Background
Robust and clinically convenient cancer biomarkers are of great importance for the successful delivery of precision medicine and better clinical care for cancer patients. Firstly, the latency for cancers is long, and cancer patients usually exhibit symptoms at advanced stages when curative treatment may no longer be available. Therefore, screening and early detection of cancers at asymptomatic and/or curable stages are especially critical in improving patients’ survival and quality of life [ 1 – 3 ], as well as reducing the burden of healthcare system. Secondly, the dynamics of tumorigenesis is the main hurdle to effective treatments in cancer patients [ 4 ] and remains a grand challenge for cancer precision medicine. Tumor heterogeneity and clonal evolution are the two major consequences of the dynamics of tumorigenesis, which can simultaneously drive tumor evolution [ 5 ], posing challenges in the selection of anticancer drugs as well as the optimal doses of these drugs. Thirdly, patients with metastatic cancers of unknown primary sites usually have poor prognosis and dismal survival rate because site-specific targeted therapies are not effective [ 6 , 7 ]. To reduce cancer mortality and improve the overall quality of healthcare outcomes and well-being of the patients, there is an urgent need to develop minimally-invasive biomarkers that are sensitive and specific enough for clinical applications such as early cancer detection, longitudinal surveillance of dynamic tumor progression under drug treatment, and the selection of targeted therapies for cancers of unknown primary sites. In this review, we overview the current states of molecular biomarkers in cancer diagnosis and prognosis, as well as discuss the potential benefits and limitations of the new approaches for detecting a novel class of epigenetic biomarkers, 5-hydroxymethylcytosine (5hmC) biomarkers, in cell-free DNA (cfDNA). Furthermore, we provide an update and current perspectives for 5hmC alterations in patients with different cancer types and discuss their potential role as cancer biomarkers. Finally, we point out the improvements and future work required to make these biomarkers effective in the clinic.
Current strategies of cancer biomarker discovery
Tissue biopsy versus liquid biopsy
Tumor tissues are the gold-standard sources for identifying cancer-specific biomarkers. However, tissue biopsy has some intrinsic limitations as it can be invasive and clinically risky. Tissue biopsy usually requires surgical resection to obtain tumor tissues, and surgery entails risks such as bleeding and infection [ 8 , 9 ]. In addition, information acquired from a single-region tissue biopsy only provides a spatially limited snap-shot of a tumor and might fail to reflect the intra-tumor heterogeneity [ 4 , 10 ]. This could lead to an inaccurate diagnosis of cancer type and stage or unreliable prognostic assessment of relapse risk and survival [ 11 ]. Multi-region tissue biopsy offers an alternative way to capture the intra-tumor heterogeneity [ 10 , 12 ]; however, its clinical application is limited due to the volume and accessibility of tumor tissues [ 9 ]. Furthermore, drug treatment confers selective pressure on tumor cells, resulting in adaptive clonal evolution and possibly subsequent drug resistance [ 4 , 13 , 14 ]. While longitudinal profiling of tumor heterogeneity provides valuable information in treatment response evaluation and therapy optimization, this again is of limited clinical feasibility given the difficulty of obtaining tumor tissues for multiple time points [ 15 ].
In contrast, liquid biopsy is emerging as a minimally invasive tool with great potential in cancer management. It uses circulating materials such as cfDNA, circulating tumor cells (CTCs), and exosomes to detect molecular alterations that are indicative of cancer progression. cfDNA is fragmented cellular DNA released into the bloodstream by cells undergoing apoptosis and necrosis, and possibly through active secretion [ 16 , 17 ]. In healthy individuals, cfDNA is primarily derived from apoptotic hematopoietic cells [ 18 ], whereas, in cancer patients, cfDNA can be of tumor tissue and tumor microenvironment origin, reflecting the genetic and epigenetic alterations of tumor tissues and their corresponding microenvironment [ 19 ]. As a result, cfDNA has been extensively studied recently, especially in the field of early cancer detection, cancer staging and subtyping, and disease surveillance. The origins of biomarkers for liquid biopsy are summarized in Fig. 1 . Fig. 1 The origins of biomarkers for liquid biopsy. Molecular alterations of circulating materials, such as cell-free DNA, circulating tumor cells, exosomes, and circulating nucleosomes and RNA, can be used as biomarkers for cancer diagnosis and prognosis. Genetic biomarkers can be identified via mutational profiling of nucleic acids extracted from circulating tumor cells. Epigenetic biomarkers can be obtained via methylation profiling and nucleosome foot printing of cell-free DNA, circulating tumor cells, and circulating nucleosomes and RNA
Genetic biomarkers versus epigenetic biomarkers
Clinical trials based on mutational profiling of circulating tumor DNA (ctDNA; cfDNA of tumor origin) are increasing. Several studies have demonstrated the feasibility of applying ctDNA analysis to capture the heterogeneity of cancer evolution, therefore, making ctDNA analysis a tool for cancer surveillance and care [ 20 ]. For example, longitudinal sampling of ctDNA levels not only indicated the presence of cancer but also had good association with therapy response and disease progression in pancreatic cancer [ 21 ]. Additionally, driver mutations identified in ctDNA samples showed a high concordance with that from matched tumor samples in prostate cancer [ 19 ] and multiple myeloma [ 22 ]. In metastatic breast cancer, estrogen receptor 1 ( ESR1 ) mutations are responsible for the resistance to aromatase inhibitors [ 23 ]. The mutational profile of ESR1 in ctDNA was associated with that in tumor tissues and therefore reflected the clonal evolution of breast cancer under the treatment of aromatase inhibitors [ 23 ]. This evidence indicates that ESR1 mutations in ctDNA are potential biomarkers in treatment monitoring. Despite several studies have demonstrated that the mutational signatures in ctDNA were consistent with those in corresponding tumor tissues, there is currently insufficient evidence of clinical validity and utility for the majority of ctDNA-based mutational assays in advanced cancer, and there is no evidence supports that they can be applied to early cancer detection [ 20 , 24 , 25 ].
In addition, merely relying on the identification of tumor-derived driver mutations in ctDNA cannot capture the whole complexity of tumor biology [ 26 ]. Unlike mutations, the reversible epigenetic modifications are more plastic and can reflect the changes of tumor microenvironment and tissue of origin [ 27 , 28 ]. Epigenetic modifications such as DNA methylation may represent a novel and promising analytical tool for biomarker discovery with broad potential applications in risk assessment, early cancer detection, prognosis, and prediction of response to therapy [ 29 – 31 ]. To date, DNA methylation-based assay, Epi proColon, has been approved by the US Food and Drug Administration (USFDA) for colon cancer detection [ 32 ]. In the early stages of carcinogenesis, many epigenetic changes have occurred in normal tissues before somatic mutations and histopathological changes can be detected [ 33 ]. Therefore, epigenetic analysis of cfDNA combined with mutation-based analysis may contribute to a better understanding of the interplay across molecular alterations in the cancer genome, epigenome, and tumor microenvironment in tumor heterogeneity and clonal evolution [ 27 , 28 , 30 , 34 – 36 ].
Despite the promises, the applications of these genetic or epigenetic biomarkers in population screening and very early stage cancer detection can still be challenging. Like traditional biomarkers, they also suffer from the same issues of low sensitivity and specificity because of the limited amount of circulating materials and the noise in the detection [ 37 ].
Advances in epigenetic cancer biomarker discovery in liquid biopsy
The most extensively studied epigenetic feature for cancer biomarker discovery in cfDNA is DNA methylation, especially the 5-methylcytosine (5mC) modification at CpG dinucleotides [ 29 , 35 , 38 – 41 ]. In hepatocellular carcinoma, 5mC biomarkers derived from ctDNA showed better diagnostic and prognostic values than currently used indicators (such as serum-based alpha-fetoprotein [AFP] and TNM staging) [ 35 ]. In addition, repetitive elements such as long intersperse nucleotide element 1 (LINE-1) and Alu are known proxies for global DNA methylation [ 42 ]. In diffuse large B cell lymphoma, LINE-1 methylation in cfDNA has been shown to be strongly associated with clinical outcomes, demonstrating its potential as a prognostic biomarker [ 43 ]. Another approach in 5mC biomarker discovery is to identify tissue-specific methylation haplotypes as biomarkers to estimate tumor burden and tissue-of-origin in cfDNA [ 40 ]. These multi-CpG haplotypes have been shown to outperform the traditional single-CpG methylation biomarker in cancer classification [ 40 ]. Recently, other epigenetic features such as 5-hydroxymethylcytosine (5hmC) and nucleosome positioning and occupancy on cfDNA have also been utilized to infer tissue of origin and cancer progression [ 27 , 44 – 47 ]. Although genome-wide nucleosome distribution of cfDNA provides valuable information in the deconvolution of pooled cfDNA to infer tissue of origin, its clinical application has not been extensively studied [ 27 , 48 ]. In this review, we summarize the advances in the genome-wide profiling of 5hmC dynamics in cfDNA for cancer biomarker discovery based on the unique features and distinct biological functions of modified cytosines.
5-Hydroxymethylation
In the human genome, 5mC is the most abundant and well-known DNA methylation variant that plays an important role in the regulation of gene expression [ 49 ]. The 5mC-associated methylation patterns are usually tumor- and tissue-specific, reflecting the origin of the metastatic tumors and their altered epigenomes [ 36 ]. With the mediation of ten-eleven translocation (Tet) proteins, 5mC can be further oxidized to 5hmC, 5-formylcytosine (5fC), and 5-carboxylcytosine (5caC) [ 50 ]. Among them, 5hmC is the most abundant and stably oxidized product [ 51 ]. In contrast to about 8% of cytosine that is methylated in the human genome, only 0.5%–1% of cytosines are hydroxymethylated. The scarcity of 5hmC in the genome may pose challenges in distinguishing true signals from noise upon sequencing. However, it also has the potential to improve the statistical power in biomarker discovery because of the reduced multiple hypothesis burden [ 39 , 52 – 54 ]. Unlike the uniform distribution of 5mC outside of the promoter regions, satellites, and repeat DNA sequences [ 55 ], 5hmC has distinct distributions across different functional regions, and its abundance varies across different tissues and cell types [ 56 , 57 ], with tissue type playing a dominant role in determining the distribution patterns of 5hmC [ 58 ]. 5hmC is enriched primarily in the distal regulatory regions, gene bodies of actively expressing genes and promoters, indicating its connection with active transcription [ 59 ]. Genome-wide analysis of 5mC has indicated the global hypomethylation pattern in tumor tissues, whereas depletion of 5hmC has also been associated with the hypermethylation of gene bodies in various cancers [ 58 , 60 , 61 ]. Significant enrichment of 5hmC is observed in both tissue-specific and cancer-specific differentially methylated regions as compared with that of 5mC [ 62 ]. Thus, genome-wide analysis of 5hmC dynamics can further refine our understanding of the relationship between cancer and methylome.
Enabling technologies for profiling 5hmc
Because 5hmC and 5mC dynamics can be informative of tumorigenesis, epigenome-wide analysis of cfDNA has also been conducted to identify minimally-invasive cfDNA-derived biomarkers for better cancer management [ 35 , 45 , 46 , 63 – 66 ]. An overview of the most common genome-wide 5hmC quantification methods are summarized in Table 1 . Table 1 Overview of genome-wide 5hmC quantification methods Technique Advantages Disadvantages Whole-genome TAB-Seq [ 70 ] (direct measurement of 5hmC) Whole-genome Single nucleotide resolution Absolute methylation value High sequencing depth required Input > 100 ng DNA Whole-genome oxBS-Seq [ 71 , 74 ] (inference of 5hmC) RRBS TAB-Seq [ 91 ] (direct measurement of 5hmC) Single nucleotide resolution Absolute methylation value Targets CpG islands Input > 500 ng DNA RRBS oxBS-Seq [ 52 ] (inference of 5hmC) ~ 60 million PE reads/sample Input > 500 ng DNA Nano-hmC-Seal [ 53 ] (direct measurement of 5hmC) Input 5 ng DNA Excellent genomic coverage High specificity for 5hmC ~ 30 million PE reads/sample Lower resolution 5hmC, 5-hydroxymethlcytosine; TAB-Seq, Tet-assisted bisulfite sequencing; oxBS-Seq, oxidative bisulfite sequencing; PE, paired-end
Table 1
Overview of genome-wide 5hmC quantification methods
5hmC, 5-hydroxymethlcytosine; TAB-Seq, Tet-assisted bisulfite sequencing; oxBS-Seq, oxidative bisulfite sequencing; PE, paired-end
| Technique | Advantages | Disadvantages |
| Whole-genome TAB-Seq [ 70 ] (direct measurement of 5hmC) | Whole-genome Single nucleotide resolution Absolute methylation value | High sequencing depth required Input > 100 ng DNA |
| Whole-genome oxBS-Seq [ 71 , 74 ] (inference of 5hmC) | ||
| RRBS TAB-Seq [ 91 ] (direct measurement of 5hmC) | Single nucleotide resolution Absolute methylation value | Targets CpG islands Input > 500 ng DNA |
| RRBS oxBS-Seq [ 52 ] (inference of 5hmC) | ~ 60 million PE reads/sample | Input > 500 ng DNA |
| Nano-hmC-Seal [ 53 ] (direct measurement of 5hmC) | Input 5 ng DNA Excellent genomic coverage High specificity for 5hmC ~ 30 million PE reads/sample | Lower resolution |
Bisulfite-based whole-genome sequencing and reduced representation bisulfite sequencing (RRBS) are conventional methods for methylation profiling and have been applied in biomarker discovery with cfDNA [ 40 , 67 ]. However, there are several limitations of applying bisulfite-based methods on cfDNA. First, traditional bisulfite-based methods cannot distinguish between 5mC and 5hmC, and thus are not capable of capturing the dynamics of these two distinct modification types [ 68 , 69 ]. Second, although modified bisulfite-based methods such as Tet-assisted bisulfite sequencing (TAB-seq) [ 70 ] and oxidative bisulfite sequencing (oxBS-seq) [ 71 ] can detect and quantify 5hmC at nucleotide resolution, they require an amount of DNA input (> 100 ng) that is not feasible for cfDNA from blood samples (1–2 ng cfDNA from 2 to 3 mL of plasma). This makes its application in early cancer detection challenging [ 72 ]. Third, bisulfite-based methods require a high sequencing depth, which is costly and further limits their application in the clinical setting. Restriction enzyme-based methods like reduced representation 5-hydroxymethylcytosine profiling rely on the efficiency of enzymes. As a result, their detection capacity can be limited as inefficient digestion might result in the loss of information on certain 5hmC sites [ 73 ]. In contrast, enrichment-based methods such as hME-Seal [ 53 ] and the nano-hmC-Seal [ 59 ] rely on the selective chemical binding of 5hmC to enrich for 5hmC-containing DNA fragments. These fragments are then sequenced to obtain genome-wide information of 5hmC [ 53 , 59 ]. Because the size of these chemically selected fragments is a major determinant of the resolution of 5hmC mapping [ 56 ], these enrichment-based methods can provide good coverage and high specificity despite low resolution [ 74 ]. Given that these enrichment-based approaches can achieve a balance between cost and detection capacity, they are likely to have great potential to be applied widely in the cfDNA-based liquid biopsy in cancer management for large cohorts in the clinical setting [ 75 ]. However, lack of a consensus computational framework to analyze these enrichment-based sequencing data may also hamper the interpretation of the results from the clinicians’ end. Although 5hmC-Seal technology is the state-of-the-art in genome-wide profiling of 5hmC dynamics in cfDNA, it relies on relative abundance to infer absolute modification levels and cannot provide single-base resolution 5hmC information. Development of statistical approaches to infer base resolution 5hmC modification levels from enrichment counts would provide more insights into the dynamics of 5hmC. In addition, many of the intrinsic features of 5hmC enrichment-based sequencing data must be taken into consideration to build this computational framework. GC content, copy number variation, strand-specific and asymmetric 5hmC distribution need to be statistically corrected before downstream analysis. Therefore, an integrative computational framework must be established for 5hmC-enrichment sequencing data to increase the sensitivity and specificity of 5hmC-derived biomarkers in cancer research and enable their easy clinical application in the future. The computational analysis of 5hmC enrichment-based sequencing data can be decomposed into the following components: (1) modeling (to model 5hmC enrichment signal taking into account factors such as enrichment bias, local density bias, and copy number variations); (2) quantification (to infer modification level from normalized regional count-based enrichment data [ 76 ]); (3) construction of cancer prediction models (to identify differentially modified sites or regions among different conditions [ 76 ]). Another approach is to construct a model to predict cancer status or stage based on the estimated proportions and the tissue-of-origin of tumor-derived cfDNA in the blood sample [ 39 ].
Advances in 5hmc-based cancer biomarker discovery
Several recent studies have demonstrated that 5hmC signatures in cfDNA are reliable and sensitive epigenetic markers that are indicative of types and stages of cancers [ 44 – 46 ]. These cfDNA-derived 5hmC biomarkers are found to achieve higher detection sensitivity than classical biomarkers [ 45 ].
Colorectal cancer (CRC)
CRC is a commonly diagnosed cancer [ 1 , 77 ]. Despite recent declines in CRC incidence in the United States and other developed countries, the incidence and mortality continue to increase in the rest of the world [ 1 ]. This reduction in CRC incidence is largely attributed to screening via colonoscopy and to cancer prevention efforts [ 1 , 77 ]. However, the invasive nature of colonoscopy leads to poor patient compliance. Stool and blood DNA methylation assays based on candidate genes show great diagnostic and prognostic values, but the sensitivities and specificities of the assays vary and are usually inconsistent [ 78 ]. In a Chinese cohort study evaluating cfDNA-derived 5hmC analysis in 80 colorectal cancer patients and 90 healthy individuals [ 45 ], a total of 989 differentially methylated 5hmC loci in gene bodies were selected as biomarkers to train the machine learning algorithm for cancer classification. The classifier achieved 83% sensitivity and 94% specificity (area under curve [AUC] = 0.95) in the validating dataset (24 patients and 35 controls) and 88% sensitivity and 89% specificity (AUC = 0.94) in another independent validating dataset (32 patients and 37 controls) for cancer classification [ 45 ]. The discriminatory performance of these cfDNA-derived 5hmC biomarkers was not only comparable to that of 5hmC tissue biomarkers but also significantly outperformed the current USFDA approved blood-based methylation test Epi proColon [ 45 , 79 ]. Epi proColon relies on the methylation status of the single gene septin 9 ( SEPT9 ) to infer the presence of cancer and can only achieve a detection sensitivity of 0.48 [ 79 ]. To be noted, the sensitivity and specificity of the cancer classifier often vary with cancer stages. Because the majority of colorectal cancer patients in this study are at TNM stages III and IV, the performance of this classifier cannot be over-interpreted [ 45 ].
Gastric cancer (GC)
GC is a common digestive cancer with 26,240 new cases and 10,800 deaths estimated in 2018 in the United States [ 1 ]. Early-stage GC is asymptomatic and exhibits high genomic heterogeneity, making endoscopic or surgical biopsy-based molecular testing rather inaccurate and non-representative [ 80 ]. Recent studies suggested the potential role of cfDNA-based molecular profiling in future clinical applications such as diagnosis and targeted therapy selection. However, the epigenetic alterations on cfDNA remain understudied [ 81 – 83 ]. A pilot study has explored the 5hmC alterations in cfDNA from GC patients as compared to that from healthy individuals in a Chinese cohort [ 45 ]. Patients with GC and controls were divided into discovery (7 patients and 18 controls) and validation groups (25 patients and 35 controls) [ 45 ]. In total, 1431 differentially methylated 5hmC loci in gene bodies were identified and trained on the cancer classifier, and the classifier achieved 92% sensitivity and 91% specificity (AUC = 0.93) in the validating dataset and 90% sensitivity and 97% specificity (AUC = 0.97) in another independent validating dataset [ 45 ]. Again, consistent with the study in CRC, these cfDNA-derived 5hmC biomarkers performed better than classical early diagnosis biomarkers, such as carcinoembryonic antigen (CEA) and cancer antigen 19-9 (CA19-9), and other epidemiological factors, such as smoking and alcohol [ 45 ]. In both CRC and GC, the classifiers derived from cfDNA are disease-, clinical stage-, and cancer type-specific, suggesting their potential values as diagnostic cancer biomarkers [ 45 ].
Esophageal cancer (EC)
EC is among the five leading causes of cancer-related death in male patients between age 40 and 59 [ 1 ]. Similar to GC, evidence for the effects of molecular biomarkers on diagnosis and treatment guidance are limited because of the challenges in detecting genomic alterations or clinical symptoms at early stages and in resolving the discrepancy of genomic profiling between primary tumors and metastatic lesions [ 81 ]. One study conducted genome-wide 5hmC profiling in 150 newly diagnosed EC patients and 177 healthy controls in China [ 46 ]. The classifier achieved 93.75% sensitivity and 85.71% specificity with an average AUC of 0.947. Consistent with the observations in CRC and GC, the 5hmC signatures on cfDNA were indicative of clinical stages of EC. The probability of predicting cancer based on the 5hmC classifier increased with the progression of cancer stage. In addition, EC patients with lymph node metastases were predicted to have significantly higher cancer probability as compared with patients without lymph node metastases [ 46 ].
Lung cancer (LC)
LC is the leading cause of cancer-related death in both genders with 83,550 estimated deaths in males and 70,500 estimated deaths in females in 2018 [ 1 ]. Unlike the steady increase of 5-year survival rate in most cancer types, the 5-year survival rate of LC patients is the second lowest with little improvement over the past years [ 1 ]. Minimally-invasive early LC detection is urgently needed to improve the survival rate because the low-dose computed tomography screening not only fails to provide low false positive results with high predictive values but also requires additional invasive testing procedures afterwards [ 66 , 84 ]. One study evaluated the diagnostic value of 5hmC signatures in 15 LC patients from China and 40 healthy controls from the United States and found that global cell-free 5hmC levels were gradually depleted during the development from early non-metastatic to late metastatic stages [ 44 ]. Another independent study on a Chinese cohort identified 2459 genes with differential 5hmC levels by comparing the genome-wide 5hmC profiles among 66 non-small cell lung cancer patients and 67 healthy individuals [ 85 ]. Candidate 5hmC biomarker panel derived from this study was observed to achieve better detection sensitivity as compared with that of known clinical biomarkers such as CEA, carbohydrate antigen 125 (CA125) and neuron-specific enolase (NSE).
Multiple myeloma (MM)
MM is the second most common hematological malignancy with profound disruption of epigenomes. A recent proof-of-concept study applied the nano-hmC-Seal technology to explore the potential of cfDNA-derived 5hmC biomarkers to improve the current invasive and expensive bone marrow biopsy-based biomarkers [ 86 ]. This study profiled the genome-wide 5hmC modifications in 9 MM patients, 5 patients with premalignant precursor condition defined as monoclonal gammopathy of undetermined significance, 5 patients with another premalignant precursor condition defined as smoldering multiple myeloma, and 19 newly diagnosed, treatment-naïve European American patients as controls. In total, 183 genes containing differential 5hmC loci were identified, and MM patients were separated from the patients in precursor conditions utilizing these 5hmC signatures. In addition, MM patients with different relapse statuses could also be separated using 5hmC signatures [ 86 ]. These preliminary findings again highlighted the values of 5hmC signatures as independent diagnostic and prognostic biomarkers in MM.
Conclusions and future directions
With the biological understanding that 5hmC is at the nexus of glucose metabolism and cancer epigenetics [ 61 ], recent studies have demonstrated the clinical prospects of using genome-wide 5hmC dynamics on cfDNA to improve cancer management. However, several challenges remain, limiting the translational success of cfDNA-derived 5hmC biomarkers in clinical settings. Current studies of 5hmC biomarker discovery have been mostly focused on genic regions, partly because of their genomic enrichment pattern and putative gene regulatory role. Expanding 5hmC biomarker discovery beyond genic regions to for example other cis -regulatory elements or unbiased genome-wide scans will potentially provide opportunities to identify the optimal clinically useful 5hmC biomarkers and enhance our knowledge of their biological relevance. Another challenge is that circulating cfDNA can originate from various sources; therefore, the genome-wide analysis of cfDNA is possibly hampered by the effect of genetic or epigenetic heterogeneity. A comprehensive cross-tissue comparison will be needed to establish highly tissue-specific 5hmC features in cfDNA [ 87 ]. Well-controlled animal models such as the patient-derived xenograft mouse model can also be utilized to evaluate experimentally the tumor relevance of 5hmC signals in cfDNA [ 45 , 88 ]. Moreover, emerging technologies such as single-cell epigenetic assays could help determine the contributions of cfDNA from various sources [ 89 ]. Furthermore, considering their distinct biological functions and genomic distributions, integrating both 5mC and 5hmC modification markers together with nucleosome foot printing in the future would be promising to maximize the detection sensitivity in early-stage cancer and tissue-of-origin [ 90 ]. Finally, since the current 5hmC cancer biomarker discovery studies are generally small in sample size, future better statistically powered case–control and longitudinal studies will be necessary to identify more reliable 5hmC biomarkers, and evaluate the relationships between 5hmC and various potential confounding factors as well as the dynamic changes of 5hmC in patients (e.g., after treatment), thus facilitating the clinical applications of this promising tool in cancer precision medicine.
Abbreviations
circulating tumor DNA
cell-free DNA
5-methylcytosine
5-hydroxymethlcytosine
5-formylcytosine
5-carboxylcytosine
circulating tumor cell
alpha-fetoprotein
tumor staging based on primary tumor (T), reginal lymph nodes (N) and distant metastasis (M)
ten-eleven translocation
differentially methylated region
tet-assisted bisulfite sequencing
oxidative bisulfite sequencing
reduced representation bisulfite sequencing
reduced representation 5-hydroxymethylcytosine profiling
colorectal cancer
area under the receiver operating characteristic curve
gastric cancer
carcinoembryonic antigen
cancer antigen 19-9
carbohydrate antigen 125
neuron-specific enolase
esophageal cancer
lung cancer
low-dose computed tomography
non-small-cell lung cancer
multiple myeloma
paired-end
Authors’ contributions
All authors contributed substantially to the writing of this review. All authors read and approved the final manuscript.
Acknowledgements
We would like to thank the Taishan Scholars Program of Shandong Province, P. R. China.
Competing interests
W.Z. advises Shanghai Epican Genetech Co., Ltd., which holds a license of the 5hmC-Seal technology. Other authors declare that they have no competing interests.
Availability of data and materials
Not applicable.
Consent for publication
Not applicable.
Ethics approval and consent to participate
Not applicable.
Funding
This work was partly supported by a grant from the National Institutes of Health P30 C060553 Career Development Fund (to W.Z.).
References
1. Siegel RL Miller KD Jemal A Cancer statistics, 2018 Cancer J Clin 2018 68 1 7 30 10.3322/caac.21442 29313949
2. Bax C Taverna G Eusebio L Sironi S Grizzi F Guazzoni G Innovative diagnostic methods for early prostate cancer detection through urine analysis: a review Cancers. 2018 10.3390/cancers10040123 29670060 PMC5923378
3. Martin LR Williams SL Haskard KB DiMatteo MR The challenge of patient adherence Ther Clin Risk Manag 2005 1 3 189 199 18360559 PMC1661624
4. Sabaawy HE Genetic heterogeneity and clonal evolution of tumor cells and their impact on precision cancer medicine J Leuk (Los Angel). 2013 1 4 1000124 10.4172/2329-6917.1000124 24558642 PMC3927925
5. Gerlinger M Rowan AJ Horswell S Larkin J Endesfelder D Gronroos E Intratumor heterogeneity and branched evolution revealed by multiregion sequencing N Engl J Med 2012 366 1 10.1056/nejmoa1113205 22397650 PMC4878653
6. Varadhachary GR Raber MN Cancer of unknown primary site N Engl J Med 2014 371 8 757 765 10.1056/NEJMra1303917 25140961
7. Kato S Krishnamurthy N Banks KC De P Williams K Williams C Utility of genomic analysis in circulating tumor DNA from patients with carcinoma of unknown primary Cancer Res 2017 77 16 4238 4246 10.1158/0008-5472.CAN-17-0628 28642281 PMC5729906
8. Loeb S Vellekoop A Ahmed HU Catto J Emberton M Nam R Systematic review of complications of prostate biopsy Eur Urol 2013 64 6 876 892 10.1016/j.eururo.2013.05.049 23787356
9. Ilie M Hofman P Pros: can tissue biopsy be replaced by liquid biopsy? Transl Lung Cancer Res. 2016 5 4 420 423 10.21037/tlcr.2016.08.06 27655109 PMC5009092
10. Gerlinger M Horswell S Larkin J Rowan AJ Salm MP Varela I Genomic architecture and evolution of clear cell renal cell carcinomas defined by multiregion sequencing Nat Genet 2014 46 3 225 233 10.1038/ng.2891 24487277 PMC4636053
11. Hoffman RM Gilliland FD Adams-Cameron M Hunt WC Key CR Prostate-specific antigen testing accuracy in community practice BMC Fam Pract 2002 3 19 10.1186/1471-2296-3-19 12398793 PMC137591
12. Zhang J Fujimoto J Zhang J Wedge DC Song X Zhang J Intratumor heterogeneity in localized lung adenocarcinomas delineated by multiregion sequencing Science 2014 346 6206 256 259 10.1126/science.1256930 25301631 PMC4354858
13. Friedman R Drug resistance in cancer: molecular evolution and compensatory proliferation Oncotarget. 2016 7 11 11746 11755 10.18632/oncotarget.7459 26909596 PMC4914245
14. Diaz LA Jr Williams RT Wu J Kinde I Hecht JR Berlin J The molecular evolution of acquired resistance to targeted EGFR blockade in colorectal cancers Nature 2012 486 7404 537 540 10.1038/nature11219 22722843 PMC3436069
15. Khan KH Cunningham D Werner B Vlachogiannis G Spiteri I Heide T Longitudinal liquid biopsy and mathematical modeling of clonal evolution forecast time to treatment failure in the PROSPECT-C phase II colorectal cancer clinical trial Cancer Discov 2018 10.1158/2159-8290.cd-17-0891 30166348 PMC6380469
16. Jahr S Hentze H Englisch S Hardt D Fackelmayer FO Hesch RD DNA fragments in the blood plasma of cancer patients: quantitations and evidence for their origin from apoptotic and necrotic cells Cancer Res 2001 61 4 1659 1665 11245480
17. Wan JC Massie C Garcia-Corbacho J Mouliere F Brenton JD Caldas C Liquid biopsies come of age: towards implementation of circulating tumour DNA Nat Rev Cancer 2017 17 4 223 238 10.1038/nrc.2017.7 28233803
18. Lehmann-Werman R Neiman D Zemmour H Moss J Magenheim J Vaknin-Dembinsky A Identification of tissue-specific cell death using methylation patterns of circulating DNA Proc Natl Acad Sci USA 2016 113 13 E1826 E1834 10.1073/pnas.1519286113 26976580 PMC4822610
19. Wyatt AW Annala M Aggarwal R Beja K Feng F Youngren J Concordance of circulating tumor DNA and matched metastatic tissue biopsy in prostate cancer J Natl Cancer Inst 2017 10.1093/jnci/djx118 29206995 PMC6440274
20. Ossandon MR Agrawal L Bernhard EJ Conley BA Dey SM Divi RL Circulating tumor DNA assays in clinical cancer research J Natl Cancer Inst 2018 110 9 929 934 10.1093/jnci/djy105 29931312 PMC6136923
21. Park G Park JK Son DS Shin SH Kim YJ Jeon HJ Utility of targeted deep sequencing for detecting circulating tumor DNA in pancreatic cancer patients Sci Rep. 2018 8 1 11631 10.1038/s41598-018-30100-w 30072705 PMC6072791
22. Mishima Y Paiva B Shi J Park J Manier S Takagi S The mutational landscape of circulating tumor cells in multiple myeloma Cell Rep 2017 19 1 218 224 10.1016/j.celrep.2017.03.025 28380360 PMC5439509
23. Schiavon G Hrebien S Garcia-Murillas I Cutts RJ Pearson A Tarazona N Analysis of ESR1 mutation in circulating tumor DNA demonstrates evolution during therapy for metastatic breast cancer Sci Transl Med. 2015 7 313 313ra182 10.1126/scitranslmed.aac7551 26560360 PMC4998737
24. Merker JD Oxnard GR Compton C Diehn M Hurley P Lazar AJ Circulating tumor DNA analysis in patients with cancer: American Society of Clinical Oncology and College of American Pathologists Joint Review J Clin Oncol 2017 10.1200/jco.2017.76.8671 29504847
25. Pantel K Blood-based analysis of circulating cell-free DNA and tumor cells for early cancer detection PLoS Med. 2016 13 12 e1002205 10.1371/journal.pmed.1002205 28027295 PMC5189941
26. Rachiglio AM Esposito Abate R Sacco A Pasquale R Fenizia F Lambiase M Limits and potential of targeted sequencing analysis of liquid biopsy in patients with lung and colon carcinoma Oncotarget. 2016 7 41 66595 66605 10.18632/oncotarget.10704 27448974 PMC5341823
27. Snyder MW Kircher M Hill AJ Daza RM Shendure J Cell-free DNA comprises an in vivo nucleosome footprint that informs its tissues-of-origin Cell 2016 164 1–2 57 68 10.1016/j.cell.2015.11.050 26771485 PMC4715266
28. Hoadley KA Yau C Hinoue T Wolf DM Lazar AJ Drill E Cell-of-origin patterns dominate the molecular classification of 10,000 tumors from 33 types of cancer Cell 2018 173 2 291 304 10.1016/j.cell.2018.03.022 29625048 PMC5957518
29. Leygo C Williams M Jin HC Chan MWY Chu WK Grusch M DNA methylation as a noninvasive epigenetic biomarker for the detection of cancer Dis Markers 2017 2017 3726595 10.1155/2017/3726595 29038612 PMC5605861
30. Liu L Toung JM Jassowicz AF Vijayaraghavan R Kang H Zhang R Targeted methylation sequencing of plasma cell-free DNA for cancer detection and classification Ann Oncol 2018 29 6 1445 1453 10.1093/annonc/mdy119 29635542 PMC6005020
31. Gordevicius J Krisciunas A Groot DE Yip SM Susic M Kwan A Cell-free DNA modification dynamics in abiraterone acetate-treated prostate cancer patients Clin Cancer Res 2018 24 14 3317 3324 10.1158/1078-0432.ccr-18-0101 29615462
32. Koch A Joosten SC Feng Z de Ruijter TC Draht MX Melotte V Analysis of DNA methylation in cancer: location revisited Nat Rev Clin Oncol. 2018 10.1038/s41571-018-0004-4 29666440
33. Peltomaki P Mutations and epimutations in the origin of cancer Exp Cell Res 2012 318 4 299 310 10.1016/j.yexcr.2011.12.001 22182599
34. Statham AL Taberlay PC Kelly TK Jones PA Clark SJ Genome-wide nucleosome occupancy and DNA methylation profiling of four human cell lines Genom Data. 2015 3 94 96 10.1016/j.gdata.2014.11.012 26484155 PMC4535833
35. Xu RH Wei W Krawczyk M Wang W Luo H Flagg K Circulating tumour DNA methylation markers for diagnosis and prognosis of hepatocellular carcinoma Nat Mater 2017 16 11 1155 1161 10.1038/nmat4997 29035356
36. Wen Y Wei Y Zhang S Li S Liu H Wang F Cell subpopulation deconvolution reveals breast cancer heterogeneity based on DNA methylation signature Brief Bioinform 2017 18 3 426 440 10.1093/bib/bbw028 27016391
37. Fiala C Diamandis EP Utility of circulating tumor DNA in cancer diagnostics with emphasis on early detection BMC Med. 2018 16 1 166 10.1186/s12916-018-1157-9 30285732 PMC6167864
38. Zeng H He B Yi C Peng J Liquid biopsies: DNA methylation analyses in circulating cell-free DNA J Genet Genom 2018 10.1016/j.jgg.2018.02.007 29706556
39. Feng H Jin P Wu H Disease prediction by cell-free DNA methylation Brief Bioinform 2018 1 1 10.1093/bib/bby029 PMC6556903 29672679
40. Guo S Diep D Plongthongkum N Fung HL Zhang K Zhang K Identification of methylation haplotype blocks aids in deconvolution of heterogeneous tissue samples and tumor tissue-of-origin mapping from plasma DNA Nat Genet 2017 49 4 635 642 10.1038/ng.3805 28263317 PMC5374016
41. Kang S Li Q Chen Q Zhou Y Park S Lee G CancerLocator: non-invasive cancer diagnosis and tissue-of-origin prediction using methylation profiles of cell-free DNA Genome Biol 2017 18 1 53 10.1186/s13059-017-1191-5 28335812 PMC5364586
42. Barchitta M Quattrocchi A Maugeri A Vinciguerra M Agodi A LINE-1 hypomethylation in blood and tissue samples as an epigenetic marker for cancer risk: a systematic review and meta-analysis PLoS ONE 2014 9 10 e109478 10.1371/journal.pone.0109478 25275447 PMC4183594
43. Wedge E Hansen JW Garde C Asmar F Tholstrup D Kristensen SS Global hypomethylation is an independent prognostic factor in diffuse large B cell lymphoma Am J Hematol 2017 92 7 689 694 10.1002/ajh.24751 28378885
44. Song CX Yin S Ma L Wheeler A Chen Y Zhang Y 5-Hydroxymethylcytosine signatures in cell-free DNA provide information about tumor types and stages Cell Res 2017 10.1038/cr.2017.106 28820176 PMC5630676
45. Li W Zhang X Lu X You L Song Y Luo Z 5-Hydroxymethylcytosine signatures in circulating cell-free DNA as diagnostic biomarkers for human cancers Cell Res 2017 10.1038/cr.2017.121 28925386 PMC5630683
46. Tian X Sun B Chen C Gao C Zhang J Lu X Circulating tumor DNA 5-hydroxymethylcytosine as a novel diagnostic biomarker for esophageal cancer Cell Res 2018 10.1038/s41422-018-0014-x 29467383 PMC5951904
47. Ivanov M Baranova A Butler T Spellman P Mileyko V Non-random fragmentation patterns in circulating cell-free DNA reflect epigenetic regulation BMC Genom 2015 16 Suppl 13 S1 10.1186/1471-2164-16-S13-S1 PMC4686799 26693644
48. Lehmann-Werman R Neiman D Zemmour H Moss J Magenheim J Vaknin-Dembinsky A Identification of tissue-specific cell death using methylation patterns of circulating DNA Proc Natl Acad Sci 2016 113 13 E1826 E1834 10.1073/pnas.1519286113 26976580 PMC4822610
49. Laird PW The power and the promise of DNA methylation markers Nat Rev Cancer 2003 3 4 253 266 10.1038/nrc1045 12671664
50. Ito S Shen L Dai Q Wu SC Collins LB Swenberg JA Tet proteins can convert 5-methylcytosine to 5-formylcytosine and 5-carboxylcytosine Science 2011 333 6047 1300 1303 10.1126/science.1210597 21778364 PMC3495246
51. Branco MR Ficz G Reik W Uncovering the role of 5-hydroxymethylcytosine in the epigenome Nat Rev Genet 2011 13 1 7 13 10.1038/nrg3080 22083101
52. Skvortsova K Zotenko E Luu P-L Gould CM Nair SS Clark SJ Comprehensive evaluation of genome-wide 5-hydroxymethylcytosine profiling approaches in human DNA Epigenet Chromat 2017 10 1 16 10.1186/s13072-017-0123-7 PMC5397694 28428825
53. Song CX Szulwach KE Fu Y Dai Q Yi C Li X Selective chemical labeling reveals the genome-wide distribution of 5-hydroxymethylcytosine Nat Biotechnol 2011 29 1 68 72 10.1038/nbt.1732 21151123 PMC3107705
54. Doerks T Copley RR Schultz J Ponting CP Bork P Systematic identification of novel protein domain families associated with nuclear functions Genome Res 2002 12 1 47 56 10.1101/gr.203201 11779830 PMC155265
55. Song CX Yi C He C Mapping recently identified nucleotide variants in the genome and transcriptome Nat Biotechnol 2012 30 11 1107 1116 10.1038/nbt.2398 23138310 PMC3537840
56. Yu M Hon GC Szulwach KE Song CX Zhang L Kim A Base-resolution analysis of 5-hydroxymethylcytosine in the mammalian genome Cell 2012 149 6 1368 1380 10.1016/j.cell.2012.04.027 22608086 PMC3589129
57. Nestor CE Ottaviano R Reddington J Sproul D Reinhardt D Dunican D Tissue type is a major modifier of the 5-hydroxymethylcytosine content of human genes Genome Res 2012 22 3 467 477 10.1101/gr.126417.111 22106369 PMC3290782
58. Thomson JP Meehan RR The application of genome-wide 5-hydroxymethylcytosine studies in cancer research Epigenomics. 2017 9 1 77 91 10.2217/epi-2016-0122 27936926
59. Han D Lu X Shih AH Nie J You Q Xu MM A highly sensitive and robust method for genome-wide 5hmC profiling of rare cell populations Mol Cell 2016 63 4 711 719 10.1016/j.molcel.2016.06.028 27477909 PMC4992443
60. Chen K Zhang J Guo Z Ma Q Xu Z Zhou Y Loss of 5-hydroxymethylcytosine is linked to gene body hypermethylation in kidney cancer Cell Res 2016 26 1 103 118 10.1038/cr.2015.150 26680004 PMC4816137
61. Vasanthakumar A Godley LA 5-hydroxymethylcytosine in cancer: significance in diagnosis and therapy Cancer Genet 2015 208 5 167 177 10.1016/j.cancergen.2015.02.009 25892122
62. Li X Liu Y Salz T Hansen KD Feinberg A Whole-genome analysis of the methylome and hydroxymethylome in normal and malignant lung and liver Genome Res 2016 26 12 1730 1741 10.1101/gr.211854.116 27737935 PMC5131824
63. Warton K Mahon KL Samimi G Methylated circulating tumor DNA in blood: power in cancer prognosis and response Endocr Relat Cancer 2016 23 3 R157 R171 10.1530/ERC-15-0369 26764421 PMC4737995
64. Widschwendter M Zikan M Wahl B Lempiainen H Paprotka T Evans I The potential of circulating tumor DNA methylation analysis for the early detection and management of ovarian cancer Genome Med. 2017 9 1 116 10.1186/s13073-017-0500-7 29268796 PMC5740748
65. Mastoraki S Strati A Tzanikou E Chimonidou M Politaki E Voutsina A ESR1 methylation: a liquid biopsy-based epigenetic assay for the follow-up of patients with metastatic breast cancer receiving endocrine treatment Clin Cancer Res 2018 24 6 1500 1510 10.1158/1078-0432.ccr-17-1181 29284708
66. Hulbert A Jusue-Torres I Lung cancer recurrence epigenetic liquid biopsy J Thorac Dis. 2018 10 1 4 6 10.21037/jtd.2017.11.124 29600008 PMC5863116
67. Chan KC Jiang P Chan CW Sun K Wong J Hui EP Noninvasive detection of cancer-associated genome-wide hypomethylation and copy number aberrations by plasma DNA bisulfite sequencing Proc Natl Acad Sci USA 2013 110 47 18761 18768 10.1073/pnas.1313995110 24191000 PMC3839703
68. Huang Y Pastor WA Shen Y Tahiliani M Liu DR Rao A The behaviour of 5-hydroxymethylcytosine in bisulfite sequencing PLoS ONE 2010 5 1 e8888 10.1371/journal.pone.0008888 20126651 PMC2811190
69. Jin SG Kadam S Pfeifer GP Examination of the specificity of DNA methylation profiling techniques towards 5-methylcytosine and 5-hydroxymethylcytosine Nucleic Acids Res 2010 38 11 e125 10.1093/nar/gkq223 20371518 PMC2887978
70. Yu M Hon GC Szulwach KE Song CX Jin P Ren B Tet-assisted bisulfite sequencing of 5-hydroxymethylcytosine Nat Protoc 2012 7 12 2159 2170 10.1038/nprot.2012.137 23196972 PMC3641661
71. Booth MJ Ost TW Beraldi D Bell NM Branco MR Reik W Oxidative bisulfite sequencing of 5-methylcytosine and 5-hydroxymethylcytosine Nat Protoc 2013 8 10 1841 1851 10.1038/nprot.2013.115 24008380 PMC3919000
72. Plongthongkum N Diep DH Zhang K Advances in the profiling of DNA modifications: cytosine methylation and beyond Nat Rev Genet 2014 15 10 647 661 10.1038/nrg3772 25159599
73. Petterson A Chung TH Tan D Sun X Jia XY RRHP: a tag-based approach for 5-hydroxymethylcytosine mapping at single-site resolution Genome Biol 2014 15 9 456 10.1186/s13059-014-0456-5 25248841 PMC4212096
74. Booth MJ Branco MR Ficz G Oxley D Krueger F Reik W Quantitative sequencing of 5-methylcytosine and 5-hydroxymethylcytosine at single-base resolution Science 2012 336 6083 934 937 10.1126/science.1220671 22539555
75. Cui L Chung TH Tan D Sun X Jia XY JBP1-seq: a fast and efficient method for genome-wide profiling of 5hmC Genomics 2014 104 5 368 375 10.1016/j.ygeno.2014.08.023 25218799
76. Bock C Analysing and interpreting DNA methylation data Nat Rev Genet 2012 13 10 705 719 10.1038/nrg3273 22986265
77. Arnold M Sierra MS Laversanne M Soerjomataram I Jemal A Bray F Global patterns and trends in colorectal cancer incidence and mortality Gut 2017 66 4 683 691 10.1136/gutjnl-2015-310912 26818619
78. Mojtabanezhad Shariatpanahi A Yassi M Nouraie M Sahebkar A Varshoee Tabrizi F Kerachian MA The importance of stool DNA methylation in colorectal cancer diagnosis: a meta-analysis PLoS ONE 2018 13 7 e0200735 10.1371/journal.pone.0200735 30024936 PMC6053185
79. Church TR Wandell M Lofton-Day C Mongin SJ Burger M Payne SR Prospective evaluation of methylated SEPT9 in plasma for detection of asymptomatic colorectal cancer Gut 2014 63 2 317 325 10.1136/gutjnl-2012-304149 23408352 PMC3913123
80. Battaglin F Naseem M Puccini A Lenz HJ Molecular biomarkers in gastro-esophageal cancer: recent developments, current trends and future directions Cancer Cell Int 2018 18 99 10.1186/s12935-018-0594-z 30008616 PMC6042434
81. Pectasides E Stachler MD Derks S Liu Y Maron S Islam M Genomic heterogeneity as a barrier to precision medicine in gastroesophageal adenocarcinoma Cancer Discov 2018 8 1 37 48 10.1158/2159-8290.Cd-17-0395 28978556 PMC5894850
82. Gao J Wang H Zang W Li B Rao G Li L Circulating tumor DNA functions as an alternative for tissue to overcome tumor heterogeneity in advanced gastric cancer Cancer Sci 2017 108 9 1881 1887 10.1111/cas.13314 28677165 PMC5581520
83. Wang H Li B Liu Z Gong J Shao L Ren J HER2 copy number of circulating tumour DNA functions as a biomarker to predict and monitor trastuzumab efficacy in advanced gastric cancer Eur J Cancer 2018 88 92 100 10.1016/j.ejca.2017.10.032 29207318
84. Bennett CW Berchem G Kim YJ El-Khoury V Cell-free DNA and next-generation sequencing in the service of personalized medicine for lung cancer Oncotarget. 2016 7 43 71013 71035 10.18632/oncotarget.11717 27589834 PMC5342606
85. Zhang J Han X Gao C Xing Y Qi Z Liu R 5-Hydroxymethylome in circulating cell-free DNA as A potential biomarker for non-small-cell lung cancer Genom Proteom Bioinf 2018 10.1016/j.gpb.2018.06.002 PMC6076378 30010036
86. Chiu BYQZZ Stepniak L Zhang X Chernoff M Zimmerman TL He C Zhang W 5-Hydroxymethylcytosine of circulating cell-free DNA in plasma: a novel non-invasive marker for progression and prognosis in multiple myeloma 2017 ASH Atlanta
87. Tang W Wan S Yang Z Teschendorff AE Zou Q Tumor origin detection with tissue-specific miRNA and DNA methylation markers Bioinformatics 2018 34 3 398 406 10.1093/bioinformatics/btx622 29028927
88. Yada E Wada S Yoshida S Sasada T Use of patient-derived xenograft mouse models in cancer research and treatment Future sci OA 2017 4 3 FSO271 10.4155/fsoa-2017-0136 29568561 PMC5859329
89. Karemaker ID Vermeulen M Single-cell DNA methylation profiling: technologies and biological applications Trends Biotechnol 2018 10.1016/j.tibtech.2018.04.002 29724495
90. Shen SY Singhania R Fehringer G Chakravarthy A Roehrl MHA Chadwick D Sensitive tumour detection and classification using plasma cell-free DNA methylomes Nature 2018 10.1038/s41586-018-0703-0 30429608
91. Hahn MA Li AX Wu X Pfeifer GP Single base resolution analysis of 5-methylcytosine and 5-hydroxymethylcytosine by RRBS and TAB-RRBS Methods Mol Biol 2015 1238 273 287 10.1007/978-1-4939-1804-1_14 25421665 PMC4340590