William Check, PhD
When Margaret L. Gulley, MD, started doing in situ hybridization to detect Epstein-Barr virus by RNA expression, about 20 years ago, she made an important observation. “Early on, we were getting samples from all over, since we were one of the few labs doing RNA testing on paraffin sections,” says Dr. Gulley, professor of pathology and laboratory medicine at the University of North Carolina-Chapel Hill School of Medicine. Dr. Gulley’s laboratory always ran a so-called housekeeping gene transcript as an internal quality control to check for specimen quality. “When we didn’t detect EBV, we knew it was a true negative result if we could find housekeeping RNAs in the specimen,” she says. Dr. Gulley noticed that specimens from one facility routinely yielded poor RNA preservation. In quizzing those at the facility about their tissue preparation procedures, she and colleagues found they didn’t want to smell formalin fumes at the grossing bench, so they were dropping the cassettes of fresh tissue into a water bath and only later transferring them to formalin.
“We thought that was the explanation for poor RNA quality—a major delay in fixation. That shows a major problem. There is no real standard for preparing paraffin blocks.” Many preanalytic variables that are not controlled for affect the quality of nucleic acid, she told CAP TODAY.
For most of the history of formalin fixation and paraffin embedding, standardization wasn’t critical. Today, when molecular assays are a routine part of laboratory testing, standardization has become an issue. “Clinicians are demanding that we do a lot more molecular testing on paraffin blocks, so we need to pay more attention to the quality of nucleic acid,” Dr. Gulley says. She shared the lessons she has learned in two decades of RNA testing, in October at the Association for Molecular Pathology annual meeting on genomic medicine in a session called “Solving Challenges in Genomic Testing of FFPE [formalin-fixed paraffin-embedded] Specimens.”
Paraffin embedding is the most practical way to preserve tissue, Dr. Gulley told attendees. “We don’t need to change pathology practice in a dramatic way. I don’t want to replace formalin fixation,” she says. “We can make formalin fixation better.”
Among the many steps for assessing and improving the quality of RNA in FFPE specimens, Dr. Gulley spoke of so-called spike-in controls—known RNA molecules added to the specimen at an early step in nucleic acid procedures. She cited in particular the work of the External RNA Controls Consortium (ERCC), set up in 2003 to develop spike-ins that can be added to the reaction mix and assayed downstream to evaluate performance of RNA-based assays. The ERCC accomplished its mission in 2010 with the publication of a set of 96 RNA standards, which many vendors now make available commercially.
Marina N. Nikiforova, MD, was a co-moderator of the session at which Dr. Gulley spoke. “Work in molecular diagnostics typically involves fixed tissue samples. It is not a perfect specimen,” Dr. Nikiforova, associate professor of pathology and director of the molecular anatomic pathology laboratory at the University of Pittsburgh Medical Center, told CAP TODAY. She says fresh frozen tissue is a much better source of RNA and DNA, with formalin fixation making molecular work more difficult. But fixed tissue has one big advantage, she notes: “It is widely available to surgical pathologists from specimens sent for surgical analysis.”
There are other advantages. “We can look at H&E-stained slides and make sure we are selecting the best area of tumor for molecular analysis and we can assess the number of tumor cells available for testing.”
Then, too, direct examination of fixed tissue slides makes it possible for the pathologist to exclude necrotic tissue, she says. And “it allows us to microdissect under the microscope or by laser capture.” Fixed specimens, Dr. Nikiforova concludes, are valuable to molecular diagnostics and can be analyzed by both conventional techniques and next-generation sequencing. “We currently use a commercially available panel for targeted NGS at our lab at UPMC on fixed tissue with great success,” she says. “The panel allows us to test small tumor biopsies for 46 cancer genes and help oncologists determine treatment strategy.”
In the molecular diagnostics service at Memorial-Sloan Kettering Cancer Center, a variety of tests are done on both DNA and RNA extracted from FFPE tissues, says Cyrus Hedvat, MD, attending pathologist, acting chief of the molecular diagnostics service, and laboratory director of diagnostic molecular pathology. “That applies particularly to solid tumors, since blood and bone marrow samples are received fresh,” he notes. In Dr. Hedvat’s service, testing for RNA typically involves looking for abnormal RNA transcripts in solid tumors, such as sarcoma fusion gene transcripts, to establish a definitive diagnosis or a differential diagnosis, for example, Ewing’s sarcoma. “We also use RNA assays to guide treatment, such as by expression of the mutant form of the gene for EGFR in glioblastoma,” he told CAP TODAY.
For these assays, Dr. Hedvat follows the principle that molecular assays must be adapted to the nucleic acids extracted from FFPE specimens. “What we have done,” he explains, “is to recognize that RNA from these samples is a combination of fragments and degraded molecules. Both RNA and DNA suffer fragmentation from fixation, RNA to a greater degree.” RNA and DNA fragments from FFPE specimens range from 150 to 200 base pairs in length. Primers used in PCR assays to detect nucleic acids are designed to be optimal for this size. “This way,” Dr. Hedvat says, “we can maximize the amount of amplifiable RNA or DNA present in FFPE samples.”
Dr. Gulley’ls AMP talk was titled “Tips to Improve RNA Profiling of Paraffin-Embedded Tissue.” In an interview, she explains why her focus was on RNA profiling. “What we are finding is that, while proteins have traditionally been thought to be the functional elements in cells, RNAs, and particularly noncoding RNAs such as microRNAs, are at least as important as proteins in activities that reflect normal versus disease status, as well as the kind of disease. We are realizing that testing RNAs is a valuable way to analyze disease states.”
Dr. Gulley began her talk at the meeting with a few simple but important lessons she has learned in two decades of doing in situ hybridization:
- RNA is variably preserved in fixed tissues.
- Fixation tends to inactivate nucleases.
- Acid decalcification is problematic for RNA extraction.
- Controlled analysis has showed that formalin fixation from 12 to 24 hours maximizes RNA yield; any longer diminishes recovery of nucleic acids.
To prepare fixed tissue for RNA extraction, Dr. Gulley deparaffinizes it in xylene and dips the tissue twice in 100 percent ethanol, then does macrodissection to enrich for the lesion. “Some pathologists worry that, with RNA profiling, the microscope will become antique,” she notes. The opposite is true: “Morphology is not going away,” she insists. “In fact, it’s crucial to knowing what you’re putting into your assay.”
She turned next to the critical area of controlling for RNA quality. One tool is the set of RNA controls from the ERCC—96 different RNA sequences, all nonhuman and nonpathogen—that are commercially available and certified by the National Institute of Standards and Technology. (Further information on the ERCC control set can be found in the Clinical and Laboratory Standards Institute document MM16-A.) In addition to assessing the quality of an RNA specimen—during extraction, storage, analysis, and detection—combinations of these spiked RNA molecules can be used to identify specimens to be analyzed together. “I am forever grateful [to the ERCC] for the work they did to develop QC materials that are now commercially available to help us ensure that the assays we are using perform as expected,” Dr. Gulley told CAP TODAY. “Over the last two years multiple vendors have made those RNA spike-in reagents available.”
The Microarray QC Project is another quality control effort, one that Dr. Gulley and her colleagues summarized in a review (Tang W, et al. J Molec Diagn. 2012;14:1â€“11).
Of multiplex RNA profiling and its demand for a more complex approach to QC, she says: “There is a real trend in molecular diagnostics to test many things in parallel to get more value than is provided by the single tests we tend to do now. When we create an expression profile we get information about dozens or thousands of RNAs expressed in tissues. In this way we can overcome the deficits that any one analyte might have in terms of specificity and sensitivity for the outcome of interest and get a better idea of what’s going on inside a cell or tissue.”
Dr. Gulley cites two examples of RNA profiling that are designed to be done on FFPE tissues. One is the FDA-approved Pathworks Diagnostics Tissue of Origin test, a 1,550-gene expression profile that attempts to match a given tissue to one of 15 common cancer types by its RNA expression profile. A second example is the Oncotype DX 21-gene real-time PCR assay to classify invasive breast cancer. “This assay is not FDA approved,” Dr. Gulley notes, “but it has been used in many trials and is well accepted in clinical practice.”
Dr. Gulley showed two RNA profiling assays that she and her colleagues are developing. One uses 50 transcripts to classify breast cancer cases into five types: luminal A, luminal B, normal, HER2, and basal. (For an illustration, see Fig. 3 in Tang, et al., cited above.) A second uses RNA profiling to analyze EBV transcripts in tumors (Tang W, et al. Infect Agent Cancer. 2012;7:21). “In this assay we looked at 96 different RNAs at once, including three ERCC spike-ins, 20 viral RNAs, and 73 human RNAs, demonstrating that we can simultaneously examine both human and pathogen RNAs, as well as QC analytes, in the same panel,” Dr. Gulley told CAP TODAY.
This assay might be used clinically in virus-related cancers, such as gastric cancer and lymphoma. Both assays are works in progress.
“RNA gene expression profiling is robust and can add value to pathologic consultations,” Dr. Gulley concluded. In this context, software supplied by manufacturers is crucial to understanding how well the assay is working, particularly since these assays can measure hundreds of thousands or even millions of analytes. One example she showed was an Agilent visual representation of RNA quality. “As a pathologist I like visual display,” she says. For breast cancer profiles, which include transcripts from ER, PR, and HER2 genes, the results from profiling can be compared with immunostaining for these same analytes, as an additional QC check. Dr. Gulley advised her AMP audience to take advantage of the redundancy offered by arrays to do additional QC. For instance, one can measure the same RNA multiple times or analyze it from both the 3’ and 5’ ends. She recommended that pathologists normalize RNA expression profiles to transcripts from housekeeping genes that are chosen for the particular specimen type and intended use of the assay.
In summary, doing QC of RNA profiles requires a new approach. “When testing so many things at once, it requires a new paradigm for doing QC,” Dr. Gulley says. “You can’t possibly have negative and positive controls for every analyte in a large panel. So what we are controlling instead is the overall process of classifying the sample into one of the important categories of outcomes, such as a molecular subtype of breast cancer.”
In her interview with CAP TODAY, Dr. Nikiforova, too, had thoughts on improving the quality of assays on RNA and DNA from FFPE tissue. She agrees that tissue should not be fixed in formalin for too long, certainly not for more than 48 hours. In addition, fixatives other than formalin might make nucleic acids inaccessible for extraction. In this regard, she cites the heavy metal fixative B5. Another typical error: sending tissue that was in the process of decalcification. “Decalcification solutions cause extensive DNA fragmentation,” she says.
In her laboratory, Dr. Nikiforova does nucleic acid assays for rearrangments in solid tumors, such as the genes RET plus PTC in thyroid cancer. Other genes she tests for are PAX8 and PPAR-γ. For these assays she uses real-time PCR.
Soon Dr. Nikiforova will do assays using next-generation sequencing. The Ion Torrent and Illumina MiSeq platforms are already designed for FFPE tissue, she says, adding that they allow sequencing of very short fragments of DNA and are successful. “We just introduced the Ion Torrent AmpliSeq panel for detection of 739 mutations in 46 different cancer-related genes,” Dr. Nikiforova says. “We evaluated it in more than 100 tumor samples, all from FFPE tissues, and had very good reproducibility, accuracy, and sensitivity of mutation detection relative to Sanger sequencing and, for lower-level amounts of DNA, PCR. We found hardly any failed specimens. And it requires a minimal amount of DNA, 10 ng, which is easily extractable from any tumor specimen, even small biopsies.”
At Memorial Sloan-Kettering, Dr. Hedvat says his laboratory mostly performs assays for mutations in solid tumors, such as EGFR in lung cancer, KRAS in lung and colon cancer, and BRAF in colon cancer. They use two reference genes for QC of assays in sarcomas, the TATA binding protein and actin. “These are internal controls to ensure we are getting sufficient amplifiable RNA from the sample,” he says. He has seen a phenomenon similar to Dr. Gulley’s experience of two decades ago: “We get a lot of outside samples. Among samples processed here over the last few years we have a 90 percent success rate, while 70 percent of samples from outside are successful in amplification of the housekeeping genes. So outside samples have a greater extent of degradation.”
Other controls that Dr. Hedvat uses are transcripts from a cell line known to have the gene variant he is looking for (a positive patient sample can also serve this function) and a cell line known to lack the aberrant gene.
Another method used to determine the quality of extracted RNA is direct analysis prior to doing the assay, but Dr. Hedvat and his colleagues haven’t found it useful or predictive of success. “Our best approach is to run the sample and look for amplification of the housekeeping gene,” he says. In some circumstances he can see using an Agilent bioanalyzer to check RNA quality before doing a reaction. “For example,” he says, “an Affymetrix test itself is fairly expensive, up to $1,000. You might do a more sophisticated reading of those samples to see how they will perform. Our assay is not so expensive and we have not found those types of pre-test measures useful.”
If an assay is not successful, as judged by amplification of the housekeeping gene, Dr. Hedvat says, “we will usually repeat the assay.”
Dr. Hedvat, too, is in the process of setting up next-gen sequencing. About a year ago his laboratory acquired an Illumina MiSeq benchtop sequencer. He has been validating this platform with Illumina’s TruSeq Amplicon cancer panel, which looks at 212 amplicons covering 48 genes.
In the same vein of working on multiplex gene panels, Dr. Hedvat is working on a leukemia panel using Raindance technology that detects abnormalities in 30 genes. “I can absolutely see [NGS replacing current methods] in the near future,” he says. “We are currently defining data analysis and processes so that I can be comfortable that we can do it reproducibly. Various professional groups are starting to develop standards. Standards really didn’t exist until a couple of months ago.
“All of the assays we are developing, such as the Illumina cancer panel, are designed to work with FFPE tissues,” Dr. Hedvat says. He highlights what he calls “the critical point”: “What makes them work for FFPE is that they are designed for the base pair length of the nucleic acid fragments.”
To illustrate this point, Dr. Hedvat talked about Raindance technology. “Raindance generates microdroplets, each containing one nucleic acid fragment.” These microdroplets are the input material for PCR; the microdroplets allow multiplex PCR assays with up to 1,000 different primer pairs in a single reaction.
Raindance has two products of this type. “The original product was designed because of the cost of primers,” Dr. Hedvat says. “They wanted to use as few primers as possible, so they designed the primers for fragments 800 base pairs long. This allowed them to use fewer primers to cover the region of interest. However, this product was not compatible with nucleic acids extracted from FFPE. It would not give adequate amplification because of the shorter fragments—150 to 200 base pairs—from FFPE tissue.
“So now Raindance has a DeepSeq product that works with shorter fragments, but it requires more primers, so it is more expensive.”
Dr. Gulley summarized this whole issue succinctly in the Q&A that followed her AMP talk. “FFPE is the best source of tissue for molecular assays,” she said. “Changing practice is hard and we don’t want to go looking for new fixatives. We need to adapt our molecular methods to FFPE tissues.” And that’s just what’s happening.
William Check is a writer in Ft. Lauderdale, Fla.