CD83 CAR T provides an alternative strategy to CD19-targeting in B cell malignancies

Abstract

Background

Chimeric antigen receptor T cells (CAR T) targeting CD19 can cure patients with aggressive B cell malignancies. Relapse due to CD19 loss and infections from B cell aplasia remain major causes of death after CD19 CAR T. We demonstrate that CD83 is a clinically relevant target for CAR T therapy in B cell malignancies, overcoming limitations of CD19 CAR T.

Methods

We evaluate CD83 expression among B cell acute lymphoblastic leukemia (ALL), chronic lymphocytic leukemia (CLL), mantle cell lymphoma (MCL), diffuse large B cell lymphoma (DLBCL), and circulating healthy B cells, compared with CD19, CD20, and CD22. We test the activity of CD19 and CD83 CAR T against Raji and Nalm6 target cells using in vitro and in vivo preclinical models. Functional and immunophenotypic aspects of CD19 and CD83 CAR T are investigated to compare potency. Further, we evaluate CD19 and CD83 expression on B cell ALL after prior CD19-directed therapy, revealing a propensity for CD19 antigen loss. To simulate CD19 antigen loss, we test CD19 and CD83 CAR T against CD19 knock-out Nalm6 cells.

Results

CD83 is expressed on B-cell ALL, MCL, and DLBCL, with negligible expression on circulating B cells and CLL. B cell ALL expresses CD83 at diagnosis and on relapse following CD19-directed therapy, when CD19 is decreased. CD83 CAR T kills B cell leukemia/lymphoma cell lines in vitro, providing efficacy after CD19 antigen loss. We show CD83 CAR T is effective as monotherapy in mice. Exploratory experiments suggest CD83 CAR T may offer enhanced efficacy when administered sequentially after CD19 CAR T.

Conclusions

Our preclinical evidence justifies further translational investigation of CD83 CAR T in B cell leukemia and lymphoma. The benefits of CD83 CAR T include a low risk for B cell aplasia and efficacy after exposure to CD19 CAR T.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Commercial CD19 CAR T cells for B cell leukemia and lymphoma are limited by B cell aplasia and secondary infectious disease complications, as well as CD19 antigen loss.

• CD83 CAR T cells have demonstrated preclinical efficacy in treating acute myeloid leukemia and graft-versus-host disease.

• While others have shown CD83 is targetable in Hodgkin lymphoma with monoclonal antibody therapy, this is the first preclinical demonstration of CD83 CAR T in non-Hodgkin’s lymphoma and B cell acute lymphoblastic leukemia (ALL).

WHAT THIS STUDY ADDS

• We demonstrate that B cell malignancies are susceptible to CD83 CAR T.

• CD83 CAR T cells remain effective after prior exposure to CD19-directed therapy or against CD19-deficient B cell ALL.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Relapse after commercial CD19 CAR T is associated with poor survival.

• While retreatment of patients with CD19 CAR T is limited by antigen escape, CD83 CAR T cells offer a potent alternative.

• Our phase I trial of CD83 CAR T in relapsed/refractory acute myeloid leukemia is open (NCT06871410). We expect the safety data from the trial will inform subsequent clinical studies of CD83 CAR T in B cell malignancies.

Background

Chimeric antigen receptor (CAR) T cell therapy has revolutionized the treatment of B cell leukemia and lymphoma. CD19-targeted CAR T products have earned a second-line therapeutic indication for diffuse large B cell lymphoma (DLBCL) and have displaced the early use of high-dose chemotherapy and autologous hematopoietic cell transplantation.^{1 2} Indeed, CD19 CAR T can cure more than a third of treated patients with aggressive large cell lymphoma.³ However, relapse remains a major cause of treatment failure.⁴ Additionally, CD19 CAR T can induce lasting B cell aplasia and secondary hypogammaglobulinemia.⁴ As such, serious infections are a leading cause of non-relapse mortality after CD19 CAR T.⁴ These limitations of CD19 CAR T therapy underscore a critical unmet need for patients with high-risk B cell leukemia and lymphoma.

CD83, a member of the immunoglobulin superfamily, is normally expressed on a subset of mature dendritic cells, macrophages, and monocytes. In the setting of cancer, we have shown that it is also present on acute myeloid leukemia (AML).^5–8 Importantly, CD83 is absent from steady-state peripheral B cells from healthy individuals, healthy circulating neutrophils, and hematopoietic stem cells in the bone marrow.⁵ In preclinical models, CD83 CAR T effectively kills AML cell lines and patient-derived leukemia.^{5 6} A phase I clinical trial evaluating the safety of CD83 CAR T therapy in relapsed or refractory AML is open (NCT06871410). Beyond AML, CD83 is a clinically relevant target in mantle cell lymphoma (MCL) and the Reed-Sternberg cells of Hodgkin lymphoma (HL).^{9 10} In this work, we investigate the expression and antigen density of CD83 in patients with a variety of B cell malignancies. We study the potency and efficacy of CD83 CAR T as a monotherapy in preclinical models. Moreover, we characterize the durability of CD83 expression in the setting of CD19 antigen loss in a group of patients with acute lymphoblastic leukemia (ALL) who relapsed following CD19-directed therapy. Based on the persistence of CD83 expression, we explore the activity of sequential administration of CD19 CAR T followed by CD83 CAR T as a strategy to mitigate relapse. Overall, we provide compelling data justifying the translational study of CD83 CAR T in aggressive B cell malignancies.

Methods

Patient samples

Peripheral blood or bone marrow samples were obtained from patients with chronic lymphocytic leukemia (CLL, n=33), ALL (ALL, n=40), MCL (n=5), or DLBCL (n=2) who consented to an Institutional Review Board-approved biospecimen collection study at Moffitt Cancer Center, University of Minnesota, or Roswell Park Comprehensive Cancer Center. Circulating B cells from consented healthy donors were used as a control (n=9).

CAR T cell production

The CD19 and CD83 CAR T cells were generated as previously described from healthy apheresis donors.⁵ The CAR T constructs were designed from a SFG retroviral backbone that contained an anti-human CD19 or anti-human CD83 scFv, CD8α signal peptide and transmembrane, a hinge domain, and completed with a 41BB and CD3ζ co-stimulatory domain.⁵ The number of untransduced T cells used in the described experiments was normalized to the CAR T transduction efficiency.

CAR T potency and function

Cytokine analysis

CD19 CAR, CD83 CAR, or untransduced T cells (1×10⁵) were cocultured with Raji target cells (effector to target (E:T) ratio=2:1) for 24 hours. Supernatants were harvested and IFNγ, IL-2, IL-6, and TNFα were quantified using a human Luminex assay kit (R&D Systems) on an Ella instrument (Biotechne) according to manufacturers’ instructions.⁶

CAR T killing of target cells

Normalized amounts of CD19 or CD83 CAR T cells (1×10⁵ cells) were cultured with Raji or Nalm6 target cells at E:T ratios 2:1, 1:1, and 1:2 in duplicate. Where indicated, CD19 CAR, CD83 CAR, and untransduced T cells were evaluated against CD19 knockout Nalm6 cells (CD19 expression <1%), to model clinical antigen loss (E:T ratio 3:1).¹¹ Autologous untransduced T cells were used as controls. The cytotoxicity experiments were performed using an xCELLigence RTCA (real-time cell analysis) instrument (ACEA Biosciences).⁶

Fluorochrome-conjugated antibodies and flow cytometry

Clinical samples

Acquired peripheral blood or marrow samples were thawed and surface stained with live/dead Aqua, CD3, CD5, CD10, CD19, CD20, CD22, CD33, CD34, CD45, CD79a, and CD83. CD4, and CD83. CD19, CD20, CD22, and CD83 expression was determined by gating on live events, based on disease phenotype; CLL: CD3⁻, CD5⁺, CD79a⁻; ALL lymphoblasts: CD3⁻, low side scatter, CD45^low/−, CD34⁺, CD79a⁺; MCL: CD3⁻, CD5⁺, CD79a⁺; DLBCL: CD3⁻, high forward scatter, CD45⁺, CD5⁻, CD79a⁺. Fluorescence minus one (FMO) staining controls were used for each fluorochrome. The frequencies of CD19⁺, CD20⁺, CD22, and CD83⁺ cells were recorded. In a subset of ALL lymphoblast (n=13) and CLL cell (n=12) samples, CD19, CD20, CD22, or CD83 antigen density at time of diagnosis or CD19 and CD83 expression at time of diagnosis versus relapse (n=6) after having received a CD19-directed therapy, including brexucabtagene autoleucel or blinatumomab, was evaluated. A sample at time of diagnosis was unavailable for one patient treated with CAR T. Online supplemental table 1 details the fluorochrome-conjugated antibodies used.

Human cell lines

CD83 expression was evaluated on Nalm6, Daudi, and Raji cell lines by flow cytometry. Resting human T cells served as a CD83 negative/low staining control. The cell lines were acquired from ATCC. All cell lines were confirmed to be mycoplasma-free and of a low number of passages.

Animals

Immunodeficient Nod Scid Gamma (NSG) mice (ages 6–12 weeks old, male or female sex) were transplanted with luciferase-transduced Nalm6 cells (5×10⁵). Bioluminescence imaging (BLI) was performed to monitor tumor engraftment and subsequent progression.⁶ Once engraftment was confirmed by BLI (~day +5), the mice received CD19 CAR, CD83 CAR, or untransduced T cells (3×10⁶). The dose of CAR T was normalized to the CAR transduction efficiency. In addition to BLI, mice were also monitored for clinical markers of well-being. Mice were euthanized once they met endpoint at the conclusion of the experiment (day +60) or if premoribund, exhibited measurable tumor growth >15 mm, or impaired locomotion from tumor mass. Experiments were conducted with 8–12 mice per group. In separate, identically performed short-term experiments, mice were euthanized at day +21 and spleens were harvested to enumerate and characterize Nalm6 and CAR T cells. 8–9 mice per group. Mouse allocation was randomized and treatment blinding was performed according to ARRIVE standards.¹² Mice were purchased from the Roswell Park Comprehensive Cancer Center Comparative Oncology Shared Resource. Mice were housed in BSL2 cages of 4–5 animals based on weight in an immunodeficient mouse unit.

CAR T memory phenotype

CD19 CAR T, CD83 CAR T, or untransduced T cells (UT) were harvested from treated mice on day +21. The T cells were evaluated by flow cytometry for memory phenotype, characterized as central memory (CM: CD45RA⁻, CCR7⁺), effector memory (EM: CD45RA⁻, CCR7^-), effector memory re-expressing CD45RA (EMRA: CD45RA⁺, CCR7⁻), and naïve (CD45RA⁺, CCR7⁺). CAR T were identified by anti-G4S linker staining. Where indicated, T cells were evaluated for CD39, PD-1, or TIM-3 expression. CountBright Absolute Counting Beads (Thermo Fisher Scientific) were used to enumerate cells.

Statistics

Data are reported as mean values (±SEM). Analysis of variance was used for group comparisons, with the corresponding multiple comparisons test noted in the figure legends. For comparison of survival curves, a log-rank test was used. A mixed-effects analysis was used to evaluate changes in CD19 and CD83 expression at diagnosis and relapse of B cell ALL. The type of CD19-directed therapy was the independent variable, and change in CD19 or CD83 expression from diagnosis to relapse was the outcome. The statistical analysis was conducted using Prism software V.10.4.1 (GraphPad). Statistical significance was defined by a two-tailed p less than 0.05.

Results

CD83 is a clinically relevant target in human B cell malignancies

We first surveyed CD83 expression on a range of human B cell leukemia and lymphoma cell lines, including Nalm6, Daudi, and Raji cells. Resting human T cells served as a CD83 negative/low control.^{5 6} Nalm6 is derived from precursor B cell ALL, while Daudi and Raji cell lines originate from aggressive Burkitt lymphoma. CD83 expression across the three cell lines was in excess of 90% (figure 1A). Having established robust CD83 expression in immortalized cell lines, we next evaluated CD83 expression on clinical samples from patients with ALL, CLL, MCL, and DLBCL. Additionally, we compared CD83 expression on these B cell malignancies to CD19, CD20, and CD22 (figure 1B). Circulating B cells from healthy donors served as a control group. In contrast to CD19, CD20, or CD22, CD83 expression was significantly lower on healthy B cells compared with cancer cells, though expression was similar between healthy B cells and CLL cells (figure 1B). However, in terms of CLL, CD19 was the only antigen consistently expressed at a high level among those tested. CD83 expression rivaled CD19, CD20, and CD22 in MCL and significantly exceeded CD20 expression in DLBCL (figure 1B). Overall, these data support that CD83 is a relevant target of engineered cell therapy for B cell ALL, MCL, and DLBCL. Moreover, the low expression of CD83 on healthy circulating B cells reduces the risk of B cell aplasia, an important clinical complication of CD19 CAR T and likely a risk for CAR T targeting CD20 or CD22.⁵

Figure 1. CD83 is a clinically relevant target of human B cell malignancies. (A) CD83 expression is shown among resting CD83 negative T cells, Nalm6, Daudi, and Raji cell lines, depicted as filled, open, blue, and red histograms, respectively. Horizontal bars depict gated CD83 positivity. (B) CD19, CD83, CD20, and CD22 expression is demonstrated among healthy peripheral B cells (n=9) versus B cell malignancies; chronic lymphocytic leukemia (CLL, n=33), B cell acute lymphoblastic leukemia (ALL, n=23), mantle cell lymphoma (MCL, n=5), and diffuse large B cell lymphoma (DLBCL, n=2), two-way ANOVA, Dunnett’s test compared CD83 expression versus CD19, CD20, or CD22 across all groups. (C) Graph shows CD19, CD20, CD22, and CD83 antigen density among B cell ALL (n=13) and CLL (n=12) patient samples, two-way ANOVA, Tukey’s test compared antigen expression means between ALL and CLL groups. MESF:molecules of equivalent soluble fluorochrome. *p<0.05, **p=0.001–0.01, ***p=0.0001–0.001, and ****p<0.0001. NS, not significant. ANOVA, analysis of variance.

We next investigated the antigen density of CD83 on ALL lymphoblasts and CLL cells, compared with CD19, CD20, and CD22 (figure 1C). CD83 antigen density on ALL lymphoblasts was statistically similar to CD20 and CD22, but decreased compared with CD19 (figure 1C). In CLL, CD83 antigen density was significantly lower than CD19 or CD20. The aggregate CD83 expression and antigen density data support the investigation of CD83 CAR T in B cell malignancies like B cell ALL, but not CLL.

CD83 CAR T effectively kill human B cell lymphoma and leukemia in vitro

To investigate the potency of CD83 CAR T cytotoxicity against B cell malignancies, we cultured CD83 CAR, CD19 CAR, or untransduced T cells with Raji or Nalm6 cells at various E:T ratios (figure 2A–F). The CD19 and CD83 CAR T cells were produced as described, which consisted of anti-human CD19 or CD83 scFvs, a CD8α hinge and transmembrane domain, and completed with a 41BB co-stimulatory domain and CD3ζ.⁵ Using an Xcelligence platform, the CD83 CAR T demonstrated robust Raji or Nalm6 killing in vitro, similar to clinically relevant CD19 CAR T (figure 2A–F). Importantly, these experiments show CD83 CAR T provides rigorous anti-tumor activity against B cell malignancies.

Figure 2. CD83 CAR T effectively kill human B cell lymphoma and leukemia in vitro. (A–F) Anti-human CD19 CAR, CD83 CAR, or untransduced T cells (UT) were co-cultured with Raji or Nalm6 targets (both express CD19 and CD83) at various effector to target (E:T) ratios. Raji or Nalm6 cells alone served as a control. The CD19 CAR and CD83 CAR T cells consisted of a CD8α signal peptide and transmembrane plus a hinge domain, followed by a 41BB co-stimulatory domain and CD3z. Target cell killing was determined by an Xcelligence assay over 80 hours. Three healthy T cell donors are shown for each condition, with error bars representing SEM. Significance was determined by one-way ANOVA, Tukey’s test compared means between groups at the 80-hour time point. *p<0.05, **p=0.001–0.01, ***p=0.0001–0.001, and NS, not significant. ANOVA, analysis of variance.

CD83 CAR T produce IFNγ, IL-2, and TNFα in response to Raji target cells

Cytokine production is an important metric of CAR T potency.¹³ CD83 CAR T, CD19 CAR, or untransduced T cells were cultured with Raji targets at an E:T ratio of 2:1, and supernatants were analyzed for IFNγ, IL-2, IL-6, and TNFα after 24 hours (figure 3A–D). compared with untransduced T cells, CD83 CAR T produced significant amounts of IFNγ, IL-2, and TNFα (figure 3A–C, p<0.01). When compared with CD19 CAR T, CD83 CAR T exhibited donor variability that led to no consistent direction of change in IFNγ, IL-2, or TNFα production (figure 3A–C). However, the amount of IL-6 generated by CD83 CAR T tended to be less than CD19 CAR T in three of the four tested donor products (figure 3D). This observed pattern of robust IFNγ, IL-2, TNFα production, yet low IL-6 could have clinical benefit, as IL-6 is a critical mediator of cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome.¹⁴

Figure 3. CD83 CAR T produce IFNγ, IL-2, and TNFα in response to Raji target cells. CD19 CAR, CD83 CAR, or untransduced T cells (UT) were co-cultured with Raji target cells at an effector to target (E:T) ratio of 2:1. Cytokines were quantified from culture supernatants after 24 hours. (A–D) Graphs show the mean amounts of IFNγ, IL-2, TNFα, and IL-6 produced by CD19 CAR, CD83 CAR, or UT. A total of four independent healthy T cell donor experiments are shown. Two-way ANOVA, Tukey’s test compared the means of each T cell group with each other. *p<0.05, **p=0.001–0.01, ****p<0.0001, and NS, not significant. ANOVA, analysis of variance.

Prior exposure to CD19-directed therapy leads to CD19 antigen loss in ALL, but CD83 expression is robust

CD19-negative relapse occurs in 30%–70% of patients with B cell malignancies after CD19-directed therapy, including CAR T and T cell engager treatments.¹⁵ Antigen loss eliminates the clinical benefit of retreating patients with CD19-directed strategies.^16–18 We compared CD19 or CD83 expression after relapse post brexucabtagene autoleucel or blinatumomab for ALL. In this sample set (n=6) of diagnostic and relapse samples, we observed a significant reduction in the frequency of CD19+B cell ALL lymphoblasts at the time of relapse (figure 4A). However, the percentage of CD83+B cell ALL lymphoblasts during relapse was comparable to levels at diagnosis (figure 4B). The mean fluorescence intensity of CD19 and CD83 on B cell ALL lymphoblasts followed similar patterns with disease recurrence following CD19-directed therapy, with the CD19 gMFI on lymphoblasts being significantly decreased on relapse (figure 4C,D). This suggested that while retreatment with CD19 CAR T may have diminished efficacy in pretreated, relapsed ALL, targeting such disease with CD83 CAR T could be effective. We observed a similar effect in vitro, where co-culture with CD19 CAR T, but not with untransduced or CD83 CAR T, led to decreased CD19 expression on Raji cells over 24 hours (figure 4E). CD83 expression on Raji cells remained intact whether exposed to CD19 CAR or CD83 CAR T (figure 4B). To study CD19 antigen loss in a preclinical model, we used CD19 knockout Nalm6 cells (CD19 expression <1%).¹¹ CD83 CAR T demonstrated significantly greater cytotoxicity against CD19KO Nalm6 cells compared with CD19 CAR T (figure 4F). CD19 CAR T were ineffective against the CD19KO Nalm6 cells (figure 4F). These data demonstrate that CD19-directed therapy can reduce the durability of CD19 antigen expression in B cell malignancies. Moreover, even when CD19 CAR T cells are rendered less effective by CD19 antigen loss, CD83 CAR T cells offer potent tumor killing.

Figure 4. Prior exposure to CD19-directed therapy leads to CD19 antigen loss in ALL, but CD83 expression is robust. (A–D) CD19 and CD83 antigen expression and density are shown on B ALL lymphoblasts (n=6), comparing expression at the time of diagnosis to that of relapse after receiving CD19-directed therapy, including brexucabtagene autoleucel (CD19 CAR T) or blinatumomab (CD19/CD3 bispecific T cell engager), Mann-Whitney U test. (E) Contour plots from a representative experiment show CD19 versus CD83 expression on Raji cells after a 24-hour co-culture with CD19 CAR, CD83 CAR, or untransduced (UT) T cells. (F) Graph shows CD19 CAR, CD83 CAR, or UT killing of CD19 knockout (CD19 expression <1%) Nalm6 cells (effector to target ratio 3:1) by an Xcelligence assay. Significance was determined by one-way ANOVA with Tukey’s multiple comparisons test between groups at the 80-hour time point. Data from five independent CAR T cell donors are shown. *p<0.05, **p=0.001–0.01, ****p<0.0001, and NS, not significant. ALL, acute lymphoblastic leukemia; ANOVA, analysis of variance.

CD83 CAR T exhibit preclinical activity against B cell ALL in vivo

CD83 CAR T were compared with CD19 CAR T against luciferase-transduced Nalm6 in immunodeficient NSG mice. CD83 CAR T significantly improved survival, compared with untreated mice or those dosed with untransduced T cells (figure 5A). Mouse survival was intermediate for those treated with CD83 CAR T, compared with untransduced or CD19 CAR T cells (figure 5A–C). Compared with untreated mice, CD83 CAR T delayed maximum tumor progression by approximately 15 days (figure 5B,C). Overall, these experiments demonstrate the in vivo activity of CD83 CAR T using an aggressive, clinically relevant B cell ALL model.

Figure 5. Antitumor activity of CD83 CAR T against Nalm6 in vivo. NSG mice were injected with 500 000 luciferase-transduced Nalm6 cells, and then 3×10⁶ CD19 CAR, CD83 CAR, or untransduced T cells (UT) a week later, once tumor engraftment was confirmed. Mice were followed for survival until day +60 and tumor growth was monitored by bioluminescence imaging (BLI). Graphs show (A) survival and (B) Raji radiance per BLI assessments, log-rank test (A) and two-way ANOVA (B), Dunnett’s test compared means between groups across all times. (C) BLI trends are shown across three independent experiments. Pooled data from three independent experiments is shown (A–C) with 12 mice per group. **p=0.001–0.01, ***p=0.0001–0.001, and ****p<0.0001. ANOVA, analysis of variance.

Characteristics and performance of CD83 CAR T early after infusion

NSG mice bearing Nalm6 tumors were treated with UT, CD19 CAR T, or CD83 CAR T as described, and humanely euthanized on day +21 to harvest residual tumors and CAR T or UT within the recipient spleens. Overall, CD83 CAR T and CD19 CAR T similarly killed Nalm6 tumors in vivo early after infusion, with UT offering ineffective antitumor activity (figure 6A). Nalm6 tumors from mice treated with UT, CD19 CAR T, or CD83 CAR T showed a reduction in CD19 expression relative to controls (figure 6B). While CD83 expression intensity is lower than CD19 among Nalm6 tumors from untreated mice, those that were treated with UT or either CAR T cells showed intact CD83 expression (figure 6C). Overall, this effect by responding T cells resulted in significant CD19 antigen loss compared with CD83, consistent with our in vitro observations (figure 6B). There were significantly fewer CD4⁺ or CD8⁺ CD83 CAR T cells acquired from CD83 CAR T treated mice, compared with those injected with CD19 CAR T (figure 6D,E). However, the CD8/CD4 ratio was favorably increased among CD83 CAR T treated mice, compared with CD19 CAR T bearing animals (figure 6F). CD4⁺ CD83 CAR T did show evidence of exhaustion at day +21, expressing significantly high levels of CD39 and PD-1, compared with CD4⁺ CD19 CAR T (figure 6G). CD8⁺ exhaustion markers were statistically similar among CD19 CAR T and CD83 CAR T cells (figure 6H). CD4⁺ CD83 CAR T significantly favored a central memory phenotype, compared with CD19 CAR T (figure 6I). However, CD8⁺ CD83 CAR T were primarily effector memory and EMRA (figure 6J). Overall, CD83 CAR T demonstrated potent early antitumor activity and stable CD83 antigen expression on its target cell.

Figure 6. Characteristics and performance of CD83 CAR T early after infusion. NSG mice were injected with 500 000 luciferase transduced Nalm6 cells, and then 3×10⁶ CD19 CAR, CD83 CAR, or untransduced T cells (UT) a week later once tumor engraftment was confirmed. Mice were humanely euthanized on day +21 and spleens were harvested to enumerate and characterize residual Nalm6 and resident CAR T or UT cells. (A) Graph shows number of CD22⁺ Nalm6 cells from mouse spleens. One-way ANOVA, Tukey’s test compared the Nalm6 means by treatment group. (B) Graph depicts fold change in CD19 or CD83 negative Nalm6 cells recovered from mouse spleens, reflecting antigen loss. Two-way ANOVA, Tukey’s test compared the means of each treatment group with each other. (C) Representative histograms show CD19 or CD83 expression among residual Nalm6 cells from mouse spleens at day +21. Open histogram=FMO, filled histogram=Nalm6 alone/untreated, blue=UT, red=CD19 CAR T, and green=CD83 CAR T groups. (D, E) Graphs demonstrate the number of CD4⁺ and CD8⁺ UT or CAR T cells harvested from the recipient spleens. (F) Graph depicts the ratio of CD8/CD4 T cells from harvested spleens for each group. One-way ANOVA, Tukey’s test compared the means by treatment group. Graph shows the frequency of CD39, PD-1, or TIM3 expression (G, H) or memory subsets (I, J) among CD4⁺ or CD8⁺ UT or CAR T cells from the recovered spleens. Memory phenotype was characterized as central memory (CCR7⁺, CD45RA⁻), effector memory (CCR7⁻, CD45RA⁻), naïve (CCR7⁺, CD45RA⁺), or effector memory re-expressing CD45RA (CCR7⁻, CD45RA⁺). Two-way ANOVA, Tukey’s test compared the means by T cell group within each phenotype (G, H) or memory (I, J) subset. Pooled data from two independent experiments are shown with 8–9 mice per group.*p<0.05, **p=0.001–0.01, ***p=0.0001–0.001, and ****p<0.0001. NS, not significant. ANOVA, analysis of variance.

Sequential administration of CD19 CAR followed by CD83 CAR T provides enhanced efficacy

Based on the activity of CD19 and CD83 CAR T against B cell lymphoma and leukemia, we performed an exploratory investigation of the sequential administration of CD19 CAR T followed by CD83 CAR T in a luciferase-transduced Raji model. This approach recapitulates recent clinical trial designs of CAR T targeting multiple lymphoma antigens, where cell therapy is given in sequential fashion.^{19 20} A reduced dose of 1×10⁶ CAR T was used to avoid overpowering the system with a large dose of CAR T, permitting the detection of potential benefit with sequential CD19 CAR and CD83 CAR T therapy. We observed that both the CD19 CAR and CD83 CAR T significantly improved the survival of tumor bearing NSG mice compared with the untreated group, with similar median survival (online supplemental figure 1A). However, when CD83 CAR T was administered sequentially after CD19 CAR T, separated by a week, the treated mice experienced improved overall survival compared with the monotherapy groups (online supplemental figure 1A). The sequential CD19 CAR plus CD83 CAR T strategy also resulted in the lowest tumor burden of Raji cells (online supplemental figure 1B,C). These data suggest that sequential administration of CD83 CAR T, after an initial infusion of CD19 CAR T, may improve clinical outcomes in terms of depth and durability of response (online supplemental figure 1B,C).

Discussion

We demonstrate that CD83 is a clinically relevant antigen to therapeutically target ALL, MCL, and DLBCL. Others have demonstrated that anti-human CD83 antibody and derivative antibody-drug conjugates can kill MCL and HL.^{9 10} We have previously shown that targeting CD83 with CAR T spares normal hematopoiesis and maintains circulating B cell numbers in preclinical murine models.⁵ Conversely, CD19 CAR T cells often induce deep and long-lasting B cell aplasia.⁴ This is of major clinical importance, given that B cell aplasia and subsequent serious infections are responsible for over half of all non-relapse deaths after CAR T infusion.⁴ Our data suggest that CAR T targeting CD20 or CD22 will have similar deleterious effects on peripheral B cells, given their high expression along with CD19. In fact, CD20 CAR T has been tested in the application of forced B cell eradication as an experimental strategy to clear HIV reservoirs.²¹ Elimination of circulating B cells was also observed in a phase I trial of CD22 CAR T in B cell ALL.²² A recent phase II trial of CD22 CAR T for large B cell lymphoma, evaluating a different cell product, was recently closed due to an unfavorable risk/benefit profile.²³ Therefore, CD83 CAR T, which has been previously shown to preserve peripheral B cells, directly addresses a critical unmet need for patients with ALL, MCL, or DLBCL at risk of serious infections and the sustained morbidity of hypogammaglobulinemia with CD19 CAR, CD20 CAR, or CD22 CAR T.^{4 5 16–18 22}

Our data from several independent human tissue microarrays also show negligible expression of CD83 on healthy tissue, reducing the risk of on-target, off-tumor toxicity.^{5 6} Preliminary data from others suggest CD83 antibody is safe in nonhuman primates, where the antibody preserves hematopoietic stem cells, T cells, and B cells.²⁴ Anti-CD83 antibody exerts reductions, but not outright absence of, dendritic cells in treated non-human primates.²⁴ We observed a similar effect in humanized mouse models, where CD83 CAR T decreased mature CD83⁺ dendritic cells, but we verified that T cell responses to viral peptides remained intact.⁶ We initially developed CD83 CAR T for AML, and have since found it to be promising in other hematologic malignancies, like B cell leukemia/lymphoma in this present work.^{5 6} CD83 antigen density among B cell ALL is similar to levels observed in AML.⁵ Our phase I clinical trial of CD83 CAR T for relapsed/refractory (R/R) AML (NCT06871410) is open and successfully infusing patients. Ultimately, the phase I trial of CD83 CAR T in R/R AML will provide insights on how CD83 CAR T, rather than an antibody, impacts T and B cell populations in treated patients and inform further translation of these engineered cells in broader hematologic malignancies.

The cytokine profile of CD83 CAR T cells provides further insight into their safety. We observed that CD83 CAR T stimulated with target cells produce ample IFNγ, IL-2, and TNFα, which are necessary for cytotoxic potency, but little IL-6 compared with CD19 CAR T. This difference in IL-6 production could be influenced by nuances in CD19 versus CD83 target cell antigen density. Since therapeutic targeting of the IL-6 receptor with tocilizumab is a well-known strategy to treat CRS, and engineering approaches have been pursued to either program CAR T to secrete anti-IL-6 scFv or edit out IL-6 production capability from CAR T to reduce CRS,^{25 26} we surmise that CD83 CAR T may have unique properties to limit the risk of CRS. This includes the inherently low levels of CD83 CAR T IL-6 production paired with the fact that mature proinflammatory monocytes and macrophages express CD83, produce IL-6 required for the propagation of CRS, and will be susceptible to killing by the CD83 CAR T.^{14 25} The onset of CRS and usage of tocilizumab and glucocorticoids post CD83 CAR T infusion will be important secondary endpoints of the upcoming phase I AML clinical trial.

While commercial CD19 CAR T can cure an estimated 30%–40% of patients with select B cell malignancies, CD19 antigen loss remains a major cause of relapse.³ This observation is further compounded by prior exposure to protein therapies targeting CD19, which can substantially diminish later responses to CAR T against the same antigen.²⁷ Therefore, identification of targets that render B cell malignancies susceptible to CAR T after commercial CD19 CAR T is a critical need. Attempts to address antigen loss have largely included simultaneous combinations of CAR T targeting CD19, CD20, CD21, and/or CD22.^{15–18 22} While promising in preclinical studies, early phase studies have yet to demonstrate clinical activity with dual targeting CAR T in B cell malignancies.^{15–18 22} Antigen loss is a recurring issue, even in dual-specific CAR T. In a pilot group of patient samples, we investigated CD83 expression at the time of ALL diagnosis and after relapse post CD19-targeted therapy. CD19 expression was significantly reduced after relapse. However, CD83 expression was still present on malignant cells after ALL relapse. This observation justified our in vivo studies, where we sequentially administered CD19 CAR T followed by CD83 CAR T. Rather than designing a dual specific CAR, the pragmatic strategy of administering separate CAR T products offered simplicity in engineering and production. The sequential approach rendered improved survival in a clinically relevant mouse model, providing at least an additive effect.

In a direct comparison of CD83 vs CD19 CAR T in de novo untreated Nalm6 tumors, CD19 CAR T had an advantage in terms of survival. However, CD83 CAR T still offered significantly better survival compared with UT or untreated controls. While CD19 and CD83 CAR T provide similar tumor clearance early after infusion, the late disease progression after CD83 CAR T monotherapy could be related to CD4⁺ CAR T exhaustion as noted by increased CD39 and PD-1 expression. Importantly, we expect the role for CD83 CAR T will be in the relapsed setting, following prior exposure to CD19 CAR T. While CD19 CAR T exposed B cell leukemia/lymphoma cell lines are subject to CD19 antigen loss, CD83 expression is stable to increase after CD19 or CD83 CAR T. We are actively investigating strategies to incorporate CD83- with CD19-targeting for B cell leukemia and lymphoma. While sequential delivery of CD19 CAR T followed by CD83 CAR T is advantageous in terms of engineering simplicity, we are considering alternative approaches. First, we are developing a dual-targeting CD19/CD83 CAR T using loop and tandem designs. Further, we have developed IL-18 and IL-36γ cytokine-armored variants of the CD83 CAR T which we expect will enhance monotherapy potency. IL-18 and IL-36γ armors bolster tumor clearance by recruiting endogenous T effector cells.^{28 29} Having the ability to enlist and activate endogenous T cell reinforcement could overcome suboptimal CD83 antigen expression at the time of relapse. While CD83 CAR T will target mature dendritic cells (DC), we have previously shown that CD83 CAR T do not induce DC aplasia.⁶ CD83 negative DC antigen presentation remained intact in the presence of CD83 CAR T, facilitating CD8 cytotoxic T cell responses against pathogenic viral peptides.⁶ Thus, we surmise epitope spreading will remain intact and possibly amplified by using cytokine-armored versions of the CD83 CAR T. Cytokine-armored CD83 CAR T could also eliminate the need for lymphodepleting chemotherapy in heavily pre-treated patients who experience post-CD19 CAR T relapse.²⁸ Last, further refinement of the sequential delivery of CD19 CAR followed by CD83 CAR T includes possible dose and temporal modifications of the cell therapy. The unique properties of CD83, including its preserved expression after CD83 CAR T exposure and persistence following relapse post-CD19 CAR T therapy, justify further investigation of CD83 CAR T for B cell leukemia and lymphoma.

Data availability statement

Data are available on reasonable request.