Chimeric antigen receptor (CAR) T-cell therapy has had a substantial effect on the treatment landscape of advanced B-cell cancers. CAR T-cell products are approved for the treatment of B-cell acute lymphoblastic leukemia, B-cell lymphomas, and multiple myeloma. Ongoing trials are evaluating CAR T-cell therapy in solid cancers and in a large variety of autoimmune diseases.

CAR T-cell manufacturing involves T-cell apheresis, CAR transduction, and expansion of transduced autologous or allogeneic T cells. To date, approved CAR T-cell products rely on lentiviral transduction methods, but concerns have been raised that integration events may dysregulate gene expression and cause malignant transformation of T cells. A recent study derived from the adverse event reporting system of the Food and Drug Administration (FDA) showed that T-cell non-Hodgkin lymphomas were identified in 17 of 536 reports (3.2%) regarding secondary cancers, which represented 17 of 12,394 (0.1%) of all reports of CAR T-cell therapy. Here, we describe a case of aggressive CAR+ T-cell lymphoma and investigate its lymphomagenesis.

The majority of children with T-cell ALL can be treated with standardized chemotherapy regimens, but those with induction failure or with elevated minimal residual disease after consolidation generally proceed to allogeneic stem-cell transplantation. For patients who relapse after allogeneic stem-cell transplantation, the prognosis is poor, with less than 15% long-term survival. In this context, cellular immunotherapy could potentially lead to deep leukemic clearance, but CAR T strategies against T-cell cancers have been challenging, partly because of the fratricide effects between T cells and the serious attendant risks of protracted T-cell lymphopenia. Fratricide can be avoided during manufacturing in situations in which an antigen is down-regulated after activation of CAR T cells or when subsets of antigen-negative T cells can be harnessed. Alternatively, preemptive cell-engineering steps have included the expression of inhibitory proteins to sequester target antigens such as CD7 and the disruption of CD7 expression by genome editing. The latter has also been combined with simultaneous disruption of T-cell-receptor gene expression and a molecule critical to the synthesis of major histocompatibility complex class II in an attempt to generate universal CAR T cells.

A variety of genome-editing platforms have reached clinical-phase applications for therapeutic T-cell engineering. Protein-guided zinc-finger nucleases, homing endonucleases, and transcription activator–like effector nucleases (TALENs) have been investigated for ex vivo T-cell modification, as have RNA-guided systems based on CRISPR–Cas9. Editing with nucleases, whether to knock out genes by DNA breakage and nonhomologous end-joining repair, or to target transgene insertion by template-mediated homology-directed repair, is associated with chromosomal aberrations. A trial of TALEN-edited CAR19 T cells was recently halted while translocation events involving chromosome 14q (the site of the T-cell receptor alpha constant [TRAC] locus) were investigated. Vector-related genotoxicity (damage to a patient’s DNA caused by viral vectors) has not been a major issue in T-cell therapies, although complications due to malignant transformation appear to have arisen in a trial of transposon-modified CAR19 T cells.

CARs are synthetic receptors that redirect the specificity, function, and metabolism of T cells (Figure 2). CARs consist of a T-cell activating domain (typically including the zeta chain of the CD3 complex) and extracellular immunoglobulin-derived heavy and light chains to direct specificity. These minimal structures, termed first-generation CARs, recognize antigen independently of HLA but do not direct sustained T-cell responses, owing to their limited signaling capability. Chimeric costimulatory receptors, which enhance proliferation and afford antiapoptotic functions in human primary T cells, paved the way for dual-signaling CARs that could effectively direct the expansion of functional T cells on repeated exposure to antigen. These receptors, termed second-generation CARs, enabled the generation of the persistent “living drugs” that are the foundation of current CAR T-cell therapy.

We chose CD19 as our first target not only because of its frequent expression in B-cell leukemias and lymphomas but also because of its broader and higher expression relative to other potential targets, such as CD20 or CD22. Its expression in normal tissues, which is confined to the B-cell lineage, predicted that on-target and off-tumor activity would be limited to B-cell aplasia, a side effect that can be mitigated with immunoglobulin-replacement therapy. We further reasoned that B-cell depletion may preempt a potential antibody response to the CAR, especially its murine components. A single infusion of human peripheral-blood T cells engineered with a CD19-specific CAR was shown to eradicate established lymphomas and leukemias in mice, which prompted pursuit of the clinical translation of CD19 CAR therapy.

Clinical implementation required a reproducible T-cell manufacturing platform, which hinges on effective gene-transfer tools and T-cell culture conditions. Research to restore immune function in patients with human immunodeficiency virus infection or the acquired immunodeficiency syndrome led to the development of reproducible culture techniques. The FDA approved the first applications for an investigational new drug for CD19 CAR therapy in 2007. The first protocols used either gamma-retroviral or lentiviral vectors that encoded CARs that included either CD28 or 4-1BB costimulatory domains.

Figure 2:
Caption: Structure of CARs and T-Cell Receptors.
Description: Panel A shows the structure of a T-cell receptor, which consists of heterodimeric and antigen-specific α and β chains that closely associate with the invariant ε, δ, γ, and ζ chains of the CD3 complex. The T-cell receptor binds to the HLA allele that has a bound peptide derived from a tumor antigen on the target cell. Panel B shows the CAR, which includes the single-chain variable fragment (scFv) that binds to tumor antigens, fused to a spacer and transmembrane domain. The intracellular domain contains costimulatory domains, such as CD28 and 4-1BB and the CD3ζ chain, which drive signal activation and amplification of CAR T cells. S–S denotes disulfide bond.

This is not the first report of a CAR T-cell–derived lymphoproliferative process occurring in the gut after treatment with cilta-cel. However, we now describe a lymphoma harboring a CAR transgene integration into a common cancer-associated gene.

We identified integration of a lentiviral vector in an antisense orientation in the first intron of TP53. Lentiviral vectors show preference for insertion in transcription units, which may disrupt regulatory sequences necessary for gene expression or generate alternate isoforms through splice donor or acceptor sites in the retroviral vector sequence. The results of immunohistochemical analysis and single-cell RNA sequencing were consistent with decreased TP53 transcription and reduced p53 expression in clonal T cells bearing the TP53-integrated CAR vector.

Whether a monoallelic TP53 CAR vector integration confers a predisposition to transformation is unclear. An additional integration in TP53 was detected as a single event in the duodenal-biopsy samples obtained from the patient but was not associated with clonal expansion. Furthermore, in a study involving 58 patients treated with CD19 CAR T cells, 19 unique integration sites were found in the TP53 transcription unit, and none were associated with clonal expansion, which suggests that monoallelic TP53 integration alone is not sufficient to transform CAR T cells. Thus, we sought to identify alternative factors that may have contributed to oncogenesis.

A second dominant insertion was detected in TANGO2, a gene with poorly described function that is associated with developmental disorders but has no known role in oncogenesis. Its transcriptional expression was maintained in our clone.

We further detected a mutation in SOCS1, a known tumor suppressor that regulates cytokine signaling through JAK–STAT pathways. The substitution occurred in the degradation-regulating “SOCS box” domain at a site that has previously been implicated in large B-cell lymphoma. Immunohistochemical analysis confirmed JAK–STAT pathway activation, and the patient had a notable clinical improvement after initiating treatment with ruxolitinib, a JAK2 inhibitor.

Immunohistochemical analysis confirmed JAK–STAT pathway activation, and the patient had a notable clinical improvement after initiating treatment with ruxolitinib, a JAK2 inhibitor. Both TP53 and SOCS1 alterations are commonly identified and can co-occur in patients with peripheral T-cell lymphoma; owing to limitations of available biospecimens, we were not able to conclusively differentiate the relative contribution of TP53 and SOCS1 alterations to the malignant transformation in this patient.

Of note, a preexisting DNMT3A mutation that was found in peripheral-blood samples was not enriched in duodenal-biopsy samples. Mutations in genes that are associated with clonal hematopoiesis, such as the gene encoding tet methylcytosine dioxygenase 2 (TET2) and DNMT3A, have been identified in persisting CAR T-cell clones as well as in post–CAR T-cell lymphomas associated with or without CAR integration. This case shows that preexisting clonal mutations may not necessarily contribute to and are not required for all scenarios of CAR T-cell malignant transformation.

In addition to gene disruption by means of insertional mutagenesis, a CAR could potentially contribute to clonal expansion and transformation by providing sustained activating and prosurvival signals. We previously found that some CARs may cooperate with loss of TET2 to promote highly proliferative clones in an MYC-dependent manner. In these studies, the propensity to sustain lymphoproliferation depended on the CAR design and its costimulatory domain, which raises the possibility that different CAR signaling domains may differ in their potential to cooperate with preexisting mutations or insertional disruptions to promote malignant transformation. For instance, in the context of TET2 disruption, CARs with 4-1BB costimulatory domain had a greater likelihood of association with clonal expansion than the CD28 costimulatory domain.

CAR T-cell–derived lymphomas are rare. It is notable that this patient did not have any obvious masses or radiologic or cytologic abnormalities to initially suggest leukemia or lymphoma, despite a chronic and severe clinical syndrome with persistent diarrhea and weight loss.