Adeno-associated virus delivered CXCL9 sensitizes glioblastoma to anti-PD-1 immune checkpoint blockade

Abstract

There are numerous mechanisms by which glioblastoma cells evade immunological detection, underscoring the need for strategic combinatorial treatments to achieve appreciable therapeutic effects. However, developing combination therapies is difficult due to dose-limiting toxicities, blood-brain-barrier, and suppressive tumor microenvironment. Glioblastoma is notoriously devoid of lymphocytes driven in part by a paucity of lymphocyte trafficking factors necessary to prompt their recruitment and activation. Herein, we develop a recombinant adeno-associated virus (AAV) gene therapy that enables focal and stable reconstitution of the tumor microenvironment with C-X-C motif ligand 9 (CXCL9), a powerful call-and-receive chemokine for lymphocytes. By manipulating local chemokine directional guidance, AAV-CXCL9 increases tumor infiltration by cytotoxic lymphocytes, sensitizing glioblastoma to anti-PD-1 immune checkpoint blockade in female preclinical tumor models. These effects are accompanied by immunologic signatures evocative of an inflamed tumor microenvironment. These findings support AAV gene therapy as an adjuvant for reconditioning glioblastoma immunogenicity given its safety profile, tropism, modularity, and off-the-shelf capability.

Fig. 1. Chemokine signature of glioblastoma tumors.

a Immunoblots of GBM samples showing signal intensity of 31 chemokines (n = 8). CXCL9 (undetected) is outlined in red. b Box-whisker plots of cumulative relative protein expression of immunoblots shown in panel (a). c Recombinant AAV6 vector design encoding CXCL9 and the fluorescent reporter gene RFP. d 3D IHC of RFP-labeled GL261 tumor tissue collected 1 week following AAV6-EGFP injection. The top row depicts 3D rendering at 10 x magnification, scale bar 200 µm. AAV6 transduced cells are shown in green, GFAP in red, RFP+ tumor cells in gray, and DAPI nuclear stain in dark blue. 2nd and 3rd rows depict 2D digital zoom as outlined by the yellow dashed line in the top row to enhance cellular resolution. Voxel-based co-localization between AAV6 and GFAP (2nd row) and AAV6 and tumor cells (3rd row) is shown as a separate channel (yellow or pink). Representative images selected from n = 5. e 3D IHC of AAV6-EGFP transduction in age-matched naïve control mice. The top row depicts 3D rendering at 10x magnification, scale bar 200 µm. AAV6 transduced cells are shown in green, GFAP in red, and DAPI nuclear stain in dark blue. 2nd row depicts 2D digital zoom as outlined by the yellow dashed line in the top row to enhance cellular resolution. Voxel-based co-localization between AAV6 and GFAP is shown as a separate channel (yellow). Representative images selected from n = 3. f Box-whisker quantitative summary of voxel-based AAV6 co-localization with either tumor (GL261, n = 5) or astrocytes in each tumor-bearing (n = 3) and naïve mice (n = 3). g Box-whisker plot of flow cytometry quantitation of AAV6 (EGFP + ) co-localization with either tumor cells (GL261, RFP + ) or astrocytes (GFAP + , RFP-) at 3-, 5-, and 7 days post AAV6 transduction as illustrated in the schematic. Two-way ordinary ANOVA statistical analysis performed comparing percent transduction between tumor and astrocytes across matched time points, n = 6 per time point. P-values ≤ 0.05 are considered statistically significant. Box-whisker plots display the box ranging from the first to the third quartile, the center median value, and the whiskers extend from each quartile to the minimum and maximum values. Source data are provided as a Source Data File.

Fig. 2. AAV6 tumor tropism.

3D IHC displaying geospatial distribution of AAV6 encoded transgene (BFP) in (a) GL621 and (b) KR158 tumors collected 1 week following in vivo transduction (green), n = 2 per model, scale bar 150 µm. DRAQ5 nuclear dye (pink) is used to identify tumor borders, as outlined by the white dashed line. c Intra-tumor AAV6 treatment schematic for protein detection of AAV6 encoded CXCL9. ELISA detection of CXCL9 protein in serum at one and 2 weeks following AAV6-CXCL9 or AAV6-EGFP control intracranial injection in (d) GL261 and (e) KR158 models, n = 3 per time point, per group. Age-matched naïve controls used to establish baseline CXCL9 levels indicated by dashed black line. ELISA detection of CXCL9 protein in brain tissues isolated at 1 and 2 weeks following AAV6-CXCL9 or AAV6-EGFP control intracranial injection in (f) GL261 and (g) KR158 models. Left and right hemispheres lysed separately to reflect tumor-bearing and contralateral (focal and distal) signal detection. Statistical analyses performed using two-way ANOVA analysis with Tukey’s multiple comparisons test. Age-matched naïve brain, and sham (saline) injected tumors included as negative control and tumor baseline control, with the latter represented by the dashed black line, n values shown. P-values ≤ 0.05 are considered statistically significant. Bar graphs depict group mean with error bars representing standard error of the mean. Source data are provided as a Source Data File.

Fig. 3. AAV6-CXCL9 directed lymphocyte chemotaxis.

a Diagrammatic overview of in vitro competitive T lymphocyte chemotaxis assay. b Lymphocyte (CTV + , blue) chemotaxis in AAV6-EGFP (control, green) transduced or AAV6-CXCL9 (RFP + , red) transduced GL261 field at 1- and 24 h following co-culture. Statistical analyses performed by two-way ANOVA with Sidak’s multiple comparisons test, n = 3 per time point, per group. Representative images shown. Dashed white line represents the lymphocyte-tumor border at assay start. c Competitive chemotaxis measured as described in (b) in C8-D1A astrocytes field at 1- and 24 h following co-culture. d Schematic outlining combination AAV6 and PD-1 ICB treatment and tissue collection and survival analysis. Flow cytometric detection of tumor-infiltrating CD8+ lymphocytes in control, single or combination AAV6-CXCL9 plus anti-PD-1 ICB treatment in (e) GL261 and (f) KR158 models. Fold-change normalization based on sham values (dashed black line). Statistical analysis performed using ordinary one-way ANOVA with Fisher’s least significant difference (LSD) test for multiple comparisons, n = 3–6 per treatment group, individual values shown. Flow cytometric detection of tumor-infiltrating CD4+ lymphocytes in control, single or combination AAV6-CXCL9 plus anti-PD-1 ICB in (g) GL261 and (h) KR158 models. Fold-change normalization based on sham values (dashed black line). Statistical analysis performed using ordinary one-way ANOVA with Fisher’s LSD test for multiple comparisons, n = 3–6 per treatment group, individual values shown. i 3D IHC of lymphocytes (EGFP + , green) in sham tumor control (scale bar 100 µm) and combination AAV6-CXCL9 plus anti-PD-1 ICB treated GL261 tumors (scale bar 70 µm) isolated day 15. Tissues counterstained with DAPI (blue) and GFAP (astrocytes, red). Images representative of n = 3 per treatment group. j Representative 3D IHC of a region of lymphocyte infiltration (EGFP + , green) in combination treated GL261 tumors, n = 3, scale bar 30 µm. Tissues were counterstained with DAPI nuclear dye (blue) and GFAP (astrocytes, red). Voxel-based co-localization between EGFP and GFAP indicate areas of convergence, ‘cell junctions’, between astrocytes and T cells, shown as a separate channel (white). 3D surface renderings of astrocytes and lymphocytes to visualize cell junctions in greater detail. P-values ≤ 0.05 are considered statistically significant. Bar graphs depict group mean with error bars representing standard deviation. Box-whisker plots display the box ranging from the first to the third quartile, the center median value, and the whiskers extend from each quartile to the minimum and maximum values. Source data are provided as a Source Data File.

Fig. 4. AAV6-CXCL9 sensitizes GBM tumors to anti-PD-1 immunotherapy.

Survival analysis in (a) GL261 and (b) KR158 tumor-bearing mice treated with control, AAV6-CXCL9, and anti-PD-1 ICB alone and in combination. Median survival for each treatment group shown, n = 8 per group. Statistical analysis was performed using Log-rank (Mantel-Cox) test comparing individual treatment groups. c Tile-stitch 10x 3D IHC imaging of GL261 tumors resected from sham control and combination AAV6-CXCL9 plus anti-PD-1 ICB treated GREAT mice, n = 3 per group, scale bar 300 µm. DAPI nuclear dye (blue) used to identify tumor area outlined by the dashed white line. Digital magnification of regions outlined in the far-left panel to show higher image resolution for each treatment, scale bar 50 µm. EYFP (green) correlates with IFNγ expression. d 3D IF of tissue from GL261 tumor tissue as shown in (c) immunolabeled for CD45 expression (red), scale bar 20 µm. Digital zoom of region outlined in the far-right panel shows co-localization between CD45 and IFNγ, indicating these are immune cells, n = 3, scale bar 5 µm. e Diagrammatic summary of combination treatment strategy with concomitant CD4 or CD8α antibody depletion. f Survival analysis in GL261 tumor-bearing mice treated with combination AAV6-CXCL9 plus anti-PD-1 monoclonal antibody, with or without anti-CD8α depletion. Statistical analysis performed using Log-rank (Mantel-Cox) test comparing individual treatment groups, n = 8 per group. g Survival analysis in GL261 tumor-bearing mice treated with combination AAV6-CXCL9 plus anti-PD-1 monoclonal antibody, with or without anti-CD4 depletion. Statistical analysis performed using Log-rank (Mantel-Cox) test comparing individual treatment groups, n = 7–8 per group. h Survival analysis in long-term survivors from combination treated animals re-challenged with tumor at day 55 of study (n = 5). Age-matched naïve control mice were orthotopically implanted with GL261 as survival control arm. Statistical analysis was performed using Log-rank (Mantel-Cox) test comparing individual treatment groups. P-values ≤ 0.05 are considered statistically significant. Source data are provided as a Source Data File.

Fig. 5. Immunological landscape of GBM tumors treated with AAV6-CXCL9 and anti-PD-1 immunotherapy.

a Summary of UMAP cell clusters. UMAP of cell types clustered by scRNA transcriptional analysis of 52,344 CD45+ cells isolated from GL261 tumor bearing mice treated with (b) sham (saline) or (c) combination AAV6-CXCL9 + aPD-1 treated GL261 tumors, n = 3 mice per group. Summary circle chart depicting cell cluster population frequency detected for each treatment included alongside each UMAP. Quantitative change in population frequency of (d) CD8 + T cells, (e) Treg cells, (f) monocytes, and (g) non-classical (n–c) monocytes across treatment groups. Statistical analyses performed using ordinary one-way ANOVA with Fisher’s LSD test for multiple comparisons, n = 3 per group, individual values shown. h UMAP projection of Cxcr3 transcript expression (TPM) detected across all cell clusters. i UMAP of T cell clusters. j Summary of Cxcr3 transcript expression across defined T cell clusters. Box-whisker plots display the box ranging from the first to the third quartile, the center median value, and the whiskers extend from each quartile to the minimum and maximum values. Source data are provided as a Source Data File.

Fig. 6. AAV6-CXCL9 and anti-PD-1 immunotherapy stimulates CD8 lymphocyte activation.

a Venn Diagram representing differentially expressed genes affiliated with each treatment. b Heatmap depicting scRNA-seq-derived cell-cell communication networks enriched or decreased in response to combination AAV6-CXCL9 + aPD-1 as compared to AAV6-CXCL9 + IgG2 treatment across identified cell clusters. c Heatmap depicting scRNA-seq-derived cell-cell communication networks enriched or decreased in response to combination AAV6-CXCL9 + aPD-1 as compared to AAV6-EGFP + aPD-1 treatment across identified cell clusters. d Waterfall summary plot of scRNA-seq-derived signaling pathways enriched in CD8 + T cells following combination AAV6-CXCL9 + aPD-1 as compared to AAV6-CXCL9 + IgG2 treatment. e Waterfall summary plot of scRNA-seq-derived signaling pathways enriched in CD8 + T cells following combination AAV6-CXCL9 + aPD-1 as compared to AAV6-EGFP + aPD-1 treatment. f Heatmap representation of gene expression analysis derived from all cell clusters using the nCounter® Immune Exhaustion Panel (nanoString) following AAV6-CXCL9 gene therapy with or without PD-1 ICB. CD8 + T cell populations outlined in black for each treatment group. g–l Quantification of common pathways found to be differentially regulated in CD8 + T cells in response to treatment, derived from n = 669, 674, 1099, 912, and 1656 single cells pooled from n = 3 individual samples per treatment group, graphically presented from left to right. Statistical analyses performed using Kruskal–Wallis test followed by Dunn’s multiple comparisons, with individual values shown. P-values ≤ 0.05 are considered statistically significant. Box-whisker plots display the box ranging from the first to the third quartile, the center median value, and the whiskers extend from each quartile to the minimum and maximum values. Source data are provided as a Source Data File.

Fig. 7. Inflammatory signature of preclinical GBM treated with AAV6-CXCL9 and anti-PD-1 ICB.

a Heatmap summary of scRNA-seq-derived CCL-CXC expression in CD8 + T cells isolated from GL261 tumors in response to AAV6-CXCL9 and anti-PD-1 ICB treatment created using GraphPad Prism. b–i Quantification of CCL-CXC genes found to be differentially expressed in CD8 + T cells in response to treatment derived from n = 669, 674, 1099, 912, and 1656 single cells pooled from n = 3 individual samples per treatment group, graphically presented from left to right. Statistical analyses performed using Kruskal–Wallis test followed by Dunn’s multiple comparisons, with individual values shown. j Representative immunoblots depicting chemokine and cytokine protein expression detected in GL261 tumors resected following treatment with AAV6-CXCL9 with and without PD-1 ICB (n = 3 for sham, rAAV6-EGFP + IgG2a, and rAAV6-EGFP + aPD-1; n = 4 for rAAV6-CXCL9 + IgG2a and rAAV6-CXCL9 + aPD-1). -4 per group. k Heatmap summary of CCL-CXC relative protein expression found to be differentially expressed in response to AAV6-CXCL9 with and without PD-1 ICB, created using GraphPad Prism. l Circos interactome analysis of detected differentially expressed proteins and predicted receptors. P-values ≤ 0.05 are considered statistically significant. Box-whisker plots display the box ranging from the first to the third quartile, the center median value, and the whiskers extend from each quartile to the minimum and maximum values. Source data are provided as a Source Data File.

Fig. 8. Diagrammatic summary of findings.

Intra-tumor delivery of AAV6 encoding CXCL9 results in robust transduction of tumor-reactive astrocytes, creating a chemotactic gradient of secreted CXCL9. This improves lymphocyte trafficking in combination with anti-PD-1 ICB through chemokine-receptor engagement between CXCL9 in the TME and CXCR3 expression on lymphocytes. CD8 + T cells are required for durable survival response to treatment, indicating that tumor cell killing is mediated by the adaptive arm of immunity. Combination treatment also transforms the inflammatory milieu of tumors, creating a pro-inflammatory environment evidenced by the presence of cytokines and chemokines that further promote innate and adaptive immune activation.