NEW YORK – Startup Bluestar Genomics showed in a recently published study that its 5-hydroxymethylcytosine (5hmC)-based liquid biopsy assay, which tracks epigenetic modifications linked to gene regulation, holds promise for the detection of early-stage pancreatic duct adenocarcinoma (PDAC).
The startup, however, has loftier ambitions, with plans to use the same technology to launch an early-stage, pan-cancer, DNA methylation-based assay as a laboratory-developed test (LDT) out of an envisioned CLIA-certified lab in San Diego.
Bluestar was cofounded in 2017 by Stephen Quake, a bioengineering professor at Stanford and scientific advisor to the firm; CEO Patrick Arensdorf; and advisor Mattias Westman. According to Samuel Levy, chief scientific officer of Bluestar, the spinout has since expanded to 20 employees between its two offices in San Diego, which serves as its wet lab and sequencing hub, and San Francisco, which serves as a location for biostatistics, bioinformatics, and additional activities.
Bluestar's 5hmC workflow begins by collecting up to 10 ml of a patient's blood sample before purifying 4 ml of plasma. After extracting about 10 ng of circulating tumor DNA (ctDNA), the firm's platform modifies hydroxymethyl groups on cytosine by attaching biotin tags to the bases. The team then amplifies the biotinylated DNA fragments by using streptavidin-coated magnetic beads, which allow a "pull-down" assay to isolate DNA molecules that have 5hmC from those do not have the biomarker.
The firm then sequences the targeted fragments using an Illumina next-generation sequencing instrument, generating data that is analyzed in a bioinformatic step to produce an epigenetic signature. The group can identify regions or peaks of 5hmC accumulation in the cfDNA through a predictive model and establish biomarkers for cancer detection.
While the method has not changed since the researchers initially published the proof-of-concept study in 2018, Levy said that his team has automated the entire process and that "a lot of the chemistry" has significantly improved. The firm has also optimized the assay's bioinformatics portion to develop an algorithm that it believes could be incorporated into a commercial test. He noted that the 5hmC assay requires three days to produce diagnostic results.
In the new study published last month in Nature Communications, Levy and his colleagues initially collected and isolated plasma samples from a cohort of 41 PDAC patients and 38 non-cancer control subjects. The team then enriched for and sequenced genetic regions of interest.
After applying a regression model on the sequencing data, Levy's team identified thousands of genes that are hydroxymethylated in the samples. The group built a predictive model using 37 genes that Levy said "are the most valuable in partitioning" PDAC and control patients.
In the training set, Levy and his colleagues found the assay had a clinical sensitivity of 98 percent and a specificity of 63 percent, with an area under the curve (AUC) of 0.92.
The researchers then applied the 37-gene predictive model on two external cohorts, including samples from 30 PDAC patients and 215 controls. They saw that the 5hmC assay had an AUC range of 0.92 to 0.94 between the two cohorts.
Levy believes the study's results indicate that the hydroxymethylated changes detected in cfDNA can reflect the known disease biology of different cancers, such as early-stage PDAC.
"The fact you can take those changes … and essentially derive specific biomarkers that can be used to identify pancreatic cancer for all cell-free DNA leads to the application of testing [for] the disease in asymptomatic individuals," Levy said. "This is ultimately where we want to go [with the technology]."
Levy also highlighted that "some of the thousands of genomic changes" in areas that do not code for proteins identified using the assay — such as transposal and enhancer regions — may also be valuable sources of biomarkers. He pointed out that the changes in the hydroxymethylome not only track genes, but also track a combined set of biological features that encompass gene regulation and transcription.
Since publishing the study, Levy's team has validated the 5hmC assay on a cohort of pancreatic cancer patients that "almost doubled the size" of patients in the prior study. At a specificity of 98 percent, Levy said the assay had a sensitivity of 75 percent. The group plans to publish the study early next year.
The team has also begun exploring applying the 5hmC technology to diagnose other cancers. In a Medrxiv preprint published in January, which has not been peer-reviewed, the researchers evaluated whether they could use the 5hmC assay to identify biomarkers to detect breast, lung, pancreatic, and prostate cancer.
Using a cohort of 48 breast, 55 lung, 32 prostate, 53 PDAC cancer patients, and a control group of 180 non-cancer patients, the team found that found that the methylation-based assay had an AUC of 0.89, 0.84, 0.95, and 0.83 for the different cancers respectively.
Commercial expansion
Bluestar eventually aims to develop a pan-cancer diagnostic assay that "is agnostic to [tumor stage]" and acts a screening tool, Levy said. However, the firm is still finalizing how many and which cancers it will include on the panel besides PDAC.
"So far, there's only a few screening tests for individuals [suspected of] having cancer," Levy said. "We'd like to be able to come up with a molecular test that would accomplish those different kinds of individual tests, but also screen for cancers that aren't screened for currently."
In addition to the original Quake Lab patent issued earlier this year, Bluestar has filed six patents related to technology for measuring epigenomic changes in the bloodstream. Levy said that some of the pending patents are also based on biomarker sets for specific cancer types.
Bluestar plans to initially offer the assay as an LDT out of an envisioned CLIA-certified lab in San Diego. However, Levy said that the company will also engage in discussions with the US Food and Drug Administration to determine the validation data needed to apply for a 510(k) clearance. The firm is also finalizing additional aspects of the test, including its price and anticipated launch.
While Levy declined to disclose how much Bluestar has raised since its inception, he noted that the firm will continue to raise funds as it begins future validation studies and develops the LDT version of the assay.
Several firms and academic groups are developing or offer methylation-based liquid biopsy assays to detect multiple cancers in their earliest stages. Startups like Grail, Lexent Bio, and Thrive Earlier Detection have been acquired by larger in vitro diagnostic firms seeking to enter or expand their market share in the liquid biopsy space.
Researchers at the University of Toronto have developed a methylation-based technique they describe as "cell-free methylated DNA immunoprecipitation and high-throughput sequencing" to detect early stage pancreatic cancer.
Levy argued that Bluestar's DNA methylation analysis method distinguishes itself from competitors because it allows researchers to examine the cancer biology and regions on the genome that are undergoing dynamic demethylation. He believes the 5hmC-based method allows his team to establish very specific biomarkers for different cancer subtypes.
Levy highlighted that the method is cheaper to run over other whole-genome sequencing analysis-based approaches.
"You're looking at a fraction of genome that's undergoing active demethylation by measuring the hydroxymethylome," Levy said. "We've shown, through the studies that we have been able to publish, that this allows an enriched signal … in plasma samples [and that] you can identify biomarkers that have commercial value."