Primary liver cancer represents one of the most prevalent malignant tumors worldwide and poses a significant global health challenge. Globally, it ranks as the sixth most commonly diagnosed cancer and the third leading cause of cancer-related death (1). According to 2022 global epidemiological data, approximately 865,000 new liver cancer cases and 758,000 deaths occur annually (2). As reported by China’s National Cancer Center, 368,000 new cases occurred in China in 2022, accounting for approximately 42% of the global burden and establishing the country as the most heavily impacted by liver cancer (3). A recent 2025 global projection predicts that without intervention, new annual cases will surge to 1.52 million by 2025, while deaths will rise sharply to 1.37 million (4). HCC comprising 90% of liver cancer cases, primarily arises from risk factors including hepatitis B and C infections. Notably, non-infectious etiologies such as non-alcoholic steatohepatitis linked to metabolic syndrome or diabetes are becoming increasingly prevalent causal factors in Western nations (5).Clinical management faces substantial hurdles due to limitations in early diagnosis, specifically the persistent lack of highly sensitive and specific serum biomarkers for screening (6), alongside the tendency for ultrasound and imaging modalities to miss small lesions against cirrhotic backgrounds (7). Consequently, most patients present with advanced disease at diagnosis, rendering them ineligible for curative surgery. Although tyrosine kinase inhibitors (e.g., sorafenib, lenvatinib) and immune checkpoint inhibitors(ICIs) targeting PD-1/CTLA-4 have advanced treatment paradigms, high tumor heterogeneity in HCC restricts their efficacy to subsets of patients. Frequent development of therapeutic resistance further leaves many without viable options (8, 9). The intrinsically aggressive nature of HCC which is marked by high malignancy, recurrence, and metastasis, compounds these challenges, establishing it as a major therapeutic challenge globally. Nevertheless, the rapid evolution of multi-omics technologies and intensified research into novel early diagnostic biomarkers offer promising avenues (10). Metabolites, protein signatures, genetic markers, and circulating cells represent emerging candidates to enhance early diagnostic accuracy (11, 12). Since its advent, scRNA-seq has enabled the resolution of gene expression at individual cell resolution, overcoming tissue-level heterogeneity limitations. This breakthrough allows unprecedented exploration of cellular behavior, underlying mechanisms, and tissue (13). Tumor heterogeneity is the core to HCC treatment resistance which is now being deconvoluted through scRNA-seq. By delineating the heterogeneity of the HCC tumor TME, this technology provides unprecedented insights into tumor growth, metastasis, and intercellular crosstalk mechanisms (14). Collectively, these advances offer new hope for transformative therapeutic strategies in HCC. In recent years, researchers have been attempting to summarize the roles and clinical value of various cell types in the tumor microenvironment of liver cancer (15–18). Standing on the shoulders of giants, we have synthesized a large number of review studies and experimental articles to further explore the application value of scRNA-seq in HCC. We systematically integrated the mainstream subclustering of five types of immune cells in scRNA-seq studies within the tumor microenvironment. We emphasize that the TME is not composed of static cell types, but rather a continuous spectrum of functional states that continuously shift under tumor stress, metabolic stress, and treatment exposure. Based on this foundation, we utilized a new analytical framework of marker genes - cellular developmental/functioning states - resistance mechanisms/therapeutic value translation to provide new perspectives for precise and personalized tumor interventions.

NK cells, as innate lymphocytes, serve as primary responders initiating immune activation. NK cells play pivotal roles in host defense and immune surveillance. These cells directly eliminate tumor cells not only through cytotoxic effector functions, but also by acting as regulatory lymphocytes, which secrete cytokines and interact with both innate and adaptive immune cells (32). Traditional classification stratifies NK cells by CD56 expression density into cytotoxicity-dominant CD56^dim and cytokine-secreting CD56^bright subpopulations. Recent studies using scRNA-seq refine this framework by delineating developmental trajectories and functional states, identifying critical subsets including cytotoxic NK cells, regulatory NK cells, and memory-like NK cells (33, 34).

Perforin-1 (PRF1), encoded by the PRF1 gene, is a pore-forming protein stored in secretory vesicles. PRF1 acts as a key cytotoxic molecule in NK cells and plays an essential role in their tumor-killing function (35). During target cell engagement, PRF1 is released and forms pores in the target cell membrane, disrupting intracellular calcium balance. This enables Granzyme B (GZMB) and other cytotoxic agents to enter the cytoplasm, triggering an enzymatic cascade that ultimately induces tumor cell death (36). Given the critical function of cytotoxic NK cells in directly lysing cancer cells, their antitumor efficacy becomes severely limited when NK cells exhibit an exhausted phenotype. Thus, research on stimulating sustained secretion of PRF1 and GZMB by NK cells holds significant importance (37). Adoptive cell therapy using chimeric antigen receptor NK cells offers unique advantages over chimeric antigen receptor T cells therapy. It avoids triggering severe cytokine release syndrome, graft-versus-host disease, or neurotoxicity. It also permits diverse cell sourcing and enables cytotoxic enhancement through genetic engineering. Its favorable safety and efficacy profiles have been consistently validated across multiple hematologic and solid tumors (38).

Regulatory NK cells exhibit high XCL1 gene expression, with the chemokine XCL1 being a critical component in the NK cell-mediated antitumor signaling network (39). As the primary source of XCL1, NK cells precisely recruit conventional type 1 dendritic cells (cDC1) into the TME via XCL1 secretion. This initiates a coordinated “NK-DC-T cell” immune attack axis, where cDC1 infiltration density directly dictates T-cell activation efficacy and subsequent tumor cell clearance (40). Functioning as cellular communication hubs, NK cells deficiency renders the TME an “immune-deserted island”. It permits tumor immune escape through failed immune cell activation. Notably, XCL1 is a signature gene of immunoregulatory NK cells which is aberrantly expressed on certain tumor cells, potentially contributing to immune evasion and chemotherapy resistance (41). Given the pivotal role of the XCR1-XCL1 axis in antitumor responses, modulating NK-cDC1 crosstalk to boost cDC1 recruitment. Intratumoral XCL1 injection enhances infiltration of antigen-specific CD8⁺ T cells and NK cells, thereby suppressing tumor growth (42). XCL1 expression is significantly positively correlated with the number of tumor-infiltrating CD8⁺ T cells and PD-L1 expression, potentially inducing PD-1/PD-L1 interactions and dysfunction of CD8⁺ T cells through the XCL1-XCR1 axis, thereby predicting response to anti-PD-1/PD-L1 therapy (43). Spatial transcriptomics has confirmed the colocalization characteristics of XCL1⁺ CD8⁺ T cells with immune cells, suggesting that targeting this cell population may enhance the efficacy of immunotherapy for hepatocellular carcinoma (44).

Memory-like NK cells highly express the immunostimulatory receptor NKG2C (encoded by KLRC2), a type II transmembrane protein (45). NKG2C transmits activation signals upon binding HLA-E on target cells, promoting NK cell degranulation and cytokine secretion to enhance infected cell elimination. NKG2C⁺ NK cells demonstrate potent antibody-dependent cellular cytotoxicity (ADCC) facilitating sustained clearance of infected cells. Chronic infections drive clonal expansion of NKG2C⁺ NK cells, forming persistent “memory-like” populations with augmented effector functions (46). The competitive binding of NKG2C and the inhibitory receptor NKG2A to HLA-E dynamically regulates NK cell reactivity balance (47). In the TME, HLA-E overexpression often impairs T/NK cell effector function. To address this, novel chimeric receptors termed “NKG2A/C-swapping receptors” have been engineered. These receptors confer enhanced and specific cytotoxicity against tumors with moderate-to-high HLA-E expression while sparing normal tissues, representing a promising therapeutic strategy (48).

Current NK cell classification relies predominantly on CD56 and CD16, lacking resolution for the complexity of NK cell development and stress-responsive states. Broader tissue coverage remains limited. Integrating spatial transcriptomics can comprehensively map tissue-specific NK cell distribution and function. Future efforts should leverage epigenomic and transcriptomic datasets within spatiotemporal multi-omics approaches to dissect real-time microenvironmental regulation of NK plasticity, thereby refining subtype classification and advancing precision immunotherapies.

DCs represent the most potent professional antigen-presenting cells in the human immune system and serve as the exclusive activators of the naïve T cells. Functioning as critical bridges between innate and adaptive immunity, DCs play pivotal roles in initiating and sustaining T-cell antitumor responses. DCs involve in antigen presentation in tumor and lymph nodes, primary T-cell activation, and maintenance of T-cell survival and effector function within the TME. Deficiency or functional impairment of DCs severely limits antitumor immunity, thereby facilitating tumor progression (49). DCs are broadly classified into conventional DCs (cDCs, encompassing cDC1 and cDC2) and plasmacytoid DCs (pDCs).

cDCs constitute central hubs of the cancer immunity cycle. cDC1s, as master cross-presenting cells, drive CD8⁺ T cell responses against malignancies via MHC-I–restricted tumor antigen presentation,. While cDC2s predominantly prime CD4⁺ T-cell responses through MHC-II and modulate Th1/Th2/Th17/Treg immunity (50). cDC1s uniquely express XCR1 which is the sole receptor for XCL1 enables their recruitment to inflammatory/neoplastic sites (51). Engineered T cells overexpressing XCL1 have been shown to increase DC abundance, significantly inducing antigen spreading and endogenous polyclonal T-cell responses to amplify antitumor immunity (52).

Signal regulatory protein alpha (SIRPα), a hallmark of cDC2s, interacts with CD47 to maintain immune homeostasis by restraining DC overactivation (53). SIRPα⁺ DC2s reduce cross-presentation efficiency and impair CD8⁺ T-cell responses through CD47 signaling (54). In the TME, they mature into regulatory DCs that phagocytose apoptotic tumor cells via the AXL-SIRPα axis, upregulate PD-L1, suppress CD8⁺ T cell function, and promote Treg differentiation which critically enforcing immune tolerance. Tumor cells overexpressing CD47 evade immune clearance by engaging SIRPα on DCs to inhibit phagocytosis (55). Anti-SIRPα antibodies blocking this axis restore DC function, enhance antigen presentation and tumor phagocytosis by macrophages. Dual targeting of SIRPα and PD-L1 synergistically activates T cells, offering novel combination immunotherapy strategies (56).

pDCs, historically named for their plasma cell-like morphology, pDCs are specialized DC subsets that function as primary producers of type I interferons (IFN-I) and modulate antitumor immunity by regulating NK/T-cell activity in the TME (57). Blood dendritic cell antigen 2 (BDCA2), a C-type lectin receptor exclusively expressed on pDCs, marks this population (57). BDCA2⁺ pDCs exhibit robust proinflammatory functions through IFN-α/TNF secretion, they can also drive immune suppression by inducing Tregs via the IDO1 pathway to dampen Th1/Th17 responses (58). Current BDCA2 research prioritizes autoimmune diseases, focusing on IFN-I overproduction as a key pathophysiology in systemic lupus erythematosus. Therapies crosslinking BDCA2 potently inhibit IFN-I production by pDCs, presenting a promising treatment strategy for systemic lupus erythematosus (59, 60). Beyond systemic lupus erythematosus, BDCA2 also emerges as a compelling therapeutic target for other immune disorders and hematologic malignancies.

Collectively, advances in single-cell profiling of DC subsets illuminate key mechanisms of immune regulation and therapeutic resistance in HCC.

T cells constitute a critical lymphocyte subset within the TME, playing a central role in antitumor immunity and serving as the foundation for most immunotherapies. CD4⁺ T cells encompass subsets including TH1, TH2, TH17, TFH, and Treg. They are all differentiated from naïve CD4⁺ T cells while each exhibiting distinct effector functions essential for shaping adaptive immune responses (61).

CC chemokine receptor 7 (CCR7), a lymph node-homing receptor expressed on dendritic cells (DCs) and T cells, guides antigen-presenting DCs and T-cell subsets to lymphoid organs via interactions with ligands CCL21/CCL19 to initiate immune responses (62). CCR7 is highly expressed in naïve CD4⁺ T cells (63), with elevated levels observed in patients with lymph node-metastatic cancers (64). CCR7 facilitates lymphatic metastasis across malignancies as CCR7-expressing tumor cells mimic immune cell behavior by following chemokine gradients toward lymph nodes (65). Anti-CCR7 antibodies are being explored to inhibit CLL migration to lymph nodes, though clinical efficacy remains under investigation (66).

CXCR3 is a G protein-coupled receptor enriched in Th1 cells (67). CXCR3 mediates targeted migration and immune responses through ligand binding on endothelial cells, playing key roles in infections, autoimmune disorders, and tumor immunity (68). The CXCR3 axis comprises ligands CXCL9, CXCL10, and CXCL11 and shows strong cancer relevance (69). Notably, CXCR3 drives immunogenic cell death in HCC, where its expression positively correlates with T-cell infiltration. HCC patients with high CXCR3 expression display improved responses to immunotherapy (70).

CCR4, initially characterized as a TH2-selective chemokine receptor, is now recognized across multiple T-cell subsets (71). Its interaction with the atypical ligand CXCL12 drives cellular migration and proliferation in physiological and cancerous contexts (72). CXCR4 overexpression in over 20 cancer types is associated with poor prognosis and cancer stemness (73), while CCR4 antagonists suppress distant metastasis in liver cancer models (74).

The interleukin-23 receptor (IL-23R), primarily expressed on TH17 cells, transmits signals enhancing TH17 differentiation while promoting IL-22 secretion from CD4⁺ T cells (75, 76). IL-23R signaling induces multiple highly proinflammatory cytokines and serves as a key pathogenic driver in chronic inflammation and autoimmune diseases (77). Expanding beyond their immunological roles, TH17 cells exhibit tumor-promoting properties (78). During acute hepatic inflammation, these cells contribute to inflammatory cascades and induce HCC recurrence through dual mechanisms: facilitating cancer cell migration/invasion and enhancing tumor stemness (79).However,IL-23R involvement in this process warrants investigation.

CXCR5 features selective expression on follicular helper T (TFH) cells. Together with its ligand CXCL13, this receptor-ligand pair governs TFH development and functional specialization (80). TFH cells are indispensable for germinal center formation and maintenance (81), thereby modulating humoral immunity. The CXCL13-CXCR5 axis contributes to tumor progression through multifaceted mechanisms involving direct tumor stimulation, immune cell recruitment, and TME modulation (82).

FOXP3 serves as the master transcriptional regulator for regulatory T cells (Tregs), orchestrating genes controlling their differentiation and immunosuppressive activity (83, 84). FOXP3⁺ Tregs employ diverse mechanisms, including secretion of anti-inflammatory cytokines (TGF-β, IL-10, IL-35) and perforin/granzyme-mediated cytolysis (85). FOXP3⁺ Tregs suppress antitumor immunity and facilitate tumor escape (86). Clinically, elevated FOXP3⁺ Treg infiltration correlates with metastatic progression in HCC patients (87). Therapeutic inhibition enhances IFN-γ production in CD8⁺ T cells and restricts tumor growth (88). A recent paradigm-shifting approach demonstrates that FOXP3⁺ -overexpressing chimeric antigen receptor T cells exhibit potent antitumor efficacy, which lack immunosuppressive functions (89).

As an immune checkpoint molecule, CTLA-4 is expressed on activated T cells and Tregs. It both inhibits T-cell activation and delivers essential co-stimulatory signals to Tregs (90). Functionally, it suppresses immune responses through intrinsic negative regulation of effector T cells and extrinsic modulation via Treg-mediated suppression (91, 92). CTLA-4 overexpression compromises antitumor immunity to enable immune escape (93) In HCC, CTLA-4 blockade with neutralizing antibodies enhances tumor clearance (94), while genetic silencing augments antitumor responses (95). Therapeutically, the dual immune checkpoint regimen combining CTLA-4 inhibitor ipilimumab with PD-1 blocker nivolumab has achieved landmark status, which becomes China’s first approved first-line dual-immunotherapy for HCC.

CD8⁺ T cells serve as the terminal effectors of adaptive immunity, functioning as primary cytotoxic executors in antitumor immunity (96). Research focuses on their effector mechanisms, with increasing attention to CD8⁺ T cell exhaustion in tumor responses.

LEF1 demonstrates elevated expression in naïve CD8⁺ T cells (97). Collaborating with TCF1, it provides sustained surveillance of CD8⁺ T cell identity and function by promoting T-lineage genes while suppressing non-T-lineage programs (98). Notably, LEF1 enhances self-renewal capacity, drug resistance, dedifferentiation, and invasiveness in HCC cells (99). For instance, it confers lenvatinib resistance by advancing Epithelial-mesenchy maltransition, migration, and invasion in HCC cells, thereby establishing LEF1 as a novel therapeutic target to overcome acquired lenvatinib resistance (100).

Chemokine receptor CX3CR1 express across multiple immune cell types.CX3CR1⁺CD8⁺ T cells exhibit potent antitumor effector functions (101) and manifest exceptional cytotoxicity (102).CX3CR1 expression displays a graded transcriptional pattern reflecting CD8⁺ T cell differentiation status, with levels positively correlating with increased GZMB, perforin 1, and granzyme A expression (103). While their antitumor efficacy is validated in diverse cancers, further experimental studies are required to delineate their role in HCC.

CD69, a C-type lectin characteristically expressed on CD8⁺ tissue-resident memory T cells, inhibits T-cell egress from tissues and participates in tissue retention alongside other lectins (104). Tissue-resident memory CD8⁺ T cells provide infection protection at barrier sites (105) and local immune defense against tumor rechallenge (106). Studies indicate that CD69 expression modulates CD8⁺ T-cell exhaustion. Within the TME, CD69⁺CD8⁺ T cells progressively lose effector functions, adopting an exhausted phenotype. Blocking CD69 or its signaling pathways enhances their antitumor activity (107). Notably, CD69⁺ tissue-resident memory CD8⁺ T cells expressing unique signature genes were identified in HCC patient tumors, exhibiting correlation with patient survival outcomes (108).

The granzyme K gene (GZMK) is enriched specifically in innate-like lymphocytes and certain CD8⁺ T-cell subsets. GZMK⁺ T cells typically represent central and effector memory T-cell populations (109). GZMK⁺ T cells display inflammatory potential, correlating positively with plasma IL-6, TNFα, and IL-8 levels (110). Recent research reveals GZMK can instigate inflammation through complement activation mediation (111). In HCC, GZMK⁺CD8⁺ T cells demonstrate cytotoxicity (14), and higher GZMK expression associates with favorable prognosis across multiple HCC cohorts (112).

Lymphocyte-activation gene 3 (LAG3) is expressed on CD4⁺ T cells, CD8⁺ T cells, and Tregs (113), functioning as an immune checkpoint inhibitory receptor (114). Combined LAG-3 and PD-1 signaling drives T-cell exhaustion while impeding IFN-γ-mediated autocrine antitumor immunity (115). LAG3 engagement with MHC class II and other ligands transduces T-cell inhibitory signals leading to dysfunction, which notable upregulation on exhausted T cells within the TME (116).LAG3 correlates with poor prognosis in HCC patients when expressed at high levels (117). Whilst anti-LAG3 monotherapy yields suboptimal efficacy, combination strategies particularly with PD-1 inhibitors show superior therapeutic promise (118). Indeed, dual LAG-3/PD-1 blockade demonstrates manageable safety and objective efficacy in advanced HCC treatment (119).

Layilin (LAYN), a transmembrane protein functioning as a C-type lectin. It participates primarily in cellular adhesion and modulates diverse including immune cell activation and T-cell subset differentiation, exerting pro-tumorigenic roles in HCC (120). LAYN is upregulated in tumor-infiltrating CD8⁺ T cells and Tregs within HCC and associates with their suppressive functions (97). LAYN-overexpressing CD8⁺ T cells display hallmark exhaustion features and diminished antitumor capacity. Patients with high LAYN levels exhibit poorer overall survival (121).

T-cell subset profiling remains a sustained research focus. Despite considerable efforts on CD8⁺ T cells due to their prominence in tumor immunity, this field nears investigational saturation with diminishing novel discoveries. Emerging paradigms now emphasize exploring untapped CD4⁺ T cell subsets. Notably, Tregs—whose discoverers received the 2025 Nobel Prize in Physiology or Medicine,show exceptional promise as a prime candidate for next-generation investigations. Their discovery underscores the transformative potential of regulatory T cells in cancer immunology.

Landmark advances in macrophage functional studies originated from the 1990s-established M1/M2 polarization model. In this model, IFN-γ/LPS-induced classically activated M1 macrophages exhibit pro-inflammatory/antitumor functions, while IL-4/IL-13-induced alternatively activated M2 macrophages display anti-inflammatory/pro-repair activities. This paradigm, inspired by Th1/Th2 differentiation concepts, significantly advanced our understanding of macrophage roles in tumorigenesis and infections (122), revealing polarization’s centrality in immune regulation. However, scRNA-seq and spatial transcriptomics have exposed limitations in this binary model. It oversimplifies in vivo macrophage heterogeneity, depicts context-dependent polarization states, and fails to incorporate developmental origin diversity. The M1/M2 polarization framework has essential differences in conceptual assumptions and research methods compared to single-cell scRNA-seq. The M1/M2 polarization model originates from in vitro stimulation experiments. It assumes that macrophages differentiate into two relatively stable, opposing functional states under specific stimuli (e.g., IFN-γ/LPS or IL-4/IL-13) to explain immune regulation during inflammation or repair processes (123). However, this framework implies the premises of discrete states and single dominant signals, making it difficult to reflect the true cellular landscape shaped by multiple overlapping signals, metabolic reprogramming, and developmental origin differences in complex in vivo microenvironments (124). In contrast, scRNA-seq is a data-driven, hypothesis-free single-cell analysis method. By simultaneously measuring the expression levels of thousands of genes at single-cell resolution, it directly reveals the continuous spectrum of macrophage functional states, dynamic transition trajectories, and spatial correlations within tissues (125, 126). At the methodological level, M1/M2 relies on a limited number of markers and artificially defined polarization conditions, while scRNA-seq characterizes a multidimensional, highly heterogeneous network of transcriptional states through clustering analysis, pseudotime inference, and cell communication modeling. Contemporary research integrates multi-dimensional approaches including single-cell subpopulation definition, spatial functional mapping, and pseudotime trajectory inference (127, 128), providing insights beyond static classifications. Current representative categories include inflammatory macrophages, angiogenic macrophages, interferon macrophages, regulatory macrophages, lipid-associated macrophages, and tissue-resident macrophages (124).

CCL3 (macrophage inflammatory protein-1α, MIP-1α), a CC-chemokine engaging CCR1/CCR5/CCR9 receptors, is highly expressed by macrophages upon inflammatory stimulation (129). This pro-inflammatory chemokine actively recruits monocytes/macrophages to inflammatory sites (130). In human HCC, CCL3 orchestrates antitumor immunity, including enhancing phagocytic activity, upregulating MHC molecules for antigen presentation, recruiting/activating T cells, and reprogramming the TME to restore adaptive immunity (131). Therapeutic CCL3 administration enhances TME immunogenicity. Mice with CCL3-enriched tumors showed delayed growth post anti-PD-1 monoclonal antibody treatment, suggesting combinational αPD-1/CCL3 strategies may augment clinical efficacy (132).

Secreted phosphoprotein 1 (SPP1), a pleiotropic phosphoglycoprotein, regulates immune responses while promoting tumor proliferation, invasion, and therapy resistance (133). Although produced in multiple organs, SPP1 expression is restricted to osteoblasts, fibroblasts, macrophages, dendritic cells, lymphocytes, and monocytes. Cancer cells also express SPP1, and elevated circulating SPP1 or tumor SPP1 correlates with poor prognosis across cancers (134). SPP1⁺ macrophages promote tumor angiogenesis and hypoxic microenvironments by secreting extracellular matrix proteins (e.g., CD44, MMPs) and activating HIF-1 signaling (122). In CRC, SPP1⁺ macrophage/FAP⁺ fibroblast interactions through TGF-β and IL-17 signaling exacerbate pro-angiogenic and immunosuppressive milieus. SPP1⁺ TAMs across tumor types overexpress angiogenesis-related genes (e.g., VEGF pathway members) and correlate with adverse outcomes (135). Compared with normal tissues, tumors exhibit significantly enriched SPP1⁺ macrophage infiltration. High SPP1 expression induces resistance to anti-PD-L1 immunotherapy (122) and potentiates liver cancer stemness, growth, migration, and chemoresistance (136). Mechanistically, SPP1^-CD44 interactions activate multiple exhaustion-associated pathways in CD8⁺ T cells, with SPP1 transcriptionally regulating T cell exhaustion (137), thereby facilitating tumor immune evasion. High expression of SPP1 is significantly associated with poor prognosis in various cancers, including colorectal cancer and liver cancer. Anti-SPP1 combined with PD-1 inhibitors (e.g., NCT05230901) can reverse the immunosuppressive functions of TAMs (138). SPP1⁺ TAMs secrete MMP9 and collagen, promoting tumor invasion and metastasis, and the expression level of SPP1 can predict resistance to PD-1 therapy in esophageal squamous cell carcinoma (139). Despite the enormous potential of SPP1 as a therapeutic target in cancer, fibrosis, and other diseases, and with several drug discovery projects underway, there are currently no approved SPP1-targeted drugs globally.

ISG15 (Interferon-Stimulated Gene 15) encodes the first identified ubiquitin-like protein modifier, functioning both as a free intracellular/extracellular molecule and as a post-translational modifier during ISGylation (140). During viral infections, IFN-I induction is a key component of innate immunity which triggers ISG15 expression, which participates in cancer cell apoptosis via IFN production (141). While ISG15 upregulation enhances IFN-mediated macrophage phagocytosis and antiviral activity (142), tumor-secreted ISG15 acts as a TME factor that induces M2-like polarization, promotes tumor progression, and suppresses cytotoxic T-lymphocyte responses (143). Consequently, ISG15 exhibits context-dependent roles in cancer. It functions as either a tumor suppressor or oncogene by altering distinct biological pathways across cancer types. Elevated ISG15 expression in HCC cell lines and clinical specimens correlates with cell cycle progression, cancer cell proliferation/migration, and poor 5-year survival (144).

Arginase 1 (Arg1), encoded by the ARG1 gene, hydrolyzes arginine as a rate-limiting enzyme of the urea cycle. Produced by tumor-infiltrating myeloid cells including macrophages, granulocytes, DCs, immature progenitors. High Arg1 activity depletes L-arginine in TME, suppressing T-cell proliferation and reducing effector T-cell populations (145). Macrophage polarization toward M2 phenotypes involves metabolic reprogramming in which Arg1 plays key roles (146). Arg1-expressing macrophages inhibit CD4⁺ T-cell proliferation/cytokine production (147), thereby supporting tumor growth. Notably, tumor cells supply arginine to macrophages to enhance polyamine biosynthesis, driving immunosuppressive polarization and CD8⁺ T-cell dysfunction (148). Targeting Arg1 effectively reprograms macrophages; Arg1-derived peptide vaccines activate Arg1⁺ CD4⁺ T cells, promoting inflammatory macrophage phenotypes and generating antitumor immune responses to inhibit growth (149).

Apolipoprotein E (APOE), a plasma cholesterol-transport protein and Alzheimer’s disease (AD) risk factor, has been predominantly studied in AD pathology (150). Highly expressed in macrophages, macrophage-derived APOE exerts immunomodulatory functions beyond lipid transport (151). Macrophage APOE critically regulates extracellular vesicle production, influencing T-lymphocyte proliferation, activation, and IFN-γ secretion (152). Enriched in M2 macrophages, exosomal APOE mediates macrophage-gastric cancer crosstalk to promote metastasis (153). APOE⁺ macrophages interact with exhausted CD8⁺ T cells, attenuating immune checkpoint inhibitor efficacy, whereas APOE blockade enhances immunotherapy responses (154). Pan-cancer analyses reveal differential APOE expression between tumors and normal tissues, correlating with clinical phenotypes and serving as a prognostic biomarker across cancers (155). APOE is a strong genetic risk factor associated with Alzheimer’s disease, and APOE genotyping can be used for risk stratification and early diagnosis (156). APOE may serve as a biomarker for HER2-negative breast cancer, aiding in predicting responses to immunotherapy. APOE⁺ TAMs promote the expression of genes related to lipid metabolism and are associated with an immunosuppressive tumor microenvironment and combination immunotherapy may enhance anti-tumor immune responses (157). While research on APOE has primarily focused on Alzheimer’s disease, its role in tumor biology remains highly controversial and faces multiple methodological challenges.

LYVE-1 (Lymphatic Vessel Endothelial Hyaluronan Receptor-1), a marker of lymphangiogenesis, is expressed on lymphatic endothelial cells and macrophages (158). Associated with tumor lymphangiogenesis and TME modulation (159), macrophage-expressed LYVE-1 mediates leukocyte docking/transmigration (160) and cancer cell-endothelial adhesion. In melanoma metastasis studies, LYVE-1 blockade enhanced pro-inflammatory states in pre-metastatic livers, altering TME to reduce hepatic metastasis (161). LYVE-1⁺ macrophages represent a protumorigenic, anti-inflammatory subset promoting hyaluronan remodeling in TME, and their depletion delayed mammary tumor growth (162). Anti-LYVE-1 monoclonal antibodies inhibit lymphangiogenesis, suppressing both primary tumor formation and metastasis, highlighting LYVE-1 as a promising therapeutic target (163).

Although the M1/M2 paradigm remains influential, growing evidence urges its abandonment in single-cell contexts (164, 165). This binary model is established from in vitro polarization—oversimplifies tissue macrophage states, ignoring their continuous, dynamic, and multidimensional in vivo landscapes. Persisting with rigid classifications obscures critical biological insights and impedes precision therapeutics. As scRNA-seq becomes increasingly accessible, multidimensional interrogation of functional subsets will provide comprehensive blueprints for tumor immunotherapy.

B cells, long recognized as core components of the adaptive immune system, play pivotal roles within the TME. Beyond antibody secretion, antigen presentation, and cytokine production, they modulate immune responses to influence tumor progression (166). Based on developmental trajectories and functional states, major B-cell subsets include naïve B cells, memory B cells, germinal center (GC) B cells, plasma cells, and regulatory B cells (Bregs) (167, 168).

TCL1A, a marker of naïve B cells, regulates signaling during early B-cell development; its downregulation signifies B-cell maturation (169). TCL1A promotes proliferation by activating stem cell expansion pathways (170), enhancing KI67 expression and modulating the cell cycle (reducing G1 phase, increasing S/G2 phases). It also upregulates CR2 expression on B cells (154), enabling efficient antigen presentation (155). Notably, TCL1A promotes tertiary lymphoid structure formation (171), which correlates with positive immunotherapy responses and survival outcomes (172–174), underscoring its value as a therapeutic target.

CD27, a member of the TNF receptor superfamily and key memory B-cell marker, interacts with its ligand CD70 to regulate humoral immunity and tolerance (175). CD27 sustains memory B-cell survival by activating NF-κB signaling to inhibit apoptosis and guiding migration to secondary lymphoid organs for survival factors. Upon antigen re-exposure, CD27-CD70 engagement activates NF-κB and MAPK/ERK pathways, driving rapid memory B-cell activation and proliferation. Subsequently, CD27 signaling induces plasma cell differentiation via downstream transcriptional activators, facilitating high-affinity antibody secretion during secondary immune responses (176). The CD27-CD70 axis represents a promising target across hematologic and solid tumors: the agonistic anti-CD27 antibody varlilumab shows clinical efficacy (177), soluble CD27 predicts anti-PD-1 monotherapy response (178), and serves as an HCC risk biomarker (179).

The transcriptional repressor BCL6, predominantly expressed in GC B cells, is essential for GC formation and B-cell differentiation (180). By repressing differentiation-related genes, it maintains GC B cell proliferation/survival and governs TFH-dependent somatic hypermutation/class-switch recombination, ultimately generating high-affinity antibody-secreting plasma cells and memory B cells (181). BCL6 also acts as an HBV promoter suppressor in hepatocytes, enhancing chemokine production and immune cell liver infiltration (182). However, BCL6 is oncogenic: it epigenetically represses pro-apoptotic genes via promoter binding to promote cancer cell survival (183), making its targeted degradation therapeutically valuable (184).

MZB1 (Marginal Zone B and B1 cell-specific protein 1), upregulated during plasma cell differentiation (185), encodes an ER-resident co-chaperone critical for antibody secretion capacity (186). MZB1 deficiency impairs plasmablast migration (via β1-integrin dysregulation) and terminal plasma cell differentiation (187). Overexpressed in hematologic malignancies, MZB1 sustains malignant B-cell protein folding and chemoresistance in multiple myeloma (188, 189). As a key regulator of ER homeostasis and antibody production, MZB1 targeting may modulate plasma cell function to mitigate inflammation (190).

Regulatory B cells (Bregs) suppress immunity primarily through IL-10 production (191). CD38high Bregs secrete IL-10/TGF-β to convert naïve TH cells into FoxP3⁺ Tregs (192), while suppressing antigen-specific T-cell effector functions and Th17 differentiation (193),collectively facilitating tumor immune evasion. CD38 governs multiple inflammatory processes (migration, adhesion, phagocytosis, antigen presentation/release) and is a therapeutic target in hematologic malignancies (194). Anti-CD38 antibodies (e.g., isatuximab, daratumumab) induce antibody-dependent cytotoxicity, with clinical efficacy validated in multiple myeloma, NK-cell lymphoma, and CD19^- B-cell malignancies (195, 196).

While B-cell biology in hematologic tumors is well-characterized, their roles in solid tumors (breast, lung, colorectal cancers) remain limited. B-cell subsets exhibit context-dependent plasticity across tumor types, disease stages, and treatments. Current insights represent only partial functional delineations; deeper investigations into TME-specific mechanisms and subset-targeted immunomodulation are warranted.

In the cascade of antitumor immune responses, which spanning antigen presentation, phagocytosis, cytotoxicity, and antibody secretion. NK cells, DCs, T cells, macrophages, and B cells each execute distinct roles in tumor cell elimination. However, this process is frequently hijacked by tumors to facilitate their growth and immune evasion. Beyond these, neutrophils, endothelial cells, and fibroblasts also exert critical functions within the TME. Critically, these cell types are not independent entities; they engage in intense intercellular crosstalk through diverse signaling networks. Thus, elucidating cell-cell communication is paramount for decoding TME complexity. Research on cell interactions supported by scRNA-seq is no longer limited to average signals from specific cell types (197); it can precisely identify individual cells’ gene expression, protein secretion, and signaling characteristics, thereby highlighting cellular heterogeneity (198). The integration of spatial transcriptomics further reveals the spatiotemporal dynamics and functional diversity of these interactions. Researchers have used scRNA-seq and spatial transcriptomics to discover that MRC1⁺ tumor-associated TAMs interact with cancer-associated fibroblasts through WNT5A and HGF signaling to promote the metastasis of hepatocellular carcinoma (124). Microfluidic chip analyses combined with scRNA-seq secretion profiling showed that only a subset of CD8⁺ T cells secretes IFN-γ after contacting tumor cells, while most cells enter an exhausted state due to PD-1/PD-L1 signaling (199).Additionally, researchers found through ligand-receptor analysis that NK cells recruit cDC1 to the tumor core by secreting XCL1 and CCL5, forming a positive feedback loop that enhances anti-tumor immunity (200). In summary, single-cell technologies have transformed the study of cell interactions from a blurry average relationship to a finely regulated network with spatial, temporal, and state resolution. While not all discussed cell subtypes have been extensively studied in HCC, mechanisms observed in other cancers may exhibit comparable or opposing effects in HCC which finds awaiting further experimental validation. Meanwhile, we have listed the subgroups and marker genes discussed in detail in Table 1 and some potential subgroup marker genes that are not mentioned in Table 2. scRNA-seq enables precise characterization of transcriptional profiles that define cellular states. Yet “signature gene” expression is inherently multifaceted: genes may be expressed across multiple cell types or exhibit functional reversals during different developmental stages of the same lineage. Under pathological stresses, which including tumor hypoxia, inflammatory stimulation, metabolic stress, or therapeutic exposure—virtually any functional gene may be induced, silenced, or reprogrammed, thereby sculpting novel cellular states with unique phenotypic and biological significance. The exploration of functional subtypes seems to hold near-limitless potential (Supplementary Figure S2).