Anti-CD21 Chimeric Antigen Receptor (CAR)-T Cells for T Cell Acute Lymphoblastic Leukaemia (T-ALL)

CAR-T cell therapy against CD19 has changed the treatment landscape in relapsed/refractory (r/r) B-ALL. R/r T-ALL has a dismal prognosis, with an unmet need for effective targeted therapies. Several unique challenges mean that CAR-T cell therapy has yet to be successfully translated to T-ALL. Most strategies have targeted pan-T cell antigens (CD7, CD5) but these are limited by T cell aplasia and fratricide, requiring elimination of CAR-T antigen surface expression during manufacture. An ideal target would be exclusively or largely confined to the malignant T cell component but published examples of these (CD1a and TRBC1) are expressed in only minor T-ALL subsets.

We previously showed that CD21 is expressed in a NOTCH-dependent manner in T-ALL (Leukemia. 2013, 27:650) and have developed it as a potential immunotherapy target, being primarily expressed on normal B cells, with minimal expression on mature T cells.

70% of human T-ALL cell lines (9/16) expressed surface CD21 by flow cytometry (FACS), with a median antigen density in positive lines of 2545/cell. In primary T-ALL, 57% of presentation samples (n=58) expressed CD21 (median antigen density 1168/cell). 45% of relapse (n=11) and 20% of primary refractory cases (n=30) expressed CD21, with a similar antigen density to presentation samples. CD21 positivity varied by maturational stage, with highest expression in cortical T-ALL (80% of cases) followed by pre-T (72%), mature (67%), ETP (25%) and pro-T (17%).

Healthy donor blood (n=14) showed CD21 expression limited to B cells and a low proportion (11%) of T cells (10-fold lower intensity v B cells, 316 antigens/cell). T cell CD21 expression was not up-regulated upon activation with CD3/CD28 antibodies (n=6) and was not associated with markers of differentiation/exhaustion.

To target CD21, DNA gene-gun vaccination of rats with a plasmid encoding full-length CD21, followed by phage display was performed and multiple anti-CD21 scFvs isolated. These were cloned into 4-1BBζ CARs and expressed in primary T cells but failed to kill or secrete cytokines in response to CD21+ SupT1 cells.

CD21 is a bulky molecule, with 15/16 sushi repeats in the extracellular domain. All isolated scFvs were found to bind membrane-distal domains. We hypothesized that ineffective signalling due to inadequate synapse formation was responsible for poor performance of anti-CD21 CAR-T, and that binders to membrane-proximal epitopes would signal more efficiently. We re-vaccinated rats with the first 5 sushi repeats of CD21 and generated a library of binders which bound CD21 at this membrane-proximal region.

Multiple candidate binders expressed as CARs were functional, with cytotoxicity and interferon-γ secretion in response to CD21+ target cells. However, non-specific background cytokine secretion was seen against CD21 negative cells, and no IL-2 secretion was seen. Re-cloning binders into a fragment antigen binding (Fab)-CAR architecture yielded constructs capable of specific cytotoxicity, IFN-γ and IL2 secretion against a CD21+ cell line but not its CD21 negative counterpart (n=6). Our lead anti-CD21 candidate CAR specifically proliferated in vitro, without fratricide or premature exhaustion/ differentiation, and was active against low-density CD21-positive cell lines (n=3) and primary cells from 2 T-ALL patients. Improved functionality of Fab v scFv-based CAR was not driven by higher affinity binding or CAR surface expression.

We tested anti-CD21 CAR in murine models of T-ALL. NSG mice were injected with SupT1-luciferase cells and treated with aCD19 or aCD21 CAR-T on day +5. At 2 weeks post treatment, markedly lower disease burden was seen in CD21 CAR-T v CD19 recipients by bioluminescence imaging (median radiance 71700 v 790000 p=0.0079). Further, we injected primary T-ALL blasts in another cohort, treating with aCD19 or aCD21 CAR-T on D+20. Serial bleeds from day 27 post CAR-T showed tumour control in aCD21 CAR treated mice (p=0.024) with an overall survival advantage (median OS 44 days vs undefined, HR = 19.8, p = 0.0069, n=4/group).

In summary, we propose CD21 as a novel target for CAR-T cell therapy in T-ALL. Its expression is largely restricted to the malignant T cell compartment, overcoming issues with fratricide and on-target off-tumour effects seen in many T-ALL CAR-T strategies to date. Despite the complexity of the target, we have successfully generated an aCD21 CAR that is functional both in vitro and in vivo.