Charna Albert
February 2024—When multimodality testing reveals discordant biomarker results, which method is correct? Annette S. Kim, MD, PhD, and JinJuan Yao, MD, PhD, in a CAP23 session last fall used their cases to share strategies for resolving discrepancies—or, in some cases, what look like discrepancies.
“All pathologists must be prepared to understand the underlying causes of the discordance to advise clinicians appropriately,” said Dr. Kim, Henry Clay Bryant clinical professor of pathology and director of the Division of Diagnostic Genetics and Genomics at the University of Michigan. “This may mean resolving differences between positive and negative results or interpreting noncategorical data, such as weak or equivocal.”
Dr. Kim is vice chair of the CAP Personalized Health Care Committee. She presented with fellow committee member Dr. Yao, who is assistant attending at Memorial Sloan Kettering Cancer Center. Two of the four cases they presented are reported here. For more on this topic, see the CAP’s CME/CE program (https://shorturl.at/yAOZ6).
Case No. 1 is that of a 46-year-old woman with a midline intraventricular mass with possible invasion of the corpus callosum. Fig. 1 is composed of the low-power view, showing areas of viable and necrotic tumor (top), as well as the high-power view, showing the high-grade histopathologic features of the tumor (below), with an estimated tumor purity of 90 percent.
The mass was diagnosed as a high-grade glioma.
Testing for several genetic alterations must be performed on diffuse gliomas to diagnose these tumor types for proper clinical care, according to guidelines published in 2022 (Brat DJ, et al. Arch Pathol Lab Med. 2022;146[5]:547–574). On diffuse gliomas that are wildtype for IDH1/2 and H3, BRAF mutation testing (V600) may be performed. Though the evidence to support this recommendation was assessed as low, the recommendation was influenced by the availability of targeted therapy, in addition to a limited number of studies.
Immunohistochemistry for the R132H variant of IDH1—the most frequent IDH1 mutation—was found to be negative, Dr. Kim said. The IHC for BRAF V600E was positive.
“BRAF p.V600E is found at varying prevalence in gliomas,” she said. Though it’s more common in low-grade disease, “even in high-grade gliomas, up to eight or nine percent can have a BRAF mutation, so it would not be unreasonable to have a BRAF-positive high-grade glioma.” She and her colleagues elected to confirm the BRAF result with orthogonal testing by digital droplet PCR (ddPCR).
In the assay they used, each droplet contained, in addition to a single template molecule of the patient’s DNA, fluorescence resonance energy transfer probes, in which a fluorophore is attached to a quencher through an oligonucleotide. The oligonucleotide is complementary either to the wildtype strand of DNA for the locus of interest or the mutant strand. As polymerase is added, exonuclease activity releases the fluorophore from the quencher. “Your droplets will turn color, and you can measure the color of the fluorescence through a flow cytometer,” she said, denoting whether the DNA is mutant or wildtype.
The patient was negative for BRAF p.V600E by ddPCR. “So the IHC was positive, but the molecular was negative.”
“Our initial assumption was that something might be wrong with the molecular result,” she said. “We thought the molecular was a false-negative.” In the ddPCR assay validation, “we did find this particular assay would detect BRAF p.V600K”—the second-most common variant at that locus—“but at a lower fluorescence level, because there was a different affinity for the probe than the typical p.V600E variant.” They debated whether the tumor could be harboring a novel V600 variant, detectable by IHC but not by the molecular test. “One of the reasons we hypothesized that the molecular might be a false-negative is because through the CAP, we’ve published data showing that dinucleotide variants of BRAF at the V600 codon can be missed by many molecular assays.”
In the CAP study, laboratory-developed testing methods for BRAF performed better than FDA-approved companion diagnostics, with LDTs demonstrating 96.6 percent acceptability and FDA-approved tests 93 percent (Moncur JT, et al. Arch Pathol Lab Med. 2019;143[10]:1203–1211). The main cause of the discrepancy was p.V600K analysis, with LDT acceptability rates at 88 percent and FDA-approved assays at 66.1 percent.
“When challenged with a dinucleotide variant such as V600K, the FDA-approved assays did significantly worse,” Dr. Kim said.
The FDA-approved platform with the lowest pickup rate, she said, has a sensitivity of about 10 percent neoplastic cellularity to detect p.V600E, but “the probe designed to bind to V600E does not bind well at all to V600K. As a result, the threshold to be able to pick up this variant is 60 percent neoplastic cellularity, and in that case the variant would be misidentified as a V600E. This explains why we had qualms about some of these molecular assays that might involve hybridization of a mutation-specific oligo,” she said. It seemed possible that their molecular assay, too, could have missed some other novel or uncommon dinucleotide variant of V600.
Next-generation sequencing, which is a methodology that should identify all the different variants at this site agnostically, also was negative. “So the conclusion from this study was actually that the IHC was a false-positive,” she said. The IHC revealed lysosomal staining with a strong granular staining pattern (Fig. 2), “as opposed to a BRAF-positive control, which tends to have a diffuse, almost muddy-looking staining.”
One review of the literature reported high but not perfect sensitivity and specificity in IHC for BRAF (Ritterhouse LL, et al. Semin Diagn Pathol. 2015;32[5]:400–408). Compared with molecular testing in melanoma, thyroid, and colon cancer, IHC had a sensitivity range of 71 to 100 percent and specificity range of 62 to 100 percent. “So there was a significant false-positive rate,” she said. According to the review, a contribution to the wide range in specificity was from the results of one study that reported high numbers of false-positives, in part because of tumors scored positive when as little as 10 percent of the tumor demonstrated moderate to strong staining. “In which case it might be positive by IHC but negative by molecular,” Dr. Kim said.
High background staining, too, can cause issues, as with the lysosomal background staining in the false-positive IHC for BRAF from the case. “Another thing that should be taken into consideration is that BRAF IHC assays almost always are validated against melanomas,” she said. “They don’t take into account the background noise you might get in nontumor cells in some of these other tissue types.”
The final resolution: The mass was a diffuse glioma, IDH-wildtype and BRAF-negative. “And although the ddPCR for BRAF was initially thought to be a false-negative, in fact the IHC was a false-positive.”
Case No. 2 involved a partially obstructive rectal mass, discovered in a colonoscopy for blood in stool.
The patient, a 52-year-old woman, had a complicated past medical history. She was diagnosed at age 44 with POEMS (polyneuropathy, organomegaly, endocrinopathy, monoclonal plasma cell disorder, skin changes) syndrome after a two-year history of bilateral weakness, skin hyperpigmentation, thrombocytosis, and monoclonal gammopathy with elevated vascular endothelial growth factor levels.
“The patient did undergo an autologous stem cell transplant at age 45 and was in hematologic remission but still had a lot of neuropathy,” Dr. Kim said. She was diagnosed with Barrett’s esophagitis at age 49.
Initial radiology staging was T4N2M1. The liver MRI identified a segment 7 (0.7 cm) and 4a (3.2 × 3.8 cm) small liver metastases. The pelvic MRI found a T4N2 tumor involving the peritoneum and uterus.
The patient was diagnosed with invasive rectal adenocarcinoma by rectosigmoid biopsy and subsequent resection (post-neoadjuvant therapy). Residual tumor remained after therapy (stage ypT3NO) (Fig. 3).
As in all colorectal cancer cases, Dr. Kim said, a hereditary cancer syndrome could have been at play, of which Lynch syndrome is the most common.
Of the genes associated with hereditary colorectal carcinoma, MLH1, MSH2, MSH6, and PMS2 usually are examined by IHC. Mutations in the epithelial cell adhesion molecule, located near the MSH2 gene, also can result in loss of MSH2. “Those are the ones we typically evaluate up front in Lynch syndrome,” she said.
The patient’s mother had colon cancer and melanoma, an uncle had lung cancer, and a paternal grandmother also had colon cancer.
Testing methods for Lynch syndrome and other hereditary cancer predispositions range from single-gene to massive multigene testing. “There is HER2 IHC with reflex to FISH. There is also IHC for PD-L1, MLH1, PMS2, MSH2, and MSH6, and that can then result in reflex testing, either to BRAF or MLH1 methylation testing,” Dr. Kim said. KRAS/NRAS expanded mutation testing (exons 2, 3, and 4), too, is another. Microsatellite instability, PCR, NGS, and other modalities for germline testing have been included in the most recent NCCN guidelines, and NGS for other somatic mutation testing and for mismatch repair signatures, MSI-specific loci, tumor mutational burden, and other tests also may be done. Members of the CAP Personalized Health Care Committee are reviewing methods for examining MSI by NGS and preparing a paper for publication, Dr. Kim noted.
Rectal adenocarcinoma is typically diagnosed on biopsy and then treated with chemoradiation before resection, so tissue for testing is often scant at diagnosis. “As a result, every institution has its own algorithm for microsatellite instability or mismatch repair deficiency,” she said. At the University of Michigan, the algorithm begins with IHC for mismatch repair protein expression. “So that was what we performed. And all four markers”—MLH1, PMS2, MSH2, and MSH6—“were completely intact in this patient” (MLH1 only shown as an example, Fig. 4).
With mismatch repair IHC, “typically we are looking for loss of individual markers or particular patterns of loss to indicate specific germline mutations.” For example, “MLH1 and PMS2 together form the mutLα complex. Its function is lost cooperatively.” If all four markers are intact, the MMR pathway is typically assumed intact, “though there can be rare exceptions to this.” But if either of these markers is lost, the two are typically lost in tandem. “And that would then lead you to look for a BRAF mutation or somatic MLH1 promoter methylation. Rarely, it can be due to an MLH1 germline mutation.”
MSH2 and MSH6 together form the mutSα complex. “That complex is lost collaboratively as well,” she said. Typically, the loss is driven by an MSH2 germline mutation that causes loss of function in both markers. Finally, in some cases, PMS2 or MSH6 alone could be lost as a germline mutation, which would result in loss of function in only that marker. In Lynch syndrome, 40 percent of mutations involve MLH1, 34 percent involve MSH2, 18 percent involve MSH6, and eight percent involve PMS2.
False-negatives can occur in MMR IHC. “Sometimes there’s variable or faint MMR protein staining,” Dr. Kim said. True subclonal loss of expression, likely representing tumor heterogeneity, can also occur, which can confound testing on small samples in particular. “It might look like it’s intact or it might look like it’s absent, depending on where you biopsy.” And overall expression can be reduced after chemoradiation, “especially with regard to MSH6, and that can look like a false-negative as well.”
False-positives, too, can occur. “One uncommon cause for a false-positive might be missense mutations that occasionally give rise to intact expression of a functionally inactive protein.” But atypical staining patterns are the more common cause. “For instance, MLH1 can have a granular or dot-like pattern, and that should be interpreted as negative.” Finally, suboptimal specimen processing, inadequate validation, or poor quality control can cause either false-negatives or false-positives.
Based on the patient’s intact IHC staining for all the common mismatch repair proteins, “we would have assumed this was a sporadic case with intact mismatch repair function,” Dr. Kim said. But when MSI testing was performed, it was positive for MSI-high.
One function of the mismatch repair system is to correct single mismatches, she said. “Proliferating cell nuclear antigen and the MSH6/MSH2 complex bind to the identified mismatch, and then the MLH1 and PMS2 complex initiates exonuclease activity” that cleaves out the mismatched region. “DNA polymerase fills in the gap and puts in the correct nucleotide, and then there’s a DNA ligase step.” Defects in the mismatch repair system lead to high tumor mutational burden.
The mismatch repair system also corrects errors caused by replication slippage, often referred to as “stutter.” Stutter occurs in homopolymer regions when polymerase’s attempt to replicate a sequence results in insertions or deletions. Stutter occurs in vivo and in vitro. “If you have defects in your mismatch repair system, you can have a high number of insertions and deletions,” referred to as indels, she noted.
Replication slippage can result in a range of sizes of homopolymer regions, both by PCR and physiologically. “Intact MMR systems can correct the typical replication slippage, but MMR deficiency results in microsatellite instability.” In MSI testing, to identify slippage or stutter, PCR primers are designed to flank the area of a mononucleotide repeat. “Then we look at the size of the PCR amplicons,” she said. Using the Bethesda criteria, five loci are examined. “If two or more show this particular type of stutter—and this is found in about 10 to 20 percent of sporadic cancers, but also in Lynch syndrome—that would be called MSI-high.” One of five loci would be considered MSI-low, which typically behaves clinically as microsatellite stable, and zero of five would be considered microsatellite stable.
In the case of the 52-year-old woman, five of five mononucleotide loci (BAT-25, BAT-26, NR-21, NR-24, MONO-27) demonstrated microsatellite instability.
After an MSI-high finding, BRAF testing is typically performed, and in this case testing by capillary electrophoresis was performed with allele-specific PCR targeting p.V600E and p.V600K. The result: “Stone-cold negative for V600E and V600K,” Dr. Kim said.
Sixty percent of sporadic MSI-high colorectal cancers have a BRAF p.V600E mutation. “What this means is the identification of a mutation is incredibly informative if present,” she said. “However, if it’s negative, it doesn’t mean anything. It could be sporadic, or it could be germline. And this is one of those cases where the result was negative, and so we don’t know.” With IHC markers intact, positive MSI testing, and an uninformative BRAF, “does this patient have Lynch syndrome or not?” she asked. “And what is the cause for this discrepancy?”
They performed next-generation sequencing. “Seventy-four variants came through the pipeline. Twenty-eight were insertions or deletions, and 23 of those were found at homopolymer indels. So this is the perfect signature for microsatellite instability.”
Looking at each nucleotide that was altered, including those flanking the altered nucleotide on either side, there were many C to T changes, manifest as a single base signature or SBS6—a signature strongly associated with mismatch repair deficiency. “So it’s not just which genes are mutated and what kind of a mutation it is, it’s also the pattern of signatures that can be used to define MMR.” Another pattern associated with mismatch repair deficiency—the ID7 pattern—was present as well. And there’s a loose association between tumor mutational burden and the homopolymer indel burden in cases of mismatch repair deficiency.
Germline colorectal cancer panel testing was performed. An MLH1 point mutation, MLH1 c.113A>G (p.Asn38Ser), considered pathogenic in ClinVar, was identified. The patient did have Lynch syndrome. And the discrepancy was caused by a rare MLH1 missense mutation that resulted in defective MMR protein function but intact MLH1 protein expression.
“IHC is a screening test,” she said. “In cases of compelling family history, use your clinical judgment to consider germline or additional testing. Many of the common algorithms clinical labs use could miss this case entirely, which would have implications for therapy and for family testing.”
There was no “false” assay result in this case, she said, although the results appear discrepant. “They are true results with complicated underlying biology and individual assay limitations.”
Charna Albert is CAP TODAY associate contributing editor.