Microglia in the peri-infarcted cortex upregulated STING after ischemic stroke

We first examined the expression pattern of STING at different time points after photothrombotic stroke. At 3, 7, and 14 days post injury (dpi), the protein expression levels of STING were significantly upregulated relative to sham-treated group, peaking at 7 dpi (F(4, 15) = 19.58, P < 0.0001, Fig. 1A, B). We also detected an increment in cGAS expression levels at 3 and 7 days after stroke (F(4, 15) = 7.626, P = 0.0015, Additional file 1: Fig. S1A, B). IFNβ is the primary downstream effector of STING signaling [35], therefore, we detected the expression levels of IFNβ by ELISA around the infarcted cortex and found that IFNβ was also increased at 7 and 14 dpi (F(4, 24) = 6.729, P = 0.4813, Fig. 1C). Next, we investigated the cellular distribution of STING by immunofluorescence staining after stroke. We found obvious colocalization of STING and microglia marker IBA1 both in sham-treated group and stroke-injured group (Fig. 1D), while little or no colocalization was detected between STING and NeuN (Fig. 1H) or GFAP (Fig. 1I), indicating that STING was primarily expressed by microglia under both physiological and stroke-induced pathological conditions. Moreover, not all microglia expressed STING. From 3 days after stroke, the percentage of STING-positive microglia was dramatically increased (F(4, 38) = 25.42, P < 0.0001, Fig. 1E). The intensity of STING and the percentage of STING-positive area were also upregulated at 7 dpi (F(4, 38) = 17.38, P = 0.0076, Fig. 1F; F(4, 38) = 11.91, P < 0.0001, Fig. 1G), which was compatible with Western Blotting experiments (Fig. 1A, B). Together, these results showed that microglia was the primary cell type that expressed STING, and photothrombotic stroke could result in elevated expression levels of STING in microglia during the subacute phase of stroke.

Fig. 1

After stroke, STING was primarily expressed and upregulated in microglia. A Representative bands of STING and β-ACTIN using Western Blotting experiments. B The quantification of STING expression levels with β-ACTIN as the internal reference. The expression levels were relative to sham-treated group. n = 4 mice per condition. C The protein levels of IFNβ detected by ELISA. n = 5–6 mice. Each dot represented an individual mouse. D Representative immunofluorescent micrographs of STING and IBA1 at different time points after photothrombotic stroke. E The quantification of STING-positive microglia. Microglia were labeled by IBA1. F The mean fluorescence intensity (MFI) of STING relative to sham-treated group. G The quantification of STING-positive area. n = 7–9 field of view (FOV) from 3 mice per condition. Each dot represented a FOV. H and I Representative immunofluorescent micrographs of STING and neuronal marker NeuN (H) and astrocytic marker GFAP (I). Boxed regions were enlarged for analyzing colocalization. Scale bar = 20 μm (D and I). Scale bar = 10 μm (H). Data were presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001

STING inhibitor H151 effectively suppressed STING activation and promoted motor function recovery after stroke

H151 is a highly potent and selective small-molecule covalent antagonist of STING [36]. H151 was intraperitoneally administrated for 8 consecutive days, commencing 1 h after photothrombotic stroke. 7 days after injury, increased levels of phosphorylated TBK1 was observed, which indicated STING signaling activation [35]. H151 treatment could reduce TBK1 phosphorylation, while the total levels of TBK1 remained unchanged (F(3, 20) = 7.538, P = 0.0015, Fig. 2A and B). Similarly, IFNβ expression level was also suppressed by H151 (F(3, 19) = 16.10, P < 0.0001, Fig. 2C), accompanied by a reduction in the mRNA levels of interferon-stimulated genes (Oasl2, Isg15, Ifit3) after stroke (Oasl2: F(3, 18) = 10.13, P = 0.0004; Isg15: F(3, 17) = 7.613, P = 0.0019; Ifit3: F(3, 17) = 6.534, P = 0.0039, Fig. 2D). The STING protein levels were also significantly reduced by H151 treatment (F(3, 16) = 81.32, P < 0.0001, Additional file 3: Fig. S3A, B). These results showed that STING signaling activation could be potently inhibited by H151 administration. We then investigated the effects of H151 on stroke-induced motor function deficits. Grid walking test (Fig. 2E) and cylinder test (Fig. 2F) were used to evaluate the forelimb motor impairments and forelimb-use asymmetries, respectively [37, 38]. After PT injury, mice showed obvious deficits in these behavioral tests, while H151 treatment could improve their performance in grid walking test at 3 dpi (group main effect, F(3, 135) = 348.9, P < 0.0001, Fig. 2E) and in cylinder test (group main effect, F(3, 128) = 159.1, P < 0.0001, Fig. 2F) at 1 and 7 dpi. We further conducted adhesive removal test for assessing sensory-motor impairments (Fig. 2G, H). Injured forelimb tended to spend longer time to sense and remove the adhesive tape after stroke, while H151 treatment could significantly shorten the removal time (group main effect, F(3, 135) = 32.93, P < 0.0001, Fig. 2H). Altogether, these results showed that the inhibition of STING signaling activation by H151 could significantly enhance the post-stroke recovery of mice.

Fig. 2

H151 inhibited STING activation and promoted post-stroke recovery of motor function. A Representative bands of phospho- and total-TBK1, as well as β-ACTIN using Western Blotting experiments. B The relative expression levels of phospho- and total-TBK1. β-ACTIN was used as the internal reference. The results were relative to sham + vehicle group. C The protein levels of IFNβ detected by ELISA. D The relative mRNA levels of interferon-stimulated genes (Oasl2, Isg15, and Ifit3). The results were relative to sham + vehicle group. n = 4–7 mice per condition. Each dot represented an individual mouse. E The quantification of foot fault rate in grid walking test at different time points after photothrombotic stroke. F The quantification of forelimb asymmetry in cylinder test. G The time spent by the injured forelimb to sense the adhesive tape in adhesive removal test. H The time to remove tape in adhesive removal test. n = 12–13 mice for behaviour assessment. Data were presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ns: not significant

STING inhibition by H151 could suppress the overactivation of microglia after stroke

We next investigated whether H151 affected microglia activation by immunostaining IBA1 and CD68 (Fig. 3A), which is considered as a marker for microglia pro-inflammatory activation [39,40,41,42]. From 3 dpi after PT modeling, the number of IBA1-positive microglia significantly increased (F(4, 53) = 37.42, P < 0.0001, Fig. 3B), accompanied by an increased percentage of CD68-positive area (F(4, 53) = 14.59, P < 0.0001, Fig. 3C). The average fluorescence intensity of CD68 began to up-regulate from day 1 after PT modeling (F(4, 53) = 80.61, P < 0.0001, Fig. 3D). The number of CD68-positive microglia also upregulated from 3 dpi (F(4, 53) = 43.71, P < 0.0001, Fig. 3E), indicating that the microglia around the infarcted cortex was activated after stroke from 3 dpi and remained activated during the subacute phase of stroke. However, after consecutive H151 administration, microglia number was significantly decreased (F(2, 23) = 116.9, P < 0.0001, Fig. 3F, G), and CD68 expression level was also reduced in microglia (F(2, 23) = 36.30, P < 0.0001, Fig. 3H; F(2, 23) = 47.48, P < 0.0001, Fig. 3I; F(2, 23) = 122.4, P < 0.0001, Fig. 3J). This indicated that H151 administration could inhibit microglia activation.

Fig. 3

H151 administration reduced CD68 expression in microglia after stroke. A Representative immunofluorescent micrographs of IBA1 and CD68 after sham treatment or at different time points after stroke injury. B The number of IBA1-positive cells in each FOV in sham or PT group. C The percentage of CD68 positive area in each FOV. The results were relative to sham group. D The MFI of CD68 after sham or injury. E The number of CD68- and IBA1-double positive cells after sham or injury. n = 10–12 FOV from 3 mice. F Representative images of IBA1 and CD68 after administration of H151 or vehicle in sham-treated or PT-injured mice. Boxed regions were enlarged for analyzing colocalization. G–J The number of IBA1-positive cells (G), the percentage of CD68 positive area (H), CD68 MFI (I), and CD68/IBA1-double positive cells were quantified after H151 or vehicle administration at 7 days after sham or stroke. n = 8–9 FOV from 3 mice. Scale bar = 20 μm. Data were presented as mean ± SEM. **P < 0.01, ***P < 0.001

We further analyzed the morphological changes of IBA1-positive microglia after H151 treatment (Fig. 4A). 7 days after stroke, microglia showed typical activated morphology with a reduced number of cellular processes (F(2, 2439) = 324.5, P < 0.0001, Fig. 4B; F(2, 48) = 18.30, P < 0.0001, Fig. 4C), enlarged cell body (Fig. 4D), and decreased process length (F(2, 48) = 38.28, P < 0.0001, Fig. 4E), while after H151 treatment microglia exhibited a more ramified morphology (Fig. 4A–E). We also evaluated the effects of H151 on inflammatory molecules after stroke and found that STING inhibition could significantly reduce the mRNA levels of Nlrp3 (F(3, 18) = 21.20, P < 0.0001, Fig. 4F), Caspase 1 (F(3, 17) = 7.969, P = 0.0016, Fig. 4G), Il1β (F(3, 18) = 14.32, P < 0.0001, Fig. 4H), and Tnfα (F(3, 17) = 21.93, P < 0.0001, Fig. 4I). Altogether, these results showed that STING inhibition by H151 could reduce the number of CD68-positive microglia, restore microglia morphology to a more ramified state, and suppress neuroinflammation after stroke.

Fig. 4

STING inhibition affected the morphological features of microglia and reduced the levels of inflammatory cytokines. A Representative skeletonization of IBA1-labelled microglia. Scale bar = 5 μm. B–E Sholl analysis at 7 days after surgery for intersection numbers (B), endpoints (C), IBA1-positive area (D), and mean process length (E) per cell after consecutive administration of vehicle or H151. n = 17 cells from 3 mice per condition. Each dot represented an analyzed cell. F–I mRNA levels of Nlrp3 (F), Caspase1 (G), Il1β (H), and Tnfα (I) were detected by qRT-PCR at 7 days after surgery. n = 4–6 mice per condition. Each dot represented an individual mouse. Data were presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001

Microglia-mediated synapse phagocytosis was inhibited after STING inhibition

Previous studies have shown that microglia are involved in synapse elimination during both the acute and subacute phases after stroke [5, 6]. Similarly, we also found obvious colocalization of IBA1 and post-synaptic marker PSD95 (Fig. 5A) or pre-synaptic marker SYP (Fig. 5C) at 7 days after PT injury. However, after consecutive H151 treatment, there was a significant reduction in synapse engulfment by microglia (F(2, 89) = 62.62, P < 0.0001, Fig. 5B; F(2, 81) = 28.49, P < 0.0001, Fig. 5D). We further conducted immunofluorescence triple staining for IBA1, PSD95, and CD68 to show that the synaptic materials might be digested by lysosomes of microglia (Additional file 2: Fig. S2A). Similarly, increased colocalization of IBA1, PSD95, and CD68 was observed after stroke injury, which could be significantly decreased after STING inhibition (F(2, 160) = 75.45, P < 0.0001, Additional file 2: Fig. S2B). Consist with previous studies [6], we also found obvious synaptic loss around the infarcted cortex (Fig. 5E). However, there was more puncta numbers of SYP (F(2, 21) = 6.589, P = 0.0060, Fig. 5F), PSD95 (F(2, 20) = 15.51, P < 0.0001, Fig. 5G), and their colocalization (F(2, 20) = 39.72, P < 0.0001, Fig. 5H) in PT + H151 group, indicating much synapses were preserved after STING inhibition. We further verified these using STING knockout mice (STING-KO mice, Fig. 6A). Similar to H151 treatment, the microglia in STING-KO mice engulfed less presynaptic and post-synaptic elements (SYP: F(2, 141) = 48.38, P < 0.0001, Fig. 6C; PSD95: F(2, 77) = 27.44, P < 0.0001, Fig. 6D) after stroke, accompanied by increased synaptic density around the infarcted region (F(2, 25) = 16.31, P < 0.0001, Fig. 6B, E). Together, these experiments showed that STING inhibition could suppress the phagocytotic ability of microglia against stroke-affected synapses and promote the recovery of synaptic density.

Fig. 5

STING inhibition by H151 suppressed microglial phagocytosis of synapses after stroke. A Representative micrographs and 3D reconstructions of IBA1 and PSD95 at 7 days after surgery. B Quantitative data of PSD95 puncta number in IBA1-positive microglia. C Representative micrographs and 3D reconstructions of IBA1 and SYP after vehicle or H151 treatment. D Quantification of SYP puncta number in IBA1-positive microglia. E Representative images of SYP and PSD95. Boxed regions were enlarged for analyzing colocalization. F Relative SYP puncta number. G Relative puncta number of PSD95. H The quantification of SYP/PSD95 colocalized puncta number. The results were relative to sham + vehicle group. In (A–D), n = 23–40 cells from 3 mice per condition. Each dot represented an analyzed cell. In (E–H), n = 6–10 FOV from 3 mice per condition. Each dot represented a FOV. Scale bar = 5 μm. Data were presented as mean ± SEM. **P < 0.01, ***P < 0.001

Fig. 6

Genetically deleting STING inhibited microglial phagocytosis of synaptic elements after stroke. A The protein expression levels of STING in wildtype mice and STING-knockout mice. B Representative 3D reconstructions of IBA1 and synaptic elements, along with representative images of SYP and PSD95 at 7 days after surgery. C and D The quantification of SYP puncta number (C) and PSD95 puncta number (D) in IBA1-positive microglia. E The quantification of SYP/PSD95 colocalized puncta number. The results were relative to sham-treated wildtype mice. In (C, D), n = 11–74 cells from 3 mice per condition. Each dot represented an analyzed cell. In (E), n = 9–10 FOV from 3 mice per condition. Each dot represented a FOV. Scale bar = 5 μm. Data were presented as mean ± SEM. *P < 0.05, ***P < 0.001

STING inhibition down-regulated several phagocytosis-related molecules

We moved on to explore how STING was involved in microglial phagocytosis. Complement family had been proven to mediate the phagocytic recognition of targeted synapses [43], we therefore analyzed the mRNA levels of complement components. 7 days after PT injury, C1qa, C1qb, C3, C3ar1, C5ar1, and Itgb2 was upregulated relative to sham-treated group (Fig. 7A). Among these targets, C1qa, C1qb, C5ar1, and Itgb2 could be significantly decreased after STING inhibition. H151 administration did not affect the levels of these molecules in sham-treated mice (C1qa: F(3, 18) = 26.32, P < 0.0001; C1qb: F(3, 17) = 27.71, P < 0.0001; C3: F(3, 18) = 9.906, P = 0.0004; C3ar1: F(3, 19) = 7.103, P = 0.0022; C5ar1: F(3, 17) = 11.06, P = 0.0003; Itgb2: F(3, 17) = 11.60, P = 0.0002, Fig. 7A). In addition to complement components, we also measured the levels of the known phagocytic receptors expressed in microglia [44, 45]. Of the detected receptors, Cd36 and Fc family receptors (Fcgr1, Fcer1g, Fcgr2b, Fcgr3, and Fcgr4) were sensitive to STING inhibition (Cd36: F(3, 18) = 4.878, P = 0.0118; Dap12: F(3, 18) = 13.31, P < 0.0001; Trem2: F(3, 18) = 13.15, P < 0.0001; Fcgr1: F(3, 17) = 14.37, P < 0.0001; Fcer1g: F(3, 17) = 15.53, P < 0.0001; Fcgr2b: F(3, 17) = 20.69, P < 0.0001; Fcgr3: F(3, 17) = 28.46, P < 0.0001; Fcgr4: F(3, 17) = 29.87, P < 0.0001, Fig. 7B). The process of phagocytosis also requires the conduction of intracellular signaling molecules [46,47,48]. We therefore detected the intracellular phagocytosis-related signaling molecules, and found that the transcriptional levels of Rac2, Pld4, and Hmox1 could be significantly down-regulated after STING inhibition (Rac2: F(3, 17) = 17.30, P < 0.0001; Hmox1: F(3, 17) = 14.45, P < 0.0001; Pld4: F(3, 17) = 41.74, P < 0.0001, Fig. 7C). These results showed that many phagocytosis-related molecules were under the regulation of STING signaling.

Fig. 7

Antagonizing STING activation decreased the levels of phagocytosis-related molecules. A The relative mRNA levels of complete components after vehicle or H151 treatment at 7 days after surgery. B Heat map of the relative mRNA levels of phagocytic receptors. Darker color represented higher fold change. #P < 0.05, ##P < 0.01, ###P < 0.001, compared with sham + vehicle group; *P < 0.05, ***P < 0.001, compared with PT + vehicle group. C The fold change of mRNA levels of Rac2, Hmox1, and Pld4 after vehicle or H151 treatment. The results were relative to sham + vehicle group. n = 4–6 mice per condition. Data were presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001

We next asked how STING achieved this regulation. We hypothesized that these phagocytosis-related molecules might be regulated by one or more shared transcriptional factors, and some of the transcriptional factors were the downstream effectors of STING signaling. Using the database of Cistrome Data Browser, several candidate transcriptional factors were predicted to regulate the majority of the above-mentioned molecules (Fig. 8A). After further verification by qRT-PCR and Western Blotting, STAT1 was chosen for further investigation because it was the only transcriptional factor that could be down-regulated in both mRNA (Spi1: F(3, 17) = 26.25, P < 0.0001; Stat1: F(3, 17) = 8.481, P = 0.0012, Fig. 8B) and protein levels (STAT1: F(3, 20) = 33.87, P < 0.0001, Fig. 8C, D; SPI1: F(3, 20) = 8.047, P = 0.0010, Fig. 8E) after STING inhibition. It is known that the nucleus translocation of phosphorylated STAT1 is required to evoke STAT1-dependent gene expression [49]. For assessing the nuclear distribution of phosphorylated STAT1, we first performed immunostaining against IBA1, phospho-STAT1, and Hoechst. We found that stroke induced obvious nuclear translocation of phospho-STAT1 in microglia (Fig. 8F), which could be significantly inhibited by H151 treatment (F(2, 150) = 50.17, P < 0.0001, Fig. 8G). For more accurate quantification, we extracted the nuclear fraction and used for Western Blotting. Low levels of GAPDH were detected in the nuclear extracts, indicating little contamination from cytoplasmic fractions under our experimental conditions (Additional file 4: Fig. S4A). We found increased protein levels of phospho-STAT1 in the nuclear fraction at 7 days after stroke, and H151 administration reduced the nuclear distribution of phospho-STAT1 (F(3, 16) = 6.907, P = 0.0034, Fig. 8H, I). Altogether, these results indicated that STING intimately regulated various phagocytosis-related molecules and their transcriptional factor STAT1.

Fig. 8

H151 inhibited STAT1 expression and nucleus translocation. A Chordal graph showing the regulatory relationship between transcriptional factors and phagocytosis-related molecules based on the database of Cistrome Data Browser. B Heat map of the mRNA levels of the transcriptional factors that potentially regulated phagocytosis. Darker color represented higher fold change. ##P < 0.01, ###P < 0.001, compared with sham + vehicle group; *P < 0.05, compared with PT + vehicle group. n = 4–6 mice per condition. C Representative bands of STAT1 and SPI1 in each experimental condition. D, E Quantitive analysis of protein expression levels STAT1 (D) and SPI1 (E). n = 6 mice per condition. F Representative immunofluorescent micrographs and 3D reconstructions of phosphorylated STAT1, IBA1, and Hoechst. Scale bar = 5 μm. G The relative p-STAT1 puncta number in the nucleus of IBA1-positive cell. n = 19–74 cells from 3 mice. Each dot represented an analyzed cell. H Representative bands of phospho-STAT1 and Lamin B1. The nuclear fractions were used for Western Blotting. I Quantitive analysis of protein expression levels of phospho-STAT1. Lamin B1 was used as internal control for nuclear protein. Data were presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ns: not significant