The Evolution of Epigenetic Biomarkers

No part of the diagnostic route to cancer is pleasant – the initial symptoms, the scans, the waiting, the uncertainty. But of all the undesirable experiences, it’s likely that tissue biopsy ranks highest on the list of things patients hope they never need to repeat. And yet, tissue biopsy certainly offers important insights into an individual patient’s disease, which is why we continue to request such invasive, and occasionally risky, tests. But tissue biopsy presents problems beyond just obtaining the sample. It’s a painful procedure for patients and, depending on the tumor’s location and characteristics, may require general anesthetic, specialist imaging, or other interventions. This lack of accessibility can make retesting difficult and put the quality of the sample at risk. And a successful sample isn’t the end of the story; some may be too small for all of the required tests, or too difficult to preserve well. Still others may not contain examples of every cell type and mutation in a heterogeneous tumor – and there’s no way of finding that out by examining the sample we get.

A new kind of sample

By now, most – if not all – pathologists have heard of liquid biopsy, and some of you may already be using it in your work. What makes liquid biopsy a better option? The approach measures biomarkers in bodily fluids, which are limitless resources and, in most cases, accessible in a minimally invasive manner. Blood, for instance, requires a simple draw, whereas saliva and urine are even easier to obtain. It’s true that certain fluids, such as cerebrospinal fluid or pleural effusions, are trickier to access, but even those carry fewer risks than traditional tissue biopsy or tumor resection. As a result, the liquid biopsy procedure is not only safer, but also less expensive for both patients and healthcare systems. It’s not completely free of limitations – for instance, target material often appears in only small concentrations in the sample, meaning that tests may be less sensitive or require advanced sampling techniques – but the advantages are promising.

From liquid biopsy samples, we isolate or interrogate circulating tumor cells (CTCs), circulating tumor DNA (ctDNA), cell-free DNA (cfDNA), and exosomes so that we  can analyze biomarkers such as gene fusions, point mutations, methylation changes, circulating mRNA, and exosomes. This variety of options means that we can screen patients at risk of disease (or even, one day, perhaps the entire population). After diagnosis, we can stratify them for treatment, monitor their response to therapy, and keep an eye out for minimal residual disease (MRD) – all without the need for invasive surgical procedures.

My analyte in liquid biopsy is cfDNA – short (150–180 bp) fragments of DNA. cfDNA tends to be present in higher abundance than CTCs and can be assessed from frozen plasma, which broadens the available testing options. But it’s not only tumors that release cfDNA into the blood; all cells can do it, which means we can examine genetic and epigenetic changes in tissue symptomatic of other diseases as well.

There are a number of tests on the market to measure genetic alterations in blood samples. The first, approved by the US Food and Drug Administration in 2016, searches for EGFR exon 19 deletions and exon 21 substitutions in non-small cell lung cancer to select those who may benefit from inhibitor treatment (1). Newer tests are not limited to one or two mutations; in fact, they aren’t even limited to one or a few cancers. One recent test reported in the literature looks for 16 gene mutations in cfDNA, as well as abnormalities in eight circulating proteins, to detect eight different common cancers (2). Its overall sensitivity ranges from under 40 percent to nearly 100 percent, and its specificity sits around 99 percent. It is important to note, however, that non-cancer samples were not analyzed in this study, so this specificity needs further investigation. Nevertheless, it’s clear that these tests have huge potential – and that’s why they hold such promise for ovarian cancer.

Tackling ovarian cancer

Ovarian cancer is the third most common gynecological malignancy worldwide (3), with nearly 300,000 new cases diagnosed each year. Unfortunately, it’s also the second most lethal, claiming nearly 185,000 lives – almost two thirds of those diagnosed. This is partly because the disease is so often detected late, and partly because it frequently shows resistance to standard chemotherapy.

At the moment, CA-125 is the gold standard circulating protein biomarker for the management of ovarian cancer. But despite its status as the main single biomarker in diagnostic testing, it’s not elevated in up to half of early cancers – and it can be elevated in a number of non-cancerous conditions, including benign tumors, follicular cysts, endometriosis, infection, and even pregnancy. As a result, it has been deemed unsuitable for screening, and its capacity for cancer diagnosis is limited. Potential new tests reported in the literature, evaluating liquid from Pap smears, show promise; one that analyzed 18 genes and aneuploidy was able to detect 33 percent of ovarian cancer cases, including early-stage disease (4). When combined with ctDNA, the detection rate increased to 63 percent – still low sensitivity, but with 100 percent specificity.

It’s this potential that prompted me to investigate liquid biopsy’s potential in high-grade serous cancer (HGSC), the most common – and most aggressive – form of ovarian cancer. Recent research shows that HGSC actually begins in the distal fallopian tube as serous tubal intraepithelial carcinoma (STIC) before seeding to the ovary (5). My colleagues and I developed a bespoke clinical cohort of six patients with matched tissue from normal fallopian tube, STIC, ovarian cancer, omentum, and normal ovary, with the goal of identifying novel epigenetic biomarkers.

Epigenetic advances

Why epigenetics? These markers boast several benefits, such as stability and frequency of occurrence. People often think that mutations are the easiest cancer-associated change to test; however, this is not always the case. Although p53 is mutated in approximately 95 percent of ovarian cancers, the specific mutation is not always the same, and potential mutations are spread over kilobases of DNA sequence. Epigenetic changes tend to be more conserved. They also offer noninvasive accessibility (for instance, via blood) and reversibility – all of which mean they may be not only diagnostic, but also potential treatment targets. We opted to look specifically at DNA methylation because it’s the most widely studied genetic alteration, tends to be highly concentrated in CpG islands within and near promoters, and has known behavior: hypermethylation is associated with chromatin condensation and gene silencing.

How do you identify methylation markers worthy of investigation? Genetics and omics studies can reveal them in specific tissues. Once a candidate gene is identified, it is validated in its native tissue and then migrated to liquid biopsy. An example is the SEPT9 gene, which is hypermethylated in many colorectal cancers. A discovery project looking for a blood test for these cancers examined tumor tissue, normal colon, non-colonic control tissue, and peripheral blood lymphocytes to identify this candidate gene (6), which was then further validated by a retrospective trial in symptomatic patients. After prospective trials that compared methylation testing with colonoscopy and the fecal immunochemical test for population screening (7,8), an epigenetic screening test for colorectal cancers was approved in 2016.

One approach to epigenetic biomarker identification is the “candidate gene” approach in a disease with known biology. The gene is typically a novel oncogenic driver or tumor suppressor. Its altered expression signifies its potential importance in disease, leaving investigators with questions like: how is its expression regulated? Could it be used as a biomarker? If so, what kind – diagnostic, prognostic, or predictive? My colleagues and I recently identified just such an alteration, present in up to 15 percent of ovarian cancer cases, and are currently investigating its possibilities as a methylation-based biomarker for the disease.

Another approach is the “discovery” approach, in which a methylation array highlights differentially methylated CpG islands, whether or not they are easily associated with a specific gene. Regardless of how these potential methylation marks are identified, they must then be validated using assays such as bisulfite pyrosequencing. This test determines methylation levels by converting unmethylated cytosine to uracil; methylated cytosine remains unchanged. Thus, differential levels of methylation in control and tumor tissue can be measured to validate potential epigenetic biomarkers. In the case of the marker we discovered, its sensitivity is comparable to the gold standard CA-125, but its specificity is much higher. In combination with CA-125, it yields improvements in both sensitivity and specificity.

The translation equation

It’s clear that these biomarkers have real potential for use in the diagnosis and management of ovarian cancer. But how do we make this a reality? We need to translate our current knowledge to liquid biopsy, but that won’t be easy. Bisulfite conversion is very harsh on DNA and pyrosequencing lacks sensitivity – so we’ll need a lot of material to analyze.

There are no standard methods for processing blood or extracting DNA, although ASCO/CAP guidelines recommend collecting samples in cell stabilization or EDTA tubes and processing within six hours. But is there enough ctDNA in a standard blood sample – especially if the tumor itself is small? If we assume that the tumor is approximately 1 cm in diameter (thus consisting of about 500 million cells) and 0.01 percent of genome equivalents are present in circulation, then there will be between 17 and 20 tumor genomes (34 to 40 copies of any sequence) per milliliter of plasma. The limits of methylation detection are reported to be as low as one or two copies (9), so the assay should work. However, it does need to be very specific to differentiate normal from tumor DNA.

Are there any other assay options? Methylation-specific restriction digestion may offer an alternative, although it would be necessary to ensure the optimal conversion of unmethylated DNA. Some techniques also offer improved sensitivity over pyrosequencing: methylation-specific PCR (with droplet digital PCR), PCR with high-resolution melting, and COLD-PCR (co-amplification at lower denaturation temperature). All of these assays hold promise and have the required level of sensitivity – a sign that, despite the technical challenges we have yet to fully overcome, it’s time to explore our options in the realm of liquid biopsy.