Reprogramming CD8+ T-cell Branched N-Glycosylation Limits Exhaustion, Enhancing Cytotoxicity and Tumor Killing

Abstract

T-cell therapies have transformed cancer treatment. Although surface glycans have been shown to play critical roles in regulating T-cell development and function, whether and how the glycome influences T cell-mediated tumor immunity remains an area of active investigation. In this study, we show that the intratumoral T-cell glycome is altered early in human colorectal cancer, with substantial changes in branched N-glycans. We demonstrated that CD8+ T cells expressing β1,6-GlcNAc-branched N-glycans adopted an exhausted phenotype, marked by increased PD1 and Tim3 expression. CRISPR-Cas9 deletion of key branching glycosyltransferase genes revealed that Mgat5 played a prominent role in T-cell exhaustion. In culture-based assays and tumor studies, Mgat5 deletion in CD8+ T cells resulted in improved cancer cell killing. These findings prompted the assessment of whether MGAT5 deletion in anti-CD19 chimeric antigen receptor (CAR) T cells could enable this therapeutic modality in a solid tumor setting. We showed that MGAT5 knockout anti-CD19-CAR T cells inhibited the growth of CD19-transduced tumors. Together, these findings show that MGAT5-mediated branched N-glycans regulate CD8+ T-cell function in cancer and provide a strategy to enhance the antitumor activity of native and CAR T cells.

Figure 1:. A T cell glycosylation switch towards β1,6-branched N-glycans occurs during progression to malignancy, associated with loss of function on T cells.

A) Histochemistry on normal colon (N), sessile serrated lesions (SSL), SSL with dysplasia (SSL-D), and serrated colorectal cancer (Ser CRC) samples using the lectin L-PHA that recognized β1,6-branched N-glycans, as well as immunohistochemistry for CD3 and CD8, and the respective quantifications. L-PHA quantification was performed on ImageJ; CD3 and CD8 semi-quantification was performed by 2 independent evaluators (0=<25%, 1=25–50%, 2=50–75%, and 3=>75%). Normal colon: n=7 (L-PHA), n=5 (CD3, CD8); SSL: n=21; SSL-D: n=22; Ser CRC: n=12. Image magnification 400x; scale bar, 50 μm. B) Imaging mass cytometry (IMC) representative images on normal colon (N), sessile serrated lesions (SSL) and serrated colorectal cancer (Ser CRC) samples depicting Keratin and E-cadherin in blue, CD8 in red and Granzyme B (GzB) in green. C) Number of CD8⁺ T cells and D) number of GzB producing CD8⁺ T cells. Data representative of 2 donors/4 regions of interest for normal colon and 3 donors/6 regions of interest for SSL and Ser CRC. The regions of interest from the same donor are represented in the graphs with the same color. E) Relative expression of MGAT5 from FFPE samples of normal colon (n=4), SSL (n=37), SSL-D (n=23) and Ser CRC (n=12). All samples were normalized to 18S housekeeping gene expression. F) Pearson correlation of MGAT5 mRNA expression versus CD3 mRNA expression or G) CD8 mRNA expression, considering all FFPE samples. H) Expression of L-PHA on CD8⁺ T cells from fresh colonic biopsies from healthy donors (n= 14) and SSL patients (SSL without and with dysplasia; n=30). I) Expression of GNA on CD8⁺ T cells from fresh colonic biopsies from healthy donors (n= 6) and SSL patients (SSL without and with dysplasia; n=30). J) Ratio of L-PHA and GNA expression on CD8⁺ T cells from fresh colonic biopsies from healthy donors (n= 6) and SSL patients (SSL without and with dysplasia; n=16). K) Pearson correlation of L-PHA expression and % of IFNγ-producing CD8⁺ T cells or L) % of GzB-producing CD8⁺ T cells from SSL fresh colonic biopsies (SSL without and with dysplasia; n=15). M) Pearson correlation of MGAT5 mRNA expression versus CTLA4 or N) TIGIT mRNA expression from a public CRC serrated dataset (GEO: GSE76987). Kruskal-Wallis test with uncorrected Dunn’s test: (A), (C), (D) and (E); Pearson correlation: (F), (G), (K), (L), (M) and (N); Unpaired Mann-Whitney test: (H), (I) and (J); *, p < 0.05; **, p < 0.01; ***, p < 0.001 and ****, p < 0.0001; error bars represent SEM of different biological samples. See also Figure S1.

Figure 2:. Exhausted T cells presented an increased expression of β1,6-branched N-glycans.

A) Schematic representation of the experimental design: MC38-OVA cells were subcutaneously inoculated on C57BL/6 WT mice and when the tumors become visible, naïve OT-I T cells were adoptive transferred to mice. Afterwards, TILs were isolated in three different timepoints (day 5, 9 and 14) and characterized. B) Characterization of L-PHA expression on OT-I TILs based on the expression of PD1, Tim3 and CD39 in all the timepoints (n=6 for D5; n=8 for D9 and n=6 for D14) and representative histograms. C) Schematic representation of the experimental design: Naïve CD8⁺ T cells were isolated from C57BL/6 WT mice and activated with αCD3/αCD28 Dynabeads in the presence of IL-2 for 24 hours. After activation, cells are plated with or without continuous TCR stimulation (via αCD3/αCD28 Dynabeads) under normoxic (20% O₂) or hypoxic (1.5% O₂) oxygen conditions for five days. D) PD1 and Tim3 expression on CD8⁺ T cells in the different conditions (AN; AH; CN and CH). Representative flow cytometry plots with a gate on PD1/Tim3 percentage of cells are shown (n=4). E) L-PHA expression on CD8⁺ T cells (n=4). Repeated measures One-way ANOVA with uncorrected Fisher’s LSD test: (B), (D) and (E); *, p < 0.05; **, p < 0.01; ***, p < 0.001 and ****, p < 0.0001; error bars represent SEM of different biological sample. Acute activation in normoxia (AN); Acute activation in hypoxia (AH); Chronic activation in normoxia (CN); Chronic activation in hypoxia (CH). See also Figure S2.

Figure 3:. Removal of β1,6-branched N-glycans is able to decrease exhaustion phenotypes in T cells, increasing its activation.

A) Schematic representation of the protocol performed for CRISPR/Cas9 KO. Briefly, naïve OT-I T cells were cultured with αCD3/αCD28 and IL2. After 24h, Cas9 Nuclease V3 was mixed with sgRNA to form the RNP complex, CRISPR/Cas9 transfection was performed and cells were expanded for 7 days before performing the experiments. B) Schematic representation of the N-glycosylation pathway highlighting the synthesis of complex branched N-glycans. C) L-PHA expression on non-target T cells (NC) vs. Glycoengineered T cells (Mgat1 (n=5), Mgat2 (n=7) and Mgat5 KO (n=9)) and representative histograms. D) PD1 and Tim3 expression. Representative flow cytometry plots with a gating on PD1 and Tim3 percentage of cells among live CD8⁺ T cells are shown (n=4 for Mgat1 KO, n=6 for Mgat2 KO and n=8 for Mgat5 KO). E) Tox expression and F) Tcf1 expression on NC vs. Glycoengineered T cells (n=4/5 for Mgat1 KO, Mgat2 KO and n=4/6 for Mgat5 KO, respectively) and the respective representative histograms. G) Tbet and Eomes expression. Representative flow cytometry plots with a gating on Tbet and Eomes percentage of cells among live CD8⁺ T cells are shown (n=5 for Mgat1 KO, n=4 for Mgat2 KO and n=6 for Mgat5 KO). H-I) seahorse metabolic assay with sequential administration of oligomycin (O), FCCP (F), 2DG, and rotenone/antimycin A (R/A) to measure H) oxygen consumption rate (OCR) and I) extracellular acidification rate (ECAR) on NC vs. Glycoengineered T cells (n=4/group). J) CD69 expression after αCD3/αCD28 5 hours restimulation (n=4 for Mgat1 KO, n=6 for Mgat2 KO and n=8 for Mgat5 KO) and representative histograms. K) CD69 expression on CD8⁺ T cells from Rag1^CreMgat1^fl/fl, Rag1^CreMgat2^fl/fl and Mgat5 KO mice vs. the respective controls (n=5) upon 48h αCD3/αCD28 activation, as described in Figure S3C, and representative histograms. L) Percentage of proliferative cells (CFSE) on Rag1^CreMgat1^fl/fl (n=3), Rag1^CreMgat2^fl/fl (n=3) and Mgat5 KO mice (n=4) vs. the respective controls upon 48h αCD3/αCD28 activation. M-P) On day 7, cells (NC vs. Glycoengineered CD8⁺ T cells) were restimulated with αCD3/αCD28 for 1, 5, 15, 30 or 60 min. Western blotting was then conducted on lysates from stimulated cells. M) Representative blots for all proteins are shown; with quantification for N) p-ERK1; O) p-Zap70 and P) p-Plcγ (n=3/group). Repeated measures One-way ANOVA with uncorrected Fisher’s LSD test: (C)-(F), (H)-(J); One-way ANOVA with uncorrected Fisher’s LSD test: (K) and (L); Repeated measures Two-way ANOVA with uncorrected Fisher’s LSD test: (G) and (N)-(P). *, p < 0.05; **, p < 0.01 and ****, p < 0.0001; error bars represent SEM of different biological sample/ independent experiments. See also Figure S3.

Figure 4:. Mgat5 KO glycoengineered T cells show increased cytotoxicity, promoting tumor killing.

A) Schematic representation of the experimental design: NC or Glycoengineered T cells (Mgat1, Mgat2 and Mgat5 KO) were cocultured with MC38-OVA to evaluate its tumor killing capacity over time. B) PD1 and Tim3 expression. Representative flow cytometry plots with a gating on PD1 and Tim3 percentage of cells among live CD8⁺ OT-I T cells are shown (n=7). C) CD69 expression in coculture with MC38-OVA for 5h (n=7) and representative histograms. D) Day 7 OT-I CD8⁺ T cells (NC vs. Mgat5 KO) were cocultured with MC38-OVA cells at a 1:16 effector:target ratio in an Agilent xCELLigence RTCA DP system, to monitor target cell killing. XY plot on the left shows % Cytolysis ± SEM. E) Day 7 NC vs. Mgat5 KO OT-I CD8⁺ T cells were cocultured with MC38-OVA cells at a 1:1 effector:target ratio and CD107a expression on OT-I CD8⁺ T cells was assessed and representative cytometry plots with a gating on CD107a percentage of cells among live CD8⁺ T cells are shown (n=6). F) Day 7 EV (empty vector) vs. Mgat5 overexpressing OT-I CD8⁺ T cells were cocultured with MC38-OVA cells at a 1:16 effector:target ratio in an Agilent xCELLigence RTCA DP system to monitor target cell killing. XY plot on the left shows % Cytolysis ± SEM. Two-way ANOVA with uncorrected Fisher’s LSD test: (B) and (D); Paired t-test: (C) **, p < 0.01 and ****, p < 0.0001; error bars represent SEM of different biological sample. See also Figure S4.

Figure 5:. Mgat5 KO glycoengineered T cells present an effector function phenotype with increased cytotoxicity, promoting tumor killing in vivo.

A) Schematic representation of the experimental design: Mgat5 KO Glycoengineered OT-I T cells were adoptive transferred in MC38-OVA-bearing mice and 1) tumor growth and survival were assessed over time and 2) tumor infiltrated lymphocytes (TILs) were isolated after 20 days of tumor growth. B) Tumor growth curves and C) overall survival of Mgat5 KO Glycoengineered OT-I T cells comparing to non-target cells (NC) and PBS. D) IFNγ and TNFα expression on OT-I TILs upon SIINFEKL restimulation for 5 hours. Representative flow cytometry plots with a gating on IFNγ and TNFα percentage of cells among live OT-I TILs are shown (n=13 for NC vs. n=15 for Magt5 KO). E) Percentage of GzB-producing OT-I TILs upon SIINFEKL restimulation for 5 hours. Representative flow cytometry plots with a gating on GzB percentage of cells among live OT-I TILs are shown (n=13 for NC vs. n=15 for Magt5 KO). F) PD1 and Tim3 expression. Representative flow cytometry plots with a gating on PD1 and Tim3 percentage of cells among live OT-I TILs are shown (n=16 for NC vs. n=18 for Magt5 KO). G) Percentage of CD39 expressing OT-I TILs and representative flow cytometry plots with a gating on CD39 percentage of cells among live OT-I TILs (n=16 for NC vs. n=18 for Magt5 KO). H) Volcano plot depicting the differential expressed genes in OT-I TILs NC vs. Magt5 KO. I) Heatmap from RNA sequencing of NC and Mgat5 KO OT-I TILs, regarding the expression of cellular respiration; T cell effector functions and T cell memory/ stemness related genes. J) Pathway analysis of gene sets enriched in Mgat5 KO OT-I TILs. Data represent three (B-E) or four independent experiments (F-G). Two-way ANOVA with uncorrected Fisher’s LSD test: (B) and (D); Unpaired t-test: (E) - (G); *, p < 0.05; **, p < 0.01; ***, p < 0.001 and ****, p < 0.0001; error bars represent SEM of different biological samples. See also Figure S5.

Figure 6:. MGAT5 KO improves CAR-T cell therapy against solid tumors.

A) Schematic representation of the experimental design: CD3⁺ T cells were isolated from healthy PBMCs donors and activated with αCD3/αCD28 dynabeads for 48 hours. Afterwards, CRISPR/Cas9 transfection and CAR transduction was performed and cells were expanded for 12 days before performing the experiments. B) Day 12 CAR-T cells and untransduced (UTD) T cells were cocultured with A549-hCD19 cells at 1:1 effector:target ratio in an Agilent xCELLigence RTCA DP system to monitor target cell killing. XY plot on the left shows % Cytolysis ± SEM. C-D) Day 12 CAR-T cells and UTD T cells were cocultured with A549-hCD19 cells at a 1:1 effector:target ratio and CD107a expression was assessed (n=4) on C) CD8⁺ T cells and D) CD4⁺ T cells. Representative cytometry plots with a gating on CD107a percentage of cells among live CD8⁺ T cells and CD4⁺ T cells are shown. E) A549-hCD19 tumor growth curves and F) survival of NSG mice who had received human CD19-CAR-T cells or UTD T cells. Data represent one experiment using one human donor. Two-way ANOVA with uncorrected Fisher’s LSD test: (B) and (E); Repeated measures One-way ANOVA with uncorrected Fisher’s LSD test: (C) and (D); *, p < 0.05; **, p < 0.01; and ****, p < 0.0001; error bars represent SEM of different biological sample. See also Figure S6.

Figure 7:. Reprogramming CD8⁺ T Cell N-Glycosylation as an innovative strategy for enhancing tumor cell killing.

Cellular therapies, such as CAR-T cells have revolutionized treatment of blood cancers, yet remain ineffective in solid tumors. While T cell surface glycans play essential roles in the regulation of T cell activity and function, the biological relevance of the T cell glycome in tumor immunity remains largely unexplored. Here, we show that the intratumor T cell glycome composition is regulated in early stages of human colorectal cancer, displaying significant alterations in MGAT5-mediated branched N-glycans expression, which imposes suppressive functions to T cells. Mechanistically, CD8⁺ T cells expressing β1,6-GlcNAc branched N-glycans exhibit an exhausted phenotype with increased PD1 and Tim3 expression. The deletion of Mgat5 glycogene on CD8⁺ T cells by CRISPR/Cas9, revealed a critical role for Mgat5-mediated branched N-glycans in T cell exhaustion. We show that MGAT5 deletion in CAR-T cells results in marked enhancement of antitumor properties with increased cytotoxicity (increased degranulation by CD107a expression). Together, our findings establish the role of MGAT5-mediated branched N-glycans as a checkpoint that regulates T cell activity in cancer with promising impact in immunotherapy through innovative Glyco-CAR-T cells.