Redirecting T-Cells Against AML in a Multidimensional Targeting Space Using T-Cell Engaging Antibody Circuits (TEAC)

Background

T cell redirection strategies, such as CAR-T and bispecific antibodies (bsAb), are rapidly changing the way in which we approach and treat cancer. While CAR-T and bsAb have shown impressive clinical efficacy in a limited number of cancers, both strategies are ultimately limited by on-target toxicity that currently restricts application to B-cell lineage tumors as the number of genuinely tumor-specific surface antigens is extremely limited. BsAb also suffer from off-target toxicity relating their ability to directly active T-cells severely restricting the therapeutic window.

We sought to solve these inherent problems with the current generation bsAb by re-designing the molecule to alter the mechanism of T-cell activation. By splitting the T-cell engaging VHVL antibody paratope between two separate molecules we created two molecules that formed an active T-cell engaging unit through protein domain complementation following proteolytic activation. Each antibody could target independent surface antigens vastly increasing targeting permutations. Thus, these two antibodies functioned as an "antibody circuit" permitting Boolean type logic to precisely control T-cell activation in multi-dimensional targeting space. We selected AML as model cancer to develop T-cell Engaging Antibody Circuits (TEACs) due to the highly characterized surface antigen landscape and the clear challenges and limitations of single-antigen targeting approaches.

Results

We first screened 10 AML cell lines for candidate surface antigens based upon prior studies of surface antigen display (Perna F et al, 2016) and identified CD33, CD123, CD49d, CD70, CD71, CD38, CLEC12A, Flt3, CD24, CD244, TIM3 and CCR1 as promising targets. We developed a secondary TEACs screening assay where the two TEAC molecules contained either a FITC or biotin binding domain and paired these to commercial FITC or biotin conjugated antibodies targeting the antigens above. We screened 72 TEAC pairs against the 10 cell lines which identified optimal antigen target combinations which included CD33xCD123, CD33xCLEC12A, CD33xCD49d and CD33xCD24. Using a FRET-based fluorescent peptide assay to identify peptide linkers susceptible to proteases we found MMP2 to be highly expressed in AML samples and thus designed all our TEACs with this cleavage site.

We next generated IgG4 format TEACs targeting CD33, CD123, CLEC12A and CD24 that included the MMP2 cleavage activation site and tested these as TEAC pairs in vitro. This screen identified the CD123xCD33 as the most active TEAC pair which was active in 9/10 cells lines. To assess potential safety concerns, we tested TEACs and CD123 and CD33 BiTEs individually and as pairs on PBMCs and on plate-immobilized molecules. These data demonstrated that BiTEs were extremely active against healthy monocytes and also activate T-cells non-specifically once plate-immobilized. In contrast CD123xCD33 IgG TEACs pairs did not activate T-cells when plate-immobilized and did not target healthy monocytes.Finally, we examined the activity of both CD33 BiTEs and CD123xCD33 TEACs on primary patient AML samples. We conducted FRET based assays which confirmed high activity of MMP2 cleavage site on all primary AML samples. When we examined T-cell activation, CD123xCD33 TEACs were active in all CD123+ CD33+ AML samples evaluated with an EC50 of 30ug/ml.

Conclusion

These data suggest T-cell engaging antibody circuits is a new approach that could be safely applied toward AML. TEAC agents do not directly activate T-cells and CD123xCD33 TEAC pairs do not activate PBMC or monocytes. However, CD123xCD33 TEACs show strong activity against AML cell lines and primary CD123+CD33+ AML cells.