NEW YORK – Researchers at Dartmouth-Hitchcock Medical Center have developed a method to analyze copy number alterations in brain tumors that could provide a diagnosis while a patient is still on the operating table.
In a standing-room-only talk at last week's annual meeting of the Association for Molecular Pathology, Chun-Chieh Lin, a pathologist at Dartmouth, described his team's nanopore sequencing-based approach to making copy number calls in genes commonly used to classify brain tumors, such as FGFR1 or EGFR, and detecting chromosomal aberrations such as 1p/19q co-deletion.
The key piece of the puzzle is a genomic DNA concatenation method, dubbed irreversible Sticking Compatible Overhang to Reconstruct DNA (iSCORED), which enables copy number variant (CNV) calls to be made by sampling the tumor genome with reads from long DNA fragments that are randomly concatenated from 100 bp to 150 bp DNA fragments. The researchers described this method in a preprint posted to MedRxiv last month.
Using a special cocktail of restriction enzymes, genomic tumor DNA is digested and the fragments are then ligated into molecules 1 kb to 2 kb in length. This improves the speed of acquiring data, as the one-step prep can be done in 30 minutes and the long fragments reduce the amount of time that a sequencing pore spends loading a new DNA molecule. "That's actually the more time-consuming part," Lin said.
Combining iSCORED with the real-time data acquisition of the Oxford Nanopore Technologies (ONT) MinIon sequencer, the preprint reported the ability to deliver a molecular tumor classification within two to two-and-a-half hours.
In a retrospective study of 26 malignant brain tumors, they were able to show 100 percent concordance in copy number variant detection when compared to sequencing and microarray-based tests used in the clinical lab.
With a more powerful PromethIon sequencer and bioinformatics running on a graphics processing unit setup, that diagnosis can now be available in less than 1.5 hours, Lin told GenomeWeb, with reagent costs of $125 per sample, including a $70 MinIon flow cell that can be reused for six additional samples.
Getting a diagnosis during surgery is important because it may allow the surgeon to adjust their strategy, said Jeroen de Ridder, a researcher at the Netherlands' University Medical Center Utrecht who has developed a similar, methylation-based classifier for brain tumors. "[They] may opt for a more radical or more conservative resection," he said. "This can be based on the prognosis or on the availability of alternative treatments for that particular subclass."
The iSCORED method is just the latest attempt to provide faster classification of tumors using nanopore sequencing from ONT, which sponsored the AMP workshop in which Lin presented his data.
In 2019, researchers at the University of Southern California published SMURF-seq (sampling molecules using re-ligated fragments), which also concatenates fragments to do copy number analysis. While it can also acquire the data to make these calls in about an hour, the sample prep takes longer. The authors of that paper said it takes 90 minutes, though Lin said it could take up to two hours. The genomic resolution is also lower, he said, at 60 kb to 150 kb, and genomic input requirements are higher, at 2 micrograms to 3 micrograms, compared to 200 nanograms to 400 nanograms for iSCORED.
Lin's method also joins Jeroen de Ridder's nanopore-based method for intraoperative, methylation-based brain tumor classification, published last month in Nature. Lin noted that methylation data can also be obtained for fragments prepped with iSCORED, although the ability to correctly classify tumors this way was reliant on high tumor purity in the sample.
"In some cases the methylation signal may be inconclusive, or there may be doubt between subclasses, in which case observing very specific CNVs could be used as a tiebreaker," such as 1p/19q codeletion, de Ridder said. "However, to me, it is unclear how often this is really expected to be the case; in our recent paper, we describe one example where this was the case but we had to specifically retrieve it from the archives."
In the preprint, Lin's team reported that one hour of sequencing provided around 350 Mb of data, representing approximately 1.4 million mapped fragments. This is "roughly equivalent to one mapped fragment every 1.5 kb to 3 kb in the genome" he said. In post-sequencing analysis, his team binned fragments into 60 kb bins and counted the numbers of mapped fragments in each bin. "This is to decrease variability, and 60 kb per bin is an established 'high resolution' for CNV assessment," he said.
By normalizing the data, the team was able to detect small CNVs 1 Mb to 3Mb in size, "which is critical and clinically relevant because most oncogene amplification falls into this category," Lin said, such as EGFR, MYCN, and ERBB2.
These are calls that can't be made with previous rapid copy number detection methods, he added, as they don't acquire enough data in the timeframe needed for intraoperative diagnosis. The iSCORED approach produces approximately three times the data of SMURF-seq, for instance, he said.
The Dartmouth team also looked at the ability of iSCORED to provide calls from tissue samples of varying tumor content. "We were able to detect CCNE1 amplification in samples with as low as 5 percent tumor [content] using only 500,000 mapped fragments," they wrote. "Additionally, low copy number gain (22 copies) was also reliably detected with the same parameter, albeit at a higher tumor percentage and with more fragments — 20 percent and 1.5 million fragments, respectively."
For the concordance study, sample were run by the Illumina TruSight Tumor 170 assay, whole-exome sequencing, or Affymetrix OncoScan chromosomal microarrays. The preprint authors noted that they have applied for a patent on iSCORED.
De Ridder said he would like to see a systematic analysis of copy number profiling, compared to methylation-only classification. "How often is the CNV profile essential for classification?" he said.
The iSCORED method, like any tool, has limits. It is unlikely to detect gene fusions, Lin said, and detecting homozygous deletions is challenging. The method has yet to be used during an actual surgery, he noted.
But the information could be useful even if it's not delivered during an operation.
"Waiting for a diagnosis is a significant burden for the patient and their loved ones, especially when it comes to central nervous system tumors where the prognosis can be very poor for some cancer types," de Ridder said.
And the information could help patients get on targeted treatments faster. "A lot of those oncogenes we detect are targetable," Lin said.