VEGF MIMIC PEPTIDES: POTENTIAL APPLICATIONS IN CENTRAL NERVOUS SYSTEM THERAPEUTICS

VEGF is expressed in central nervous system and its expression increases in hypoxia and in inflammatory brain disorders. A wealth of data suggests that VEGF may exert neuroprotective activities and promote neuroregeneration in disease status. Moreover, the risk of developing certain neurological disorders may be 2 dependent on dysfunction in the VEGF system. Therefore, a strong rationale does exist to suggest that VEGF-based therapeutics could be implemented in conditions such as stroke or amyotrophic lateral sclerosis and experimental data supporting this hypothesis have been obtained. However, the unfavorable pharmacokinetic profile of this growth factor, and concerns on its safety have limited the development of VEGF as a therapeutic tool for neurological disorders. In this review, we discuss why a new class of VEGF mimic peptides holds promises to become a safer, cheaper and more easily manageable tool for central nervous system therapeutics.


VEGF As A Tool For Neuroprotection And Neuroregeneration
A great number of excellent reviews have been published on VEGF structure and biological activity that can help understanding why and how VEGF could help in CNS disorders [5,6]. Here we will just quickly remember a few points that we believe essential for the issue of this review i.e. the design and potential applications of VEGF mimic peptides in neurological disorders.
VEGF is a 34-42 kDa homodimeric glycoprotein belonging to the protein family of the cystine-knot growth factors. The VEGF family is composed of several isoforms (Table 1), VEGF-A, VEGF-B, VEGF-C VEGF-D and PlGF that differ in their ability to bind to VEGF receptors. Most of the biological activity in CNS is exerted by VEGF-A and in the present review we will focus on this isoform.
At least six different VEGF-A isoforms are originated through the differential splicing of the vegf gene: VEGF 121 , VEGF 145 , VEGF 165 , VEGF 183 , VEGF 189 , and VEGF 206 . They differ in the C-terminal region where the binding determinants for heparin, encoded for by exons 6 and 7, are located ( Fig. 1).Consequentely, the different VEGF isoforms differ in their ability to bind heparin and in their mobility in the extracellular space. More specifically, VEGF 121 is freely diffusible because it does not bind heparan-sulfate whereas VEGF 189 strongly binds to this molecule and is, therefore, retained in the extracellular matrix and on cell membranes; the main VEGF isoform, VEGF 165, has intermediate properties [7][8][9]. Splicing isoforms with anti-angiogenic bioactivity have also described (VEGF 165b and VEGF-Ax) but their role and activity are still controversial [10][11][12][13].
The crystal structure of VEGF  has been determined [14,15]. This protein has a cystine knot structure similar to PDGF but, unlike this growth factor, VEGF has a α-elical N-terminus [14] (Fig. 2). VEGF monomers consist of a total of seven β-strand segments (β1 to β7) and two α-helical segments (α1 and α2) [15]. VEGF molecule has a central portion made by four antiparallel β sheets arranged in two pairs: β1 and β3, and β5 and β6. The cystine knot is made by two disulfide bonds forming a covalently linked ring structure between β3 and β7 plus a third disulfide bond that connects β1 and β4 penetrating the ring. Two monomers assemble in an antiparallel fashion to make a dimer. Two symmetrical disulfide bonds between Cys51 and Cys60 keep the monomers together. Dimeric conformation is stabilized by hydrofobic interactions among loop β1 to β3 and segment β5 to β6 of one monomer, and the N-terminal α helix of the other monomer. The two poles of the dimer contain the cystine knot regions.
Three different VEGF receptors, VEGFR-1 (Flt-1), VEGFR-2 (KDR, Flk-1), and VEGFR-3 (Flt-4) have been cloned. VEGFA binds to VEGFR-1 and VEGFR-2, VEGF-B and PIGF to VEGFR-1 whereas VEGF-C and VEGF-D bind to VEGFR-2 and VEGFR-3 (Table 1). Neuropilin receptors NRP1 and NRP2 act as coreceptors for VEGFR-1 and VEGFR-2 and can bind VEGF-A, VEGF-B and PlGF. The three different receptor types mediate different VEGF biological actions being VEGFR-2 responsible for the mitogenic, angiogenic and permeability-enhancing effects of VEGF and VEGFR-3 involved in lymphoangiogenesis [16]. The role of VEGFR-1 is more controversial because it    [233][234][235][236][237][238][239][240][241][242][243][244][245][246][247][248][249][250][251] is believed to act mainly as a VEGF decoy although it can also cooperate with VEGFR-2 in mediating some of VEGF biological effects [17]. VEGFRs are composed of seven immunoglobulin-like extracellular domains joined to the intracellular kinase domains by a short transmembrane region. Mutational and crystallographic studies suggest that the two poles of the dimer represent the portion of the VEGF molecule taking part to receptor binding. More specifically, VEGFR binding to VEGFR-2 receptors involves two identical symmetrical sites each composed by residues of β2 (Ile-46) and β5 (Gln-79, Ile-83, Pro-84, Lys-85) sheets, of the α2-β2 loop (Ile-43) of one monomer, and by residues of the N-terminal helix (Phe-17) and of the β3-β4 loop (Glu-64) of the other [14]. Upon VEGF binding, VEGF receptors dimerize and become auto-phosphorylated at the intracellular kinase domain on specific Tyr residues that act as docking sites for adaptor molecules [18,19]. The assembly of this complex ends with the activation of multiple transduction pathways that include the MEK-MAPK cascade via PLCγ, PI3K-Akt pathway via Shb, Src via TSAd and p38 [20]. An important player in signal transduction through VEGFR-2 is eNOS that is activated through PLC-γ and PI3K-dependent phosphorylation [21]. After activation, VEGFR-2 is internalized and recycled. VEGF and its receptors are expressed both in prenatal and in postnatal neurons [5,22,23] and their expression greatly increases after injury, for instance after stroke (see the next section). Distinct roles have been identified for VEGF both in the development of central nervous system and in the adult brain but their analysis goes beyond the scope of this paper and the interested reader can find details on these issues in several excellent recent reviews [5,24,25]. Here we will briefly review the biological effects that make VEGF valuable as a potential tool for CNS therapeutics: neuroprotection and neuroregeneration.
As we will review with more details in the next section, VEGF exerts neuroprotective activity in a number of experimental models of neurodegenerative diseases both in vitro ed in vivo such as, for instance, the exposure to amyloid peptides, to glutamate or the combined deprivation of oxygen and glucose [26][27][28][29][30]. Most of this neuroprotective activity can be explained by the activation via VEGF-R2 of the prosurvival cascade involving PI3-K and Akt [27,28,31]. An additional mechanism that could have a role in VEGF-mediated neuroprotection is the inhibition of voltage gated K + channel activity in neurons. It has been demonstrated, indeed, that the efflux of K + is an early event in neuronal death and that its pharmacological blockade can spare neurons from dying [32,33]. Xu et al. [34] showed that VEGF suppresses K + currents in acutely dissociated neurons from rat hippocampal slices. This effect was very fast and rapidly reversible. Because confocal imaging of the acutely dissociated neurons showed the presence of VEGFR-1 but not of VEGFR-2 it was suggested that this effect was entirely dependent on the activation of VEGFR-1 receptors [34]. VEGF effect on K + currents may depend on the phosphorylation of key tyrosine residues on specific K + channel subunits. Specifically, an increase in tyrosine phosphorylation of K V 1.2 subunits has been observed upon VEGF exposure in SH-SY5Y cells after hypoxia and glucose deprivation [35]. VEGF-induced K V 1.2 phosphorylation may involve a PI3-K dependent cascade because it is abolished by the PI3-K inhibitor wortmannin [35]. Interestingly, mutant K V 1.2 channels lacking key tyrosine residues have a higher activity than wild type channels suggesting that tyrosines have a role in the negative control of this channel type [36]. Importantly, the enhanced phosphorylation of K V 1.2 channels upon VEGF treatment has been observed also ex vivo in brain slices obtained from rats in which experimental brain ischemia was induced by the transient occlusion of the middle cerebral artery [37]. Other ion channels are modulated by VEGF in a way that could explain part of the neuroprotection elicited by this growth factor. It has been demonstrated, indeed, that in cultured neurons, VEGF reduces the influx of Ca 2+ evoked by KCl or glutamate through the inhibition of high voltage activated Ca 2+ channels [38]. By acting through PLC-γ VEGF also induces a slow and small increase in intracellular Ca 2+ dependent on the release of this ion from intracellular stores; this effect could have a role both in the physiological effects of this growth factor, such as the promotion of neuronal plasticity, and in neuroprotection [39]. Finally, VEGF shifts leftward the activation curve of voltage gated Na + channels hence causing a decrease in their activity and a reduction in neuronal excitability [40].
VEGF beneficial effects in neurological diseases also depend on the ability of this growth factor to promote neuroregeneration. VEGF is indeed the key factor of the so called neurovascular niche. This term designes a set of different cell types including neurons, their precursors neuroblasts, astrocytes, endothelial cells and pericytes that cooperate in a highly coordinate way to mediate adult neurogenesis and to repair tissue damage in the CNS [41,42]. Not only it becomes activated and plays a crucial role in repair after ischemic lesions [43][44][45][46][47] but it also involved in the reparative response in inflammatory CNS diseases [48,49]. Angiogenesis has a central role in the activation of the neurovascular niche. Both ischemic and inflammatory stimuli trigger the formation of new blood vessels thorough the sprouting of preexisting blood vessels and the homing of circulating endothelial precursor cells and VEGF is the prime responsible of this response [50,51]. In addition VEGF acts as a chemoattractant for sub-ventricular zone neuroblasts and promotes their differentiation [52]. Moreover, VEGF induces endothelial cells to release other chemoattractant proteins such as Ang-2 and GROα [53]. This process is amplified by the release of more VEGF and of matrix metalloproteinases 2 and 9, BDNF and stromal derived factor 1 by newly formed endothelial cells. By their side, recruited neuroblasts further promote the entire process by releasing pro-angiogenic factors like angiopoietin 2 and FGF [50,51].
The data that we just reviewed suggest that VEGF may exert neuroprotective and neuroregenerative effects in CNS. We also mentioned that VEGF synthesis increases after injury and this raises the possibility that this growth factor could be part of an endogenous adaptive system whose activation aims to reduce CNS damage. The major stimulus for the transcription of the VEGF gene in CNS is hypoxia acting through HIF-1 [52,53]. As such, a substantial increase in VEGF synthesis and release will occur in CNS diseases in which tissue perfusion is impaired [54,55]. Moreover, it is expected that in patients with an impaired response of VEGF to hypoxia because, for instance, of polymorphisms in the regulatory regions of the vegf gene, hypoxia will induce more serious damage than in normal subjects. We will review some examples of this variable susceptibility in the next section. It is very important to metion here that VEGF increases not only in ischemic brain disorders but also in the presence of inflammation that as we mentioned before, is a strong stimulus for the activation of the neurovascular niche. This may explain why this growth factor is actually involved in conditions in which tissue hypoperfusion is not the main causative factor. Several mechanisms that have been extensively reviewed by Szade et al. [56], can explain how the synthesis of VEGF and, more in general, the whole process of angiogenesis are activated by inflammation. Briefly, a certain degree of tissue hypoxia often occurs in inflamed tissues, HIF-1 may be activated in an hypoxia-indepent way by NF-kB in response for instance to free oxygen radicals and, possibly, by NO [56].
It has also to be considered that inflammatory cells such as monocytes that penetrate from the blood into the inflammatory focus produce and release VEGF hence contributing to increase the local concentrations of this growth factor [56].
Before moving to the next section where we will examine more in detail the mechanisms of VEGF-mediated neuroprotection in selected neurodegenerative diseases, we would like to mention that several compounds with proven neuroprotective properties such as polyphenols, may affect VEGF release and activity. Protective effects of polyphenols have been reported in different neurodegenerative diseases including Parkinson's and Alzheimer's disease, and in stroke [57][58][59]. Multiple mechanisms seem to converge in conferring neuroprotective activity to these natural compounds including the ability to chelate iron, to exert antioxidant and anti-inflammatory effects, to interfere with amyloid, tau and alpha-synuclein aggregation, to prevent of apoptosis and mitochondrial dysfunction, to promote vasodilation in brain vessels and to modulate sirtuin activity [58,[60][61][62][63]. The effect of polyphenols on VEGF and, more extensively, on angiogenesis is very complex. Indeed, although there is a great interest on these compounds as tools for cancer chemoprevention [64] because they exert antiangiogenic effects in cancer by inhibiting HIF-1, by blocking metalloproteinases and by activating AMPK in endothelial cells [65][66][67][68][69], evidence for polyphenol-induced proangiogenic effects has also been reported. For instance, it has been shown that quercetin activates an angiogenic pathway by inhibiting HIF-prolyl hydroxylase, by stabilizing HIF-1 alpha and by promoting its nuclear localisation [70] and that (-)-epicatechin-3-gallate activates HIF-1α and increases VEGF expression in T47D human breast carcinoma cells [71] and increases HIF-1α protein level and activates HIF-1-mediated gene transcription even under normoxic conditions in human prostate cancer cells [72]. These effects together with the proven ability to reduce the oxidative stress that is a major factor in causing cell demise in neurodegeneration [73][74][75], may account for polyphenol-induced neuroprotection.

In Which neurological disorders could VEGF be helpful?
The role of endogenous VEGF and the possible beneficial effects of the administration of exogenous VEGF have been evaluated in many neurological disorders leading to the proposal that a neurovascular approach should be implemented for the treatment of neurodegenerative disease [76]. Excellent reviews on the possible therapeutic applications of VEGF have been recently published [5,6,51,[77][78][79]. We will just remember some issues on a few neurological disorders that we believe salient in the perspective of the present paper, whereas the interested reader may find additional information on these diseases and on others that will be not discussed here such as neurotrauma or demyelinating diseases, in the review papers that we cited before.

Stroke
After stroke, oxygen tension in the ischemic brain tissue dramatically drops to very low levels and remains low until reperfusion is established. This represents a strong signal for the activation of HIF-1dependent transcription. Consequently, an increase in VEGF transcription and protein expression is expected to occur as it was observed in different experimental stroke models [80]. Kovács et al. [81] evaluated VEGF expression in the brain after the permanent occlusion of the middle cerebral artery (pMCAO) in the rat. They found a significant increase in both VEGF 121 and VEGF 165 isoforms that was already detectable 18 h after the surgical procedure. A different time course of VEGF expression was observed in macrophages and neurons, where it peaked after 18 h to rapidly decrease thereafter, and in glial cells where it only peaked 5 days after ischemia. Interestingly, the highest levels of VEGF were found in macrophages. A late expression of VEGF was also observed in endothelial cells where it progressively increased with reparative angiogenesis to peak at about 1 week from surgery. A high expression of VEGFR-1 was observed in endothelial but not in other cell types. After pMCAO, Marti et al. [55] observed in the ischemic penumbra an increase not only in VEGF, and VEGFR-1 but also in VEGFR-2 expression. The evidence that the increase in VEGF was accompanied by a parallel increase in its receptors suggests that, not only, there is more VEGF in the hypoxic brain but tissue responsiveness to this growth factor is potentially also enhanced. VEGF expression also increases after the transient occlusion of middle cerebral artery (tMCAO), an experimental model of brain ischemia that resembles more closely than pMCAO to human stroke because it also reproduces the reperfusion phase. Hayashi et al. [82] observed a very rapid increase in the expression of both VEGF 121 and VEGF 165 after tMCAO that was already evident 1 h after the surgical procedure, peaked after 3 h and decreased after 1 d. Interestingly, VEGF immunoreactivity was also strongly increased. After bilateral clamping of the carotids, Lee et al. [83] observed a transient increase in VEGF mRNA that occurred earlier in hippocampal neurons starting after 12 h and decaying after 1 d, and later in hippocampal, striatal and cortical astrocytes. An increase in VEGF expression was also observed in rats after the creation carotid artery-jugular vein fistula [65]. VEGF 164 was the VEGF isoform preferentially expressed after this procedure whereas the levels of VEGF 188 , and VEGF 120 were very low; VEGF 164 expression was detectable 24 h postoperatively, peaked by 7 days, decreased by 21 days, and returned to basal levels by 90 days [84]. Also in this experimental model of brain hypoxia, the increase in VEGF expression was paralleled by an increase in the expression of VEGFR-1 whereas VEGFR-2 levels did not change [84].
To establish whether the observed increase in VEGF expression after stroke could have any protective role against the ischemic damage, the effect of this growth factor was explored in experimental models in vitro of neuronal ischemia. More specifically, VEGF protected HN33, an immortalized hippocampal neuronal cell line, from the death induced by the combined deprivation of oxygen and glucose [27]. Because these cells only express VEGFR-2 and NP1 but not VEGFR-1, and VEGF effect was not replicated by the neuropilin-1 ligand placental growth factor-2, it was concluded that neuroprotection was exerted through the activation of VEGFR-2. Pharmacological experiments provided arguments to suggest that the main transduction pathway involved was PI3/Akt [27]. VEGF was also neuroprotective against combined oxygen and glucose deprivation in primary cultures of cortical neurons [26] and protected hippocampal neuronal cultures from the excitotoxic damage induced by either glutamate or N-methyl-D-aspartate [29,30]. Moreover, exogenously added VEGF reduced cell death and prevented the impairment of LTP and preserved spontaneous synaptic potentials and excitability in rat CA1 neurons cultured in vitro and exposed to combined oxygen and glucose deprivation [66]. Importantly, the VEGFR-2 inhibitor SU5416 abolished VEGF-induced neuroprotection suggesting that this class of receptors were involved [85].
As described in the previous section, a wealth of experimental data accumulated in the decade after the publication of this first evidence led to the identification of VEGF as the key activating factor of the neurovascular niche and provided a clear rationale for the neuroprotective effect of this growth factor in stroke. Consistent with the hypothesis that the increase in endogenous VEGF and in its receptors in the ischemic brain could drive a neuroprotective and reparative response is the evidence that VEGF inhibition worsens the ischemic damage. Such a worsening was observed both after the intracerebroventricular administration of antisense oligonucleotides against VEGF-A [86], of anti-VEGF antibodies [87] or after the intraperitoneal administration of the VEGFR-2 inhibitor SU146 [88].
If endogenous VEGF may protect against ischemic damage, strategies that increases the synthesis of VEGF or of its receptors should be able to reduce brain damage after stroke. Wei and her collaborators reported an elegant demonstration of this hypothesis. They first showed that whiskers stimulation enhances angiogenesis in the barrel cortex following focal ischemia in mice [89] and later demonstrated that it induces the expression of VEGFR in neuroblasts that migrate towards the ischemic lesion where they exert neuroprotective and neuroregenerative effects [90]. Importantly, these effects were VEGFmediated as indicated by their abrogation by the VEGFR-2 inhibitor SU5416 [71]. The evidence that VEGF dysfunction could be a risk factor for stroke in humans is limited. Two metanalyses on the associations of VEGF-A gene polymorphisms have been recently published [91,92]. Based on the findings of ten different studies Wu et al. [92] concluded that the +936 C > T (rs3025039) polymorphism could represent a risk factor for stroke, especially in Asians, whereas the polymorphism −1154 G > A (rs1570360) is not associated with this disease. Qiu et al. [91] evaluated five studies with a total of 2904 cases and 2824 controls and concluded that the +1192 C > T (rs2305948) and +1719 T > A (rs1870377), but not the −604T > C (rs2071559) VEGFR2 polymorphisms could be potential risk factors for stroke.
The evidence reviewed above that VEGF expression increases after stroke and that exogenous VEGF protects neuronal cells from ischemic death in vitro, prompted studies aiming to establish whether the pharmacological administration of exogenous VEGF could reduce brain damage and improve the outcome in experimental stroke. The scenario emerging from the results of the studies performed to address this question was, however, more complicated than expected. Briefly, while neuroprotection can be elicited with VEGF, it only occurs with the exact timing of its administration and the right choice of its dosage; otherwise, VEGF may actually be detrimental and worsen the ischemic damage. The main reason why VEGF can be dangerous in stroke is that it can increase blood vessel permeability ultimately worsening ischemia-induced brain edema. We will come back to this important point in Section 4 because it represents a major obstacle for the implementation of VEGF therapeutics in CNS disorders. However, now we will quickly review the main experimental evidence that led to realize the risks of VEGF therapy in stroke, starting from the seminal paper of Zhang et al. [93]. Because brain edema occurs early after stroke whereas reparative angiogenesis is a later event, they supposed that the risk of increasing blood vessel permeability could be restricted to a short time window after stroke, whereas the proangiogenic effect should last longer. Therefore, VEGF should be dangerous only if given immediately after stroke. To test this hypothesis they compared the effect of the early (1 h) and late (48 h) postischemic administration of VEGF. The main finding of the study was that the late VEGF administration reduced ischemic brain damage and improved the neurological recovery in a rat model of focal cerebral embolic ischemia whereas early administration worsened the outcome and increased edema and hemorrhages. However, the idea that the beneficial effect of VEGF in stroke are only due to the promotion of new vessel formation appears quite limiting. Indeed, Sun et al. [94] demonstrated that when intracerebroventricularly injected 1 or 3 days after stroke, VEGF also promotes neuronal survival in the penumbra and neurogenesis in the subventricular zone. The evidence that VEGF has neuroprotective effects in brain ischemia in vivo raises the important question whether delaying too much to avoid its permeabilizing effects could actually waste part of its beneficial neuroprotection. Data reported by Harrigan et al. [95] suggest that not only the timing but also the dosage of VEGF injection could determine whether BBB permeabilization will be predominant over neuroprotection. They demonstrated, indeed, that when VEGF was intraventricularly injected at a concentration of 5 μg/ml it promoted neoangiogenesis with no BBB permeabilization in normal rats [95], and reduced the ischemic damage after tMCAO with no worsening in brain edema [96]. Conversely, at the higher dose of 25 μg/ml, VEGF induced ventriculomegaly in normal rats presumably by increasing leakage through BBB [95] The dose dependence of VEGF effect in stroke was also confirmed by Manoonkitiwongsa et al. [97] who infused for one week into the internal carotid artery of a group of rats, three different doses of this growth factor (2, 8 and 60 μg) after a 4 h occlusion of MCAO and found neuroprotection only for the low and intermediate doses.
In conclusion, available evidence suggests that VEGF could be beneficial in stroke but only if given not too early and at not too high doses. These conditions seriously limit the implementation of VEGF therapeutics for stroke in clinical practice.

Amyotrophic lateral sclerosis
Amyotrophic lateral sclerosis (ALS) is one of the neurological diseases in which the involvement of VEGF has been more strongly demonstrated. The first evidence linking ALS and VEGF was reported by the group of Carmeliet that studied the consequences of the genetic ablation of HIF-responsive elements in the VEGF promoter in transgenic mice [98]. These mice (Vegf δ/δ mice) had very low viability and only few of them survived till adulthood; those surviving more than 5 months developed symptoms suggestive of a motor neuron disease with severe muscle weakness and atrophy. Pathological analysis showed denervation-like changes in muscles and prominent signs of motor neuron degeneration in ventral horn of the spinal cord. A decrease in blood vessel density was observed in spinal cord and laser Doppler flowmetry showed a decreased baseline neural blood flow. These data suggest that motor neuron degeneration in Vegf δ/δ mice could have been caused by spinal cord hypoperfusion. However, the evidence that primary motor neurons cultures from these animals showed a reduced survival when compared with those from their wild type littermates, and that this low in vitro viability was rescued by exogenous VEGF 165 clearly indicates that the loss of VEGF neuroprotective effect on neurons could also be implicated. VEGFR-2 and NRP1 receptors are probably involved as suggested by the inefficacy of VEGF 121 to restore to normal motor neuron viability in vitro considering that this VEGF isoform does not bind to NRP1. Moreover, anti-NRP1 and anti-VEGFR-2 antibodies worsened the survival deficit in motor neurons from Vegf δ/δ mice.
The finding that the loss of hypoxic VEGF regulation induced an ALS-like phenotype in Vegf δ/δ mice raised the important question of whether VEGF was implicated just in this very unusual ALS model or it could have a more general role in this disease. An important argument in support of the second possibility came from the study of G93A SOD1 mice. This is a very popular animal model of ALS that recapitulates the best characterized familial form of the human disease that is caused by inactivating mutations in superoxide dysmutase (SOD) [99,100]. A first evidence of the possible involvement of VEGF in this animal model was the finding of lower levels of VEGF in their spinal cord as compared with controls [101,102]. Moreover, G93A-SOD1 mice showed a spinal cord motor neuron loss after experimental spinal cord ischemia that was more severe than in wild type mice and that was rescued by the intraperitoneal administration of VEGF 165 (2 μg) [101]. Evidence has been reported of blood flow-metabolism uncoupling in the spinal cord of these mice even before the appearance of neurological symptoms [103] and, importantly, a significant increase in HIF immunoreactivity in the spinal cord was observed as the disease progressed hence confirming that this structure is chronically hypoxic. This increase in HIF-1 was not paralleled by a similar increase in the level of the target gene VEGF [104] suggesting that somehow the SOD1 G93A mutation affected VEGF synthesis or release. A mechanism that has been proposed to explain this effect of mutant SOD is that it could affect VEGF mRNA stability. It has been shown indeed that different from wild type SOD1, mutant G93A-SOD 1 has a high affinity for adenine/ uridine rich elements (ARE) in the 3'-untranslated region (UTR) of the VEGF mRNA [105]. ARE sequences normally regulate mRNA stability through the binding to specific RNA stabilizing proteins such as members of the embryonic lethal abnormal visual (ELAV). What is supposed to happen in G93A-SOD 1 mice is that mutant SOD prevents the binding of these regulatory proteins hence reducing VEGF mRNA stability and ultimately causing a decrease in VEGF synthesis [106].
An additional evidence of the involvement of VEGF in ALS came from a model in vitro of this disease based on the mouse NSC-34 motor neuron cell line. When cerebrospinal fluid (CSF) from patients with sporadic ALS is added to the culture medium, the expression of VEGFR-2 is downregulated and these cells degenerate and die [107]. This process can be prevented by the addition of exogenous VEGF.
Collectively these data suggest that a defect either in VEGF or in VEGFRs could be responsible for some forms of ALS by causing perfusion deficits and neuronal degeneration in the ventral horn of the spinal cord. Therefore, it is expected that exogenous VEGF could be beneficial in ALS by counteracting these deficits. Consistent with this hypothesis beneficial effects have been observed in G93A-SOD1 transgenic mice upon treatment either with intracerebroventricularly administered VEGF 165 [108] or with VEGF gene carrying lentiviruses retrogradly transported to motoneurons [109].
PI3-K/Akt activation is probably the primary mechanism responsible for VEGF neuroprotection in ALS. For instance, it was demonstrated that in rat primary motor neuron cultures transfected with an adenoviral vector encoding a mutant SOD1, G93A-SOD1, VEGF attenuated G93A-SOD1-induced caspase-3 cleavage and DNA fragmentation by activating PI3K/Akt signaling via VEGFR2 [110]. Additional mechanisms have been proposed to explain VEGF neuroprotection in ALS. For instance, evidence has been reported that VEGF could decrease the Ca 2+ permeability of AMPA receptors by promoting the expression of GluR2 subunits [111]. This could represent a neuroprotective mechanism in ALS considering that Ca 2+ influx through AMPA receptors induces motor neuron degeneration and causes an ALS-like condition in experimental animals [112][113][114][115].
The evidence that VEGF is implicated in the genesis of ALS in several animal models raised the question of whether the VEGF system could be impaired in human ALS too. Low levels of VEGF in CSF were observed early in the course of ALS [116]. In addition, case reports of ALS developing in patients taking anti-VEGF drugs have been reported [117]. Several studies evaluated the hypothesis that defects in hypoxic VEGF regulation similar to those induced in the Vegf δ/δ mice by genetic manipulation in vitro do also occur spontaneously in human ALS because of polymorphisms of the VEGF gene promoter. Lambrechts et al. [101] performed a metanalysis on over 900 individuals from Sweden and over 1000 individuals from Belgium and England to investigate the possible association of three common vegf gene polymorphisms (22578 C > A (rs699947), 1154 G > A (rs1570360), 2634 G > C (rs2010963)) with ALS. They found that people with two haplotypes in the promoter region of VEGF (-2,578 A/-1,154 A/-634 G and -2,578 A/-1,154 G/-634 G) had a 1.8 times greater risk of developing ALS than control subjects. These findings, however, were not confirmed in a later larger metanalysis that provided evidence only of a weak association of the −2578AA genotype with ALS risk in males [118]. Other studies gave conflicting results as some of them reported a possible association between VEGF promoter gene polymorphisms and ALS [119,120] and others did not [121][122][123][124][125]. The search for mutations in the promoter region of VEGFR2 in ALS patients also gave negative results [126].
In conclusion, whereas the evidence that a deficit in VEGF could have a role in several ALS animal models that could also benefit of the treatment with this growth factor, the definite proof of VEGF involvement in human ALS is still missing.

Alzheimer disease
Even though Alzheimer's disease (AD) has been traditionally considered as a disease that primarily involves neurons, it is now clear that it also has an important vascular component in the form of an ischemic damage that contributes to potentiate and extend the primary neuronal degeneration [127,128]. Alterations in cerebral vascular architecture have been observed in animal models of this disease [129] and perfusion deficits that begin in posterior cingulate and precuneus to later become more widespread and extend to the hippocampus, have been documented in AD patients [130,131]. The basic mechanism of this decrease in perfusion is a specific form of amyloid angiopathy with small sclerotic vessels that accumulate amyloidogenic peptides in their wall [132][133][134]. Considering that hypoperfusion and the ensuing chronic hypoxia are strong stimuli for the synthesis of VEGF it is not surprising that this growth factor could be somehow involved in AD pathophysiology. Early studies reported high VEGF levels in AD CSF and brain tissue [135][136][137]. For instance, Tarkowski et al. [136] found high levels of this peptide in CSF of AD patients and in patients with vascular dementia. Immunohistochemical analysis of autoptic samples from AD patients showed higher VEGF levels than in age matched controls, mainly localized in clusters of reactive astrocytes in the neocortex [135]. High VEGF levels were also found in brain microvessels [137] and in the white matter [138] of AD patients. Importantly, evidence was reported for a hypoxic drive of VEGF hyperproduction in AD white matter because the level of this growth factor strongly correlated with the ratio of the myelin-associated glycoprotein that is highly hypoxia-sensitive, to proteolipid protein 1 that is hypoxia resistant [138]. Later investigations, however, did not confirm these early data and reported, instead, decreased VEGF levels making the issue of how do VEGF concentration change in AD still somehow controversial [139,140]. The early findings of lower CSF VEGF concentrations in AD patients as compared with normal subjects also prompted larger scale studies to investigate the hypothesis that this growth factor could represent a potential biomarker for AD. These investigations yielded conflicting results. Plasma VEGF concentration was higher than in controls in AD patients in the series of Yu et al. [141] and lower in that of Mateo et al. [142]. Lower than normal VEGF levels in CSF were also found in studies performed with multianalyte platforms for the simultaneous detection of multiple candidate biomarkers [143,144]. In the light of accumulating evidence that VEGF could represent a prognostic marker of disease evolution, more recently the interest moved from differences in CSF VEGF levels between normal subjects and AD patients to differences among different patients. Hohman et al. [145] performed a longitudinal study in 130 patients with mild cognitive impairment, in 59 with AD and in 90 agematched normal controls to evaluate whether CSF VEGF levels could predict the evolution of the disease. They found that higher VEGF levels are consistently associated with a slower progression of hippocampal atrophy and of cognitive decline. Interestingly, VEGF interacted both with tau and αβ1-42 in predicting disease progression. As a whole, the data collected in humans seem to suggest a scenario in which AD microangiopathy causes a hypoxic drive for VEGF synthesis and release; the more effective will be the VEGF response to this stimulus the less aggressive will be disease evolution. There have been also reports on the association of specific VEGF polymorphisms and the risk of developing AD [146,147] although they were not confirmed in other studies leaving this question still open [148]. A recent metanalysis by Liu et al. [149] showed that the risk is related to ApoE status with the −22578C > A (rs699947) polymorphism associated to higher and the −1154 G > A (rs1570360) polymorphism to a lower risk of developing AD only in patients with APOE є4 negative status. Conversely, no association between the risk of AD and VEGF promoter polymorphisms was found in another recent metanalysis [150].
Whereas the question of how VEGF concentrations change in AD is still open, consistent evidence suggests that this growth factor does not work as effectively in the AD as in the normal brain. Indeed, AD patients seem to be resistant both to the VEGF angiogenic and vasodilatatory effect. A prevalent hypothesis is that αβ oligomers could directly induce a status of cell unresponsiveness to VEGF. It has been demonstrated, indeed, that they cause a reduction in cerebral blood flow [151] and impair angiogenesis by cerebral endothelial cells in vivo and in vitro [152,153]. These effects could be the consequence of a direct antagonism of VEGF by αβ oligomers on endothelial VEGF receptors. Patel et al. [154] demonstrated that αβ 1−42 inhibits the phosphorylation of VEGFR-2 and Akt-1 in HUVEC cells and that this effect may be reverted by excess VEGF. Moreover, cell-free immunosorbent binding assay experiments showed that αβ 1−42 oligomers prevent VEGF binding to VEGFR2 because they bind to the extracellular region of these receptors [154]. The presence of VEGF resistance in AD and the evidence that it can be counteracted by excess VEGF suggests that this growth factor could exert beneficial effects in this disease. Proofs of concept of this hypothesis have been obtained in AD animal models in which an improvement in memory and cognition can be obtained by increasing VEGF levels in the brain through diverse delivery strategies such as the intraperitoneal injection of VEGF [155], transplantation of VEGF overexpressing bone marrow mesenchymal stem cells in the hippocampus [156], or the neuron specific hyperexpression of VEGF 165 [157].

Parkinson's disease
VEGF improves the survival of mesencephalic neurons cultured in vitro [158] and of dopaminergic neurons in organotypic cultures of the mesencephalon [159]. This evidence prompted to investigate its possible protective effect in different in vitro and in vivo Parkinson's disease (PD) models including the exposure to 6-OH-dopamine or to MPP + . VEGF was protective against both of these neurotoxins: it increased the survival of primary mesencephalic neurons treated in vitro with these compounds, reduced dopaminergic neuron loss and promoted reactive astrogliosis in vivo and improved neurological symptoms including amphetamine-induced rotational behavior [160][161][162]. The beneficial effect of VEGF was, however, observed only at lowmoderate doses whereas at higher doses it was lost. This was demonstrated both in vitro by exposing cultured neurons to different concentrations VEGF (10 or 100 ng/ml) and in vivo in rats implanted with small or big microcapsules containing VEGF-secreting BHK cells. The lower neuroprotective potential in vivo of high VEGF doses could be due to blood vessel permeabilization and edema [161].
Other more physiological ways of increasing brain VEGF have been evaluated in experimental models of PD. In particular, based on the observation that physical exercise can increase brain concentration of VEGF [163,164], the possible beneficial effect of exercise has been evaluated in rats with experimental PD. In rats the density of microvessels in the substantia nigra decreases with ageing. Villar-Cheda et al. [165] demonstrated that this decrease in vascularization is associated with a parallel decrease in nigral VEGF mRNA levels. Intriguingly, both the deficit in vessel density and in VEGF concentration could be partially corrected by physical exercise. Even though these findings are intriguing, a scenario in which physical exercise could rescue dopaminergic cell from death in human PD appears unlikely mainly because in humans affected with PD, microvessel density and VEGF levels in substantia nigra are actually increased and not decreased as in aged rats [166,167].
A role has been suggested for VEGF also as a mediator of the therapeutic effect of deep brain stimulation of the subthalamic nucleus [168].
Cui et al. [169] investigated the mechanism responsible for VEGFinduced neuroprotection in MPP + -exposed rat cerebellar granule cells. They showed that VEGF reversed the inhibition of PI3-K/Akt pathway, but further enhanced the activation of ERK induced by MPP(+).
Also in PD the role of VEGF and its receptors as a biomarker begins to be considered of potential value. Janelidze et al. [170] measured these molecules alongside with other angiogenic proteins in CSF of 100 patients with PD and 38 elderly controls. They found higher levels of VEGF and its receptors in PD patients with dementia than in controls. Importantly, high levels of this and other angiogenic factors strongly correlated with the presence of BBB permeabilization as evaluated with the CSF/plasma albumin ratio, with white matter lesions, with cerebral microbleeds and with gait difficulties. The high VEGF levels in PD can be interpreted either as the compensatory response to the repeated brain ischemic insults caused by the severe orthostatic hypothension that may occur in this disease or as the consequence of cytokineinduced release from astrocytes and microglia [171]. VEGF-induced BBB permeabilization could have a role in PD progression by damaging dopaminergic neurons [172]. Among PD patients the highest VEGF levels have been observed in patients with L-dopa induced dyskinesias that were reduced by candesartan [173]. Consistent with the finding of high VEGF levels in the CSF of PD patients, high microvessel density and VEGF expression have been observed in the mesencephalon of patients affected with this disease [166,167].
The graft of mesencephalic neurons in the striatum held great promises for the treatment of PD. However, low survival of grafted cells and poor vascularization of the graft may represent a substantial obstacle for the success of this procedure. Based on the evidence that VEGF has both proangiogenic and neuroprotective activity Pitzer et al. [158] investigated the possible use of this growth factor to improve mesencephalic graft survival in the rat. They found that the direct injection of VEGF in the striatum promoted the formation of new vessels and the neovascularization of the graft. However, surprisingly, the survival of grafted dopaminergic neurons was not improved respect to control.

Epilepsy
One of the first evidence that VEGF could have a role in seizures came from a study not directly related to epilepsy. Newton et al. [174] performed, indeed, a gene microarray analysis to identify the genes whose expression was modulated in the rat brain after an electroconvulsive shock (ECS) with the aim of obtaining information on the mechanism responsible for the beneficial effects of this procedure in psychiatric disorders such as major depression. They found that the expression of genes encoding for neuroprotective and angiogenic factors was strongly increased, and, of note, that VEGF was one of the most sensitive among them. Interestingly, the strongest VEGF induction was observed in the hippocampus and in choroid plexus. This finding was confirmed also at the protein level by Nicoletti et al. [175] in the classical pilocarpine experimental model of temporal lobe epilepsy. The well known neuroprotective effect of this growth factor suggested that its induction could represent a mechanism activated to protect hippocampal neurons from the seizure-induced damage that has a major role in causing persistent epilepsy after status epilepticus. In light of this hypothesis, exogenous VEGF would be expected to further decrease pilocarpine-induced hippocampal damage as was actually observed in rats implanted with VEGF-containing micropumps that released this growth factor in the dentate hilus [175]. As in the other neurodegenerative diseases considered in the previous paragraphs, also in epilepsy most of the neuroprotective effects of VEGF seem to be mediated by VEGFR-2 and the consequent activation of PI3K/Akt cascade. An elegant demonstration of the role of VEGFR-2 was provided by Nikitidou et al. [176] who showed that transgenic mice overexpressing these receptors develop less serious epilepsy after electrical kindling, a widely used experimental method to induce epilepsy that is based on the repeated electrical stimulation of specific brain areas such as the amygdala. Importantly, this protective effect occurred in the absence of any change in blood vessel or glia. Additional beneficial effects of VEGF in epilepsy depend on effects directly exerted by this growth factor on neuronal excitability. McCloskey et al. [177] showed that in rat hippocampal slices VEGF reduced the amplitude of responses elicited by Schaffer collateral stimulation. Although this growth factor did not affect the epileptiform activity evoked by GABA receptor blockade with bicuculline in brain slices form normal animals, it significantly reduced the spontaneoud discharges that occur in hippocampal brain slices from pilocarpinetreated animals, an experimental model of temporal lobe epilepsy [177]. Moreover, exogenous VEGF reduced both the ictal and interictal discharges evoked by the potassium channel blocker 4-aminopyridine or by the removal of Mg 2+ from the extracellular fluid in mouse hippocampal slices [178].
The beneficial effects played by VEGF on neurons could be partially counteracted by its vascular effects. As mentioned in Section 4 exogenous VEGF induces blood vessel permeabilization and increases the permeability of the blood brain barrier as well. This could be highly unwanted in epilepsy also considering that, according to a prevalent pathogenetic model, the loss of blood brain barrier tightness could be one of the triggering factor of epilepsy [179,180]. In this perspective, Morin-Brureau et al. [181] investigated whether also the endogenous VEGF that is released during seizures may increase BBB permeability. To this aim they studied the blood vessels of organotypic hippocampal cultures that were exposed to 25 μM kainate to induce epileptic events in vitro and looked to blood vessel density and to the expression of BBB markers such as zonulin-1, claudin and occludin. They found that seizure induction by kainate causes an increase in vascular density and decreased and disorganized the expression of zonulin-1.
In conclusion, whereas VEGF seems to be beneficial in epilepsy it could also worsen this disease by inducing BBB permeabilization.

Pharmacokinetics issues complicating the implementation of VEGF therapeutics and possible solutions
The data that we quickly reviewed in the previous section provide arguments to suggest that VEGF could have potential therapeutic applications for several neurological disorders. However, VEGF pharmacokinetics hinders the implementation of its use in clinical practice. Some of these problems appeared already evident when VEGF was considered as a potential drug for therapeutic angiogenesis in cardiology [1] and are similar to those that emerged when other growth factors were considered for CNS therapeutics [182].
As mentioned in the introduction, VEGF is a glycoprotein and, as such, a potentially unfavorable pharmacokinetic profile is expected. As a general rule, protein drugs have indeed a problematic pharmacokinetics: they have to be administered by parenteral route, may be degraded in plasma or tissues and may have significant difficulties in distributing to peripheral organs. Because of the interest in implementing its use as a pharmacological tool for cardiac revascularization after myocardial ischemia, the pharmacokinetics of exogenous recombinant VEGF has been studied with some detail. Eppler et al. [183] performed a pharmacokinetic noncompartmental analysis in patients receiving rhVEGF by intracoronary and intravenous injection. The main findings were that this growth factor has dose-dependent nonlinear pharmacokinetics. Terminal half life was 13.4 ± 8.3 min after the administration of a low dose (17 ng/kg/min) and 33.7 ± 13 min when a high dose (50 ng/kg/min) was given, whereas the steady state distribution volume was 1520 ± 900 and 960 ± 260 ml/kg, respectively. Clearance ranged around 39.1 ± 17 ml/min/kg after the low dose and decreased to 19.1 ± 5.7 after the high. These findings were explained assuming that VEGF removal from plasma depends on two different mechanisms: 1) the interaction with high affinity receptors that are believed to be mainly represented by endothelial VEGFR and 2) low affinity high capacity elimination systems that could be contributed for both by low affinity receptor sites or by classical elimination organs such as the kidney. According to these data, VEGF rapidly disappears from plasma because it is removed in a diffuse and "untargeted" way by endothelial cells through their receptors. A crucial role of clearance by endothelial cells was also demonstrated in modeling studies that addressed the pharmacokinetics of endogenously released VEGF, an issue of major practical interest for rational dose adjustment of anti-VEGF drugs. These models assume that VEGF is released by parenchimal cells in the interstitium. Then, it, in part, passively crosses the endothelium in a way predicatable on the basis of its Stokes-Einstein radius, i.e. of the radius of an ideal sphere that diffuses at the same rate, and, in part, is cleared by the endothelial cells through abluminal VEGFRs [184]. Once that it reaches the plasma it is cleared by peripheral endothelial cells through luminal VEGFR [185]. Therefore, the basic model involves two compartments, blood and tissues, separated by an endothelial barrier that bilaterally operates a significant clearance. Interestingly, according to this model, changes in tissue permeability do not modify the concentration of tissue VEGF but can significantly affect the concentration in plasma of this growth factor [184]. To describe VEGF pharmacokinetics in patients with solid tumors, the model was further complicated introducing a third high permeability compartment made by the tumor [186]. An additional factor that can modify plasma and tissue VEGF concentration is its degradation both in plasma and in peripheral tissues. The evidence that when incubated in vitro with plasma in the absence of any cellular element, VEGF has a short halflife (τ 1/2~4 0 min) suggested that it is degraded with a rate constant of 2.3-2.8·10 −4 s −1 [187]. Most of this degradation is operated by plasmin that cleaves VEGF 165 at the Arg 110 /Ala 111 residues yielding to two fragments VEGF  and VEGF  . The former retains the binding regions for VEGFR-1 and −2 but has a low biological activity as compared to native VEGF [188]. Interestingly, modified forms of VEGF of higher stability and bioactivity were obtained by substitution of Arg 110 with Ala 110 or Gln 110 , and Ala 111 with Pro111 [189]. VEGF can also be proteolytically processed by tissue metalloproteinases in multiple different ways that differentally contribute to the fine tuning of its biological activity in tissues [187]. Collectively, the available evidence on VEGF pharmacokinetics that we just reviewed, suggests that once given intravenously this growth factor will evenly distribute to peripheral tissues being, however, largely captured by the endothelial cell lining that sequester it and prevent its further diffusion deep in these tissues. Given this pharmacokinetic profile, a low CNS penetration for peripherally administered VEGF is expected to occur similar to what has been described for other growth factors such as LIF or NGF [190]. The poor penetration of VEGF in CNS was confirmed by Yang et al. [191] who observed that after the intravenous injection of 125 I-VEGF in rats the majority of radioactivity localized to peripheral tissues whereas it was very low in the brain. Considering this unfavorable pharmacokinetic profile in most of the studies that used VEGF as a neuroprotective agent this growth factor was given by invasive administration routes allowing its direct penetration in the brain. The simplest strategy that has been used is the intracerebroventricular administration in which the drug is directly injected in the ventricular system, usually in lateral ventricules, through cannulae that are stereotaxically positioned after drilling a hole in the skull. Such an administration route has been used in some of the studies exploring the effect of exogenous VEGF on stroke, PD, AD, ALS that we reviewed already in the previous section. The intracerebroventricular (icv) administration has several important limitations that include its invasiveness, the risk of serious neurosurgical complications and, importantly, the need of repeated administration when short lived drugs are administered. Indeed, icv-administered drugs are expected to rapidly disappear from cerebrospinal fluid and, therefore, if a prolonged pharmacological effect is required, repeated injections or the implantation of reservoirs that slowly release the molecule through a stably implanted cannula. This strategy proved successful, for instance, in animal models of ALS. Storkebaum et al. [108] implanted in SOD1 G93A rats a minipump that released VEGF in the lateral ventricles for more than 100 days and were able to document both the transportation of the growth factor to the spinal cord and a significant amelioration of the symptoms of the disease. Quantitative data are missing to firmly establish how much of the administered VEGF really reaches the neurons or brain endothelial cells but it is commonly believed that a large part of what is injected is actually lost because it is sequestered by the ependymal cells lining the ventricles [192]. To overcome the limitation of the short duration of action and, possibly, of the low intracerebral bioavailability several strategies have been developed and tested in animal models of neurodegenerative diseases [182]. Among them of special interest appears encapsulation. VEGF, as other growth factors, may be included in microspheres that once implanted in the brain slowly release their content. VEGF has been encapsulated in alginate microgels obtaining its sustained release [193,194]. However, this formulation was not used for the intracerebral administration that was, instead, experimented with poly(lactic-co-glycolic acid) (PLGA) microspheres. These microspheres appear well suited for sustained drug release because once that they are injected they are gradually hydrolyzed and slowly release their content; in addition, they seem reasonably safe and well tolerated [195]. Moreover, the speed of their hydrolysis may be finely adjusted by modifying the ratio of poly(lactic acid) to poly(glycolic acid). VEGF was originally incapsulated in PLGA nanoparticles to obtain a sustained release in cardiovascular disorders. For instance, Golub et al. [196] showed that they promoted revascularization in a mouse model of femoral artery ischemia and Formiga et al. [197] observed a substantial improvement in vasculogenesis and tissue remodeling in a rat model of cardiac ischemia-reperfusion. VEGFcontaining PLGA nanoparticles have been also used in animal models of neurological disorders. Herrán et al. [198] instilled VEGF containing PLGA nanospheres on the cortical surface of APP/Ps1 mice, an experimental model of AD, through a craniotomy window, after opening the dura mater. Animals treated with the nanospheres performed much better than untreated controls in behavioral tests showing no difference from wild type littermates. Moreover, the histopathological analysis of their brains collected three months after the procedure showed a much smaller amyloid plaque load, a reduced neuronal loss and an increase in angiogenesis. Evidence for enhanced neurogenesis was also obtained [199]. The same research group also evaluated the efficacy of VEGF plus GDNF containing PLGA nanospheres in rats made parkinsonian with the intrastriatal injection of 6-OH dopamine showing both an improvement in amphetamine-induced rotation behavior and an increase in the density of dopaminergic neurons in substantia nigra [200,201]. A surgical approach based on the implantation in the core of the ischemic lesion of a scaffold made by hyaluronic acid (HA)-based biodegradable hydrogel, mixed with PLGA microspheres containing VEGF and angiopoietin-1, and antibodies against Nogo receptor has been evaluated in experimental stroke [202].
Encapsulation techniques have a crucial relevance also for cellular therapies of neurodegenerative diseases that are based on the surgical transplantation of cells genetically engineered to stably release VEGF. However, these therapeutic modalities are beyond the scopes of the present review. We will, instead, examine in more detail another delivery strategy that can be used to direct VEGF in a noninvasive way, the intranasal administration.

Intranasal VEGF administration
Intranasal administration represents a promising alternative to the invasive approaches that we just described. This delivery strategy uses the nasal cavities to get a direct access to the brain bypassing the general circulation [203]. This is made possible by the anatomy of nasal cavities. Briefly, they consist of two main regions: a smaller region, the vestibule, and a larger cavity, the main nasal cavity. Vestibule is the outer part and receives the air crossing the nostrils. The main nasal cavity consists of two different portions: the respiratory and the olfactory regions. The respiratory region is lined by a mucus secreting epithelium with cilia. Because of the presence of turbinates and of microvilli the total surface of this portion of nasal cavities is larger than expected. The cilia move the mucus towards the pharynx and, remarkably, the speed of mucus flux is much slower than in the respiratory tree. This epithelium is highly vascularized which makes it potentially suitable for systemic absorption of drugs. However, the respiratory region of the nasal cavities is also innervated by fibers of the trigeminal nerves that can be used by inhaled drugs as paths to reach the brain [190][191][192]. This process can also take place in the olfactory region of the nasal cavity. It is made by the cribriform plate that is crossed by the fibers of the olfactory nerves that make contact with olfactory receptors in the olfactory mucosa. This is a pseudostratified nonbeating mucous epithelium. Drugs may cross the olfactory epithelium and be transported perineurally along the olfactory fibers towards the olfactory bulbs hence crossing the BBB. Additional mechanisms such as perivascular transportation or direct diffusion to the cerebrospinal fluid may also contribute to drug transportation to CNS. Centrally-acting drugs such as antiepileptic drugs, general anesthetics, analgesic and antipsychotic drugs have been administered through the intranasal route and in some cases this is also a relevant route of administration in clinical practice [204][205][206][207][208][209][210][211][212][213][214][215][216][217]. Because drugs given intranasally bypass, at least in part, the general circulation and get direct access to the brain, this administration route appears as an interesting option for the administration of neuroprotective growth factors [218][219][220][221]. The advantage of directing therapeutics to the brain avoiding peripheral distribution and degradation does not seem to be limited to small drugs and peptides and, intriguingly, the use of the intranasal route has been also explored for the administration of umbilical stem cells to promote neurovascular repair in the ischemic brain [222,223]. The ability of VEGF to penetrate into the brain after intranasal administration was demonstrated Yang et al. [191]. They gave 125 I-VEGF to rats by the intranasal route and looked for the appearance of radioactivity in the brain 30 min thereafter. A strong signal was observed especially in the trigeminal nerve, the optic nerve, olfactory bulbs, olfactory tubercle, striatum, medulla, frontal cortex, midbrain, pons, appendix cerebri, thalamus, hippocampus, cerebellum. Conversely peripheral distribution of intranasally administered VEGF was negligible was and substantial if it was given intravenously.
It was shown that when it was given by this route, VEGF promoted the formation of new blood vessels in the boundary of the ischemic lesion after MCAO in rats and reduced the neurological impairment [224]. An unusual bell-shaped dose dependence curve was observed for the neuroprotective effects of VEGF after tMCAO with the highest protection after the intranasal administration of 200 μg/ml and the lowest after either 100 or 500 μg/ml [225]. Intranasal VEGF was also effective in a rat model of global cerebral ischemia reducing cognitive impairment as evaluated in behavioral tests performed 15 days after stroke [85]. Whereas these data demonstrate that intranasally administered VEGF is effective in eliciting neuroprotection, high doses of this growth factor were required suggesting that this administration route is probably highly inefficient. Although, at the best of our knowledge, the nasal bioavailability of VEGF has not been determined yet, values between 0.01-0.1% have been suggested as a general rule, for intranasally administered drugs [226]. Several factors could contribute to lowering the bioavailability of intranasally administered drugs including, for instance, their transfer in the farynx where they are swallowed, their degradation by enzymes of the nasal mucus and of the nasal epithelium and the short residency time on the absorbing surface of the mucosa [227]. It remains to be established whether and to which extent permeabilizing effects of VEGF on nasal mucosa could compensate for these factors. It has been reported, indeed, that this growth factor permeabilizes the nasal mucosa about 10 6 times more potently than histamine on a molar base [228].

Safety concerns on VEGF therapeutics and possible solutions
Potential concerns can be raised on the use of VEGF in CNS therapeutics. Some of them are not related to the specific indication for neurological disorders and are similar to those that could take place if VEGF was used for the treatment of heart or limb ischemia, namely hypotension, whereas, others are specifically observed in the context of CNS diseases.
Because of the well established role of VEGF in tumor growth, an increased risk of cancer has been considered as a likely, seriously worrying potential complication of VEGF treatment. However, the clinical evidence accumulated up to date in trials evaluating therapeutic angiogenesis with VEGF did not substantiate this worry [229][230][231][232][233][234]. Moreover, in experimental animals hyperexpressing VEGF, an high prevalence in hemangiomas but not of cancers was observed [235][236][237][238][239].
The most serious concern about the use of VEGF in CNS therapeutics is related to the effect of this growth factor on BBB permeability. In normal conditions the tightness of BBB is strictly regulated and its loss occurs in many neurological disorders including stroke, traumatic brain disease, AD, PD, ALS, multiple sclerosis and viral encephalitis [240][241][242][243][244][245][246][247]. The loss of the integrity of BBB has an important role in the progression of the aforementioned disorders because it contributes to extend brain damage by multiple mechanisms that include, among the others, the impairment of the clearance of toxic products such as amyloid peptides and the influx into brain parenchyma of toxic substances, of cytokines and blood proteases such as coagulation factors, and of inflammatory cells [247,248]. VEGF increases blood vessel permeability and it was actually initially identified because of this biological activity and not because of its proangiogenic activity [249]. As such it is a well known BBB permeabilizing agent. Different mechanism contributes to VEGF BBB-permeabilizing activity. Specifically, this growth factor promotes transcytosis [250,251]. In addition, it induces the dissociation of tight junction by downregulating the expression of zonulin-1, claudin and occludin [252][253][254][255].
Current evidence suggests that endogenous VEGF could be one of the causative factors of BBB permeabilization in CNS diseases. It has been demonstrated, indeed, that reactive astrocytes release this growth factor that can diffuse from the brain parenchyma to the endothelium and permeabilize the BBB [256]. Mice with targeted disruption of VEGF in astrocytes showed a reduced BBB breakdown in response to the injection of recombinant adenovirus expressing IL-1 and milder neuropathology after induction of experimental allergic encephalomyelitis [256]. Interestingly, in control mice, most of the detrimental effects of VEGF on BBB integrity were lost after the pharmacological blockade of eNOS suggesting that this growth fact was acting by inducing the synthesis of NO. An additional mechanism that could increase VEGF levels in the damaged brain is activated in brain ischemia and involves the tissue plasminogen activator (tPA) [257,258]. It has been demonstrated, indeed, that tPA induces the expression of VEGF in neurons, endothelial cells and bone marrowderived myeloid cells by a mechanism that could involve the activation of HIF-1 dependent transcription [257,258]. A convincing argument of the involvement of endogenous VEGF in BBB dysfunction after stroke is the evidence that it can be reduced with either anti-VEGF antibodies or blockers of VEGFR-2 [257,259]. The hypothesis that not only VEGF levels but also the sensitivity to this growth factor increased after stroke, was evaluated in diabetic mice and in normal controls. After stroke, a marked increase in the expression of VEGFR-2 was observed in the periinfartual area in both these groups although it was much stronger in diabetic than in normal mice [260]. These data suggested that a heightened responsiveness to endogenous VEGF could have a role in BBB permeabilization after stroke. This hypothesis was confirmed by the evidence that shutting down the expression of VEGFR-2 in conditional adult knock-out mice or pharmacologically inhibiting their activity with the VEGFR-2 inhibitor SU5416 reduces BBB leakage in experimental stroke. Surprisingly, however, the protective effects were observed only in diabetic and not in normal mice suggesting that the potentiation of VEGFR response is probably a causative factor of BBB dysfunction only in diabetes.
As a whole the data that we reviewed show that in several neurological diseases endogenous VEGF actually contributes to brain damage by inducing BBB dysfunction. This raises serious concerns about the rationale of using this growth factor to treat CNS disorders because its beneficial effect could be reverted by the detrimental effects on BBB leakage. As a matter of fact there is a rich scientific literature showing that VEGF can actually act a double edge sward in many neurological disorders. Probably stroke is the condition in which this issue appeared as more urgent.

VEGF-mimic peptides as a new tool for VEGF therapeutics in disorders of the central nervous system
The data reviewed in the previous sections show that although there is a solid rationale and convincing evidence that VEGF could be helpful in several neurological disorders, this growth factor is far from having a good pharmacological profile for its development in clinical practice. Indeed, as detailed before, it has an unfavorable pharmacokinetics and toxicological profile. In addition, it is not a molecule easy to prepare on large scale for pharmaceutical purposes considering the stringent requirements with respect to impurity identification and reproducibility of the batch. A first structural characteristic that complicates its preparation is that the main VEGF isoform is constituted by two monomers, each 165 amino acids long, covalently linked by a two intermolecular disulfide bonds. Furthermore, the VEGF structure is characterized by a cystine knot motif which is formed by three intramolecular and interconnected disulfide bonds. The presence of these specific disulfide bonds make the preparation of VEGF, in a reproducible and standardized protocol, challenging. VEGF expression in bacteria and eukaryotic cells has been reported [261][262][263][264] as well as the chemical synthesis by native chemical ligation [265]. All these techniques present advantage and drawbacks. For example, VEGF expression in bacteria and chemical synthesis require a refolding step respectively from inclusion body and after assembly of the full length protein. The refolding procedure is time and reagent consuming, reduces the yield of final protein, and introduces variability in the preparation. Collectively these considerations suggest that an additional factor besides the unfavorable pharmacokinetics and toxicology, complicating VEGF implementation in CNS therapeutics is its complicated and expensive synthesis.
An interesting strategy that is emerging in the last years to overcome these limitations is to prepare smaller and simpler peptides that could interact with VEGF receptors and trigger their activation hence reproducing VEGF biological activity. These peptides are known as VEGF mimic or VEGF mimetic peptides. The rationale behind their synthesis is that the interaction of VEGF with its receptors does not involve the whole VEGF molecule but only part of it that can be reproduced in isolation and will, hopefully, retain the ability to stimulate the receptor itself. Because there is the need of maintaining the intrinsic activity at the receptor, it is much more challenging to design such molecules than peptides that simply bind to the receptors but do not have agonistic properties. As a matter of fact, a larger number of VEGF-antagonist peptides has been synthesized and tested as potentially useful tools to inhibit pathological angiogenesis in tumors [266][267][268][269][270] whereas only few peptides with VEGF biological properties have been synthesized [271]. The first of them was reported in 2005 by D'Andrea et al. [272]. It was named QK and it is a 15 mer αhelical peptide that was designed to reproduce the α-helical region of VEGF 17-25 using a structure-based approach. The choice of this region derived from the analysis of the published crystal structure of VEGF bound to domain 2 of VEGFR1 [273]. This domain together with domain 3 is known to be the only of the seven extracellular domains that is really critical for VEGF binding [243]. This crystal study showed that the interaction between VEGF and VEGFR1 involves three discontinuous portions of the VEGF molecule: the N-terminal helix (residues 17-25), the loop joining strand β3 and β4 (residues 61-66) of one VEGF monomer, and the β-hairpin encompassing strand β5 and β6 (residues 79-93) of the other VEGF monomer. The QK peptide was, therefore, designed to reproduce a portion of the VEGF molecule critically involved in receptor binding as it includes 5 (Phe-17, Met-18, Tyr-21, Gln-22, and Tyr-25) out of 21 residues located at a distance less than 4.5 Å from the receptor surface [272]. Specific residues of the 17-25 VEGF sequence were substituted in QK with the aim of either increase the stability of the α-helix conformation in water or of the increase the activity at VEGFR. For instance, Trp was substituted for Phe-17 to increase hydrophobic interactions, Gln was substituted for Met-18 to enhance hydrogen bond interactions at the receptor site, Asp-19 was replaced by Glu, and Ser-24 with Lys to increase helix propensity [272]. The whole structure was further stabilized by introducing N-and C-capping extra-sequences. Circular dichroism analysis and NMR spectroscopy showed that peptide assumes in water a high populated helical conformation (Fig. 3). The structural determi-nants of QK structure were studied in water by NMR and molecular dynamics revealing an unusual thermodynamic stability [274]. The molecular determinants of peptide stability were also characterized [275,276]. The ability of QK to bind to VEGFR was verified in a radioassay experiments on isolated EC membranes [272], and by flow cytometric analysis on A375 cells [276]. The VEGF agonistic activity of the QK peptide was demonstrated using both models in vitro and in vivo. More specifically, it was observed that, in cultured endothelial cells, QK activates VEGFR2 inducing the phosphorylation of its intracellular kinase domain [242], and the activation of Akt, Erk1/2, eNOS [277]. Moreover, like VEGF, QK also induced the upregulation of vegfr2 gene in these cells [277]. QK was also able to reproduce the major biological effects of VEGF on endothelial cells inducing their proliferation, migration and survival [272,277]. QK retained VEGF proangiogenic properties as demonstrated ex vivo on aortic rings [277] and in vivo in matrigel plugs [272,277]. Although we cannot conclusively exclude that part of the effect of QK or of other VEGF-mimic peptides could be exerted intracellularly, this appears very unlikely and current evidence all converge in suggesting that VEGF mimic peptides act at the level of VEGF plasmamembrane receptors.
Because of its ability to promote new vells formation, the possibility that QK could be effective as a tool for therapeutic angiogenesis appeared likely and was experimentally tested. The first disease models in which it was evaluated were hindlimb ischemia and wound healing [278]. Hindlimb ischemia was induced in rats by the permanent closure of the right common femoral artery. QK or VEGF or a scrambled peptide as control, were infused intraperitoneally for two weeks. Not differently from VEGF 165 , QK improved the perfusion of the ischemic hindlimb as evaluated at angiography and increased the density of capillaries in the tibialis anterior muscle of the ischemic hindlimb [278] (Fig. 4). To test the effect on wound healing a 20 mm in diameter open wound was surgically made on the dorsum of anesthetized rats and QK, VEGF 165 or a scrambled peptide were applied for 5 days as a pluronic acid gel on the wound. Both QK and VEGF accelerated wound healing as compared with the scramble peptide [278]. QK showed regenerative properties also in an experimental model of gastric ulcer accelerating healing in a iNOS-dependent manner [279] and improved the course of experimental diabetic peripheral neuropathy [280]. The chemical and biological properties of peptide QK stimulated the interest of several groups working on biomaterial for applications in regenerative medicine [281][282][283][284][285][286][287][288][289][290] ratifying the importance of this molecule as VEGF substitute for applications where a proangiogenic response is required.
Collectively the accumulated data suggested that QK could be a reasonable alternative to VEGF in proangiogenic and regenerative therapies. The effect of QK was, therefore, also tested in a pMCAO model of experimental brain ischemia. In the first experiments, QK was administered intracerebrovetricularly using a microinfusion pump that released a solution containing a concentration of 3 μg/ml QK at the rate of 1 μl/h the peptide and that was implanted before surgery [291]. Ischemic damage evaluated three days after surgery was significantly smaller in rats given QK than in controls. However, because of its small size, QK should easily distribute in tissues and reach the ischemic tissue even when given intravenously. This hypothesis was tested by intravital pial microvessel videoimaging, a technique that visualized pial vessels after the intravascular administration of fluorescent probes in living animals through a window made in the skull. Using this Fig. 3. Structure of the VEGF mimic peptide QK. A) NMR structure of the QK peptide. Note the α-helix conformation and the two caps at the N-and C-terminals of the peptide. B) Model reconstruction of QK binding to VEGF receptors (Flt-1D2) (Reproduced from D'Andrea et al., [242], no permission needed). Fig. 4. The QK peptide promotes a regenerative response in hindlimb ischemia. Panel a shows representative sections of the tibialis anterior muscle obtained from animals treated with the inactive peptide VEGF 15 , with VEGF or with QK that were continuously infused in the femoral artery for 14 days all at the concentration of 10 −7 Mol/L (Magnification ×40; bar =10 µm). The sections were stained with lectin GBS-I to visualize the capillaries. Panel b shows the quantification of the data as mean ± SEM. Note that muscle capillary density is significantly higher in rats treated with either VEGF or QK than in animals that received VEGF 15  technique, we observed that after intravenous administration QK rapidly adhered to the endothelial lining of pial vessels and thereafter extravasated in brain parenchyma on the venular side (Fig. 5). This finding suggests some important difference from VEGF (that remains, however, to be quantitatively tested) because, as we mentioned in Section 3, this growth factor is in great part captured inside endothelial cells and this capture limits its diffusion beyond vessel endothelial lining. The ability to diffuse outside the capillary lumen into the brain parenchyma, suggested that QK could be effective even if given intravenously. This hypothesis was tested giving different QK doses (1.2, 12, 40, 120, 360 and 1200 μg/kg) through an indwelling femoral catheter to different groups of rats one hour after pMCAO. A significant reduction of the infarct area was observed in rats receiving 40 or 120 μg/kg QK whereas higher or lower doses were ineffective. Because of its small size, QK was also expected to easily diffuse into the brain when given intranasally. We tested this hypothesis using a fluorescent derivative of this peptide and performing confocal imaging on brain slices obtained three hours after its intranasal administration. A strong fluorescence was observed in several brain areas including the cerebellar cortex, the hippocampus and striatum and the olfactory bulbs whereas the signal was fainter in cerebral cortex. This distribution profile was similar to what has been described before for other intranasally administered peptides. The evidence that QK diffuses well into the brain parenchyma suggested that this administration route could be used in brain ischemia. This hypothesis was confirmed by experiments performed in rats that received QK intranasally three hours after pMCAO. The ischemic damage evaluated after three days was 40% lower in these animals that in controls receiving vehicle and the beneficial effect of the treatment persisted also in animals sacrificed 7 days after surgery. Interestingly, in experiments performed with all of the three administration routes that we used, we noticed that there was no apparent increase in the volume of the brain of QK-treated rats as compared with controls. This important finding establishes a major difference respect to VEGF because it suggests that different from VEGF, this peptide could not induce edema formation. To further investigate this point QK and VEGF were compared for their ability to permeabilize pial vessels as evaluated from FITC-dextran extravasation in intravital pial microvessel videoimaging experiments. The results obtained showed that fluorescent dextran significantly extravasated after the administration of VEGF given either by intracarotid injection or by topical application but remained inside blood vessel when QK was given by either these routes (Fig. 6). Collectively the results of our study showed that QK could be an extremely interesting option for CNS therapeutics because it has neuroprotective properties but does not permeabilize the BBB. Much work remains to be done with this promising peptide. For instance, its possible activity in experimental model of other CNS disorders remains to be tested. Moreover, a mechanistic explanation for the lack of BBB permeabilization after QK administration remains to be established although several hypotheses could explain this property of the peptide. For instance, we have mentioned in Section 4 that different transduction mechanisms are responsible for the proangiogenic and prosurvival effect of VEGF on one side and its vessel permeabilizing activity on the other. Specifically, vessel permeabilization seems to be largely dependent on the activation of the src cascade. Therefore, it is tempting to speculate that, somehow, the QK peptide could selectively activate the beneficial VEGF transduction cascades but not those leading to vessel permeabilization. Interestingly, a similar divergence in the ability to activate postreceptor events has been described for VEGF 121 and VEGF 165 that in the skin differentially activate src and the MAPK cascades [292]. Another possible explanation is in the different structure of QK and VEGF. As we mentioned at the beginning of this section, the binding site of VEGF to VEGFR has a multimodular structure with different components from different portions of the VEGF dimer and the 17-25 sequence is only part of it. It is conceivable, therefore, that the QK peptide could be less potent than VEGF on a molar base and that, therefore, it could be safer than VEGF as far as BBB permeabilization is concerned simply because this effect is observed at high VEGF doses [95][96][97]. Under many respect the QK peptide can be considered a prototypical neuroprotective peptide that showed the feasibility of a therapeutic approach using only part and not the whole VEGF molecule. After its synthesis further efforts have been directed to synthetise novel and possibly better molecules. We just pointed out that QK only reproduces one of the different portions of the VEGF monomer that take part to VEGF biding site to its receptors. The formal demonstration that VEGF receptors can also be activated by synthetic peptides reproducing these The QK peptide freely diffuses in brain parenchyma after intracarotid injection. The images were obtained by acquiring sequential images by pial vessel intravital videomicroscopy through a window opened in the skull of a rat before (A) and after (B,C and D) the intracarotid injection of fluorescent QK. Note that after injection fluorescence quickly moves from the arteriolar (B) to the venular compartment (C) to completely translocate, thereafter, into the brain parenchyma (Reproduced from Pignataro et al., [261], no permission needed). other interaction modules was obtained with the synthesis of HPLW, a 17-mer peptide that was designed on the β-hairpin region 87-101 of PlGF [293]. When dissolved in water, HPLW assumes a well formed βhairpin conformation in a two-step folding process [294]. The peptide HPLW is able to bind to VEGFRs as demonstrated by NMR studies on purified domain 2 of VEGFR-1 [293,295] and in presence of ECs overexpressing VEGFRs [296]. VEGFR binding was also confirmed in vitro by citofluorimetry using ECs [293]. The biological activity of peptide HPLW was determined in vitro and in vivo. HPLW activates VEGF-dependent intracellular pathways, induces proliferation and rescue from apoptosis in ECs, and promotes capillary formation in an in vivo angiogenesis assay [293]. Because both QK and HPLW reproduce only one interaction module of VEGF with its receptors it is conceivable that more potent peptides could be obtained by the combination of multiple modules in the same molecule. The design of such multimodular molecules raises specific problems because these different regions have to maintain their relative spatial arrangement to interact with the receptors. De Rosa et al. [297] were successful in designing a series of second generation VEGF mimic peptides that include not only the α -helix 17-25 of QK but also a region similar to the β -hairpin 79-93, joined by variable linkers. The effect of the most interesting of these peptides, named EP3, EP5 and EP6 were tested in vitro for their ability to induce the phosphorylation of VEGF receptor and the activation of the ERK1/2 pathway, to promote endothelial cell migration and proliferation in vitro and angiogenesis both in vitro and in vivo. The results showed that these peptides have biological effects similar to VEGF. In particular, the most active of them, EP6, showed a biological activity comparable to QK [297]. It will be interesting to test in future experiments whether these peptides yield any advantage respect to QK in experimental models of neurodegenerative diseases.

Conclusion
In conclusion, the data that we reviewed in the present paper show that VEGF has a great potential as a therapeutic tool for neurodegenerative diseases. However, the development of such a VEGF-based therapeutics for practical clinical applications is hampered by the unfavorable pharmacokinetics profile of this molecule and by concerns on its safety. New VEGF-mimic peptides could represent a cheaper and easier to administer alternative to VEGF. Only QK, a prototype of first generation compounds that reproduce only one of the VEGF binding molecules to VEGFR, has been tested in neurological disorders and the data obtained suggested that it could be safer than VEGF. New, second generation peptides, designed to reproduce multiple VEGFR interac-tion modules, hold promises of being more effective than QK and studies with these molecule in CNS disorders are eagerly awaited.

Bench
The analysis of the effects of small peptides reproducing one or more regions of VEGF may help understanding basic issues on the structure-activity relationship of this growth factor.

Bedside
Small VEGF-mimic peptides could have better pharmacokinetic and/or toxicological properties than VEGF and be, therefore, potentially suitable for use in human diseases.
Community VEGF-mimic peptides-based therapeutics could help reducing the burden of severe neurodegenerative disorders that cannot be efficiently treated with currently available drugs.

Funding
This work was not supported by any external funding. Fig. 6. The QK peptide does not induce BBB permeabilization. The tightness of BBB was evaluated by measuring the leakage of 70 kDa fluorescent dextran, a high molecular weight fluorescent probe that is retained in the capillary lumen unless the endothelium is permeabilized. Panel A shows representative images obtained before and after the intracarotid injection of QK (top row) or VEGF (bottom row). Note that the parenchyma become much brighter after VEGF than after QK injection. The bar graphs in panel B show the mean values of the fluorescence of brain parenchyma after the administration of different doses of QK (on the left) or VEGF (on the right). Note that fluorescence becomes significantly higher after VEGF than after QK injection. (Reproduced from Pignataro et al., [261], no permission needed).