Reduced microglia activation following metformin administration or microglia ablation is sufficient to prevent functional deficits in a mouse model of neonatal stroke

Neonatal stroke is a devastating insult that can lead to life-long impairments. In response to hypoxic–ischaemic injury, there is loss of neurons and glia as well as a neuroinflammatory response mediated by resident immune cells, including microglia and astrocytes, which can exacerbate damage. Administration of the antidiabetic drug metformin has been shown to improve functional outcomes in preclinical models of brain injury and the cellular basis for metformin-mediated recovery is unknown. Given metformin’s demonstrated anti-inflammatory properties, we investigated its role in regulating the microglia activation and used a microglia ablation strategy to investigate the microglia-mediated outcomes in a mouse model of neonatal stroke. Hypoxia-ischaemia (H-I) was performed on post-natal day 8. Metformin was administered for one week, starting one day after injury. Immunohistochemistry was used to examine the spatiotemporal response of microglia and astrocytes after hypoxia-ischaemia, with or without metformin treatment. To evaluate the effects of microglia depletion after hypoxia-ischaemia, we delivered Plexxikon 5622 for 1 or 2 weeks post-injury. The regional pattern of microglia and astrocyte depletion was assessed through immunohistochemistry. Motor behaviour was assessed with the righting reflex, hindlimb suspension, grip strength and cylinder tests. Herein, we revealed a spatiotemporally regulated response of microglia and astrocytes after hypoxia-ischaemia. Metformin treatment after hypoxia-ischaemia had no effect on microglia number and proliferation, but significantly reduced microglia activation in all regions examined, concomitant with improved behavioural outcomes in injured mice. Plexxikon 5622 treatment successfully ablated microglia, resulting in a > 90% depletion in microglia in the neonatal brain. Microglia rapidly repopulated upon treatment cessation of Plexxikon. Most interesting, microglia ablation was sufficient to reduce functional deficits after hypoxia-ischaemia, mimicking the effects of 1 week of metformin treatment post-injury. These results highlight the importance of regulating the neuroinflammatory response after neonatal stroke to promote recovery.


Introduction
Neonatal stroke is an important cause of brain injury that occurs in ~ 1 in 2300 infant births [12,17,45]. Children who suffer from this form of hypoxic-ischaemic (H-I) brain injury exhibit long-term neurologic complications ranging from mild behavioural deficits to severe seizures, developmental delays and/or cerebral palsy [19,20,40,59]. Neural injury results from excitotoxicity, oxidative stress and inflammation, leading to a loss of oligodendrocytes and neurons, both of which are particularly sensitive to hypoxic-ischaemic insult [1-4, 10, 29, 37]. Neonatal stroke, despite its prevalence and significant impact on the health care system, is less well studied than adult stroke. The current treatment for neonatal stroke is hypothermia; however, the window of efficacy severely limits its potential to reduce the damage that results from the insult [41,51]. The need for novel interventions to limit the damage and promote neural repair is clear.
Metformin, a commonly used drug to treat type-2 diabetes, has been shown to have pleiotropic effects in the brain. Our group, and others, have shown that metformin administration leads to expansion of the endogenous neural precursor pool in their subependymal zone (SEZ) niche, migration of neural precursor cells into strokeinjured parenchyma, contribution to neurogenesis and oligodendrogenesis and improved motor and cognitive outcomes [10,48]. Metformin has also been shown to promote neurogenesis and reduce neuroinflammation in a model of juvenile cranial radiation [11,38], similar to other post-natal and adult models of brain injury where metformin's anti-inflammatory properties have been demonstrated by regulating the AMPK-NF-Kbdependent pathway [23,57]. The promising effects of metformin to support brain repair were further demonstrated in a study that examined metforminmediated recovery in a model of cranial radiation and a pilot clinical trial that demonstrated the safety and feasibility of metformin for children with acquired brain injury [3]. Given the pleiotropic effects of metformin, the cellular basis for the improved outcomes following perinatal brain injury and metformin administration is unclear.
Microglia are the resident immune cells of the central nervous system and during brain maturation, microglia perform several functions, such as synaptic pruning and supporting axon myelination [22,43,54,60]. In the uninjured post-natal brain, microglia adopt a ramified morphology with dynamic process movements, allowing them to sense the microenvironment [39]. It has been shown in the uninjured adult brain that microglia exhibit regional differences in terms of cell density, as well as inflammatory profiles. For example, cortical and striatal microglia display higher degrees of ramification, are less proliferative and less phagocytic than SEZ microglia [34,46,55,56,58]. Following injury, microglia become activated, a process that includes proliferation and pronounced de-ramification to amoeboid shaped cells, as well as increased cytokine release and phagocytosis, all of which are processes that can exacerbate brain damage [18,31,42,63].
The neuroinflammatory response after injury has also been shown to be dependent on astrocytes. Similar to microglia, astrocytes have pro-inflammatory and immunoregulatory (neuroprotective) subpopulations [52]. Astrocyte activation results in their hypertrophy, upregulation of GFAP and in severe insult, can result in their proliferation and the formation of a glial scar. Once activated, they can exacerbate injury by inhibiting neural progenitor homeostasis, impairing synapse homeostasis and increasing cytokine release [50,64]. Notably, astrocytes activation has been shown to be dependent on microglia activation both in vivo and in vitro [24,27,30].
In this study, we aimed to shed light on metformin's cellular targets following neonatal stroke. We characterized the microglia and astrocyte response, with and without metformin treatment, using a wellestablished model of neonatal stroke. Herein, we show that the microglia response is regionally distinct at early times post-stroke and metformin treatment reduces microglia activation, as measured by morphology and protein expression (CD68). Metformin treatment after neonatal stroke did not affect astrocyte activation. Using a novel model of microglia ablation, we show that the loss of microglia does not result in motor behaviour impairments in early post-natal mice. Most interesting, the loss of microglia protects against the functional deficits that result following neonatal stroke, mimicking the metformin-mediated outcomes after 1 week of metformin treatment post-stroke. Our work provides novel insight into metformin's anti-inflammatory effects after neonatal stroke and further highlights the role of microglia in the observed motor impairments following injury to the early post-natal brain.

Tissue preparation and immunohistochemistry Tissue preparation
Animals were sacrificed on P9, P12, P14, P15, P21 and P22 with an overdose of Avertin. Transcardial perfusions were performed using cold PBS followed by 4% paraformaldehyde (PFA; P6148, Sigma-Aldrich). The brains were removed and post-fixed in 4% PFA overnight, then cryoprotected in 30% sucrose solution and stored at 4 °C until cryosectioning. Brain sections (coronal, 20um) were mounted on Superfrost Plus microscope slides Fisher Scientific) and kept at − 20 °C until processing.

Imaging and quantification
Slides were imaged using an Axiovert 200 M inverted microscope (Zeiss, Jena, Germany) and an Axio Observer Z1 inverted motorized microscope (Zeiss, Jena, Germany) with a high-speed spinning disk CSU-X1 confocal scanner unit (Yokogawa, Texas, USA) and an Axiocam 506 high-resolution camera (Zeiss, Jena, Germany). Images were acquired with the AxioVision 4.8 and Zen software (Zeiss, Germany). Three to four sections were counted per brain. Three regions of interest were imaged in the cortex and striatum (each 350 × 300 um; 0.105 mm 2 ) and the lateral wall of the lateral ventricles (150 × 300 um; 0.045 mm 2 ) (including the dorsolateral corner), in the ipsilateral and contralateral hemispheres (relative to the ischaemic insult). Areas chosen extended from the crossing of the corpus callosum anteriorly to the crossing of the anterior commissure, posteriorly.
Amoeboid microglia were defined as DAPI + cells with a rounded cell body morphology with and a lack of processes (≤ 1 process less than 5um in length) extending from the soma. Cells in the cortex were manually counted using the Cell Counter plugin in Fiji (Schindelin et al., 2012). These counts were expressed as fold changes and percentages of cells per unit area for each of the regions.
To measure microglia intensity and soma size, images were exported as TIFF files and sections were converted to 8 bits and filtered (filter = MaxEntropy). Pixel counts and soma length measurements were generated for Iba1 + cells (1) in the cortex (each 350 × 300 um; 0.105 mm 2 ), (2) in the striatum (each 350 × 300 μm; 0.105 mm 2 ) and (3) and in the dorsolateral corner of the lateral ventricle (150 × 300 um; 0.045 mm 2 ) using the pixel inspection tool and measurement tool in Fiji, respectively, from > 3 sections per brain.

Behavioural assessments
Mice were handled on the day prior to behavioural testing in all tasks. On the day of testing, mice were given 30 min to acclimatize to the room before testing.

Righting reflex
The righting reflex was performed 1 h post-H-I [16]. Mice were placed in the supine position and the time needed to turn into prone position was recorded (maximum 60 s) and averaged over 3 trials.

Hindlimb suspension
On P12, mice performed the hindlimb suspension test, which assesses hindlimb strength [10,16]. Mice were placed head-down on the edge of a 50-ml falcon tube, with tissue at the bottom. Mice were suspended on the edge of the tube using their hindlimbs and the latency to fall was recorded for 3 trials (maximum 60 s). The mean latency of the trials was calculated per mouse.

Grip strength
On P15, mice performed the grip strength test [16]. This test assesses the paw strength of all 4 paws simultaneously. Mice were placed on an 11 × 7 inch grid, and it was rotated from a horizontal to a vertical position. The angle at which the mouse fell from the grid was recorded and averaged over 3 trials.

Cylinder
On P22, mice performed the cylinder test which assesses functional asymmetry. In this test, mice are placed in a Plexiglas transparent cylinder and recorded for 4 min or until each paw touches the cylinder at least 10 times. Forepaw preference was measured as the number of touches from the left, right or both forepaws. Percent forepaw preference was established using the following equation:

Statistical analysis
Statistical analyses were conducted using GraphPad Prism 9 (GraphPad Software, San Diego, CA, USA). All data are graphed as mean ± standard error of the mean (SEM). Statistical analysis consisted of unpaired t-tests, one-way ANOVA or two-way ANOVA. Where appropriate, Tukey's post hoc analysis was used. Statistical significance was established as p < 0.05. Experimenters were blinded to the experimental conditions during analyses.

Microglia increase in numbers after H-I, with or without metformin, in the cortex and striatum.
Our group has previously demonstrated that one week of metformin treatment is sufficient to reduce motor functional deficits following neonatal stroke [10,33], Additional file 1: Fig. S1A, B). We sought to investigate the neuroinflammatory response in the neonatal brain post-H-I and ask whether metformin regulated microglia and astrocyte activation in this model. Experiments were performed in GFAP-GFP mice. On P8, mice underwent H-I, which consisted of the ligation of the left common artery, followed by a one-hour rest period and subsequent onehour period of hypoxia. One hour following the entirety of the H-I injury, mice were assessed on the righting reflex to confirm that the H-I injury resulted in described impairments in motor function [10]. As expected, H-Iinjured mice showed a significant increase in the latency to prone (p = 7.34 × 10 -5 ) (Additional file 2: Fig. S2A, B). We characterized the neuroinflammatory cells, microglia and astrocytes, on P9, P12, P14 and P21, in control and H-I-injured mice, with and without metformin treatment from P9 to P14 (Fig. 1A). Our analyses focused on brain regions associated with motor function, namely, the motor cortex and striatum, based on previous work demonstrating that H-I-induced motor impairments are rescued by metformin administration (Dadwal et al., [33], Additional file 1: Fig. S1A, B). We also examined the subependymal zone (SEZ) which contains neural precursor cells (NPCs) that are known to be activated following H-I with enhanced proliferation and migration to the injured parenchyma where they give rise to new neurons and glia following metformin treatment [10].
We first assessed the total numbers of DAPI labelled cells per unit area in each of the cortex, striatum and SVZ of brains from the following groups of mice: Sham (uninjured), Sham + metformin, H-I only and H-I + metformin. We observed no difference in cell density across regions over time (P9, P12, P14 and P21) or between groups (Additional file 2: Fig. S2C-E). Next, we assessed the numbers of Iba1 + microglia in each region at P9, 12, 14 and 21. We found a significant increase in Iba1 + microglia in H-I-injured brains relative to Shams at P12 in both the cortex (2.4-fold increase; p = 0.0063) (Fig. 1B-F) and striatum (2.7fold increase; p = 0.032) ( Fig. 1 H-L) which returned to control levels by P14 (Fig. 1B-M). There was no change in numbers of Iba + microglia in the SEZ following H-I ( Fig. 1N-R). The increased numbers of microglia after H-I in the cortex and striatum were coincident with a significant increase in proliferating microglia (Iba1 + EdU + cells) in the cortex (p = 0.0064) and striatum (p = 0.0085) relative to Shams (Fig. 1G, M). Notably, in all regions, metformin treatment alone did not alter the numbers or the proliferation of microglia when compared to Sham-treated mice (Fig. 1F, G, L, M, R). Moreover, metformin treatment after H-I did not alter the magnitude of the microglia expansion (2.1-fold increase in the cortex, p = 0.073; 3.0-fold increase in the striatum, p = 0.29) of microglia when compared to H-I only-injured mice (Fig. 1F, M). Similarly, the percentage of Iba1 + Edu + cells was unchanged between H-I and H-I + metformin-treated mice in the cortex and striatum (Fig. 1G, M). Hence, H-I results in a temporally and a regionally distinct expansion in microglia, irrespective of metformin treatment.

Microglia morphology and protein expression is modified following H-I in the presence of metformin
While the numbers of microglia were not modified by metformin treatment, we predicted that changes in the activation state of the microglia would be seen based on metformin's anti-inflammatory effects after injury [11,21,32,38,57]. We examined two key features of the microglia activation after injury: morphology (i.e. amoeboid phenotype) and CD68 protein expression, a marker of phagocytosis. We quantified the relative percentage of Iba1 + microglia with an amoeboid phenotype, as defined by a rounded cell body morphology with a lack of processes extending from the soma. In the cortex, H-I increased the proportion of amoeboid microglia at P12 (43% increase from Sham; p = 8.43 × 10 -5 ), a time when increased numbers of microglia were observed ( Fig. 2B-E, F). A similar pattern of microglia activation was observed in the striatum, with significantly increased proportions of amoeboid microglia post-H-I and this activation was seen at both P9 (27% increase from Sham; p = 8.94 × 10 -9 ) and P12 (41% increase from Sham; p = 1.87 × 10 -11 ) ( Fig. 2I-L, M). Interestingly, while no changes in the numbers of Iba1 + cells were seen in the SEZ after H-I, we observed a significant increase in the relative percent of amoeboid microglia as early as P9 (38.73% increase from Sham; p = 8.85 × 10 -5 ) and at P12 (48.42% increase from Sham; p = 1.06 × 10 -5 ) ( Fig. 2P-S, T). Strikingly, H-I + metformin-treated mice showed a significant reduction in amoeboid microglia in the cortex at P12 (43% decrease from H-I; p = 0.00011), back to control (Sham) values (Fig. 2E, F). The same was true in the striatum, at P9 (26% decrease from H-I; p = 8.49 × 10 -7 ) and P12 (40% decrease from H-I; p = 3.66 × 10 -10 ), where metformin treatment after H-I resulted in a reduction in amoeboid microglia back to control (Sham injured) levels ( Fig. 2L, M). Identical to what was observed in the parenchyma, H-I + metformin-treated mice had significantly reduced SEZ amoeboid microglia at P9 (25% decrease from H-I; p = 0.0052) and P12 (43% decrease from H-I; p = 0.00012) that was not significantly different from Shams (Fig. 2S, T). We did not observe a difference in the soma size of Iba1 + cells between treatment groups in any region assessed (cortex, p = 0.15; striatum, p = 0.36; SEZ, p = 0.48) (Fig. 2G, N, U); however, we did observe a significant increase in the intensity of 77. G Quantification of percent proliferating microglia (Iba1 + EdU + /Iba1 +) in the cortex at P12 across treatment groups. A significant increase was observed after H-I (7.50 ± 2.86% Iba1 + EdU + cells/unit area in Sham vs. 45.98 ± 11.45% Iba1 + EdU + cells/unit area in H-I, p = 0.0064) and Met treatment did not prevent this expansion (45.98 ± 11.45% Iba1 + EdU + cells/unit area in H-I vs. 32.38 ± 2.07% Iba1 + EdU + cells/unit area in H-I + Met, p = 0.37). There was no significant difference between Met-treated mice and H-I + Met-treated mice (8.38 ± 4.13% Iba1 + EdU + cells/unit area in Met vs. 32.38 ± 2.07% Iba1 + EdU + cells/unit area in H-I + Met-treated mice, p = 0.056). H Representative image of microglia (Iba1 +) and proliferating microglia (Iba1 + EdU +) at P12 in the striatum in Sham mice (×20 magnification). Proliferating microglia are depicted by yellow arrows. I Representative image of microglia (Iba1+, red) and proliferating microglia (Iba1 + (red) EdU + (white) double-positive cells) at P12 in the striatum in Met-treated mice (×20 magnification). Proliferating microglia are depicted by yellow arrows. J Representative image of microglia (Iba1 +) and proliferating microglia (Iba1 + EdU +) at P12 in the striatum after H-I (×20 magnification). Proliferating microglia are depicted by yellow arrows. K Representative image of microglia (Iba1 +) and proliferating microglia (Iba1 + EdU +) at P12 in the striatum after H-I + Met treatment (×20 magnification). Proliferating microglia are depicted by yellow arrows. L Quantification of the number of microglia (Iba1 +) in the striatum across time points and groups expressed as fold change. Microglia were significantly increased after H-I at P12 relative to Sham 13. M Quantification of proliferating microglia (Iba1 + EdU + /Iba1 +) in the striatum at P12 across groups. Significant increases were observed between Sham and H-I (7.99 ± 2.41% Iba1 + EdU + cells/unit area in Sham vs. 33.58 ± 7.02% Iba1 + EdU + cells/unit area in H-I, p = 0.0085) and Met-treated mice vs. H-I + Met-treated mice (6.12 ± 1.16% Iba1 + EdU + cells/unit area in Met vs. 47.19 ± 2.64% Iba1 + EdU + cells/ unit area in H-I + Met, p = 0.15). Quantification of proliferating microglia (Iba1 + EdU + /Iba1 +) in the striatum at P12 across groups. Significant increases were observed between Sham (7.99 ± 2.41% Iba1 + EdU + cells/unit area) vs. H-I (33.58 ± 7.02% Iba1 + EdU + cells/unit area) (p = 0.009) and Sham (7.99 ± 2.41% Iba1 + EdU + cells/unit area) vs. H-I + Met (47.19 ± 2.64% Iba1 + EdU + cells/unit area) (p = 0.006). N Representative image of microglia (Iba1+) and proliferating microglia (Iba1 + EdU +) at P12 in the SEZ in Sham mice (×20 magnification). Proliferating microglia are depicted by yellow arrows. O Representative image of microglia (Iba1+, red) and proliferating microglia (Iba1 + (red) EdU + (white) double-positive cells) at P12 in the striatum in Met-treated mice (×20 magnification). Proliferating microglia are depicted by yellow arrows. P Representative image of microglia (Iba1 +) and proliferating microglia (Iba1 + EdU +) at P12 in the SEZ after H-I (×20 magnification). Proliferating microglia are depicted by yellow arrows. Q Representative image of microglia (Iba1 +) and proliferating microglia (Iba1 + EdU +) at P12 in the SEZ after H-I + Met treatment (×20 magnification). Proliferating microglia are depicted by yellow arrows. R Quantification of the number of microglia (Iba1 +) in the SEZ across time points and groups expressed as fold change. No significant differences were found between groups (p = 0. 15 the Iba1 + microglia after H-I that was consistent across brain regions (p = 0.0017 in the cortex, p = 0.046 in the striatum, and p = 0.0084 in the SEZ) relative to sham animals ( Fig. 2H, O, V). This finding is consistent with previous reports of cerebral ischaemia increasing the intensity of Iba1 + microglia at acute time-points after injury [25]. Interestingly, metformin treatment significantly reduced the intensity of Iba1 + microglia to control levels (cortex, p = 0.023,striatum, p = 0.021; SEZ, p = 0.0012) (Fig. 2H, O,V). Hence, metformin administration reduces microglia activation post-H-I. We further assessed microglia activation using CD68 protein expression, a marker of microglia phagocytosis, using immunohistochemistry (Additional file 3: Fig. S3). . This increase was prevented by HI + Met (10.59 ± 1.32% amoeboid Iba1 + cells in Met-treated mice vs. 14.11 ± 1.68% amoeboid Iba1 + cells in H-I + Met-treated mice, p = 0.999 relative to Met and p = relative to H-I + Met). The percentage of amoeboid microglia was significantly increased after H-I at P12 relative to Sham (18.61 ± 2.53%, amoeboid Iba1 + cells in Sham mice vs. 59.48 ± 4.05% in H-I-injured mice, p = 1.87 × 10 -10 ). This increase was prevented by HI + Met (7.89 ± 2.36% amoeboid Iba1 + cells in Met-treated mice vs.19.57 ± 3.67% amoeboid Iba1 + cells in H-I + Met-treated mice, p = 0.99).N Quantification of the soma length of microglia (Iba1 +) in the striatum across groups. There was no difference in soma length across treatment groups (7.98 ± 0.25 µm cell soma length in Sham 6.39 ± 0.68 µm after H-I; 7.68 ± 0.26 µm soma length in Met vs. 8.75 ± 1.60 µm in H-I + Met, p = 0.36). O Quantification of the pixel intensity of microglia (Iba1 +) in the striatum across groups. There was a significant increase in the intensity of Iba1 + microglia after H-I relative to Sham (59.5 ± 17.37 pixels in Sham vs. 105.7 ± 5.29 pixels after H-I, p = 0.046). There was no difference in pixel intensity after H-I + Met relative to Met (60.1 ± 5.29 pixels in Met vs. 105.7 ± 5.29 pixels after H-I, p = 0.92), but Met treatment significantly decreased in the intensity of Iba1 + microglia after injury (105.7 ± 5.29 pixels in H-I vs. 51.2 ± 7.51 pixels in H-I + Met, p = 0.021). Representative image of amoeboid microglia (Iba1 +) at P9 in the SEZ in Sham mice (×40 magnification). Image is representative of the first time point of change in morphology. Representative image of amoeboid microglia (Iba1 +) at P9 in the SEZ in Met-treated mice (×40 magnification). Image is representative of the first time point of change in morphology. Representative image of amoeboid microglia (Iba1 +) at P9 in the SEZ in H-I mice (×40 magnification). Image is representative of the first time point of change in morphology. Representative image of amoeboid microglia (Iba1 +) at P9 in the SEZ in H-I + MET mice (×40 magnification). Quantification of the amoeboid microglia (Iba1 +) in the SEZ across time points and groups. The percentage of amoeboid microglia was significantly increased after H-I at P9 relative to Sham (15.43 ± 1.54% amoeboid Iba1 + cells in Sham mice vs. 54.16 ± 8.53% in H-I-injured mice, p = 8.85 × 10 -5 ). This increase was prevented by HI + Met (15.54 ± 1.04% amoeboid Iba1 + cells in Met-treated mice vs. 22.87 ± 3.93% amoeboid Iba1 + cells in H-I + Met-treated mice, p = 0.841). The percentage of amoeboid microglia was significantly increased after H-I at P12 relative to Sham (18.62 ± 2.5%, amoeboid Iba1 + cells in Sham mice vs. 67.04 ± 8.81% in H-I-injured mice, p = 1.06 × 10 -5 ) and again, this increase was prevented by HI + Met (9.13 ± 2.07% amoeboid Iba1 + cells in Met-treated mice vs. 24.44 ± 1.02% amoeboid Iba1 + cells in H-I + Met-treated mice, p = 0.763). Quantification of the soma length of microglia (Iba1 +) in the SEZ across groups. There was no difference in soma length across treatment groups Coincident with the times we observed an increase in amoeboid microglia following H-I, we found a significant increase in the proportions of Iba + CD68 + /Iba1 + cells in the cortex at P12 (35% increase from Sham; p = 0.013) (Additional file 3: Fig. S3A-F); striatum at P9 (38% increase from Sham; p = 0.042) and P12 (46% increase from Sham; p = 0.0010) (Additional file 3: Fig. S3G-L); and SEZ at both P9 (46% increase from Sham; p = 0.0060) and P12 (47% increase from Sham; p = 0.023) (Additional file 3: Fig. S3M-R). Most striking, we observed a decrease in the proportion of Iba1 + CD68 + /Iba1 + cells in H-I + metformin-treated mice in all regions examined relative to H-I only mice (cortex, 32% decrease at P12, p = 0.022; striatum, 47% decrease at P9 and 42% decrease at P12, p = 0.013 and p = 0.002, respectively; SEZ, 53% decrease at P9 and 52% decrease at P12, p = 0.0024 and p = 0.013, respectively) (Additional file 3: Fig. S3F, K, L, Q, R). Together, these data demonstrate that metformin administration reduces the activation of microglia following H-I by reducing the presence of amoeboid, CD68 + expressing microglia.

Astrocytes increase in numbers after H-I, with or without metformin, in the cortex and striatum
Astrocytes are responsive to brain injury and recent work in the adult brain has revealed that microglia play a role in astrocyte activation [30,49]. Moreover, astrocytes in the adult brain are responsive to metformin in a model of Parkinson's Disease [30,49]. Based on these reports revealing the interplay between microglia and astrocytes and our observation that metformin regulates the microglia response post-H-I, we examined the numbers of GFAP expressing astrocytes in the cortex and striatum of GFAP-GFP mice in the same groups as described above (Fig. 3A. First, we examined the astrocyte response post-H-I only. In the cortex, we observed a significant increase in GFP + cells as early as P9 post-H-I (2.3-fold increase from Sham; p = 0.0026 and this was concomitant with an increase in GFP + EdU + /GFP + cells at P9 (p = 0.035) (Fig. 3B-G). In the striatum, GFP + cells were significantly increased at P12, the same time as microglia activation post-H-I (9.9-fold increase from Sham; p = 6.69 × 10 -9 ); however, this increase in numbers was not coincident with increased EdU + GFP + /GFP + cells (Fig. 3H-M), suggesting that the proliferation occurred prior to the time of EdU injection at P12.
Metformin administration following H-I resulted in the same increase in GFP + cells at P9 in the cortex (1.5-fold increase from Met, p = 0.0044), with a trend towards an increase in proliferating astrocytes (8% increase from Met; p = 0.27) (Fig. 3B-G). In the striatum, H-I + metformin treatment resulted in increased numbers of GFP + cells (4.1-fold increase from Met; p = 6.69 × 10 -9 ) (Fig. 3H-M). Again, this increase in numbers was not coincident with increased EdU + GFP + /GFP + cells, suggesting that the proliferation occurred prior to the time of EdU injection at P12. Taken together, these data suggest regional heterogeneity in terms of astrocyte activation and additionally, unlike what is observed in the adult injured brain, metformin administration does not dampen the astrocyte response in the neonatal brain.

Plexxikon 5622 successfully ablates microglia in neonates without affecting motor behaviour
Towards the goal of determining the cellular basis for metformin's positive effects on motor outcomes post-H-I and G Quantification of proliferating astrocytes (GFP + EdU + /GFP +) in the cortex at P9 across groups. Astrocyte numbers were significantly increased after H-I at P9 relative to Sham (2.64 ± 2.27% GFP + EdU + cells in Sham mice vs. 15.36 ± 2.37% in H-I-injured mice, p = 0.035). There was no significant difference in GFP + EdU + cells in Met-treated mice relative to and H-I + Met (7.63 ± 3.25% GFP + EdU + cells in Met-treated mice vs. 15.28 ± 2.77% GFP + EdU + cells in H-I + Met-treated mice, p = 0.27). H Representative image of astrocytes (GFP +) and proliferating astrocytes (GFP + EdU+, yellow arrows) at P12 in the striatum in Sham mice (×20 magnification). I Representative image of astrocytes (GFP+, green) and proliferating astrocytes (GFP + (green) EdU + (white) double-positive cells; yellow arrows) at P12 in the striatum in Met-treated mice (×20 magnification). J Representative image of astrocytes (GFP +) and proliferating astrocytes (GFP + EdU+, yellow arrows) at P12 in the striatum after H-I injury (×20 magnification). K Representative image of astrocytes (GFP +) and proliferating astrocytes (GFP + EdU+, yellow arrows) at P12 in the striatum after H-I + Met (×20 magnification). L Quantification of the number of astrocytes (GFP +) in the striatum across time points and groups expressed as fold change. Astrocyte numbers were significantly increased after H-I at P12 relative to Sham (1.00 ± 0.63-fold change of GFP + cells in Sham mice vs. 9.95 ± 2.46 in H-I-injured mice, p = 6.69 × 10 -9 ). This increase was not prevented by HI + Met based on our findings that metformin significantly reduced microglia activation post-H-I, we asked if microglia ablation prior to H-I would result in improved functional outcomes. We sought to establish a microglia ablation paradigm in which microglia would be selectively depleted within the neonatal brain at the time of injury and coincident with the time when metformin administration leads to improved functional outcomes [10], Additional file 1: Fig. S1). We used a pharmacological approach to deplete microglia using Plexxikon 5622 (PLX) is a colony stimulating factor 1 (CSF1) receptor antagonist, which inhibits microglia survival (Riquer and Sollars 2020; [53]. We administered daily i.p. injections of PLX to neonatal mice starting on P8. The number of Iba1 + microglia was assessed in the cortex, striatum and SEZ at P15 (Additional file 4: Fig. S4A). Throughout this study, pup weight was monitored daily as a measure of overall health. PLX treatment had no effect on weight (Additional file 4: Fig. S4B). We found that one week of 50ug/g PLX treatment (P8-P15 resulted in a significant depletion of Iba1 + microglia in the cortex (5.5-fold decrease from vehicle; p = 0.033 (Additional file 4: Fig. S4C-F), striatum (3.3-fold decrease from vehicle; p = 0.0034) (Additional file 4: Fig. S4G-J) and SEZ (3.3-fold decrease from vehicle; p = 0.0038) (Additional file 4: Fig. S4K-N). In the SEZ, previous studies have shown that SEZ microglia can regulate the behaviour of SEZ neural stem cells by enhancing their proliferation [35,46]. Hence, we predicted that the loss of microglia would reduce the number of neural stem cell-derived colonies (neurospheres) in vitro. The numbers of neurospheres were assessed at P15 following PLX treatment (when the greatest depletion of microglia is observed) and we observed no change in the numbers of neurospheres between vehicle-and PLX-treated mice (Additional file 4: Fig. S4O). Hence, microglia ablation in neonates does not result in changes in astrocytes and neural stem cell numbers.
Previous work in adult rodents has demonstrated that PLX treatment cessation results in a rapid recovery of microglia [14,53]. To determine whether a similar observation occurred in the neonatal brain, the numbers of Iba1 + cells were assessed in mice that received 50 ug/g PLX treatment for one week (P8-P15) and brains were examined at P22 (a time when functional recovery was observed following H-I + metformin treatment) [10] ( Fig. 4A-V). As shown in Fig. 4, complete repopulation of Iba1 + cells was observed by P22 after one week of (See figure on next page.) Fig. 4 Treatment cessation following microglia ablation, which had no effects on astrocytes or sham motor behaviour, causes rapid microglia repopulation. A Experimental timeline. Mice received PLX administration from P8 to P15 (SAC on P15 or P22) or P8 to P22 (SAC on P22). Immunohistochemistry (IHC) was performed on P15 and P22. Hindlimb suspension was performed on P12 and Grip Strength on P15. B Representative image of microglia (Iba1+, red) at P15 in the cortex of vehicle-treated mice (×20 magnification). C Representative image of microglia (Iba1+, red) at P15 in the cortex of mice treated with PLX from P8 to P15 (×20 magnification). D Representative image of microglia (Iba1+, red) at P22 in the cortex of vehicle-treated mice (×20 magnification). E Representative image of microglia (Iba1+, red) at P22 in the cortex of mice treated with PLX from P8 to P15. F Representative image of microglia (Iba1+, red) at P22 in the cortex of vehicle-treated mice (×20 magnification). G Representative image of microglia (Iba1+, red) at P22 in the cortex of mice treated with PLX from P8 to P22. H Quantification of the number of microglia (Iba1 +) in the cortex across groups expressed as fold change. There was a significant decrease in microglia after PLX treatment relative to vehicle at P15 (1.00 ± 0.11-fold change of Iba1 + cells in vehicle-treated mice vs. 0.19 ± 0.05 in H-I-injured mice, p = 0.0008). There was no difference in microglia numbers between vehicle vs. PLX at P22 following treatment from P8 to P15 (1.00 ± 0.11-fold change of Iba1 + cells in vehicle-treated mice vs. 0.80 ± 0.10 in H-I-injured mice, p = 0.646). There was a significant decrease in microglia after treatment relative to vehicle at P22 following two weeks of treatment (PLX from P8 to P22) (1.00 ± 0.14-fold change of Iba1 . J Representative image of microglia (Iba1 +) at P15 in the striatum of PLX-treated mice (×20 magnification). K Representative image of microglia (Iba1+, red) at P22 in the striatum of vehicle-treated mice (×20 magnification). L Representative image of microglia (Iba1+, red) at P22 in the striatum of mice treated with PLX from P8 to P15. M Representative image of microglia (Iba1+, red) at P22 in the striatum of vehicle-treated mice (×20 magnification). N Representative image of microglia (Iba1+, red) at P22 in the striatum of mice treated with PLX from P8 to P22. O Quantification of the number of microglia (Iba1 +) in the striatum across groups expressed as fold change. There was a significant decrease in microglia after PLX treatment relative to vehicle at P15 (1.00 ± 0.08-fold change of Iba1 + cells in vehicle-treated mice vs. 0.31 ± 0.04 in H-I-injured mice, p = 0.005). There was no difference in microglia numbers between vehicle vs. PLX at P22 following treatment from P8 to P15 (1.00 ± 0.17-fold change of Iba1 + cells in vehicle-treated mice vs. 0.91 ± 0.05 in H-I-injured mice, p = 0.982). There was a significant decrease in microglia after treatment relative to vehicle at P22 following two weeks of treatment (PLX from P8 to P22) . Q Representative image of microglia (Iba1 +) at P15 in the SEZ of PLX-treated mice (×20 magnification). R Representative image of microglia (Iba1+, red) at P22 in the SEZ of vehicle-treated mice (×20 magnification). S Representative image of microglia (Iba1+, red) at P22 in the SEZ of mice treated with PLX from P8 to P15. T Representative image of microglia (Iba1+, red) at P22 in the SEZ of vehicle-treated mice (×20 magnification). U Representative image of microglia (Iba1+, red) at P22 in the SEZ of mice treated with PLX from P8 to P22. V Quantification of the number of microglia (Iba1 +) in the SEZ across groups expressed as fold change. There was a significant decrease in microglia after PLX treatment relative to vehicle at P15 (1.00 ± 0.04-fold change of Iba1 + cells in vehicle-treated mice vs. 0.31 ± 0.01 in H-I-injured mice, p = 0.001). There was no difference in microglia numbers between vehicle vs. PLX at P22 following treatment from P8 to P15 (1.00 ± 0.13-fold change of Iba1 + cells in vehicle-treated mice vs. 1.20 ± 0.09 in H-I-injured mice, p = 0.463). There was a significant decrease in microglia after treatment relative to vehicle at P22 following two weeks of treatment (PLX from P8 to P22) (1.00 ± 0.04-fold change of Iba1 + cells in vehicle-treated mice vs.  (Fig. 4D, E, H), striatum (Fig. 4K, L, O) and SEZ (Fig. 4R, S, V). Finally, in a separate cohort of mice, PLX was administered for two weeks (P8-P22) in order to maintain the Ib1 + depletion and/or achieve a greater depletion than recorded for one week of treatment, and the numbers of Iba1 + cells were quantified on P22. In all regions, Iba1 + cells were significantly depleted but not significantly different compared to one week of PLX treatment in the cortex (3.3-fold decrease from vehicle; p = 0.0033 (Fig. 4F-H), striatum (3.2-fold decrease from vehicle; p = 0.005) (Fig. 4M-O) and SEZ (2.0-fold decrease from vehicle; p = 0.016) (Fig. 4T-V). Hence, PLX treatment to neonatal pups leads to a significant loss of microglia, which are rapidly repopulated upon cessation of the PLX treatment.
Given the documented interplay between astrocytes and microglia, we asked whether microglia ablation resulted in an astrocyte response in the neonatal brain. We counted the numbers of astrocytes (GFP + cells) in the cortex and striatum following 1 or 2 weeks of PLX treatment and found no change in GFP + cells (Additional file 5: Fig. S5A). We next examined the activation profile of astrocytes after H-I and in the absence of microglia (i.e. H-I + PLX treatment). At P15, a time when H-I + PLX-treated mice had the greatest depletion of microglia and no behavioural deficits, there was no significant chance in astrocytes (GFP + cells) in the cortex (p = 0.12) (Additional file 5: Fig. S5B-F) or striatum (p = 0.22) (Additional file 5: Fig. S5G-K) relative to any treatment group. Hence, microglia ablation in neonates does not result in changes in astrocytes numbers.
While microglia ablation does not lead to functional impairments in uninjured adults, this has not been explored in neonates [14,53]. We therefore asked if microglia ablation impacted motor behaviour using functional assays to evaluate gross motor function during and at the end of PLX treatment. PLX administration started on P8 and the hindlimb suspension test was performed at P12 and the grip strength test at P15 in vehicle-and PLX-treated mice. In both tasks, PLXtreated mice behaved similarly to vehicle-treated controls (Fig. 4W, X). These findings reveal that microglia ablation following PLX treatment does not result in motor impairments in uninjured mice.

Microglia ablation mimics the effects of metformin treatment following H-I
To ask if the loss of microglia activation post-H-I was sufficient to reduce motor impairments following H-I, mice received H-I on P8 and PLX treatment until P15 (P8-15) (Fig. 5A). We performed immunohistochemistry for Iba1 + cells at P15 and found that similar to uninjured mice treated with PLX, there was a significant depletion in microglia in the cortex (2.4-fold decrease from H-I; p = 0.0076) (Fig. 5B-E), striatum (3.4-fold decrease from H-I; p = 0.00060) (Fig. 5F) and SEZ (3.0-fold decrease from H-I; p = 0.0022) (Fig. 5G) of mice given H-I and PLX. Hence, PLX treatment is sufficient to ablate microglia in both the uninjured and H-I injured neonatal brain.
Based on our findings that (1) metformin treatment after H-I reduces microglia activation and (2) metformin administration for 1 week post-H-I are sufficient to reduce motor impairments in the cylinder test on P22, we predicted that metformin's anti-inflammatory effects on microglia may be mediating the improved functional outcomes seen after 1 week of metformin treatment (P8-15) and motor assessment on P22 [10] (Additional file 1: Fig. S1, Fig. 5A). Thus, we examined the motor outcomes following H-I + microglia ablation (PLX). Mice received H-I and PLX treatment from P8 to P15 to ablate microglia over the same time course that metformin administration is effective at reducing functional impairments in previous studies [10], Additional file 1: Fig. S1). Mice that received H-I only or H-I + PLX showed a significant increase in the latency to prone position in the righting reflex task when compared to uninjured mice (Fig. 5H). This validated the injury in the H-I-treated mice. On P12, the hindlimb suspension test was performed and only mice that had received H-I displayed a significant motor impairment (Fig. 5I). Similarly, the grip strength test on P15 revealed a deficit in H-I injured mice with no deficit observed in H-I + PLX mice which were not significantly different from uninjured groups (Fig. 5J). Most striking, H-I + PLX-treated mice were not impaired in the cylinder task at P22, identical to what is observed in H-I + metformin-treated mice that have metformin-mediated reduced microglia activation (Fig. 5K). Together, these findings reveal that a reduction in microglia activation, either through metformin treatment or ablation, is sufficient to eliminate the motor impairments that result following neonatal H-I injury.

Discussion
In this study, we characterized the spatiotemporal response of microglia and astrocytes, in the neonatal brain following H-I, and modified microglia activation using two paradigms (Metformin and PLX). First, we asked whether the drug metformin, which is known to have pleiotropic effects on cells in the CNS, impacts microglia activation, with the goal of uncovering the cellular targets that underlie the beneficial outcomes observed following neonatal stroke and metformin treatment. Our findings reveal that microglia and astrocyte activation are regionally and temporally distinct within the injured brain and further that microglia activation is reduced during the time that metformin is administered post-H-I. We next developed a novel microglia ablation paradigm (PLX treatment) in neonatal mice and confirmed that the loss of microglia in the early post-natal brain is sufficient to protect against functional impairments following H-I. We found that reducing the activation of microglia-demonstrated by reduced proliferation, amoeboid morphology, CD68 expression and total numbers of microglia-through metformin treatment or microglia ablation resulted in a complete lack of motor impairments following H-I. The microglia ablation timeline mimicked the metformin administration that resulted in improved functional outcomes, implicating microglia as an important target for promoting recovery following neonatal injury.
Our study investigated the effects of modifying the activation state of microglia, as measured by proliferation, morphology and protein expression, and the impact on behavioural outcomes. A number of reports have shown that metformin has anti-inflammatory properties and can improve the microenvironment after injury by modulating the release of cytokines [15,65]. Delivery of metformin has been shown to inhibit the inflammatory response 24 h post-injury in the early post-natal rodent brain and there have been reports of fewer Iba1 + microglia and GFAP + astrocytes in the cortex of after H-I + Metformin treatment compared to after HI-alone [15]. Interestingly, decreased expression of IL-6 and TNF-α was observed in the cortex of injured brains after metformin treatment and it was shown to be attenuated by the suppression of the TLR4/NF-κB signaling pathway, and notably, TLR4 is widely expressed on microglia [61]. Additional support for reduced levels of TNF-α, and IL-6, in addition to IL-1β, comes from studies of chronic metformin pre-treatment in a rodent model of adult stroke [65]. Metformin treatment ameliorated the microglial and astrogliotic response observed in mice that received stroke only [65]. Further, microglia numbers, proinflammatory morphology and expression of CD68 were also reduced after neonatal H-I when metformin treatment was delayed post-stroke [33]. Taken together, this suggests that understanding the factor(s) that modify the microenvironment and regulate microglia behaviour is potential targets and important considerations for brain repair strategies.
Moreover, the effects of metformin pre-treatment on microglia in the neurogenic niche, when metformin has been given prior to juvenile cranial radiation, has also revealed reduced microglia activation as measured by microglia numbers and CD68 expression [11]. Reduced microglia activation was measured using morphology and protein expression (CD68) and we observed a rapid increase in amoeboid microglia after H-I, which is consistent with previous studies [6,18]. Both measures of activation were significantly reduced following metformin treatment. Interestingly, astrocytes, another key player in the neuroinflammatory response, were similarly activated by H-I and did not show a reduction in numbers in the presence of metformin. More recent work has shown that pro-inflammatory cytokines released by microglia are required for astrocytes activation [30], hence, we had predicted that astrocyte activation would occur after the microglia activation following H-I. This prediction was not supported by our findings, but instead, we observed a regionally distinct astrocyte activation that was earlier than microglia activation in the cortex of neonatal injured mice. These findings may highlight the differences between the neonatal and mature injured brain in response to injury.
We tested the hypothesis that neonatal H-I would trigger a microglia response and that metformin would dampen this response. Our observation that metformin did not reduce the microglia population after H-I was not predicted based on reports that metformin administration in the sub-acute period following early post-natal injury [15] reduces the absolute numbers of microglia. The different injury models and species (rat versus mouse) may account for the difference. In addition, the age of the mice at the time of injury, treatment and when the microglia are assessed can impact the outcomes observed. With regard to age, for example, microglia isolated from neonatal (2 days old) and weaned mice (21 days old) after LPS treatment revealed that only neonatal microglia showed increased mRNA expression of inflammatory markers, suggesting that microglia are much more responsive to injury in the neonate than in older animals [9]. These age-related differences are further supported by the fact that following H-I and sub-acute metformin treatment, the absolute numbers of microglia were decreased at P28 (i.e. adult microglia) [33]. The observed change in microglia and astrocyte numbers was not reflected in a change in the total number of DAPI + cells. This is consistent with concomitant with increases in other cell populations after metformin treatment, such as increased numbers of neurons and oligodendrocytes, as previously reported following neonatal stroke metformin administration [10]. Hence, our finding that H-I and metformin alter proliferation at early times post-H-I in mice and may reflect age-and speciesdependent aspects of the neuroinflammatory process.
Metformin significantly reduced the numbers of amoeboid microglia after H-I. Microglia from H-I and metformin-treated brains failed to display an amoeboid phenotype which could be explained by (1) metformin preventing microglia from taking on an amoeboid phenotype or (2) H-I inducing the amoeboid phenotype and metformin reverts the activated microglia to their "resting" ramified phenotype. Given the early time point after injury at which we recorded a change in morphology after H-I (i.e. 1 day post-injury), our data support the hypothesis that metformin prevents the amoeboid phenotype. Microglia shape has been shown to be dependent on the Par1b/MARK2 pathway [13] (i.e. polarity-regulating kinase partitioning-defective pathway 1b/microtubule affinity-regulating kinase 2). Interestingly, Par1 activates NF-kb in lung endothelial cells following thrombin stimulation in vitro [7,44]. Accordingly, metformin may act on the Parb/MARK 2 and NF-kb pathway to inhibit microglia from becoming amoeboid. Another role of metformin is to alter the inflammatory profiles of microglia (i.e. between proinflammatory and anti-inflammatory states) [32]. Along these lines, metformin treatment after MCAO resulted in decreased staining for CD16/32 (proinflammatory) and increased staining for CD206 (antiinflammatory at 14 days when compared to saline controls [26]. Interestingly, the cellular response to metformin is age dependent as have been seen with neural precursor cells [48] as well as with microglia [28] where metformin treatment led to increased proportions of CD206 + microglia and importantly, altered microglia morphology by increasing the number and complexity of their ramifications [28]. While others have examined the effects of different CSF1R inhibitors (BLZ945, PLX3397) on microglia in postnatal mice [36,62], as well as in the context of H-I injury [62], our study builds on a study of PLX5622 in postnatal rats [47]. We have investigated the effects of PLX5622-mediated microglia depletion in neonatal mice on motor behaviours, in uninjured and stroke injured neonatal mice. Consistent with previous work, we observed depletion of microglia following the delivery of the CSF1R inhibitor and eventual repopulation after delivery of PLX5622 ceased [36]. Importantly, our experiments have shown that one week of Plexxikon 5622 treatment is sufficient to cause significant (~ 80%) depletion of microglia in the cortex, striatum and SEZ, without affecting astrocytes, NPCs or motor behaviour. Although microglia ablation has shown no motor or cognitive deficits in adults, we hypothesized that this may not be the same for neonates given the importance of microglia in early postnatal development [22,43,54,60]. The effect of microglia ablation on fine motor and cognitive development in neonates remains unknown and more research is needed. Further highlighting the differences between adult and neonatal microglia ablation paradigms, Bellver-Landette and colleagues (2019) demonstrated worsened tissue and functional outcomes in the absence of microglia in the context of adult spinal cord injury [5]. Herein, we have shown that microglia ablation after H-I results in a lack of functional impairment on a battery of motor tests, which was correlated with a lack of functional impairment following metformin treatment after H-I. Therefore, we propose that metformin prevents microglia activation, and this is sufficient to prevent motor deficits following neonatal H-I.