Monocytes Are Required for Both Optimal Anti-Leukemic Efficacy and the Cytokine Release Syndrome By CAR-T Cells: Lessons from an Innovative Xenotolerant Mouse Model

Background: Chimeric antigen-receptor (CAR)-engineered T cells promise to cure chronic and acute leukemias refractory to standard treatments. Before this promise is fulfilled, however, two crucial issues need to be solved: i) how to circumvent the emergence of secondary resistance (e.g. due totarget-antigen loss; leukemic lineage switch); ii) how to manage associated toxicities (e.g. the cytokine release syndrome, CRS; lineage aplasias). Unfortunately, all these issues cannot be addressed pre-clinically in currently available NSG mouse models, because they lack human hematopoiesis and, furthermore, ultimately develop xenograft-versus-host disease (X-GVHD), preventing the evaluation of long-term effects.

Methods: We have developed an innovative xenotolerant model by transplanting human hematopoietic stem cells (HSCs) intraliver in newborn NSG mice triple transgenic for human SCF, GM-SCF and IL-3 (SGM3). Differently from "classical" NSG, SGM3 mice reconstituted high levels of human T cells (>1000 cells per microL at week 8), which, once transferred in secondary recipients, persisted up to 200d without causing X-GVHD, even after irradiation. Robust and specific xenotolerance was confirmed by in vitrohyporesponsiveness to NSG, bot not to C57/Bl6 antigens (irradiated splenocytes) or human HLAs (PBMCs). Secondary transfer experiments in leukemic and/or HSC-humanized SGM-3 mice have been then designed for studying the determinants of CAR-T cell efficacy and associated toxicities in the absence of confounding xenoreactivity.

Results: SGM3-derived T cells were transduced ex vivo with either a CD19 or a CD44v6 CAR (both having a CD28 2G design) after activation with CD3/CD28-beads and IL-7/IL-15, resulting in a preferential and functional CD45RA+/CD62L+/CD95+ stem memory T cell (T_SCM) phenotype. Once transferred in secondary recipients previously engrafted with a CD19+/CD44v6 leukemic cell line, CD19 or CD44v6 CAR-T cells equally mediated rapid tumor clearance both in low and high tumor-burden settings, in the absence of malaise or elevated human IL-6 levels in vivo. At later time points (after 100d), however, approximately 50% of responding mice relapsed despite significant CAR-T cell persistence in vivo (>50 cells per microL). A significant fraction of leukemia relapses were characterized by post-transcriptional down-regulation of CD44v6 expression or CD19 loss, respectively. Conversely, secondary transfer of SGM3-derived CAR-T cells in leukemic SGM3 mice that had been previously humanized with HSCs resulted in the development of a clinical syndrome similar to the CRS observed in clinical trials (high fevers, elevated IL-6, TNF-alpha and serum amyloid A levels - mouse analog of C-reactive protein in humans), resulting in 30% lethality. This CRS was anticipated and shortened for CD44v6 compared with CD19 CAR-T cells and worse in the case of 4-1BB compared with the original CD28 2G CAR designs. Strikingly, mice recovering from the CRS benefited from durable leukemic remissions, yet experienced long-lasting CD19+ B-cell or CD44v6+ monocyte aplasias. Deepness of remission was confirmed in "tertiary" recipients, which did not develop leukemia after the infusion of bone-marrow cells from mice in remission 150d since CAR-T cell infusion. Interestingly, in this model, tocilizumab administration at the time of either CD19 or CD44v6 CAR-T cell infusion efficiently prevented the CRS, but did not interfere with their comparable and long-term anti-leukemic effects. Conversely, depleting monocytes/macrophages before therapeutic CAR-T cell infusion by either lyposomal clodronate or by the prophylactic CD44v6 CAR-T cells inhibited CRS development, but also resulted in significantly worse leukemia-free survival (at 250d, 0% vs 80%, P<0.0001).

Conclusions: A number of lessons can be learned from this innovative xenotolerant mouse model of CAR-T cell immunotherapy: monocytes are required for both i) optimal anti-leukemic efficacy, and ii) the occurrence of CRS; iii) tocilizumab prevents the CRS without interfering with efficacy; iv) monocyte aplasia induced by CD44v6 CAR-T cells does not impact on their efficacy, at least in the theraeputic setting, and may ameliorate CRS toxicity. As for CD44v6 CAR-T cells, this model could be used for effectively predicting the efficacy and associated toxicities of new CAR-T cell therapies, speeding up their clinical development.