Neurons continue to be generated in the subventricular zone (SVZ) throughout postnatal development and adulthood in rodents. Whereas in adults, virtually all neuroblasts migrate tangentially to the olfactory bulb via the rostral migratory stream (RMS), in neonates, a substantial fraction migrate radially through the corpus callosum (CC) to the cortex. Mechanisms of radial cortical migration have remained unknown. We investigated this by taking recourse to enhanced green fluorescent protein (EGFP)–labeled neuroblasts in the CC and deep cortical layers of neonatal mice and found that they are frequently located adjacent to vasculature. Using time-lapse 2-photon microscopy in acute brain slices, we demonstrate that EGFP-labeled neuroblasts migrate along blood vessels. Although in close proximity to blood vessels, migrating neuroblasts are separated from endothelial cells by 1–2 layers of astrocytic processes, as revealed by electron microscopal studies of retrovirally labeled postnatally born cells. We propose that 2 factors could contribute to the decline of radial migration to the cortex during postnatal development, namely the establishment of a glial sheath delineating the RMS and a gradual decrease in the density of blood vessels in the CC. Together, our data provide evidence for a new mode of radial cortical migration of SVZ-generated neurons involving vasculature and astrocytes.
During embryonic development in mammals, most cortical glutamatergic neurons are generated in the ventricular zone of the dorsal telencephalon and migrate radially to their final location in the cortex (Rakic 1972). Cortical γ-aminobutyric acid (GABA)ergic interneurons, on the other hand, are generated primarily in the ganglionic eminence and the preoptic area of the ventral telencephalon. From there, interneuron progenitors migrate tangentially to the cortex (Marin and Rubenstein 2003; Wonders and Anderson 2006; Batista-Brito and Fishell 2009; Gelman et al. 2009). In rodents, neurogenesis persists postnatally in 2 distinct neurogenic niches: the subventricular zone (SVZ) and the subgranular zone (Temple and Alvarez-Buylla 1999). Newborn cells generated in the SVZ migrate tangentially via the rostral migratory stream (RMS) to the olfactory bulb (OB), where they become GABAergic granule cells or dopaminergic, glutamatergic, and GABAergic periglomerular neurons (Lois and Alvarez-Buylla 1994; De Marchis et al. 2007; Brill et al. 2009).
We recently described the generation of transgenic mice expressing enhanced green fluorescent protein (EGFP) under the control of the 5HT3 receptor promoter (5HT3-EGFP mice). In addition to the expected expression of EGFP in GABAergic interneurons located mainly in the upper cortical layers, we observed that in these mice, most neuroblasts in the RMS express EGFP. Furthermore, in neonatal and juvenile 5HT3-EGFP mice, we noticed that in addition to the expected tangential migration to the OB, EGFP-positive neuroblasts leave the RMS prematurely and migrate through the corpus callosum (CC) into the cortex where they integrate mainly in the lower cortical layers (Inta et al. 2008). Radial neuroblast migration gradually decreases over the first 4 postnatal weeks and is eventually barely detectable. Cells migrating to the cortex become mature GABAergic interneurons endowed with unique anatomical and electrophysiological features (Le Magueresse et al., in press). Thus, neuroblasts generated in the neonatal SVZ show radial migration to the cortex in addition to the tangential migration toward the OB that has been extensively studied in adult animals. The mechanisms underlying radial cortical migration in neonates are unknown.
Studies on the migration of newborn cortical pyramidal cells have shown that during the embryonic period, postmitotic glutamatergic neurons cells arising in the ventricular zone (VZ) migrate to their final location in the cortex using 2 modes of migration. The first postmitotic neurons migrate by somal translocation to form the preplate, which in turn gives rise to Cajal–Retzius cells and subplate neurons (Miyata et al. 2001; Nadarajah et al. 2001; Nadarajah and Parnavelas 2002). At subsequent developmental stages, elongated fibers of radial glia guide neuroblasts in migration to their final location in the cortex (Rakic 1972, 2007). GABAergic interneurons generated embryonically in the ganglionic eminence migrate tangentially but eventually migrate also radially in the developing cortex to reach their final laminar position (Nadarajah and Parnavelas 2002; Ang et al. 2003). Radial migration of GABAergic interneurons takes place along radial glia and is connexin 43 dependent, thus sharing some similarities with radial glia–guided migration of newborn glutamatergic neurons generated in the VZ (Elias et al. 2007; Elias et al. 2010). During the perinatal period, radial glial cells undergo transformation into mature astrocytes (Cameron and Rakic 1991; Chanas-Sacre et al. 2000; Tramontin et al. 2003). They are hence unlikely to support radial glia–guided migration is unlikely to occur beyond the first postnatal days. There is, however, increasing evidence that vasculature can provide structural organization and molecular cues that guide neuroblast migration in the developing and adult brain. In adult animals, blood vessels parallel the RMS and guide neuroblasts toward the OB (Snapyan et al. 2009; Whitman et al. 2009). Furthermore, neuroblasts in progression toward their final location within the OB migrate radially along blood vessels (Bovetti et al. 2007). Recent studies have also provided evidence for a close association of neuroblasts and blood vessels in the embryonic cortex (Javaherian and Kriegstein 2009; Stubbs et al. 2009).
We hypothesized that vasculature may provide a substrate for radial cortical migration of interneuron progenitors generated neonatally in the SVZ. We used a combination of techniques including retroviral fluorescent labeling, time-lapse 2-photon microscopy on acute brain slices, electron microscopy, and immunostainings to investigate the underpinnings of SVZ-generated neuroblast migration to the cortex. We show that astrocytes associated to blood vessels in the CC constitute a scaffold for cortical migration of postnatally generated cortical neurons in neonatal mice. Our results indicate that structural changes occur in the RMS and CC in the course of development. The glial sheath associated with the RMS in adults is absent in neonates, and the density of blood vessels in the CC of neonatal animals is higher, which may foster radial cortical migration. Thus, we described radial migration of immature neurons to the cortex during the early postnatal period, which is fundamentally different from radial glia–mediated radial cortical migration occurring in the embryonic brain. The process described here may be reenacted in the diseased adult brain, since neuroblasts generated in the SVZ that are derouted to the periinfarct site following stroke have been found adjacent to blood vessels.
Materials and Methods
The generation of the 5HT3-EGFP transgenic mice and the correct expression of the transgene have been described previously (Inta et al. 2008). For immunostainings, adult mice were overdosed with ketamin/xylazin and intracardially perfused with 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) (0.1 M, pH 7.4). The brains were removed and fixed overnight in 4% PFA. Pups (P4–P12) were killed by decapitation, the brain was removed and fixed overnight in 4% PFA. All procedures were approved by the Local Animal Care Committee.
Retroviral Labeling of Neurons Born Postnatally in the SVZ
We used a replication-deficient murine Moloney leukemia virus–based retroviral vector expressing RFP under the control of the CAG promoter, kindly provided by Dr F. H. Gage (Salk Institute, La Jolla, CA). The concentrated viral solution (108 cfu/mL) was produced in the packaging cell line human embryonic kidney 293 (HEK 293), as previously described (Laplagne et al. 2006). Viral solutions were injected in the brain of wild-type pups at postnatal day 4 (P4) to label newborn cells. Injections were performed as follows: 1 μl of viral solution was injected through a glass micropipette into the SVZ of 4-day-old wild-type mice using the following coordinates from bregma: 0.5 mm anterior, 1.5 mm lateral, and 1.5 mm ventral. Pups were returned to their mother and killed after 5–6 days.
Acute Slice Preparation
To label blood vessels in acute brain slices, we added the red fluorescent dye DiI (initial concentration 1% in dimethyl sulfoxide, final concentration 0.02%) to warm (31–32 °C) artificial cerebrospinal fluid (ACSF) containing (in mM): 120 NaCl, 3.5 KCl, 2.5 CaCl2, 1.3 MgSO4, 1.25 NaH2PO4, 25 NaHCO3, 25 glucose, bubbled with 95% O2/5% CO2. 5HT3-EGFP mouse pups aged P4-P8 were overdosed with ketamin/xylazin and intracardiacally perfused with DiI-containing ACSF. Subsequently the brain was swiftly removed from the skull and placed in ice-cold DiI-free ACSF. 250–300 μm-thick parasagittal brain sections were prepared using a vibratome (HR2, Sigmann Elektronik) and kept submerged in ACSF at 32 °C for 30 min, then at room temperature (22–25 °C). Sections were subsequently transferred to the recording chamber and continuously superfused with ACSF at room temperature.
Two-Photon Imaging on Acute Slices
Imaging of fluorescent labeled cells in acute slices was carried out on a TCS SP5 microscope (Leica) equipped with a 20× (1 NA) water immersion objective. A picosecond laser (Mai-Tai, Spectra Physics) was used to excite EGFP and DiI at 900 nm. Images (512 × 512 pixels) were acquired at 1–1.5 μm per pixel resolution in the xy dimension and 0.5–2 μm steps in the z dimension. Movies were made from 3D stacks acquired sequentially every 4 to 15 min. Maximal intensity projections were subsequently aligned in ImageJ (NIH). To measure the speed of migration, moving neuroblasts were tracked manually using Mtrack J plugin for ImageJ (E. Meijering). The duration of a saltatory period and the distance covered by the neuroblast during that period were calculated only during migration, starting with the first time point where movement of the cell body was initiated. The speed of migration was extracted from these 2 values.
Brain slices (60 μm) were cut from fixed brains using a vibratome (VT1000S, Leica). Free-floating sections were first permeabilized in 0.2% Triton X-100 in PBS for 30 min and blocked in 3% bovine serum albumin in PBS for 1 h before incubation with the primary antibody at 4 °C overnight. After 3 washes in PBS, the sections were incubated with the secondary antibody for 2 h at room temperature, washed 3 times with PBS, and mounted. Primary antibodies were rabbit anti-EGFP (1:10 000, Invitrogen), chicken anti-EGFP (1:1000, Abcam), mouse anti-CD31 (PECAM-1) (1:500, BD Pharmingen), mouse anti-GFAP (1:1000, Sigma), goat anti-doublecortin (1:500, Santa Cruz), and rabbit anti-DsRed (1:1000, Clontech Living Colors). Secondary antibodies were anti-rabbit Alexa 488 (1:1000, Invitrogen), anti-chicken Alexa 488 (1:1000, Invitrogen), anti-mouse Cy3 (1:1000, Jackson Immuno Research Laboratories), anti-goat Cy3 (1:1000, Jackson Immuno Research Laboratories), anti-rabbit Cy3 (1:1000, Jackson Immuno Research Laboratories), and anti-mouse Alexa 647 (1:1000, Invitrogen). Sections were analyzed on a confocal microscope LSM510 (Zeiss).
Image Analysis and Quantifications
In z-projections of confocal image stacks, the average distance of randomly distributed cells to a blood vessel was estimated using a bootstrap analysis. To ensure unbiased sampling, a grid consisting of 9 rows and 9 columns was superimposed to the maximal intensity projection of the confocal picture. We measured the distance between each intersection located out of the RMS and its closest blood vessel. To avoid counting from intersections located at the border of the image, that might be close to blood vessels outside of the field of view, we first calculated an estimate of the average distance between blood vessels (D), using exclusively blood vessels located at the center of confocal images. We then built the grid such that borders of the grid were more than D + 2 standard deviations away from the border of the image (Supplementary Fig. 1).
Subsequently, the distance from each EGFP-positive neuroblast within the borders of the grid to its nearest blood vessel was calculated in every confocal image stack using maximal intensity projections (Supplementary Fig. 1). Differences between the theoretical distance of randomally distributed cells to their nearest blood vessel and the real distance were compared using unpaired Student's t-test. All values are given as mean ± standard error of the mean (SEM).
Morphological Reconstructions and Quantification
Confocal image stacks were acquired in the CC using sagittal brain sections from 5HT3-EGFP mice immunostained for PECAM-1 and EGFP. The morphology of blood vessels in the CC was reconstructed in z-projections using the Neurolucida system (MBF Bioscience). Data on length and surface of blood vessel were collected using Neuroexplorer (MBF Bioscience) and normalized to the size of the area covered by the z-projection. Differences between neonates and adults were compared using unpaired Student's t-test. All values are given as mean ± SEM.
For visualization of anti-GFP primary antibodies, slices were incubated in biotinylated anti-rabbit antibody, followed by avidin-biotinylated enzyme complex and diaminobenzidene (DAB) enhancement, osmification, and epon resin flat embedding. Serial 60-nm sections were collected on pioloform-coated copper grids and contrasted with lead citrate. The electron micrographs were taken on an EM 10 Zeiss CR electron microscope.
Primary astrocyte cultures were prepared from young postnatal or adult mice as previously described (Banker and Goslin 1991). Briefly, brains from either P5/6-old or 5-week-old mice were removed in cold PBS glucose and sliced in sagittal sections (∼400 μm) with a blade. The area containing the RMS + CC was dissected out, minced, and placed into 15 mL tubes on ice. The tissue was incubated with papain (0.08%)/DNase I (0.001%) for 3 min at 37 °C to obtain single cells suspensions which were plated at a density of 100 000 cells/mL in Dulbecco's modified Eagle's medium (DMEM) (Invitrogen) supplemented with 10% fetal bovine serum (FBS) and penicillin/streptomycin. In order to distinguish astrocytes from young and adult brains, cultures from adult tissue were infected with adeno-associated virus (AAV) expressing tomato protein and kept in culture for 5–6 weeks to obtain an adequate amount of cells for the assays. Media were exchanged every week.
To test whether neuroblasts migrate preferentially along RMS/CC astrocytes, a mix of astrocytes from young mice and HEK cells (30 000 astrocytes + 5000 HEK cells per well) were plated on 24-well coverslips coated with poly-L-lysine (1 mg/mL, Sigma) in DMEM 10% FBS 1 day before preparing the explants. On the following day, the antimitotic 5-fluoro-2-deoxyuridine (FDX 2 mg/L + uridine 5 mg/L, Sigma) was added to the medium in each well to prevent the overgrowth of both cell types. The SVZ was microdissected from ∼500 μm thick coronal sections of P9 5HT3-EGFP transgenic mice brains, gently minced in small pieces (∼400 μm), and placed on top of astrocyte and HEK cell cocultures. Explants were maintained on the cell culture mix for 3 days, fixed, and stained with anti-GFAP and anti-EGFP antibodies as well as with 4′,6-diamidino-2-phenylindole (DAPI) (1 μM, Sigma).
To test whether neuroblasts migrate preferentially along astrocytes from young or adult brain, we followed the same procedure but plated astrocytes from young RMS/CC and AAV tomato–infected astrocytes from adult RMS/CC (20 000 cells from each age per well).
Confocal pictures were taken for each explant using a Zeiss LSM 700 microscope. We measured the distance from each EGFP+ cell to the closest astrocyte for all the neuroblasts that had migrated out of the explants (further away from explant-derived astrocytes). To obtain values from a random distribution, we superimposed a grid to each picture and measured the distance from each intersection point to the nearest astrocyte.
Vasculature as a Scaffold for Radial Migration through the Corpus Callosum in Neonates
On sagittal sections from young (P12) 5HT3-EGFP mice, we noticed that neuroblasts labeled with antibodies to EGFP and doublecortin (DCX) in the CC and lower cortical layers were often located in close proximity to areas devoid of DAPI staining, suggesting that they might be adjacent to blood vessels (Fig. 1A). To probe the association between vasculature and EGFP-positive neuroblasts, we performed immunostainings with antibodies to platelet endothelial cell adhesion molecule (PECAM-1) to label blood vessels in 5HT3-EGFP mice aged P4. PECAM-1-stained blood vessels were loosely aligned with the RMS (Fig. 1B), as reported previously in newborn and adult mice (Snapyan et al. 2009; Whitman et al. 2009; Nie et al. 2010). In the CC, vessels were oriented in different directions, including many radially oriented blood vessels (Fig. 1B). While most EGFP-positive cells were located in the RMS, many could be seen in the CC and lower cortical layers, often in direct apposition to radially oriented blood vessels (Fig. 1C). We determined whether the proximity of 5HT3-EGFP cells and blood vessels in the CC was higher than predicted by chance. On maximal intensity projections of confocal stacks, we found that the somata of EGFP-expressing cells were on average at a distance of 16.9 ± 1.0 μm from the closest blood vessel (3891 measurements in 32 confocal stacks from 5 mice), which is significantly lower than the mean theoretical distance assuming random distribution of neuroblasts (25.0 ± 1.0 μm; 2208 measurements in 32 confocal stacks from 5 mice; P < 1 × 10−6, see Materials and Methods, Supplementary Fig. 1 and Fig. 1D). Forty-nine percent of cell somata examined (32 confocal stacks) were at a distance < 10 μm from a blood vessel versus 31% predicted if the distribution were random. Conversely, 51% of cells were at a distance > 10 μm from a blood vessel (Fig. 1E). In this latter population, some of the cells in fact contacted blood vessels via their leading process. Others, however, did not seem to be adjacent to vasculature and may be cells that have stopped migrating and begun differentiation or be cells that have just initiated migration out of the RMS toward a capillary. Another possibility is that they migrate independently of vasculature. To investigate whether differentiated cells were as likely to be in the proximity of blood vessels as neuroblasts, we performed triple immunostainings with antibodies to DCX, EGFP, and PECAM-1 to label immature neuroblasts and blood vessels in 5HT3-EGFP mice aged P4 (Fig. 2A). At that age, all DCX+ cells also expressed EGFP (Fig. 2A′). In the CC and cortical layer VI, 7.5 ± 0.5% of EGFP+ cells were DCX− (15 confocal stacks examined from 3 mice). We found that EGFP+DCX+ are significantly closer to vasculature than EGFP+DCX− cells, indicating that vasculature exerts a stronger influence on neuroblasts than on cells already further ahead in the neuronal differentiation pathway (Fig. 2B,C).
To directly demonstrate migration of EGFP-labeled neuroblasts along blood vessels, we labeled vasculature using intracardiac perfusion of DiI-containing ACSF in 5HT3-EGFP pups (P4–P9) and subsequently performed time-lapse 2-photon imaging experiments in acute slices for 2–10 h. In keeping with our results from PECAM-1-immunostained fixed tissue, EGFP-positive neuroblasts in acute slices were often located in close proximity to DiI-labeled blood vessels (Fig. 3A,B). One hundred and two migrating cells were identified by time-lapse 2-photon imaging (11 movies from 5 mice). Time-lapse imaging revealed that EGFP-labeled cells often migrate along radially oriented blood vessels located in the CC (Fig. 3B′ and Supplementary Movie 1). Their migration was saltatory, that is, cell movements were preceded and followed by periods of immobility. Most neuroblasts had unbranched leading processes, whose outgrowth always preceded somal translocation. Migration was bidirectional. On average, 59.4% (28/48) of radially migrating cells along blood vessels migrated toward the cortex and 41.6% (20/48) migrated back to the RMS. During locomotion, cells migrating toward the cortex along blood vessels had a speed of 43.7 ± 5.2 μm/h (n = 28). Cells migrating back to the RMS that had a comparable speed, namely 45.2 ± 5.4 μm/h (n = 20). A significant fraction of migrating EGFP+ cells (54 of 102, i.e., 52.9%) did not have their soma in the direct vicinity of a blood vessel during migration. However, 22 of these cells contacted blood vessels via their leading process during migration. Of a total of 102 migrating cells, only 32 (31.4%) migrated independently of blood vessels. The speed of migration of cells migrating along blood vessels or independently of vasculature was not different (44.5 ± 3.7 and 39.7 ± 4.0 μm/h, respectively). The duration of saltatory periods was also similar (33.9 ± 2.4 and 33.7 ± 2.7 min, respectively). We cannot exclude, however, differences pertaining to other characteristics of migration, for example, smoothness of movement due to short periods of immobility, given that pauses between saltatory events can be fairly short—less than 8 min (Snapyan et al. 2009)—and would be missed using our acquisition protocol. Conversely, acute slices prevent the comparison of the frequency or duration of very long periods of immobility, which we observed in both populations of migrating neuroblasts. The duration of such pauses can exceed the duration of a movie and might in the end affect neuroblast migrating along blood vessels or independently thereof differentially, if followed over longer periods. This, however, cannot be studied in vitro.
Neuroblasts generated in the adult SVZ migrate in chains to the OB (Lois et al. 1996). Here, we observed both migration in chains and migration of individual neuroblasts along radially oriented blood vessels in the CC.
Taken together, these results demonstrate that neuroblasts in the neonatal CC and cortex are closer to blood vessels than predicted by chance and migrate radially to the cortex using vasculature as a scaffold.
Neuroblasts Migrate Radially along Astrocytes Lining Blood Vessels
We next examined the relationship between migrating neuroblasts and blood vessels in the CC and deep cortical layers of 5HT3-EGFP mice aged P12 at the ultrastructural level. We observed EGFP+ neuroblasts in the vicinity of blood vessels in the CC and lower cortical layers. On electron micrographs, DAB-stained EGFP+ cells featured a narrow rim of cytoplasm and an irregularly shaped nucleus, often with prominent invaginations. Of note, although in close proximity to blood vessels, the migrating cells were always separated from endothelial cells by astrocytic processes (Fig. 4A–D). In addition to newborn cells generated perinatally and postnatally in the SVZ, a fraction of EGFP-expressing cells are generated embryonically in the caudal ganglionic eminence (Inta et al. 2008; Vucurovic et al. 2010). To verify that neuroblasts migrating along astrocytes surrounding blood vessels in the CC and cortex are indeed generated postnatally, we injected a retroviral vector in the SVZ of wild-type mouse pups (P4) to selectively express GFP in dividing newborn cells. Five to seven days after retroviral infection, cells in migration through the CC and lower cortical layers could be identified using anti-GFP immunostainings (Fig. 4E). GFP+ cells were found in close proximity to blood vessels in the CC and lower cortical layers. Glial cells are also generated in the SVZ during the early postnatal period but migrate along radial glia. Therefore, retrovirally labeled GFP+ cells found next to blood vessels are unlikely to be immature glial cells (Zerlin and Goldman 1997; Kakita and Goldman 1999; Marshall and Goldman 2002). In line with our results from young 5HT3-EGFP mice, retrovirally labeled cells were separated from endothelial cells by 1–2 layers of astrocytic processes in all electron micrographs examined (Fig. 4F). Therefore, neuroblasts do not physically contact blood vessels during radial migration but are adjacent to astrocytic processes that tightly enwrap blood vessels. Similar results were reported in the adult OB, where blood vessels are always surrounded by a thin sheet of astrocyte process along which neuroblasts migrate (Bovetti et al. 2007). In the adult RMS, there is also evidence for astrocyte processes lining blood vessels, but some neuroblasts in direct contact with endothelial cells were also observed (Snapyan et al. 2009; Whitman et al. 2009). Thus, ultrastructural analysis confirms that postnatally generated neuroblasts en route to the cortex migrate in proximity to blood vessels and suggests that the association of blood vessels, astrocytes, and migrating newborn neurons is widespread throughout postnatal development and adulthood.
The Structure of Vasculature and Glial Sheath Differ between Neonates and Adults
Migration patterns differ in young and adult mice. While in neonates, many neurons leave the RMS before reaching the OB and migrate radially to the cortex, in adult mice, virtually all SVZ-derived neuroblasts migrate to the OB. In adult animals, astrocytic processes enwrap chains of migrating neuroblasts. It has been suggested that this glial sheath may isolate neuroblasts from the surrounding parenchyma, provide molecular cues promoting migration to the OB, and mechanically prevent cells from migrating out of the RMS (Lois et al. 1996). Blood vessels, on the other hand, are aligned with the RMS in adult mice (Snapyan et al. 2009; Whitman et al. 2009). Since astrocytes and vasculature are also closely associated to neuroblasts in the neonatal RMS and CC, we investigated whether changes in the structure of vasculature and glia may underlie different patterns of migration in young and adult animals. We examined sagittal brain sections from neonatal and adult 5HT3-EGFP mice immunostained with antibodies to PECAM-1 and to glial fibrillary acidic protein (GFAP) to visualize blood vessels and astrocytes, respectively.
The structure of vasculature in the CC differed between neonates and adults. PECAM-1-stained blood vessels appeared to be thicker and to cover a larger proportion of CC in neonates. To quantify the length, thickness, and density of blood vessels in the CC, we reconstructed their morphology in z-projections from confocal stacks. Confocal image stacks were taken from the dorsal part of RMS. The RMS was visible in all stacks, in order to ensure that the same brain area was examined in each animal (Fig. 5A,B). We found that the total length of blood vessel normalized to the surface of z-projection was not significantly different in neonates and adults, although there was a trend toward higher length in adults (11.1 × 10−2 ± 0.76 × 10−2 μm blood vessel/μm2 of z-projection, n = 46 stacks from 5 neonates and 13.2 × 10−2 ± 0.61 × 10−2 μm blood vessel/μm2 of z-projection, n = 47 stacks from 5 adult mice, respectively; P = 0.051; data not shown). The density of blood vessels (surface occupied by blood vessels per surface of z-projection) was 0.32 ± 0.02 μm2 blood vessel/μm2 of z-projection in neonates and hence significantly higher than in adults where it was 0.25 ± 0.02 μm2 blood vessel/μm2 of z-projection (n = 46 and n = 47, respectively; P < 0.03). This was correlated with higher thickness of blood vessels in neonates, as evidenced by the higher surface/length ratio of blood vessels in neonatal mice (29.7 ± 1.2 μm, n = 46 and 18.1 ± 1 μm, n = 47, respectively; P < 2 × 10−9; Fig. 5C). These results indicate that increased radial migration in the CC of neonatal mice correlates with thicker blood vessels and higher ratio of vasculature surface to surface of CC. Thicker blood vessels in neonates may provide a larger scaffold for migration out of the RMS and therefore promote radial cortical migration in young animals. In support of this hypothesis, we found a significant correlation between the number of EGFP+ cells directly apposed to a blood vessel and blood vessel thickness (Fig. 5D).
In 4-day-old mice, GFAP could not be detected in immunohistochemical experiments, in line with previous results from neonatal rats and mice (Peretto et al. 2005). We performed anti-GFAP immunostainings in sagittal brain sections of 5HT3-EGFP mice aged P8. At that age, antibodies to GFAP labeled a network of small intermingled astrocytic processes in the CC. Interestingly, far fewer labeled processes were observed in the RMS, where EGFP-positive neuroblasts did not appear to be associated with astrocytic processes (Fig. 6). These results indicate that the glial sheath is not formed in young mice, confirming previous data (Peretto et al. 2005). In contrast, GFAP immunoreactivity in 8-week-old mice was associated with thicker astrocytic processes clearly separated from each other in the CC. At this stage, numerous GFAP-positive processes could be observed in close association with EGFP-positive neuroblasts in the RMS (Fig. 6), although not forming a clear border delineating the RMS, which corroborates a recent description of the glial sheath in adult rodents (Whitman et al. 2009). Therefore, it is possible that in young animals the absence of proper glial sheath enables neuroblasts to migrate radially along blood vessels, while in adult animals, the glial sheath confines migrating cells in the RMS. To provide further evidence for the role of glia in the guidance of neuroblasts from the RMS, we performed in vitro migration assays.
Previous in vitro studies demonstrated that astrocytes promote the migration of SVZ-derived neuroblasts by releasing yet unidentified factors (Mason et al. 2001; García-Marqués et al. 2010). To determine whether astrocytes affect migration of 5HT3-EGFP neuroblasts, we plated SVZ-derived explants from 5HT3-EGFP mice on cocultures of RMS/CC astrocytes and HEK cells. Our results demonstrate that EGFP+ neuroblasts migrated in close proximity to glial cells also when other cells were present in the culture (Fig. 7A–D). Therefore, astrocytes constitute a preferred substrate for migration rather than providing only a physical substrate.
Since neuroblasts migrate radially in the CC of young mice but not of adult mice, we asked whether adult RMS/CC astrocytes might have higher migration-promoting effect on neuroblasts as compared with astrocytes derived from early postnatal period. To verify this possibility, we cultured SVZ-derived explants from 5HT3-EGFP neonatal mice on a mixture of astrocytes derived from either the RMS/CC of early postnatal mice or the RMS/CC of adult mice. Both types of astrocytes attracted 5HT3-EGFP neuroblasts in a similar manner (Fig. 8A,B), suggesting that differences between young and adult astrocytes do not account for the difference in migration in the CC of young and adult mice.
The presence of glial cells within and around the adult RMS, often referred to as glial sheath, was described in previous studies but their functional role has remained elusive. It was speculated that they might provide a permissive environment for migration and directional cues or conversely confine migrating cells to the RMS (Lois et al. 1996). Taken together, our results suggest that the glial sheath forms a tract that guides neuroblasts toward the OB. In the absence of a proper glial sheath in young animals, other chemical cues in the CC and/or cortex, together with a high density of astrocytes and vasculature, might promote radial migration out of the RMS.
We demonstrated that neuroblasts from the postnatal RMS use blood vessels located in the CC as a scaffold along which they migrate radially to the cortex. Furthermore, we showed that postnatally generated neuroblasts do not migrate directly along blood vessels but are physically separated from endothelial cells by 1–2 layers of astrocytic processes that tightly enwrap blood vessels. In combination with the presence of a larger RMS in young mice (Peretto et al. 1999, 2005; Tramontin et al. 2003), we identified here 2 additional factors that may contribute to the existence of radial cortical migration in newborn and juvenile mice: 1) higher density of blood vessels in the CC of neonatal mice, providing a larger scaffold for migration than in adult animals and 2) absence of glial sheath in young mice. These results show the existence of a novel type of radial cortical migration of newborn neurons, which does not depend on radial glial cells but vasculature and astrocytes as a scaffold.
Link between Vasculature, Neurogenesis, and Neuroblast Migration
There is increasing evidence for a link between vasculature, neurogenesis, and neuroblast migration. Seminal studies showed that in brain regions associated with continuous generation of new neurons, angiogenesis is tightly linked to neurogenesis. In the subgranular zone, a neurogenic niche in adult rats, newborn dividing cells were found in clusters grouped around small blood vessels (Palmer et al. 2000). In the higher vocal center of the adult songbird, testosterone-induced angiogenesis promotes neurogenesis (Louissaint et al. 2002). Recent studies in adult mice did not find evidence of significant angiogenesis in the SVZ. However, unique characteristics of vasculature in the SVZ foster neurogenesis. Thus, stem cells and transit-amplifying cells are in close contact to SVZ vasculature lacking astrocyte end-feet and pericyte coverage (Tavazoie et al. 2008). Proliferation of SVZ progenitors can be affected by perturbing their adhesion to endothelial cells using an α6 integrin–blocking antibody, which demonstrates that interaction with vasculature is crucial for SVZ neurogenesis (Shen et al. 2008). Not only does the generation of new neurons depend on the presence of blood vessels, but it is now clear that newly generated neurons can maintain this close association with blood vessels on their way to their final destination in the brain. This was first observed in the adult OB, where neuroblasts generated in the SVZ migrate radially along astrocytic processes that enwrap blood vessels (Bovetti et al. 2007), and more recently in the adult RMS (Snapyan et al. 2009; Whitman et al. 2009). Thus, our results add to the body of evidence suggesting that vasculature not only promotes neurogenesis but also constitutes a scaffold for neuronal migration. Studies demonstrating migration of neuroblasts along blood vessels have been conducted on adult-generated neuroblasts, which give rise mainly, though not exclusively, to GABAergic interneurons. Together with our results, this may support the idea that immature neurons with a GABAergic fate may preferentially migrate along blood vessels. However, 2 recent studies show that neuroblasts expressing Tbr2, a marker of intermediate progenitor cells that give rise to pyramidal cells, are located in closer proximity to vasculature than predicted by chance in the embryonic telencephalon (Javaherian and Kriegstein 2009; Stubbs et al. 2009). Progenitor cells divide near vascular branch point, suggesting that vasculature contributes to the existence of a neurogenic niche embryonically. Neuroblast neurites were found along labeled blood vessels in the SVZ and VZ, indicating that migration might be influenced by embryonic vasculature. Therefore, the influence of vasculature on neuronal progenitor migration may be widespread and not restricted to a particular neuronal subtype or to the postnatal brain.
Specific Features of Vasculature-Guided Radial Cortical Migration
The novel mode of migration described here is distinct from previously delineated types of neural migration toward and within the cortex. In brief:
During early embryonic development, glutamatergic cells generated in the VZ use somal translocation to migrate radially and form the preplate (Miyata et al. 2001; Nadarajah et al. 2001; Nadarajah and Parnavelas 2002).
In the embryonic brain, immature GABAergic interneurons migrate tangentially from the ganglionic eminence to the developing cortex. Several lines of evidence suggest that migrating interneuron precursors may be able to move from one substrate to another. Immature interneurons have been show to migrate along TAG-1-expressing axons of the developing corticofugal system in the intermediate zone (Denaxa et al. 2001; Morante-Oria et al. 2003). At the end of their tangential journey, interneuron precursors migrate radially along radial glia in the developing cortex (Elias et al. 2010).
In mice, the perinatal SVZ is a secondary source of cortical astrocytes and oligodendrocytes (Kakita and Goldman 1999; Marshall and Goldman 2002). SVZ-generated astrocyte and oligodendrocyte precursors in radial migration to the cortex have been found in the proximity of vimentin-positive radial glia, suggesting that radial glia may also guide their migration in the forebrain shortly after birth. Immature astrocytes guided by radial glia first contact blood vessels with their end-foot and progressively wrap around them (Zerlin and Goldman 1997). These astrocytes might in turn provide the physical substrate for the radial migration of SVZ-derived neuroblasts reported here.
Glioma cells migrate along the abluminal surface of blood vessels (Farin et al. 2006). Thus, glioma cells do not migrate on astrocytic processes lining blood vessels but instead insert their process between the blood vessel and the perivascular astrocyte end-feet to migrate in direct contact with endothelial cells.
Thus, the mechanisms of migration described in the present report share some characteristics of other types of cell migration toward or within the cortex, in particular, the saltatory migration of small immature cells with a leading process but differ from all other modalities in that postnatally generated interneuron progenitors use astrocytes enwrapping vasculature as a substrate to migrate radially. This mechanism is strikingly similar to the vasculature-guided migration of neuroblasts described in the OB and RMS of adult mice, where migrating neuroblasts are also separated from endothelial cells by astrocytic processes (Bovetti et al. 2007; Snapyan et al. 2009; Whitman et al. 2009). Interestingly, migration of neuroblasts along blood vessels in the OB occurs in a single direction toward the external part of the OB (Bovetti et al. 2007). Our results show that in the CC of young pups, neuroblasts migrate radially not only toward the cortex but also backwards to the RMS.
Molecular Cues Fostering Vasculature-Guided Migration
It is likely that chemoattractants and chemorepellents, possibly secreted by astrocytes and/or endothelial cells, are involved in the guidance of SVZ-generated cells toward the cortex. Brain-derived neurotrophic factor (BDNF), in particular, may contribute to vasculature-guided radial migration. BDNF secreted by endothelial cells is known to promote neuronal recruitment and survival of newborn neurons (Leventhal et al. 1999). It has been shown recently that BDNF binds to p75NTR on neuroblasts and fosters migration. In turn, GABA released by migrating neuroblasts elicits insertion of TrkB receptors at the surface of astrocytes, which traps free extracellular BDNF. Neuroblasts deprived of BDNF subsequently enter the stationary phase (Snapyan et al. 2009). Since we observed a close relationship between neuroblasts, astrocytes, and blood vessels in the CC of young mice, it is possible that this mechanism also controls the switching between migrating and immobile modes during radial migration. Vascular production of stromal-derived factor 1 (SDF1) and angiopoietin 1 (Ang1) have also been reported to promote neuroblast migration to the periinfarct site following stroke (Ohab et al. 2006). However, it is not clear whether these factors contribute to neuronal migration in nonpathological conditions.
Other factors may contribute to the control of radial migration in neonates, independently of vasculature. In vitro studies showed that prokineticin 2 and GDNF are secreted by the OB and have the ability to attract neuroblasts generated in the SVZ (Ng et al. 2005; Paratcha et al. 2006). However, surgical removal of the OB in vivo does not prevent neuroblast migration toward the bulb (Jankovski et al. 1998; Kirschenbaum et al. 1999), indicating that secretion of prokineticin and GDNF in the OB is not sufficient to account for tangential migration and that other chemical cues may be involved in the confinement of neuroblasts in the RMS. Slit, in particular, is a chemorepellant secreted from the choroid plexus and septum that forces neuroblasts to remain in the adult RMS (Wu et al. 1999; Sawamoto et al. 2006). Newborn neurons also express Slit-1 (Nguyen-Ba-Charvet et al. 2004) and signal to astrocytes expressing Robo receptors (Kaneko et al. 2010). This interaction is crucial for the formation and maintenance of the glial sheath in adult mice (Kaneko et al. 2010). Changes of Slit-1 and/or Robo receptor expression during development might thus alter the structure of the RMS and of the glial sheath.
Following stroke in the striatum or cortex of mice, neurogenesis is increased in the SVZ and SVZ-derived neuroblasts are recruited to the site of insult (Jin et al. 2001; Arvidsson et al. 2002; Kreuzberg et al. 2010). Analysis of postmortem brain tissue also revealed that neurogenesis is increased in the SVZ of adult ischemic patients (Macas et al. 2006). Neuroblasts en route to the poststroke area were found in close proximity to blood vessels (Yamashita et al. 2006; Zhang et al. 2009). Migration along blood vessels was directly investigated by time-lapse imaging in brain slices after ischemic injury, proving that neuroblasts generated in the SVZ that are derouted after stroke use blood vessels as a scaffold (Kojima et al. 2010). Thus, stroke in the adult brain may recapitulate a process that occurs physiologically in the young animal. Following stroke induced by medial cerebral artery occlusion, only 0.2% of dead striatal neurons are replaced by SVZ-generated new neurons 6 weeks after insult (Arvidsson et al. 2002). Improving migration to the infarct site might increase this proportion substantially. This would require a better understanding of the mechanisms governing neuroblast migration. Investigating in more detail the molecular determinants of blood vessel–guided migration to the cortex in neonatal mice may help understand the basis for neuroblast migration in the diseased brain.
Supplementary material can be found at: http://www.cercor.oxfordjournals.org/
C.L.M. was supported by a Postdoctoral Grant from Medical Faculty Heidelberg. J.A. was supported by a Marie Curie Postdoctoral Fellowship from European Union (EU). H.M. was supported by the Schilling Foundation, the Sonderforschungsbereich 488 (project D3, ME 1985/1-1), and the EU Synapse (grant LSHM-CT-2005-019055).
We thank U. Amtmann, R. Hinz-Herkommer, and I. Preugschat-Gumprecht for technical assistance, F. H. Gage for the retroviral vector, and E. Brinkmann and A. Ortlieb for help with Neurolucida reconstructions and quantification. Conflict of Interest : None declared.