CAR T-cell Resistance to Oncogenic Transformation Available to Purchase

In the last decade, six chimeric antigen receptor T-cell (CART) products have been approved by the regulatory agencies for clinical use in the United States and worldwide. All the CART products target lymphoid malignancies, i.e., diseases that originate either from the B cells (lymphomas and leukemias) or plasma cells (multiple myeloma). Anti-CD19 and anti–BCMA-CART have generated deep and often prolonged responses in patients with otherwise resistant chronic lymphocytic leukemia, acute lymphoblastic leukemia (ALL), non–Hodgkin lymphomas, and multiple myeloma. In particular, several studies have shown that CART19 can cure about 30% to 40% of patients with large B-cell lymphoma and follicular lymphoma and that CART-BCMA induce long-term responses in multiple myeloma. Nevertheless, CART therapies are associated with short-term toxicities such as cytokine-release syndrome and immune effector cell–associated neurotoxicity syndrome, which can be severe in some patients (1). Given this prolonged survival, patients might experience mid- to long-term effects that are typically minimal but may include cytopenias related to CART activity, inflammation, and consequent infections.

In November 2023, the FDA first mentioned a potentially increased risk of developing T-cell lymphomas (TCL) after FDA-approved anti-CD19 and anti–BCMA-CART products. A follow-up comment by the FDA indicated that 22 cases of T-cell malignancies were reported in the FDA Adverse Event Reporting System and were diagnosed after five of the six CART products; molecular analyses in three cases showed the CAR transgene in the malignant clones (2). Indeed, an abstract from the American Society of Hematology (3) described a case of TCL occurring after the CART-BCMA ciltacabtagene autoleucel that also expressed the CAR in tumor cells. Molecular analyses revealed the presence of genetic mutations (TET2, NFKB2, PTPRB, and a germline activating mutation of JAK3), already present prior to CART manufacturing and the CAR insertion in PBX2 locus. We consequently analyzed the risk of second primary malignancies (SPM) in all the commercially treated patients with CART19 and CART-BCMA at the University of Pennsylvania (n = 449; ref. 4). At a median follow-up of 10.3 months, we identified 3.6% of the patients with SPM, which occurred at a median time of 26.4 and 9.7 months after CART infusion for solid and hematologic malignancies, respectively. The projected 5-year cumulative incidence was 15.2% for solid and 2.3% for hematologic malignancies. We also identified a case of TCL occurring 3 months after axicabtagene cilolecucel in the context of a lung cancer; the TCL comprised CD8⁺ cytotoxic cells but, of note, was CAR19 negative (4).

At this point, the reported cases of SPM are observational and do not take into account the phenomenon of immortal time bias (5). Although we are still gathering more evidence on the actual incidence of TCL post-CART and the actual occurrence of CAR positivity, in this opinion piece, we sought to discuss the current evidence of the potential risk and mechanisms coming into play in the potential CART-to-TCL transformation and highlight potential future strategies to mitigate this risk. It is critical to stress the crucial difference in the occurrence of CAR-expressing TCL, which so far has been rare, and that of CAR-negative TCL, which is likely unrelated to the CART manufacturing/administration and has been described in heavily pretreated patients with cancer after chemotherapy and radiotherapy.

Any genetic modification of human cells using viruses such as retro/lentiviruses or gene-editing systems carries an intrinsic risk for malignant transformation through multiple mechanisms. For example, the human T-lymphotropic virus type 1 is a retrovirus that can cause cancer in humans, specifically adult T-cell leukemia/lymphoma. Retroviruses can induce cancer through several mechanisms, including the integration of viral DNA into the host genome, which may lead to the activation of oncogenes or the inactivation of tumor suppressor genes, inducing chronic inflammation that creates a favorable environment for cellular transformation and disruption of normal cellular regulation by viral proteins that promote malignant transformation. Despite these mechanisms, the transformation of target cells by retroviruses is a rare event and involves a complex interplay of viral and host factors. A formative example of retrovirally mediated transformation is the gene editing of hematopoietic stem cells (HSC) from patients with X-linked severe combined immunodeficiency or Wiskott-Aldrich Syndrome through retroviral vector-mediated gene transfer. A subset of the treated cohort subsequently developed T-cell leukemia or myelodysplasia/myeloid leukemia because of the vector’s integration proximal to the LMO2 or MECOM oncogenes, respectively. This transformation risk has led to the refinement of vector technologies, fostering the advent of lentiviral vectors.

Lentiviruses, a subclass of retroviruses, have been widely adopted in gene therapy and research because of their ability to integrate into the genome of nondividing cells, which is a valuable trait for stable gene expression. The risk of transformation by lentiviruses is considered lower than by retroviruses, given the minimal likelihood of insertional mutagenesis. In particular, self-inactivating lentiviral vectors contain modifications to the long terminal repeats that significantly reduce the promoter activity of the integrated DNA, decreasing the risk of activating nearby oncogenes. Moreover, lentiviral integration profiles tend to be less prone to target gene-dense regions and regulatory elements compared with γ-retroviral vectors, which further mitigates the risk of insertional oncogenesis.

Most recently, transposons and nuclease-based gene editing, such as CRISPR–Cas9, base editing, and zinc fingers have reached clinical testing. These systems are fundamentally different from viral-based systems given the fact that they do not necessarily require a virus for ex vivo delivery. Transposons are sequences of DNA that can move from one location in the genome to another. This ability makes them intriguing tools for genetic research and potential therapeutic applications. In clinical and therapeutic contexts, transposon systems like the Sleeping Beauty transposon system and the piggyBac transposon system have garnered attention for their potential to stably integrate therapeutic genes into the genome of patient cells, bypassing the need for a viral vector. However, in 2021, two cases of CAR+ TCL following infusion of piggyBac transposon–engineered CD19-directed CART were described. Analysis of these cases showed a CAR-expressing CD4⁺ TCL with extremely high transgene and altered genomic copy numbers and point mutations. In both cases, the post-CART TCL progressed and resulted in the death of one of the patients, whereas the other patient successfully received an allogeneic transplant (6).

Most recently, nuclease-based gene editing, such as CRISPR–Cas9, base editing, and zinc fingers have reached clinical testing. These systems are fundamentally different from viral-based systems given the fact that they do not necessarily require a virus for ex vivo delivery. CRISPR–Cas9 is renowned for its ability to induce double-strand breaks (DSBs) at specific genomic locations, allowing for precise gene editing. However, the process of repairing these DSBs, primarily through nonhomologous end joining or homology-directed repair, can lead to insertions or deletions that might activate oncogenes or inactivate tumor suppressor genes. Additionally, off-target effects, in which CRISPR–Cas9 edits unintended parts of the genome, represent a significant oncogenic risk if these edits disrupt regulatory elements or critical genes involved in cell cycle control or apoptosis. Zinc finger nucleases are engineered proteins that create DSBs in DNA at specific locations, facilitating targeted genome editing. Compared with CRISPR-Cas9, zinc finger nucleases tend to be designed and optimized for higher specificity, potentially reducing the risk of off-target effects. Base and prime editing, derivatives of the CRISPR–Cas9 system, allows for the direct conversion of one DNA base pair into another without inducing DSBs. This potentially reduces the risk associated with error-prone repair mechanisms; however, base editors can still have off-target activity, potentially leading to unwanted base conversions elsewhere in the genome. The potential oncogenic risk associated with genome-editing technologies such as CRISPR-Cas9 is an area of ongoing research and concern within the scientific community. There is no evidence so far of the risk of transformation related to these systems. However, preclinical reports have stressed the importance of the timing of gene editing during CART manufacturing, with earlier editing of resting T cells leading to decreased chromosomal alterations. Furthermore, studies have shown that CRISPR–Cas9 induces double-stranded DNA breaks and enrichment for cells deficient in p53, potentially leading to chromosomal instability and the risk of transformation. Lentiviruses and now CRISPR–Cas9 have been widely adopted in the clinical setting of gene therapy for monogenic disorders such as beta thalassemia and sickle cell disease (four lentivirally and one CRISPR–Cas9 engineered FDA-approved HSC-edited products), CART immunotherapy (four of six FDA-approved products), and are now being tested for in vivo gene addition and genome editing in multiple settings.

Ultimately, the factors that determine the risk of transformation mediated by retroviruses, transposons, and nucleases involve specific insertion site biases, off-target gene editing, and the number of copy integrations. For example, we previously described a patient with chronic lymphocytic leukemia treated with CART19 (CTL019) who experienced abnormal CART expansion that included 94% T-cell clonality. Clonal CART cells were characterized by lentiviral insertion of the CAR transgene in the TET2 gene, resulting in its disruption and consequent promotion of CART proliferation (7). In another unique report, a pediatric patient with ALL treated with CD22-directed CART showed significant clonal expansion of the CART cells because of lentiviral insertion of the CAR in the CBL locus (8). The aforementioned patient receiving CART-BCMA and then developing CAR+ TCL had a single insertion site in the PBX2 gene present in the vast majority of CAR+ TCL cells, thereby suggesting an active role of that specific gene in transformation (3). In a recent study, genes at integration sites enriched in CART19 responders were commonly found in cell-signaling and chromatin modification pathways, suggesting that insertional mutagenesis in these genes promoted therapeutic T-cell proliferation.

Furthermore, another mechanism that could potentially trigger T-cell proliferation in CAR-transduced cells is tonic signaling. Indeed, tonic signaling might promote the growth of a neoplastic clone that is already prone to survive and proliferate. Of note, some costimulatory domains, such as CD28, lead to short persistence and, therefore, might be associated with lower risk of CAR+ lymphoma compared with long-persisting costimulatory domains (e.g., 4-1BB).

Lastly, gene editing of specific cell types might lead to a higher or lower risk of transformation. For example, HSCs have high plasticity and potency that might facilitate transformation, whereas macrophages or NK cells are typically short-lived and are unlikely to transform. As mentioned above, retrovirally mediated transduction of HSCs led to the development of T-ALL in children with X-linked severe combined immunodeficiency. There are no reports so far on the risk of transformation associated with lymphotropic virus in HSCs, although more patients are being treated, and close monitoring will be required to assess the actual risk.

Together with the gene editing causes of transformation, several preexisting or cell-extrinsic factors can drive or facilitate T-cell lymphomagenesis. Focusing on CART immunotherapy, the occurrence of SPM, including TCL, in patients with a previous diagnosis of hematologic malignancies has been reported (9). Furthermore, simultaneous or metachronous occurrences of two hematologic cancers have been described. Indeed, chemotherapy and radiotherapy cause DNA damage that can drive neoplastic transformation, typically leading to myeloid neoplasms. Patients receiving autologous or allogeneic stem cell transplantation have an even higher risk of SPM.

The recent reports of the development of CAR+ TCL post CART do not necessarily imply that the neoplastic clone derived after the transduction with the CAR vector. Indeed, although the CAR construct can integrate within DNA regions that can lead to transformation, typically cancer generates after multiple hits. Therefore, it is plausible to assume that a preneoplastic or neoplastic T-cell clone carrying mutations preexisted the development of CAR+ TCL. For example, in the report from Australia, the mutations present in the CAR+ TCL were already present in the peripheral blood before CART treatment (3). Circulating cancer cells might survive undetected for several months or years before diagnosis. Therefore, it is likely that a preexisting clone was collected during apheresis and accidentally transduced during manufacturing. Interestingly, the CAR insertion site in the CAR+ TCL was a single one involving the PBX2 gene. Therefore, either a single preexisting neoplastic cell was transduced, or of the several preexisting clonal cells, the one receiving that specific integration site transformed. In this latter situation, the insertion site would have an active role in driving or promoting the transformation. Therefore, a risk factor for transformation could be underlying clonal hematopoiesis related to older age and prior exposure to chemotherapy. Indeed, patients with lymphoma and myeloma have high rates of clonal hematopoiesis, including up to 20% to 60% of patients receiving CART (10, 11). Moreover, in our published case of a CD8⁺ TCL after CART19 previous exposure to seven cycles of pembrolizumab and the consequent development of autoimmunity might have played a role in stimulating a possible preexisting TCL clone (12).

Moreover, as mentioned, inflammation and signaling might play a role in driving oncogenesis of preexisting clones. Clonal hematopoiesis and chronic inflammation are closely related and can contribute not only to the accumulation of mutations but also the support of cancer survival in an inflamed environment. T cells with different CAR costimulatory domains drive different cytokine release patterns and levels of inflammation (overall higher in CD28- vs. 4-1BB–costimulated CART) that might contribute to progression of preneoplastic clones. Furthermore, in our published case of CD8⁺ TCL occurring 3 months after CART19 infusion (4), the TCL were spatially close to a lung cancer which might hint at the possibility of reactivity against the NSCLC as a stimulating factor. Morevoer, despite the absence of high levels of CAR transgene in the TCL, the timing and the fact that it displayed an uncommon cytotoxic immunophenotype might suggest a contribution of T-cell activation during CART manufacturing or post-CART inflammation. More studies are needed to determine whether T-cell activation and expansion during manufacturing could drive the transformation of JAK3-mutated T cells that also receive TCR stimulation (13).

The issue of the potential increased risk of TCL after CART still requires more investigation into the actual incidence, the causal mechanisms, including culprits discussed above and summarized in Fig. 1. It is important to recall that associations do not imply causality. In our cohort at the University of Pennsylvania we did not identify an increased risk of TCL post CART as compared with a comparable population of patients with cancer receiving chemotherapy. The occurrence of CAR+ TCL is still anecdotal and we need peer-reviewed reports and prospective studies. Although the presence of CAR transgene in the majority of the TCL cells would suggest a driving role, deep analyses are needed to fully understand the relative contribution of the CAR transduction versus germline mutations and previous chemotherapy. There is certainly a need for transparency (2, 4, 14) and large independent databases to assess the actual risk. We also would like to emphasize the importance of continued research, biobanking, transparent reporting, and participation in post-authorization safety studies.

The risk of viral-induced oncogenesis necessitates a careful risk–benefit analysis in the context of gene therapy applications, especially for nonlife-threatening conditions, such as autoimmune disease. Ongoing research aims to further understand lentiviral integration mechanisms and to develop safer vectors, such as those targeting specific genomic loci for integration, to minimize the risk of insertional mutagenesis. Additionally, comprehensive preclinical testing and long-term follow-up of patients treated with lentiviral vector–mediated therapies are critical to ensure their safety and efficacy.

To address these oncogenic risks, significant efforts are being made to improve the specificity and efficiency of genome editing tools. Advanced computational models, high-fidelity Cas9 variants, and enhanced base editor versions are being developed to minimize off-target effects. Moreover, comprehensive in vitro and in vivo studies, along with careful monitoring of treated patients, are essential for assessing long-term safety and oncogenic potential. Despite the theoretical risks, the precise and targeted nature of these genome editing technologies presents a favorable risk–benefit profile for many potential applications, particularly for diseases with no other effective treatments. Ongoing research and clinical trials continue to refine these methods to maximize their therapeutic potential while minimizing associated risks.

Nevertheless, strategies to reduce SPM and TCL are warranted and include treatment of patients earlier in the course of their disease to avoid repetitive exposure to chemotherapy and/or radiation, surveillance for clonal hematopoiesis, reduction of actionable risk factors (e.g., smoking), and close clinical monitoring for possible occurrence. For example, strategies to reduce the intensity of lymphodepleting chemotherapy might be beneficial also in reducing the risk of SPM, although there are no data available (15). The insights garnered from these adverse events have significantly influenced the trajectory of gene therapy research, underscoring the imperative for rigorous, long-term surveillance of gene therapy recipients and the innovation of vector systems that mitigate the risk of insertional oncogenesis. Furthermore, to address the potential risk of CAR+ TCL, the design of next-generation products might include suicide systems, such as iCasp9 that can be triggered in the rare case of transformation. In this regard, we previously hypothesized that the CAR itself could become a specific target for transformed CART cells and that an anti–CAR CART product can effectively kill CART cells. Ultimately, it has not been determined whether these systems would be effective in transformed CAR T cells and therefore, more preclinical studies are needed.

In conclusion, the observed very low incidence of secondary TCL should provide reassurance to the scientific community about the safety of commercially available autologous CART products (5, 14). The safety of allogeneic CAR T cells requires further study (6). This aligns with the FDA’s assertion that “[...] the overall benefits of these products continue to outweigh their potential risks for their approved uses [...]” (2).

M. Ruella reports grants and personal fees from viTToria Bio, nonfinancial support from Beckman and ORFLO, personal fees from GSK, Bristol Myers Squibb, Sana, NanoString, GLG, and Guidepoint, and grants from AbClon and ONI during the conduct of the study, as well as multiple patents on CART Immunotherapy managed by the University of Pennsylvania that are pending, issued, licensed, and with royalties paid from viTToria Bio, Novartis, and Tmunity. C.H. June reports other support from Kite Gilead and Novartis outside the submitted work, as well as patent in the Field of CAR T Cells issued, licensed, and with royalties paid from Novartis and Kite.

C.H. June: Conceptualization, writing, reviewing, editing. M. Ruella: Conceptualization, writing, reviewing, editing.

The authors would like to thank all patients and colleagues. We extend our appreciation to the Institutional Review Boards, the Institutional Biosafety Committee, and the Clinical Cell and Vaccine Production Facility at the University of Pennsylvania Center for Cellular Immunotherapies. This study was supported by NIH/NCI R37-CA-262362-02 (M. Ruella) and P01 CA214278 (M. Ruella and C.H. June); the Leukemia & Lymphoma Society SCOR grant (C.H. June); the Laffey-McHugh Foundation (M. Ruella); Parker Institute for Cancer Immunotherapy (C.H. June); and the Berman and Maguire Funds for Lymphoma Research at Penn (M. Ruella).