GSK-3 inhibitor elraglusib enhances tumor-infiltrating immune cell activation in tumor biopsies and synergizes with anti-PD-L1 in a murine model of colorectal cancer

Abstract

Inhibition of GSK-3 using small-molecule elraglusib has shown promising preclinical antitumor activity. Using in vitro systems, we found that elraglusib promotes immune cell-mediated tumor cell killing, enhances tumor cell pyroptosis, decreases tumor cell NF-κB-regulated survival protein expression, and increases immune cell effector molecule secretion. Using in vivo systems, we observed synergy between elraglusib and anti-PD-L1 in an immunocompetent murine model of colorectal cancer. Murine responders had more tumor-infiltrating T-cells, fewer tumor-infiltrating Tregs, lower tumorigenic circulating cytokine concentrations, and higher immunostimulatory circulating cytokine concentrations. To determine the clinical significance, we utilized human plasma samples from patients treated with elraglusib and correlated cytokine profiles with survival. Using paired tumor biopsies, we found that CD45+ tumor-infiltrating immune cells had lower expression of inhibitory immune checkpoints and higher expression of T-cell activation markers in post-elraglusib patient biopsies. These results introduce several immunomodulatory mechanisms of GSK-3 inhibition using elraglusib, providing a rationale for the clinical evaluation of elraglusib in combination with immunotherapy.

Statement of significance: Pharmacologic inhibition of GSK-3 using elraglusib sensitizes tumor cells, activates immune cells for increased anti-tumor immunity, and synergizes with anti-PD-L1 immune checkpoint blockade. These results introduce novel biomarkers for correlations with response to therapy which could provide significant clinical utility and suggest that elraglusib, and other GSK-3 inhibitors, should be evaluated in combination with immune checkpoint blockade.

Figure 1.. Elraglusib triggers pyroptosis and sensitizes tumor cells to increase immune-mediated cytotoxicity in a co-culture model.

Co-cultures were treated with drug concentrations as indicated. A 1:1 effector to target (E:T) ratio was used with a 24-hour co-culture duration. EthD-1 was used to visualize dead cells, 10X magnification, scale bar indicates 100 μm. (A) SW480 and TALL-104 T cell co-culture assay images at the 24-hour timepoint. 24-hour tumor cell pre-treatment with 5 μM elraglusib, followed by 24-hour co-culture. (B) Quantification of co-culture experiment using the percentage of dead cells out of total cells (n=3). (C) Quantification normalized by cell death observed with drug treatment alone (n=3). (D) SW480 and donor-derived CD8+ T cell co-culture assay images at the 24-hour timepoint. 24-hour tumor cell pre-treatment with 5 μM elraglusib, followed by 24-hour co-culture. (E) Quantification of co-culture experiment using the percentage of dead cells out of total cells (n=3). (F) Quantification normalized by cell death observed with drug treatment alone (n=3). (G) The number of HCT 116 GFP+ cells were quantified after 48 hours of culture with DMSO, 5 μM elraglusib, and/or 5000 TALL-104 cells (n=3). (H) The number of HT-29 GFP+ cells were quantified after 48 hours of culture with DMSO, 5 μM elraglusib, and/or 5000 TALL-104 cells (n=3). (I) The number of HCT 116 GFP+ cells were quantified after 48 hours of culture with DMSO, 5 μM elraglusib, and/or 5000 NK-92 cells (n=3). (J) The number of HT-29 GFP+ cells were quantified after 48 hours of culture with DMSO, 5 μM elraglusib, and/or 5000 NK-92 cells (n=3). (K) 40X images were collected with a Molecular Devices ImageXpress^® Confocal HT.ai High-Content Imaging System. White arrows indicate pyroptotic events. Western blot analysis of (L) HCT-116 and (M) HT-29 CRC cells for expression of indicated proteins after treatment with indicated cytokines or drugs. Quantification of IFN-γ secretion (pg/mL) post-DMSO or elraglusib treatment for 24 hours in (N) TALL-104 cells and (O) NK-92 cells (n=3). Error bars represent the mean +/− standard deviation. Statistical test: one-way ANOVA with Tukey’s test for multiple comparisons. P-value legend: * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001.

Figure 2.. Elraglusib treatment induces apoptosis and suppresses survival pathways in tumor cells.

Western blot analysis of (A) HCT-116 and HT-29 CRC cells for expression of indicated proteins after increasing durations of elraglusib treatment (0–72 hours). (B) Western blot analysis of Mcl-1 expression in HCT-116 CRC cells after increasing durations of elraglusib treatment. CRC cells were treated with 1 μM elraglusib for 24 hours and treated versus untreated samples were compared in triplicate. Microarray analysis results were visualized using volcano plots for (C) HCT-116, (D) HT-29, and (E) KM12C CRC cell lines. Top down- and up-regulated genes post-elraglusib treatment as compared to controls are shown. Results were calculated using a FC cutoff of >1.5, <−1.5, and a p value of <0.05. (F) The number of genes up- or downregulated in each of the three cell lines within several signaling pathways of interest. Green indicates downregulation and red indicates upregulation of gene expression. (G) A Venn Diagram was used to compare the 3,124 genes that were differentially expressed post-treatment with elraglusib in the three colon cancer cell lines (HCT-116, HT-29, KM12C). (H) Tumor cells (HCT-116, HT-29) were treated with 1 μM elraglusib for 48 hours and cell culture supernatant was analyzed with the Luminex 200. Red indicates a positive FC and green indicates a negative FC (n=3). (I) Tumor cells (HCT-116, HT-29) were treated with 5 μM elraglusib for 48 hours and cell culture supernatant was analyzed with the Luminex 200. Red indicates a positive FC and green indicates a negative FC (n=3).

Figure 3.. Elraglusib treatment increases effector molecule secretion and induces an energetic shift in cytotoxic immune cells.

Western blot analysis of (A) TALL-104 and (B) HT-29 cytotoxic immune cells for expression of indicated proteins after increasing durations of elraglusib treatment (0–72 hours). (C) Proposed model for non-canonical NF-κB pathway activation: increased NIK expression indicates non-canonical NF-κB pathway activation which enhances the expression of chemokines and cytokines (CCL11, TNF-α, GM-CSF) and subsequently leads to increased recruitment and proliferation of cytotoxic immune cells (CD8+ T, CD4+ T, NK cells). Immune cells were treated with 1 μM elraglusib for 24 hours and treated versus untreated samples were compared in triplicate. Microarray analysis results were visualized using volcano plots for (D) NK-92 and (E) TALL-104 immune cell lines. (F) A Venn Diagram was used to compare the 124 genes that were differentially expressed post-treatment with elraglusib in the two immune cell lines (TALL-104, NK-92). (G) 10X single-cell sequencing analysis of immune cells treated with elraglusib. TALL-104 and NK-92 cells were treated with 1 μM elraglusib for 24 hours and aggregate data was visualized using a t-SNE plot. (H) Immune cells show differential expression of mitochondria-encoded genes (MT) and ribosomal genes (RB) post-treatment with elraglusib. (I) Heatmap comparing gene expression post-elraglusib treatment as compared to control. Red indicates a positive FC and green indicates a negative FC. (J) Immune cells (TALL-104, NK-92) were treated with 1 μM elraglusib for 48 hours and cell culture supernatant was analyzed with the Luminex 200. Red indicates a positive FC and green indicates a negative FC (n=3).

Figure 4.. Elraglusib enhances immune cell tumor-infiltration to prolong survival in combination with anti-PD-L1 therapy in a syngeneic murine model of MSS colon carcinoma.

(A) Experimental model overview of the syngeneic murine colon carcinoma BALB/c murine model using MSS cell line CT-26. (B) Kaplan–Meier estimator curves for all treatment groups as indicated. Statistical significance was determined using a Log-rank (Mantel-Cox) test. (C) Overview of cell lineage markers used for flow cytometric immunophenotyping analysis. 14-days post-treatment initiation immune cell subpopulations were analyzed in the spleen and tumor. (D) Splenic T cells, (E) Tumor-infiltrating T cells, (F) Splenic NK cells, and (G) Tumor-infiltrating NK cells were compared between responders (R, n=3) and non-responders (NR, n=18). NK cell subsets based on the expression of CD11b and CD27 were compared in the spleen and visualized via (H) bar graph and (I) pie chart. NK cell subsets based on the expression of CD11b and CD27 were also compared in the tumor and visualized via (J) bar graph and (K) pie chart. T cell ratios were compared in the (L) Spleen and (M) Tumor. Statistical significance was determined using two-tailed unpaired T tests. P-value legend: * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001.

Figure 5.. Responders have a more immunostimulatory tumor microenvironment as compared to non-responders.

IHC analysis of tumors 14 days post-treatment initiation or tumors from long-term mice. Non-responders (NR) and responders (R) were compared. 20X images, scale bar represents 100 μm. (A-B) CD3, (C-D) Granzyme B, (E-F) Ki67, (G-H) PD-L1, and (I-J) cleaved-caspase 3 (CC3) were compared at the 14 days post-treatment initiation timepoint, and the long-term timepoint, respectively. Statistical significance was determined using two-tailed unpaired T tests (n=6). Serum from long-term mice sacrificed was analyzed via cytokine profiling for (K) CCL21, (L) VEGFR2, (M) CCL7, (N) CCL12, (O) BAFF, (P) VEGF, (Q) IL-1 β, (R) IL-6, (S) CCL22, (T) GM-CSF, (U) CCL4, (V) TWEAK, and (W) CCL2. Responders (red) and nonresponders (black) were compared. A Kruskal-Wallis test was used to calculate statistical significance followed by a Benjamini-Hochberg correction for multiple comparisons. p values are shown for analytes that were significantly different between responders and non-responders and are ordered by significance. P-value legend: * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001.

Figure 6.. Patient plasma concentrations of cytokines correlate with progression-free survival (PFS), overall survival (OS), and in vivo response to therapy results.

Plasma samples from human patients with refractory solid tumors of multiple tissue origins enrolled in a Phase 1 clinical trial investigating a novel GSK-3 inhibitor elraglusib (NCT03678883) were analyzed using a Luminex 200 (n=19). (A) Baseline analyte concentrations in pg/mL were plotted against PFS. (B) 24-hour post-dose analyte concentrations in pg/mL were plotted against PFS. (C) Baseline analyte concentrations in pg/mL were plotted against OS. (D) 24-hour post-dose analyte concentrations in pg/mL were plotted against OS. Simple linear regressions were used to calculate significance. R squared and p values were reported. p values less than 0.05 were reported as statistically significant. Graphs are ordered from most to least significant starting at the upper left. Heatmaps show linear regression slope values, R squared values, and p values ordered by significance starting from the top. (E) Table summarizing demographics. (F) Cytokines grouped by function. FC is shown where green indicates a negative (<0) FC compared to the baseline (pre-dose) value and red indicates a positive (>0) FC. (G) Table comparing murine and human circulating biomarker trends. Red boxes indicate that an analyte concentration positively correlated with response to therapy/PFS/OS while green boxes indicate that an analyte concentration negatively correlated with response to therapy/PFS/OS.

Figure 7.. Spatial profiling of patient tumor biopsies reveals a more immunostimulatory tumor microenvironment post-treatment with elraglusib.

Patient samples were analyzed using NanoString GeoMx Digital Spatial Profiling (DSP) technology. (A) Pie charts showing biopsy timepoint, primary tumor type, metastatic biopsy tissue type, and paired/unpaired biopsy sample information breakdowns. (B) A representative region of interest (ROI) showing PanCK+ and CD45+ masking. Green indicates CK, red indicates CD45, and blue indicates DAPI staining. (C) A Sankey diagram was used to visualize the study design where the width of a cord in the figure represents how many segments are in common between the two annotations they connect. The scale bar represents 50 segments. Blue cords represent CD45+ segments and yellow cords represent panCK+ segments. (D) Heatmap of all areas of interest (AOIs). Patient IDs, immune cell locations, biopsy timepoint, biopsy tissue, primary tumor location, and segment identity information are color coded as indicated in the legend. (E) PanCK+ ROI CD39 expression plotted against time-on-study (TOS). Points are color-coded by time on study (TOS) / time on treatment with darker blue points indicating a shorter TOS or time on treatment. (F) CD45+ ROI CD163 expression plotted against TOS. (G) Volcano plot showing a comparison of CD45+ region protein expression in post-treatment biopsies and pre-treatment biopsies regardless of timepoint. Grey points are non-significant (NS), blue points have p values < 0.05, and red points have false discovery rate (FDR) values less than 0.05. The size of the point represents the log2 UQ Signal-to-noise ratio (SNR). (H) Volcano plot showing a comparison of tumor-infiltrating CD45+ immune cell segment protein expression in pre- versus post-treatment biopsies. Grey points are nonsignificant (NS), blue points have p values < 0.05, and red points have false discovery rate (FDR) values less than 0.05. The size of the point represents the log2 UQ Signal-to-noise ratio (SNR).