CD70-CAR-Ts: a second wind for high-risk myeloma Free

In this issue of Blood, Kasap et al¹ present an innovative chimeric antigen receptor T-cell (CAR-T) design targeting CD70 that may offer a new therapeutic axis for patients with high-risk multiple myeloma (MM) refractory to BCMA-directed therapies. Despite remarkable initial response rates to salvage therapies, relapsed/refractory MM remains an incurable disease, particularly dire in patients harboring high-risk cytogenetic abnormalities such as t(4;14), 1q gain, or del(17p), whose survival is markedly shorter despite modern therapeutic advances.² The advent of BCMA-targeted immunotherapies, including antibody-drug conjugates, bispecific T-cell engagers, and CAR-T therapies, has transformed the treatment landscape. Nevertheless, durable responses are limited: up to 50% of patients relapse within 5 to 6 months of BCMA CAR-T infusion, especially those with extramedullary disease, plasma cell leukemia, or elevated inflammatory markers.³ 

Relapse mechanisms found in patients who received BCMA-directed therapy are multifactorial. Tumor-intrinsic events include biallelic deletions or mutations of the TNFRSF17 gene, γ-secretase-mediated shedding of surface BCMA, and clonal selection of antigen-low variants under therapeutic pressure. In parallel, T-cell dysfunctions, such as senescence, exhaustion, and reduced memory-like subsets, are common in heavily pretreated patients and may limit CAR-T persistence and expansion. Finally, the bone marrow microenvironment imposes additional barriers through suppressive immune cells and inhibitory cytokines.^4,5 

The dismal prognosis after relapse following BCMA-directed CAR-T therapy highlights a critical gap in current treatment algorithms. In real-world cohorts, median progression-free survival in this setting often falls below 4 to 6 months.⁶ Although novel antigens such as GPRC5D, FcRH5, SLAMF7, and CD229 have entered the therapeutic arena, most are variably expressed across disease stages and patient subgroups.² 

The search for optimal targets remains a major challenge in myeloma. True tumor-specific antigens are rare. Most candidates fall into the category of tumor-associated antigens, whose expression in healthy tissues may result in on-target/off-tumor toxicity, as seen in other settings. Furthermore, some targets such as CD70 are also present on activated T cells, raising concerns of fratricide, unless gene editing is used. The discovery pipeline remains slow and validation intensive, which highlights the value of rational and structure-informed approaches such as that used by Kasap et al.

In this setting, CD70 has emerged as a promising immunotherapeutic target. The authors demonstrate that CD70 is preferentially expressed in high-risk MM subtypes such as t(4;14) and 1q gain, but not in TP53-mutated cases, suggesting a selective expression profile that may allow therapeutic windowing. Importantly, CD70 was virtually absent in healthy hematopoietic progenitors and in precursor plasma cell disorders, reinforcing its therapeutic safety. CD70 was also retained in anti-BCMA-resistant cell lines.

Mechanistically, CD70 is regulated by NSD2, a histone methyltransferase often upregulated in t(4;14) disease, and by the transcription factor TFAP2A, both validated via CRISPR-based knockouts. Notably, hypomethylating agents such as azacitidine can upregulate CD70 expression, suggesting a role for epigenetic priming. Still, transient expression on immune cells necessitates CD70 knockout in T cells to avoid fratricide.⁷ Preclinical models have also shown that antigen-driven exhaustion is a risk in this setting, reinforcing the need for engineering refinements.

A defining feature of this study lies in the structure-guided engineering of the CAR. Instead of a conventional single-chain variable fragment, they used the extracellular domain of CD27, CD70’s natural ligand. This design produced 80- to 100-fold superior expansion in vivo, without compromising specificity. Domain dissection confirmed that the CRD1 subdomain was essential for function, a finding divergent from prior assumptions. Ligand-based CARs may reduce immunogenicity, improve synapse formation, and produce more predictable expansion, consistent with recent data in acute myeloid leukemia and lymphoma.⁸ 

Although this expansion is biologically favorable, it may increase the risk of cytokine release syndrome, neurotoxicity, or off-tumor effects, especially given basal CD70 expression in activated immune cells.⁸ In the CTX130 trial, a CRISPR-edited allogeneic CD70 CAR-T, cytokine release syndrome occurred in 67% of patients, though no specific tissue toxicities were noted.⁹ CD27-based CARs, by virtue of their potent expansion and persistence, may require tightly controlled dosing, safety switches, or affinity tuning. Interestingly, their robust expansion occurred even without endogenous CD27 expression, suggesting intrinsic signaling advantages yet to be fully defined.

To further address the challenge of antigen escape, the authors developed a bicistronic CAR construct targeting both CD70 and BCMA. This approach preserved efficacy in models expressing either or both antigens and may preempt clonal escape mechanisms well described after BCMA monotherapy.¹⁰ Unlike tandem CARs or pooled products, bicistronic constructs offer coexpression in every cell, simplifying manufacturing and ensuring redundancy. The nonoverlapping expression of CD70 and BCMA among patients suggests broader applicability, especially for those previously treated with BCMA-targeted agents.

The work by Kasap et al provides a compelling preclinical rationale to bring CD70-targeted CAR-Ts into the clinic, particularly for patients with high-risk MM who relapse after BCMA-directed therapy, a population with limited treatment options and poor survival outcomes. The structure-based design, favorable expansion kinetics, and dual-targeting strategy position this CAR platform as a potential next-generation cellular immunotherapy.

However, several important limitations must be acknowledged. First, the absence of validation in immunocompetent or humanized models precludes definitive conclusions about safety, persistence, and potential off-tumor toxicity. Second, although the dual-target CAR demonstrates impressive preclinical efficacy, the relative contribution of each specificity, and potential dominance or interference, has yet to be disentangled. Third, the requirement for gene editing (CD70 knockout) during manufacturing adds complexity and may limit rapid scalability unless integrated into standardized good manufacturing practice.

From a clinical standpoint, future trials will need to address several key questions: how durable are responses in patients with CD70-positive but BCMA-low disease? What is the optimal dosing strategy to balance efficacy and toxicity? Can CD70 expression be reliably used as a biomarker to guide patient selection? And, finally, will epigenetic modulation or other strategies be required to upregulate antigen expression in vivo?

Ultimately, this study underscores the value of rational target selection and structure-guided CAR engineering in expanding the boundaries of cellular therapy. As CAR-T platforms move beyond single-antigen targeting, precision immunotherapy for genetically and epigenetically defined MM subgroups may soon become a clinical reality.

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