NEW YORK – Researchers are expanding the limits of what is possible using nanopore sequencing, with two new studies exploring its applications in cell-free DNA (cfDNA) methylation profiling to identify cancer-specific epigenetic changes.
A group at Stanford University School of Medicine developed a technique to characterize the epigenetic signatures of cell-free DNA at the single-molecule level, while a collaborating team from the Hebrew University of Jerusalem and the Institute for the Study and Prevention of Cancer (ISPRO) in Italy developed a way to identify cancer-specific methylation changes and cellular origins.
In a preprint study currently available on BioRxiv, Stanford researcher Billy Lau and his colleagues used an amplification-free method to measure native epigenetic signatures with potential applications in early disease detection, therapy response, and minimal residual disease monitoring, among others. Their technique, adapted for Oxford Nanopore's PromethIon platform, also improved the number of reads sequenced per sample by an order of magnitude over the company's conventional protocol.
"We were very interested in using [nanopore sequencing] for looking at the methylomes of cell-free DNA," Lau said. "We knew [that] one concern was that the yield of sequencing reads would not be as high as Illumina, [but] we wanted to see what information you could tease out and whether this could be feasible for [things] like cancer disease burden and if you could track [that] longitudinally."
This, Lau explained, could also facilitate treatment response monitoring or even early cancer detection.
Critical to such methylation analysis is the fact that nanopore technology enables a more direct look at the native state of the isolated cfDNA by avoiding PCR amplification.
After validating their approach in cfDNA that they fragmented in the lab, the Stanford researchers analyzed blood and tissue samples from 20 colorectal cancer patients, distinguishing distinct gene-level methylation profiles that tracked with known cancer-related phenomena, such as increased methylation of immunologic marker genes and decreased methylation of tumorigenic modulating genes.
Statistically significant methylation differences between healthy and cancer-related samples identified numerous genes in the oncogenic Myc pathway indicating that the methylation changes were cancer-specific.
Quentin Gouil, a senior research officer who studies epigenetics at Australia's Walter and Eliza Hall Institute of Medical Research but was not involved with the current study, noted in an email that reducing costs and coverage as the Stanford team describes can lead to a very sparse dataset.
"That means methylation estimates at individual sites are not reliable," he said. "By comparison, methylation arrays often used for biomarkers, like epigenetic clocks, achieve high coverage at targeted sites in the genome, so they produce robust estimates per site. Therefore, for the same 'modality' — DNA methylation as a biomarker — the two technologies provide very different data and need to be analyzed differently."
Lau explained that this may be less of a limitation than stated by Gouil, as his team's low-pass method covers more methylated sites than can be expected to be informative.
"In our opinion," he said, "sequencing many sites is more informative than precisely determining the methylation of a single site. Because DNA methylation is correlated across the genome and reflective of overall transcriptional programs of malignant cells, by capturing more and more informative sites we can get a better estimation of tumor burden."
Responding to concerns about whether the low starting input might result in insufficient library sizes to run single samples, Lau said that the PromethIon instrument "doesn't complain if you underload, so loading smaller amounts is not a big issue."
His group routinely performs nanopore sequencing runs with well under 50 to 150 femtomoles of cfDNA per flow cell without issue.
"In the logical extreme," he added, "it even makes sense to run single samples on PromethIon as opposed to running on a single MinIon chip because you have 5X to 8X the number of pores on a PromethIon chip, so even if you underload you will get more data."
Although the Stanford team adapted its library preparation protocol work with Oxford Nanopore's LSK110 kit, Lau said that their solution is flexible with future nanopore adapters.
Such forward thinking may prove useful, as Gouil noted that particular chemistry has since been superseded by the LSK112 and LSK114 kits, both of which include a high-capture adapter to further reduce the amount of library that one must load into a flow cell.
"By combining this new high-capture adapter to the protocol tweaks developed in this manuscript, it would be very interesting to see how low-input we can go," he said.
Lau agreed, saying that "the really exciting thing about Kit 14/R10.4.1 (LSK114) chemistry is the fact that you have the 'Marathon' flow cells which can run for much longer and give a higher data yield."
Lau noted that Pacific Biosciences also has a methylation detection workflow for applications such as this one.
"Their error rates are much better by virtue of their HiFi chemistry, which is nice," he said, adding that he found Oxford Nanopore’s platforms easiest to work with in this case.
"There really isn't any strong reason why studies like this can't technically be done with PacBio," he said, "although in my opinion Oxford Nanopore has a much clearer technology roadmap for increased throughput."
Lau and his colleagues currently have no plans to commercialize their technology, although he commented that there have been some early discussions of the topic. "Nothing's set in stone," he said.
Commenting on the Stanford study, Benjamin Berman, a professor at the Hebrew University of Jerusalem in Israel whose lab also studies cfDNA methylation using nanopore sequencing, said "I think it's very exciting." According to Berman, one of the technical challenges for the field is to achieve consistent sequencing coverage across samples with similar DNA inputs.
That said, he thinks the method described in the Stanford study, which is optimized to deliver higher sequencing coverage and a more consistent number of reads, is "really important."
Cell of origin
Last month, Berman's team, in collaboration with researchers from the Institute for the Study and Prevention of Cancer (ISPRO) in Italy, also published a study in Genome Biology using nanopore sequencing to detect cell-of-origin and cancer-specific methylation features in cfDNA.
In his study, Berman and his collaborators retrospectively analyzed the sequencing data of the cfDNA samples from six metastatic lung adenocarcinoma patients and seven healthy controls, which previously underwent shallow nanopore whole-genome sequencing at approximately 0.2X coverage.
While the sample size of the study is too small to determine the limits of sensitivity and specificity of nanopore cfDNA methylation sequencing compared with Illumina bisulfite sequencing, the results showed that samples reliably preserved both cell type-specific and cancer-specific methylation features, as well as copy number alterations and cancer-specific fragmentation features.
Lau said that while the Hebrew University researchers were looking more at biological features, he and his team at Stanford are aiming for a more translational assay. But, he added, it's good to see multiple groups testing the limits of what can be done with nanopore sequencing technology.
Compared with Illumina bisulfite sequencing, Berman thinks nanopore has some "very attractive features" for profiling cfDNA methylation, especially for use in the clinic.
For one, he said the simplicity of sample processing, which circumvents bisulfite conversion and PCR, makes the workflow easier to be adopted for routine clinical use. In addition, because bisulfite treatment tends to significantly degrade the DNA, cell type-specific fragmentation determined by the chromatin structure "can be severely biased."
Still, Berman noted a few challenges in using nanopore sequencing to detect cfDNA methylation. The biggest hurdle, according to him, is that since cfDNA is normally short in size, around 150 bp to 200 bp long, it poses an obstacle to analyzing the methylation profile using nanopore sequencing, which is more suited for detecting methylation in longer DNA molecules.
Additionally, although Berman considers the fast-evolving nanopore sequencing "an extremely exciting technology," the rapid update of the sequencing chemistry has made it somewhat difficult for researchers to catch up.
"It feels like you are on the bleeding edge … you could have big batch effects from these changes," he said. "I definitely think they have to stabilize for clinical applications."
In terms of error rate, Berman said nanopore sequencing's error rate for detecting methylation is "quite concordant" with that of Illumina bisulfite sequencing. However, beyond methylation, when it comes to identifying single-nucleotide mutations in cfDNA, nanopore sequencing's accuracy is still "a bit poorer" compared with Illumina or Pacific Biosciences sequencing, he added.
That said, Berman noted that PacBio sequencing has the potential to become "a big competitor" for nanopore sequencing for cfDNA analysis given the former has a "very high base quality," which is advantageous for detecting point mutations in cfDNA.
Longer fragments
In addition to high accuracy, PacBio sequencing has also demonstrated its utility in detecting ultra-long cfDNA molecules. In a study published in May in Clinical Chemistry, Dennis Lo, a researcher at the Chinese University of Hong Kong, and his collaborators demonstrated that single-molecule sequencing by Pacific Biosciences could achieve long cfDNA detection and direct methylation analysis for cancer patients.
For that study, Lo's team applied PacBio single-molecule real-time (SMRT) sequencing to plasma samples from 13 patients with hepatocellular carcinoma, 13 patients with chronic hepatitis B virus (HBV) infection, and 15 healthy controls, followed by fragment size and methylation analysis.
The study reported that PacBio SMRT sequencing was able to detect a long cfDNA population in cancer samples that were previously unidentified, with the longest tumoral cfDNA being 13.6 kb. While direct methylation analysis of these long DNA molecules enabled tissue-of-origin analysis for individual plasma DNA molecules, the researchers also said they were able to leverage multiple CpG sites in these long plasma DNA molecules to further discern patients with and without cancer, "unlocking new possibilities for long cfDNA-based cancer diagnostics."
However, some researchers think more data are still needed to better understand the biological relevance of these long cfDNA molecules. "I do think that a lot more work would have to be needed in terms of understanding the sample input and what exactly are they capturing from these studies," said Bodour Salhia, a researcher at the University of Southern California who has been studying cfDNA methylation for almost a decade.
Despite the boom of long-read sequencing technologies for profiling cfDNA methylation, Salhia, whose lab has been using Illumina bisulfite sequencing, does not yet see a strong edge of the long-read sequencing-based methods compared with the short-read approaches.
"It's exciting to see new methods emerging to study DNA methylation because it's one of those fields that always lagged behind genomics," she said. "But I think that bisulfite sequencing using short reads is still perfectly acceptable and does capture a good degree of sample complexity."
Especially given the natural size distribution of cfDNA, which tends to congregate around 160 bp in length, short-read sequencing is the "appropriate" technology to use, Salhia said.
Still, Salhia thinks the bisulfite-free and PCR-free nature of the long-read-based methods are advantageous. "The one advantage really that I see [for] the long-read sequencing right now, if it could truly work, is that it may be a simplified workflow," she said, "but I don't see that it's better than or superior than regular short-read sequencing right now."
Commenting on the emerging long-read sequencing-based approaches, Stephen Quake, a bioengineering professor at Stanford University and head of science at the Chan Zuckerberg Initiative, noted that "it's nice to see there are alternative ways to read [cfDNA] methylation; the field has needed new approaches."
While Quake, whose lab has also been studying cfDNA methylation using Illumina sequencing, said the long read-based studies present interesting data academically, he noted that "it is a little too early to know how [these methods] will translate or if they will translate" for clinical use.
Meanwhile, short-read sequencing-based cfDNA methylation detection assays seem to already have a head start in the clinical diagnostic realm, manifested by Grail's 2021 launch of Galleri, a pan-cancer liquid biopsy assay that identifies abnormal methylation patterns in blood cfDNA to achieve multi-cancer early detection.
In addition to Grail, which is now part of Illumina, a plethora of other companies, such as the University of Vienna-spinout HealthBioCare and Redwood City, California-based Guardant Health, are also developing early cancer detection diagnostic tests tapping into cfDNA methylation signals.
"The field sort of took off in the last couple of years primarily because of Grail," said Salhia, who also has a spinout startup named CpG Diagnostics aiming to develop a cfDNA blood test to diagnose ovarian cancer. "They chose methylation over other genomic approaches, and that was a turning point in the field."
In terms of cost, although Salhia said whole-genome bisulfite sequencing is still "a bit cost prohibitive," she considers the targeted methylation detection approach using short-read sequencing, such as the panel-testing strategy adopted by Grail, "highly cost-effective."
However, Quake, who cofounded Bluestar Genomics, a company developing hydroxymethylation-based liquid biopsy assays for cancer detection, said he is not so sure how financially practical it will be to measure targeted cfDNA methylation signatures or the whole methylome in the long run for diagnostics, given so much of the genome is methylated.
Instead, he argues that hydroxymethylation, which is indicative of active gene regulation, can potentially be a more cost-effective epigenetic signal with which to develop cfDNA methylation assays.
"When you are measuring hydroxymethylation, you are not measuring genes that are on, you are not measuring genes that are off, you are measuring genes that are in the process of being turned on," he explained, adding that hydroxymethylation "really captures something interesting dynamically about the tumor."
Because only a fraction of the genome is hydroxymethylated at any point in time, Quake reasoned that the signal can be useful for lowering the cost of methylation-based liquid biopsy tests. "If you can look at only the hydroxymethylated DNA, the sequencing costs go way down, [and] you can do a global unbiased measurement without having to make a panel but do it in a way that's very practical and cost-effective," he said.
"It's always good to have multiple [contestants] in the field," said Stanford's Lau, "because one, it could be friendly competition and two, because it advances the field."