Mechanisms of resistance to bispecific T-cell engagers in multiple myeloma and their clinical implications

The prognosis of patients with relapsed/refractory multiple myeloma (RRMM) with prior exposure to immunomodulatory drugs (IMiDs), proteasome inhibitors, and anti-CD38 monoclonal antibodies (mAbs) remains poor.¹ In this population, chimeric antigen receptor (CAR) T cell and bispecific T-cell engagers (TCEs)–targeting B-cell maturation antigen (BCMA) represent a new standard of care. Idecabtagene vicleucel (ide-cel; an anti-BCMA CAR T cell) has been approved based on an overall response rate (ORR) of 73% and a median progression-free survival (PFS) of 8.8 months in patients with heavily pretreated triple-class exposed myeloma.² Ciltacabtagene autoleucel (cilta-cel, another anti-BCMA CAR T cell) has also been approved in this population based on an ORR of 97.9% and a median PFS of 34.5 months.³ Despite this favorable efficacy profile, accessibility and manufacturing process still represent a limitation for broad use of CAR T cells in MM.⁴ Bispecific TCEs are readily available off-the-shelf mAbs capable of binding to an antigen on tumor cells and to another antigen on T cells to redirect these lymphocytes toward malignant cells.⁵ To date, 2 bispecific TCEs, teclistamab and elranatamab, targeting CD3 on T cells and BCMA on myeloma cells, have been approved for the treatment of triple-class exposed RRMM. In the MajesTEC-1 study, teclistamab led to an ORR of 63% and a median PFS of 11.3 months in triple-class exposed patients who received a median number of 5 prior lines.⁶ In the Magnetismm-3 study (cohort A), elranatamab led to an ORR of 61% and a median PFS of ∼15 months in triple-class exposed patients who received a median number of 5 prior lines.⁷ Bispecific TCE targeting other tumor antigens (ie, Fc Receptor Like 5 (FCRL5) and G-protein coupled receptor family C group 5 member D [GPRC5D]) also demonstrated promising activity in relapsed MM.⁸ Recently, talquetamab, another bispecific TCE targeting GPRC5D, has also been approved for patients with RRMM, based on the results of the MonumenTAL-1 study.^9,10 In a population of patients with advanced, T-cell–redirecting agent-naive myeloma (n = 145; 69% triple-class refractory, median of 5 prior lines), talquetamamb (0.8 mg/kg biweekly) single agent demonstrated an ORR of 72% and a median PFS of 14 months. In patients previously exposed to a T-cell–redirecting agent (n = 51), talquetamab resulted in an ORR of 64%, with a median duration of response of 11.9 months. Despite this favorable efficacy profile, nearly one-third of patients do not respond to bispecific therapy (primary resistance). Moreover, most responding patients treated with bispecific antibodies will finally develop disease progression (acquired resistance). This review aims at describing the tumor-intrinsic and tumor-extrinsic mechanisms leading to bispecific TCE resistance.

To date, the BCMA-targeting TCEs teclistamab and elranatamab are approved for the treatment of patients with RRMM who received at least 3 lines of prior therapy and are triple-class exposed. Data from clinical trials identified several baseline clinical characteristics as predictors of poor response to BCMA-targeted TCEs, including presence of extramedullary disease (EMD), International Staging System (ISS) stage III, and refractory status. In Magnetismm-3, patients with EMD had an ORR of 38.5% to elranatamab, compared with 71.4% in patients without EMD.⁷ Patients with ISS III (vs ISS I-II) and penta-refractory disease (vs not penta refractory) also had an inferior response rate to elranatamab. In Majestec-1, ORR to teclistamab was also significantly inferior in patients with EMD or ISS III.⁶ Lower response rate to teclistamab in patients with EMD could be related to higher level of soluble BCMA (sBCMA) in this population.¹¹ High tumor burden was also associated with lower response rate to elranatamab (bone marrow [BM] plasma cells ≥ 50%) and teclistamab (BM plasma cells ≥ 60%).^6,7 In contrast, high cytogenetic risk was not found to significantly affect response rate to these 2 drugs.^6,7 Talquetamab is, to date, the only approved TCE targeting GPRC5D. In MonumenTAL-1, the presence of EMD was the only baseline clinical characteristic found to significantly influence response rate, with a median ORR of 48.5% and 43.2% in the weekly and biweekly cohorts in patients with EMD vs 81.8 and 88% in the weekly and biweekly cohorts in patients without EMD, respectively.⁹ ISS and cytogenetic and refractory status did not significantly affect response to talquetamab in this study.

Whole-genome sequencing of myeloma cells before BCMA-targeting TCE therapy and at relapse identified genetic inactivation of Tumor Necrosis Factor Receptor Superfamily Member 17 (TNFRSF17) gene (encoding BCMA protein) as a common tumor-intrinsic resistance mechanism. Truger et al reported the first case of BCMA antigen loss due to a homozygous deletion of TNFRSF17 gene.¹² More recently, Lee et al analyzed 14 patients with disease progression on BCMA-targeting TCE therapy and revealed biallelic TNFRSF17 inactivation in 6 cases (42.8%), by homozygous deletion (n = 1) or monoallelic loss with mutation (n = 5).¹³ Two patients displayed convergent evolution, with the emergence of several resistant clones harboring distinct TNFRSF17 alterations, highlighting the strong selective pressure imposed by TCE. TNFRSF17 mutations involved hot spots in the extracellular domain of BCMA, with 1 missense p.Arg27Pro mutation and 2 in-frame deletions p.Pro34del (found in 3 patients) and p.Ser30del (in 2 patients). Mutant proteins were still recognized by polyclonal anti-BCMA antibodies and retained the ability to bind APRIL (a proliferation-inducing ligand) and activate the prosurvival NF-κB signaling. However, BCMA extracellular domain mutations abrogated TCE binding and TCE-induced cell death. Importantly, TNFRSF17 mutations conferred distinct sensitivities to different anti-BCMA TCEs. In vitro, cells harboring p.Arg27Pro and p.Pro34del mutations were resistant to teclistamab and elranatanab but remained sensitive to alnuctamab, whereas cells with p.Ser30del mutation were resistant to teclistamab but remained sensitive to elranatanab and alnuctamab. These data in cell lines need to be confirmed in vivo but suggest that myelomas resistant to 1 anti-BCMA TCE might still be sensitive to another targeting a different epitope.

Tumor-intrinsic mechanisms of resistance to the GPRC5D-targeting TCE talquetamab were assessed in 2 synchronous studies.^13,14 Combining deep whole-genome sequencing and single-cell multiomics, Derrien et al reported convergent evolution in a patient with a clonal 12p deletion (encompassing GPRC5D locus) in the pretreatment sample. Seven resistant subclones emerged at relapse, each having acquired a distinct second hit in GPRC5D (3 frameshift indels, 2 nonsense mutations, 1 in-frame deletion, and a large deletion encompassing the transcription start site) leading to the complete loss of GPRC5D protein at the cell surface.¹⁴ Similarly, Lee et al reported 4 talquetamab-resistant cases with biallelic GPRC5D inactivation due to homozygous deletion or monoallelic deletion with mutation (1 frameshift indel, 1 missense, and 2 nonsense mutations).¹³ The mutation landscape of GPRC5D mostly involves truncating mutations distributed all along the protein sequence, in sharp contrast with the hot spot mutations in TNFRSF17 that alter TCE recognition while preserving BCMA-mediated prosurvival signaling (Figure 1). Finally, Derrien et al reported 2 talquetamab-resistant cases with a loss of GPRC5D expression due to the long-range epigenetic silencing of its promoter and enhancer regions in absence of any genetic alteration.¹⁴ This is a proof-of-concept that epigenetic remodeling alone can induce TCE resistance by silencing the transcription of the antigen. Overall, resistance to GPRC5D-targeting TCEs usually involved a complete inactivation of the target, suggesting that myeloma cells better tolerate the loss of GPRC5D than the loss of BCMA. Consistently, reduced or lost GPRC5D expression was observed in 6 of 6 patients who relapsed after anti-GPRC5D CAR T-cell therapy,¹⁵ whereas the loss of BCMA expression was rare after anti-BCMA CAR T-cell therapy (3/71 [4%]).¹⁶ BCMA promotes the growth of MM cells, protects them from apoptosis, and promotes immunosuppression in the BM microenvironment.^17,18 These prosurvival effects may prevent the selection of clones with BCMA inactivation, even in the presence of anti-BCMA treatment.

Using single-cell RNA sequencing and TCR tracing, Friedrich et al explored the dynamic response of T cells in patients with myeloma treated with anti-BCMA TCE.¹⁹ TCE response was driven by the clonal expansion of effector CD8⁺ T cells but also naive T cells. Importantly, major histocompatibility complex (MHC) class I interaction with tumor cells and MHC class I: TCR costimulatory signaling were required for the functional recruitment and priming of naive T-cell clones. Several lines of evidence highlighted the loss of MHC class I as a potential tumor-intrinsic mechanism of TCE resistance beyond the loss of the target antigen. First, the expression of MHC class I (HLA-E and HLA-C) and class II genes (CD74) was deregulated in response to TCE treatment. Second, loss of MHC class I surface expression was identified at relapse by flow cytometry in some patients. However, the frequency and causal mechanism of this loss of MHC class I expression remain to be established.

The response to bispecific TCE treatment is affected by several tumor-extrinsic factors including the preexisting T-cell landscape, its evolution, and the immunosuppressive tumor microenvironment created by myeloma cells and related to previous treatments.^19-21 In a preclinical study, Verkleij et al showed that talquetamab-mediated killing of MM cells is impaired by an increased proportion of several T-cell populations, including T cells expressing the exhaustion marker programmed cell death 1 (PD-1), activated T cells expressing human leukocyte antigen (HLA)-DR, and regulatory T cells (Treg).²⁰ In the transplantable Vk∗MYC MM mouse model, T cells upregulated PD-1 expression in response to anti–BCMAxCD3 bispecific TCE and diminished in functionality over time, leading to systematic relapse after treatment.²¹ Interestingly, the addition of pomalidomide, an IMiD, increased the expansion of lytic T cells and short-term efficacy of the TCE, but also induced important toxicity and exacerbated T-cell exhaustion, leading to only marginal survival benefit in this preclinical model. In contrast, a combination of the BCMA-targeting TCE with cyclophosphamide was safe and allowed for long-term myeloma control by reducing tumor burden, depleting Tregs, and preventing TCE-induced T-cell exhaustion. In line with these preclinical studies, Friedrich et al found that the abundance of exhausted CD8⁺ T-cell clones predicts response failure to BCMAxCD3 bispecific TCE in patients with MM.¹⁹ Consistently, van de Donk et al reported baseline immune characteristics predicting unfavorable response to the same TCE, including lower T-cell numbers, higher T cells expressing PD-1, TIM-3, or CD38, increased Tregs and CD38⁺ Tregs, and lower proportion of naive T cells.²² These studies stressed the importance of the preexisting T-cell repertoire in the response to bispecific TCE therapy. Other factors generate an immunosuppressive environment in MM and may contribute to TCE resistance, including the interaction beween MM and bone marrow stromal cells (BMSCs), inhibitory cytokines (transforming growth factor β, interleukin-6, or interleukin-10) and myeloid cells.^23,24 The interaction between MM and BMSCs has been shown to protect MM cells from T-cell cytotoxicity.^25,26 In vitro, the addition of BMSCs impaired the talquetamab-mediated lysis of MM cell lines.²⁰ This protective effect involved cell-cell contact but not BMSC-derived soluble factors nor a reduction in T-cell activation, suggesting the induction of tumor cell-intrinsic resistance mechanisms. Inhibitor myeloid cells such as myeloid-derived suppressor cells and plasmacytoid dendritic cells have also been reported to drive an immunosupressive environment favoring MM progression.^27-30 Their potential role in TCE resistance remains to be explored in patients.

Tumor-intrinsic and tumor-extrinsic mechanisms of TCE resistance are summarized in Figure 2.

The identification of molecular mechanisms underlying TCE resistance provides valuable insights to guide future immunotherapy in MM (Table 1). Before treatment, molecular characterization of the target antigens in tumor cells and of the immune repertoire may help select the most appropriate immunotherapy for each patient. At relapse, understanding the molecular cause of resistance will be instrumental in choosing the next treatment line.

TCE resistance by loss of the target antigen requires the inactivation of the 2 copies of the gene. Preexisting deletions or mutations of TCE targets may thus favor the emergence of resistance. A representative example is the talquetamab-resistant case published by Derrien et al in which a preexisting 12p deletion (encompassing GPRC5D) allowed the emergence of 7 resistant subclones, each harboring a distinct second hit.¹⁴ Similarly, Lee et al described 3 patients harboring pretreatment 16p (encompassing TNFRSF17) or 12p deletions who developed subclones resistant to BCMA-targeting (respectively GPRC5D) TCE after acquisition of second hits.¹³ Screening of target alterations in large cohorts of TCE treatment-naive MM revealed recurrent heterozygous deletions of TNFRSF17 (3%-8%), GPRC5D (13%-15%), or CD38 (10%).^12,13,31,32 Of note, patients with 16p deletion (encompassing TNFRSF17) have increased deletion frequencies of other chromosomes and may be more vulnerable to the biallelic loss of other genes.³² Altogether, heterozygous deletion of 1 target occurs in ∼30% of MM. Other targets such as FCRL5 and SLAMF7, located on chromosome arm 1q, are recurrently gained in RRMM. In addition to deletions, rare somatic mutations of GPRC5D were identified in TCE-naive MM,¹² as well as somatic (1.1%) and germline (0.7%) TNFRSF17 mutations, including a recurrent p.Pro33Ser germline variant notably encountered in a patient with primary refractory disease to anti-BCMA TCE.¹³ Screening these events may improve TCE response by prioritizing target genes with 2 intact copies in MM cells, although the predictive value of monoallelic target alterations at baseline remains to be demonstrated in clinical series. In addition to their genomic status, the baseline expression of target antigens may influence TCE response. The 2 talquetamab-resistant cases with epigenetic GPRC5D silencing belonged to the t(11;14) molecular group¹⁴ that displays the lowest GPRC5D messenger RNA expression.²⁰ In vitro, the efficacy of talquetamab was superior in patient-derived MM cells with high GPRC5D expression. Whether a low baseline expression may facilitate acquired TCE resistance by epigenetic silencing of the target, for example, by extension of inactive chromatin marks, will need to be examined in large clinical cohorts.

The abundance of exhausted-like T-cell clones was associated with TCE response failure, providing a rationale for immune monitoring before treatment.⁶ This could be done by cytometry or single-cell RNA/variable diversity joining (VDJ) sequencing. The feasibility of integrating single-cell RNA sequencing analyses in clinical trials was already demonstrated in MM⁸ and could allow for monitoring the evolution of T-cell clones as well as their phenotypic trajectories. Similarly, investigation of the other BM microenvironmental components, both soluble (cytokines) and cellular (Tregs, BMSC, myeloid-derived suppressor cells, and plasmacytoid dendritic cells) and their status may provide information about creating a permissive environment for optimal clinical activity of TCE. However, additional studies are required to establish straightforward measures and cutoffs on specific cell populations that could be used in clinical practice.

Resistance mechanisms also inform the strategy of immunotherapy sequencing in MM. To date, limited clinical data regarding TCE sequencing are available. In patients receiving anti-BCMA TCE as first subsequent therapy after talquetamab (n = 19), the ORR was 57.9%, which is close to ORR in Majestec-1 or Magnetismm-3 studies.³³ In MonumenTAL-1, patients receiving talquetamab as subsequent therapy after BCMA TCE (n = 18), the ORR was 44.4%, in comparison with 71.7% in prior TCE-naive patients (0.8 mg/kg cohort).¹⁰ Complete inactivation of a target, for example, by homozygous deletion, likely precludes response to other immunotherapies targeting the same antigen. For example, in the case reported by Truger et al, biallelic loss of TNFRSF17 after BCMA-targeting TCE led to an absence of response to subsequent treatment with an anti-BCMA antibody-drug conjugate.¹ By contrast, mutations in the extracellular domain of BCMA can impair the binding of 1 TCE but not another.¹³ Patients may thus benefit from sequential or combined TCE targeting different BCMA epitopes. TNFRSF17 mutations were less frequent in resistant MM after BCMA-targeting CAR T cell than BCMA-targeting TCE. This could be an argument for CAR T-cell use in first immunotherapy line, especially with the advent of more efficient CAR platforms such as T-Charge.³⁴ More than half of the 14 relapses after BCMA-targeting TCEs studied by Lee et al did not involve TNFRSF17 genetic alterations.¹³ Part of these cases are likely explained by tumor-extrinsic factors such as T-cell exhaustion.¹⁹ Such mechanisms may also prevent response to subsequent TCE targeting other antigens. By contrast, all sequenced relapses after GPRC5D-targeting TCE were driven by genetic or epigenetic GPRC5D inactivation,^13,14 which should not impair the efficacy of immunotherapies targeting other antigens. Consistent with these predictions, immune and genome profiling of a few patients with sequential immunotherapies suggests that T-cell exhaustion precludes response to subsequent immunotherapy lines, whereas genetic inactivation of an antigen does not impair response to another immunotherapy targeting another antigen or epitope.³⁵ Importantly, the frequency of heterozygous deletions encompassing TCE target gene loci increases significantly between newly diagnosed MM (NDMM) and RRMM.¹² Myeloma cells also contribute to creating an immunosuppressive BM by several means.^29,36-38 As a result, the immune microenvironment gets compromised during MM progression,³⁹ and RRMM display features of T-cell exhaustion.⁴⁰ Both effects are likely to limit the efficacy of TCEs in advanced disease and argue for their use in early treatment lines.

Both tumor-intrinsic and tumor-extrinsic resistance mechanisms of TCE resistance have been elucidated, but they were so far analyzed separately. Joint analyses of tumor and immune cell responses to TCE treatment will be useful to understand how tumor cells may influence T-cell response and to estimate the proportion of cases in which resistance remains unexplained. Interestingly, some post-TCE relapses displayed subclonal target inactivations affecting most but not all tumor cells.^13,14 Treatment escape in the remaining subclones may involve undetected target alterations (eg, mutations in very small clones) or other resistance mechanisms yet to be characterized. Epigenetic inactivation was investigated for GPRC5D¹⁴ but not TNFRSF17. In addition, γ-secretase can shed BCMA protein from the cell surface and release sBCMA into the blood.⁴¹ An activation of this process may allow for TCE escape through the removal of the target antigen from MM cells and interference of the drug with sBCMA. High sBCMA levels were associated with increased tumor burden, EMD, and lower response to anti-BCMA TCE.^11,42,43 In vitro, high sBCMA levels decreased the binding of anti-BCMA antibodies to MM cells⁴⁴ and the efficacy of anti-BCMA CAR T cells and TCEs.⁴⁵ Interestingly, structural genomic rearrangements leading to the overexpression of BCMA and higher sBCMA levels were identified in MM after anti-BCMA CAR T-cell/TCE treatment.⁴⁵ Finally, 2 studies reported down-modulation of interferon gamma signaling as an acquired mechanism of resistance to HER2-targeting TCE in gastric and breast cell lines, conferring resistance to killing by active T lymphocytes.^46,47 Inhibition of interferon gamma signaling has not been reported so far in TCE- treated MM.

To limit target-related mechanisms of resistance, strategies using multitarget TCEs are currently being investigated. One option is to combine bispecific antibodies. The phase 1b study RedirecTT-1 evaluated the combination of biweekly talquetamab and teclistamab in patients with RRMM (median of 4 prior lines; 79.6% triple-class refractory). The combination demonstrated high efficacy with an ORR of 86.6%, including 40.2% complete response, and a median PFS of 20.9 months.⁴⁸ Another option for multitarget approach is the use of trispecific antibodies, targeting CD3 and 2 distinct targets on plasma cells. A phase 1 study is currently evaluating a trispecific antibody targeting CD3 × BCMA × GPRC5D in patients with relapsed myeloma (ClinicalTrials.gov identifier: NCT05652335). To limit tumor-extrinsic mechanisms of resistance (ie, T-cell exhaustion) several ongoing studies evaluate bispecific antibodies in combination with IMiDs (ie, lenalidomide and pomalidomide), anti CD38 antibodies (ie, daratumumab), or immune checkpoint inhibitors (ie, cetrelimab) that have been shown to promote T-cell activity.^49,50 Initial efficacy and safety results of a talquetamab plus pomalidomide combination were promising in the MonumenTAL-2 study.⁵¹ Phase 1b studies combining bispecific antibodies with anti-CD38 mAbs demonstrated promising response rates.^52,53 The ongoing phase 1 to 2 study TRIMM-3 evaluates the efficacy and safety of teclistamab or talquetalab in combination with anti–PD-1 cetrelimab (ClinicalTrials.gov identifier: NCT05338775). Moreover, ongoing clinical trials evaluated the combination of BCMA × CD3 bispecific antibodies with gamma secretase inhibitors (GSI) to decrease soluble BCMA levels implicated in BCMA bispecific resistance (ClinicalTrials.gov identifier: NCT04722146). Combination of TCE plus the GSI nirogacestat led to a promising response rate, but high-grade immune events were reported in the cohort with early administration of GSI during teclistamab priming doses.⁵⁴ Combining bispecific TCEs with cyclophosphamide may also improve T-cell persistence and function, as demonstrated in preclinical models.²¹ Altogether, combinations with various therapeutic classes hold great promise to improve the efficacy of TCEs in myeloma, notably by limiting T-cell exhaustion.

This study was supported by the Région Pays de la Loire (Connect Talent program), Nantes Métropole (RPH2114NNA), CHU Nantes, the I-SITE NExT (ANR-16-IDEX-0007), HéMA-NExT research cluster and the SIRIC ILIAD (INCa-DGOS-INSERM-ITMO Cancer_18011).

Contribution: All authors performed a literature review, and wrote and edited the manuscript.

Conflict-of-interest disclosure: P.M. participates in the advisory boards of and receives honoraria from Janssen, Celgene, AbbVie, Pfizer, Amgen, Sanofi, and Takeda. M.S. participates in the advisory boards of AbbVie and NCGM. C.T. participates in the advisory boards of and receives honoraria from Janssen. The remaining authors declare no competing financial interests.

Correspondence: Eric Letouzé, Nantes Université, INSERM, CRCI2NA, 8 quai Moncousu 44000 Nantes, France; email: eric.letouze@inserm.fr; and Cyrille Touzeau, University Hospital Hôtel-Dieu, 1 Place Alexis Ricordeau, 44000 Nantes, France; email: cyrille.touzeau@chu-nantes.fr.