CSF1R inhibition at chronic stage after spinal cord injury modulates microglia proliferation

Traumatic spinal cord injury (SCI) induces irreversible autonomic and sensory‐motor impairments. A large number of patients exhibit chronic SCI and no curative treatment is currently available. Microglia are predominant immune players after SCI, they undergo highly dynamic processes, including proliferation and morphological modification. In a translational aim, we investigated whether microglia proliferation persists at chronic stage after spinal cord hemisection and whether a brief pharmacological treatment could modulate microglial responses. We first carried out a time course analysis of SCI‐induced microglia proliferation associated with morphological analysis up to 84 days post‐injury (dpi). Second, we analyzed outcomes on microglia of an oral administration of GW2580, a colony stimulating factor‐1 receptor tyrosine kinase inhibitor reducing selectively microglia proliferation. After SCI, microglia proliferation remains elevated at 84 dpi. The percentage of proliferative microglia relative to proliferative cells increases over time reaching almost 50% at 84 dpi. Morphological modifications of microglia processes are observed up to 84 dpi and microglia cell body area is transiently increased up to 42 dpi. A transient post‐injury GW2580‐delivery at two chronic stages after SCI (42 and 84 dpi) reduces microglia proliferation and modifies microglial morphology evoking an overall limitation of secondary inflammation. Finally, transient GW2580‐delivery at chronic stage after SCI modulates myelination processes. Together our study shows that there is a persistent microglia proliferation induced by SCI and that a pharmacological treatment at chronic stage after SCI modulates microglial responses. Thus, a transient oral GW2580‐delivery at chronic stage after injury may provide a promising therapeutic strategy for chronic SCI patients.


| INTRODUCTION 1.| Microglia at chronic stage after spinal cord injury
Spinal cord injury (SCI) that affects almost 1 million new persons annually worldwide (Kumar et al., 2018) induces motor, sensory, and autonomic deficits.SCI occurs at a mean age of 39.8 ± 12.2 years and the mortality ranges from 0% to 15% (Kumar et al., 2018).There is thus large number of patients with chronic SCI (Middleton et al., 2012) and no curative treatment for any of the induced symptoms is available.After SCI, microglia, the immunocompetent cells of the central nervous system, undergo highly diverse activation processes, including proliferation, and play a critical role on functional recovery through the release of detrimental and beneficial factors to their surrounding cells (for review see Gaudet & Fonken, 2018).In mammalians, the overall dynamics of the microglial response to SCI is characterized by an early proliferation and migration toward the lesion site (for review see Perez et al., 2021).Microglia then participate in the stabilization of the astroglial barrier that surrounds the lesion core and forms the glial scar.Few studies have investigated the microglial response at chronic stage following SCI.In rats, microglia number is still increased 28 days after spinal cord over-hemisection (Schwab et al., 2001) and contusion/compression injury (Zendedel et al., 2012).
Thirty-three days after spinal cord contusion microglia are still activated (Hains & Waxman, 2006) and 60 days after impaction a higher number of microglia as compared to control is reported (Akhmetzyanova et al., 2022).

| Modulation of the microglial response in spinal cord injury
Manipulating the microglial response is an attractive approach to promote axonal regrowth and functional recovery after SCI.CSF1 (colony stimulating factor 1) regulates microglial proliferation, differentiation, and cell survival.In mice, the use of a CSF1 receptor (CSF1R) inhibitor that completely eliminates microglia, before and after a SCI, either did not improve motor recovery (Xia et al., 2022) or even worsens it (Bellver-Landete et al., 2019;Brennan et al., 2022).In agreement, Ganciclovir-induced microglia depletion prior and after severe SCI in mice did not improve motor recovery (Jakovcevski et al., 2021).Conversely, we have shown that transient inhibition of microglia proliferation by oral administration of GW2580 (a CSF1R inhibitor that only inhibits microglia proliferation; Conway et al., 2005) after a lateral hemisection of the spinal cord promotes motor recovery and preserves tissue in both adult mice and nonhuman primates.Indeed, we first demonstrated in mice that GW2580-treatment between 4 weeks prior to SCI and up to 6 weeks post-lesion decreases microglia proliferation and improves fine motor recovery (Gerber et al., 2018).
Second, we have shown in mice and in nonhuman primates that a transient treatment after injury (1 and 2 weeks for mice and nonhuman primates, respectively) promotes motor function recovery and modulates myelinated fibers reorganization (Poulen, Aloy, et al., 2021).Strikingly, increasing the duration of the inhibition (6 weeks) does not further improve recovery (Poulen, Bartolami, et al., 2021).
In this study, we show for the first time in mice that SCI-induced microglia proliferation persists at 6 and 12 weeks after SCI and that a transient GW2580-delivery at chronic stages reduces microglia proliferation, changes microglia morphology toward homeostatic phenotype, and modifies outcomes on myelin dynamics.

| Ethic committee approval
Experimental procedures followed the European legislative, administrative, and statutory measures for animal experimentation (EU/Directive/2010/63 of the European Parliament and Council) and the ARRIVE guidelines.Experiments were approved by the Veterinary Services Department of Hérault, the regional ethic committee n 36 for animal experimentation, and the French Ministry of National Education, Higher Education and Research (n 34118).

| Animals and spinal cord injury
All animals used in this study were adult (90-95 days old) heterozygous CX3CR1 +/eGFP transgenic female mice.They were originally obtained from Dr. Dan Littman (Howard Hughes Medical Institute, Skirball Institute, NYU Medical Center, New York, NY, USA) and were maintained on C57BL/6 genetic background (The Jackson Laboratory, Bar Harbor, ME, USA).The fractalkine receptor CX3CR1 is highly expressed in microglia and macrophages (Jung et al., 2000).As CX3CR1 +/eGFP mice express enhanced green fluorescent protein (eGFP) under the Cx3cr1 promoter, microglia and macrophages can be visualized using fluorescent microscopy.Animals were housed in controlled conditions of temperature and hygrometry with a 12 h light/dark cycle and ad libitum access to food and water.
Lateral hemisection of the spinal cord was performed as previously described by Noristani et al. (2015).Briefly, mice were anesthetized using 1%-1.5% isoflurane gas (Vetflurane ® , Virbac, France) and eye gel was applied to the cornea.A midline incision of the back skin was done followed by separation of the underlying muscles.The vertebral lamina was removed at Thoracic 9 level (T9) and the dura mater was locally incised to give access to the spinal cord.
Lateral hemisection was performed using a micro-scalpel (10315-12, FST, Heidelberg, Germany).Sutures were then applied to muscles and skin over the injury site.Mice were monitored for 2 h after the procedure before returning to their home cage.Bladders were emptied by hand twice every day until full recovery of sphincter control.A total of 57 mice underwent SCI in this study.Animals were sacrificed at 1, 3, 7, 14, 42, or 84 dpi for tissue analysis.An uninjured group of three mice was age-matched to the 84 dpi group (180 days old).

| GW2580-delivery
The CSF1R inhibitor GW2580 was used to inhibit microglia proliferation during the chronic stage of SCI (150 mg/kg/day for each animal).
As previously described by Gerber et al. (2018), 0.1% GW2580 (LC laboratories, Woburn, MA, USA) was added to the mice regular chow (A04, maintenance diet, SAFE diets, Augy, France).Briefly, chow was mixed with water to obtain a dough, allowing GW2580 incorporation.Pellets were then reconstituted and dehydrated for 48 h at 37 C. GW2580 chow was given to injured mice during 1 week, either between 28 and 35 dpi or between 70 and 77 dpi.Injured untreated (control) mice received the same processed food without addition of GW2580.

| Ex vivo diffusion-weighted magnetic resonance imaging
Dissected spinal cords were stored in 1% PFA until ex vivo MRI acquisition.For acquisition, samples were immerged in Fluorinert FC-770 liquid (3M™ Electronic Liquids, Saint Paul, USA) in a glass tube surrounded by a custom-made solenoid coil dedicated to spinal cord investigations (Coillot et al., 2016;Noristani et al., 2018).The coil was placed in a 9.4 T apparatus (Agilent Varian 9.4/160/ASR, Santa Clara, CA, USA) associated with a VnmrJ Imaging acquisition system (Agilent, USA).
Segmentations were done manually using Myrian Software (Intrasense, Montpellier, France), as described previously (Poulen, Aloy, et al., 2021).Longitudinal apparent diffusion coefficient (LADC) was quantified on a 2 cm segment centered on the lesion site.Quantifications were Diffusion gradients were applied in three directions including the rostro-caudal axis and two directions perpendicular to the spinal cord.
Acquisitions without applying diffusion gradient (Gs = 0 G/m À1 ) were done on the same images.
Cryosections from mice spinal cords were washed for 10 min, then incubated 20 min in 20 mM lysine (Thermo Fisher Scientific, Waltham, USA).Sections were washed 2Â 10 min and blocked during 2 h in a blocking buffer containing 1% bovine serum albumin and 0.1% Triton X-100 (both from Sigma Aldrich, Darmstadt, Germany).
Sections were incubated with the primary antibodies for 48 h at 4 C.Sections were then rinsed 3 Â 10 min and incubated with the corresponding fluorescent secondary antibody for 2 h at room temperature, then rinsed 3 Â 10 min.Mounting was done with fluorosave (Dako, Glostrup, Denmark).For BrdU and eGFP co-labeling, a preincubation step of 30 min in 2 N hydrogen chloride (HCl) was added for DNA denaturation, followed by 3 Â 10 min washes in 0.1 M sodium borate at pH 8.5 (Sigma Aldrich, Darmstadt, Germany) before carrying on with the immunofluorescence protocol described above.
For DAPI (Sigma Aldrich, Darmstadt, Germany, μg/mL) staining, 10 min incubation was done at the end of immunohistochemistry protocol, then rinsed 2 Â 10 min in PBS.
Secondary antibodies used were Alexa 594 donkey anti-rabbit, Alexa 594 donkey anti-rat and Alexa 488 goat anti-chicken (1:1000, Life Technologies, Carlsbad, USA for the three latter).

| Fluorescent microscopy
Axio Imager 1 microscope (Zeiss, Oberkochen, Germany, lens Â20) was used to acquire 12 fields per section at 1.89 and 3.15 mm, rostral and caudal to the epicenter.Equivalent locations were acquired in uninjured animals.The same exposure settings were kept for all acquisitions.Ki67 + (or BrdU + ) cells were manually quantified on separated fluorescent emission (Texas Red in red) and density was determined.
Proliferative microglia (Ki67 + eGFP + , or BrdU + eGFP + ) were counted on the merged emissions (GFP in green and Texas Red in red) and the density determined.Then the density of Ki67 + eGFP + over the density of Ki67 + ratio has been calculated (similar quantification method was used for BrdU + ).Coherent, France) as previously described (Poulen, Aloy, et al., 2021;Poulen, Gerber, et al., 2021).We imaged axial spinal cord sections (14 μm) at three spinal cord levels (lesion epicenter, 3.15 mm rostral and 3.15 mm caudal to the lesion).At each level, three images (dorsal, medial, and ventral) were taken in the lateral white matter, both ipsilateral and contralateral to the lesion side.

| Coherent anti-stokes Raman scattering
Pictures were taken at 14, 42, and 84 dpi.For uninjured animals, images at equivalent locations were taken.A Â20 water immersion lens (W Plan Apochromat DIC VIS-IR) with the following characteristics was used: 1024 Â 1024 pixels frame size, scan speed of 8 (Pix-elDwell 1.27 μs/scan) and a Â3.5 zoom.CARS excites the lipids CH 2 vibrational mode at 2845 cm À1 (Mytskaniuk et al., 2016).Excitation wavelengths were 836 and 1097 nm (synchronized Tisaphire and OPO, respectively) and the signal was detected at 675 nm (filter from 660 to 685 nm).Pictures are a stack of 3 μm (three slices).Density of intact myelin sheaths was manually quantified by an experimenter blind to the conditions as previously described (Poulen, Gerber, et al., 2021).For some acquisitions, eGFP signal was also acquired to observe microglia/macrophages location relative to myelin fibers.

| THUNDER imager 3D
For mosaics of the sections and fluoromyelin, images acquisitions were performed on a THUNDER imager 3D (Texas Red filter, Leica, Wetzlar, Germany, lens Â20 and Â63, respectively) at 1.89 mm caudal from lesion epicenter for injured animals and at an equivalent location for uninjured animals as previously described (Bringuier et al., 2023;Poulen, Aloy, et al., 2021).
For myelin fibers quantification, in each animal, a 5 μm-thick/24 slices stack (z-step: 0.21 μm) of a 200 μm Â 200 μm field was acquired in the lateral funiculus on both sides.Stacks were cleared using the THUNDER imager 3D Large Volume Computational Clearing process and a single-slice field (mid-stack) of 40 μm Â 40 μm was then exported for quantification.Intact myelin fibers were counted by an experimenter blind to the conditions on ImageJ software (National Institutes of Health, USA).Moreover, for quantification of phagocytic microglia engulfing myelin debris eGFP signal was also acquired.This quantification was performed directly on the merged channels (5 μmthick stacks; GFP in green and Texas red in red) to assess the internalization of myelin in the three dimensions.

| Semi-automated microglia morphology analysis
Two fluorescent microscopy images taken in the lateral white matter 1.89 mm caudal to the lesion site were used for morphological analysis of microglia for each animal.Two home-developed macros of Fiji (ImageJ, National Institutes of Health, USA; Schindelin et al., 2012) were applied.
The "soma" macro automatically segmented the cell bodies by saturating the signal, and treating the images with auto-threshold "Minimum" (histogram is iteratively smoothed using a running average of size 3, until there are only two local maxima).To eliminate left little processes, a morphological opening operation was applied.The size of each cell body was then computed with "Analyze Particles" with a 50 pixels minimum size.This resulted in automatic detection and measure of all somas of microglia in the treated images.Packed cells were discarded from the analysis by applying a maximum size threshold of 300 μm 2 .Leftover cell processes were also discarded from the analysis, by applying a minimum size threshold of 16.7 μm 2 based on previous studies (Kozlowski & Weimer, 2012).
The skeletonization macro was used on five randomly selected cells in each image."Tubeness" plugin (part of Fiji) was applied as a preprocessing step.Each cell was then segmented by auto-threshold "Triangle" (geometric method).This was followed by skeletonization using "Skeletonize (2D/3D)" plugin and measure using "Analyze Skeleton (2D/3D)" plugin.This resulted in automatic measure of the longest path in the skeleton of each selected cell (representative of microglial processes length).Cells with no processes were discarded from the analysis by applying a minimum length threshold of 3 μm.

| Semi-automated myelin analysis
Eight fluorescent microscopy images taken in the lateral white matter 1.89 mm caudal to the lesion site (four ipsilateral and four contralateral to the lesion) from Fluoromyelin stained spinal cord were used for morphological analysis of myelin for each animal.A macro was developed in Fiji (Schindelin et al., 2012) to quantify myelin fibers parameters (g-ratio, myelin thickness).After manually pointing inside all myelin fibers (corresponding to the axon) a threshold using the algorithm of Phansalkar (Phansalkar et al., 2011) is applied to the image (local threshold).It creates a binary image where the surrounding myelin appears in white.The Wand tool is used to set up the "black reference" corresponding to the manually selected point.Constant black region corresponds to the axon that is further measured.The black region (axon, area 1) is then enlarged to overlap the surrounding white area (myelin, area 2) until it meets another black region in the vicinity (that corresponds either to another axon or to background).
Myelin is then defined as the subtraction of area 1 to area 2. If myelin is not perfectly segmented, the user can validate or not the selected fiber (Figure 6).

| Manual microglia morphology analysis
Images used for semi-automated microglia morphology analysis were further manually assessed.Based on previous described criteria (Franco-Bocanegra et al., 2021), 10 randomly selected microglia (including the five that were analyzed with the macro) per photograph were classified into three morphological categories; homeostatic (Figure S2A), intermediate (Figure S2B), and amoeboid microglia (Figure S2C).

| Statistics
Statistical analysis was done using Graphpad Prism (Graphpad software 5.03, USA).Differences were significant for p ≤ .05. Results are expressed as mean ± standard error of the mean (SEM).For timepoints comparisons of cell proliferation and morphology, one-way ANOVA with Tukey's post hoc tests were used.One-way ANOVA was performed to compare the effect of time on the variable.Tukey's post hoc tests were done for multiple comparisons.Tukey's post hoc tests significance is presented on the graphs always using the # symbol.
Student's t-tests with Welch's correction were used for all unpaired comparisons (between groups), while paired t-tests were used for all paired comparisons (ipsilateral to contralateral, within the same group).Levels of significance are always presented with the * symbol.
To analyze the repartition of cell body areas, Kolmogorov-Smirnov tests were done on the online tool "Quest Graph™ Kolmogorov-Smirnov (K-S) Test Calculator" (AAT Bioquest, Inc., Pleasanton, CA, USA).
Uninjured mice displayed sparse surveilling microglia (Figure 2a, insets 1 and 2).Consistently with the peak of microglia proliferation (Figure 1f), the density of microglia increased at 3 dpi (Figure 2b).At chronic stages (42 and 84 dpi; Figure 2c,d, respectively), the density of microglia displaying an amoeboid shape progressively increased on the ipsilateral side of the injured spinal cord (Figure 2c,d, inset 1).
We therefore quantified morphological changes of microglia over time.We selected 3 dpi (Figure 2b), since it corresponds to the peak of proliferation, and chronic stages (42 and 84 dpi; Figure 2c,d, respectively) because we ultimately want to investigate the effect of a transient inhibition of microglia proliferation at chronic stage after SCI.
Quantifications were done using two home-developed automated macro in Fiji at 1.89 mm caudal to the lesion on both the ipsilateral and contralateral white matter and include at least 20 microglia per analysis for cell morphology assessment (Figure 2e) and at least 60 microglia soma (15 per animal) for cell body measurement (Figure 2h).SCI induced a decrease in the mean of longest microglia process path at alltime points ipsilateral to the lesion (uninjured: 63.4 ± 5.7 μm; 3 dpi: 25.0 ± 2.2 μm; 42 dpi: 45.8 ± 5.1 μm; 84 dpi: 35.9 ± 4.3 μm, one-way ANOVA, overall significance ###p ≤ .001,F = 10.44; Figure 2f).
Together, these results show that at chronic stage after SCI microglia are highly dynamic in terms of proliferation and morphology, microglia displaying a persistent amoeboid phenotype at chronic phase ipsilateral to the lesion.
F I G U R E 2 Microglia morphology at chronic stages after spinal cord injury.Fluorescent micrographs of axial thoracic spinal cord sections from CX3CR1 +/eGFP transgenic mice showing microglia (eGFP + , green) in an uninjured spinal cord (a), 3 dpi (b), 42 dpi (c), and 84 dpi (d).Analyzed images (Â20) were taken 1.89 mm caudal to the lesion epicenter.For each image, two insets show representative microglia from the ipsilateral (1) and the contralateral side (2) of the spinal cord.Example of single cell region of interest (ROI) used to generate a mask and to skeletonize the cell (e).Histograms displaying quantification of the longest path among all protrusions of the skeletonized microglia, at all time-points ipsilateral (f) and contralateral (g) to the lesion site.Example of micrograph used to generate a cell body mask and selection of soma from microglia with (arrows) and without (arrowheads) protrusion (h).Histograms displaying quantification of microglial body area at all time-points ipsilateral (i) and contralateral (j) to the lesion site.Number of mice: 3-6 per group.Number of microglia analyzed: at least 20 per group for morphological measurement and at least 60 per group for cell body measurement.Thresholds: 3 μm (f and g) and 16.7 μm 2 (i and j).Data are expressed as mean per group ± SEM.Statistics; one-way ANOVA followed by Tukey's multiple comparison post hoc tests (relative to uninjured), #p ≤ .05,##p ≤ .01,###p ≤ .001;ns: not significant (f, g and i, j).Scale bars: 500 μm (a-d), 20 μm (e), and 50 μm (insets 1 and 2 from a to d, h).
3.2 | Myelinated fibers are modified for at least 3 months after SCI Acute injury-induced demyelination is followed by remyelination that commences within the first few weeks after SCI and persists for months (for review, see Pukos et al., 2019).Due to its high content of lipids, myelin integrity can be assessed with CARS microscopy (Poulen, Gerber, et al., 2021).Thus, using CARS, we quantified the density of spinal cord intact myelinated fibers (Figure 3j) at three time points after SCI (14, 42, and 84 dpi; Figure 3) on axial sections at three F I G U R E 3 Coherent anti-stokes Raman scattering (CARS) microscopy a label free method to quantify the density of intact myelin fibers.Schematic drawing of a hemisected spinal cord, images were taken on the ipsilateral and contralateral sides of the lesion at the epicenter as well as 3.15 mm rostral and caudal to the lesion (i).Representative CARS axial photographs of myelin obtained from CX3CR1 +/eGFP mouse spinal cord at 84 dpi contralateral (a-c) and ipsilateral (d-f and o-q) to the lesion epicenter.Representative images taken at 84 dpi rostral (a, d) and caudal (c, f) to the lesion as well as at the epicenter (b, e).Curves displaying quantification of intact myelinated fibers density on the contralateral (g) and the ipsilateral (h) side of the lesion.For all time-points (uninjured, 14, 42, and 84 dpi) quantification of intact myelin fibers was done at the epicenter, rostral and caudal to the lesion site.Three images per location per animal were quantified (18 pictures per mouse) and the mean value per location per animal was taken.Representative image of an intact (j) and degraded myelin (k).Representative images taken in an uninjured condition (l-n) and 84 dpi on the ipsilateral side at the epicenter (o-q).Photographs showing myelin (CARS, l and o), microglia (eGFP, m and p) and merge (n and q).In all images, arrows point to intact myelinated fibers while arrowheads point to degraded myelin.Number of mice: three per group.Data are expressed as mean per group ± SEM.Statistics; unpaired t-test with Welch correction, *p ≤ .05,**p ≤ .01,***p ≤ .001(g and h).Scale bars: 20 μm (a-f and l-q) and 5 μm (j and k).
different levels (epicenter, 3.15 mm rostral and caudal to the lesion site; Figure 3i).Myelin fibers were also quantified in uninjured mice (Figure 3g,h,l-n).In all locations and groups, we acquired three images in the white matter (dorsal, medial, and ventral) ipsilateral and contralateral to the lesion side (Figure 3i).We quantified intact myelinated fibers (Figure 3j and arrows in all panels).Not surprisingly, fiber density was strongly decreased at the epicenter (Figure 3g,h) in all injured groups, where numerous degraded myelin were present (Figure 3k and arrowheads in all panels).Contralateral (Figure 3a-c) to the lesion, both rostral (Figure 3a) and caudal (Figure 3c) to the epicenter (Figure 3b), the density of fibers decreased at 14 and 42 dpi but returned to uninjured value at 84 dpi (Figure 3g).Ipsilateral (Figure 3d-f) to the lesion, a decrease in the density of intact fibers was seen at all time-points rostral to the lesion (Figure 3d,h).Interestingly, myelinated fiber density returned to uninjured value at 84 dpi caudal to the lesion (Figure 3f,h).Moreover, taking advantage of eGFP expression in microglia/macrophages in CX3CR1 +/eGFP mice, we confirmed that at epicenter, ipsilateral to the lesion, microglial morphology remained activated (Figure 3p,q) as compared with the uninjured context (Figure 3m,n).
Together, these results show that remyelination processes are dynamic at chronic stage after SCI.

| CSF1R inhibition at chronic stages after SCI reduces microglia proliferation
We have shown that microglia are still highly dynamic at chronic stage after SCI, we thus investigated whether a transient 1-week depletion of microglia proliferation at chronic stages would represent an attractive translational approach.We orally administered GW2580 (a CSF1R inhibitor that selectively inhibits microglia proliferation) for 1 week to CX3CR1 +/eGFP mice either 28 or 70 days after a lateral T9 hemisection of the spinal cord (Figure 4d).Daily intraperitoneal injection of BrdU was done concomitantly with the GW2580-delivery.We then examined the effect of GW2580-delivery on microglial proliferation using immunohistochemistry (Figure 4a-c) 1 week after the end of the GW2580 oral delivery (i.e., 42 and 84 dpi; Figure 4d).
We then studied whether a transient GW2580-delivery at chronic stage after SCI has an effect on microglia morphology.As for the time course analysis, we assessed two morphological parameters (i.e., longest microglia process path Figure 2e) and cell body area Figure 2h) at 1.89 mm distance caudal to the lesion (Figure 5).At 42 dpi, the mean of longest microglia process was similar on both sides of the spinal cord (ipsilateral and contralateral in control and GW2580-treated mice (Figure 5a, insets 1 and 2 and Figure 5c).At 84 dpi, controls displayed similar longest microglia process values on both sides conversely to treated mice that presented longer processes on the contralateral side as compared with the ipsilateral side of the spinal cord (57.60 ± 6.51 μm vs. 33.42± 3.34 μm, respectively, p ≤ .01; Figure 5b, insets 1 and 2 and Figure 5d).No difference between control and treated animals were seen on either side of the lesion at 42 (Figure 5c) and 84 dpi (Figure 5d).At 42 dpi (Figure 5e) and 84 dpi (Figure 5f), microglia cell body surface was lower on the contralateral side in control and GW2580-treated groups.No difference between control and treated animals were observed on either side of the lesion at 42 (Figure 5e) and 84 dpi (Figure 5f).We analyzed the distribution of the cell bodies area in control and GW2580-treated groups at 42 (Figure S4A,C) and at 84 dpi (Figure S4B,D).Kolmogorov-Smirnov test indicated that the repartition of cell bodies area was different between groups at 42 dpi (p ≤ .05, Figure S4E,E') but not at 84 dpi (Figure S4F,F').Manual classification of the repartition of microglia morphologies (Figure S2G) show that the proportion of homeostatic microglia decrease after SCI without difference between control and treated groups (Figure S2H).Proportion of amoeboid microglia increased up to 42 dpi regardless of groups; 84 dpi proportion of amoeboid microglia returned to baseline value in the control group only, but no difference between control and treated animals were seen at either 42 or 84 dpi (Figure S2I).
Together, these results suggest that the GW2580-mediated inhibition of microglia proliferation at chronic stages after SCI modulates microglial response.

| CSF1R inhibition at chronic stages after SCI modifies myelinated fibers
To study the consequence of GW2580-delivery at chronic stages on lesion size and spinal cord microstructure, in particular myelin, we used ex vivo DW-MRI.Ex vivo DW-MRI did not reveal a difference between treated and control groups at 42 or 84 dpi neither in lesion extension (Figure S5A,B) nor in LADC (Figure S5C,D).To further investigate putative consequences of GW2580-delivery on myelin, we used fluoromyelin staining and quantified the density of myelinated fibers (Figure 6).At 42 dpi, myelin fibers density was decreased by the injury (*p ≤ .05,30,833 ± 2256 and 18,125 ± 1499 fibers/mm 2 for uninjured and injured, respectively), conversely myelin fibers density returned to uninjured values in the GW2580-treated group (23,938 ± 830 fibers/mm 2 ; Figure 6a).At 84 dpi, myelin fibers density returned to uninjured value in both injured groups (29,453 ± 4976 and 31,188 ± 4245 fibers/mm 2 for control and GW2580-treated, respectively, Figure 6a).Comparison of myelin fiber density between ipsilateral and contralateral sides of the spinal cord at 84 dpi highlights a difference only in the GW2580-treated group with a higher density F I G U R E 4 Effect of GW2580-delivery at chronic stages after spinal cord injury on microglia proliferation.Immunofluorescence microscopy of thoracic spinal cord sections showing microglia (a, eGFP + , green), proliferative cells (b, BrdU + , red), and proliferative microglia (c, eGFP + BrdU + , yellow, arrows) in GW2580-treated CX3CR1 +/eGFP mouse at 84 dpi.Scheme of the two chronic GW2580 oral administration protocols (d).For cell quantification, images were taken in the grey and the white matters, both rostral and caudal at two distances (3.15 and 1.89 mm) to the lesion epicenter.Quantification of the density of proliferative microglia (eGFP + BrdU + ) per mm 2 of tissue in spinal cord sections from uninjured as well as 42 and 84 dpi in control and GW2580-treated mice (e).Direct comparisons of the density of proliferating microglia (eGFP + BrdU + /mm 2 ) between the ipsilateral and contralateral sides of the spinal cord in control and GW2580-mice at early (42 dpi) and late (84 dpi) chronic stages after injury (f).Percentage of proliferative microglia relative to proliferative cells (eGFP + BrdU + /BrdU + cells) in spinal cord sections from uninjured as well as 42 and 84 dpi in control and GW2580-treated mice (g).Direct comparisons of the percentage of proliferative microglia (eGFP + BrdU + / BrdU + cells) between ipsilateral and contralateral sides of the spinal cord in control and treated mice at early (42 dpi) and late (84 dpi) chronic stages after injury (h).Data from three uninjured mice are shown with plain black circles (e, g).Number of injured mice: 4-6 per group.Data are expressed as mean per mouse ± SEM.Statistics; unpaired t-test with Welch correction, *p ≤ .05,**p ≤ .01,***p ≤ .001;ns: not significant (e-h), paired t test (ipsi vs. contra comparisons), *p ≤ .05,**p ≤ .01;***p ≤ .001;ns: not significant (f, h).Scale bar: 50 μm (a-c).
To further analyze the effect of GW2580-delivery on myelin, we measured myelin thickness.Quantifications were done using a home-developed automated macro in Fiji.Four images were taken at 1.89 mm caudal to the lesion in both the ipsilateral and contralateral white matter and analysis include at least 200 myelinated fibers per condition.Only intact fibers were analyzed (Figure 6g).42 dpi myelin was thicker in the GW2580-treated group as compared with the control group.No difference was observed at 84 dpi (Figure 6h).Lastly, we investigated consequences of the GW2580-delivery on the phagocytose of myelin debris by microglia.We therefore quantified phagocytic microglia engulfing myelin debris (Figure 7a) in control (Figure 7d-g) and GW2580-treated CX3CR1 +/eGFP mice (Figure 7h-k).
Together, these results suggest that GW2580-mediated inhibition of microglia proliferation at chronic stage after SCI modulates myelination processes and decreases the number of phagocytic microglia.

| DISCUSSION
Together, our results show that at chronic stage after SCI microglia are still highly dynamic in terms of proliferation and morphology.We then show that a transient inhibition of microglia proliferation using GW2580-delivery at chronic stages reduces microglia proliferation, modifies microglia morphology toward homeostatic phenotype, and changes outcomes on myelin dynamics.( Menassa et al., 2022).In mice and humans, a proliferation wave also occurs during postnatal development (Menassa et al., 2022;Nikodemova et al., 2015), then microglia number remains stable across lifespan (Askew et al., 2017), displaying low (<1%) basal proliferation rate (Lawson et al., 1992;Tay et al., 2017).Nonetheless, microglia turnover has been estimated to 28% per year in humans, suggesting complete renewal of the population during a lifetime (Reu et al., 2017).In physiological contexts, microglia are highly dynamic and constantly surveil the environment through their branched processes (Nimmerjahn et al., 2005).
Microglia responses to pathologies are diverse, including migration to the lesion site (Davalos et al., 2005;Nimmerjahn et al., 2005), morphological modifications, phagocytosis, and proliferation (for review see Garden & Möller, 2006;Sierra et al., 2019).Microglia proliferation occurs during chronic phases of CNS diseases such as infection (Trzeciak et al., 2019), neurodegeneration (Langmann, 2007;Olmos-Alonso et al., 2016), prion diseases (De Lucia et al., 2016;G omez-Nicola et al., 2014), demyelination (Hagan et al., 2020;Nowacki et al., 2019), andepilepsy (Di Nunzio et al., 2021).After SCI in mice, an early microglia proliferation has been reported during the first week after injury, peaking between 3 and 7 dpi (Bellver-Landete et al., 2019;Lytle & Wrathall, 2007;McDonough et al., 2013;Noristani et al., 2017; for review see Perez et al., 2021), but to our knowledge, this is the first report of a persistent proliferation at chronic stage of SCI, that is, 3 months after traumatism.Indeed, using Ki67 and BrdU immunodetection, we found an increase in microglia proliferation at 84 dpi as compared with uninjured; only BrdU staining additionally revealed an increased proliferation at 42 dpi.This discrepancy may be due to the difference between Ki67 and BrdU markers, the first labeling cells in the G1, S, G2, and M phases and the second labeling cells during the S phase.Combination of the two methods therefore shows that microglia proliferation persists at least up to 84 dpi.Moreover, chronic microglia proliferation is associated with a persistent amoeboid morphology ipsilateral to the lesion.

| Inhibition of microglia proliferation at chronic stage after SCI modifies microglia response
Few studies investigated the effect of the inhibition of microglia proliferation at chronic disease stage.In a mouse model of Alzheimer disease, a 3-month GW2580 treatment starting at 6 months of age leads to a shift of microglia profile from an inflammatory to an anti-inflammatory phenotype (Olmos-Alonso et al., 2016).This was correlated with an improvement in memory as well as a reduction of synaptic degeneration (Olmos-Alonso et al., 2016).In the SOD1 G93A mouse model of amyotrophic lateral sclerosis, GW2580-treatment between 8 and 16 weeks of age induces protective effects on muscle innervation, reduces microglia proliferation, diminishes motoneuron death (Martinez-Muriana et al., 2016).It results in a slow-down of disease progression and an increase in survival (Martinez-Muriana et al., 2016).In a mouse model of prion disease GW2580 administration between 14 and 18 weeks after infection restores neuronal differentiation and decreases neurogenesis (De Lucia et al., 2016).
Finally, in a chemically induced mouse model of epilepsy, GW2580-mediated inhibition of microglia proliferation during the early disease stage has no effect on status epilepticus and subsequent epilepsy but induces neuroprotection in the hippocampus (Di Nunzio et al., 2021).Conversely, spontaneous seizure is reduced when GW2580 is administered to chronic epileptic mice, that is, a 2 weeks treatment starting 72 days after induction of status epilepticus (Di Nunzio et al., 2021).This study demonstrated that microglia proliferation may play different roles over the course of epilepsy through contribution to neural cell death in the early phase and to seizure in chronic phases (Di Nunzio et al., 2021).In chronic SCI, we have shown that a prolonged GW2580-treatment, that partly covered chronic stage (starting 4 weeks prior to SCI and ending 6 weeks post-lesion) improves motor recovery (Gerber et al., 2018).Interestingly, when microglia proliferation is inhibited after the lesion, a prolonged 6 weeks treatment (Poulen, Bartolami, et al., 2021) does not further improve recovery as compared to a transient 1-week administration restricted to the acute phase following injury (Poulen, Aloy, et al., 2021).
In this study, we now show that inhibition of microglia proliferation initiated at two chronic stages after SCI modulates microglial response, presumably through limitation of secondary inflammation as reflected by microglia morphology.In a previous study using RNAseq analysis, we have shown that a transient 1-week GW2580-treatment, given just after SCI, leads to a down regulation in microglia of gene associated with cell proliferation, cell migration, inflammatory response, and immune response (Poulen, Aloy, et al., 2021).Moreover, comparison of genes that are deregulated in microglia by the injury (uninjured/SCI; Noristani et al., 2017) with genes that are deregulated in microglia by the GW2580-treatment (SCI-untreated/SCI-GW2580; Poulen, Aloy, et al., 2021) further substantiate the decrease in the inflammatory response of microglia.Further investigation is required to assess whether GW2580-delivery plays similar role at chronic stages.

| Inhibition of microglia proliferation at chronic stage after SCI modifies myelin dynamics
A recent study demonstrated persistent remyelination and demyelination processes at chronic stages after T9 spinal cord contusion (Pukos et al., 2023).Indeed, axons became wrapped with new myelin up to 6 months after SCI, with a peak at 3 months.Concomitantly, chronic demyelination was also observed up to 6 months (Pukos et al., 2023).In agreement with this, we show that remyelination processes occurs at chronic stage after SCI.Wallerian degeneration in the CNS is very slow and can take months to years (for review, see Vargas & Barres, 2007), therefore inhibition of microglia proliferation at chronic stage may modulate Wallerian degeneration and modify phagocytosis and thus myelin clearance.Our data suggest that GW2580 may act on remyelination dynamics since myelin thickness increase in the GW2580-group at 42 dpi and intact myelinated fibers density returns to baseline values by 84 dpi in the control group and at 42 dpi in the GW2580 one.
Our study demonstrates for the first time a prolonged and robust microglia proliferation at chronic stages after SCI.We then show that transient GW2580-delivery at late stages after lesion reduces microglia proliferation, modifies microglia morphology toward homeostatic phenotype, and modulates myelin dynamics.Therapeutical strategies at chronic stage to improve recovery are rarely studied.Thus, further studies to investigate whether a transient inhibition of microglia proliferation at chronic stage after SCI improves functional recovery will be of utmost importance in the aim of translation to the clinic.

AUTHOR CONTRIBUTIONS
Jean-Christophe Perez performed majority of the experiments, analyzed the data and contributed to the writing of the manuscript.Gaetan Poulen participated in the analysis of immunohistochemistry.
Maida Cardoso performed MRI acquisition.Hassan Boukhaddaoui participated in the design of CARS acquisition and analysis.Chloé Marie Gazard participated in the analysis of microglia morphology.
Gilles Courtand designed the macro for morphological analysis.

4. 1 |
Microglia proliferation persists at chronic stage after spinal cord injury Microglia, one of the most dynamic cell types of the central nervous system (CNS) repeatedly exhibit proliferative phase.After entering the CNS from the embryonic yolk sac, progenitors quickly amplify microglia numbers by waves of successive proliferation and death F I G U R E 5 Effect of GW2580-delivery at chronic stages after spinal cord injury on microglia morphology.Fluorescent micrographs of thoracic spinal cord sections from GW2580-treated CX3CR1 +/eGFP mice showing microglia (eGFP + , green) at 42 dpi (a) and at 84 dpi (b).For each image, two insets show representative microglia from the ipsilateral (1) and the contralateral side (2) of the spinal cord.Analyzed images (Â20) were taken 1.89 mm caudal to the lesion epicenter.Comparison of the longest path among all microglial protrusions between the ipsilateral and contralateral sides of the lesion at 42 dpi (c) and 84 dpi (d).Comparison of the microglial body area between the ipsilateral and contralateral sides of the lesion at 42 dpi (e) and 84 dpi (f).Number of mice: 4-6 per group.Number of microglia analyzed: at least 20 per group for longest path and 60 for cell body area.Thresholds: 3 μm (c and d) and 16.7 μm 2 (e and f).Data are expressed as mean per group ± SEM.Statistics; unpaired t-test with Welch correction, ns: not significant; paired t test (ipsi vs. contra comparisons) **p ≤ .01,***p ≤ .001;ns: not significant (c-f).Scale bar: 500 μm (a and b).

F
I G U R E 6 Effect of GW2580-delivery at chronic stages after spinal cord injury (SCI) on myelin density.Quantification of intact myelinated fibers density in uninjured mice, GW2580-treated and control mice at 42 and 84 days after SCI (a).Direct comparison of myelin fibers density between the ipsilateral and contralateral side of the spinal cord in control and GW2580-treated mice at 42 and 84 dpi (b).Fluorescent micrographs of thoracic spinal cord sections stained with fluoromyelin from control (c, d) and GW2580-treated CX3CR1 +/eGFP mice (e, f) showing myelinated fibers (arrows) at 84 dpi ipsilateral (c and e) and contralateral (d and f) to the lesion.Quantifications were done 1.89 mm caudal to the epicenter.Myelin thickness was quantified in intact fibers located 1.89 mm caudal to the lesion epicenter.For semi-automated quantification a mask was generated to identify fibers and myelin thickness was measured (g).Violin plots displaying quantification of myelin thickness in all groups ipsilateral and contralateral to the lesion site (h).Data from three uninjured mice are shown with plain black circles (a).Number of injured mice: 4-5 per group.Data are expressed as mean per mouse ± SEM (a, b) or as mean per group ± SEM (h).Statistics: unpaired t test with Welch correction, *p ≤ .05(a, h) and paired t test, *p ≤ .05(b).Scale bar: 10 μm (c-f) and 5 μm (g).

F
I G U R E 7 GW2580-delivery at chronic stages after spinal cord injury decreases microglial phagocytosis of myelin.Fluorescent micrographs of axial sections of thoracic spinal cord showing microglia (eGFP + , green), myelin (Fluoromyelin, red), and merge.Orthogonal view showing myelin debris engulfed by phagocytic microglia (a).Quantification of phagocytic microglia density in the white matter caudal to the lesion epicenter in uninjured mice and at 42 and 84 dpi in GW2580-treated and control mice (b and c).Quantifications were done ipsilateral (b) and contralateral (c) to the lesion site at 1.89 mm caudal to the epicenter.Merged micrographs of control (d-g) and GW2580-treated CX3CR1 +/eGFP mice (h-k) showing phagocytic microglia (arrows) at 42 dpi (d, e, h, i) and 84 dpi (f, g, j, k).Images were taken in the white matter on the ipsilateral (d, h, f, j) and the contralateral (e, i, g, k) sides of the lesion.Data from three uninjured mice are shown with plain black circles (b and c).Number of injured mice: 4-5 per group.Data are expressed as mean per mouse ± SEM.Statistics: unpaired t test with Welch correction, *p ≤ .05,**p ≤ .01(b).Scale bars: 10 μm (a) and 20 μm (d-k).