Dissecting the Transcriptomic Profile of CD163⁺ Macrophages in Hepatocellular Carcinoma with VETC⁺ Angiogenesis Provides Potential Targets of Treatment

Abstract

Introduction

Hepatocellular carcinoma (HCC) is characterized by a highly vascularized tumor microenvironment (TME), with VETC being a distinctive, frequent, and prognostically unfavorable type of this vascular TME. VETC⁺ HCCs show lymphocyte deprivation and enrichment with large, foamy macrophages distributed close to endothelial cells.

Methods

A total of 7 resected HCCs were retrospectively analyzed according to VETC status. Spatial transcriptomic profiling was performed on formalin-fixed, paraffin-embedded samples using NanoString GeoMx™ DSP, followed by differential expression and pathway analyses. SPP1 expression was further evaluated by immunohistochemistry and multiplex immunofluorescence.

Results

This study aimed to characterize these macrophages with a spatial transcriptomic approach. We demonstrated that they exhibit a pro-tumor, M2-like tumor-associated macrophages (TAMs) signature (SPP1, ACP5, GPNMB, and FABP5), distinct from those in VETC⁻ cases. They are enriched with specific pathways involved in immunosuppression and angiogenesis. SPP1, ADAM9, and MMP9 were among the genes mostly upregulated. Immunohistochemistry analysis confirmed strong SPP1 expression in TAMs spatially associated with endothelial cells in VETC⁺ cases.

Conclusion

Our findings suggest that M2-like TAMs in VETC⁺ HCCs contribute to both angiogenesis and immunosuppression. SPP1 and ADAM9 could represent novel therapeutic strategies to reshape TAMs and suppress VETC onset.

Introduction

A progressive enrichment of the vascular tumor microenvironment (TME) characterizes hepatocarcinogenesis, and hepatocellular carcinoma (HCC) typically presents as a hypervascular lesion [1]. Vessel that encapsulate tumor cluster (VETC) is a peculiar form of this vascular TME, in which CD34⁺ endothelial cells (EC) enwrap HCC cells, consistently associated with poor outcome [2]. VETC has been reported in approximately 40% of resectable HCC, 50% of post resection recurrence and up to 60% transplanted HCC, invariably conferring an adverse prognostic impact [1–3]. Moreover, VETC exerts a predictive value both for locoregional [4, 5] and systemic anti-angiogenic therapies [6, 7].

VETC⁺ HCCs are characterized by a low degree of immune infiltration [8], particularly by T-lymphocyte deprivation and reduced CD8⁺ T-cell infiltration [9]. Even when present, the immune infiltrate appears inefficient, with reduced TCR signal transduction and IFN-γ signaling [10], and lack of tumoral PD-L1 expression [11]. Consistent with these findings, VETC⁺ HCCs are enriched for activation of the WNT/β-catenin pathway [9, 11], which has been strongly associated with immune exclusion in HCC [12, 13]. On the other hand, WNT/β-catenin-activated HCC can exhibit focal VETC⁺ areas accompanied by abundant immune infiltration [14]. Collectively, these observations suggest an interplay between the immune TME and the development of the VETC pattern. This hypothesis is further supported by evidence regarding angiogenic factors. VETC formation is associated with high expression level of ANGPT2, VEGFA, and FGF2 [8, 9]. However, increased expression of these angiogenic mediators does not result in VETC formation in the presence of significant lymphocyte infiltration and an inflammatory/angiostatic milieu [9]. It is, therefore, possible that antitumor immunity, beyond exerting direct cytotoxic effects on tumor cells, may indirectly suppress VETC formation [9]. In keeping with this hypothesis, T helper 1 (Th1)/cytotoxic T lymphocyte-type inflammatory cytokines and chemokines, produced during effective antitumor immune response, function as potent angiostatic factors, inhibiting tumor angiogenesis and promoting vascular normalization [15, 16].

The complex interplay between vascular and immune components of the TME in HCC becomes even more complex when considering macrophages (MΦ) [17, 18]. Tumor-associated macrophages (TAMs) are broadly categorized into two functionally distinct phenotypes: classically activated (M1-like) and alternatively activated (M2-like) MΦ [19]. M1-like TAMs generally exert antitumor effects, whereas M2-like TAMs suppress T-cell-mediated immune responses, promote tumor angiogenesis, and facilitate tumor progression and metastasis [20]. In HCC, M2-like TAMs deplete activated CD8⁺ T cells, reduce the secretion of inflammatory mediators such as IFN-γ and granzyme B in T cells, and activate the immune escape and immunosuppression [21]. Conversely, M1-like TAMs recruit B cells via the CXCL12/CXCR4 signaling axis and enhance cytotoxic T cell through the expression of PD-L1 regulated by CXCR4 [22]. With respect to vascular interactions, M2-like TAMs secrete VEGF and multiple pro-angiogenic factors that drive angiogenesis in HCC [23, 24], whereas VEGF depletion favors MΦ polarization toward the M1 phenotype [25]. In addition, angiogenetic factors such as COX-2 and Tie-2 can modulate TAM recruitment and infiltration in HCC [26, 27]. The intrinsic plasticity of MΦ allows them to switch polarization states in response to changes in the TME or therapeutic interventions [19].

We recently reported that VETC⁺ HCCs are enriched in a distinctive morphological subset of MΦ presenting as large, irregular in profile and with a foamy cytoplasm [28]. Interestingly, these morphological features are in keeping with that of TAMs observed in liver metastases from colorectal cancer [29]. Moreover, these MΦ are spatially close to ECs in VETC⁺ HCC but not in HCC with a VETC⁻ pattern. Herein, using a spatial transcriptomic approach, we unveiled the profile of MΦ in VETC⁺ tumors. We demonstrate that these cells exhibit an M2-like phenotype with a distinct transcriptional signature (SPP1, GPNMB, TREM2, FABP5, MMP19, MGLL, ACP5, LPL, FABP4) and upregulation of pathways involved in HCC angiogenesis and immunosuppression.

Materials and Methods

Case Selection

We retrospectively collected seven cases of HCC among patients underwent liver surgery at IRCCS Humanitas Clinical and Research Hospital between 2023 and 2024. For each patient, we evaluated the histotype and grade according to WHO (5th edition) and VETC according to Renne et al. [2]. We then recorded the etiology of HCC, and we observed it was heterogeneous both in VETC⁺ and VETC⁻ cases. Among VETC⁺ cases, one case exhibited a metabolic etiology, one a viral etiology (HBV/HCV), and one a combined metabolic/exotoxic etiology (for the last VETC⁺ case, clinical data were not available). Among VETC⁻ cases, one case showed a viral etiology (HCV) and two cases a metabolic etiology. Then, we selected a single formalin-fixed, paraffin-embedded block per case.

NanoString Digital Spatial Profiling (GeoMx™)

GeoMx™ profiling was performed on all the 7 patients affected by resectable HCC. A tissue from patients were cut, placed on a single SuperFrost Plus slide, and deparaffinized by baking the sections for 15 min at 60°C followed by three 5 min xylene washes, two 5 min 100% EtOH, one 5 min 95% EtOH, and one 1 min wash in 1X PBS wash. Antigen retrieval was performed by incubating the sections for 20 min at ∼99°C in 1X Tris-EDTA (pH 9.0) and then incubating the sections in 95°C EDTA/Tris buffer for 20 min. Sections were then incubated for 15 min at 37°C with 0.1 μg/ml proteinase K in 1X PBS and then washed with 1X PB. This step was followed by one 5-min wash with 10% neutral buffered formalin, two 5 min washes with NBF stop buffer, and one wash with 1X PBS. Finally, the section slides were hybridized overnight with the GeoMx Human Whole Transcriptome Atlas (WTA) oligonucleotide probe mix (NanoString Technologies, Inc). The tissue was then double-labeled with an anti-CD163 antibody conjugated to Alex 594 (1:50 dilution Abcam 282,114) and Syto13 (1:10 dilution NanoString GeoMx Nuclear Stain Morphology Kit) according to the Nanostring Manual Slide Preparation guide (MAN-10150-02).

A total of 44 regions of interest (ROIs, 660 μm × 785 μm) were profiled: 28 intratumoral (14 VETC⁺/14 VETC⁻) and 16 peritumoral (7 VETC⁺/9 VETC⁻), with a median of 10 ROIs selected from each slide. Each ROI was then illuminated with a UV light so the oligonucleotides probe barcodes present within each tissue were photocleaved and collected individually into a single well of a 96-well collection plate. NanoString GeoMx Digital Spatial Profiling (DSP) libraries were sequenced using the Illumina Nextseq2000 platform. Raw BCL files were converted to FASTQ format using BCLconvert (version 3.8.2). For each area of illumination, digital count conversion (.dcc) files were created from the FASTQ files using the GeoMx NGS Pipeline (version 3.1.1). DCC files from each tissue sample were imported into a NanoStringGeoMxSet object via the readNanoStringGeoMxSet function from the GeomxTools R package (v3.5) in R (v4.3.3) [30].

Sequencing quality was assessed for each segment, and segments with a high proportion of genes below the limit of quantification were excluded, resulting in a total of 43 retained segments. For downstream analysis, we focused on the 9,075 genes detected in at least 5% of the segments. Following filtering, gene expression data were normalized using Q3 normalization. Differential expression analysis was conducted between MΦ VETC⁺ HCC and VETC⁻ HCC within both intratumoral and peritumoral regions using a linear mixed-effects model without a random slope, accounting for tissue subsampling by including slide identity as a fixed effect. Genes were considered significantly up- or downregulated based on a false discovery rate threshold of 0.1 and an absolute log2 fold change greater than 0.5. To interpret differential expression results in a pathway context, pre-ranked gene set enrichment analysis was performed using the fgsea R package (v1.28) [31], with curated gene sets from MSigDB (Hallmark, Reactome, and WikiPathways). Enrichment of TAM signatures in intratumoral ROIs was assessed using the GSVA R package (v1.50) [32] on log-normalized expression values, with parameters mx.diff = TRUE and kcdf = “Gaussian.” Clustering of ROIs based on GSVA scores was performed using Euclidean distance and Ward’s D2 linkage method.

SPP1 Immunohistochemical Evaluation

We performed SPP1 immunohistochemistry on a Tissue Micro Array (TMA) including a retrospective series of 210 HCC (VETC⁺ = 41; VETC⁻ = 169); each TMA block was cut to obtain 2u sections for immunostaining. The SPP1 primary monoclonal antibody (SPP1 Osteopontin Antibody, 22952-1-AP, Proteintech) was used at 1:250 dilution. Two liver pathologists (LDT; CDC) evaluated all cases, blind to VETC status, and reviewed discordant results to reach a consensus. In addition, one section was also stained with H/E. The results were analyzed with chi-squared test statistical analysis.

Sequential Immunofluorescence Analysis

FFPE tissue samples were cut into 3-µm sections and subjected to heat-induced antigen retrieval using DIVA Antigen Decloaker solution (Biocare Medical, USA). Immunostaining was performed at room temperature, following the manufacturers’ instructions, with mouse monoclonal antibodies against CD34 (1:200; Biocare Medical), CD163 (1:200; Novocastra, Italy), and osteopontin (1:300; Santa Cruz, USA), each diluted in antibody diluent/blocking buffer (Akoya Biosciences, USA). Opal fluorophore reagents were prepared in Plus Amplification Diluent buffer (Akoya Biosciences), using Opal 520 for CD34, Opal 620 for CD163, and Opal 690 for osteopontin. Sections were counterstained with DAPI (Akoya Biosciences) and mounted with Fluorescence Mounting Medium (Dako, Italy). Whole-slide images were acquired at 20× objective magnification with a Zeiss Axioscan Z1 automated slide scanner (Zeiss, Italy).

Results

MΦ Infiltrating VETC⁺ HCC Display a Pro-Tumor (M2-Like) TAMs Signature

We first compared the histological features of MΦ in intratumoral tissue obtained from VETC⁺ with that of MΦ from VETC⁻ HCC (Fig. 1a). Consistent with our previous morphological observations [28], MΦ infiltrating VETC⁺ HCC were more abundant and exhibited a distinctive spatial distribution along ECs (Fig. 1a, (A, C)). In contrast, CD163⁺ TAM in VETC⁻ cases were less frequent and randomly distributed within tumoral tissue (Fig. 1a (d, f)). Building upon these morphological findings, we investigated transcriptomic differences between MΦ from VETC⁺ and VETC⁻ HCC using Nanostring spatial technique (Fig. 1b). Differential gene expression analysis yielded 89 upregulated and 97 downregulated genes, confirming a distinct profile of MΦ according to the VETC status (Fig. 1c, d). Few differences were also observed in the peritumoral liver parenchyma (online suppl. Fig. 1; for all online suppl. material, see https://doi.org/10.1159/000552648).

Fig. 1.

MΦ infiltrating VETC⁺ HCCs are characterized by a peculiar morphology and specific gene signature. a Morphology of CD163⁺ TAMs and relationship with VETC. A, C VETC⁺ HCCs are characterized by clusters of malignant cells enveloped by a continuous layer of CD34⁺ ECs (A, ×20), abundant CD163⁺ TAMs (B, ×20) which are close to ECs (C, ×40). D, F In VETC⁻ cases, CD34⁺ ECs retain the regular distribution of sinusoids (D, 20×), CD163⁺ TAM are less frequent (E, ×20) and randomly distributed in the tumoral tissue (F, ×40). b Representative NanoString GeoMx spatial image from an HCC specimen, showing the selection of peritumor and intratumor ROIs. Enlarged views of representative peritumor and intratumor ROIs are shown on the right. Fluorescence channels are indicated as follows: Syto13 (nuclear stain), blue; CD163 (macrophages), yellow. Green dots indicate CD163-positive cells identified during ROI segmentation. c Heatmap depicting the transcriptome of TAMs in intratumoral ROIs according to the VETC. d Volcano plot showing differential gene expression analysis according the VETC phenotype, yielding 89 upregulated (red dots) and 97 downregulated (green dots) gene.

To determine whether MΦ in VETC⁺ HCC resemble pro-tumor TAM populations previously described in primary or metastatic liver tumors, we performed gene set variation analysis (GSVA) using the GSVA R package [32]. We applied a TAM signature comprising GPNMB, TREM2, SPP1, FABP5, MMP19, MGLL, ACP5, LPL, FABP4 genes previously associated with pro-tumor MΦ in colorectal liver metastases [33] and HCC [10, 34]. The GSVA enrichment score for this TAM signature was significantly higher in MΦ from VETC⁺ than in MΦ from VETC⁻ tumors, both in intratumoral (p = 0.031) and peritumoral (p = 0.0093) regions. No significant differences were observed between intratumoral and peritumoral TAM within the same VETC category (Fig. 2a). This result indicates that MΦ in VETC⁺ HCC exhibit a pro-tumor TAM phenotype. At the single-gene level, intratumoral VETC⁺ TAMs significantly overexpressed SPP1 (p = 0.0091), ACP5 (p = 0.012), and GPNMB (p = 0.0067) compared with MΦ from VETC⁻ tumors (Fig. 2b).

Fig. 2.

MΦ from VETC⁺ HCCs have a profile of pro-tumor TAMs and are characterized by specific pro-angiogenic pathways. a Boxplots describing GSVA enrichment score of TAMs in VETC⁺ (red) and VETC⁻ (green) HCC, for intratumoral and peritumoral areas. b Violin plots for single gene belonging to the TAM signature in VETC⁺ (red) and VETC⁻ (green) HCC, both for intratumoral and peritumoral areas. c GSEA barplot of significantly enriched angiogenetic pathways. d Violin plots for TAMs-genes included in the most upregulated angiogenetic pathway (reactome signaling by VEGF), in VETC⁺ (red) and VETC⁻ (green) HCC.

VETC⁺ M2-like TAMs Are Characterized by Specific Pro-Angiogenic Pathways

To further elucidate the functional role of TAMs in VETC⁺ HCC, we explored the functional pathways differentially enriched by gene set enrichment analysis. A total of 480 pathways were significantly enriched (p adjusted <0.05). The 50 most enriched pathways, ranked by normalized enrichment score, are shown in online supplementary Figure 2a. Given the distinctive vascular nature of VETC, we specifically examined angiogenesis-related pathways. Twelve angiogenic pathways remained significantly enriched after normalized enrichment score normalization (Fig. 2c). Among these, the most upregulated pathway was reactome signaling by VEGF. Within this pathway, SOD2, ADAM9, RPL27, and PTMA were significantly more expressed in VETC⁺ TAMs (Fig. 2d). When focusing on the three most upregulated vascular pathways (reactome signaling by VEGF; WP VEGFAVEGFR2 signaling pathway; WP angiogenesis) three genes, MMP9, PLIN, SOD2, were consistently upregulated in intratumoral TAMs from VETC⁺ HCC (online suppl. Fig. 2b). These data support a pro-angiogenic functional polarization of TAMs within VETC⁺ tumors.

VETC⁺ HCC M2-Like TAMs Are Enriched with SPP1, Overlapping with Defined HCC MΦ Subsets and Representing a Possible Target for TAM Reshaping

A recent study exploring the molecular characteristic of advanced HCC responsive to atezolizumab/bevacizumab identified four TAMs subpopulations: CXCL10⁺, TREM2⁺, SPP1⁺, and MT1G⁺ [35]. Using GSVA R package [32], we assessed the enrichment of these subpopulation signatures in our ROIs to explore potential associations with therapeutic responsiveness.

Three subpopulations – CXCL10⁺, TREM2⁺, SPP1⁺ – were significantly enriched in VETC⁺ tumors. Among these, CXCL10⁺ and SPP1⁺ signatures were also enriched in peritumoral tissue (Fig. 3a). Unsupervised clustering analysis demonstrated that these three subsets clustered together and displayed a gradient pattern strongly associated with VETC positivity (Fig. 3b). Taking into consideration the robust association observed with SPP1, we evaluated SPP1 protein expression by immunohistochemistry in a retrospective cohort of 210 HCC (VETC⁺ = 41). Large, irregularly shaped TAMs located in close proximity to ECs exhibited strong SPP1 expression in 71% (29/41) of VETC⁺ tumors. In contrast, only scattered SPP1⁺ cells were observed in 31% of cases (53/169) of VETC⁻ tumors (Fig. 3c, d). This difference was highly significant (p < 0.001), supporting SPP1 as a potential biomarker of VETC-associated MΦ (Fig. 3c, d).

Fig. 3.

M2-like TAMs associated with VETC⁺ express SPP1, a possible target for TAM reshaping. a Three recently reported TAMs subpopulations (macro CXCL10, macro TREM2 and macro SSP1 [33]) were enriched in VETC⁺ cases (red boxplots) as compared to VETC⁻ cases (green boxplots). Macro MT1G subpopulation is not significantly different. b Unsupervised clustering analysis highlighted that the three subpopulations cluster are increasingly more frequent in VETC⁺ intratumoral regions. c Spatial distribution of SPP1⁺ TAMs in VETC⁺ (A–D) and VETC⁻ case (E, H). A continuous layer of CD34⁺ ECs surrounding cluster of HCC cells consistent with a diagnosis of VETC⁺ HCC (A); several CD163⁺ TAMs lie down beneath ECs (B); SPP1⁺ staining overlaps that observed with CD163, suggesting that most TAMs are CD163⁺/SPP1⁺ (C); a multiplex staining highlights the presence of several CD163⁺/SPP1⁺ TAMs (yellow) and their spatial proximity to ECs (white) (D); CD34⁺ ECs retained the capillary distribution in VETC⁻ HCC (E); CD163⁺ TAMs, even if present, are randomly distributed without any spatial relationship with ECs (F); SPP1⁺ TAMs are extremely rare (G); a multiplex staining highlights scattered SPP1⁺ TAMs (red), without any relationship with CD163⁺ TAMs (green) and CD34⁺ endothelial cells (white) (H). The four sections for both VETC⁺ and VETC⁻ HCC have been cut consecutively every at 3 μm; for immunohistochemistry DAB was used as chromogen; for multiplex white highlighted CD34⁺ endothelial cells, green CD163⁺ TAMs, red SPP1⁺ TAMs; yellow TAMs co-expressing CD163 and SPP1. d VETC⁺ HCC are significantly enriched with SPP1⁺ TAMs (chi-square statistical test). e TAM-SPP1⁺ rich VETC⁺ HCC ROIs are characterized by a microenvironment inhibitory to effective T-cell recruitment and activity. Boxplots show GSVA enrichment scores for TAM-driven immunosuppressive gene signature across VETC⁺ (red) and VETC⁻ (green) regions. Each point represents an individual ROI. p value by Wilcoxon rank-sum test.

To further investigate whether SPP1⁺ TAMs are associated with impaired T-cell tumor infiltration, we designed a macrophage-specific signature capturing lymphocyte recruitment, T-cell activation, and immunosuppressive regulation (including CCL2, CCL5, CCL22, CCL20, CXCL12, ICAM1, VCAM1, PECAM1, SPP1, CD68, LGALS3, CSF1R, NT5E, ENTPD1, IL10, TGFB1, CD274, PDCD1LG2, MMP9, and MMP14), and compared the signature GSVA enrichment scores between VETC⁺ and VETC⁻ regions. The signature was significantly enriched in intratumor VETC⁺ regions, indicating a macrophage-rich and immunosuppressive microenvironment, consistent with a context that is unfavorable for effective T-cell infiltration and function (Fig. 3e).

Discussion

A close spatial and functional interplay between MΦ and EC is a key regulator of physiological and pathological angiogenesis. In multiple contexts, including normal adrenal tissue, glioblastoma, breast cancer, and HCC, distinct MΦ subsets localize near EC and modulate vascular integrity, permeability, and development [36–38]. In HCC, FOLR2⁺ MΦ have been shown to induce oncofetal reprogramming of ECs, highlighting the existence of a specialized immune-vascular niche [32]. MΦ-EC interaction may be further shaped by pro-angiogenic signals within HCC TME: hypoxia-driven HIF-1α promotes M2 polarization, whereas VEGFA inhibition favors M1 differentiation [23–25]. TAMs reshape alters their spatial relationship with EC and, by disrupting this spatial-functional unit, may suppress neoangiogenesis and restore antitumor immunity [39, 40].

We recently reported that VETC⁺ HCC are enriched in large, foamy MΦ closely aligned with ECs [28]. In the present study, we demonstrate that these MΦ display a pro-tumorigenic TAM profile, characterized by enrichment of genes involved in immunosuppression and angiogenesis. Notably, they exhibit high expression of SPP1 (osteopontin), a multifunctional protein implicated in tumor progression and therapeutic resistance [41]. SPP1⁺ TAMs have been associated with T-cell exhaustion across multiple tumor types, including glioblastoma, ovarian cancer, renal cell carcinoma, and metastatic disease [37, 42–44]. In HCC, SPP1⁺ TAMs have recently been shown to accumulate at the periphery of tumor cluster, forming a physical tumor immune barrier (TIB) that limits T-cell infiltration. TIB formation contributes to immune exclusion, reduced response to immunoherapy and poor prognosis [39]. The morphology of TIB closely overlaps with the VETC⁺ HCC cluster surrounded by M2 TAM described in our current and previous work [28], as shown in online supplementary Figure 3. Furthermore, both VETC⁺ HCCs and TIB+ HCC are characterized by immune exclusion and an unfavorable prognosis [2, 39]. This near-complete morphological and functional overlap suggests that TIB⁺ HCC and M2-like TAM surrounding VETC⁺ HCC cluster may represent manifestations of the same biological phenomenon that drive immune deprivation and adverse prognosis. Consistent with this interpretation, macrophage-coated tumor cluster HCC has recently been associated with worse clinical outcome [40]. The study by Liu et al. [39] also provides important therapeutic insights. Using murine models, the authors demonstrated that SPP1 blockade enhances the efficacy of anti-PD-1 immunotherapy. They proposed a biomarker-driven approach in which SPP1-negative HCC may be treated with immunotherapy alone, whereas SPP1-positive tumors could benefit from a sequential or combinatorial strategy targeting SPP1 to dismantle the macrophage barrier prior to immune checkpoint inhibition. A similar strategy has shown promise in experimental glioblastoma models, where depletion of SPP1⁺ TAMs restored T-cell functional balance and improved responsiveness to immune checkpoint inhibitors [45]. Additional preclinical evidence further supports the concept that targeting SPP1, in combination with anti-PD1 agents, can restore T-cell infiltration and enhance the efficacy of immune checkpoint blockade [45, 46]. Our data indicate that M2-like TAMs enrichment is already detectable exclusively in the peritumoral liver parenchyma surrounding VETC⁺ HCC. This finding is in keeping with the concept that non-neoplastic tissue actively facilitates the onset of HCC. This notion, usually referred as “cancer field effect” and originally postulated for skin cancer [47], finds its foundation for HCC in a study showing that the genetic profile of peritumoral but not that of tumoral tissue, correlated with HCC outcome [48]. The concept was later strengthened by proving that peritumoral tissue might develop a pro-oncogenic niche that predispose to HCC [49]. Finally, a comprehensive analysis showed that cirrhotic tissue can exhibit, in up to 50% of patients, few specific immune-mediated signatures associated with a greater tendency to develop HCC, the highest being observed for the “immunosuppressive” [50]. Interestingly, mice treated with nintedanib or aspirin and clopidogrel have a reduction of this immune-mediated cancer field effect and developed fewer and smaller HCC [50]. These findings paved the way to use chemopreventive approaches in subjects at risk of developing neoplastic conditions. In this scenario, our results raise the possibility to test serum SPP1 in cirrhotic patients to identify those at high risk of developing HCC who will likely benefit from these preventive strategies. In addition, they allow the speculation of that the “field effect” does not simply promote the onset of HCC but may play a role in defining the intrinsic characteristics, such as the presence or absence of VETC.

Beyond SPP1, VETC-associated M2-like TAMs overexpress several angiogenesis-related genes, including ADAM9 and MMP9. ADAM9 expression has been associated to VEGFA and ANGPT2 [51] – two of the major player of VETC onset – as well as to the upregulation of pathways involved in vasculogenesis, inflammation and immune response [52]. Therapeutic targeting of ADAM9 can suppress carcinogenesis and exert antitumor effects on HCC [53]. SPP1⁺/MMP9⁺ TAMs have consistently been associated with poor prognosis in HCC [54, 55]. In particular, they have been described in close spatial and functional relationship with cancer stem cells [54] as well as with physical exclusion of T cells from HCC core [54] and pro-tumoral immunosuppressive functions [55]. Notably, VEGFA can induce MMP9 expression in SPP1⁺ TAMs [52], supporting the existence of a feed-forward pro-angiogenic loop.

Our work lays the foundation for further studies aiming at investigating the predictive and therapeutic relevance of SPP1⁺ TAMs in VETC⁺ HCC. The present study is primarily descriptive and is subject to intrinsic limitations, including the relatively small number of cases analyzed by spatial transcriptomic, which limits generalizability and the ability to fully account for clinical heterogeneity. In addition, functional validation studies were not feasible due to the lack of robust murine models recapitulating the VETC phenotype.

Despite these limitations, our data offer mechanistic insight into the interaction between SPP1⁺ TAMs and ECs within VETC⁺ HCC. We propose that SPP1⁺ TAMs may restrict lymphocyte infiltration and promote exhaustion of T cells already present within the tumor. Concurrently, the production of pro-angiogenic mediators may amplify VEGFA and ANGPT2 signaling in a paracrine feedback loop, thereby reinforcing VETC formation [8].

In conclusion, we define the transcriptomic and spatial characteristics of TAMs in VETC⁺ HCC, highlighting their dual contribution to angiogenesis and immune suppression. The consistent enrichment of SPP1⁺ TAMs at both transcriptomic and protein levels underscores their potential biological and clinical relevance, and provides a rationale for combinatorial therapeutic strategies targeting angiogenesis, macrophage polarization, and immune checkpoints pathways.

Statement of Ethics

This study was performed in accordance with the Declaration of Helsinki. This human study was approved by the Local Ethics Committee (Comitato Etico Territoriale Lombardia 5 – Approval No.: 52/24. Written informed consent was waived due to the retrospective nature of the study, as stated by Comitato Etico Territoriale Lombardia 5.

Conflict of Interest Statement

L.R. reports consulting fees from AbbVie, AstraZeneca, Basilea, Bayer, BMS, Eisai, Elevar Therapeutics, Exelixis, Genenta, Hengrui, Incyte, Ipsen, Jazz Pharmaceuticals, MSD, Nerviano Medical Sciences, Roche, Servier, Taiho Oncology, Zymeworks; lecture fees from AstraZeneca, Bayer, BMS, Eisai, Guerbet, Incyte, Ipsen, Roche, Servier; travel expenses from AstraZeneca and Servier; and research grants (to Institution) from AbbVie, AstraZeneca, BeiGene, Exelixis, Fibrogen, Incyte, Ipsen, Jazz Pharmaceuticals, MSD, Nerviano Medical Sciences, Roche, Servier, Taiho Oncology, TransThera Sciences, Zymeworks. A.L. has received consulting fees from Advanz Pharma, GSK, Ipsen, Gilead, Dr Falk, AlfaSigma, and Takeda; speaker fees from Gilead, Abbvie, MSD, Advanz Pharma, AlfaSigma, GSK, Astra Zeneca, Ipsen, and Incyte; and travel support from Ipsen and Falk Fundation and grant support (to Humanitas Research Hospital) from Mirum, GSK, Ipsen, Dr. Falk, and Gilead. G.T. and L.R. were members of the journal’s Editorial Board at the time of submission.

Funding Sources

The research leading to these results has received funding from AIRC under IG 2020 – ID 25087 project – P.I. Di Tommaso Luca”; CKYN is supported by the AIRC Start-Up (Grant No. 30787); E.V. is supported by the AIRC Grant No. IG 2020-ID, 24,858; The PhD students M.V. and R.P. were supported by the PhD program in Experimental Medicine of the University of Milan. R.P. was supported by an AIRC fellowship for Italy. The sponsors had no active role on study design, execution and analysis, and manuscript conception, planning, writing, and decision to publish.

Author Contributions

Concept and design: L.D.T., F.M., C.D.C., R.A., L.R., A.L., L.T., and E.V. Case selection and database management: C.D.C., F.P., G.C., G.T., and A.R.P. Experiments and procedures: G.B., S.M., A.R.P., B.D., M.V., R.P., C.S., B.F., C.K.Y.N., S.P., and F.G., Writing of article: L.D.T., C.D.C., A.R.P., and G.B.

Funding Statement

The research leading to these results has received funding from AIRC under IG 2020 – ID 25087 project – P.I. Di Tommaso Luca”; CKYN is supported by the AIRC Start-Up (Grant No. 30787); E.V. is supported by the AIRC Grant No. IG 2020-ID, 24,858; The PhD students M.V. and R.P. were supported by the PhD program in Experimental Medicine of the University of Milan. R.P. was supported by an AIRC fellowship for Italy. The sponsors had no active role on study design, execution and analysis, and manuscript conception, planning, writing, and decision to publish.

Data Availability Statement

The authors confirm that the data supporting the findings of this study are available within the article and its supplementary materials. Further inquiries can be directed to the corresponding author.