HCC progression is known to be associated with HBV or HCV infection and liver cirrhosis. However, evidence from the CheckMate 040 and KEYNOTE-224 trials indicated that viral load or immune responses to HBV/HCV may not necessarily influence T cell activation and subsequent antitumor activity (14,15). Furthermore, the results of a meta-analysis revealed that neither HBV nor HCV affected the tumor immune microenvironment, and the presence or absence of viral infection was not an effective criterion for the selection of patients for programmed death 1 (PD-1)/programmed death-ligand 1 (PD-L1) immunotherapy (16).

Obesity and being overweight are considered to be risk factors for numerous diseases, including cancer (17). Obesity caused by a high-fat diet impairs CD8⁺ T cell infiltration and function, which alters the immune microenvironment in mice and enhances tumor growth (18). In contrast, another study revealed that patients with advanced HCC that had higher body mass indices (BMI; >25) appeared to have an improved prognosis following immunotherapy (19).

In clinical practice, the serum AFP level represents a primary indicator used in diagnosis and for monitoring the effectiveness of liver cancer treatment (20). The AFP levels are increased in ~two-thirds of patients with HCC (21). The expression levels of certain immune checkpoint proteins, such as SIGLEC15, CTLA4, CD274, PDCD1LG2, PDCD1, TIGIT, LAG3 and HAVCR2, have been revealed to differ with regards to the AFP level (22). It has been suggested that AFP could be used as a prognostic biomarker for HCC immunotherapy. For example, Spahn et al (23) observed that baseline AFP concentrations <400 µg l^-1 before the start of treatment were associated with increased rates of PR or CR and reduced rates of progressive disease (PD). However, in the CheckMate 459 trial, patients with high baseline AFP levels (>400 ng/ml) had an increased overall survival (OS) (11). The objective response rate (ORR) was revealed to be positively associated with the early stages of AFP reduction therapy and PD-1 blockade, while progression-free survival (PFS) and OS were also increased (2,24). Therefore, the combination of AFP with other serum markers deserves further investigation to improve diagnostic accuracy. For example, previous studies have indicated that the C-reactive protein (CRP) and AFP in immunotherapy (CRAFITY) score, which combines CRP with AFP, can be used to predict treatment outcomes and treatment-associated adverse events in patients with HCC undergoing immunotherapy (25,26). However, there is still disagreement over whether AFP can serve as a prognostic biomarker for immunotherapy (25,27,28).

Blood inflammatory biomarkers are both affordable and useful for the early identification of disease. It has been suggested that a neutrophil-lymphocyte ratio (NLR) ≥5 and a platelet-lymphocyte ratio (PLRs) ≥300 are independent prognostic factors for OS, predicting reduced OS, PFS, ORR and an increased risk of mortality in patients receiving immunotherapy (29,30). Similarly, a multicenter study revealed that the NLR could predict PFS in patients with unresectable HCC treated with Atezolizumab plus Bevacizumab, particularly in patients with modified albumin-bilirubin grade 1 or 2a (31).

Jeon et al (32) revealed that the numbers of classical monocytes (such as CD14⁺CD16) increased on day 7 in patients with durable clinical benefit compared with that in patients with non-durable clinical benefit. The CRP level, an indicator of inflammation, has also been revealed to have good prognostic value in lung and renal cell cancer (33-35). The baseline CRAFITY score, developed by Scheiner et al (26), has been demonstrated to be effective for the assessment of patients receiving immunotherapy. Specifically, the median OS was revealed to be 27.6, 11.3 and 6.4 months in the CRAFITY-high (2 points), CRAFITY-intermediate (1 point) and CRAFITY-low (0 points) groups, respectively, and the best radiological response ratio [CR/PR/stable disease (SD)/PD] is stratified based on the CRAFITY score. This use of the score was also supported by a retrospective study conducted in Japan where the OS and PFS of 297 patients that received Atezolizumab and Bevacizumab treatment were associated with AFP and CRP (25). However, the prediction model is currently only applicable to patients that receive Atezolizumab and Bevacizumab; additional validation is required for other immunotherapy medications.

According to recent studies, the gut microbiota serves an important role in the development and occurrence of liver cancer (36-38). The underlying mechanism involves the gut-liver axis and is associated with dysbiosis, intestinal permeability and bacterial metabolites. Dysbiosis and intestinal permeability make it easier for bacterial metabolites to reach the liver. Bacterial products such as lipopolysaccharides (LPS) can cause inflammation and cancer in the liver (39,40). In addition, Toll-like receptor 4 (TLR-4), which is widely distributed on the surfaces of various liver cells and has been demonstrated to mediate hepatic carcinogenesis, is the specific recognition receptor for LPS (41). In a study by Chung et al (36) the stools of eight antibiotic-treated patients were collected for microbiota analysis. Patients receiving Nivolumab demonstrated no alterations in the diversity and composition of their gut microbiota. However, a skewed Firmicutes/Bacteroidetes ratio and a low Prevotella/Bacteroides ratio were revealed to predict a poor immunotherapy response in patients with liver cancer, while the presence of Akkermansia species suggested a positive prognosis (36). Another study on 167 patients with hepatobiliary cancer treated with immunotherapy, revealed that a number of bacteria, such as Lachnospiraceae bacterium-GAM79, were associated with an improved OS and PFS after treatment, while other bacteria, such as Veillonella, were associated with an increased risk of immune-associated side effects (37). There is also an association between the diversity of the gut microbiota and the levels of aspartate aminotransferase and alanine aminotransferase, which reflect liver function (38). Stool samples from patients that responded to anti-PD-1 therapy contained a greater taxonomic abundance compared with those of non-responders. The characterization of the dynamic changes in the gut microbiome can be useful for an earlier prediction of anti-PD-1 treatment outcomes in HCC. With the rapid development of microbial multi-omics, analysis of the gut microbiota has potential as a predictive biomarker for liver cancer immunotherapy. It has been reported that fecal microbiota transplantation from donors that achieved CR/PR on long-term anti-PD-1 therapy to patients failed to respond to immunotherapy can increase the intra-tumoral lymphocyte infiltration (42).

ICIs may be immunogenic and recognized by the human immune system, which could lead to the induction of the humoral immunity and subsequent adverse ADA responses (43). Different monoclonal antibodies are associated with different rates of ADA development, with Atezolizumab having the highest rate (~30%) compared with others (5-10%) (44). ADAs may affect the pharmacokinetics and pharmacodynamics of therapeutic antibodies, and may even neutralize the therapeutic antibodies (45). A cohort study by Kim et al (46) reported that increased ADA levels at the second Atezolizumab injection (day 1 of chemotherapy treatment cycle 2) may be associated with poor clinical outcomes. Reducing Atezolizumab exposure in patients with advanced HCC, Atezolizumab and Bevacizumab administration and an established ADA level >1,000 ng/ml can accurately predict the curative effect (46). Anti-Atezolizumab antibody-positive patients did not demonstrate a reduction in the frequency or severity of adverse events (44). However, a meta-analysis of 11 clinical trials, based on studies using Atezolizumab monotherapy or combination therapy, demonstrated that unadjusted descriptive analyses could not identify a clear association between the ADA status and the frequency or severity of adverse events. Furthermore, any ADA impact was not driven by neutralizing activity (47). The most distinctive feature of ADA assays is their lack of accurate quantification, as there is no reliable calibration reference standard for ADAs (48). Currently, there is no effective method for predicting which drugs may cause ADAs. Table II provides a brief overview of the host-associated biomarkers that are used for HCC immunotherapy.

PD-1 is an immunosuppressive transmembrane protein that is expressed on the surface of cells such as T, B and myeloid cells. By binding to PD-L1, it inhibits T cell activation and proliferation, negatively stimulates T cells, blocks the T cell receptor, and negatively impacts how the immune system combats cancer (49,50). Inhibition of PD-1/PD-L1 prevents the interaction between PD-1 on T cells and PD-L1 on tumor cells, thus, restoring the T cell-mediated antitumor immune response (51). However, anti-PD-1/PD-L1 therapies are only effective in 20-40% of patients (52). Numerous studies have demonstrated that the expression level of PD-L1 on immune and tumor cells is associated with the anti-PD-1 treatment response in HCC (14,15,53,54). However, these studies varied in their detection techniques and methods used to measure PD-L1, so there is no universal standard for the detection and quantification of PD-L1(55). The most commonly used methods for measuring PD-L1 are the tumor proportional score (TPS) and the combined positive score (CPS) (56).

The KEYNOTE-244 study retrospectively analyzed the association between PD-L1 expression levels and response to Pembrolizumab treatment, finding a treatment response to Pembrolizumab when PD-L1 was quantified using CPS but not when it was quantified using TPS (15). However, a study on the response to Nivolumab demonstrated different outcomes. Patients that tested positive for PD-L1 (TPS ≥1%) had an increased median OS (28.1 months) compared with those that tested negative for PD-L1 (median OS of 16.6 months) (13). Recently, a meta-analysis of nine cohort studies (seven PD-L1 and three PD-1) demonstrated that PD1/PDL-1 was a marker of poor survival rate regardless of OS, HR, CI, disease-free survival (DFS) and other evaluation methods (54). High PD-1/PD-L1 expression levels were associated with aging, multiple tumors, high α-fetoprotein levels and an advanced Barcelona Clinic liver cancer stage (14,53). In addition, PD-L1, as measured by CD274 (a PD-L1 messenger RNA) expression levels in the IMbrave150 trial, were revealed to be increased in patient with CR/PR compared with that in patients with SD/PD. Patients with high CD274 levels also demonstrated an increased PFS compared with those with low expression levels (54). However, PD-L1 expression levels are influenced by various factors. PD-L1 can be induced by IFN-γ, hypoxia or TLR-mediated pathways (57). Tumor heterogeneity and the tumor interstitium were observed to be the primary causes of inconsistent outcomes, followed by differences in detection methods (58). Thus, the value of the PD-L1 expression level as a predictive biomarker for immune checkpoint blockade therapy in HCC has been reduced.

The CTNNB1 gene encodes the intracellular signaling transducer β-catenin, which is essential for embryonic development, cell fate determination, proliferation and migration (59). One of the key signaling pathways that control liver regeneration, homeostasis and tumorigenesis is the Wnt/β-catenin cascade (60,61). In a mouse model of HCC, activation of this pathway promoted immune evasion and conferred resistance to anti-PD-1 therapy (62,63). Similar outcomes were observed in liver cancer. Harding et al (64) reported that all 10 patients with mutations in components of the Wnt-β-catenin pathway demonstrated PD and a reduced median survival rate compared with patients without mutations. This implies that the Wnt-β-catenin pathway is a marker of immunotherapy sensitivity (62). Additionally, patients with HCC with mutations in CTNNB1 were revealed to have increased OS and PFS compared with patients with no mutations in CTNNB1. Thus, CTNNB1 may serve as an independent prognostic factor in HCC following immunotherapy (65,66).

Another dysregulated signaling pathway is the transforming growth factor-β (TGF-β) pathway which is involved in inflammation, fibrogenesis and immunomodulation in the HCC microenvironment (67). Increased TGF-β signaling may lead to T cell exhaustion through the upregulation of PD-1 signaling, while inhibition of TGF-β signaling may increase the anti-tumor immunity in HCC (68). Studies using mouse models have indicated that a combination of blocking TGF-β signaling and anti-PD-L1 antibodies could reduce TGF-β signaling, promote T cell infiltration into the tumor environment, and reshape the immune microenvironment, thus, stimulating effective anti-tumor immune responses and tumor regression (69,70).

Numerous studies are investigating cancerous genes in this era of precision medicine. Least absolute shrinkage and selection operator regression analysis of data from The Cancer Genome Atlas and International Cancer Genome Consortium dataset and the International Cancer Genome Consortium database revealed nine genes (ANP32B, BMI1, ASF1A, CDK5, BUB1, CBX3, CBX2, CDK1 and BCORL1) to be independent predictors of HCC prognosis (71). Another study identified 11 immune-associated genes, NDRG1, MAPT, FABP6, CACYBP, HSP90AA1, ISG20L2, NRAS, BRD8, OSGIN1, CD320 and PSMD14, that were used to predict immune cell infiltration and construct a prognostic index for the prediction of immunotherapy efficacy (72).

The number of somatic mutations per DNA megabase (Mb), known as the TMB, is used to quantitatively evaluate the mutations carried by tumor cells (73). Greater numbers of neo-antigens, indicated by increased TMB, increases the likelihood that T cells will be recognized, which is clinically associated with improved ICI outcomes. Thus, the TMB is regarded as a reliable marker for estimating the effectiveness of immunotherapy in HCC. Data on 17 types of cancer were collected in a study by Samstein et al (74) confirming the initial finding that a high TMB is associated to immunotherapy effectiveness. Based in part on data from the KEYNOTE-158 study, the FDA approved the use of Pembrolizumab for solid tumors with 10 or more mutations/Mb in June 2020. However, there is not a fixed value of TMB for all types of cancer as the number of mutations defining TMB-high status varies with the type of cancer (74,75). Liver cancer has a median number of 4 mutations/Mb (n=755), with only 0.8% of patients having TMB-high tumors.

There are numerous studies on the role of the TMB in HCC (76,77). In a phase I clinical study, Xu et al (76) assessed the safety and efficacy of the combination of SHR-1210 (an anti-PD-1 antibody) and Apatinib in the treatment of patients with HCC. It was revealed that patients with a high-TMB had a worse prognosis compared with patients with a low TMB (mean, 8.53 vs. 1.44 mutations/Mb). Additionally, patients with a high-TMB had a reduced PFS with a reduction of 0.9 months compared with patients with a low TMB (76). However, only 1 patient (TMB, 15 mutations/Mb) in a fraction case series (total n=17) experienced a prolonged CR to ICI therapy. The TMB did not differ between responders and non-responders, highlighting the need for larger clinically annotated datasets to analyze outcome prediction (77).

Mismatch repair (MMR) in clinical practice is assessed largely by the reactions of four representative MMR-associated proteins (MLH1, MSH2, MSH6 and PMS2). One of the missing proteins is called DNA mismatch repair deficiency (dMMR) (78,79). MSI occurs during DNA replication, leading to alterations in the length or base composition of the MS, mainly as a result of dMMR. The MSI status of a tumor can be categorized as stable (MSS), high instability (MSI-H) or low instability (80). Perbolizumab was given FDA approval in 2017 to treat MSI-H/dMMR solid tumors that are unresectable or metastatic, have progressed after prior therapy and for which there are no adequate alternative treatment options. The first pan-cancer marker identified, MSI-H/dMMR, is now being used to direct tumor immunotherapy, and has been demonstrated to have clinical value for the treatment of tumors (81,82). Even though the incidence of the MSI-H phenotype in HCC is low at only ~2%, inflammation-mediated MMR pathway dysfunction may be to blame for the accumulation of mutations observed during hepatitis-associated tumorigenesis (78,83,84). According to several reports, Pembrolizumab treatment completely reverses MSI in patients with advanced HCC (84,85). However, a study revealed that out of 50 patients, only one (2.0%) was identified as MSI-H with high TMB, CD8⁺ lymphocyte infiltration, and low VEGF expression levels, and that patient did not experience as dramatic a response to Pembrolizumab treatment as suggested by other reports (86). MSI/dMMR is frequently used as a measure of the efficacy of immunotherapy for colorectal cancer (87). The most recent clinical study on neoadjuvant therapy for colorectal cancer included 12 patients with MSI-H/dMMR, and it revealed that all patients that finished treatment with checkpoint blockade had a clinically CR, without any reported adverse events of grade 3 or higher (88). However, another study that compared the OS of patients with resected colorectal cancer liver metastases between patients with MSS and MSI revealed that patients with MSI had a reduced OS, indicating a poor prognosis (89). The low proportion of patients with high TMB or MSI in HCC compared with gastric and colon cancer, and the sparse and contradictory information available, mainly from a small number of case reports or case series, make it impossible to determine predictive accuracy (78,86).

The TME describes the area surrounding the tumor, containing various cell types, such as endothelial, immune cells and fibroblasts. Extracellular components, such as cytokines, the extracellular matrix, growth factors, hormones and peripheral blood vessels, are associated to the development and metastasis of tumors (90). In addition to these, the TME in liver cancer also contains pit cells, Kupffer cells, hepatic stellate cells, liver sinusoidal endothelial cells and hematopoietic stem cells (91). As CD8⁺ lymphocytes are the most common T cell subset, the present review focuses on them. In several tumor types, high expression levels of CD8⁺ tumor-infiltrating lymphocytes (TILs) are associated with a favorable prognosis (92). High intra-tumoral CD8⁺ TIL levels have been associated with longer OS and DFS in a meta-analysis involving a total of 3,509 patients (93). Nevertheless, according to the experimental data from the CheckMate 040 trial, increased CD3⁺ or CD8⁺ tumor-infiltrating T cells were associated with improved survival rates and treatment responses, although this association was not apparent (94). Additionally, prognosis was not revealed to be associated to macrophage markers (14). Exhausted CD8⁺ T cells also exhibit a lack of cytotoxicity, decreased release of proinflammatory cytokines, such as IL-2, IL-12, IFN-γ and TNF-α, increased expression levels of inhibitory receptors, such as PD-1 and CTLA-4, and transcriptional and epigenetic changes (95). Additionally, compared with other types of cancer, HCC has an increased concentration of PD-1(Hi) CD8⁺ T cells that express exhaustion-associated inhibitory receptors, such as PD-1 and CTLA-4 on the surface of T cells, which is indicative of a poor prognosis (96). Immunohistochemistry (IHC) has demonstrated a strong association between an increased proportion of CD38⁺ cells and an improved response to ICIs (97). An unfavorable prognosis was revealed to be predicted by the upregulation of the LDHA, BFSP1, PPAT, NR0B1 and PFKFB4 genes, as demonstrated by a tissue microarray analysis (98). Thus, the TME can be used as a biomarker for the precise identification of patients who are sensitive to immunotherapy. However, the clinical use of TME components as biomarkers to predict the response to immunotherapy in HCC appears challenging. There is a need for the standardization and validation of test methods, test timing and test interpretation.

Evaluation of the treatment of patients with liver cancer should be performed throughout the treatment course, with the need for convenient, rapid and reproducible methods. It is evident that repeated multiple invasive biopsies of tumor tissue are unacceptable to patients, and the detection of circulating tumor cells (CTCs) in the blood via liquid biopsy would be more convenient for clinical use. Peripheral blood can be used to detect circulating biomarkers such as exosomes, circulating tumor DNA (ctDNA), CTCs and metabolites (99). Single- or double-stranded DNA that responds to tumor heterogeneity forms ctDNA, which is derived from tumor cells (100). According to a study by Cabel et al (101), synchronous changes in the ctDNA levels and the tumor size at 8 weeks after immunotherapy were predictors of DFP and OS in non-small cell lung cancer (NSCLC) and colorectal cancer. However, the plasma contains only trace amounts of ctDNA, which also fluctuates dynamically, resulting in a fluctuating detection threshold and false negatives (102). Another possible circulating biomarker is CTCs. The potential of HCC-CTCs expressing PD-L1 as prognostic and predictive biomarkers was investigated in a study by Winograd et al (103), which revealed that PD-L1-positive CTCs were typical of advanced HCC. Immunotherapy led to a good therapeutic response in patients with PD-L1⁺ CTCs (103). According to a different study, patients with 20% PD-L1-positive CTCs had an increased OS (median not reached vs. 8.9 months) and PFS (median 6.1 vs. 2.9 months) compared with patients with <20% PDL1-positive CTCs (76). These findings suggest that baseline ctDNA and high CTC levels might be used as predictors to select patients for immunotherapy and that dynamic changes in measured CTCs might be used as an indicator of treatment response in liver cancer, although this is still at an early stage. Further research is required on other circulating biomarkers such as extracellular vesicles and circulating RNA.