Combining De-Glycosylating Agents with CAR-T Cells for Targeting Solid Tumors and Reducing Toxicity

Background: The adoptive transfer of CAR-T cells have shown impressive results against B-cell malignancies, but still limited efficacy against solid tumors. The discovery of the key factors regulating the activity of CAR-T cells is required to improve their antitumor potency and modulate toxicities. Since solid tumors display a wide range of glycosylation alterations, including increased N-glycan branching, we hypothesized that peptidic epitopes may be masked by glycans from CAR-T cell targeting, especially in richly glycosylated proteins.

Results: To investigate if sugar chains may be sterically hulking for CAR-T cell targeting, we generated N-glycosylation-defective pancreatic tumor cell lines. This aim has been achieved by knocking-out the expression of the glycosyltransferase Mgat5, a key enzyme involved in the process of N-glycan branching, using the CRISPR-Cas9 technology. As model antigens for CAR targeting, we focused on CD44v6 and CEACAM-5 (CEA) since they are both heavily glycosylated proteins over-expressed on a wide variety of solid tumors, including pancreatic adenocarcinoma. Strikingly, the impairment of N-glycosylation resulted in a dramatic increase of tumor targeting by both CD44v6 (4-fold, p<0,001) and CEA CAR-T cells (10-fold, p<0,001). This effect associated with improved CAR-T cell activation, suggesting more proficient antigen engagement. To exploit this mechanism in order to increase the efficacy of CAR-T cells against solid tumors, we sought to block tumor N-glycosylation with the clinical-grade glucose/mannose analogue 2-Deoxy-D-glucose (2DG). Similarly to genetically induced glycosylation blockade, treatment with 2DG also sensitized tumor cells to recognition by CAR-T cells, significantly increasing their elimination (CD44v6: 3-fold, p<0,01; CEA: 13-fold, p<0,001). Notably, 2DG alone proved to be ineffective as mono-therapy, suggesting a synergistic effect with CAR-T cells. To get more insights on this mechanism, we took advantage of previous studies reporting that 2DG interference with N-glycosylation can be reverted by the addition of exogenous mannose. Of notice, mannose did revert the synergy between 2DG and CAR-T cells (p<0,05), implying that blockade of N-glycosylation rather than glycolysis is the crucial mechanism involved. These findings were further confirmed by using the N-glycosylation inhibitor tunicamycin (CD44v6: 2,5-fold; CEA: 5-fold, p<0,01) and by Western blot, looking at the presence of de-glycosylated proteins on tumor cell surface after 2DG treatment. Next, we challenged the combined approach in a pancreatic adenocarcinoma xenograft mouse model. Accordingly with in vitro data, mice receiving CAR-T cells highly benefited from 2DG administration (5-fold less tumor at 7d, p<0,05), which conversely was unable to mediate any antitumor effect alone. Interestingly, improved antitumor activity was accompanied by a decrease in the frequency of CAR-T cells expressing one or more exhaustion and senescence markers, such as TIM-3, LAG-3, PD-1 and CD57 (SPICE software analysis, p=0,0105). Finally, thanks to metabolic deregulation (Warburg effect), 2DG is expected to selectively accumulate in cancer cells compared to healthy tissues, supporting the safety of the combined approach. Accordingly, we observed that the same doses of 2DG able to enhance tumor cell recognition by CAR-T cells failed to increase the elimination of healthy cells, such as keratinocytes.

Conclusions: Our results indicate that i) the glycosylation status of tumor cells regulates the efficacy of CAR-T cells, especially when targeting highly glycosylated antigens, and ii) combining CAR-T cells with the de-glycosylation agent 2DG, which preferentially accumulates in tumor masses, may pave the way for a successful immunotherapy against solid tumors.