Generation of a Highly Sensitive CD22 CAR

Background:

CD22 is an attractive target for Chimeric Antigen Receptor (CAR) since it is expressed by most B-cell malignancies. However, CAR targeting of CD22 is challenging since CD22 has a long ectodomain of 300 Å, which is several times wider than the optimal immune synapse. Further, the ectodomain is rigid, so even targeting of membrane proximal domains may not allow effective synapse formation. Further still, CD22 is expressed at low density and dropping of antigen density has been described as a mode of escape. A previously described CD22 CAR based on the M971 antibody required a very short linker between single-chain VH-VL for sensitive function, presumably to allow CAR concatenation but which comes at the cost of basal signalling. Our aim was to engineer an anti-CD22 CAR which triggered not only killing, but proliferation in response to less than 1000 CD22 molecules per cell, without basal signalling. We conducted a functional screen for stable / high affinity binders recognizing the CD22 membrane proximal domains that function in a CAR against cells expressing low levels of CD22 target.

Methods:

Binders were generated in two campaigns. Hyperimmune mice were vaccinated with recombinant CD22 protein lacking the highly immunogenic N-terminal V-type domain. Similarly, Wistar rats were immunized via genetic vaccination using a modified gene encoding the 4 membrane proximal domains of CD22, cloned into a DNA vaccine plasmid designed to stimulate a humoral immune response. A total of 18 binders passed biophysical screening properties of stability, affinity and proximal binding. Binding affinity ranged between 1-36.5nM; the binders were cloned into 41BB-Zeta signalling CARs. We generated a panel of CD22 expressing SupT1 cell lines. Using expression cassettes with amber stop codons preceding the CD22 frame, we generated stable SuptT1 cell lines which express CD22 below the limit of amplified flow cytometry detection (<100 copies, CD22^Und) and a cell line with 255 copies / cell (CD22^Low). In addition, using conventional means, we generated cell lines expressing intermediate and high levels of CD22 expression (CD22^Mid=3,444, CD22^High=78,916 copies/cell).

Results:

The 18 CAR constructs went through an initial round of screening using primary T-cells as effectors and engineered SupT1 cells as targets. A threshold of >50% killing of CD22^Und target cells at an effector:target (E:T) ratio of 1:4 at 72 hour was set for further characterization. Only one CAR derived from the hyperimmune murine binders, and two derived from the rat binders met this threshold. Further characterization of killing at lower E:T ratios, cytokine release and proliferation revealed that CAR based on the rat binder 9A8 had superior function in response to low density target cells compared with the other two CARs and was characterized further and compared with the M971 CAR. This comparison showed approximately 2-fold better killing of CD22^Und 9A8, and approximately 4-fold increased proliferation with no basal signalling. Next, we tried to correlate binder characteristics with sensitivity to low CD22 density. No correlation was found between either affinity, epitope, or stability and sensitivity. In fact, two of the best candidates could be paired with CARs which recognized the same CD22 domain and had similar binding kinetics and stability but had much lower sensitivity to CD22. This suggests that some other as yet unappreciated binder characteristic leads to sensitivity and we are currently investigating this via a structural analysis of binder / target interaction. Finally, we are currently testing 9A8 co-expressed with the CAT19 CD19 CAR (AUTO1) (Blood 2017 130:806) in small animal models of B-ALL.

Conclusion:

A highly sensitive CD22 CAR could be selected by empiric screening of CARs derived from a set of binders which target the proximal domains of CD22. CAR concatenation was not required for a high-degree of sensitivity to CD22 antigen. No correlation was found between binder characteristics and function.