Microglia-derived FGF21 as a modulator of astrocytic phenotype and cerebral ischemia injury

Background: Fibroblast growth factor 21 (FGF21) is an important neuroprotective factor in the central nervous system (CNS), and it has been reported that FGF21 can protect against cerebral ischemia during the acute phase. However, the possible effects of FGF21 on ischemic brains and the interactions between FGF21 and nonneuronal cells have not been examined. Thus, the aim of this study was to elucidate the protective effects of endogenous FGF21 in ischemic brains. Methods: In this study, in vivo ischemia/reperfusion injury mouse model established by transient middle cerebral artery occlusion (MCAO)/reperfusion and in vitro cell models of oxygen/glucose deprivation (OGD)/reoxygenation (R) were used. Western blot analysis, RT-PCR, double immuno�uorescence staining, immunohistochemistry, 2,3,5-triphenyltetrazolium chloride (TTC) staining, hematoxylin-eosin (H&E) staining, neurobehavioral tests, cell counting kit-8 (CCK-8) assay and high-throughput gene sequencing were employed to explore the mechanism by which FGF21 unleash neuroprotective effort of astrocyte phenotype shifts in ischemic stroke. Results: We found that cortical FGF21 expression signi�cantly increased after MCAO/reperfusion, peaking at 7 d. Ischemia-activated microglia were the main sources of endogenous FGF21 in brain tissue. However, FGF21 de�ciency aggravated brain injury and slowed neurological functional recovery in FGF21 knockout mice. The in vitro and vivo studies revealed that FGF21 could activate astrocytes and mediate astrocytic phenotype. FGF21-activated astrocytes contributed to neuronal survival and synaptic protein upregulation after ischemia. Conclusion: Collectively, our data indicate that FGF21 plays vital roles in alleviating ischemic brain by mediating the manifestation of potentially pro-recovery astrocytic phenotypes. Therefore, modulation of FGF21 is a potential target strategy for stroke.


Background
Ischemic stroke is a common neurological injury and the third leading cause of disability and death in China [1].Unfortunately, the physiopathological mechanism of ischemic stroke is complex, and recombinant tissue-type plasminogen activator (rt-PA) is considered to be the most effective treatment agent [2,3].However, intravenous thrombolysis fails to show excellent therapeutic effects against ischemic stroke.After cerebral blood ow is restored, the postischemic in ammatory response and several detrimental factors, including excitotoxic neurotransmitters and cytokines, cause further neuronal injury [1].Elucidation of the mechanisms of neuronal injury and repair in order to reduce the severity of neurological diseases and discovery of other e cacious treatments targeting neuronal protection are urgently needed.
Currently, endogenous neuronal protection is a promising strategy for neurological recovery and regeneration after ischemic injury [4].The effects and functions of nonneuronal cells, such as microglia and astrocytes, supporting neuronal survival should not be neglected [5].Accumulating evidence suggests that circulating factors released by glia subjected to ischemia play important roles in the pathogenesis of stroke [6].Microglia, the brain-resident phagocytes, can be activated in response to brain injury and diseases and are involved in neuroin ammation in ischemic brains [7,8].Microglia, especially those of the M2 phenotype, can promote brain repair by producing trophic factors, which are potentially helpful for neurons and can remove apoptotic neurons to balance the microenvironment in ischemic brains [9].For instance, interleukin-10 (IL-10) from M2 microglia attenuates local in ammatory reactions by downregulating the expression of proin ammatory cytokines and chemokines after ischemic stroke [10].In addition, Ma et al. further demonstrated that microglia exert a protective role by releasing neurotrophic factors, such as brain-derived neurotropic factor (BDNF), insulin-like growth factor 1 (IGF-1), and nicotinamide phosphoribosyl transferase (NAMPT) [11,12].Similarly, astrocytes, the most abundant glial cells in the central nervous system (CNS), can be induced to develop different phenotypes depending on the damage and the phase [13].In contrast to A1 astrocytes, which upregulate classic complement cascade genes and are destructive to synapses, A2 astrocytes participate in the regulation of neurodevelopment and the promotion of synapse formation by producing neurotrophic factors [14].Previous studies have shown that growth and nutritional factors or proteins (DJ-1, MEGF10), which can sustain astrocytic neuroprotection, are synthesized and secreted by A2 astrocytes [15].Recently, DJ-1 has been found to induce neuroprotection by mediating the Nrf2/ARE pathway [16].In addition, Chung et al. have reported that astrocytes actively contribute to synapse elimination in the developing CNS via astrocytic MEGF10 [17].Furthermore, other endogenous trophic factors, such as FGF10 and FGF2, produced by ischemic brains are essential for brain repair, as they inhibit NF-κB-dependent neuroin ammation and upregulate the expression of PDGFRβ [18,19].However, more work is needed to elucidate the roles of other factors from glia in ischemic brains.
FGF21 is a unique member of the FGF superfamily [20], and its role in type 2 diabetes related to the promotion of insulin synthesis and secretion has been studied widely [21].It has also been revealed that exogenous FGF21 administration ameliorates ischemia-reperfusion injury in cardiomyocytes [22], improving the antioxidant capacity of endothelial cells [23].Moreover, a growing body of evidence supports the notion that FGF21 can be released by brain tissue [24].Under pathological conditions, FGF21 acts as a therapeutic factor, improving insulin sensitivity and glucose uptake to attenuate metabolic stress-induced Alzheimer disease pathology [25].Chen et al demonstrated that FGF21 protects the blood-brain barrier by forming an FGF21/FGFR1/β-klotho complex and activating PPAR-γ after traumatic brain injury [26].However, animal model studies on neurological diseases have paid no attention to the effects and detailed mechanisms of endogenous FGF21, and the potential e cacy of endogenous FGF21 under physiological conditions is not clear.Hence, we hypothesized that endogenous FGF21 is an urgent neuroprotective modulator serving to ameliorate brain damage induced by ischemia/reperfusion.The aim of the present study was to examine the functions of endogenous FGF21 during long-term brain rehabilitation after ischemic stroke and to explore more neuroprotective mechanisms between FGF21 and nonneuronal cells.

Animals models
All animal experiments were authorized by the Laboratory Animal Ethics Committee of Wenzhou Medical University, and the procedures were conducted in compliance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.Wild-type male C57BL/6 mice weighing 20-25 g were obtained from the Animal Center of the Chinese Academy of Sciences (Shanghai, China), and FGF21 knockout (KO) male C57BL/6 mice were supplied by Chinese-American Research Institute for Diabetic Complications (Wenzhou Medical University, China).These mice were subjected to transient middle cerebral artery occlusion (MCAO)/reperfusion as previously described [27].In brief, each animal was anesthetized with iso urane, and its rectal temperature was maintained at 36.5 ℃-37.5 ℃.The right common carotid artery (CCA) was isolated, and the internal carotid artery (ICA) and external carotid artery (ECA) were exposed.Then, a silicone rubber-coated nylon mono lament was inserted into the right middle cerebral artery (MCA).A laser Doppler owmeter (MedWrench, Brentwood, USA) was used to detect cerebral blood ow (CBF), and mice with CBF reductions of more than 80% compared with the baseline were selected.The occluding lament was withdrawn after 80 min to allow reperfusion.
Recombinant human FGF21 (rhFGF21) (Wenzhou Medical University, China) was dissolved in a sterile 0.9% normal saline solution.rhFGF21 (1.5 mg/kg) or vehicle control (sterile 0.9% normal saline solution, the volume of vehicle is the same as that of rhFGF21 solution) were injected intraperitoneally at the end of reperfusion, and were administered at 24 h intervals for 7 d.

Immunohistochemistry and image analysis
On 7 d, brains were removed and post-xed in 4% PFA at 4 °C overnight.Subsequently, the brains were embedded in para n and sectioned into 5 μm slices.Followed by xylene depara nation, ethanol rehydration, and incubated with 3% H 2 O 2 (diluted in PBS solution) for 15 min.Then, the sections were boiled in citrate buffer for 2 min to retrieve antigen, and nonspeci c antibody binding was blocked by 5% bovine serum albumin (BSA) for 30 min at 37 °C.Rabbit monoclonal anti-FGF21 (1:250, ab171941, Abcam) were incubated with tissue sections at 4 °C overnight, followed by incubation in goat anti-rabbit secondary antibody (1:800, ab6721, Abcam) for 2 h at 37 °C.Sections were colored with DAB-kit (ZSGB-BIO, Beijing, China), and then counter stained with hematoxylin (Beyotime Institute of Biotechnology, Shanghai, China).Finally, the resulting images were examined under white light using a Nikon ECLIPSE 80i microscope (Nikon, Tokyo, Japan) and analyzed using ImageJ software.The percent of ipsilateral FGF21-positive area increase was calculated using the following formula: [1-(area of ipsilateral FGF21 staining/area of contralateral FGF21 staining)] ×100%.

Neurobehavioral tests mNSS scores
To evaluate neurological de cits, modi ed neurological severity scores (mNSSs) were performed over 14 d after MCAO/reperfusion as reported previously [28].Brie y, mNSSs were graded on a scale of 0-18, mice were assessed for motor, re ex and sensory performance.The scores were awarded when mice were unable to perform the tests or lacked the tested re exes as mentioned.The higher the scores represented strict damage.

Corner turning test
The corner test was used to access sensorimotor abnormalities as previous described [29].A corner was made with an angle of 30° using two cardboard pieces, and then mice entered the corner spontaneously.
Normal mice turned either left or right, while MCAO/reperfusion induced-mice preferentially turned toward the left (ischemic) side.The percentage of left over 20 trials were calculated.

Forelimb grip strength
Forelimb grip strength of mice was tested by a recording grip meter according to reported [30].In brief, mice were held by tail and allowed to grasp a wire ring in diameter connected to a strain gauge.The force required to pull the subject away from the wire was recorded as the measure of the grip strength.

Rotarod test
To judge motor balance and coordination, the rotarod test was performed using a rotarod apparatus (Comerio, Solbiate Olona, Italy) on the basis of JK et al [31].After placed onto the rungs of the accelerating rotarod, in which the speed increased slowly from 5 to 40 rpm over 4 min, the time that mice remained on the rotarod was measured.The mean duration (in seconds) over three individual trials was recorded.
Infract volume measurement 3 d after MCAO/reperfusion, the brains were quickly removed and sliced into 2-mm-thick sections.The brain slices were immersed in cold 1% 2,3,5-triphenyltetrazolium chloride (TTC, Sigma-Aldrich, Missouri, USA) for 15 min at 37 °C in dark, and then xed in 4% paraformaldehyde at 4 °C overnight.The infarct area in white of each section was measured by Image J software.Percentage of infarct volume was calculated using the following equation: [(contralateral hemisphere volume-uninfracted area of ipsilateral hemisphere volume)/contralateral hemisphere volume×2] ×100% Histological staining Hematoxylin-eosin (H&E) staining was performed to evaluate the histopathological changes of brains based on the previous protocol [32].Brain sections were dried at 60-65 ℃ for 30 min, brain sections were subjected to H&E solution (Beyotime Institute of Biotechnology) after depara nized and rehydrated.
Images were captured using a Nikon Eclipse 80i microscope (Nikon) and analyzed by ImageJ software.The residual volume was calculated using the following formula: (ipsilateral hemisphere volume /contralateral hemisphere volume) ×100%.

Primary microglia and astrocyte cultures
Cortical microglia and astrocyte cultures were isolated and puri ed from 2-day-old neonatal Sprague-Dawley rats as reported previously [34].In brief, after digesting and ltering, the mixed cells were plated on 75 cm 2 asks and cultured in DMEM/F12 medium containing 10% FBS and 1% Pen/Strep for 14 d.Mixed medium with a trypsin solution (0.25% trypsin diluted 1:4 in DMEM/F12) were induced for 30 min, microglia still remained attached to the bottom of the well, while other cells were detached [35].
Mixed glial cultures were shake at 220 rpm for 4 h, and at 180 rpm overnight to remove non-astrocytic cells.Astrocytes were dissected and reseeded on poly-L-lysine-coated plates.After the cells reached 80-90% con uence, cultures were treated with vehicle-or 100 nM of recombinant human FGF21 (rhFGF21) (Wenzhou Medical University, China) for 4 h, and cultures were switched to neurobasal medium containing 2% B27, 0.5 mM L-glutamine and 1% Pen/Strep subsequently.Another 24 h later, astrocytic conditioned medium was collected, and were transferred to neurons during the reoxygenation for further experiments.
Oxygen glucose deprivation (OGD)/Reoxygenation (R) For OGD/R treatment, the cell cultures were subjected as previously reported [36].After 30 min of OGD, the primary cultures (neurons, microglia and astrocyte) were replaced with normal medium, and cells were allowed to reoxygenation for 24 h for protein expression assay.On the other hand, neuron cultures were transferred to an anaerobic chamber for 2 h.At the end of OGD treatment, conditioned medium from astrocytes were added, and the cultures were placed in normal atmosphere for 24 h.

Cell counting kit-8 (CCK-8) assay
Cellular viability was subjected to CCK-8 (Beyotime Institute of Biotechnology) assay.After cells were treated, the CCK-8 solution was added to the culture medium (CCK-8 diluted 1:10 in medium), which were then incubated in the dark for 4 h at 37 °C.The optical density (OD) was measured at 450 nm (BioTek, Winooski, VT, USA).The percentage of cellular viability was calculated using the following equation: (OD value/control OD value) ×100%.

RNA isolation and Real time-PCR analysis
Total RNA from astrocytes and brain tissues isolated using Trizol reagent (Roche, Basel, Switzerland), and complementary DNA (cDNA) were obtained after reverse transcription (Prime Script RT Reagent Kit, Takara, Tokyo, Japan).Then Real time-PCR reactions were performed as follows: incubation at 50 °C for 2 min, denaturing at 95 °C for 10 min, followed by 40 cycles of 95 °C for 15 sec, and 60 °C for 1 min.
Primers were designed by Sangon Biotech (Shanghai, China) and the PCR primer sequence for each gene was provided in Table 1.High-throughput gene sequencing Three harvested vehicle-or rhFGF21-treated astrocyte samples from each group were subjected to highthroughput gene sequencing by Sangon Biotech (Shanghai, China).Brie y, total RNA was isolated using a Total RNA Extraction Kit, and a Qubit RNA Assay Kit and a Qubit 2.0 Fluorometer (Life Technologies, CA, USA) were used to measure the purity, concentration, and integrity of the RNA.Then, 0.1-1 μg of the total RNA was used as input material, and a VAHTSTM mRNA-seq V2 Library Prep Kit for Illumina® was used to generate sequencing libraries.PCR ampli cation and mRNA sequencing library construction were performed with fragmented mRNA.After the library quality was evaluated, mRNA sequencing was performed and sequencing reports generated by Sangon Biotech (Shanghai, China).

Statistical analysis
All value analysis was presented as the mean ± SEM of at least three independent experiments, and were performed using GraphPad Prism 7.0 (GraphPad Software Inc., San Diego, CA, USA).Two-way ANOVA for more than two groups.One-way ANOVA followed by Tukey's tests when analyzing more than two groups.Non-paired Student's t-test was used for the comparisons of two groups.A value of P<0.05 was considered statistically signi cant.

Ischemia induces FGF21 expression by microglia in the cortical area
The levels of endogenous FGF21 in the brain at different time points after MCAO/reperfusion were tested.We discovered that the mRNA levels of FGF21 were mainly concentrated in the cortical area but that FGF21 levels in the striatum signi cantly increased at 14 d after MCAO/reperfusion (Fig. 1a).In the cortical penumbras of mice subjected to MCAO/reperfusion, FGF21 protein levels increased and reached high levels from 1 d to 7 d before gradually declining to the levels found in the sham-operated mice over the next 7 d (Fig. 1b, c).Furthermore, compared with that in the contralateral cortex, the protein expression of FGF21 in the ipsilateral cortex was signi cantly elevated (by 2-fold) at 7 d (Fig. 1d, e).Moreover, immunohistochemistry further con rmed the increased FGF21 levels in ischemic cortices (Fig. 1f-i).Thus, 7 d after MCAO/reperfusion was selected as the optimal time point for the follow-up experiment.
Double-immuno uorescence staining of brain sections was performed to explore which types of cells were the main sources of endogenous FGF21 in ischemic cortices.The number of FGF21immunoreactive puncta was greater in the ischemic core than in the boundary zone (Fig. 1j, k).FGF21positive puncta were largely colocalized with Iba1-marked microglia, while few NeuN-positive neurons and GFAP-positive astrocytes were observed in the ischemic core (Fig. 1j).Additionally, some FGF21 was observed in microglia and astrocytes in the boundary zone (Fig. 1k).
Next, primary neuron, astrocyte and microglia cultures were subjected to OGD/R, and FGF21 protein expression was detected in vitro.As anticipated, Western blotting showed that FGF21 expression was extremely upregulated (by 5-fold) in microglia after 30 min of OGD and 24 h of reoxygenation (Fig. 1l, m).In contrast, FGF21 expression was not induced in the neuron or astrocyte cultures (Fig. 1l).Thus, our data strongly suggested that ischemia induced FGF21 generation by microglia in ischemic cortices.

FGF21 de ciency increases brain atrophy and aggravates neurological de cits
Given the amelioration of brain atrophy after administration of exogenous rhFGF21 [33], we hypothesized that endogenous FGF21 was capable of attenuating neurological de cits.There were no differences in appearance between wild-type (WT) and FGF21 KO mice (Fig. 2a).Determination of the cerebral infarct volume by TTC staining did not reveal any differences caused by the absence of FGF21 at 3 d after ischemia (Fig. 2b, c).However, FGF21-de cient mice exhibited larger residual brain volumes than WT mice at 14 d after MCAO/reperfusion (Fig. 2d, e).These ndings veri ed that endogenous FGF21 could not attenuate the brain infarct volume during the acute stage of stroke but could ameliorate neuronal tissue loss over the long term.
Moreover, successive functional recovery of the brain after rhFGF21 treatment has previously been reported in a neonatal cerebral hypoxia/ischemia model [33], so we subsequently determined whether FGF21 de ciency induced detrimental functional outcomes.A battery of neurobehavioral tests was carried out during both the early and late periods of MCAO/reperfusion.As shown in Figure 2f, g and i, the sensorimotor outcomes were similar between WT and FGF21 KO mice during the acute stage; however, protracted deterioration was observed in FGF21 KO mice from 7 d to 14 d after ischemic injury.The latency to walk on a rotarod was greater in FGF21 KO mice than in WT mice, and the results for forelimb strength were the same as those for rotarod walking latency (Fig. 2f, g).Similarly, a corner-turning test suggested delayed recovery of FGF21 KO mice compared with WT mice during the late stage (Fig. 2i).However, FGF21 KO mice exhibited much worse outcomes than WT mice during the early stage in the modi ed neurological severity score (mNSS) test (Fig. 2h).Collectively, the behavioral assay data demonstrated that the absence of FGF21 led to persistent impairment of neurological function at late stages after ischemic injury.
rhFGF21-mediated astrocytes contribute to neuronal survival and upregulate synaptic protein expression FGF21 attenuates neuronal apoptosis in the penumbra after MCAO/reperfusion through the FGFR1/PI3K/caspase-3 signaling pathway [37].Here, we investigated whether FGF21 also promotes bene cial neuronal outcomes by mediating glial activity.After we treated astrocytes with rhFGF21 (100 nM), we transferred the conditioned medium to OGD-induced neurons (Fig. 3a).There were no visible morphological changes in neurons after treatment with the conditioned medium or with rhFGF21 alone (Fig. 3b).Notably, these media were protective; in particular, the conditioned medium from rhFGF21treated astrocytes could protect neurons against OGD.In addition, normal neuronal viability gradually decreased by 40% after 2 h of OGD and 24 h of reoxygenation (Fig. 3c); however, rhFGF21-treated astrocyte conditioned medium (21ACM) effectively increased neuronal viability (Fig. 3c).rhFGF21 alone also exhibited a protective effect against OGD-induced neuronal death (Fig. 3a).
In addition, we hypothesized that conditioned medium from astrocytes would promote synaptic remodeling of neurons, and the results of presynaptic and postsynaptic expression assessment were similar to those of neuronal viability assessment.As shown in Figure 3e and f, the expression levels of synaptotagmin 1 (SYT1) and synaptophysin after OGD/R onset were signi cantly promoted in the normal astrocyte conditioned medium (ACM)-and 21ACM-treated neurons compared with control neurons, but rhFGF21-activated astrocytes exhibited upregulated postsynaptic density 95 (PSD95) expression (Fig. 3d).In addition, treatment with rhFGF21 alone was as effective as treatment with 21ACM in upregulating synaptic protein expression after OGD/R (Fig. 3d-f).All these results proved the bene cial effects of rhFGF21-treated astrocytes on neuronal survival and neuroplasticity.
rhFGF21 actives astrocytes and modulates the astrocytic phenotype The above cellular ndings veri ed that rhFGF21 could attenuate neuronal damage by in uencing astrocytes.However, the mechanism of this effect on astrocytes remained unclear; thus, we examined the effect of rhFGF21 on astrocyte physiology.When astrocytes were treated with rhFGF21 for 24 h, the morphology and cellular viability did not change (Fig. 4a, b).Then, gene sequencing was performed, and the gene expression pro les of the astrocytes in the Norm and rhFGF21 groups were compared.The differentially expressed genes between the Norm and rhFGF21 groups are delineated in a volcano plot (Fig. 4c).A total of 2380 differentially expressed genes were detected by gene sequencing bioinformatic analyses, of which 1350 genes were upregulated and 1030 genes were downregulated in the rhFGF21 group.A Venn diagram was used to visually demonstrate the numbers of expressed genes in the different groups as well as the overlapping expression relationships (Fig. 4d).The overlapping region represents the expressed genes (18448) that were shared between the two groups.In addition, gene expression pattern clustering analysis of genes with signi cant differences was employed to discover commonalities in the expression of different genes (Fig. 4e).

rhFGF21 upregulates astrocyte-induced expression of trophic factors
As Christopherson et al previously described, A2 astrocytes induce synaptogenesis by secreting several trophic proteins [38].After establishing that rhFGF21 altered A2 gene expression in astrocytes, we sought to investigate the changes in key factors released by rhFGF21-activated astrocytes.As shown in the heatmap of the reactive transcripts, the production of trophic factors such as BDNF, transforming growth factor-β (TGF-β), FGF2, and thrombospondin-1 (Thbs1) was markedly increased by rhFGF21 treatment (Fig. 5a).Subsequently, elevated mRNA levels of BDNF, Thbs1 and TGF-β were observed, whereas IGF-1 and vascular endothelial growth factor (VEGF) mRNA expression was not changed (Fig. 5b).Similarly, the protein levels of TGF-β, BDNF and Thbs-1 were upregulated after treatment with rhFGF21 for 24 h, while those of VEGF and IGF-1 were not altered (Fig. 5c, d).Taken together, these data revealed that rhFGF21 guided astrocytes into potentially bene cial phenotypes that are expected to accelerate neurobehavioral recovery by producing neurotrophic factors.
rhFGF21 modulates the astrocytic phenotype and upregulates trophic factors in vivo The above evidence suggested that rhFGF21 promoted astrocytes activation thereby contributed to the protection of neurons from OGD injury.Therefore, we next sought to explore whether rhFGF21 was associated with astrocytic phenotype shifts in an animal model of ischemic stroke.When the mice were subjected to MCAO/reperfusion, rhFGF21 were administered intraperitoneal after the arterial occlusion, and then RT-PCR was used to analyze the astrocytic phenotype gene expression change in ipsilateral brain cortex at 7 d after ischemia.We found that several A1 and A2 phenotype genes (Psmb8, Clcf1, Tgm1 and Cd109) expression increased after MCAO/reperfusion (Fig. 6a).Meaningfully, there was a signi cant increase in the expression of A2 markers (Clcf1, Tgm1, Pptgs2, Cd109 and S100α10) after rhFGF21 treatment, while rhFGF21 also induced Serping1, Rt1-s3 and Psmb8, prototypic A2 genes (Fig. 6a).
One of the central roles of A2 astrocytes-derived BDNF and Thbs1 in the ischemic brain is to protect brain cells via promoting synaptogenesis [39].After establishing that rhFGF21 can alter gene expression in astrocytes, we next examined rhFGF21 could enhance the production of BDNF and Thbs1 in the brain.As shown in Fig 6 b and c, the protein levels of BDNF and Thbs1 were signi cantly higher in the rhFGF21 group than those in the Sham and MCAO groups at 7 d after ischemia injury (Fig. 6b, c).

Discussion
FGF21, which has been recognized as a prime lipid catabolic factor, plays an important role in regulating energy balance [20].In recent years, it has been found that FGF21 could be an early biomarker for dysregulated mitochondrial dynamics and endoplasmic reticulum (ER) stress conditions and that it induces glial cells to promote remyelination in response to CNS disorders [40,41].Following cerebral ischemia/reperfusion injury in rats, endogenous FGF21 production by neurons in the penumbra is effectively increased [37].Consistent with this information, prolonged increases in the mRNA and protein levels of FGF21 in ischemic cortices were observed from days 1 to 7 after MCAO/reperfusion in the current study, suggesting an emergency response to brain ischemia.At 7 d after MCAO/reperfusion, FGF21-positive puncta were mainly localized in Iba1-positive microglial cells in vivo, and in vitro, microglia, but not neurons or astrocytes, signi cantly expressed FGF21 in response to OGD/R, identifying microglia as the sources of FGF21 in ischemic brain tissue.These data further support the suggestion by Salminen et al that endogenous FGF21 is released by different cell types after different degrees of brain damage to alleviate related tissue injury [42].
It has been reported that FGF21 reduces caspase-3 activity and cell death by activating the PI3K/Akt/Bcl-2-dependent cell death signaling cascade [43].For example, Ye et al reported that exogenous FGF21 treatment can inhibit neuronal apoptosis by regulating the PI3K/Akt signaling pathway to ameliorate brain injury [33].In addition, administration of exogenous FGF21 downregulates ER stress-related protein expression by activating the ERK1/2, PI3K/Akt, and JNK signaling pathways, thus protecting neurons from ischemic injury [44].Consistent with these ndings, we have demonstrated a role for endogenous FGF21 in brain tissue recovery and functional outcome improvement after stroke.A greater extent of neuronal tissue loss was observed in FGF21 KO mice than in WT mice at 14 d after MCAO/reperfusion, but TTC staining showed no signi cant difference in infarct volume between FGF21 KO mice and WT mice at the early stage after surgery.Furthermore, FGF21 de ciency signi cantly prolonged motor de cits and results in degeneration of long-term behavioral performance after ischemia.On the other hand, FGF21 has been reported to ameliorate cognitive functioning in rats after hypoxic/ischemic brain injury [33].However, the effects of endogenous FGF21 on cognitive performance need further exploration in a future study.The present data strongly suggest that FGF21 stably promotes long-term neurological recovery; thus, targeting of FGF21 is a promising therapeutic strategy after stroke.
In response to brain damage, microglia are capable of producing neurotrophic and anti-in ammatory cytokines, such as IL-10, IL-4, IGF-1 and TGF-β, that help inhibit in ammation and promote tissue repair during the late stage [45].Here, we have demonstrated that microglial-derived FGF21 is a bene cial signal that delays neuronal death and reduces brain atrophy over the long term.Additionally, TNF, IL-1α, and IL-6 released by lipopolysaccharide (LPS)-activated microglia are together su cient to induce the A1 phenotype in astrocytes [46,47].However, studies on A2 phenotype modulation in astrocytes are still rarely reported.We hypothesized that FGF21 from activated microglia participates in astrocytic phenotype modulation.Consistent with this hypothesis, we observed an A2 phenotype shifts after rhFGF21 treatment in the current study, and we found that the levels of A1-like markers such as Serping1, Rt1-s3, and Psmb8 were decreased.Thus, our data clearly show that rhFGF21 induces the A2 phenotype, and the observed reduction in the A1 phenotype strongly supports the contribution of rhFGF21 to astrocyte modulation in astrocyte cultures and ischemic brain.Furthermore, A2 astrocytes are now known to play bene cial roles in neuronal survival under pathological conditions [48,13].After brain damage, A2 astrocytes promote synapse formation, support blood-brain barrier function, and release nutritional factors such as BDNF, Thbs-1, and IGF-1, which have been suggested to be protective in models of stroke [49].BDNF produced by rhFGF21-activated astrocytes is well known to play extremely important roles in the CNS, contributing to neuronal survival as well as the maintenance of homeostatic neuroplasticity [50].
Astrocytic release of high Thbs-1 levels can enhance synaptic and dendritic plasticity [51].Our results con rm these ndings, as conditioned media from astrocytes containing these trophic factors supported the survival of neurons and helped with neuroplasticity.
Our current study focused on the effects of rhFGF21 on astrocyte phenotype under normal or pathological conditions.However, in the context of CNS disease, activated microglia not only exhibit a protective phenotype but also produce the in ammatory cytokines IL-6, IL-1β and TNF-α, the collective effects of FGF21 and circulating factors on astrocyte modulation warrant further investigation [5].

Conclusions
In summary, our study has illustrated, for the rst time, that FGF21 is mainly produced by microglia in response to ischemia.Under conditions simulating ischemia, FGF21 attenuated brain tissue loss and stabilized behavioral outcomes during the recovery phase.Subsequently, FGF21 also mediated the astrocytic phenotype and upregulated astrocytic expression of neurotrophic factors.Finally, FGF21 stimulated trophic factor production via astrocyte modulation, promoting neuroprotection and neuroplasticity in OGD/R-induced neurons.These ndings will contribute to the use of FGF21 as a promising agent for stroke treatment.Proposed framework for the neuronal protective effects of FGF21 against ischemic injury.FGF21 from injured microglia ameliorates ischemic injury, and rhFGF21-activated astrocytes modulate pro-recovery phenotypes that are neuroprotective and promote neuroplasticity.

Figure 3 Effects
Figure 3