low = bottom 67%; n = 12,247 tumor cells from five HCC patients, including 362 CopyKAT-confirmed malignant cells from HCC 02 [VETC-positive]). VETC-high tumor cells exhibited significantly elevated LAMTOR2 expression compared with VETC-low cells (**** indicates p < 0.0001). d Violin plot showing that VETC module scores are significantly higher in LAMTOR2-high tumor cells compared with LAMTOR2-low tumor cells (**** indicates p < 0.0001), supporting the reciprocal direction of the VETC–LAMTOR2 association observed in (c). e Gene ontology (GO, left) and Kyoto Encyclopedia of Genes and Genomes (KEGG, right) pathway enrichment plots of differentially expressed genes from VETC-high, LAMTOR2-high tumor cells exhibit significant upregulation of pathways including fatty acid degradation, lipid catabolic process, metabolism of xenobiotics by cytochrome P450, and drug metabolism–cytochrome P450. (UMAP Uniform Manifold Approximation and Projection, VETC + HCC vessels encapsulating tumor clusters-positive hepatocellular carcinoma, VETC vessels encapsulating tumor clusters)

To further dissect the molecular phenotype of VETC-positive HCC, we applied a scoring approach based on DEGs identified from spatial transcriptomic analysis of VETC-positive versus VETC-negative HCC regions (log₂ fold change > 0.75, Supplementary Fig. 6a, b). Uniform Manifold Approximation and Projection (UMAP) embedding of HCC cells followed by VETC signature scoring revealed clear stratification of tumor cells into VETC-high and VETC-low subpopulations (Supplementary Fig. 5e), which recapitulated gene expression patterns observed in the spatial transcriptomics data (Supplementary Fig. 5f–h).

LAMTOR2 expression was significantly higher in the VETC-high HCC group compared to the VETC-low group (p < 0.0001, Fig. 4c), further supporting its association with this vascular phenotype. Additionally, cells with high LAMTOR2 expression exhibited significantly elevated VETC scores compared to the LAMTOR2-low group (p < 0.0001, Figs. 4d, Supplementary Fig. 6d–f), further confirming this association.

Subsequently, we isolated VETC-high HCC cells with high LAMTOR2 expression and performed differential gene expression and pathway enrichment analysis. This revealed significant upregulation of pathways associated with lipid degradation and detoxification processes, including fatty acid degradation, lipid catabolic process, metabolism of xenobiotics by cytochrome P450, and drug metabolism–cytochrome P450 (Fig. 4e).