Identification of persister cells and evolution of tumor clones during immunotherapy and at minimal residual disease timepoints decoded by single-cell whole-genome sequencing in multiple myeloma

Abstract Background: Mutations in cancer driver genes and target antigens are associated with poor prognosis in multiple myeloma (MM), but their role in minimal residual disease (MRD) and in driving progression under immunotherapy remains unclear. To detect the cells responsible for relapse, we used longitudinal single-cell whole-genome sequencing (scWGS) to isolate and characterize tumor cells - including rare MRD cells - from patients treated with quadruplet therapy (Daratumumab, Lenalidomide, Bortezomib, Dexamethasone (DRVd)), anti-BCMA-directed T-cell-redirecting therapy (Teclistamab), or anti-BCMA CAR-T cell therapy (Cilta-cel). In addition, we present a method to assess MRD cells using peripheral blood (PB) in conjunction with bone marrow (BM) biopsies. Methods Tumor plasma cells from 11 patients with high-risk smoldering MM (HR-SMM, n=9) or relapsed/refractory MM (RRMM, n=2) were isolated using 9-color fluorescence-activated cell sorting (FACS) after magnetic bead enrichment for CD138 on BM and PB samples. For scWGS, ResolveDNA (Bioskryb Genomics) was used to amplify DNA prior to library preparation. All DNA libraries were prepared using Ultra II FS (New England Biolabs), and sequencing was performed at a depth of 60X (WGS) or 10X per single cell (scWGS). Subclonal dynamics were analyzed using the PhylogicNDT suite, and phylogenetic trees were inferred with the Sequoia method. Results We performed longitudinal WGS at baseline, MRD timepoints, and/or relapse in patients receiving DRVd (n=6), Teclistamab (n=4), or Cilta-cel (n=1). In patients treated with Daratumumab, CD38 expression was low in tumor samples collected during treatment (n=3) but high in those collected after the end of treatment (n=3). In one of the 3 on-treatment samples, a heterozygous CD38 gene deletion was identified in 3 out of 3 single tumor cells, from 2 independent clonal branches. No genomic alterations at the CD38 locus were detected in the remaining samples, suggesting that they could be rechallenged with CD38-targeted therapies irrespective of clinical response and clonal selection. Notably, the CD38 deletion was detected at an MRD timepoint and was later confirmed by bulk sequencing at disease progression. In contrast, scWGS of another hyperdiploid patient who remained MRD+ revealed unexpected subclonal chromosomal heterogeneity (including within trisomies 9 and 15) across all 22 single cells analyzed at the baseline timepoint - findings that were not detectable by bulk sequencing. After 12 cycles of therapy, the patient achieved a very good partial response (VGPR), and we profiled 12 tumor cells as part of the planned MRD assessment, which were CD38 low/negative. The tumor cells branched out of the clone found at baseline and harbored additional copy-number abnormalities (del(16q) and del(6q)). Across the five patients treated with Teclistamab (n=4) or Cilta-cel (n=1), BCMA expression remained high throughout treatment despite 4 of them not responding or relapsing after treatment. The fifth patient received Teclistamab, we profiled 4 single cells collected after 6 months of therapy (MRD negative 10⁻⁵, positive 10⁻⁶). We identified a new heterozygous loss of chromosome 16 (encompassing BCMA) in the single BCMA+ tumor plasma cell, while 3 other candidate cells were found to be normal plasma cells (false positives by FACS). The patient continued therapy and achieved CR and MRD negativity (10⁻⁶), suggesting that even cells with a single copy of BCMA remaining can still be eliminated post-therapy. Lastly, a patient receiving Cilta-cel achieved CR with MRD positivity (10⁻⁶). We performed WGS of 7 and 30 BCMA+ single circulating tumor cells at 10 and 18 months post-infusion, respectively, prior to clinical progression. Our analysis revealed genomically mature, persistent tumor cells lacking BCMA mutations, branching from the baseline tumor and acquiring additional mutations. Relapsed disease at 18 months originated from several of the clones already present at 10 months. Conclusion Using scWGS, we profiled persistent tumor cells during immunotherapy with minimal disease burden available. Subclonal analysis suggested that these persister cells primarily branched from a genomically mature, late clone - some acquiring additional mutations targeting the antigen as early as 6 months into therapy - rather than from an ancestral population.