SAN FRANCISCO (GenomeWeb) – The Mayo Clinic has developed and launched internally an RNA sequencing assay to identify gene fusions relevant for cancer therapy selection, prognostics, and diagnostics.
The assay, which involves transcriptome sequencing and a bioinformatics filtering approach that focuses on identifying fusions in 571 genes, was described this week in the Journal of Molecular Diagnostics.
Kevin Halling, the senior author of the study and a medical geneticist at the Mayo Clinic, said that the assay includes a mixture of fusions that have therapeutic, diagnostic, and prognostic significance. Mayo Clinic has offered it clinically since January, and Halling said that the team is working on developing another RNA-seq assay specifically for hematological malignancies, and is also working on one that is compatible with formalin-fixed paraffin-embedded tissues so that oncologists outside of the Mayo Clinic can order the test.
In the published study, the team evaluated 76 samples, including 15 normal tissue samples and 61 tumor samples from a variety of different cancer types. Sequencing was performed on an Illumina HiSeq 2500 instrument and identified fusions were confirmed using RT-PCR and Sanger sequencing. In addition, the researchers made use of a customized bioinformatics pipeline, MAP-RSeq, to detect fusions.
Of the 61 tumor samples, 39 had previously been analyzed for fusions and rearrangements and there were 43 known fusions. Twelve tumor samples had not been previously analyzed.
The RNA-seq assay failed in two of the 39 specimens, containing two known fusions. Among the remaining 37 samples, with 41 known fusions, it identified 38 fusions. The assay had zero false positives in normal samples, so it had a sensitivity and specificity of 93 percent and 100 percent, respectively. The false negative results occurred in a B-cell lymphoma sample, a colon cancer sample, and an angiofibroma sample. The authors hypothesized that the RNA-seq assay did not detect the fusion in the B-cell lymphoma sample due to analytical sensitivity limitations. The fusion had previously been detected with a hybrid capture technique. For the colon cancer cell line sample and the angiofibroma sample, the RNA-seq assay did generate supporting reads for the fusion, but not enough to surpass the threshold of five supporting reads for the fusion to be called.
Among the 39 tumor samples that had previously been analyzed, the RNA-seq assay also identified an additional six fusions in five samples that had not been called before, which were validated by other methods. In addition, from the 22 samples that had not previously been analyzed, RNA-seq identified 16 fusions in 12 samples.
Of the 22 fusions that had not previously been detected, 18 were novel, and all 22 were confirmed by RT-PCR, giving the RNA-seq assay an overall positive predictive value of 100 percent.
Aside from determining the assay's overall accuracy, Halling said, the team wanted to evaluate it on different sample types. Although formalin-fixed paraffin-embedded tissue samples are the most common sample type, the fixation process tends to degrade nucleic acids, making fusions harder to detect. "The RNA is much more chopped up in FFPE," Halling said.
In order to evaluate the effect of degradation on RNA-seq results, the researchers chemically degraded RNA to varying degrees. Halling explained that the Mayo evaluates RNA samples using a metric called RIN, for RNA integrity number. High RINs mean higher quality samples. They analyzed a tumor universal human reference RNA cell line mix and artificially degraded the sample to RINs of 8.4, 7.6, 5.9, 4.9, and 3.9. As expected, the researchers noted that RNA-seq quality declined with a decrease in RIN, but noted that the biggest impact was seen the further away from the 3' end the gene fusion was located. For instance, the BCR-ABL1 fusion is around 5 kilobases away from the polyA tail, and at RINs lower than 8.4, the fusion was not detected.
Halling said that at the Mayo, pathologists use a method known as snap freezing to preserve clinical samples for genetic testing. That helps keep RIN values higher than in FFPE tissue, he said, but even then they can be as low as 3, depending on how long the tissue sample sits out before being frozen.
Nonetheless, he said that in order to offer the assay outside of the Mayo Clinic, the researchers are working to validate a method that can work on FFPE samples. To do that, he said, the team is working on developing a targeted RNA sequencing approach as opposed to a transcriptome assay, which will enable higher coverage of the specific genes.
However, Halling said, the best way to preserve RNA integrity is to actually culture the tumor cells. "Then you have a supply of live cells," he said, and RIN values from those specimens are often above 9.
A cell culture may be another option for outside institutions to send samples to the Mayo for testing, Halling said. Some institutions already routinely culture tumor samples, he said, so "if we can get the ordering physician to also order a culture, then that would circumvent problems with transporting frozen samples."
The clinical RNA-seq assay for gene fusion detection has been live since January, and although the total number of samples that have been analyzed clinically is still small, at about 15, the team has already discovered three new fusions, Halling said.
The assay has a list price of around $2,000 to $2,500, he said, and the Mayo Clinic will bill patients' insurance companies.
Mayo also offers an NGS-based oncology test to detect rearrangements, called MP-seq for mate pair sequencing. That test also detects fusions, Halling noted, "but won't tell you if they are expressed." In addition, he said, the RNA-seq assay "can detect down to the nucleotide level what's going on," while MP-seq uses shallow sequencing, so does not pick up any point mutations. "You have to design PCR primers to figure out the actual sequence at the base pair level," Halling said.
However, the MP-seq assay will detect some rearrangements that the RNA-seq assay will not. For instance, he said, there is a known clinically relevant translocation that results in the IGH gene being placed next to the MYC oncogene, but that doesn't actually create a fusion between the two genes. However, since IGH is a promoter, it upregulates MYC.
"We're still trying to get a good feel over when one [assay] should be used over the other, and we're doing a lot of work to figure out which test is best for which situation," Halling said.
In the future, he said, the Mayo team is also looking to validate the RNA-seq assay not just for identifying gene fusions in the 541 genes, but to also use the accompanying gene expression data.
He said one reason why the team based the assay on transcriptome sequencing, despite focusing its clinical analysis on just 541 genes, was to give it flexibility to validate other capabilities.
Several years ago when oncologists first began asking about an RNA sequencing assay, they were primarily interested in it for gene expression analysis. However, Halling said, a gene expression RNA sequencing assay is more challenging to validate clinically because there is not a lot of literature that shows how best to manage patients based on gene expression. Two exceptions however, are Agendia's MammaPrint and Genomic Health's OncotypeDx tests, which both use gene expression to predict whether breast cancer patients can avoid chemotherapy. On the other hand, there is a lot of data demonstrating the clinical utility of gene fusions, he said.
In the future, the Mayo team hopes to demonstrate how to better use and interpret gene expression data. Already, he said, there is an IRB-approved protocol that enables the gene expression data to be mined, although that data is not returned in the patient's clinical report.
Halling added that RNA sequencing is also useful for confirming that mutations identified in somatic DNA are actually being expressed. "If it's not expressed, it may not be targetable," he said.