Cell-free DNA methylation as a potential biomarker in brain disorders

The diagnosis of many brain disorders is challenging due to the nonspecific clinical presentation and/or the lack of accurate biomarkers. In addition, many neurological disorders show slow progression with a long asymptomatic period followed by a stage with mild clinical symptoms. This limits the potential for early detection with effective therapeutic treatment only occurring late in the course of the disease. Clearly, developing noninvasive and accurate diagnostic tools for detecting brain disorders is an unmet medical need. Altered methylation patterns in circulating cell-free DNA (cfDNA) have been implicated as useful tools for noninvasive cancer detection, prenatal diagnostics and organ transplantation assessment. Given the critical roles of DNA methylation in many brain disorders, unusual DNA methylation alternations in cfDNA could be a promising biomarker for disease diagnosis. This commentary presents an overview of investigational applications of cfDNA methylation signatures as potential molecular markers to diagnose various brain disorders. The authors' views on the technologies for cfDNA methylation analysis based on next-generation sequencing are also discussed.

Brain disorders include conditions or disabilities that affect the brain, including those caused by illness, genetics and traumatic injury. The diagnosis of brain disorders is one of the most difficult challenges for modern medicine due to the complexity of the nervous system and the lack of accurate biomarkers [1]. Occasionally, brain imaging is used to study the structural and biochemical changes that come with diseases. However, several studies have demonstrated that symptoms of chronic neurodegenerative diseases (e.g., Alzheimer's disease [AD], Parkinson's diseases, Huntington diseases and vascular and frontotemporal dementias) cannot be definitely diagnosed until the patient is in the final stages of the disease, making therapeutic treatment extremely difficult [2]. Cerebrospinal fluid obtained by lumbar puncture is helpful for the early detection of several neurodegenerative diseases, but the procedure is invasive, painful and potentially dangerous. Therefore, a rapid, safe and accurate diagnostic method is needed to save or significantly improve the lives of patients.

There is growing interest in the use of extracellular DNA in bodily fluids to obtain diagnostic information. Liquid biopsy techniques, such as cfDNA are quickly emerging as noninvasive methods for detecting fetal cfDNA in the plasma of pregnant women, detecting donor-derived DNA in the plasma of transplantation recipients and detecting circulating tumor DNA in the plasma of patients with cancer [3]. CfDNA refers to all nonencapsulated DNA in the bloodstream. They are nucleic acid fragments (usually 160–180 bp) that enter the bloodstream upon cell death caused by either regular cell turnover or pathogenesis, indicating cfDNA is a mixture of DNA fragments from multiple tissues and cell types. If certain organs or tissues in the body experience a pathology with enhanced cell death, higher quantities of cfDNA from the affected area are present in circulation [4]. In addition, the molecular weight and size distribution of cfDNA may indicate its source. For example, the cfDNA fragments from apoptosis are around 180 bp, whereas necrosis results in larger fragments [5]. After capturing and sequencing the cfDNA, unusual genetic alternations are identified by comparing the samples with cfDNA from normal individuals (e.g., the detection of tumor-specific genetic variants and the detection of fetal genetic abnormalities) [6–9]. The utilization of cfDNA can dramatically reduce the difficulties of obtaining cancer biopsies and minimize potential risks for serious fetal injury. However, these methods are used to identify the presence of high levels of genetic variants, as seen in cancer and germline. For the detection of disorders that are not usually associated with an increased rate of genetic alterations, nongenetic approaches, such as epigenetics, are necessary [4].

Nongenetic signatures of cfDNA have been studied to broaden the diagnostic applications of circulating DNA beyond the analysis of genetic markers alone, including cfDNA methylation variation, the profile of cfDNA fragmentation and topology of cfDNA [3]. Among these, cfDNA methylation profiles provide a rich source of information for the deconvolution of tissue of origin, indicating its usefulness in locating the site of pathology and inferring cell death in the source organ [6–8,10]. Each tissue/cell type in the body carries unique methylation features associated with gene expression, which are useful to define cell/tissue identity. Consequently, the diagnostic potential of cfDNA methylation status has significantly increased in recent years for tumor progression, noninvasive prenatal and organ transplantation assessments. For instance, pregnant women show an increased placental contribution in cfDNA compared with the plasma of nonpregnant healthy controls, determined by genome-wide DNA methylation profiles [6]. In a pregnant woman diagnosed with follicular lymphoma during early pregnancy, methylation deconvolution revealed a grossly elevated contribution from B cells to the plasma DNA pool, and the localized B cells were the origin of the copy number abnormalities observed in plasma [6]. DNA methylation markers can be used for cancer detection by targeting individual hypermethylated tumor suppressor genes. One recent study reported an accuracy of over 90% for identifying the tissue of origin using a targeted methylation approach in 359 patients with cancer [9]. Deconvolution of the healthy plasma methylome also revealed a signal from neurons, accounting for ∼2% of cfDNA [7], making cfDNA methylation status a promising diagnostic tool in a diverse range of brain disorders. For many brain disorders, altered DNA methylation patterns can be detected in the early stages of pathogenesis. Growing evidence has implicated the utility of cfDNA as a potential biomarker for brain tumors and various neurological diseases, as shown in Table 1.

Table 1. . Studies exhibiting the diagnostic potential of cell-free DNA methylation in brain disorders.

Condition	cfDNA source	Observations	Ref.
Brain tumor	A total of 220 patient samples (70 IDH mutant gliomas, 52 IDH wild-type gliomas, 60 meningiomas, 9 hemangiopericytomas, 14 low-grade glial-neuronal tumors)	CfDNA methylation signatures can distinguish different primary brain tumors	[11]
	A total of 22 glioma patients (15 IDH mutant, 7 IDH wildtype)	Identified glioma-specific cfDNA methylation features	[12]
	A total of 32 patients with different grades of brain tumors	CfDNA MCPH1 methylation levels are associated with tumor grade	[13]
MS	A total of 59 RRMS patients	Increased cfDNA levels; significant alternations of methylation in 56 gene promoters	[14]
	A total of 20 RRMS patients	Increased levels of demethylated MOG in cfDNA	[15]
	A total of 24 RRMS patients	Elevated methylation status of select LINE-1 promoter CpG sites	[16]
ALS	A total of 20 ALS patients	Elevated methylation levels of selected CpG sites in RHBDF2 gene in cfDNA	[17]
	A total of 28 ALS patients	Increased proportion of skeletal muscle components in cfDNA methylation profiles	[18]
AD	A total of 27 AD patients	Elevated methylation levels of select CpG sites in LHX2 gene	[19]
TBI	A total of 15 TBI patients	Increased levels of cfDNA unmethylated ‘Brain1’	[10]

AD: Alzheimer's disease; ALS: Amyotrophic lateral sclerosis; cfDNA: cell-free DNA; MS: Multiple sclerosis; RRMS: Relapsing-remitting multiple sclerosis; TBI: Traumatic brain injury.

Brain tumors

Brain tumors are abnormal masses of brain tissue with uncontrolled and abnormal growth of brain cells. Brain imaging often provides a wide differential diagnosis, ranging from indolent low-grade tumors to aggressive cancers, making the accurate diagnosis of lesions a major challenge in the management of intracranial pathologies [11]. Sampling tissue for tumor diagnosis through invasive neurosurgery introduces risk. However, many cancers release tumoral materials into body biofluids, making the noninvasive diagnosis of tumors possible. The use of liquid biopsies for cancer detection and management has gained prominence in the last ten years. DNA methylation abnormalities are biologically important in brain tumors and can be applied to categorize tumors into different prognostic groups. Diaz Jr. et al. found that the load of plasma cfDNA correlated with tumor staging and prognosis [20]. Four years later, Shen et al. developed cell-free methylated DNA immunoprecipitation sequencing (cfMeDIP-seq) for profiling of genome-wide bisulfite-free plasma DNA methylation patterns for extracranial cancers. The study suggested that the methylation changes in plasma cfDNA directly relate to extracranial cancers with divergent cells of origin in a noninvasive and cost-effective manner, despite the low amount of cfDNA [8]. The same research group later found that highly specific signatures of cfDNA methylation could be used to detect and accurately distinguish different primary brain tumors using cfMeDIP-seq [11]. They used the identified DNA methylation profiles to distinguish glioma samples from patients with systemic cancers and healthy controls. Moreover, the circulating methylome signatures could discriminate brain tumors with similar cell-of-origin lineages, which were indistinguishable using standard-of-care MRI [11]. The results highlighted the application of plasma methylomes for the accurate diagnosis of common primary intracranial tumors. Using the Infinium Human Methylation 850K array, the DNA methylation profile of glioma cfDNA was found to be distinct from nontumor specimens as well as other neoplasms [12]. Deconvolution analysis for the methylome found that glioma cfDNA methylomes contained more neuronal cell-related methylation signatures compared with nontumor samples. The identified glioma-specific noninvasive epigenetic LB (Glioma-eLB) signatures reflected a tumor signature and can be used for glioma prediction [12].

In addition to the profiling of cfDNA methylomes, cfDNA can also be utilized to determine the methylation status of some molecular marker candidates for brain tumors. For example, the MCPH1 [21] gene is primarily expressed in the fetal brain and plays a critical role in fetal brain development. Promoter methylation of the MCPH1 gene, which leads to inactivation of gene expression, was found in over 90% of the tissues from patients with brain tumors [22]. Using cfDNA, MCPH1 methylation was detected in more than half (54%) of samples. The methylation levels exhibited a significant association with tumor grade. Therefore, detecting MCPH1 methylation using cfDNA might be a strategy for the assessment of brain tumors [13]. In line with the identification of Glioma-eLB using cfDNA, some promoter methylated markers associated with genes critical for glioma tumorigenesis were revealed, such as PVT1 and CXCR6 [12]. These encouraging results indicate that cfDNA-based epigenetic signatures are promising markers for brain tumor diagnosis and the tracking of tumor progression.

Neurological disorders

Aberrant DNA methylation can mediate neuronal cell death, making it relevant to many neurodegenerative diseases. In recent years, the exploration of cfDNA methylation signatures using liquid biopsy-based approaches has continuously increased for neurological disorders. MS is an unpredictable neurodegenerative disease that disturbs the communication between the brain and body and within the brain. Understanding the mechanism of abnormal DNA methylation in MS could clarify the pathology and lead to the discovery of new therapeutic targets. Liggett et al. found cfDNA of RRMS exhibited unique disease- and state-specific features [14]. Promoter methylation in cfDNA of 56 genes was explored using a microarray-based assay for RRMS patients and healthy controls. DNA methylation patterns of promoters were significantly different between patients and healthy controls and could distinguish RRMS patients that were either in remission or in exacerbation. Moreover, the methylation pattern used to discriminate RRMS patients in exacerbation from those in remission had over 70% specificity and sensitivity. These results indicate that epigenetic biomarkers based on cfDNA methylation are feasible for MS detection and could be useful for other neurodegenerative diseases. MS is often characterized by the demyelination of axons in the CNS. The demyelinated lesions in the brain and spinal cord are caused by oligodendrocyte (ODC) cell death [15]. However, no molecular biomarkers for ODC death in MS have been clinically approved. CfDNA of patients with active RRMS showed increased demethylation levels of the MOG gene when compared with inactive disease and healthy controls using a CpG methylation-specific quantitative PCR assay [14]. The MOG gene is uniquely expressed in ODCs as an integral part of the myelin sheath. Therefore, differential cfDNA methylation patterns in MOG could be a biomarker for MS diagnosis and prognosis. In addition, the L1PA2 subfamily fragments of LINE-1 repetitive elements in cfDNA displayed a higher promoter methylation level in RRMS than in controls through bisulfite sequencing [16]. Real-time PCR assays revealed that the methylation levels of some CpG loci within the LINE-1 promoter were increased in cfDNA from RRMS patients, suggesting that the methylation patterns of CpG sites in LINE1 elements may be another biomarker for noninvasive MS detection.

ALS is a progressive neurological disorder that leads to the death of upper and lower motor neurons. When compared with healthy controls, altered DNA methylation signatures were discovered in either postmortem brain tissues or in the spinal cord samples of patients with ALS, including a set of differentially methylated CpG sites and promoter hypomethylation of genes (e.g., APEX1, APTX, OGG1 and PNKP) involved in DNA repair functions. These results highlight the crucial roles of DNA methylation in ALS pathogenesis. Using blood DNA samples, elevated 5mC levels have been observed in patients with ALS versus control groups, along with the promoter hypermethylation of the C9orf72 gene [23,24]. Using cfDNA, bisulfite pyrosequencing was used to assess the methylation levels of two distinct CpGs (CpG1 and CpG2) located at the promoter-enhancer region of the RHBDF2 gene, an activator of the EGF receptor signaling pathway involved in the pathogenesis of ALS [17]. Both sites showed elevated methylation levels in patients with ALS compared with controls, and they were positively correlated. More recently, Caggiano et al. developed the CeFiE algorithm to identify cell-type proportions in cfDNA using whole-genome bisulfite sequencing (WGBS) information [18]. The analysis identified a significantly higher proportion of skeletal muscle components in the cfDNA of patients with ALS compared with healthy controls. These results warrant further whole-genome cfDNA methylation analysis in patients to discover novel epigenetic-based biomarkers for ALS.

AD is a progressive neurodegenerative disease that leads to atrophy and brain cell death. AD is a major cause of dementia affecting older people worldwide. DNA methylation is known to be altered in AD, including 5mC and 5hmC. AD-associated methylation signatures have been found across multiple brain regions, such as the hippocampus, middle frontal gyrus, middle temporal gyrus, as well as cortex. However, the global methylation levels vary among brain regions. Hypermethylation was discovered for some genes critical for AD pathogenesis, including APP, TREM2 and ANK1. Using blood samples, elevated global DNA methylation was revealed in peripheral blood mononuclear cells (PBMCs) of late-onset AD samples [25]. Blood tests of genes closely related to AD showed altered DNA methylation levels in patients with AD. For example, increased methylation and expression of the APP gene in PBMCs of AD samples was observed in comparison to the healthy twin in monozygotic twins discordant for AD. The PIN1 gene displayed hypomethylation, while the methylation changes of PSEN1 and APOE genes were not different [26]. Interpreting the results of this study is challenging because of confounding factors. However, methylation profiling based on the cfDNA of AD patients is limited. One study, using 36 adult participants (nine nondemented controls and 27 patients with AD) and patients with mild AD, detected increased circulating nucleic acids in AD plasma [19]. CfDNA in patients with AD contained neuronal tissue-specific methylated LHX2 at CpG sites 1 and 5, suggesting the neuronal cell death caused by β-amyloid toxicity. This finding emphasizes cfDNA methylation status as a potential biomarker in AD prediction. Another recent study by Zac et al. defined CG and non-CG (CH) methylation that is unique to neurons of the dorsolateral prefrontal cortex, and brain region-specific mCG features [27]. With this information, neuron-derived cfDNA and cerebellum cfDNA within acute neurotrauma and chronic neurodegeneration were identified using targeted multiplex next-generation bisulfite sequencing (tNGBS). Cognitive decline was also associated with a decrease of dorsolateral prefrontal cortex cfDNA. The study provides a framework for the investigation of brain-derived molecular biomarkers for neurodegenerative diseases.

Noninvasive diagnostic approaches for acute brain disorders (stroke, brain traumatic injury and infections) are needed as well. Traumatic brain injury (TBI) is often caused by an external force to the head, with the occurrence of neuronal injury in combination with disruption of the blood–brain barrier. Lehmann-Werman et al. identified a set of tissue-specific DNA methylation markers in cfDNA using an Illumina Infinium Human Methylation 450K array, which was able to assess cell death in the tissues of interest [10]. The brain-derived (neuronal or glial) cfDNA methylation markers were detected in patients with severe TBI following cardiac arrest and MS, providing the first evidence of brain-derived cfDNA during neurodegeneration. A cluster of CpG sites around the locus of CG09787504 (termed ‘Brain1’ in the paper) showed higher levels of unmethylated Brain1 in cfDNA in brain damage samples. However, this preliminary study was unable to determine an association between the levels of the brain damage biomarker NFL and brain-specific cfDNA. Further studies are needed to identify the correlation between cfDNA methylation patterns and TBI.

Conclusion & perspective

Proof-of-concept studies indicate that cfDNA methylation patterns could be useful biomarkers for brain disorders. Several techniques relying on DNA methylation profiling can be performed to discover new specific methylation features for different neurological disorders, including WGBS, reduced-representation bisulfite sequencing, MeDIP-seq, methylated CpG tandem amplification and sequencing, tNGBS and 5hmC profiling. However, there are no current clinical applications of cfDNA methylation biomarkers for brain disorders. Many challenges and questions remain. First, at what point during disease progression will brain-derived cfDNA need to be explored? It is largely unknown whether potential DNA methylation biomarkers originate from downstream factors of dying neuronal cells or from upstream events. Second, bisulfite treatment can cause DNA degradation and lead to the loss of some critical DNA regions. Capture-based sequencing could also lead to the loss of some critical DNA methylation sites. TET-assisted pyridine borane sequencing was described as an ideal method for cfDNA methylation studies, with the advantages of mild and bisulfite-free treatment for cfDNA [28]. Further, third-generation, or long-read sequencing technologies, such as single-molecule real-time sequencing by Pacific Biosciences and nanopore sequencing by Oxford Nanopore Technologies which allow direct interrogation of single-base DNA modifications without any chemical conversions if DNA prior to sequencing, could be ideal tools for exploring cfDNA methylation signatures [29]. Third, identification of the tissue of origin of cfDNA helps to pinpoint the site of pathology. However, existing data indicate that brain-derived DNA methylation is only a small fraction of the cfDNA methylome. The selection of methylation markers used for deconvolution analysis must be refined and optimized; more sensitive computational methods of methylation profiling and deconvolution are needed to improve the accuracy of biomarker identification using cfDNA. Extensive validation of scientific findings using large sample cohorts will be necessary before recommendation for clinical application. Taken together, the progress summarized here emphasizes the clinical potential of cfDNA methylation as molecular biomarkers to predict and monitor various brain disorders. Additional efforts to investigate the DNA methylation and cfDNA underlying these disease mechanisms will be worthwhile.