The angiogenic factor vascular endothelial growth factor A (VEGF) has been shown to have a role in neurogenesis, but how it affects adult neurogenesis is not fully understood. To delineate a role for VEGF in successive stages of olfactory bulb (OB) neurogenesis, we used a conditional transgenic system to suppress VEGF signaling at the adult mouse sub-ventricular zone (SVZ), rostral migratory stream (RMS) and OB, which constitute the respective sites of birth, the migration route, and sites where newly born interneurons mature and integrate within the existing OB circuitry. Following the development of fluorescently tagged adult-born neurons, we show that sequestration of VEGF that is constitutively expressed by distinct types of resident OB neurons greatly impaired dendrite development in incoming SVZ-born neurons. This was evidenced by reduced dendritic spine density of granule cells and significantly shorter and less branched dendrites in periglomerular neurons. Notably, the vasculature and perfusion of the SVZ, RMS and OB were not adversely affected when VEGF suppression was delayed until after birth, thus uncoupling the effect of VEGF on dendritogenesis from its known role in vascular maintenance. Furthermore, a requirement for VEGF was specific to newly born neurons, as already established OB neurons were not damaged by VEGF inhibition. This study thus uncovered a surprising perfusion-independent role of VEGF in the adult brain, namely, an essential role in the maturation of adult-born neurons.
Vascular endothelial growth factor (VEGF; VEGFA – Mouse Genome Informatics) is the key factor promoting and coordinating most, if not all, processes of blood vessel formation in the embryo and adult. In addition to its angiogenic role, VEGF is also important for maintaining vascular homeostasis by way of adjusting the vasculature to meet dynamic changes in oxygen supply and demand, controlling vascular permeability, and maintaining certain organ-specialized vascular phenotypes (e.g. the blood-brain barrier or endothelial fenestrations). Functions of this pleiotropic factor, however, extend beyond the vascular system and include essential roles in hematopoiesis, bone formation, hepatocyte proliferation and others (for a review, see Ferrara, 2004). An increasing body of evidence implicates VEGF in neuronal processes in the adult brain. Ectopic VEGF augments proliferation of adult hippocampal neurons and enhances hippocampus-dependent learning (Cao et al., 2004; Jin et al., 2002; Schanzer et al., 2004). Conversely, inhibition of VEGF abolishes the increase in basal neurogenesis that takes place in response to exercise or an enriched environment (Cao et al., 2004; Fabel et al., 2003). Neurogenesis was also shown to increase in the adult sub-ventricular zone (SVZ) in response to VEGF (Jin et al., 2002; Schanzer et al., 2004). A specific neurogenic process in which the role of VEGF was further analyzed is the seasonal addition of new neurons to the high vocal center (HVC) of male songbirds (Louissaint et al., 2002). VEGF is also neuroprotective, as shown both in vitro (Jin et al., 2000; Matsuzaki et al., 2001) and in animal models of ischemic neuronal injury (Sun et al., 2003). Remarkably, hypoexpression of Vegfa in the spinal cord leads to neuronal cell death and the development of an amyotrophic lateral sclerosis (ALS)-like disease that can be rescued by exogenous VEGF (Azzouz et al., 2004; Lambrechts et al., 2003). VEGF has also been suggested to play a role in neuronal migration during brain development (Schwarz et al., 2004) or following injury (Wang et al., 2007) and has been shown to promote neuronal migration in vitro (Balenci et al., 2007).
Canonically, neurogenesis in the adult brain takes place in two niches: the subgranular zone of the hippocampus and the SVZ, which continuously supplies new interneurons to the olfactory bulb (OB). The latter niche is more suitable for discerning successive stages of the overall process in which VEGF might play a role, owing to the fact that neuronal birth, migration, differentiation and integration are spatially and temporally separated. Briefly, neuroblasts destined to become OB interneurons leave their respective sites of origin in the SVZ and migrate rostrally (tangential migration) along the rostral migratory stream (RMS). Upon reaching the OB, cells migrate radially and, upon reaching their destinations at the granule cell layer (GCL) or glomerular layer (GL), differentiate into granule cells (GCs) or periglomerular cells (PGNs), respectively, becoming fully functional interneurons (see Fig. 1C). Tangential migration in the mouse OB is completed within 2-7 days following cell division, radial migration within 5-7 days, and differentiation to GCs or PGNs within 2 or 4 weeks, respectively, from birth (Petreanu and Alvarez-Buylla, 2002). In the next period of 45 days, sensory inputs determine whether newly added interneurons will thrive or regress. Approximately 50% of newly integrated GCs survive beyond this critical period and function for ∼100 additional days on average, prior to being replaced by new cells (Ninkovic et al., 2007; Petreanu and Alvarez-Buylla, 2002). Whereas adult-born GCs comprise a sustained fraction (10%) of all GCs, newly added PGNs have a better survival rate, resulting in a linear increase in their number, reaching 30% at the age of 9 months (Mizrahi et al., 2006; Ninkovic et al., 2007). Others have reported that the population of newborn GCs increases with time to reach ∼60% at the age of 12 months (Imayoshi et al., 2008).
The processes of neuronal proliferation, migration, differentiation and survival are governed by complex cues that are mediated by both intrinsic and extrinsic factors that only recently started to be revealed (reviewed by Lledo et al., 2006). Here, we found that Vegfa is constitutively expressed in the OB in a unique spatial pattern that cannot be simply explained by its known vascular roles, which prompted us to examine a possible role for VEGF in OB adult neurogenesis. Using a bi-transgenic mouse system that we devised for conditional knockdown of VEGF signaling in the relevant regions of the adult brain, we indeed uncovered an unexpected role of VEGF in the maturation of adult-born neurons.
MATERIALS AND METHODS
All animal procedures were approved by the animal care and use committee of the Hebrew University. Transgenic mouse lines used in this study were as follows. The CamKIIα-tTA brain-specific driver line was purchased from Jackson Labs (Mayford et al., 1996). Vegfa-lacZ-KI reporter was a generous gift of A. Nagy (Miquerol et al., 1999). Flk1-lacZ-KI reporter was a generous gift of J. Rossant (Shalaby et al., 1997). The pTet-sVEGF-R1 responder line was as described previously (May et al., 2008). For switching off I-sVEGF-R1, water was supplemented with 500 mg/l tetracycline (Tevacycline, Teva) and 3% sucrose. For switching on the transgene, tetracycline-supplemented water was replaced with fresh water for the desired time. BrdU (Sigma, 100 mg/kg body weight) was injected intraperitoneally for the durations indicated. Pimonidazole (Hypoxyprobe, Chemicon, 60 mg/kg body weight) was injected intraperitoneally 30 minutes prior to sacrifice. Dextran-FITC (MW 500,000, Sigma) was injected into the tail vein 5 minutes prior to sacrifice (2 mg per mouse).
Lentivirus, retrovirus and stereotaxic injection
Subventricular zone virus injection was as described (Mizrahi, 2007). Briefly, a commercial lentivirus (TranzVector, Tranzyme) or Moloney murine retrovirus (a generous gift of Carlos Lois, Picower Institute of Learning and Memory, MIT, Cambridge, MA, USA), both expressing GFP, were used to label newborn neurons. Mice were anaesthetized and a skull hole was drilled overlying the SVZ (coordinates relative to bregma: anterior, –1 mm; lateral, –1 mm; depth, 2.2 mm). Each mouse received two injections (∼0.75 μl) on both the left and right SVZ. For histological analysis, mice were sacrificed, perfused with 4% paraformaldehyde and processed as described below.
Two-photon imaging and analysis
Obs were fixed with 1% paraformaldehyde, rinsed in 30% sucrose and dissected to 350 μm coronal sections. Imaging was carried out on an Ultima microscope (Prairie Technologies, Middleton, WI, USA) equipped with a 40× (0.8 NA) LUMPlan F1 water-immersion objective (Olympus). A femtosecond laser (Mai-Tai Spectra Physics) was used to excite GFP at 900 nm. Images (512 × 512 pixels) were acquired at 0.25 μm/pixel resolution in the xy dimension and in 0.9-1 μm steps in the z dimension. Reconstructions were carried out strictly on the raw data images. Reconstructions were performed manually from the complete three-dimensional image stacks using Neurolucida (MicrobrightField, Colchester, VT, USA). Quantitative parameters included the number of branch points and total dendritic branch length. Since 350 μm-thick slices were used, we could not assure that the whole dendritic tree was captured. In addition, in fixed slices the quality of the optical signal normally deteriorates when imaging deeper than 175 μm from the surface. For these two reasons, our data (although composed of a consistent sample) do not capture the complete dendritic tree. Notably, all reconstructions were conducted blind to the experimental group.
RNA in situ hybridization
In situ hybridization of formalin-fixed, paraffin-embedded tissue sections with 35S-labeled antisense riboprobes was performed as described (Dor et al., 2001).
Formalin-fixed, paraffin-embedded tissues were dissected to 3 μm sections. Staining was performed with the following: anti-VEGFR1 (1:100, Santa Cruz), anti-BrdU (1:100, Amersham), anti-NRP1 (1:50, R&D), anti-NRP2 (1:100, Cell Signaling), anti-CD31 (1:50, BD Pharmingen), anti-VEGF (1:100, Calbiochem), anti-p44/42 (1:200, Cell Signaling), Hypoxyprobe (1:100, Chemicon) and anti-cleaved caspase 3 (1:100, Cell Signaling). The secondary reagents used were: Cy3 anti-goat, biotin anti-rat and Cy5 streptavidin (Jackson ImmunoResearch), Alexa-555 anti-rabbit (Invitrogen), FITC anti-mouse (Zymed) and ImmPress Universal (Vector Laboratories).
For BrdU-NeuN immunostaining, frozen 50 μm sections were pretreated with 50% formamide/2×SSC (0.3 M NaCl, 0.03 M sodium citrate) followed by 2 M HCl and 0.1 M boric acid. Incubation with rat anti-BrdU (1:200, Serotec) and mouse anti-NeuN (1:400, Chemicon) was followed by biotinylated rabbit anti-rat (Vector Laboratories), Cy5-conjugated extravidin and Cy3 anti-mouse (Jackson ImmunoResearch). Confocal microscopy was performed using a Nikon Eclipse 9i. z-stacks were done using a ×60 oil-immersion objective and 0.5 μm steps.
Tissue was fixed with 4% paraformaldehyde for 15 minutes on ice and then washed twice in rinse solution (PBS containing 2 mM MgCl2, 0.01% sodium deoxycholate and 0.02% Nonidet P40). Staining was by incubation overnight at 37°C with rinse solution containing 5 mM ferrocyanide, 5 mM ferricyanide and 1 mg/ml X-Gal.
VEGF is constitutively expressed in three types of OB neurons, whereas VEGF receptors are mostly expressed in nearby endothelial cells.
To determine whether Vegfa is naturally expressed in the SVZ-RMS-OB axis and to identify Vegfa-expressing cells, we examined brain sections from adult (2-3 months old) mice harboring a Vegfa promoter-lacZ reporter transgene. In this mouse strain, a lacZ cassette has been inserted into the 3′UTR of the endogenous Vegfa gene such that β-galactosidase (β-gal) activity faithfully recapitulates natural patterns of Vegfa expression (Miquerol et al., 1999). Patterns of Vegfa expression were also determined by mRNA in situ hybridization and immunostaining using a specific riboprobe and antibodies, respectively. In the SVZ, Vegfa expression was evident as weak and diffuse β-gal staining or as a weak mRNA in situ signal. By comparison, the adjacent choroid plexus was strongly labeled, as described previously (Breier et al., 1992) (see Fig. S1 in the supplementary material). In the RMS, Vegfa expression was not detected by any of three detection methods (see Fig. S1 in the supplementary material). Remarkably, a unique spatial pattern of Vegfa expression was evident in the OB, mostly distributed in two concentric cell layers (Fig. 1A,B). Vegfa-expressing cells were subsequently identified as mitral cells within the inner layer and as juxtaglomerular cells (tufted and periglomerular cells) within the outer cell layer (Fig. 1A, middle, Fig. 1C). In the GCL, Vegfa was also expressed by some astrocytes, as evident by co-immunostaining with GFAP (Fig. 1A, right). This unique pattern of expression was unlikely to merely reflect a vascular maintenance function of VEGF, as in that case Vegfa would have been expected to be more uniformly expressed within the OB, rather than being restricted to these specific locales. The finding that the four types of VEGF-producing cells reside close to destination sites of incoming new neurons (see Fig. 1D) prompted us to examine a possible role for VEGF in neuronal maturation or integration.
As a prelude to examining a putative role for VEGF in neuronal maturation, we wished to determine which cells within the OB might respond to the constitutively expressed VEGF by virtue of expressing VEGF receptors. Considering that VEGFR2 (Flk1; Kdr – Mouse Genome Informatics) was previously shown to be expressed on neural cells both in vitro (Jin et al., 2006) and in vivo (Jin et al., 2002; Krum et al., 2002) and to mediate certain neuronal functions (Jin et al., 2002; Maurer et al., 2003), we first examined its pattern of expression by mRNA in situ hybridization (ISH) and through the use of an Flk1 promoter-lacZ reporter mouse (Shalaby et al., 1997). Consistent with previous studies (Schanzer et al., 2004), Flk1 expression was indeed detected by both methods in neuroepithelial cells of the SVZ ependyma and in the choroid plexus (see Fig. S2 in the supplementary material). In the OB proper, however, Flk1 expression was only detected on endothelial cells residing in the GCL and GL, but not on neuronal cells or astrocytes in the same layers (see Fig. 2A,B). A similar analysis was extended to VEGFR1 (Flt1), previously reported to be expressed in astrocytes of the SVZ and RMS (Krum et al., 2002; Wittko et al., 2009). Focusing on the OB proper, however, a Flt1 ISH signal was only detected in endothelial cells of the GL and GCL layers, but not in neurons or astrocytes in the same regions (Fig. 2C).
Next, we determined patterns of OB expression of neuropilin 1 (Nrp1) and neuropilin 2 (Nrp2) by in situ immunostaining. Consistent with previously reported patterns (Cloutier et al., 2002), olfactory sensory neuron (OSN) axons were found to be strongly immunoreactive. However, no expression was detected in endothelial cells or in GCs or PGNs within the OB (Fig. 2D,E). Together, these results suggest that neuronal cells within the OB might not be directly responsive to VEGF (see Discussion).
A transgenic system for conditional knockdown of VEGF in the SVZ-RMS-OB axis
To examine a possible role for VEGF in successive stages of OB neurogenesis, we devised a transgenic system to conditionally knockdown VEGF signaling at the relevant regions of the adult brain. The system is based on conditional induction and, in turn, de-induction of a secreted VEGF decoy receptor. Briefly, transgenic mice expressing a tetracycline-regulated transactivator protein (tTA) from a calmodulin kinase IIα (CamKIIα; Camk2a – Mouse Genome Informatics) promoter (Mayford et al., 1996) were mated with transgenic mice harboring a transgene encoding a chimeric tetracycline-regulated protein comprising an IgG1-Fc tail fused to the five Ig-like loops of the extracellular domain of VEGFR1. The induced secreted receptor (abbreviated as I-sVEGF-R1) efficiently binds and sequesters VEGF, thereby precluding its signaling (Kendall et al., 1996) (see Fig. 3A).
In all experiments described below, double-transgenic mice were selected for VEGF modulations, whereas littermates that inherited only one of the two transgenes served as controls. The onset of I-sVEGF-R1 activation in double-transgenic animals and the duration of its expression were tightly controlled by including or omitting tetracycline from the drinking water (‘off’ and ‘on’ modes, respectively) (Fig. 3A,B).
Under the dictate of the CamKIIα promoter, which was chosen because of its increased activity in brain areas of increased plasticity, I-sVEGF-R1 was robustly induced upon tetracycline withdrawal in the SVZ, RMS and in the GCL and GL layers of the OB, i.e. in the respective sites of neurogenesis, neuronal migration and neuronal maturation and integration (Fig. 3B,C). I-sVEGF-R1 was also strongly induced in olfactory glomeruli formed by coalescence of OSN axons. Indeed, intense I-sVEGF-R1 immunoreactivity was found in OSN cell bodies in the olfactory epithelium (Fig. 3C, right). The CamKIIα promoter also drove I-sVEGF-R1 expression in additional areas of the adult brain, such as in the cortex and hippocampus, although areas outside the SVZ-RMS-OB axis were not examined in this study. Importantly, the induced VEGF-sequestering receptor is a secreted protein such that VEGF is efficiently titrated regionally, irrespective of the exact identity of the producer cell. This provided us with the opportunity to determine which process, if any, might require VEGF.
Uncoupling perfusion-dependent and -independent requirements for VEGF in the developing and adult brain
Uncovering non-vascular roles of VEGF using different methods of Vegfa inhibition, such as Cre-mediated deletion or using a hypomorphic Vegfa allele (Haigh et al., 2003; Raab et al., 2004), has been hampered by the fact that these procedures may negatively impact on tissue perfusion. It has been difficult, therefore, to rule out the possibility that observed phenotypes are merely a secondary consequence of hypoxia or nutrient deprivation or result from a deficit in other crucial circulating factors. For example, it is currently unknown whether the severe developmental brain defects in Vegfa hypomorphic mice (Haigh et al., 2003) reflect a non-vascular role of VEGF in the brain or are secondary to impaired perfusion. VEGF is known to be required not only for neovascularization, but also for sustaining newly formed vessels, which become refractory to VEGF only upon their maturation (Benjamin et al., 1999). We reasoned, therefore, that beyond a critical developmental period, VEGF blockade would no longer lead to vessel regression or to a compromised vascular function. To test this, we used the same CamKIIα-driven conditional system for inducing I-sVEGF-R1 in the mouse brain at different embryonic and postnatal days.
The severity of the resulting brain defects (e.g. the magnitude of cortical thinning and neuronal apoptosis) was indeed markedly reduced as the onset of VEGF inhibition was further delayed (data not shown). When the onset of I-sVEGF-R1 induction was delayed until embryonic day 13.5, animals survived to adulthood and gross brain abnormalities were no longer detected at birth. Yet, a continued VEGF blockade led to involution of the OB, culminating in rudimentary bulbs and even their total absence (Fig. 4). Inspection of bulbs undergoing involution revealed a significant vascular loss (Fig. 4C,D), widespread regions of tissue hypoxia [visualized by staining with an antibody against pimonidazole (Hypoxyprobe) that selectively marks hypoxic microenvironments] and extensive cell apoptosis (Fig. 4F,H). Premature VEGF withdrawal also led to collapse of the RMS vascular support, ensuing hypoxia and massive apoptosis of SVZ-born neurons along their migration route towards the OB (Fig. 4E,G). In fact, as a result of the failure to cross the RMS, whether owing to the lack of a proper vascular scaffold or to the loss of their target, migrating neuroblasts piled up amidst the RMS and eventually died (Fig. 4E,G).
In sharp contrast to the VEGF dependence of the RMS and OB vasculatures at earlier times, delaying the onset of VEGF blockade until after birth no longer resulted in damage to the, by then, already mature vascular network anywhere along the SVZ-RMS-OB axis. When VEGF blockade was initiated in the adult brain (Fig. 5), no evidence for excessive endothelial or neuronal cell death was noted (as examined by immunostaining for cleaved caspase 3; Fig. 5A). More importantly, organ perfusion was not affected, as evidenced by CD31 (PECAM1 – Mouse Genome Informatics) immunostaining for endothelial cells in the RMS (Fig. 5C) and in the OB (Fig. 5D,F), as well as by tail vein injection of fluorescently tagged dextran, which revealed no change in the density and distribution of patent vessels (Fig. 5E,G). As expected, there was also no evidence for hypoxia (Fig. 5B). These findings allowed us to rigorously rule out the possibility that any of the neuronal OB phenotypes described below, as induced by VEGF inhibition at the indicated times after birth, might be secondary to a compromised perfusion.
We nevertheless wished to obtain a positive indication that inducing I-sVEGF-R1 in the adult does induce VEGF blockade in the OB and eventually leads to altered signaling in neuronal cells. To this end, we examined the phosphorylation status of p42/44 (ERK1/2 – Mouse Genome Informatics), a known downstream target of VEGF (Gliki et al., 2001) as well as of other neurogenic factors (Fu et al., 2002; Obrietan et al., 2002; Wu et al., 2007). Phosphorylation of p42/44 in GCs is induced by activity (odor stimulation) (Miwa and Storm, 2005). Switching on I-sVEGF-R1 in the adult brain for 3-6 weeks indeed led to a reduction in phospho-p42/44 in GCs from ∼1000 to ∼400 positive cells/mm2 in the GCL area (see Fig. S3 in the supplementary material). We therefore conclude that VEGF loss-of-function (LOF) in the adult OB, although not adversely changing perfusion, may alter GC signaling, even though this is not necessarily via a direct mechanism.
Influx of SVZ-born neurons into the OB and deployment therein are not affected by VEGF blockade
To identify a particular step in the overall process that might be compromised by VEGF inhibition, we first compared the number of newly born neurons that reach the OB using BrdU labeling experiments. I-sVEGF-R1 was induced in 2-month-old mice and 1 week later a short BrdU pulse was given. Animals were maintained in the ‘on’ mode for an additional 45 days to allow sufficient time for neuronal migration, maturation and integration and were then sacrificed for analysis. Newly added neurons were identified as cells double positive for BrdU and NeuN (NEUNA60 – Mouse Genome Informatics), and their numbers in the GCL and GL were compared with those detected in control littermates (see Fig. 6A for representative examples and Fig. 6C for quantification). A similar number of BrdU+/NeuN+ cells were detected in the GCL of control and VEGF-inhibited mice (663 and 611 cells/mm2, respectively). Likewise, the number of newly added neurons was not significantly different in the GL of control and VEGF-inhibited mice (188 and 228 cells/mm2, respectively).
Considering that the number of newly added neurons in the OB measures the net outcome of SVZ neurogenesis, migration through the RMS and survival, this result on its own does not rule out a role for VEGF in any of the individual processes. Nevertheless, the fact that the number of SVZ-born neurons reaching the OB and integrating therein was not reduced in the face of VEGF blockade enabled us to independently examine a putative role of VEGF in neuronal maturation.
VEGF inhibition leads to reduced dendritic spine density in newly acquired GCs
The finding of constitutive Vegfa expression in close proximity to integration sites of incoming neurons (Fig. 1) prompted us to examine a possible requirement for VEGF during the processes of neuronal maturation and integration. We first compared newly integrated GCs in control and VEGF-inhibited OBs with respect to their morphology. Briefly, I-sVEGF-R1 was induced in adult mice (with mono-transgenic littermates serving as controls), followed by SVZ injection of a lentivirus encoding green fluorescent protein (GFP) to tag SVZ-born neurons. OBs were retrieved for analysis 45 days later. Initial two-photon analysis of 350 μm-thick coronal sections did not reveal a significant difference in gross dendritic morphology of GCs (Fig. 7B). However, further analysis of 40 μm slices by confocal microscopy showed a marked decrease in the density of dendritic spines in mice that had been subjected to VEGF inhibition (Fig. 7C,D). Comparing the cumulative number of spines along dendritic segments exceeding 5 μm in length revealed that mice in which I-sVEGF-R1 was induced had only half the number of spines as control mice (0.21 versus 0.42 spines/μm, respectively). This reduction was also reflected in direct density measurements, where the average distance between adjacent spines was found to be significantly greater in VEGF-inhibited mice (Fig. 7D). We conclude that although VEGF is not required for GC dendritogenesis, it is required for proper development of the complete array of synaptic connections of these cells.
VEGF inhibition impairs dendritogenesis of newborn PGNs
A similar type of analysis to that described above for GCs was extended to the more peripheral PGNs. GFP-expressing PGNs were imaged in thick slices by two-photon microscopy. This allowed us to capture the dendritic trees of newly added PGNs, which were then reconstructed in 3D. PGNs from VEGF-inhibited mice displayed markedly fewer developed dendritic trees than control mice (Fig. 7E). This was evidenced by a 2-fold reduction in total dendritic length (364.8 μm compared with 762.9 μm in controls) (Fig. 7F, left), as well as significant reduction in the number of dendritic branching points (3.2 compared with 9.3 in controls) (Fig. 7F, right; but see reservations it terms of whole-cell morphology in Materials and methods). These results indicate that VEGF is required for proper dendritogenesis of adult-born PGNs.
Dendritogenesis of established GCs and PGNs is not impaired by VEGF inhibition
We next wished to determine whether VEGF is also required for the maintenance of dendritic spines in mature GCs and for the maintenance of proper dendritic trees of mature PGNs. We turned to SVZ injection of GFP-expressing retrovirus that allows for a more accurate birthdating of SVZ-born cells. Following the injection of a GFP-encoding retrovirus in the SVZ, animals were divided into two groups. In the first group, animals were kept in the ‘off’ mode for 30 days to allow sufficient time for integration and maturation of GFP-tagged incoming neurons. I-sVEGF-R1 was then switched ‘on’ for an additional 30 days, the OB resected and dendritic trees and spines analyzed as described above. In the second group, I-sVEGF-R1 was induced concomitantly with SVZ injection of the same GFP-encoding retrovirus and animals were maintained in the ‘on’ mode throughout the 60-day duration of the experiment (see Fig. 8A). This experiment allows adult-born neurons that were deprived of VEGF during development and onwards (‘developing’) to be distinguished from those deprived of VEGF after their maturation (‘established’). In accordance with the results obtained with the lentivirus vector (Fig. 7, above), withdrawal of VEGF during development resulted in GCs with fewer dendritic spines and in PGNs with poorly developed dendritic trees (Fig. 8). By sharp contrast, no impairments of spine formation or dendritogenesis were detected when VEGF was withdrawn after maturation and integration of GCs and PGNs had taken place (Fig. 8). Further, VEGF knockdown at this stage of maturation resulted in a significant increase in GC spine density.
These results indicate that whereas VEGF is essential for normal spine formation and dendritogenesis of adult-born neurons, it is not required for maintaining already formed dendrites and spines. Nevertheless, the increase in spine density in VEGF-inhibited mature GCs suggests that it might have a role in those cells (see Discussion).
Vascular and neuronal cells may interact in different ways in the adult brain. Not only are constant adjustments of the vasculature required to match perfusion demands imposed by dynamic changes in neuronal activity, but the vasculature may also provide essential neurotrophic factors and participate in generating a proper niche for adult neurogenesis (Palmer et al., 2000; Shen et al., 2008; Tavazoie et al., 2008). Because VEGF is the key factor responsible for vascular homeostasis in the brain, it is also the prime candidate for a mediator of vascular/neuronal interactions and possibly also as a direct effector of non-vascular cells within the adult brain. VEGF has indeed been shown to enhance activity-induced hippocampal neurogenesis (Cao et al., 2004; Fabel et al., 2003). However, how VEGF might affect OB neurogenesis and which particular events in the neurogenic process might rely on VEGF remain unknown. Here, we uncovered a new and surprising role of VEGF in the maturation of OB interneurons. Specifically, VEGF is important for proper dendritogenesis of newly added PGNs and for dendritic spine formation of newly added GCs. Whether the new roles ascribed for VEGF by this study also apply to adult neurogenesis in the hippocampus, and perhaps even to general structural plasticity, remains to be determined.
Traditionally, it has been difficult to distinguish neuronal phenotypes resulting from impaired perfusion, which is an anticipated consequence of VEGF LOF, from perfusion-independent effects. The facility to conditionally knockdown VEGF in the adult brain in a highly localized and spatially restricted manner provided by our unique transgenic system allowed us to uncouple the two in the context of the natural in vivo setting. This is particularly significant considering that different vascular beds within the brain may differ with regard of their dependency on VEGF. Furthermore, as demonstrated here for the RMS and OB vasculatures, the transition from a VEGF-dependent to a VEGF-independent phase may take place with a different developmental or postnatal timing. Clearly, our ability to exercise VEGF LOF only after securing that VEGF is no longer required for vascular maintenance (Fig. 5) is a prerequisite for exploring perfusion-independent functions of VEGF in the adult brain. Maintaining vascular integrity in the face of VEGF inhibition is also crucial in the context of the SVZ ‘vascular niche of neurogenesis’ and when utilizing blood vessels as a scaffold for neuronal migration in the OB (Bovetti et al., 2007).
Our results showing that adult VEGF blockade does not reduce the number of SVZ-born neurons entering the OB (Fig. 6) ought to be discussed with reference to studies demonstrating a role for VEGF in SVZ neurogenesis and in neuronal migration, as both are successive steps contributing to the overall process of populating the OB with new neurons. Previous studies that demonstrated enhanced proliferation of SVZ neurons by VEGF were based on VEGF gain-of-function (Jin et al., 2002; Wang et al., 2007), and currently there is no evidence for a natural role of endogenous VEGF in SVZ neurogenesis, despite high expression of Flk1 in the ependyma (as confirmed here; see Fig. S2 in the supplementary material). Our results suggest, albeit not prove, that whereas vessels are needed for RMS migration (Bovetti et al., 2007), ongoing VEGF signaling might not be required. A requirement for VEGF signaling in neuronal migration has been elegantly demonstrated for the process of facial nerve development in which VEGF signaling was directly implicated in proper somata positioning (Schwarz et al., 2004). Yet, it remains to be determined whether VEGF also plays a direct role in neuroblast migration in the context of OB adult neurogenesis.
A comparable steady-state level of newly integrated neurons (Fig. 6) is compatible with the notion that VEGF is not required to sustain the survival of incoming neurons, a conclusion supported by our failure to detect increased cell death in the VEGF-inhibited OB (Fig. 5).
The fact that overall, sufficiently high numbers of SVZ-born neurons matured and integrated in the OB in the face of VEGF inhibition enabled us to uncover the roles of VEGF in dendritogenesis and dendritic spine formation. Whereas a number of proteins are known to promote dendritogenesis and synaptogenesis in adult-born cells in the dentate gyrus of the hippocampus (Danzer et al., 2008; Sato et al., 2007; Segura et al., 2007), almost nothing is known about factors that affect these processes in the OB, of which VEGF might be a first example.
In the adult hippocampus, VEGF has been shown to be involved in activity-induced neurogenesis but not to affect the rate of basal neurogenesis (Cao et al., 2004; Fabel et al., 2003). In the adult OB, it has been reported that environmental factors, such as odor deprivation by nostril occlusion, reduce GC dendritic spine density (Rehn et al., 1988; Saghatelyan et al., 2005). Interestingly, it was shown in these studies that odor deprivation only affects spine density in newborn GCs, and not in mature GCs. Our findings that VEGF sequestration results in reduced spine density in newborn, but not mature, GCs (Fig. 8) are also supported by a recent report describing a similar critical period for synaptic development of new GCs with respect to sensory inputs (Kelsch et al., 2009). Therefore, it is tempting to extrapolate that VEGF might mediate physiological responses to environmental changes. In fact, we found that whereas VEGF inhibition led to reduced dendritogenesis in developing cells, spine density in already mature cells was increased. Although we do not have a definitive explanation for this puzzling observation, several possibilities exist. First, GC dendritic spines are known to remain highly dynamic even after integration (Mizrahi, 2007) and VEGF might play a role in their continuous plasticity (such as in the control of spine pruning). Second, the increase in spine density of resident neurons might reflect an as yet unknown compensation mechanism for poor dendritogenesis of incoming new neurons. Third, this increase might reflect a compensation mechanism for other possible effects of VEGF inhibition in the OB that are yet to be discovered.
An open question concerns the functional consequences of the observed impediments to the maturation of new neurons in response to VEGF inhibition. A major effect is not expected in light of a recent report in which mice performed normally in several olfaction tasks despite genetic ablation of most interneurons (Imayoshi et al., 2008). Unexpectedly, however, in preliminary experiments we noticed compromised olfaction in VEGF-inhibited mice (data not shown), raising the possibility that, in addition to its effect on interneurons, VEGF might also be required for other cells participating in odor perception.
A basic unanswered question is whether the effects of VEGF on dendritogenesis are direct, i.e. take place via VEGF binding to neuronally expressed VEGF receptors, or indirect, i.e. via inducing endothelial cells or astrocytes to secrete another factor that acts on neuronal cells. A precedent for an indirect effect is the VEGF-induced seasonal addition of new neurons to the HVC of male songbirds (Louissaint et al., 2002). Here, testosterone-induced VEGF was found to stimulate nearby endothelial cells to secrete brain-derived neurotrophic factor (BDNF), which, in turn, promotes recruitment of neurons from the HVC ventricular zone. That neuronal cells may directly respond to VEGF is supported by expression of the high-affinity VEGF receptors VEGFR1 and VEGFR2 and of the auxiliary non-signaling receptors NRP1 and NRP2 on neuronal cells and on astrocytes (Jin et al., 2002; Jin et al., 2000; Mani et al., 2005; Maurer et al., 2003). We could not detect expression of any of the VEGF receptors on adult-born interneurons in the natural in vivo context (Fig. 2). Although we cannot exclude the possibility of low-level expression that is below our detection threshold, the data better argue for an indirect, paracrine effect of VEGF mediated by endothelial cells, which are shown here to express VEGF receptors. An autocrine VEGF effect on the resident neurons that produce it is yet another possibility, although autocrine VEGF effects have so far been described only in endothelial cells (Lee et al., 2007). Supporting a direct mechanism are findings showing that VEGF can induce neurite growth in cultured cortical neurons (Jin et al., 2006; Khaibullina et al., 2004; Rosenstein et al., 2003). In vivo, however, only the study of Schwarz et al., cited above in the context of migration, has convincingly demonstrated a direct action of VEGF on neuronal cell bodies (Schwarz et al., 2004). A rigorous distinction between direct and indirect mechanisms may necessitate neuronal-, glial- and endothelial-specific ablation (or functional knockdown in vivo) of each of the four VEGF signaling and auxiliary receptors. Whether direct or indirect, our findings clearly show that VEGF function is essential for proper development of the dendritic trees and dendritic spines specifically of maturing interneurons.
We thank Janet Rossant and Andras Nagy for mice and Carlos Lois for viral reagents. This study was supported by the Israel Science Foundation and the German-Israeli Foundation (to E.K.) and by a Career Development Award from the International Human Frontier Science Program Organization and the Abisch Frenkel Foundation (to A.M.).
Competing interests statement
The authors declare no competing financial interests.
Supplementary material for this article is available at http://dev.biologists.org/lookup/suppl/doi:10.1242/dev.039636/-/DC1
- Accepted November 5, 2009.
- © 2010.