Genetic and epigenetic alterations in HCC cells can contribute to an immune-deserted microenvironment. Notably, activation of the Wnt/β-catenin signaling pathway is observed in 30–50% of HCCs and is strongly associated with the exclusion of T cells from the tumor (10,39). β-catenin activation in tumor cells downregulates the expression of C-C motif chemokine ligand (CCL) 4, which impairs DC recruitment and thereby prevents the effective priming of anti-tumor T cells. This creates an immunologically cold tumor niche. Clinically, HCC tumors with catenin β-1 gene mutations that activate β-catenin rarely respond to ICIs, making this pathway a recognized driver of immune resistance (40).

Other oncogenic pathways frequently altered in HCC also contribute to immunosuppression; for example, MYC oncogene overexpression, which is detected in ~50% of HCCs, upregulates PD-L1 in tumor cells and modulates cell metabolism to favor an immunosuppressive milieu (39). Similarly, the loss-of-function of phosphatase and tensin homologs or activation of PI3K-AKT signaling can promote an immunoresistant environment by increasing the expression of inhibitory molecules and recruiting suppressive myeloid cells (41).

HCC cells actively upregulate immune checkpoint ligands and other immunoinhibitory molecules. HCCs frequently express PD-L1 on their surfaces of their cells, especially those with upregulated MYC expression or enriched in progenitor-like traits. PD-L1 can directly bind to PD-1 on T cells, thereby promoting T-cell exhaustion (16). In addition, some HCC cells express galectin-9, an immunosuppressive lectin that interacts with TIM-3 on T cells, further contributing to T-cell exhaustion (42).

The immune infiltrate in HCC is skewed toward cell types that sustain an immunosuppressive, cold TME (28). Tregs are often enriched in the blood and tumor tissues of patients with HCC, where they suppress antitumor immunity by secreting IL-10 and TGF-β, and by consuming IL-2, which limits the proliferation of effector T cells (43,44).

TAMs of the M2 phenotype produce high levels of IL-10 and prostaglandins, and express immune checkpoint molecules such as PD-1, which suppress phagocytosis and innate immune responses (45). These TAMs also recruit Tregs via CCL22 and promote tissue remodeling, which impairing or inactivating effector T cell functions. Myeloid-derived suppressor cells (MDSCs), a heterogeneous population of immature monocytes and neutrophils with immunosuppressive activity, are also present in HCC. MDSCs release arginase and inducible nitric oxide synthase, which metabolically starve T cells of L-arginine and produce nitric oxide, thereby causing T-cell dysfunction (44).

The TME of HCC is enriched with immunosuppressive cytokines (40,46). TGF-β, often abundant due to cirrhosis and CAF activity, is a key driver of immune exclusion. It inhibits the proliferation and effector functions of T cells and natural killer (NK) cells, while promoting the differentiation of Tregs and M2 macrophages (46). A high-TGF-β signature in HCC is associated with an immune-excluded phenotype and resistance to immune checkpoint blockade (40).

HCC tumors are frequently hypoxic due to their rapid growth and abnormal vasculature (47). Hypoxia induces hypoxia-inducible factor-1α, which upregulates adenosine production and VEGF expression, both of which suppress antitumor immunity. Adenosine accumulates in the TME via CD39/CD73 ectonucleotidase activity in cancer and stromal cells. It potently inhibits T- and NK-cell activity through A2A adenosime receptors, while promoting the expansion and functional activation of Tregs and the polarization of macrophages to the M2 phenotype (48).

The unique immune environment of the liver plays a central role in the immune evasion of HCC. Under normal conditions, the liver is constantly exposed to food antigens and microbial products from the gut via the portal circulation. Therefore, non-immunogenic (tolerogenic) mechanisms have evolved in the liver to prevent unnecessary immune activation. Kupffer cells, which are liver-resident macrophages, and liver sinusoidal endothelial cells (LSECs) constitutively express inhibitory molecules, including PD-L1 and Fas ligand, and secrete anti-inflammatory cytokines such as IL-10 and TGF-β, which induce T-cell tolerance (49–51). Naïve T cells encountering antigens in the liver often become anergic or differentiate into Tregs (52,53).

HCC exploits these tolerogenic pathways: Kupffer cells may present tumor antigens in a tolerogenic manner, while LSECs can induce the deletion of tumor-reactive T cells, thereby preventing effective antitumor immune priming (54). In addition, cirrhotic livers exhibit expanded populations of immunosuppressive myeloid and stellate cells, contributing to an immunosuppressive fibrotic microenvironment, even before malignant transformation (55).

During HCC development, immune editing can occur, whereby the adaptive immune system initially eliminates highly immunogenic tumor cell clones, leading to the selection of tumor cells that are less detectable by T cells (56). Over time, HCC tumors may exhibit the downregulation or loss of certain tumor-associated antigens. For example, tumor variants that no longer express common antigens, such as α-fetoprotein (AFP) or glypican-3 (GPC3), may evade T-cell- or antibody-based therapies targeting these antigens (57). Similarly, the loss of major histocompatibility complex (MHC) class I molecule expression due to β2-microglobulin mutations or defects in the antigen presentation machinery can impair cytotoxic T-cell recognition, although this may potentially increase the susceptibility of the tumor to NK cells (58,59).

ICIs include anti-PD-1 antibodies, such as nivolumab and pembrolizumab, anti-PD-L1 antibodies, such as atezolizumab and durvalumab, and anti-CTLA-4 antibodies, such as ipilimumab and tremelimumab. These agents form the foundation of immunotherapy strategies in HCC, and are able to activate T-cell responses if primed T cells are already present in the TME. However, single-agent ICIs have shown limited overall response rates in HCC (~15%), indicating that numerous tumors lack preexisting active T cells (60,61). Therefore, ICIs are commonly combined with other agents that can initiate or amplify antitumor immunity. One such approach is dual checkpoint blockade, such as PD-1 plus CTLA-4 inhibition, where CTLA-4 blockade primarily acts on the lymphoid organs to expand T-cell clonal diversity and reduce Treg-mediated suppression, whereas PD-1/PD-L1 blockade reactivates exhausted T cells in the tumor (62).

In HCC, the combination of nivolumab and ipilimumab has demonstrated manageable toxicity and a notably higher response rate compared with nivolumab monotherapy, indicating that dual blockage can convert a subset of immunologically cold tumors into responsive ones (6). Similarly, a combination of tremelimumab and durvalumab was evaluated in the phase III HIMALAYA trial, using a single priming dose of tremelimumab followed by ongoing durvalumab treatment. This regimen induced an immunogenic boost in patients with unresectable HCC that improved survival compared with that of patients treated with sorafenib (63).

Abnormal blood vessels in HCC contribute to immune evasion by promoting hypoxia and serving as physical barriers to immune cell trafficking. Anti-angiogenic therapies, including the anti-VEGF antibody bevacizumab and multi-kinase inhibitors, such as sorafenib, lenvatinib, cabozantinib and regorafenib, not only inhibit tumor angiogenesis but also modulate the TME (37). VEGF blockade has immunomodulatory effects; it can normalize tumor vasculature, thereby enhancing lymphocyte infiltration and alleviating the VEGF-mediated suppression of DCs and T cells (37).

The combination of PD-L1 inhibitor atezolizumab with bevacizumab in the IMbrave150 phase III trial significantly improved response as well as overall survival (OS) and progression-free survival (PFS) in patients with unresectable HCC compared with those of patients treated with sorafenib (60). This exemplifies how combining immune checkpoint blockade with angiogenesis inhibition can convert a cold tumor into a more immune-active one. However, the results of the IMbrave050 trial, a follow-up study of adjuvant treatment with atezolizumab plus bevacizumab after resection or ablation in high-risk HCC showed that the benefit in recurrence-free survival was not sustained, and the effect on OS remained non-significant (64). The IMbrave150 trial succeeded because advanced HCC presents abundant neoantigens and ongoing immune activity that respond to PD-L1 and VEGF blockade, whereas the adjuvant IMbrave050 setting involved minimal residual disease and immune quiescence, limiting checkpoint and angiogenesis inhibition efficacy in preventing recurrence. It is suggested that bevacizumab enhances T-cell access to the tumor and reduces VEGF-induced immunosuppression, thereby allowing atezolizumab-activated T cells to function more effectively (65).

Traditional locoregional treatments for HCC, including TACE, radioembolization with yttrium-90 and thermal ablation, including radiofrequency or microwave ablation, can exert immunological effects that may be exploited to ignite antitumor immunity (66). These treatments cause tumor cell death and promote the release of tumor antigens and damage-associated molecular patterns into the environment, a process known as immunogenic cell death (66).

TACE induces ischemic and chemotherapeutic stress in tumor cells, potentially triggering a surge in neoantigen and proinflammatory cytokine release that attracts immune cells (67). Evidence suggests that TACE or ablation can lead to transient increases in T-cell activation and clonal expansion targeting tumor antigens (68). Radiation therapy upregulates MHC class I expression on tumor cells and increases chemokine levels for T-cell recruitment, while also inducing an abscopal effect, whereby localized radiation leads to systemic antitumor immunity (69).

Clinical trials have combined ICIs with locoregional therapies. For example, tremelimumab combined with ablation has been shown to enhance T-cell responses and induce objective tumor regression in patients with advanced HCC (67). Ongoing trials, such as EMERALD-1 and −2, are evaluating the PD-L1 inhibitor durvalumab in combination with TACE for patients with intermediate-stage HCC (70,71). Selected clinical trials and combination immunotherapy strategies for advanced HCC are summarized in Table II (6,60,67,72–77).

Another approach to convert cold tumors into more immunogenic ones is to deplete or reprogram immunosuppressive cells in the TME. Colony stimulating factor-1 receptor (CSF-1R) inhibitors, such as pexidartinib, can reduce TAM numbers or alter TAMs from an immunosuppressive M2-like phenotype toward a more proinflammatory, antitumor M1-like phenotype (78). In HCC mouse models, the inhibition of CSF-1/CSF-1R signaling, which is critical for monocyte-to-macrophage differentiation and survival, decreases TAM infiltration and promotes a shift toward a pro-inflammatory environment, thereby sensitizing tumors to anti-PD-1 therapy (79).

C-C motif chemokine receptor 2 (CCR2) antagonists also show promise by preventing the recruitment of monocytes with high expression of lymphocyte antigen 6 complex, locus C, a key surface glycoprotein marker used to distinguish functional subsets of monocytes, that are precursors for MDSCs and TAMs (80). The inhibition of CCL2-CCR2 signaling reduces the accumulation of intratumoral macrophages and leads to T-cell-dependent tumor regression in preclinical HCC models (81).

Tregs are more challenging to target selectively, but low-dose cyclophosphamide or anti-CD25 antibodies have been used in solid tumors such as ovarian, breast cancer, prostate cancer, renal cell carcinoma, and melanoma to transiently deplete Treg populations (82).

Finally, TGF-β pathway inhibitors are another class of compounds used to remove a major suppressive influence. Inhibitors such as galunisertib, a TGF-β receptor kinase inhibitor, or fresolimumab, an anti-TGF-β antibody, have been investigated in patients with HCC (83,84). Although these agents have not yet been tested in phase III trials in combination with ICIs, preclinical analysis suggests that blocking TGF-β signaling enhances the penetration of T cells into tumors and is a promising mechanism for use in combinations that aim to convert cold tumors into hot ones (85).

Oncolytic viruses (OVs) are engineered or naturally occurring viruses that selectively replicate in and lyse cancer cells while stimulating an antitumor immune response (86). This lytic process releases tumor antigens and viral pathogen signals, effectively transforming the tumor into a vaccine depot. Several OVs have been evaluated for the treatment of HCC.

Talimogene laherparepvec (T-VEC), a modified herpes simplex virus type-1 that encodes granulocyte-macrophage (GM)-CSF, has been approved for melanoma treatment and tested via intratumoral injection in liver tumors (87). Early trials combining T-VEC with ICIs have shown increased local immune cell infiltration in HCC and demonstrated the feasibility of generating a localized autitumor immune response (88).

Pexastimogene devacirepvec (Pexa-Vec; JX-594), an oncolytic vaccinia virus expressing GM-CSF, has also been evaluated in HCC. A phase II trial showed signs of improved survival in a subset of patients (89), In addition, a phase III trial (PHOCUS), evaluating Pexa-Vec combined with sorafenib, was conducted in advanced HCC (86). The PHOCUS phase III trial showed that adding Pexa-Vec to sorafenib does not improve overall survival in advanced HCC, leading to early study termination, though Pexa-Vec was well tolerated with manageable safety (90).

Active immunization strategies aim to stimulate the immune system to recognize HCC-specific antigens, thereby enhancing T-cell priming against the tumor. Although past vaccine trials in HCC targeting antigens such as AFP (91), GPC3 (92) or multi-peptide mixtures (93) have shown only modest success, they all have demonstrated immunogenicity, indicating that T cells targeting these antigens can be expanded in patients. A summary of emerging immunotherapeutic approaches for HCC is provided in Table III (21,23,35,67,79,89,94–101). Certain approaches-particularly cancer vaccines, oncolytic viruses, and certain adoptive cell therapies-have shown evidence of inducing tumor-specific immune responses in clinical or preclinical studies, whereas others (metabolic, microbiome, and TAM-targeting strategies) are still at the investigational or early-trial stage and remain to be validated for immunogenic effects.

Combining cancer vaccines with ICIs or other TME modulators is a rational strategy. For example, DC vaccines pulsed with HCC tumor antigens can generate a wave of tumor-specific T cells, and concurrent PD-1 blockade can prevent their exhaustion, allowing them to penetrate and exert cytotoxic activity within tumors (102).

Adoptive cell therapy supplies effector cells directly to patients. In HCC, efforts have focused on engineering T cells to target tumor antigens. Chimeric antigen receptor (CAR) T cells specific for GPC3, an oncofetal antigen expressed in HCC cells, have been tested in early-phase trials (102). A phase I study of CAR-GPC3 T cells in advanced HCC showed some partial responses; however, it also revealed that the immunosuppressive TME limited CAR-T persistence and function (100).

A preclinical study used murine and xenograft HCC models to show that CAR-T cells engineered to secrete IL-12 upon antigen engagement could remodel the TME and enhance systemic antitumor activity (103). Researchers isolated and characterized high-affinity TCRs specific for AFP peptides presented by HLA-A*02:01. These engineered TCR-T cells recognize AFP-expressing HCC cells and exhibited potent, antigen-specific cytotoxicity in vitro and in mouse xenograft models. Follow-up studies using the same AFP TCR construct led to a first-in-human phase I trial (NCT03971747) in HCC patients (104). These TCR-T cells can recognize intracellular antigens presented on MHC molecules, potentially broadening targetable proteins beyond surface markers.

An emerging strategy to enhance antitumor immunity involves modulating epigenetic or metabolic pathways to render tumor and immune cells more immunogenic. Epigenetic drugs, including DNA methyltransferase inhibitors such as decitabine or histone deacetylase (HDAC) inhibitors such as entinostat, can restore the expression of silenced tumor-associated antigens and increase MHC molecule presentation on cancer cells, thereby enhancing their recognition by T cells (105).

In addition to their effects on tumor cells, epigenetic modulators can also alter the differentiation of immune cells. For example, HDAC inhibition has been shown to promote a pro-inflammatory macrophage phenotype and reduce the accumulation of MDSCs, effectively boosting immune responses (106). Furthermore, in HCC models, combining HDAC inhibitors with PD-1 blockade increased intratumoral T-cell infiltration and led to tumor regression (106).

The gut microbiome has a major influence on systemic immunity and has been shown to affect the response to ICIs in melanoma, non-small cell lung cancer and renal cell carcinoma. Therefore, researchers are interested in modulating the gut microbiota to shift the TME in HCC to a hotter immune state (95). Certain bacterial metabolites in the gut can enhance antitumor immunity. Fecal microbiota transplantation from ICI-responsive patients has been shown to improve therapeutic responses in melanoma and may hold promise for application in HCC (95).

The first wave of phase III trials of single-agent ICIs for advanced HCC has yielded mixed results. In the CheckMate-459 trial, nivolumab showed durable responses in a subset of patients when used as a first-line treatment compared with sorafenib, but it did not significantly improve OS in the primary analysis (107). Similarly, in the second-line setting of the KEYNOTE-240 trial, pembrolizumab yielded a modest improvement in survival compared with placebo in the second-line setting, however this was not significant (108).

The IMbrave150 trial demonstrated that combining atezolizumab with bevacizumab significantly improved OS and the objective response rate (ORR) in treatment-naïve patients with advanced HCC. The median OS was 19.2 months in the combination arm compared with 13.4 months in patients treated with sorafenib, while the ORR was ~ 27% compared with ~12%, respectively (60). Numerous responders showed substantial tumor shrinkage, and some achieved durable remission (60).

In the phase III HIMALAYA trial, a single high dose of tremelimumab (300 mg) combined with chronic dosing of durvalumab achieved a superior OS compared with that of sorafenib in front-line advanced HCC (63). This Single Tremelimumab Regular Interval Durvalumab (STRIDE) regimen capitalized on an initial wave of broad T-cell activation from CTLA-4 blockade, followed by sustained PD-L1 blockade. The median OS was 16.4 months compared with 13.8 months in patients treated with sorafenib. A subset of patients treated with the STRIDE regimen achieved long-term survival >2 years (63).

In the CheckMate-040 cohort 4 phase II study, three different regimens combining nivolumab with ipilimumab were evaluated in patients who had progressed on sorafenib. An ORR of 32% and median OS of 22 months was obtained in the arm with higher ipilimumab dosage (6). Approximately 8% of patients achieved a CR (6).

A pilot trial combining the CTLA-4 inhibitor tremelimumab with subtotal radiofrequency ablation in patients with advanced HCC whose tumors were suitable for partial ablation reported that 26% of patients had a partial response in non-ablated lesions, and disease control was achieved in 89% of patients (67). Notably, an increase in intratumoral CD8⁺ T cells and a reduction in viral load in hepatitis B virus (HBV)-infected patients were observed after treatment, indicating a systemic immune effect (67). These findings support the concept that local tumor destruction can synergize with checkpoint blockade.

Retrospective analyses also suggest that patients receiving radiotherapy or TACE shortly before or during nivolumab therapy have improved response rates compared with those receiving nivolumab alone (109). For example, one study found an abscopal response in ~20% of patients who received radiotherapy for a single lesion while receiving nivolumab treatment compared with <10% in those treated with nivolumab alone (110).

A first-in-human study of T cells engineered with a TCR against an AFP peptide presented by HLA-A2 in patients with AFP-positive HCC showed that the treatment was safe and resulted in transient reductions in serum AFP levels and tumor shrinkage in certain patients (112). Although no objective responses were observed according to RECIST criteria, one patient exhibited a measurable decrease in tumor burden that does not meet the threshold for a partial response (PR), and others achieved stable disease (112).

A randomized phase II trial of patients with early-stage HCC compared a tumor lysate-activated autologous DC vaccine with a control group of patients treated with best supportive care following curative resection or ablation (113). The vaccinated group experienced a significantly longer median time to recurrence than the control group (36 vs. 25 months, respectively) and a higher recurrence-free survival at 1 year (75.6 vs. 61.1%, respectively) (113).

Several mouse model studies have directly demonstrated the benefits of reprogramming the HCC microenvironment:

A study of mice with diet-induced NASH revealed that PD-1 blockade did not protect against liver cancer and instead promoted the incidence and progression of HCC due to activated, but dysfunctional, CD8⁺ T cells producing TNF-α. However, when those CD8⁺ T cells were depleted or TNF-α was neutralized, checkpoint blockade efficacy improved (15).

In a chemically-induced HCC mouse model, it was revealed that tumor-derived OPN recruits TAMs through the CSF-1 pathway. The blockade of CSF-1R not only reduced TAM infiltration but also led to increased CD8⁺ T-cell activity and responsiveness to anti-PD-L1 therapy (79).

A study of lenvatinib demonstrated that the inhibition of FGFR4, a protein that is overexpressed in some HCC tissues, not only affected tumor cell growth but also reduced intratumoral Treg accumulation in mouse xenografts and PD-L1 expression in tumor cells. This dual effect was associated with improved anti-PD-1 antibody efficacy and tumor rejection in mouse models (114).

In an orthotopic HCC model in mice, PD-1 gene-deleted CAR T cells targeting GPC3 outperformed conventional CAR T cells, as they were more persistent and infiltrated the tumors more deeply. In addition, these modified CAR T cells resulted in complete tumor eradication in a higher proportion of the mice (35).

Preclinical findings have revealed new targets, including CSF-1R (78), CXCR2 (115) and adenosine (116), and confirmed mechanisms such as Wnt-mediated exclusion, that can be therapeutically targeted (116).

As our understanding of HCC immune biology deepens, there is growing recognition that not all patients should receive the same immunotherapy regimen. Future treatment algorithms will likely incorporate biomarkers to stratify patients. For example, tumors with Wnt/β-catenin mutations or an immune-excluded phenotype may be prioritized for trials evaluating Wnt pathway inhibitors or other T-cell-recruiting strategies (39).

A robust immune score for HCC could be developed, analogous to the Immunoscore used in colorectal cancer (117). This could help to determine the required intensity of immunotherapy. Liquid biopsies, including circulating tumor DNA and immune cell profiling assays, may enable the early assessment of whether a tumor is becoming hotter or remains immunologically cold, allowing for adaptive treatment adjustments.

While PD-1/PD-L1 and CTLA-4 are the current targets, several other inhibitory pathways are being explored in clinical trials and could be integrated into HCC treatment (5). Antibodies targeting lymphocyte-activation gene 3 (LAG-3), known as relatlimab, were recently approved by the US Food and Drug Administration for the treatment of melanoma and are of interest in HCC, as LAG-3 is upregulated in exhausted T cells in the liver (118).

T-cell immunoreceptor with Ig and ITIM domains (TIGIT) is another checkpoint protein expressed on T and NK cells; anti-TIGIT antibodies, such as tiragolumab, combined with anti-PD-1 therapy are in trials for HCC following promising results in other cancer types (119). These emerging checkpoint inhibitors may help to revive exhausted T cells in HCC when PD-1 blockade alone is inadequate.

Building on the success of doublet regimens, such as atezolizumab + bevacizumab and durvalumab + tremelimumab, the next wave of studies is testing triplet combinations to further convert cold tumors into a hot state. For example, the COSMIC-312 trial has evaluated a triplet of PD-1 + CTLA-4 + tyrosine kinase inhibitors, namely nivolumab + ipilimumab + cabozantinib, to simultaneously target T cells, Tregs and tumor angiogenesis (120). The COSMIC-312 phase III trial reported that cabozantinib + atezolizumab significantly improved progression-free survival vs. sorafenib (6.8 vs 4.2 months; HR 0.63) but did not improve overall survival (121). However, the nivolumab + ipilimumab + cabozantinib triplet remains under investigation, with no published efficacy results yet for this regimen.

However, as toxicity and cost increase with each added agent, future research must identify which patient subsets truly require multilayered therapy. Sequential approaches, such as an induction phase followed by a maintenance phase, rather than concurrent triple therapy, may also mitigate toxicity.

Although most immunotherapy trials have focused on advanced HCC, there is a strong rationale for applying these strategies in earlier stages of the disease, when immune function is less impaired and the tumor burden is lower. Neoadjuvant immunotherapy prior to surgery is being investigated, and small studies administering ICIs or combinations prior to resection have shown that some patients achieve significant tumor necrosis and robust pathologic responses, which might translate to a lower recurrence risk (40).

In the adjuvant setting, after curative resection or ablation, the goal is to eradicate residual disease and prevent recurrence. The CheckMate-9DX phase III trial of adjuvant nivolumab did not meet the primary endpoint of recurrence-free survival (122), possibly due to trial design considerations or insufficient treatment duration. However, other adjuvant trials are ongoing, including the KEYNOTE-937 trial of pembrolizumab and the EMERALD-2 trial of durvalumab ± bevacizumab (123,124).

As already noted, NASH-related HCC has a suppressive immune milieu that may require tailored therapeutic strategies (15). One future approach is to treat underlying liver conditions in parallel with the cancer itself. In NASH, this could involve therapies that reduce hepatic inflammation, such as farnesoid X receptor agonists, acetyl-CoA carboxylase inhibitors or anti-IL-1β agents, administered in conjunction with immunotherapy, to prevent the non-productive activation of T cells. For HBV-associated HCC, continuing antiviral therapy is crucial, and additional strategies, such as therapeutic HBV vaccines or TCR-based approaches targeting HBV antigens, may boost antitumor immunity, given that some HCC tumor cells express HBV-derived proteins.

Accessibility of the liver via the hepatic artery allows for innovative strategies for the delivery of immunotherapies directly to the tumor or liver, thereby maximizing local immune activation effects while minimizing systemic toxicity. Ongoing trials are currently examining the locoregional delivery of IL-12 plasmids (125), toll-like receptor 9 agonists (126), and CAR T cells through hepatic artery infusion (127). Nanotechnology-based approaches may enable nanoparticles carrying immune stimulants, such as stimulators of IFN gene agonists, or small interfering RNAs targeting immune checkpoints, to be delivered selectively to liver tumors (96). Such approaches can convert the tumor site into an immune-reactive zone without exposing the whole body to high cytokine or drug levels, thereby potentially reducing systemic side effects.

As immunotherapy strategies are intensified to convert cold tumors into hot ones, the risk of collateral damage to normal liver tissue increases. Future protocols will require robust monitoring procedures for immune-related adverse events and potentially prophylactic measures. For example, patients with cirrhosis are at risk of decompensation if immunotherapy triggers autoimmune hepatitis (128,129). Gut microbiome manipulation is being considered as a means of improving efficacy and also mitigating toxicity, since certain microbiome profiles are associated with the risk of colitis (130,131). Researchers are also exploring selective immunosuppressants capable of controlling toxicity without completely compromising antitumor immunity (132,133).

As data from genomic, transcriptomic, imaging and clinical sources accumulate, artificial intelligence and machine learning will likely play a role in the identification of features that define cold and hot tumors and in guiding treatment selection. For example, deep learning applied to histological images can quantify immune infiltrates and their spatial distribution, thereby potentially predicting outcomes (115). Multiomics analysis may identify new therapeutic targets, for example, by identifying that a certain chemokine is the key factor limiting T-cell infiltration in a subset of patients, which could be targeted by drugs. Systems biology approaches may be used to model complex interactions within the HCC immune ecosystem, simulate the performance of a given combination and guide the rational design of treatment regimens.