Article Text

Download PDFPDF

Liquid biopsy for liver diseases
  1. Jelena Mann1,
  2. Helen L Reeves2,
  3. Ariel E Feldstein3
  1. 1 Newcastle Fibrosis Research Group, Institute of Cellular Medicine, Faculty of Medical Sciences, Newcastle University, Newcastle upon Tyne, UK
  2. 2 Northern Institute for Cancer Research, Newcastle University, Newcastle upon Tyne, UK
  3. 3 Department of Pediatrics, University of California, San Diego, California, USA
  1. Correspondence to Dr Ariel E Feldstein, Division of Pediatric Gastroenterology, Hepatology and Nutrition UCSD, San Diego, CA 92103-8450, USA; afeldstein{at}


With the growing number of novel therapeutic approaches for liver diseases, significant research efforts have been devoted to the development of liquid biopsy tools for precision medicine. This can be defined as non-invasive reliable biomarkers that can supplement and eventually replace the invasive liver biopsy for diagnosis, disease stratification and monitoring of response to therapeutic interventions. Similarly, detection of liver cancer at an earlier stage of the disease, potentially susceptible to curative resection, can be critical to improve patient survival. Circulating extracellular vesicles, nucleic acids (DNA and RNA) and tumour cells have emerged as attractive liquid biopsy candidates because they fulfil many of the key characteristics of an ideal biomarker. In this review, we summarise the currently available information regarding these promising and potential transformative tools, as well as the issues still needed to be addressed for adopting various liquid biopsy approaches into clinical practice. These studies may pave the way to the development of a new generation of reliable, mechanism-based disease biomarkers.

  • clinical trials
  • chronic liver disease
  • hepatocellular carcinoma
  • liver biopsy

Statistics from


Liver biopsy remains the gold standard procedure for the diagnosis of various chronic liver diseases and for assessing the presence, pattern and severity of liver injury, inflammation and the stage of fibrosis. Outside of specific causes of liver disease such as viral hepatitis, currently available non-invasive blood tests lack sensitivity and specificity and have limited utility in general for establishing both a specific diagnosis as well as the severity of liver damage. While various imaging modalities such as MRI-based and ultrasound-based elastography are widely used for assessment of liver steatosis and fibrosis, these techniques lack sensitivity and specificity for early stages of fibrosis and are not useful in the determination of inflammation and hepatocellular injury. Biomarkers that can fill these gaps might provide significant advances in the approach to diagnosis and monitoring of disease progression/regression in the clinical setting. These biomarkers should fulfil a number of attributes, including high sensitivity and specificity, and be able to determine prognosis and predict response to therapy.

In the context of hepatocellular carcinoma (HCC), the role of liver biopsy is currently restricted to the diagnosis of cancer in the absence of established cirrhosis, or diagnosis of radiologically atypical lesions in the presence of cirrhosis.1 The lack of more ‘routine’ use of biopsy to confirm HCC diagnosis is often criticised and regularly debated. In patients with cirrhosis, radiological diagnosis in the presence of typical features has high specificity, and biopsy, which is associated with patient morbidity and small risks of haemorrhage and tumour seeding, is unnecessary. On the other hand, the lack of tissues for research studies lies at the heart of failures, progressing improvements in mortality associated with this condition.2 While HCC tumour heterogeneity may limit the usefulness of biopsy beyond confirming diagnosis, as a single biopsy is often poorly representative,3 there is little doubt that as the choice of first-line and second-line therapies for HCC increases,1 biomarkers to guide and monitor treatments are urgently required. The hope is that in combination with liquid biopsy tools, more robust, reproducible biomarkers, unrestricted by tumour heterogeneity, will emerge.4 For surveillance, diagnosis and monitoring of treated patients, rapid serum or plasma-based tests which can be repeated have great potential. Circulating tumour cells (CTCs) complementing studies of nucleic acids and other peripheral blood cells may also have key roles in the management of cancers,5 including HCC.6

Extracellular vesicles

Extracellular vesicles (EVs) are small membrane vesicles released by cells in the extracellular environment as part of normal physiology or during pathological developments, a process conserved throughout evolution from bacteria to humans and plants.7 EVs are effective communicators that are generated by a cell of origin or parenteral cell and can act on a number of target cells in the environment they are released into, as well as at distant sites acting as long-range signals.8 EV cargo may reflect the cell of origin as well as the specific stress that induces their formation and release. They transport a variety of bioactive molecules, including mRNA, miRNAs, proteins and lipids, that can be transferred among cells, regulating various cell responses.7 EVs can be generally classified based on their size and mode of biogenesis into two main categories: exosomes or ectosomes (also called microparticles (MPs) or microvesicles). Exosomes are approximately 30–100 nm in size and are intraluminal vesicles formed by the inward budding of endosomal membrane during maturation of multivesicular endosomes (MVEs), which are intermediates within the endosomal system, and secreted on fusion of MVEs with the cell surface. This process depends on molecular components and mechanisms of the endosomal machinery, including the endosomal sorting complex required for transport proteins. Ectosomes—also called MPs or microvesicles—are approximately 100–1000 nm in size and are released from outward budding and fission of the plasma membrane and the subsequent release of vesicles into the extracellular space in a process dependent on caspase and stress kinases activation.

A growing number of studies have provided evidence for a key pathophysiological role of EVs in various aspects of liver injury including inflammation, and fibrogenesis particularly in disease processes such as non-alcoholic steatohepatitis (NASH) and alcoholic liver disease (ALD).9 10 These findings, in conjunction with the fact that EVs are released into the systemic circulation and are remarkably stable in this environment, support the concept that assessment and quantitation of EVs in blood might represent a novel form of liquid liver biopsy (figure 1). Different approaches have been used, including analysing changes in EV subpopulations and cell-specific EVs by using selective surface markers or EV cargo, as well as untargeted comprehensive approaches, to assess EV composition such as protein, lipids or RNA (figure 1). Most studies have focused on the two former approaches. Kornek and colleagues11 12 reported on the profile of blood MPs using fluorescence-activated cell sorting (FACS) analysis with selective leuco-endothelial surface markers in patients with non-alcoholic liver disease (NAFLD) and chronic hepatitis C (CHC), and in healthy controls. They observed an increase in MPs enriched in surface markers from monocyte and invariant natural killer T cells and a decrease in MPs derived from neutrophils and leuco-endothelial cells in patients with NAFLD compared with both patients with CHC and healthy controls.12 While MPs derived from both CD4+ and CD8+ T cells were elevated in both patients with CHC and NAFLD compared with controls, their levels could not discriminate between these two diseases.12 Using similar leuco-endothelial cell markers, Rautou et al analysed the profile of circulating MPs in patients with cirrhosis secondary to CHC and ALD.13 14 Although the levels of MPs enriched in leuco-endothelial cell markers were higher in patients with cirrhosis compared with healthy controls, and in general increased with the severity of cirrhosis assessed using Child-Pugh scores, there was significant overlap in the values among patients with Child-Pugh A through C.13 Indeed myeloid cell-derived EVs have been shown to be released in circulation in a wide variety of inflammatory conditions, which could be predicted to impact their utility as biomarkers for liver disease. In order to improve the specificity, several groups have tried to identify liver-specific protein markers for blood EV quantitation, such as asialoglycoprotein receptor 1 (ASGPR1), cytokeratin-18 and vanin-1.14–18 We recently examined the proteome of circulating EVs in mouse models of NAFLD/NASH and demonstrate that these vesicles carry a selective antigenic composition.19 Moreover, hierarchical clustering analysis allowed for discrimination of mice with established NASH from those with isolated steatosis and normal livers with high accuracy, suggesting that a potential protein signature in blood EVs might be used to diagnose NASH non-invasively. While we are currently validating these results in patients with NASH, Arbelaiz and colleagues20 recently reported a similar approach to compare serum EVs from patients with primary sclerosing cholangitis (PSC), cholangiocarcinoma (CCA) and HCC and from healthy controls. The circulating EV proteome profiles by mass spectrometry revealed multiple differentially expressed proteins among groups. Using individual proteins identified to be differentially expressed, they showed similar area under the receiver operating characteristic curve to that from established tumour markers. Unfortunately, the diagnostic values of various combinations of the top differentially expressed proteins were not reported, and future studies are needed to further determine whether proteomic signatures in serum EV can be developed as a useful diagnostic tool for patients with CCA, PSC and HCC.

Figure 1

Extracellular vesicles (EVs) are released from a variety of different cell types in the liver under physiological and pathophysiological conditions into the extracellular space where they interact with neighbouring target cells, as well as into the systemic circulation where they can be isolated and quantified using a variety of different methodologies. EV surface markers and cargo may reflect the cell of origin as well as the specific stress that induces their formation and release.

Besides proteins another EV payload that has become of interest for the development of EV-based liver biomarkers is microRNAs. These small non-coding RNAs are released into the extracellular space where they are protected from degradation either by association with different proteins such as those from the Ago2 family or protection within EVs. MicroRNA profiling provides the advantage of their tissue specificity. In the case of the liver, particular attention has been focused on miR-122 and miR-192 due to liver specificity and enrichment representing over 70% of the total microRNA liver population.21 Notably, the partitioning of miR-122 in the circulation to either the protein-rich or EV compartments might represent a disease-specific or liver damage pattern-specific phenomenon. Indeed, Pirola et al 22 demonstrated that as opposed to healthy individuals where miR-122 is present in the circulation only in the Ago2 complex fraction, in patients with NAFLD the majority of serum miR-122 circulates in Ago2-free forms. The specific compartment was not further assessed in this study. In this context, using experimental models of NAFLD, we have found that circulating miR-122 is enriched in the vesicular fraction in plasma,19 the levels are dynamic, increasing over time and correlating with histological features of disease severity. Further support for the importance of compartmentation was provided by work from Bala et al.23 These authors elegantly showed that the enrichment of miR-122 in one particular compartment in circulation was dependent on the type of liver injury, being predominantly in EVs in models of ALD and in protein-enriched fractions in acetaminophen-induced liver injury. Future studies addressing the importance of changes on both concentration and compartmentalisation of circulating microRNAs and the potential occurrence of differential packaging of circulating microRNA in different forms of liver diseases may have significant impact on the development of EV-associated microRNAs as liver biomarkers. Using an untargeted approach by profiling the microRNA composition from EVs released by hepatocytes isolated from the intragastric infusion model of ALD, we recently identified a number of differentially expressed microRNAs.24 Furthermore, a specific hepatocyte-derived microRNA signature including miR-29a, miR-340, let7f and miR-30a was contained in circulating EVs of these mice and could accurately identify the presence of ALD versus controls. In addition, this signature was disease-specific as it allowed for discrimination of ALD versus NAFLD, NASH and cholestatic liver disease. Finally, we observed this signature in blood EVs from ambulatory patients with mild ALD in a pilot proof-of-concept study.24 In summary, growing evidence from both various experimental models of liver injury as well as from patients with different types of liver diseases is pointing to the potential role of EVs as a novel liquid liver biopsy approach and opens the possibility of a new era of precision medicine where disease-specific and liver injury-specific signatures can be accurately identified in a non-invasive manner.

Circulating nucleic acids

One of the novel trends in the development of non-invasive biomarkers for liver disease is the use of circulating nucleic acid found in the serum or plasma of patients, which comprises fragments of genomic DNA, frequently referred to as cell-free DNA (cfDNA), as well as RNA. Both cfDNA and cfRNA fractions found in the serum or plasma originate from dying cells, thereby providing a valuable source of material which can instruct about the nature of biological and pathological processes within its cellular source. In liver disease, circulating nucleic acids serve several purposes in terms of liquid biopsy, which can provide diagnostic and prognostic tools that screen and inform about the state of the underlying aetiology of disease. Importantly, they can also be useful in the screening of HCC, which often develops on the background of liver cirrhosis. Utility of cfDNA and cfRNA ranges from risk prediction for the development and early detection of cancer, to treatment response, recurrence monitoring and tumour molecular profiling. In addition to cfDNA, in HCC there is a fraction of genomic DNA that is derived from tumour cells, thus being termed circulating tumour DNA (ctDNA). Each of the nucleic acid moieties mentioned has a prominent role in different aspects of diagnosis or prognosis, and will be briefly considered in turn, with some of the latest developments mentioned in more detail (see figure 2).

Figure 2

Cell-free nucleic acids (RNA and DNA) are released from a variety of dying hepatic cells; these nucleic acids then enter systemic circulation where they can be isolated from serum or plasma and quantified using a number of methods. Screening cell-free nucleic acids for specific epigenetic modifications, such as DNA methylation, as well as sequence alterations can aid fibrosis grading and early HCC detection and/or recurrence .

Cell-free DNA

cfDNA was originally discovered in 1948. It has since been shown that the fragments of cfDNA have an average size of ~180 base pairs (bp), roughly coinciding with the length of DNA found in the smallest unit of chromatin, the mononucleosome. It is therefore most likely that the cfDNA fragments are released by cells undergoing apoptosis. In some conditions, fragments of longer or significantly shorter length have been found, suggesting that necrosis may also contribute to the pool of cfDNA.25 26 It was originally thought that the higher level of cfDNA in the blood may be an indicator of tumour growth; however, it has since been realised that many other conditions, including autoimmune diseases, result in similar cfDNA increase. The focus of more recent studies has therefore shifted towards epigenetic characteristics of cfDNA, such as its methylation signature (table 1). Methylation pattern is a prominent characteristic of DNA and one that is exquisitely cell type-specific. During development and progression of all types of liver disease, the pattern of DNA methylation changes over time in multiple hepatic cell types. In terms of disease pathology, increasing grade of liver fibrosis was shown to be the single most defining feature linked to disease progression. In the absence of non-invasive tests that can discern intermediate grades of fibrosis, release of methylated cfDNA from dying hepatocytes or other liver cells poses a possibility to track fibrosis grades by detecting altered methylation signatures in plasma. A study by Hardy and colleagues27 demonstrated one of the first examples of this approach. NAFLD is an increasingly common disease affecting up to one-third of the population, particularly in North America and Europe. Although research into non-invasive diagnostic approaches and staging of NAFLD via blood tests or imaging techniques has gradually advanced, one-third of patients with NAFLD cannot be diagnosed or staged accurately with current methods. Hardy et al 27 showed that cfDNA methylation at the PPARγ gene promoter in a patient’s plasma rises in correlation with hepatic fibrosis stage. The CpG methylation was lowest in control patients and highest in a cohort of patients with NAFLD who had biopsy-proven severe fibrosis. Similar trends were also observed in ALD cirrhosis. Interestingly, the source of plasma cfDNA harbouring higher CpG methylation at the PPARγ promoter appeared to be hepatocyte nodules rather than myofibroblast-rich scar tissue. In addition, a test based on plasma DNA methylation at the PPARγ gene promoter outperforms the commonly used NAFLD fibrosis score. A follow-up study confirmed that high CpG methylation at PPARγ promoter is associated with severe fibrosis/cirrhosis in NAFLD and in HBV, suggesting that this phenomenon is not limited to metabolic liver diseases.28 Furthermore, this methylation signature remains unaffected by the presence of HCC. Intriguingly, the signature appears to be specific to hepatic fibrosis as patients with diffuse systemic sclerosis, where a number of organs are affected by severe fibrosis, but not the liver itself, have very low level of DNA methylation at the PPARγ promoter.28

Table 1

Summary of cell-free DNA epigenetic modifications identified in liver fibrosis and HCC

Changes in DNA methylation have been shown to be pivotal in the development of HCC, particularly within the CpG islands of tumour suppressor genes. A number of studies have analysed methylation patterns in HCC tumour resections and reported hypermethylation of p15, CDKN2A (encoding for p16), glutathione S-transferase P1 (GSTP1), Ras association domain family 1A (RASSF1A), APC, SOCS1, SOCS3, TIMP3, blood vessel epicardial substance (BVES) and Homeobox A9 (HOXA9) genes and hypomethylation of long interspersed element-1 (LINE-1) repetitive sequence within the tumours.6 29–35 However, the same methylation pattern was detectable in only a subproportion of the reported genes found in cfDNA/ctDNA; for example, hypermethylation of GSTP1 was detected in 50% of the cfDNA, RASSF1A was found in 70%–93% (depending on the study), whereas hypomethylation of LINE-1 was present in 66.7% of cfDNA in the sera of patients with HCC.32 33 GSTP1 methylation in cfDNA was analysed in a number of studies, collectively showing specificity of 70%–91% and sensitivity of 50%–75% as a diagnostic marker of HCC, thus outperforming alpha-fetoprotein (AFP) which is the current diagnostic utility often used alongside contrast-enhanced cross-sectional imaging to confirm the presence of a tumour.29 32 34 RASSF1A methylation in cfDNA showed a positive correlation with tumour size, while LINE-1 hypomethylation was associated with tumour progression and poor prognosis. Combination of LINE-1 and RASSF1A methylation measurement in cfDNA was able to predict recurrence of disease following a resection of HCC.33 A recent study using ctDNA from a cohort of 1098 patients with HCC and 835 controls obtained 10 methylation markers in cfDNA with which a diagnostic model was constructed. Once applied, this model had a sensitivity of 85.7% and specificity of 94.3% for HCC in algorithm training cohort comprising 560 normal samples and 715 HCC.36 When applied to a validation cohort of 383 HCC and 275 normal samples, this algorithm achieved a sensitivity of 83.3% and a specificity of 90.5%, thereby demonstrating that this approach differentiated HCC from normal control with an area under the curve (AUC) of 0.966.36

Current data are largely gathered in proof-of-concept studies, which collectively suggest that cfDNA/ctDNA may in near future provide a superior tool for early tumour detection.6 29 30 Such optimistic outlook is based on several factors, including the use of highly sensitive techniques such as digital PCR and NanoString to measure vanishingly small amount of cfDNA in the serum or plasma.37–39 Indeed, these approaches may revolutionise early detection given that a study published a decade ago found that aberrant methylation of cfDNA fragments was detectable up to 9 years before standard methods diagnosed HCC.40 Furthermore, cfDNA is not restricted to blood, but can also be readily isolated from urine.38 A recent study showed that monitoring of methylation changes in GSTP1 and RASSF1A cfDNA in the urine of patients following HCC resection allowed early detection of recurrence, which was identified up to 9 months before the disease was detected by MRI.38

In addition to using cfDNA in HCC for detection of changes in methylation pattern, the fragments have also been sequenced to monitor for the presence of tumour-associated mutations.41–43 In this context, it is the ctDNA released by the dying tumour cells that provides valuable material for molecular analysis as they contain genetic defects found in the tumour, including gene rearrangements, gene copy number variations and a spectrum of mutations.37 41 43 Analysis of genetic defects allows for tracking molecular as well as cellular heterogeneity, which can be particularly important once multiple metastatic sites develop that may not be initially radiologically detectable. Delineating the particular set of mutations present in a patient can lead to personalised medicine screening, particularly in situations where distinct drug resistance mechanisms may be developing. Although there are no currently approved clinical tests employing cfDNA mutational screening in liver disease of HCC, a precedent has been set with clinical screening tests for EGFR mutational status in non-small-cell lung cancer which provide treatment indication for anti-EGFR therapeutics.44

Cell-free non-coding RNA

Cell-free non-coding RNA comprises long and short species of RNA, both of which are involved in the regulation of gene expression. The long non-coding species (lncRNAs) are RNA molecules with a length of >200 bp. The role of lncRNA in liver diseases and fibrosis has not yet been fully defined; however, there are a number of studies that are attempting to define the longevity of lncRNAs in plasma or serum, with main technical difficulties relating to collection, processing and accurate measurement of lncRNAs. However, better progress has been made in analyses of lncRNAs in the plasma of patients with HCC, where at least 10 different lncRNA species have been reported to exhibit detectable levels. MALAT1 and lncSPRY4-1T1 levels increase with grade of HCC, with resection and removal of HCC bulk leading to their decrease in circulation.45 46 Plasma levels of LINC00152, XLOC014172 and RP11-160H22.5 are able to distinguish HCC from either chronic hepatitis, cirrhosis or normal liver. When combined with AFP, these three lncRNAs have been shown to generate a much more sensitive test for HCC, with an AUC of 0.986 and 0.985.47

The short non-coding RNA species are less than ~28 bp long, of which micro RNAs (miRs) are the most studied category. miRs generally target 3’UTR of mRNA, inducing its degradation, or alternatively they can act to block translation. miRs play numerous roles in a variety of biological processes, including development, cell differentiation, metabolism and cell death. Their importance in hepatology has been based on differential expressions between diseased and normal liver, as well as HCC and non-tumour tissue. There have been a number of recently published reviews that cover the topic of circulating cell-free miR (cf-miRs) in the context of liver diseases and fibrosis.48–51 The majority of the findings relate to patients with HBV and HCV, with fewer studies carried out in NAFLD and ALD. Figure 3 shows the identity of upregulated or downregulated cf-miRs listed by the disease aetiology; the data shown have been gathered from recent reviews. There are numerous miRs that change their circulating level as a result of a particular aetiology. Overlap between circulating miRs in different aetiologies suggests that the presence of hepatic condition/fibrosis per se may result in alterations in particular miR signatures. Indeed, two studies suggest the use of an algorithm based on several cf-miR levels to track the grade of disease. A very recent study shows that the use of circulating miR-122 level in combination with Wisteria floribunda agglutinin-positive Mac-2 binding protein achieves 47% sensitivity, 87% specificity and overall accuracy of 74% in detecting advanced HBV fibrosis.52

Figure 3

(A) Upregulated miRs in the plasma of patients with various liver disease aetiologies. (B) Downregulated miRs in the plasma of patients with various liver disease aetiologies. The figure is based on summarised data in several recent reviews. ALD, alcoholic liver disease; miRs, micro RNAs; NAFLD, non-alcoholic liver disease.  

Circulating miRs have been extensively studied in HCC, with results showing promising diagnostic and prognostic value. Initial studies uncovered a panel of cf-miRs including miR-21, miR-26a, miR-27a, miR-122, miR-192, miR-223 and miR-801 with high diagnostic accuracy for HCC and ability to discriminate patients with HBV, cirrhosis or even control patients, measuring an AUC of 0.84, 0.88 and 0.94, respectively.53 54 More recently, a panel of miR-29b, miR-122 and miR-885 in combination with AFP showed high diagnostic accuracy in detecting HCC within normal population, while a combination of AFP with miR-22, miR-122, miR-221 and miR-885 was superior at the early diagnosis of HCC on the background of cirrhosis. miR-224 and miR-500 were found to reflect the tumour burden, as both decreased following HCC resection.55 Meta-analysis of 24 previously published studies identified miR-21, miR-122 and miR-199 as the most reliable diagnostic markers for HCC.56 A very recent study used RNAseq on plasma samples from patients with HCC, identifying miR-101–3p, miR-106b-3p and miR-1246 with best prognostic performance and AUC of 0.9.57 The same panel of cf-miRs was then tested comparing HCC with healthy controls, producing an AUC of 1.00. Finally, these miRs were measured in the serum of patients with HCC versus patients with cirrhosis, where once again they proved to have very high diagnostic accuracy (AUC=0.96).57

Circulating tumour cells

For cancers to metastasise, whole living cells need to escape the primary tumour, survive in the circulation and then successfully establish themselves in a new ‘metastatic niche’. The biology of metastatic spread most likely involves interaction with other circulating cells as well as with the target environment (figure 4) but is poorly understood. It is the rarity of CTCs, in the region of one per billion blood cells in patients with metastatic disease,58 that has hampered both mechanistic and candidate biomarker studies.

Figure 4

Despite the rarity of circulating tumour cells (CTC), in the region of one per billion blood cells in patients with metastatic disease, advances in technologies now enable their detection and isolation, supporting both biomarker and mechanistic biology studies. HCC, hepatocellular carcinoma.

Earlier PCR-based methods detecting tumour-specific mRNA in blood cells, sometimes after depletion of peripheral blood mononuclear cells, were proposed as surrogates for CTCs that had prognostic benefit. In patients with HCC, the initial focus was on the detection of the established tumour biomarker, AFP.59 Technical issues with methodologies ultimately limited the usefulness of this approach, as did inherent difficulties associated with using AFP—a biomarker elevated mostly in advanced stage disease and in a subset of patients only. Using combinations of biomarkers did not help, and as cell isolation methods have advanced and the presence of circulating tumour nucleic acids has been realised, the focus has turned to detection of whole tumour cells or cancer-specific abnormalities in ctDNA or ctRNA in the serum or plasma.60 61

From the late 1990s, the methods for isolating CTCs have steadily evolved. There are two general categories of these, basing detection on either known biological or physical properties of tumour cells. Biological methods presently exploit known tumour-specific epitopes for which antibodies have been developed, in combination with a means of enrichment. The CellSearch system (Veridex) employs antibodies recognising EpCAM and cytokeratin coated onto immunomagnetic beads to capture CTCs and has been approved by the US Food and Drug Administration for use in patients with breast, prostate or colorectal cancers.62 This method though is reliant on tumour expression of the target proteins and is of limited use for tumours which do not typically express them. A number of studies indicate that EpCAM or cytokeratin-positive CTCs are present in only a third of patients with HCC.63 64 To try to combat this, some research groups have broadened target epitopes to include candidates such as ASGPR, Hepar 1 and carbamoyl phosphate synthetase 1.65 The physical features used for CTC selection commonly include size, density, deformability and electric charge, employing centrifugation and filters or flow devices with channels of varying size or nature. These physical methods are not reliant on the expression of known cancer epitopes but may be less cancer-specific. They may also be subject to selection bias. Selection based on larger size, for example, will miss smaller CTCs.

At the present time, beyond the technical success of ‘detection’, studies in the CTC HCC field have centred largely on improving sensitivity and specificity. Some researchers have combined broader selection with genetic analyses of isolated cells, such as HER-2 amplification or TP53 deletion,66 cytogenetic abnormalities,67 or RNA expression signatures,68 while others have used downstream immunohistochemistry, for example with the HCC morphogen promoting Wnt signalling, glypican-3.69 While the explosion in technologies supporting CTC detection, isolation and characterisation is exciting, the wide range being adopted by clinical researchers, each with its own advantages and disadvantages, is a hindrance in some respect as few studies have been reproduced by more than one research group. For incorporation into clinical trials and practice, standardised protocols with reproducible results are needed and are currently lacking in the HCC field.

In combination, though, clear associations between enumeration and HCC stage, as well as prognosis, within defined groups of patients are starting to emerge.64 66 70 Similarly, predictive associations between enumeration and recurrence postresection71 have been reported. However, as these methods are all costly, the use of CTC enumeration as a useful clinical tool would need to be firmly established as advantageous over and above the evidence-based prognostic or stratification tools currently used, which may be an unrealistic goal. Beyond enumeration, phenotypic characterisation of CTC may be much closer to clinical utility, as currently tissue-based predictive biomarkers that aid treatment stratification or choice are lacking. It is also clear from a number of studies now, including our own,64 that CTCs are a heterogeneous population within the circulation and are therefore more likely than tissue biopsy to reflect ‘the whole tumor’. As phenotypic characterisation can be chosen to study key targets, it is here that academic and industrial partnerships might have the greatest impact. As proof of principle, the CTC extracellular signal-regulated kinases (ERK/pAKT) phenotype has recently been reported to predict sensitivity to sorafenib.72 Going forward, phenotyped CTCs may be surrogates for guiding enrichment trials with therapies targeting c-myc or FGFR4, for example. Methods to capture living CTCs from patients with HCC and culture them into three-dimensional spheroid-like structures have also been developed, taking the opportunities for personalised medicine to a whole new level. In the study by Zhang et al,73 individual sensitivity to sorafenib and oxaliplatin was explored, but screening multiple candidates in this fashion, perhaps in combination with patients’ CTC-derived xenografts,74 may become a future reality.

Liquid biopsy also offers the opportunity to characterise CTC interactions with circulating immune cells, and the combination of elevated EpCAM (mRNA+) CTC and regulatory CD4+ T cells has recently been reported to be indicative of early HCC recurrence and poor clinical outcome.75 As the oncology community as a whole embraces cancer immunotherapies, the opportunities to define circulating tumour immune cell interactions as predictive tools have potential, as does their coculture for mechanistic study.

While the integration of CTCs as clinical tools guiding clinical decision making has huge potential, perhaps the most innovative contributions of CTC studies will be in advancing our understanding of the biology of metastatic disease, alongside the development of strategies to combat it. Technological advances in next-generation sequencing have been used to detect and analyse CTCs at the molecular level, confirming the heterogeneity of cancer cell populations and potentially identifying new driver mutations responsible for the tumour metastasis and resistance of tumours to chemotherapies.76 Beyond these landmark developments in transcriptomics, where the full potential will not be realised for some time, other key physical and functional attributes are being appreciated. Sun et al 77 have recently characterised and compared CTCs isolated from peripheral veins and arteries, as well as the portal vein taking blood into the liver and the hepatic vein taking blood out of the liver, prior to HCC resection. The greatest number of CTCs was detected in the hepatic vein going out of the liver, with a dramatic reduction in peripheral vessels after passage through the lungs.77 Furthermore, single cell characterisation of CTCs confirmed phenotypic heterogeneity across vascular sites, with a predominantly epithelial phenotype at release into the hepatic vein, versus an epithelial mesenchyme transformed (EMT) one—associated with Smad2 and β-catenin signalling—in CTCs isolated in the peripheral veins. These cells either survived longer or underwent EMT after release into the circulation. The CTC burden, as well as the presence of CTC clusters, in both hepatic and peripheral veins predicted lung and liver metastases on follow-up studies. Meanwhile, single cell polarity has recently been shown to be critical for adhesion, attachment, migration and metastasis of CTCs, with manipulation of polarity proposed as a novel therapeutic target.78 Maintaining the adhesion theme, Talin1 is a component of focal adhesions, and its expression in CTCs has also been proposed as a mechanism facilitating CTC attachment to fibronectin-rich liver endothelial cells promoting metastatic spread.79 Talin1 depletion or inhibition may also be worthy of therapeutic exploitation.

Finally, CTCs are often detected in association with white blood cells.64 Immune surveillance is a key part of our innate cancer defence, and while interactions with cytotoxic T cells may be encouraging, increasingly it is realised that tumour cells acquire the means to escape the immune system. Interactions with T regulatory cells, which can suppress cytotoxic responses and promote more myeloid-derived suppressor cells, may protect the CTCs and promote their survival. In response to infections, neutrophils release neutrophil extracellular traps (NETs), and it appears that expression of β1-integrins on CTCs and NETs can facilitate CTC capture by NETs and support the development of a metastatic niche.80 Platelet receptors are also involved in tumour cell-induced platelet aggregation, and are proposed to be an essential immune surveillance escape mechanism of CTCs, contributing to tumour cell extravasation facilitation of metastatic colonisation of distant organs.81

This short overview of CTCs is by no means a comprehensive one, but rather highlights the excitement, interest and rapid pace of scientific discovery in the liquid biopsy CTC field that is relevant to patients with liver disease.

Summary and clinical implications

With the progress in the availability of novel therapies for various chronic liver diseases and liver cancer, the development of liquid biopsy tools for precision medicine has evolved into a key priority in the field of hepatology. Non-invasive reliable biomarkers that can supplement and eventually replace invasive liver biopsy are crucial for adequate patient selection and monitoring of response to treatment. Similarly, detection of liver cancer at an earlier stage of the disease, potentially susceptible to curative resection, can be critical to improve patient survival. Circulating EVs, cfDNA, cfRNA and CTC have emerged as attractive liquid biopsy candidates because they fulfil many of the key characteristics of an ideal biomarker. Although these various liquid biopsy approaches are promising and potential transformative tools, several key issues still need to be addressed, and most of the data gathered have been within proof-of-concept studies. Larger and independent studies as well as the development of protocols for standardisation of the various methodology used, including adequate sample preparation, quantitation and data normalisation techniques, are needed. These studies may pave the way to the development of a new generation of reliable, mechanism-based disease biomarkers.



  • Contributors JM, HLR and AEF contributed equally. AEF developed the main concept of the manuscript. All authors contributed to review and revision and approved the final manuscript.

  • Funding This work was supported in part by grants from the US National Institutes of Health (R01 DK113592, U01 AA024206) to AEF, the UK Medical Research Council (grant MR/K10019494/1), the National Institute on Alcohol Abuse and Alcoholism (NIAAA) (grant UO1 AA018663), and the National Institute for Health Research Newcastle Biomedical Research Centre based at Newcastle Hospitals NHS Foundation Trust and Newcastle University to JM, and the Bobby Robson Foundation, CR UK Newcastle Experimental Cancer Medicine Center award C9380/A18084 and CR UK grant C18342/A23390 to HLR.

  • Competing interests None declared.

  • Patient consent Not required.

  • Provenance and peer review Commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.