Cell-Free DNA from Blood or Bone Marrow as Alternatives to Bone Marrow Cells for Molecular Diagnostics and Monitoring of Multiple Myeloma

Introduction: Molecular profiling of bone marrow (BM) tumor cells in multiple myeloma (MM) can be limited by low tumor cell content for analysis by multiple assays. We therefore evaluated the concordance of molecular profiles of cancer cells enriched from BM versus cell-free DNA (cfDNA) derived from peripheral blood (PB) and BM plasma collected at diagnosis. Use of cfDNA has the potential to enable the prioritization of scarce cancer cells for single-cell RNA-sequencing or other applications that require intact cells.

Methods: The MM Molecular Monitoring (M4) Study is a prospective cohort study of newly diagnosed MM patients (pts) receiving standard-of-care treatment at 12 sites across Canada. From 27 pts in this study, we analyzed diagnostic BM-derived CD138+ enriched plasma cell (PC) DNA (n=27, 15-20 mL BM per pt), BM plasma cfDNA (n=26 same tube as BM sampling), PB cfDNA (n=27, 15-20 mL), and PB buffy coat DNA (n=27, 4-6mL) using a combination of a targeted next generation sequencing (NGS) panel (10,000X coverage) and shallow (0.1-1X) whole-genome sequencing (sWGS). PB and BM samples collected in STRECK and EDTA tubes, respectively incurred overnight shipping to a central lab for initial processing. The targeted panel includes exons from 37 genes of potential clinical relevance, all immunoglobulin (Ig) loci, and 5 MM-specific translocation hotspots. CD138+ PCs were profiled by clinical FISH for recurrent MM translocations and copy number variations (CNVs).

Results: All PB and BM cfDNA samples yielded sufficient DNA quantity for analyses (>20ng), yet only 19/26 BM cfDNA samples passed quality control and were sequenced. Various strategies such as shearing, dilution, and the use of heparinase were not effective in salvaging the failed BM cfDNA. However, the high success rate in a separate local study (24/25 BM cfDNA samples) suggests shipping in non-cell stabilizing tubes may have contributed to failures.

The average DNA yield for PB and BM cfDNA samples was 1408ng (sd = 2752ng) and 2462ng (sd = 4301ng), respectively. We estimated tumor fraction (TF) by examining the proportion of genome altered by CNVs using ichorCNA in samples with sWGS. 14/24 (58%) of PB cfDNA samples and 10/18 (56%) of BM cfDNA samples had a high TF (>5%). TF in PB and BM cfDNA were significantly correlated (rho = 0.85, p < 0.001, n = 18 pairs), but were consistently higher in PB cfDNA (p = 0.03, n = 18 pairs).

We first compared mutations between BM-derived PCs and cfDNA. We found 12 distinct somatic mutations (variant allele fraction >1%) that were previously reported to have a pathogenic effect in 11/27 (41%) PC samples. This consisted of 10 single nucleotide variants (SNVs) and 2 deletions across NRAS, BRAF, KRAS, CYLD and PRDM1. We found 9/12 mutations in matched PB cfDNA samples [sensitivity (SN)=75%] and 4/8 mutations in matched BM cfDNA (SN = 50%).

We next searched for FISH-validated chromosome rearrangements among enriched BM PCs. From these, we correctly detected 5/8 (63%) fusions using the targeted panel that were called out by clinical FISH. These were further detected in matched PB cfDNA (4/5) and BM cfDNA (2/4) samples. Recurrent MM CNVs (amplification at chromosome 1q or deletions at chromosomes 1p, 13p or 17p) were identified in 18/27 (67%) PC samples by sWGS. These CNVs were successfully detected in 8/9 matched PB cfDNA samples with high TF (by sWGS) and in 2/7 samples with low TF. In BM cfDNA, CNVs were successfully detected in 7/8 high TF samples and 1/6 low TF samples.

Lastly, we evaluated BCR Ig rearrangements. We detected a clonal Ig rearrangement in 26/27 (96%) BM-derived PC samples. These clones were detected in 25/26 (96%) matched PC to PB cfDNA pairs and in 16/18 (89%) PC to BM cfDNA pairs. The proportion of fragments at Ig loci which corresponded to productive Ig rearrangements was significantly correlated with TF in PB cfDNA (rho = 0.62, p = 0.002, n = 23) but not in BM cfDNA (rho = 0.37, p = 0.14, n = 17).

Conclusions: PB cfDNA yielded 78% of the aberrations detected in BM-derived PCs, whereas BM cfDNA yielded 62%. The lower rate of detection in BM cfDNA may reflect DNA contamination from dying white blood cells as a result of shipping BM samples in non-cell stabilizing preservative tubes. BM and PB cfDNA contain highly concordant CNV calls to PCs when cfDNA TF estimates were >5%. Tumor Ig rearrangements were more readily detected in cfDNA compared to other sources of variation and may provide a less-invasive alternative for disease monitoring.