Granulocyte-colony stimulating factor is neuroprotective in a model of Parkinson's disease


Address correspondence and reprint requests to Dr Armin Schneider, Department Molecular Neurology, Axaron Bioscience AG, Im Neuenheimer Feld 515, 69120 Heidelberg, Germany.
E-mail:; or Dr Jochen H. Weishaupt Department of Neurology, University Hospital Göttingen, Robert-Koch-Str. 40, 37075 Göttingen, Germany.


We have recently shown that the hematopoietic Granulocyte-Colony Stimulating Factor (G-CSF) is neuroprotective in rodent stroke models, and that this action appears to be mediated via a neuronal G-CSF receptor. Here, we report that the G-CSF receptor is expressed in rodent dopaminergic substantia nigra neurons, suggesting that G-CSF might be neuroprotective for dopaminergic neurons and a candidate molecule for the treatment of Parkinson's disease. Thus, we investigated protective effects of G-CSF in 1-methyl-4-phenylpyridinium (MPP+)-challenged PC12 cells and primary neuronal midbrain cultures, as well as in the mouse 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) model of Parkinson's disease. Substantial protection was found against MPP+-induced dopaminergic cell death in vitro. Moreover, subcutaneous application of G-CSF at a dose of 40 μg/Kg body weight daily over 13 days rescued dopaminergic substantia nigra neurons from MPTP-induced death in aged mice, as shown by quantification of tyrosine hydroxylase-positive substantia nigra cells. Using HPLC, a corresponding reduction in striatal dopamine depletion after MPTP application was observed in G-CSF-treated mice. Thus our data suggest that G-CSF is a novel therapeutic opportunity for the treatment of Parkinson's disease, because it is well-tolerated and already approved for the treatment of neutropenic conditions in humans.

Abbreviations used

body weight


3, 4-dihydroxyphenylacetic acid


enhanced green fluorescent protein


granulocyte-colony stimulating factor


homovanillic acid


3,4, dihydroxy-l-phenylalanine






Parkinson's disease


tyrosine hydroxylase


5-hydroxyindole acetic acid

The symptoms of Parkinson's disease (PD), including resting tremor, rigidity and bradykinesia (de Lau et al. 2004), are caused by a slowly progressive degeneration of dopaminergic neurons in the substantia nigra, with an ensuing dopaminergic deficit in the striatum due to loss of dopaminergic nigrostriatal projections. The main therapeutic strategies for treating PD are centered around the replacement of dopamine in the brain, either by L-DOPA, or dopamine agonists. However, none of these drugs halts or delays the disease process (Oertel and Quinn 1997; Silver and Ruggieri 1998), and not all patients show satisfying response to symptomatic treatment.

In contrast to dopaminergic treatment, neuroprotective therapeutic approaches are aimed at preventing dopaminergic neuronal cell loss. In this context, neurotrophic growth factors have received major attention during the last years. The most well-studied is the glial cell line derived neurotrophic factor (GDNF) (Tomac et al. 1995; Gash et al. 1996; Kordower et al. 2000). Although experimentally successful, transfer into the clinic by intracerebroventricular delivery of GDNF demonstrated a range of adverse reactions (nausea, anorexia, weight loss, paresthesias, and hyponatremia; Nutt et al. 2003) and failed to show efficacy (Nutt et al. 2003), likely due to insufficient penetration to the target zones. Small trials using direct intraparenchymatic delivery presented encouraging results (Gill et al. 2003; Love et al. 2005; Patel et al. 2005; Slevin et al. 2005), however, a large randomized phase II trial was terminally halted by the sponsor due to inefficacy and safety concerns (e.g. discussed in Peck 2005). Although numerous internet sources are available, including press releases of the sponsoring company (AMGEN; homepage:, the study has not formally been published in a scientific journal. Thus the severity of the safety issues (4 patients developed antibodies against GDNF; cerebellar lesions were observed in a concomitant monkey study) are difficult to judge so far (summarized in Peck 2005).

Cell replacement approaches are experimentally feasible in PD models (e.g. Kim et al. 2002), but human trials are burdened by a number of complicating problems (Olanow et al. 2003), and the current neurosurgical approach is also questionable in view of the high prevalence of PD.

Thus, there is a need for neuroprotectants that slow disease progression by preventing or slowing down dopaminergic cell death (Schapira and Olanow 2004) and which ideally have been already introduced into the clinic for other applications, thereby allowing more rapid transfer to phase II clinical trials by potentially omitting extensive phase I trials (Reichmann 2002; LeWitt 2004). In addition, failure of drugs due to safety issues is less likely.

The cytokine G-CSF drives generation of neutrophilic granulocytes in the hematopoietic system, based on a reduction of apoptosis and promotion of differentiation of granulocyte precursors (Begley et al. 1986). As such, G-CSF has been in common clinical use for many years for the treatment of neutropenia, and meanwhile has been given to over 3 million patients with few adverse events (Frampton et al. 1994; Crawford 2003). We and others have recently discovered that G-CSF has neuroprotective properties in stroke models, and appears favorable in experimental allergic encephalomyelitis in vivo (Zavala et al. 2002; Schäbitz et al. 2003; Six et al. 2003; Shyu et al. 2004; Gibson et al. 2005; Komine-Kobayashi et al. 2005; Schneider et al. 2005). The protective effect appears as a direct one, mediated by the G-CSF receptor present on neurons (Schneider et al. 2005). Moreover, apart from its anti-apoptotic activity, G-CSF also dampens systemic inflammatory reactions (Hartung 1998). This property may be of additional benefit in neurodegenerative conditions.

We therefore asked whether G-CSF, that indeed fulfils several criteria for a neuroprotectant mentioned above, would be protective in a model of PD, a chronic neurodegenerative disorder which pathophysiologically involves both neuronal apoptosis and inflammatory responses (Czlonkowska et al. 2000; Beal 2003; Teismann and Schulz 2004). We describe expression of the G-CSF receptor in the nigral dopaminergic system, and find neuroprotective effects of G-CSF in models of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine/1-methyl-4-phenylpyridinium (MPTP)/(MPP+)-induced toxicity, both in vitro and in vivo. Moreover, we could demonstrate that G-CSF application attenuates MPTP-induced striatal dopamine depletion, suggesting that G-CSF is a candidate molecule for neuroprotective treatment strategies of PD. Because G-CSF already has a proven favorable clinical safety profile, it could possibly be transferred rapidly to phase II/III clinical trials.

Materials and methods

MPTP and drug administration in mice

10–12 month old male mice (C57BL/6JOla Hsd) were purchased from Harlan Winkelmann (Borchen, Germany). As a spontaneous C57BL/6 α-synuclein deletion strain with identical strain specification, albeit from a different breeding colony, has been described (Specht et al. 2001), α-synuclein expression was confirmed by Western blotting of brain lysates. Animals (n = 7–10 per experimental group) were housed in plastic cages. They were kept under a 12 h light/12 h dark cycle with free access to food and water. For MPTP intoxication, mice received 5 intraperitoneal injections on 5 consecutive days. The control group was injected with PBS as vehicle. Granulocyte-Colony Stimulating Factor (G-CSF) (Neupogen®, AMGEN, Thousand Oaks) treatment was given at 40 µg/kg bodyweight diluted in the appropriate buffer [10 mm sodium acetate, 250 mm sorbitol, and 0.004% Tween 80 (pH 4)] as daily subcutaneous injections during the whole experiment. Treatment was started one day before the first MPTP intoxication and continued daily 6 h before the MPTP injections.

Dosing was adjusted according to a non-linear dose/bodyweight relationship. Usually, according to the standard chronic MPTP paradigm, a dose of 30 mg/kg body weight (BW) is applied for mice aged 8 weeks (Dauer and Przedborski 2003). Mice at this age usually weigh about 20 g, resulting in an average MPTP dose of 0.6 mg per animal. This dose was proportionally increased according to the respective BW2/3. Consequently, given the body weight of our 10–12 month old animals (BWaged mice) and the body weight of 8–week old animals (BWyoung mice; per definition set to the typical 20 g), we applied the following doses:


For example, the body weight of a 10–12 month old mouse of 30 g results in a dose of 0.786 mg MPTP, corresponding to 26.2 mg/kg BW MPTP. Most 10–12 month old mice weighed between 27 and 32 g. Experimental groups were matched for body weight distribution. Mean body weights were 28.4 ± 0.7 g, 28.3 ± 0.5 g and 28.1 ± 0.4 g for the vehicle, MPTP and MPTP/G-CSF treatment group in the neuroprotection experiment, respectively.

Appropriate guidelines were followed in handling MPTP, and MPTP injected mice (Przedborski et al. 2001). Animal experiments were approved by the appropriate German authorities (Bezirksregierung Braunschweig).

Brain preparation and immunohistochemistry

Seven days after the last dose of MPTP, mice were anesthetized with chloralhydrate and perfused intracardially with saline, followed by 4% paraformaldehyde (PFA) in PBS (pH 7.4). Brains were quickly dissected, additionally post-fixed in PFA overnight at 4°C, cryoprotected in 30% sucrose in PBS for 24 h at 4°C, snap frozen, and stored at −80°C until sectioning. For each group frontal cryostat-cut sections (50 µm) through the substantia nigra were collected free-floating.

After several rinses in PBS, sections were incubated in 10% normal goat serum for 30 min at room temperature, followed by incubation with a polyclonal anti-tyrosine hydroxylase antibody (1 : 1000; CliniScience) diluted in 0.1% Triton-X100 and 1% normal goat serum in PBS for 24 h at 4°C. All brains were processed with the same antibody lot. After three rinses in PBS, sections were incubated with Cy3-conjugated affinity purified IgG (goat-antirabbit; 1 : 200; Dianova) for 1 h. Finally, sections were DAPI stained, washed in PBS, dehydrated and coverslipped. The total number of tyrosine hydroxylase (TH) stained SNpc neurons was counted manually in every second section. Only cells with identified nucleus were counted. The total number of cells in the respective SNpc was calculated by multiplying the number of counted neurons with a factor of 2. The same method was applied for all groups by a blinded observer, which further argues against the possibility of a bias.

The tyrosine hydroxylase (TH) promoter (TH-EGFP) mouse line used for G-CSF receptor staining was originally created by micro-injection of the DNA construct into fertilized C57BL/6 J × DBA/2 J F2 eggs, and then crossed to an inbred C57BL/6 J background (Matsushita et al. 2002). In our hands, the line was maintained by breeding with wild-type animals of the C57BL/6JOla Hsd substrain (Harlan Winkelmann). Immunohistochemical staining of G-CSF receptor protein in the TH-EGFP transgenic mice was performed similar to the TH staining, but using 20 µm cryostat sections. A rabbit polyclonal G-CSFR primary antibody was used (1 : 400 dilution; SantaCruz), followed by incubation with Cy3-conjugated affinity purified IgG (goat-antirabbit; 1 : 200; Dianova). Co-localization of EGFP and Cy3 fluorescence was imaged with a Leica confocal microscope, using a 63x oil immersion objective and corresponding filter setting. For negative controls, the primary anti-serum was omitted and single stainings served as control for any fluorescence cross-talk between detection channels.

For G-CSF receptor staining in Wistar rats (250 g; Charles River, Sulzfeld, Germany), coronar sections of paraffin-embedded rat brain (2 µm) were de-paraffinated and microwaved (citrate buffer at 500 W for 10 min). Afterwards, sections were incubated at 4°C over night with the rabbit G-CSFR antibody (1 : 100) and a mouse tyrosine hydroxylase antibody (1 : 100; R & D Systems). Following incubation with a biotinylated antirabbit secondary antibody (Vector Laboratories; 1 : 200), detection was carried out by incubation with Streptavidin-TRITC (Dianova; 1 : 200) and an antimouse FITC-conjugated antibody (1 : 200, Dianova). Nuclear staining was performed by using Hoechst33342.

HPLC measurements

On the day of the assay (12 days after the first MPTP injection), tissue samples (8–10 animals per experimental group) were sonicated in 20 µl of 0.1 m perchloric acid/mg of striatal tissue. After centrifugation (14 000 g, 30 min, 4°C), 20 µl of supernatant was injected onto a C18 reverse-phase HR-80 catecholamine column (ESA, Bedford, MA, U.S.A). Dopamine, 3, 4-dihydroxyphenylacetic acid (DOPAC), and homovanillic acid (HVA) were quantified by HPLC with electrochemical detection. The mobile phase (pH = 2.9) consisted of 90% 75 mm sodium phosphate, 275 mg/L octane sulfonic acid solution, and 10% methanol. Flow rate was 0.8 mL/min. Peaks were detected by an ESA Coulochem II with a model 5010 detector (E1 = 20 mV, E2 = 320 mV). Data were collected and processed using the Chromeleon computer system (Dionex, Idstein, Germany).

Primary midbrain neuronal cultures and immunocytochemistry

The mesencephalic floor plate was dissected from E14 Wistar rat embryos and further processed for establishing dissociated cell cultures as previously described (Lingor et al. 2000). Cells were seeded on poly L-ornithine/laminin (Sigma)-coated coverslips (12 mm diameter) at a density of 175 000 cells/cm2. Cultures were maintained at 37°C in a humidified atmosphere and 5% CO2 in DMEM/F12 plus the N1 supplements and antibiotics for 6 days. Medium was changed 24 h after establishing cultures, on day in vitro (DIV) 1, and subsequently every second day.

On DIV4 G-CSF (Neupogen®, AMGEN, Thousand Oaks) was added at concentrations of 30 ng/mL and 50 ng/mL. After 12 h (beginning of DIV5), MPP+ (4 µm) was applied, and cells fixed after 24 h of MPP+ treatment with 4% PFA in PBS (pH 7.4) for 15 min. In order to identify dopaminergic neurons in cultures prepared from rat embryos, immunocytochemical staining against tyrosine hydroxylase was performed. To reduce unspecific binding diluted goat serum (10% in PBS) was applied for 10 min at 23°C, followed by incubation with a polyclonal anti-TH antibody (rabbit, CliniScience) at 1 : 1000 dilution. For fluorescence labelling a secondary Cy-3-labeled goat-antirabbit antibody was used. Coverslips were mounted on glass slides and observed with a Zeiss Axiophot microscope. For G-CSF receptor staining, a GCSF-R antibody (rabbit polyclonal, SantaCruz) was used at a 1 : 500 dilution.

All chemicals were freshly prepared and dissolved in cell culture medium immediately prior to application.

TH-positive neurons were counted at 20-fold magnification in six different areas per well. In cell counts, only TH-positive neurons possessing neurites were considered, to avoid distortion by non-functional yet TH-positive cell debris. Cell survival is expressed as percentage standardized to the unchallenged and untreated control condition (100%). All experiments were performed at least in duplicate with at least three wells in parallel.

MPP+ toxicity assay in PC12 cells

PC12 cells were seeded at a density of 40 000 in 96-well plates in the following medium: DMEM low glucose, 10% horse serum +5% FCS, 1% Pen/Strep. MPP + (Sigma) was added to yield final concentrations of 50–6250 µm. Vehicle or G-CSF (Neupogen®, AMGEN, Thousand Oaks) at 50 ng/mL were added. After 24 h, cell viability was determined by a fluorimetric cell viability assay (CellTiter-blue assay, Promega) using a Fluostar platereader (BMG). Eight wells were assayed for each data point. Data are given as mean relative cell viability ± SEM normalized to 100% viability in respective controls.

PCR for the G-CSF receptor

RNA was prepared from substantia nigra tissue from mouse using the Chomczynski- Sacchi method (Chomczynski and Sacchi 1987) followed by Qiagen RNA-Easy columns. Reverse transcription was performed using oligo dT primers and Superscript II reverse transcriptase (Stratagene). PCR was carried out using the following primers: mG-CSFR_2582 s (TGTGCCCCAACCTCCAAACCA) and mG-CSFR_2817as (GCTAGGGGCCAGAGACAGAGACAC) with an expected product length of 235 bp. Annealing was carried out at 64°C, and PCR was run for 45 cycles in a Lightcycler thermal cycler (Roche Diagnostics). For detection of the G-CSF receptor mRNA in rat PC12 cells, the following primers were used: ratGCSFR-frag32s (CCATTGTCCATCTTGGGGATC) and ratGCSFR-frag265as (CCT GGA) with an expected product length of 233 bp. PCR was performed over 50 cycles using 66°C as annealing temperature.

Statistical analysis

All values are expressed as mean ± SEM The two-sided t-test, or anova followed by Student-Newman-Keuls post-hoc test was used to determine significance as appropriate (NCSS Software, NCSS, Kaysville, Utah, USA). A p-value < 0.05 was considered statistically significant.


G-CSF receptor expression in dopaminergic substantia nigra neurons

While performing immunohistochemistry for the expression of the G-CSF receptor (G-CSFR) in the CNS, we noted immunopositive signals in the substantia nigra pars compacta (SNpc) (Fig. 1a). Many of the G-CSFR-positive cells showed a clear neuronal morphology (Fig. 1a, insert). We confirmed this result on the mRNA level by RT-PCR analysis of dissected mouse substantia nigra tissue (Fig. 1b). By double-immunofluorescence analysis in the rat, we could locate the G-CSF receptor signal to cells that expressed the established marker for dopaminergic neurons, TH. Indeed, all TH-positive cells in the dorsal or ventral part of the rat substantia nigra pars compacta were positive for the G-CSF receptor (Fig. 1c-h). To confirm the results on G-CSF receptor expression at the protein level in mice, we used a transgenic mouse line EGFP under the control of the tyrosine TH promoter (TH-EGFP mouse line; Matsushita et al. 2002). In this mouse line, TH positive cells, i.e. dopaminergic neurons, are labelled by EGFP fluorescence with a sensitivity and specificity of more than 90% in the SNpc, as shown by comparison with TH immunohistochemistry in a previous study (Matsushita et al. 2002), allowing precise identification of dopaminergic neurons with better signal-to-noise-ratio compared with immunohistochemical TH stainings. We found G-CSFR protein to be expressed by the majority (84.1 ± 0.7%) of EGFP fluorescent SNpc cells, but also in the majority of dopaminergic ventral tegmental area neurons (Fig. 1i-k). Few other cells than dopaminergic neurons expressed G-CSFR (Fig. 1i-k). However, the sensitivity of EGFP expression for dopaminergic neurons is not 100% in these mice (Matsushita et al. 2002). Therefore, the respective cells could represent either the few dopaminergic neurons that failed to express EGFP, or be truly non-dopaminergic but G-CSF receptor-positive cells. So far, we are unable to differentiate between these two possibilities. The immunohistochemical data obtained by direct co-staining from the rat however, suggest that there are indeed neurons in the SN that express G-CSFR but are not TH-positive.

Figure 1.

G-CSF receptor expression in the rodent substantia nigra. (a) Robust G-CSFR immunoreactivity was found in the rat substantia nigra, but also in the ventral tegmental area. Insert shows a typical cell with neuron-like morphology. (b) RT-PCR on RNA isolated from the substantia nigra of mice. There is a strong specific band for the G-CSF receptor with the appropriate size (expected size 235 bp). M: DNA marker; SN: cDNA prepared from substantia nigra; neg. negative control without addition of reverse transcriptase. (c- h) Coronal paraffin sections of rat brain in the area of the substantia nigra were stained with an antibody against the G-CSF receptor (Cy3-labelled secondary antibody, red) and tyrosine hydroxylase (TH; FITC labeled, green). In merged pictures, nuclei are additionally stained with Hoechst33342. (c-e) dorsal part of the SNpc. (f-h) ventral part of the SNpc. All neurons that expressed TH were also positive for the G-CSF receptor (see merged images). Occasional, unidentified cells, were positive for G-CSFR, but not TH. (i- k) Mice dopaminergic SNpc neurons are labeled by transgenic EGFP expression under the control of the TH promoter. Overlapping G-CSFR immunoreactivity (Cy3-labelled secondary antibody, red) is shown in (j) and (k; merged). Note that most but not all EGFP fluorescent neurons express G-CSFR (arrowhead). Dopaminergic (i.e. EGFP fluorescent) cells expressed G-CSF receptor protein (arrow). SN: substantia nigra pars compacta; VTA: ventral tegmental area.

Thus, as a pre-requisite for a direct neuroprotective action of G-CSF, the vast majority of dopaminergic neurons express the G-CSF receptor.

G-CSF protects against MPP+ toxicity in PC12 cells and primary neuronal midbrain cultures

After we had found G-CSFR expression by substantia nigra dopaminergic neurons in vivo, we consequently proceeded to study potential neuroprotective effects of G-CSF on dopaminergic neurons.

PC12 cells is a rat clonal pheochromocytoma cell line (Greene and Tischler 1976) that can synthesize, metabolize, and transport dopamine (Rebois et al. 1980; Tuler et al. 1989). We found that PC12 cells expressed the receptor for G-CSF as verified by RT-PCR (Fig. 2a). Therefore, as a first screening experiment, we determined efficacy of G-CSF in MPP+-induced neurotoxicity in PC12 cells. PC12 cells were treated with increasing concentrations of MPP+ (50, 250, 1250 and 6250 µm MPP+) which produced a dose-dependent decrease in cell viability. After treatment with 50 µm MPP+, cell viability was found to be reduced to 75.9 ± 1.2% compared to control cultures, and further decreased to 6.3 ± 0.1% at the highest MPP+ dose tested (6250 µm). Concomitant treatment with G-CSF (50 ng/mL) fully protected against toxicity of 50 µm MPP+ after 24 h (101.4 ± 2.1% cell viability compared to unchallenged control cultures; Fig. 2b). With higher MPP+ concentrations, the protective effect relatively diminished, but G-CSF significantly enhanced cell viability at all concentrations tested, even slightly at 6250 µm MPP+ (8.7 ± 0.7% cell viability with 6250 µm MPP+ and G-CSF treatment; Fig. 2b).

Figure 2.

G-CSF protects PC12 cells and rat primary dopaminergic midbrain neurons against MPP+ toxicity. (a) RT-PCR on RNA isolated from PC12 cells. There is a strong specific band for the G-CSF receptor with the appropriate size (expected size 233 bp). M: DNA marker; PC12: cDNA prepared from substantia nigra; neg: negative control without addition of reverse transcriptase. (b) PC12 cells were treated with increasing concentrations of MPP+, the active metabolite of MPTP. Cell survival decreased with higher MPP+ doses [vehicle, grey bars; all values expressed as percent survival compared with unchallenged control cultures (set to 100%)]. Concomitant G-CSF treatment at a concentration of 50 ng/mL significantly enhanced cell survival after 24 h at all MPP+ concentrations [p < 0.05 (*); unpaired two-tailed Student's t-test], but the relative protective effect decreases with increasing MPP+ concentrations (G-CSF; black bars). (c- f) Compared with control cultures (c, f), MPP+-treatment for 24 h resulted in a marked reduction of TH-immunoreactive neurons in rat primary neuronal midbrain cultures (d, f). Treatment with G-CSF (e, f) attenuated MPP+ toxicity. The MPP+/vehicle group was compared to the untreated control group and both MPP+/G-CSF treatment groups (30 ng/mL and 50 ng/mL G-CSF) by anova followed by Student-Newman-Keuls post-hoc test; F3,23 = 21.5; (*): p < 0.05 (*).

However, the PC12 cell line must be regarded as an artificial in vitro system. Due to their immortalized character, and the different anatomic origin of these cells their predictive relevance for the nigral dopaminergic system is unclear. Therefore, we next proceeded to rat primary neuronal midbrain cultures to confirm the neuroprotective effects of G-CSF. Immunohistochemically, G-CSF receptor expression was found in the majority of TH positive neurons (data not shown). After the rat primary neuronal midbrain cultures were exposed to 4 μM MPP+ for 24 h, quantification of surviving TH immunoreactive cells demonstrated a marked toxicity with a reduction of surviving dopaminergic neurons to 26.2 ± 2.9% in MPP+ treated cultures compared with untreated controls (Figs 2c.d.f). Survival was substantially enhanced by G-CSF treatment at a concentration of 50 ng/mL and 30 ng/mL, with 30 ng/mL showing a slightly lower efficacy (57.2 ± 9.7% and 53.7 ± 10.9% surviving dopaminergic neurons, respectively; Figs 2e.f).

G-CSF protects against MPTP toxicity in older mice

Next, to ask whether our in vitro findings were of relevance to an in vivo model for PD, we chose the subacute MPTP-model in mice (MPTP injection on 5 consecutive days), which is known to induce apoptosis of dopaminergic neurons (Tatton and Kish 1997; Eberhardt et al. 2000). PD has a mean onset of disease at around 60 years of age, and we therefore chose aged mice (10–12 month old), in contrast to the usually 8–12 week old animals used in most previously published MPTP experiments. Because older C57Bl/6 mice are known to be more susceptible to MPTP than younger mice (Irwin et al. 1992), we chose a non-linear body weight/dose relation (see material and methods). Mice were treated with G-CSF for 13 days at a dose of 40 µg/kg bodyweight per day subcutaneously, starting one day before the first MPTP dose. As expected, MPTP treatment caused a significant loss of TH-positive neurons in the SNpc 12 days after the first MPTP injection (4908 ± 180 and 2154 ± 92 dopaminergic neurons comparing vehicle/saline versus vehicle/MPTP; Figs 3a.b.d). In agreement with our in vitro results, G-CSF substantially protected dopaminergic neurons from MPTP-induced cell death in the SNpc (3772 ± 66 dopaminergic neurons; Figs 3c.d).

Figure 3.

G-CSF protects against MPTP toxicity in vivo. 10–12-month-old mice were injected with MPTP or saline control on 5 consecutive days according to a non-linear dose/body weight dosing regime (see material and methods), and dopaminergic neurons were stained with a primary antibody against TH. The secondary antibody was labeled with Cy3. Panels a-c show representative pictures of the SNpc. While MPTP injections resulted in a significantly reduced number of SNpc dopaminergic neurons (b, d) compared with vehicle injected control animals (a, d), subcutaneous injection of G-CSF (40 μg/kg BW daily over 13 days) rescued dopaminergic neurons (c, d). Significance between MPP+ group and both other groups was assessed by anova followed by Student-Newman-Keuls posthoc test; F2,20 = 24.2; *: p <0.05.

G-CSF attenuates MPTP-induced striatal dopamine depletion

Finally, we asked whether G-CSF application would result not only in increased morphological survival of dopaminergic SNpc neurons, but would also protect against MPTP-induced depletion of striatal dopamine levels. Using the same MPTP dosing regime and aged C57Bl/6 mice as described above, we found a decrease in striatal levels of both dopamine (4.0 ± 0.3 ng/mg compared with 22.7 ± 1.5 ng/mg tissue weight in control animals) as well as the dopamine metabolites 3, 4-dihydroxyphenylacetic acid (DOPAC) (0.8 ± 0.1 ng/mg vs. 2.4 ± 0.3 ng/mg) and homovanillic acid (HVA) (0.5 ± 0.1 ng/mg versus 1.3 ± 0.3 ng/mg), as measured by HPLC 12 days after the first MPTP injection. G-CSF treatment significantly attenuated MPTP-induced loss of dopamine, DOPAC and HVA levels in the striatum (5.77 ± 0.27 ng/mg, 1.22 ± 0.12 ng/mg and 1.07 ± 0.15 ng/mg tissue weight, respectively; Fig. 4 shows percentage values normalized to control animals). Confirming the specificity of our finding, we did not detect significant changes in serotonin levels or its metabolite 5-hydroxyindole acetic acid (5-HIAA) upon MPTP application or G-CSF treatment (Fig. 4).

Figure 4.

Effect of G-CSF administration on MPTP-induced depletion of dopamine and its metabolites in the mouse striatum. Values are given in percent, normalized to concentrations in vehicle-treated control mice. G-CSF treatment significantly reduced MPTP-induced dopamine and dopamine metabolite depletion, while serotonin and its metabolite 5-HIAA remained unaltered. Note that a different scale was chosen for dopamine, as relative dopamine depletion by MPTP administration substantially exceeded reduction of DOPAC and HVA levels. Statistical significance between the respective two groups (MPP+/vehicle vs. MPP+/G-CSF treatment) was assessed by an unpaired two-tailed Student's t-test. *: p < 0.05.


This study demonstrates for the first time neuroprotection of G-CSF in a model for a chronic neurodegenerative disease, with regard to both morphological maintenance of neuronal cell bodies and preservation of striatal dopamine levels.

We aimed at finding a neuroprotective molecule that, given to be effective in our disease models, could be immediately transferred to clinical studies. G-CSF is already in clinical use for many years with a good safety record for the treatment of neutropenia (Crawford 2003). Physiologically, G-CSF regulates survival, differentiation, and proliferation of the neutrophil leukocyte lineage (Sakamoto et al. 2003; Tehranchi et al. 2003; Mangan and Reddy 2005), partly due to a direct anti-apoptotic effect. This anti-apoptotic activity appears preserved in neurons (Schneider et al. 2005). Together with its neuroprotective efficacy in models for cerebral ischemia, this prompted the hope that G-CSF might be a potential therapeutic molecule to treat chronic neurodegenerative diseases, for example Parkinson's disease. Importantly, we have recently shown that G-CSF has the ability to penetrate the intact blood–brain barrier in rodents (Schneider et al. 2005).

Here, we describe G-CSF receptor expression in both rat and mouse dopaminergic SNpc neurons, and show that G-CSF can prevent neuronal cell death in models of Parkinson's disease. We found increased numbers of morphologically preserved dopaminergic neurons challenged with MPTP or MPP+, both in a dopaminergic cell line and in primary neuronal midbrain cultures, as well as in vivo. In our in vivo experiments, we tested only one dose regime, and can not exclude that an even stronger benefit could have been observed with other treatment protocols. Theoretically, the observed decrease in TH positive cells in the SN, and the respective therapeutic effect of G-CSF, could also be due to a transient cellular atrophy and regulation of TH expression rather than representing true cell death and neuroprotection. However, we have addressed this possibility already in an earlier study by retrograde Fluorogold-labelling of dopaminergic neurons (Eberhardt et al.; 2000). Using the same MPTP dosing regime, the number of Fluorogold-labelled cells was reduced similarly (by 45%) compared to the reduction of TH positive cells in Eberhardt et al. (about 50%) or in the present study (44%). Thus, after MPTP treatment, loss of TH positive neurons was due to cell death and not just a result of TH down-regulation in atrophic neurons. Accordingly, the observed G-CSF effect can not be explained by TH up-regulation. Moreover, our in vitro data also support a true neuroprotective effect. In particular, the data from PC12 cells can not be due to such a bias, as cell survival is not assessed by TH staining in this case.

In contrast to our results, a very recent study failed to detect protective effects of G-CSF in dopaminergic neurons in primary neuronal midbrain cultures (Henze et al. 2005). This apparent discrepancy could be due to the lower G-CSF concentrations used in that study compared with our work. While Henze et al. tested a concentration range of 0.1–10 ng/mL, we found robust neuroprotection at 30 ng/mL and 50 ng/mL.

G-CSF passes the intact blood–brain barrier (Schneider et al. 2005), and we found that TH-positive neurons express the G-CSF receptor. Moreover, we observed neuroprotection in the G-CSF receptor expressing PC12 cell line, where a contribution of glial cells or immune mechanisms is excluded. Hence, our findings strongly argue in favor of a direct receptor-mediated anti-apoptotic neuroprotective action of G-CSF in the MPTP model. However, at our present knowledge we cannot exclude that more systemic, or at least indirect, effects of G-CSF also play a role in protection in the MPTP model. Anti-inflammatory strategies can ameliorate the action of MPTP (e.g. (Choi et al. 2005)), and G-CSF has prominent systemic anti-inflammatory properties (Hartung 1998) that might contribute to its protective action in this model. Indeed, G-CSF previously proved to be effective in experimental allergic encephalomyelitis (EAE) (Lock et al. 2002; Zavala et al. 2002).

Importantly, G-CSF treatment also attenuated MPTP-induced striatal dopamine and dopamine metabolite depletion. This indicates preservation of axonal projections. Because of the short time between MPTP treatment and HPLC measurements, it is unlikely that the observed effects are due to sprouting, as first markers of sprouting have been observed only from day 14 after MPTP treatment onwards (Ho and Blum 1998). Without such a protective effect on striatal dopamine metabolism, PD patients would be unlikely to benefit from neuroprotective therapy in terms of clinical outcome. The rescuing effect of G-CSF on the levels of striatal dopamine and its metabolites was moderate, and far from the protection of dopamine levels reported before with other therapeutic strategies in this model. However, many of these studies included therapeutic approaches that will not be clinically applicable in the near future, e.g. experimental compounds without any clinical safety record or genetic manipulations (Przedborski et al. 1996; Yang et al. 1998; Klivenyi et al. 2000; Vila et al. 2001). In addition, it is unclear if quantitative data from differing experimental settings of the MPTP model are directly comparable (e.g. there are few data available with aged mice). Importantly, many anti-apoptotic or anti-inflammatory strategies failed to show any protective effect on dopaminergic terminals (Liberatore et al. 1999; Dehmer et al. 2000; Eberhardt et al. 2000; Klivenyi et al. 2000; Crocker et al. 2003). The aim of our study was to provide proof of principle that G-CSF, which is already in clinical use, is beneficial in a model of PD. The use of G-CSF for long-term treatment in a more chronic paradigm, e.g. in PD patients, may also be protective at the level of dopaminergic terminals, but its functional impact, especially in a clinical setting, will have to be clarified by further investigations beyond the scope of this study.

Using the model of MPTP toxicity, we tested G-CSF in a widely accepted in vivo paradigm for PD (Dauer and Przedborski 2003) with, according to most recent data, still increasing evidence concerning its relevance for PD (Fornai et al. 2005). In contrast to the vast majority of previous studies using the MPTP model, we established the MPTP paradigm with a modified MPTP dosing regime in 10–12 month old mice. This modified model further supports the relevance of our findings, as older mice are more appropriate for studies on potential new PD treatment strategies than the commonly used 8–10 week old animals, which most likely do not reflect the situation in PD patients with their usually advanced age. Nevertheless, a caveat has to be expressed, as the G-CSF dosing regime, lesion status of the substantia nigra at clinical diagnosis, time-course of cell death at clinical diagnosis and time cause of cell death could still differ between our MPTP model and PD patients. Thus, also mechanisms of all cell death and protection might be at least partially different.

Our concept of G-CSF as a neuroprotectant parallels the recent use of erythropoietin (EPO), which physiologically regulates red blood cell numbers, in neurological disease models and clinical stroke studies (Ehrenreich et al. 2002). Also EPO receptor expression has recently been found in the CNS, and application of recombinant human EPO (rhEPO) prevented neuronal cell death (Siren et al. 2001; Weishaupt et al. 2004). Enthusiasm for rhEPO as a potential neuroprotective therapeutic could be tempered, however, as it also increases red blood cell counts and platelet aggregability. Although phase II studies for G-CSF in neurodegenerative diseases are lacking to date, G-CSF could turn out to have a more advantageous safety profile when compared to EPO. As both cytokines activate different anti-apoptotic signalling cascades (Digicaylioglu and Lipton 2001; Celik et al. 2002; Schäbitz et al. 2003; Schneider et al. 2005), a potential combination of these two cytokines in neurodegenerative diseases might be advantageous from a theoretical point of view, should these proteins indeed turn out to have synergistic effects. However, the intention of this study was to give evidence for a therapeutic potential of G-CSF for the treatment of PD. Further work will be needed to delineate the precise cellular and molecular pathways underlying this effect, although most recent data suggest that principle aspects of G-CSF signal transduction might be conserved between the myeloid system and neuronal cells (Schneider et al. 2005).

Taken together, G-CSF is a neurotrophin that, on the one hand, has robust neuroprotective properties for dopaminergic neurons and striatal dopamine levels in a model of PD, and, at the same time, has already been established with a good safety profile in clinical settings for many years. Therefore, this protein fulfils the criteria for a novel neuroprotectant mentioned in the introduction. It is advantageous over other neurotrophic factors by an unproblematic delivery to the brain and a decreased risk to encounter major unexpected adverse effects in Parkinson patients. Nevertheless, a safety record for the chronic treatment of PD with G-CSF, also with respective dose-finding studies in humans, will have to be performed in phase IIa/b studies. Potential safety issues mostly relate to the known actions of G-CSF in the hematopoietic system, however, safe applications of G-CSF even over years have been reported (Dale et al. 1993; Bonilla et al. 1994). So far, we have no indications to assume that it might be possible to construct derivatives of G-CSF that lack systemic effects, as has been done for EPO (Erbayraktar et al. 2003; Leist et al. 2004), where a specific neural receptor composition is suggested (Brines et al. 2004). Quite to the contrary, it might be even desirable to maintain systemic immunomodulatory functionality of G-CSF in Parkinson patients. For long-term development, theoretically attractive alternatives for clinical practice might be orally available G-CSF receptor agonists (Tian et al. 1998), or pegylated G-CSF forms with longer plasma half-life times (Kinstler et al. 1996), should these be able to pass into the brain.


We thank Christine Poser, Frank Herzog, Claudia Heuthe, and Ulrike Bolz for excellent technical assistance, Nicole Schmidt for dissecting tissue, and Cathy Ludwig for critically reading the manuscript. This study was partially supported by the DFG Research Centre ‘Molecular Physiology of the Brain’ (CMPB).