G-CSF's natural function of mobilizing stem cells from the bone marrow triggered initial explorations of its potential usefulness in stroke with the idea that mobilized stem cells may home into the injured brain 9. A series of preclinical investigations in animals using G-CSF for the therapy of ischemic stroke was initiated to answer the question whether mobilized bone marrow cells contribute to improved outcome 91011. The capacity of bone-marrow derived cells to restore function in the injured brain has indeed been demonstrated (for review see 12), but the mechanism of their advantageous action remains unclear. The proposed transdifferentiation of bone marrow derived cells into neural cells that induce functional and structural recovery poststroke was recently doubted by several studies (e.g. 1314). The assumption that G-CSF mobilized bone marrow cells might have caused the observed functional improvements was also propagated by Shyu et al 11. However, dividing cells in the ischemic hemisphere, mainly seen in the subventricular zone, were presumably originated from adult neural stem cells concordantly with results from other groups 710. Komine-Kobayashi and colleagues 15 subjected chimeric mice with EGFP-expressing bone marrow-derived cells to transient occlusion of the middle cerebral artery. The authors report that indeed migration of bone-marrow derived monocytes was not increased at all after G-CSF treatment, but rather decreased. So far not evidence proven is another bone marrow cell mediated mechanism. G-CSF may induce invasion of bone-marrow-derived stem cells into the infarcted brain which could contribute to enhanced neuro- and angiogenesis by secretion of neurotrophines and other trophic factors 12. In conclusion the recent evidence from animal experiments cast doubt on the perception that mobilization of stem cells is the sole or even most important mechanism of action for functional recovery after G-CSF treatment.

A first indication of a potential direct effect on cells of the brain came from the observation that G-CSF had a direct protective effect in cultured neurons against glutamate-induced cell death 6. After cerebral ischemia, endogenously released G-CSF is presumably active on the upregulated G-CSF receptor in periischemic regions at risk, the so called penumbra, and may provide protection against apoptotic cell death in neurons (figure 1). Schneider showed that after interaction with its receptor, G-CSF activates through JAK signalling, three independent anti-apoptotic pathways: The signal transducer and activation of transcription (STAT)-3, the extracellular-signal-regulated kinase (ERK) and the phophatidylinositol 3-kinase (PI3K)-Akt pathway 7. Komine-Kobayashi also found antiapoptotic effects of G-CSF on neurons after cerebral ischemia through the JAK/STAT signaling pathway and subsequent activation of Bcl-2 15. Moreover, G-CSF increased cIAP2 levels in the ischemic cortex and thereby decreased the activation of caspase 3, an important trigger of apoptotic processes 16. In a rat model of intracerebral hemorrhage G-CSF's antiapoptotic activity in cells in the perihematomal area was revealed by a TUNEL assay, which detects less DNA fragments as a result from apoptotic signaling cascades after G-CSF treatment 17.

Neural progenitor cells reside for a lifetime in certain areas of the brain, particularly the subventricular zone (SVZ), the olfactory bulb and the hippocampus. Certain conditions such as stroke induce the generation of new neurons from precursor cells, a phenomenon which may potentially be utilized to restore brain function. G-CSF's most striking effect regarding neurogenesis was seen in the dentate gyrus, where the number of newly generated neurons under ischemic conditions 71011 but also in nonischemic, sham-operated animals was increased. In the striatum there was only a trend toward an enhanced neurogenesis after G-CSF treatment which was not statistically significant and the number of newly generated cells was rather small 7. This finding is not surprising, since the striatum is known to habor only a low number of neuronal precursors 1819. Generation of new differentiated cells from endogenous stem cells is an intricate interplay among different components such as proliferation, differentiation, and selective survival. In vivo experiments revealed that G-CSF promotes neurogenesis in all of these components. The number of newly generated cells was increased, the cells differentiate towards a neuronal fate and anti-apoptotic pathways are activated 7. The in vivo findings of an increased neurogenesis after G-CSF treatment were confirmed by in vitro experiments. It was shown that adult neural stem cells isolated from the rat SVZ or hippocampal region that grow as neurospheres in culture express the G-CSF receptor 720. G-CSF dose-dependently induced maturation of cultured progenitor cells towards a neuronal phenotype and increased the population of the differentiated cells 7.

Angiogenesis is a process where new vessels arise from pre-existing ones 21. Future treatment strategies in stroke focus on optimisation of this process in the ischemic boundary zone 22. However, the contribution of angiogenesis to functional recovery after stroke is still unclear 232425. Lee and colleagues showed that G-CSF enhanced angiogenesis in a rat stroke model measured by endothelial cell proliferation, the vascular surface area, the number of branch points, and the vascular length 26. The G-CSF effect was more pronounced when treatment was initiated earlier. But even when treatment was delayed up to seven days after the induction of ischemia an increased angiogenesis accompanied by an enhanced long-term functional recovery could be observed 26. Expression of the vascular marker von Willebrand factor in BrdU positive cells after G-CSF treatment demonstrated the generation of new endothelial cells 11. Taguchi found an accelerated angiogenesis measured by an angiographic score without an enhanced functional outcome 27. However, the results of this study have to be interpreted with caution since immunodeficient mice were used in which G-CSF may not exert its immunomodulative properties. The immunomodulative effects are presumably important for post-stroke functional recovery, as described below.

Recent research revealed that interactions between cerebral ischemia and the immune system are exceptionally relevant for the functional outcome of stroke patients 28. Stroke induced immunodepression can cause infections, such as pneumonia, a frequent complication in stroke patients. However, immunodepression may potentially improve stroke outcome by alleviating the autoaggressive responses due to ischemia-induced exposure of central nervous system-specific antigens to the immune system 293031. Thus, immunomodulation and an increase of in immunocompetence may also be responsible for the rather acute effects of G-CSF32 Indeed, a reduced infiltration of neutrophils and microglia in the ischemic hemisphere after G-CSF treatment was observed 1526, whereas our group could not detect such a difference between the placebo group and the G-CSF group 6. A further analysis of inflammatory cells in the ischemic hemisphere revealed a decreased activation of inducible nitric oxide synthase (iNOS)-positve microglia in animals treated with G-CSF 15. As a consequence of the iNOS inhibition, a reduced nitrotyrosine production as a marker for nitrosadative stress was detected in NeuN positive cells 15. In contrast to these immunohistochemistry and western blot findings there was no G-CSF induced reduction of iNOS on the mRNA level 33. Administration of interleukin-1 beta is known to deteriorate cerebral ischemia and an interleukin-1 beta receptor antagonist may neutralize this effect 34. Thus, the reduction of the ischemia induced interleukin-1 beta upregulation by G-CSF may contribute to infarct size reduction 1633. Recently, our group showed that G-CSF suppresses MMP-9, which is known to mediate inflammation, blood-brain barrier breakdown with subsequent edema formation and tissue injury in acute stroke 35.

Successful testing of a candidate stroke drug in animal models does not firmly predict efficacy in clinical studies 36. As a result of many failed clinical stroke trials the Stroke Academic Therapy Industry Roundtable (STAIR) established recommendations for the preclinical evaluation of stroke drugs 37. The STAIR criteria postulate that the efficacy of a new drug should be demonstrated in a variety of stroke models performed in different species and by different laboratories. Indeed, G-CSF showed efficacy in different species and different stroke models such as transient ischemia in mouse 9153839 and rat 6711 as well as permanent ischemia in mouse 1038 and rat 78. Moreover, G-CSF's efficacy was investigated in animals with comorbitiy, such as diabetes and hypertension, which is important, since stroke patients usually exhibit those conditions 4041. As recommended by the STAIR functional outcome in animal should be tested besides measuring infarct size reduction. G-CSF demonstrated an improvement in short-term 71538 and long-term 78101139 functional neurological deficits. Systemic parameters relevant for stroke pathophysiology such as blood pressure or oxygen saturation were not influenced by G-CSF as measured by physiological monitoring of animals subjected to stroke 6715. The overall high methodological quality of preclinical G-CSF stroke studies was corroborated by a recent systemic analysis 42. Philip and colleagues found that animal experimental stroke studies of G-CSF had the highest quality in a STAIR guideline derived quality score compared to all other neuroprotective agents that are currently investigated in clinical phase II or III trial 42.

To enhance the chance of a successful transfer of preclinical data in clinical trials beyond the application of the STAIR criteria, systematic meta-analyses of candidate neuroprotectants in animal experiments were conducted 43444546. To get an overall impression of G-CSF's efficacy in the recently published preclinical studies and for potential guidance of further clinical studies, we have performed a meta-analysis and meta-regression analysis of G-CSF in animal models of focal cerebral ischemia 47. The meta-analysis showed that G-CSF effectively reduced both infarct volumes and sensorimotor deficits. Infarct sizes were reduced by 42%. The reduction of infarct volumes in G-CSF-treated animals was proportional to the infarct volumes of placebo-treated animals as indicated by the L'abbé plot 47. This proportional infarct size reduction demonstrates G-CSF's efficacy in milder stroke models as well as in severe hemispheric stroke models. Sensorimotor deficits which were categorized in three subgroups (Rotarod running, neuroscore, limb function) were improved between 24% and 40%. Our meta-regression, which was the first meta-regression analysis of a neuroprotective drug in animal stroke models, identified higher doses of G-CSF to be associated with significantly smaller infarct volumes for doses between 10 and 60 μg/kg body weight (infarct size reduction 0.8% per one μg/kg body weight increase in dose when applied within the first 6 hours and 2.1% per one μg/kg body weight increase in dose when applied later than 6 hours after induction of ischemia). Time on Rotarod was significantly extended by 2.1% and 2.2% per one μg/kg body weight increase in dose for early and late treatment initiation, respectively. Also, limb function and neuroscore improved significantly when G-CSF dose was increased. This dose-response relationship is particularly important finding since conclusive experimental dose finding data deriving from a singular stroke study are currently not available. Also a critical aspect of stroke drug development is the therapeutic time window. For G-CSF effects on infarct size the time window is at least 24 hours in the transient suture occlusion model in rodents 911. Regarding functional outcome Zhao 48 reported a beneficial effect of G-CSF when administered more than three month after the onset of ischemia. Using a meta-regression technique we found that a delayed treatment was as effective as an early treatment initiation and may even lead to smaller infarct sizes 47. This result is particularly interesting since the time window for most candidate neuroprotectants closes early after symptom onset 49. The potential of a much longer time-window of G-CSF compared to other stroke drugs might be explained by the above mentioned multimodal actions consisting of neuroprotective and particularly proregenerative properties.