Carbon Monoxide Modulation of Microglia-Neuron Communication: Anti-Neuroinflammatory and Neurotrophic Role

Microglia, the ‘resident immunocompetent cells’ of the central nervous system (CNS), are key players in innate immunity, synaptic refinement and homeostasis. Dysfunctional microglia contribute heavily to creating a toxic inflammatory milieu, a driving factor in the pathophysiology of several CNS disorders. Therefore, strategies to modulate the microglial function are required to tackle exacerbated tissue inflammation. Carbon monoxide (CO), an endogenous gaseous molecule produced by the degradation of haem, has anti-inflammatory, anti-apoptotic, and pro-homeostatic and cytoprotective roles, among others. ALF-826A, a novel molybdenum-based CO-releasing molecule, was used for the assessment of neuron-microglia remote communication. Primary cultures of rat microglia and neurons, or the BV-2 microglial and CAD neuronal murine cell lines, were used to study the microglia-neuron interaction. An approach based on microglial-derived conditioned media in neuronal culture was applied. Medium derived from CO-treated microglia provided indirect neuroprotection against inflammation by limiting the lipopolysaccharide (LPS)-induced expression of reactivity markers (CD11b), the production of reactive oxygen species (ROS) and the secretion of inflammatory factors (TNF-α, nitrites). This consequently prevented neuronal cell death and maintained neuronal morphology. In contrast, in the absence of inflammatory stimulus, conditioned media from CO-treated microglia improved neuronal morphological complexity, which is an indirect manner of assessing neuronal function. Likewise, the microglial medium also prevented neuronal cell death induced by pro-oxidant tert-Butyl hydroperoxide (t-BHP). ALF-826 treatment reinforced microglia secretion of Interleukin-10 (IL-10) and adenosine, mediators that may protect against t-BHP stress in this remote communication model. Chemical inhibition of the adenosine receptors A2A and A1 reverted the CO-derived neuroprotective effect, further highlighting a role for CO in regulating neuron-microglia communication via purinergic signalling. Our findings indicate that CO has a modulatory role on microglia-to-neuron communication, promoting neuroprotection in a non-cell autonomous manner. CO enhances the microglial release of neurotrophic factors and blocks exacerbated microglial inflammation. CO improvement of microglial neurotrophism under non-inflammatory conditions is here described for the first time.


Background
Microglia are the main immune cell population in the central nervous system (CNS) [1]. Originating from a primitive myeloid precursor [2], microglia populate the brain and spinal cord, where they account for 5-10% of total glial population with crucial biological functions [1,3]. As the brain first line of defence, microglia are sentinels that constantly screen the brain parenchyma. When encountering potentially pathogenic or toxic agents that endanger the CNS integrity, microglia initiate an inflammatory response [1,3]. This response involves major functional and morphological changes, namely alteration of microglia expression of surface receptors, intracellular enzymes and secreted molecules [1]. Release of inflammatory mediators, such as reactive oxygen species (ROS) [4,5] and nitric oxide (NO) derivatives [3,6,7], cytokines [3,6,7], metabolites and extracellular vesicles [8] induce cell death on compromised cellular populations and/or potentiate the microglial phagocytosis of bacteria, viruses and debris [1,9,10]. Additionally, microglia recruit other cells and help set up the adaptive immune response [11]. Conversely, microglia also have a relevant role in neurotrophic support [12], as well as brain development, removing immature and apoptotic populations [13]. Microglia also participate in the regulation of synaptic plasticity and neuronal activity [14][15][16][17], and stimulate synaptogenesis and synaptic maturation [18,19]. Several of these functions are reliant on microglia paracrine mechanisms, through secretion of specific factors like thrombospondin, insulin growth factor 1 (IGF-1) [12,[20][21][22][23][24], brain-derived neurotrophic factor (BDNF), adenosine, interleukin-10 (IL-10) as well as low levels of inflammatory cytokines TNF-α (tumour necrosis factor α) and IL-1β [25][26][27][28]. Being both the hub of neuroinflammation and crucial for neuronal development, survival and activity, microglia function is under tight regulation. Deficient microglial function causes imbalances in brain tissue homeostasis: by affecting synaptic plasticity and the overall neuronal function [1] and/or by triggering exacerbated inflammation, which is a common feature of both chronic and acute CNS disorders [29,30].
Known for its toxicity, carbon monoxide (CO) is a gaseous molecule, endogenously produced by the degradation of haem catalysed by haem oxygenase (HO) [31]. This reaction yields CO, along with free iron (Fe 2+ ) and biliverdin [31,32]. HO has two described isoforms: the constitutive form HO-2 and HO-1, an inducible isozyme, whose expression is activated by a wide array of stress signals [32]. CO is a signalling molecule, and the CO/HO-1 axis is protective in several biological contexts, modulating inflammation, apoptosis, cell differentiation and metabolism [33][34][35] CO-derived ROS production primes the anti-inflammatory player peroxisome proliferator-activated receptor gamma co-activator 1 α (PPAR-γ) in macrophages [36] and lung [37], decreasing downstream pro-inflammatory mediators. CO also regulates the activities of mitogen-activated protein kinase (MAPK), resulting in lower secretion of key cytokines, which limits pro-inflammatory stimulation [38][39][40]. In CNS, CO has been described as neuroprotective by preventing neuronal cell death in vitro [41,42] and in ischaemia and reperfusion in vivo models [43,44]. In glial cells, CO is cytoprotective in astrocytes, namely CO inhibited apoptosis by regulating mitochondrial membrane permeabilization [45], by improving oxidative metabolism [46] and by modulating P2X7 receptors [47]. Likewise, vasodilation is promoted by CO via astrocytic glutamate receptors [48]. In microglia, CO reduces inflammatory response induced by LPS (lipopolysaccharide), thrombin or interferon γ (IFN-γ) [38,49]. This anti-inflammatory effect occurs by chemically inhibiting PI3K (phosphoinositide 3-kinase) an ERK (extracellular-signal-regulated kinase) [38,49]. CO also decreases neuroinflammation in microglia via modulation of cell metabolism, namely by enhancing mitochondrial respiration [50]. In vivo models of CNS disorders have highlighted the potential of CO as a regulator of local brain immunity. In fact, CO treatment ameliorates tissue inflammation and limits cell loss and disease progression in models for autoimmune encephalomyelitis [51] and haemorrhagic stroke [52].
Herein, we aimed to assess how CO alters microglia function and paracrine communication with the surrounding cellular milieu, particularly with neurons. In fact, CO's role on microglia-neuron direct communication has not been an object of study to the best of our knowledge. Thus, the present work targets the underlying molecular mechanisms of this intercellular communication primarily by the analysis of neuroinflammatory and neurotrophic factors in microglial secretome. Then, it is explored how secretome from CO-treated microglia affects neuronal survival, morphology and function.
The therapeutic advent for CO is supported by the increasing number of applications [32] and can be clinically implemented by the development of CO-releasing molecules (CORMs) to circumvent several limitations regarding CO's pharmacological delivery, namely safety and tissue specificity [53][54][55]. Herein, a novel CORM, ALF-826, is used to assess the impact of CO on microglial neuroinflammation and neurotrophism, and on microglia-neuron communication. ALF-826 is a molybdenum carbonyl complex which spontaneously releases CO in biological media, and it has been recently demonstrated that ALF-826 is anti-inflammatory in BV2 microglial cells when applied after LPS challenge [56]. Moreover, molybdenum is a trace element that is essential for the function of at least three human enzymes: xanthine oxidase, sulphite oxidase and aldehyde oxidase.

Materials
All cell culture plastics were purchased from Corning (Corning, NY) and Sarstedt (Sarstedt, Germany). Chemicals of analytical grade were obtained from Sigma-Aldrich (St Louis, MI), unless stated otherwise. All used antibodies and their respective technique are described in Supplementary Table 1.

Primary Cell Cultures
Animals were housed in standard laboratory conditions with free access to water and standard rodent chow. All animal experiments were performed with the approval of the i3S Animal Ethics Committee and in accordance with European Union Directive on the protection of animals used for scientific purposes (2010/63/EU) and the Portuguese Law that transposes it (Decreto-Lei n.º 113/2013). All efforts were made to minimise animal suffering. Neuron hippocampal cell cultures were isolated from embryonic day 18 Wistar rats. Pregnant females were euthanized in a CO 2 chamber; their abdominal region was pulverised with 70% ethanol and opened using surgical scissor and tweezers. Embryos were retrieved, decapitated, and the brains were placed in a Petri dish containing Hank's Balanced Salt Solution (HBSS, Thermo-Fisher Scientific) supplemented with 100 µg/mL gentamicin (Lonza). The brains were dissected under a magnifying glass, and the hippocampi were collected. Hippocampal tissue was incubated for 15 min at 37 °C with 1.5 mg/mL of trypsin (Thermo-Fisher Scientific), and after allowing the hippocampi to sediment, the supernatant was removed and a 10% FBS (Thermo-Fisher Scientific) solution was added. Following gentle agitation, the supernatant was discarded, and a washing step was performed with fresh HBSS. The supernatant was again removed, and neurobasal medium (Thermo-Fisher Scientific) supplemented with 25 µM glutamate (Sigma-Aldrich), 0.5 mM L-glutamine (Thermo-Fisher Scientific), 100 µg/mL gentamicin (Lonza) and 1% B27 supplement (Thermo-Fisher Scientific) was added, and the tissue was dissociated by thorough pipetting. The suspension was then put through a 70-µm cell strainer and collected, and cells were counted with a Neubauer counting chamber. Neurons were plated on 24 multi-well plates (100,000 cells/well), on top of poly-D-lysine (PDL, 1 mg/mL, Sigma-Aldrich) coated glass coverslips. Cells grew in vitro for 10 days before any analysis was carried forward, with fresh medium being added after7 days in culture.
Mixed glial cell (MGC) cultures were obtained from cortices of 2-day old Wistar pups. Animals were quickly sacrificed, with brains being removed and dissected consequently in HBSS (Thermo-Fisher Scientific) with penicillin (100 units/mL) and streptomycin (100 μg/mL) (Thermo-Fisher Scientific). White matter and meninges were discarded, and the remaining tissue was collected and mechanically homogenised. The cell suspension was passed through a 25-gauge syringe and incubated with 0.1 U/mL DNAse I (Zymo) and 1.5 mg/mL trypsin (Thermo-Fisher Scientific) for 15 min at 37 °C. After incubating, trypsin was inactivated by adding DMEM GlutaMAX (Thermo-Fisher Scientific) supplemented with penicillin (100 units/ mL), streptomycin (100 μg/mL) (Thermo-Fisher Scientific) and 10% FBS (Thermo-Fisher Scientific), and the suspension was centrifuged afterwards (550 g, 15 min). The resulting supernatant was carefully discarded, and the pellet was resuspended in the same medium and put through a 100µm cell strainer. The obtained suspension was distributed equally into 75-cm 2 t-flasks (2 brain per flask) to make a total media of 10 mL per flask. Prior to adding the cells, the t-flasks were coated with 1 mg/mL PDL (Sigma-Aldrich) for 1 h. The cell medium was changed every other day.
Primary microglia cells were obtained from 10-day-old MGC cultures, which were placed on an orbital shaker incubator for 2 h at 200 rpm. The cell suspension was collected and centrifuged (260 rcf, 10 min), and the supernatant was discarded. The pellet was resuspended in DMEM/F12 (Thermo-Fisher Scientific), and cells were counted using a Neubauer counting chamber. Microglia cells were always obtained from either 10 (shake 1) or 17 days in vitro (DIV, shake 2) MGC cultures. Older cell cultures were not considered for this. Microglial cells were plated on 24 multi-well plates, with DMEM/F-12 GlutaMAX (Thermo-Fisher Scientific) and 4% L-glutamine (Sigma-Aldrich), penicillin (10 units/mL) and streptomycin (10 μg/mL) (Thermo-Fisher Scientific) and 10% FBS (Thermo-Fisher Scientific) supplement.
For neurotrophic assay, BV2 cells (10 × 10 4 cells per well) were treated for 48 h with ALF-826 (50 µM), then medium was collected and centrifuged (5 min, 500 g). The conditioned medium (microglial culture supernatant) was added into 48 h-differentiated CAD neuronal culture. t-BHP at 7.5 and 10 µM was added into the CAD neuronal culture to promote cell death for 24 h, with or without the presence of BV2 microglia-conditioned medium, when mentioned (Fig. 7A).
To assess the effect of purinergic signalling on neuron-microglia communication, 48 h-differentiated CAD neurons were treated, simultaneously, with microglia-conditioned media and adenosine receptor antagonists SCH-58261 (1 and 5 µM) and Dipropylcyclopentylxanthine (DPCPX, 25 and 50 µM). Primary culture of microglia derived from 10 or 17 DIV MGC cultures was collected and plated on 24 multi-wells (7 × 10 5 cells/well). Twenty-four hours after seeding, cells were treated with 50 µM of ALF-826 for 24 h then with 10 ng/mL of LPS for another 24 h. Microglia medium was collected and centrifuged at 500 g for 5 min, and its supernatant was used as conditioned medium. Fifty percent volume of the neuronal medium of 10 DIV hippocampal neurons cultured in 24 multiwell plates was replaced by microglia-conditioned media. Neuronal cells were analysed following 24 h.

Reagents and Solution Preparation
ALF-826 was provided by Proterris (Portugal) Lda. Stock solutions of 2.5 mM were prepared by dissolving the compound in Dimethyl sulfoxide (DMSO, Sigma-Aldrich) and diluting it 1/10 in a NaHCO 3 solution (0.1 mM, pH 8.3). ALF-826 was then filtered, aliquoted and stored at − 80 °C at a final concentration of 2.5 mM. ALF-826 has a half-life of 37 min in Hepes 7.4 buffer. Its cytotoxicity is negligible at 100 µM (98% survival of RAW246.7 macrophages at 24 h), and its CO effective load is ca. 3 equivalents of CO per mol (information provided by Proterris (Portugal) Lda.).
For the preparation of the depleted form of ALF-826 (iALF-826), the compound was dissolved in DMSO, left for 24 h at room temperature to allow for the dissociation of the CO groups from the molecular scaffold and subsequently diluted in 1/10 in a NaHCO 3 solution (0.1 mM, pH 8.3). iALF-826 was then filtered, aliquoted and stored at − 80 °C at a final concentration of 2.5 mM.
Fresh CO gas solutions of CO were prepared by saturating PBS, bubbling 100% of CO compressed gas (Linde, Germany) for 30 min, yielding a 10 −3 M stock solution. CO concentration in solution was measured spectrophotometrically, by the conversion of deoxymyoglobin to carbon monoxymyoglobin, in an assay previously described by Motterlini et al. [53].

Neuronal Cell Viability Assay
CAD neurons were collected by trypsinisation and stained with 1 ng/mL of propidium iodide (PI, Thermo-Fisher Scientific) for 15 min at 37 °C. Cell viability was analysed by flow cytometry using the FACS Canto II (BD Biosciences). A 488-nm laser line was used for excitation, and PI was read in the FL-3 channel in the linear scale. Appropriate controls, such as positive staining controls and unstained samples, were always carried out. Data analysis was performed with the FlowJo software (BD Biosciences).

Microglia Reactivity Assay
BV2 cells were collected by scrapping and stained with 500 ng/mL of PE/Cy7 (Phycoerythrin/Cyanine 7) antimouse CD11b antibody (BioLegend) for 15 min at 37 °C. Microglia reactivity was analysed by flow cytometry using FACS Canto II (BD Biosciences). A 488-nm laser line was used for excitation, and reactivity levels were determined by reading sample's PE/Cy7 Median Fluorescence Intensity in the FL-4 channel in the linear scale. Appropriate controls, such as positive staining controls and unstained samples, were always carried out. Subsequent data analysis was performed using FlowJo software (BD Biosciences).

Enzyme-Linked Immunosorbent Assay
Enzyme-linked immunosorbent assay (ELISA) was performed to measure tumour necrosis factor α (TNF-α), interleukin-1 β (IL-1β) and IL-10 microglia supernatant levels, using the respective Standard ABTS ELISA Development Kits (PeproTech). For glial cell-derived neurotrophic factor (GDNF) and brain-derived neurotrophic factor (BDNF), different kits were used: (GDNF ELISA Kit (Abcam) and Human/Mouse BDNF DuoSet ELISA (R&D)). Microglia culture media was collected and centrifuged to remove cellular debris, and the pellet was discarded. The resulting supernatant was kept at − 80 °C until analysis. All experiments were performed in accordance with the respective manufacturer's instructions. Absorbance values were measured at 415 nm, with wavelength correction set at 560 nm, using an Infinite F200 PRO microplate reader (Tecan).

Griess Reaction Assay
Nitrite of microglia supernatant was quantified by Griess reaction colorimetric test. Microglia culture media was collected and centrifuged to remove cellular debris and subsequently incubated with Griess reagent (1:1 ratio, Sigma-Aldrich) for 10 min at room temperature (RT), protected from light. Absorbance was measured at 540 nm using an Infinite F200 PRO microplate reader (Tecan). Nitrite concentration was calculated with reference to a standard curve generated with known concentrations of sodium nitrite (NaNO 2 , Sigma-Aldrich).
Equal amounts of protein were separated by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) on a polyacrylamide gel (10% gel unless stated otherwise), with a NZYColour Protein Marker II (NZYtech) being used as band size reference. Proteins were then electrically transferred onto an Amersham Protran 0.45 NC nitrocellulose membrane (GE LifeSciences) and blocked with 5% (m/v) BSA in T-TBS for 1 h. Membranes were labelled consecutively with primary and secondary antibodies as indicated in Supplementary Table 1. Immunoblots were exposed to ECL Clarity Western Detection Reagent (Bio-Rad) 5 min, and the reactive bands were detected after the membranes were exposed to X-ray film (Chemidoc Touch Imaging System, Bio-Rad). The resulting area and intensity of the bands were quantified with ImageLab software (Bio-Rad). β-actin was used as internal loading control unless otherwise is stated.

ROS Generation Assay
ROS, in particular hydrogen peroxide generation, were measured following the conversion of 2′,7′-dichlorofluorescein diacetate (H2DCFDA) (Invitrogen) to fluorescent 2′,7′-dichlorofluorescein (DCF). BV2 microglia cells were seeded ( 9 × 10 3 cells/well) on 96-well black flat bottom plates and treated with ALF-826 and/or LPS, as described previously. After treatment, microglia supernatant was removed, and cells were washed twice with ice-cold PBS and treated for 15 min with 5 μM of H2DCFDA. Cell extracts were subsequently washed, and fluorescence intensity was measured using a Tecan Infinite F200 PRO microplate reader (λ ex 485 nm/λ em 530 nm). ROS generation was calculated as an increase over baseline levels, determined for untreated cells and normalised to total protein quantification for each condition.

Neuronal Morphology Analysis
Neurite tracing was done using the NeuronJ Fiji plug-in (Erik Meijering) and following instructions present in an online manual provided by the developer (https:// image scien ce. org/ meije ring/ softw are/ neuro nj/ manual/). Neurites were labelled and clustered as primary (originating from cell soma), secondary (from a primary neurite) and tertiary (from a secondary neurite) processes. Software calculation determined total number of neurites per field as well as individual neurite length. Similarly, neuron skeleton reconstruction and Sholl analysis (quantification of intersections at concentric spheres originating from cell soma to 150 µm, with radius every 5 µm) were performed using the Simple Neurite Tracer Fiji plug-in, in accordance to instructions present in online tutorials (https:// imagej. net/ Simple_ Neuri te_ Tracer). For each individual experiment, four to six images were analysed for each glass slide.

High-Performance Liquid Chromatography
To quantify microglial adenosine, supernatant from BV2 cells was collected from each well, deproteinized in PCA 3 M and centrifuged at 4 °C, 10 min and 13,000 g. Then the supernatants neutralised (pH 6.8-7.2) with KOH/Tris (Sigma-Aldrich) as previously described by Conde and Monteiro (2004). Adenosine was quantified by HPLC with UV detection at 254 nm. The HPLC system consisted of an LC 10-AD pump, SIL-20AC autosampler, SPD-20 A/ AV UV-VIS wavelength detector and Class VP software to analyse the chromatograms (Shimadzu, Kyoto, Japan). The analytical column was a Lichrospher 100 RP-18 (125 4 mm, i.d., particle size 5 μm, Merck, Madrid, Spain) protected by LichroCART 4-4 guard-columns (Merck, Madrid, Spain). The mobile phase was 100 mm KH2PO4 with 15% methanol, pH 6.5, run at a flux of 1.75 ml min −1 . External standards were prepared under the same conditions as the biological samples and adenosine identification, and quantification was made against the standards. Adenosine supernatant levels were normalised to total protein microglial content.

Statistical Analysis
Results are presented as mean ± standard error of the mean (SEM), with at least three biological replicates being performed for all experiments. All statistical analyses were performed using the Prism 6.0 software (GraphPad). Kolmogorov-Smirnov test was used to check data normality, and subsequent parametric or non-parametric tests were used to assess data statistical significance.

Conditioned Medium from CO-Treated Microglia Provides Neuroprotection to Neuronal Culture Against Cell Death
In order to assess the role of carbon monoxide (CO) on remote neuron-microglia communication, a conditioned medium approach was used, which allows to evaluate the impact of microglia secretome on neuronal function and survival. BV2 microglia were pre-treated with 50 µM ALF-826 for 24 h and then incubated or not with 500 ng/mL of LPS (lipopolysaccharide), a TLR-4 (toll-like receptor) agonist, classically used to elicit an inflammatory-like response. After 1 day, microglia medium was collected and added into the catecholaminergic neuronal cell line CAD culture for 24 h, challenging cell survival (Fig. 1A). Neuronal cell survival assessment was based on membrane integrity (propidium iodide internalisation) by flow cytometry (Fig. 1B). Viability of CAD neuronal cells exposed to conditioned media derived from inflammatory microglia decreased, and this effect was partially prevented whenever BV2 microglia were pre-treated with ALF-826 (Fig. 1B). In addition, expression of neuronal cleaved caspase-3 (Fig. 1C), which is a classical biomarker of apoptosis, was quantified by western blot. Accordingly, supernatant derived from CO-treated microglia also tends to decrease the levels of cleaved caspase-3 in neurons. These data indicated that CO decreases microglia release of neurotoxic factors, limiting neuronal cell death. Fig. 1 Carbon monoxide (CO) treatment in microglia attenuates neuronal cell death induced by conditioned medium from LPS-activated microglia. BV2 microglia were treated with ALF-826 (50 µM) for 24 h and with LPS (500 ng/mL) for another 24 h. 'Activated' microglia are cells treated with LPS. CAD neuronal culture was subsequently incubated with microglia-conditioned medium (A). Cell viability was assessed by propidium iodide staining quantification in flow cytometry (B) or by cleaved caspase 3 protein expression by western blotting (C). For 1B, n = 6-7, error bars represent mean ± SEM. One-way ANOVA test with Tukey's multiple comparison test was performed (p value = 0.0163, F = 5.38). The densitometry of cleaved caspase-3 (average of fragments 17 and 19 kDa densitometry) expression was normalised for β-actin and is presented as percentage relative to the control. Representative membranes for the experiments are displayed (C). n = 4, error bars represent mean ± SEM. Kruskal-Wallis test with Dunn's multiple comparison was performed (p value = 0.0214, Kruskal-Wallis statistic = 6.006). CO provides neuroprotection in primary hippocampal neurons via modulation of primary microglia. Rat hippocampal neurons were challenged with conditioned medium from primary microglia cells for 24 h and immunostained with βIII-tubulin primary antibody (green) and DAPI (nuclei in blue), scale bar = 25 µm (D). Cell viability was assessed by counting total and apoptotic neurons. White arrows highlight the nucleus of live neurons, and yellow arrowheads highlight the nucleus of dead neurons. The live/total nuclei ratio was registered and is presented in E. n = 9; error bars represent mean ± SEM. One-way ANOVA test with Tukey's multiple comparison test was performed (p value = 0.0011, F = 9.229) ◂ Molecular Neurobiology (2022) 59:872-889 878 The data obtained in BV2 microglia-CAD neuron conditioned media protocol were then validated in a more physiologically relevant model: hippocampal neuronal and microglia rat primary cultures. Primary cultures of microglia were treated with ALF-826 (50 µM, 24 h) and LPS (10 ng/mL, 24 h), and the supernatant was collected, centrifuged and subsequently incubated 1:1 (volume) with neuronal media to challenge 10 days in vitro (DIV) hippocampal neurons. To assess neuronal cell death in primary cultures, neurons were immunolabelled with an antibody against βIII-tubulin (green), the nuclei were stained with DAPI (blue) and cells were imaged using a fluorescence microscope (Fig. 1D). Apoptosis was quantified by counting chromatin-condensed nuclei (Fig. 1E). In accordance with data from Fig. 1B, neurons cultured with inflammatory conditioned media presented a significant decrease in cell survival, which was prevented whenever microglia were pre-treated with ALF-826.

Conditioned Medium from CO-Treated Microglia Improves Neuronal Morphology
Neuronal morphology of differentiated CAD cell line was analysed for indirect evaluation of neuronal function (Fig. 2). Neurons exposed to inflammatory microglia media presented a decrease in cellular complexity parameters, such as average neurite length ( Fig. 2A) and number of neurites per cell (Fig. 2B). Additionally, data from Sholl analysis, which is a quantitative method used for the analysis of neuronal arborization, further indicated the existence of a decrease of the overall neuronal complexity whenever neurons were challenged with inflammatory conditioned media (Fig. 2C, D). Conditioned medium derived from CO-treated microglia prevented the decrease of neurite length and the number of neurites per cell and maintained neurite arborization at similar levels as the non-treated control neurons. Thus, microglial ALF-826 treatment has a paracrine neuroprotective effect on neuronal cells, potentially achieved through modulation of microglial release of neurotoxic proinflammatory factors.

CO Has an Anti-Inflammatory Role on LPS-Challenged BV2 Microglia
Since CO appears to provide neuroprotection by regulating neuron-microglia remote communication, thus secretome factors were evaluated (Fig. 3). The extracellular levels of the pro-inflammatory factors nitric oxide (NO), tumour necrosis factor α (TNF-α) and interleukin-1β (IL-1β) were quantified in microglial supernatant. Griess colorimetric assay, which allows for the quantification of nitrites, an indirect measure of NO, indicated an increase in nitrites concentration following LPS exposure (Fig. 3A). Thus, ALF-826 treatment partially inhibited LPS-induced nitrite release. The decrease of nitrite secretion was dependent on ALF-826 concentration ( Figure S1A). TNF-α was quantified by ELISA; CO also prevented the secretion of TNF-α pro-inflammatory cytokine in LPS BV2 (Fig. 3B).
To confirm that the effect of ALF-826 was due to the CO release and not of the scaffold of the molecule, nitrite levels were also assessed in BV2 cells following treatment with a CO-saturated PBS solution (50 µM). In fact, the saturated solution also reverted NO secretion in LPSstimulated microglia ( Figure S1B). To further confirm the effect of the CO-releasing molecule, LPS-treated microglia were also incubated with either ALF-826 or with the CO depleted form (iALF-826). Here, while ALF-826 inhibited secretion of NO ( Figure S1C) and TNF-α ( Figure S1D), iALF-826 did not produce such effect.
It was next addressed whether CO anti-neuroinflammatory function is present in primary cultures of microglia. Microglial medium was collected for indirect quantification of NO levels by Griess colorimetric assay and of TNF-α and IL-1β levels by ELISA. For NO and IL-1β secreted factors, LPS triggered a significant increase, which was prevented in the presence of ALF-826 pretreatment ( Fig. 3C and E). For TNF-α, ALF-826 decreases its secretion, but not in a statistically different manner (Fig. 3D). Contrary to BV2 cell line (data not shown), in microglial primary cultures, LPS elicited elevated levels of IL-1β, suggesting a strong NLRP3 (NLR Family Pyrin Domain Containing 3) inflammasome reaction. Altogether, these results validate that CO limits exacerbated inflammation, by reducing the levels of secreted pro-inflammatory factors.
To further confirm the anti-inflammatory effect of ALF-826 treatment on BV2 function, we quantified the expression of specific microglial reactivity markers: intracellular ROS production and surface receptor CD11b. The role of CO on intracellular ROS production was evaluated by H(2)DCF-DA staining that increases its fluorescence in the presence of hydrogen peroxide. Likewise, CO limited LPS apparent increase on microglia-derived oxidative stress (Fig. 4A). Finally, expression of surface receptor CD11b was quantified via flow cytometry. CD11b immunoreactivity appears to increase when LPS was administered to microglia and this was reverted by in ALF-826 pre-treatment (Fig. 4B).
Overall, the results suggest that ALF-826 is a strong pharmacological anti-inflammatory agent due to CO release. In fact, ALF-826 limits the expression and secretion of several pro-inflammatory agents. Thus, CO-promoted non-cell autonomous neuroprotective effect might result from limiting the inflammatory response of microglial cells.

Medium-Derived from 'Resting' Microglia Limits Apoptosis in CAD Neuronal Cells Exposed to Pro-Oxidant Stimulus and Improves Neuronal Morphological Complexity.
Other than being the CNS immune guardians, microglia have a crucial role in neuronal basal function, that ranges from modulation of proliferation/differentiation balance during development and trophic support, morphological and synaptic plasticity [19,57,58]. Moreover, to the best of our knowledge, there are no data showing CO's role in the modulation of microglia neurotrophic functions. To test the potential microglia-induced neurotrophic effect and neuroprotection, neuron-microglia-conditioned media approach was used. BV2 microglial cells were treated or not with 50 µM ALF-826 for 48 h, then, microglial supernatant was collected for treating differentiated CAD neurons, which were simultaneously challenged with a pro-oxidant agent (tert-Butyl hydroperoxide, t-BHP) to induce neuronal cell death for 24 h (Fig. 5A). Conditioned medium from microglia and from CO-treated microglia, both reduced levels of neuronal death at 7.5 and 10 µM of t-BHP (Fig. 5B). These results indicate that microglia-conditioned medium, by itself, has a neuroprotective effect against oxidative stress, while CO does not further improve microglia-induced neuroprotection.
As microglia have been described to modulate neuronal network dynamics [13], it was next assessed whether CO affects neuronal morphology under basal conditions. In fact, neuronal morphology is a manner to assess neuronal function, and thus an indirectly manner to evaluate neuroprotection. CAD neurons were incubated with conditioned media from control and CO-treated microglia and stained with anti-βIII-tubulin (green) and with DAPI (nuclei staining) for morphological analysis. Microglial CO treatment modulates neuronal morphology by increasing the number of neurites per cell (Fig. 6A) and neurite length (Fig. 6B).
Overall, CO appears to stimulate microglial neurotrophic function by improving neuronal morphology under basal conditions.

CO Stimulates Microglial Neurotrophic Response
Lastly, the microglia secretome was analysed to disclose which molecular players were involved in CO-mediated microglia neuroprotection and neuronal morphological modulation. Several candidates described as key players in neurotrophism, and neuronal support were assessed: BDNF (brain-derived neurotrophic factor), GDNF (glial cell-derived neurotrophic factor) and the anti-inflammatory cytokine IL-10. In fact, CO-stimulated microglial IL-10 secretion was demonstrated by ELISA (Fig. 7A). However, CO administration did not increase BDNF or GDNF secretion (data not shown). This result could point towards a neurotrophic role of CO. In fact, IL-10 overexpression is not only relevant for resolution of inflammation, but also for regulation of cell survival, neurite dynamics, development and synaptogenesis [59,60]. Moreover, our lab has shown in the past that CO affects astrocyte-to-neuron paracrine communication, promoting neuroprotection via purinergic signalling [61]. Thus, we assessed the hypothesis that CO's regulation of microglianeuron remote communication is partially modulated by a similar mechanism (Fig. 7B). High-performance liquid chromatography (HPLC) showed that CO administration appears to increase microglial adenosine levels in the supernatant (Fig. 7C). Also, the protein levels of surface enzyme CD73, which converts purine mononucleotides into nucleosides, were also increased in CO-treated microglia (Fig. 7D). Hence, CO may reinforce neuron-microglia communication via purinergic mechanisms.

Chemical Inhibition of Microglia-to-Neuron Purinergic Communication Reverts CO's Neuroprotective Properties
To validate whether the CO neuroprotective and morphogenic effect are caused by an increase in levels of adenosine in the microglial supernatant, neurons were treated with A 2A R antagonist SCH-58261 (1 or 5 µM) or with A 1 R antagonist 8-Cyclopentyl-1,3-dipropylxanthine (DPCPX) (25 or 50 µM) and simultaneously incubated with microgliaconditioned media (Fig. 5A). Then, neuronal morphometric parameters and neuronal survival were analysed. Inhibition of adenosine receptors did not have any effect on neuronal survival (data not shown). Nevertheless, the increase in neurite length observed whenever CAD cells were treated with conditioned media from CO-treated microglia was reverted whenever an adenosine receptor antagonist was used (Fig. 8B), indicating that the CO effect is mediated via purinergic communication. Interestingly, this effect was not A 1 or A 2A R specific, as both compounds reverted the COderived effect. Also, treating microglia with the inactivated scaffold form of ALF-826 did not have an effect on either number of neurites per cell or neurite length (Fig. 8).

Discussion
In this study, we have shown that carbon monoxide (CO) modulates microglia function and impacts on microglia-toneuron remote communication. Using a conditioned media protocol, it was demonstrated that carbon monoxide ultimately promotes indirect neuroprotection: (i) by limiting microglial pro-inflammatory reactivity and (ii) by enhancing basal microglial neurotrophism. The non-cell autonomous effect of CO acting on microglia and promoting neuroprotection has never been described before. Furthermore, a novel molybdenum-based CO-releasing molecule: ALF-826 was used herein. This molecule presents low cytotoxicity in BV2 microglia cell line [56] and delivers a large CO load (personal information of Proterris (Portugal) Lda.). The anti-inflammatory feature of CO has been extensively studied in macrophages [40,62]. Likewise, microglia treated with carbon monoxide-releasing molecule-3 (CORM-3) also secreted lower levels of tumour necrosis factor α (TNF-α), interleukin-1β (IL-1β) and nitrites after being triggered with different stimuli (lipopolysaccharide, thrombin and interferon γ) [38,39]. Although there are data demonstrating the anti-neuroinflammatory role of CO in microglia, not much is known regarding direct CO modulation of neuron-microglia communication in an inflammatory setting, and how it affects neuronal survival and function.
Herein, we demonstrated that conditioned media derived from microglia treated with a pro-inflammatory stimulus (LPS) promoted neuronal cell death. Neurons challenged with inflammatory microglial supernatant partially lost morphological complexity. Moreover, microglial pre-treatment with ALF-826 rescued changes in neuronal morphology and partially reverted neuronal death. Neuronal structural degradation is a marker of cell senescence and cell death in general and can be a consequence of exacerbated inflammation [63]. Alterations in neuronal network complexity are a common phenotype of neurodegenerative disorders and play a role in defective synaptic connectivity [64][65][66][67]. It has been shown that exposure of primary embryonic neurons to high levels of exogenous TNF and IL-1β alters cell morphology, affecting total arborization, number and length of dendrites, growth cone dynamics as well as dendritic spine density and maturity [68]. Thus, loss of neuronal morphological complexity can also be partially a consequence of proinflammatory pathways disruption cytoskeleton-signalling mechanisms, which CO could alleviate by inhibiting microglial inflammatory output. These results indicate that LPS-treated microglia drive a noxious response that causes severe neuronal damage, which can be prevented by ALF-826 microglial pre-treatment.
We assessed microglial production of soluble inflammatory factors in the presence and absence of ALF-826 administration. LPS increased secretion of soluble inflammatory factors TNF-α and nitrites, which were attenuated by CO administration in microglia. In contrast, LPS did not promote IL-1β release in BV2 microglia cell line (data not shown).
Despite microglia modulation of inflammation and CNS immunity, these myeloid cells also regulate brain development, synaptogenesis and neuronal turnover and provide strong neurotrophic support to neuronal cells [12, 19-24, 57, 58, 69, 70]. Interestingly, CO modulation of microglial neurotrophic response has never been explored. Herein, we showed that conditioned media derived from non-activated microglia protected neurons from oxidative damage by partially preventing neuronal cell death. Microglia neuroprotection can result from secretion of neurotrophins (BDNF, neurotrophin-3, nerve growth factor) [20,71], inflammatory cytokines [24] or survival growth factors (IGF-1) [12], which impact cell fate. In the present work, CO did not increase microglial production and secretion of BDNF or GDNF, while IL-10 release was stimulated by CO. Microglial IL-10 is a key anti-inflammatory regulator, anti-apoptotic and an important synaptogenic mediator in vivo [72][73][74]. Accordingly, using in vivo acute inflammation model and in vitro LPS-stimulated macrophages, CO promoted IL-10 production [40]. Under physiological conditions without inflammatory stimulation, this is the first work showing CO-induced production of IL-10 in microglia. Furthermore, IL-10 promotes neurite outgrowth and synapse formation in OGD (oxygen-glucose deprivation)-challenged primary cortical neurons [60]. Herein, conditioned medium from CO-treated microglia increased number of neurites per cell and neurite length. Because IL-10 has a neuromorphogenic role [60], one may speculate that IL-10 is involved in neuronal morphology improvement by reinforcing neurotrophism.
Moreover, we unveiled that CO acts as a stimulator of microglia purinergic signalling, which affects neuronal function and survival. Purinergic signalling is important for cellular cross-talk and both microglial environmental sensing and reactivity [75], as well as neuronal survival and function [76]. Activation of adenosine receptor A 2A has been reported to promote axonal elongation and dendritic branching in cortical neurons [77]. A recent breakthrough paper has shown that microglia can sense neuronal ATP, convert it into adenosine, which acts as Fig. 6 Administration of CO in 'resting' microglia increases neuronal morphological complexity after exposure to conditioned medium. CAD neurons treated with BV2 microglia (previously treated or not with ALF-826 at 50 µM) conditioned media for 24 h were labelled with a βIII-tubulin antibody and DAPI, and fluorescence microscopy was performed. Micrographs were analysed, and morphological features, A number of neurites per cell and B neurite length, were assessed from five independent experiments (10-15 cells per condition). C Representative cell pictures for each condition with cells immunostained with βIII-tubulin primary antibody (scale bar = 25 µm) and D representative tracings of 2D acquired images. n = 5; error bars represent mean ± SEM. For 6A, one-way ANOVA test with Tukey's multiple comparison test was performed (p value < 0.0001, F = 22.94), for 6B, Kruskal-Wallis test with Dunn's multiple comparison was performed (p value = 0.0327, Kruskal-Wallis statistic = 6.269) Molecular Neurobiology (2022) 59:872-889 884 an important suppressant of neuronal activity and overall animal behaviour [17]. We have previously shown similar results concerning astrocyte to neuron communication. Namely, CO-treated astrocytes provided neuroprotection and increased neuronal viability in a co-culture model by enhancing release and conversion of ATP into adenosine [61].
Interestingly, previous reports have also shown upregulation of ectonucleotidases following brain ischemia, providing cerebral protection [78]. In microglia, CD39 and CD73 are crucial for microglial ramification, phagocytosis and motility [79]; and in vitro silencing of CD73 in BV2 microglia results in a clear pro-inflammatory profile [80]. Furthermore, CD73 expression is relevant for microglial polarisation and inflammatory attenuation in a spinal cord injury model [80]. This data could indicate that there is a paracrine feedback loop in microglia in which secreted ATP is converted into adenosine and recognised by the self, thus regulating microglial inflammatory activity. Herein, chemical inhibition of neuronal adenosine indicates that microglial Fig. 7 CO stimulates 'resting' microglial secretion of anti-inflammatory factor IL-10 and adenosine. BV2 microglia supernatant (derived from microglia treated or not with ALF-826 at 50 µM) was collected, and secretion levels of IL-10 was quantified by ELISA (A). Microglia communicate with neuronal cells via purinergic signalling (B). For the quantification of adenosine in the microglial media, the supernatant was collected, treated and analysed via HPLC (C). BV2 microglia CD73 protein expression was quantified by western blot-ting (D). CD73 densitometry was normalised for β-actin. Results were normalised to total microglial protein levels. All results are presented as percentage relative to the negative control. n = 4-5; error bars represent mean ± SEM. For 7A, unpaired t test was performed (p value = 0.0377, t = 2,486), for 7C, Mann-Whitney U test was performed (p value = 0.1, Mann-Whitney U = 0.0), for 7D Mann-Whitney test was performed (p value = 0.0286, Mann-Whitney U = 0.0) adenosine has a direct effect on neuronal function, but we cannot exclude that CO also regulates microglial CD73 expression, increasing adenosine levels in the supernatant, priming microglia to release higher levels of other neuroactive compounds. This would be in accordance with existing results, that claim that CD73 activity drives microglia towards an M2 profile by autocrine A 2B agonism. Further experiments are needed to assess this hypothesis.
In summary, neurotrophic factors, anti-oxidant molecules or even anti-apoptotic players can be involved in microglia-induced neuroprotection under non-inflammatory conditions.

Conclusion
Taken together, ALF-826 is a promising new CO releasing molecule that limits microglial reactivity and secretion of inflammatory mediators, and that improves the release of neurotrophic factors. Both effects protect neurons against cell death and improve neuronal morphology. In fact and to the best of our knowledge, this is the first report showing that CO affects basal microglia function, by stimulating neurotrophism. Hence, CO regulates neuron-microglia remote communication and promotes neuroprotection in a non-cell autonomous manner by both limiting inflammation and improving neurotrophism. The novel ALF-826 has also proven to be a useful CORM for the reproducible and non-toxic delivery of CO enabling the study of intercellular processes and with potential for future clinical applications.
Author Contribution NLS designed the study, conducted the experiments, helped with data analysis and evaluation, wrote the manuscript and approved the final version of this manuscript. IP conducted the experiments, helped with data analysis and approved the final version of this manuscript. JB helped with primary cultures and approved the final version of this manuscript. CSFQ designed part of the study, conducted some of the experiments and approved the final version of this manuscript. BFM and SVC helped with HPLC experiments and data analysis and approved the final version of this manuscript. CCR provided the CORM ALF-826 (which is not available commercially), helped with data analysis and approved the final version of this manuscript. TS designed part of the study, helped with data analysis and approved the final version of this manuscript. HLAV designed the study, helped with data analysis and evaluation, wrote the manuscript and approved the final version of this manuscript.
Funding This work was financed by FEDER-Fundo Europeu de Desenvolvimento Regional funds through the COMPETE 2020-Operational Programme for Competitiveness and Internationalization (POCI),

Data Availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on a reasonable request.

Declarations
Ethics Approval and Consent to Participate All animal experiments were performed with the approval of the i3S Animal Ethics Committee and in accordance with European Union Directive on the protection of animals used for scientific purposes (2010/63/EU) and the Portuguese Law that transposes it (Decreto-Lei No. 113/2013). All efforts were made to minimise animal suffering.

Consent for Publication
All authors agreed to the publication of this manuscript.