Whole genome sequencing of cell-free DNA for assessment of minimal residual disease in high-risk smoldering multiple myeloma

Abstract Introduction In multiple myeloma (MM), assessment of minimal residual disease (MRD) is limited by a need for serial bone marrow (marrow) biopsies, adding discomfort and logistical complexity. Furthermore, single-site assessment of marrow MRD may be limited in settings of extramedullary/patchy disease. There is a need to develop peripheral assessment of MRD. Target-capture panels, whole-exome, and low-pass whole genome sequencing (WGS) to identify somatic variants in plasma cell-free (cf)-DNA generally fail to achieve adequate detection limits in low tumor fraction (TF) settings (i.e., MRD). Tumor-informed approaches, where baseline marrow tumor is profiled and somatic variants are tracked in cfDNA, have thus far achieved the lowest limits of detection (LOD). We therefore tested the MRD performance of tumor-informed WGS of cfDNA in a low tumor-burden setting using longitudinal samples from an interventional trial for high-risk smoldering MM (NCT01572480; carfilzomib, lenalidomide, and dexamethasone). Methods MRDetect is an ultra-sensitive tumor-informed detection approach for MRD. First, 80X WGS of baseline marrow tumor samples (with 40x germline match) was performed on CD138+ plasma cells to obtain high-confidence ground-truth single nucleotide variant (SNV) catalogues. Serial blood plasma (Streck tubes) from timepoints corresponding to pre-treatment baseline, marrow MRD assessment, and progression of disease (PD) were then subjected to 40x WGS and cfDNA tumor-supporting reads matching SNV from marrow WGS were used to estimate TF in blood. Patient-specific LODs were computed by comparing detection rates in plasma to control SNV compendium detection rates derived from tumor-supporting reads in unrelated plasma samples (i.e., contamination/artifact). Results were compared to flow MRD in marrow, which was performed at end of combination therapy and annually, thereafter (EuroFlow; LOD 10-5). Results We assayed 71 plasma samples and 25 baseline marrow samples from 25 patients (n=96 WGS; median follow-up 66 months). The median TF of baseline plasma samples was 6x10-4 (range; 1x10-4-9x10-3). The median LOD across patients was 1.2x10-4 (range 2x10-4-9x10-5). Tumors with genomic complexity (e.g., chromothripsis and APOBEC activity) had higher baseline TF (p = 0.007). In line with this, a higher TF at baseline was associated with eventual PD (n=7, p<0.001). In longitudinal analysis, cfDNA WGS recapitulated MRD+ for 14/15 (93%) marrow MRD+ assays. Representing a limitation in cases with low mutational burden, one t(11;14) case with the lowest SNV count (~2000) could not be detected in plasma despite marrow MRD-positivity. However, 15/21 (71%) marrow MRD- timepoints were seen to instead be MRD+ in plasma. Most discrepancies occurred in cases where marrow MRD+ conversion or PD later occurred, or in cases where flanking (sequential) MRD marrow assessments were MRD+ (10/15, 75%), suggesting increased resolution of plasma over marrow assay. To support these discrepancies as true MRD+ in cfDNA, we demonstrated that in these cases: 1) the proportion of clonal variants in marrow tumor was similar to those supported in plasma; 2) APOBEC mutational signatures were present in cfDNA tumor-supporting variants suggesting MM as the source rather than a plasma cell/B-cell common ancestor; and 3) fragmentomic analysis revealed that tumor-supporting reads (i.e., fragments) were shorter (density peaks at 145-150 base pairs) than reference (i.e., normal; 167 base pairs), supporting malignant origin. Additional sequential samples in process will be presented at the meeting. For cases with marrow MRD+ conversion or PD, plasma collected three and six months prior detected early MRD resurgence in 4/6 (75%), representing a logistical advantage. Furthermore, incremental growth of TF could be tracked over serial measurements and foreshadowed PD. Finally, 5/7 patients with PD had MRD+ in plasma at end of combination therapy, but only 2 were marrow MRD+ at the same time point. No concordant plasma/marrow MRD- patients experienced PD (HR; inf, p = 0.035). Discussion Compared to marrow flow, cfDNA WGS consistently identified MRD. Despite the deeper LOD of localized marrow flow, cfDNA WGS detected MRD+ when marrow flow did not, representing an advantage of the peripheral approach. Furthermore, cfDNA WGS serially tracked TF and was prognostic. Finally, LOD is expected to be deeper in newly diagnosed and relapsed MM, where SNV burden for disease tracking is higher.