LysMCre/IRF2BP2flox mice that ablate IRF2BP2 in the myeloid lineage (Chen et al., 2015) were bred into the C57BL6 background, fed with regular chow, and randomly assigned to experimental groups. The animal care and use committee of the University of Ottawa approved all procedures performed with these mice.

The photothrombotic focal stroke method was performed as described by Watson et al. (1985). In brief, 2 month-old male mice were anesthetized in an induction chamber ventilated with 5% isoflurane gas, and then transferred to a stereotaxic frame and maintained at 1.5% isoflurane using a nose mask. A midline incision of the scalp was made to expose a 1 × 1 mm region of the skull. Five minutes after administration of Rose Bengal (10 mg/ml in PBS, made fresh by vortexing and filtered, 100 mg/kg, i.p.), a collimated green laser (532 nm wavelength at 20 mW, Beta Instruments) was placed 2 cm above the skull, positioned 0.7 mm anteroposterior and 2 mm mediolateral relative to the Bregma and used to illuminate the skull for 10 min, to initiate photothrombosis. Cyanoacrylate glue was used to close the wound, and mice were monitored until they regained consciousness. For IFNβ treatment, 30 min after photothrombosis, mouse recombinant IFNβ (10,000 units, Sigma) was delivered by intravenous injection via the tail vein in 100 μl saline.

Infarct volumes were determined by cresyl violet staining to reveal Nissl bodies of living neurons in 10 coronal sections (20 μm thick) sampled every 10 sections over a 2 mm distance overlapping the area of infarction, as described previously (Schock et al., 2008). Infarct volume was calculated by stacking infarct areas in serial sections using the ImageJ software (NIH). Lesion volumes were normalized to brain volumes of the corresponding sections to control for edema. Investigators were blinded to genotype.

Microglia from IRF2BP2KO and littermate control mouse pups were cultured as described (Valdearcos et al., 2014). Briefly, cortices from postnatal mice (day 1–3) were stripped of meninges, minced into small pieces, and vortexed in DMEM (Wisent) to kill neurons before being sieved through a 70 μm strainer. Cell suspensions from each pup were seeded separately onto poly-D-Lysine (Sigma) coated T-75 tissue culture flasks (Falcon) in complete DMEM medium with 10% FBS (Wisent). Culture medium was gently replaced every 2 days until the microglia were visible on the astrocyte monolayer (~7–10 days). Loosely attached microglia were then shaken off at 250 rpm for 30 min and cultured in DMEM supplemented with 10% FBS for 3 h. After microglia were attached, medium was changed to Neurobasal (GIBCO) with N1 supplement (Sigma) and cells were then ready for further treatment with LPS (100 ng/ml, Sigma) or IL4 (20 ng/ml, Sigma) for 48 h or for the phagocytosis assay (see below). The purity of these cells upon plating was in excess of 95% as confirmed by CD11b staining via FACS (data not shown).

Primary cultured microglia (8 × 104 cells/well; 24 well plate with coverslips) from IRF2BP2KO and littermate control mice were incubated with Flash-Red conjugated microspheres (Bangs Laboratories, Inc.) in the presence or absence of 100 ng/ml LPS for 24 h. Cells were washed three times with cold 1× PBS, fixed with 4% paraformaldehyde and washed several times before incubated with Iba1 antibody (#019-19741, Wako), followed by Cy2-conjugated secondary antibody (#711-225-152, Jackson ImmunoResearch). Internalization of these microspheres was visualized on a Zeiss Z1 AxioImager fluorescence microscope. The number of microspheres per cell was counted as an index of phagocytic activity. Investigators were blinded to genotype.

Isolation of microglia after stroke was done as described previously, with modifications (Ren et al., 2011). Following mechanical dissociation of the brain, a single cell suspension was obtained by chopping the tissue with tweezers and squeezing the tissue through a 20 gauge needle. The suspension with cells was centrifuged at 1500 g for 5 min and passed through a 70 um filter. The cells were then layered on top of 30% percoll (Sigma) and 70% percoll. The gradient was centrifuged at 800 g for 40 min at room temperature without brake. The cells at the 70%–30% percoll interface was collected and washed in culture media for 10 min at 1500 rpm at 4°C. The cells were re-suspended in PBS and stained with the following antibodies in the presence of CD16/32 blocking antibody (BD, 553142). Samples were read on the BD Fortessa or Moflo Astrios.

The gating for microglia was done in the following manner to separate resident and incoming cell populations. Resident microglia: CD45^intLy6c⁻780 Viability dye⁻, CD11b^lo/+; migratory cells: CD45⁺Ly6c⁺ 780 Viability dye⁻, CD11b^lo/+. The number of cells expressing MHCII and CD206 (from MHCII negative cells) were calculated using FlowJo, Treestar Inc. Antibodies for FACS analysis were as follows: MHCII-PECy7 (25-5321-82), CD11b-PerCP-cy5.5 (45-0112-82), Ly6c-APC (17-5932-82), CD45-eflour 450 (48-0451-82) all purchased from eBioscience and CD206-PE (141706) from Biolegend. Fixable Viability Dye eFluor^® 780 was also from eBioscience.

BMDM were cultured as we described (Chen et al., 2015) and treated with IFNβ (100 unit/ml) for 2 h prior to stimulation with LPS (10 ng/ml, Sigma) for 4 h. RNA was isolated for RT-qPCR analysis.

Total RNA from microglia, BMDM or brain tissue was extracted using the Qiagen RNeasy Mini Kit. Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) was conducted as described previously (Pandey et al., 2013), and the results were normalized to GAPDH or actin. Primers used for qPCR: Ccl2: (F) 5′-CACTCACCTGCTGCTACTCATTC-3′, (R) 5′-TCTTTGGGACACCTGCTG-3′. iNos: (F) 5′-AGCCCT CACCTACTTCCTG-3′, (R) 5′-TCTCTGCCTATCCGTCTC-3′, TNFα: (F) 5′-CCACCACGCTCTTCTGTCTAC-3′, (R) 5′-AGGGTCTGGGCCATAGAACT-3′. IL1β: (F) 5′-CAGGCTC CGAGATGAACAA-3′, (R) 5′-CCCAAGGCCACAGGTATTT-3′. Arg1: (F) 5′-TCACCTGAGCTTTGATGTCG-3′, (R) 5′-CTG AAAGGAGCCCTGTCTTG-3′. CD206: (F) 5′-CAAGGAAGGT TGGCATTTGT-3′, (R) 5′-CCTTTCAGTCCTTTGCAAGC-3′. YM1: (F) 5′-GGGCATACCTTTATCCTGAG, (R) 5′-CCACT GAAGTCATCCATGTC-3′. GAPDH: (F) 5′-TGTTCCTACCC CCAATGTGT-3′, (R) 5′-TGTGAGGGAGATGCTCAGTG-3′. actin: (F) 5′-CCTTCTGACCCATTCCCACC, reverse 5′-GCTTCTTTGCAGCTCCTTCG-3′.

Protein extraction and Western blot analysis were performed as described (Schock et al., 2008). A custom rabbit antibody against IRF2BP2 was described previously (Teng et al., 2011). Cryostat sections (20 μm) were subjected to cresyl violet staining and immunofluorescence. Immunofluorescence images were acquired on a Zeiss Z1 fluorescent microscope. Primary antibodies used and their dilutions are: MHCII (Biolegend #107602, anti-rat 500× dilution), CD68 (Santa Cruz, sc70761, anti-mouse, 500×), CD206 (Santa Cruz, sc34577, anti-goat, 500×), Iba1 (WAKO, #019-19741, anti-rat, 500×), NeuN (Millipore MAB377, anti-mouse, 500×), and GFAP (Santa Cruz, sc170, anti-goat, 500×). Cy2-, cy3-, cy5-conjugated secondary antibodies (Jackson Labs) were used at 1000× dilution. For immunofluorescence images, three independent fields at 20× magnification from six sections were imaged and CD206+ and CD68+ cells counted using imageJ.

Sensory-motor function was measured by removal of adhesive tape placed on the palms of mice as described online-only supplement, as described (Bouet et al., 2009). Mice in their home cage were placed in a testing room for about 30 min to allow habituation. Before testing, the mice were transferred to a new clean cage and the home cage was used as the testing cage. For testing, an adhesive tape cut into squares enough to cover the hairless part of the paw was placed and pressed gently into the left and right forelimb paw of a restrained mouse. Then mouse was placed in the testing cage and timed immediately by stopwatch with a maximum of 2 min per trial period. Data collected includes left and right contact time and left and right removal time. Mice were trained 6 days before stroke and testing was done 4 days post-stroke. Investigators were blinded to genotype.

All results are presented as means ± SEM. For between-group comparisons of fold changes, values were normalized by log transformation whereas percentages were normalized by arcsin transformation and a two-tailed Student’s t test was applied. Differences in means were considered significant at p < 0.05.

The LysMCre transgene expresses Cre recombinase under the control of the endogenous LysM promoter in macrophages (Ye et al., 2003) and microglia (Frieler et al., 2011; Hamner et al., 2015). We reported previously that LysMCre/IRF2BP2flox (IRF2BP2KO) mice ablate IRF2BP2 in their macrophages (Chen et al., 2015). To confirm that IRF2BP2 expression is also ablated in brain microglia, we cultured primary microglia isolated from IRF2BP2KO and littermate control (WT) neonatal mice. Immunoblot analysis revealed that IRF2BP2 protein expression was ablated in microglia of KO mice (Figure 1A). Trace residual IRF2BP2 protein likely reflects the few non-microglia cells that contaminate these primary cultures.

We next compared how loss of IRF2BP2 affects microglia polarization in response to LPS and IL4 stimuli. Both microglia and macrophages can become polarized to a pro-inflammatory M1 (induced by interferon gamma or LPS) or an anti-inflammatory M2 subtype (induced by IL4). M1 markers include inducible nitric oxide synthase (iNOS) and the pro-inflammatory cytokines Ccl2, Tumor Necrosis Factor-α (TNF-α) and IL1β and cell surface marker MHCII, whereas M2 markers include Arginase 1 (Arg1), CD206 (aka Mrc1, Mannose receptor, C type 1), YM1 and Fizz1 (Raes et al., 2002; Cherry et al., 2014; Okuneva et al., 2015).

In contrast to what we observed previously in BMDMs, primary microglia from KO mice did not express higher levels of the inflammatory cytokines TNFα and IL1β at baseline (NT, non-treated; Figure 1B), although this may reflect differences in culture conditions for macrophages and microglia. However, in response to LPS stimulation, a much more robust M1 activation of iNOS, Ccl2, TNFα and IL1β was detected in IRF2BP2KO microglia (Figure 1B). On the other hand, IL4-induced activation of the alternative (anti-inflammatory, M2 subtype) markers Arg1, YM1 and Fizz1 was impaired in KO microglia (Figure 1C). Interestingly, IL4 induces robust expression of CD206 in the IRF2BP2-deficient microglia (Figure 1C), unlike in IRF2BP2-deficient macrophages (Chen et al., 2015).

We next tested how mice lacking IRF2BP2 in microglia would respond to ischemic brain injury to the left sensorimotor cortex induced by photothrombosis (Figure 2). Cresyl violet staining revealed a similar-sized infarction lesion in WT and KO mice 1 day after stroke (Figures 2A,B). Immunofluorescence staining at 4 days showed a clear demarcation of NeuN+ neurons, absent within the ischemic core (IC; Figure 2C). By 4 days, lesion volumes regressed by 25% in littermate control mice, but failed to regress in KO mice (Figure 2D). Thus, IRF2BP2 ablation in microglia/macrophages appears to delay tissue repair.

Injured or dead neurons activate local microglia that secrete chemokines and cytokines to recruit more immune cells to help clear the dead cell debris for tissue repair. To compare whether a deficit in this process may contribute to delayed regression of lesion 4 days after stroke, immunostaining was used to examine different subtypes of microglia recruited to the peri-infarct area after stroke. We found 50% fewer CD68+ cells, a marker for microglia/macrophages, present in the peri-infarct area of KO mice compared to littermate control mice 1 day after photothrombosis (Figure 3A). An in vitro phagocytosis assay showed that KO microglia engulf fluorescent-labeled microparticles just as efficiently as WT microglia (Figure 3B), suggesting that the early deficit in phagocytic cell numbers rather than their ability to phagocytose may account for delayed infarct repair.

Immunostaining showed that significant numbers of CD206+ M2 microglia are recruited to the peri-infarct area 1 day after stroke. However, fewer CD206 cells were detected in KO mice at 1 and 4 days post-stroke, compared to littermate controls (Figure 3C). FACS analysis of microglia isolated 4 days after stroke confirmed fewer CD206+ cells on the ipsilateral (stroke) side in KO compared to WT mice (Figure 3D). Moreover, FACS analysis revealed a higher percentage of M1 (MHCII+) microglia on the stroke side (ipsilateral) compared to the contralateral side in WT mice. Interestingly, KO mice showed a similar high percentage of MHCII+ inflammatory microglia in both hemispheres after stroke (Figure 3D).

We next compared M1 and M2 gene expression 4 days after photothrombosis in the cortical hemisphere ipsilateral to the stroke vs. the contralateral side of the cortex by qRT-PCR. Consistent with immunostaining and FACS analysis that showed fewer anti-inflammatory M2 microglia but more inflammatory M1 microglia at the peri-infarct area of KO mice (Figure 3), lower expression of an M2 gene (Fizz1) and higher levels of M1 genes (Ccl2, TNFα, IL1β) were detected in the ipsilateral cortex of KO mice (Figure 4).

To compare the functional outcomes after focal ischemic injury of the sensory-motor cortex, mice were subjected to the adhesive removal test 1 week prior to photothrombosis and then again at 4 days after stroke to evaluate sensory-motor functional deficits. Photothrombosis in the left sensory-motor cortex impaired both sensory and motor function of the right forelimb in WT and KO mice, as indicated by the increased time required to detect (contact, i.e., sensory response, Figure 5A) and to remove the adhesive (i.e., motor function, Figure 5B), compared to the contralateral side (left forelimb). Sensory and motor functional deficits were more severe in KO mice.

Although a beneficial effect of IFNβ to limit stroke injury remains controversial (Maier et al., 2006; Inacio et al., 2015; Kuo et al., 2016), IFNβ is known to limit inflammatory cytokine expression via IRF2 (Hinkerohe et al., 2005; Buttmann et al., 2007). Since there are three corepressors of IRF2, including IRF2BP2, IRF2BP1 and IRF2BPL, it was not clear whether loss of one corepressor IRF2BP2 would affect IFNβ-dependent attenuation of inflammatory cytokine activation. Our prior microarray analysis had shown that IRF2BP2 deficient macrophages express the two other corepressors of IRF2, showing no change in IRF2BP1 but 1.6-fold elevated expression of IRF2BPL (see online supplement Table III in Chen et al., 2015) that could have compensated for loss of IRF2BP2. Thus, we first compared how loss of IRF2BP2 would affect the ability of IFNβ to suppress IL1β expression. Indeed, we found that LPS-mediated induction of IL1β was suppressed by IFNβ in WT BMDM. However, IFNβ failed to suppress the inflammatory response in IRF2BP2 KO macrophages, where an even higher activation of IL1β was observed after LPS stimulation (Figure 6A). Thus, IRF2BP2 is the major repressor of IRF2 to mediate the suppressive effect of IFNβ.

To test whether IFNβ can limit stroke injury following focal ischemia, IFNβ was injected 30 min after photothrombosis and the infarct volume was measured after 4 days. Consistent with the inflammation-damping effect of IFNβ we observed in cultured WT macrophages, infarct volume was reduced by IFNβ in WT mice. However, this protective effect of IFNβ to limit infarction injury was absent in IRF2BP2 KO mice (Figures 6B,C). Thus, IRF2BP2 in microglia/macrophages is necessary to mediate the anti-inflammatory and protective effect of IFNβ on stroke injury.