The subventricular zone (SVZ) of the lateral ventricle is the major site of neurogenesis in the adult brain. Neuroblasts that are born in the SVZ migrate as chains along the rostral migratory stream (RMS) to the olfactory bulb. Little is known about the mechanisms that control interactions between neuroblasts during their migration. Here we show that migrating neuroblasts express β1integrins and that the integrin ligand laminin is localized to cell chains. Using genetically modified mice and time-lapse video recordings of SVZ explants, we demonstrate that β1 integrins and laminin promote the formation of cell chains. Laminin also induces the aggregation of purified neuroblasts. We conclude that the formation of cell chains in the RMS is controlled in part by β1 integrins via binding to laminin. In addition, we provide evidence that β1 class integrins are required for the maintenance of the glial tubes and that defects in the glial tubes lead to the ectopic migration of neuroblasts into the surrounding tissue.
Neurons are generated in the adult mammalian brain throughout life from neural stem cells that are localized in specialized niches (Alvarez-Buylla and Lim, 2004; Kempermann et al., 2004). The major site of neurogenesis in the adult mammalian brain is the subventricular zone (SVZ) of the lateral ventricle (Alvarez-Buylla and Lim, 2004; Lledo et al., 2006). Slow dividing stem cells (type B cells) in the SVZ give first rise to transit amplifying cells (type C cells), which generate fast dividing neuroblasts (type A cells) that migrate tangentially within glial tubes along the rostral migratory stream (RMS) to the olfactory bulb (OB). In the OB, the cells switch to a mode of radial migration and differentiate into interneurons of the granular and periglomerular layers (Luskin, 1993; Lois and Alvarez-Buylla, 1994).
Several secreted signaling molecules control the directional migration of neuroblasts in the RMS to ensure that appropriate numbers of cells arrive, in a timely manner, in appropriate cell layers of the OB to replace preexisting interneurons. These include diffusible chemoattractants such as netrin-1, prokineticin-2, and glial cell line-derived neurotrophic factor, as well as chemorepellents such as slit1/2 (Wu et al., 1999; Murase and Horwitz, 2002; Nguyen-Ba-Charvet et al., 2004; Ng et al., 2005; Paratcha et al., 2006). The CSF flows parallel to the direction of migration, and its flow is required for the formation of gradients of guidance molecules in the RMS (Sawamoto et al., 2006), providing a potential mechanism controlling directional neuroblast migration. In addition, receptor tyrosine kinases of the Eph and ErbB families and their respective ephrin and neuregulin ligands have been implicated in controlling chain migration in the RMS (Conover et al., 2000; Anton et al., 2004; Ghashghaei et al., 2006). Subsequent to arrival in the OB, signaling molecules such as reelin, tenascin R, and prokineticin-2 induce detachment of neuroblasts from cell chains (Hack et al., 2002; Saghatelyan et al., 2004; Ng et al., 2005), controlling the switch from tangential to radial migration.
Ultimately, secreted signaling molecules must control the activity of cell surface receptors that regulate interactions of neuroblasts with each other and with cues in their environment. The adjustment of these interactions is essential for directional movement of cell chains at defined speed, dispersal of cell chains on arrival within the OB, and migrational arrest of the cells in their final target areas. The neural cell adhesion molecule (NCAM) is expressed on type A neuroblasts and, although not essential for chain formation, controls interactions of neuroblasts with the stationary environment through which the cells migrate. Defects in this process affect the motility of cell chains and the size of the OB (Tomasiewicz et al., 1993; Cremer et al., 1994; Ono et al., 1994; Hu et al., 1996; Chazal et al., 2000). It has also been proposed that neuroblast motility is controlled by integrins. Integrins are heterodimeric cell surface molecules consisting of one α and one β subunit that bind to extracellular matrix (ECM) components and counter receptors on adjacent cells. In vertebrates, genes for 18 α and 8 β subunits have been identified. These subunits assemble into 24 different receptors that fall into subfamilies such as the β1 integrins, which form by heterodimerization of the β1 subunit with 12 α subunits (Hynes, 2002). β1 integrins have been reported to control the motility of neuroblast, but the findings have been conflicting. One study has reported that β1 integrins are expressed in the RMS until postnatal day 10 (P10) and that antibodies to β1 slow down migration before but not after P10 (Murase and Horwitz, 2002). A second study has provided evidence that the integrin α6β1 is expressed even in the adult RMS and that injection of β1 antibodies disrupts the adult RMS (Emsley and Hagg, 2003). In both studies, it has remained unclear whether the antibodies affected the migration of neuroblasts directly or indirectly through effects on other cells or other aspects of cell behavior such as proliferation and differentiation.
Despite the identification of receptors and signaling molecules that control the migration of neuroblasts, little is known about the mechanisms regulating the interactions between neuroblasts during chain migration. Our findings now reveal an unexpected function for β1 integrins in this process. We show here that neuroblasts express integrins containing the β1 and β5 subunits. Using genetic and cell biological approaches, we demonstrate that β5 integrins are dispensable for chain migration, whereas β1 integrins promote cell–cell interactions that link neuroblasts into chains. The β1 integrin ligand laminin is recruited to the cell surface of migrating neuroblasts and induces chain formation in SVZ explants and the aggregation of purified neuroblasts. We conclude that β1 integrins and their laminin ligands promote the formation of cell chains in the adult RMS. We also provide evidence that β1 class integrins are required for maintaining the integrity of the glial tubes.
Materials and Methods
Itgb1–CNSko mice were generated by crossing Itgb1flox/flox mice with Itgb1flox/+nestin–Cre+/− mice (Graus-Porta et al., 2001). Itgb5–null mice, and mice carrying mutations in the Lama2 (laminin alpha 2) and Lama4 (laminin alpha 4) genes have been described (Sunada et al., 1994; Xu et al., 1994; Patton et al., 2001). Mutants and control littermates were used for all experiments.
Fluorescence-activated cell sorting.
The SVZ and the distal part of the RMS at the level of the OB were dissected in cold HBSS without Ca2+ and Mg2+ (Invitrogen, Carlsbad, CA) from four wild-type C57BL/6 mice at P60. Cells were dissociated for 6 min at 37°C with 5 mg/ml trypsin (Worthington, Lakewood, NJ) and 0.1% Dnase I (Worthington), dissociated mechanically in sorting buffer (HBSS without Ca 2+ and Mg2+ without phenol red, 1 mm EDTA, 25 mm HEPES, pH 7.0, and 1% FBS), and filtered through a 40 μm nylon cell strainer (BD Biosciences, San Jose, CA). Cells were stained with an anti-polysialic acid (PSA) antibody for 15 min at 4°C in sorting buffer, rinsed, and then stained with an anti-mouse IgM FITC-conjugated antibody (Jackson ImmunoResearch, West Grove, PA) for 15 min at 4°C. PSA-positive cells were sorted directly into the lysis buffer using a fluorescence-activated cell sorting (FACS) Vantage DiVa (BD Biosciences) and processed for RNA purification. Unstained cells and cells stained with the secondary antibody only were used as negative controls. Two independent experiments were performed for both regions (SVZ and RMS) to collect two samples of 30,000 cells from the SVZ and two samples of 100,000 cells from the RMS.
Gene expression sample and microarray processing.
Total RNA was examined on an Agilent (Palo Alto, CA) Bioanalyzer, and the 28S/18S ratio exceeded 2.0 for all samples. RNA quantity was estimated from this analysis. Total RNA (20 ng) was reverse transcribed and amplified with three rounds of amplification as described previously (Scherer et al., 2003). The in vitro transcription kit used for the final amplification was from Enzo Diagnostics (Farmingdale, NY). The same individual processed all reactions and arrays at the same time. Fragmented and labeled cRNA was hybridized to custom Affymetrix (Santa Clara, CA) arrays as described below. Arrays were washed using a custom Genomics Institute of the Novartis Research Foundation microarray washer for the low-stringency washes and otherwise used standard Affymetrix protocols. Arrays were scanned with a G2500A (Affymetrix) scanner. MAS5.0 (Affymetrix) quality control measures of “Average Background” and “Percentage Present” were examined from .CHP and .RPT files produced by default setting of the MAS5.0 algorithm. All arrays passed the criteria of no more than two SDs from the mean for both of these measures simultaneously.
Measurement data and specifications.
Probe values were condensed from probe sets using MAS5 software (Affymetrix). Arrays were globally scaled to an average intensity value of 200. Detectable expression was defined as at least two times above the average intensity (across all arrays) for the average value in at least one experimental group. Significant fold change values were defined as being at least a fold change of 1.5 and, when the coefficient of variation, across replicates, was less than a value of 0.6 in both experimental groups.
A custom mouse microarray, GNF1M, (Affymetrix) was used for these experiments. A description of the contents of this array has been described previously (Su et al., 2004).
Quantitative reverse transcription-PCR analysis.
Three brains from 2-month-old C57BL/6 mice were dissected, and coronal sections (∼500 μm thick) were prepared with a scalpel. The SVZ and granule cell layer from the OB were dissected, total RNA was isolated using RNeasy kit (Qiagen, Valencia, CA), and RNA (1 μg) was retrotranscribed into cDNA using oligo-dT with Superscript Strand II cDNA synthesis kit (Invitrogen). Real-time quantitative PCR was performed on Chromo4 (MJ Research via Bio-Rad, Hercules, CA) using SYBR green chemistry (Applied Biosystems, Foster City, CA) as described previously (Teyssier et al., 2005). Primers were as follows: glyceraldehyde-3-phosphate dehydrogenase (GAPDH) forward, 5′-TCAACGGGAAGCCCATCA-3′; GAPDH reverse, 5′-CTCGTGGTTCACACCCATCA-3′; Itgb1 forward, 5′-TGTTGGTCAGCAACGCATATC-3′; Itgb1 reverse, 5′-TCCCATCTCCAGCAAAGTGAA-3′; Itgb3 forward, 5′-CAACGTCCTCCAGCTCATTGT-3′; Itgb3 reverse, 5′-TTCCGGCAGGTCACGTACTT-3′; Itgb5 forward, 5′-GCTGCTGTCTGCAAGGAGAAG-3′; Itgb5 reverse, 5′-CACCCAGTTTTCCATCCAGTG-3′; Itga1 forward, 5′-CGTCCTAGATGGCTCCAACAG-3′; Itga1 reverse, 5′-CTATTTTGTTTGCTGCCACCA-3′; Itga2 forward, 5′-GCTTGATGCGGACGGACATT-3′; Itga2 reverse, 5′-CACACGCAGATCCAAAGGGT-3′; Itga3 forward, 5′-CTACTCCCTGAGCGGGAAGAT-3′; Itga3 reverse, 5′-GTACAAAGCGCAGGGTCCAAC-3′; Itga6 forward, 5′-GAATTGACCTCCGCCAGAAGT-3′; Itga6 reverse, 5′-ATACTTTGGCTCGGAACCTTG-3′; Itga7 forward, 5′-CGATGGACGCTTGGTCTGTG-3′; and Itga7 reverse, 5′-CTGATGCCTGAGGTCATCCA-3′. Relative mRNA levels for each integrin subunit was determined using the comparative threshold cycle method (Applied Biosystems) and have been normalized to GAPDH mRNA levels. Results are expressed as number of copies of mRNA relative to 1000 copies of GAPDH mRNA. The experiments have been performed three times in duplicate for each set of genes. Values are given as mean ± SEM. A Student's t test was performed.
Histology and immunohistochemistry.
The analyses were performed with at least three mutant mice (Itgb1–CNSko, Itgb5–null, laminin α2/α4 double mutants) and control littermates. LacZ and Nissl stainings were performed as described previously (Graus-Porta et al., 2001). For immunohistochemistry, mice were perfused with 4% paraformaldehyde in PBS. Brains were dissected, postfixed overnight at 4°C and embedded in 3.5% agarose and 8% sucrose. Sagittal vibratome sections (50 μm) were prepared and stained as floating sections in 48 well plates. Sections were incubated with PBS containing 10% goat serum (Invitrogen) (blocking buffer) for 1 h at room temperature (RT) and incubated with primary antibody in blocking buffer for 1 h at RT or overnight at 4°C. For some antibodies, 0.5% Triton X-100 was included to permeabilize the cells. Sections were rinsed with blocking buffer and incubated for 1 h at RT with secondary antibodies coupled to fluorochromes (Invitrogen). Nuclei were counterstained for 10 min at RT with 4′,6′-diamidino-2-phenylindole (DAPI) (1 μg/ml; Sigma, St. Louis, MO). Sections were rinsed and mounted with ProLong Gold antifade mounting medium (Invitrogen). Stainings were analyzed with a confocal microscope (Fluoview-LSM; Olympus Optical, Melville, NY) or a deconvolution microscope (Deltavision; Applied Precision, Issaquah, Washington). The following antibodies were used: anti-β1 integrins rabbit polyclonal (Graus-Porta et al., 2001), GFAP rabbit polyclonal (DakoCytomation, Glostrup, Denmark), PSA-NCAM mouse IgM (kindly provided by G. Rougon, Université de la Mediterranée Aix-Marseille 2, Marseille, France), bromodeoxyuridine (BrdU) mouse IgG (BD Biosciences), laminin rabbit polyclonal (kindly provided by L. Sorokin, University of Uppsala, Uppsala, Sweden), calretinin, polyclonal rabbit (Chemicon, Temecula, CA), caspase 3 polyclonal rabbit (Asp175; Cell Signaling Technology, Beverly, MA). Staining with mouse anti BrdU IgG was performed as described previously (Belvindrah et al., 2002). Mice were administered with BrdU (10 mg/ml in PBS, 50 mg/kg of body weight). Mice were analyzed either 1 h later (to study cell proliferation) or after 7 or 21 d after the single BrdU pulse (to study cell migration). To test cell proliferation, nuclei that stained positive for BrdU were counted in the SVZ, using a standardized counting area: BrdU-positive nuclei from three adjacent sections (at the same anatomical level on the side of the SVZ bordering the lateral ventricle) were counted from three Itgb1–CNSko mice and three control littermates. Staining with DAPI did not reveal any changes in cell density. To test cell migration, nuclei that stained positive for BrdU in the proximal and distal RMS, and the granule cell layer in the OB (noted as regions 1, 2, and 3 in Fig. 5a) were counted in a 3600 μm2 square in three adjacent sections from three adult Itgb1–CNSKo and littermate controls for each experiment (7 and 21 d chase). All values are given as mean ± SEM. A Student's t test was performed.
OB volume quantification.
To estimate the OB volume, serial coronal sections were prepared with a Vibratome (Leica, Nussloch, Germany) after intracardiac perfusion with paraformaldehyde (PFA) and postfixation of brains overnight at 4°C. A total of three Itgb1–CNSko mice and three control littermates were processed. Serial sections were stained with Nissl as described previously (Belvindrah et al., 2002). The surface of the OB was determined using MetaMorph Software (Molecular Devices, Palo Alto, CA). We therefore measured the OB area in each section (A1 − An) and determined the final volume by summing up the surface area and multiplying it with the thickness of the sections [∑ (A1… An) × 50 μm; result expressed as cubic millimeters]. Values are given as mean ± SEM. A Student's t test was performed.
Mice were perfused with 2% PFA/2% glutaraldehyde in PBS and postfixed overnight at 4°C. Brains were cut into 50-μm-thick sections. Sections were washed 30 min in 0.1 m Na cacodylate buffer, pH 7.3, and postfixed for 1 h with 1% OsO4 in 0.1 m Na cacodylate buffer. Sections are dehydrated with increasing concentration of ethanol (50, 70, 90, and 100%). Samples are incubated in propylene oxide and impregnated overnight in Epon/Araldite resin before polymerization at 60°C. Thin sections were stained with lead citrate before observation in the electron microscope.
Matrigel time-lapse video microscopy.
Matrigel experiments were performed as described previously (Wichterle et al., 1997). Briefly, P0–P4 mouse brains were dissected, coronal sections (∼500 μm thick) were prepared with a scalpel, and the SVZ was carefully dissected out from the wall of the anterior horn of the lateral ventricle and cut into pieces of 100–300 μm in diameter. The explants were mixed into 70% of Matrigel (BD Biosciences), 2.4 mg/ml collagen type I (BD Biosciences), or 2.4 mg/ml collagen type I plus 300 μg/ml laminin-1 (Sigma), kept at 37°C/5% CO2 for 7 min (for Matrigel experiments) and 1 h (for collagen experiments) without medium for polymerization, and then cultivated for 48 h at 37°C/5% CO2 in Neurobasal medium supplemented with B27 (Invitrogen) and penicillin/streptomycin. After 48 h in culture, explants were examined using a microscope, pictures were taken using a digital camera (Coolview; Nikon, Tokyo, Japan), and distances were measured using MetaMorph software as described previously (Hu and Rutishauser, 1996; Hu et al., 1996; Chazal et al., 2000). Migration distance was calculated as the distance in micrometers between the edge of the explant and the border of the cell migration front. The experiments were performed three times independently, with at least three explants per condition. For each explant, eight measurements were performed, and the data were pooled and are presented as cumulative frequency plots. For live video imaging, explants were cultivated in a humidified chamber at 37°C/5% CO2, and pictures were taken every 5 min using a digital camera (Hamamatsu, Shizouka, Japan) and analyzed with MetaMorph Software.
Cell aggregation assays.
The SVZ and rostral part of the RMS from 10 adult mice was dissected, and PSA–NCAM positive cells were isolated by FACS (see above). On average, 100,000 cells were recovered and used in cell aggregation assays. Cells were washed in 5 ml of buffer H (20 mm HEPES, 140 mm NaCl, and 4 mm KCl), resuspended in buffer H, and incubated in hanging droplets. When indicated, buffer H was supplemented during the aggregation assay with 2 mm CaCl2 and 100 μg/ml laminin-1 (Sigma). The hanging droplets were incubated for 20 min at 37°C/5% CO2, turned over, and incubated for 2 h at 37°C/5% CO2. Pictures were taken with a digital camera (Coolpix 990; Nikon). The number of single cells and the number of cells in aggregates were quantified. All values are derived from three independent experiments and are given as mean ± SE.
β1 and β5 integrins are expressed in migrating neuroblasts in the RMS
To identify integrin subunits that are expressed in migrating type A neuroblasts, we isolated by FACS PSA-NCAM-positive cells from two distinct areas of the CNS (Fig. 1a). First, we isolated cells from the SVZ, in which cell proliferation is high and cells initiate migration. Second, we isolated cells from the distal RMS close to the OB, in which cells are actively migrating. RNA was prepared and subjected to microarray analysis (for statistical analysis, see Material and Methods). As expected, PSA-NCAM-positive cells in the SVZ expressed markers for cell proliferation (e.g., PCNA) and for type A neuroblasts [e.g., Tau, Dcx (doublecortin), Dlx2 (distal-less homeobox 2), and mCD24 (the mouse cluster of differentiation 24)] (Fig. 1b). PSA-NCAM-positive cells from the distal part of the RMS showed a similar expression pattern, but some genes that are associated with proliferation in the SVZ (PCNA, mCD24, and Dlx2) were expressed in reduced amounts, whereas markers for differentiation such as GABA receptor 2 were upregulated (Fig. 1b). Genes specific for hematopoietic precursors or glial cells were not expressed, further confirming the quality of the FACS-sorted material (data not shown).
We next analyzed the expression of integrin subunits (Fig. 1c,d). Of the eight integrin β subunits, only β1 and β5 were expressed at detectable levels (Fig. 1c). β1 can heterodimerize with the integrin α1–α11 subunits and with αυ, whereas β5 forms heterodimers with αυ only (Hynes, 2002). Most of these α subunits were detected in the SVZ and RMS (Fig. 1d). The expression of α1, α6, and α7 appeared to increase in the RMS when compared with the SVZ. These subunits heterodimerize with β1 to form laminin receptors (Hynes, 2002), suggesting that the expression of several laminin receptors is increased during migration.
To confirm the expression of integrin subunit genes in type A neuroblasts and to obtain more reliable quantitative data for gene expression levels, we performed quantitative reverse transcription-PCR. We focused the analysis on the β1, β3, and β5 subunits, as well as on those α subunits that heterodimerize with β1 to form functional laminin receptors (α1, α2, α3, α6, and α7) (Fig. 1e). Consistent with the results from microarray analysis, we observed expression of the integrin β1, β5, α1, α3, α6, and α7 subunits but very low levels of the β3 and α2 subunits. The data also revealed increased expression for the α6 and α7 subunits in the distal part of the RMS when compared with the SVZ.
To evaluate expression of β1 integrins in type A neuroblasts of adult mice at the protein level, we performed Western blots and immunohistochemistry experiments. β1 protein was detected in extracts from the microdissected SVZ, OB, and control cerebral cortex (Fig. 1f). In sagittal sections from P60 mice, β1 integrin expression was detected along the RMS from the SVZ to the OB (Fig. 2). Many β1-positive cells were assembled into chains and expressed PSA-NCAM (Fig. 2; higher magnification in inset in b), confirming their identity as migratory type A neuroblasts. β1 integrins was also expressed in some GFAP-positive glial cells along the RMS and in the corpus callosum (Fig. 2c). Unfortunately, antibodies against the murine integrin β5 subunit were not of high enough quality to determine its distribution (data not shown).
Defects in the RMS of mice lacking β1 but not β5 integrins in the CNS
To determine the function of the β1 and β5 integrins in the RMS, we took a genetic approach. The integrin β5 subunit gene (Itgb5) is not required for the viability of mice (Huang et al., 2000), allowing us to study the RMS in adult Itgb5–null mice. We analyzed three mutant mice by histology at different levels along the rostral migratory stream and within the OB. We could not detect a defect in the organization of the RMS and OB in the mutants (Fig. 3a,b and data not shown). We next analyzed integrin β1-deficient mice. Because mice with a null mutation in the integrin β1 gene (Itgb1) die during embryogenesis, we took advantage of a mouse line carrying an Itgb1flox allele (Graus-Porta et al., 2001) and a nestin–CRE mouse line (Tronche et al., 1999). We have shown that the nestin–CRE mouse line expresses CRE in neural progenitors in the developing CNS (Graus-Porta et al., 2001). To evaluate the activity of the CRE transgene during adult neurogenesis, we crossed nestin–CRE mice with a reporter mouse line that carries a Rosa26lacZflox transgene. In this mouse line, LacZ expression is activated by CRE recombination (Mao et al., 1999). Staining for LacZ confirmed that the majority of cells in the SVZ and RMS had undergone CRE-mediated recombination, and double staining for LacZ and PSA-NCAM confirmed that recombination occurred in most, if not all, type A neuroblasts (data not shown). Complete recombination in all cells in the RMS, including neuroblasts and glial cells was expected because the nestin gene is expressed in the precursors of neurons and glia in the developing and adult brain (Lendahl et al., 1990; Graus-Porta et al., 2001; Merkle et al., 2004; Kuo et al., 2006).
We next crossed Itgb1+/−nestin–CRE+/− mice with Itgb1flox/flox mice. To confirm that the Itgb1 gene was inactivated in the mutant offspring (referred to as Itgb1–CNSko mice), we stained sagittal sections from adult mouse brains with antibodies to integrin β1. As expected, β1 expression was abolished in the RMS of Itgb1–CNSko mice (Fig. 1g and data not shown). We next stained sections with Nissl. In control mice, neuroblasts in the RMS formed compact chains (Fig. 3c–e). In Itgb1–CNSko mice, the RMS was present, but it appeared less compact and the cells were no longer assembled into chains (Fig. 3c,d). This was particularly evident in high-magnification views (Fig. 3e). Together, our findings show that the RMS is severely perturbed in the absence of β1 but not β5 integrins.
Defects in the size of the OB and in cell migration in Itgb1–CNSko mice
The defect in the morphology of the RMS in Itgb1–CNSko mice suggests that neuroblasts may traverse the RMS less efficiently, which could affect the number of cells that reach the OB. In addition, cell proliferation or survival may be affected. We therefore analyzed the size and morphology of the OB in wild-type and mutant animals, and we analyzed cell proliferation, survival, and migration. First, we prepared serial sections of the OBs and determined its volume (Fig. 4). The OB in Itgb1–CNSko mice was significantly reduced in size (OB volume: wild type, 6.02 ± 0.14 mm3; Itgb1–CNSko, 4.61 ± 0.39 mm3), but the organization of different cell layers was not changed (Fig. 4a), suggesting that neuroblasts that reached the OB switched normally from tangential to radial migration and differentiated into interneurons. Next, we analyzed cell death and proliferation. Staining with antibodies to caspase 3 revealed only very few cells undergoing apoptosis, and there was no apparent increase in the number of caspase 3-positive cells in the mutant mice (data not shown). To evaluate cell proliferation, sagittal sections of the brain were prepared 1 h after injection with BrdU. The number of BrdU-positive cells in the SVZ, the main proliferative zone, was determined. Care was taken to quantify cells at the same histological level in wild-type and mutant mice. We did not observe a difference in the number of BrdU-positive cells in the SVZ of Itgb1–CNSko mice (BrdU-positive cells: wild type, 125.5 ± 9.1; Itgb1–CNSko, 125 ± 5.79). We conclude that defects in cell proliferation and apoptosis are primarily not responsible for perturbations in the RMS and OB of Itgb1–CNSko mice.
To analyze cell migration, we determined the distribution of BrdU-positive cells 7 and 21 d after a single BrdU pulse (Fig. 5). We quantified the number of BrdU-positive cells in the proximal and distal RMS and in the granule cell layer in the OB (noted as regions 1, 2, and 3 in Fig. 5a–c,e). At 7 d after the BrdU pulse, the vast majority of BrdU-positive cells in wild-type mice had entered the RMS but had not reached the OB (Fig. 5b,c). In Itgb1–CNSko mice, there was an approximate threefold reduction of BrdU-positive cells in the RMS, indicating that a reduced number of cells had migrated into the RMS (Fig. 5b,c). In addition, many BrdU-positive cells had migrated out of the RMS into the surrounding tissue (Fig. 5b, arrows). The absolute number of the ectopically migrating cells was difficult to quantify but likely contributed significantly to the reduction of BrdU-positive cells within the RMS. At 21 d, the majority of BrdU-positive cells in wild-type mice had left the RMS and reached the granule cell layer in the OB (Fig. 5e). In Itgb1–CNSko mice, reduced number of cells had reached the granule cell layer, and increased numbers of BrdU-positive cells remained in the RMS (Fig. 5e). This was also evident in coronal sections. Few BrdU-positive cells were present in the RMS of wild-type mice (Fig. 5d), whereas many had reached the granule cell layer (Fig. 5d, inset). In contrast, BrdU-positive cells were prominent in the RMS of Itgb1–CNSko mice (Fig. 5d) and few in the granule cell layer (Fig. 5d, inset). Together, these findings suggest that cell migration from the SVZ to the OB is impaired but not abolished in Itgb1–CNSko mice. Interestingly, staining of OB sections with calretinin antibodies revealed radially migrating cells, containing long leading process oriented perpendicular to the surface of the OB (Fig. 5f). Although we cannot exclude that β1 integrins may control some aspects of radial migration, the findings show that cells that have migrated through the RMS and reached the OB switch from tangential to radial migration and populate appropriate cell layers in the OB.
Formation of cell chains and the glial tubes
We next determined the cause for the defect in cell migration in Itgb1–CNSko mice. Type A neuroblasts in the RMS migrate as chains within glial tubes (Lois et al., 1996). The expression of β1 integrins in neurons and glia suggested that they control the formation/maintenance of cell chains and/or the glial tubes. Staining of sections with antibodies to PSA-NCAM demonstrated that cell chains were altered in the mutants. In sagittal sections, neuroblasts still seemed to cluster, but cells were not aligned into chains extending along the rostrocaudal axis of the RMS (Fig. 6a). In coronal sections, PSA-NCAM-positive cells in wild-type mice were visible as clusters, whereas cells in the mutants were scattered (Fig. 6b). Staining with GFAP revealed that the glial tubes were also disrupted in the mutants. In wild-type mice, GFAP-positive glial cells were in close apposition to the RMS, but they were more widely distributed in the mutants (Fig. 6c,d).
These findings were confirmed at the ultrastructural level on coronal sections of the RMS (Fig. 7a–c). Type A neuroblasts can be identified by their small cytoplasm and big dark nuclei occupying the majority of the cell body; astrocytes are characterized by their light cytoplasm containing dense bodies (Doetsch et al., 1997). In the RMS of control mice, type A neuroblasts were attached to each other and formed dense clusters that were often associated with one or two astrocytes (Fig. 7a,b). In Itgb1–CNSko mice, neuroblasts were dispersed and only occasionally formed contacts with each other (Fig. 7a,b), but they maintained contacts with astrocytes. This was particularly evident in high-magnification views (Fig. 7c). As reported previously, electron-dense material of uncharacterized composition is present at the contact zone between neuroblasts in wild-type mice (Doetsch et al., 1997). Electron-dense patches were rarely observed in Itgb1–CNSko mice, whereas tight junctions between neuroblasts and astrocytes were preserved (Fig. 7c). These findings suggest that the electron-dense patches between neuroblasts may be integrin-dependent adhesion complexes that are specifically affected in Itgb1–CNSko mice. However, the overall integrity of the glial tubes is perturbed in Itgb1–CNSko mice, suggesting that contacts between glial cells are also affected. Overall, the findings suggest that defects in both the formation of cell chains and the glial tubes likely contribute to the migratory defect in the RMS of Itgb1–CNSko mice.
Functions of β1 integrins in the formation of cell chains
The expression of β1 integrins in migrating neuroblasts suggest that they control their migratory behavior directly. To test for functions of β1 integrins in the formation and movement of cell chains, we cultivated SVZ explants in Matrigel. In this assay, cells from wild-type mice emigrate as cell chains independently of a glial scaffold (Wichterle et al., 1997). Consistent with previous observations, we observed cell chains in explants derived from wild-type mice (Fig. 8a,b). In contrast, cells in explants prepared from Itgb1–CNSko mice did not form chains and migrated as single cells (Fig. 8c,d). A quantification of the distance traveled by 48 h revealed no difference for cells in explants from wild-type and mutant mice (Fig. 8e). We next monitored the dynamics of cell migration by time-lapse video microscopy. Wild-type cells migrated along each other in tightly linked chains, in which the leading cell developed one or multiple processes (Fig. 9a) (supplemental Movie 1, available at www.jneurosci.org as supplemental material). In contrast, cells from mutant animals migrated as single cells and formed one or several leading processes ending in growth cones that appeared to explore the environment (Fig. 9b) (supplemental Movie 2, available at www.jneurosci.org as supplemental material). After a phase of leading process extension, the cell bodies translocated toward the tips of the processes (Fig. 9b, cell marked with an arrowhead). Unlike wild-type cells, which appeared to migrate directionally away from the explant, the movement of β1-deficient cells appeared more random, and cells even reversed the direction of migration (Fig. 9b, cell marked by arrowhead). Together, our data show that β1 integrins are not required for the migration of neuroblast into Matrigel. Instead, β1 integrins control the mode of migration. In the absence of β1 integrins, most of the cells appear to migrate in a mode similar to the leading cell in chains, but individual cells no longer form stable contact with each other to form cell chains.
Formation of cell chains depends on laminin
Integrins may control chain formation by binding to counter receptors on adjacent neuroblasts or by binding to ECM glycoproteins that are deposited between neuroblasts. Because integrins that interact with laminin are upregulated during migration (Fig. 1), we analyzed the distribution of laminin within the RMS. Different laminin isoforms have been described that are formed by the assembly of different α, β, and γ subunits into heterotrimeric molecules (Miner and Yurchenco, 2004). In our microarray analysis, we observed prominent expression of several laminin α, β, and γ subunits in the RMS, including the α1, α2, and α4 subunits (data not shown). To confirm the expression of laminins in the RMS, we stained sagittal sections with available antibodies that recognize the laminin α1 and α2 subunits. We observed prominent staining in the SVZ and RMS of adult mice (Fig. 10a,b). Higher-magnification views showed that the laminin subunits were concentrated around the surface of PSA-NCAM-expressing neuroblasts in cell chains (Fig. 10c–e). Together, our microarray analysis and immunohistochemistry data show that several laminin subunits are expressed in the RMS and are appropriately localized to mediate interactions between neuroblasts.
To analyze whether laminins are required for the formation of cell chains, we next analyzed the RMS in mice with mutations in laminin subunit genes. We focused our studies on laminin α2- and α4-deficient mice, because both α subunits are expressed in the adult RMS (Fig. 10 and data not shown). Furthermore, the mutant mice survive to postnatal ages (unlike other mice with mutations in laminin subunit genes such as laminin α1-, β1-, and γ1-deficient mice) (Miner and Yurchenco, 2004), allowing an analysis of the RMS in adult mice. Mice with mutation in either the laminin α2 or α4 subunit genes did not show any defects in the formation of the RMS (data not shown). However, α2/α4 double-mutant mice showed defects in the RMS. Although the RMS could be detected in the mutants (Fig. 10f,g) and cell chains were visible in some animals, the overall organization of the RMS was far less compact (Fig. 10h,i). Many of the cells did not assemble into cell chains (Fig. 10j,k). Because the laminin α2 subunit is a component of the laminin-2, -4, and -12 isoforms whereas the α4 subunit is a component of the laminin-8 and -9 isoforms (Miner and Yurchenco, 2004), our findings suggest that several of these isoforms have redundant functions in the formation of cell chains in the RMS.
To further define the function of laminin for the formation of cell chains, we analyzed chain formation in SVZ explants. Previous studies have shown that the formation of cell chains that emigrate from SVZ explants is dependent on the in vitro environment. Cell chains form in Matrigel (which contains laminin) but not in collagen gels (which do not contain laminin) (Wichterle et al., 1997). Unfortunately, laminin alone does not polymerize into three-dimensional matrices. Therefore, to test whether laminin can induce chain migration, we explanted the SVZ from wild-type mice into mixed collagen/laminin gels. We used for this purpose laminin-1, which is expressed in the RMS (Fig. 10) and readily available as a purified molecule. As a control, we prepared SVZ explants in collagen gels without laminin-1. Cells in collagen gels migrated as single cells and showed a similar morphology and migrational behavior to integrin β1-deficient cells in Matrigel (Fig. 11a). The cells formed single or multiple leading processes that were tipped with growth cones. In contrast, cells in mixed collagen/laminin-1 gels migrated as chains (Fig. 11b). Although the chains were similar in morphology to those in Matrigel, the chains seemed less compact. In addition, cells in mixed collagen/laminin-1 gels migrated at reduced speed when compared with cells in collagen gels, suggesting that the formation of cell chains reduced the speed of migration (Fig. 11c).
To test directly whether laminin-1 promotes cell–cell interactions between neuroblasts, we performed short-term cell aggregation assays. PSA-NCAM-positive neuroblasts were isolated by FACS, and cell aggregation was quantified 2 h after the addition of laminin-1 (Fig. 11d,e). Because interactions of integrins with laminin-1 are dependent on divalent cations, we performed the experiments in the presence or absence of Ca2+. FACS-sorted PSA-NCAM-positive cells in buffer without Ca2+ formed few if any aggregates. When the cells were incubated in buffer containing Ca2+, aggregates formed that consisted predominantly of two cells. In the presence of both Ca2+ and laminin-1, the cells formed many more and larger aggregates. Aggregates were visible after 20 min (data not shown) but were more prominent after 2 h (Fig. 11d,e). Together, our experiments provide strong evidence that laminin is critical for the formation of cell chains in the RMS.
We provide here several lines of evidence that β1 integrins and their laminin ligands control the formation of cell chains in the RMS of adult mice. First, we show that β1 integrins and several laminins are expressed in the appropriate pattern to support chain formation. Second, functional data show that β1 integrins are required for chain formation in vivo and in vitro. Third, laminin induces the aggregation of PSA-NCAM-positive neuroblasts and it promotes chain migration in reconstituted collagen/laminin matrices, and laminin α2/α4 double-mutant mice show altered migration in the RMS. Surprisingly, although β1 integrins are required for the formation of cell chains, β1-deficient neuroblasts still migrate from the SVZ to the OB. These results suggest that β1 integrins and their laminin ligands promote the formation of cell chains, whereas additional receptors are important for cell motility independently of chain formation. In addition, the glial tubes are disrupted in Itgb1–CNSko mice, suggesting additional functions for integrins and their laminin ligands, for example in mediating interactions between glial cells.
β1 integrins and the formation of cell chains
Previous studies have identified several receptor molecules that control chain migration in the RMS, such as PSA-NCAM, which controls interactions of neuroblasts with the stationary environment through which the cells migrate (Tomasiewicz et al., 1993; Cremer et al., 1994; Ono et al., 1994; Hu et al., 1996; Chazal et al., 2000) and members of the EphB and ErbB receptor family that affect neuroblast proliferation and migration (Conover et al., 2000; Anton et al., 2004; Ghashghaei et al., 2006). However, the molecules that link neuroblasts into chains have remained elusive. The data presented here now provide strong evidence that β1 integrins are important to link neuroblasts into chains. β1 integrins are expressed on the surface of migrating neuroblasts in cell chains, and chain formation is disrupted in Itgb1–CNSko mice. At the light microscopic level, β1-deficient neuroblasts are in close proximity to each other, but they are not assembled into chains along the rostrocaudal axis of the RMS. Ultrastructural studies confirm that contacts between migrating neuroblasts are disrupted in the absence of β1 integrins. In contrast, contacts between neuroblasts and astrocytes are maintained. In vitro data provide evidence that β1 integrins act at least in part in neuroblasts to link them into chains: although β1-deficient cells still migrate from SVZ explants into Matrigel, the cells no longer form chains and instead migrate as single cells.
We also provide first insights into the mechanism by which β1 integrins control chain formation. β1 integrins are receptors for ECM molecules including laminins. Several integrins that bind laminins (α1β1, α2β1, α3β1, α6β1, and α7β1) are expressed in migrating neuroblasts. Likewise, several laminins are expressed in the RMS, and the formation of cell chains is perturbed in laminin α2/α4 double-mutant mice. In vitro, purified laminin can induce the aggregation of neuroblasts, and neuroblasts migrate as chains in mixed collagen/laminin gels, whereas collagen alone promotes migration but not chain formation. The simplest interpretation of our results is that β1 integrins in neuroblasts bind to laminin to link the cells into chains. This could be achieved by different mechanisms. Although the integrins α3β1, α6β1, and α7β1 bind to the same site in laminin, the integrins α1β1 and α2β1 bind to an additional site (Miner and Yurchenco, 2004). Therefore, neuroblasts could be linked to each other by integrins on the surface of adjacent cells that bind to distinct sites in a single laminin molecule. Alternatively, because laminin assembles into networks (Miner and Yurchenco, 2004), integrins on adjacent neuroblasts may interact with multiple binding sites within a macromolecular laminin matrix. This latter model shows striking parallels to the mechanisms by which integrins control cellular interactions in Drosophila, in which different integrins on muscle and tendon link the cells via an ECM network to each other (Bokel and Brown, 2002).
Although laminin is able to restore chain migration of wild-type cells in collagen gels, the chains appear less compact and move with reduced speed when compared with cell chains in Matrigel. These findings suggest that, although β1 integrins and laminins initiate chain formation, other receptors and signaling molecules cooperate with β1 integrins and laminins to control the behavior of neuroblasts in chains. As one possibility, additional ECM glycoproteins may be required for chain formation. Our data already suggest that laminin isoforms containing the laminin α2 and α4 subunits are involved, but our data also show that additional laminin subunits are expressed in the RMS. Several isoforms may cooperate with each other or with additional ECM molecules to achieve compaction of cells chains. Alternatively, integrins may cooperate with other secreted signaling molecules and cell surface receptors to control chain formation. Intriguingly, EphB and ErbB receptor tyrosine kinase control not only chain migration (Conover et al., 2000; Anton et al., 2004; Ghashghaei et al., 2006) but also the function of β1 integrins in cell adhesion and migration (Adelsman et al., 1999; Huynh-Do et al., 1999; Davy and Robbins, 2000; Miao et al., 2000; Colognato et al., 2002). Therefore, the cooperative activation of integrins by laminin and receptor tyrosine kinases by their ephrin and neuregulin ligands may control the overall behavior of cells in chains. As one possibility, cooperative signaling by tyrosine kinases and integrins may be required to activate cellular signal transduction pathways that control the assembly of the actin cytoskeleton as described in other cell types (Schwartz and Ginsberg, 2002). In addition, integrins in epithelial cells control the activity of cadherins (Arregui et al., 2000; Yano et al., 2004), which may act downstream of integrins in the compaction of neuroblasts in cell chains. Likewise, defects in the formation of cell chains have been observed in mice deficient in the secreted chemorepellents slit1 and slit2 (Hu, 1999; Wu et al., 1999; Nguyen-Ba-Charvet et al., 2004), suggesting that integrins, laminins, and slit molecules may act in a common molecular pathway. We can at present only speculate how slits, integrins, and laminins cooperate to affect chain migration. An intriguing possibility is that laminins may bind slit molecules and help to establish a slit gradient. Alternatively, integrins may recruit laminin/slit complexes to the surface of neuroblasts to concentrate the slit ligand and facilitate the activation of slit receptors within cells. We are currently performing systematic perturbations of various molecular pathways to define the signals that cooperate with β1 integrins in the formation and movement of cell chains.
Surprisingly, although Wichterle et al. (1997) have shown that cell migration is observed in Matrigel (which contains laminin and collagen) but not on polylysine or in collagen gels, Mason et al. (2001) have reported that chain migration occurs on polylysine even in the absence of Matrigel. These findings suggest that laminin may not be essential for chain migration. However, under the culture conditions used by Mason and colleagues (but not by Wichterle et al.), glial fibers emanated from the explants and neurons appeared to migrate along these fibers. Therefore, molecules secreted from the glial cells and including laminin may provide the necessary environment for chain migration under the culture conditions used by Mason and colleagues.
β1 integrins, laminin, and glial tubes
Although our in vitro assays suggest that β1 integrins control interactions between neuroblasts, our in vivo data suggest that integrins affect the RMS in different ways as well. Most prominently, the glial tubes are disrupted in Itgb1–CNSko mice. These defects could be a secondary consequence of perturbations in interactions between neurons or could affect glial cells more directly. Although the glial tubes are not essential for neuronal chain migration in the in vitro Matrigel assay (Wichterle et al., 1997), we observed an increasing number of integrin β1-deficient neuroblasts leaving the RMS prematurely on route to the OB. The glial tubes may provide a physical boundary lining the migratory pass of neuroblasts, but it may also have a more active role to control migration. An interesting hypothesis would be that the glial tubes generate a corridor for migration by secreting signaling molecules, affecting the formation of chains and controlling migrational speed. This model is consistent with the findings reported here. The data also suggest that the reduced size of the OB in Itgb1–CNSko mice is a consequence of perturbations in the formation of both cell chains and the glial tubes, affecting migrational speed and allowing some cells to leave the RMS prematurely. Overall, this leads to a net loss of the number of cells that reach the OB and therefore to a reduction in the size of the OB.
β1 integrins and cell migration in the CNS
Previous studies have addressed the function of β1 integrins in neuronal precursors. These studies that were mainly performed under well defined in vitro conditions have provided compelling evidence that integrin signaling controls the adhesion, maintenance, and migration of neuronal precursors (Campos et al., 2004; Tate et al., 2004; Andressen et al., 2005). However, the function of β1 integrins in neuronal precursors and in differentiating neurons in the more complex in vivo environment is less well understood. Studies over the past decades have demonstrated that neuronal precursors and differentiating neurons in the CNS migrate by different modes, including radial migration along glial fibers, tangential migration along neuronal processes, and migration in cell chains (Alvarez-Buylla and Garcia-Verdugo, 2002; Hatten, 2002; Marin and Rubenstein, 2003; Kriegstein and Noctor, 2004). β1 integrins were thought to control cell motility in several of these modes of migration (Clegg et al., 2003), but our findings suggest that β1 integrins have a much more specific role than previously appreciated. Our previously published data (Graus-Porta et al., 2001; Blaess et al., 2004) provide strong evidence that many of the migratory events that are important for the development of the cerebral and cerebellar cortex progress normally in the absence of β1 integrins. Similarly, although studies with function blocking antibodies have suggested that β1 integrins control cell motility of neuroblasts in the RMS, for example by controlling leading process extension (Murase and Horwitz, 2002; Emsley and Hagg, 2003), our data now show that β1-deficient neuroblasts in the RMS still elaborate extensive leading processes and are able to migrate directionally toward the OB. However, the cells migrate with reduced speed, suggesting that integrins control some aspects of migration. In fact, directional cell motility is a complex process and likely depends on the cooperative action of several cell surface receptors. Likewise, once the cells reach the OB, chains are disassembled and cell switch to a radial mode of migration. The secreted signaling molecules reelin, tenascin R, and prokineticin-2 induce detachment of neuroblasts from cell chains (Hack et al., 2002; Saghatelyan et al., 2004; Ng et al., 2005), but the function of integrins in this process is unclear. We observed radially migrating cells in Itgb1–CNSko mice, but, because cell migration within the RMS already was perturbed, we could not quantify whether integrins are required for reelin, tenascin R, or prokineticin-2 signaling. Additional experiments will be necessary to address this possibility.
This work was funded by National Institutes of Health Grants RO1-NS046456 and U01-MH078833 (U.M.). We thank the following: G. Rougon, Harold Cremer (both at Université de la Mediterranée Aix-Marseille 2, Marseille, France) and L. Sorokin (University of Uppsala, Uppsala, Sweden) for antibodies; Dr. K. Hodivala-Dilke (Cancer Research UK, London, UK) and Dr. D. Sheppard (University of California, San Francisco, San Francisco, CA) for tissue from Itgb5-null mice; Dr. K. S. Spencer (The Scripps Research Institute, La Jolla, CA) for help for live imaging; Dr. M. Wood (The Scripps Research Institute) for help with electron microscopy; the flow cytometry core facility of The Scripps Research Institute for FACS; Susan Sutton (Genomics Institute of the Novartis Research Foundation, San Diego, CA) for microarray studies; and members of the laboratory for critical discussion.
- Correspondence should be addressed to Ulrich Müller, Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA 92073.