The NFκB/COX2 Pathway Mediates the Inhibition of CAR-T Cell Proliferation and Function By Bone Marrow Stroma

Chimeric Antigen Receptor (CAR)-T cells have proven highly successful in the treatment of some haemopoietic neoplasms. However, in many cases the treatment fails and/or disease relapses. Amongst the different factors that impact on a negative clinical outcome, an unfavourable immunosuppressive microenvironment has been proposed as playing a critical role. One of the key molecules in the microenvironment that have been identified as impairing anti-tumor immunity is prostaglandin-E2 (PGE2). We hypothesised that bone marrow mesenchymal stromal cells (BM-MSC) acquire the ability to secrete PGE2 after exposure to CAR-T cells activated against leukemic cells. We used a CAR-T cell consisting of a CD33 binder and CD28z-OX40 costimulatory moiety. BM-MSC were exposed to the CAR following in vitro activation on CD33-coated plates for 24 hours. To model the different ways by which stromal cells can license their immunosuppressive activity, BM-MSC were exposed to the supernatant harvested from the cultures of activated CARs or in direct contact with the activated CARs themselves. After incubation with the CAR-conditioned medium (CAR-cm), BM-MSCs were investigated for their transcriptome by RNASeq. Gene ontology analysis of the transcriptome revealed a strong NF-κB and STAT1 signature in BM-MSC exposed to the CAR-cm compared to those which were not exposed. Amongst the most upregulated molecules, we identified several chemokines (CXCL10, CCL8, CXCL8, and CXCL9), PDL1 and CECAM1, IL-6, indoleamine-2,3 dyoxigenase (IDO), and most prominently PTGES and PTGS2. When CAR-cm licensed BM-MSC were incubated with CD33-activated CAR-T cells, we observed that CAR proliferation, cytotoxicity and cytokine production was significantly inhibited. CAR function was restored if Rel-A – a REL-associated protein critically involved in NF-κB heterodimer formation, nuclear translocation and activation – was inhibited in BM-MSC by shRNA before exposure to CAR-cm. We previously described that BM-MSC undergo apoptosis when in contact with activated cytotoxic T cells and that this also triggers their immunosuppressive capability. CD33-activated CAR-T cells were incubated with BM-MSC and the proportion of annexin-V positive BM-MSC was assessed after 4 hours. We observed that a large fraction of BM-MSC underwent apoptosis and that this was dependent on caspase activation and the release of cytolytic granules in a contact-dependent fashion. The supernatant of apoptotic BM-MSC but not the terminally apoptotic and neither the live cells exhibited a strong inhibitory activity on CAR-T cell proliferation, cytotoxicity and effector cytokine production. Similarly to what described with the CAR-cm induced profile, BM-MSC immunosuppression required NF-κB and was effected via the COX2 pathway. We finally tested the impact of apoMSC in vivo in a xenograft model of AML (Fig. 1). CD33-positive KG1 cells were injected in NSG mice and 2 weeks later, CD33-specific CAR-T were administered with or without (controls) apoMSC. Peripheral blood samples were collected 7, 14, and 21 days after. We observed an inhibitory effect of apoMSc on the in vivo expansion of CAR-T cells which was accompanied by disease progression. The addition of celecoxib at the time of infusion of apoMSC, restored the ability of CAR-T cell to expand in vivo and control leukemic cell increase in the peripheral blood.

We conclude that BM-MSC can be induced to acquire immunosuppressive properties by activated CAR-T cells through 2 different but overlapping mechanisms, both dependent on NF-κB and COX2 signalling pathway. Our work provides critical information to overcome resistance to CAR-T cell treatment by neutralising the negative effects of the stromal microenvironment.