Liquid biopsy in hepatocellular carcinoma: Challenges, advances, and clinical implications

Biology and characteristics

CTCs refer to cells that have detached from the tumor mass and circulated in the bloodstream for distant metastases. As malignant tumors grow, matrix metalloproteinases are secreted, breaking down the basement membrane and allowing the tumor cells to enter blood vessels or lymphatic vessels [6]. Once entering the blood vessels, tumor cells become CTCs, which undergo epithelial-to-mesenchymal transition (EMT) process. This process equips CTCs with mesenchymal properties to survive immune attacks, shear stress, and anoikis - a form of programmed cell death that occurs in anchorage-dependent cells when they detach from the surrounding extracellular matrix [7]. Only a few CTCs escape immune system surveillance and extravasate to metastasize to other organs. Given the nature of CTCs, their presence is regarded as a biomarker for both cancer detection and prognostication.

Technical challenges and technological advancements

CTCs are extremely rare in circulation, with a prevalence of one CTC per 10⁶–10⁷ white blood cells (WBCs) [8,9]. As such, efficient enrichment of CTCs from blood is crucial for downstream analysis. There are two main categories of technologies to enrich CTCs: immunoaffinity-based and physical property-based CTC enrichment techniques. The immunoaffinity-based techniques use antibodies to target antigens on the cell membrane, which can be employed for either positive selection to detect CTCs or negative selection to filter out WBCs [10]. For positive selection of CTCs, the most used marker is epithelial cell adhesion molecule (EpCAM), as many cancers originate from epithelium. The CellSearch^® System is the first and only EpCAM-based method approved by Food and Drug Administration (FDA) for enrichment and enumeration of CTCs in metastatic breast, colorectal, and prostate cancers [11-13]. This system uses ferromagnetic beads coated with anti-EpCAM antibodies to capture CTCs, followed by magnetic-activated cell sorting for isolation. However, cancer cells often undergo EMT, and HCC CTCs are derived from hepatocytes rather than epithelial cells. Therefore, using anti-EpCAM antibodies alone may not be sufficient for effective enrichment of HCC CTCs [14]. To address this limitation, cocktails of different antibodies targeting asialoglycoprotein receptor (ASGPR), glypican‐3 (GPC3), and vimentin have been developed, demonstrating increased capture rates of HCC CTCs [15]. Microfluidic devices, such as CTC-Chip^TM [16-19] and NanoVelcro Chip^TM [20-23], which can also be used with different antibodies, have been developed to improve the capture efficiency and the purity of HCC CTCs. For example, nanosubstrates that mimic Velcro® and the overlying chaotic mixer of NanoVelcro Chip^TM significantly increase the contact frequency and affinity between CTCs and the chip surfaces, achieving >80% capture efficiency in HCC patients [15]. Compared to the positive selection strategy mentioned above, negative selection utilizes anti-CD45 antibodies to target and deplete hematopoietic cells from the background [24], thereby enriching CTCs. However, negative selection is typically used in conjunction with positive selection, physical property-based isolation, or microfluidic devices, as the purity of enriched CTCs from negative selection alone is usually low [10]. The third generation of CTC-Chip^TM, CTC-iChip [17,19,25], and its updated version combine either positive selection or negative selection with an arrayformat, inertial-enhanced design to fulfill high-throughput (10⁷ cells/s) and enrich all CTC populations with high purity. A 10-gene signature in HCC CTCs enriched by CTC-iChip was quantified as a surrogate for the presence of CTCs and demonstrated as a promising biomarker for early detection of HCC [19].

The second CTC enrichment method, physical property-based isolation, leverages the differences in size, mechanical plasticity and dielectric mobility of CTCs [26]. Techniques in this category include microfiltration, density-based gradient centrifugation, dielectrophoresis, and microfluidic devices. As these methods are antigen-independent, they can address the issue of EpCAM loss during EMT and the heterogeneous antigen expression of CTCs. They also offer the advantages of relatively low cost and ease of downstream analysis, as there is no need to separate CTCs from coated antibodies. The Parsortix^® PC1 system is an example of microfluidic device recently cleared by FDA to enrich CTCs in metastatic breast cancer patients based on cell size and deformability [27].

Despite enrichment using the aforementioned technologies, more than 1,000 WBCs per CTC still inevitably remain in the background [8]. Therefore, immunofluorescence staining is typically applied to distinguish CTCs from the background WBCs. CTCs are usually defined as cells with DAPI+ oval nuclei, CK+ cytoplasma, CD45- cytoplasma, and positive staining of targeted surface markers such as EpCAM, vimentin, or ASGPR, particularly for HCC [15,28]. The identified CTCs could either be enumerated or collected for single-cell analysis [28,29]. In addition to manual enumeration and identification of CTCs, bioinformatics analysis was proposed to select a panel of genes that are highly expressed in CTCs but lowly expressed in background WBCs [19,22,30]. This approach enables “bulk” genetic analysis of the samples after CTC enrichment, streamlining the entire analysis process for further clinical applications like HCC detection and prognostication.

Advantages and disadvantages of CTCs

Compared to other types of liquid biopsy, CTCs preserve the most complete molecular information due to the protection of their cell membranes. With the advancement of CTC isolation technologies and single-cell analysis, CTCs can be released from the captured substrates with high cell viability [28]. This enables downstream analyses of DNA, RNA and proteins to investigate the mechanisms of tumor progression, EMT, and metastasis, as well as to surrogate clonal evolution of the cancer. In addition, in vitro proliferation of the CTCs allows for functional tests to assess drug resistance and to understand the interactions between CTCs and epithelial cells, platelets, and immune cells during EMT, providing insights into the blood microenvironment [28].

However, CTCs also have several disadvantages. First, no standardized pre-analytical protocols exist for CTC studies.9 While the two FDA-cleared devices, CellSearch^® and Parsortix^® PC1 Systems, each have their own protocols and corresponding blood collection tubes (BCTs), they are not approved for use in HCC patients. Additionally, the choice of BCT varies based on the downstream CTC analysis. 9 For example, the CellSave BCT is used for CTC enumeration, 31 while acid citrate dextrose solution A (ACD-A) and ethylenediaminetetraacetic acid (EDTA) BCTs are preferred for CTC mRNA analysis [32,33]. With appropriate BCTs and storage at 4°C, CTCs and their molecular contents are generally stable for up to 3 days [34,35]. However, for molecular analyses, it is recommended to verify the stability of the target molecular panels over the processing interval to ensure study reliability. Second, despite advancements in automated reading system and artificial intelligence/deep learning that assist in the identification of CTCs [36,37], enumeration still heavily relies on manpower and requires significant experience, which is inevitably subjective to some degree. Third, CTCs are heterogenous as their originated tumor cells [38,39]. While continuously profiling these heterogeneous CTCs could help unravel the dynamic changes in the tumor and its evolving biology [24], this poses challenges for the CTC enrichment process. This is particularly true for immunoaffinity-based methods, as different CTC populations may express varying surface antigens. Lastly, due to their nature, CTCs are relatively rare in patients with early-stage cancer, which makes their utility as a biomarker for early detection of HCC challenging.

Biology and characteristics

CfDNA are small fragments of degraded DNA (<200 bp) that originate from disrupted or apoptotic cells and circulate in the bloodstream [40]. CfDNA derived from malignant cells is also known as circulating tumor DNA (ctDNA). In normal conditions, the concentration of cfDNA is typically low, but it increases in certain conditions such as exercise, inflammation, diabetes, and malignancy [41,42]. In HCC patients, cfDNA exhibits both quantitative changes, such as elevated levels of cfDNA, and qualitative changes, including copy number alterations (CNVs), single-nucleotide variations (SNVs), methylation patterns, 5-hydroxymethylcytosine (5hmC), and fragment features [40]. For CNVs in HCC cfDNA, arm-level amplifications of 1q and 8q, and deletions of 1p, 4q, and 8p, are frequently observed [43]. SNVs in TP53, CTNNB1, and the promoter region of TERT are common in cfDNA from HCC patients, offering potential for cancer detection [44,45]. DNA methylation is known to be involved in DNA regulation and carcinogenesis, notably appearing even in the early stage of tumor development. Therefore, cfDNA methylation has been regarded as a promising biomarker for early-stage detection of HCC [46]. Recent studies suggested that at least 10 to 50 cfDNA methylation markers are required to robustly distinguish HCC from controls [47,48], while around 200 cfDNA methylation markers are needed to differentiate between different cancers [49]. Alternatively, incorporating relevant clinical variables such as age, sex, AFP, AFP-L3%, and DCP along with a few cfDNA methylation markers is a common approach to enhance the sensitivity for HCC detection [47,48]. In addition to mutation and methylation patterns, cfDNA fragmentation has attracted the attention of many researchers. Genome-wide fragmentation profiles, fragmentation size, nucleosome footprint, DNA 5’ end motif, jagged ends, and endpoint locations were shown to provide valuable information to differentiate HCC patients [50].

Technical challenges and technological advancements

It Is estimated that ctDNA could be <1% of the total amount of cfDNA, with levels even lower at around 0.01% in early-stage cancer [51]. Unlike the various methods available for isolating CTCs, there are currently no reliable methods for enriching ctDNA from cfDNA. As the size of cfDNA in HCC patients tends to be shorter, approximately 144 bps [52], a study suggested applying fragment-size selection of <150 bp to increase the proportion of ctDNA [53]. However, this approach is not commonly used in most cfDNA-based biomarker studies. Instead, researchers are focusing on developing PCR and sequencing technologies to either increase sensitivity or broaden the genomic coverage to detect subtle changes in ctDNA. In general, PCR techniques are used to detect known SNVs and methylation markers, offering high sensitivity and quick results [54]. Among the promising commercial cfDNA platforms for HCC detection [55], Oncoguard^® Liver by Exact Science has developed a PCR-based technology, the Target Enrichment Long probe Quantitative Amplified Signal (TELQAS) assay [56], to measure selected methylation markers. The TELQAS assay combines three key features: 1) a multiplexed pre-amplification process, 2) longer and high-melting-temperature probes, and 3) a second round of amplification of individual target loci, which enables sensitive and specific detection of low-abundance methylation targets in cfDNA without requiring whole-genome pre-amplification.

The other methods, besides PCR-based technologies, are next-generation sequencing (NGS), including whole-genome sequencing, whole-exome sequencing, and target-panel sequencing. While NGS provides a comprehensive view of the entire genomic map, those methods are more expensive and time-consuming [54]. For the detection of low-frequency SNVs in HCC ctDNA, Lee et al. [45] designed an algorithm that condenses hopped indexes for barcode-based targeted sequencing to eliminate PCR and sequencing errors, thereby achieving ultra-high depth and enhanced sensitivity. Similarly, HCCscreen^TM by GENETRON uses barcode-based targeted sequencing with rapid amplification of cDNA ends and multiple primers to detect the SNVs in TP53, CTNNB1, and AXIN1, the promoter region of TERT, and the HBV integration breakpoint, which are included in a panel for HCC detection. The researchers further developed the Mutation Capsule Plus (MCP) technology [57], which enables 1) parallel profiling of mutation and methylation changes in the same reaction, 2) repeated profiling of a single cfDNA sample without sacrificing sensitivity, and 3) genome-wide screening for unidentified methylation markers. For methylation sequencing of cfDNA, HelioLiver^TM’s ECLIPSE platform employs a gentler enzymatic conversion approach, yielding a greater quantity and quality of converted DNA compared to traditional bisulfite conversion, which is well-known for causing substantial DNA and signal damage [58]. Additionally, the ECLIPSE platform incorporates a bioinformatics pipeline named Cell Heterogeneity-Adjusted cLonal Methylation (CHALM) [59], which is able to assess the heterogeneity of sequenced bulk cfDNA, thereby enhancing the accuracy of quantifying differentially methylated markers. More recently, a research team led by Zhou developed a cost-effective assay named cell-free DNA Methylome Sequencing (cfMethyl-Seq) to profile CpG-rich regions in cfDNA [49]. Since these regions comprise only 3% of the whole genome, the cost of cfMethyl-Seq is significantly reduced while still covering over 90% of CpG islands and achieving more than 12× enrichment compared to whole-genome bisulfite sequencing. Leveraging cfMethyl-Seq, cancer-specific and tissue-specific hypermethylation and hypomethylation cfDNA markers were identified and demonstrated excellent performance in detecting and locating cancers, including HCC.

Advantages and disadvantages of cfDNA

Unlike the absence of standardized guidelines for the collection, processing, and storage of CTCs, National Cancer Institute’s (NCI) Biorepositories and Biospecimen Research Branch (BBRB) has introduced a cfDNA-specific guideline [60] to harmonize the pre-analytical processes and minimize potential cfDNA degradation, analytical artifacts, and contamination. This guideline includes evidence-based recommendations on BCTs, temperature and duration of ambient storage or transport, centrifugation speed and duration for blood processing, as well as cfDNA extraction and storage. Notably, if blood is collected in Streck tubes, processing can be delayed for up to 3 days. In addition, the pre-analytical centrifugation requirement involves only two steps within an hour and does not require specialized centrifugation equipment, making it feasible in standard laboratories. These advantages facilitate multicenter clinical trials investigating cfDNA’s potential for HCC detection and prognostication and simplify the establishment of institutional biorepositories for future studies. Indeed, compared to CTCs and EVs, the genomic landscape of HCC ctDNA is the most extensively profiled, and several cfDNA-based assays are commercially available and validated in large-scale clinical studies.

Despite its advantages, cfDNA-based assays are relatively expensive and time-consuming, especially for NGS-based assays. Since cfDNA is fragmented, some chromosomal analyses like karyotyping and fluorescence in situ hybridization cannot be performed. Additionally, as there are no reliable methods to isolate ctDNA from cfDNA, it is challenging to specifically attribute identified genetic or epigenetic characteristics to HCC. Finally, cfDNA heterogeneity may interfere with quantification results. However, several of the aforementioned studies [49,56,58] are underway to improve cost-effectiveness and minimize the impact of cfDNA heterogeneity.

Biology and characteristics

EVs are a group of lipid-bilayer “cargo” vesicles containing proteins, lipids, RNA, DNA and non-coding RNA released by cells into the extracellular environment [61,62]. Their major functions are intercellular communication, transfer of “cargo” cell-to-cell, and modulating physiologic conditions. In HCC, EVs from various cells influence the tumor microenvironment [63] and drive cancer progression by modulating cancer cell proliferation and migration, angiogenesis, extravasation, immune escape, and metastasis [64,65]. Recent evidence also shows that EVs derived from HCC can promote tumorigenesis and cancer stemness [66,67]. As such, EVs are regarded as promising HCC biomarkers for early detection of cancer, prognostication, and monitoring cancer progression.

Technical challenges and technological advancements

Multiple EV isolation methods are endorsed by The International Society for Extracellular Vesicles (ISEV) and its latest guideline, “Minimal information for studies of extracellular vesicles (MISEV2023): from basic to advanced approaches.” [62] Among these methods, differential ultracentrifugation, performed at speeds of 10,000–20,000g for 10 to 90 minutes followed by 100,000–200,000g for 45 to 150 minutes, is the most commonly used for EV isolation in research. However, this method is time-consuming, and ultracentrifugation machines are not commonly available, which hinders its application in clinical settings. Precipitation using hydrophilic polymers, such as polyethylene glycol, is also commonly used to isolate EVs due to its low cost, speed, simplicity, and high yield [62,68]. But this method typically results in low-purity EVs, often co-precipitating with vesicle-free miRNA and proteins like albumin and lipoproteins [69]. Filtration leverages membranes with different pore sizes to isolate EVs, allowing not only for the separation of EVs from proteins but also for the harvesting of EVs of specific sizes. Compared to precipitation, EVs enriched through filtration are purer, though the yield is slightly lower. Additional advantages include the lack of requirement for special equipment and simplicity, although filter clogging and the co-retention of proteins of similar size remain inevitable. Size exclusion chromatography separates nanoparticles based on their size as they pass through the column packed with numerous porous polymers and buffers. Because EVs are larger than the pore size, they cannot enter the pores like smaller molecules, allowing them to flow through the column more quickly and be separated effectively. This method can achieve higher purity of isolated EVs compared to filtration. However, it is more time-consuming and may require an additional step to concentrate the diluted EV samples for downstream analysis.

Although the aforementioned methods are effective to isolate EVs and urine and plasma samples of HCC patients, they cannot distinguish between HCC-derived EVs from those of non-HCC origins, making it challenging to detect tumor-specific molecular information within HCC-derived EVs. To overcome this limitation, immunoaffinity-based isolation methods using antibodies to target specific EV surface molecules have been investigated to capture the HCC-associated EVs [70,71]. Julich-Haertel et al. [72] demonstrated that AnnexinV+ EpCAM+ ASPGR1+ EVs are elevated in HCC patient’s sera and decrease after curative treatment, indicating that utilizing these markers might be a potential method to enrich HCC-associated EVs. Assuming that the surface markers of EVs resemble those of the parental tumor cells due to their shared cytomembrane, Sun et al. [71] targeted EpCAM, ASPGR1, CD147, and GPC3 to purify HCC-associated EVs. The research team developed methyltetrazine-modified microbeads, termed Click Beads, which trigger a click chemistry reaction upon encountering EVs coated with antibodies modified with trans-cyclooctene. Compared with conventional streptavidin-biotin-mediated immunoaffinity-based approaches, the click chemistry reaction enables faster and irreversible purification [71,73], thereby improving capture efficiency and reducing nonspecific EV capture. In combination with duplex real-time immuno-PCR, the authors enabled sensitive quantification of EV subpopulations positive for both the aforementioned HCC-associated markers (EpCAM, ASPGR1, CD147, and GPC3) and EV-related markers (CD63 and CD9) [62], which have shown promise for detecting early-stage HCC. Microfluidic devices leveraging immunoaffinity and physical characteristics are also a novel strategy to purify HCC-associated EVs [62,68,74]. For example, the same research team developed EV Click Chips [70,75], which integrates the following features: 1) an antibody cocktail (EpCAM, ASPGR1, and CD147) to enrich HCC-associated EVs, 2) a chaotic mixer and nanostructured substrate to significantly increase the surface area as EVs flow through, thereby enhancing capture efficiency, and 3) click chemistry-mediated EV capture and disulfide cleavage-driven EV release, which further increases the purity of the targeted EVs for downstream analyses. Although immunoaffinity-based isolation methods, including microfluidics technology, can isolate HCC-associated EVs and achieve high EV purity, they have some limitations [62,68,74], including low EV yield, the cost of antibodies, and the need for separation between EVs and antibodies or devices if further functional studies are required.

Currently, there is no universal pre-analytical protocol for blood collection, processing, and storage in blood-based EV studies. Acknowledging that the initial quality of plasma and serum can significantly affect study results, the ISEV Blood Task Force developed the Minimal Information for Blood EV Research (MIBlood-EV) guideline to help researchers transparently report crucial pre-analytical information, thereby improving the reproducibility of results. Most recently, Dhondt et al. [76] conducted the first large‐ scale benchmarking study to comprehensively compare the impact of BCTs and blood processing intervals on blood sample quality and downstream analyses. In brief, plasma is preferred over serum for analyses, as less soluble platelet activation is observed in plasma, reducing potential confounding of the results. ACD-A, citrate, or EDTA BCTs are recommended for plasma collection, with ACD-A and EDTA approved for a blood processing interval of up to 8 hours after collection. Unlike CTC and cfDNA, which remain stable after 72 hours, significantly elevated levels of platelet-derived EVs and erythrocyte-derived EVs due to platelet activation and hemolysis are observed after 72 hours from blood collection, potentially affecting the study results. The Streck BCT can also be used for plasma collection in EV studies, although it offers no clear advantage over ACD-A, citrate, or EDTA BCTs regardless of the blood processing interval.

Urine is the second most commonly used biofluid in EV studies and offers the advantages of easy and noninvasive collection. The ISEV Urine Task Force has developed guidelines [77,78] for collection, processing, and storage to standardize and facilitate urinary EV studies. Similar to blood samples, urine samples should be processed within 8 hours after collection to prevent microbial growth. Further detailed recommendations for each processing step can be found in the quick reference card [78].

Advantages and disadvantages of EVs

Similar to CTCs, EVs are protected by a lipid bilayer and contain not only DNA but also RNA, proteins, and lipids, providing additional molecular information, while being easier to collect and preserve. Given the roles of EVs in HCC progression, functional studies of EVs can be conducted to reveal the underlying mechanisms when new EV-based biomarkers are discovered. Although blood samples for plasma collection in EV studies are recommended to be processed within 8 hours, which can pose challenges for fresh sample transfer, plasma can be obtained through a simple two-step centrifugation process that is feasible in both clinical and research laboratories. The plasma can then be stored and transported at –80°C for subsequent clinical EV studies. However, due to the lack of a standardized pre-analytical protocol for EV studies, factors such as BCTs, processing and storage times, and the number of freeze-thaw cycles may introduce technical variations. Unlike CTCs and cfDNA, research and biomarker exploration in the field of HCC EVs are still in the early stages of development. For example, given the small amounts of the HCC-derived EVs enriched using the novel technologies [70,71], more sensitive methods are required to detect and quantify scarce molecules, such as RNA, within these EV subpopulations. Lastly, larger scale clinical trials are needed to validate the findings of HCC EV biomarker studies.

Early detection of HCC using CTCs

Earlier efforts to detect CTCs used immunoaffinity against EpCAM, one of the commonly used membrane-associated proteins, not expressed by normal blood cells. However, EpCAM positive rate was only about 35% in CTC [95]. This result is likely because its expression is low in the early-stage HCC and is often lost during the EMT process. To overcome these issues, research has been conducted using HCC-specific markers such as GPC3, a cell membrane-anchored protein, more frequently observed in moderately or poorly differentiated HCC tumor cells and ASGPR, a transmembrane protein exclusively expressed on the surface of hepatocyte [96-98]. While they did not show excellent detection performance as an individual marker, combination of EpCAM, GPC3 and ASGPR antibodies using microfluidic synergistic chip achieved a detection rate of up to 97% in patients with HCC at various stages [99]. On the other hand, analyzing various types of CTCs, such as epithelial, mesenchymal, and mixed type, can help with enhanced accuracy for early HCC detection [100]. Chen et al. [100] enumerated all the CTCs, including epithelial, mesenchymal, and mixed type, which showed better performance in discriminating between HCC and non-malignant liver disease compared to AFP.

More research has been conducted by combining CTCs with other biological markers to overcome the low sensitivity of CTCs. Combining CTCs with AFP resulted in better diagnostic yield compared to CTCs alone or AFP alone (0.821 vs. 0.774 vs. 0.669) [101]. Guanine nucleotide-binding protein subunit beta-4 (GNB4) and Riplet gene methylation showed high sensitivity and specificity when they were combined with CTCs [102]. Despite advancements in CTC technology to detect HCC, CTCs for HCC surveillance are not endorsed due to high error rates, low detection rates in early stages, and difficulties in developing standardized testing methods [103].

Early detection of HCC using cfDNA

In earlier cfDNA research, the primary focus was on the quantitative changes in cfDNA as a biomarker for early-stage HCC detection. Following the results of the Iizuka group, which showed that cfDNA concentration was higher in HCV-associated HCC patients compared to the HCV carrier group [104], several other studies reported that the concentration of cfDNA in HCC patients was 2 to 4 times higher than in those with chronic hepatitis, and nearly 20 times higher than in healthy individuals [105,106]. Additionally, tumor size, stage, and shorter survival time are correlated to cfDNA concentration. However, since cfDNA concentration can be influenced by liver inflammation, it would not be a reliable indicator for HCC, making it challenging to accurately determine the status of HCC [40].

Despite the clear evidence that analysis of alteration in cfDNA is highly beneficial for HCC detection, individual cfDNA alterations alone may lack sufficient diagnostic value for HCC [107]. On the other hand, the development of NGS technology has improved the understanding of the genetic alteration landscape in HCC. Recent studies using NGS have been able to detect most of the mutations present within tumors, and in some cases, the mutation detection rate was reported to be higher in liquid biopsy than in tissue biopsy. This indicates that intratumoral heterogeneity can be effectively identified, making it a powerful tool for improving early-stage diagnosis [108,109]. Guo et al. analyzed the cfDNA mutation profile using a targeted gene panel. High frequency of HCC-related gene mutations was observed in patients with chronic hepatitis B or liver cirrhosis. The area under a receiver operating characteristic curve (AUROC) of the gene mutation burden calculated from a panel of genes for HCC detection was 0.92 [110]. These results indicate that somatic mutation detected in cfDNA may be an excellent biomarker for the early detection of HCC. Also, genetic alteration of ctDNA provides valuable diagnostic value in combination with conventional serum protein tumor markers. Cohen et al applied CancerSEEK test, which is composed of protein biomarkers (CA-125, CEA, CA 19-9, hepatocyte growth factor, prolactin etc.) and 16 gene mutations of ctDNA (NRAS, CTNNB1, PIK3CA, FBSW7, APC, EGFR, BRAF, CDKN2A, PTEN, FGFR2, HRAS, KRA, AKT1, TP42, PPP2R1A, GNAS) to detect 8 different solid tumors, including liver cancer. Overall sensitivity of liver cancer was about 98% and specificity was over 99% [111]. These remarkable results suggest that cfDNA might play a crucial role in the early detection of HCC.

CfDNA methylation pattern also can provide valuable information to detect HCC [112]. Analyzing the methylation pattern of the SEPT9 promoter, an important tumor suppressor, in cfDNA demonstrated a potential role for HCC detection among cirrhotic patients [113]. HCC-specific circulating DNA methylation profiles identified 283 CpGs with methylation differences, enabling differentiation between HCC and non-HCC samples with AUROC of 0.957 (sensitivity=90%, specificity=97%) [114]. Blinded multicenter validation study using HelioLiver^TM test, which is generated using methylated cfDNA markers, serum protein markers and patient demographics, demonstrated better overall detection ability of HelioLiver^TM test (AUROC=0.944) compared to AFP (AUROC=0.851) and GALAD (gender, age, AFP-L3, AFP, and des-carboxy-prothrombin) score (AUROC=0.899). HelioLiver^TM test also showed better sensitivity (75.7%) for detecting early-stage HCC (I/II) compared to AFP (≥20 ng/mL, 57%) or GALAD score (≥−0.63, 65%) [58]. Another multicenter study used mt-HBT (multitarget HCC blood test), which combined 3 methylation markers (HOXA1, TSPYL5 and B3GALT6), AFP and patient sex, demonstrated superior sensitivity for early-stage HCC (I/II, 82%) detection compared to AFP (≥20 ng/mL, 40%) or GALAD (≥–0.63, 71%) [115]. Another study also demonstrated that combining cfDNA methylation markers with traditional serum protein biomarkers provides excellent accuracy for HCC detection, with an AUROC of 0.87 [116].

CfDNA fragment pattern also provides valuable information for detecting HCC. Non-tumor cfDNA fragments are, on average, 167 bp in size, while ctDNA fragments are typically shorter, around 145 bp [117]. This result is potentially due to the highly active transcriptional state of specific regions. A multi-national study that used whole genome cfDNA fragmentome profile was conducted to detect HCC patients using a machine learning model. The sensitivity for detecting cancer was 88% in an average-risk population at 98% specificity and 85% among high-risk individuals at 80% specificity.118 Another study analyzed cancer-related mutations and fragment length profile classification model using three fragmentomic features of ctDNA through deep sequencing of 13 genes associated with HCC. They demonstrated an AUROC of 0.88, a sensitivity of 89%, and a specificity of 82% for HCC detection in the discovery cohort [119]. Chen et al. [120] applied a combined profile of 5-Hydroxymethylcytosine/Motif/Fragmentation/nucleosome footprint using NGS technology and showed 95% sensitivity and 97% specificity for detecting HCC from cirrhotic patients. A recent meta-analysis, which analyzed 33 papers, reported an overall sensitivity of 72.2% and a specificity of 82.3% for 7 quantitative studies and an overall sensitivity of 56.8% and a specificity of 88.2% for 26 qualitative studies [121].

Prognostication using CTCs

Several studies have shown that the number of CTCs is correlated with tumor size, portal vein tumor thrombosis, tumor stage, differentiation, and shorter survival [15,143]. Our team developed the HCC-CTC mRNA scoring system for the prognosis of HCC. This system is composed of 2 steps: enrichment of CTCs from blood collected from HCC patients using the NanoVelcro CTC assay and lysis of the enriched CTCs to extract mRNA; second, reverse transcription to obtain HCC-CTC-derived cDNA. The extracted cDNA is reacted with the HCC-CTC RS panel, a final set of 10 genes with prognostic value, using the Nanostring nCounter technique. Based on the results, patients were categorized into low-risk and high-risk groups to compare their prognosis. In the validation cohort, the overall survival (OS) for the high-risk group showed significantly worse results (P=0.02), HCC-CTC panel was an independent predictor of survival (HR 5.7; 95% CI 1.5–21.3; P=0.009), along with BCLC stage [30]. A long-term study tracking patients with locally advanced or metastatic HCC reported that a significant increase in CTC count is associated with worsening treatment response [144]. The mesenchymal phenotype of CTCs (M-CTCs) is also associated with poor prognosis, including not only tumor aggressiveness and portal vein tumor thrombosis but also reduced OS [145,146].

Prognostication using cfDNA

CfDNA holds strong prognostic value for HCC due to its short half-life and reflection of tumor burden and biology. One study demonstrated that SNV and CNV of cfDNA correlate with tumor burden and predict prognostic outcomes better for recurrence-free survival (RFS) and OS than traditional protein biomarkers such as AFP, AFP-L3% and DCP [147]. A similar study showed that poor disease-free survival is observed in HCC patients with cfDNA mutations in their post-operative plasma [148]. Oh et al. analyzed the prognosis of HCC patients using cfDNA concentration and the I-score, which indicates genomic instability. cfDNA is an independent prognostic factor for OS (HR 2.51; 95% CI 1.62–3.89; P<0.001), while the I-score is an independent prognostic factor for both time to progression (TTP) (HR 1.71; 95% CI 1.13–2.58; P=0.011) and OS (HR 1.85; 95% CI 1.16–2.96, P=0.010) [149].

Multiple studies revealed the association between epigenetic alteration of cfDNA and HCC prognosis. One study developed a combined prognostic score (cp-score) from 8 methylation markers and revealed that the high-risk group based on cp-score is an independent predictor for survival (HR 1.548; 95% CI 1.246–1.924; P<0.001) [48]. A recent multicenter study validated 20 differentially methylated regions (DMRs), which predict a high recurrence risk (HR 3.33; 95% CI 1.87–5.92; P<0.001) [150]. A systematic review analyzing eight studies (5 studies with pre-treatment samples, 2 studies with pre & post-treatment samples, and 1 study with post-treatment samples) reported that the mutation of ctDNA was independently associated with reduced disease-free survival (HR 3.01; 95% CI 1.23–11.30) [151]. A study using data from participants in the SORAMIC trial showed that cfDNA levels significantly correlated with metastases and survival. In addition, cfDNA kinetics over time revealed a trend with the clinical history of the patients, supporting its use as a biomarker to monitor therapeutic response [152].

Prognostication using EVs

Several studies demonstrated the prognostic value of EV in HCC. High EV miRNA-21 levels positively correlated with the stage of HCC and linked to the PTEN/Akt signaling pathway, which enhances EMT and tumor proliferation. That suggests that EV miRNA-21 could be a potential predictive prognostic marker of HCC [153]. The EV miRNA-665, which is linked to the Mark/ERK anti-apoptotic pathway, also showed a correlation with HCC progression and shorter survival time [154]. Some studies revealed that the prognostic impact of EV is mediated by angiogenesis. EV miRNA-210 suppresses the expression of SMAD4/STAT6, therefore strengthening tumor angiogenesis [155]. Another study has reported that EV miR-103 inhibits VE-cadherin, p120, and zonula occludens, weakening endothelial cell junctions and ultimately leading to tumor metastasis [156]. On the other hand, the downregulation of tumor suppressor EV miRNAs can also indicate poor prognosis. Studies have reported that decreased EV miRNA-125b and miRNA-638 levels are associated with shorter time to recurrence (TTR) and OS [157,158]. EV LncRNA-ATB, which induces tumor invasion, EMT, and metastasis, is positively correlated to advanced HCC stage, portal vein thrombosis, leading to reduced OS and progression-free survival (PFS) [159]. EV circular RNAs also showed prognostic value for HCC. Upregulation of circPTGR1, which acts in decreasing interaction between miRNA-449 and MET, is associated with advanced cancer stage including metastasis and poor prognosis [160]. EV may predict the treatment response to immunotherapy. The high expression of circUHRF1 is associated with a decreased NK cell proportion in blood (P<0.001), which inhibits NK cell-derived IFN-γ and TNF-α secretion in HCC, ultimately leading to immunosuppression. Also, high expression of EV circUHRF1 is an independent prognostic factor for recurrence (HR 1.762; 95% CI 1.172–2.428; P=0.019) [161], suggesting that EV cirUHRF1 may induce poor response to anti-PD1 immunotherapy.