Abstract
Innate immunity is an ancestral process that can induce pro- and anti-inflammatory states. A major challenge is to characterise the transcriptional cascades that modulate the response to chronic and acute inflammatory challenges. The Drosophila melanogaster Gcm transcription factor represents an interesting candidate for its potential anti-inflammatory role. Here we explore its evolutionary conservation and its mode of action. We found that the murine ortholog Gcm2 (mGcm2) is expressed upon aging, which is considered as a state of chronic inflammation. mGcm2 is found in a subpopulation of microglia, the innate immune cells of the central nervous system (CNS). Its expression is also induced by a lyso-phosphatidylcholine (LPC)-induced CNS demyelination (acute inflammation) and mGcm2 conditional knock out mice show an increased inflammatory phenotype upon aging or LPC injection. In agreement with the role of this transcriptional cascade in inflammation, the human ortholog hGCM2 is expressed in active demyelinating lesions of Multiple Sclerosis (MS) patients. Finally, Drosophila gcm expression is induced upon aging as well as during an acute inflammatory response and its overexpression decreases the inflammatory phenotype. Altogether, our data show that the inducible Gcm pathway is highly conserved from flies up to humans and represents a potential therapeutic anti-inflammatory target in the control of the inflammatory response.
Introduction
The immune response is one of the oldest processes of living organisms. It goes from simple enzymatic reactions in bacteria to cellular and humoral pathways in more complex animals (1, 2). The main requirement for a successful immune response is the recognition of non-self through specific receptors that activate different immune pathways. The majority of both the receptors and their downstream pathways are highly conserved throughout evolution. The immune response is controlled by either pro- or anti-inflammatory cues. The pro-inflammatory JAK/STAT, Toll and NF-κB pathways are found in both insects and mammals (3–7). The TGF-β signalling pathway is associated with promotion of anti-inflammatory properties in vertebrates upon resolution of inflammation, with similar molecules being present in flies (8, 9). One of the most challenging issues is to discover transcription factors that coordinately block the inflammatory response.
Glial cells missing/Glial cell deficiency (Gcm/Glide, Gcm throughout the text) is expressed in the haemocytes of Drosophila melanogaster, functional orthologs of the vertebrate macrophages (10). gcm silencing in the fly macrophages does not on its own produce an overt phenotype, but it enhances the inflammatory phenotype triggered by the constitutive activation of the JAK/STAT pathway (11–13). While gcm is only expressed early and transiently in the haemocytes, its silencing also enhances the response to an acute inflammatory challenge performed well after gcm expression has ceased. Thus, Gcm in flies modulates the acute and chronic inflammatory responses, through mechanisms that are not fully understood.
Gcm is an atypical zinc finger transcription factor that is structurally conserved throughout evolution. The two mammalian orthologs are named hGCM1 and hGCM2 in humans, mGcm1 and mGcm2 in mice. Gcm1 is required for the differentiation of trophoblasts in the developing placenta and its mutation is associated with pre-eclampsia (14). Gcm2 is expressed and required mainly in the parathyroid glands, where it is necessary for the survival and the differentiation of the precursor cells but no role was described in the immune cells (15–17). Since the fly gcm gene, but not its orthologs, is also necessary in glia (18, 19), and since the fly glial cells constitute the immune cells of the nervous system, we speculated that this transcriptional pathway may have an anti-inflammatory role in microglia, the resident macrophages of the vertebrate nervous system. Microglia are not only responsible for the development and the homeostasis of the central nervous system (CNS) but are also involved in neuroinflammation (20–24). Moreover, recent studies suggest that the increased inflammation in the aged brain is attributed, in part, to the resident population of microglia (25). Interestingly, the heightened inflammatory profile of microglia in aging is associated with a ‘sensitised’ or ‘primed’ phenotype, which might be triggered by transcriptional pathways controlling the inflammatory response. This phenotype includes differences in morphology and gene expression of aged versus young microglia (26–30).
Here we show that the murine ortholog mGcm2 starts being expressed in a subset of microglia upon aging. Loss of mGcm2 enhances the aging phenotype in terms of microglia morphology and expression of pro-inflammatory markers, corroborating the hypothesis that this transcription factor has an anti-inflammatory role during chronic inflammation. Furthermore, mGcm2 expression is induced in demyelinated lesions triggered by lysophosphatidylcholine (LPC) injection in the spinal cord and in mGcm2 inducible conditional knock-out animals, the same challenge triggers a much stronger inflammatory reaction. Therefore, mGcm2 is involved in chronic and acute responses. Most importantly, the human ortholog hGCM2 is expressed in active demyelinating lesions of Multiple Sclerosis (MS) patients. Finally, chronic and acute challenges also induce the expression of gcm in flies. De novo expression of gcm counteracts the inflammatory phenotype, explaining its mode of action and highlighting its anti-inflammatory potential. Altogether, our data demonstrate that Gcm constitutes a highly conserved immune transcriptional cascade from flies up to humans and represents a novel potential therapeutic target in the control of the inflammatory response.
Material and Methods
Mouse lines
Cx3cr1-Cre mice were obtained from TAAM Orléans and bred to maintain them on the C57Bl/6J background. The mGcm2flox/flox line was created at the Institute Clinique de la Souris (ICS, Strasbourg) (Figure 1A). Conditional knock-out (cKO) mice with their littermate controls derived from Cx3cr1-Cre/+;mGcm2flox/+ males were crossed with mGcm2flox/flox females. The open field behavioural test was conducted at the ICS.
To generate tamoxifen-inducible conditional knock-out mGcm2 mice specifically in microglia, Cx3cr1CreER/+ knock-in mice (Parkhurst et al., 2013; JAX stock #021160) were crossed with mGcm2flox/flox mice. Cx3cr1CreER/+;mGcm2flox/+ mice were crossed with mGcm2flox/flox animals to generate Cx3cr1CreER/+; mGcm2flox/flox mice (called icKO thereafter) and Cx3cr1CreER/+; mGcm2flox/+ (control). WT and mGcm2flox/flox were also used as controls for demyelinating lesion experiments. All mice were maintained on a normal diet in a 12-hour light/dark cycle. 5 to 10 mice were used per group (per gender, age and genotype).
Genotyping
DNA were extracted according to the Jacks Lab protocol. For primers, we used: CX3Cr1cre GTTCGCAAGAACCTGATGGACA and CTAGAGCCTGTTTTGCACGTTC, mGcm2flox CAATAGGGAAGTGATCCCTAGAGTC and GGGAAACTTGTCTGTTCTTTCACACAG and mGcm2WT CAATAGGGAAGTGATCCCTAGAGTC and GGGAAACTTGTCTGTTCTTTCACACAG, forward and reverse, respectively. Primer sequences for Cx3cr1CreER genotyping were: CX3 F-CTTCTTGCGATTCTTGCAGG; CX3 R -CACTACCTCATCATCCATGA; CX3CreER1- CACGGGGGAGGCAGAGGGTTT; CX3CreER2-GCGGAGCACGGGCCACATTTC.
Tamoxifen treatment
To induce Cre recombination, adult Cx3cr1CreER/+; mGcm2flox/flox and Cx3cr1CreER/+; mGcm2flox/+ mice (10 to 16week-old) were treated with tamoxifen (100mg/kg; Sigma) by intraperitoneal injections, during 5 consecutive days prior LPC-induced demyelination.
LPC-induced demyelination of the mouse spinal cord
C57Bl6/J 12week-old females from Janvier were used for focal spinal cord demyelinated lesions. To evaluate microglial response and oligodendrocyte differentiation, we used tamoxifen treated Cx3cr1CreER/+;mGcm2flox/flox, Cx3cr1CreER/+; mGcm2flox/+, Gcm2flox/flox and WT mice. Twenty minutes before anaesthesia induction with 3% Isoflurane, animals were injected with Buprenorphine (0.1mg/kg). After induction, isoflurane concentration was increased to 2% for the surgical phase. Animals were placed on the stereotaxic frame, and small incision was made at the level of thoracic vertebrae (T8 and T9). Using a Hamilton syringe connected with a glass capillary, 1% lyso-phosphatidylcholine (LPC, sigma) was injected into the dorsal funiculus of the spinal cord. The injection site was marked with an active charcoal, and internal and external sutures were made. After surgery, animals were injected subcutaneously with Buprenorphine during 2 consecutive days and then treated ad libitum with a solution of Buprenorphine in the drinking water.
P1 primary CNS cultures
Postnatal day 1 (P1) cultures were produced as previously described (31). The cultures were kept in the incubator at 37°C and 5 % CO2 for 14 days. The medium was changed at day 1 and day 3. In vitro cultures were fixed with 4 % PFA (Electron Microscopy Sciences) in PBS 0.1M and then proceed to the immunolabeling.
Tissue dissections
Mice were anesthetised by a solution of Ketamine (100mg/ml)/Xylazine (Rompun, 20mg/ml): 130mg/kg of ketamine + 13mg/kg of xylazine, transcardially perfused with ice-cold PFA 4% in 0.1M PBS, and then the brains, the spinal cord, lungs and adipose tissue were dissected. The tissues were fixed overnight with 4 % PFA in 0.1M PBS and then the brains were cut into left and right hemisphere, the rest of the tissues were cut in half. Half of the samples were embedded with paraffin and the other half with cryomatrix. 8 μm and 50 μm thick sections were used for paraffin labelling and for cryo-section, respectively.
For LPC demyelinated lesions, mice were euthanised several days post LPC-injection (dpi), after lethal anaesthesia with Xylazine (10mg/kg) and pentobarbital sodium (150mg/kg), and then transcardially perfused 4% PFA in 0.1M PBS. Spinal cords were dissected and post-fixed 2 h in 4% PFA. After post-fixation, spinal cords were cryoprotected in 20% sucrose solution O/N and then frozen in O.C.T. compound (Thermo Fisher) at −60°C in isopentane. Coronal spinal cords were sections (12μm thickness) were performed at cryostat (Leica) and slides were kept on −80°C until use.
Immunolabelling in mouse samples
For immunolabeling, samples were permeabilized with PTX (0.1M PBS, 0.1 % Triton-X100) for 30 min and incubated with blocking buffer for 1 h at room temperature (RT). The samples were incubated with primary antibodies overnight at 4°C, then incubated with the appropriate secondary antibodies. Finally, they were incubated with DAPI, (Sigma-Aldrich)) to label the nuclei and the samples were mounted with Aqua Poly/Mount (Polysciences). Primaries and their appropriate secondary antibodies are on supplementary table 1.
For peroxidase immunolabelling of Iba1, sections were incubated with the primary antibody incubation overnight and then washed extensively in 0.1M PBS,0.1% Triton. Slides were incubated for 1 h in RT with biotinylated secondary antibody for 1h, washed extensively and then incubated with the avidin–biotin–peroxidase (ABC) complex (Vector Laboratories). After washes, slides were incubated with the chromogen 3,3’ diaminobenzidine tetrahydrochloride (DAB; Sigma–Aldrich) until desired labelling intensity developed and counterstained with haematoxylin.
Oil Red O staining
For Oil Red O (ORO) that labels macrophages containing myelin debris, spinal cord sections were dried at RT, rinsed in 60% isopropanol, then stained with freshly prepared and filtered 0.01% ORO solution. After 20 min, slides were rinsed in 60% isopropanol, counterstained with haematoxylin, rinsed in water and mount in aqueous mounting medium.
MS tissue samples
Snap frozen post-mortem brain and cerebellar samples from MS and control patients were obtained from the UK MS tissue bank (Imperial College, London, approved by the Wales Research Ethics Committee, ref. no. 18/WA/0238). For this study, we used 4 MS and 1 control samples (Supplementary Table 2). 12μm-thick sections were cut on a cryostat, and lesions were classified as active (N=2), chronic active (N=1) and chronic inactive (N=2), using Luxol fast blue/major histocompatibility complex class II (MHCII) staining, as previously described (32). For immunohistochemistry, sections were post-fixed 20 min in 2% PFA and immunofluorescence labelling was performed as mentioned above for mouse experiments.
Quantitative RT-PCR analysis
Animals were euthanised by CO2 inhalation, and samples of spinal cords around of the injection site was dissected. Total RNA, from 4 dpi LPC- and saline-injected spinal cords, were purified using RNeasy Mini Kit (Qiagen 74104). RNA concentrations were measured using Nanodrop. Reverse transcription was performed by using High-capacity cDNA reverse transcription kit with RNase inhibitor (Applied Biosystem 4374967). qPCR was performed with the TaqMan Fast Advanced Master Mix (Thermo Fisher 4444556) with specific probes for Hprt, mGcm2, iNOS, TLR2, ARG1, Il-4ra, and CD16 (Supplementary Table 3). qPCR reactions with the same concentration of cDNA were run in duplicates using LightCycler 96 (Roche). Hprt was used as a housekeeping gene and used as endogenous control. ΔCt values were used to determine the relative gene expression change.
RNAscope multiplex assay
RNAscope in situ hybridisation (RNA ISH) was performed on fixed frozen sections of mouse and human post-mortem samples. Preparations, pre-treatment and RNA ISH steps were performed according to the manufacturer’s protocols. All incubations were at 40°C and used a humidity control chamber (HybEZ oven, ACDbio). For mouse experiments, probe mixes used were as follows: mCX3CR1 (Cat No.314221-C2), mGcm2 (Cat No. 530481-C3), mGcm1 (Cat No. 429661-C1), Polr2a (Cat No.312471, used a positive control) and dapB (No. EF191515; used a negative control). For RNA ISH on human tissues, probes used were: hGCM2 (Cat.871081), hCD68 (Cat.560591) and Polr2a (Cat No.310451, used as positive control); on Drosophila, we used the gcm probe (Cat No.1120751-C1)
Tyramide dye fluorophores (Cy3, Cy5: TSA plus; and Aykoya) or Opal fluorophores (Opal 520, Opal 570, Opal 690, Aykoya) were used diluted appropriately in RNAscope Multiplex TSA dilution buffer. Slides were also counterstained with DAPI.
Imaging
Fluorescent and brightfield imaging were performed under a 20X objective using Axioscan (Zeiss) and Nanozoomer (Hammamatsu) for all quantitative analysis, and representative images in the figures were made using an Apotome (Zeiss).
Leica Spinning Disk microscope equipped with 20, 40 and 63X objectives was used to obtain confocal images with a step size of 0.2-1 μm. For the quantifications, five or more fields per sample were used with more than 50 cells in total.
Image Analysis
Image analysis was performed with the Fiji image analysis program and with Imaris. Fiji was mainly used to produce images with sum of Z-projections. In all images, the signal was set to the same threshold in order to compare the different genotypes. Imaris (version 9.5.1) was used to analyse the morphology of microglial cells during aging using a semi-automatic protocol. The p-values were estimated after comparing control to cKO cells by two-way ANOVA test followed by bilateral student test. To analyse the activation state of microglia in LPC lesions at 7 and 14 dpi, we used the Visiopharm software. We selected 2 parameters: the number of branching points (ramifications) and roundness. Data were analysed by two-way ANOVA, followed by Tukey post hoc test.
Transcriptome analysis of gcmKD haemocytes from Drosophila embryos and larvae
Haemocytes were sorted by FACS from stage 16 embryos of the following genotypes: srp(hemo)Gal4/+;UAS-RFP/+ for the control and srp(hemo)Gal4/+;UAS-RFP/UAS-gcm-RNAi (BDRC #31519) for the gcmKD (33). Haemocytes were also sorted from wandering third instar larvae of the following genotypes: srp(hemo)Gal4/HmlΔRFP (control) and srp(hemo)Gal4/hmlΔRFP;UAS-gcm-RNAi/+ (BDRC #31519) for the gcmKD. 20000 to 50000 cells were sorted for each replicate, three replicates were done for each condition. The RNA were extracted using Tri-reagent (SigmaAldrich) according to the manufacturer protocol. RNAseq libraries were prepared using the SMARTer (Takara) Low input RNA kit for Illumina sequencing. All samples were sequenced in 50-length Single-Read. At least 40×106 reads were produced for each replicate. Data analysis was performed using the GalaxEast platform (http://www.galaxeast.fr/, RRID:SCR_006281) as described in (33, 34). The differential expression analysis was carried out using HTseq-Count (RRID:SCR_011867) and DESeq2 (RRID:SCR_015687) (35). The gene ontology analysis was done with ShinyGO 0.76 (36). The graphs were plotted using the packages pheatmap (RRID:SCR_016418) and ggplot2 (RRID:SCR_014601) in R (version 3.4.0) (R Core Team, 2017). The RNAseq data were deposited in the ArrayExpress database at EMBL-EBI (www.ebi.ac.uk/arrayexpress) under accession number E-MTAB-8702.
Expression profile of the microglial markers in adult Drosophila haemocytes and glia
The dataset GSE79488 for adult haemocytes and GSE142788 for adult glia were retrieved from GEO database, mapped using RNA-STAR and compared with DESeq2 as described above (37). The microglial genes conserved across evolution were determined by Geirsdottir (38). The Drosophila orthologs were determined using DIOPT (39). The heatmap was drawn using pheatmap (RRID:SCR_016418).
Tracing gcm expression
To assess the induction of gcm expression in Drosophila upon wasp infestation, a lineage tracing system was expressed under the control of gcm-specific promoters and a thermosensitive inhibitor. The detailed genotype is UAS-FLP/+;act5c-FRT,y+,FRT-Gal4,UASmCD8GFP/gcmGal4,tubulinGal80TS;UAS-FLP,Ubi-p63E(FRT.STOP)Stinger/6KbgcmGal4 (gcm>g-trace) for the experiment and UAS-FLP/+;act5c-FRT,y+,FRT-Gal4,UASmCD8GFP/+;UAS-FLP,Ubi-p63E(FRT.STOP)Stinger/+ for the control (Supplementary Figure 8A). Embryos and larvae were raised at 18°C (tracing off) until the second instar larval stage, to avoid revealing the embryonic expression of gcm. Second instar larvae were infested with the parasitoid wasp Leptopilina boulardi. 20 female wasps were used to infest 100 Drosophila larvae for 2 h at RT. Infested larvae were then transferred at 29°C (tracing on) and let develop until wandering third instar larval stage for further analyses.
Wasp infestation assays
Wasp eggs were collected upon bleeding 50 infested larvae in PBS 1X added with some N-Phenylthiourea crystals (Sigma) to prevent haemocyte melanisation. Wasp eggs were fixed in 4% PFA/PBS 1X for 30 min, washed with PTX (0.3% Triton X-100 in PBS 1X), incubated 1h in PTXN (NGS 5% in PTX, Vector Laboratories), incubated overnight at 4°C in primary antibodies, washed with PTX, incubated 1h in secondary antibodies, 1h with DAPI and TRITC-phalloidin (Sigma), then mounted with Vectashield mounting medium (Vector Laboratories). Haemocytes were labelled as previously published (12). Primaries and their appropriate secondary antibodies are on supplementary table 1.
For the gain and loss of function assays, we used control, gain of function (GOF) or loss of function (LOF) flies upon crossing w;hmlΔGal4/CyoGFP;tubGal80TS animals with white-1118 (control), UASTgcmF18A (gain of function, GOF) or UASgcmRNAi (loss of function, LOF) animals, respectively (40,12). The infestation was conducted as described above and the tumour size was counted as previously described (12).
Statistical analysis
Variance analysis using bilateral student tests for unpaired samples was used to estimate the p-values for microglia ramifications, coverage area, iNOS, Arg1-positive microglia. In each case, at least five animals were counted. In all analyses, “ns” means not significant, “*” for p-value < 0.05; “**” for p-value < 0.01; “***” for p-value < 0.001.
For cell quantification in LPC lesions, 3-5 spinal cord sections per animal were analysed, with at least N=3/group. Quantification of Olig2+ and CC1+ cells was performed using ZEN (Zeiss) and QuPath software. Statistical analyses were performed by using R and GraphPad (Prism) Software using Two-way ANOVA with Tukey post hoc comparison tests.
Study approval
All animal experiments were conducted according to the European law for the welfare of animals. All animal procedures were reviewed and approved by the “Comités d’Ethiques en Expérimentation Animale” of IGBMC-ICS and of the Paris Brain Institute - ICM.
Results
Expression of the murine Gcm genes in vitro
We assessed whether the mGcm1 and mGcm2 genes are expressed in microglia by characterising CNS primary cultures (41, 42). Microglia were labelled with the broad hematopoietic marker CD45 (43). Microglia adopt one of the three morphologies both in vitro and in vivo: ramified microglia with small cell body and long ramifications, round cells with a small cell body and no ramifications, amoeboid microglia with a big cell body and no ramifications (44–46). Round cells are considered as activated microglia, the other two as resting cells (47). mGcm2, but not mGcm1, is expressed in 30% of round shaped microglia (Figure 1B and Supplementary figure 1A). We ascertained the specificity of the mGcm2 immunolabeling in a conditional knock out (cKO throughout the text) mGcm2flox/flox mouse line crossed it with the Cx3cr1-Cre line expressed in microglia. The mGcm2flox/flox line was produced by introducing two LoxP sites upstream and downstream of the Gcm exons 2,3 and 4 (Figure 1A). mGcm2 labelling is absent in the cKO microglia, proving by the same token the specificity of the antibody and the efficiency of our cKO model (Figure 1B).
In sum, mGcm2 is expressed in active microglia of CNS cultures.
mGcm2 is expressed in the microglia of aged animals
Based on the above results, we analysed neonatal microglia at postnatal day 14 (P14) but found no mGcm2 expression (Figure 1C). This could be explained by the fact that, in vivo, microglia display major differences from microglial cultures (48). The finding that mGcm2 expression is specific to activated microglia in vitro, prompted us to ask whether its expression is induced in an inflammatory condition such as aging. Brain and spinal cord were collected at different ages: P14, 2, 12 and 24 months. As cell-specific markers, we used the microglia-specific cocktail (CD11b, CD68, F4/80) that distinguishes microglia from the meningeal macrophages (49). mGcm2 is expressed in microglia at 12 and 24 months but not in 2month-old animals in different areas of the brain including the cortex (Figure 1C). Moreover, the number of mGcm2-positive microglia increases over time: 12 and 24month-old cortices show 25% and almost double mGcm2-positive microglia (48%), respectively (Figure 1D).
We also performed RNA ISH on brain sections and confirmed that mGcm2 is expressed in aged brains (18month-old animals), while mGcm1 is not expressed in either control or cKO brains (Supplementary figure 1B,C). This further validates the efficiency of our flox/flox line and shows that mGcm1 does not compensate for the lack of mGcm2 expression. Finally, to evaluate the expression in other resident macrophages, lung and adipose tissues were dissected from 24month-old animals and cryo-sections were labelled for CD45 and mGcm2. No mGcm2 labelling was observed in the resident macrophages associated with those tissues (Supplementary figure 1D).
Thus, mGcm2 is expressed only in a subset of aged microglial cells.
Microglia of mGcm2 cKO mice have an activated morphology in homeostatic conditions
To explore the role of mGcm2 in microglia in vivo, we characterised their morphology, which is tightly linked to the activation state of microglia: resting cells have long ramifications while activated cells have shorter ramifications (50). Sections of control and cKO brains from different age groups were labelled with the microglial marker Iba1 (Figure 2A-C) (51, 52). Our criteria include the number of ramifications and the coverage area of the cell, that is, the area where microglia extend their ramifications at, which thus represents the area they can survey. The former parameter is a direct measurement of their activation state while the latter is an indirect indicator, as activated microglia have shorter or less ramification, which results in a decreased coverage area.
We found that microglia morphology changes over time, but more in the cKO animals (Figure 2A-C). More specifically, the number of ramifications per cell decreases during aging in both genotypes (Figure 1A). However, there is a significant decrease in the number of ramifications in mGcm2 cKO compared to control microglia at 24 months (Figure 2B). The same trend is also visible for the coverage area (Figure 2C tendency to decreased coverage area in the mGcm2 cKO).
Altogether, these results reveal for the first time that the lack of mGcm2 has an impact on microglia morphology, indicative of a pro-inflammatory phenotype.
mGcm2 cKO animals display a pro-inflammatory profile
We complemented the morphological data by labelling 2, 12 and 24month-old brains with pro- and anti-inflammatory markers. Microglia/macrophage activation states are classified as pro-inflammatory (or M1) or anti-inflammatory (or M2). We chose the markers inducible nitric oxide synthase (iNOS) for the M1 and Arginase-1 (Arg1) for the M2 state, which were already used to characterise microglia in vivo (53). Labelling from different age groups shows that iNOS is present mostly in 12 and 24month-old animals and its expression is increased in both groups as they age (Figure 3A,B). Upon quantifying the iNOS-positive microglia, we found that mGcm2 cKO animals specifically display an increased number of iNOS-positive microglia compared to control animals by 24 months (Figure 2B).
Next we evaluated the number of Arg1-positve cells (Figure 3C,D). 2month-old control animals have a higher number of Arg1-positive microglia compared to 12 and 24month-old animals (Figure 3D). Likewise, mGcm2 cKO animals show a significant decrease in the number of Arg1-positive microglia, from 2 (74%) to 12 months (47%). Importantly, only the percentage of Arg1-positive microglia in the cKO animals further decreases from 12 to 24 months, resulting in a significant decline. Thus, the microglia lacking mGcm2 display a stronger progression toward a pro-inflammatory phenotype compared to control microglia.
The state of other CNS populations does not seem overtly affected in the mutant animals. The size of astrocytes increases in inflammatory conditions (54, 55), a phenomenon called astrogliosis that can be measured via GFAP labelling (56). We found no difference between the GFAP labelling of 24month-old control and mutant animals (Supplementary figure 2A,B). Similarly, we found no difference in neuronal cell death by co-labelling the cell death marker caspase 3 with the pan neuronal marker NeuN (Supplementary figure 2C-E). No difference in oligodendrocyte number was found either (Supplementary figure 2F,G). Accordingly, mutant and control animals show similar behavioural habits in an open field test (Supplementary figure 2H-J), which assays general locomotor activity levels, anxiety, and willingness to explore.
The morphological and the molecular data show that mGcm2 has an anti-inflammatory role in murine microglia during chronic inflammatory conditions such as aging.
mGcm2 is expressed in acute LPC lesions and hGcm2 in active MS lesions
Since mGcm2 expression is induced upon aging, we asked whether it is also expressed in immune cells following CNS injury, which represents a condition of acute inflammation. We induced acute demyelination by LPC injection in the mouse dorsal white matter spinal cord and analysed mGcm2 expression profile. LPC lesions show limited amount of inflammation, usually associated with myelin debris removal. The mGcm2 protein was specifically detected in a subset of CD45+ immune cells (Figure 4A). RNAscope ISH assays also showed mGcm2 labelling in very few microglia specifically within the lesions from 2 to 21 dpi (Figure 4B).
To assess the relevance of these findings in humans, we analysed the expression of hGCM2 in acute and chronic MS lesions, as well as in the normal appearing white matter from MS and non-neurological control cases (Supplementary Table 1). MS lesion subtypes were first characterised using Luxol Fast Blue and MHCII staining (Figure 5A) and classified as active, chronic active and chronic inactive (32). The hGCM2 protein was specifically detected in few MHCII+ immune cells located in active MS lesions, which have traditionally been defined as showing demyelination with inflammatory infiltrates, whereas chronic lesions show demyelination with little or no activity (Figure 5B). We next performed double RNA ISH for hGCM2 and hCD68 and found hGCM2 expression in few hCD68-positive microglia only in active lesions and in the active rim of chronic active lesions (Figure 5C).
In sum, hGCM2 and mGcm2 gene expression is upregulated in demyelinating lesions, in a subset of cells belonging to the microglia/macrophage lineages.
Loss of mGcm2 in microglia promotes a pro-inflammatory response after demyelination
To decipher the functional role of mGcm2 in microglia/macrophages in LPC lesions, we generated an inducible cKO mouse line. The mGcm2 floxed strain (Figure 1A) was crossed with the Cx3cr1CreER/+ mouse line (31). The deletion was induced in the F1 generation, specifically in microglia by tamoxifen injection (Figure 6A). We refer to this mouse strain as Cx3cr1CreER/+;Gcm2flox/flox (inducible cKO, icKO). Cx3cr1CreER/+;Gcm2flox/+ heterozygous littermates (named control thereafter), Gcm2flox/flox and WT animals were used as controls for LPC lesions. 10-16 week-old mice were treated with tamoxifen during 5 consecutive days prior to LPC induced lesions. We then assessed the impact of mGcm2 deletion in microglia on the different steps of the remyelination process, including OPC recruitment, differentiation and remyelination (Figure 6B). mGcm2 deletion was first confirmed by qPCR analysis of RNA extracted from dissected spinal cord lesions at 4 dpi. mGcm2 relative mRNA expression is indeed severely reduced in LPC demyelinated spinal cords of icKO mice with respect to controls (Figure 6C).
Once the tamoxifen-inducible Cre deletion of mGcm2 in Cx3cr1-expressing cells had been validated, we monitored the microglia response to LPC lesions in mGcm2 icKO mice. Immunohistochemistry for iNOS combined with the microglia cocktail markers (CD11b, F4/80 and CD68) revealed a drastic increase of microglial cells in a pro-inflammatory state in demyelinated lesions of mGcm2 icKO mice as compared to controls (Figure 6D). To further confirm that the mGcm2 deletion promotes a microglial pro-inflammatory state after LPC-induced demyelination, we performed qPCR analysis of M1 and M2 marker relative expression. In agreement with the immunohistochemistry data as well as the aging data, the expression of the M1 gene iNOS (Figure 6E) and TLR2 (Figure 6F) revealed a trend increase after spinal cord demyelination in the icKO mouse strain with respect to WT, mGcm2flox/flox and control mice. We also evaluated the relative gene expression of the M2 markers Arg1 (Figure 6G), Il-4ra (Figure 6H) and CD163 (Figure 6I) in mGcm2 icKO and control animals, however, our data did not reveal any significant difference.
To further corroborate these findings, we examined microglial cell morphology in conditional mGcm2 icKO and control mice in LPC demyelinated lesions, in vivo. Spinal cord sections throughout the demyelinated lesions were labelled with the microglial marker Iba1 and microglia morphology was evaluated with VisioPharm (57). In line with the above data, the number of ramifications point per microglial cell decreases significantly in mGcm2 icKO compared to controls, specifically in demyelinated lesions at 7 and 14 dpi (Figure 6J). As an additional morphological parameter of activated microglia, we evaluated the roundness of the cells at the same time points. The number of microglial cells exhibiting a round morphology increases significantly in mGcm2 icKO compared to controls (Figure 6K), further indicating that mGcm2 loss-of-function in microglia favours a pro-inflammatory state.
The above findings support a role of mGcm2 as a new anti-inflammatory transcription factor in mouse microglia under pathological conditions.
Loss of mGcm2 in microglia delays oligodendrocyte differentiation in demyelinated lesions
Microglia activation state plays a critical role in the regulation of demyelination and remyelination (58). Therefore, we asked whether mGcm2 loss-of-function in microglia hampers myelin debris clearance and oligodendrocyte differentiation in LPC induced demyelinated lesions. Oil-Red O staining was performed to evaluate the density of microglia containing myelin debris (Supplementary Figure 3A). The Oil-RedO-positive aeras as well as the percentage of Oil-RedO-positive aeras in LPC lesions do not differ significantly between mGcm2 icKO and control mice at similar time point post-demyelination (Supplementary Figure 3B,C), suggesting that mGcm2 deletion in microglia does not affect the ability of these cells to phagocyte myelin debris.
We next examined whether oligodendrocyte differentiation could be hampered. Spinal cord sections of demyelinated lesions from icKO and control mice were immunolabeled for Olig2, a pan-oligodendrocyte marker, together with CC1, a specific marker of differentiated oligodendrocytes (Supplementary Figure 4A). Quantification of the percentage of Olig2+CC1+ mature oligodendrocytes and Olig2+CC1-OPCs revealed a significant increase of differentiated oligodendrocytes in both groups at 7 and 21 dpi (Supplementary Figure 4B, D). Nevertheless, oligodendrocyte differentiation is delayed in LPC lesions of the mGcm2 icKO strain with respect to controls, between 7 and 14 dpi (Supplementary Figure 4B,C). It is worth noting that the overall number of Olig2+ oligodendroglial cells in demyelinated lesions increases in both groups but was significantly lower in the icKO mice compared to the controls at 21dpi (Supplementary Figure 4D), suggesting that the pro-inflammatory state of mGcm2 icKO microglia hampered oligodendroglia cell survival or apoptosis.
gcm expression is induced upon aging in Drosophila melanogaster
Since the Gcm pathway is induced in chronic (aging) and acute (LPC induced) inflammatory conditions in mice, we asked whether this is an ancestral process and evaluated it in aging Drosophila brains. In physiological conditions, Gcm is expressed in a subpopulation of haemocytes, in which it has an anti-inflammatory role (12). Gcm expression is confined to the haemocytes derived from the first hematopoietic wave that occurs in the procephalic mesoderm of the embryo (19). These haemocytes cease to express Gcm by the end of embryogenesis and survive to the adult where they coexist with those derived from the second wave occurring in the larval lymph gland, which is Gcm independent (59).
In the adult, haemocytes are mostly associated with peripheral tissues (60), we therefore first assessed the number of brain associated haemocytes over time by labelling dissected fly brains with macrophage markers (P1/NimC4 and Hemese) at different ages, from week 1 (young animals) to week 6 (old animals), (Figure 7A,B). The number of brains that display associated haemocytes increases over time and by 6 weeks all brains are associated with haemocytes (Figure 7C). The number of brain-associated haemocytes observed in each animal varies, which likely also depends on the dissection protocol that may dissociate from the brain migratory cells as haemocytes. We next assessed the expression profile of Gcm in the adult brain upon aging, by performing in situ hybridization with a RNAscope probe. In young animals, Gcm is expressed only in two neuronal clusters located in the central brain (Supplementary Figure 5A) at lateral and dorsal positions (61). In 6week-old animals, however, Gcm labelling is also present at ectopic positions, indicating de novo expression (Figure 7A,D). As in the case of the haemocyte markers, gcm labelling is located at the surface of and not within the brain, indicative of cells that are associated with but do not belong to the tissue itself.
gcm expression is induced upon acute challenge and counteracts the inflammatory state
gcm expression in the old haemocytes might be a transient process. Since the in situ assay only identifies cells that express gcm at the time of the dissection, this approach may miss cells that have expressed gcm earlier. For this reason, we performed lineage tracing using the g-trace tool and followed all the cells that express and/or have expressed gcm at some point (62). This may reveal more gcm-positive cells than the in situ assay. To make sure that we specifically look at de novo expression and not at remnant expression from the embryonic haemocytes, we activated the g-trace only after adult eclosion (Figure 8A,B). The results showed gcm-positive cells that are associated with the brain and do not express glial or neuronal markers. Of note, the number of gcm-positive cells that are associated with old brains is not higher than those revealed by in situ. Surprisingly, these cells do not express the P1/Hemese pan-haemocyte markers either (Figure 7E and Supplementary Figure 5B). This may indicate an uncharacterised population of haemocytes present at that location or these cells may have not yet acquired a proper haemocyte identity. The production of additional tools will be necessary to distinguish amongst these possibilities, but clearly non-neural gcm expressing cells become associated with the aged brain.
Based on the above results, we asked whether gcm is reactivated upon an acute inflammatory challenge. In Drosophila, the most studied acute inflammatory response is induced by wasp parasitisation. In brief, the parasitoid wasp Leptopilina boulardi is allowed to infest and lay eggs in Drosophila larvae. This leads to extensive haemocyte proliferation and activation, which consists in resting haemocytes transdifferentiating into lamellocytes (activated haemocytes) in the infested larvae (63). Lamellocytes are huge cells able to encapsulate the wasp egg, preventing it from hatching and hence allowing Drosophila to escape the infestation. Upon activating the g-trace tool only during the larval life, no gcm expression can be detected in normal conditions in third instar larvae (Supplementary Figure 5C). Upon wasp infestation, however, gcm (g-trace)-positive cells surround the wasp eggs, revealing a de novo expression of gcm following the acute challenge (Figure 8C). We also found gcm expression in circulating lamellocytes (Figure 8D). Thus, gcm expression is induced by an acute inflammatory challenge in flies.
Since gcm is expressed de novo upon wasp infestation, we asked how gcm gain or loss of function (GOF or LOF, respectively) would affect the response to wasp infestation (Figure 9A). To this purpose, we specifically induced (GOF) or silenced (LOF) gcm expression in the larval haemocytes after wasp infestation (Figure 8B) and evaluated the so-called tumour phenotype as a readout of the inflammatory response. The infested animals of the three genotypes (control, LOF and GOF) carry tumours, but their number and/or size varies. Large tumours contain the wasp eggs encapsulated by the fly haemocytes, while small/medium size tumours are due to haemocyte aggregations. LOF animals have in average more tumours than control and GOF animals (Figure 9B). This is mostly due to a very large increase in the number of small tumours (Figure 9C,D). GOF animals, on the other hand, show a decreased number of large tumours compared to control animals.
In sum, silencing the de novo expression of gcm aggravates the inflammatory phenotype and inducing gcm expression de novo ameliorates it.
gcm downregulation triggers a pro-inflammatory state in the Drosophila haemocytes
Altogether, our data indicate that the conserved Gcm pathway is induced in response to a wide variety of challenges and counteracts acute as well as chronic inflammation. It seemingly has a priming role: the cKO mice are fully viable, fertile and do not display an overt inflammatory phenotype, much like the mutant flies (12). To investigate the molecular mechanisms underlying this priming process we proceeded to a high throughput analysis in flies, given the simplicity of this animal model. Since Gcm is also involved in gliogenesis and its mutation is embryonic lethal, we analysed the transcriptome of srp(hemo)Gal4;UAS-gcm-RNAi (gcmKD) animals in which Gcm is specifically affected in haemocytes. gcmKD haemocytes present overall increased levels of expression of immune related genes, such as anti-microbial peptides and components of major immune pathways (Figure 10A). In the embryo, the different expression is restricted mostly to STAT92E and Toll, but in the larva more genes appear to have different expression levels between gcmKD and control haemocytes. The highest difference is found in Gene Ontology (GO) terms associated with immune regulation, such as regulation of immune response to bacteria and signalling pathway of recognition of peptidoglycans (Figure 10B).
Thus, silencing gcm triggers a pro-inflammatory state.
Discussion
The present study identifies a novel and evolutionarily conserved anti-inflammatory transcriptional pathway. The Gcm transcription factor was known to regulate the development of fly immune cell populations (glia and haemocytes) and to modulate the inflammatory response in flies. Here we demonstrate that the expression of fly and murine Gcm genes is induced upon challenge, which helps counteracting the inflammatory state. Moreover, the human Gcm2 ortholog is expressed in MS lesions. The identification of this conserved pathway opens exciting perspectives to study neuro and inflammatory diseases.
mGmc2 and its anti-inflammatory role in the microglia of aged mice
Aging results in gradual loss of normal function, due to changes at the cellular and molecular level (64). Microglia exhibit an exaggerated pro-inflammatory response during aging, a phenomenon referred to as microglia priming (26, 27). Morphologically, aged microglia have enlarged processes, cytoplasmic hypertrophy and a less ramified appearance (28, 65). They also express higher levels of activation markers than control microglia (66). All these changes lead to impaired remyelination and are associated with decreased number of M2 microglia, cells in an anti-inflammatory state (58).
Our data indicate that mGcm2 expression helps keeping the inflammatory state under control during aging. Aged microglia of cKO mice have a pro-inflammatory morphology with less ramifications and coverage area. Furthermore, they show a statistically significant increase of the pro-inflammatory marker iNOS and a significant decrease of the anti-inflammatory marker Arg1 between 12 and 24 months. The most parsimonious hypothesis is that mGcm2 acts as a regulator of the activation state in microglia. Loss of its expression leads to an uncontrolled stimulation and thus a higher pro-inflammatory profile even under basal conditions. It is widely accepted that targeting anti-inflammatory (M2) regulators of microglia could lead to new therapeutic targets for aging, mGcm2 may represent one of them.
Vertebrate Gcm2 is a new anti-inflammatory transcription factor in CNS demyelinating lesions
The mGcm2 protein is present in few CD45-positive immune cells after acute demyelination induced by LPC injection, indicating that mGcm2 expression is restricted to a subset of inflammatory cells. Moreover, using double RNA ISH for mGcm2 and cx3cr1, we provided compelling evidence supporting the expression of mGcm2 in a subset of microglia/macrophages, specifically in demyelinated lesions. Acute inflammation is one of the features of LPC-induced demyelination (67, 68). Interestingly, hGCM2 expression was also detected in a small fraction of MHCII/CD68-double positive microglia/macrophages located specifically in active lesions and in the active rim of chronic active lesions of MS, thus further supporting a conserved function of Gcm2 in microglial/macrophage lineage cells in humans. In order to evaluate the role of mGcm2 specifically during an acute inflammatory condition and not as a constitutive deletion, as we did during aging, we used the icKO model. Our data gained in the mGcm2 icKO, specifically in microglia/macrophages, clearly indicate that mGcm2 favours an anti-inflammatory M2 state, and could therefore promote indirectly myelin regeneration and repair by modulating OPC differentiation after acute demyelination (58). This emerging anti-inflammatory role of Gcm2 may open new therapeutic interventions, targeting this transcription factor in demyelinating diseases.
Gcm: glia, haemocytes and microglia
Our data explain what was considered as a conundrum. The fly gcm gene was initially described for its developmental role in glia and haemocytes (10, 18). Surprisingly, while the genes necessary for neuronal differentiation are functionally conserved in evolution, the orthologs of Gcm, the fly glial promoting factor, are neither expressed nor required in the differentiation of vertebrate glial cells (69, 70). Moreover, none of the early transcription factors expressed/required in fly glial cells and target of Gcm such as the panglial Reverse Polarity (Repo) transcription factor is expressed in vertebrate glia and this gene is not even conserved in the vertebrate genome. The gliogenic pathway seems therefore not evolutionarily conserved, in line with the hypothesis that glia may have evolved several times, adapting to the needs of the organism (71). The glia of simple organisms play a neural role, as they control axon ensheathment, synapse activity and insulate the brain through the blood brain barrier (BBB). In addition, they act as the resident immune cells of the nervous system. By contrast, in complex organisms, the role of brain resident immune cells is taken by a new cell type of non-neural origin, the microglia, that infiltrate the nervous system before the BBB is formed. This division of labour guarantees a better and more targeted response to inflammatory challenges. We speculate that the haemocytes of simple, fast developing, animals migrate along the nervous system and contribute to remove dying cells, but as the BBB forms, they are excluded from the tissue, the immune function being taken up by the glial cells. According to this hypothesis, we should find orthologs of microglial markers expressed in fly haemocytes as well as glia. A recent and elegant study characterised the microglia transcriptional programs across ten species spanning more than 450 million years of evolution (72). By using these data with the RNAseq data from adult fly haemocytes and glia, we created a heatmap of the microglia orthologs in Drosophila (37, 73). These orthologs are mostly shared between the two fly cell populations (Figure 10C,D). One microglial gene whose ortholog is shared by haemocytes and glia is Peli2, a member of E3 Ubiquitin ligases controlling the Toll signalling pathway (74, 75). The Pellino (Pli) fly gene antagonises Toll-mediated innate immune signalling by controlling MyD88 turnover in macrophages. Future studies will determine whether the pathway is also conserved in glia. In addition to the shared genes, some are specific to glia (spn42Da, moody, CG42709) or haemocytes (CG7882, pgant9, eya and CG30345). Interestingly, moody is one of the most known markers of the Drosophila BBB glia and its ortholog in mammals (Gpr84) is a well-known pro-inflammatory maker that is highly up regulated in microglia upon nerve injury (76). Revisiting the role of moody during neuroinflammation could shed light into the function of this gene in both species.
As a corollary of the above and based on multiple pieces of evidence, we propose that repo constitutes the bona fide fly gliogenic gene. Accordingly, Repo misexpression in the mesoderm suppresses haematopoiesis and its lack triggers the expression of haemocyte markers in the nervous system (77). We speculate that the Gcm pathway has an ancestral, conserved, role is in immunity and may have been coopted in the differentiation of fly glia. One of the future challenges will be to characterize the Gcm-positive cells associated with the aged brain.
The conserved role of the Gcm pathway in immune processes
Together with the inhibitory role of Gcm on the JAK/STAT pathway and the increased response to wasp infestation, which relies on the Toll cascade, the present data strongly suggest that Gcm controls different inflammatory conditions, including aging (12). This is also in line with the transcriptomic data of the gcmKD animals, which confirm the induction of the JAK/STAT pathway observed in vivo and extend the inhibitory role of Gcm to other pathways such as IMD. Furthermore, gcm GOF can ameliorate the inflammatory phenotype in flies, thus paving the way for new therapeutic strategies against autoimmune diseases, such as MS, where the inflammatory response needs to be contained. We hypothesise that the induction of the Gcm cascade reduces the intensity of the inflammatory response and hence has a protective function. This hypothesis is corroborated by recent data obtained in other organisms. A peculiar macrophage population called pigment cells is present in the sea urchin (78). Such cells are involved in the immune defence by the production of a pigment that has anti-microbial properties. Morpholino antisense oligonucleotides for Spgcm (gcmMO) injection showed that gcmMO animals are less resistant to challenging environmental conditions portrayed by decreased survival rate compared to the control (79–81). A recent study showed that a gcm ortholog is also expressed in glia-like cells of the freshwater crayfish (Pacifastacus leniusculus) upon an acute inflammatory response (82). Moreover, the expression of a planaria gcm ortholog was found to be induced upon regeneration in a subset of cells close to the wound; its silencing has no effect in homeostatic conditions but impairs neoblast repopulation upon wounding (83, 84). Thus, more and more studies highlight an evolutionarily conserved mechanism.
In sum, we report here the discovery of an anti-inflammatory transcriptional cascade that is conserved from flies to humans. Given the strong potential of transcription factors in coordinating the expression of several genes and the scarce number of known transcription factors with a similar function, this work represents a major contribution to understand the molecular mechanisms controlling the inflammatory response. It also lays the ground for studying novel therapeutical targets for neuro-inflammatory diseases in humans.
Author contributions
AP, SM, PC, RP, BNO and AG designed the experiments and co-wrote the manuscript. AP, SM, CR, RP performed the experiments in mice, RP performed the experiments in humans, SM and AP performed the assays in Drosophila melanogaster, YY was responsible for the mGcm2flox/flox production and PC analysed the RNAseq data.
Acknowledgements
We thank the Imaging Center of the IGBMC for technical assistance, as well as the Mouse Clinic (ICS) for producing of the mGcm2flox/flox strain. We also thank the ICM mouse facility (ICMice), the ICM histology (Histomics) and the cellular imaging (ICM-Quant) facilities. AP was supported by the ARSEP Foundation and the grant from Laboratoires d’excellence (LabEx INRT), SM by CEFIPRA and the FRM foundations, YY by the ARSEP foundation. RP was funded by the ARSEP foundation and NeurATRIS.
This work was supported by INSERM, CNRS, UDS, Ligue Régionale contre le Cancer, Hôpital de Strasbourg, ARC, CEFIPRA, ANR grants, the CNRS/University LIA Calim, The French MS foundation ARSEP, the Investissements d’Avenir ANR-10-IAIHU-06 (IHU-A-ICM) and ANR-11-INBS-0011 (NeurATRIS). The IGBMC was also supported by a French state fund through the ANR labex. We are grateful to the UK MS tissue Bank (Imperial College, London, UK) for providing post-mortem MS brain samples and to C. Linnington and I. Ando for providing antibodies.
REFERENCES
- 1.↵
- 2.↵
- 3.↵
- 4.
- 5.
- 6.
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.
- 22.
- 23.
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.
- 81.↵
- 82.↵
- 83.↵
- 84.↵