OUP user menu

Powerful beneficial effects of macrophage colony-stimulating factor on β-amyloid deposition and cognitive impairment in Alzheimer's disease

Vincent Boissonneault, Mohammed Filali, Martine Lessard, Jane Relton, Gordon Wong, Serge Rivest
DOI: http://dx.doi.org/10.1093/brain/awn331 1078-1092 First published online: 17 January 2009


Alzheimer's disease is a major cause of dementia in humans. The appearance of cognitive decline is linked to the overproduction of a short peptide called β-amyloid (Aβ) in both soluble and aggregate forms. Here, we show that injecting macrophage colony-stimulating factor (M-CSF) to Swedish β-amyloid precursor protein (APPSwe)/PS1 transgenic mice, a well-documented model for Alzheimer's disease, on a weekly basis prior to the appearance of learning and memory deficits prevented cognitive loss. M-CSF also increased the number of microglia in the parenchyma and decreased the number of Aβ deposits. Senile plaques were smaller and less dense in the brain of M-CSF-treated mice compared to littermate controls treated with vehicle solution. Interestingly, a higher ratio of microglia internalized Aβ in the brain of M-CSF-treated animals and the phagocytosed peptides were located in the late endosomes and lysosomes. Less Aβ40 and Aβ42 monomers were also detected in the extracellular protein enriched fractions of M-CSF-treated transgenic mice when compared with vehicle controls. Finally, treating APPSwe/PS1 mice that were already demonstrating installed Aβ pathology stabilized the cognitive decline. Together these results provide compelling evidence that systemic M-CSF administration is a powerful treatment to stimulate bone marrow-derived microglia, degrade Aβ and prevent or improve the cognitive decline associated with Aβ burden in a mouse model of Alzheimer's disease.

  • Alzheimer's disease
  • β-amyloid
  • microglia
  • M-CSF


Alzheimer's disease is a neurodegenerative disorder that represents the most important cause of dementia in humans. Extracellular deposits of β-amyloid peptides (Aβ), often termed senile plaques, and formation of intracellular neurofibrillary tangles of hyperphosphorylated tau protein are the two principal hallmarks of this disease. Aβ aggregates are known to induce synaptic dysfunction, and thus are linked with learning and memory deficits in both human and mouse models of the disease, making Aβ deposits a target for prevention or treatment (Citron et al., 1996; Selkoe, 2001; Mattson, 2004).

Microglia are attracted to Aβ aggregates and decorate plaques. Such a phenomenon has been observed in both human and transgenic mouse model of Alzheimer's disease, suggesting an important role for these cells in the CNS (Dickson et al., 1988; Haga et al., 1989; Itagaki et al., 1989; Perlmutter et al., 1992; Ard et al., 1996; Sheng et al., 1997; Frautschy et al., 1998; Wegiel et al., 2001, 2003, 2004; Nicoll et al., 2003; Malm et al., 2005). While some studies have demonstrated that in some circumstances, microglia can be detrimental in the Alzheimer's disease brain (Giulian et al., 1996; Kim and de Vellis, 2005; Walker and Lue, 2005), many others have supported the theory that they are actually beneficial (Nguyen et al., 2002; Turrin and Rivest, 2006). Indeed, stimulation of the immune system reduced Aβ burden (Rogers et al., 2002; Malm et al., 2005). Aβ is phagocytosed, delivered to the lysosome (Paresce et al., 1997; Chung et al., 1999) and degraded in microglia (Morgan et al., 2000; Rogers and Lue, 2001; Rogers et al., 2002; Hartman et al., 2005; Liu et al., 2005). However, primary culture of mouse microglia indicates that these resident immune cells of the CNS need to be activated prior to gaining the ability to clear Aβ from the brain.

Macrophage colony-stimulating factor (M-CSF) is a cytokine that promotes proliferation and activation of microglia (Giulian and Ingeman, 1988), allowing their differentiation into macrophage-like cells (Santambrogio et al., 2001; Monsonego and Weiner, 2003). Interestingly, low M-CSF levels were recently found in patients with presymptomatic Alzheimer's disease or mild cognitive impairment (MCI), which together with low levels of other haematopoietic cytokines predicted the rapid evolution of the disease toward a dementia state 2–6 years later (Ray et al., 2007). In vitro, exposing mouse microglia to M-CSF enables the acidification of their lysosomes and subsequently, the degradation of internalized Aβ (Majumdar et al., 2007). To investigate whether M-CSF treatment could have beneficial effects on cognitive impairment and amyloid burden in a mouse model Alzheimer's disease, we treated APPSwe/PS1 mice with M-CSF as the disease pathology progressed and found that M-CSF efficiently prevented cognitive impairment and Aβ burden in the CNS.

Materials and Methods

Animals and injections

The animals used in this study were all male transgenic mice bearing a chimeric human/mouse β-amyloid precursor protein (APPSwe) gene and the human presenelin 1 gene (A246E variant). These mice were purchased from The Jackson Laboratory [Strain: B6C3-Tg(APP695)3Dbo Tg(PSEN1)5Dbo/J] and maintained on a C57BL/6J background. Animals were injected i.p. weekly with either NaCl 0.9% (130 μl) or a solution containing 10 ng/μl mouse M-CSF (final injection: 40 μg/kg, 130 μl).

All animals were acclimated to standard laboratory conditions (14-h light, 10-h dark cycle; lights on at 06:00 and off at 20:00 h) with free access to rodent chow and water. Protocols were conducted according to the Canadian Council on Animal Care guidelines, administered by the Laval University Animal Welfare Committee.

Behavioural analyses

In this study, we used a battery of behavioural tasks to test the general and cognitive health. Mice were weighed every week throughout the experimental period. The behavioural and general health of mice was regularly performed using a modified version of the primary observational screen described in the SHIRPA protocol (Rogers et al., 1997). The experimenter who observed and recorded the behaviour was not aware of treatment and genotype of the tested animals [WT, APPSwe/PS1 + NaCl 0.9% (APP + Saline, n = 15), APPSwe/PS1 + M-CSF 40 μg/kg (APP + M-CSF, n = 13)]. Baseline data were obtained at 3 or 6 months of age, whereas the effects of the treatment were determined at 6 or 9 months (See experimental timeline Figs 1A and 8A).

Figure 1

M-CSF-injected transgenic APPSwe/PS1 mice show normal behaviour in the T-water maze test. (A) Timeline of preventive experiment using M-CSF injection in transgenic mice. (B) Trials in the reversal phase in T-water maze task for WT (n = 15) and 6-month-old APPSwe/PS1 mice treated with saline (n = 15) or M-CSF (n = 13) from age 2 months. Also displayed are the escape latencies during spatial learning in T-water maze task. Data are expressed as mean ± SEM (One-way ANOVA: **P < 0.01, ***P < 0.001). (C) Scores for the nesting behaviour measured before (pre-drug) and after (post-drug) treatment with M-CSF. (D) Mouse body weights were taken every week and the mean of each group was plotted.

The T-water maze was used to measure the spatial learning and memory of the mice. The apparatus was made of transparent plastic (length of stem 64 cm, length of arms 30 cm, width 12 cm and height of walls 16 cm) and filled with water (23 ± 1°C) at a height of 12 cm. A small escape platform (11 × 11 cm) was placed 1 cm beneath water level. When reaching the platform, mice were allowed to stay on it for 20 s. During the acquisition phase, the platform was placed either in the right or the left arm for all trials (animals in each group were equally distributed between the two sides). Mice were repeatedly placed at the entrance of the T-maze until they successfully located and climbed onto the hidden platform five times successively. The number of trials required until each animal met this criterion and the latency to find the platform was recorded. Whenever the mice failed to reach the escape platform within the maximally allowed time of 60 s, it was gently steered in that direction and placed on it.

During the reversal phase, 48-h later, the location of the platform was reversed to the other side of the T-maze. The latency and number of trials required until each animal successfully learn the new location of the platform were again recorded.

Thereafter, the nesting behaviour was used to test for changes in emotional status (e.g. apathy). Reduced nesting has been observed in hippocampal lesioned mice and mouse models of Alzheimer's disease (Deacon, 2006). Animals were individually housed in a cage containing sawdust and in which a 5 × 5 cm piece of cotton was introduced to allow nesting behaviour. One day later, the quality of the nest was determined according to a five-point scale as described by Deacon (2006): 1—Nestlet apparently untouched, 2—Nestlet partially torn up, 3—Nestlet mainly shredded but no apparent presence of nesting site, 4—Observable flat nest, 5—Observable (near) perfect nest.

Finally, one trial step-through procedure was investigated to measure the effect of M-CSF on retention of weaker memory formation acquired after one single shock in the passive avoidance paradigm, as previously described in Richard et al. (2008). Briefly, the apparatus (Ugo Basile) used for this test consisted of two sections, one being illuminated (which is also the start compartment) and the other one being in full darkness (known as the escape compartment). The floor for each section consisted of a metallic grid. Using a generator, the dark compartment's grid was electrified. On the first day (training day), mice were placed in the illuminated compartment for a period of 60 s and the door separating both sections was opened. The latency to enter to the dark compartment was then recorded. Upon entering the dark section, the animal received a mild foot shock (0.5 mA, 2 s) and kept in it for 10 s before returning to its home cage. On the following day, animals were once again placed in the lighted compartment, and the step through latency before re-entering the dark side was measured up to 300 s.

APPSwe/PS1-GFP chimeric mice creation by irradiation and bone marrow transplantation

Two-month-old APPSwe/PS1 mice were irradiated with a total of 10 grey in a cobalt-60 source (Theratron-780, MDS Nordion). Six hours later, mice were injected by the tail vein with an amount of 5 × 106 fresh green fluorescent protein (GFP) positive bone marrow cells obtained by flushing extracted femur from GFP mice in a sterile environment. Cells were harvested using Dulbecco's PBS (DPBS) complemented with 2% foetal bovine serum (FBS). To remove clumps from the extract, the preparation was filtered using a 40 μm nylon filter and peleted. The supernatant was removed and cells re-suspended in DPBS before being chased through a 25 g syringe needle. Cells were centrifuged and re-suspended in fresh DPBS for a final concentration of 5 × 106 cells for a volume of 200 μl. Mice used in this experiment were housed in sterile cages with irradiated food. One week before the irradiation, and up to 2 weeks after the transplantation, mice were given drinking water complemented with 0.2 mg/ml trimethoprim and 1 mg/ml sulfamethoxazole.


Free-floating sections (25 µm thick) were incubated for 30 min in KPBS containing 4% goat serum, 1% BSA and 0.4% Triton X-100. Using the same buffer solution diluted 1/2 in KPBS, the sections were incubated for 120 min in primary antibody (rabbit anti-Iba1, 1:3000, Wako Chemicals; mouse anti-Aβ, clone 6E10, 1:3000, Chemicon International; rat anti-LAMP-2, 1:500, Stressgen Bioreagents) at room temperature. The sections were then rinsed 3 × 10 min in KPBS, followed by a 120 min incubation in fluorochrome-conjugated goat secondary antibody (anti-rabbit Alexa 488, 1:1000, Molecular Probes; anti-mouse Cy3, 1:1000, Jackson Immuno Research Laboratories; anti-mouse Alexa 546, 1:1000, Invitrogen; anti-mouse Alexa 488, 1:1000, Molecular Probes; anti-rat Alexa 568, 1:1000, Invitrogen). Sections were rinsed 3 × 10 min in KPBS, mounted onto SuperFrost slides (Fisher Scientific, Nepean, ON, Canada), stained with DAPI (2e−4%; Molecular Probes) and coverslipped with Fluoromount-G (Southern Biotech).

Plaques and microglia quantification

Immunostained brain sections (12th section through the hippocampus region) were analysed using a stereological apparatus (n = 4 for each group). Slices were stained for microglia (anti-Iba1, 1:3000, Wako Chemicals), Aβ (anti-Aβ, clone 6E10, 1:3000) and nuclei (DAPI, 2e−4%). A total of four animals per group were analysed. Two sections were chosen at –1.70 and –3.08 mm from the bregma according to a stereotaxic atlas (Paxinos and Franklin, second edition). Real-time images (1600 × 1200 pixels) were obtained using a Nikon C80i microscope equipped with both a motorized stage (Ludl) and a Microfire CCD color camera (Optronics). Such an apparatus was operated using the Stereo Investigator software designed by Microbrightfield. Both cortex and hippocampus areas were traced using a Cintiq 18S interactive pen display (Wacom). The sampling method was previously tested in a pilot study to ensure correct estimation in the number of quantified elements. First, the Stereo Investigator software was used to delimit counting frames of 670 × 500 μm, and located at each 2000 μm in the x-axis and at every 1000 μm in the y-axis of the previously selected cortex and hippocampus regions. Aβ plaques were traced and microglia counted for each frame using the pen display, a 40× Plan Apochromat objective (NA 0.95) and a triple-band filter (DAPI/FITC/TRITC, Chroma Technology). Iba-1-immunoreactive cells were counted only when their nuclei (DAPI labelled) were in the dissector area (20 mm) and not in contact with the two forbidden contours of the counting frame. Plaques were counted and encircled only when they were not intersecting forbidden lines. Although our method did not follow all the principles of unbiased stereological procedure, it generated semi-quantitative data highly representative of the general state of the animal. Moreover, this method was designed using a pilot experiment that ensured that this type of quantification was representative of the total amount of plaques and microglia determined by unbiased stereological quantification (Simard et al., 2006). Results are expressed as mean ± SEM.

Confocal laser scanning microscopy

Confocal laser scanning microscopy was performed using a BX-61 microscope equipped with the Fluoview SV500 imaging software 4.3 (Olympus America, Inc, Melville, NY). Images were obtained, stacked and analysed by sequential scanning optimized by a two-frame Kalman filter. Scans were performed on layers 100 nm apart and pictures were adjusted for optimal viewing via Photoshop CS2 (Adobe system).

Protein extraction and detection

Proteins from hemi-forebrains were extracted using a modified method of the procedure published by Lesne et al. (2006). All manipulations were done on ice to minimize protein degradation. One hemi-forebrain was placed in a 1 ml syringe with a 20 G needle. Five hundred microlitres of buffer A [50 mM Tris–HCl pH 7.6, 0.01% NP-40, 150 mM NaCl, 2 mM EDTA, 0.1% SDS, 1 mM phenylmethylsulfonyl fluoride (PMSF), protease inhibitor cocktail] were added and 10 up and down were made to homogenize the tissue, followed by a 5 min centrifugation at 3000 r.p.m. at 4°C. The supernatant (extracellular proteins enriched fraction) was then collected and frozen at −80°C. The insoluble pellet was suspended in 500 μl TNT-buffer (Buffer B; 50 mM Tris–HCl pH 7.6, 150 mM NaCl, 0.1% Triton X-100), followed by a 90 min centrifugation at 13 000 r.p.m. at 4°C. The supernatant (cytoplasmic proteins enriched fraction) was then collected and frozen at –80°C. The pellet was suspended in 500 μl buffer C (50 mM Tris–HCl pH 7.4, 150 mM NaCl, 0.5% Triton X-100, 1 mM EGTA, 3% SDS, 1% deoxycholate, 1 mM PMSF, protease inhibitor cocktail) and incubated at 4°C, 50 r.p.m. for 1 h. The samples were centrifuged for 90 min at 13 000 r.p.m. at 4°C and the supernatant (membrane proteins enriched fraction) was collected and frozen at –80°C. The remaining pellet was suspended in 20 μl of 70% formic acid followed by addition of 380 μl Tris–HCl 1 M pH 8.0. Samples were incubated at 4°C, 50 r.p.m. for 1 h, centrifuged for 90 min at 13 000 r.p.m. at 4°C and the supernatant (insoluble proteins enriched fraction) was collected and frozen at –80°C. Protein concentration of each fraction was determined using the Quantipro BCA assay kit (Sigma) according to the manufacturer protocol.

For Aβ42 and total Aβ detection, 30 μg of extracellular, cytoplasmic and membrane proteins fractions were separated on a 16% Tris–Glycine polyacrylamide gel and on a 10–20% gradient precast tricine polyacrylamide gel (Biorad), respectively. Separated proteins were then transferred to a nylon membrane and detected by Western blotting. Rabbit polyclonal anti-Aβ 1–42 (1:1000, Abcam) was used for Aβ42 detection and mouse anti-amyloid beta protein monoclonal antibody (1:500, Chemicon International) for total Aβ. Quantification was done by determining density of the bands using the ImageJ software (http://rsb.info.nih.gov/ij/). Optical values were normalized according to the actin (mouse anti-actin monoclonal antibody, 1:3000, Chemicon International) loading control. Results are expressed as mean ± SEM.

Statistical analysis

The normality of our samples was assessed using the d’Agostino-Pearson omnibus test. All statistical analyses were performed using the Prism 5 software (GraphPad Software). Behavioural data were subjected to ANOVA followed by Bonferroni's post hoc tests. Intergroup differences were analysed using standard two-tailed unpaired t-test and results expressed as mean ± SEM.


M-CSF treatment prevents the progression of social withdrawal and cognitive impairment

To investigate whether injection of M-CSF could prevent the appearance of neurological impairment, saline and M-CSF injected APPswe/PS1 and their littermate controls mice were tested at 3 and 6 months of age for spatial learning in a T-water maze and nesting behaviour (See experimental time line Fig. 1A). We recently reported that APPSwe/PS1 mice have normal performance at 3 months of age and impaired performance by 6 and 9 months of age when compared with wild-type (WT) control mice during the reversal phase of the T-water maze and in the memory retention of the passive avoidance task (Richard et al., 2008). These tasks are therefore quite sensitive to the behavioural impairments associated with Aβ accumulation in the brain of APPSwe/PS1 mice.

In the T-water maze, spatial memory was assessed by the escape latency in the hidden platform test of blocks of four trials and by the mean of trials to reach the criterion. The escape latency or the mean of trials to criterion during the acquisition and reversal session were not significantly altered in 3-month-old mice (Fig. 1B). The initial test during the acquisition phase revealed no significant differences in terms of swimming performance or motivation to escape from the water in 6-month-old mice. However, the analysis of escape latencies during reversal training revealed significant effects for the group factor [F(2,80) = 19.4, P < 0.0001], trial block factor [F(3,120) = 78.0, P < 0.0001] and the interaction (F(6,120) = 2.6, P = 0.021) in 6-month-old APPSwe/PS1 mice. Escape latencies of WT and APP + M-CSF mice decreased across trial blocks. A Bonferroni post hoc analysis indicated that escape latency in blocks 6, 7 and 8 for the APP + Saline was higher than that of WT mice (P < 0.05) and that M-CSF treatment significantly prevented the increase of escape latency in the 6-month-old transgenic mice. Moreover, APP + Saline mice had higher trials to criterion than WT and APP + M-CSF groups (Fig. 1B, t28 = 3.6, P = 0.0012 and t26 = 3.3, P = 0.0029, respectively). Both groups (WT and APP + M-CSF) exhibited intact spatial memory, because the escape latencies and the trials to criterion improved across training. M-CSF treatment prevented cognitive decline in APPSwe/PS1 transgenic mice.

The nesting behaviour was conducted to investigate the effect of M-CSF on the social withdrawal and apathy associated with Alzheimer's disease. As depicted by the Fig. 1C, the nest scores were similar in both 3-month-old WT and APPSwe/PS1 groups of mice. However, 6-month-old saline-treated APPSwe/PS1 mice had a lower quality of nest building compared with both WT and APP + M-CSF groups (Fig. 1C, t28 = 3.5, P = 0.0017 and t26 = 4.3, P = 0.0002 respectively). These results indicated that APPSwe/PS1 mice treated with M-CSF exhibited a significant decrease in apathy as indicated by better nesting scores.

Body weight was monitored weekly to assess the health of all animals. Interestingly, using a two-way ANOVA, we could observe a significant effect of time and treatment on the average body weight between saline (NaCl 0.9%; n = 15) or M-CSF (40 μg/kg; n = 13) injected animals (Fig. 1D) (P < 0.001). Thus, the M-CSF treated animals exhibited a small increase in body weight compared with the saline-injected controls. In addition the general health of each animal was determined using the Shirpa primary screening test (Rogers et al., 1997), showing no difference between each group. These results indicate that M-CSF treatment significantly prevented the behavioural deficit associated with Aβ pathology without causing obvious side effects on the general health.

M-CSF injection increases the number of microglia

M-CSF is a key cytokine implicated in the proliferation of monocytic cell lineages such as macrophages and microglia. To determine the number of microglia, 25-μm thick brain sections were labelled using a polyclonal rabbit anti-Iba1 (anti-ionized calcium binding adaptor molecule 1) antibody. Iba1-immunoreactive cells were more numerous in the brain of M-CSF-treated mice than saline-treated animals (Fig. 2A). They were counted in the cortex and hippocampus using a stereological apparatus and the total number reported on their respective area was multiplied by 100 to provide a general index of microglia load (Fig. 2B). Compared to saline-treated mice, both cortex (0.01873 ± 0.00072% versus 0.04488 ± 0.00284%) and hippocampal (0.01319 ± 0.00171% versus 0.03499 ± 0.00270%) areas of M-CSF-treated animals exhibited a significant increase of 2.39- and 2.65-fold in microglial cell number, respectively. These data demonstrate that long term M-CSF treatment augmented the population of microglia in the CNS of APPSwe/PS1 mice.

Figure 2

M-CSF stimulates microglia proliferation and infiltration in the brain of APPSwe/PS1 mice. Immunofluorescence using a primary antibody directed against Iba1 in conjunction with an Alexa 488-conjugated secondary antibody was used to label microglia. (A) The general pattern of microglia in brain sections of saline- and M-CSF-injected transgenic APPSwe/PS1 mice (n = 6 for each group) is depicted by confocal microscopic images (Scale bar is 50 μm). (B) The number of microglia in the cortex and the hippocampus were determined by unbiased stereology procedure. Data are expressed as mean ± SEM (Student t-test: ***P < 0.001). (C) GFP-positive microglia in the brains of 6-month-old APPSwe/PS1-GFP chimeric mice that were treated with either saline or M-CSF from age 2 months. Please note the marked accumulation of bone marrow-derived microglia in the brain of APPSwe/PS1 chimeric mice following chronic M-CSF administrations. (First line: Scale bar is 100 μm, Second line: Scale bar is 50 μm).

To determine whether the increased number in microglia number was due to infiltration of bone marrow derived cells in the CNS, we created 2-month-old APPSwe/PS1-GFP chimeras and treated them with either NaCl 0.9% (n = 7) or M-CSF 40 μg/kg (n = 5) for a period of 4 months. Interestingly, most treated animals showed massive infiltration of bone marrow derived microglia both in the cortex and hippocampus, in comparison to the saline controls (Fig. 2C). Although this observation does not rule out the possibility for self-renewal of microglia, it provides direct evidence that a large proportion of microglia derives from the bone marrow donor cells in response to systemic M-CSF treatment.

M-CSF-injected mice exhibited less Aβ deposits than control animals

Cognitive impairment is correlated with the appearance of Aβ deposits in the cortex and hippocampus of APPSwe/PS1 mice. In this model, plaques are known to be well established and visible at 6 months of age. We hypothesized that since M-CSF-treated mice performed well in the T-water maze task, they should have less Aβ deposits in the cortex and hippocampus. Brain sections for both groups were stained using the mouse anti-Aβ (clone 6E10) antibody. Both immunofluorescence (data not shown) and laser scanning confocal microscopy indicated less Aβ deposits (Red) in the brain of M-CSF/APP than Saline/APP mice (Fig. 3A). Aβ load was quantified using a stereological apparatus and the ratio between the area occupied by Aβ deposits and the total area of the analysed region was multiplied by 100 (Fig. 3B). Aβ load was significantly lower (P = 0.0003) in the brain of M-CSF-treated animals (0.3010 ± 0.0808%) than littermate controls (1.121 ± 0.109%), resulting in a 4.0-fold decrease in the cortex. Similar results were obtained in the hippocampus (P = 0.0117) where a 3.76-fold decrease was quantified in M-CSF-treated mice (0.2360 ± 0.0721%) compared to the Saline/APP group (0.8883 ± 0.2035%).

Figure 3

Chronic M-CSF treatment decreases Aβ deposits in the brains of APPswe/PS1 mice. (A) Brain sections were stained using a primary antibody directed against Aβ (6E10) in conjunction with an Alexa 546-conjugated secondary antibody. Senile plaques are depicted by confocal microscopic images (n = 6 for each group) (Scale bar is 50 μm). (B) The relative area occupied and number of Aβ plaques were determined by unbiased stereology in the cortex and hippocampus of saline- and M-CSF-injected transgenic APPswe/PS1 mice. Data are expressed as mean ± SEM (Student t-test: *P < 0.05, ***P < 0.001).

The total number of plaques was also determined and normalized according to the total area screened by our approach (Fig. 3B). The total area of Aβ deposits decreased in M-CSF-treated animals, and the relative number of plaques significantly decreased by 2.06-fold in the cortex (P = 0.0146) and 3.09-fold in the hippocampus (P = 0.0297). Ratios varied from 4.096e−4 ±2.254e−5 to 1.992e−4 ± 4.589e−5 and from 4.307e−4 ± 1.144e−4 to 1.395e−4 ± 1.1151e−5 for each region, respectively. These data provide very clear evidence that early and chronic M-CSF treatment prevents the formation and/or clears Aβ deposits in the CNS.

M-CSF enhanced microglial association with Aβ deposits and internalization

We next wanted to determine the relationship between microglia and Aβ deposits as previously reported by Simard and colleagues (2006). Microglia (Green) and Aβ deposits (Red) were immunostained using antibodies directed against Iba1 and Aβ (clone 6E10), respectively. Immunofluorescence revealed a close association between microglia and Aβ plaques in the cortex and the hippocampus of APP mice (Fig. 4A). Interestingly, M-CSF-treated animals had less dense plaques and these Aβ deposits were often devoid of microglia (data not shown). Indeed, the ratio of microglia present in Aβ deposits was higher in the control group (Fig. 4B). In the cortex, microglia represented 11.45 ± 1.10% of total Iba1 immunoreactive cells in saline-treated mice, whereas they represented 4.22 ± 1.19 in M-CSF-treated mice (P = 0.0042). A 3.24-fold decrease was found in the hippocampus (P = 0.0437). It is interesting to note that no differences were found between saline or M-CSF groups when the number of plaque associated microglia was compared to the total analysed area (data not shown). This suggests that the same amount of microglia was recruited to Aβ deposits in both conditions. However, plaques were less abundant in the brain of M-CSF-treated animals, which resulted in a higher ratio of microglia associated to Aβ plaques. Such a relative abundance was increased by 3.64 (P = 0.0003) and 2.65 (P = 0.0003) fold in the cortex and hippocampus of M-CSF-treated APPSwe/PS1 transgenic mice, respectively (Fig. 4B).

Figure 4

The number of Aβ associated microglia is increased following repeated M-CSF injections in a mouse model of Alzheimer's disease. (A) Aβ deposits and microglia were immunostained using an anti-Aβ (6E10) (Red) and anti-Iba1 (Green) antibodies on coronal sections of APPswe/PS1 mice treated with saline or M-CSF. Plaques associated (A) microglia were observed using confocal microscopy (n = 6 for each group) (Scale bar is 50 μm). (B) Plaques associated microglia load were measured using a stereological apparatus on immunofluorescent brain sections. Data are expressed as mean ± SEM (Student t-test: *P < 0.05, ***P < 0.001).

Laser scanning confocal microscopy allowed us to visualize Aβ into microglia showing that the number of Aβ-positive microglia is increased by M-CSF in the cortex and hippocampus (Fig. 5A and B). A 1.62 (P < 0.0001) and 1.30 (P = 0.0001) fold increase was measured in both regions, respectively, with values ranging from 45.15 ± 2.63% to 74.77 ± 1.25% and 54.15 ± 1.07% to 70.61 ± 1.49%. Taking into account the total area analysed, the presence of Aβ-containing microglia also markedly increased in the brain of M-CSF-treated animals. The ratios in whole cortex and hippocampus were significantly higher in those mice, ranging from 8.664e−3 ± 7.652e−4 to 3.347e−2 ± 1.920e−3 and from 7.093e−3 ± 8.107e−4 to 2.465e−2 ± 2.072e−3, revealing a 3.86 (P < 0.0001) and a 3.48 (P = 0.0002) fold increase.

Figure 5

The number of Aβ-containing microglia increased following repeated M-CSF injections in a mouse model of Alzheimer's disease. (A) Aβ deposits and microglia were immunostained using an anti-Aβ (6E10) (Red) and anti-Iba1 (Green) antibodies on coronal sections of APPswe/PS1 mice treated with saline or M-CSF. (A) Aβ-containing microglia were observed using confocal microscopy (n = 6 for each group) (Scale bar is 50 μm). (B) Aβ containing microglia load were measured using a stereological apparatus on immunofluorescent brain sections. Data are expressed as mean ± SEM (Student t-test: ***P < 0.001).

The distribution of microglia and their association with Aβ deposits were clearly changed in response to the M-CSF treatment. Although the ratio of Aβ deposits associated with microglia was higher in Saline/APP animals, more microglia were associated with fibrillar Aβ in the brain of M-CSF-treated APPSwe/PS1 mice. Moreover, this latter group had more Aβ-containing microglia suggesting a pronounced effect of M-CSF on the elimination of Aβ by microglia.

Brains of M-CSF injected animals contain less Aβ monomers

Using a modified method previously published by Lesne et al. (2006), extracts enriched in extracellular, cytoplasmic and membrane proteins were obtained from hemi-forebrains of APPSwe/PS1 mice. Our first attempt was to determine the relative abundance of Aβ isoforms, especially the monomer, mainly consisting of Aβ40 and Aβ42 peptides. Anti-Aβ (Clone 6E10) was used to reveal all isoforms by Western blotting. As depicted in Fig. 6A, less Aβ monomers were detected in the extracellular fraction from the brains of M-CSF-treated transgenic animals compared to vehicle controls. Quantification by densitometry revealed a 75 and 55% decrease in the monomer levels after chronic M-CSF treatment in the extracellular (P = 0.0034) and membrane fraction (P = 0.0407), respectively (Fig. 6A). However, monomer levels in the cytoplasmic fraction remained unchanged by the haematopoietic cytokine.

Figure 6

The brains of M-CSF-treated APPSwe/PS1 mice contain less extracellular Aβ. Extracts enriched from extracellular (Extra), cytoplasmic (Cyto) and membrane (Mem) proteins were separated on a 10–20% Tris–Tricine denaturing polyacrylamide gel to reveal Aβ isoforms (anti-Aβ antibody, clone 6E10) present in the brains of saline- (S) and M-CSF- (M) injected APPSwe/PS1 mice. (A) Aβ monomer levels were quantified by densitometric analysis (n = 4 for each group, **P = 0.0054 and *P = 0.0407 for extracellular and membrane fractions respectively). (B) Aβ42 levels were then determined by Western blot and quantified by densitometry for both extracellular and cytoplasmic fractions extracted from saline- (S) and M-CSF- (M) treated brains (n = 4 for each group, ***P = 0.0009 for extracellular fraction).

We then determined Aβ42 levels in extracellular and cytoplasmic fractions using a specific anti-Aβ42 antibody (Fig. 6B). Again, Aβ42 levels were decreased by 31% in the forebrain of M-CSF-treated animals compared with their littermate saline-treated controls (P = 0.0009). Aβ42 levels were similar in the cytoplasmic fractions extracted from the brains of both groups of transgenic mice. These analyses of protein levels show that M-CSF treatment is effective at targeting Aβ where it accumulates in the extracellular space.

Aβ is internalized by microglia and is present in the late endosome

M-CSF injection stimulated microglia proliferation, decreased extracellular Aβ and increased Aβ-containing microglia in APPSwe/PS1 transgenic mice. As previously demonstrated (Simard et al., 2006), a sub-population of microglia has the ability to internalize Aβ making them very good candidates to clear the toxic protein from the CNS (Fig. 7A). Of great interest are the recent data that M-CSF causes acidification of microglial lysosomes, thus promoting Aβ degradation (Majumdar et al., 2007). The cell body of microglia, Aβ (Red) and LAMP-2 (Blue) highly co-localized (Purple), as determined by confocal microscopy (Fig. 7B). We concluded from this observation that microglia may be important players in the clearance of Aβ via endocytosis and lysosome-mediated degradation following M-CSF treatment in APPSwe/PS1 mice.

Figure 7

Aβ is internalized by microglia and is present in late endosomes and lysosomes. (A) Brain sections of 6-month-old saline and M-CSF-administered transgenic mice were stained for Iba1 (Green), Aβ (6E10 antibody) (Red) and DAPI (Blue). Colocalization of Aβ with microglia was determined by confocal microscopy, (Scale bar is 10 μm). (B) Brain sections were then immunostained for Aβ (Red) and Lamp-2 (Blue), a late endosome and lysosome marker. Colocalization of amyloid deposits in endosomes of microglia cells was confirmed by confocal microscopy (Scale bar is 10 μm).

M-CSF treatment reverses the neurological behavioural abnormalities of mice already harbouring Alzheimer's disease-like pathology

Since M-CSF injection enabled us to prevent the development of Aβ deposits and the apparition of behavioural impairment in 6-month-old APPSwe/PS1 transgenic animals, we next wanted to determine whether the use of this haematopoietic cytokine could exert similar effects in animals already exhibiting behavioural defects and Alzheimer's disease-like pathology. To do so, we compared APPSwe/PS1 mice that were treated from 6 to 9 months. Thus we studied the effects of M-CSF in APPSwe/PS1 mice after the onset of Aβ pathology and behavioural changes that are detectable by the T-water maze, nesting activity and passive avoidance task (Fig. 8).

Figure 8

APPSwe/PS1 mice with installed Aβ deposits and treated with M-CSF exhibit a stabilized cognitive state. (A) Timeline of therapeutic experiment using M-CSF injection in transgenic mice. (B) Trials in the reversal phase in T-water maze task for 6- and 9-month-old APPSwe/PS1 mice treated with saline (n = 12) or M-CSF (n = 13). Also displayed are the escape latencies during spatial learning in T-water maze task. Data are expressed as mean ± SEM (one-way ANOVA). (C) Scores for the nesting behaviour measured before (pre-drug) and after (post-drug) treatment with M-CSF. Data are expressed as mean ± SEM (One-way ANOVA). (D) Step-through latencies latencies measured during the passive avoidance before (pre-drug) and after (post-drug) the treatment with M-CSF. Data are expressed as mean ± SEM (One-way ANOVA). (E) Efficacy of M-CSF treatment on both M-CSF and saline groups based on the difference in trials needed in the T-water maze. (F) Overall behavioural performance of the treated animals based on a minimum difference of six trials when compared with their assessments before the treatment.

No significant difference was measured between the three groups of 6-month-old mice during the acquisition phase of the T-water maze. However the analysis of escape latencies during reversal training revealed significant effects for the group factor [F(2,76) = 11.8, P < 0.0001], trial block factor [F(3,114) = 36.6, P < 0.0001] and the interaction between the two variables [F(6,114) = 3.2, P = 0.006]. Escape latencies of WT mice decreased across trial blocks, suggesting better performance on the spatial task. Indeed APPSwe/PS1 transgenic animals had higher trials to criterion than WT (Fig. 8B, t27 = 3.9, P = 0.0005 and t26 = 3.6, P = 0.0012, respectively). These results indicate that spatial memory during the reversal phase of the T-water maze was significantly impaired in APPSwe/PS1 mice.

There was no difference during the acquisition phase between the three groups after 3 months of treatment, but the analysis of escape latencies during reversal training revealed significant effects for the group factor [F(2,76) = 13.4, P < 0.0001], trial block factor [F(3,114) = 42.1, P < 0.0001] and interaction [F(6,114) = 3.6, P = 0.03]. Escape latencies of WT and APP + M-CSF mice decreased across trial blocks, whereas those of APP + Saline mice increased. A Bonferroni multiple comparison tests indicated that escape latency in blocks 6, 7 and 8 for the APP + Saline was higher than that of WT mice (P < 0.05). Post hoc tests analysis also indicated that escape latency in blocks 7 and 8 for the APP + Saline was higher than APP + M-CSF mice (P < 0.05). Moreover, APP + Saline mice had higher trials to criterion than APP + M-CSF mice and WT mice (Fig. 8B, t23 = 2.4, P < 0.03 and t26 = 4.73, P < 0.0001, respectively). Taken together, these results indicate that spatial memory during the reversal phase of the T-water maze was significantly impaired in APP + Saline treated mice compared to APP + M-CSF treated mice and WT mice. Our results demonstrate that such deficit of spatial memory can be alleviated by M-CSF treatment.

The overall performance in T-water maze was compared for each animal. Final and initial performances in the reversal phase of the task were plotted (Fig. 8E). All animals were classified in one of the three categories based on the difference between their final and initial performance in the reversal phase of the T-water maze task. If animals had a decrease of at least six trials, they were placed in the improvement category and they were considered stable if they had no difference of more or less six trials. They were placed in the deteriorated category if they had more trials than their initial states (Fig. 8F). Although APP + Saline mice did not exhibit any improvement of the cognitive impairment, 38.5% of the M-CSF treated group performed better than their initial states. Moreover, 23% of APP + M-CSF mice were placed in the deteriorated category and 38.5% were considered as being stable. Together these data provide evidence that M-CSF treatment is very efficient to alleviate the cognitive deficit associated with Aβ neuropathology in aging APPSwe/PS1 transgenic animals.

The quality of nest building was significantly lower in 6-month-old APPSwe/PS1 mice (Fig. 8C, t27 = 3.87, P = 0.0006 and t26 = 3.2, P = 0.0034). Such a behaviour was improved in this mouse model of AD after 3 months of M-CSF treatment (Fig. 8C, t26 = 4.5, P = 0.0001 and t23 = 3.86, P = 0.0008). Both M-CSF and WT groups had similar nesting activity (Fig. 8C, t27 = 0.37, P = 0.72) suggesting a better social interaction and more motivation for daily living activity.

The passive avoidance is another indicator for non-spatial memory function. As depicted by the Fig. 8D, retention latencies were lower in 6-month-old APPSwe/PS1 mice than WT animals (t27 = 3.32, P = 0.0026 and t26 = 2.8, P = 0.011, respectively). Three months later, APP + Saline mice exhibited a significant poorer retention score when compared to WT and APP + M-CSF animals (Fig. 8D, t26 = 4.3, P = 0.0002 and t23 = 3.1, P = 0.0056, respectively). A similar performance was measured in both APP + M-CSF and WT groups (Fig. 8D, t27 = 1.1, P = 0.3) suggesting a strong beneficial effect of M-CSF on memory consolidation.


M-CSF treatment was extremely effective in preventing the cognitive decline associated with Aβ burden in APPSwe/PS1 mice at a critical age of the disease. Indeed, mice treated weekly with M-CSF (40 μg/kg) performed as well as WT mice in the T-water maze task. Of clinical importance is the lack of obvious side effects of repeated injections with the cytokine, which did not seem to be detrimental to the general health of the animals as measured by the acquisition capacity in the T-water maze task and the Shirpa screening test. Taken together these observations are very exciting and open new possibilities for a potential and effective new treatment for Alzheimer's disease.

The number of microglia greatly increased in the cortex and hippocampus of M-CSF-treated animals, suggesting a direct effect of the haematopoietic cytokine on the proliferation of these cells. Microglia have the ability to proliferate during immune stimuli and CNS injuries, but the origin of these progenitors still remains to be elucidated. We propose that these new microglia largely derive from the bone marrow in our model for the following reasons:

  1. A large number of GFP-positive microglia were found in the brain of chimeric APPSwe/PS1 mice following the chronic treatment with M-CSF;

  2. Bone marrow-derived microglia (BMDM) are quite efficient at internalizing and phagocytozing Aβ in this mouse model of Alzheimer's disease (Simard et al., 2006), in contrast to the resident microglia;

  3. Inhibition of BMDM exacerbates the disease progression and Aβ burden;

  4. The pharmacokinetic profile of 40 μg/kg M-CSF strongly indicated a Cmax and exposure consistent with a systemic effect on the haematopoeitic cells rather than a direct effect within the CNS (data not shown); and

  5. Gene deletion of CCR2 in BMDM accelerates the cognitive decline in APPSwe/PS1 mice (Naert G et al., unpublished results).

CCR2 is a key receptor for the diapedesis of myeloid cells and breeding of CCR2-deficient mice with APP transgenic animals impairs microglial accumulation and accelerates disease progression (El Khoury et al., 2007). Together these data point toward haematopoietic progenitors as being a primary target for M-CSF.

Concomitant with increased microglial cell number, the number of Aβ plaques and the area occupied by such deposits were significantly lower in the brain of M-CSF-treated mice. Their plaques were in most cases smaller and obviously less dense than those of littermate controls, suggesting that injecting M-CSF either prevented the synthesis of Aβ or favoured the immediate clearance of this short peptide. We also demonstrated that M-CSF highly diminished the monomeric form of Aβ40 and Aβ42 peptide in the extracellular fraction. These data together with the presence of Aβ in the late endosomes and lysosomes of microglia provide strong evidence that chronic M-CSF treatment stimulated the clearance of the toxic amyloid peptides from the extracellular milieu. More microglia were associated with the senile plaques and the amount of Aβ-containing microglia greatly increased in response to the haematopoietic cytokine. It has previously been showed that microglia are unable to eliminate internalized Aβ, and might actually release it and participate in plaque formation. However, Majumdar et al. (2007) recently reported that M-CSF acidifies microglial lysosomes and stimulates Aβ degradation. This could explain why less Aβ is present in the brains of our treated mice. Treatment of our animals with this cytokine could promote the differentiation, proliferation and activation of a population of microglia able to degrade Aβ and thus, resulting in a decrease in the Aβ pathology and behavioural impairment.

It is quite interesting to note the powerful effects of M-CSF to prevent and improve cognitive impairment in APPSwe/PS1 mice without noticeable side effects. M-CSF-injected mice exhibited a healthy state beside a slight increase in body weight, which could simply be attributed to the fact that their average body weights were higher than saline controls at the beginning of the experiment. Moreover, treated animals presented no sign of encephalitis or vasculitis, ruling out detrimental inflammatory effects of M-CSF injection at the concentration used in the present study. Previous studies have observed detrimental effects of microglia stimulated with M-CSF. These include higher secreted levels of IL-1 and IL-6 by M-CSF-stimulated BV-2 microglial immortalized cell line when exposed to Aβ (Murphy et al., 1998). Moreover, microglia near Aβ deposits exhibit higher expression of the receptor for M-CSF (M-CSFR) in the AβPPV717F mouse model of AD (Murphy et al., 2000). The increased expression of this receptor in the BV-2 cell line may induce inflammatory responses, which can be detrimental for neuronal elements (Mitrasinovic et al., 2001). However these data were generated in immortalized cell lines and organotypic co-cultures, which differ quite substantially from in vivo models.

Although Alzheimer's disease patients may respond differently to chronic M-CSF treatment, we did not observe any visible detrimental effects in APPSwe/PS1 mice. This was also the case in mice treated from age 6–9 months. However, injection of haematopoietic cytokines raises the question of possible long-term side effects. All three CSFs are implicated in the development and progression of some inflammatory and autoimmune disorders, such as arthritis and cancers (Hamilton, 2008). Injecting M-CSF on a regular basis could result in the appearance or accentuation of some inflammatory disorders, but our data in APPSwe/PS1 mice failed to show any inflammatory responses to the chronic treatment as determined by a careful visual evaluation of organ morphology during necropsy. However, we did not perform a detailed histological analysis of the organs, and evidence of inflammation and/or systemic immunological dysfunction would need to be looked for in future studies in mice and in any human study. In fact, Alzheimer's disease patients may actually have a defect in their innate immune abilities to clear Aβ and prevent cognitive decline associated with accumulation of the toxic peptide in the extracellular milieu. Recent data have shown that anti-inflammatory drugs not only failed to improve cognitive functions, but may actually be detrimental in some patients (see the news and views in Nat Med 2008; 14: 916). Strategies aiming to inhibit the innate capacity of microglia to clear Aβ are clearly not the direction to go for treating this disease.

These data together may explain the ability of M-CSF to stabilize and improve the cognitive and behavioural functions of mice that received the injections from 6 months of age. These transgenic animals exhibited significant signs of cognitive decline and Aβ load in the CNS before receiving weekly injections of the cytokine or its control solution. Therefore, systemic M-CSF treatment is not only effective at the level of the prevention, but also at the level of stabilization and improvement once the disease is well installed. These results are very promising for the development of a new therapeutic strategy for patients exhibiting mild cognitive impairment.

In conclusion, our results strongly suggest that injecting M-CSF can reduce amyloid load and prevent behavioural impairment and memory losses associated with the Alzheimer's disease-like pathology in APPSwe/PS1 mice. Moreover, such a treatment largely prevented Aβ burden and consequently the incidence of the pathologies. Since this cytokine has been tested in humans to stimulate the hematopoietic system, it is conceivable to propose its use as a new treatment for Alzheimer's disease.


The Canadian Institutes in Health Research; Neuroscience Canada (Brain repair program); Alzheimer Society of Canada and Fonds de la recherche en santé du Québec (PhD fellowship to V.B.). S.R. holds a Canadian Research Chair in Neuroimmunology.


  • Abbreviations:
    β-amyloid precursor protein
    Swedish β-amyloid precursor protein
    bone marrow derived microglia
    bovine serum albumin
    colony-stimulating factors
    Dulbecco's phosphate buffered saline
    foetal bovine serum
    green fluorescent protein
    lysosomal-associated membrane protein-2
    mild cognitive impairment
    macrophage colony-stimulating factor
    macrophage colony-stimulating factor receptor
    phosphate buffered saline


View Abstract