Tumor-associated high endothelial venules are associated with enhanced lymphocyte infiltration and favorable prognosis in resected hepatocellular carcinoma Authors: Shu Aoyama, Takehiro Noda, Hirofumi Akita, Yosuke Mukai, Kazuki Sasaki, Shinichiro Hasegawa, Daisaku Yamada, Yoshito Tomimaru, Tadafumi Asaoka, Hidenori Takahashi, Shogo Kobayashi, Junzo Shimizu, Azumi Ueyama, Takuro Saito, Kansuke Kido, Eiichi Morii, Hisashi Wada, Yuichiro Doki, Hidetoshi Eguchi Published: 2025 ABSTRACT BackgroundTumor-infiltrating lymphocytes (TILs) are essential to effective immunotherapy for hepatocellular carcinoma (HCC). Among factors regulating TILs, tumor-associated high endothelial venules (TA–HEVs) have been identified in various cancers and may contribute to TIL recruitment. This study aimed to investigate the clinical significance of TA–HEVs in HCC and their association with TILs.Methods156 patients with HCC who underwent hepatic resection were analyzed. TA–HEVs were identified using double immunohistochemical staining for MECA-79 and CD31. Clinical outcomes and TIL density were compared between TA–HEV–present and –absent cases. Multicolor flow cytometry and RNA sequencing of tumor tissues were also performed.ResultsTA–HEVs were observed within tertiary lymphoid structures (TLSs) surrounding tumors in 31 of the 156 cases. The proportion of TLSs containing TA–HEVs increased with TLS maturation. TA–HEV–present cases had significantly better disease-free survival (3 years: 79.8% in TA–HEV–present cases vs 60.1% in TA–HEV–absent cases, p = 0.011). Multivariate analysis identified TA–HEVs as an independent favorable prognostic factor. Moreover, TA–HEV–present cases had significantly higher densities of intra-tumoral CD4⁺ and CD8⁺ T cells. RNA sequencing revealed enhanced pathways related to immune response and lymphocyte activation in TA–HEV–present cases, with high expression of cytotoxic lymphocyte markers. Flow cytometry confirmed elevated proportions of activated CD8⁺ T cells in TA–HEV–present cases.ConclusionsTA–HEVs in HCC were associated with significantly better prognosis and may contribute to immune activation within the tumor microenvironment.Supplementary InformationThe online version contains supplementary material available at 10.1007/s00262-025-04265-z. Background HCC accounts for over 80% of primary liver cancers and is the fourth leading cause of cancer-related mortality worldwide [1]. Despite local treatments, such as hepatectomy and radiofrequency ablation, tumor recurrence occurs in 35–80% of cases, leading to a poor prognosis [2]. Recently, the development of immune-combination therapies for HCC, such as atezolizumab (anti-PD-L1 antibody) with bevacizumab (anti-VEGF antibody) and durvalumab (anti-PD-L1 antibody) with tremelimumab (anti-CTLA-4 antibody), has been remarkable [3, 4]. Still, response rates for these therapies remain limited to approximately 20–30%, highlighting the need to better understand factors affecting treatment efficacy. The presence of immune cells, especially CD8+ T cells, within the tumor plays an important role in determining the efficacy of immunotherapy in many types of cancer [5–7]. However, in HCC, the factors that regulate the migration and infiltration of immune cells into the tumor are not fully understood. Tertiary lymphoid structures (TLSs) and tumor-associated high endothelial venules (TA–HEVs) are thought to be involved in lymphocyte recruitment within the tumor microenvironment and have been the focus of several studies. TLSs are ectopic lymphoid formations that develop in inflamed, infected, or tumor-adjacent tissues [8]. TLSs in tumor-adjacent tissues are thought to facilitate immune cell infiltration into tumor sites and are associated with favorable prognosis in various cancers, including HCC [9–12]. High endothelial venules (HEVs) are specialized mammalian segments of vasculature present in secondary lymphoid organs. The endothelial cells of HEVs express L-selectin ligands and facilitate the constant infiltration of immune cells into lymph nodes through lymphocyte rolling [13]. Tumor-associated TLSs contain vasculature that resembles HEVs, which have been designated as TA–HEVs [14]. TA–HEVs are found at the tumor margin and are believed to facilitate transport of naïve T cells [15]. Previous studies have reported that the presence of TA–HEVs is associated with high densities of tumor-infiltrating lymphocytes (TILs) and a favorable prognosis in various cancers [16–18]. However, the significance of TA–HEVs and the relation between TA–HEVs and TILs in HCC are still unclear. The aim of this study was to elucidate the prognostic significance of TA–HEVs in HCC. In addition, we investigated the quantitative relationship between TA–HEVs and TILs and the characteristics of TIL subsets that are associated with TA–HEVs. Methods Patients and tissue specimens Two distinct cohorts were enrolled for this study, representing different time frames and different sample collection and storage. Cohort 1 included patients undergoing curative resection for HCC at The University of Osaka Hospital from January 2015 to August 2022. Exclusion criteria were as follows: a history of preoperative treatments, such as trans-arterial chemoembolization and systemic therapy; combined hepatocellular–cholangiocarcinoma; and an R2 resection. Clinicopathological data and prognostic information for these cases were retrospectively collected from medical records. Formalin-fixed, paraffin-embedded (FFPE) samples from resected specimens were used for the evaluation of TA–HEVs, TLSs, and TILs. Some frozen tissue samples were used for RNA sequencing. Cohort 2 included patients who underwent curative resection for HCC from January 2022 to July 2025, with fresh tumor tissue available for collection. Clinicopathological characteristics were collected, and fresh tumor samples were used for flow cytometry. FFPE samples were used for TA–HEV assessment, and some frozen tissue samples were analyzed by RNA sequencing. Assessment of TLSs and TA–HEVs TLSs were identified as aggregates of lymphocytes > 5000 μm2 at the boundary between the HCC and non-tumor areas within 1000 μm outward from the tumor border [19]. These clusters were observed under a microscope in FFPE slides stained with hematoxylin and eosin (H&E). For identification of TA–HEVs, single immunohistochemical staining was performed using anti-MECA-79 (sc-19602, Santa Cruz; 1:200 dilution), anti-CK7 (ab183344, Abcam; 1:400 dilution), anti-CD31 (ab9498, Abcam; 1:1000 dilution), and anti-MAdCAM-1 antibodies (66,594-1-Ig, Proteintech; 1:500 dilution). The staining procedure followed a previous report [20]. Briefly, FFPE blocks were sectioned into slices 3.5 μm thick and antigen retrieval was carried out using citrate buffer. Endogenous peroxidase activity was blocked with 0.3% hydrogen peroxide in methanol, followed by incubation with each primary antibody. Peroxidase labeling was performed using a peroxidase labeling reagent (Histofine series, Nichirei Bioscience), and staining was conducted with DAB (#8059, Cell Signaling). Supplemental Table S1 lists the antibodies used for immunohistochemical staining. Next, the presence of TA–HEVs was assessed using double immunohistochemical staining as follows: procedures up to endogenous peroxidase blocking were carried out as in single immunohistochemical staining. Sections were then incubated with anti-MECA-79 antibody. Biotin labeling was performed using a biotinylation reagent (BA-4000-1.5, Vector Laboratories), followed by alkaline phosphatase labeling (#434,322, Thermo Fisher Scientific). Staining was carried out using PermaBlue (K058, Diagnostic Biosystems). Thereafter, sections were incubated with anti-CD31 antibody. Peroxidase labeling was then performed using a peroxidase labeling reagent (#414,321, Nichirei Bioscience), and staining was conducted with DAB. TA–HEVs were identified by manual microscopic inspection as CD31+ MECA-79+ vascular structures within TLSs. For additional validation, double immunohistochemical staining with anti-CK7 antibody and anti-MECA-79 antibody was also performed using the same method. TA–HEVs were also identified as CK7− MECA-79+ vascular structures within TLSs. The maturation of TLSs was evaluated based on their morphological features and immunohistochemical staining with anti-CD21 antibody (NCL-L-CD21-2G9, Leica; 1:200 dilution) and anti-CD23 (ab16702, Abcam; 1:1000 dilution) as previously reported [9, 19]. TLSs were classified into three categories: early TLSs (E-TLSs), indistinct lymphocyte aggregates lacking CD21 and CD23 signals; primary follicle-like TLSs (PFL–TLSs), round or oval clusters of lymphocytes positive for CD21 but negative for CD23; and secondary follicle-like TLSs (SFL–TLSs), follicles with a germinal center composed of lymphocytes with clear cytoplasm and positive for both CD21 and CD23. TLS assessment was performed in all cases on samples from Cohort 1, and TA–HEV assessment covered all samples from Cohort 1 and 2. For pathological assessments in this study, including the presence of TA–HEVs, slides were reviewed and confirmed by board-certified pathologists. Assessment of TILs For assessment of TILs, serial sections of FFPE slides were stained with anti-CD4 antibody (ab213215, Abcam; 1:200 dilution), anti-CD8 antibody (ab17147, Abcam; 1:400 dilution), and anti-Foxp3 antibody (ab20034, Abcam; 1:500 dilution). TIL density was evaluated using image analysis software (HALO; Indica Labs). Background regions were removed using the software’s classifier function. Tumor areas and areas extending 1000 μm outward from the tumor border (hereafter referred to as tumor margin areas) were then manually annotated. The tumor border was defined as the outermost part containing tumor cells, and the capsular area was considered non-tumoral tissue. The number of positively stained cells for each antibody was automatically quantified in both regions. Detection thresholds for each antibody were adjusted based on staining intensity to ensure accurate analysis. TIL assessments were performed in the resected HCC specimens of Cohort 1. Genetic sequence data RNA sequencing was performed on frozen tumor samples obtained from surgically resected HCC specimens. Tissue samples were homogenized in QIAzol (Qiagen), and total RNA was extracted using a QuickGene AutoS RNA Cultured Cell Kit (KURABO). Library preparation was conducted with a TruSeq Stranded RNA Library Prep Kit (Illumina). Sequencing was performed on a NovaSeq Xplus sequencer (Illumina) in paired-end mode (2 × 101 nt). Post-sequencing, adapter trimming was carried out using Trimmomatic, and reads were mapped to the reference genome (hg19) using TopHat version 2.1.1. Fragments per kilobase million were calculated with Cufflinks version 2.2.1, and raw counts were determined using featureCounts version 2.0.0. Pathway enrichment analysis was performed using the web-based tool iDEP 2.0 (https://bioinformatics.sdstate.edu/idep/), with Gene Ontology pathways serving as the reference. Furthermore, the immune-related cell composition in each sample was inferred using the Microenvironment Cell Populations (MCP) counter based on the expression of marker genes. These marker genes were aggregated into scores representing the relative abundance of immune-related cell types [21]. The raw data for RNA sequences have been submitted to Sequence Read Archive (submission no. PRJNA1227701). For RNA sequencing, the most recently collected cases from the two cohorts were prioritized for analysis. Among them, 10 cases with both TLSs and TA–HEVs and 10 cases without either TLSs or TA–HEVs were selected based on FFPE-based evaluation. Flow cytometry Multicolor flow cytometry was performed to evaluate TIL subsets associated with TA–HEV presence in fresh HCC tissues from resected specimens associated with all cases in Cohort 2. The methodology followed previously reported protocols [22, 23]. Tumor dissociation and cell isolation were carried out using the Tumor Dissociation Kit (#130-095-929, Miltenyi Biotec). For cell-surface staining, cells were incubated in a single step with an antibody cocktail comprising fluorochrome-conjugated antibodies targeting cell-surface antigens, including CTLA-4. This was followed by cell fixation and intracellular staining using a separate antibody cocktail for intracellular antigens. Two distinct multicolor staining panels were used for cell staining (see Supplemental Table S2 for antibodies). Stained cells were analyzed using the NovoCyte Quanteon Flow Cytometer and NovoExpress Software (Agilent). In this study, regulatory T cells (Tregs) were defined as CD45RA− Foxp3+ CD4+ T cells and effector CD4+ T cells as CD45RA− Foxp3− CD4+ T cells. Statistics Descriptive statistics were presented as mean ± SD or median with IQR. Categorical variables were analyzed using Fisher’s exact test, and continuous variables were compared using either the t test or the Brunner–Munzel test. OS and DFS were estimated with the Kaplan–Meier method and the log-rank test. Prognostic factors were evaluated through univariate and multivariate analyses with a Cox proportional hazards model. In the multivariate analysis, covariates were selected based on a significance threshold of p < 0.05 in univariate analysis. Statistical significance was defined as a two-sided p value < 0.05. Statistical analyses and data visualizations were performed using R (Version 4.3.2; The R Project for Statistical Computing). Results TA–HEVs are present within TLSs at the HCC margin Cohort 1 included 156 primary HCC cases. Patient characteristics for this cohort are presented in Supplemental Table S3. The median age overall was 74 years, and 76.9% were male and 23.1% female. Chronic hepatitis B was present in 14.1% of cases, chronic hepatitis C was observed in 33.3%, and liver cirrhosis was observed in 9.6% of cases. A total of 96.8% of cases were classified as Child–Pugh A. Median tumor size was 3.0 cm, and 85.3% of cases involved a single tumor. TA–HEVs were present in 31 cases (19.9%). The representative H&E-stained images are shown in Fig. 1A. TLS formation was observed in 59 cases (37.8%) in Cohort 1 (Fig. 1A, upper panel, yellow arrowhead). TLSs were seen mainly on the non-tumor side adjacent to the tumor border. We performed immunohistochemical staining of MECA-79 to identify TA–HEVs in HCC tissues and stained for CK7 and CD31 to discriminate endothelial venules from biliary epithelium, which also expresses MECA-79 in liver tissues [24] (Fig. 1B). MECA-79-positive tubular structures were detected within TLSs at the tumor margin and showed negative expression of CK7 and positive expression of CD31 and MAdCAM-1, a gut-associated or immature HEV marker [25, 26] (Fig. 1B, upper panel). The tubular structures of CD31+ MECA-79+ were identified as TA–HEVs. MECA-79-positive tubular structures that showed positive expression of CK7 and negative expression of CD31 and MAdCAM-1 were considered to be bile ducts (Fig. 1B, lower panel). Double immunostaining for MECA-79 and CD31 identified CD31+ MECA-79+ tubular structures within TLSs, confirming them as TA–HEVs (Fig. 1C, red arrows). In this staining, bile ducts that were positive only for MECA‑79 were also identified (Fig. 1C, yellow arrowheads). Furthermore, double immunostaining for MECA-79 and CK7 confirmed that these TA–HEVs were negative for CK7 expression (Fig. 1D, red arrows). The experiments described below were based on the results of the double MECA-79 and CD31 immunostaining for identification of TA–HEVs at the HCC tumor margin.Fig. 1A H&E images in the tumor boundaries: top, case with TLS; bottom, case without TLS. Yellow arrowheads indicate TLS (100 ×). B Immunohistochemical staining images of MECA-79, CK7, CD31, and MAdCAM-1 in the same regions: top, case with TA–HEVs; bottom, case without TA–HEVs (100 ×). C Representative images of double immunohistochemical staining of MECA-79 and CD31: left, 20 × ; right, magnified view (100 ×). MECA-79 was stained with a blue substrate and CD31 with a brown substrate. Red arrows indicate TA–HEVs and yellow arrowheads indicate bile ducts. D Representative images of double immunohistochemical staining of MECA-79 and CK7: left, 20 × ; right, magnified view (100 ×). MECA-79 was stained with a blue substrate and CK7 with a brown substrate. Red arrows indicate TA–HEVs and yellow arrowheads indicate bile ducts. E Representative images of TLSs in each maturation category as evaluated by H&E and CD21/CD23 immunohistochemical staining. F Numbers of TLSs classified by each maturation category and stratified by the presence of TA–HEVs. G Number of cases classified by the presence of TA–HEVs and stratified by the maturation category of the most mature TLS in each case. All scale bars: 100 μm. *p < 0.05, **p < 0.001. Abbreviations: H&E; hematoxylin and eosin; TLS, tertiary lymphoid structure; TA–HEV, tumor-associated high endothelial venule; E-TLS, early tertiary lymphoid structure; PFL–TLS, primary follicle-like tertiary lymphoid structure; SFL–TLS, secondary follicle-like tertiary lymphoid structure A H&E images in the tumor boundaries: top, case with TLS; bottom, case without TLS. Yellow arrowheads indicate TLS (100 ×). B Immunohistochemical staining images of MECA-79, CK7, CD31, and MAdCAM-1 in the same regions: top, case with TA–HEVs; bottom, case without TA–HEVs (100 ×). C Representative images of double immunohistochemical staining of MECA-79 and CD31: left, 20 × ; right, magnified view (100 ×). MECA-79 was stained with a blue substrate and CD31 with a brown substrate. Red arrows indicate TA–HEVs and yellow arrowheads indicate bile ducts. D Representative images of double immunohistochemical staining of MECA-79 and CK7: left, 20 × ; right, magnified view (100 ×). MECA-79 was stained with a blue substrate and CK7 with a brown substrate. Red arrows indicate TA–HEVs and yellow arrowheads indicate bile ducts. E Representative images of TLSs in each maturation category as evaluated by H&E and CD21/CD23 immunohistochemical staining. F Numbers of TLSs classified by each maturation category and stratified by the presence of TA–HEVs. G Number of cases classified by the presence of TA–HEVs and stratified by the maturation category of the most mature TLS in each case. All scale bars: 100 μm. *p < 0.05, **p < 0.001. Abbreviations: H&E; hematoxylin and eosin; TLS, tertiary lymphoid structure; TA–HEV, tumor-associated high endothelial venule; E-TLS, early tertiary lymphoid structure; PFL–TLS, primary follicle-like tertiary lymphoid structure; SFL–TLS, secondary follicle-like tertiary lymphoid structure To investigate factors associated with TA‑HEV presence within TLS, we assessed the association between TLS maturation and TA‑HEV presence. As shown in Fig. 1E, TLSs were classified as E‑TLS, PFL‑TLS, or SFL‑TLS based on follicle formation and CD21 and CD23 expression. In Cohort 1, there were 235 TLSs in total among the 59 patients, and 105 of these TLSs (44.7%) contained TA‑HEVs. E‑TLSs accounted for more than half of all TLSs, numbering 143 (60.9%), while PFL‑TLSs and SFL‑TLSs numbered 31 (13.2%) and 59 (25.1%), respectively. TA‑HEVs were present even in E‑TLSs (48 of 143, 33.6%), but increased significantly with TLS maturity, to 15 of 31 (48.4%) in PFL‑TLSs and 49 of 59 (83.1%) in SFL‑TLSs (p < 0.001) (Fig. 1F). On a per-case basis, of 59 cases with TLSs, TA–HEVs were present in 31 cases (52.5%). Figure 1G shows the phenotype of the most mature TLS in each case; TA‑HEV–present cases had a significantly higher proportion of mature TLSs than TA‑HEV–absent cases (TA‑HEV–present cases, n = 31: PFL‑TLS 6 cases (19.4%), SFL‑TLS 17 cases (54.8%) vs. TA‑HEV–absent cases, n = 29: PFL‑TLS 2 cases (6.9%), SFL‑TLS 1 case (3.4%), p < 0.001). These results indicated an association between TLS maturation and TA–HEV formation. HCC cases with TA–HEVs are associated with significantly favorable DFS Table 1 gives a comparison of the patient characteristics between those with and without TA–HEVs in Cohort 1. The two subgroups did not differ in clinical or pathological factors.Table 1Patient characteristics according to TA–HEV status, Cohort 1TA–HEV absent n = 125TA–HEV present n = 31pAge (years)74 (67–79)73 (70–79)0.901Sex > 0.99 Male96 (76.8%)24 (77.4%) Female29 (23.2%)7 (22.6%)HBsAg16 (12.8%)6 (19.4%)0.388Anti-HCV antibody41 (32.8%)11 (35.5%)0.833Child–Pugh classification > 0.99 A121 (96.8%)30 (96.8%) B4 (3.2%)1 (3.2%)Liver cirrhosis12 (9.6%)3 (9.7%) > 0.99ASA-PS0.721 111 (8.8%)4 (12.9%) 290 (72.0%)22 (71.0%) 324 (19.2%)5 (16.1%)Alpha-fetoprotein (ng/ml)6 (4–67)15 (5–102)0.445Des-gamma-carboxy prothrombin (mAU/ml)83 (28–981)106 (45–507)0.913Tumor size (cm)3.0 (2.2–5.0)2.7 (2.0–4.5)0.275Number of tumors0.802 Single107 (85.6%)26 (83.9%) Multiple18 (14.4%)5 (16.1%)Histological type0.173 Well/moderately differentiated96 (76.8%)20 (64.5%) Undifferentiated/poorly differentiated29 (23.2%)11 (35.5%)Microscopic vascular invasion26 (20.8%)5 (16.1%)0.802Microscopic intrahepatic metastasis12 (9.6%)1 (3.2%)0.467Continuous values are presented as medians with IQRs, and categorical values are presented as numbers (%)Abbreviations: ASA–PS, American Society of Anesthesiologists Physical Status; TA–HEV, tumor-associated high endothelial venule Table 1Patient characteristics according to TA–HEV status, Cohort 1TA–HEV absent n = 125TA–HEV present n = 31pAge (years)74 (67–79)73 (70–79)0.901Sex > 0.99 Male96 (76.8%)24 (77.4%) Female29 (23.2%)7 (22.6%)HBsAg16 (12.8%)6 (19.4%)0.388Anti-HCV antibody41 (32.8%)11 (35.5%)0.833Child–Pugh classification > 0.99 A121 (96.8%)30 (96.8%) B4 (3.2%)1 (3.2%)Liver cirrhosis12 (9.6%)3 (9.7%) > 0.99ASA-PS0.721 111 (8.8%)4 (12.9%) 290 (72.0%)22 (71.0%) 324 (19.2%)5 (16.1%)Alpha-fetoprotein (ng/ml)6 (4–67)15 (5–102)0.445Des-gamma-carboxy prothrombin (mAU/ml)83 (28–981)106 (45–507)0.913Tumor size (cm)3.0 (2.2–5.0)2.7 (2.0–4.5)0.275Number of tumors0.802 Single107 (85.6%)26 (83.9%) Multiple18 (14.4%)5 (16.1%)Histological type0.173 Well/moderately differentiated96 (76.8%)20 (64.5%) Undifferentiated/poorly differentiated29 (23.2%)11 (35.5%)Microscopic vascular invasion26 (20.8%)5 (16.1%)0.802Microscopic intrahepatic metastasis12 (9.6%)1 (3.2%)0.467Continuous values are presented as medians with IQRs, and categorical values are presented as numbers (%)Abbreviations: ASA–PS, American Society of Anesthesiologists Physical Status; TA–HEV, tumor-associated high endothelial venule Patient characteristics according to TA–HEV status, Cohort 1 Continuous values are presented as medians with IQRs, and categorical values are presented as numbers (%) Abbreviations: ASA–PS, American Society of Anesthesiologists Physical Status; TA–HEV, tumor-associated high endothelial venule The association between the presence of TA–HEVs and postoperative prognosis was examined. DFS was significantly better in the TA–HEV–present group (3year rates: 79.8% with TA–HEVs present vs. 60.1% with TA–HEVs absent; 5 year rates: 74.5% vs. 40.0%, p = 0.011) (Fig. 2A, left panel). The TA–HEV–present group also showed a nonsignificant trend to better OS (3 year OS rates: 96.7% vs. 86.2%; 5 year OS rates: 91.6% vs. 77.7%, p = 0.065) (Fig. 2A, right panel).Fig. 2A DFS (left) and OS (right) in Cohort 1, stratified by TA–HEV presence. B DFS analysis in Cohort 1, stratified by TLS presence (left) and by combined TA–HEV/TLS presence (right). C DFS (left) and OS (right) in Cohort 1, stratified by non-tumoral HEV presence. Abbreviations: DFS, disease-free survival; OS, overall survival; TA–HEV, tumor-associated high endothelial venule; TLS, tertiary lymphoid structure; HEV, high endothelial venule A DFS (left) and OS (right) in Cohort 1, stratified by TA–HEV presence. B DFS analysis in Cohort 1, stratified by TLS presence (left) and by combined TA–HEV/TLS presence (right). C DFS (left) and OS (right) in Cohort 1, stratified by non-tumoral HEV presence. Abbreviations: DFS, disease-free survival; OS, overall survival; TA–HEV, tumor-associated high endothelial venule; TLS, tertiary lymphoid structure; HEV, high endothelial venule In this study, TA–HEVs were defined as CD31⁺ MECA-79⁺ vascular structures within TLSs; thus, TLSs were present in all TA–HEV–present cases and they could also be present in TA–HEV–absent cases. The presence of TLSs, however, did not significantly influence DFS in Cohort 1 (3 year DFS rates: 70.1% for TLS–present cases vs. 60.0% for TLS–absent cases; 5 year DFS rates: 59.7% vs. 38.4%, p = 0.103) (Fig. 2B, left panel). In addition, DFS was similar between TA–HEV–absent cases with TLSs and TA–HEV–absent cases without TLSs (3 year DFS rates: 60.4% in TLS–present cases vs 60.0% in TLS–absent cases; 5 year DFS rates: 45.1% vs. 38.4%, p = 0.729) (Fig. 2B, right panel). These findings suggest that TLSs alone may not contribute to prognosis but that the formation of TA–HEVs within TLSs could have an effect.We also evaluated the association between non-tumoral HEV presence and prognosis, and found no significant relationship with DFS or OS (3 year DFS rates: 77.4% in non-tumoral HEV-present cases vs. 61.8% in non-tumoral HEV–absent cases, p = 0.729; 3-year OS rates: 88.9% vs. 87.9%, p = 0.410) (Fig. 2C). These findings indicate that non-tumoral HEVs are unlikely to function as effective mediators of immune surveillance against tumor recurrence. To further investigate the prognostic impact of TA–HEVs, univariate and multivariate analyses of DFS were performed in Cohort 1, and the results are summarized in Table 2. Univariate analysis identified three factors as prognostically significant in this cohort: number of tumors (multiple), presence of microscopic intrahepatic metastasis, and absence of TA–HEVs. Multivariate analysis singled out TA–HEVs as the only independent prognostic factor (p = 0.029; HR 2.405, 95% CI 1.094–5.288). These findings strongly suggest that the absence of TA–HEVs was an unfavorable prognostic factor.Table 2Univariate and multivariate Cox regression analyses for DFS after hepatectomy in Cohort 1Univariate analysisMultivariate analysisHR95% CIpHR95% CIpAge ≥ 75 (vs < 75)1.0670.671–1.6970.784Sex Male (vs female)1.4330.811–2.5320.216Child–Pugh classification B (vs A)2.5570.798–8.1910.114Alpha-fetoprotein > 10 ng/ml (vs ≤ 10 ng/ml)1.4220.898–2.2520.134Des-gamma-carboxy prothrombin > 400 mAU (vs ≤ 400 mAU)1.3050.790–2.1550.298Tumor size > 2 cm (vs ≤ 2 cm)1.2370.700–2.1830.464Number of tumors Multiple (vs single)2.2511.181–4.2930.0141.6850.798–3.5600.171Histological type Poorly (vs moderately/well) differentiated0.8330.413–1.6830.612Microscopic vascular invasion + (vs − )1.3850.783–2.4520.263Microscopic intrahepatic metastasis + (vs − )2.2391.071–4.6820.0321.4770.628–3.4740.371TA–HEV − (vs + )2.6391.211–5.7470.0152.4051.094–5.2880.029Abbreviation: TA–HEV, tumor-associated high endothelial venule Table 2Univariate and multivariate Cox regression analyses for DFS after hepatectomy in Cohort 1Univariate analysisMultivariate analysisHR95% CIpHR95% CIpAge ≥ 75 (vs < 75)1.0670.671–1.6970.784Sex Male (vs female)1.4330.811–2.5320.216Child–Pugh classification B (vs A)2.5570.798–8.1910.114Alpha-fetoprotein > 10 ng/ml (vs ≤ 10 ng/ml)1.4220.898–2.2520.134Des-gamma-carboxy prothrombin > 400 mAU (vs ≤ 400 mAU)1.3050.790–2.1550.298Tumor size > 2 cm (vs ≤ 2 cm)1.2370.700–2.1830.464Number of tumors Multiple (vs single)2.2511.181–4.2930.0141.6850.798–3.5600.171Histological type Poorly (vs moderately/well) differentiated0.8330.413–1.6830.612Microscopic vascular invasion + (vs − )1.3850.783–2.4520.263Microscopic intrahepatic metastasis + (vs − )2.2391.071–4.6820.0321.4770.628–3.4740.371TA–HEV − (vs + )2.6391.211–5.7470.0152.4051.094–5.2880.029Abbreviation: TA–HEV, tumor-associated high endothelial venule Univariate and multivariate Cox regression analyses for DFS after hepatectomy in Cohort 1 Age ≥ 75 (vs < 75) Sex Male (vs female) Child–Pugh classification B (vs A) Alpha-fetoprotein > 10 ng/ml (vs ≤ 10 ng/ml) Des-gamma-carboxy prothrombin > 400 mAU (vs ≤ 400 mAU) Tumor size > 2 cm (vs ≤ 2 cm) Number of tumors Multiple (vs single) Histological type Poorly (vs moderately/well) differentiated Microscopic vascular invasion + (vs − ) Microscopic intrahepatic metastasis + (vs − ) TA–HEV − (vs + ) Abbreviation: TA–HEV, tumor-associated high endothelial venule In TA–HEV–present HCC, T cells are highly concentrated within the tumor and tumor margin areas Next, we analyzed the association between T cell density (CD8+ T cells, CD4+ T cells, and Foxp3+ T cells) within tumor areas and at tumor margin areas in 97 consecutive cases from Cohort 1. TA–HEVs were identified in 19 cases (19.6%). Figure 3A shows representative figures with the tumor and tumor margin areas. At the tumor margin, CD8+ and CD4+ T cell densities were significantly higher in TA–HEV–present cases (CD8+ T cells: 253.5/mm2 in TA–HEV–present cases vs. 141.5/mm2 in TA–HEV–absent cases, p = 0.011; CD4+ T cells: 309.6/mm2 vs. 146.4/mm2, p = 0.004) (Fig. 3B). Similarly, in tumor areas, the densities of CD8+ T and CD4+ T cells were significantly higher in TA–HEV–present cases (CD8+ T cells: 185.5/mm2 vs. 108.5/mm2 in TA–HEV–absent cases, p = 0.023; CD4+ T cells: 169.5/mm2 vs. 86.4/mm2, p = 0.040) (Fig. 3C). In contrast, Treg (Foxp3+ T cells) densities were low with or without TA–HEV presence in tumor or tumor margin areas.Fig. 3A Representative images of immunohistochemical staining for TA–HEVs and immune cells Top panel: low-magnification (20 ×) images. The tumor margin area is enclosed by dotted lines. The blue square indicates a part of the tumor margin area, and the red square indicates a part of the tumor area. Middle panel: high-magnification (100 ×) images of the tumor margin area. Red arrows indicate TA–HEVs. Bottom panel: high-magnification (100 ×) images of the tumor area. The scale bar represents 500 μm in the top panel and 100 μm in the middle and bottom panels. B Immune cell density in the tumor margin area. C Immune cell density in the tumor area. *p < 0.05. Error bars represent SE. Abbreviations: TA–HEV, tumor-associated high endothelial venule; Treg, regulatory T cell A Representative images of immunohistochemical staining for TA–HEVs and immune cells Top panel: low-magnification (20 ×) images. The tumor margin area is enclosed by dotted lines. The blue square indicates a part of the tumor margin area, and the red square indicates a part of the tumor area. Middle panel: high-magnification (100 ×) images of the tumor margin area. Red arrows indicate TA–HEVs. Bottom panel: high-magnification (100 ×) images of the tumor area. The scale bar represents 500 μm in the top panel and 100 μm in the middle and bottom panels. B Immune cell density in the tumor margin area. C Immune cell density in the tumor area. *p < 0.05. Error bars represent SE. Abbreviations: TA–HEV, tumor-associated high endothelial venule; Treg, regulatory T cell HCC cases with TA–HEVs show higher expression of T cell marker genes To further examine the relationship between TA–HEVs and TILs, we conducted RNA-seq analysis using frozen HCC specimens. RNA-seq data from 10 tumor tissues with both TA–HEVs and TLSs and 10 tumor tissues without either TA–HEVs or TLSs were analyzed. The two subgroups in this analysis did not differ significantly in patient characteristics (Supplemental Table S4). As shown in Fig. 4A, principal component analysis indicated a trend toward separation of TA–HEV–present and TA–HEV–absent cases, suggesting an association of TA–HEVs with differences in gene expression profiles. In the pathway enrichment analysis referencing Gene Ontology, the most enriched pathways in TA–HEV–present subgroup were immune response and immune system process, followed by leukocyte activation, cell activation, lymphocyte activation, and regulation of immune system process. These results implicate TA–HEVs in specific pathways related to immune activity and lymphocyte activation (Fig. 4B).Fig. 4A PCA plot of RNA sequencing data. Blue dots represent TA–HEV–present cases, and red dots represent TA–HEV–absent cases. B Pathway enrichment analysis using gene ontology as the reference, comparing between cases with and without TA–HEVs. Pathways are arranged in descending order of FDR. Circle size represents the number of referenced genes, and color indicates the enrichment score. C Heatmap showing expression levels of marker genes for immune cells in each case, as determined by the MCP counter. Data are standardized across cell types. D Expression scores for each immune cell in the MCP counter. E Fold changes of gene expression in TA–HEV–present cases compared to TA–HEV–absent cases. *p < 0.05, **p < 0.001. Error bars represent SE. Abbreviations: PCA, principal component analysis; PC, principal component; TA–HEV, tumor-associated high endothelial venule; GO, gene ontology; FDR, false discovery rate; MCP, microenvironment cell population; NK, natural killer A PCA plot of RNA sequencing data. Blue dots represent TA–HEV–present cases, and red dots represent TA–HEV–absent cases. B Pathway enrichment analysis using gene ontology as the reference, comparing between cases with and without TA–HEVs. Pathways are arranged in descending order of FDR. Circle size represents the number of referenced genes, and color indicates the enrichment score. C Heatmap showing expression levels of marker genes for immune cells in each case, as determined by the MCP counter. Data are standardized across cell types. D Expression scores for each immune cell in the MCP counter. E Fold changes of gene expression in TA–HEV–present cases compared to TA–HEV–absent cases. *p < 0.05, **p < 0.001. Error bars represent SE. Abbreviations: PCA, principal component analysis; PC, principal component; TA–HEV, tumor-associated high endothelial venule; GO, gene ontology; FDR, false discovery rate; MCP, microenvironment cell population; NK, natural killer Using the MCP counter, we estimated the abundance of immune-related cells based on expression levels of marker genes specific to each cell type. The resulting heatmap shows distinct expression patterns between TA–HEV–present and TA–HEV–absent cases (Fig. 4C). Indeed, the expression scores of marker genes for T cells, CD8+ T cells, and cytotoxic lymphocytes were significantly higher in TA–HEV–present cases (T cells: 5.70 in TA–HEV–present cases vs. 2.12 in TA–HEV–absent cases, p < 0.001; CD8+ T cells: 5.23 vs. 0.96, p = 0.003; cytotoxic lymphocytes: 3.01 vs. 1.54, p = 0.031) (Fig. 4D). The expression scores for myeloid dendritic cells were also significantly elevated in TA–HEV–present cases (3.27 vs. 1.11 with TA–HEVs absent, p = 0.008). Based on the above findings, the genetic analysis also indicated increased TILs in TA–HEV–present cases. Our results suggested an association between TA‑HEVs and the recruitment of immune cells to the tumor, and we conducted an analysis of genes involved in chemotaxis. TA‑HEV–present cases showed increased expression of molecules associated with T‑cell recruitment and activation (Fig. 4E). In particular, the chemokines CCL21 and CXCL12, which are central to lymphocyte trafficking, were significantly elevated. We also observed significantly higher levels of the cytokine IFNG and the chemotaxis regulator STAT1, both of which are linked to T‑cell activation and retention. These findings suggest a potential association between TA‑HEVs and a molecular environment favorable for T‑cell recruitment. Activated TILs are increased in HCC with TA–HEVs Next, we used multicolor flow cytometry to evaluate subsets of TILs enriched in TA–HEV–present cases. This analysis was conducted in Cohort 2, which included 49 patients. Patient characteristics of this cohort were similar to Cohort 1 (Table 3). Based on double immunohistochemical staining in FFPE samples, there were nine TA–HEV–present cases and 40 TA–HEV–absent cases in Cohort 2. The two subgroups in this cohort did not differ significantly in patient characteristics (Supplemental Table S5). Figure 5A illustrates the gating strategy and shows a representative flow cytometry result from one TA–HEV–present case. Briefly, the gating strategy began with live CD45+ cells, followed by the selection of lymphocytes, singlet lymphocytes, and then T cells. CD8+ and CD4+ subsets then were identified. CD8+ T cells were analyzed based on CD45RA expression, PD-1 expression, and Tim-3 expression. CD4+ T cells were divided into effector and regulatory subsets, and cell-surface CTLA-4 expression was assessed in each subset.Table 3Patient characteristics, Cohort 2n = 49Age (years)78 (73–82)Sex Male36 (73.5%) Female13 (26.5%)HBsAg7 (14.3%)Anti-HCV antibody9 (18.4%)Child–Pugh classification A48 (98.0%) B1 (2.0%)Liver cirrhosis5 (10.2%)ASA-PS 12 (4.1%) 241 (83.7%) 36 (12.2%)Alpha-fetoprotein (ng/ml)7 (4–128)Des-gamma-carboxy prothrombin (mAU/ml)162 (39–1876)Tumor size (cm)4.0 (2.4–6.5)Number of tumors Single42 (85.7%) Multiple7 (14.3%)Histological type Well-differentiated19 (38.8%) Moderately differentiated21 (42.9%) Poorly differentiated9 (18.4%)Microscopic vascular invasion8 (16.3%)Microscopic intrahepatic metastasis1 (2.0%)Continuous values are presented as medians with IQRs, and categorical values are presented as numbers (%)Abbreviations: ASA–PS, American Society of Anesthesiologists Physical StatusFig. 5A Representative images of flow cytometry analysis and gating strategy for a TA–HEV–present case. B Comparison of immune cell proportions stratified by TA–HEV presence in Cohort 2. *p < 0.05, **p < 0.001. Error bars represent SE. Abbreviations: Lym, lymphocytes; Treg, regulatory T cells Table 3Patient characteristics, Cohort 2n = 49Age (years)78 (73–82)Sex Male36 (73.5%) Female13 (26.5%)HBsAg7 (14.3%)Anti-HCV antibody9 (18.4%)Child–Pugh classification A48 (98.0%) B1 (2.0%)Liver cirrhosis5 (10.2%)ASA-PS 12 (4.1%) 241 (83.7%) 36 (12.2%)Alpha-fetoprotein (ng/ml)7 (4–128)Des-gamma-carboxy prothrombin (mAU/ml)162 (39–1876)Tumor size (cm)4.0 (2.4–6.5)Number of tumors Single42 (85.7%) Multiple7 (14.3%)Histological type Well-differentiated19 (38.8%) Moderately differentiated21 (42.9%) Poorly differentiated9 (18.4%)Microscopic vascular invasion8 (16.3%)Microscopic intrahepatic metastasis1 (2.0%)Continuous values are presented as medians with IQRs, and categorical values are presented as numbers (%)Abbreviations: ASA–PS, American Society of Anesthesiologists Physical Status Patient characteristics, Cohort 2 Continuous values are presented as medians with IQRs, and categorical values are presented as numbers (%) Abbreviations: ASA–PS, American Society of Anesthesiologists Physical Status A Representative images of flow cytometry analysis and gating strategy for a TA–HEV–present case. B Comparison of immune cell proportions stratified by TA–HEV presence in Cohort 2. *p < 0.05, **p < 0.001. Error bars represent SE. Abbreviations: Lym, lymphocytes; Treg, regulatory T cells The TA–HEV–present cases had a significantly higher proportion of T cells among lymphocytes (82.0% for TA–HEV–present cases vs. 70.5% for TA–HEV–absent cases, p = 0.036) (Fig. 5B). The proportions of CD8+ and CD4+ T cells among T cells were comparable between the two subgroups. Among CD8+ T cells, the proportion of CD45RA− cells was significantly higher in TA–HEV–present cases (89.1% vs. 78.2% with TA–HEV absent, p = 0.036), as was the proportion of PD-1⁺ cells within CD8⁺ T cells (79.2% vs. 61.4%, p = 0.010). Of note, Tim-3⁻PD-1+ CD8+ T cells, which are considered activated CD8⁺ T cells, were significantly more prevalent in TA–HEV–present cases (66.7% vs. 53.7%, p = 0.019). The proportions of Tim-3+ CD8+ T cells and terminally exhausted Tim-3⁺PD-1+ CD8+ T cells, however, were similar between the subgroups. Among CD4+ T cells, the PD-1 positivity rate of Foxp3−CD4+ T cells was significantly higher in TA–HEV–present cases (71.7% vs. 52.5%, p < 0.001). Interestingly, although the proportions of effector CD4+ T cells and Tregs among CD4+ T cells were comparable, cell-surface CTLA-4 positivity in both effector CD4+ T cells and Tregs was significantly higher in TA–HEV–present cases (effector CD4⁺ T cells: 5.4% vs. 2.4%, p = 0.011; Tregs: 27.7% vs. 10.1%, p < 0.001). The above results indicating a quantitative increase in TILs and qualitative changes in their activation, mediated by TA–HEVs, are suggested to contribute to improved prognosis in HCC cases with TA–HEVs. Discussion This study represents the first report to highlight TA–HEVs in HCC. TA–HEVs were observed in 19.9% of resected samples of HCC, and the presence of TA–HEVs was associated with significantly better DFS. Moreover, cases with TA–HEVs had significantly higher TIL densities and an increased proportion of activated T cells among TILs. These results suggest that quantitative and qualitative alterations in TILs, mediated by TA–HEVs, might contribute to improved DFS in HCC patients. In this study, DFS was significantly better in TA–HEV–present cases and OS trended higher in TA–HEV–present cases. In the multivariate analysis for DFS, the presence of TA–HEVs was identified as an independent prognostic factor. These findings are consistent with previous studies that demonstrated an association of TA–HEVs with favorable prognosis in breast [14], gastric [16], and colorectal cancers [17]. Martinet et al. first reported the presence of TA–HEVs in human solid tumors. In that study focusing on breast cancer, TA–HEVs were observed in most cases, and breast cancer samples with high TA–HEV density also had high densities of TILs, such as CD8+ T cells and CD20⁺ B cells. Moreover, high TA–HEV status was an independent favorable prognostic factor for both DFS and OS [14]. Similarly, Park et al. analyzed over 400 cases of gastric cancer and reported that TA–HEVs were frequently observed in the cancer tissues. TA–HEV–high tumors had a high density of TILs and high TA–HEV status was an independent favorable prognostic factor for both OS and DFS [16]. In the present study, the frequency of TA–HEVs in HCC was relatively low, at about 20%, a divergence from previous reports that may be associated with the liver-specific immune suppressive microenvironment. However, in TA–HEV–present tumors, high densities of CD8+ and CD4+ TILs were observed, and as noted, the presence of TA–HEVs was identified as an independent favorable prognostic factor for DFS. These findings suggest that, as in other studied cancers, TA–HEVs in HCC may play an important role in the recruitment of TILs within the tumor microenvironment and serve as a prognostic indicator. Although the prognostic relevance of non-tumoral HEVs has not yet been reported, a previous study in HCC demonstrated that ectopic lymphoid-like structures which were indicated as non-tumoral TLSs could function as microniches for tumor progenitor cells. Hepatic lymphoid-like structures contributed to late recurrence after resection and indicated poor prognosis [27]. Another report showed that intratumoral TLSs were linked to favorable outcomes and immunotherapy response, while peritumoral TLSs would promote tumor progression [28]. Given the potential role of TA–HEVs in immune surveillance in the residual liver and influencing recurrence after HCC resection, we examined the association between the presence of non-tumoral HEVs and prognosis. Our data showed that non-tumoral HEVs were identified in 14.1% of cases, but no association with prognosis was observed. In this study, the observation period after curative resection differed from that of the previous report, and differences in patients’ clinical factors and evaluation methods for non-tumoral TLSs might influence the clinical outcomes. Nevertheless, our results suggest that the function of HEVs in non-tumoral TLSs may differ from that in TA–HEVs, suggesting functional heterogeneity of HEVs in tumor immunity, and that non-tumoral HEVs are unlikely to contribute substantially to antitumor immune responses. In the present study, cases with TLSs accompanied by TA–HEVs exhibited significantly better DFS, whereas DFS was comparable between cases with TLSs lacking TA–HEVs and those without TLSs. This suggests that the presence of TA–HEVs is necessary for the prognostic relevance of TLSs in HCC. In animal models, TA–HEV induction has been shown to be closely associated with TLS development [29]. However, the association between TLS maturation and the presence of TA–HEVs remains unclear. In HCC and several types of malignancy, it was reported that TLS maturation assessed by germinal center formation with CD21/23 signals was associated with patients’ prognosis [9, 19, 30]. Therefore, we evaluated TLS maturation by morphological features and CD21/CD23 signals and examined the relation with the presence of TA–HEVs. As a result, TA‑HEVs were more frequently observed depending on the maturation of TLSs: 34% in E‑TLSs, 48% in PFL‑TLSs and 83% in SFL‑TLSs. These findings suggest a significant correlation between TLS maturation and TA–HEV induction, consistent with previous studies indicating that TLS formation and TA–HEV induction would share regulatory mechanisms and TA–HEVs were induced at early stages of TLS development [31]. In our results, RNA-seq analysis revealed higher expression of markers in T cells and cytotoxic lymphocytes in cases with TA–HEVs, further supporting the link between TA–HEVs and increased TIL density. In addition, pathway analysis of RNA-seq data revealed upregulation of immune-related pathways. In line with their earlier report [16], Park et al. analyzed a large number of gastric cancer samples and reported that TA–HEV–high cases showed upregulation of immune-related pathways, such as positive regulation of immune system. Moreover, consistent with our findings, that study also demonstrated upregulation of dendritic cell-related genes in TA–HEV–high cases, suggesting that TA–HEVs may contribute to enhanced antigen presentation activity within the tumor. Taken together, these findings indicate the presence of a highly active immune microenvironment in HCC cases with TA–HEVs. The HCC cases with TA–HEVs in the current work had higher proportions of three TIL subsets: CD45RA− CD8+, Tim-3⁻ PD-1+ CD8+, and PD-1+ Foxp3− CD4+ T cells. PD-1 is an inhibitory receptor and often regarded as a T-cell exhaustion marker. PD-1 expression itself does not indicate irreversible dysfunction, however, and previous studies have shown that PD-1+ T cells can retain effector functions and a functionally active phenotype [32]. Tim-3 expression in PD-1+ CD8+ T cells has been recognized as an important indicator of functional state, and Tim-3+ PD-1+ CD8+ T cells exhibit severely impaired cytokine production and are considered terminally exhausted [33]. However, Tim-3−PD-1+ CD8+ T cells have been shown to maintain high cytokine-producing capacity [34]. Taken together, a population of Tim-3− PD-1+ CD8+ T cells in TA–HEV–present HCC tumors could represent activated T cell status with effector potential, leading to an immune-hot status and a favorable prognosis. The accumulation of CD45RA− CD8+ T cells observed in TA–HEV–present cases suggests a shift in differentiation status of tumor-infiltrating T cells toward a memory- or effector-like phenotype. Memory/effector T cells can mount rapid and sustained antitumor responses upon antigen re-exposure [35]. Collectively, the intratumoral recruitment of functionally active T cells and memory/effector CD8+ T cells via TA–HEVs may lead to enhanced tumor immune surveillance, thereby promoting the elimination and suppression of tumor cells and contributing to improved DFS. We found that the proportions of PD-1–positive cells among Foxp3− CD4+ and CD8+ T cells and of cell-surface CTLA-4 positive cells among effector CD4⁺ T cells and Tregs were elevated in TA–HEV–present HCC cases. These immune checkpoint molecules are key targets of immunotherapies for HCC [4, 36]. Although an association between TA–HEVs and immunotherapy response in HCC has not been reported, previous studies in human melanoma have shown higher response rates and favorable outcomes with anti-PD-1 and -CTLA-4 therapies in cases with TA–HEVs [15]. One possible mechanism involves the formation of an immune niche surrounding TA–HEVs, in which progenitor-exhausted CD8+ T cells are recruited via TA–HEVs [37], and these progenitor-exhausted cells undergo a proliferative burst and differentiate into cytotoxic, terminally differentiated CD8+ T cells upon administration of immune checkpoint inhibitors [38, 39]. Collectively, these findings suggest that TA–HEVs might serve as a predictive biomarker for responsiveness to immunotherapy in HCC. CTLA-4 in T cells is predominantly localized within intracellular vesicles and is commonly evaluated by intracellular staining in flow cytometry [40]. However, previous studies have suggested that both the quantity of cell-surface CTLA-4 and the kinetics of its expression are key determinants of the efficacy of anti-CTLA-4 immunotherapy, as only the surface-expressed fraction is accessible to therapeutic antibody binding [23]. Therefore, we evaluated cell-surface CTLA-4 expression by flow cytometry in this study. The same report demonstrated that highly immunosuppressive effector Tregs express higher levels of cell-surface CTLA-4 [23]. In the present study, the proportion of CTLA-4+ Treg cells was elevated in TA–HEV–present cases, suggesting that Tregs in these cases may possess higher immunosuppressive function. At the same time, TA–HEV-present cases also showed higher proportions of CD8+ T cells and conventional CD4+ T cells with activated phenotypes. Given that TA–HEV–present cases exhibited significantly better DFS, it is likely that these antitumor immune responses outweighed the immunosuppressive effects of CTLA-4+ Treg cells. The identification of TA–HEVs in the liver is controversial. MECA-79 is widely used as an HEV marker in various cancers, acting as a carbohydrate epitope and a ligand for L-selectin [41]. However, hepatic bile ducts also present the carbohydrate epitope of MECA-79, and bile ducts have been detected by immunohistochemistry using anti-MECA-79 antibody [24]. Indeed, in our immunohistochemical analysis, MECA-79+ CK7+ tubular structures suggestive of bile ducts were detected separately from TA–HEVs. Therefore, in this study, we identified TA–HEVs as CD31+ MECA-79+ vascular structures within TLSs based on double immunohistochemical staining with an antibody to the vascular endothelial marker CD31 and an anti-MECA-79 antibody. This approach enabled us to accurately identify TA–HEVs without misinterpreting them as bile ducts, allowing for a precise evaluation of their association with prognosis and TILs. Recent work and the present study have shown that TA–HEVs play a critical role in facilitating lymphocyte trafficking into the tumor [26]. Preclinical studies have indicated that therapeutic interventions such as immune checkpoint inhibitors and anti-angiogenic agents can induce TA–HEV formation, enhancing lymphocyte infiltration and antitumor responses [15, 42]. In immunosuppressive tumors such as HCC, induction of TA–HEVs by these strategies can represent potential interventions to shift the tumor status from “immune cold” to “immune hot.” Moreover, recent advances in spatial transcriptomics and in situ imaging technologies have enabled detailed analysis of the molecular characteristics and microenvironmental interactions of TA–HEVs [37]. These emerging tools are expected to reveal the heterogeneity and plasticity of TA–HEVs and open new ways for developing TA–HEVs as therapeutic targets or predictive biomarkers. In HCC, functional studies and interventional research using these technologies may contribute to the establishment of novel TA–HEV-based strategies for immune modulation. This study has several limitations. First, this study was conducted on a small cohort from a single institution. In particular, the limited number of HCC cases with TA–HEVs restricted the flow cytometry and RNA-seq analyses to a small sample size. Nevertheless, the present study was able to demonstrate consistent findings. Second, although we examined the relationship between TA‑HEVs and TILs in human clinical samples, we did not address the temporal dynamics or underlying mechanisms. Thus, our findings remain descriptive, and it is unclear whether TA‑HEVs actively induce changes in TILs or instead arise as a consequence of TIL alterations driven by other factors. Further studies using improved models are warranted to advance this research. In conclusion, we have demonstrated the clinical significance of TA–HEVs using resected human HCC specimens. Cases with TA–HEVs exhibited significantly better DFS. Furthermore, these cases showed a significantly higher density of TILs and an increased proportion of activated T cells within TILs. These findings suggest that TA–HEVs may contribute to tumor immune activation and, consequently, to improved prognosis in HCC. Supplementary Information Below is the link to the electronic supplementary material.Supplementary file1 (DOCX 34 KB) Supplementary file1 (DOCX 34 KB)