PDF
Citation
Print
INTRODUCTION
Multiple myeloma (MM) is a malignancy of differentiated, mature lymphoid-derived plasma cells (PCs) characterized by significant genetic heterogeneity, mutational burden, and high-risk genomic lesions, including t(4;14) 14q32, del(17p), and del(13q) [1, 2]. Almost all MMs evolve from a premalignant plasma cell clonal disorder termed monoclonal gammopathy of undetermined significance (MGUS); however, this precursor condition remains undetected in several clinical cases [3]. Smoldering MM (SMM) is an indolent stage between MGUS and MM, with a 10-fold higher risk of progression to MM compared with that of MGUS [4].
Genetic alterations are crucial to MM progression. Next-generation sequencing (NGS) has helped identify recurrent mutations in cancer-related genes, including RAS and TP53, with clonal heterogeneity [5, 6]. Specifically, mutations in the RAS/RAF/MAPK pathway genes (BRAF, NRAS, or KRAS) occur more frequently in MM than in SMM [7]. However, molecular profiles of MM genomes derived from single-site bone marrow (BM) biopsies represent only a small fraction of the BM compartment. Large-gauge needles used for these biopsies cause patient discomfort and complications, such as bleeding and infection. Additionally, monitoring using positron emission tomography/computed tomography (PET/CT) imaging may also fail to capture the intrinsic spatial and temporal heterogeneities of MM [8]. The invasive nature of BM biopsies and the radiation exposure associated with PET/CT hinder the frequent application of these procedures during therapy. Therefore, a less invasive technique capable of capturing clonal heterogeneity in clinical settings is required.
Circulating tumor DNA (ctDNA) consists of small DNA fragments released from apoptotic or necrotic tumor cells [9]. Genetic alterations present in tumor tissues can be accurately recapitulated using liquid biopsies targeting ctDNA [10]. ctDNA levels are higher in MM than in other tumors, and high concordance has been observed between mutations in BM aspirates and ctDNA [11, 12]. Recent studies have explored ctDNA analysis for treatment outcomes and tumor load monitoring in MM [13, 14]; however, ctDNA-based liquid biopsy has not yet been extensively validated in clinical settings.
In this study, we evaluated the clinical implications and prognostic relevance of liquid biopsies targeting ctDNA in patients with MGUS, SMM, or newly diagnosed MM (NDMM). Notably, we performed targeted deep sequencing of ctDNA collected from treatment-naive patients, as well as an exploratory analysis using serially collected ctDNA from patients with MM before and after treatment.
MATERIALS AND METHODS
Patients and samples
Patients newly diagnosed as having PC dyscrasia, including MGUS, SMM, and MM, between March 2019 and September 2021 were eligible for this study. We retrospectively identified 118 eligible patients who agreed to participate and provided peripheral blood (PB) and BM samples at the time of the diagnosis of PC dyscrasia at Seoul St. Mary’s Hospital, Korea. Among them,16 patients were excluded because of inadequate QC for targeted NGS analysis. The final cohort included seven patients with MGUS, six with SMM, and 89 with MM (Supplemental Data Fig. S1). We analyzed data in March 2023. This study was approved by the Institutional Review Board of Seoul St. Mary’s Hospital (KC24TISI0257) and conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all patients. Detailed methods on ctDNA extraction, NGS protocol, identification of somatic variants, DNA copy number analysis, subcohort selection, and statistical analysis are described in the Supplemental Data section. The raw FASTQ files are available from the corresponding authors upon reasonable request.
Definition of clinical features
The diagnosis of MGUS, SMM, or MM was established based on the International Myeloma Working Group criteria [15]. Hypercalcemia was defined as a serum calcium level >11.0 mg/dL, anemia as an Hb level <10.0 g/dL, and renal failure as a serum creatinine level >2 mg/dL [15]. Paramedullary myeloma was defined as plasmacytomas originating from the bone in continuity with the BM, whereas extramedullary myeloma referred to soft tissue plasmacytomas or PC infiltration at anatomical sites distant from the BM. All patients with SMM and MM underwent PET/CT screening with [18F]Fluorodeoxyglucose (FDG) to identify bone lytic lesions and paramedullary or extramedullary myeloma. PC leukemia was defined as ≥5% circulating PCs in the PB [16]. Myeloma-extended sites, as described in previous reports [17, 18], included paramedullary myeloma, extramedullary myeloma, and PC leukemia. NDMM staging was performed according to the International Staging System [19]. High-risk cytogenetic abnormalities were defined as the presence of one or more of the following based on FISH results using BM samples collected at diagnosis: del(17p), t(4;14), t(14;16), and 1q gain/amplification [20, 21].
RESULTS
Patient characteristics
Mean ages, sex distributions, and clinical features such as renal insufficiency, bone lytic lesions, and MM subtypes are summarized in Table 1. Kappa chain-type MM was more prevalent than lambda chain-type MM. Myeloma-extended sites, including paramedullary myeloma, extramedullary myeloma, and PC leukemia, were identified in 31.5% of patients with MM.
Genetic alterations in ctDNA samples
We used targeted deep sequencing (Supplemental Data Table S1) to analyze ctDNA samples from 102 patients at the time of diagnosis, including those with MGUS (N=7), SMM (N=6), and MM (N=89). The median ctDNA concentrations for MGUS, SMM, and MM were 10.8 ng/mL (range, 6.5–199.5), 9.6 (range, 1.9–13.1) ng/mL, and 15.3 ng/mL (range, 2.8–523.5), respectively; however, the higher concentration in patients with MM than in those with MGUS or SMM was not statistically significant (Supplemental Data Table S2). The median sequencing depth across the entire target genome was 2,163× (range, 674× –3,045×; Supplemental Data Table S2).
In total, we identified 226 somatic mutations, including 174 single nucleotide variants and 52 indels, across the 102 patients (Supplemental Data Table S3). Non-silent mutations in at least one target gene, such as KRAS, BRAF, DNMT3A, and TP53, were detected in 76.5% of patients (78/102) [5, 22, 23]. Non-silent mutation positivity rates in BM samples were 100% for both MGUS and SMM and 94% for MM. In contrast, rates in ctDNA were 57%, 67%, and 79%, respectively, suggesting an association between ctDNA mutation positivity and disease progression (P=0.041).
The average number of ctDNA non-silent mutations increased with disease progression: 1.0 in MGUS, 1.8 in SMM, and 1.9 in MM. When comparing BM and ctDNA mutation profiles, none of the patients with SMM (0 out of 2) had shared mutations, whereas 25.0% of patients with MGUS (1 out of 4) and 68.9% of patients with MM (42 out of 61) had at least one shared mutation between BM and ctDNA. Among the 102 patients, 11 (10.8%) harbored at least one copy number alteration (CNA). These CNAs were identified in 10 patients with MM and one patient with MGUS (Supplemental Data Table S4). Recurrent copy number gains were observed in MYC (N=3), MCL1 (N=3), NIPA2 (N=3), and STK11 (N=2), whereas other CNAs were observed as singleton events.
Differences in ctDNA profiles according to disease progression
Recurrent genetic alterations were identified in 10 genes across multiple ctDNA samples (>5%; Fig. 1A; Supplemental Data Table S5): KRAS (N=15, 14.7%), NIPA2 (N=13, 12.7%), TP53 (N=11, 10.8%), GNAS (N=11, 10.8%), ZFHX3 (N=9, 8.8%), NRAS (N=9, 8.8%), NOTCH1 (N=8, 7.8%), MLH1 (N=8, 7.8%), BRAF (N=7, 6.9%), and MYC (N=6, 5.9%). Notably, 29 (32.6%) patients harbored at least one RAS/RAF (KRAS, NRAS, and BRAF) activating mutations. Among KRAS mutations, p.G12D/C/S/V (N=8) was the most frequent, followed by p.Q61R/L/H (N=4), p.G13D (N=3), and p.Y64D (N=2) (Fig. 1B). Singleton events accounted for the remaining KRAS mutations. For NRAS, p.Q61R/H/K (N=6) was the most common mutation, followed by p.G13V/R (N=2) and p.G12D (N=1) (Fig. 1C). Among BRAF mutations, p.V600E (N=3) was the most frequent, followed by p.D594G/N (N=2), p.N581I (N=1), and p.K483E (N=1) (Fig. 1D). RAS/RAF-activating mutations were significantly enriched in MM genomes but were not detected in MGUS or SMM genomes (P=0.017).
Recurrent mutations in TP53 were scattered across the entire coding region (Fig. 1E) and were exclusively detected in MM genomes (12.7%; N=13; Fig. 1A). These findings suggest that mutations in KRAS, NRAS, BRAF, and TP53 may contribute to the progression of MM from precursor diseases. In contrast, mutations in GNAS were detected not only in MM (9.0%; N=8) but also in MGUS (14.3%; N=1) and SMM (33.3%; N=2) genomes (Fig. 1A), indicating that GNAS mutations may occur during the early stages of MM progression.
Clinical correlation between ctDNA at diagnosis with MM
In ctDNA from patients with MM, recurrent genetic alterations were detected in 12 genes (occurring in >5% of patients, Supplemental Data Table S5): KRAS (16.9%), NIPA2 (12.4%), TP53 (12.4%), NRAS (10.1%), GNAS (9.0%), ZFHX3 (9.0%), NOTCH1 (7.9%), MLH1 (7.9%), BRAF (7.9%), MYC (5.6%), APC (5.6%), and TET2 (5.6%). We investigated the association between these 12 genes and clinicopathological characteristics, including hypercalcemia, renal insufficiency, anemia, bone lytic lesions, amyloidosis, myeloma-extended sites, neutropenia, and thrombocytopenia. Simple correlation analysis revealed that genetic alterations in five genes were positively related to specific clinicopathological characteristics (Supplemental Data Table S6): hypercalcemia and TET2 (P=0.006), renal insufficiency and NRAS (P=0.012), bone lytic lesions and TP53 (P=0.050), paramedullary myeloma and TP53 (P=0.020), and extramedullary myeloma and NRAS (P=0.007), MYC (P=0.012), or APC (P=0.012). Using Fisher’s exact test, the relationships between hypercalcemia and TET2 (P=0.048), renal insufficiency and NRAS (P= 0.031), paramedullary myeloma and TP53 (P=0.030), and extramedullary myeloma and NRAS (P=0.032) remained significant (Table 2). Multivariate analysis of clinical variables with a univariate P-value <0.05 indicated that none were associated with clinicopathological characteristics (Supplemental Data Table S7).
Prognostic implication of ctDNA
We performed a survival analysis of patients with MM to investigate the prognostic impact of genetic alterations in ctDNA. After excluding five patients who did not receive treatment because of loss to follow-up or early death, the median follow-up duration was 28.4 months [95% confidence interval (CI), 25.7–32.4] from treatment initiation. The 2-yr overall survival (OS) and progression-free survival (PFS) rates were 89.5% (95% CI, 82.8–96.7) and 54.9% (95% CI, 44.9–67.1), respectively.
Univariate analysis helped identify significant associations between poor OS and four genes: APC (P=0.006), NIPA2 (P= 0.009), TET2 (P=0.011), and NRAS (P=0.030) (Fig. 2A and Supplemental Data Table S8). High lactate dehydrogenase levels (P=0.020) and ineligibility for transplantation (P=0.007) were also significantly associated with poor OS in univariate analysis (Supplemental Data Table S8). However, multivariate analysis, including candidate genes and clinical variables that were significant in the univariate analysis, demonstrated that none were independent prognostic factors for OS (Table 3).
Univariate analysis revealed that patients with NRAS (P= 0.035) and TET2 (P<0.001) alterations showed significantly poorer PFS than that of those without NRAS and TET2 alterations (Fig. 2B and Supplemental Data Table S8). Additionally, five clinical variables—transplant ineligibility (P<0.001), older age (P=0.004), high-risk cytogenetic profile (P=0.01), fewer treatment regimens (P=0.026), and lambda chain-type (P= 0.037)—were significantly associated with poorer PFS (Supplemental Data Table S8). Multivariate analysis showed that the presence of TET2 alterations [hazard ratio (HR)=7.11; 95% CI, 1.92–26.24; P=0.003] and three clinical variables [transplant ineligibility (HR=4.001; 95% CI, 2.117–25; P=0.001), lambda chain-type (HR=2.07; 95% CI, 1.1–3.9; P=0.025), and high-risk cytogenetic profile (HR=2.65; 95% CI, 1.14–5.7; P=0.023)] were independent prognostic factors for poor PFS (Table 3).
Monitoring genetic alteration profiles in ctDNA during treatment
We analyzed the genetic alteration profiles of serial ctDNA samples (N=20) from seven patients with MM to evaluate the clinical utility of liquid biopsies during therapy. In two patients whose PFS was >12 months from the first-line treatment (MM510 and MM429), the baseline levels of variant allele frequency (VAF) of detected mutations decreased over 20 months (Fig. 3A). No newly detected mutations were observed during first-line therapy in these patients. They exhibited favorable treatment responses to the second-line therapy, with MM510 achieving a very good partial response (VGPR) and MM429 achieving a complete response (CR) following second-line treatment.
In contrast, five patients with PFS ≤ 12 months showed increasing VAFs until relapse (Fig. 3B). Newly acquired mutations were detected in four patients (80%) during the course of therapy, indicating that the acquisition of new genetic alterations during treatment is relatively common in patients with poor PFS. For instance, in MM564, the newly acquired KRAS p.G12V (VAF=40.0%) and p.A11P (VAF=39.3%) after achieving CR to bortezomib-melphalan-prednisone (VMP) might be related to relapse observed 2 months after completing first-line therapy. Similarly, MM347 was initially treated with lenalidomide-dexamethasone (Rd). Although this patient achieved VGPR, MM progressed within 9 months of the initiation of Rd, and extramedullary MM occurred 5 months after Rd termination. During this period, AKT1 mutations disappeared, while EGFR p.V786M (VAF= 51.8%) and SPOP p.S298S (VAF=48.4%) mutations were newly acquired in ctDNA at the time of the first relapse, suggesting clonal evolution from BM or extramedullary sites.
Notably, MM233 received frontline treatment with VMP, followed by second-line Rd for the first relapse. Before initiating third-line treatment, the VAF of the TP53 p.T125T mutation increased more than three-fold compared with baseline at diagnosis. Subsequently, the patient received carfilzomib-dexamethasone (Kd); however, the additional acquisition of the RAC1 p.D38N mutation and an increased VAF of the GNAS mutation resulted in disease progression with extramedullary MM within 4 months of Kd initiation.
Fourth-line daratumumab monotherapy was administered, resulting in PR, but early disease progression occurred 4 months after daratumumab initiation. Consistent with the observed clinical progression, the VAFs of all three mutations decreased during the first 3 months of daratumumab treatment but increased again at the fourth relapse.
DISCUSSION
The advent of NGS and improved accessibility to these platforms has advanced the understanding of mutational characteristics in ctDNA from patients with MGUS and SMM, as well as explored their evolution into symptomatic MM. Our data support three major conclusions: first, the number of mutations detected in ctDNA increases with disease progression from MGUS to SMM/MM; second, MM exhibits a higher frequency of shared mutations between BM and ctDNA compared with that in precursor diseases, with frequent RAS/RAF and TP53 mutations; and third, serial ctDNA profiling is feasible as a minimally invasive method in patients with MM.
ctDNA burden correlates with disease status across MGUS, SMM, NDMM, post-treatment MM, and MM at relapse [13, 24]. Consistent with these findings, we observed an average of 1.0 mutations in ctDNA from MGUS and 1.9 mutations in SMM/MM. Additionally, a high concordance rate of mutations (68.9%) between BM and ctDNA was observed in MM, consistent with a reported 70.9% concordance in patients with MM [25]. However, we did not find any significant differences in VAF among disease groups. RAS/RAF and TP53 mutations were exclusively detected in MM genomes but not in precursor diseases. The Myeloma XI trial reported that up to 50% of CD138+ sorted BM samples from patients with NDMM harbored mutations in genes involved in the RAS signaling cascade, including KRAS, NRAS, and BRAF [26]. TP53 dysregulation also plays a pivotal role in the pathogenesis of various cancers, including MM. Collectively, mutations detected in ctDNA from precursor diseases are, both quantitatively and qualitatively, compared with ctDNA in MM and do not sufficiently represent mutations in malignant PCs within the BM. The acquisition of additional driver mutations in RAS/RAF or TP53 may serve as ctDNA markers to identify the progression from precursor diseases to symptomatic MM.
Interestingly, recurrent NIPA2 mutations were detected in patients with MM. Although NIPA2 has previously been reported as recurrently mutated in prostate cancer [27], to the best of our knowledge, no study has demonstrated a direct link between NIPA2 mutations and the pathogenesis of MM. NIPA2 is involved in mitophagy regulation [28], and recent research has highlighted the involvement of mitophagy-related genes in MM pathogenesis [29, 30]. The potential connection between NIPA2 and MM pathogenesis remains an intriguing avenue for future investigation.
Targeted NGS has identified RAS mutations as a common feature in patients with extramedullary myeloma [31]. TP53 mutations have been associated with paramedullary myeloma, a condition indicative of advanced bone lytic disease independent of extramedullary myeloma. These findings provide additional evidence supporting our results [32, 33]. To the best of our knowledge, our study is the first to identify statistically significant associations between ctDNA mutations and clinical presentations in a relatively large real-world cohort.
Several studies have actively explored the potential of ctDNA profiles as prognostic biomarkers for NDMM [25, 34, 35]; however, limited evidence exists for its utility alongside established clinical prognostic variables in large cohorts. We investigated the prognostic significance of ctDNA profiles in conjunction with clinical variables in a real-world MM cohort. Mutations in TET2 were correlated with adverse effects on the 2-yr PFS rate, persisting even after adjusting for other clinical variables, such as age, transplant eligibility, cytogenetic risk, and treatment intensity. This finding offers compelling evidence for the utility of ctDNA as a prognostic biomarker in MM.
Among individuals undergoing serial ctDNA monitoring, two with sustained long-term responses exhibited reduced ctDNA VAFs and lacked driver mutations, specifically RAS/RAF mutations. Conversely, all five patients progressing to relapsed/refractory MM (RRMM) within a short period showed increased VAFs in RAS/RAF. In three cases of RRMM in which TP53 trajectory data were available (MM564, MM543, and MM233), a similar trajectory pattern of increasing TP53 mutation VAFs was observed with disease progression. A recent study reported an increased frequency of RAS/RAF and TP53 mutations in RRMM compared with those in NDMM [25], highlighting their association with advanced disease status. Additionally, the acquisition of mutations beyond RAS/RAF and TP53 is a well-documented characteristic of RRMM [36, 37]. Our data further support the notion that changes in these mutations serve as potential biomarkers for predicting a deep response and favorable response duration.
This study has some limitations. First, continuous ctDNA profile monitoring from MGUS to SMM to MM was not performed in the same patient. Second, despite confirming the prognostic significance of TET2 mutation in the MM cohort, we could not distinguish between ctDNA and clonal hematopoiesis of indeterminate potential. Further prospective trials involving ctDNA profiling together with leucocyte sequencing in a larger MM cohort will provide clinically applicable biomarkers related to disease development and/or prognosis. Third, the increase in the total ctDNA mutation number, which indicates the overall mutation burden, is considered a biomarker for MM [14, 38]. However, the specific mutations required for standardization in this context remain unknown [38]. Similarly, we noted a highly heterogeneous appearance of mutations beyond RAS/RAF and TP53 in the ctDNA trajectory data. The development of personalized ctDNA assays, such as NGS panels or droplet digital PCR covering individual mutational profiles, is required to overcome this heterogeneity.
In summary, our findings suggest that profiling genetic alterations in ctDNA using a minimally invasive approach during the precursor disease stage is feasible for predicting the development of MM and its specific clinical presentation. ctDNA profiling may enhance conventional risk stratification in NDMM and serve as a dynamic tool for identifying significant genetic events during treatment. Ultimately, ctDNA assessment may provide valuable insights into the optimal timing for precise assessments, including BM examinations and PET/CT, to assist in confirming MM progression and the necessity for treatment.
XML Download