Promoter hypermethylation as a reversible mechanism of resistance to GPRC5D-directed therapy in multiple myeloma

GPRC5D-directed immunotherapies, such as the GPRC-5D×CD3 bispecific antibody talquetamab (TAL) and the anti-GPRC5D chimeric antigen receptor (CAR) T-cell products MCARH109 and Arlocabtagene autoleucel (Arlo-cel), have shown significant clinical activity in heavily pretreated multiple myeloma (MM) patients.^1-3 Most patients however eventually develop resistance, and while biallelic loss of GPRC5D on chromosome 12p13.1 is now established as a resistance mechanism in GPRC5D-exposed patients,^4-6 Derrien and colleagues have recently described long-range chromatin closure and loss of promoter accessibility at the GPRC5D locus as a novel form of epigenetic resistance in talquetamab-resistant patients.⁷ Similarly, Ma and colleagues demonstrated that altered GPRC5D promoter methylation results in antigen loss in patients relapsing after anti-GPRC5D CAR T-cell therapy.⁸ Although both studies offered novel mechanistic insights, direct functional implications of GPRC5D promoter hypermethylation as a reversible and putatively druggable mechanism of target evasion have yet to be demonstrated.

Building on our previously published findings,⁹ we first analyzed an expanded whole-genome sequencing (WGS) dataset comprising 402 TAL-naïve patients. All patients provided informed consent in accordance with the Declaration of Helsinki (ethics vote 8/21, University of Wuerz-burg). Monoallelic GPRC5D alterations (including deletions, frameshift or missense mutations) were detected in 12.9% of cases, whereas the frequency of GPRC5D loss, defined as either deletion or copy-neutral loss of heterozygosity (CN-LOH), was substantially enriched at 61.5% (8/13) in a second cohort of TAL-exposed patients with more extensive pretreatment (Figure 1A). Of these TAL-exposed patients, four of eight patients exhibited clonal biallelic GPRC5D alterations (biallelic deletions or CN-LOH with a clonal mutation), whereas the remaining four patients carried isolated monoallelic deletions or monoallelic loss combined with subclonal events.

Figure 1.Clinical and functional evidence of monoallelic GPRC5D loss. (A) Bar chart demonstrating frequency of GPRC5D alterations in talquetamab (TAL)-naïve (N=402) versus TAL-exposed multiple myeloma (MM) patients (N=13). (B) 23.3 Mb monoallelic deletion on chromosome 12p including the G-protein coupled receptor class C group 5 member D (GPRC5D) gene and 3 subclon-al GPRC5D mutations detected in MM1 by whole-genome sequencing. (C) Immunohistochemistry showing complete loss of GPRC5D at relapse. (D) Quantitative polymerase chain reaction analysis showing a 94% reduction of GPRC5D mRNA expression at TAL relapse versus baseline.

From the subset of patients with monoallelic GPRC5D alterations, we here report on a 58-year-old patient (MM1) with penta-refractory high-risk MM, including del17p and high-risk gene expression profiling by Sky-92 testing.¹⁰ In his tenth line of therapy, the patient was treated with TAL and daratumumab as part of the phase Ib TRIMM-2 clinical trial (clinicaltrial gov. Identifier: NCT04108195) and achieved a stringent complete remission (sCR) according to International Myeloma Working Groupcriteria lasting for a total of 15 months. At relapse, WGS of bone marrow-derived CD138⁺ plasma cells revealed an acquired monoallelic 12p deletion spanning 23.3 Mb, which included the GPRC5D gene, along with three subclones carrying GPRC5D mutations (p.Tyr12*, variant allele frequency [VAF] 6%; p.Trp126*, VAF 9%; p.Trp199*, VAF 5%) (Figure 1B). Immunohistochemistry (IHC) analysis demonstrated complete loss at protein level (Figure 1C) and a 94% reduction of GPRC5D mRNA expression (P<0.0001) by quantitative polymerase chain reaction (qPCR) (Figure 1D). Given the subclonal nature of the single-nucleotide variants (SNV) which affected the remaining allele, we concluded that these additional genomic events were unlikely to fully account for the near-complete loss of GPRC5D expression, but rather support a scenario in which a second epigenetic hit, e.g., by promoter methylation, may have contributed to inactivation of the residual allele.¹¹

Figure 2.Functional validation of promoter hypermethylation and therapeutic intervention in vitro. (A) Whole-genome bisulfite sequencing (WGBS) in patient MM1 at relapse versus baseline visualizes increase in G-protein coupled receptor class C group 5 member D (GPRC5D) promoter methylation. (B) Quantitative polymerase chain reaction (qPCR) analysis comparing GPRC5D mRNA expression in OPM2 with wild-type versus monoallelic versus biallelic knockout status. (C) Dual-luciferase reporter assay using CpG-free constructs containing the unmethylated (-) GPRC5D promoter region versus methylated construct (+) versus empty vector (EV). (D) qPCR analysis for GPRC5D mRNA expression in AMO1 cells treated with the DNA methyltransferase 1 inhibitor (DNMTi) GSK-3484862 after 3 and 7 days of treatment. (E) Promoter methylation in AMO1 cells treated with GSK-3484862 after 3 and 7 days of treatment. (F) Proposed mechanism of GPRC5D regulation in patients with pre-existing monoallelic alterations, suggesting promoter hypermethylation as a second hit and potential target for DNMTi. TAL: talquetamab; DMSO: dimethyl sulfoxide. Figure 2F created in BioRender under publication license https://BioRender.com/hv0ij0d. Statistical significance is indicated as follows: ***P<0.001; ****P<0.0001.

To further explore this hypothesis, we assessed promoter methylation in MM1 using whole-genome bisulfite sequencing (WGBS). This analysis revealed a 26.5% increase in GPRC5D promoter methylation in the TAL relapse sample of MM1 compared to baseline (Figure 2A). In further validation of this observation, we generated CRISPR-Cas9 OPM2 MM cell line models with either monoallelic or biallelic GPRC5D knockouts (KO). Biallelic KO models showed an expected 91.5% reduction in GPRC5D mRNA levels, whereas monoallelic alterations reduced GPRC5D expression to 57.2% of wild-type (WT) (42.8% reduction) (Figure 2B), confirming that monoallelic gene loss was indeed not sufficient to fully abrogate GPRC5D expression. We next investigated if altered promoter methylation may account for the loss of GPRC5D expression in MM1. To this end, regulatory promoter activity was studied by Firefly/ Renilla dual-luciferase assay¹² using a CpG-free vector (CpGL) containing the GPRC5D promoter and the differentially methylated segment for chromosome 12 (12,952,081-12,952,935). In vitro methylation reduced the reporter activity by ~278-fold as compared to the unmethylated construct (Figure 2C), providing direct evidence that hypermethylation of this regulatory element suppresses the transcriptional activity of GPRC5D. Next, AMO1 MM cells, intrinsically characterized by a hypermethylated GPRC5D promoter, were treated with the pre-clinical non-covalent DNA methyltransferase 1 inhibitor (DNMTi) GSK-3484862. DNMT1 inhibition led to a 2.2-fold increase in GPRC5D expression by day 3 and a 5.9-fold increase at day 7 after treatment (Figure 2D) along with a significant decrease of promoter methylation (mean methylation difference: -8.6%; P<0.001; Figure 2E).

Taken together, the functional and clinical data presented in this report support an epigenetic second-hit model, in which promoter hypermethylation contributes to GPRC5D down-regulation under the selective pressure of GPRC5D-targeted therapy.

Over the past years, cellular plasticity and epigenetic adaptation, either through permissive chromatin accessibility or enhancer rewiring, have emerged as key mechanisms of resistance in cancer which enable rapid cellular adaptation without the need for genetic alterations.^13,14 In an elegant study, Derrien and colleagues previously combined WGS with single-nucleus RNA- and ATAC-sequencing to describe loss of promoter accessibility together with long-range chromatin closure (~1.15 Mb) around the GPRC5D locus in TAL-resistant patients. These observations indicate that epigenetic repression of the promoter may contribute to genetic silencing and subsequent antigen loss.⁷ Beyond chromatin accessibility, Ma and colleagues provided further single-base-resolved DNA methylation data to describe multiple hypermethylated CpGs across GPRC5D regulatory regions in post-CAR T relapses lacking biallelic genetic lesions.⁸ The methylation levels of these regions showed an inverse correlation with GPRC5D mRNA expression across MM cell lines. However, the specific regulatory elements demethylated by treatment, as well as a direct causal relationship at the promoter site, were not identified. Our study adds to this gap in knowledge by demonstrating a direct link between epigenetic alterations at the GPRC5D promoter and reduced GPRC5D expression. Using CRISPR gene edited mono- versus biallelic GPRC5D KO models, we further delineate that monoallelic GPRC5D loss does not fully abolish expression but rather implicates the need for a second epigenetic hit for full transcriptional loss. These findings propose a model in which MM patients with baseline monoallelic GPRC5D deletions may be particularly susceptible to secondary promoter hypermethylation (Figure 2F). Moreover, our data provide first evidence that promoter hypermethylation can be overcome by targeted inhibition of DNA methyltransferases. This co-treatment strategy, which aims at preventing target evasion, may have broader implications for overcoming epigenetic resistance to immunotherapies across other targets, including the BCMA-coding gene TNFRSF17¹⁵, as well as for other malignancies beyond MM.

Footnotes

• Received September 11, 2025
• Accepted December 23, 2025

Correspondence

^#JMW and KMK contributed equally as senior authors

K.M. KORTÜM - kortuem_m@ukw.de

Disclosures

HE discloses research funding and other support from Sanofi, Takeda, Novartis, BMS, GSK, Amgen, and Janssen. CH is a shareholder and employee of MLL Munich Leukemia Laboratory. LR discloses honoraria from Sanofi, Amgen, GSK, Janssen, Pfizer, and BMS. JMW discloses honoraria from GSK, Menarini-Stemline, Pfizer, Abbvie, Sanofi, BMS, Oncopeptides and Janssen. JMW discloses consultancy for Johnson & Johnson, Sanofi, Takeda, Pfizer, Oncopeptides, Menarini-Stemline, Skyline Dx, Abbvie, Amgen and GSK; and research support from BMS. KMK discloses honoraria from GSK, Menarini-Stemline, Pfizer, AbbVie and BMS; and grants and honoraria from Janssen. All other authors have no conflicts of interest to disclose.

Contributions

Funding

This work was funded by the National Institute for Cancer Research (Program EXCELES, project ID number LX22NPO5102), and supported by the European Union Next Generation EU Program. KMK received support from the DFG (TRR387 and KFO 5001), the Stifterverband, and Deutsche Krebshilfe via MSNZ. TS was funded by the DFG (project number 442740310). JMW was supported by the BMBF/INTERACT Advanced Clinician Scientist Program at University of Würzburg. HE received funding from the German Federal Ministry of Education and Research (BMBF, grant number 13N15986).

Acknowledgments

We thank all patients and their families who participated in this study. This work was supported by the Genomnet Hematology (Genomnetzwerk Hämatologie).