Transitioning from FISH to NGS for immunoglobulin translocation detection in multiple myeloma

Abstract

Abstract The International Myeloma Society and International Myeloma Working Group recently updated the definition of high-risk multiple myeloma and for the first time this definition includes genomic features that can only be detected by DNA sequencing. As a consequence, the working group recommended clinical testing should transition from FISH to DNA sequencing-based technologies that can detect the translocations, copy number alterations, and coding mutations used to calculate risk along with beta-2-microglobulin levels. To support this transition, we characterized the translocation breakpoints in 68 commercially or publicly available myeloma cell lines to create a gold-standard reference set for the community. This was then used to evaluate a series of different structural callers or dedicated immunoglobulin translocation calling tools. Additionally, as part of the clinical validation of a rapid whole genome sequencing platform, we established the accuracy, limit of detection and precision of our platform and analytical workflow for the detection of immunoglobulin translocations. All cell lines and patient samples used for clinical validation were sequenced on a NovaSeq X Plus using PCR-free libraries sequenced to 30-40x and >118x, respectively. To confirm complex events in the cell lines, we performed long-read sequencing using a PacBio Revio with libraries selected to have inserts exceeding 15kb. We tested the ability of Manta v1.6, IgCaller v1.3, DRAGEN v4.4.6, and SCITAV v0.6.5 to call the individual derivative chromosomes or at least any one derivative from a given translocation. Within the panel of myeloma cell lines, we detected 160 junctions from 92 distinct balanced or unbalanced translocation events in 66 cell lines between a common target gene in myeloma (NSD2, CCND3, MYC, MAFA, CCND1, CCND2, MAF, MAFB) and one of the immunoglobulin loci (IgH, IgK, IgL). Of the individual junctions, we found 9 with breakpoints proximal but outside of the immunoglobulin loci, 15 with insertions of a tertiary part of the genome ranging from 42bp-338kb and one with an immunoglobulin breakpoint containing 565bp of novel sequence. Approximately, 5.6% of derivative chromosome junctions will be undetectable by standard short-read sequencing approaches given the 8/15 events with insertions exceeding a typical WGS library or the one with unmappable sequence. Although 7/15 events with insertions under 260bp could likely be detected with some whole genome sequencing tests, we limited our comparison of bioinformatics approaches to the 135 junctions from 82 translocations that are within the immunoglobulin loci and do not contain defined insertions at the junctions. Using each calling tool in a tumor-only mode, the individual derivative call rate ranged from 66.7% (Manta) to 97.8% (SCITAV), with IgCaller detecting 76.3% while DRAGEN detected 88.2%. Since the majority of translocations are balanced events with two detectable derivatives, we assessed which percentage of translocations are detectable when at least one derivative is detected. This raised the detection rate range from 81.7% (Manta) to 100% (SCITAV), with IgCaller detecting 87.8% and DRAGEN 91.5%. A common theme in undetected breakpoints was immunoglobulin windows with low mapping quality resulting in read pairs being distributed between homologous regions, which resulted in the same event being called multiple times or these reads being ignored, leading to a missed call. For clinical validation we compared the WGS results with gold standard FISH assays from each patient and SCITAV had a sensitivity and specificity of 100%, while DRAGEN had a sensitivity of 94.4% and specificity of 100%. The limit of detection for SCITAV was established to be 6% VAF by diluting tumor DNA into matched normal DNA and additional precision replicates at this VAF resulted in a 96% recovery of calls across all replicates. We have shown that the majority of immunoglobulin translocations are detectable with WGS, but comprehensive detection requires optimized approaches to overcome mapping quality issues existing within immunoglobulin loci. These gold standards and particularly the difficult to detect events can be used for the establishment of clinical sequencing assays for multiple myeloma, which will provide more uniform and complete risk assessments and hold the promise of impacting patient care by identifying novel therapeutic options or predicting if specific immunotherapies will be effective.