Infarct size reduction
Here we demonstrate a robust infarct-reducing effect of a 4 h treatment delay with G-CSF in a severe hemispheric stroke model (MCAO), viewed as the gold standard for testing neuroprotective effects. The end point for determining infarct volume was 24 h following onset of ischemia, a standard timepoint for neuroprotective studies. We know, however, that the effect of G-CSF is not transient as has been discussed with some experimental drugs, but is also detectable when examining the brain at 72 h post ischemia after a combined CCA/distal MCA occlusion . The achieved total reduction in infarct volume of 34.5% is an extension of previous studies in the same model where treatment was initiated 30 min  and 2 h  after onset of ischemia. We also demonstrate for the first time a clear treatment effect both on cortical as well as subcortical areas with a meaningful sample size per group. This is important, as the effect of a number of neuroprotective drugs has been shown to be limited to the cortex, e.g. . The neuroprotective acitivity of G-CSF in a broader range of brain regions can be explained by the broad distribution of the G-CSF receptor in the brain . However, the effect on lesion volume in the cortex is more pronounced than in the subcortex, which is either explained by the relative distribution of infarct core and presumptive penumbra to striatum and cortex in the MCAO filament model, or indeed demonstrates a stronger protective effect in the cortex.
Support for a delayed window of opportunity for G-CSF in acute stroke models comes from three other independent studies. Treatment effects were investigated in comparatively mild ischemia models producing small hemispheric or cortical infarctions appropriate to assess recovery effects rather than to test the neuroprotective potential of a drug in the acute phase. Six et al. have shown a stunning 55% infarct reducing effect when treatment was delayed for 24 h after transient MCAO in the mouse with a relatively small sample size of n = 6 per group . Komine-Kobayashi et al. demonstrated a decreased infarct size in the same model after initiation of G-CSF treatment 24 h and 72 h after the infarct . Shyu et al. reported effects on infarct size when treatment was initiated as late as 24 h after the insult in a rat combined CCA/distal MCA model with 90 min occlusion of the distal MCA, which produces purely cortical infarcts . Although these studies have used milder models, it appears probable from the cumulated evidence that the window of opportunity in a severe hemispheric stroke model could be also considerably longer than 4 h, especially in view of the still sizable treatment effect observed at this point. It is therefore desirable to further explore this acute time window and define the timepoint where efficacy is lost.
By comparing the differences of total infarct volumes without and with edema correction, our data indicate an absolute reduction in edema volume in favor for G-CSF (compare 51.9 mm3 to 72.5 mm3 in the vehicle treated group, Figure 1). This seems to be in agreement to the data generated by Gibson et al  who have seen edema reduction in both transient and permanent MCAO after G-CSF treatment. There, the authors have used a direct method to determine edema by measuring brain total water content by the dry weight method. However, as it is well known that larger infarcts produce more vasogenic edema, it is important to relate the edema volume to the infarct volumes in the respective groups to exclude edema reduction simply as a consequence of reduced infarct volume. Here, the edema volume is in similar proportion to the total infarct size in both the vehicle (21.7%) and the G-CSF group (23.2%). Thus, there is no evidence for a separate effect on edema extent by G-CSF treatment. As there are no data in the work mentioned above to estimate the relative edema extent to infarct sizes, which are decreased by G-CSF treatment, it is difficult at the moment to judge whether the data can be truly interpreted as an independent edema-reducing effect of G-CSF.
We have previously already demonstrated significant activity of G-CSF towards enhancing post-stroke recovery in the photothrombotic model in conjunction with stimulation of neurogenesis by using a diligent assessment of sensorimotor function . In this previous experiment we have applied a five-day treatment of G-CSF at 15 μg/kg bodyweight/day starting with the first dose 1 h after induction of ischemia. Therefore it is possible that part of the recovery effect seen is due to the acute antiapoptotic property of G-CSF. To minimise any acute effects on infarct size, and define time windows for G-CSF's pro-regenerative functions after stroke, we delayed the initial treatment to 24 and 72 h post ischemia. In addition, we chose a dosage scheme that appears clinically safe and feasible for a treatment period of weeks, 10 days of 10 μg/kg bodyweight daily. This application scheme is similar to the one used in the previous recovery experiment, but has a longer duration and lower single doses of G-CSF. This scheme is also readily translatable to an explorative study in stroke patients for testing the pro-regenerative potential in man. For exploring the therapeutic time window for initiation of a recovery-enhancing effect post-stroke we have here concentrated on a robust parameter, time to fall off the rotarod device. The course of the curves for all ischemic groups neatly demonstrates that the groups only start to separate in the second week of the experiment, and highlight that the initial value post-stroke is absolutely identical, indicating that there was no difference in the extent of the initial damage in this model relevant to sensorimotor performance. Thus, we have reached our goal to uncouple acute neuroprotective effects of G-CSF from effects on long-term recovery.
Surprisingly, a recovery-enhancing effect at the 72 h delayed treatment initiation was still clearly detectable, and not significantly different from the initiation at 24 h. This indicates that the window of opportunity for pure recovery enhancement with G-CSF may not be time-dependent in the same way as therapeutic effects in the acute phase after stroke. Indeed, there is no statistical difference between the 24- and 72-h initiation of treatment. A substantial window of opportunity for improvements in functional outcome by G-CSF treatment is also suggested by Lees et al. , who report effects in rotarod measurements particularly when treatment was delayed to 1 day and longer after MCAO.
It will now be very interesting to further delay treatment and to define the limits of G-CSF efficacy in such models. There is a surprising scarcity of data relating to the characterisation of susceptible post-stroke time periods for recovery processes in the rodent. In light of the neurogenesis-enhancing effects of G-CSF it is highly interesting that, at least for ischemia-induced neurogenesis, there is no clear time dependence from the insult, and increased neurogenesis may keep up for months after stroke , thereby likely providing a long period of susceptibility to therapeutic interference by neurogenesis-enhancing substances such as G-CSF.
Implications for the clinic
Here, we have further delineated a truly bimodal efficacy of G-CSF both for acute effects on infarct size, as well as on chronic enhancement of post-stroke recovery. Together with a substantial body of evidence for efficacy of G-CSF in varied stroke models (reviewed in ) our data further strengthen confidence in the potential of G-CSF for the treatment of human stroke.
In addition to a rational mode-of-action within stroke pathophysiology, any candidate for further stroke drug development should be explored according to the STAIR criteria . G-CSF fulfills these criteria well: blood-brain-barrier penetration, neuroprotective activity in different stroke models including permanent ischemia demonstrated by independent groups, activity shown in different species, well-known pharmakocinetics, as well as functional outcome data. Likewise, data on the potential therapeutic time window are important. We have recently shown that G-CSF is neuroprotective until at least 2 h post-stroke in a severe focal cerebral ischemia model . As indicated by the present results, G-CSF is effective when treatment is delayed as late as 4 h post-stroke. Although conversions to the human situation are delicate, a 4-h time window in this stroke model may relate to a therapeutic time window in humans of several hours.
The effect on functional outcome appears as a separable activity of G-CSF, independent of its neuroprotective potential. This makes G-CSF a highly attractive candidate for recovery enhancement in the subacute or chronic phase of a stroke and increases the likelihood for detecting clinical outcome improvement in acute stroke trials.