Capture Rate of V(D)J Sequencing for Minimal Residual Disease Detection in Multiple Myeloma

Survival has improved significantly in multiple myeloma with the introduction of more efficacious treatments leading to deeper responses. Response assessment using minimal residual disease (MRD) negativity has been shown to be a strong predictor for outcome in multiple myeloma (1–4). Currently, MRD assessment is performed in the majority of clinical trials in multiple myeloma and is increasingly being incorporated into routine clinical care (5–7). MRD can be assessed using several different techniques, where multicolor flow cytometry and next-generation sequencing (NGS) are the most commonly used (6–9).

The NGS method takes advantage of the unique V(D)J sequences created within the immunoglobulin heavy and light chain loci during B-cell development (10–15). These V(D)J sequences can be used to determine clonality and can be tracked over time to determine the size of the malignant clone (11, 12). One of the known challenges with NGS for V(D)J rearrangements is the vast diversity of possible V(D)J sequences, resulting in a need for a multiplex PCR and a sophisticated computational approach (15). In addition, somatic hypermutation may affect the annealing of primers and decrease the capture rate. The NGS MRD assay developed by Adaptive, the research version of clonoSEQ, targets all possible V(D)J combinations in one reaction and has been used in several recent large randomized trials in multiple myeloma (4, 12, 16–18). The reported ∼80% capture rate of the first version of the Sequenta/Adaptive 1.3 assay limited the ability to track MRD status post therapy. The assay has recently been updated to increase resilience to somatic hypermutation and include inline controls to assess the quality and quantity of genomic DNA within the sample. As there is limited data on predictive factors for V(D)J clonotype capture success, we were motivated to assess V(D)J clonotype capture and tracking over time in the clinical setting using the NGS MRD assay.

Bone marrow samples were collected prospectively from patients with multiple myeloma seen at Memorial Sloan Kettering Cancer Center (MSKCC) between the years 2010 and 2017. A total of 159 samples from patients with newly diagnosed multiple myeloma (n = 101) and relapse/refractory multiple myeloma (n = 58) were identified and included in the study (Table 1). In addition, we identified 31 posttreatment and follow-up samples from 14 patients who had a complete response or very good partial response at the time of follow-up sampling.

After bone marrow collection, patient samples were sorted for mononuclear cells using Ficoll density gradient separation. A subset of samples with high levels of mononuclear cells was sorted for CD138⁺ plasma cells using a MACSR Separator (n = 6) per MSKCC's standard operating procedure. Samples underwent DNA extraction using the Qiagen dual DNA/RNA extraction kit.

The study was approved by the MSKCC Institutional Review Board committee and all patients had consented to MSKCC's institutional tissue acquisition protocol (06–107, 09–141, 14–276, and 15–017). Written informed consent was obtained from each subject for the tissue acquisition protocol. The studies were conducted in accordance with the Declaration of Helsinki.

After DNA extraction, 500 ng genomic DNA from each sample was sequenced using the NGS MRD assay, the research application of clonoSEQ, developed by Adaptive Biotechnologies (Seattle, Washington). All samples were analyzed in a blinded manner without access to clinical or outcome information. Locus-specific primer sets were used to amplify DNA encompassing the CDR3 region of the IGH complete (IGH), IGH incomplete (IGH_D), and light chain IGK and IGL loci (8, 15). Through Illumina NGS, the V(D)J clonotypes as well as the frequencies of the captured V(D)J clonotypes were obtained as previously described (12, 16). Trackable sequences were identified based on the following four criteria: (i) the sequence needed to represent >3% of “like” sequences, i.e., for the IGK locus the sequence must represent >3% of all IGK sequences; (ii) the sequence must represent >0.2% of the total nucleated cell population; (iii) the sequence must reside within at least 40 cells; and (iv) the sequence must be discontinuously distributed defined as there not being >5 sequences within the next log decile of sequence frequencies. The identified clonotype sequences underwent quality control per the assay protocol and all unique clonotypes that passed quality control were used for MRD tracking.

The baseline samples were sequenced for genomic events using our multiple myeloma–specific targeted NGS panel (called myTYPE; ref. 19). This is a custom capture assay, which includes the IGH locus, where the vast majority of chromosome 14 translocations occur as well as genome wide single nucleotide polymorphisms (SNP), one per three Mb, to assess for hyperdiploidy and other copy-number alterations (20–22). In addition, the panel includes 120 genes (exons and selected introns) that are recurrently mutated, amplified, or deleted in multiple myeloma, e.g., oncogenes, tumor suppressor genes, and genes in signaling pathways or actionable drug targets (23–26).

For each sample, 100 to 200 ng of gDNA was used for library construction. Exon capture by hybridization (Nimblegen SeqCap, Madison, Wisconsin) was done on barcoded sequence libraries. Samples were sequenced to a target depth of 600x using Illumina HiSeq 4000 to generated paired-end 100 base pair reads. The bioinformatic algorithms CaVEMan, Strelka, muTect 2.0, and pindel were used to call somatic mutations (19, 27–30). CNVkit, BRASS, and Delly were used to call copy-number alterations and structural variations (31–33). Sixteen unmatched normal bone marrow samples were used for comparison. All variant calling underwent manual curation and were annotated as oncogenic, likely oncogenic, unknown, artifact, or SNP based on previously reported criteria (19). Five of the 159 samples failed myTYPE sequencing. Samples were considered myTYPE-positive if one or more high confident genomic aberration was detected through targeted sequencing with myTYPE.

Univariate analyses were carried out for V(D)J clonotype detection in relation to bone marrow plasma cell percentage in bone marrow aspirates and biopsies, M-spike level, immunoglobulin isotype (IgG or IgA), light chain type (kappa or lambda), age, and sex. We also analyzed the clonotype capture rate in relation to presence of genomic aberrations using the targeted myTYPE panel. A multivariate analysis was performed including the significant predictors from the univariate analysis, bone marrow plasma cell percentage in bone marrow aspirates and biopsies, M-spike level, and myTYPE positivity. Univariate and multivariate analyses were carried out using logistic regression and results are presented as ORs with 95% confidence intervals (CI). All statistical analyses were performed in R version 3.4.3.

The data generated in this study are publicly available in the European Variation Archive (PRJEB31370).

The V(D)J clonotype capture rate was 93% (N = 70/75) in baseline multiple myeloma samples that were myTYPE-positive, indicating that these samples had sufficient amounts of detectable tumor DNA. The overall clonotype capture rate was 79% (N = 129/159) when all samples, regardless of DNA content, were analyzed (Fig. 1). Correspondingly, the V(D)J clonotype capture rate was higher in samples with a higher tumor cell content measured indirectly through estimates of bone marrow plasma cell infiltration on aspirate and biopsy morphology. There was no significant difference in V(D)J clonotype capture rate between samples collected at diagnosis or relapse. In univariate analysis, samples with a genomic aberration detected by myTYPE had a significant 7-fold higher clonotype detection rate (OR, 7.3; 95% CI, 2.8–22.6). The ORs for clonotype detection were 1.7 (95% CI, 1.3–2.5) and 1.5 (95% CI, 1.2–1.8) for every increase in 10% bone marrow plasma cells on aspirate and biopsy, respectively. For every 1 g/dL increase in M-spike, the OR of clonotype detection was 1.6 (95% CI, 1.2–2.2). The relationship between OR of V(D)J clonotype capture, myTYPE positivity, and plasma cell percentage on bone marrow aspirates is presented in Fig. 2. Conversely, age, gender, type of immunoglobulin heavy chain (IgG or IgA), or light chain type (kappa or lambda) had no significant effect on the V(D)J clonotype detection rate (Table 2). The different genomic alterations and V(D)J capture is shown in Fig. 3.

In multivariate analysis, myTYPE positivity was an independent predictor of clonotype detection (Fisher exact test P < 0.001), with an OR for clonotype capture success of 4.4 (95% CI, 1.4–16.9; P = 0.016). The ORs were not significant for an increase in 10% bone marrow plasma cells on aspirate (OR, 1.4; 95% CI, 1.01–2.1; P = 0.062) or an increase of 1 g/dL increase in M-spike (OR, 1.3; 95% CI, 0.9–2.1; P = 0.15; Table 2).

A median of three trackable clonotypes (range 1–7) per patient was identified. IGK was the locus where the majority of dominant clonotype sequences were found, followed by IGH, IGH_D, and IGL (Table 3). Also the majority of patients with lambda light chain disease had a dominant V(D)J clonotype in the IGK locus.

Thirty-one posttreatment samples from 14 patients, who were in a complete response or very good partial response and had successful baseline V(D)J clonotype capture, were identified and included in the follow-up cohort. There were between 1 and 6 available follow-up samples per patient and all 31 posttreatment samples passed quality control for V(D)J sequencing. Of the 31 included samples, 19 were MRD-positive and 12 were MRD-negative by the NGS MRD assay. Through clinical assessment using multicolor flow cytometry using a different vial (earlier pull), 9 were MRD-negative, 20 were MRD-positive, and clinical flow cytometry was missing in two follow-up samples. Four samples were negative by flow cytometry and positive by the MRD NGS assay while one sample was positive by NGS and negative by flow cytometry.

In all patients, the clonotypes identified at baseline were present in all MRD-positive follow-up samples. Figure 4 illustrates tracking of clonotypes in four patients over time. Patient 1 had varying disease burden over time with several samples obtained at a time of relapse disease (Fig. 4A), Patients 2 and 3 had a decrease in disease burden over time (Fig. 4B and C), and Patient 4 was MRD-negative in the follow-up samples (Fig. 4D). The last sample from Patient 4, in whom three independent dominant clonotypes were tracked, had a higher count of the IGK clonotype at the last time point. In this case, this particular IGK clonotype has a less unique V(D)J rearrangement, meaning that the probability of it being independently recreated within a nonmalignant cell is high and that the results therefore would fall below the limit of confident identification detection.

MRD assessment and MRD negativity are becoming increasingly important in the management of multiple myeloma. Substantial evidence exists showing that MRD negativity after induction treatment translates into a longer progression-free survival (3, 4). Reliable and sensitive methods for MRD detection are therefore essential. In our study, the V(D)J clonotype capture rate using Adaptive's NGS MRD assay was high, at 93% in baseline multiple myeloma samples with sufficient tumor DNA. Importantly, MRD assessment using NGS was feasible in all samples where a baseline V(D)J sequence was detected and the V(D)J clonotypes remained stable during follow-up.

In order to use NGS testing for MRD status, the unique clonotype(s) must be assessed pretreatment to allow for MRD tracking (11). We found that baseline V(D)J capture was higher in samples with confirmed presence of tumor DNA, i.e., samples where at least one somatic genomic aberration was detected using the targeted myTYPE panel. Detectable tumor DNA using myTYPE was the strongest predictive factor with a 4.4-fold increase in the OR of V(D)J clonotype capture success. While the expectation would be that all samples would have at least one genomic aberration, we only found this in 49% (N = 75) of the samples. In the remaining samples, either no genomic aberrations were detected, or the variants fell below the level where we could confidently determine real variants from artifacts. Correspondingly, we found a high baseline capture in samples with a high tumor burden, i.e., samples with a high bone marrow aspirate plasma cell infiltration or an associated high M-spike level. Additional clinical variables such as patient age, sex, and type of multiple myeloma, e.g., IgG versus IgA or kappa versus lambda light chain expression, did not affect V(D)J clonotype capture even though the study was sufficiently powered to confidently detect such potential differences. A possible reason for the lower capture rate when all samples were included is that this study was based on samples collected on a research biospecimen protocol. At collection, the study samples were aspirated after the clinical samples, thus from the third or fourth aspirate pull, which increases the risk of hemodilution. In addition, the majority of samples were not CD138⁺ sorted. When used in the clinical setting, the samples drawn in the first or second aspirate pull will have a higher plasma cell content, and the chance of V(D)J capture success will likely be even higher (4).

The NGS MRD assay identifies unique V(D)J clonotypes within the CDR3 region of the immunoglobulin genes and all unique clonotypes are used for tracking of MRD. DNA is a stable marker and NGS provides an objective quantification method for disease burden with a sensitivity of 10^–6 (4, 11, 34, 35). The assay includes multiplex primers directed towards all possible immunoglobulin rearrangements so that all rearrangements present in the plasma cell clone are amplified and sequenced. Using the IGH, IGH_D, IGK, and IGL genes, up to eight unique V(D)J clonotypes can be identified in a B-cell or plasma cell clone. The majority of the dominant V(D)J clonotype sequences were found in the IGK locus even in the patients who had lambda light chain multiple myeloma. This likely reflects the process of light chain recombination during B-cell development, which starts by default with recombination of the IGK locus and continues to the IGL locus if the IGK rearrangement does not result in a functional kappa light chain (36).

All trackable V(D)J clonotypes identified at baseline were present in MRD-positive follow-up samples. We found identical clonotype sequences at baseline and follow-up in MRD-positive samples supporting previous findings that the V(D)J rearrangement remains stable in post-germinal center mature B-cell malignancies such as multiple myeloma, regardless of further subclone development in terms of somatic mutations and structural variations (37). Furthermore, the NGS MRD assay has been updated to overcome sequencing failures caused by somatic hypermutation, something that has been of concern in previous V(D)J assays. Somatic hypermutation is present to a high degree in post-germinal center B cells and can interfere with the annealing of PCR primers. Although we cannot fully rule out a possible effect of somatic hypermutation, there was a high baseline detection rate in samples with detectable tumor DNA as well as in MRD-positive follow-up samples, suggesting that suboptimal detection due to somatic hypermutation was not of major concern in this study. Our findings underscore the stability of V(D)J region over time and the utility of the V(D)J for disease tracking (11).

The majority of studies on V(D)J sequencing and MRD tracking have used bone marrow samples for analysis. Although blood based assays are more convenient, circulating myeloma cells in the periphery have been shown to be 10- to 100-fold lower than in the marrow and the clinical relevance of monitoring in blood is an active area of investigation (38–41). Recent efforts to assess disease burden and MRD in blood through liquid biopsies for circulating tumor DNA (ctDNA) have been promising in several malignancies including lymphoma and solid tumors (42–44). Similar studies on ctDNA for V(D)J sequencing in multiple myeloma have, however, shown a poor correlation with findings in bone marrow and blood (45, 46). Interestingly, emerging highly sensitive protein-based techniques such as MALDI-TOF and Q-TOF have shown promising results for tracking of low-level disease in blood (47). Nevertheless, until these techniques are fully validated, bone marrow tests remain standard for MRD assessment in multiple myeloma.

In summary, complete response with MRD negativity, currently at the sensitivity level of 10^–6 (1 myeloma cell in 1,000,000 cells), is the goal of current multiple myeloma treatments and is increasingly being used to inform clinical decision-making (4, 48). Achieving MRD negativity is an important prognostic marker and is associated with a longer progression-free and overall survival and is incorporated as an endpoint in many clinical trials (3, 4, 49, 50). Our results confirm the high baseline V(D)J clonotype capture in baseline samples using Adaptive's NGS MRD assay and, as expected, presence of sufficient tumor DNA was the primary factor in determining sensitivity of the assay. MRD tracking was feasible in all samples with a detected baseline V(D)J clonotype, and in our study, the clonotypes remained stable with no further rearrangement over time regardless of disease burden. Thus, NGS for V(D)J rearrangements offers a reliable and sensitive method for baseline clonality detection and MRD tracking in multiple myeloma in the clinical setting.

M. Hultcrantz reports grants from Swedish Research Council, MMRF, Swedish Cancer Society, MSKCC Core Grant, and Swedish Blood Cancer Foundation during the conduct of the study, as well as other support from Curio Science and Intellisphere, LLC outside the submitted work. A. Jacob reports to be an employee and shareholder of Adaptive Biotechnologies. N. Korde reports other support from MedImmune, Amgen, and Janssen outside the submitted work. S. Mailankody reports other support from Takeda Oncology, Janssen Oncology, Bristol Myers Squibb, and Allogene Therapeutics, as well as personal fees from PleXus Education and Physician Education Resource outside the submitted work. A.M. Lesokhin reports other support from Bristol Myers Squibb, Genmab, Boehringer Ingleheim, Amgen, Genentech, Trillium, Janssen, and Serametrix, Inc. outside the submitted work. H. Hassoun reports grants from Janssen, as well as personal fees from Novartis outside the submitted work. E.L. Smith reports personal fees from BMS, Fate Therapeutics, Eureka Therapeutics, Chimeric Therapeutics, and Precision Biosciences outside the submitted work; in addition, E.L. Smith has a patent for CAR T cells for MM licensed and with royalties paid from BMS. O.B. Lahoud reports personal fees from MorphoSys outside the submitted work. G.L. Shah reports other support from Janssen and Amgen outside the submitted work. M. Scordo reports personal fees and other support from Angiocrine Bioscience, Inc. and Omeros Corporation, as well as personal fees from McKinsey & Company, Kite - A Gilead Company, and i3-Health outside the submitted work. S. Giralt reports personal fees and other support from Celgene, Janssen, BMS, Sanofi, Actinium, Amgen, Pfizer, GSK, and Jazz, as well as other support from Omeros outside the submitted work. O. Landgren reports grants and personal fees from Amgen, Celgene, Janssen, and Takeda; personal fees from Karyopharm, Adaptive Biotech, Binding Site, Bristol Myers Squibb, Cellectis, Oncopeptides, and Pfizer; grants from Multiple Myeloma Research Foundation, Perelman Family Foundation, National Cancer Institute, and FDA; other support from Theadex, Merck, and Janssen outside the submitted work. No disclosures were reported by the other support authors.

M. Hultcrantz: Conceptualization, resources, data curation, formal analysis, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. E.H. Rustad: Conceptualization, data curation, formal analysis, validation, methodology, writing–review and editing. V. Yellanpantula: Data curation, formal analysis, methodology, writing–review and editing. A. Jacob: Conceptualization, data curation, formal analysis, writing–review and editing. T. Akhlaghi: Data curation, formal analysis, methodology, writing–review and editing. N. Korde: Data curation, writing–review and editing. S. Mailankody: Data curation, writing–review and editing. A.M. Lesokhin: Data curation, writing–review and editing. H. Hassoun: Data curation, writing–review and editing. E.L. Smith: Data curation, writing–review and editing. O.B. Lahoud: Data curation, writing–review and editing. H.J. Landau: Data curation, writing–review and editing. G.L. Shah: Data curation, writing–review and editing. M. Scordo: Data curation, writing–review and editing. D.J. Chung: Data curation, writing–review and editing. S. Giralt: Data curation, writing–review and editing. E. Papaemmanuil: Conceptualization, data curation, validation, investigation, methodology, writing–review and editing. O. Landgren: Conceptualization, resources, data curation, supervision, validation, investigation, writing–review and editing.

We acknowledge the use of the MSKCC Integrated Genomics Operation Core, funded by the NCI Cancer Center Support Grant (CCSG, P30 CA08748). We thank the Perelman Family Foundation, the Multiple Myeloma Research Foundation, Swedish Research Council, and the Swedish Blood Cancer Foundation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.