SIRT1 Mediates Melatonin’s Effects on Microglial Activation in Hypoxia: In Vitro and In Vivo Evidence

Melatonin exerts direct neuroprotection against cerebral hypoxic damage, but the mechanisms of its action on microglia have been less characterized. Using both in vitro and in vivo models of hypoxia, we here focused on the role played by silent mating type information regulation 2 homolog 1 (SIRT1) in melatonin’s effects on microglia. Viability of rat primary microglia or microglial BV2 cells and SH-SY5Y neurons was significantly reduced after chemical hypoxia with CoCl2 (250 μM for 24 h). Melatonin (1 μM) significantly attenuated CoCl2 toxicity on microglia, an effect prevented by selective SIRT1 inhibitor EX527 (5 μM) and AMP-activated protein kinase (AMPK) inhibitor BML-275 (2 μM). CoCl2 did not modify SIRT1 expression, but prevented nuclear localization, while melatonin appeared to restore it. CoCl2 induced nuclear localization of hypoxia-inducible factor-1α (HIF-1α) and nuclear factor-kappa B (NF-kB), an effect contrasted by melatonin in an EX527-dependent fashion. Treatment of microglia with melatonin attenuated potentiation of neurotoxicity. Common carotid occlusion was performed in p7 rats, followed by intraperitoneal injection of melatonin (10 mg/kg). After 24 h, the number of Iba1+ microglia in the hippocampus of hypoxic rats was significantly increased, an effect not prevented by melatonin. At this time, SIRT1 was only detectable in the amoeboid, Iba1+ microglial population selectively localized in the corpus callosum. In these cells, nuclear localization of SIRT1 was significantly lower in hypoxic animals, an effect prevented by melatonin. NF-kB showed an opposite expression pattern, where nuclear localization in Iba1+ cells was significantly higher in hypoxic, but not in melatonin-treated animals. Our findings provide new evidence for a direct effect of melatonin on hypoxic microglia through SIRT1, which appears as a potential pharmacological target against hypoxic-derived neuronal damage.


Introduction
Melatonin (5-methoxy-N-acetyltryptamine) is an endogenous neurohormone produced primarily by the pineal gland and mainly involved in the regulation of circadian rhythms. Aside from its classical Table 1. Primers used for real-time polymerase chain reaction (PCR) amplification.

Immunohistochemistry and Cell Count
At P8, intracardiac perfusion was performed as described previously [40]. Animals were anesthetized with intraperitoneal administration of ketamine 40 mg/kg and xylazine 4 mg/kg, and perfused intracardially with 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4. Brains were removed and post-fixed in the same fixative solution for 2 h at RT and then immersed overnight at 4 • C in 0.1 M phosphate buffer, pH 7.4. Then, the brains were dehydrated through an ascending ethanol series and separately included in paraffin wax.
The number of Iba1 immunoreactive cells in different brain areas was determined in two slides/brain and at least eight counting frames per animal. For each photo, cells were counted in two squares of 200 µm by 200 µm. The number of Iba1+ cells with nuclear SIRT1 localization in the CC was determined in two slides/brain and at least six counting frames per animal.

Statistical Analysis
All in vitro data were from at least three independent experiments run at least in triplicate. For primary microglia, experiments were from separate isolations. In vivo experiments were carried out on at least three animals/group. All experimental values are presented as the mean ± SEM. Statistical analyses were performed, as appropriate, by Student's t-test or one-way ANOVA followed by Neuman-Keuls post-hoc test using GraphPad Prism Software (GraphPad Software, San Diego, CA, USA). P < 0.05 was the criterion for statistical significance. CoCl 2 toxicity was initially tested on differentiated neuronal-like cells SH-SY5Y. A concentrationresponse curve was carried out by the MTT assay, testing effects at early (5 h) and late (24 h) time points (Figure 1a). The results confirmed the time and concentration-dependent neurotoxic effects of CoCl 2 , which appeared significant starting at 250 µM. Based on these data, this concentration was used in all subsequent experiments. The MTT assay on both primary rat microglial cells ( Figure 1b) and BV2 cells (Figure 1c) showed significant reduction of survival after 24 h of exposure to CoCl 2 . CoCl 2 toxicity was confirmed by trypan blue exclusion assay both on SH-SY5Y cells (viable cells: CTR 84.71% ± 1.08 vs. CoCl 2 58.38% ± 3.39) and BV2 cells (viable cells: CTR 96± 0.65 vs. CoCl 2 69.47 ± 3.74). To explore whether and how microglia affected neuronal susceptibility to hypoxia-induced injury, neuronal-like cells were co-cultured with primary microglia plated on transwell inserts. In the presence of microglia, CoCl 2 resulted slightly but significantly in more toxicity, as revealed by the MTT assay ( Figure 1d). A similar effect was observed when neuronal-like cells were exposed to conditioned medium (CM) from BV2 cells pulsed with CoCl 2 for 3 h, followed by an 18 h-recovery to obtain CM devoid of residual CoCl 2 (Figure 1e). CoCl2 toxicity was initially tested on differentiated neuronal-like cells SH-SY5Y. A concentration-response curve was carried out by the MTT assay, testing effects at early (5 h) and late (24 h) time points (Figure 1a). The results confirmed the time and concentration-dependent neurotoxic effects of CoCl2, which appeared significant starting at 250 μM. Based on these data, this concentration was used in all subsequent experiments. The MTT assay on both primary rat microglial cells ( Figure 1b) and BV2 cells (Figure 1c) showed significant reduction of survival after 24 h of exposure to CoCl2. CoCl2 toxicity was confirmed by trypan blue exclusion assay both on SH-SY5Y cells (viable cells: CTR 84.71% ± 1.08 vs. CoCl2 58.38% ± 3.39) and BV2 cells (viable cells: CTR 96± 0.65 vs. CoCl2 69.47 ± 3.74). To explore whether and how microglia affected neuronal susceptibility to hypoxia-induced injury, neuronal-like cells were co-cultured with primary microglia plated on transwell inserts. In the presence of microglia, CoCl2 resulted slightly but significantly in more toxicity, as revealed by the MTT assay ( Figure 1d). A similar effect was observed when neuronal-like cells were exposed to conditioned medium (CM) from BV2 cells pulsed with CoCl2 for 3 h, followed by an 18 h-recovery to obtain CM devoid of residual CoCl2 (Figure 1e). CoCl2 toxicity on SH-SY5Y in coculture with primary microglia (+pMG; d) or treated in conditioned medium (+CM) from CoCl2pulsed BV2 cells (e). Bars are mean ± SEM of at least three independent experiments. *p < 0.05 vs. own zero (a) or vs. respective control (c) and °p < 0.05 vs. other groups by one-way ANOVA followed by Newman-Keuls test for statistical significance.

Melatonin Protects Microglia after Hypoxia via SIRT1 Activation
The expression of the selective MT1 receptor in microglia was tested prior to investigating the effects of melatonin. Western blot analysis was carried out to ascertain the specificity of the antibody used ( Figure 2a  . CoCl 2 toxicity on SH-SY5Y in co-culture with primary microglia (+pMG; d) or treated in conditioned medium (+CM) from CoCl 2 -pulsed BV2 cells (e). Bars are mean ± SEM of at least three independent experiments. * p < 0.05 vs. own zero (a) or vs. respective control (c) and • p < 0.05 vs. other groups by one-way ANOVA followed by Newman-Keuls test for statistical significance.

Melatonin Protects Microglia after Hypoxia via SIRT1 Activation
The expression of the selective MT1 receptor in microglia was tested prior to investigating the effects of melatonin. Western blot analysis was carried out to ascertain the specificity of the antibody used ( Figure 2a  In addition, the effects of melatonin on CoCl2-induced inflammatory challenges were preliminarily determined. To this end, mRNA expression of interleukin (IL)-6, IL-1β, and tumor necrosis factor (TNF)-α was analyzed by real-time quantitative PCR after 5 h of exposure to CoCl2. The results confirmed that CoCl2 increased cytokine transcription (5.23 ± 1.27; 3.64 ± 0.9 and 4.5 ± 1.44fold change vs. C for IL-6, IL-1β, and TNF-α, respectively). These effects were prevented by melatonin (0.76 ± 0.14; 0.44 ± 0.04 and 1.18 ± 0.21-fold change vs. C for IL-6, IL-1β, and TNF-α, respectively). Accordingly, the direct addition of 1 μM melatonin in combination with CoCl2 slightly but significantly increased survival of both primary microglia ( Figure 3a) and BV2 cells (Figure 3b). The involvement of SIRT1 in melatonin's action was addressed in these conditions by selective pharmacological inhibition with EX527 (5 μM), which was shown to prevent melatonin's protective effects in both primary microglia ( Figure 3a) and BV2 cells ( Figure 3b). Additionally, blockade of AMP-activated protein kinase (AMPK) pathway, classically linked to SIRT1 activation, with selective inhibitor BML-275 (2 μM) abolished melatonin's protective effects against CoCl2 (Figure 3b). Given the highly superimposable responses of primary microglia and BV2 cells, the latter were used for all further investigations. In addition, the effects of melatonin on CoCl 2 -induced inflammatory challenges were preliminarily determined. To this end, mRNA expression of interleukin (IL)-6, IL-1β, and tumor necrosis factor (TNF)-α was analyzed by real-time quantitative PCR after 5 h of exposure to CoCl 2 . The results confirmed that CoCl 2 increased cytokine transcription (5.23 ± 1.27; 3.64 ± 0.9 and 4.5 ± 1.44-fold change vs. C for IL-6, IL-1β, and TNF-α, respectively). These effects were prevented by melatonin (0.76 ± 0.14; 0.44 ± 0.04 and 1.18 ± 0.21-fold change vs. C for IL-6, IL-1β, and TNF-α, respectively). Accordingly, the direct addition of 1 µM melatonin in combination with CoCl 2 slightly but significantly increased survival of both primary microglia ( Figure 3a) and BV2 cells ( Figure 3b). The involvement of SIRT1 in melatonin's action was addressed in these conditions by selective pharmacological inhibition with EX527 (5 µM), which was shown to prevent melatonin's protective effects in both primary microglia ( Figure 3a) and BV2 cells ( Figure 3b). Additionally, blockade of AMP-activated protein kinase (AMPK) pathway, classically linked to SIRT1 activation, with selective inhibitor BML-275 (2 µM) abolished melatonin's protective effects against CoCl 2 ( Figure 3b). Given the highly superimposable responses of primary microglia and BV2 cells, the latter were used for all further investigations.
The responses of BV2 cells were then analyzed at an early time point (8 h) with the intention to precede massive cell death. Analysis of SIRT1 expression by Western blot (Figure 4a) showed that SIRT1 protein levels remained unchanged early (8 h) after exposure of BV2 cells to either CoCl 2 alone or in combination with melatonin (1 µM). However, the deacetylase's subcellular localization appeared modified by CoCl 2 , as shown by immunocytochemical imaging (Figure 4b; green). SIRT1 was highly nuclear in control conditions, prevalently cytosolic after exposure to CoCl 2 , and restored into the nucleus by 1 µM melatonin. Notably, such an effect was precluded when SIRT1 was selectively inhibited with EX527 (5 µM). The responses of BV2 cells were then analyzed at an early time point (8 h) with the intention to precede massive cell death. Analysis of SIRT1 expression by Western blot (Figure 4a) showed that SIRT1 protein levels remained unchanged early (8 h) after exposure of BV2 cells to either CoCl2 alone or in combination with melatonin (1 μM). However, the deacetylase's subcellular localization appeared modified by CoCl2, as shown by immunocytochemical imaging (Figure 4b; green). SIRT1 was highly nuclear in control conditions, prevalently cytosolic after exposure to CoCl2, and restored into the nucleus by 1 μM melatonin. Notably, such an effect was precluded when SIRT1 was selectively inhibited with EX527 (5 μM).   The responses of BV2 cells were then analyzed at an early time point (8 h) with the intention to precede massive cell death. Analysis of SIRT1 expression by Western blot (Figure 4a) showed that SIRT1 protein levels remained unchanged early (8 h) after exposure of BV2 cells to either CoCl2 alone or in combination with melatonin (1 μM). However, the deacetylase's subcellular localization appeared modified by CoCl2, as shown by immunocytochemical imaging (Figure 4b; green). SIRT1 was highly nuclear in control conditions, prevalently cytosolic after exposure to CoCl2, and restored into the nucleus by 1 μM melatonin. Notably, such an effect was precluded when SIRT1 was selectively inhibited with EX527 (5 μM). CoCl2 prevents SIRT1 nuclear translocation in microglia. In (a), Western blot analysis of SIRT1 expression in BV2 cells exposed to CoCl2 (250 μM for 8 h) alone, or in combination with either melatonin (1 μM) or melatonin + EX527 (5 μM). In (b), immunostaining of SIRT1 (green) and nuclear counterstaining with DAPI (blue) in BV2 cells treated with CoCl2 alone or in combination with . CoCl 2 prevents SIRT1 nuclear translocation in microglia. In (a), Western blot analysis of SIRT1 expression in BV2 cells exposed to CoCl 2 (250 µM for 8 h) alone, or in combination with either melatonin (1 µM) or melatonin + EX527 (5 µM). In (b), immunostaining of SIRT1 (green) and nuclear counterstaining with DAPI (blue) in BV2 cells treated with CoCl 2 alone or in combination with melatonin (1 µM) or melatonin + EX527 (5 µM). Bars are mean ± SEM of three independent experiments. Representative images are shown. Scale bar = 20 µm.

Melatonin Modulates Microglial Hypoxic and Inflammatory Markers Indirectly Affecting Neurons during Hypoxia
The distinctive hypoxic marker HIF-1α was analyzed by immunostaining in BV2 cells exposed to CoCl 2 for 8 h. Results showed that CoCl 2 increased nuclear localization compared to the control, where Biomolecules 2020, 10, 364 9 of 20 only a faint signal was detectable ( Figure 5). In the presence of melatonin (1 µM), nuclear HIF-1α was reduced, an effect that appeared sensitive to SIRT1 inhibition with EX527 (5 µM; Figure 5). melatonin (1 μM) or melatonin + EX527 (5 μM). Bars are mean ± SEM of three independent experiments. Representative images are shown. Scale bar = 20 μm.

Melatonin Modulates Microglial Hypoxic and Inflammatory Markers Indirectly Affecting Neurons during Hypoxia
The distinctive hypoxic marker HIF-1α was analyzed by immunostaining in BV2 cells exposed to CoCl2 for 8 h. Results showed that CoCl2 increased nuclear localization compared to the control, where only a faint signal was detectable ( Figure 5). In the presence of melatonin (1 μM), nuclear HIF-1α was reduced, an effect that appeared sensitive to SIRT1 inhibition with EX527 (5 μM; Figure 5). Expression of the p65 subunit of the inflammatory marker NF-kB was analyzed by Western blot at the same time point (8 h). CoCl2 induced a significant increase in NF-kB expression compared to control. This effect was attenuated by melatonin but restored with the addition of EX527 (Figure 6a). Consistent with this result, immunostaining nuclear localization of NF-kBp65 was increased by CoCl2 exposure, and such effect prevented by melatonin in an EX527-sensitive manner (Figure 6b). Expression of the p65 subunit of the inflammatory marker NF-kB was analyzed by Western blot at the same time point (8 h). CoCl 2 induced a significant increase in NF-kB expression compared to control. This effect was attenuated by melatonin but restored with the addition of EX527 (Figure 6a). Consistent with this result, immunostaining nuclear localization of NF-kBp65 was increased by CoCl 2 exposure, and such effect prevented by melatonin in an EX527-sensitive manner (Figure 6b).
Finally, we assessed the indirect outcome of melatonin's modulation of hypoxic microglia on neuronal vulnerability to hypoxia. To this end, a conditioned medium (CM) protocol was chosen where BV2 cells were pulsed with CoCl 2 , alone or in combination with drugs, and recovered in fresh medium devoid of drugs and thus enriched only with released factors (see the Materials and Methods section for detailed protocol). Results showed that exposure of neuronal-like cells to CoCl 2 (24 h) in the presence of CM, collected from BV2 microglia pre-pulsed with CoCl 2 (BV2-CM-CoCl 2 ), exacerbated neuronal death at the MTT assay (Figure 7). Such potentiation of neuronal damage was mitigated in the presence of CM from BV2 cells pre-treated with CoCl 2 +melatonin (BV2-CM-CoCl 2 +MEL; Figure 6). Finally, CM derived from BV2 pre-pulsed with CoCl 2 + melatonin + EX527 (BV2-CM-CoCl 2 + MEL + EX) restored full toxicity (Figure 7).     Finally, we assessed the indirect outcome of melatonin's modulation of hypoxic microglia on neuronal vulnerability to hypoxia. To this end, a conditioned medium (CM) protocol was chosen where BV2 cells were pulsed with CoCl2, alone or in combination with drugs, and recovered in fresh medium devoid of drugs and thus enriched only with released factors (see the Materials and Methods section for detailed protocol). Results showed that exposure of neuronal-like cells to CoCl2 (24 h) in the presence of CM, collected from BV2 microglia pre-pulsed with CoCl2 (BV2-CM-CoCl2), exacerbated neuronal death at the MTT assay (Figure 7). Such potentiation of neuronal damage was mitigated in the presence of CM from BV2 cells pre-treated with CoCl2+melatonin (BV2-CM-CoCl2+MEL; Figure 6). Finally, CM derived from BV2 pre-pulsed with CoCl2 + melatonin + EX527 (BV2-CM-CoCl2 + MEL + EX) restored full toxicity (Figure 7).  Melatonin modifies the microglial response to CoCl 2 , reducing its effects on hypoxic neuronal damage. BV2 cells were pulsed with CoCl 2 (250 µM) alone or in combination with melatonin (1 µM) or melatonin+EX527 (5 µM) for 3 h, followed by washing and recovery in fresh medium devoid of any drug. Conditioned medium (BV2-CM) was collected after further 18 h of incubation and transferred to SH-SY5Y during challenge with CoCl 2 (250 µM for 24 h). Neuronal-like cell viability was then tested by the MTT assay. Bars are mean ± SEM of three independent experiments. * p < 0.05 vs. own control, • p < 0.05 vs. CoCl 2 + MEL + EX527. Another group by one-way ANOVA followed by Newman-Keuls test for statistical significance.

Melatonin Differently Affects Microglia in the Cortex and Hippocampus of Hypoxic Rats
Seven-day old rats were subjected to common carotid artery occlusion followed by hypoxia, as described in detail in the Materials and Methods section. Melatonin (10 mg/kg) was injected i.p. after induction of anoxia, and animals sacrificed 24 h later. Expression of selective microglial marker Iba1 was analyzed by Western blot in ipsilateral hippocampal and cortical protein extracts of sham-operated or hypoxic (HI) animals, treated or not with melatonin (MEL). The results showed no effects on Iba1 expression in the cortex (Figure 8a), while a trend of the increase was evident in the hippocampus, although it did not reach statistical significance (Figure 8b). Immunostaining for Iba1 was carried out on whole-brain sections, and the number of microglial cells was determined by counting positive cells in the ipsilateral areas of interest. Microglia in the hippocampus appeared highly branched, and their number increased significantly in hypoxic animals compared to sham-operated groups (Figure 8c). Melatonin did not appear to modify such an increase (Figure 8c). Interestingly, a small population of microglial cells displaying an activated amoeboid phenotype was evident selectively in the CC. Their number did not significantly differ between treatment groups (not shown). Expression of SIRT1 in microglial cells was investigated by double immunostaining with Iba1 on whole brain sections. Microglia in the hippocampus and cortex did not appear to express clearly detectable levels of SIRT1. Remarkably, however, amoeboid microglia in the CC (green, Figure 9) Figure 8. Melatonin does not affect Iba1+ cell number in rats subjected to CCAO. Seven-day-old rats were subjected to sham surgery or to ligation of the right carotid artery followed by hypoxia. Melatonin-treated groups were subsequently injected with melatonin (10 mg/kg) and sacrificed after 24 h. Iba1 expression in ipsilateral cortical (a) and hippocampal (b) protein extracts was evaluated by Western blot analysis. Microglia were labeled by immunohistochemical staining of Iba1 (green, c) with DAPI nuclear counterstaining (blue, c). Double positive Iba1+/DAPI+ cells were counted in the CA1 area of the hippocampus (HC; graph in c). Bars are mean ± SEM of at least three animals/group. * p < 0.05 vs. sham by one-way ANOVA followed by Newman-Keuls test for statistical significance. Representative images of blots and of ipsilateral HC are shown. Scale bar = 100 µm.

Expression of SIRT1 Is Selectively Modified in Amoeboid Microglia of the Corpus Callosum in Hypoxic Rats
Expression of SIRT1 in microglial cells was investigated by double immunostaining with Iba1 on whole brain sections. Microglia in the hippocampus and cortex did not appear to express clearly detectable levels of SIRT1. Remarkably, however, amoeboid microglia in the CC (green, Figure 9) appeared clearly positive for SIRT1 (red; Figure 9). Moreover, SIRT1 localization was prevalently cytoplasmic in the hypoxic group (HI), while it appeared with a more defined nuclear localization in melatonin-exposed animals, both sham and hypoxic (sham + MEL and HI + MEL; Figure 9). This was confirmed by counting the number of amoeboid Iba1+ cells displaying nuclear SIRT1, which was significantly lower in hypoxic animals but restored in the presence of melatonin (graph in Figure 9).
Biomolecules 2020, 10, x FOR PEER REVIEW 13 of 21 Figure 9. Melatonin promotes nuclear SIRT1 localization in amoeboid microglia of the corpus callosum in rats subjected to CCAO. Seven-day-old rats were subjected to sham surgery or to ligation of the right carotid artery followed by hypoxia (HI). Melatonin (MEL)-treated groups were subsequently injected with melatonin (10 mg/kg) and sacrificed after 24 h. Double immunohistochemical staining of Iba1 (green) and SIRT1 (red) with DAPI nuclear counterstaining (blue) is shown. The graph reports the percentage of nuclear SIRT1+, Iba1+ microglial cells in the area. Representative images of ipsilateral CC are shown. Insets show magnification of cells indicated by the arrows. Scale bar = 50 μm. Bars are mean ± SEM of at least three animals/group. *p < 0.05 vs. sham and **p < 0.05 vs. HI by one-way ANOVA followed by Newman-Keuls test for statistical significance.
Immunohistochemical analysis of pNF-kBp65 in the same area showed a defined nuclear localization in amoeboid microglia of hypoxic animals (HI), compared to both sham-operated and hypoxic melatonin-treated (sham and HI + MEL) groups ( Figure 10). The number of NF-kB+/Iba1+ microglia was significantly increased in hypoxic animals compared to the other groups, together with the number of cells where it was localized in the nucleus. Figure 9. Melatonin promotes nuclear SIRT1 localization in amoeboid microglia of the corpus callosum in rats subjected to CCAO. Seven-day-old rats were subjected to sham surgery or to ligation of the right carotid artery followed by hypoxia (HI). Melatonin (MEL)-treated groups were subsequently injected with melatonin (10 mg/kg) and sacrificed after 24 h. Double immunohistochemical staining of Iba1 (green) and SIRT1 (red) with DAPI nuclear counterstaining (blue) is shown. The graph reports the percentage of nuclear SIRT1+, Iba1+ microglial cells in the area. Representative images of ipsilateral CC are shown. Insets show magnification of cells indicated by the arrows. Scale bar = 50 µm. Bars are mean ± SEM of at least three animals/group. * p < 0.05 vs. sham and ** p < 0.05 vs. HI by one-way ANOVA followed by Newman-Keuls test for statistical significance.
Immunohistochemical analysis of pNF-kBp65 in the same area showed a defined nuclear localization in amoeboid microglia of hypoxic animals (HI), compared to both sham-operated and hypoxic melatonin-treated (sham and HI + MEL) groups ( Figure 10). The number of NF-kB+/Iba1+ microglia was significantly increased in hypoxic animals compared to the other groups, together with the number of cells where it was localized in the nucleus. Figure 10. Melatonin antagonises nuclear NF-kB localization in amoeboid microglia of the corpus callosum in rats subjected to CCAO. Seven-day-old rats were subjected to sham surgery or to ligation of the right carotid artery followed by hypoxia (HI) alone or with subsequent injection of melatonin (10 mg/kg; HI + melatonin (MEL). Animals were sacrificed after 24 h. Double immunohistochemical staining of Iba1 (red) and NF-kB (green) with DAPI nuclear counterstaining (blue) is shown. The graph reports the percentage of nuclear SIRT1+, Iba1+ microglial cells in the area. Representative images of ipsilateral CC are shown. Arrows indicate representative Iba+ cells with nuclear NF-kB; arrowheads indicate representative Iba1+ cells with extranuclear NF-kB staining. Asterisks indicate cells reported in the insets at higher magnification. Scale bar = 40 μm. Bars are mean ± SEM of at least three animals/group. *p < 0.05 vs. sham and **p < 0.05 vs. HI by one-way ANOVA followed by Newman-Keuls test for statistical significance.

Discussion
Melatonin is an endogenous hormone characterized by neuroprotective activities, exerted through multiple mechanisms. Its production is regulated by light-dark cycles, with peaks of release during the dark period and suppression by light [1]. Melatonin has been largely shown to protect against hypoxic injury with the reduction of the infarct area [5,41,42]. Noteworthy, increased mortality following hypoxia-ischemia has been reported in MT1 receptor knockout mice [17]. Melatonin's ability to easily cross the blood-brain barrier and its safety profile make it an ideal candidate, especially for pediatric use against long term neurological deficits that develop following prenatal/perinatal hypoxia. Hypoxic-ischemic encephalopathy is, in fact, a still frequent condition responsible for neonatal morbidity and mortality [26]. Brain or whole-body hypothermia is currently the only therapeutic option offering amelioration in the prognosis, as from preclinical and clinical data [26,[43][44][45]. In addition, evidence of a benefit from adjunctive therapy with melatonin has been described [46,47], and a recent study on melatonin safety, pharmacokinetics (PK), and dosage in the neonatal population reported positive results [25]. Finally, it is worth noting that a significant increase of melatonin was described following hypoxia both in experimental models and in human stroke, suggesting even the existence of an endogenous protective response on its part [48]. The detailed comprehension of melatonin's mechanisms of action is hence of paramount importance. So far, studies on the protective activity of the hormone in hypoxia have mainly focused on direct neuronal effects, whereas evidence regarding melatonin's involvement in the modulation of inflammatory Figure 10. Melatonin antagonises nuclear NF-kB localization in amoeboid microglia of the corpus callosum in rats subjected to CCAO. Seven-day-old rats were subjected to sham surgery or to ligation of the right carotid artery followed by hypoxia (HI) alone or with subsequent injection of melatonin (10 mg/kg; HI + melatonin (MEL). Animals were sacrificed after 24 h. Double immunohistochemical staining of Iba1 (red) and NF-kB (green) with DAPI nuclear counterstaining (blue) is shown. The graph reports the percentage of nuclear SIRT1+, Iba1+ microglial cells in the area. Representative images of ipsilateral CC are shown. Arrows indicate representative Iba+ cells with nuclear NF-kB; arrowheads indicate representative Iba1+ cells with extranuclear NF-kB staining. Asterisks indicate cells reported in the insets at higher magnification. Scale bar = 40 µm. Bars are mean ± SEM of at least three animals/group. * p < 0.05 vs. sham and ** p < 0.05 vs. HI by one-way ANOVA followed by Newman-Keuls test for statistical significance.

Discussion
Melatonin is an endogenous hormone characterized by neuroprotective activities, exerted through multiple mechanisms. Its production is regulated by light-dark cycles, with peaks of release during the dark period and suppression by light [1]. Melatonin has been largely shown to protect against hypoxic injury with the reduction of the infarct area [5,41,42]. Noteworthy, increased mortality following hypoxia-ischemia has been reported in MT1 receptor knockout mice [17]. Melatonin's ability to easily cross the blood-brain barrier and its safety profile make it an ideal candidate, especially for pediatric use against long term neurological deficits that develop following prenatal/perinatal hypoxia. Hypoxic-ischemic encephalopathy is, in fact, a still frequent condition responsible for neonatal morbidity and mortality [26]. Brain or whole-body hypothermia is currently the only therapeutic option offering amelioration in the prognosis, as from preclinical and clinical data [26,[43][44][45]. In addition, evidence of a benefit from adjunctive therapy with melatonin has been described [46,47], and a recent study on melatonin safety, pharmacokinetics (PK), and dosage in the neonatal population reported positive results [25]. Finally, it is worth noting that a significant increase of melatonin was described following hypoxia both in experimental models and in human stroke, suggesting even the existence of an endogenous protective response on its part [48]. The detailed comprehension of melatonin's mechanisms of action is hence of paramount importance. So far, studies on the protective activity of the hormone in hypoxia have mainly focused on direct neuronal effects, whereas evidence regarding melatonin's involvement in the modulation of inflammatory responses [49], as also shown by our data, prompted us to turn our attention to the effects on microglia that indirectly influence neurons. Microglia are, in fact early responders to hypoxic insults and can deeply affect neuronal vulnerability [50]. Furthermore, we aimed to explore the intriguing possibility that SIRT1, a deacetylase endowed with manifold protective actions and extensively studied in neurons [22], could be a mediator of melatonin's effects on microglia against hypoxic insults.
Melatonin has been reported to act both on selective G-protein coupled receptors and as an antioxidant. This latter function can be both direct, due to melatonin's scavenger properties targeted in particular to mitochondria, or indirect through modulation of the expression of antioxidant enzymes [51][52][53]. In addition, melatonin has been shown to directly interact with target proteins modulating their function [4,54]. In our work, we first confirmed the expression of melatonin receptor MT1 in both primary and immortalized microglia. MT1 is coupled to Gi and Gq proteins, and its activation inhibits the formation of cyclic AMP, activates AMP-activated protein kinase (AMPK) signaling, and inhibits phospho-CREB and protein-kinase A signaling [55][56][57]. This isoform, in particular, was here chosen because of its known involvement in the neuroprotective actions of melatonin, as from experimental models of newborn hypoxic-ischemic brain injury [17]. In addition, MT1 was previously shown to mediate activation of SIRT1 in other tissues, such as the liver [58].
In the present study, we used a previously established in vitro model of chemical hypoxia with CoCl 2 , which stabilizes the α subunit of the transcription factor HIF-1, dampening its degradation [59]. The main cellular responses between cobalt and low oxygen-induced hypoxia have been reported to be significantly similar [60]. HIF-1 upregulation takes place strictly during hypoxia to activate selective response genes [61][62][63][64]. Interestingly, a cell-type-specific role has been suggested for HIF-1, whose upregulation in glial cells appears to trigger detrimental effects [65][66][67][68][69]. In agreement, in our model, nuclear HIF-1 was highly expressed in microglia following CoCl 2 exposure, and microglial viability was significantly reduced. Previous in vitro studies have shown microglial cell death in CoCl 2 -induced hypoxic conditions [70,71]. In line with this negative role for HIF-1, its myeloid-specific knock-out in mice has been shown to reduce neuronal and microglial death following hypoxia, lowering inflammation, and improving behavioral recovery [72]. Likewise, the knockdown of HIF-1 was sufficient to elicit an anti-inflammatory effect in BV2 cells exposed to CoCl 2 [72].
Melatonin was directly protective against microglial damage, and this effect proved to be dependent on SIRT1 and its downstream AMPK pathway [73][74][75], as shown by the use of selective pharmacological inhibitors. Interestingly, we did not detect any variation in SIRT1 levels, but rather found a shift between its cytoplasmic or nuclear localization in different conditions. While the detrimental effects exercised by hypoxia coincided with a cytoplasmic localization, melatonin promoted SIRT1 presence in the nucleus, both under basal and hypoxic conditions. Hence, the involvement of SIRT1 in our conditions is supported by its activation state, as here determined by analysis of its localization, together with the effects of pharmacological inhibition. Consistent with this notion, the subcellular localization of SIRT1 actually accounts for its differential actions: in particular, nuclear SIRT1 is endowed with protective anti-inflammatory action [76], while cytoplasm-localized SIRT1 was shown to enhance apoptosis in different cancer cell lines [77,78]. Despite previous reports indicating that SIRT1 levels declined following in vivo hypoxia and were restored by melatonin, this was observed selectively in neurons [21,79,80].
In order to determine if improvement of microglial survival induced by melatonin/SIRT1 was coupled with a direct action on the microglial inflammatory phenotype, we analyzed the effects of treatment on expression/localization of transcription factors HIF-1α and NF-kB. As previously described for HIF-1, NF-kB was also selectively detected in the nuclei in hypoxic conditions, consistent with its activation and polarization of microglia to sustain an incipient inflammatory reaction [81]. It is important to consider that both HIF-1 [82] and NF-kB [83] are direct targets of SIRT1, which keeps them inactivated by deacetylation. The antagonistic cross-talk between SIRT1 and both NF-kB and HIF-1 has been demonstrated to play a role in inflammation and energy metabolism; perturbations in this signaling can lead to chronic inflammation [84]. In light of this, the opposite compartmentalization we found for HIF-1 and NF-kB on one side and SIRT1, on the other, is perfectly in line with their predicted state of activation and downstream effects.
In order to further assess the role of microglia as a target of melatonin action, we extended our investigation to an in vivo model of hypoxia. The permanent ligation of the common carotid artery represents a validated and widely used model of perinatal hypoxia. Postnatal 7-day rats were chosen since, at this time, the developmental stage is histologically similar to that of a 32-to 34-week gestation human fetus or newborn infant [85]. Aiming to analyze the effects of melatonin early during the development of cell damage, we chose a 24 h time-point previously shown to correspond to the initial phases of microglial activation, preceding full neuronal damage [86]. Among the different areas examined for a microglial response, only in the CA1 area of the hippocampus, we detected an increase in the number of microglia, an effect that, however, was not modified by melatonin. Iba1-positive cells of the hippocampus in our conditions did not show the typical amoeboid phenotype of activated microglia, but rather displayed long thin processes and did not express detectable levels of SIRT1. Modulation of hippocampal SIRT1 by melatonin has been previously described, but it was confined to the neuronal cell population [79]. On the contrary, intense staining for SIRT1 was evident in the population of amoeboid microglia of CC, characterized by round cell bodies devoid of ramifications. Even more remarkably, the enzyme was nuclear in melatonin-treated, but not in hypoxic animals. Amoeboid microglia of the CC have been previously described as a transient population of active microglia physiologically involved in the early developmental stages of periventricular white matter [87]. Our finding in the model of hypoxia appears even more compelling in light of the high vulnerability to hypoxic injury described, especially for this area, in the newborn [7,88]. Amoeboid microglia, in particular, has in fact been shown to be implicated in the early (present at 3 h following injury) inflammatory response induced by hypoxia in this area of the neonatal brain, leading to periventricular white matter and neuronal damage as wells as disruption of the immature blood-brain-barrier [88][89][90][91]. On these bases, we focused our attention on this microglial population to confirm the ability of melatonin to affect the inflammatory response as observed in the in vitro model. Accordingly, the prevalently nuclear localization of NF-kB in the hypoxic group was indicative of a state of inflammation, a condition that was counteracted by melatonin. Our data are in agreement with the suggested antagonistic crosstalk between NF-kB and SIRT1 [84] and suggest that the fine balance between the two factors may mediate the response of CC amoeboid microglia to melatonin to contrast the hypoxia-induced inflammatory phenotype. In line with this, recent evidence from a model of perinatal hypoxia suggested that microglial HIF-1, through miRNA-210, targets SIRT1, reducing NF-kB deacetylation, and rescuing its inflammatory activity [92].

Conclusions
In conclusion, we have shown in vitro that microglial SIRT1 plays a major role in response to melatonin following hypoxic injury and appears able to dampen the neuroinflammatory event. In vivo, at the early time point examined, this occured in a specific microglia subpopulation localized in the CC that is selectively vulnerable to hypoxia early during brain development. These findings overall confirm microglia as an additional target for the action of melatonin in hypoxia with a mechanism involving SIRT1.