Liquid biopsy in hepatocellular carcinoma: circulating tumor cells and circulating tumor DNA

In recent decades, CTCs and ctDNA have been intensively detected in HCC patients. Tables 1 and 2 summarize the previously demonstrated characteristics of CTCs and ctDNA.

CTCs are widely believed to be a significant determinant of metastasis and recurrence in cancers and are not recommended as an independent HCC diagnostic tool [96]. Therefore, CTCs may serve as a potential biomarker for prognosis.

Currently, EpCAM⁺CTCs have been intensively investigated in HCC, although the knowledge about their clinical relevance in HCC is lagging behind other major tumor types, such as breast cancer, prostate cancer and lung cancer [97]. The overall survival (OS) and disease-free survival (DFS) were significantly shorter in the CTC-positive cohort with HCC [34] and also associated with poor clinical characteristics [40]. However, the definition of CTC positivity is still conflicting. Some researchers defined CTC positivity as “≥1 CTC” or “≥2 CTC”, but others used “≥5 CTC” to analyze results. However, one thing for sure is that the more CTCs are detected, the poorer prognosis the patients will have.

In addition, EpCAM⁺CTCs may serve as a real-time parameter for monitoring HCC recurrence. A preoperative CTC (7.5 ml) of ≥2 is a novel predictor for tumor recurrence in HCC patients after surgery, especially in patient subgroups with AFP levels of ≤400 ng/mL or low tumor recurrence risk [48]. Nevertheless, only approximately 35% of HCC cases express EpCAM, limiting the clinical application of EpCAM⁺CTCs in predicting prognosis. Some other methods have tried to overcome this dilemma. For instance, detection of EpCAM+CTCs with co-existing CD4 + CD25 + Foxp3+ Treg cells indicated HCC recurrence. Zhou et al. [98] discovered that patients with high CTC/Treg levels showed a significantly higher risk of developing postoperative HCC recurrence than those with low CTC/Treg levels (66.7% vs. 10.3%, P < 0.001). Apart from this attempt, other subtypes of CTCs for early prediction of HCC recurrence were also explored. Zhong et al. [99] found that of the 62 HCC patients, mesenchymal CTCs and mixed CTCs in the recurrence group were significantly higher than in the nonrecurrence group, and mesenchymal CTC positivity (HR = 3.453, P = 0.007) was an independent risk factor for early recurrence. A similar study [35] also found that CTC count ≥16 and mesenchymal-CTC percentage ≥ 2% before resection were significantly associated with early recurrence, multi-intrahepatic recurrence, and lung metastasis. Altogether, different markers and combinations in CTCs can serve as a tool for predicting HCC recurrence and prognosis, but those markers may not be specific to HCC and limit the use of CTCs in early diagnosis in some way.

In general, ctDNA mainly shows great diagnostic value in HCC. ctDNA is superior to that of previously described plasma biomarkers in terms of higher sensitivity and better clinical correlation.

In recent years, the “methylation pattern” of ctDNA has been the most intensively investigated hotspot [13]. Since methylation changes in ctDNA occur early in tumorigenesis and are potentially reversible, changes in methylation may offer the best hope for early cancer detection. Furthermore, methylation patterns are unique to each cell type and are highly stable under physiologic or pathologic conditions [100]. Therefore, the recognition of different methylation patterns may have the potential to serve as discriminatory tools for the detection and diagnosis of HCC. Wong and colleagues [81] detected for the first time that concurrent p15 and p16 methylation was positive in the plasma/serum of 92% (11 of 12) of HCC patients. Following this study, many researchers investigated the cfDNA methylation profile in HCC patients. For instance, Ras association domain family 1A (RASSF1A) promoter hypermethylation [77] was detected in 90% of the HCC group and could differentiate HCC patients from healthy controls and chronic HCV infection alone with an overall predictive accuracy of 77.5 and 72.5%, respectively. Therefore, aberrant promoter methylation in ctDNA could be evaluated as a screening tool for HCC patients, especially for small HCC among high-risk populations at an early stage.

Moreover, several hot methylated genes were combined to diagnose HCC. For example, p16, p15 and RASSF1A were explored in the serum of 50 HCC patients and provided an overall predictive accuracy of 89% with sensitivity and specificity of 84 and 94%, respectively [73]. A panel of four genes (APC, GSTP1, RASSF1A, and SFRP1, [68]) could discriminate HCC from normal controls with 92.7% sensitivity and 81.9% specificity. Additionally, three abnormally methylated genes (APC, COX2, RASSF1A) and one miRNA (miR-203) were combined to establish a predictive model by which nearly 75% of patients, who could not be diagnosed with AFP at 20 ng/mL, were detected [56]. Recently, Xu et al. [101] identified an HCC-specific methylation marker panel including ten markers and constructed a diagnostic prediction model that showed high diagnostic specificity (90.5%) and sensitivity (83.3%) superior to AFP (AUC 0.696 vs. 0.816), differentiating HCC patients from those with liver diseases or healthy controls. Interestingly, normal controls with positive detection likely to had etiological risk factors of HCC such as HBV infection and/or alcohol drinking history. In conclusion, hypermethylation in promoter regions is now recognized as an important early event in carcinogenesis. The combinations of different methylated tumor suppressor genes were absent or very low in normal tissues DNA. Therefore, detection of those DNA is specific to HCC and can diagnose HCC with high specificity and sensitivity.

Currently, the most commonly used serum marker AFP has limited diagnostic value because of low sensitivity of only 50% in HCC. cfDNA levels could discriminate HCC and HCV carriers at the optimal cutoff value of 73.0 ng/ml with a sensitivity of 69.2% and a specificity of 93.3% [76]. In addition, the combined AFP and cfDNA detection can improve HCC diagnosis. A study [102] showed an elevated diagnostic value with 95.1% sensitivity and 94.4% specificity in discriminating HCC from normal controls when using the combined detection of cfDNA and AFP. A similar study [103] found that the sensitivity of the combined detection of cfDNA with one marker (AFP or AFU) and cfDNA with two markers (AFP and AFU) was 71.8, 87.2 and 89.7% vs. 56.4, 53.8 and 66.7% for cfDNA, AFP and AFU when used alone, respectively. In AFP-negative HCC, the levels of plasma hTERT DNA in HCC patients with AFP ≤ 20 ng/ml were significantly higher than in HBV patients [69], indicating that hTERT DNA may serve as a novel complementary tool for AFP in the screening and detection of HCC. In a word, quantitative analysis of cfDNA is sensitive and feasible, and the diagnostic value of cfDNA is superior to AFP or AFU. Combined detection of cfDNA with AFP or AFU or both could improve diagnostic sensitivity of HCC. Furthermore, it is intriguing to imagine quantitation of cfDNA together with somatic mutation analysis to screen malignancy in the further. However, the definition of elevated cfDNA level is hard to decide and the optimal cut-off point is controversial in different researches because of various detection methods used.

Apart from its association with the malignancy, cfDNA could also be used as a potential biomarker to stratify liver fibrosis in patients with non-alcoholic fatty liver disease (NAFLD). Hardy, T. et al. [104] found that plasma DNA methylation of the peroxisome proliferator-activated receptor gamma (PPARγ) gene promoter increased with fibrosis severity and could be used as an independent predictor of fibrosis severity in NAFLD.

Except for early diagnosis, ctDNA also plays a vital role in HCC prognosis. For untargeted ctDNA, the levels of cfDNA are always negatively associated with DFS and OS [105, 106] and may serve as an independent prognostic factor for HCC recurrence and extrahepatic metastasis [107].

Targeted ctDNA detection can also reflect intratumoral heterogeneity and predict poor prognosis. Hot spot mutants, such as TP53, CTNNB1, and TERT, were the most preferred targets for the detection of ctDNA aberrations. Using droplet digital PCR, 27 of 48 HCC patients (56.3%) were found to have at least one kind of circulating mutant, with the mutant allele frequency ranging from 0.33 to 23.7% [58], indicating intra-tumoral heterogeneity in HCC. Another study used the MiSeq™ system to detect the same mutations, finding that ctDNA with mutations could be detected more easily in patients who suffered vascular invasion (P = 0.041) and predicted a shorter DFS time (P < 0.001, [57]). In addition, allelic imbalance (AI) on chromosome 8p at D8S264 in circulating plasma DNA was closely associated with the 3-year DFS (P = 0.014), and combined detection with cfDNA levels was independently associated with DFS (P = 0.018) and OS (P = 0.002) of patients with HCC [75]. In general, continuous detection of tumor-associated mutations in ctDNA can overcome the limitations of tumor heterogeneity and predict prognosis of HCC. However, there is still a problem because the circulating mutants are not all derived from tumor cells since germline mutation might also be detected in plasma. Therefore, it is necessary to screen corresponding tumor samples to identify their origins.

Surgical resection remains the primary treatment for HCC. ctDNA and CTCs are also used to assess the effect of surgical resection and monitor tumor burden. For instance, Wong et al. [79] found that the median p16INK4a methylation induced in plasma and buffy coat concordantly decreased 12- and 15-fold after surgical resection, respectively; Fan and his colleagues [108] used an orthotopic model by in vivo flow cytometry to detect CTCs in HCC, finding that the number of CTCs and early metastases decreased significantly after the resection, concluding that the resection prominently restricted hematogenous disseminating and distant metastases.

Subtypes of CTCs may associate with therapeutic response and could serve as a supplement for molecular subtype in HCC. Nel et al [109] investigated 11 patients with different therapies (watch and wait: 3; selective internal radiation therapy: 5; transarterial chemoembolization (TACE): 1; resection: 1; Nexavar: 1), detected a remarkable variation of CTCs with epithelial, mesenchymal, liver-specific, and mixed characteristics and found that different subgroups varied significantly among patients. Importantly, these CTC subgroups were associated with the therapeutic outcomes, and an increase in epithelial cells was associated with a worse treatment outcome in HCC patients. In addition, pERK+/pAkt- CTCs were more sensitive to sorafenib and remained an independent factor associated with a good prognosis (hazard ratio = 9.389, P < 0.01) in HCC patients [110]. Moreover, in vitro released CTCs were also evaluated for their sensitivity to chemotherapeutic drugs. Zhang at el [38] used a microfluidic chip to isolate and release viable CTCs, and the CTCs-treated sorafenib (5 samples) or oxaliplatin (5 samples) formed a significantly less number of spheroids compared to the control. To conclude, individual profiling of CTCs may have distinct clinical implications, which might help to predict outcome and potentially to select the appropriate treatment.

Importantly, somatic mutation frequency of ctDNA reflected clinical dynamics corresponding to sequential therapy and might provide a possible solution for monitoring tumor burden and prognosis. Cai et al. [111] monitored an HCC patient during the course of comprehensive therapy and observed an increased circulating level of 8 somatic mutations even before imaging diagnosis and the increase of AFP levels after the first TACE treatment. In this process, 1 nonsynonymous somatic mutation (HCKp. V174 M) was identified after the first TACE treatment and then became undetectable after the second surgery, while it sharply increased during the second recurrence. Therefore, ctDNA could track the change of different mutants and therapeutic responses in real-time longitudinal monitoring.

Moreover, somatic mutations detected in ctDNA can guide therapy. Ikeda et al. [54] evaluated 14 patients with advanced HCC. A patient with a CDKN2A-inactivating and a CTNNB1-activating mutation received palbociclib (CDK4/6 inhibitor) and celecoxib (COX-2/Wnt inhibitor) treatment and found low levels of AFP at baseline at 2 months. Another patient with a PTEN-inactivating and a MET-activating mutation received sirolimus (mechanistic target of rapamycin inhibitor) and cabozantinib (MET inhibitor) and found AFP declined by 63% (8320 to 3045 ng/mL). In conclusion, ctDNA derived from noninvasive blood tests can provide exploitable genomic profiles in patients with HCC and guide therapy in some ways (Fig. 2).

In general, liquid biopsy is an exciting method that is noninvasive, overcomes tumor heterogeneity and can monitor tumor progression, recurrence or therapeutic response in real time. There is also an ongoing clinical trial (NCT02973204) from the US National Laboratory of Medicine (NIH) for CTCs and ctDNA in HCC, aiming at predicting which patients require special monitoring and individualized therapy and exploring the value of these tests in supporting clinical decision-making. However, there are challenges that researchers face in translating liquid biopsy from bench to bedside that still stand in the way.

First, the biological basis of CTCs or ctDNA is controversial; for example, the exact mechanisms by which ctDNA is released into the blood remains to be clarified. Apoptosis and necrosis are the most discussed origins of cfDNA [112]. Autophagy and hypoxia might account for ctDNA release. Malignant cells could activate autophagy as a survival mechanism, and autophagic activity regulates apoptosis; as a result, ctDNA is released [113]. Exposures to intermittent hypoxia (IH) could increase the shedding of ctDNA into the circulation [114]. Therefore, it is important to gain a better understanding of the biological characteristics of the liquid biopsy.

The descriptions of CTCs and ctDNA in this review are mainly focused on HCC. However, many other types of cancers may detect the same mutations in genes, such as TP53, KRAS, or BRAF. Cancer-associated mutations occur with increasing age, raising a problem that cannot be neglected: how could cfDNA be tissue-specific? The epigenetic biomarkers of cfDNA may answer this question [115]. DNA methylation is the most favored epigenetic modification whose profile is highly tissue-specific and shows the potential to determine the tissue origin of cfDNA [116‐119]. In addition, the nucleosome occupancy and the ending pattern of cfDNA molecules also reflect the original tissues. By deep sequencing cfDNA, Snyder et al. generated maps of genome-wide in vivo nucleosome occupancy and found nucleosome spacing patterns, providing the information of the tissues of origin for cfDNA [120]. There were millions of plasma ctDNA end coordinates across the genome, and HCC-associated ctDNA had preferred end signatures different from other cfDNA [121], making it possible to trace the origin of ctDNA.

Moreover, liquid biopsy, especially ctDNA, shows considerable potential in classifying molecular subtyping of certain malignancies. Plasma ctDNA genotyping can classify transcriptionally defined tumor subtypes of diffuse large B cell lymphoma (DLBCL). Biopsy-free plasma genotyping of ctDNA in 41 DLBCL patients was highly consistent with tumor tissue biopsy classification (88%) [122], which may facilitate individualized therapy. Martínez-Ricarte F et al developed a sequencing platform to simultaneously and rapidly genotype seven genes (IDH1, IDH2, TP53, TERT, ATRX, H3F3A, and HIST1H3B) in CSF ctDNA, allowing the subclassification of diffuse gliomas, the most common primary tumor of the brain, having different subtypes with diverse prognoses [123].

Nevertheless, liquid biopsy has its own limitations. The techniques of collecting “liquid”, and the isolation, enrichment or detection of CTCs and ctDNA must be standardized. Most of the current studies used different technologies or assays to detect CTCs or ctDNA, resulting in diverse sensitivity and specificity. Although ctDNA has high specificity in diagnosis, a multi-marker analysis may offer a more comprehensive insight into cancer specificity [55, 56, 101]. In addition, more multicenter, larger and longer-term studies are urgently needed for the implementation of the liquid biopsies clinically, including clinical trials. Currently, most of the data gathered have been within proof-of-concept studies and lack of credibility. Altogether, the liquid biopsy is a critical part of precise medicine, and is believed will become a clinical reality in near future.