Metastatic tumor cells residing in lymph nodes are surrounded by a large number of leukocytes and stroma cells that do not contain genetic alterations, resulting in low tumor fraction. Furthermore, tumor genetic profile can evolve over time under various selective pressures, leading to tumor heterogeneity [27, 29, 30]. Consequently, actionable mutations can be at low VAFs, and assays with poor VAF limits of detection can exhibit clinical false negatives that deprive patients from optimal targeted therapies. To mitigate this problem, clinical pathology labs may enrich tumor content through labor- or capital-intensive macrodissection or laser microdissection. However, these approaches cannot overcome tumor subclonal heterogeneity developed by various tumor evolution mechanisms. Thus, an assay that detects a range of BRAF mutations with LOD below VAF of 1% will likely produce higher clinical sensitivity than current clinical practice. With the BDA Sanger approach, a range of BRAF mutations are detected in FFPE tissue-derived DNA with VAF down to 0.26%.
Analytical and clinical validation of BRAF V600E mutation. We first validated the performance of the BDA BRAF assay on detecting and quantitating the most common BRAF V600E mutation using Horizon Discovery reference materials and synthetic DNA strands. We ran the BRAF assays on reference materials with VAF values ranging from 0.1% to 100%. Two reactions were performed for each sample, one for selective variant enrichment and the other for input quantitation. The normalized result from taking the difference of the two Ct values (Cq) is independent of DNA input and can be used to determine the mutation’s VAF. Higher VAFs in the reference material were reflected as earlier amplification and thus lower Cq values (Fig. S1a).
Cq values and log VAFs exhibited linear correlation with an R2 greater than 0.99 for V600E. The equation derived from the linear fitting was used to quantitate unknown VAFs in the original sample (Fig. S1b). Median Cq value of 0.1% reference sample was 10.2, and that of WT sample was 11.8, and value of Cq for WT was always at least 1.0 higher than that of the V600E sample at 0.1% VAF (Fig. S1b). The Sanger sequencing trace of the qPCR amplicon product showed V600E at roughly 50% after BDA enrichment (Fig. S1c), confirming the identity of the mutation.
Next, we applied the BDA BRAF assay to 12 FFPE SLNB or CLND samples from seven metastatic melanoma patients. For 5 of the patients, we prepared paired tumor-enriched/not enriched FFPE samples; paired tissues for other two patients were not available (Table. 1). Results showed that 7 FFPE specimens from 4 patients (patient 1-4) had V600E mutations with VAFs ranging from 0.26% to 38.30%. The two specimens from patient 6 were identified as V600-WT. The remaining two specimens from patient 5 had two separate G>A variants (Fig. S2d) that are characteristic of deamination damage associated with FFPE treatment and storage. Because the incidence of two G>A mutations appearing simultaneously in the same sample in such close proximity is low, we do not believe these are real mutations. More generally, identification of multiple G>A or C>T mutations in the same sample in close proximity are likely to be hallmarks of unrepaired FFPE damage.
We also performed ddPCRTM BRAF V600 screening kit from BioRad, anti-BRAF V600E IHC staining and NGS on matched tumor samples from the same individuals (Table. 1). For a low tumor fraction sample of patient 1 (Fig. 2a), macrodissection was able to increase tumor fraction as seen in both BDA quantification results (Fig. 2b, 2d) and ddPCR results (Fig. 2c, 2e). The lowest VAF detected by BDA assay was 0.26% in sample 129292, which had VAF of 0.36% assayed by ddPCR. Sanger trace of the same sample showed that the ultra-low level variant was enriched by BDA assay to approximately 50% (Fig. 2f-h), allowing Sanger to visualize low level mutations. For all V600E-positive specimens identified by BDA assay, ddPCR results were all positive and showed high quantitative concordance with BDA assay even at VAF levels lower than 1% (Table. 1, Fig. 3). ddPCR did not detect any variant molecule in samples with no BRAF V600 mutation according to BDA assay. Nevertheless, suspected FFPE damages at nearby loci were not detected by ddPCR as it could not detect mutations outside codon 600 (Fig. S2d).
For samples with mutation VAFs greater than 1%, BDA Sanger, ddPCR, and IHC all had high quantitative concordance. For example, sample 129288 had VAF of 38.30% identified by BDA, which is consistent with a heterozygous mutation in 81.05% of cells identified through IHC (Table.1) It is also worth noting that IHC could not be used to find V600E positive cells for patient 6, because it was pigmented and melanin produced similar brown color that would be mistaken as positively stained in DAB (3,3’-diaminobenzidine) detection system (Fig. 2i). All three molecular diagnostic approaches (BDA Sanger, ddPCR, and NGS) do not have this limitation (Fig. 2j-k).
The BRAF mutation status for these samples were qualitatively reported by MD Anderson Molecular Diagnostic Laboratory, which applied one of three separate targeted NGS panels on the Ion Torrent or Illumina NGS platforms . These panels typically have a limit of detection of about 5% VAF, but given the importance of the BRAF V600E mutation, samples with sub-threshold V600E VAFs were re-analyzed by a second NGS panel using a different sequencing platform. Because Illumina and Ion Torrent are based on different detection principles (optical fluorescence vs. pH) and have different error profiles, it was unlikely that the same false positive variants would appear in both platforms. Thus, if a sample analyzed by both NGS platforms contained reads supporting the V600E mutation, then the mutation was called regardless of implied VAF.
Analytical and clinical characterization of BRAF non-V600E mutations. Based on the COSMIC database, more than 10% of BRAF mutations in melanoma are non-V600E mutations (Fig. 4a) and some (i.e., V600R, L597Q/R/S, K601E) have been recently reported to be associated with efficacy of BRAF inhibitor therapy . The BDA Sanger assay is in principle capable of detecting these mutations, and the lack of corresponding mutation peaks indicates that the 12 samples tested are likely negative for all other BRAF mutations in this region.
To confirm the presumed result, we performed analytical validation of the BDA Sanger assay using spike-in reference samples with the L597R, L597Q, L597S, V600K, V600R, and K601E mutations (Fig. 4b-4g). The limit of detection for all tested mutations were no worse than 0.5% VAF, and for most mutations 0.1% VAF would be confidently called based on qPCR Cq value alone (Fig. 4b-4g). The quantitation formula generated from linear fitting varied for different mutations, suggesting that the variant enrichment performance was mutation-specific.
The BRAF V600K mutation is covered by the ddPCR BRAF V600 screening kit; consequently, the BDA Sanger and ddPCR assays were concordant for sample 129538 from patient 7 (Fig. 4h). However, because ddPCR is limited to a single fluorescence color channel for reporting all mutations in the kit, the V600K mutation in patient 7 could not be distinguished from the V600E mutations in patients 1 through 4.