Spatial and functional targeting of intratumoral Tregs reverses CD8+ T cell exhaustion and promotes cancer immunotherapy

Abstract

Intratumoral Tregs are key mediators of cancer immunotherapy resistance, including anti-programmed cell death (ligand) 1 [anti-PD-(L)1] immune checkpoint blockade (ICB). The mechanisms driving Treg infiltration into the tumor microenvironment (TME) and the consequence on CD8+ T cell exhaustion remain elusive. Here, we report that heat shock protein gp96 (also known as GRP94) was indispensable for Treg tumor infiltration, primarily through the roles of gp96 in chaperoning integrins. Among various gp96-dependent integrins, we found that only LFA-1 (αL integrin), and not αV, CD103 (αE), or β7 integrin, was required for Treg tumor homing. Loss of Treg infiltration into the TME by genetic deletion of gp96/LFA-1 potently induced rejection of tumors in multiple ICB-resistant murine cancer models in a CD8+ T cell-dependent manner, without loss of self-tolerance. Moreover, gp96 deletion impeded Treg activation primarily by suppressing IL-2/STAT5 signaling, which also contributed to tumor regression. By competing for intratumoral IL-2, Tregs prevented the activation of CD8+ tumor-infiltrating lymphocytes, drove thymocyte selection-associated high mobility group box protein (TOX) induction, and induced bona fide CD8+ T cell exhaustion. By contrast, Treg ablation led to striking CD8+ T cell activation without TOX induction, demonstrating clear uncoupling of the 2 processes. Our study reveals that the gp96/LFA-1 axis plays a fundamental role in Treg biology and suggests that Treg-specific gp96/LFA-1 targeting represents a valuable strategy for cancer immunotherapy without inflicting autoinflammatory conditions.

Keywords: Cancer immunotherapy; Chaperones; Immunology; Oncology; T cells.

Figure 1. Treg-specific gp96 deletion results in tumor regression and prolonged survival in mice.

(A) Experimental schema for primary implantation and rechallenge of MC38 tumor cells in Foxp3^{eGFP-Cre-ERT2} Hsp90b1^WT/WT (WT) and Foxp3^{eGFP-Cre-ERT2} Hsp90b1^fl/fl (KO) mice. For primary implantation, WT or KO mice (8–10 weeks old; n = 9/group) received tamoxifen for 10 days (75 mg/kg, i.p.; days –10 to 0), followed by a single s.c. injection of MC38 tumor cells (2 × 10⁶ cells/mouse; day 0) into their right flank. Tumor volumes were measured daily or every 2 days (length × width in mm) using a digital caliper, starting from day 5 after tumor cell implantation. For rechallenge, all tumor-regressed KO mice and age-matched tumor-naive WT mice were rechallenged s.c. on the opposite flank with 2 × 10⁶ MC38 tumor cells 60 days after primary tumor cell implantation. Tumor growth was monitored as described. (B–D) Growth curves depict primary implantation and rechallenge with 2 × 10⁶ MC38 (B), 1 × 10⁶ MB49 (C), and 2.5 × 10⁵ B16-F10 (D) tumor cells in WT and KO mice. MB49 and B16F10 tumors were implanted following the same scheme as that used for MC38 tumor cells. n = 6–10/group. (E–G) Survival curves following primary inoculation and rechallenge with MC38 (E), MB49 (F), and B16-F10 (G) tumor cells in WT and KO mice. Mice were euthanized when tumors reached more than 16 mm in diameter. n = 6–10/group. Results are representative of more than 3 independent experiments. Tumor growth curves were analyzed by repeated-measures, 2-way ANOVA (B–D); survival incidence analysis was performed by log-rank (Mantel-Cox) test (E–G); ****P < 0.0001 (KO vs. WT).

Figure 2. Treg-specific gp96 deletion preserves immune homeostasis in mice.

Both WT and KO mice received a 10-day tamoxifen treatment (days –10 to 0). Ninety days later (D90), the mice were euthanized and specified tissues were collected for analysis. (A) Representative flow plots illustrate the distribution of CD3⁺ T cell subsets in murine SPLs and pLNs from both groups (n = 5/group), with values indicating the percentages of the specified subsets in total CD3⁺ T cells. (B–G) Absolute numbers of total lymphocytes (total LN cells) (B), CD4⁺Foxp3⁺ Tregs (C), CD4⁺Foxp3^– Teffs (D), CD44^hiCD62L^loCD4⁺ Teffs (E), CD8⁺ T cells (F), and CD44^hiCD62L^loCD8⁺ T cells (G) in SPLs and pLNs from mice of both groups. (H) Serum levels of IL-6, IFN-γ, and IL-10 were measured on day 90 following tamoxifen administration (days –10 to 0) in WT and KO mice using ELISA. n = 6–9/group. CTRL, positive control. (I) Representative H&E-stained images of the indicated organs from WT and KO mice (day 90; n = 5–8/group). Scale bars: 100 μm. Results are representative of more than 3 independent experiments. Data are shown as the mean ± SEM. *P < 0.05, ***P < 0.001, and ****P < 0.0001 (KO vs. WT). For statistical analyses, a 2-tailed, unpaired Student’s t test was performed (B–H).

Figure 3. Gp96 regulates CD11a/CD18 (LFA-1) integrin expression in Tregs and facilitates their infiltration into the TME.

(A) WT and KO mice (n = 5–8/group) were pretreated with tamoxifen (days –10 to 0) and s.c. implanted with MC38 tumors on day 0. TILs were harvested on days 7, 9, and 11. Representative flow cytometric plots and summary graphs show the percentages of CD4⁺Foxp3⁺ Tregs in specified tissues at these time points. (B) Rag2^–/– recipient mice (n = 5/group) were implanted s.c. with 2 × 10⁶ MC38 cells on day 0. TdTomato-expressing Tregs from SPLs of R26^{STOP-tdTomato} Foxp3^{eGFP-Cre-ERT2} Hsp90b1^WT/WT (TdTomato-WT) or R26^{STOP-tdTomato} Foxp3^{eGFP-Cre-ERT2} Hsp90b1^fl/fl (TdTomato-KO) donor mice were collected, preactivated, and transferred (2 × 10⁶ cells/mouse; n = 5/group) into recipient mice on day 3. On day 10, SPLs and tumors were harvested for flow cytometry. Representative flow cytometric plots and summary graphs indicate the percentages of infiltrating TdTomato⁺Foxp3⁺ Tregs among CD45⁺ cells. (C–F) WT and KO mice (n = 3/group) received tamoxifen (days -10 to 0), and splenic Tregs’ integrin expression was assessed using flow cytometry at designated time points (D–8, day –8; D–7, day –7; D–6, day –6; D–4, day –4; D10, day 10). Representative flow cytometric plots (day 10) and summary graphs (time course) show frequencies of indicated surface integrins on splenic Foxp3⁺ Tregs in WT and KO mice. (G–I) Naive CD4⁺ T cells from SPLs of C57BL/6 mice were differentiated into iTregs under Treg-skewed conditions for 2 days (days –2 to 0) followed by CRISPR/Cas9 KO of indicated integrins on day 0; cells were cultured for 3 more days (days 0–3). MC38 tumor–bearing Rag2^–/– mice (n = 3–6/group) received specific integrin-KO or nontargeting control iTregs on day 3 post-tumor implantation; SPLs and tumors were collected on day 10 for flow cytometry. Representative flow cytometric plots (G) and summary graph (H) show relative number of Foxp3⁺ Tregs (in CD45⁺ cells total). (I) Absolute numbers of Tregs in SPLs and TILs. Results represent 3 independent experiments. Data indicate the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 (KO vs. WT), by 2-tailed Student’s t test used for comparisons of different experimental groups (A–F) and 1-way ANOVA with Dunnett’s multiple-comparison test for multiple-comparison analyses (H and I).

Figure 4. LFA-1 blockade prevents Treg infiltration into the TME.

(A) Experimental scheme illustrates the process of LFA-1 blockade using anti–LFA-1 or IgG2a isotype-matched control Abs in C57BL/6 mice implanted with MC38 tumor cells. Anti–LFA-1 or isotype Abs were administered every 2 days starting from day 4 after MC38 tumor cell implantation on day 0; TIL analysis was conducted on day 9 (n = 7/group) and day 16 (n = 7/group). (B–D) Spectral flow cytometric analysis of CD45⁺ TILs from day-9 MC38 tumors treated with anti–LFA-1 or isotype Abs. LFA-1 blockade significantly reduced the frequency of cluster 1 (NK cells), cluster 2 (including CD3⁺CD4⁺Foxp3^– non-Tregs and CD3⁺CD4⁺Foxp3⁺ Tregs), and cluster 12 (CD8⁺ T cells) (highlighted in blue). As a subset of CD4⁺ T cells, Tregs expressed high levels of Foxp3 and were located at the bottom of cluster 2. (C) UMAP visualization shows the distribution of the indicated markers. (D) edgeR analysis indicating CD45⁺ TILs clusters with significant changes to frequency following treatment with anti–LFA-1 (left) versus isotype Abs (right). (E) Representative flow cytometric plots and graph depict the percentages of Foxp3⁺ Tregs in CD45⁺ TILs from day-9 MC38 tumors. (F–H) Spectral flow cytometric analysis of CD45⁺ TILs from day-16 MC38 tumors treated with anti–LFA-1 or isotype Abs. Similar to results in B, cluster 2 (NK cells), cluster 3 (including CD3⁺CD4⁺Foxp3^– non-Tregs and CD3⁺CD4⁺Foxp3⁺ Tregs), and cluster 4 (CD8⁺ T cells) exhibited reduced abundance following LFA-1 blockade (highlighted in blue). (F) FoxP3⁺ Tregs were localized at the bottom of cluster 3. (G) UMAP visualization shows the distribution of the indicated markers. (H) edgeR analysis indicating CD45⁺ TILs clusters with significant changes to frequency following anti–LFA-1 versus isotype Ab treatment. (I) Representative flow cytometric plots and graph depict the percentages of Foxp3⁺ Tregs in CD45⁺ TILs from day-16 MC38 tumors. Results are representative of 3 independent experiments. Data are shown as the mean ± SEM. ****P < 0.0001 (anti–LFA-1 vs. isotype), by 2-tailed Student’s t test for comparisons of different experimental groups (E and I). FC, fold change.

Figure 5. Gp96 deletion in Tregs reduces their CD25 expression but preserves their responsiveness to IL-2/p-STAT5 signaling activation.

(A) Volcano plot depicting DEGs in splenic Tregs from KO versus WT mice after 10-day tamoxifen treatment (days –10 to 0; n = 4/group). The x axis indicates log2 fold change, and the y axis represents –log₁₀ (corrected P value). Gray dots (NA) indicate no significant difference; blue dots (down) indicate downregulated genes in KO Tregs (adjusted P < 0.05; Ilr2a highlighted); red dots (up) indicate upregulated genes (adjusted P < 0.05). (B) GSEA of T cell activation genes between gp96-KO and WT Tregs shows FDR q = 0.03 and NES = –1.45. (C) GO enrichment scatter plots show the top 20 enriched pathways for DEGs in WT vs. gp96-KO Tregs, with dot size representing gene counts and GeneRatio indicating DEG ratios. (D) Heatmap of relative expression of selected genes encoding transcription factors, surface markers, and intracellular molecules in Tregs from WT and KO mice. Red indicates a high expression level; blue indicates a low expression level. Expression of Ilr2a (highlighted in blue) was significantly downregulated in gp96-KO Tregs. (E) UMAP visualization of splenic Tregs from WT and KO mice (top; n = 6–7/group). Mice received either PBS or IL-2/JES6-1 complex at specific time points, concurrently with tamoxifen treatment (days –10 to 0). Cluster 5 (CD25^hi Tregs) and cluster 9 (CD25^lo Tregs) were enriched in gp96-KO Tregs (highlighted in red). (F) Heatmap shows marker expression levels by cluster, with clusters 5 and 9 highlighted. (G) edgeR analysis indicates clusters significantly altered in splenic Tregs from WT (right) versus KO (left) mice receiving PBS or IL/JES6-1. Clusters 5 and 9 are highlighted. (H) Representative flow cytometric plots and summary graphs depict Foxp3 and CD25 expression in splenic Tregs from WT and KO mice on day 0, as described in E. Numbers indicate the frequencies of CD25⁺ subset in Foxp3⁺ Tregs. (I) Representative flow cytometric plots and graph depicting the ex vivo levels of p-STAT5 in splenic Tregs from WT and KO mice (n = 4–5/group) on day 0, as described in E. Numbers indicate frequencies of the p-STAT5⁺ subset in Foxp3⁺ Tregs. Results represent 3 independent experiments. Data are shown as the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, by 1-way ANOVA with Dunnett’s T3 multiple-comparison test (H and I).

Figure 6. Upon gp96 deletion, the absence of infiltrating Tregs promotes CD8⁺ TIL accumulation and activation, leading to repression of MC38 tumors.

(A) Spectral flow cytometric analysis of CD45⁺ TILs collected from day-11 MC38 tumors (implanted at 2 × 10⁶ cells on day 0) from WT and KO mice pretreated with tamoxifen (days –10 to 0; n = 6/group). Cluster 4 (CD8⁺ T cells; highlighted in red) expressing both CD3 and CD8 was significantly enriched in KO mice. Expression distribution of the indicated markers is shown in the bottom plots. (B) Representative flow cytometric plots (day 11) and summary graph (days 7, 9, 11, and 14) show the percentages of CD8⁺ T cells in CD45⁺ TILs from MC38 tumors from WT and KO mice (n = 5–8/group). (C) Representative flow cytometric plots (day 9) and summary graphs (days 7, 9, 11, and 14) show the frequencies of the CD44^hiCD62L^lo population within CD8⁺ TILs from MC38 tumors grown in WT and KO mice (n = 5–8/group). (D) Experimental schema depicts the administration of CD8-depleting Ab (anti-CD8a Ab) or matched isotype Ab treatment in MC38 tumor–bearing WT and KO mice (2 × 10⁶ MC38 cells/mouse, s.c.; n = 6–7/group) that received tamoxifen before treatment (days –10 to 0). (E) Tumor growth curves of MC38 cells grown in WT and KO mice receiving anti-CD8a or isotype Ab treatment. (F) Experimental schema outlines the ACT of TdTomato⁺ WT or KO Tregs and/or CD8⁺ T cells into Tcrbd^–/– mice (recipient) following MC38 tumor cell implantation (0.6 × 10⁶ cells, s.c.). TdTomato⁺ WT or gp96-KO Tregs from SPLs of TdTomato⁺ WT and TdTomato⁺ KO donor mice were isolated, preactivated, and adoptively transferred into mice bearing day-2 MC38 tumors. In parallel, CD8⁺ T cells were isolated from dLNs of C57BL/6 mice bearing day-12 MC38 tumors, stimulated in vitro for 3 days, and transferred into recipient mice on day 4. (G–I) MC38 tumor growth curves among the indicated 5 groups of mice (n = 4–7/group). Results are representative of 3 independent experiments. Data are shown as the mean ± SEM. **P < 0.01, ***P < 0.001, and ****P < 0.0001 (KO vs. WT), by 2-tailed Student’s t test for comparisons of different experimental groups where multiple comparisons were not performed (B and C) and repeated-measures 2-way ANOVA for analysis of tumor growth curves (E, H and I).

Figure 7. Upon gp96 deletion, the lack of infiltrating Tregs inhibits TOX-mediated CD8⁺ TIL exhaustion in MC38 tumors.

(A) Spectral flow cytometric analysis of CD8⁺ TILs collected from day-14 MC38 tumors grown in WT and KO mice (n = 7/group). Cluster 11 (CD44^hiCD62L^loICOS⁺CD8⁺ Teffs; highlighted in red) increased in the KO group; clusters 8 and 19 (CD44^hiCD62L^loTcf1^–TOX⁺PD-1⁺Lag3⁺Tim3⁺CD8⁺ subsets; highlighted in blue) decreased in the KO group. Expression distribution of the indicated markers is shown in the bottom plots. (B) Heatmap of marker expression by cluster. (C) Representative flow cytometric plots and summary graphs compare the percentages of the Tcf1^–TOX⁺ and Tim3⁺Lag3⁺ population within CD44^hiCD62L^loCD8⁺ TILs from day-7, -9, -11, and -14 MC38 tumors grown in WT and KO mice pretreated with tamoxifen (days –10 to 0; n = 5–8/group). (D) Representative flow cytometric plots and summary graph indicating the percentages of IFN-γ⁺TNF-α⁺ CD8⁺ TILs from day-14 MC38 tumors grown in WT and KO mice (n = 6/group) after a 5-hour ex vivo TCR stimulation with anti-CD3 and anti-CD28 Abs. Results are representative of more than 3 independent experiments. Data are shown as the mean ± SEM. ****P < 0.0001 (KO vs. WT), by 2-tailed Student’s t test for comparisons of different experimental groups (C and D).

Figure 8. Depletion of Tregs attenuates TOX-mediated CD8⁺ TIL exhaustion in MC38 tumors from Foxp3^DTR mice.

(A) Experimental schema illustrates the process of DT or PBS treatment in Foxp3^DTR mice (n = 7–8/group) implanted with MC38 tumor cells (day 0; 2 × 10⁶ MC38 cells/mouse). (B) Tumor growth curves of MC38 tumor cells grown in Foxp3^DTR mice receiving DT or PBS treatment. (C) Spectral flow cytometric analysis of CD8⁺ TILs collected from day-17 MC38 tumors grown in Foxp3^DTR mice. Cluster 3 (CD44^hiCD62L^loTcf1^–TOX⁺PD-1⁺Lag3⁺Tim3⁺CD8⁺ subset; highlighted in blue) decreased in the DT group. Expression distribution of the indicated markers is shown in the bottom plots. (A) Heatmap visualization of marker expression by cluster. (E–H) Representative flow cytometric plots and summary graph indicate the percentages of Foxp3⁺ Tregs (within CD4⁺ TILs) (E), the CD44^hiCD62L^lo population (within CD8⁺ TILs) (F), the Tcf1^–TOX⁺ subset (within CD44^hiCD62L^loCD8⁺ TILs) (G), and Tim3⁺Lag3⁺ cells (within CD44^hiCD62L^loCD8⁺ TILs) (H) among groups. Results are representative of 3 independent experiments. Data are shown as the mean ± SEM. ****P < 0.0001 (DT vs. PBS), by repeated-measures 2-way ANOVA for tumor growth curves (B) and 2-tailed Student’s t test for comparisons of different experimental groups (E–H).

Figure 9. In the absence of Tregs, IL-2 blockade induces TOX-mediated CD8⁺ TIL exhaustion, promoting MC38 tumor growth.

(A) Experimental schema depicts the administration of IL-2–blocking Abs (S4B6-1 and JES6-1) or matched isotype Abs in MC38 tumor–bearing WT and KO mice (2 × 10⁶ MC38 cells/mouse, s.c.; n = 6–8/group). (B) Representative flow cytometric plots and summary graphs show the frequencies of Tcf1^–TOX⁺ and Tim3⁺Lag3⁺ cells within CD44^hiCD62L^loCD8⁺ TILs from day-11 MC38 tumors grown in WT and KO mice (n = 6–7/group). (C) Growth curves of MC38 tumors grown in WT and KO mice that received IL-2 blockade (group 4; 120 μg for each Ab) or isotype Ab treatment (group 1). Tumor sizes were measured. (D) Spectral flow cytometric analysis of CD8⁺ TILs isolated from day-11 MC38 tumors grown in WT and KO mice treated with IL-2 blockade (group 4) or isotype Ab (group 1). Cluster 11 (CD44^hiCD62L^loTcf1^–TOX⁺PD-1⁺Lag3⁺Tim3⁺CD8⁺ subset), shown as circles, is highlighted in green. Expression distribution of the indicated markers is also shown in the plots on the right. (E) Heatmap of the indicated markers by cluster. (F) edgeR analysis indicates significant differences in CD8⁺ TIL clusters between KO and WT groups (cluster 11 is highlighted). (G) Significant differences in cluster 11 enrichment among the indicated groups. Green indicates low abundance of cluster 11; purple indicates high abundance of cluster 11. Results are representative of 3 independent experiments. Data are shown as the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 (IL-2 blockade vs. isotype). A 2-tailed Student’s t test was performed for comparisons of different experimental groups (G) where multiple comparisons were not performed; for analyses where multiple comparisons were performed, a 1-way ANOVA with Šidák’s multiple-comparison correction was performed for comparison of treatment groups across genotypes (B); a 1-way ANOVA analysis with Dunnett’s multiple-comparison correction was performed for comparison within groups for treatments compared with the respective isotype (B). Tumor growth were curves were analyzed by repeated-measures 2-way ANOVA (C).