CD19/CD20 Bispecific Chimeric Antigen Receptor (CAR) in Naive/Memory T Cells for the Treatment of Relapsed or Refractory Non-Hodgkin Lymphoma

Effective chimeric antigen receptor (CAR)-T cell therapy exerts a strong selective pressure against cancer cells that express the CAR-targeted antigen, and downregulation or loss of expression is the natural escape route for target antigens that are not critical to cell survival. Accurate quantification of relapse rate attributable to CD19 antigen escape is complicated by lack of tissue acquisition following relapse, and reported CD19-negative relapse frequencies range from 27% to 100% of relapsed cases among patients with leukemia and lymphoma (1–5). The frequency of cases with CD19-negative relapse demonstrates the susceptibility of CD19 to antigen loss and points to the identification of alternative target antigens that are more resistant to gene expression downregulation as a potential remedy.

To address the problem of CD19 antigen escape, we developed a CD19/CD20 bispecific CAR-T cell therapy, and previously demonstrated its ability to eradicate B-cell lymphoma with heterogeneous CD19 expression and prevent relapse in mouse models of human lymphoma (6, 7). CD19/CD20 bispecific CAR-T cells outperformed single-input CD19 CAR-T cells in achieving long-term, progression-free survival (PFS) in a lymphoma xenograft model (6, 7). CD20, like CD19, is a pan–B-cell marker, and the first-line therapy for B-cell malignancies typically includes an anti-CD20 antibody such as rituximab (8). In fact, rituximab is commonly included in each subsequent line of chemotherapy administered to patients with NHL, yet CD20 antigen loss is a low-frequency event despite repeated cycles of CD20-targeted therapies (9), suggesting CD20 may be a suitable CAR target with a low propensity for antigen escape. However, the clinical outcomes of CD20 CAR-T cell therapy have been uneven to date (10–13), resulting in more limited clinical advancement compared with CD19 CAR-T cell therapy. We hypothesized that simultaneously targeting CD19 and CD20 would enable both a high initial response rate and increased resistance to antigen escape. Importantly, dual targeting of CD19 and CD20 would not increase on-target, off-tumor toxicity compared with either CD19 or CD20 single-input CAR-T cell therapy because both CD19 and CD20 are B cell–specific markers, thus limiting the off-tumor toxicity to healthy B cells whose aplasia is a clinically manageable condition (14).

Multiple trials have reported an association of CD19-positive relapses with T-cell exhaustion and lack of CAR-T cell persistence, and patients with durable response to therapy have relatively elevated levels of naive and memory T (T_N/MEM) cells (4, 5, 15–18). The concept of infusing T cells enriched in naïve and memory phenotypes is consistent with prior reports indicating T_N/MEM cells enhance in vivo CAR-T cell function (19–21). Therefore, we hypothesized that engineering T_N/MEM cells to express our molecularly optimized CD19/CD20 bispecific CAR would yield a therapy, termed CART19/20, with strong efficacy coupled with improved CAR-T cell persistence and reduced CAR-T cell exhaustion. Here, we report the dose-escalation phase I trial evaluating CART19/20 in adults with relapsed/refractory NHL and demonstrate durable responses and strong safety profiles in patients treated with CART19/20 cells.

We generated a CD19/CD20 bispecific CAR consisting of a single-chain variable fragment (scFv) derived from the anti-CD20 monoclonal antibody Leu16 fused to a second scFv derived from the anti-CD19 monoclonal antibody FMC63, followed by fusion of the scFv domains to the hinge domain of human IgG4, the transmembrane domain of human CD28, and the cytoplasmic signaling domains of human 4-1BB and CD3ζ (Fig. 1A; refs. 6, 7). The bispecific CAR was encoded by a third-generation self-inactivating lentiviral vector under the control of an elongation factor 1 alpha (EF1-α) promoter (22). We planned a phase I cell dose-escalation trial with a fixed lymphodepletion chemotherapy of fludarabine 30 mg/m² daily for 3 days on day –5 to day –3 and cyclophosphamide 500 mg/m² daily for 3 days on day –5 to day –3, followed by CART19/20 infusion on day 0 with dose levels of 50 × 10⁶ CAR⁺ T cells [dose level 1 (DL1)], 200 × 10⁶ CAR⁺ T cells (DL2), and 600 × 10⁶ CAR⁺ T cells (DL3), with each DL allowing ±30% range (Fig. 1B).

Seventeen patients were screened, 11 patients went onto leukapheresis (Fig. 1C), and 10 patients received CART19/20 infusion in two cohorts (DL1, n = 7; DL2, n = 3; Fig. 1C). The median age at the time of CART19/20 infusion was 59 years (range, 29–70; Table 1). The diagnoses were mantle cell lymphoma (MCL; n = 1), follicular lymphoma (FL; n = 3), de novo diffuse large B-cell lymphoma (DLBCL; n = 1), transformed FL to DLBCL (n = 3), primary mediastinal B-cell lymphoma (PMBCL; n = 1), and high-grade B-cell lymphoma (HGBCL; n = 1) with BCL6 and cMYC double-hit rearrangement. All patients with FL had progression of disease within 24 months after first-line treatment (POD24). The median lines of prior therapy were 3.5 (range, 2–4). One patient (patient 004) was refractory to prior anti-CD19 bispecific T-cell engager (BiTE) therapy. All patients were CAR naive and had stage 4 disease at the time of CART19/20 treatment. Nine patients were given bridging therapy prior to infusion due to progressive disease (Table 1).

As of the data cutoff on July 11, 2022, a total of 10 patients were evaluable for response. Nine patients were evaluable for dose-limiting toxicity (DLT), including six treated at DL1 and three treated at DL2. A decision was made not to escalate to DL3 based on the strong efficacy outcomes observed at the two lower dosing levels. The maximum tolerated dose was not reached.

Patient leukocytes harvested from leukapheresis were enriched for cells expressing CD62L, a marker for T_N/MEM cells, by magnetic bead–based cell separation. Leukapheresis products with greater than 5% CD14⁺ and/or CD25⁺ cells among viable singlets based on flow cytometry analysis were subjected to an additional CD14/CD25 depletion step to remove myeloid cells and regulatory T cells (Treg), respectively, prior to CD62L enrichment. T_N/MEM cells were activated with an anti-CD3/CD28 colloidal nanomatrix-based activation reagent, lentivirally transduced to express the CD19/CD20 bispecific CAR, and expanded ex vivo for a total of 12 days (n = 9), 14 days (n = 1, patient 016), or 16 days (n = 1, patient 007) in the presence of IL2 and IL15 prior to cryopreservation (Fig. 2A). One manufacturing failure occurred due to low CAR⁺ T-cell counts that did not meet dose requirements (patient 007).

The CART19/20 cell products have a high level of CD3⁺ purity, but a wide range of transduction efficiency was observed (Fig. 2B). In addition to substantial donor-to-donor variability, the specific lot of Good Manufacturing Practice (GMP)–grade lentivirus used also affected transduction efficiency, although the difference was below statistical significance (Fig. 2C; P = 0.058 by two-tailed Student t test). Despite the variability in transduction efficiency, only one manufacturing campaign failed to generate a sufficient dose for infusion. Patient 007 had a low absolute lymphocyte count (ALC) of 0.16 × 10³ cells/μL at the time of screening and a low white blood cell count (32.79 × 10³/μL) in the leukapheresis product. Furthermore, the patient's cells transduced and expanded poorly during ex vivo manufacturing (Fig. 2A and B), resulting in failure to meet the required CART19/20 cell dose for infusion.

Overall, CART19/20 cell products contained a substantial fraction of central memory T (Tcm) cells (median: 29.3%; range, 3.6%–74.9%; Fig. 2D), indicating the retention of memory phenotype in cell products manufactured from T_N/MEM cells. Of note, CAR-expressing T cells tend to have slightly higher Tcm content (median: 40.9%; range, 5.3%–80.1%) compared with the overall T-cell population (Fig. 2E). A breakdown of CD4⁺ versus CD8⁺ subtype distribution reveals that CAR⁺ T cells tend to have a higher percentage of CD4⁺ than the total T-cell population (Supplementary Fig. S1A), and CD4⁺ T cells tend to be more enriched in the Tcm phenotype compared with CD8⁺ T cells, although the difference is just under statistical significance (Supplementary Fig. S1B; P = 0.055 by a two-tailed Student t test).

We chose to incorporate CD14 depletion in order to minimize the presence of myeloid cells, which had been reported to reduce T-cell activation through phagocytosis of activation agents (23) and could potentially reduce transduction efficiency by competing with T cells for lentivirus uptake. The removal of immunosuppressive Tregs through CD25 depletion (24) aimed to further enhance the antitumor efficacy of CART19/20 products. The minimum threshold of ≥5% CD14⁺ and/or CD25⁺ cells for depletion was based on the empirical observation that up to 5% of antigen-positive cells can remain even after depletion during preclinical process development. Following this criterion, the leukapheresis products of patients 001 and 003 were not subjected to depletion and proceeded directly to CD62L enrichment (Supplementary Fig. S1C). However, the postisolation cell population for these two patients showed a notable increase in percentage of CD14⁺ (from 3% and 2% to 48% and 74%, respectively), together with an uptick in percentage of CD25⁺ (from 0.9% and 0.3% to 8% and 4%, respectively; Supplementary Fig. S1D). This unintended enrichment is consistent with the fact that a substantial fraction of CD62L⁺ cells in patient leukapheresis products is CD14⁺ and/or CD25⁺ (Supplementary Fig. S1E); thus, the CD62L enrichment step would simultaneously result in selective retention of CD14⁺ and/or CD25⁺ cells. Furthermore, the adherent nature of myeloid cells may also facilitate the retention of CD14⁺ cells during bead-based cell sorting in the absence of a depletion step. Based on these observations, the protocol was amended to trigger CD14/CD25 depletion when ≥5% of CD62L⁺ cells (as opposed to ≥5% of viable singlets) were CD14⁺ and/or CD25⁺, starting with the product for patient 004.

Despite the fact that the postisolation cell population for patients 001 and 003 contained a substantial number of myeloid cells, the CART19/20 cultures for these two patients showed typical fold expansion and viability levels during the manufacturing process (Fig. 2A). Both final products exhibited high levels of CD3⁺ purity and CAR transduction efficiency (Fig. 2B), and had similar T-cell subtype distribution at the time of cryopreservation as products made from CD14/CD25-depeleted cells (Fig. 2D and E). The only clear difference was that products generated from nondepleted starting material were CD4 dominant, whereas products generated from CD14/CD25-depleted cells were CD8 dominant (Supplementary Fig. S1A), potentially due to myeloid cells’ ability to stimulate CD4⁺ cells through MHC-II antigen presentation. Taken together, these results indicate the culture conditions succeeded in selectively expanding T cells while avoiding the negative impacts of myeloid cells observed in the manufacturing processes described in previous reports (23, 25, 26). Given that the starting percentage of CD25⁺ was <1% in both patients 001 and 003, there was minimal concern regarding Treg enrichment. Indeed, FOXP3 intracellular staining showed no clear correlation between CD25 depletion or lack thereof with the level of FOXP3 expression among CD4⁺ T cells in CART19/20 final products (Supplementary Fig. S2).

Among the subsequent products, two more were manufactured in the absence of CD14/CD25 depletion (patients 010 and 017). Both products were enriched in CD4⁺ T cells, with no distinctive pattern in ex vivo expansion rate, transduction efficiency, or T-cell subtype distribution compared with CD14/CD25-depleted products, consistent with initial observations made with patients 001 and 003 (Fig. 2).

Ten patients received CART19/20 cell products. One patient treated at DL1 (patient 003) progressed prior to the end of the DLT monitoring period and was replaced per protocol and excluded from DLT evaluation. Infusions were generally well tolerated; two patients experienced a grade 1 infusion-related reaction (IRR; Table 2). Grade 1 cytokine release syndrome (CRS) occurred in six patients; no grade 2 or higher CRS was observed in any patient. The median time from infusion to CRS was 8 days (range, 1–11), and the median duration was 2.5 days (range, 1–3). One dose of tocilizumab was given to patient 009 for grade 1 CRS lasting greater than 48 hours. There were no cases of immune effector cell–associated neurotoxicity syndrome (ICANS), and no steroids were administered in the study for CRS or ICANS management. All patients experienced grade 3 or above adverse events consisting of generalized pain (n = 2, 22%), hypotension (n = 1, 11%), anemia (n = 3, 33%), neutropenia (n = 5, 56%), and thrombocytopenia (n = 5, 56%). One episode of grade 3 hypotension occurred during bridging therapy prior to CART19/20 infusion for patient 002. The only grade ≥3 adverse events attributable to CART19/20 cells were anemia, thrombocytopenia, and neutropenia experienced by patient 009.

In contrast to the other grade ≥3 hematologic adverse events, the cytopenia in patient 009 persisted beyond the expected recovery time from lymphodepletion chemotherapy and resulted in the only DLT observed in this study. Patient 009 had received autologous stem cell transplant (ASCT) 11 months prior to receiving CART19/20 cell infusion and exhibited elevated levels of multiple cytokines before infusion (Supplementary Fig. S3A) as well as a sustained increase in C-reactive protein (CRP) and ferritin levels after infusion (Supplementary Fig. S3B and S3C). A bone marrow biopsy performed 5 months after CART19/20 cell infusion revealed an increase in frequency of a preexisting TET2 mutation and a new ASXL1 mutation, suggestive of evolution of a myeloid neoplasm in the bone marrow related to prior therapy. Patient 009 elected to cease medical treatment shortly before the 12-month follow-up assessment, and the death from grade 5 hypocellular marrow is considered possibly related to treatment and possibly related to secondary malignancy.

Primary response assessment was performed 60 days after CART19/20 infusion by PET/CT scan. Nine of 10 patients responded to therapy [overall response rate (ORR) = 90%]. Seven patients achieved a complete response (CR) by the first disease assessment at day 60 (CR rate = 70%), and two additional patients had a partial response (PR) at day 60 (Figs. 3 and 4A). Bridging therapy did not result in a significant reduction of tumor burden in the majority of cases (Fig. 4A; Supplementary Fig. S4), and there was no correlation between patient response and application of bridging therapy (Table 1). With a median follow-up of 17 months from time of CART19/20 cell infusion (range, 2–30 months), the median PFS was 18 months, and the median overall survival (OS) was not reached (Fig. 4B).

The one patient who did not respond to therapy (patient 003) was diagnosed with PMBCL (Fig. 4C) and had low ALC at the time of screening (0.17 × 10³ cells/μL). The patient exhibited marked clinical improvements, including reduced dyspnea and pain, in the week after CART19/20 cell infusion and was discharged from the hospital. However, disease progression was detected 14 days after CART19/20 infusion in a bone marrow biopsy performed as part of a routine evaluation for CAR-T cell infiltration and expansion. The screening biopsy of a supraclavicular lymph node from patient 003 was positive for CD19, CD20, CD30 (patchy), BCL2, BCL6, and cMYC (Fig. 4D), with kappa light chain restriction; the screening biopsy was negative for CD10 and BCL1. In contrast, the bone-marrow sample obtained 14 days after CART19/20 infusion expressed CD30 and weak BCL2 and was negative for CD10, CD19, CD20, BCL1, BCL6, and cMYC (Fig. 4D), indicating a clonal shift in the tumor population. A biopsy of the postrelapse lung mass analyzed by bulk RNA sequencing (RNA-seq) showed loss of gene expression for CD19, CD20, and BCL6 [reads per kilobase transcript, per million mapped reads (RPKM) = 0.12, 0.11, and 0.72, respectively], low cMYC expression (RPKM = 2.04), and a lack of CD22 expression (RPKM = 0.07). The concomitant loss of BCL6 and cMYC—two antigens not under selective pressure from the CD19/20 bispecific CAR—within 14 days of CART19/20 cell infusion suggests the possibility of a heterogeneous tumor that contained a preexisting antigen-negative subclone.

The seven patients who achieved a CR include one patient diagnosed with MCL, three patients with DLBCL (one de novo and two transformed FL), and three patients with FL (Fig. 3). All patients with FL were POD24, and the majority of patients were characterized by high tumor burdens (Supplementary Fig. S4). Among the seven patients who achieved a CR, three (patients 002, 009, and 014) had primary refractory disease and a fourth (patient 004) was refractory to anti-CD19 BiTE therapy (Table 1). In addition to anti-CD19 BiTE, patient 004 was also refractory to ROR1-targeted antibody–drug conjugate therapy and progressed through the second to fourth lines of therapy within 5 months (Fig. 3). Flow cytometry analysis of patient 004's peripheral blood at the time of screening indicated the presence of CD20⁺CD19^dim/– cells (Fig. 5A). This population was substantially reduced within 7 days of CART19/20 infusion, suggesting CART19/20's ability to target tumor cells that have downregulated or lost CD19 expression. Patient 004 achieved a CR within 60 days and remained in CR until progressive disease with biopsy-proven CD20⁺CD19⁺ FL was detected at the month-18 evaluation (Fig. 5B). Given the patient's history of therapy resistance and 18-month response to CART19/20, permission was obtained from the FDA to redose patient 004 with 126 × 10⁶ CART19/20 cells using extra aliquots of the same cell product as the first infusion. At day 60 after redosing, the patient again achieved a CR and remains in CR as of the data cutoff (Fig. 5B).

The presence of CART19/20 cells in peripheral blood after infusion was detected by both flow cytometry and droplet digital PCR (ddPCR). CAR copy number quantified by ddPCR consistently peaked at 14 days after infusion (Fig. 5C). Six of 10 patients had detectible CAR-T cells as of the last available datapoint, and all six patients with follow-up periods of 12 months or more had detectible CAR-T cells at or past 12 months (Fig. 5C). At peak expansion, up to 48% of all viable singlets in peripheral blood mononuclear cells (PBMC) were CAR⁺ T cells (median 11.6%; range, 1.5%–48.3%; Fig. 5D; Supplementary Fig. S5A and S5B). At the time of data cutoff, all seven patients who were in PR (n = 1) or CR (n = 6) remained in B-cell aplasia based on flow cytometry analysis of peripheral blood (Fig. 5E), indicating functional persistence of CART19/20 cells.

For patient 004, who remained in CR for 18 months before relapsing, CAR⁺ T cells dropped below ddPCR detection at 12 months but reemerged at the time of relapse (Fig. 5C). A second peak of CAR-T cell signal was detected by flow cytometry in the patient's peripheral blood after the second dose of CART19/20 cells (Fig. 5D). B cells were undetectable in the peripheral blood of patient 004 after the first CART19/20 cell infusion, were detected at the time of relapse, and dropped below detection again after receiving the second dose of CART19/20 cells, consistent with response to therapy (Fig. 5E).

As previously noted, CART19/20 cell products were either CD4 or CD8 dominant at the time of cryopreservation, depending on whether CD14/CD25 depletion was performed (Supplementary Fig. S1A). Regardless of the CD4:CD8 ratio in the cryopreserved cell product, both CAR-expressing T cells and total T cells rapidly became CD8 dominant after CART19/20 cell infusion in all but one patient (Supplementary Fig. S6A and S6B). The CD8 dominance declines over time, mirroring recently reported findings in a decade-long follow-up of two patients with leukemia treated with CD19 CAR-T cell therapy (27).

Consistent with clinical presentation of grade 1 CRS for over 48 hours, patient 009 showed elevated levels of most cytokines in peripheral blood compared with other patients in the trial (Fig. 5F; Supplementary Fig. S3A). As a whole, cytokine levels observed in patients treated with CART19/20 cells are similar to or substantially lower than values reported in earlier trials for single-input CD19 CAR-T cell therapies (3, 5, 28–30) and CD19/CD20 bispecific CAR-T cell therapies (18, 31). The relatively low peak cytokine levels in patients treated with CART19/20 may be a contributing factor to the strong safety profile observed in this trial to date. Taken together, results of this phase I clinical trial indicate CART19/20 cell therapy is safe and effective for the treatment of NHL (Supplementary Table S1).

CD19 CAR-T cell therapy became the first FDA-approved gene-modified cell therapy in 2017 and is making rapid progress toward incorporation in earlier lines of treatment for B-cell malignancies. However, antigen escape and lack of T-cell expansion and persistence remain key factors that limit the frequency as well as durability of response in patients treated with CD19 CAR-T cell therapy (1–5, 15–17, 32, 33). Early results of our phase I trial on CART19/20 cell therapy demonstrate that dual targeting of CD19 and CD20 using a molecularly optimized single-chain bispecific CAR (6, 7) is safe and effective in patients with relapsed/refractory NHL.

In this trial, 10 patients were treated with a median dose of 55 × 10⁶ CART19/20 cells, with no patient receiving more than 165 × 10⁶ cells. These dosing levels are substantially lower than the median dosage level used in clinical trials evaluating other CD19/CD20 bispecific CAR clinical candidates (18, 34) as well as in pivotal trials for single-input CD19 CAR-T cell products including axicabtagene ciloleucel (35, 36), tisagenlecleucel (37, 38), and lisocabtagene maraleucel (39). Despite the low dosage used, strong efficacy was observed in this trial (90% ORR; 70% CR rate) with a heavily pretreated patient population carrying a high tumor burden and highly aggressive disease. To date, all but one patient who achieved a CR have remained in CR. The only patient to experience a relapse did so after 18 months, and the patient subsequently returned to CR after receiving a second dose of CART19/20 cells. With a median follow-up of 17 months, median PFS was 18 months, and the median OS has not been reached in this trial.

Importantly, the strong efficacy observed in this trial was achieved with no ICANS of any grade and no CRS above grade 1 after CART19/20 infusion. This is in contrast to the FDA-approved single-input CD19 CAR-T cell products where the rate of grade 3 or higher ICANS reached as high as 32% in the ZUMA-1 trial (35) and the rate of grade 3 or higher CRS as high as 23% in the JULIET trial (38). To date, the vast majority of grade ≥3 adverse events experienced by patients treated with CART19/20 were toxicities attributable to bridging and lymphodepletion chemotherapies in patients with substantial previous chemotherapy exposure. One patient death (patient 009) from grade 5 hypocellular marrow is considered possibly related to treatment and possibly related to myeloid neoplasm consequent to therapies received prior to CART19/20. Although no evidence of lymphoma was detected in the patient at the time of elected transition to comfort care, this case underscores the potential benefit of providing CAR-T cell therapy as an earlier line of treatment, which could reduce the total amount of chemotherapy exposure and related toxicities to the patient. Overall, the safety profile observed in this trial compares favorably with prior reports of CD19-targeted as well as CD19/CD20-targeted CAR-T cell therapies (18, 34–40) and supports the possibility of combining strong efficiency with a high level of safety.

The trial reported here included one patient with MCL, three patients with FL, and five patients with DLBCL of various subtypes. It is plausible that different DLBCL subtypes have different response rates to CAR-T cell therapy, although no trial reported to date has been statistically powered to evaluate the responses of different DLBCL subtypes to CD19 CAR-T cell therapy (35, 38, 39). FL is often described as an indolent disease and has shown favorable response rates to therapy (37). It should be noted that in this trial, all three patients with FL were POD24, including one patient who was primary refractory (patient 002) and another patient who was refractory to the three lines of therapy immediately before CART19/20, including CD19-targeted BiTE (patient 004). All patients with FL achieved a CR after CART19/20 cell infusion. In addition to four other patients with MCL or DLBCL who also achieved a CR, one patient with primary refractory, double-hit HGBCL (patient 017) achieved a PR at day 60. Overall, CART19/20 cell therapy has shown robust efficacy in a highly pretreated patient population with challenging disease profiles.

The only patient who did not respond to CART19/20 therapy to date (patient 003) had PMBCL refractory to R-ICE (rituximab/ifosfamide/carboplatin/etoposide) salvage chemotherapy. Despite notable clinical improvement in the week after CART19/20 infusion, this patient experienced a rapid emergence of CD19^–CD20^– tumor cells that had also lost BCL6 and cMYC expression within 14 days of CART19/20 treatment. The number of protein expression changes combined with the rapidity of clonal shift suggests a preexisting tumor subpopulation that was able to swiftly expand after CART19/20 cells eliminated the originally dominant CD19⁺CD20⁺ tumor cells. The emergence of a CD19^– tumor was previously reported in the relapse of a patient with PMBCL treated with CD19 CAR-T cell therapy, and sequencing analysis indicated the likely cause was a preexisting clone that expanded under CD19-targeted selective pressure (41).

Promoting T-cell persistence and function through the generation of cell products enriched in naïve and memory cell types was a key element of the CART19/20 therapy design. CART19/20 products were generated from a T_N/MEM cell population obtained through bead-based enrichment of CD62L expression, and the final products retained substantial central memory T-cell content. Contrary to our original expectations, the presence of the CD14⁺ cells did not adversely affect our ability to successfully manufacture CART19/20 cell products with high T-cell purity and clinical efficacy. The only clearly measurable impact of CD14⁺ cell presence was the CD4:CD8 ratio of the final CART19/20 product, with the presence of CD14⁺ cells leading to CD4-dominant products, while the depletion of CD14⁺ cells from the starting material led to CD8-dominant products. However, based on the results to date, there is no correlation between the CD4:CD8 ratio and clinical outcome. Similarly, the lack of Treg depletion through CD25 did not show a measurable impact on treatment outcome, which is consistent with prior reports (42). We note that the T_N/MEM cell population isolated by CD62L enrichment could also include CD45RA⁺CD62L⁺ stem cell memory T cells (T_SCM) cells that have been shown to exhibit increased proliferation and antitumor activities (43, 44). The markers CD45RA and CD62L alone cannot distinguish between purely naive and T_SCM cells, but phenotype profiling of infusion products indicates the cryopreserved cells are predominantly CD45RA^–CD45RO⁺, consistent with exposure to activating agents during the manufacturing process (Fig. 2D and E). The 91% manufacturing success rate in this trial is comparable with those reported for other CAR-T cell therapy trials targeting CD19 or CD19/CD20 (Supplementary Table S1), demonstrating the feasibility of using T_N/MEM cells as starting material for CAR-T cell manufacturing. However, consistency in transduction efficiency is an area of improvement, with a need to reduce donor-to-donor variation as well as variability in the potency of GMP-grade lentivirus used in manufacturing.

A key question of interest is how the CART19/20 cells evaluated in this trial are able to achieve a high level of efficacy at a low dosage level and without incurring the type of toxicities observed in comparable trials. As previously reported, the CD19/CD20 bispecific CAR used here had been optimized at an amino acid sequence level to maximize efficacy (6). Safety was not a consideration in the CAR engineering process. However, the robust efficacy enabled the use of a very low cell dose to achieve therapeutic benefit, and the low cell dose may in turn have contributed to the favorable safety profile observed in this trial. The use of T_N/MEM-derived cells may further contribute to the potency and safety profile of CART19/20 cells by reducing peak cytokine levels while retaining long-term antitumor efficacy.

In summary, early results from this phase I trial indicate that autologous, CD19/CD20 dual-targeting CAR-T cells enriched in naive and memory phenotypes are safe and highly efficacious in the treatment of relapsed/refractory NHL. Our results suggest potent clinical efficacy can be achieved while avoiding severe toxicities typically associated with CAR-T cell therapy and highlight the utility of dual-antigen targeting cell-based immunotherapy.

A prospective, first-in-human phase I clinical trial assessing CART19/20 in adult patients with relapsed/refractory NHL and chronic lymphocytic leukemia (CLL)/small lymphocytic lymphoma (SLL) was initiated at the UCLA Medical Center. The study was approved by the UCLA Institutional Review Board and registered with ClinicalTrials.gov (NCT04007029). Informed written consent was obtained in accordance with the Declaration of Helsinki, the International Conference on Harmonization Good Clinical Practice (GCP), the U.S. Code of Federal Regulations for Protection of Human Subjects, the Health Insurance Portability and Accountability Act, and local regulations. Data monitoring was conducted by the UCLA Jonsson Comprehensive Cancer Center Data Safety and Monitoring Board. The primary endpoint was safety, defined by the incidence and severity of DLTs, as well as determination of the maximum tolerated dose. Secondary endpoints included clinical efficacy measures, and analysis of CAR-T-cell persistence and B-cell aplasia. The exploratory endpoint was CRS analysis.

Patients eligible for the clinical trial were ≥18 years old with DLBCL or PMBCL after ≥2 prior lines of therapy or with MCL, FL, CLL, or SLL after ≥3 prior lines of therapy. Transformed indolent lymphomas, including Richter transformation, were eligible, and previous lines of therapy were considered from the time of transformation. ASCT recipients were allowed in the study. Prior CAR-T cell therapy was an exclusion criterion, but other forms of CD19- or CD20-targeted therapies were allowed. Patients were required to have greater than 30% positivity in malignant cells of CD19 and/or CD20, as well as measurable tumor burden on PET/CT. Any NHL- or CLL/SLL-directed therapy, including cortico­steroids, within 14 days of initiation of lymphodepletion chemotherapy was exclusionary. After leukapheresis, bridging therapy was permitted at the investigator's discretion. Lymphodepletion chemotherapy, consisting of fludarabine 30 mg/m²/day and cyclophosphamide 500 mg/m²/day, was administered for 3 days from day −5 to day −3 prior to infusion. Study representation of underserved communities is summarized in Supplementary Table S2.

Adverse events were recorded for all treated patients until disease relapse or death, with incidence and severity graded using the Common Terminology Criteria for Adverse Events (CTCAE) version 5.0. CRS was graded according to the American Society for Transplantation and Cellular Therapy (ASTCT) and Lee criteria, with the former guiding treatment (45). For neurotoxic events, the ASTCT criteria were for scoring and treatment, with specific guidance to key disorders outlined by Neelapu and colleagues (46).

The clinical response in lymphoma was evaluated with the criteria defined by the Revised Cheson Response Criteria and Lugano Classification (47, 48). The ORR was defined as the total of CRs and PRs.

Fresh patient leukapheresis products were analyzed by flow cytometry to determine the CD3⁺, CD62L⁺, CD14⁺, and CD25⁺ cell frequency. When needed, cells were labeled with anti-CD14 and anti-CD25 CliniMACS microbeads and depleted using the CliniMACS Plus system (Miltenyi Biotec). The remaining cells were subsequently enriched for CD62L using the same system to yield T_N/MEM cells. T_N/MEM cells were activated with TransAct (Miltenyi Biotec) and transduced with GMP-grade lentivirus. Patient cells were expanded ex vivo for a total of 12 to 16 days in the presence of IL2 and IL15 prior to cryopreservation. Cells were never frozen prior to the manufacturing process and were only frozen and thawed once before infusion into patients.

For pre- and postisolation leukapheresis product analysis, samples were stained with antibodies for CD3, CD14, CD25, and CD62L. For CART19/20 final product analysis, cryopreserved products were thawed, washed with PBS, and stained with antibodies for CD3 and epidermal growth factor receptor (EGFR). The CD19/CD20 bispecific CAR is coexpressed with a truncated, nonsignaling EGFR (EGFRt); thus, EGFRt serves as a proxy for CAR expression. For patient peripheral blood analysis, blood samples were collected in ethylenediaminetetraacetic acid (EDTA) tubes, and PBMCs were collected using the SepMate system (STEMCELL Technologies) following the manufacturer's protocol. Isolated PBMCs were frozen until use. Thawed cells were surface stained with antibody panels for T-cell phenotype (CD3, CD4, CD8, CD62L, CD45RA, CD45RO, and EGFR) or B-cell quantification (CD19, CD20, CD56, CD3, CD14, and SYTOX Blue). Flow cytometry was performed on an Attune NxT flow cytometer (Thermo Fisher), and data were analyzed using FlowJo v.10.7.1 (FlowJo, LLC). Gating strategies are shown in Supplementary Fig. S4.

Patient peripheral blood was collected into red-top tubes containing no anticoagulant or preservative, allowed to clot for 30 minutes in the upright position at room temperature, transferred to a conical tube, and centrifuged at 900 × g for 10 minutes. The supernatant was frozen in aliquots until use. Cytokine analysis was performed by the UCLA Immune Assessment Core Facility using the Luminex 38-plex human cytokine chemokine panel following the manufacturer's protocols.

Descriptive statistics were measured by median and range for continuous variables and counts and percentages for categorical variables. Patients were censored at the time of the last follow-up. Duration of remission (DOR), PFS, and OS were estimated by the Kaplan–Meier method. The statistical software package used was IBM SPSS statistics. DOR was defined as the time of the first documented CR/PR until the first date that recurrent or progressive disease is objectively documented or until death. PFS was defined as the time of CART19/20 infusion until documentation of objective disease progression or death due to any cause. OS was measured from the date of CART19/20 infusion in the clinical trial until death.

Data involving patient confidentiality generated in this study are not publicly available due to patient privacy requirements but are available upon reasonable request from the corresponding authors. Other data generated in this study are available within the article and its supplementary data files.

S.M. Larson reports other support from TORL Biotherapeutics, 1200 Pharma, AbbVie, Bioline, Janssen, Novartis, Pfizer, Bristol Myers Squibb, Sanofi, and Kite and personal fees from Oncovalent outside the submitted work. C.M. Walthers reports grants from the Parker Institute for Cancer Immunotherapy and Jean and Stephen Kaplan during the conduct of the study; other support from Orca Bio outside the submitted work; and a patent for PCT/US2022/077283 pending. B. Ji reports grants from the Parker Institute for Cancer Immunotherapy and Jean and Stephen Kaplan during the conduct of the study; personal fees from ImmPACT Bio outside the submitted work; and a patent for PCT/US2022/077283 pending. J. Trent is a current employee of and holds equity in ImmPACT Bio. M. Roshandell reports grants from the Parker Institute for Immunotherapy and Jean and Stephen Kaplan during the conduct of the study; personal fees from Fate Therapeutics outside the submitted work; and a patent for methods for making and using therapeutic cells pending. C. Harris reports grants from the Parker Institute for Cancer Immunotherapy and Jean and Stephen Kaplan during the conduct of the study. S.B. Gosliner reports grants from the Parker Institute for Cancer Immunotherapy, the Aramont Clinical/Translational Research Program in Hematologic Malignancies, the Hornsey Foundation, Jean and Stephen Kaplan, and the Jaime Erin Follicular Lymphoma Research Consortium during the conduct of the study, as well as a patent for PCT/US2022/077283 pending to UCLA. J.M. Timmerman reports grants from Kite/Gilead, Merck, and Bristol Myers Squibb, and personal fees from Kite/Gilead and Oncovalent outside the submitted work. A. Ribas reports personal fees (past or present scientific advisory board member and stockholder) from Appia, Apricity, Arcus, Compugen, Highlight, ImaginAb, ImmPACT Bio, ImmuneSensor, Inspirna, Lutris, MapKure, Merus, PACT Pharma, Pluto, Synthe­kine, and Tango, personal fees (cofounder and stockholder) from ImmPACT Bio, and personal fees (honoraria for consulting) from Bristol Myers Squibb, Merck, and RAPT during the conduct of the study, as well as a patent for PCT/US2020/049521 issued and licensed to ImmPACT Bio. Y.Y. Chen reports grants from the Parker Institute for Immunotherapy and Jean and Stephen Kaplan during the conduct of the study; personal fees from ImmPACT Bio, Catamaran Bio, Notch Therapeutics, Prime Medicine, Sonoma Biotherapeutics, Waypoint Bio, and Pluto Immunotherapeutics outside the submitted work; and a patent for US 11,160,833 B2 issued and licensed to ImmPACT Bio, a patent for US 11,253,546 B2 issued to ImmPACT Bio, a patent for US 17/569,107 pending, and a patent for PCT/US2022/077283 pending. No disclosures were reported by the other authors.

S.M. Larson: Conceptualization, resources, supervision, investigation, project administration, writing–review and editing. C.M. Walthers: Supervision, investigation, project administration. B. Ji: Data curation, formal analysis, supervision, investigation, project administration. S.N. Ghafouri: Investigation, writing–original draft. J. Naparstek: Data curation, validation, investigation, project administration. J.M. Trent: Data curation, supervision, validation, project administration. J.M. Chen: Project administration. M. Roshandell: Supervision, investigation. C. Harris: Data curation, formal analysis, supervision, investigation. M. Khericha: Data curation, formal analysis, supervision, investigation. T. Schweppe: Data curation, formal analysis, investigation, project administration. B. Berent-Maoz: Supervision, methodology. S.B. Gosliner: Methodology. A. Almaktari: Data curation, formal analysis, investigation. M. Ayala Ceja: Data curation, investigation. M.S. Allen-Auerbach: Investigation. J. Said: Investigation. K. Nawaly: Data curation, validation. M. Mead: Investigation. S. de Vos: Investigation. P.A. Young: Investigation. C. Oliai: Investigation. G.J. Schiller: Investigation. J.M. Timmerman: Investigation. A. Ribas: Conceptualization, resources, funding acquisition, writing–review and editing. Y.Y. Chen: Conceptua­lization, resources, formal analysis, supervision, funding acquisition, visualization, methodology, writing–original draft, project administration, writing–review and editing.

We thank the study participants and all members of the Cellular Therapy Program at UCLA. This work was supported by the Parker Institute for Cancer Immunotherapy (grant no. 20163828 to Y.Y. Chen) and Jean and Stephen Kaplan (gift to Y.Y. Chen). This trial was additionally supported by the Aramont Clinical/Translational Research Program in Hematologic Malignancies and the Hornsey Foundation. A. Ribas is supported by NIH grants R35CA197633 and P30CA016042. J.M. Timmerman is supported by the Jaime Erin Follicular Lymphoma Research Consortium. This study utilized the UCLA Jonsson Comprehensive Cancer Center Flow Cytometry Shared Resource and Technology Center for Genomics and Bioinformatics, which are supported by an NIH Cancer Center Support Grant (grant no. P30CA016042 to Michael A. Teitell). We thank the UCLA Human Gene and Cell Therapy Facility for performing a Quality Assurance review of our Good Manufacturing Practice (GMP) cell-manufacturing process. We thank Dr. Bea Fernandez for her assistance in performing vector copy-number and lentivirus titer analyses. We thank Dr. Xiangzhi Meng, Dr. Ximin Chen, and Dr. Katie Campbell for assistance in performing bulk RNA-seq data analysis. We thank the UCLA Immune Assessment Core for performing cytokine analysis.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.