Upon in vitro exposure to MMF, LPS-activated microglia downregulate expression of inflammatory molecules and upregulate expression of alternative activation markers
As a pre-requisite to in vitro investigation of MMF effect on microglia, we evaluated the potential cytotoxicity of MMF, the bioactive metabolite of DMF, on N9 cells by measuring viability and proliferation upon 24-h exposure to increasing concentrations of MMF. MTT assay results show that at the selected range (1–100 μM) MMF was not toxic for N9 cells (Supplementary Fig. S1a). To select the optimal effective concentration of MMF, we analyzed the mRNA expression by MMF-treated and untreated LPS-activated N9 cells of genes representative of classical (tumor necrosis factor α, Tnf; nitric oxide synthase inducible, Nos2) and alternative activation (chemokine (C-X3-C motif) receptor 1, Cx3cr1; nuclear receptor related 1 protein, Nr4a2) phenotypes. We observed that, while Tnf and Nos2 were downregulated at all MMF concentrations tested, the lowest one, 1 μM, was the most effective in reducing the expression of Cx3cr1 and Nr4a2 to basal levels (Supplementary Fig. S1b); the effect of MMF on microglia activation was therefore tested at this concentration in all subsequent in vitro studies.
The effect of MMF on expression of molecules associated with a pro-inflammatory microglial phenotype is shown on Fig. 1a. Exposure to MMF led to a full reversal to basal expression levels or strong downregulation of the genes coding for the inflammatory cytokines, Tnf and interleukin-1β (Il1b), the myeloid transcription factor, 31 kDa-transforming protein/SFFV proviral integration 1 protein (Spi1, also known as PU.1), which is critical for the viability and effector functions of microglia, and the stress response molecules heme oxygenase-1 (Hmox1) and Nos2, all of which were strongly upregulated by N9 cells upon LPS activation (Fig. 1a). In parallel, the expression of chemokine (C-X3-C motif) receptor 1 (Cx3cr1) and CD200 receptor (Cd200r), receptors essential for the maintenance of interactions with ligands expressed by neurons and whose disruption results in highly activated neurotoxic microglia [8, 61], as well as of that of Nr4a2, an orphan nuclear receptor which functions to inhibit expression of pro-inflammatory neurotoxic mediators in both microglia and astrocytes [42], was reversed or increased over basal levels upon exposure of LPS-activated microglia to MMF (Fig. 1b). We subsequently analyzed the effect of MMF on microglial expression of other genes coding for markers associated with an alternative activation phenotype, insulin growth factor 1 (Igf1) [52], arginase 1 (Arg1) [62], and resistin-like alpha found in inflammatory zone (Retnla, also known as Fizz1) [35], and confirmed the effect of MMF in modulating the classically activated LPS-induced microglial phenotype to that of an alternatively activated phenotype, with reversion to basal levels and/or high enhancement of the expression of these factors (Fig. 1b).
These molecular data therefore strongly support our hypothesis that MMF acts on activated microglia by inducing a switch from a classically activated, pro-inflammatory to an alternatively activated, potentially neuroprotective phenotype.
MMF induces functional modifications commensurate with an alternative activation phenotype in LPS-activated microglia in vitro
To evaluate if the molecular changes observed in activated microglia upon exposure to MMF correlates with relevant functional modifications, we analyzed phagocytic function and intracellular Ca2+ concentration ([Ca2+]i), which are linked to executive microglia function such as release of pro- and anti-inflammatory cytokines, nitric oxide or trophic factors [7]. In LPS-activated N9 cells, exposure to MMF further enhanced the phagocytosis in vitro of fluorescent microbeads (Fig. 2a, b). Most importantly, this increased phagocytic activity was associated with an increase in the expression of triggering receptor expressed by myeloid cells-2 (Trem2) mRNA, which codes for a molecule involved in the clearance of apoptotic neurons and myelin debris by microglia in the absence of inflammation [21, 50], and which was downregulated upon microglial activation (Fig. 2c).
Fluorometric determination of [Ca2+]i in N9 cells incubated with the fluorescent calcium indicator Fura 2 AM demonstrated an increased [Ca2+]i upon activation with LPS, and this increase was further enhanced when the activated cells were treated with MMF (Fig. 2d).
These data indicate that the molecular phenotypic changes induced by MMF are accompanied by functional changes that are consistent with the phenotype of microglia involved in the maintenance of CNS homeostasis such as phagocytosis in the absence of inflammatory responses.
The anti-inflammatory action of MMF is mediated through signaling via HCAR2
While the mechanisms of action of DMF and of its metabolite MMF are only partially understood, MMF was recently shown to be a potent agonist of the hydroxycarboxylic acid receptor, HCAR2 [20, 51]. We therefore hypothesized that the activity of MMF on microglia may involve binding to, and activation of, this receptor. While HCAR2 was shown to be expressed by a large number of cells, including immune cells [12, 20, 29, 58], its expression in microglia has not been directly demonstrated at the protein level. We therefore ascertained its expression in murine microglia-derived N9 cells by confocal microscopy analysis upon co-staining with anti-HCAR2 (aHCAR2) and anti-CD11b antibodies. As can be seen in Fig. 3a, HCAR2 is expressed by N9 cells.
To determine if the modulating effect of MMF on microglia is mediated through binding to HCAR2, we blocked the MMF/HCAR2 interaction using aHCAR2, as specific antagonists for HCAR2 are not available, and tested the expression of genes coding for molecules associated with the classical activation phenotype, Tnf, Il1b, and Hmox1, as well as molecules associated with the alternative activation phenotype, Nr4a2, Cx3cr1, and Cd200r. Antibody blockade of HCAR2 fully reversed the effect of MMF on the mRNA expression of inflammatory and oxidative response molecules, Tnf, Il1b, and Hmox1, as well as that of Nr4a2 in activated microglia (Fig. 3b). In contrast, exposure to aHCAR2 had no effect on the upregulation of Cx3cr1 and Cd200r expression in MMF-treated activated microglia (Fig. 3b).
We noted that the expression of many of the tested genes modulated by MMF exposure were dependent on NF-κB activation, including Tnf, Il1b, Sp1, Hmox1, Nos2, and Nr4a2 [6, 25, 26, 31, 33, 37, 44, 60]. Of these gene transcripts, four that were also tested for their expression in LPS-activated MMF-treated microglia upon blockade of HCAR2, Tnf, Il1b, Hmox1, and Nr4a2 showed reversed expression levels under this condition, whereas the expression of non-NF-κB-dependent Cx3cr1 and Cd200r was unaltered (Fig. 3b); suggesting that the NF-κB pathway is implicated in the modulation of the inflammatory phenotype of microglia by MMF activation of HCAR2 signaling.
MMF modulates microglia activation phenotype through inhibition of the NF-κB pathway via the AMPK/Sirt1 axis
Based on the above demonstration that MMF induced an increase in [Ca2+]i and resulted in the inhibition of NF-κB-dependent gene expression, together with previous studies showing that activation of HCAR2 resulted in increases in [Ca2+]i [3, 20], we hypothesized that MMF was initiating an HCAR2-dependent downstream signaling pathway activated by the increase in intracellular Ca2+ levels (Fig. 4). Thus, activation of HCAR2 upon MMF binding would lead to activation of the Gi-type G protein signaling cascade; this, in turn, would result in phospholipase C activation, possibly through release of the βγ subunit [4] and, thereby, in increased [Ca2+]i. The increase in [Ca2+]i would lead to activation of AMPK through phosphorylation of threonine 172 by calcium/calmodulin-dependent protein kinase 2 (CaMKK2) [22] whose activity is enhanced by [Ca2+]i [1]. It has been shown that the NAD+-dependent protein deacetylase sirtuin-1 (SIRT 1) suppresses NF-κB by direct deacetylation at lysine 310 of NF-κB p65 [57]. As its deacetylase activity is regulated by NAD+ availability [38]; we postulated that AMPK activation would lead to an increase in NAD+ generated through phospho-AMPK (p-AMPK)-mediated induction of nicotinamide phosphoribosyltransferase (NAMPT), the rate-limiting enzyme in the NAD biosynthesis pathway [14], and thereby in activation of SIRT1 resulting in inhibition of NF-κB signaling [23, 32].
As the first step towards validating this possible pathway, we demonstrated that the increased [Ca2+]i is indeed a result of MMF binding to HCAR2, as blocking the receptor with aHCAR2 prevented the MMF-induced [Ca2+]i flux observed in Fura 2AM-treated LPS-activated N9 cells upon exposure to MMF (Fig. 5a).
We then quantified p-AMPK by Western blotting to assess AMPK activation as the first downstream component of the pathway and observed a decrease in the proportion of p-AMPK in LPS-activated N9 cells. In the presence of MMF, however, there was a clear increase in p-AMPK, and this increase was inhibited upon blocking HCAR2 (Fig. 5b).
The ensuing step of this pathway implicates an activation of NAMPT that converts NAM to NAD+, essential for the activation of SIRT1, and we measured NAD+ concentration in LPS-activated N9 cells exposed, or not, to MMF and upon antibody blockade of HCAR2. As can be seen in Fig. 5c, the expected increase in NAD+ upon exposure to MMF and its reversal in the presence of aHCAR2 support an activation of NAMPT in the HCAR2 pathway triggered by MMF.
To validate the last step of the pathway, which implicates the activation of the NAD+-dependent SIRT1 leading to deacetylation of NF-κB and therefore its inhibition, we measured the outcome of the enzymatic reaction, that is we monitored the acetylation status of NF-κB in N9 cells, upon various treatments aimed at assessing the effect not only of blocking MMF binding to HCAR2 with aHCAR2, but also of inhibiting SIRT1 itself with the SIRT1-specific inhibitor, EX527, or activating it directly with a specific activator, Resveratrol (Resv) (Fig. 5d). Densitometric quantification of Western blots demonstrated that the increase in Ac-NF-κB upon LPS activation was reversed by MMF and that this effect was abrogated in presence of aHCAR2, suggesting that SIRT1 activation is implicated in MMF effect via HCAR2 signaling. Along those lines, exposure of LPS-activated MMF-treated N9 cells to an inhibitor of SIRT1, EX527, had a similar effect as HCAR2 blockade (Fig. 5d). As a control to our study and to support the implication of SIRT1 in the effect of MMF, we exposed LPS-activated N9 cells to Resv and observed a reduction in Ac-NF-κB similar to that seen upon exposure to MMF (Fig. 5d).
Altogether, these data support the model in which the anti-inflammatory effects of MMF on microglia are mediated through activation of the AMPK/SIRT1 axis, resulting in inhibition of NF-κB, which controls the expression of multiple inflammatory cytokines and mediators.
DMF treatment leads to an increased expression of markers of alternatively activated microglia in CNS of EAE mice
To ascertain whether or not the effects of MMF that we had observed in vitro are recapitulated in vivo, we tested the therapeutic effect of its precursor, DMF, in C57Bl/6J mice with chronic EAE induced with MOG peptide. Mice were treated by daily oral gavage with 150 mg/kg body weight DMF suspended in hydroxypropyl methylcellulose (mean MMF plasma concentration ± SEM: 49.81 ± 9.57 μg/ml at 30 min after dosing; data not shown) from the day following the onset of clinical symptoms until day 30 after immunization when the chronic disease phase is well established. As seen in Fig. 6a, we confirmed the reported beneficial effect of DMF on MOG-induced EAE [28]. Thus, a significantly lower clinical severity was observed in EAE-affected mice treated with DMF from day 16, that is 4 days after the start of treatment, which remained significant until day 22, i.e., throughout the peak/early chronic phase (Fig. 6a, left panel); statistical analysis of the area under the curve (AUC) from days 15 to 30 post-immunization showed significant difference between DMF-treated and untreated mice (Fig. 6a, right panel; P < 0.05).
To determine the effect of DMF on the brain environment in EAE mice, we analyzed the mRNA expression of markers for inflammatory molecules, Tnf, Il1b, and Nos2, as well as for markers of alternative activation, Arg1, Retnla, mannose receptor C type 1 (Mrc1; also known as Cd206) and lectin galactose binding, soluble 3 (Lgals3; also known as Gal3) in brain from three DMF-treated (DMF) mice and three untreated (Vehicle) mice (Fig. 6b), at relevant time points corresponding to acute (day 15), post-peak/early chronic (day 22), and chronic (day 30) phases. While there was no significant difference in brain Tnf expression between treated and untreated mice at any time point tested, the expression of Il1b was significantly downregulated in DMF-treated mice at day 15; similarly, albeit not significant, there was a trend towards lower expression of Nos2 at the same time point. Markers of alternatively activated phenotype were significantly increased in brain of DMF-treated EAE-affected mice in the early phase of disease (Fig. 6b). By day 30, at the established chronic phase when a significant difference in disease severity was no longer detected between DMF-treated and untreated mice, the expression levels of all these markers in DMF-treated mice had returned to those of untreated mice (Fig. 6b).
These results confirm that oral administration of DMF to mice with EAE results in clinical amelioration and can induce an increase in the expression of markers of an alternatively activated phenotype, as observed in microglia exposed in vitro to MMF.
Validation of DMF effect at the synaptic level: DMF protects glutamatergic synapses
In EAE, neuroinflammation enhances glutamate transmission and promotes synaptopathy, which occurs in the early phase of disease and is associated with the release of inflammatory cytokines, such as TNFα and IL-1β, from activated microglia [9, 30]. To understand whether or not DMF might affect EAE-mediated synaptopathy, we have performed whole-cell patch clamp electrophysiological recordings from single neurons in corticostriatal slices from EAE mice, treated or not with DMF, and control mice, between 20 and 25 days post-immunization (post-peak/early chronic phase of disease). As in our previous studies [9, 40, 41], we observed pre- and post-synaptic abnormalities of glutamatergic transmission in EAE mice (Fig. 7). Thus, both frequency and kinetic properties of glutamate-mediated sEPSCs were altered in EAE mice, as compared to that in control mice, with a longer decay time accounting for increased sEPSC duration. DMF treatment normalized sEPSC frequencies (P < 0.05 as compared to vehicle-gavaged EAE-affected mice; Fig. 7a), but had no effect on the increase in sEPSC decay time and half-width (Fig. 7b, c, f). Neither rise time nor amplitude of sEPSCs was altered by EAE induction, as previously shown [9], or by DMF treatment (Fig. 7d, e).
These data might indicate that DMF directly alters sEPSCs in EAE mice by modulating basal glutamatergic transmission at central synapses. Thus, we tested the effect of MMF, the active metabolite of DMF, on spontaneous synaptic transmission. Upon exposure of corticostriatal slices from control mice to MMF, we observed a direct effect on neuronal synaptic activity. Thus, MMF at the low concentration (1 μM) significantly reduced sEPSC frequency (P < 0.01), but not amplitude (100.8 ± 1.6 vs. 100 % pre-treatment values; data not shown) in all the tested control neurons (Fig. 7g), indicating that MMF directly modulates glutamatergic transmission at the pre-synaptic level.
Alternatively, DMF might also alter sEPSCs in EAE mice by an indirect immunomodulatory mechanism through its effect on microglia. As the exposure of striatal slices from control mice to activated microglia results in altered glutamate transmission such as seen in EAE, an effect that can be attributed to TNFα released from the activated microglia [9], we directly tested the effects of MMF on microglia action at synapses. While non-activated N9 cells had no effect on the physiological properties of striatal sEPSCs recorded from control slices (Fig. 8a–c), exposure of control slices to LPS-activated N9 cells significantly increased the duration of sEPSCs (P < 0.001; Fig. 8b, d), by slowing their decay phases (P < 0.001; Fig. 8c, d). We hypothesized that the lack of protective effects of DMF on sEPSC duration in EAE mice could be related to a low central exposure of MMF after oral gavage (dose-limiting toxicity prevents testing higher chronic doses). Indeed, treatment of activated microglia with a high concentration of MMF (100 μM) fully prevented the increase of sEPSC duration, whereas their treatment with a low concentration (1 μM) had no such effect (Fig. 8b–d).
These findings support the concept that the neuroprotective effect of DMF is exerted on neurons directly at pre-synaptic terminals by modulating glutamate release, and indirectly at post-synaptic level, by modulating microglia function in a dose-dependent manner.