Acute and late administration of colony stimulating factor 1 attenuates chronic cognitive impairment following mild traumatic brain injury in mice

Abstract

Traumatic brain injury (TBI) is a leading cause of long-term neurological disability. Currently there is no effective pharmacological treatment for patients suffering from the long-lasting symptoms of TBI. We recently discovered that colony stimulating factor 1 (CSF1), an essential regulator of macrophage homeostasis, is neuroprotective and reduces neuroinflammation in two models of neurological disease in mice. Here we used a mouse model of repetitive mild TBI (mTBI) to examine whether CSF1 would attenuate cognitive deficits and improve pathological outcomes in two paradigms. In the acute paradigm, a single bolus treatment of CSF1 administered 24 hours after injury significantly reduces memory impairment and astrocyte reactivity assessed 3 months later. In the chronic paradigm, the mice were tested 3 months after mTBI when they showed cognitive deficits. The mice were then randomly assigned to receive CSF1 or PBS (as control) treatment. After one month of treatment, the PBS-treated mice remained cognitively impaired, but the CSF1-treated showed significant improvements in cognitive function. RNA-seq and Ingenuity Pathway Analysis reveals CSF1 treatment alters cognition- and memory-related transcriptomic changes and pathways. The results of this study show that acute as well as delayed CSF1 treatment attenuate chronically impaired cognitive functions and improve pathological outcomes long after mTBI. The wide therapeutic time window of CSF1, together with the fact that CSF1 is approved for human use in clinical trials, strongly supports the potential clinical usefulness of this treatment in patients with mTBI.

1. Introduction

Traumatic brain injury (TBI) is the leading cause of mortality in young adults and a major cause of death and disability across all ages in all countries (Maas et al., 2017). Moderate and severe types of TBI can have long lasting or even lifetime effects. The mild type of TBI (mTBI, also referred to as concussion) is the major type and accounts for more than 75% of all TBIs (Maas et al., 2017). Although most symptoms of mTBI resolve spontaneously within days or weeks of the injury, in 10–20% of mTBI victims some functional deficits persist, with recent estimates suggesting that as many as 44–50% of mTBI patients experience long term symptoms (Ruff, 2005; Vasterling et al., 2009). There are growing concerns about the consequences of repetitive mTBI, because recurrent brain injuries, even when mild, may induce cumulative effects and result in prolonged cognitive impairment, mood disorders, and behavioral problems (Vasterling et al., 2009). Repeated concussion has also been associated with chronic traumatic encephalopathy (CTE), a neurodegenerative disorder with progressive impairments of memory and cognition, as well as depression, anxiety, and motor abnormalities (Stern et al., 2011).

Common symptoms developed after mTBI include physical, emotional, behavioral, and cognitive, such as headache, sleep disturbance, disorders of balance, fatigue, irritability, and memory and concentration problems (Marshall et al., 2012). Amongst these, cognitive impairment is paramount, because it is the leading cause of TBI-related disability and is therefore associated with worse health-related quality of life after TBI (Gorgoraptis et al., 2019; Rabinowitz and Levin, 2014). Impairment can occur in numerous cognitive domains, including executive function, learning and memory, attention and processing speed (Rabinowitz and Levin, 2014). Currently, there are no effective pharmacological treatments for patients suffering from TBI-induced long-lasting cognitive deficits and therefore there is an urgent need to develop therapeutic options.

Many preclinical studies assessing efficacy of therapeutics for mTBI have been reported, and current targets focus specifically on reducing inflammation, oxidative stress, axonal injury, and associated neurodegenerative-like pathology (Fehily and Fitzgerald, 2017). However, despite the large number of promising neuroprotective agents identified in preclinical studies, none has been successfully translated into therapeutic intervention in humans (DeWitt et al., 2018). This is at least in part due to the fact that the underlying molecular and biochemical mechanisms of mTBI-associated impairments of memory and cognition still remain elusive. Recent studies have revealed a critical role of transcription factors in long-term synaptic plasticity and memory formation and targeting these transcription factors offers novel therapeutic opportunities (Alberini, 2009; Alberini and Kandel, 2014). Among the hundreds of molecules and transcription factors implicated in memory, cAMP response element-binding (CREB) protein has been shown to be critically involved in long term synaptic plasticity and long term memory (Kandel, 2001) and has been actively studied in TBI (Sen, 2019). It has been reported previously that moderate to severe TBI results in transient activation of CREB signaling followed by chronic signaling deficits in the hippocampus (Atkins et al., 2009; Dash et al., 1995; Hu et al., 2004). We recently developed a mouse model of repetitive mTBI, based on a controlled cortical impact (CCI) device (Luo et al., 2014). In this model, mice received three repetitive mTBIs showed a significant impairment in spatial learning and memory when tested 2–6 months after injury. A robust astrocyte reactivity and reduced levels of CREB phosphorylation were observed upon post-mortem pathological examinations (Luo et al., 2014). Interestingly, reduced levels of CREB phosphorylation were also recently observed in a weight drop model of repetitive mTBI (Rehman et al., 2019). Together results from these studies support that CREB signaling is disrupted in mTBI and that restoring CREB signaling may be a therapeutic strategy for TBI patients.

We recently discovered that Colony stimulating factor 1 (CSF1), a key growth factor in the hematopoietic system, exerts prominent neuroprotective and anti-inflammatory effects in kainic acid-induced acute excitotoxicity injury when administered systemically after injury (Luo et al., 2013). Importantly we showed that CSF1 promotes CREB phosphorylation in primary neuronal culture in vitro and restores levels of CREB phosphorylation in neurons in the hippocampus of Alzheimer model mice or after kainic acid injury (Luo et al., 2013). Therefore we hypothesized that systemic treatment with CSF1 could remedy mTBI-induced impairments in cognitive function and associated pathological changes. To test the clinical potential of CSF1, we tested CSF1 treatment in two paradigms, with CSF1 administered 24 hours or 3 months after injury.

2. Materials and Methods

2.1. Mice.

Two months old male, wildtype C57BL/6J mice were purchased from The Jackson’s Laboratory (Stock No:000664). Transgenic mice expressing firefly luciferase under the glial fibrillary acidic protein (GFAP) promoter (GFAP-luc) have been described previously (Luo et al., 2013; Luo et al., 2008; Luo et al., 2007; Luo et al., 2014). All mice were kept under a 12-hour light-dark cycle with ad libitum access to food and water. The animal studies were approved by the VA Palo Alto IACUC and performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

2.2. Closed-Head Model of repetitive mTBI.

The procedures of closed head mTBI and sham injury were previously described (Luo et al., 2014; Sahbaie et al., 2018; Zhang et al., 2016). Briefly, a benchmark stereotaxic impactor (MyNeurolab, St. Louis, MO) actuator was mounted on a stereotaxic frame (David Kopf Instruments, Tujunga, CA) at a 40° angle with a 5-mm impactor tip. After isoflurane anesthesia induction, mice were placed in a foam mold held in prone position on the stereotaxic frame and maintained under anesthesia for the duration of the procedure. The stereotaxic arm was adjusted so that the head impact was at a fixed point relative to the right eye and ear, corresponding to the S1 somatosensory cortex (Figure 1A). The impact delivered by the device to the head was 5.0 m/s with a dwell time of 0.2 seconds and impact depth of 5 mm. After impact, the mice were recovered from anesthesia on a warming pad before returning to their home cages. For repetitive injuries, three identical impact procedures were performed at an interval of 24 ± 1 hours. For sham injury, the previous procedure was performed except that the impact device was discharged in the air.

Figure 1. A single bolus of CSF1 administered 24 h post-injury attenuates associative fear memory deficits 3 months later.

(A) Schematic representation of the mouse brain viewed from above, showing the area of impact (red circle) relative to the S1 somatosensory cortex (grey area) and brain tissue sampling (dashed rectangle) used for immunohistochemical analysis. (B) Experimental design. GFAP-luc mice were randomly assigned to 3 groups (n=5 mice/group, 3 males, 2 females): sham, repetitive mTBI treated with PBS (mTBI + veh), and repetitive mTBI treated with CSF1 (mTBI + CSF1). CSF1 (800 μg/kg) or PBS were administered 24 h after the last impact. Bioluminescence imaging (BLI) was performed at 1 day before the first impact (for baseline) and day 3 after the last impact (2 days after CSF1 injection). Three months later memory was assessed by contextual fear conditioning and brains were analyzed for astrocyte reactivity and CREB phosphorylation using immunostaining. (C) Bioluminescence imaging of astrocyte reactivity in GFAP-luc mice. Representative images show increased bioluminescence signal over baseline in the head after injury (middle panel), but not in sham mice (left panel), and mTBI-induced bioluminescence signal was reduced by CSF1 treatment. (D) Astrocyte reactivity was expressed as bioluminescence fold change over baseline. (E) Results of contextual and cued fear conditioning test. n = 5 mice/group. *, P < 0.05; **, P < 0.01, ANOVA and Tukey’s multiple comparison test. In (D, E), open symbols represent female mice and closed (filled) symbols represent male mice.

2.3. CSF1 treatment.

Recombinant human CSF1 (Catalog #: 216-MCC/CF, R&D Systems, Minneapolis, MN) was injected intraperitoneally at 800 μg/kg body weight (Luo et al., 2013). The same volume of PBS was used as a control treatment. For the acute paradigm, CSF1 was injected once at 24 hours after the last impact. For the chronic paradigm, CSF1 was injected 3 times a week for a month and during the behavioral tests.

2.4. Bioluminescence imaging.

In vivo bioluminescence imaging of GFAP-luc mice were performed with the In Vivo Imaging System (IVIS Spectrum; PerkinElmer, Waltham, MA) as previously described (Luo et al., 2013; Luo et al., 2008; Luo et al., 2007; Luo et al., 2014). Mice were injected intraperitoneally with 150 mg/kg D-luciferin (PerkinElmer) 10 min before imaging and anesthetized with isofluorane during imaging. Imaging signal was quantitated as photons/s/cm2/steridian (sr) using LIVINGIMAGE software (version 4.50) (Xenogen) and integrated over 3 min. Photons were obtained from a region of interest in the head which was kept constant in area and positioning. For longitudinal comparison, baseline imaging was performed 24 h before the first impact of mTBI was initiated. The mice were imaged again at day 3 after the last impact of mTBI (2 days after CSF1 treatment). Bioluminescence (and astrocyte reactivity) was expressed as fold induction over baseline levels.

2.5. Mouse memory tests.

The two trial (forced alternation) Y-maze was performed as previously described (Pluvinage et al., 2019). In brief, the test consisted of a 5-min training trial followed by a 5-min retrieval trail, with a 1-h inter-trial interval. For the training trial, one arm of the Y maze was blocked off and mice were allowed to explore the two open arms. One hour later, the mouse was again placed in the Y maze with all three arms open and a black and white pattern placed at the end of the novel arm. Between mice and trials, the maze was wiped with ethanol to remove odor cues. For analysis, video was analyzed by a blinded observer and both number of arm entries and time spent in each arm were quantified. Mice with less than two arm entries in the first minute of the retrieval trial were excluded from the analysis.

The fear-conditioning paradigm was performed as previously described (Luo et al., 2014; Pluvinage et al., 2019; Villeda et al., 2011). In brief, mice were trained to associate cage context or an audiovisual cue with an aversive stimulus (foot shock). The test was administered over two days. On day 1, mice were placed in a fear conditioning chamber and allowed to explore for 2 min before exposed to two periods of 30 s of paired cue light and 1,000-Hz tone followed by a 2-s foot shock (0.6 mA), with a 180-s interval. Shocks were delivered through the grid floor and were controlled by FreezeScan software (Clever Sys., Inc., Reston, VA, USA). On day 2, mice were subjected to two trials. In the first trial assessing contextual memory, mice were re-exposed to the same fear conditioning chamber containing the same context as the training day, but without administration of a CS or foot shock. Freezing behavior was measured during minute 1–3 using FreezeScan. In the second trial measuring cued memory, mice were placed in a novel context that contained a different odor and exposed to the same cue light and tone from day 1 after 2 min of exploration. Freezing behavior was measured for 1–3 min following the cue. For the retention test (one month later) in the chronic paradigm, the animals were re-exposed to the same fear conditioning chamber containing the same context with no CS or foot shocks.

2.6. Tissue processing.

For histology, mice were perfused transcardially with saline and brains were extracted and fixed for 48 hours in 4% paraformaldehyde and equilibrated in 30% sucrose for histological analysis as previously described (Luo et al., 2007). The brains were coronally sectioned at 40 μm with a sliding microtome (SM2010 R, Leica, Allendale, NJ). The sections from the sampling region (Figure 1A) were serially collected in 12 tubes and stored in cryoprotective medium. For RNA-seq analysis, the brains were extracted, the ipsilateral hippocampi dissected, snap frozen and stored at −80°C.

2.7. Immunohistochemistry.

Immunohistochemistry was performed on free-floating sections following standard procedures (Luo et al., 2013; Tesseur et al., 2017). The sample region in this study (Figure 1A) covers the impact epicenter and surrounding tissue, roughly from 0.98 mm to −2.06 mm to Bregma (Luo et al., 2014). This resulted typically in 6–7 sections/tube (sections collected in 12 tubes) separated roughly by 12 × 40 μm. For each staining all sections from one tube were used. Free-floating sections were permeabilized, blocked, and stained overnight at 4°C with the following primary antibodies at the designated concentrations: Iba1 (1:500; Catalogue #: ab5076; Abcam, Cambridge, MA), CD68 (1:50; Catalogue #: MCA1957; Bio-Rad, Hercules, CA), GFAP (1:1000; Catalogue #: M0761; Agilent, Santa Clara, CA) and p-CREB (1:1000; Catalogue #: 06-519; MilliporeSigma, Burlington, MA). Sections were washed, stained with Alexa Fluorconjugated secondary antibodies (1:250), and mounted under a coverslip. Images covering the whole hippocampus (10X, for Iba1, CD68 and GFAP) or CA1 and CA3 (20X, for p-Creb) were acquired through on a confocal laser-scanning microscope (Zeiss LSM880, Carl Zeiss Microscopy, White Plains, NY). Image analysis of immunoreactivity was performed using NIH ImageJ software (Bethesda, MD) by a blinded observer. The images were first converted to 8-bit grey-scale images and then converted into binary positive/negative images by thresholding held constant for all images in a given brain region. Percent area fraction covered by the threshold was determined by ImageJ. The average of values obtained from all sections was used for each animal for statistical analysis.

2.8. Measurement of plasma cytokines.

A panel of 32 cytokines, including chemokines (Eotaxin-1/Ccl11, MCP-1/Ccl2, MIP-1α/Ccl3, MIP-1β/Ccl4, MIP-2/Cxcl2, RANTES/Ccl5, KC/Cxcl1, IP-10/Cxcl10, LIX/Cxcl5, and MIG/Cxcl9), growth factors (M-CSF/Csf1, G-CSF/Csf3, GM-CSF/Csf2, LIF/Lif), IFNs (IFN-γ/Ifng), ILs (IL-10/Il10, IL-12p40/Il12b, IL12p70, IL-13/Il13, IL-15/Il15, IL-17A/Il17a, IL-1α/Il1a, IL-1β/Il1b, IL-3/Il3, IL-4/Il4, IL-5/Il5, IL-6/Il6, IL-7/Il7, Il9/Il9, IL-10/Il10), TNF superfamily member (TNF-α/Tnf), and others (VEGF/Vegfa), was analyzed by Eve Technologies (Calgary, AB, Canada) with a mouse cytokine array/chemokine array 32-plex panel using multiplexing laser bead technology. All proteins that contained interpolated/extrapolated/out-of-range values were removed from the study. The concentration of these proteins was calculated using a standard curve and expressed in pg/mL.

2.9. Data and statistical analysis.

Data are presented as mean ± SD. Data were analyzed using 2-tailed Student’s t test for comparing two groups or ANOVA for multiple groups. Tukey post-hoc test was used to compare pairs of groups following ANOVA. Statistical analysis was performed with Prism software (version 9) (GraphPad Software, La Jolla, CA). P < 0.05 was considered statistically significant.

2.10. RNA-Seq.

In a separate set of mice after delayed treatment (treatment started at 3 months after injury and lasted one month), mice were perfused as described as above and their brains were extracted. Ipsilateral hippocampi were further dissected and subjected to RNA-seq analysis. Briefly, total RNA from the ipsilateral hippocampus was isolated using RNeasy Mini Kit (Qiagen, Germantown, MD) per the manufacturer’s instructions. The concentration of RNA was measured using a NanoDrop 8000 UV-Vis Spectrophotometer (Thermo Fisher Scientific, Waltham, MA). RNA integrity number (RIN) was measured using the Agilent Bioanalyzer (Agilent Technologies, Santa Clara, CA). Total RNA samples with RIN >9.0 and a concentration of 150ng/μl or higher were used for RNA sequencing. Sequence libraries were generated and sequenced by Novogene (Sacramento, CA). Insert size of 250–300 bp was used for cDNA library preparation. Libraries were sequenced on the Illumina platform for 150-bp paired-end reads. Reference genome and annotation files were downloaded from Ensembl, and RNA-seq data were aligned to the reference genome using the Spliced Transcripts Alignment to a Reference (STAR) software. HTSeq was used to obtain read counts of the mapped genes, and the DESeq2 package was used for differential expression analysis (https://rpubs.com/ge600/deseq2). Ingenuity Pathway Analysis (Qiagen) was used for pathway analysis. P values were adjusted by the Benjamini-Hochberg method, with padj (adjusted p-value) < 0.05 considered as significantly differential expression.

2.11. qPCR.

The same RNA samples used for RNA-seq were used for reverse transcription. A total of 1000 ng of RNA was treated with DNase I (Invitrogen) and then converted to cDNA using the SuperScript III First Strand Synthesis System (Invitrogen). cDNA was diluted 1:10 in water. Real-time quantitative PCR was performed using SYBR Green I Master mix (Roche) on a LightCycler 480 (Roche). The PCR primers are listed in Table 1. Melting curves were used to confirm the purity of the amplified product. Cycle threshold (Ct) values were normalized to actin. ΔΔCT values were used to calculate gene expression changes with actin as an internal control (Kirby et al., 2015; Tesseur et al., 2017; Villeda et al., 2014).

Table 1.

Primers for qPCR validation experiment

Gene Name	Forward	Reverse	Amplicon Size
Adcy3	CTCGCTTTATGCGGCTGAC	ACATCACTACCACGTAGCAGT	138
Bgn	TGCCATGTGTCCTTTCGGTT	CAGGTCTAGCAGTGTGGTGTC	112
C4b	ACTTCAGCAGCTTAGTCAGGG	GTCCTTTGTTTCAGGGGACAG	212
Cacna1g	TGTCTCCGCACGGTCTGTAA	AGATACCCAAAGCGACCATCTT	205
Cacna1i	GGGCGTGGCCTGTTTAGTC	TGAGGGTCTCGGAGTGCTC	172
Cacng5	ACCTGGAAGAAGGCATAATCCT	CTATGGTAAAACAGCGTCCTCG	123
Camk2g	AGTCACTCCTGAAGCTAAGAACT	GGTTTGTGGTTCCTTGACACC	283
Fgfr3	GCCTGCGTGCTAGTGTTCT	TACCATCCTTAGCCCAGACCG	217
Gng2	ACCGCCAGCATAGCACAAG	AGTAGGCCATCAAGTCAGCAG	106
Gria2	TTCTCCTGTTTTATGGGGACTGA	CTACCCGAAATGCACTGTATTCT	107
Grin2a	ACGTGACAGAACGCGAACTT	TCAGTGCGGTTCATCAATAACG	100
Itpr1	CGTTTTGAGTTTGAAGGCGTTT	CATCTTGCGCCAATTCCCG	136
Pik3r1	ACACCACGGTTTGGACTATGG	GGCTACAGTAGTGGGCTTGG	140
Plcg1	ATCCAGCAGTCCTAGAGCCTG	GGATGGCGATCTGACAAGC	105
Polr2f	GACAACGAGGACAATTTCGACG	GGAGAATCTCGACATTTTCCTGG	112
Ppm1l	ACCCGAGACGCTTTTCCTG	CGCCACATTATGGCTCTTGAA	217
Ppp3ca	GTGAAAGCCGTTCCATTTCCA	GAATCGAAGCACCCTCTGTTATT	163
Prkaca	AGATCGTCCTGACCTTTGAGT	GGCAAAACCGAAGTCTGTCAC	119
Prkcd	CCTCCTGTACGAAATGCTCATC	GTTTCCTGTTACTCCCAGCCT	181

3. Results

3.1. A single bolus of CSF1 administered 24 h post-injury attenuates associative memory deficits 3 months later.

We have previously shown that systemic administration of CSF1 reduces kainate-induced neurodegeneration and neuroinflammation when administered 6 h post-injury (Luo et al., 2013). To test if systemic administration of this factor could exert protection after mTBI, we performed bioluminescence imaging of astrocyte reactivity in a repetitive mTBI paradigm (Fig. 1B, experimental design). Brain injury is associated with astrocyte reactivity and upregulation of GFAP expression. GFAP-luc transgenic mice expressing firefly luciferase under the GFAP promoter allow us to measure bioluminescence (and astrocyte reactivity) non-invasively as a surrogate biomarker for brain injury and inflammation (Luo et al., 2013; Luo et al., 2008; Luo et al., 2007; Luo et al., 2014). We first performed baseline bioluminescence imaging of GFAP-luc mice, then induced repetitive mTBI (3 impacts over 3 days), and administered CSF1 24 h after the last impact. We performed bioluminescence imaging again 2 days later (3 days after last injury) (Fig. 1C). In agreement with previous findings (Luo et al., 2014), vehicle-treated mTBI mice showed significant increase in bioluminescence signal (and astrocyte reactivity), compared with sham animals; but CSF1-treated mice showed significant reduction of astrocyte reactivity (Fig. 1D, one-way ANOVA: F_{(2, 12)} = 22.32, P < 0.0001). To test whether the treatment led to long-term benefits on cognition, we tested these mice three months later using a fear conditioning paradigm (Fig. 1E). The CSF1 treated mTBI mice performed significantly better than vehicle-treated mice, exhibiting significantly higher contextual freezing, similar to sham-mice (Fig. 1E, one-way ANOVA: F_{(2, 12)} = 5.174, P = 0.0240). The cued memory followed the same trends, but the difference was not statistically significant.

We have previously shown that in this model cognitive impairment is accompanied with astrocyte reactivity and reduced CREB phosphorylation (Luo et al., 2014). To investigate whether the improvement in memory function was associated with reduced astrocyte reactivity and increased CREB phosphorylation, we performed immunohistochemistry with brain sections of these mice. Quantification of GFAP immunoreactivity shows significant astrocyte reactivity in the ipsilateral hippocampi of vehicle-treated mTBI mice relative to sham animals. Treatment with CSF1 significantly reduced astrocyte reactivity (Fig. 2 A–B, one-way ANOVA: F_{(2, 12)} = 7.443, P = 0.0079). On the other hand, vehicle-treated mTBI mice showed significantly reduced p-CREB immunoreactivity in the hippocampi neurons (CA3), which was restored by CSF1 treatment (Fig. 2 C–D, one-way ANOVA: F_{(2, 12)} = 16.01, P = 0.0004). Together, these results demonstrate that treatment with a single bolus of CSF1 (administered 24 h after mTBI) exerts long-term protective effects.

Figure 2. A single bolus of CSF1 administered 24 h post-injury attenuates astrocyte reactivity and restore CREB phosphorylation.

Brain sections from GFAP-luc mice were assessed for astrocyte reactivity by GFAP immunostaining (A, B) and for CREB phosphorylation (p-CREB) (C, D). n = 5 mice/group (3 males, 2 females). Scale bars in (D, F) = 25 μm. *, P < 0.05; **, P < 0.01, ANOVA and Tukey’s multiple comparison test. In (B, D), open symbols represent female mice and closed (filled) symbols represent male mice.

3.2. CSF1, administered 3 months post-injury, attenuates memory deficits after 1 month of treatment

Next, we evaluated whether CSF1 exerts protective effects in mice that already exhibit mTBI-related chronic cognitive deficits (see experimental design and outline in Fig. 3A). We first induced repetitive mTBI in wild type C57Bl/6J mice. Three months after injury, we performed the two-trial Y-maze and contextual fear conditioning tests. The two-trial Y-maze assesses the spatial recognition memory in rodents. Mice with good spatial recognition memory spend more time in the novel arm relative to the previously explored arms. In the mTBI group, the percentage of time spent visiting the novel arm was significantly lower than that of the sham group (Fig. 3B, left panel, unpaired t test, t = 2.246, df = 45, P = 0.0297), suggesting that the mTBI group had impaired spatial recognition memory (Fig. 3C, left panel, discrimination index, unpaired t test, t = 2.451, df = 45, P = 0.0182). Similarly, in the contextual fear conditioning tests the mTBI group displayed impaired contextual fear memory (Fig. 3D, left panel, unpaired t test, t = 3.476, df = 45, P = 0.0011).

Figure 3. Delayed CSF1 treatment improves memory function after traumatic brain injury.

(A) Experimental design scheme. Wildtype C57BL/6J mice (male, 3 month of age) received three mild impacts (mTBI) or underwent sham procedures (sham). Three months later mice were tested for memory functions with Y maze (B-C, left panels) and contextual fear conditioning (D, left panel). Animals were then randomized and treated with CSF1 (800 μg/kg, ip, 3 times/week) or vehicle. After one month of treatment, mice were tested again with using two trial Y maze (B-C, right panels) and contextual fear retention (D, right panel). Bars represent mean ± SD and analyzed by unpaired t test (before treatments, 2 groups, left panels) or two way ANOVA and Tukey’s multiple comparison test (after treatments, 4 groups, right panels) or. *, P < 0.05; **, P < 0.01. n = 11–12 mice/group.

After we confirmed that mice with repetitive mTBI display behavioral deficits, we randomly assigned them into vehicle (PBS)- or CSF1-treated groups. We treated the mice for 1 month and performed behavioral tests again (4 months after injury). In the two-trial Y maze test, the percentage of time spent exploring the novel arm was not significantly different between the two sham groups, whether they were treated with vehicle or CSF1. The percentage of time spent exploring the novel arm was significantly lower in the mTBI group received vehicle treatment (mTBI + veh), however, the mTBI group received CSF1 treatment (mTBI + CSF1) showed significantly higher percentage of time spent exploring the novel arm (Fig. 3B, right panel, two-way ANOVA: treatment, F_{(1, 43)} = 6.214, P = 0.0166; injury, F_{(1, 43)} = 2.947, P = 0.0932; interaction, F_(1,43) = 1.678, P = 0.2032) and significantly increased discrimination index (Fig. 3C, right panel, two-way ANOVA: treatment, F_(1,43) = 15.70, P = 0.0003; injury, F_{(1, 43)} = 0.0320, P = 0.8589; interaction, F_{(1, 43)} = 6.909, P = 0.0118), reaching to a level similar to the sham groups. Similarly, in the fear conditioning test, the (mTBI + veh) group showed reduced contextual fear memory, which was restored in the (mTBI + CSF1) group (Fig. 3D, right panel, two-way ANOVA: treatment, F_(1,43) = 1.678, P = 0.2032; injury, F_{(1, 43)} = 2.226, P = 0.1442; interaction, F_{(1, 43)} = 5.659, P = 0.0226). In summary, these results demonstrate that 4 months after injury, mice with repetitive mTBI display significant deficits in spatial recognition and fear memory, which were significantly attenuated by CSF1 treatment started at 3 months after injury.

3.3. CSF1 treatment alters plasma cytokine profile

To investigate whether plasma levels of cytokines were altered by CSF1 treatment and to discover potential biomarkers, we collected plasma at the end of the experiment and sent the samples for high-throughput multiplex cytokine analysis. Of the 32 cytokines surveyed in this study, 29 passed our initial filtering criteria. The mean expression values of cytokines in the four different groups were depicted by heatmaps, where the blue and red colors indicate the down-regulated and up-regulated expression, respectively (Fig. 4A). Compared with those of the (sham + veh) group, most cytokines of the (mTBI + veh) group showed an up-regulated pattern. Treatment with CSF1 down-regulated most of the cytokines in the (mTBI + CSF1) group. Two of the cytokines, IP10 (two-way ANOVA: treatment, F_{(1, 43)} = 6.799, P = 0.0125; injury, F_{(1, 43)} = 10.82, P = 0.0020; interaction, F_{(1, 43)} = 1.004, P = 0.3220) and IL-13 (two-way ANOVA: treatment, F_{(1, 43)} = 0.7515, P = 0.3911; injury, F_{(1, 43)} = 0.8208, P = 0.3702; interaction, F_{(1, 43)} = 5.818, P = 0.0204) showed significant differences, i.e., their levels were significantly increased in the (mTBI + veh) group compared with (sham + veh) group, but were significantly reduced in the (mTBI + CSF1) group (Fig. 4B, C). Additional significant main effects and their respective post-hoc comparisons were presented in Fig. 4D.

Figure 4. Changes of plasma cytokine/chemokine levels with delayed CSF1 treatment after mTBI.

The mice were sacrificed after the behavioral tests were performed. Plasma samples were submitted to Eve Technologies for cytokine/chemokine array. (A) Heatmap showing the levels of the 29 cytokine/chemokines. Colored bars indicate the expression levels (Red = upregulated, blue = downregulated). Rows represent cytokine/chemokines and columns represent groups. (B-C), plasma levels of IP-10 and IL-13. *, P < 0.05, by two-way ANOVA and Tukey’s multiple comparison test. n = 11–12 mice/group. (D) Additional significant main effects and their respective post-hoc comparisons by two-way ANOVA and Tukey’s multiple comparison test.

3.4. CSF1 treatment reduces microglia and astrocyte reactivity and maintains neuronal CREB phosphorylation

We next asked whether CSF1 might reduce microglia (Fig. 5) and astrocyte reactivity, and maintain or restore CREB phosphorylation (Fig. 6). Microglia reactivity, assessed by CD68 (Fig. 5B, two-way ANOVA: treatment, F_{(1, 28)} = 10.20, P = 0.0002; injury, F_{(1, 28)} = 63.02, P < 0.0001; interaction, F_(1,28) = 10.52, P = 0.0002) and Iba1 (Fig. 5D, two-way ANOVA: treatment, F_{(1, 28)} = 1.831, P = 0.1301; injury, F_{(1, 28)} = 75.06, P < 0.0001; interaction, F_(1,28) = 1.887, P = 0.1301) immunoreactivity, was significantly increased in the ipsilateral hippocampus of vehicle-treated mTBI animals relative to sham (Fig. 5, P < 0.0001, mTBI + veh group vs Sham + veh group, Tukey’s multiple comparison test). Treatment with CSF1 significantly attenuated mTBI-induced CD68 immunoreactivity relative to vehicle-treated mTBI animals (Fig. 5B, P < 0.0001, mTBI + CSF1 group vs mTBI + veh group, Tukey’s multiple comparison test), but had no significant effects on Iba1 immunoreactivity (Fig. 5D, P = 0.1406, mTBI + CSF1 group vs mTBI + veh group, Tukey’s multiple comparison test).

Figure 5. Delayed CSF1 treatment reduces microglia reactivity after mTBI.

The mice were sacrificed after the behavioral tests were performed. Brains were collected, fixed with PFA and sectioned in series. Brain sections were immunostained with antibodies against CD68 (A) or Iba1 (C). The immunoreactivity of ipsilateral hippocampus was analyzed semi-quantitatively with thresholding image analysis using ImageJ software. *, P < 0.05; **, P < 0.01, by two-way ANOVA and Tukey’s multiple comparison test. n = 8 mice/group. Scale bars in (A, C) = 25 μm.

Figure 6. Delayed CSF1 treatment reduces astrocyte reactivity and restores p-CREB immunoreactivity in the hippocampus after mTBI.

The mice were sacrificed after the behavioral tests were performed. Brains were collected, fixed with PFA and sectioned in series. Brain sections were then immunostained with an antibody against GFAP (A-C) or p-CREB (D-E). The immunoreactivity of ipsilateral hippocampus was analyzed semi-quantitatively with thresholding analysis using ImageJ software. *, P < 0.05, by two-way ANOVA and Tukey’s multiple comparison test. n = 8 mice/group. Scale bars = 50 μm in (A), = 20 μm in (D).

Consistent with previous studies (Luo et al., 2014), astrocyte reactivity assessed by GFAP immunoreactivity (Fig. 6B, two-way ANOVA: treatment, F_{(1, 28)} = 0.5219, P = 0.4210; injury, F_{(1, 28)} = 7.015, P = 0.0131; interaction, F_(1,28) = 1.049, P = 0.3144) and immunofluorescence intensity (Fig. 6C, two-way ANOVA: treatment, F_{(1, 28)} = 8.498, P = 0.0069; injury, F_{(1, 28)} = 20.82, P < 0.0001; interaction, F_(1,28) = 1.768, P = 0.1944), was significantly increased in the ipsilateral hippocampus of vehicle-treated mTBI animals relative to sham (Fig. 6, P < 0.05, mTBI + veh group vs Sham + veh group, Tukey’s multiple comparison test). Treatment with CSF1 had no effects on GFAP immunoreactivity of the sham groups, but significantly attenuated mTBI-induced GFAP immunofluorescence intensity relative to vehicle-treated mTBI animals (Fig. 5, P = 0.0270, mTBI + CSF1 group vs mTBI + veh group, Tukey’s multiple comparison test).

For CREB phosphorylation, mTBI significantly reduced p-CREB immunoreactivity in the CA3 neurons in the ipsilateral hippocampus, compared with sham, again consistent with previous studies (Luo et al., 2014). CSF1 treated mTBI mice showed significantly higher p-CREB immunoreactivity in CA3 neurons compared with vehicle-treated mTBI mice (Fig. 6 D–E, two-way ANOVA: treatment, F_{(1, 28)} = 8.937, P = 0.0655; injury, F_{(1, 28)} = 16.74, P = 0.0141; interaction, F_{(1, 28)} = 13.01, P = 0.0285. P = 0.0282, mTBI + CSF1 group vs mTBI + veh group, Tukey’s multiple comparison test). These results show that CSF1 treatment reduces microglia and astrocyte reactivity and restores CREB phosphorylation.

3.5. RNA-seq and Ingenuity Pathway Analysis reveals CSF1 treatment affects cognition and memory-related transcriptomic changes and pathways

To identify genes and pathways involved in CSF1 amelioration of mTBI-induced cognitive impairment, we performed RNA-seq analysis in a separate set of mice after delayed treatment, using total RNA isolated from the hippocampus. Through differential expression analysis of RNA-Seq data, we identified 350 differentially expressed genes (DEGs) (Padj < 0.05) across all 4 possible comparisons (Fig. 7). The volcano plots showed the up-regulated and down-regulated genes in each comparison (Fig. 7A). Vehicle-treated mTBI (mTBI + veh) mice had significantly altered expression of 92 genes, compared to Vehicle-treated sham (sham + veh) mice (Fig. 7B). CSF1 treatment of mTBI mice altered expression of 190 genes compared to vehicle-treated mTBI (mTBI + veh) mice (Fig. 7B). No DEGs were discovered from the comparison of (sham + CSF1) vs. (sham + veh). Unsupervised hierarchical clustering based on the 350 DEGs revealed that mice with the same injury and treatment status were largely clustered together (Fig. 7C).

Figure 7. RNA-seq and differential transcriptome analysis reveals changes in mRNA expression profiles between groups.

In a separate set of mice after delayed treatment, mice were perfused, and their brains were extracted. Ipsilateral hippocampi were further dissected and subjected to RNA-seq analysis. (A) Volcano plot visualizing the number of differentially expressed genes (DEGs) with respect to each comparison. The y-axis displays the -log₁₀ padj (adjusted p value) for each gene, while the x-axis displays the log₂ fold change (indicated by circle size). Horizontal dotted line indicates the cut-off criteria for DEGs with padj ≤ 5%. Red circles indicate upregulated genes while blue circles indicate downregulated genes. Black circles represent non-differentially expressed genes. (B) Venn diagrams of differentially expressed genes in the indicated comparisons. (C) Heatmap of 350 total differentially expressed genes (rows) was generated by unsupervised hierarchical clustering and resulted in clustering of the four groups. Red indicates up-regulation and blue indicates down-regulation.

We used qPCR to validate the expression of 18 genes that showed distinct expression patterns in response to injury and treatment (Table 1), according to RNA-Seq. cDNA was prepared using the same RNA that was used for RNA-seq, and actin was used as an internal control. We found a significantly positive correlation between RNA-Seq and qPCR data (Fig. 8A, R² = 0.946, F_(1,32) = 203.3, P < 0.0001), confirming the reliability of RNA-seq data. Three genes, Cacng5 (two-way ANOVA. RNA-seq: treatment, F_(1,16) = 1.294, P = 0.2721; injury, F_(1,16) = 0.1478, P = 0.7057; interaction, F_(1,16) = 9.976, P = 0.0061. qPCR: treatment, F_(1,16) = 1.060, P = 0.3186; injury, F_(1,16) = 0.0252, P = 0.8758; interaction, F_(1,16) = 12.12, P = 0.0031.), Bgn (two-way ANOVA. RNA-seq: treatment, F_(1,16) = 0.9604, P = 0.3417; injury, F_(1,16) = 2.331, P = 0.1464; interaction, F_(1,16) = 7.961, P = 0.0123. qPCR: treatment, F_(1,16) = 1.335, P = 0.2650; injury, F_(1,16) = 1.263, P = 0.2776; interaction, F_(1,16) = 5.123, P = 0.0379.), and C4b (two-way ANOVA. RNA-seq: treatment, F_(1,16) = 0.3614, P = 0.5561; injury, F_(1,16) = 17.37, P = 0.0007; interaction, F_(1,16) = 0.7724, P = 0.3925. qPCR: treatment, F_(1,16) = 0.3614, P = 0.5561; injury, F_(1,16) = 19.82, P = 0.0004; interaction, F_(1,16) = 0.1699, P = 0.6856.), representing different expression patterns, were depicted in detail in Fig. 8B.

Figure 8. Validation of RNA-seq results by qPCR.

(A) Correlation between fold changes [in comparison of (mTBI + veh) vs (sham + veh)] of qPCR and RNA-seq data. (B) Comparison of gene expression as measured by RNA-Seq (normalized counts, left panels) and qPCR (fold change, right panels). Three genes were shown to represent different expression patterns: Cacng5 (down-regulated by mTBI, but restored by CSF1 treatment), Bgn (up-regulated by mTBI, but restored by CSF1 treatment), and C4b (up-regulated by mTBI, but not restored by CSF1 treatment). n = 4–6 mice/group, *, P < 0.05; **, P < 0.01; by two-way ANOVA and Tukey’s multiple comparison test.

To elucidate the potential functions of the DEGs induced by CSF1 treatment, we performed Ingenuity Pathway Analysis (IPA) using the 190 DEGs obtained from the comparison of (mTBI + CSF1) vs (mTBI + veh). We identified 30 canonical pathways which may play important roles in CSF1 treatment, with the top 20 shown in Fig. 9A. Some of the top canonical pathways, such as synaptic long term depression, calcium signaling, CREB signaling in neurons and synaptic long term potentiation are associated with memory formation and learning. Heatmap visualization of with unsupervised clustering showed genes of these pathways that were altered in mTBI were largely restored by CSF1 treatment (upregulated genes with mTBI were downregulated with CSF1 treatment or downregulated genes with mTBI were upregulated with CSF1 treatment) (Fig. 9B).

Figure 9. Pathways affected by delayed CSF1 treatment using Ingenuity Pathway Analysis (IPA).

(A) Top canonical pathways modulated by CSF1 treatment revealed by Ingenuity Pathway Analysis of DEGs identified by comparison of (mTBI + CSF1) vs (mTBI + veh). The top x-axis indicates the ratios (orange line, calculated from the number of genes in the dataset of a given pathway divided by the total number of genes in that canonical pathway), and bottom x-axis indicates -log (P value). (B-D) Heatmaps of genes belonging to three pathways: CREB signaling in neurons (B), Synaptic long-term potentiation (C) and Synaptic long-term potentiation (D). Heatmaps were generated by unsupervised hierarchical clustering with blue and red colors on the Z-scale indicate lower and higher expression respectively.

4. Discussion

Our results demonstrate that systemic administration of CSF1 effectively attenuates mTBI-induced cognitive deficits in a concussive rodent model. We administered CSF1 at 24 hours and 3 months after injury, and remarkably, CSF1 was sufficient to attenuate the cognitive deficits in both paradigms. Since the beneficial effects were observed three and four more months after injury, our results suggest that CSF1 treatment can alleviate mTBI-induced chronic cognitive deficits and associated pathology long after injury.

Chronic cognitive deficits, including executive function, learning and memory, and attention deficits, are one of the most common consequence of TBI and can lead to significant disability and morbidity, regardless of the injury severity (Wortzel and Arciniegas, 2012). Due to the high incidence of mTBI and the common diagnosis of cognitive deficits following mTBI, enormous effort has been devoted to the development of new therapeutic interventions to improve cognitive function. However, few studies were designed to target chronic cognitive deficits after mTBI. Most previous studies targeted the acute injury responses immediately following injury characterized by a robust inflammatory response and aimed to develop strategies to dampen the inflammatory response (Morganti-Kossmann et al., 2019). Blocking the acute inflammatory responses within 24 h after injury has been shown to prevent TBI-induced cognitive deficits (Morganti-Kossmann et al., 2019). However, most drugs that have been tested in animal TBI models have short therapeutic time windows (less than 12 hours post-injury) (Mohamadpour et al., 2019) and few have been shown to improve long-term cognitive deficits (Bramlett and Dietrich, 2015). Treatments effective only within an acute time window after injury poses limitations because their optimal treatment timing may not be feasible in clinical settings (Mohamadpour et al., 2019). In the present study, we demonstrated that CSF1 has a large acute time window (24 hours post-injury) and is effective in preventing mTBI-induced long-term cognitive deficits (3 months post-injury) (Fig. 1). Our results also show that CSF1 is similarly effective in reversing mTBI-induced long-term cognitive deficits when administered at 3 months post-injury (when animals display chronic cognitive deficits) (Fig. 3). Together, our findings suggest that treatment with CSF1 may be a promising therapeutic strategy for chronic cognitive defects resulting from mTBI.

Therapeutic time window is largely determined by TBI pathology (Mohamadpour et al., 2019). That CSF1 has a large acute time window (24 hours post-injury) suggests CSF1 may target pathophysiological events that evolve relatively slowly after injury. Moreover, it is remarkable that a single injection of CSF1 leads to long-lasting protection. It suggests that the pathophysiological events that CSF1 targets are critically involved in chronic cognitive impairments. Using bioluminescence imaging of astrocyte reactivity as surrogate marker, we have shown there is massive inflammatory response right after injury in our model, followed by chronic low-grade inflammation (Luo et al., 2014). It is possible that CSF1 in the acute paradigm targets the early injury response and in the chronic paradigm, inhibits mTBI-initiated residual low-grade inflammation that remains after acute inflammation has subsided. A recent study reported that CSF1 administered intranasally 1h and 24h after injury reduced neuroinflammation at 48 h and improved water maze performance 4 weeks later in rat model of neonatal hypoxic-ischemic encephalopathy (Hu et al., 2020). These results support that early CSF1 administration attenuates neuroinflammation in the acute phase and improves chronic cognitive function. There have been several studies that have shown long-term functional and histological improvements using a single bolus of treatment in CNS trauma. For example, a single bolus of docosahexaenoic acid led to long-term functional improvement after spinal cord injury (Liu et al., 2015; Ward et al., 2010), although the therapeutic time window for docosahexaenoic acid is 30 min after injury (Liu et al., 2015). Nevertheless, these studies demonstrate that a single bolus of early treatment can lead to long-term functional and histological improvements after traumatic injury in the CNS. They highlight the importance of early intervention and therapeutic time window in the treatment of TBI.

We used RNA-Seq to probe the gene expression changes associated CSF1 treatment. These transcriptomic experiments were performed using a separate cohort of mice that was not behaviorally tested, avoiding the possible confounding effects of the learning and memory tests on gene expression. The transcriptomic changes caused by CSF1 treatment prominently affected genes with notable enrichments in pathways related to memory formation and learning, such as synaptic long term depression, calcium signaling, CREB signaling in neurons and synaptic long term potentiation, suggesting particular relevance of these pathways in CSF1 treatment.

The involvement of CREB signaling in CSF1 treatment is further supported by immunohistochemistry, where CSF1 treatment was found to restore/maintain CREB phosphorylation (Fig. 2, 6). Deficits in CREB signaling has previously been implicated in chronic mTBI (Luo et al., 2014; Rehman et al., 2019). Restoring CREB signaling has been proposed to be the underlying mechanism of many therapeutic agents in rescuing cognitive functions after TBI (Bhat et al., 2020; Rehman et al., 2019; Titus et al., 2013). Therefore, the effect of CSF1 on CREB phosphorylation may be an underlying mechanism in its improvement of mTBI symptoms.

Inflammation is an important contributor to the pathology of TBI (Corps et al., 2015; Faden et al., 2016; Gyoneva and Ransohoff, 2015). In addition to glial reactivity, TBI induces an inflammatory response involving complement activation and generation of active fragments, such as C3a and C5a anaphylatoxins, C3b, C4b, and iC3b (Huber-Lang et al., 2018). Glial reactivity as shown by the increased expression of the astrocyte marker GFAP and of microglial marker Iba1 and CD68 is a sign of neuroinflammation. In the RNA-seq analysis, many of the top genes induced by mTBI are related to inflammation (Fig. 8). Biglycan (BGN), a member of the family of small leucine-rich proteoglycans, is a ubiquitous component of extracellular matrix and a crucial proinflammatory factor (Schaefer et al., 2005; Xie et al., 2020). Together, our data of increased glial reactivity and upregulation of C4b and BGN support a chronic inflammatory state persisting long after injury.

Characterizing inflammation in the peripheral blood can help uncover the potential systemic consequences of brain injury and reveal the inflammatory status of the brain (Jeter et al., 2013; Kim et al., 2018; Kulbe and Geddes, 2016; Wang et al., 2005; Zetterberg et al., 2013). Many groups have been able to successfully detect significant increases in inflammatory markers in the blood following TBI and established associations or correlations with other injury parameters and outcomes (Jeter et al., 2013; Kim et al., 2018; Kulbe and Geddes, 2016; Wang et al., 2005; Zetterberg et al., 2013). Many of the cytokines/chemokines we measured in this study are considered pro-inflammatory (Gyoneva and Ransohoff, 2015). Their trend of up-regulation suggests a chronic, systemic inflammatory state after repetitive mTBI. Importantly, this inflammatory state was largely attenuated by CSF1 treatment (Fig. 4). Furthermore, our results suggest that plasma levels of IP-10 and IL-13 could be potential biomarkers for chronic mTBI, because they were the most prominently increased of all cytokines after injury, and were responsive to CSF1 treatment. Previous studies have shown that IP-10 and IL-13 were increased early after TBI, but few studies have assessed IP-10 and IL-13 in chronic mTBI. IP-10 (10 KDa Interferon Gamma-Induced Protein, also known as C-X-C Motif Chemokine Ligand 10, CXCL10) is a chemokine involved in T helper (Th) 1 immune responses and is consistently upregulated after TBI in gene array experiments and in samples from TBI patients (Gyoneva and Ransohoff, 2015). Further investigation is needed to characterize the cellular sources of CXCL10 in TBI.

It remains unclear whether the therapeutic effects of CSF1 is through direct impact on neurons or other cell types such as microglia and/or immune cells. CSF1R, the receptor for CSF1, is primarily expressed in all macrophages and monocytes as well as osteoclasts, and in microglia in the brain (Chitu et al., 2016; Chitu and Stanley, 2006). Microglia are crucial regulators of brain function and cognition via neuron-microglia interaction, neuronal differentiation and survival, synaptic modelling, scavenging of cellular debris and secretion of trophic factors (Prinz et al., 2019). All of these processes are modulated by CSF1 (Chitu et al., 2016; Kana et al., 2019; Wohleb et al., 2018). We and others have reported low expression of CSF1R in the neuronal lineage in the brain (Luo et al., 2013; Nandi et al., 2012) and its expression is upregulated after injury (Luo et al., 2013). Recent RNA-seq transcriptome analysis demonstrate the expression of CSF1R in astrocytes and its upregulation after traumatic spinal cord injury (Anderson et al., 2016), in ageing (Boisvert et al., 2018) and following several physiological and pathological experimental Perturbations, including Huntington’s disease and calcium signaling (Yu et al., 2020). Hence, it is possible that CSF1 may impact a mixture of cell types in the brain as well as peripheral immune function to alleviate cognitive decline.

There are several potential limitations of our study. In the acute paradigm, the number of mice per group was low and underpowered, and there was no (sham + CSF1) control group. In the chronic paradigm, only male mice were used. It is well accepted that sex/gender play a role in how the body and brain respond to injury, however, sex differences in pathophysiology and recovery of TBI are poorly understood and the effect of sex on TBI outcomes remains controversial (Gupte et al., 2019). Given the growing interest in sex as a biological variable affecting injury outcomes and treatment efficacy (Shansky, 2019), further study is needed to discover whether CSF1 improves different aspects of cognitive function in male vs female mice.

5. Conclusion

In this study we showed CSF1 has therapeutic potential against chronic cognitive dysfunction after repetitive mTBI. We evaluated the neuroprotective effects of CSF1 through a closed-head animal model of repetitive mTBI, and demonstrated that acute (initiated 24 hours post injury) or delayed (initiated 3 months post injury) treatment with CSF1 improves long-term memory and reduces pathological sequelae at late time-points in mice after repetitive mTBI. The anti-inflammatory and CREB-restorative effects of CSF1 may contribute to its neuroprotective potential in mTBI-insulted brains. Since CSF1 is approved for human use in clinical trials (Douglass et al., 2008), CSF1 may be a promising therapeutic drug for treating TBI-related chronic cognitive dysfunction. We hope that our findings may open new promising therapeutic avenues for patients that suffer from chronic cognitive deficits associated with mTBI and other neurodegenerative disorders.

• A single bolus of CSF1 administered 24 h post-injury reduces memory impairment.

• Three months after mTBI, one month of CSF1 treatment improves cognitive function.

• CSF1 treatment restores CREB phosphorylation and attenuates astrocytic reactivity.

• CSF1 treatment alters cognition-related transcriptomic changes and pathways.