Targeting Normal Myelopoiesis Negatively Affects CAR T Cell Activity

Chimeric antigen receptor (CAR) T cells have not been successful for the treatment of acute myeloid leukemia (AML) to date, and the reason for this is not clear. While CAR T cells have been traditionally studied as independent agents targeting tumor, in reality they work in the context of the endogenous host immune system, which may influence outcomes of therapy. We hypothesized that targeting normal myelopoiesis is detrimental to CAR T cell activity, and this creates a unique obstacle to effective CAR T cell therapy for AML.

We generated prototypical CAR T cells targeting B cell (CD19) and myeloid (CD33) antigens, using a standard 2 ^nd generation CAR design with 4-1BB and CD3ζ costimulatory domains. In parallel we created tumor cell lines expressing both targets by adding CD33 expression to a CD19+ B cell line (Ramos ⁺³³), and by adding CD19 expression to a CD33+ myeloid cell line (THP1 ⁺¹⁹). For ease of tracking tumor cells were labeled with GFP, and CAR T cells were labeled with RFP. CD19-targeting CAR T cells (CART19) and CD33-targeting CAR T cells (CART33) exhibited similar anti-tumor activity in standard in vitro cytotoxicity assays and in vivo NSG mouse models.

We then proceeded to add PBMCs to CAR T cell and tumor co-culture and measured tumor cell clearance and CAR T cell expansion using flow cytometry and the Incucyte Live-Cell analysis system. Addition of PBMCs enhanced the tumor clearance and CAR T cell expansion of both CART19 and CART33 in short-term co-culture assays (7-10 days). Fractionation of PBMCs revealed that CD33+ classical monocytes were able to replicate the effect of PBMCs, while the presence of CD19+ B cells did not have any effect.

Monocytes are viable in vitro for less than 7 days. To assess the long-term effects of monocytes on CAR T cells, we performed serial replating assays, in which fresh tumor cells and monocytes were added to CAR T cells every 7 days. Ongoing monocyte support further improved CART19 expansion and tumor clearance. However, the continuous presence of monocytes was detrimental to CART33 activity upon serial replating, as evidenced by suboptimal T cell expansion and loss of tumor control.

To study the effects of normal myelopoiesis on CAR T cell activity in vivo, NSGS mice (NOD/SCID/IL2Rg ^null mice with human SCF, IL-3 and GM-CSF expression) were injected with human cord blood CD34+ cells to engraft human hematopoiesis. Mice were then injected with THP1 ⁺¹⁹ tumor cells, followed by administration of untransduced (UTD) control T cells, CART19 or CART33. To prevent the confounding effect of allogeneic killing, CAR T cells were generated from the same cord blood product as the CD34+ cells. Consistent with our in vitro data, CART19 and CART33 showed better anti-tumor activity in the presence of human hematopoiesis. Notably, while the anti-tumor activity of CART19 and CART33 was equivalent in non-humanized mice, CART19 had better long-term activity in humanized mice, as mice treated with CART33 had late tumor relapses.

To directly compare the effects of myeloid cells as bystanders versus targets, we lentivirally transduced cord blood CD34+ cells with human CD19 so that they would be targets for CART19. We then differentiated these cells into myeloid progenitors (MP) by culturing in IMDM with 20% FBS and the following cytokines: SCF (100ng/ml), FLT3L (50ng/ml), IL-3 (10ng/ml), and GM-CSF (10ng/ml). After 10 days of culture >95% of MP expressed CD33, and were selected for >95% CD19 expression (Fig A). Addition of non-transduced MP to CAR T cell and Ramos ⁺³³ co-culture enhanced CART19-mediated tumor clearance without affecting CART19 expansion. However, the presence of MP ^CD19+ had a detrimental effect on both tumor control and expansion of CART19, confirming that CAR T cells are less effective when targeting myeloid cells. In contrast, both MP and MP ^CD19+ had similar effects on CART33 that mirrored the negative effects of MP ^CD19+ on CART19 (Fig B).

To summarize, we show that normal myeloid cells have beneficial effects on CAR T cell function, and targeting these same myeloid cells is detrimental to CAR T cell activity. These findings provide insight into the reason for suboptimal CAR T cell activity in AML, and show that independent of target antigen and CAR construct, the nature of the cells being targeted can change the effectiveness of this therapy.