Current data do not provide clear evidence to support CTCs as an early-stage HCC diagnostic tool (14, 49). But as for prognosis, a great deal of research supports its prospect in predicting therapeutic outcome and monitoring disease progression, particularly after resection (3). Sun et al. (27) applied CellSearch™, for the first time, to capture and analyze EpCAM+ CTCs in 123 patients before the resection of HCC and 1 month thereafter. They found that EpCAM+ CTCs≥2 per 7.5 ml of blood were proved to be the strongest prognosticator, especially in the subgroup of patients whose AFP level was lower than that of 400ng/ml. Additional similar studies proved that the expression of EpCAM+ CTCs in HCC patients showed a significant positive correlation to the serum AFP level (35), BCLC stage (38) and was associated with vascular invasion (40), disease progression (35), higher recurrence rate (33), and shorter disease-free survival (DFS) and overall survival (OS) (31). Prospective research on 135 HCC patients demonstrated that surgical resection of tumors reduced the quantity of CTC. They suggest that the postoperative CTC continued to be at a higher level (≥5), which has a better predictive validity for prognosis than AFP>400cmg/L and tumor diameter>5cm (43).

However, EpCAM was positive in less than 20% of HCC cases (50), and some of them also develop into EMT. This will reduce the detection rate of EpCAM+CTCs and limit its clinical application. The Canpatrol™ CTC analysis platform was developed for EMT, which uses microfiltration and various markers to characterize CTCs. This technology not only uses epithelial markers (EpCAM, CK8/9/19) but also adds mesenchymal markers (Vimentin and Twist) (51). Qi et al. (32) used the CanPatrol™ CTC-enrichment technique in 112 HCC patients with a positive rate of more than 90%, even for early-stage diseases. Before resection, the number of CTCs≥16 and the percentage of mesenchymal CTCs≥2% were greatly correlated with intrahepatic recurrence and distant lung metastasis. Some negative enrichment methods and combinations of different markers or technologies of CTCs have also been established to optimize the platform (29, 46, 52).

Except for the number of preoperative CTCs, the forms of CTCs also influenced the prognosis of HCC patients. Mesenchymal CTCs have more invasion and metastatic potential. Bai et al. (53) showed that high expression of CXCR4 protein in mixed CTCs was more common, which might be associated with the progression and metastasis. Another study found that the proportion of mixed and mesenchymal CTCs was also a prognostic indicator of HCC (37). In a prospective study, Sun et al (40). examined CTCs at five important sites of blood vessels in HCC patients, and showed that the presence of multi-vascular CTC cluster could prognosticate recurrence and metastasis. The CTC clusters are also been reported in some studies. Gkountela et al. (54) showed that the specific hypermethylation of the binding sites for transcription factors related to stemness and proliferation in the CTC cluster promoted metastasis. The CTC-associated white blood cell clusters of peripheral blood in patients with HCC were also related to DFS and OS (44).

CTCs may serve as an index for long-term monitoring of HCC, and its dynamic changes reflected the therapeutic response treated with locoregional therapies, such as transcatheter arterial chemoembolization (TACE), radiotherapy, and radiofrequency ablation (55–57). Rau et al. (46) used a negative selection method for the enrichment of CTCs and found that the analyzes of CTCs at a different time during the treatment is helpful to dynamically monitor the progression of HCC patients, especially those without elevated serum AFP levels. Fan et al. (58) investigated a new technique which was combined flow cytometry in vivo with orthotopic tumor models and showed that the number of CTCs decreased significantly after resection and early metastases were also reduced. The tumor size and the number of distant metastases of HCC were in accord with the dynamic change of CTCs. Moreover, Zhang et al. (59) used microfluidic chip technology to obtain CTCs and discovered that the number of spheroids formed by CTCs substantially reduced after treated with sorafenib or oxaliplatin. This reflects the potential of CTCs in the analysis of sensitivity and drug resistance of chemotherapeutic drugs.

Identifying CTCs that express specific tumor markers and drug targets enable doctors to better guide personalized therapy. In 2016, Li et al. (47) indicated that CTCs can replace tumor tissue to characterize the expression of pERK/pAkt. The pERK+/pAkt- CTCs were most sensitive to sorafenib, which is helpful to choose the appropriate treatment scheme. In recent years, immunotherapy is a research hotspot, and immune checkpoint inhibitors, including Programmed cell death protein-1 (PD-1) and programmed death-ligand 1 (PD-L1), are promising in the treatment of advanced HCC (60). Winograd et al. (61) firstly evaluated PD-L1+ CTCs in HCC. They reported that PD-L1+ CTCs could serve as a predictor of immunotherapy. Because 3 of all 6 patients who received anti-PD1 therapy showed responses to treatment and had PD-L1+ CTCs (61). In 2020, the same team conducted a larger, prospective cohort to enumerate/phenotype CTCs. They showed that PD-L1+ CTCs mainly present in advanced HCC, and related to beneficial therapeutic responses of patients with HCC who received anti-PD-1 therapy (62). The expression of PD-L1 varies with immune statuses, treatment response, and disease progression. Precise assessment of PD-L1 expression in CTCs could be used to evaluate and monitor the immune status of tumor cells in real-time.

Epigenetic changes induced by DNA methylation and DNA methylation are involved in the process of tumor occurrence and development (90). The methylation pattern of ctDNA has tremendous potential in the early diagnosis of HCC patients. Changes in methylation of multiple genes in plasma/serum, such as p15, p16, GSTP1, INK4A, RASSF1A, and so on, have been confirmed in many studies to distinguish HCC from controls. The methylation characteristics of tumor DNA in HCC are highly correlated with paired plasma ctDNA (67), so some biomarkers of DNA methylation in HCC may also be used in ctDNA. Kotoh et al. (91) developed a methylated SEPT9 assay in HCC, with 63.2% sensitivity and 90.0% specificity. They pointed out that combined diagnosis with AFP can improve the diagnosis rate of early-stage HCC. Yan et al. (92) proposed the HCC index, which was a combined diagnostic model of cfDNA, age, and AFP. The HCC index was more accurate in HCC diagnosis than cfDNA or AFP, alone. Currently, Chen et al. (81) conducted large-scale, multi-center research and constructed a diagnostic model based on HIFI (5-Hydroxymethylcytosine/motIf/Fragmentation/nucleosome footprInt) method, playing a key role in differentiating HCC from non-HCC. Both the test set and the verification set had a sensitivity and specificity of more than 95%.

In addition to the role of diagnosis, ctDNA methylation can also be served as a prognostic indicator. The study of Kotoh et al. also showed that the copy number of methylated SEPT9 was related to BCLC stage, macrovascular invasion, tumor number, and size (91). Li et al. (93) indicated that IGFBP7 promoter methylation was significantly related to OS and early tumor recurrence after hepatectomy. In a large-scale study involving 1098 HCC patients, a diagnostic prediction model was constructed by screening 10 overlapping markers of cfDNA methylation using Random Forest and Least Absolute Shrinkage and Selection Operator (LASSO) methods. The diagnostic sensitivity and specificity of this model were 85.7% and 94.3%, respectively (67). Kisiel et al. (69) identified 6 best-methylated DNA markers (MDMs) in HCC, that include ECE1, HOXA1, CLEC11A, AK055957, PFKP and EMX1, and combined Phase I Pilot, and Phase II Clinical Validation. In the diagnosis of HCC, the area-under-the-receiver-operating-curve (AUC) of this 6-marker MDM panel was 0.96, the sensitivity was 95%, and the specificity was 92%. Moreover, elevated levels of cfDNA are a positive correlation with poor prognosis. The existence and quantity of ctDNA are usually determined by detecting the mutation of cfDNA. Some hot mutants, such as TP53, CTNNB1, and TERT, are usually selected to detect ctDNA aberrations (14). Ren N et al. (94) found that allelic imbalance at D8S258 in circulating plasma DNA was also related to the poor prognosis of HCC. A recent study focused on SNVs of ctDNA and showed that the existence of MLH1 SNV, coupled with elevated ctDNA levels, can predict poor OS of HCC patients (95). In short, the different mutants and epigenetic modifications in ctDNA have significance in early diagnosis and prognostic outcomes of HCC.

Considering that liver biopsy is invasive and unnecessary in advanced HCC, ctDNA can be a reliable biomarker for dynamically monitoring tumor progression and assessing the treatment efficacy. These include identifying new mutations that drive acquired drug resistance and capturing heterogeneity between tumors. According to the study of Park et al. (96), the high level of cfDNA after radiotherapy was related to the poor outcome of treatment. Thus, ctDNA can be used as an indicator to evaluate the curative effect after radiotherapy. Recently, Zhao et al. (97) prospectively enrolled 42 patients with unresectable liver cancer. They found that TP53 mutation was related to disease progression and interventional treatment was more effective in patients without TP53 mutation. The follow-up study showed that plasma levels of CNVs and SNVs in ctDNA dynamically correlated with patients’ tumor burden in HCC (80). They decreased after surgery and increased in cases with tumor recurrence, showing that ctDNA was a feasible biomarker to monitor treatment response (80). Cai et al. (87) collected the information of postoperative ctDNA and protein biomarkers which included AFP, AFP-L3, and des-gamma-carboxy prothrombin (DCP), then evaluated the results with corresponding MRI scan images during follow-up. They confirmed that both SNVs and CNVs of ctDNA could dynamically monitor the tumor load of HCC. The combination of ctDNA and DCP could improve the detection rate of minimal/molecular residual disease (MRD) in patients undergoing hepatectomy (87).

On other hand, the detection of ctDNA mutations can guide the choice of treatment. Ikeda et al. (98) used digital ctDNA sequencing to evaluate the mutant associated with wild-type allele fraction in 14 advanced HCC patients. The level of DCP and AFP decreased with the treatment of palbociclib (CDK4/6 inhibitor) and celecoxib (COX-2/Wnt inhibitor) after two months in a patient with CDKN2A-inactivating and CTNNB1-activating mutation. And AFP declined by 63% in another patient treated with sirolimus (mechanistic target of rapamycin inhibitor) and cabozantinib (MET inhibitor). This patient had PTEN-inactivating and MET-activating mutations of ctDNA (98). Compared with tissue biopsies, cfDNA identified clinically related drug resistance changes more frequently in a prospective cohort study of patients with gastrointestinal tumors. In 78% of cases, cfDNA could detect drug resistance gene mutations that were not found in matched tumor biopsies (99). A recent study reported that the sequential mutation profiling of ctDNA can be used for molecular detection of drug resistance in HCC (100). The gene mutation of PI3K/MTOR pathway was associated with the worse PFS in HCC patients receiving tyrosine kinase inhibitors but was not associated with immunosuppression therapy (100). It shows that monitoring the mutation of the drug target gene or drug resistance gene in advanced patients can better guide the choice of personalized treatment. All in all, genomic profiles of patients with HCC can be obtained from ctDNA, which can guide the treatment to some extent.

Exosomes were thought to originate from mature sheep reticulocytes (101). Exosome belongs to extracellular vesicles (EVs), which is a nano-sized phospholipid bilayer membrane vesicle and responsible for intercell communication (102). It is secreted by living cells and formed by the separation of intracellular poly vesicles with cell membranes in the process of being released out of the cell (103). The component transported in exosomes contains several molecular biomarkers including proteins, RNA, DNA, which range in size from 50 to 140 nm (88, 104). Currently studies suggest that the cargos carried by exosomes associated with HCC, including non-coding RNAs(ncRNAs), messenger RNAs (mRNAs), and proteins, can serve as potential biomarkers for the clinical application of HCC (105). They transferred to target cells by exosomes, affecting drug resistance, tumor angiogenesis and metastasis (106–108). Exosomes show significant superiority in liquid biopsy. Exosomal substances are hard to be degraded due to the protection of phospholipid bilayer membrane (15). Its high biological stability improves the clinical applicability of exosomes, which can not only cut down the cost of short-term sample preservation but also reduce the challenges of transportation (109). Moreover, exosomes carry biological information from parental cells, so they are more typically than cfDNA (110). Exosomes are bound up with the growth and metastasis of HCC (108), tumor angiogenesis (111), and immune regulation (112).

The capture and enrichment of exosomes require the identification of their markers, such as heat shock protein 70 (HSP 70), CD9, CD63, CD81, and ALIX (113). There are some common technologies used for exosomes isolation according to their composition or physical properties, such as ultracentrifugation (UC), transmission electron microscopy (TEM), magnetic associated cell sorting (MACS), filtration, polymer-based precipitation, nanoparticle tracking analysis, fluorescence, colorimetric ELISA assays, and size exclusion chromatography (SEC). However, the eventual clinical utility of exosomes is still in its preliminary stages and needs more validation.

Exosomes contain bioactive compounds, such as RNA, DNA, proteins, and cholesterol (15), some of which can be expressed uniquely by tumor cells. MicroRNAs (miRNAs), lncRNA and circular RNAs (circRNA) have been described in many reports as noninvasive biomarker for better prediction and prognosis of HCC progression. The miRNAs, a class of conserved RNAs (usually 22-25 nucleotides), can inhibit the translation of mRNA or promote the degradation of mRNA by combining with their response element on the 3-untranslated region of the target mRNA (107). Exosomes can actively secrete miRNAs to regulate the progress of HCC (15, 131). Sohn et al (115) found that serum exosomal miR-222, miR-221, and miR-18a in patients with chronic hepatitis or cirrhosis were significantly lower than that in HCC patients. Wang et al. (132) showed that the serum exosomal miR-21 also elevated in HCC patients, which could be used to distinguish HCC from people and chronic hepatitis B patients. Ghosh et al. (130) found that the combination of four miRNAs (miR-221-3p, miR-223-3p, miR-10b-5p and miR-21-5p) showed good diagnostic ability in patients with low expression of AFP. Currently, Cho et al (116) showed that serum exo-miR-10b-5p had great potential for early diagnosis of HCC with a sensitivity of 90.7%, specificity of 75.0%, and the AUC was 0.934. The same team developed a panel that included exo-miR-4746-5p and exo-miR-4661-5p for the early diagnosis of HCC. The sensitivity, specificity, and AUC were 81.8%, 91.7%, and 0.947 respectively (118). Along with the miRNA, lncRNA and cirRNA in exosomes have also shown potential in the early detection of HCC. The expression of LINC00853, lnc85, ENSG00000248932.1, ENST00000440688.1 and ENST00000457302.2, have shown promise for the tumorigenesis prediction (128, 129, 133). Xu et al. (121) suggested that AFP combined with serum exosomal LINC00635 and ENSG00000258332.1 could discriminate HCC from chronic hepatitis B, gaining an AUC of 0.894. By comparing the level of exosomal Trna-derived small RNA (tsRNA) between the healthy and patients with liver cancer, Zhu et al. (134)found a significant increase of tsRNAs in plasma exosomes of liver cancer patients, which provided new insight into the HCC diagnostic potential of the exosome. Currently, a study suggested that the combination of three circRNAs, including circ_0004001, circ_0004123 and circ_0075792, has been served as a valuable diagnostic biomarker for HCC (127).

In addition to ncRNA, proteins are also used as biomarkers of HCC. Compared with miRNA, the study of protein in exosomes is a relatively less explored direction to detect HCC. According to a study that the proteomes of HCC, cirrhosis and healthy people have different compositions (135). Arbelaiz et al. (119) suggested that the differentially expressed exosomal proteins might be a kind of potential biomarkers for differential diagnosis. Mass spectrometry was used to detect the proteomes of exosomes in patients with HCC, primary sclerosing cholangitis, and cholangiocarcinoma. Some of these proteins including LG3BP and FIBG demonstrated superior diagnostic ability to AFP. The AUC of LG3BP for identifying HCC patients from the healthy was 0.904, and the AUC of FIBG for distinguishing intrahepatic cholangiocarcinoma from HCC was 0.894 (119). Interestingly, LG3BP and PIGR can promote the transformation, invasion, and proliferation of tumor cells, which are associated with poor prognosis (136). Fu et al. (122) found that patients with advanced HCC had higher levels of Smad3 in exosomes and suggested that combined AFP and detection of exosomes containing SMAD3 can improve the diagnosis of HCC.

Exosomes also have great potential in terms of the prognosis of HCC. Lee et al. (124) demonstrated that circulating exosomal miRNA-21 and lncRNA-ATB were associated with the T stage, the TNM stage, and portal vein thrombosis. The HCC patients who had the level of exosomal miRNA-21≥0.09 and lncRNA-ATB ≥0.0016 tend to have significantly lower OS and PFS (P<0.05). The high level of LINC00635 and ENSG00000258332.1 in HCC was related to lymph node metastasis, TNM stage, and OS (121). Some studies showed that the lower levels of exosomal miR-125b (137) and miR-638 (138) predicted poor prognosis of HCC patients. The abundance of exosomes containing SMAD3 was negatively correlated with DFS in postoperative patients of HCC (122). Moreover, circUHRF1 might be associated with resistance to anti-PD1 immunotherapy in HCC patients (126). Luo et al. (139) measured the level of exosomal circular RNA (circAKT3) from 124 patients with HCC and 100 healthy controls and found that patients with high exosomal circAKT3 tend to have a higher risk of recurrence and death.