NEW YORK (GenomeWeb) – Formalin-fixed, paraffin-embedded tissue samples are increasingly used for molecular pathology assays. Vast archives of FFPE blocks containing patient biopsies are also a potential goldmine, particularly for oncology researchers. However, lurking variability in PCR or sequencing analyses of these samples remains a problem.
To get to the bottom of this issue, a new study compared pre-PCR processing done by 13 different pathology labs with experience in BRAF and EGFR mutation detection. Researchers from the labs are listed as co-authors on the study and include six from the UK and seven from the US.
The results, published online last week in the Journal of Clinical Pathology, revealed variability in measurements of DNA quantity, purity, and integrity from standardized samples. The study cautions that pre-PCR processing of FFPE blocks may unwittingly trip up other researchers, particularly if they quantify DNA using the Nanodrop method.
The results were "a bit worrying," Rifat Hamoudi, a molecular pathology researcher at University College London and one of the organizers of the study, told GenomeWeb in an interview this week.
Hamoudi's group provided each of the labs a blinded set of eight FFPE curls from standardized blocks.
Two out the eight samples contained cells that had been engineered by UK reference standards company Horizon Diagnostics to express ratios of BRAF and EGFR mutations. The set also included a cell-negative control — essentially just a curl from a block of paraffin — and four shavings from blocks containing cells without mutations. A sample of embedded tonsil was also considered a control.
Each lab was instructed to isolate DNA from the shavings and carry out assessment of quantity, integrity, and purity using their routine protocols. Then, they were asked to do mutation detection as per their usual methods, which included assorted combinations of qPCR, Sanger, next-generation sequencing, and pyrosequencing.
Alarmingly, the study found that Nanodrop quantification overestimated the amount of extracted DNA by about five-fold, Hamoudi said.
Specifically, the study showed that 10 out of the 12 labs that used Nanodrop reported DNA yields averaging about 235 nanograms from what were, in fact, cell-negative FFPE shavings.
The Nanodrop protocol uses a spectrophotometric approach dependent on diffraction, Hamoudi explained. Discussions with the labs that had the highest false DNA quantitation suggested perhaps a processing error left some small carryover of paraffin which caused diffraction, he said. His lab used an alternative method in the study, Qubit fluorometric quantitation from Thermo Fisher's Life Technologies. It takes longer, but, because it uses an intercalating dye, it distinguishes DNA from RNA, which is useful if the initial nucleic acid extraction was imperfect. This is supported by a PLoS One study last year which also concluded NanoDrop heavily overestimated DNA concentration in FFPE samples compared to Qubit.
In quantifying the BRAF and EGFR mutations in the standard samples, the study used a digital droplet PCR platform from Bio-Rad. Hamoudi said his lab was introduced to that method through the lab at Horizon, and purchased the platform about six months ago. The method is quicker and cheaper than NGS, he said, but the quantitation is as sensitive, and they will be using ddPCR to study rare cancer mutations in the future.
The study also measured PCR inhibitors in the reference samples using an assay called SPUD. Manufactured by Asuragen, a company that also participated in the study, the method uses potato-specific DNA, Hamoudi explained. Researchers at Asuragen published a study in Genome Medicine last year that used SPUD as part of a protocol for FFPE analysis.
While the embedded engineered cells may have been fully fixed in their formalin bath, biopsy tissue fixation can be more variable, and may also differ depending on the lab. This can affect the integrity of extracted DNA.
For integrity measurements in the standard samples, Hamoudi's lab used the BIOMED-2 assay, while Asuragen used a probe-based quantitative functional index assay, or QFI, in which fewer copy numbers of a common region are amplified if the DNA is very degraded. BIOMED-2, meanwhile, is a multiplexed PCR of four genes that is then run in a gel, with the ladder indicating whether the sample DNA is high or low molecular weight. This assay takes longer than the QFI, but is more visual and gives a better sense of DNA degradation levels, Hamoudi said.
Notably, the biggest variation in DNA recovery among the different labs was in the control FFPE from tonsil. Tissue has assorted enzymes and 3D structure that may complicate processing, but is also the kind of sample most labs will be analyzing.
"Based on this, I think people need to look at DNA purity, integrity, and quantitation very carefully," Hamoudi said.
The other take-home message is that, although Nanodrop analysis is a widely used, fast, and simple method, clinical pathology labs should be cautious about placing too much confidence in its results. This is particularly true if subsequent PCR or sequencing protocols require diluting the DNA.
"If you don't know [the DNA is overestimated] and you over-dilute the sample, you're going to start with a very small amount of DNA, and that can have wide implications on your sequencing detection," Hamoudi said. Sequencing "noise" can lead to both false positive and false negative results, he said.
For urgent clinical samples, Nanodrop could still be used "as long as you know it overestimates the amount by, say, five-fold, and factor that into the calculations," Hamoudi said.
Interestingly, it was not difficult to convince a dozen other molecular pathology labs to participate in the study, Hamoudi said. "They all do mutation detection regularly, and when we spoke to them they were also interested in the same thing, because nobody has done a study ... of the source of error, or validating the accuracy of the different methods used in different sequencing labs."
The labs were also happy to have the results, he said. "I think we all learned from this the fact that the error is not really in the actual sequencing, but in what happens before."
Past mutation detection from FFPE might now benefit from "double checking" data, Hamoudi suggested. "If the lab didn't measure quantity, purity, and integrity, but just did the mutation detection [after] using Nanodrop, the chances are it is probably worth looking at this again," particularly if the detection was then done using a less sensitive method like Sanger sequencing or pyrosequencing, he said.