Type I interferon response in astrocytes promotes brain metastasis by enhancing monocytic myeloid cell recruitment

Abstract

Cancer metastasis to the brain is a significant clinical problem. Metastasis is the consequence of favorable interactions between invaded cancer cells and the microenvironment. Here, we demonstrate that cancer-activated astrocytes create a sustained low-level activated type I interferon (IFN) microenvironment in brain metastatic lesions. We further confirm that the IFN response in astrocytes facilitates brain metastasis. Mechanistically, IFN signaling in astrocytes activates C-C Motif Chemokine Ligand 2 (CCL2) production, which further increases the recruitment of monocytic myeloid cells. The correlation between CCL2 and monocytic myeloid cells is confirmed in clinical brain metastasis samples. Lastly, genetically or pharmacologically inhibiting C-C Motif Chemokine Receptor 2 (CCR2) reduces brain metastases. Our study clarifies a pro-metastatic effect of type I IFN in the brain even though IFN response has been considered to have anti-tumor effects. Moreover, this work expands our understandings on the interactions between cancer-activated astrocytes and immune cells in brain metastasis.

Fig. 1. IFN signaling is activated in brain metastasis.

a–c IFN signaling is activated in both human astrocytes and MDA231-BrM cells by coculture in vitro. N = 2 biologically independent experiments per condition. a Scheme of RNA sequencing experiment setup. b Ingenuity Pathway analysis (IPA) comparing BrM-induced changes in astrocytes and astrocyte-induced changes in BrM cells. c Heatmaps of genes in IFN signaling pathway from IPA in astrocytes and BrM cells. d, e IFN pathway is activated in brain metastatic lesions in vivo. IFIT1-GFP reporter structure is expressed in E0771-BrM cells. The reporter cells are injected into the experimental mice to track the activated IFN signaling in the brain metastasis microenvironment. d Macrometastatic lesions are isolated based on the luciferase signals in bioluminescent images (BLI), defined as BrM+ tissues. GFP expression in tdTomato⁺ BrM cells is detected by flow cytometry. e Representative images of intravital microscopy (IVM) observation of GFP expression in tdTomato⁺ BrM cells through an implanted cranial chamber. Vascular structure is outlined by orange dotted line. In the enlarged images, individual BrM cells are outlined by blue dotted line. Scale bar, 50 μm. Representative data shown of three biologically independent experiments.

Fig. 2. Low level of type I IFN signaling in brain metastasis microenvironment.

a–c Astrocytes increase the production of IFNβ after coculture with BrM cells in vitro. a Scheme of experiment setup. b RT-PCR results of IFN genes in human astrocytes. c RT-PCR results of Ifnb genes in mouse astrocytes. Al cultured alone, Co cocultured. Representative data of three biologically independent experiments for each BrM cell. d Increased production of IFNβ in metastasis-activated astrocytes in vivo. From the experimental mice developed with brain metastasis, metastatic lesions (BrM+) and metastasis-free tissues (BrM−) are isolated based on the luciferase signals in BLI. Astrocytes are purified and Ifnb expression is detected by RT-PCR. Data are from merged samples from n = 3 biologically independent experiments and presented as mean ± S.D. E0771 model (BrM−: n = 4; BrM+: n = 6). Yumm1.7 model (BrM−: n = 7; BrM+: n = 8). Source data are provided as a Source data file. e Astrocytes are the major brain stromal cells producing IFNβ in brain metastasis in vivo. E0771-BrM cells are injected into Ifnb1-YFP reporter mice. Three subpopulations in YFP⁺ stromal cells are further defined: astrocytes (ACSA2⁺), infiltrated immune cells (CD45^high), and brain resident microglia (CD45^low). Representative tSNE includes tdTomato⁺ BrM cells (in red color) and 3 subpopulations in YFP⁺ stromal cells (in 3 shades of green) from a single BrM+ lesion. The percentage of each subpopulation in total YFP⁺ stromal cells is quantified in the bar graph. Data are merged 10 samples from 3 biologically independent experiments and presented as mean ± S.D. P values are from unpaired two-tailed t test. Source data are provided as a Source data file. f, g Low level of type I IFN activation in brain metastatic lesions in vivo. IFIT1-GFP reporter structure is expressed in E0771-BrM cells. The reporter cells treated with various concentrations of IFNβ in vitro and GFP expression is quantified (Representative data of 3 biologically independent experiments). From the brain metastatic lesions established by the reporter cells, the GFP expression in tdTomato⁺ BrM cells in vivo is quantified and compared to the IFNβ-treated cells. f Three weeks after injecting the reporter BrM cells, macrometastatic lesions are isolated. The percentage of GFP⁺ cells and mean fluorescent intensity (MFI) are quantified in tdTomato⁺ BrM cells. Data are merged 14 samples from 3 biologically independent experiments and presented as mean ± SD. g One week and 2 weeks after injection, the reporter BrM cells are observed by IVM through an implanted cranial chamber. The GFP signal is quantified and normalized by tdTomato⁺ signal from the same BrM cell. One-week data are merged 59 cells from 4 experimental mice. Two-week data are merged 80 cells from 3 experimental mice. Data are presented as mean ± S.D. P values are from unpaired two-tailed t test. Source data are provided as a Source data file.

Fig. 3. Type I IFN activation in astrocytes promotes brain metastasis.

a Increased IFN response genes in astrocytes cocultured with BrM cells in vitro. Human astrocytes are cocultured with breast cancer MDA231-BrM cells. Mouse astrocytes are cocultured with breast cancer E0771-BrM and melanoma Yumm1.7-BrM. The expression levels of IFN response genes are measured by RT-PCR. Al cultured alone, Co cocultured. Representative data of 3 biologically independent experiments for each BrM cell. b–e Type I IFN signaling in astrocytes. b Human astrocytes are treated with conditioned media (CM) and type I IFN response genes are detected by RT-PCR. CM from astrocyte-BrM coculture (Co) or cultured alone (AL) are pretreated with either neutralizing antibody against IFNAR1 or the matched IgG control antibody (Ctrl). Representative data of 3 biologically independent experiments. c Astrocyte specific IFNAR1 knockout mice. d RT-PCR results of the type I IFN response gene, Ifit1, in the astrocytes cocultured with BrM cells in vitro. Primary cultured astrocytes are isolated from transgenic mice. Cre−, Gfap-Cre^−/−; Ifnar1^f/f mice; Cre+, Gfap-Cre^+/−; Ifnar1^f/f mice. Representative data of 3 biologically independent experiments for each BrM cell. e RT-PCR results of type I IFN response genes, Ifit1 and Isg15, in the astrocytes from brain metastatic lesions in vivo. From the experimental mice developed with brain metastasis, the metastatic lesions (BrM+) and matched metastasis-free tissues (BrM−) are isolated based on the luciferase signals in BLI. Data are merged samples from 2 biologically independent experiments. E0771 model (4 pairs for both Cre+ and Cre−). Yumm1.7 model (Cre−: 3 pairs; Cre+: 4 pairs). Source data are provided as a Source data file. f Type I IFN signaling in astrocytes facilitates brain metastasis. Breast cancer E0771-BrM, A7C11-BrM and melanoma Yumm1.7-BrM are injected into Cre- or Cre+ mice. Quantification of brain lesions by BLI. Data are merged from 3 biologically independent experiments and presented as mean ± S.E.M. P values are from unpaired two-tailed t test. Female mice: E0771 model (Cre−: n = 13; Cre+: n = 16); A7C11 model (Cre−: n = 8; Cre+: n = 11). Male mice: E0771 model (Cre−: n = 9; Cre+: n = 16); Yumm1.7 model (Cre−: n = 11; Cre+: n = 4). Source data are provided as a Source data file.

Fig. 4. Type I IFN activation in astrocytes facilitate monocytic myeloid cell infiltration.

a Immune cells in brain metastatic lesions in vivo. Immune cells, stained with CD45, are analyzed in brain tissues from the naive and experimental mice developed with brain metastasis. Metastatic lesions (BrM+) and metastasis-free tissues (BrM-) are isolated based on the luciferase signals in BLI. Percentages of CD45^low resident microglia and CD45^high infiltrated immune cells are quantified using flow cytometry. Data are merged samples from n = 3 biologically independent experiments presented as mean ± S.D (2 naive brain, 7 female BrM−, 24 female E0771 BrM+, 7 male BrM−, 11 male E0771 BrM+ and 6 male Yumm1.7 BrM+ samples). P values are from unpaired two-tailed t test. Source data are provided as a Source data file. b, c CD45^high infiltrated immune cells in brain metastatic lesions are further analyzed. b E0771-BrM cells are injected into experimental mice to establish brain or lung metastasis. Representative tSNE files of myeloid and lymphoid subpopulations of one brain metastasis lesion, merged lung metastatic samples from one experimental mouse and the blood sample from one brain metastasis-bearing mouse. M-MDSC, monocytic myeloid-derived suppressor cells; PMN-MDSC, polymorphonuclear MDSC. Bar graphs indicate the proportions of indicated immune subpopulations in CD45^high cells (total is 100%). c Dot plots show the representative flow profiles of MDSCs. Bar graphs show the percentages of myeloid subpopulations out of CD45^high cells. Data are merged samples from 4 biologically independent experiments (17 BrM, 19 LM and 11 blood samples). Data are presented as mean ± S.D. P values are from unpaired two-tailed t test. Source data are provided as a Source data file. d Type I IFN signaling in astrocytes facilitates monocytic myeloid cell infiltration. Breast cancer E0771-BrM and melanoma Yumm1.7-BrM cells are injected into transgenic mice with astrocyte specific IFNAR1 knock out. Cre−, Gfap-Cre^−/−; Ifnar1^f/f mice; Cre+, Gfap-Cre^+/−; Ifnar1^f/f mice. Dot plots show the representative flow profiles of MDSCs. Bar graphs show the percentages of myeloid subpopulations out of CD45^high cells. Data are merged samples from n = 3 biologically independent experiments and presented as mean ± S.D. P values are from unpaired two-tailed t test. E0771 model (Cre−: n = 17, Cre+: n = 15). Yumm1.7 model (Cre−: n = 8 -, Cre+: n = 7). Source data are provided as a Source data file.

Fig. 5. Type I IFN response activates CCL2 expression in astrocytes to facilitate myeloid cell infiltration and brain metastasis.

a CCL2 expression is activated in astrocytes cocultured with BrM cells in vitro. Mouse astrocytes are cocultured with breast cancer E0771-BrM, A7C11-BrM and melanoma Yumm1.7-BrM. The expressions of CCL2 are measured by RT-PCR. Al, cultured alone astrocytes; Co, cocultured astrocytes. Representative data of 3 biologically independent experiments for each BrM cell. b IFNβ activates CCL2 expression in astrocytes in vitro. Mouse astrocytes cells are treated with various concentrations of IFNβ and the expressions of CCL2 are quantified by RT-PCR. n = 2 biologically independent experiments. c, d Type I IFN signaling activate CCL2 expression in astrocytes. c RT-PCR results of CCL2 expression in the astrocytes coculture with BrM cells in vitro. Primary cultured astrocytes are isolated from transgenic mice: Cre−, Gfap-Cre^−/−; Ifnfar1^f/f mice; Cre+, Gfap-Cre^+/–; Ifnar1^f/f mice. Representative data of 3 biologically independent experiments for each BrM cell. d RT-PCR results of CCL2 expression in the astrocytes from brain metastatic lesions in vivo. From the experimental mice developed brain metastasis, metastatic lesions (BrM+) and metastasis-free tissues (BrM−) are isolated based on the luciferase signals in BLI. Data are merged samples from 2 biologically independent experiments. E0771 model (4 pairs for both Cre+ and Cre−). Yumm1.7 model (Cre−: 3 pairs, Cre+: 4 pairs). Source data are provided as a Source data file. e CCL2 dependent migration of monocytic myeloid cells in vitro. In the top of migration chambers, M-MDSC and PMN-MDSC sorted from spleens of brain metastasis-bearing mice are loaded. Conditioned media (CM) from astrocyte-BrM coculture (Co) or cultured alone (AL) are pretreated with either neutralizing antibody against CCL2 (anti-CCL2) or the matched IgG control antibody (Ctrl) and loaded in the bottom of migration chambers. Data are the normalized migration of M-MDSC and PMN-MDSC. M-MDSC monocytic myeloid-derived suppressor cells, PMN-MDSC polymorphonuclear MDSC. Data are merged samples from 3 biologically independent experiments and presented as mean ± S.D. P values are from unpaired two-tailed t test. E0771 model (n = 3). Yumm1.7 model (n = 4). f, h Monocytic myeloid cell infiltration facilitates brain metastasis. f BrM cells are intracardially injected into CCR2 knockout (KO) and control wildtype (WT) mice. Quantification of brain lesions by BLI. BLI images show one representative sample in each group. Data are merged samples from n = 3 biologically independent experiments and presented as mean ± S.E.M. Female mice: E0771 model (WT and CCR2KO: n = 17). Male mice: E0771 model (WT and CCR2KO: n = 10). Yumm1.7 model (WT: n = 8, CCR2KO: n = 10). g BrM cells are intravenously injected into CCR2 KO and control WT female mice. Quantification of lung lesions by BLI. BLI images show one representative sample in each group. Data are merged samples from n = 3 biologically independent experiments and presented as mean ± S.E.M. Female mice (WT: n = 12; CCR2KO: n = 13). Male mice (WT: n = 10; CCR2KO: n = 9). h Percentages of myeloid subpopulations out of CD45^high cells in brain metastatic lesions. Data are merged samples from n = 3 biologically independent experiments and presented as mean ± S.D. E0771 model (WT and CCR2KO: n = 8). Yumm1.7 model (WT: n = 5, CCR2KO: n = 2). P values are from unpaired two-tailed t test. Source data are provided as a Source data file.

Fig. 6. CCR2 is a therapeutic target to treat brain metastasis.

a Bioinformatic analyses of immune cells in clinical samples. We include breast cancer PMC6449168 and GSE125989 (paired primary and brain metastatic datasets), GSE14020 (unpaired brain and lung metastatic samples), and melanoma melanoma EGAD00001005046 (paired and unpaired brain metastatic and extracranial tumors) in the analyses. Immune scores of lymphoid and myeloid immune subpopulations analyzed by CIBERSORT. P values are the results from unpaired two-tailed t test. PMC6449168: n = 22. GSE125989: n = 16. GSE14020: n = 19 Brain, n = 18 Lung. MD Melanoma: n = 88 Brain, n = 49 Extracranial. Source data are provided as a Source data file. b Single-cell RNAseq analysis of human clinical samples. After unsupervised clustering, we identified myeloid and lymphoid clusters from GSE174401 (melanoma BrM and extracranial), GSE176078 (primary breast cancer), and GSE186344 (Multiple BrM types) using gene expression signatures. Myeloid cells (red clusters), T/NK cells (blue clusters), B/Plasma cells (green clusters), other cancer/stromal cells (gray clusters). The proportions of total immune cells, myeloid and lymphoid cells were calculated out of total cells. P values are from unpaired two-tailed t test. Data points are individual samples from 3 datasets (28 brain, 34 extracranial). c Correlations of CCL2 or CXCL12 expression and immune scores of monocytic lineage (Mono), neutrophil (Neutro), and T lymphoid (T) subpopulations in brain metastatic and extracranial tumors analyzed by MCP-counter. Breast cancer PMC6449168 and GSE125989 (paired primary and brain metastatic datasets), GSE14020 (unpaired brain and lung metastatic samples), and melanoma EGAD00001005046 (paired and unpaired brain metastatic and extracranial tumors) are analyzed. P values are from unpaired two-tailed t test. d–f CCR2/CCR5 antagonist treatment decreases brain metastasis. Cenicriviroc (CVC) or 15a are systemically applied in the female experimental mice injected with breast cancer E0771-BrM cells and male mice injected with melanoma Yumm1.7-BrM cells. DMSO, vehicle control. d CVC or 15a are applied from day 1 after cancer cell inoculation. Brain metastases are quantified by bioluminescent imaging. BLI images show one representative sample in each group. Data are merged samples from 2 biologically independent experiments and presented as mean ± S.E.M (E0771 model: 10 DMSO, 9 CEN, 9 15a samples; Yumm1.7 model: 9 DMSO, 8 CEN, 9 15a samples). P values are from unpaired two-tailed t test. e CVC is applied from day 7 after cancer cell inoculation. Brain metastases are quantified by bioluminescent imaging. BLI images show one representative sample in each group. Data are merged samples from 2 biologically independent experiments and presented as mean ± S.E.M. P values are from unpaired two-tailed t test. f CVC is applied from day 7 after cancer cell inoculation. Probabilities of survival are tracked. Data are from merged samples of 2 biologically independent experiments. P values are from log-rank test. Source data are provided as a Source data file. g Schematic summary of the effect of type I IFN activation in astrocytes on brain metastasis. Cancer activated astrocytes, together with immune cells, create a sustained low level type I IFN activated microenvironment in the brain. IFN response in astrocytes promotes brain metastasis by secreting CCL2 to recruit monocytic myeloid cells. CCR2 is a therapeutic target to decrease brain metastasis by blocking myeloid cell infiltration.