NOTCH-Induced MDSC Recruitment after oHSV Virotherapy in CNS Cancer Models Modulates Antitumor Immunotherapy

Abstract

Purpose: Oncolytic herpes simplex virus-1 (oHSV) infection of brain tumors activates NOTCH, however the consequences of NOTCH on oHSV-induced immunotherapy is largely unknown. Here we evaluated the impact of NOTCH blockade on virus-induced immunotherapy.

Experimental design: RNA sequencing (RNA-seq), TCGA data analysis, flow cytometry, Luminex- and ELISA-based assays, brain tumor animal models, and serum analysis of patients with recurrent glioblastoma (GBM) treated with oHSV was used to evaluate the effect of NOTCH signaling on virus-induced immunotherapy.

Results: TCGA data analysis of patients with grade IV glioma and oHSV treatment of experimental brain tumors in mice showed that NOTCH signaling significantly correlated with a higher myeloid cell infiltration. Immunofluorescence staining and RNA-seq uncovered a significant induction of Jag1 (NOTCH ligand) expression in infiltrating myeloid cells upon oHSV infection. Jag1-expressing macrophages further spread NOTCH activation in the tumor microenvironment (TME). NOTCH-activated macrophages increased the secretion of CCL2, which further amplified myeloid-derived suppressor cells. CCL2 and IL10 induction was also observed in serum of patients with recurrent GBM treated with oHSV (rQnestin34.5; NCT03152318). Pharmacologic blockade of NOTCH signaling rescued the oHSV-induced immunosuppressive TME and activated a CD8-dependent antitumor memory response, resulting in a therapeutic benefit.

Conclusions: NOTCH-induced immunosuppressive myeloid cell recruitment limited antitumor immunity. Translationally, these findings support the use of NOTCH inhibition in conjunction with oHSV therapy.

©2022 American Association for Cancer Research.

Figures

Fig. 1.. Notch signaling is associated with myeloid cell infiltration in human and murine brain tumors.

A-B: Classification of GBM patients according to NOTCH score (n=416, WHO CNS grade IV 2016 classification) [1]. Affymetrix data from 416 GBM patients [1] (which included 119 pro-neural, 125 mesenchymal, and 170 classical subtypes) was interrogated to define a NOTCH score. Known IDH status of these 416 GBM patients includes IDH wild type (n=195) and IDH-mutant (n=12), with the rest unknown. Based on a median NOTCH score, the patients were then divided into a NOTCH low or high group (below or above median NOTCH score). The size and color of the dots indicate the degree of correlation. C: GSEA for KEGG_NOTCH signaling in TCGA glioma (grade IV) patients (described above). D: CIBERSORT analysis of NOTCH score-high and -low glioma patients. Immune cell populations that are significantly higher are labelled in red (NOTCH score-high group) or blue (NOTCH score-low group). E: GSEA for immune cell gene signatures in metastatic brain tumor patients (detailed in computational analysis section in methods) with indicated NOTCH score (n=16). F: The indicated tumor cells were treated with rHSVQ or were left uninfected (MOI=0.4) for 12 hours. Gene expression was measured by quantitative real time PCR (n=4). G-H: Flow-cytometry analysis of tumor-bearing hemispheres after vehicle treatment (DMSO) or (oHSV) or Gamma secretase inhibitor (GSI: RO4929097) (GSI) alone or in combination with oHSV (comb) to block NOTCH signaling. Fold change in the recruitment of the indicated cell types in GL261N4- (G, n=10–15/group) and DB7 (H, n=5/group)-bearing tumor hemispheres 2 or 7 days after treatment with rHSVQ virus, respectively. Data shown are the mean ± SD.

Fig. 2.. Feed forward NOTCH activation is initiated by tumor myeloid cells after oHSV treatment.

A: Changes in macrophage expressed NOTCH pathway genes measured by RNA-Seq of Raw264.7 macrophages co-cultured with rHSVQ infected or uninfected U87ΔEGFR cells (n=4/each). B: qRT-PCR of NOTCH ligands in F4/80+ cells isolated from GL261N4-bearing brain hemispheres 2 days after treatment with rHSVQ (2×105pfu/mouse) or PBS (n=3/each group). C: Immunofluorescence staining of Jag1 in vitro in GBM12 cells (±HSVQ, MOI = 0.3 MOI) overlaid with Raw264.7 macrophages pre-stained with CellTracker Blue CMAC Dye. rHSVQ infected cells are green, Jag1 is red, and F-actin is white. Scale bar is 20μm. D-F: D: Schematic of data shown in E-F. GBM12 cells were treated with rHSVQ or control and overlaid with macrophages. Macrophages were negatively isolated using human CD56 microbeads and overlaid on RBP-Jk NOTCH reporter transduced glioma cells (E) or NOTCH reporter transduced macrophages (F) for 9 hours with or without Jag1 neutralizing antibody. Changes in luciferase activity were measured relative to uninfected cells (n=4/each). G: Immunofluorescence staining of U87ΔEGFR tumor sections. Fluorescence microscopy revealed NICD+ cells (Red) in both oHSV infected areas (ICP4+, green) and uninfected areas. In areas evident of virus infection (ICP4+), NICD+ cells aer surrounded by infected ICP4+ cells (arrowhead). H: In uninfected areas (devoid of ICP4 (green)) staining areas of NICD positivity (red in left panel) showed Iba1 and Jag 1 positivity (red in middle and last panel, respectively) in serial sections. I: DN-MAML expression in infected tumor cells could not repress oHSV-mediated Jag1 expression (n=3–4/each). Control or DN-MAML-transfected GBM12 cells were infected with rHSVQ and overlaid with Raw264.7 macrophages for 9 hours. Gene expression was analyzed by qRT-PCR. J: Dot plot from RNA-seq showing significantly changed signaling pathways in Raw264.7 macrophages co-cultured with U87ΔEGFR cells treated with or without rHSVQ infection (n=4/each). K: qRT-PCR of Jag1 in Raw264.7 macrophages treated with the indicated drugs (Pam3CSK4: 100ng/ml, FSL-1: 100ng/ml) (n=4/each). L. qRT-PCR of Jag1 in Raw264.7 macrophages co-cultured with rHSVQ infected or uninfected GBM12 cells and with the indicated drugs (DMSO or CU CPT-22: 25μM) (n=3/each). M. Schema showing tumor cell activated NOTCH contributes to macrophage recruitment to sites of viral infection and macrophage-mediated NOTCH spreading. oHSV-infected tumor cells induce Jag1 expression on macrophages through activation of TLR2 signaling. These Jag1 expressing macrophages spread into oHSV-uninfected areas and increase NOTCH signaling in uninfected tumor cells away from the site of viral infection. The statistical significance was calculated by analysis of variance with the Tukey’s post hoc test (E, F, I, K, and L). Data shown are the mean ± SD.

Fig. 3.. Macrophages recruited to oHSV treated tumors secrete CCL2.

A-C: RNA-seq analysis of murine Raw264.7 macrophages co-cultured with U87ΔEGFR cells treated with PBS or rHSVQ in the presence or absence of GSI (DAPT, 25μM) (n=4/each). A: GSEA of RNA-Seq of macrophages shows negative enrichment of cytokine production and cytokine signaling pathways upon GSI treatment. B: Venn diagram showing overlap in number of secretome-related genes in murine macrophages altered when cultured with tumor cells with oHSV ± GSI. C: Heat map of changes in expression of cytokines in murine macrophages cultured with virus infected tumor cells, treated with GSI alone, or in combination with oHSV (comb). D: Quantification of fold changes in SPP1, CCL2, and CCL22 from a cytokine array probed with conditioned medium derived from GBM12 cells co-cultured with human PBMC-derived macrophages and treated with or without GSI. Relative expression changes of significantly altered cytokines upon GSI treatment are shown relative to oHSV group (n=3/group). E. CCL2 levels released in lysates of GL261N4-bearing mouse brain hemispheres and measured by Luminex (n=4–5/group) after treatment with vehicle (DMSO), oHSV or GSI alone, or both (comb). F: CCL2 expression in GBM patients analyzed by scRNA-Seq showing myeloid cells as the major producers of CCL2. G: Heatmap representation of four cytokines in GBM patient blood monocytes, GBM TAMs, and healthy blood monocytes. H: Changes in expression of murine CCL2 in F4/80+ macrophages isolated from GL261N4-bearing immune competent mice or CD11b+ macrophages isolated from human GBM-bearing NSG mice treated with saline or oHSV (n=3/group). I: Inhibition of macrophage recruitment by CSF-1R blocking antibody reduced CCL2 secretion in vivo. Tumor-bearing mice were treated as indicated and CCL2 secretion in tumor-bearing hemispheres was measured by ELISA (n=4). J: Changes in CCL2 produced in vitro in control or oHSV infected tumor cells cultured with Raw264.7 macrophages in the presence of CU CPT22 or MyD88i (n=3/group). K: Inverse correlation between CCL2 and TLR2 in TCGA glioma dataset. Data shown are the mean ± SD.

Fig. 4.. oHSV recruited macrophages produce CCL2 and are immunosuppressive.

A: Kaplan-Meier analysis of TCGA glioma dataset with high or low CCL2 expression. B: GSEA showing high CCL2 positively is enriched for pathways relevant to MDSCs and Tregs in the TCGA GBM dataset. C: Changes in gene expression of indicated cytokines in intra-tumoral monocytes isolated from mice bearing GL261 tumors 2 days post treatment with oHSV (orange), oHSV + GSI (purple), or control (white). D: Intra-tumoral monocytes isolated from mice bearing intracranial tumors and treated with oHSV or oHSV and GSI (comb) were implanted with DB7 tumor cells in mice brains and evaluated for tumor growth (n=14–15/group). E: Kaplan-Meier analysis of wild type or CCR2 knockout mice implanted with 005 murine tumor cells. Mice were treated as indicated (n=6–10/each group). F-G: IL-10 (F) and IFNγ (G) levels released in lysates of GL261N4-bearing mouse brain hemispheres 2 days post treatment with control, oHSV, GSI alone, or GSI and oHSV (comb) as indicated in methods. H: Pearson’s correlation between CCL2 and IFNγ expression in individual mice treated as indicated (n=4–5/group). I-J: T cell activation with intra-tumoral monocytes isolated from tumor-bearing brain hemispheres (I) or subcutaneous breast cancer (J) treated with oHSV alone or in combination with GSI (comb). Monocytes were incubated with stimulated T cells for 3 days to evaluated activation. Shown are the frequency of CD44+ cells CD4+ and CD8+ T cells, as indicated (n=4/group in I and n=3/group in J, pooled samples from 2–3 mice for each sample). K: Summary schema showing recruitment of myeloid cells in the TME post virotherapy. The recruited myeloid cells upregulate Jag1 via TLR2 activation to amplify NOTCH signaling and induce CCL2 which amplifies myeloid cell recruitment and contributes to immune suppression. Data shown are the mean ± SD.

Fig. 5.. CCL2 and IL-10 induction post oHSV virotherapy in recurrent GBM patients treated with rQnestin34.5.

A: Fold change in serum CCL2 levels of recurrent GBM patients treated with rQnestin34.5. Data shown is normalized to pre-treatment values of serum CCL2 in individual patients. B-C: CCL2 levels in individual patients pre- and post-treatment. Dotted black line indicates pre-treatment median CCL2 levels in serum of these patients. D: Correlation of Immune score and NOTCH score in TCGA GBM dataset. E: GSEA of genes negatively (r

Fig. 6.. Therapeutic advantage of combining oHSV with NOTCH blockade.

A-B: Kaplan-Meier analysis of mice implanted GL261N4 (A) or DB7 (B) tumor cells. Mice were treated with rHSVQ as indicated in the methods (n=24–25/each group in GL261N4, n=8–9/each group in DB7). C: Representative MRI of a long-term survivor (>100 days) treated with oHSV and GSI. Arrow indicates tumor-injection site. D: Kaplan-Meier analysis of re-challenged mice (n=10/each group). Naïve or re-challenged mice from Fig. 6C (combination therapy group in GL261N4 bearing mice) were implanted with GL261N4 into the contralateral hemisphere. E: Representative MRI of long-time survivor (>200 days from initial tumor-implantation) in re-challenged group from Fig. 6D. Arrow and arrowhead indicate initial and second tumor injection site, respectively. F. Hematoxylin-Eosin staining of long-time survivor from Fig. 6E. G: OVA-tetramer assay in GL261N4-OVA bearing mice treated with vehicle (DMSO), oHSV or GSI alone, or both (comb). Representative image shows OVA-tetramer+ CD8+ T cells. H-I: In mice treated with control, GSI, oHSV, or GSI and oHSV (comb) (H) shows frequency of OVA+CD8+ T cells (n=7–10/group;) and (I) shows the frequency of intra-tumoral PD1+CD8+ T cells (I) as indicated (n=5/group). J: Depletion of CD8+ T cells in mice with an anti-CD8 antibody attenuated the efficacy of the combination therapy. GL261N4-bearing mice were treated with the combination therapy in the presence of isotype control, anti-CD4, or anti-CD8 depletion antibodies (n=9/each group). K: Splenocytes from long-timer survivors (LTS, from Fig. 6D) significantly induced cytotoxicity of GL261N4 cells relative to those isolated from naïve mice (n=5/naïve, n=7/LTS). L: Gene expression of effector enzymes and CTL score in NOTCH score-high and -low GBM patients in TCGA dataset. M: Schema showing the mechanism of action behind the attenuation of anti-tumor immunity via oHSV-mediated NOTCH signaling. oHSV therapy induces Jag1 expression in TAMs. Jag1 expressing TAMs further activate NOTCH signaling in the tumor and adjacent TAM cells. These NOTCH activated TAMs then induce CCL2 secretion (in a NOTCH-dependent manner) and permit monocytic MDSC recruitment to the TME. MDSCs attenuate T cell activation and limit immunotherapeutic benefit of oHSV therapy. The statistical significance was calculated by the Student’s t-test (Fig. 6K and 6L) and adjusted p-value was calculated by Benjamini and Hochberg correction (Fig. 6L), analysis of variance with the Tukey’s post hoc test (Fig. 6H and Fig. 6I) or log-rank test (Fig. 6A, 6B, 6D and 6J). Data shown are the mean ± SD.