ORLANDO, Florida (GenomeWeb) – Early-access customers of NanoString's Hyb & Seq sequencing technology and its Digital Spatial Profiling technologies described how they are using the technologies in cancer and infectious disease at the Advances in Genome Biology and Technology meeting held here this week.
As previously announced, NanoString plans to launch its DSP technology in early access in the second half of the year with full commercial launch in 2019, and it is targeting 2019 for an early-access launch of its Hyb & Seq platform.
The firm initially unveiled plans to develop the Hyb & Seq technology at the JP Morgan Healthcare Conference in 2016. At last year's AGBT, it debuted a prototype instrument, which was essentially a reconfigured nCounter machine, and demonstrated the sequencing of an 11-gene oncology panel from a formalin-fixed paraffin-embedded tissue sample.
Since then, the firm has continued to develop the technology, and last August it struck a partnership with Lam Research to help develop Hyb & Seq. During a company-sponsored workshop this week, Joe Beechem, senior vice president of research and development, said that as a result of the collaboration, the firm has now moved to using ordered arrays from random arrays, which has increased the throughput. Depending on the application, he said, the instrument will be able to generate between 20 million and 80 million reads, up from 4 million reads one year ago.
As previously described, the Hyb & Seq technology is a sequencing-by-hybridization scheme. DNA or RNA is read six bases at a time through the use of hexamer barcodes. Each barcode contains a sequencing domain that is complementary to the target molecule, as well as a barcode domain with three reporter regions: R1, R2, and R3. Each reporter region will bind a two-color reporter probe, which corresponds with two nucleotides. Thus, each 6-mer barcode is translated into the corresponding sequence two bases at a time.
During NanoString's workshop this week, Chris Mason, an associate professor at Weill Cornell Medicine, described how he designed a targeted panel to screen for 11 different microorganisms, including three gram negative bacteria, 5 gram positive bacteria, two fungi, and the RNA virus hepatitis C. The panel included both DNA and RNA detection probes. Mason said the organisms in the panel represented some commonly responsible for hospital-acquired infections, such as Escherichia coli and Staphylococcus aureus. In addition, they varied in size and GC content.
To validate the panel, he ran it on a mixture containing the reference genomes for the mirobes that was also diluted with human DNA and RNA as well as the DNA from a different virus.
The Hyb & Seq assay was able to detect the microoragnisms at less than 1,000 cells per milliliter from the same sample using a single tube. In addition, he said, "even when swamping the sample with millions of human cells, it was not a problem for detection," Mason said.
Next, he analyzed clinical samples from Weill Cornell Medicine from a variety of sites, including head epidural fluid, the spleen, bones, wounds, and lung that had all been previously diagnosed with pathogens.
Each sample was prepared for the Hyb & Seq in around 90 minutes, with no amplification or library prep steps. For each of the samples, the group analyzed them with Hyb & Seq panel and qPCR, and compared to the microbiology lab result. Hyb & Seq was 98 percent concordant with the microbiology report, with one discordant result. In the discordant sample, the microbiology report was positive for S. aureus, but both Hyb & Seq and qPCR were negative. After further analysis of the results, Mason said the group determined the result as ambiguous.
Importantly, he said, total turnaround time from sample to result was four hours, which would make it amenable for clinical applications.
Aside from infectious disease, NanoString is also positioning its technology for the clinical oncology market. One early-access customer, Anna Piskorz, a research associate at Cancer Research UK, described an oncology panel for ovarian cancer in a separate presentation at the conference. The panel looks at SNVs across the hotspot region of TP53, CNVs across three genes: TP53, MYC, and CCNE1, as well as differential gene expression from six genes: CCNE1, MYC, TP53, CDK12, TAP1, and PTEN.
Piskorz said that she was motivated to try the NanoString technology because she wanted an assay that could simultaneously evaluate SNVs, CNVs, and gene expression from clinical FFPE samples.
In addition, Piskorz said, tools that make use of known biomarkers to diagnose, determine treatment, and predict outcomes are sorely needed for ovarian cancer patients where the five-year survival rate is just 30 percent.
Ovarian cancer is actually a collection of five very different subtypes, she said. The most common subtype is also the deadliest — high-grade serous — but distinguishing between high- and low-grade serous subtypes in particular can be difficult with standard methods like evaluating cell morphology or with immunohistochemistry.
Recently, though, some studies have shown that looking at somatic mutations in the TP53 gene may be a more sensitive approach, she said, citing a study where although IHC was normal, patients actually had TP53 mutations and the high-grade serous form of the disease.
Aside from TP53 mutations, which can indicate subtype, Piskorz said other biomarkers, such as certain gene amplifications, can be used to predict treatment. As such, she said, it was important to have a test that could analyze not just somatic variants, but also copy number changes and expression differences.
Using the Hyb & Seq protocol, turnaround time from sample to result was less than 24 hours.
To validate somatic variant detection on the Hyb & Seq, the group tested it on synthetic mixtures with known mutation frequencies as well as on a cell line that had a known mutation at a frequency of less than 1 percent. In addition, Piskorv compared the Hyb & Seq for TP53 mutation detection to amplicon sequencing on the Illumina MiSeq for an FFPE sample and found that for four different variants, the Hyb & Seq resulted in error rates of 0.11 percent to 0.59 percent while the MiSeq error rates were between 0.49 percent and 0.95 percent.
In order to evaluate Hyb & Seq performance for measuring copy number changes, the technology was compared to whole-genome sequencing, and to assess performance for gene expression profiling it was compared to NanoString's nCounter system. In both cases, the technology produced results comparable with those orthogonal technologies.
"The method is accurate and reproducible," Piskorv said.
Currently, Piskorv said that the TP53 SNV portion of the assay is done in a separate tube as the gene expression profiling and copy number analysis, but said that the goal is to eventually have it as one assay done in a single tube.
In addition, she said, she would like to work to expand the assay to include additional relevant genes. For instance, she noted that the BRCA1 and BRCA2 genes could be important ones to add, since mutations in those genes can have implications for treatment.
Ultimately, "we want to use this assay in the clinic to identify potential therapies, do better patient stratification, and to enroll patients in clinical trials," she said.
Aside from Hyb & Seq, NanoString has also been developing its DSP technology, which uses oligonucleotide barcodes with photocleavable antibodies to tag proteins and genes on a slide. When light is shone on the region of interest, those oligo tags are cleaved and then sucked off the slide with a microcapillary for further analysis. Spatial identity is preserved, since the oligos from one region are removed and can be deposited into a single well of a 96-well plate. The process can be repeated for each region of interest.
"The frustrating thing about sequencing … is that it erases spatial distribution," Beechem said. The intention of DSP is to combine the benefits of microscopy with genomic and proteomic technologies.
Initially, NanoString had been developing the instrument to be compatible with its nCounter, but it has since expanded it to enable read out to be done via next-generation sequencing and has tested it using both Illumina MiSeq and its own Hyb & Seq chemistry, demonstrating comparable results on the three platforms.
Michael Davies, an associate professor at MD Anderson Cancer Center, has tested the technology on samples from melanoma patients. The technology has been useful for evaluating heterogenous tumors, he said. For instance, he described using the DSP in conjunction with the nCounter to look at PTEN deletions, which indicates resistance to checkpoint inhibitors. Using DSP, he said his group was able to see that from one FFPE slide, one portion of the tumor sample had the PTEN deletion, while another portion did not.
In addition, he's tested the technology on samples from melanoma patients with brain metastases, comparing the brain metastases to non-brain metastases. Metastatic tumors in the brain are the most common cause of death for melanoma patients, with average survival time of only four months. In addition, he said, there has been some evidence that brain metastases have unique molecular and immune features. For instance, AKT activation has been found to be higher in brain metastases versus other metastases. So, the goal was to use DSP to try to better understand these differences.
Davies said that the group currently is evaluating 98 brain metastatic samples using exome sequencing, RNA sequencing, as well as the DSP technology. "The advantage of this technology is that from a single FFPE slide we can look at multiple markers," he said. Now, he said, the group is able to do an in situ analysis of more than 1,000 RNAs in the individual regions of interest.