Blood-derived cell-free DNA is a superior biomarker for noninvasive DNA methylation profiling in multiple myeloma Open Access

TO THE EDITOR:

Multiple myeloma (MM) is an incurable hematological malignancy characterized by clonal plasma cell proliferation, with dynamic and complex genetic and epigenetic alterations that drive disease progression, therapy resistance, and relapse.^1-3 Previous studies have shown that promoter hypermethylation of distinct tumor suppressor genes such as CDKN2A, DAPK1, and CDH1 is associated with inferior prognosis in MM.⁴ More recently, at the therapeutic level, hypermethylation of the CRBN enhancer regions and PSMD5 and CD38 promoter regions have been implicated in acquired resistance to immunomodulatory drugs, proteasome inhibitors, and anti-CD38 monoclonal antibodies, respectively.^5-7 Even for bispecific T-cell engagers such as the GPRC5D-directed talquetamab, long-range epigenetic silencing of the promoter and enhancer regions has been identified as a mechanism driving resistance.⁸ Hence, characterization of the MM epigenetic profile would be highly informative for clinicians to guide and personalize the treatment trajectory. Peripheral blood–based monitoring methods by the use of “liquid biopsies” represent a powerful noninvasive approach to detect and track (epi)genetic alterations, offering significant advantages over conventional bone marrow (BM) aspirates, which are invasive and are likely to underestimate spatial tumor heterogeneity.^9-12 However, despite promising results in solid tumors and in particular for blood-derived cell-free DNA (cfDNA), evidence for DNA methylation profiling using liquid biopsies in MM is still very scarce and fragmented.^13-16 

In this proof-of-concept study, we compared the DNA methylation profiles derived from cfDNA, circulating tumor cell DNA (CTC-DNA), peripheral blood mononuclear cell DNA (PBMNC-DNA), and matched BM-DNA in 11 patients with relapsed MM. This cohort consisted of 4 males and 7 females, with a median age of 72 years. Patients received a median of 2 prior treatment lines. Additionally, both cfDNA and genomic DNA (gDNA) collected from 4 human myeloma cell lines (HMCLs; ie, OPM-2, RPMI-8226, U266, and XG-7) were used to detect previously described epigenetic alterations as verification of our workflow.¹⁷ The NEB Next Enzymatic Methyl-seq kit (New England Biolabs, Ipswich, MA) was used to analyze the methylation status of genome-wide CpG islands (CGIs) with an average genome coverage of ∼20×. Bioinformatics analysis was focused on the identification of differentially methylated regions (DMRs) in MM and HMCLs samples compared to pooled healthy control gDNA samples (n = 4). Additional information about this patient cohort and the methodology is provided in supplemental Table 1 and supplemental Methods, respectively.

In accordance with literature, we detected consistent hypermethylation of the CGIs overlapping with the promoter regions of CDKN2A, CDH1, and RASSF1A in both the cfDNA and gDNA of the HMCLs we studied (supplemental Figure 2). This indicates that cfDNA derived from the HMCL culture supernatant accurately reflects the epigenetic alterations found in matched gDNA. Upon analysis of the patient samples, we detected a total of 16 381 DMRs in the 44 MM DNA samples compared to control gDNA (supplemental Table 2). The majority of DMRs were hypomethylated and located in gene bodies and intergenic regions, consistent with what is known in the literature, whereas promoter regions, compared to other genomic features, showed a modest increase in hypermethylation⁴ (supplemental Figure 3).

We observed significant differences in CGI methylation pattern among the different DNA types we investigated (χ² test, P < .0001). As can be derived from Figure 1A, BM-DNA and PBMNC-DNA show a similar, homogeneous methylation level distribution pattern, whereas CTC-DNA and cfDNA have a more heterogeneous pattern. Principal component analysis further confirmed this heterogeneity, showing a clear distinction between cfDNA and CTC-DNA from patients with MM compared to BM-DNA and PBMNC-DNA, which clustered more closely together (Figure 1B). In contrast, the healthy control gDNA and cfDNA samples show a homogeneous methylation level distribution pattern, and after direct comparison between both control DNA types, only 97 significant DMRs (Δβ > 0.25; adjusted [adj.] P < .001) were detected (supplemental Figure 4). This finding suggests that DMRs found in patient cfDNA are likely disease specific rather than reflecting intrinsic variability in cfDNA.

cfDNA showed the highest overall DMR detectability rate, permitting detection of 14 122 of 16 381 DMRs (86.2%) found in this cohort (supplemental Table 3). Upon comparison of the circulating biomarkers (ie, cfDNA, CTC-DNA, and PBMNC-DNA) with BM-DNA, cfDNA showed the highest concordance with BM-DNA, with 5640 of 7211 DMRs (78.2%) found in BM-DNA also detected in cfDNA. This was significantly higher than the observed concordance of 53.3% and 42.4% for CTC-DNA and PBMNC-DNA, respectively (χ² test, adj. P < .0001). Furthermore, in contrast to the methylation pattern heterogeneity shown in Figure 1, a strong correlation was observed between the DMRs found in cfDNA and CTC-DNA, with 70.4% of the DMRs detected in cfDNA also found in CTC-DNA (Pearson r = 0.724). This observation reinforces the tumor-specific origin of these DNA methylation alterations. Both cfDNA and CTC-DNA permitted detection of a significant number of DMRs that were not found in matched BM-DNA, most likely reflecting the spatial epigenetic heterogeneity that has previously been reported in the BM of patients with MM.⁹ Taken together, our results suggest a superior performance of cfDNA to detect aberrant DNA methylation in MM.

Finally, to gain insight into the genetic pathways affected by aberrant DNA methylation in our patient cohort, we performed a pathway enrichment analysis on the total data set of 44 sequenced MM samples (Figure 2A-C), based on the human Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways (supplemental Tables 4-6). For the total methylation data set, we found 155 significantly enriched KEGG terms, highlighting most notably the enrichment of transcriptional misregulation, Ras/MAPK signaling, and focal adhesion pathways (adj. P < .0001). Analysis of the subset of hypermethylated DMRs (Figure 2B) showed an enrichment in cancer-related pathways such as the Fanconi anemia pathway and Hippo signaling pathway. Of note, several pathways were enriched in the hypomethylated data set but not in the hypermethylated data set. These included, most notably, the JAK-STAT, mTOR, and NF-κB signaling pathways, which are involved in MM pathogenesis, and genes associated with the extracellular matrix (ECM), previously implicated in MM-related aberrant methylation.^18,19 Because Evers et al²⁰ recently showed an association between high expression levels of specific subsets of ECM genes and inferior prognosis in MM, further research investigating the relationship between aberrant methylation and expression of these ECM genes would be of interest.

In summary, this study reports the results of, to our knowledge, the first comprehensive, comparative analysis of the applicability of currently available circulating biomarker–derived DNA sources for DNA methylation profiling in MM. Our results indicate that circulating biomarkers, especially cfDNA, reliably identify aberrant DNA methylation in key regulatory pathways in MM while revealing distinct epigenetic characteristics and differentially methylated CGI compared to BM-DNA. Although based on a limited cohort size, this study identifies cfDNA as a feasible candidate biomarker for noninvasive and comprehensive DNA methylation profiling in MM. The possibility of standardizing and automating the cfDNA extraction process is likely to facilitate its implementation in a clinical laboratory setting. These practical considerations, supported by the superior results that we obtained with cfDNA compared to other biomarkers, advocate for the use of cfDNA as the biomarker of choice for epigenetic profiling in MM. This noninvasive strategy is likely to accelerate the implementation of multiomics diagnostics in MM clinical practice in the coming years, in which it will prove to be an indispensable part in the ongoing evolution toward personalized medicine.

Acknowledgments: The authors acknowledge Veerle De Greef, Ann Heymans, Gerda Van den Brande, and the laboratory staff of Brussels Interuniversity Genomics High Throughput Core for their excellent support in the experimental work.

This study was funded by a grant from Kom op tegen Kanker, the King Baudouin Foundation, the VUB Strategic Research Program (grants SRP48 and SRP84), and the UZ Brussel Foundation. R.H. is the recipient of a PhD fellowship in Strategic Basic Research (1SE9324N) of the Research Foundation Flanders.

Contribution: R.H., I.V.R., and E.D.B. designed the study; R.H., J.S., S.A., and T.J. performed experiments; R.H., J.S., S.A., and C.O. performed data analysis; A.D.B., W.D.B., R.S., and I.V.B. were responsible for patient samples; I.V.R., E.D.B., R.S., and M.B. supervised the study; R.H. and I.V.R. drafted the manuscript; and R.H., C.O., J.S., S.A., T.J., A.D.B., W.D.B., R.S., I.V.B., M.B., E.D.B., and I.V.R. revised the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Ivan Van Riet, Department of Hematology, Universitair Ziekenhuis Brussel, Laarbeeklaan 101, 1090 Brussels, Belgium; email: Ivan.VanRiet@uzbrussel.be.