Combined Effects of Hematopoietic Progenitor Cell Mobilization from Bone Marrow by Granulocyte Colony Stimulating Factor and AMD3100 and Chemotaxis into the Brain Using Stromal Cell-Derived Factor-1α in an Alzheimer's Disease Mouse Model†‡§
Stem Cell Neuroplasticity Research Group, Kyungpook National University, Daegu, Korea
Department of Physiology, Cell and Matrix Research Institute, BSEI, World Class University Program, School of Medicine, Kyungpook National University, Daegu, Korea
Author contributions: J.W.S and J.K.L.: performed the experiments, wrote the manuscript; J.E.L.: performed the irradiation; W.K.M.: reviewed the manuscript; E.H.S: edited and reviewed the manuscript; H.K.J.: designed all experiments, supervised the project; J.B.; designed all experiments, supervised the project, edited and reviewed the manuscript. Ji-Woong Shin and Jong Kil Lee contributed equally to this article.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLSEXPRESS May 23, 2011.
Transplantation of bone marrow-derived stem cells (BMSCs) has been suggested as a potential therapeutic approach to prevent neurodegenerative diseases, but it remains problematic due to issues of engraftment, potential toxicities, and other factors. An alternative strategy is pharmacological-induced recruitment of endogenous BMSCs into an injured site by systemic administration of growth factors or chemokines. Therefore, the aim of this study was to examine the effects of therapy involving granulocyte colony stimulating factor (G-CSF)/AMD3100 (CXCR4 antagonist) and stromal cell-derived factor-1α (SDF-1α) on endogenous BM-derived hematopoietic progenitor cell (BM-HPC) recruitment into the brain of an Alzheimer's disease (AD) mouse model. To mobilize BM-HPCs, G-CSF was injected intraperitoneally and boosted by AMD3100. Simultaneously, these mice received an intracerebral injection with SDF-1α to induce migration of mobilized BM-HPCs into brain. We found that the memory deficit in the AD mice was significantly improved by these treatments, but amyloid β deposition was unchanged. Interestingly, microglial activation was increased with alternative activation of microglia to a neuroprotective phenotype. Furthermore, by generating an amyloid precursor protein/presenilin 1-green fluorescent protein (GFP) chimeric mouse, we ascertained that the GFP positive microglia identified in the brain were BM-derived. Additionally, increased hippocampal neurogenesis and improved memory was observed in mice receiving combined G-CSF/AMD3100 and SDF-1α, but not in controls or animals receiving each treatment alone. These results suggest that SDF-1α is an effective adjuvant in inducing migration into brain of the endogenous BM-HPCs, mobilized by G-CSF/AMD3100, and that the two can act synergistically to produce a therapeutic effect. This approach warrants further investigation as a potential therapeutic option for the treatment of AD patients in the future. STEM CELLS 2011;29:1075–1089
The general concept of stem cell therapy is based upon transplantation of exogenous stem cells, such as autologous bone marrow (BM) cells. It has been suggested that such a potential therapeutic approach could impact various neurological disorders, including Alzheimer's disease (AD) [1–4]. However, the safety of exogenous transplantation in human patients remains controversial because the possibility of unwanted proliferation or differentiation of the transplanted stem cells cannot be excluded, and to achieve sufficient engraftment, irradiation or other potentially toxic preconditioning must be used. An alternative approach that may prove safer, more convenient, and economical involves the stimulation of endogenous BM stem cell recruitment into damaged brain regions by systemic administration of growth factors or chemokines.
Previous studies support the possibility that activation of endogenous BM cells could be a viable therapeutic strategy for AD. For example, it has been demonstrated that differentiation of endogenous BM-derived hematopoietic progenitor cells (BM-HPCs) into microglia and their infiltration into brain can be mediated by amyloid β (Aβ), and that these recruited microglia were more effective at Aβ phagocytosis than the resident microglia in an AD transgenic mouse model [5–8]. Furthermore, peripheral administration of hematopoietic growth factors, such as granulocyte colony stimulating factor (G-CSF) or macrophage CSF (M-CSF), increased the number of BM-HPCs or microglia in the brain, and rescued the memory impairment in AD mouse models [9–11]. These studies led us to speculate that a more effective approach to recruiting endogenous BM-HPCs into brain in AD mice could be via a combined therapy involving mobilization of BM-HPCs in the periphery, together with enhanced homing of these BM-HPCs to the brain by intracerebral chemokine injection.
To recruit the endogenous stem cells from BM to an injured site, two main principles should be considered. One is mobilization of stem cells from BM to the blood. G-CSF is used extensively in clinical practice for this purpose , and it has been reported that G-CSF increases BM-derived cells in the damaged central nervous system (CNS), and promotes functional recovery [13, 14]. Moreover, the BM-HPC mobilization effect of G-CSF is enhanced synergistically by AMD3100, which is a CXCR4 antagonist [15–17]. The other crucial principle is migration of the mobilized BM-HPCs from the circulation into the injured site. Various chemokines are important for this, and among them stromal cell-derived factor-1α (SDF-1α or CXCL12) is a strong chemoattractive factor for HPCs [18, 19]. Intracerebral administration of this chemokine results in enhanced homing of BM-derived cells to the injured brain in stroke animal models .
Based on these concepts and findings, we speculated that migration of BM-HPCs, mobilized by G-CSF/AMD3100, into brain could be promoted by intracerebral injection of SDF-1α, i.e., the combination therapy might be more effective in the recruitment of BM-HPCs into brain than either therapy alone. Therefore, we treated amyloid precursor protein (APP) and presenilin 1 (PS1) double-transgenic AD mice with G-CSF/AMD3100 and SDF-1α to enhance the migration of BM-HPCs into brain. Interestingly, this treatment induced recruitment of BM-derived microglia into brain, modulated inflammatory responses, promoted hippocampal neurogenesis, and improved the spatial memory dysfunction in APP/PS1 mice, despite having no effect on Aβ deposition.
MATERIALS AND METHODS
APP/PS1 double transgenic mice and nontransgenic (NT) control littermates were generated by mating single transgenic mice expressing human mutant APP  and mutant PS1 . The APP and PS1 transgenic mice were originally obtained from Taconic (Petersburgh, NY, www.taconic.com) and Jackson Laboratory (Bar Harbor, ME, http://www.jax.org), respectively. All procedures were performed in accordance with an animal protocol approved by the Kyungpook National University Institutional Animal Care and Use Committee (IACUC). Animals were housed in a room which is maintained under controlled temperature and humidity.
APP/PS1-GFP Chimeric Mice
Six-month-old APP/PS1 mice were used as recipients to create APP/PS1-green fluorescent protein (GFP) chimeric mice. The BM donor animals were 6-week-old homozygous C57BL/6-Tg (CAG-EGFP)1osb/J mice on the C57BL/6 background  (The Jackson Laboratory, Bar Harbor, Maine). Tibias and femurs were dissected from the donor mice, BM was harvested, and single-cell suspensions were obtained using a 40-μm cell strainer (BD Biosciences, Becton-Dickinson, Franklin Lakes, NJ, www.BD.com). Recipient mice received whole body irradiation (5 Gy at dose rate of 1 Gy/minute), except to their heads, which were shielded. Following irradiation, GFP-BM (approximately 7 × 106 cells) were injected into the tail vein. From 1 week before the irradiation to until 2 weeks after the transplantation, mice were given drinking water complemented with 0.2 mg/ml trimethoprim and 1 mg/ml sulfamethoxazole. At 5 weeks after the BM transplantation, chimeric mice were confirmed by blood smears from tail clippings for the presence of GFP. Also, BM cells and peripheral blood were harvested and red blood cells and debris were examined after density centrifugation using Ficoll (Stem Cell Technologies, Vancouver, BC, Canada, www.stemcell.com). The engraftment of donor-derived BM cells and peripheral blood mononuclear cells was evaluated by flow cytometric analysis (Supporting Information Fig. S1).
Surgery for Implantation of Guide Cannula
All mice used in this study had a guide cannula implanted into the hippocampus 1 week before the first injection with SDF-1α. Briefly, after anesthesia with a combination of 100 mg/kg ketamine and 10 mg/kg xylazine, a stainless steel cannula was implanted in the hippocampus using a stereotaxic frame (David Kopf Instrument, Tujunga, CA, www.kopfinstruments.com). The guide cannula was fixed in the hippocampal brain region according to the following coordinates: 1.6 mm posterior to the bregma, 1.7 mm bilateral to the midline, and 1.2 mm ventral to the skull surface . The guide cannula allows the formation of a stable aperture in the brain through which treatments can be administrated. In the absence of treatment, the guide cannula contains an obturator that prevents the aperture from being filled in by surrounding tissue.
Eight-month-old APP/PS1 mice (n = 32) were administered human recombinant G-CSF (Peprotech, Rocky Hill, NJ, www.peprotech.com) (100 μg/kg/day; n = 16) or phosphate-buffered saline (PBS) (n = 16) intraperitoneally on 4 consecutive days. After 24 hours of the last injection, mice received an i.p. injection with AMD3100 (Sigma-Aldrich, St. Louis, MO, sigmaaldrich.com) (5 mg/kg) or PBS. Simultaneously, half of the G-CSF/AMD3100 and PBS/PBS-treated mice received stereotaxic injections of human recombinant SDF-1α (Peprotech) into the hippocampus through the implanted guide cannula on 5 consecutive days. Age-matched NT mice (n = 8) were treated with PBS in the same way. The tip of the injection cannula projected beyond the guide cannula by 1 mm. It was connected by flexible polyethylene tubing to the microinjection system, which housed a 25-μl Hamilton syringe. A total of 1 μg of SDF-1α in 3 μl of PBS or 3 μl of PBS alone was injected into the each hippocampus bilaterally, delivered at a rate of 0.8 μl/minute. Eight-month-old APP/PS1-GFP chimeric mice (n = 25) were treated by the same protocol.
Mice received i.p. injections with 5-bromo-2-deoxyuridine (BrdU; Sigma-Aldrich) dissolved in 0.9% NaCl/0.007 M NaOH solution for 13 consecutive days from the first treatment to before the behavioral testing (50 mg/kg/day).
We performed the Morris water maze (MWM) task to assess spatial memory performance as we described previously . The water maze was a white tank (1.0 m diameter, 30 cm height) located on a white curtain surrounding the pool with variety of visual extra maze cues and was filled to a depth of 20 cm with water containing white opaque nontoxic titanium dioxide to hide the target platform and maintained at 22–24°C. A submerged platform (10 cm diameter; 6–8 mm below the surface of the water) was located at a fixed position throughout the training session. The position of the platform was varied from mouse to mouse while being counterbalanced across experimental groups. The day before training began, a pretraining procedure was implemented to habituate mice to the water and to train them to escape from the water by climbing onto the platform. A training session consisted of a series of four trials per day for 10 consecutive days (total 40 trials). In each of the four trials, the animals were placed at different starting positions equally spaced around the perimeter of the pool in a random order. Each mouse was given 60 seconds to find the submerged platform. If the mouse did not mount the platform within 60 seconds, it was guided to the platform. The time to find the platform was recorded as the latency for each trial. Mice were allowed to remain on the platform for 10 seconds to recognize the target position before being returned to a home cage. On day 11, a single probe trial, in which the platform was removed, was performed after the hidden platform task had been completed, during which the proportion of time spent in the different quadrants and the number of target crossings were recorded. All trials were recorded using a charge-coupled device camera connected to a video monitor and a computer. The test was run using Image J software. All apparatus used in this study was made by O'Hara and Company (Tokyo, Japan, www.ohara-time.co.jp).
After behavioral testing, mice were sacrificed and anesthetized with 2.5% avertin in PBS. Animals underwent immediately cardiac perfusion with 4% paraformaldehyde in PBS. After the perfusion, brains were removed, postfixed overnight at 4°C, incubated in 30% sucrose at 4°C until equilibrated, and embedded in OCT compound for frozen section. Sequential 30 μm coronal sections were cut on a cryostat (CM3050S; Leica, Heerbrugg, Switzerland, www.leica.com) and stored at −20°C. APP/PS1-GFP chimeric mice were sacrificed 1 week after the first injection or at the point when the behavioral test were finished. The tissue processing of the chimeric mice was performed in the same way.
Thioflavin S Staining
Brain sections were incubated for 5 minutes at a concentration of 0.5% thioflavin S (Sigma-Aldrich) dissolved in 50% ethanol and then washed twice with 50% ethanol for 5 minutes each and then once with tap water for 5 minutes. They were then mounted with mounting medium.
Brain sections were treated with PBS containing 5% normal goat serum, 2% bovine serum albumin (BSA), and 0.4% Triton X-100 for 1 hour. In the same buffer solution, the sections were then incubated for 24 hours in primary antibodies at 4°C. The following antibodies were used: anti–ionized calcium binding adaptor molecule 1 (Iba-1) (rabbit, diluted 1:500, Wako Chemical, Osaka, Japan, www.wako-chem.co.jp), anti–acidic mammalian chitinase (anti-AMCase) (YM-1; goat, diluted 1:100, SantaCruz Biotechnology, Santa Cruz, CA, www.scbt.com), anti-interleukin-4 (IL-4) (goat, diluted 1:250, SantaCruz), anti-IL-1β (goat, diluted 1:10, R&D System, Minneapolis, MN, www.rndsystem.com), and anti–tumor necrosis factor-α (anti-TNF-α) (goat, diluted 1:20, R&D). The primary antibody was visualized by incubating with Alexa Fluor 488- or 568-conjugated secondary antibodies for 1 hour at room temperature. For double fluorescence labeling of microglia and immune-associated markers, tissue sections were stained overnight at 4°C with the specified primary antibodies diluted as described above: rabbit anti-Iba-1 (1:500) and goat anti-AMCase (1:100), goat anti-IL-4 (1:250), goat anti-IL-1β (1:10), or goat anti-TNF-α (1:20) followed by the corresponding Alexa anti-rabbit 488/anti-goat 568-conjugated secondary antibodies. For some experiments, tissue sections from APP/PS1-GFP chimeric mice were stained with combinations of the following primary antibodies: goat anti-IL-4 (1:250) and rabbit anti-Iba-1 (1:500) followed by the corresponding Cy-3 and Alexa 647-conjugated secondary antibodies. The sections were analyzed with a laser scanning confocal microscope equipped with Fluoview FSV1000 imaging software (Olympus FV1000, Tokyo, Japan, www.olympus-global.com) or with an Olympus BX51 stereology microscope.
Quantitative Real-Time PCR
RNA samples were extracted from whole brains of four individual animals per group using the RNeasy Lipid Tissue Mini kit (Qiagen, Hilden, Germany, www1.qiagen.com), and the concentration was determined using a Nanodrop ND-1000 spectrophotometer. A total of 5 μg of each RNA was converted to cDNA using the sprint RT complete-oligo(dT)18 (Clontech, Mountain View, CA, www.clontech.com) according to the manufacturer's guide. The cDNA was quantified using the QuantiTect SYBR Green PCR Kit (Qiagen). For each investigated transcript, a mixture of the following reaction components was prepared to the indicated end-concentration: forward primer (5 pM), reverse primer (5 pM), and QuantiTect SYBR Green PCR Master mix. The 10-μl master-mix was added to a 0.1-ml tube, and 5 μl volume, containing 100 ng reverse transcribed total RNA, was added as polymerase chain reaction (PCR) template. The tubes were closed, centrifuged, and placed into the Corbett research RG-6000 real-time PCR machine (Corbett Life Science, Sydney, Australia, www.corbettlifescience.com). The following primers were used: TNF-α (forward: 5′-GCTCCAGTGAATTCGGAAAG-3′, reverse: 5′-GATTATGGCT CAGGGTCCAA-3′), IL-1β (forward: 5′-CCCAAGCAATACC CAAAGAA-3′, reverse: 5′-GCTTGTGCTCTGCTTGTGAG-3′), IL-4 (forward: 5′-ATCCATTTGCATGATGCTCT-3′, reverse: 5′-GAGCTGCAGAGACTCTTTCG-3′), and YM-1 (forward: 5′-AGAGCAAGAAACAAGCATGG-3′, reverse: 5′-CTGTACCAG CTGGGAAGAAA-3′).
Methods for the isolation and purification of CD11b-positive microglia from cerebrum were previously described . Briefly, APP/PS1-GFP chimeric were sacrificed and perfused with 30 ml of PBS. Cerebrum were minced in PBS and dissociated from the tissue into single cells using RPMI 1640 (no phenol red) containing 2 mM L-glutamine, dispase, and collagenase type 3 (Sigma-Aldrich). The enzymes were inactivated by addition of 20 ml of Ca2+/Mg2+-free Hank's balance salt solution containing 2 mM EDTA and 2% fetal bovine serum, followed by trituration using pipettes of decreasing diameter. Cells were pelleted and resuspended in RPMI 1640/L-glutamine and mixed with physiologic Percoll (Sigma-Aldrich) and centrifuged at 850g for 45 minutes. The cells were then incubated with anti-mouse Cd11b-coated microbeads (Miltenyi Biotec, Auburn, CA, www.miltenyibiotec.com) for 20 minutes at 12°C. The cell-bead mix was then washed to remove unbound beads. The bead-cell pellet was resuspended in PBS/0.5% BSA/2 mM EDTA and passed over a magnetic MACS Cell Separation column (Miltenyi Biotec) following the manufacturer's instructions. CD11b-positive cells were eluted by removing the column from the magnetic holder and pushing PBS/BSA/EDTA through the column with a plunger.
Flow Cytometry Analysis (FACS)
APP/PS1-GFP chimeric mice were sacrificed at the end of MWM task and brain tissues proteolytically digested to produce a single-cell suspension. The brain suspension was incubated with various antibodies. Isotype-matched antibodies served as controls. The following antibodies were used: anti-mouse IL-4 (eBioscience, San Diego, CA, www.ebioscience.com) and anti-mouse TNF-α (eBioscience). Labeled cells were analyzed on a Becton Dickinson FACSCAria cell sorter using FACSDiva software (BD Biosciences).
Enzyme-Linked Immunosorbent Assay
APP/PS1-GFP chimeric mice were sacrificed at the end of MWM task for the quantification of IL-4 and TNF-α levels. Brain tissues from cortex and hippocampus were homogenized according to the manufacturer's recommendations. IL-4 and TNF-α protein levels were determined using mouse ELISA kits (IL-4 [Raybiotec, Norcross, GA, www.raybiotec.com] and TNF-α [USCN life, Wuhan, China, www.uscnk.com]). Standard curves were prepared using purified cytokine standards. Each experimental sample was run in duplicate and the results were averaged.
Before the BrdU histochemistry, sections were incubated in 2 N HCl for 1 hour at 37°C followed by 10-minute incubation in 0.1 M borate buffer. They were then incubated overnight at 4°C in a mixture of rat anti-BrdU (1:100; Abcam, Cambridge, MA, www.abcam.com) and mouse anti–neuronal nuclei (anti-NeuN) (1:200, Chemicon, Temecula, CA, www.millipore.com) antibodies and visualized using Alexa anti-rat 546 and Alexa anti-mouse 488 as secondary antibodies. Fluorescence was detected using a confocal microscope (Olympus FV1000, Japan).
Data are expressed as mean ± SEM unless otherwise noted. One-way analysis of variance (ANOVA) was used to evaluate mean differences among experimental groups. For the behavioral data, to compare the day-dependent decrease of the escape latencies among groups, we used two-way repeated ANOVA analysis within groups as the between-subject factor and days as the within-subject repeated measures factor. The Tukey's honestly significant difference (HSD) test was used for multiple-comparison; null hypotheses were rejected at the 0.05 level. All data were analyzed using the SAS statistical package 9.1 version.
G-CSF/AMD3100 and SDF-1α Improved Spatial Memory Impairment in APP/PS1 Mice
We examined the effects of G-CSF/AMD3100 and SDF-1α on spatial memory impairment in APP/PS1 double transgenic mice. The injection protocol is described in Figure 1A. At 2 weeks after the first injection, we performed MWM testing to evaluate improvement of spatial memory. In the MWM task, mice were trained to escape on to a submerged platform for a maximum of 60 seconds using a protocol of four trials per day for each mouse during a 10-consecutive-day period. As shown in Figure 1B, there were no significant differences in the escape latencies among the groups through day 3. However, at day 4, the escape latencies of the NT (NT control) and SDF-1α and G-CSF/AMD3100 (SG) groups were significantly reduced when compared with that of the sham-injected (PBS and PBS) groups. As expected, NT control mice improved their learning progressively by the end of MWM task; their average escape latency on day 10 was 18.5 seconds (p < .001). Notably, in the SG group, escape latency was significantly decreased when compared with the sham group except at day 7. The last day (day 10) escape latency average for this group was 31.3 seconds (p = .003). In other groups, PBS and G-CSF/AMD3100 (PG) as well as SDF-1α and PBS (SP), there were no significant differences in the escape latencies when compared with sham treated APP/PS1 controls over 10 days. Furthermore, to compare the reduction of escape latencies among the groups as each day passed, we performed two-way repeated measures using ANOVA, with group defined as the between-subject factor and days as the within-subject repeated measure factor. When compared with other groups, only the NT and SG groups showed a significant day-on-day dependent decrease in escape latency (NT: 39.6–18.5 = 53.3% reduction, SG: 41.3–31.3 = 24.2% reduction; p < .001). No significant differences in day-to-day dependent reductions occurred in the escape latencies among PG, SP, and sham groups (Fig. 1B). Figure 1C shows examples of the swimming traces in each mouse group analyzed by the MWM task on day 10. Probe analysis was performed at day 11. In the probe test, the quadrant occupancy (Fig. 1D) and the number of crossings over a platform position (Fig. 1E) were significantly decreased in the sham group when compared with that in the NT group (p < .05). However, they were recovered significantly only in the SG group (p < .05). These results show that G-CSF/AMD3100 and SDF-1α synergistically improved spatial memory dysfunction in APP/PS1 mice when compared with either compound alone.
G-CSF/AMD3100 and SDF-1α Did Not Reduce Aβ Burden in APP/PS1 Mice
To determine whether the spatial memory improvement, we observed following treatment, was associated with alterations in Aβ pathology, we performed Thioflavin S staining and examined the Aβ plaque burden in cortex and hippocampus sections from the APP/PS1 mice groups (Fig. 2A). The Aβ percentage reactive area was quantified (n = 4 per group). There was no significant difference in Aβ deposition in cortex and hippocampus among the different APP/PS1 mice groups (Fig. 2B, 2C). These results indicate that the improvement in spatial memory observed following G-CSF/AMD3100 and SDF-1α treatment of APP/PS1 mice was not related to changes in brain Aβ plaque burden.
Treatment of G-CSF/AMD3100 and SDF-1α Increased Microglia in APP/PS1 Mice
In previous studies, it has been reported that SDF-1α is a strong chemoattractive factor for several classes of HPC, including lymphocyte, myelocyte, and macrophage progenitor cells . Also, hematopoietic growth factors, such as G-CSF or M-CSF, can increase the number of brain microglia in AD mouse models [10, 11]. Therefore, we performed immunostaining with brain sections from the SG, SP, PG, and sham groups using Iba-1 antibody to examine microglial infiltration in their hippocampus and cortex (Fig. 3A). The microglial burden was quantified as percentage immune-reactive area (n = 4 per each group). The results confirm that microglia increased significantly in the SG and PG groups when compared with the sham group, both in the hippocampus and cortex (p < .01). The microglial infiltration in the SP group was also increased when compared with the sham group, but there was no significant difference. Similarly, slightly increased microglia was observed between SG and PG animals; however, there was no statistical significance between these two groups (Fig. 3B, 3C). These results suggest that the G-CSF/AMD3100 treatment is the key factor in increasing microglia in APP/PS1 mice brain, with SDF-1α producing only a minimal effect.
G-CSF/AMD3100 and SDF-1α Synergistically Enhanced Recruitment of BM-Derived Cells and Increased BM-Derived Microglia in Brain
To examine whether the increased microglia, we observed in the brain were derived from BM, we made GFP chimeric mice by irradiating 6-month-old APP/PS1 mice and intravenously injecting BM cells collected from GFP mice . The chimeric mice were then treated with G-CSF/AMD3100 and/or SDF-1α at 8-month of age (n = 3 per each group). At 1 week after the injection, brain sections were taken and the number of GFP positive cells in the hippocampal region was estimated using stereological analysis. In these mice, G-CSF/AMD3100 treatment (PG group) resulted in an increase in the number of GFP positive BM-derived cells when compared with the sham group (p < .05). Intrahippocampal administration of SDF-1α (SP group) resulted in a slight increase in GFP positive BM-derived cells, but there was no significant difference when compared with the sham group. However, interestingly, in mice that received both G-CSF/AMD3100 and SDF-1α treatment (SG group), the number of GFP positive BM-derived cells in the hippocampus and other brain regions was markedly increased (p < .01) (Fig. 4A, 4B). Next, double immunostaining was used to confirm that the positive cells were microglia derived from the BM. GFP/Iba-1-double-positive cells were counted in randomly selected confocal microscopic images from each group (Fig. 4A). Similar to the results above, local injection of SDF-1α (SP group) alone did not produce a significant change in the number of Iba-1/GFP positive cells in the hippocampus, whereas systemic G-CSF/AMD3100 treatment (PG group) did (p < .05). Notably, a highly significant increase of Iba-1/GFP positive cells was observed in the SG group. Moreover, the number of cells in the SG group was significantly increased over the PG group (Fig. 4C). These results indicate that intracerebral injection of SDF-1α is an effective synergistic strategy to induce migration of BM-derived microglia, mobilized by G-CSF/AMD3100 treatment, into the brain.
G-CSF/AMD3100 and SDF-1α Synergistically Modulated Immune Reactions in APP/PS1 Mice
Previous studies have suggested that pharmacological activation of endogenous stem cells have anti-inflammatory effects in the CNS, including in an AD mouse model [11, 26]. It has been demonstrated that Aβ induces proinflammatory cytokines, such as IL-1β and TNF-α, that can contribute to neurodegeneration in AD mouse models including the APP/PS1 mice [27, 28]. On the other hand, IL-4 downregulates Aβ-induced inflammation . In addition, alternatively activated microglia, which express YM-1 and IL-4, have a neuroprotective effect [30–32]. To evaluate the inflammatory profiles in the brains of each of our experimental groups, we examined the mRNA expression levels of these proinflammatory cytokines (IL-1β and TNF-α), and alternatively activated microglia associated markers (IL-4 and YM-1), by quantitative real-time PCR (n = 4 per each group). The SP group showed significant differences in IL-1β and IL-4 expression levels, whereas the PG group showed significance only in the YM-1 expression level when compared with the sham group (p < .05). However, only in the SG group, IL-1β and TNF-α proinflammatory cytokines levels were reduced, whereas neuroprotective IL-4 and YM-1 expression levels were significantly increased to levels equivalent to NT mice (p < .01 or p < .05 comparing SG and the sham control group) (Fig. 5A). Immunohistochemical staining confirmed the levels of these makers in the hippocampus of APP/PS1 mice (Fig. 5B–5E; examples of SG and sham groups). IL-1β and TNF-α expression was dramatically decreased in the SG group as compared to sham mice (Fig. 5B, 5C), whereas IL-4 and YM-1 were increased (Fig. 5D, 5E). Next, to determine whether there was a reciprocal relationship between activated microglia and expression levels of proinflammatory or anti-inflammatory cytokines in our treatment paradigms, we performed double immunostaining with antibodies against the proinflammatory cytokines or activated microglia markers and Iba-1 (n = 4 per each group). Interestingly, we found that the number of Iba-1 positive cells expressing IL-1β or TNF-α was increased to a greater extent in the brains of the sham group mice when compared with SG group mice (Fig. 5B, 5C). In contrast, IL-4 and YM-1 expression in Iba-1 positive cells was markedly increased in the hippocampus of the SG group (Fig. 5D, 5E). Quantitative image analysis indicated, similar to the results of real-time PCR, that only the SG group showed significant changes in the expression of proinflammatory cytokines (IL-1β and TNF-α) and alternatively activated microglia markers (IL-4 and YM-1) when compared with the sham group (p < .05) (Fig. 5F). In contrast, in other treated groups, there was no significant difference when compared with the sham group, except decreased IL-1β expression in the SP group. To confirm the immune modulation effect observed in SG group, we measured the IL-4 and TNF-α protein content in the hippocampus and cortex by enzyme-linked immunosorbent assay. As shown in the Supporting Information Figure S2, the IL-4 protein level was higher in the SG group mouse brain than in the sham group, while TNF-α was decreased. These results demonstrate that enhanced endogenous stem cell activation by G-CSF/AMD3100 and SDF-1α has a highly synergistic effect in modulating the immune reactions in the brains of APP/PS1 mice.
Furthermore, to determine whether the immune modulation effect was associated with the migration of BM-derived cells into the brain, we performed IL-4 immunostaining using brain sections from the APP/PS1-GFP chimeric mice groups (Fig. 6A). When compared with the sham group, the hippocampus of the SG group exhibited a highly significant increase (4.38-fold) in the number of IL-4/GFP positive cells (p < .01). Also, the SG group showed a significant elevation (2.06-fold) in the number of double-positive cells when compared with the PG group (p < .05; Fig. 6B). In the SP and PG groups, the number of IL-4/GFP positive cells was slightly increased as compare to the sham group, but there was no statistical significance (Fig. 6B). Interestingly, we also found that the percentage of GFP positive cells expressing IL-4 was approximately 60%, and there were no significant differences among the four groups (Fig. 6C; SG = 68.4%, SP = 56.5%, PG = 62.5%, and sham = 56.3%). Triple immunofluorescent images in APP/PS1-GFP chimeric mice using IL-4 and Iba-1 antibodies showed the BM-derived microglia expressed IL-4 (Fig. 6D).
Additionally, to examine whether these migrated, BM-derived microglia actually remained in the brain and expressed neuroprotective cytokines during the MWM task, we performed the MWM analysis using APP/PS1-GFP chimeric mice, and then carried out tissue sampling when the behavioral test was finished. In this experiment, APP/PS1-GFP chimeric mice were divided into two groups (SG and Sham group). As shown in the Supporting Information Figure S3, improved spatial memory was observed in SG treated APP/PS1-GFP chimeric mice, as was observed previously in the nonchimeric mice. Also, FACS analysis showed that IL-4 expressing and BM-derived microglia increased in SG APP/PS1-GFP chimeric mice when compared with the sham group (Supporting Information Fig. S4), and that compared with the sham group, a decreased number of TNF-α expressing microglia were observed in the SG group. These data indicate that BM-derived microglia are expressing the neuroprotective cytokine, IL-4, and that thus the recruitment of these cells into the brain by G-CSF/AMD3100 and SDF-1α contribute to the modulation of immune reactions in APP/PS1 mice.
G-CSF/AMD3100 and SDF-1α Increased Hippocampal Neurogenesis
It is well-established that both SDF-1α and G-CSF, which are neuronal ligands, induce neuronal cell proliferation and are related to neurogenesis [33–35]. Moreover, it has been reported that G-CSF treatment promotes hippocampal neurogenesis in AD mouse models [9, 11]. To examine whether the G-CSF/AMD3100 and SDF-1α treatment promoted hippocampal neurogenesis, we injected BrdU for 2 weeks before the MWM task analysis, and then performed double-immunostaining using BrdU and NeuN antibodies to estimate the number of proliferating cells in the dentate gyrus of the hippocampus using stereological methods (n = 4 per each group). The total number of BrdU-positive cells was significantly increased in the SG group (to a level equivalent to that of NT controls), when compared with the sham group (p < .05) (Fig. 7A). Also, in the NT and SG groups, BrdU/NeuN-double-positive cells showed a significant elevation when compared with the sham group (ANOVA, Tukey's HSD test, p < .05) (Fig. 7B, 7C). Next, to examine possible neuron-specific effects of the treatment, we analyzed the percentage of NeuN-positive/BrdU-positive cells. Almost all BrdU-positive cells were neuron-specific in all groups, including SG (Fig. 7D). These results demonstrate that combined G-CSF/AMD3100 and SDF-1α treatment has a highly synergistic effect in increasing hippocampal neurogenesis in APP/PS1 mice.
We report here the synergistic effects of G-CSF/AMD3100 and SDF-1α on endogenous BM-HPC recruitment and migration in an APP/PS1 model of AD, resulting in the rescue of cognitive decline and modulation of the immune response. Previous studies have observed that plasma G-CSF and SDF-1α levels [36, 37] and circulating HPCs in peripheral blood are decreased in AD patients . Moreover, it has been reported that BM-derived microglia, which differentiate from HPCs, have neuroprotective effects and play a critical role in Aβ plaque clearance [5–7]. These reports led us to speculate that manipulating levels of G-CSF and SDF-1α may enhance mobilization of blood HPCs and recruitment of BM-derived microglia into the brain, and thereby ameliorate amyloid pathology. To test this hypothesis, we induced mobilization and migration of BM-HPCs by using G-CSF/AMD3100 and SDF-1α, respectively, and examined the effects of increased HPC recruitment into brain on cognitive decline, Aβ plaque deposition, infiltration of BM-derived microglia, immunomodulation in the brain, and hippocampal neurogenesis in APP/PS1 double transgenic mice. We found that, although this combination treatment did not reduce Aβ plaque deposition, it nevertheless improved cognitive function, increased BM-derived microglia recruitment, modulated the inflammatory response, and promoted hippocampal neurogenesis.
Recent studies have demonstrated that administration of hematopoietic growth factors, such as G-CSF or M-CSF, leads to microgliosis in the AD mouse brain and these microglia are derived from the BM [10, 11]. Furthermore, intracerebral administration of SDF-1α results in enhanced homing of BM-derived microglia to the injured brain in stroke animal models . In this study, as in previous studies, we observed increased microglia in the brains of mice receiving G-CSF/AMD3100 (i.e., the SG and PG groups), but there was no significant difference between the two groups receiving combined (with SDF-1α, SG) or individual treatment (Fig. 3). However, the chimeric mice study further demonstrated that the microglia derived from BM were in fact dramatically increased in SG when compared with PG group (Fig. 4). It has been reported that G-CSF can not only recruit BM-derived microglia  but also increase resident microglia by stimulating the G-CSF receptor directly . Thus, it seems that G-CSF/AMD3100 treatment increased microglia by these two mechanisms, while SDF-1α acted mainly to promote migration of the mobilized HPCs into the brain. Therefore, the proportion of microglia derived from BM is increased in the SG group when compared with PG, although the overall microglial burden remains the same. Taken together, G-CSF/AMD3100 and SDF-1α act synergistically to specifically recruit microglia derived from BM into brain.
Radiation BM-chimeric mice have been used extensively for studies of microglial turnover and replacement [7, 41] and for evaluation of cellular gene therapy . A clear disadvantage of this model is that the recipient is usually subjected to whole-body irradiation before transplantation, which might induce microglial activation or cell death [41, 43], and could lead to aberrant recruitment and microglial transformation of BM-derived cells. In our experiments, we used head shielding to minimize microglia activation and blood brain barrier destruction by irradiation , although we cannot completely rule out these irradiation-induced effects. Although a BM-chimeric mouse per definition is not a “normal” mouse, we found that the microglial response to AD in BM-chimeric and nonchimeric mice was comparable (data not shown). These data are consistent with findings by others that irradiation does not compromise or exacerbate the innate immune response in BM-chimeric mice .
Previously, it has been suggested that BM-derived microglia are more effective in Aβ phagocytosis than the resident microglia in AD transgenic mouse model [5–8]. However, according to our results, G-CSF/AMD3100 and SDF-1α treatment did not reduce Aβ plaque burden despite the presence of increased BM-derived microglia in the brain. It is unknown whether BM-derived microglia can have positive effects on AD pathology, other than on Aβ phagocytosis , and our results indicate that this may not be the case. It has also been speculated that resident microglia may degenerate during aging or AD progression, with BM-derived microglia acting to compensate for the defective functions of senescent resident microglia [46, 47].
In this study, the improvement of spatial memory was only observed in the G-CSF/AMD3100 and SDF-1α dual treatment (SG) group (Fig. 1), but there was no significant difference in Aβ plaque deposition (Fig. 2). This result suggests that the improvement in memory dysfunction most likely involves Aβ-independent mechanisms. Amyloid imaging studies support a model in which amyloid deposition is an early event on the path to dementia, beginning insidiously in cognitively normal individuals, and accompanied by subtle cognitive decline and functional and structural brain changes suggestive of AD . End-stage symptoms of the disease are most likely caused by catastrophic synaptic damage and neuronal loss and appear uncoupled from the process of brain amyloidosis . Some researchers have suggested that there is a poor correlation between Aβ plaque load and the cognitive decline and dementia in AD [50, 51] and that end-stage symptoms are uncoupled from amyloidosis . Recently, in a phase I trial of Aβ immunization in AD patients, although Aβ plaques were reduced, this clearance did not prevent progressive neurodegeneration . Meanwhile, a number of previous studies have suggested that Aβ induces proinflammatory cytokines, such as IL-1β and TNF-α, and that these contribute to neurodegeneration in AD mouse models [27, 28, 53]. In addition, these cytokines are closely associated with the cognitive impairment of AD mouse models and conversely the downregulation of these cytokines leads to improved cognitive function [28, 54, 55]. In this study, expression levels of the proinflammatory cytokines, IL-1β and TNF-α, were downregulated by G-CSF/AMD3100 and SDF-1α treatment, while the expression level of IL-4 was elevated (Fig. 5). Furthermore, double-stained images using antibodies of Iba-1 and these cytokines (IL-1β, TNF-α, and IL-4) showed that the microglia were expressing these markers (Fig. 5B–5E). Previous reports have indicated that IL-4 attenuates the inflammation induced by Aβ and improves the IL-1β-induced impairment of long-term potentiation in rat hippocampus [29, 56]. Notably, IL-4 immunostaing using APP/PS1-GFP chimeric mice showed that 60% of the BM-derived cells were expressing IL-4 and that these BM-derived cells were elevated in the SG group when compared with the other groups (Fig. 6). Furthermore, triple labeling experiments in the APP/PS1-GFP chimeric mice demonstrated that the BM-derived microglia were expressing IL-4. These results indicate that BM-derived microglia express this neuroprotective cytokine, and thus the recruitment of these cells into the brain contributes to the modulation of immune reactions in APP/PS1 mice. It has been suggested that aging, or a chronic neuroinflammatory environment, as in AD, may alter the activation state of microglia , to either a neuroprotective or neurotoxic phenotype, identified by their gene expression profiles . The “alternative activated” microglia, which express YM-1 and IL-4, have a neuroprotective effect [30–32]. In APP/PS1 mice, it has been demonstrated that there is an age-dependent switch in microglial phenotype from alternative to classic . In our APP/PS1 mice, the IL-4 and YM-1 expression levels were decreased when compared with NT mice, consistent with prior work. However, the expression levels of these markers were restored by G-CSF/AMD3100 and SDF-1α treatment, likely because these treatments stimulated recruitment of BM-derived microglia expressing IL-4 into the brain. Therefore, in agreement with other reports, our results demonstrate that HPC recruitment by G-CSF/AMD3100 and SDF-1α treatment acts to improve cognitive function in the APP/PS1 mouse by modulating the inflammatory responses and activating a neuroprotective microglial phenotype, and that this is not dependent on Aβ.
Additionally, in AD mouse models, it has been reported that impaired capacity for hippocampal neuron replacement may contribute to the cognitive decline [59, 60]. In our 8-month-old AD mouse model, the number of newborn neurons detected in dentate gyrus was lower than in NT mice, but this was restored by cotreatment with G-CSF/AMD3100 and SDF-1α (Fig. 7). Two possible mechanisms could explain this finding. First, it has been well-established that G-CSF and SDF-1α are both neuronal ligands closely related to neuronal cell proliferation and migration [33–35, 61]. Therefore, the effect on hippocampal neurogenesis in our experiments may have resulted from direct action by these ligands on endogenous neural progenitors in the dentate gyrus. Alternatively, as inflammatory factors have varying effects on adult neurogenesis , with proinflammatory cytokines, such as IL-1β and TNF-α, produced by microglia suppressing neurogenesis  and activated microglia expressing IL-4 inducing neurogenesis , the effect may have been achieved via the immune modulatory effects of G-CSF/AMD3100 and SDF-1α treatment. Moreover, the effects of G-CSF and SDF-1α on neurogenesis have been demonstrated in other animal models of neurological disorders, including stroke [20, 34] and AD [9, 11].
Contradictory data exist regarding the role of G-CSF in rescuing memory impairment in AD mice models. Previously, it has been demonstrated that G-CSF (50 μg/kg/day for 5 days) rescues the memory impairment of 13-month-old Tg2576 mice using the MWM test, without reduction of the Aβ plaque burden . In another recent study, G-CSF (250 μg/kg/day × 6 for more than 3 weeks) treatment reversed cognitive decline and decreased Aβ plaque burden by stimulating migration of BM-derived microglia, modulating the systemic immune response, and increasing hippocampal neurogenesis in APP/PS1 double mice . However, in this study, we did not observe any improvement in memory function in the G-CSF/AMD3100 (PG group) treated group (100 μg/kg/day for 4 days) (Fig. 1). To determine whether these different results were caused by AMD3100, we repeated the MWM task in G-CSF (100 μg/kg/day for 4 days) treated APP/PS1 mice and excluded AMD3100 (n = 3). No improvement in spatial memory impairment was achieved (data not shown). One explanation for the differing results may be the use of different mice strains.
In this study, despite using only 30% of the total G-CSF dose used by Sanchos-Ramos , and a shorter injection period, with the exception of the Aβ plaque reduction, we achieved similar results by simultaneous intrahippocampal administration of SDF-1α. Over the past few decades, a considerable number of studies have suggested G-CSF-based endogenous stem cell activation as a potential therapeutic strategy for various diseases, including CNS disorders. From the viewpoint of endogenous stem cell activation, the results from our work demonstrate that local SDF-1α injection provides an effective adjuvant to improve migration of the HPCs mobilized by systemic administration of G-CSF/AMD3100. Furthermore, this enhanced endogenous cell recruitment into the brain acts to promote neuroprotective immnomodulatory mechanisms, thereby improving cognitive decline in the AD environment.
We have previously reported that bone marrow–derived mesenchymal stem cells (MSCs) can modulate immune/inflammatory responses in AD mice, activate microglia, and improve the cognitive decline. In that study, we could not observe MSCs differentiating into neurons, glial cells, or other cells, and the transplanted MSCs disappeared after 2 weeks . Therefore, we suggest that factors secreted from transplanted MSCs activated microglia to an alternative form. There is ongoing research investigating the paracrine factors secreted from MSCs that are associated with microglia activation and recruitment. With this viewpoint, we thought that activation of endogenous stem cells (e.g., hematopoietic or neural stem cells) induced by drugs might be a potential therapeutic strategy for AD. Also, this approach could be used as an alternative way to improve exogenous stem cell therapy. For example, exogenous MSC transplantation has advantages that large numbers of cells can be applied to a specific site, and the paracrine effect is prolonged after the cells are no longer present. However, a potential deleterious effect is that these transplanted cells might undergo unwanted differentiation or proliferation in the brain parenchyma. Thus, as mentioned above, the endogenous approach using drug therapy could be an alternative way for treating AD and other diseases. To effectively activate specific endogenous stem cells, various strategies should be studied in the future.
Our results suggest that endogenous stem cell activation using drugs reduces AD-like pathology and prevents cognitive decline and memory impairment. When taken together, our results provide the basis for a novel therapeutic strategy of AD through endogenous stem cell activation.
This work was supported by the grant for the Future-based Technology Development Program funded by the National Research Foundation of Korea of the Ministry of Education, Science and Technology, Republic of Korea (grant number: 2010-0020232; to J.S.B.) and a grant from Stem Cell Research Center of the 21st Century Frontier Research Program funded by the Ministry of Science and Technology, Republic of Korea (grant number: SC4170; to H.K.J.).
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.