Paper Details

ABSTRACT

Introduction

Minimal residual disease (MRD) in multiple myeloma (MM) is typically assessed using single-site bone marrow (BM) biopsies. This method fails to capture spatial heterogeneity and is inconducive to serial sampling. In response, whole-genome sequencing (WGS) of cell-free DNA (cfDNA) from peripheral blood is emerging as a systemic, spatially sensitive and minimally invasive method of MRD assessment. Previously, Illumina 40x WGS of cfDNA has been shown to deliver limits of detection (LOD) in the 10^-4 range in plasma cell neoplasms, when tracking a tumor-informed patient-specific catalogue of mutations (Baker et al., HemaSphere. 2026). While plasma MRD adds resolution to clinical trajectories, such as identifying impending progression even when BM MRD is negative (MRD-neg), deeper LODs are required for plasma-based MRD monitoring to be a competitive alternative to current invasive methods, especially in an era of profoundly efficacious T-cell redirecting therapies. Here, we utilized Ultima Genomics paired-plus-minus sequencing (ppmSeq), a duplex sequencing approach that utilizes emulsion PCR to retain both strands in the same physical space, to achieve higher duplex recovery and deeper LOD. We hypothesized that ppmSeq would improve the accuracy and depth of cfDNA-based MRD assessment, capturing systemic-level disease that more wholly reflects clinical outcome.

Methods

We queried plasma-based MRD in a set of 10 samples from 8 patients with relapsed MM receiving bispecific antibody treatment with baseline Illumina 80x CD138+ BM WGS (40x germline match). Baseline BM high confidence single nucleotide variants (SNV; i.e., ≥2 callers) were used to generate patient-specific variant catalogues for longitudinal tracking. DNA was extracted from 1-2 mL of plasma (Streck tubes). All samples were collected at timepoints corresponding to MRD assessments done in BM by either flow cytometry (FC) or Adaptive ClonoSEQ. WGS of cfDNA was then performed with ppmSeq at 100x target coverage. LOD was estimated using a binomial distribution based on the total somatic loci in each patient catalogue and the median coverage in the sample. Using a background error rate, the minimum number of mutant reads required to be 99% confident that a positive call is not due to noise was ascertained (limit of blank). The LOD was then defined for each sample as the lowest simulated tumor fraction that yielded a 95% probability of observing the required supporting reads.

Results

For the relapsed MM cohort, the tumor mutational burden ranged from 6,080-16,868. The median ppmSeq coverage was 139.5 (range 56-191). The median estimated LOD was 8.52 x 10^-6 (range 4.77 x 10^-6 – 1.35 x 10^-5) with a median number of 2 tumor-supporting reads. 4 samples were MRD-pos with a median tumor fraction of 2.8 x 10^-4 (range 3.9 x 10^-5 – 3.7 x 10^-2). Overall, cfDNA-ppmSeq demonstrated 80% concordance with BM MRD assays. 2/10 and 4/10 samples were MRD-pos by BM or cfDNA-ppmSeq, respectively. Plasma ppmSeq recapitulated both BM-positive results but detected disease in two cases with negative BM. Of these two discrepant cases, one was MRD-neg in BM (10^-6; Adaptive) but had multifocal lytic bone disease including persistent expansile lesions post-treatment and the second was MRD-neg in BM (10^-5; FC), but suffered an aggressive extramedullary relapse a year after plasma MRD-positivity. These cases highlight the spatial and temporal advantages of cfDNA-based assays over BM.

Conclusions

As the incidence of extramedullary disease (and low marrow infiltration states) rise in the era of T-cell redirection relapses, MRD detection methods must evolve to capture systemic, rather than site-restricted disease. These data support the utility of plasma cfDNA as a spatially sensitive, minimally invasive method of MRD assessment. Cohort expansion is in progress and will be presented at the meeting.