Laminar organization is a key feature of the mammalian cerebral cortex, but the mechanisms by which final positioning and “inside-out” distribution of neurons are determined remain largely unknown. Here, we demonstrate that Robo1, a member of the family of Roundabout receptors, regulates the correct positioning of layers II/III pyramidal neurons in the neocortex. Specifically, we used RNA interference in mice to suppress the expression of Robo1 in a subset of layers II/III neurons, and observed the positions of these cells at distinct developmental stages. In contrast to control neurons that migrated toward the pial surface by P1, Robo1-suppressed neurons exhibited a delay in entering the cortical plate at respective stages. Unexpectedly, after the first postnatal week, these neurons were predominantly located in the upper part of layers II/III, in contrast to control cells that were distributed throughout these layers. Sequential electroporation studies revealed that Robo1-suppressed cells failed to establish the characteristic inside-out neuronal distribution and, instead, they accumulated beneath the marginal zone regardless of their birthdate. These results demonstrate that Robo receptors play a crucial role in neocortical lamination and particularly in the positioning of layers II/III pyramidal neurons.
The mammalian cerebral cortex consists of 6 distinct layers that are largely represented by its excitatory neurons with characteristic axon projection patterns and gene expressions. Early in corticogenesis, neurons arise from the ventricular (VZ) and subventricular (SVZ) zones and move radially toward the pial surface (Pearlman et al. 1998; Nadarajah et al. 2001). Subsequent cohorts of later-born neurons migrate past their predecessors to take a more superficial position within the cortical plate (CP), resulting in an “inside-out” neuronal distribution pattern (Angevine and Sidman 1961; Rakic 1974). It is considered that the latter process involves a series of migration and positioning events such as multipolar-to-bipolar transition of neurons (Tabata and Nakajima 2003; Noctor et al. 2004; Tabata et al. 2009), radial glia-guided locomotion (Rakic 1972; O'Rourke et al. 1992; Nadarajah et al. 2001), detachment from radial glial fibers (Pinto-Lord et al. 1982; Gongidi et al. 2004; Elias et al. 2007), and terminal somal translocation (Nadarajah et al. 2001; Borrell et al. 2006; Sekine et al. 2011). The complex behavior of neurons during radial migration suggests that distinct molecular mechanisms may regulate each of the migration steps during development. Indeed, an increasing number of genes have been identified that control the early phase of radial migration (Caviness and Rakic 1978; Gupta et al. 2002; Nadarajah and Parnavelas 2002; Tsai and Gleeson 2005; Cooper 2008; Huang 2009; Honda et al. 2011). In contrast, much less is known about how the terminal positioning of neurons is regulated after they arrive at the surface of the CP. Although it has long been postulated that inside-out neuronal distribution requires a correct termination of radial migration through the local interaction between extracellular glycoprotein Reelin and its receptors (Dulabon et al. 2000; Sanada et al. 2004), confirmation of such pathway using transgenic mouse models has remained elusive (Magdaleno et al. 2002; Yoshida et al. 2006). Alternative signaling pathways may contribute to this process (Moers et al. 2008), but the positioning defects appear only in a limited number of neurons. This prompted us to investigate whether additional mechanisms may contribute to the correct positioning of the cortical neurons after they reach the upper part of the CP.
Roundabout (Robo) was first identified as a receptor for the chemorepulsive ligand Slit in the Drosophila nerve cord (Kidd, Brose et al. 1998). In the mammalian spinal cord and Drosophila nerve cord, Robo proteins play a crucial role in preventing the commissural axons from re-crossing the midline through growth cone repulsion (Kidd, Brose et al. 1998; Kidd, Russell et al. 1998; Chen et al. 2008). In the mammalian forebrain, Robo signaling has also been shown to play key roles in the axonal pathfinding (Andrews et al. 2006, 2008; López-Bendito et al. 2007). Furthermore, recent studies have shown that the inhibition of Robo1-mediated signaling can affect the proliferation and migration of the neocortical interneurons (Andrews et al. 2006, 2008; Hernandez-Miranda et al. 2011). These findings support the notion that Robo receptors may play an important role beyond axonal pathfinding in the developing neocortex.
In this study, we examined the cortical neuron subtypes that express Robo1 during development. We found that Robo1 mRNA and Robo1 protein are expressed in pyramidal neurons as they enter the CP, and in restricted zones including the upper part of layers II/III. To investigate the role of this receptor in the development of upper-layer projection neurons in vivo, we suppressed Robo1 expression in pyramidal neurons of layers II/III using RNA interference. Here, we report that Robo1-suppressed neurons exhibit a delay in entering the CP during the embryonic period. Moreover, although these neurons eventually migrate to the pial surface by the middle of the first postnatal week, they are predominantly located in the uppermost part of layers II/III after the first postnatal week. Sequential in utero electroporation of Robo1-shRNA constructs at E15.5 and E16.5 in the same cortices revealed that Robo1-suppressed cells fail to establish the typical inside-out distribution and accumulate beneath the marginal zone (MZ) regardless of their birthdates. These differences were not observed in E14.5 Robo1-shRNA-transfected neurons. Our results indicate that Robo1 is a key regulatory molecule that controls the laminar distribution and migration of layers II/III pyramidal neurons in the neocortex.
Materials and Methods
Pregnant Institute of Cancer Research (ICR) mice were purchased from CLEA Japan, Inc. The day the vaginal plug was detected was designated as E0.5. Robo1−/− knockout mice were generated by intercrossing Robo1+/− heterozygote mice as described previously (Andrews et al. 2008). All animal experiments were performed in accordance with The Animal Care and Use Committee of the National Institute of Neuroscience, RIKEN, and UCL ethics guidelines. Animals of either sex were used in our experiments.
In Situ Hybridization
Complementary DNA (cDNA) of mouse Robo1 (GenBank accession number AK040651) which had been cloned into the pFLCI vector was used to make the cRNA probe. The cDNA fragment of the unique region of Robo1 (770 bp) was obtained by PCR using the following primers: Robo1-F (5′-GGGGAATTCAATGAGTTTCAAGGAGCA-3′), Robo1-R (5′-CCCAAGCTTGCGACTGTAGGTTGTCAG-3′), annealing temperature 61°C, and subcloned into the mammalian expression vector pCMV-SPORT (Gibco BRL). Digoxigenin (DIG)-labeled antisense and sense cRNA probes were produced with T7 and SP6 polymerase, respectively, using in vitro transcription according to the manufacturer's instructions (Roche Diagnostics).
The in situ hybridization analysis was performed as described previously (Gonda et al. 2007), with a slight modification. Briefly, the brains were fixed with 4% paraformaldehyde (PFA), coronally sliced into 14 μm sections using a Cryostat (CM-3000; Leica), and mounted onto Matsunami adhesive silane (MAS)-coated glass slides (Matsunami). Hybridization of the DIG-labeled cRNA probes (1 μg/mL) on brain sections was performed at 60°C overnight, and was followed by washing twice with 50% formamide, 0.2% saline sodium citrate at 60°C for 30 min each, and once with maleic acid buffer containing 0.2% Tween 20 at room temperature for 3 min. Incubation of alkaline phosphatase-conjugated anti-DIG antibody (1:2000, Roche Diagnostics) was performed at 4°C overnight, and sections were visualized with 0.4 mM nitro blue tetrazolium and 0.35 mM 5-bromo-4-chrolo-3-indoylphosphate at room temperature overnight. The reaction was stopped by rinsing in water twice at room temperature for 5 min each, and sections were dehydrated using a graded ethanol series (70%, 80%, 90%, 95%, and 100%) for 1 min each, and cleared 3 times in xylene for 5 min each. The sections were finally mounted with Entellan (Merck).
Immunohistochemistry was performed using a floating method as described previously (Namba et al. 2009). Frozen brains fixed with 4% PFA were coronally sliced at 50 μm using a Cryostat (CM-3000), except for Robo1/Satb2 double immunohistochemistry in which sections sliced at 14 μm were used. The following primary antibodies were used: goat polyclonal anti-Robo1 (1:50; R&D System, Inc.), mouse monoclonal anti-Satb2 (1:50; Abcam), rabbit polyclonal anti-Cux1 (1:100; Santa Cruz), rabbit polyclonal anti-enhanced green fluorescent protein (EGFP) (1:1000; Invitrogen), rat monoclonal anti-EGFP (1:500; Nacalai Tesque Inc.), rabbit anti-Calbindin (1:1000; Swant), goat polyclonal anti-Brn2 (1:200; Santa Cruz Biotechnology), mouse monoclonal anti-NeuN (1:500; Chemicon), and mouse monoclonal anti-GFAP (1:1000; Sigma). For detection of goat polyclonal anti-Robo1 antibodies, sections were incubated with biotinylated donkey anti-goat antibody (1:100; Jackson ImmunoResearch Laboratories) for 2 h and processed with a conventional immunohistochemistry protocol using the tyramide signal amplification biotin system (PerkinElmer, Inc.).
For 5-ethynylil-2′-deoxyuridine (EdU) and EGFP double detection, sections were incubated overnight with anti-rabbit EGFP antibody at 4°C and subsequently processed for EdU detection using Click-iT™ EdU Imaging Kit (Invitrogen). Immunostained sections were mounted on MAS-coated glass slides and examined with a confocal laser-scanning microscope (FV1000; Olympus).
Quantification of Neuronal Density and Cortical Layer Thickness
For quantification of neuronal cell density and layers II/III thickness, serial sections (14 μm) were cut from brains of P34-P58 Robo1−/− mice and Robo1+/+ control littermates (n = 3). Sections were stained with Cux1 and 4′,6-diamidino-2-phenylindole (DAPI), and matching sections were photographed. All measurements and cell counts were performed on the primary somatosensory cortex (S1 barrel field). For cell density, layers II/III were divided into 6 equal bins according to their mean position in the Robo1+/+ neocortex and the number of Cux1/DAPI double-positive cells in each bin was counted. The same bin was applied to the Robo1−/− cortex; however, due to the decrease in the thickness of layers II/III in the Robo1−/− cortex, bin 6 includes Cux1/DAPI double-positive cells of layer IV in the Robo1−/− cortex. Total number of Cux1-positive cells in layers II/III was counted per 600 μm width of the Robo1−/− and Robo1+/+ cortices.
For plasmid-based RNA inhibition of Robo1, the complementary oligonucleotides for the following target sequence (Robo1-shRNA1: 5′-ACTCAAACCTAACGCCA-TTTA-3′; Robo1-shRNA2: 5′-AGCTGATTGTATAGCCAATTA-3′; Robo1-shRNA3: 5′-TCTCGGTAATGAAACGAAGTA-3′; Robo1-shRNA4: 5′-TCCGCTACTTTGAC-AGTTCAA-3′; and Mt-Robo1-shRNA: 5′-ACTCAAACCATTCGCCATTTA-3′) were annealed and inserted into the BamHI/HindIII sites of pSilencer 3.0-H1 (Ambion). As a control plasmid, pSilencer 3.0-H1 Negative Control (Ambion) was used. To construct the Robo1-expression vector, full-length cDNA of mouse Robo1 was inserted into the mammalian expression vector pCAGGS, which contains a modified chicken β-actin promoter with a cytomegalovirus-immediate early enhancer (CAG) promoter (Niwa et al. 1991), provided by Dr J. Miyazaki (Osaka University).
Cell Lines and Western Blot Analysis
COS7 cells were grown at 37°C in a humidified 5% CO2 atmosphere in the Dulbecco's modified Eagle's medium (Gibco BRL) containing 10% heat-inactivated fetal bovine serum (Irvine Scientific). For transient expression studies, cells were transfected with plasmid DNA using Lipofectamine Plus (Invitrogen) according to the manufacturer's protocol.
COS7 cells or brain tissues were homogenized in lysis buffer (10 mM Tris–HCl [pH 7.4], 150 mM NaCl, 1 mM ethylene diamine tetraacetic acid, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% sodium dodecyl sulfate, and protease inhibitor cocktail [Roche Diagnostics]) and incubated at 4°C for 1 h. After removing the nuclei and debris by centrifugation (2000 × g for 10 min at 4°C), protein concentration of the supernatant was determined using the Bio-Rad Protein Assay Kit (Bio-Rad). Protein samples were subjected to immunoblotting with rabbit polyclonal anti-Robo1 antibody (1:1000; Rockland Immunochemicals) or goat polyclonal anti-Robo2 antibody (1:1000; R&D Systems). Blots were developed with ECL Reagent (GE Healthcare).
In Utero Electroporation
In utero electroporation was performed as previously described (Tabata and Nakajima 2001), with a slight modification. Briefly, timed-pregnant ICR mice were deeply anesthetized with sodium pentobarbitone (Somnopentyl; Kyoritsu Pharmaceuticals), and their uterine horns were exposed. Approximately 1–2 μL of plasmid DNA solution dissolved to a final concentration of 5 mg/mL in 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid-buffered saline was injected into a lateral ventricle of each embryo with a glass micropipette, and electric pulses (30 V, 50 ms) were discharged 4 times at intervals of 950 ms with an electroporator (CUY21E; Nepa Gene). The uterine horns were then replaced in the abdominal cavity to allow the embryos to continue normal development. For sequential in utero electroporation, plasmids were electroporated into the same cortex at E15.5 and E16.5 with an interval of 24 h.
Fluorescent Activated Cell Sorting
Postnatal (P) day 1 mice transfected with EGFP-expression vector were placed on ice for anesthesia, their brains removed, and EGFP-positive regions were dissected. The dissected tissues were incubated with Liberase Blendzyme I (100 μg/mL; Roche Diagnostics) and DNase I (0.05 mg/mL; Roche Diagnostics) for 15 min at 37°C. The cells were filtered through a 40 μm nylon mesh, and incubated with propidium iodide (Becton Dickinson) (50 μg/mL in phosphate-buffered saline). Sorting and analysis were performed using a FACS Aria flow cytometer (Becton Dickinson). To isolate EGFP-positive cells, physical parameters such as the forward scatter, representing cell size, and the side scatter, representing cellular granularity, and EGFP signals detected with FITC fluorescence were used. The sorted cells were used for a reverse transcription (RT)-PCR analysis.
Total RNA was extracted from cells using an RNeasy Plus Micro Kit (QIAGEN), and cDNA was synthesized using a QuantiTect Whole Transcriptome Kit (QIAGEN) according to the manufacturer's instructions. PCR was performed using Ex Taq (Takara) and a thermal cycler (Veriti; Applied Biosystems). The thermocycle conditions were as follows: 10 s at 98°C, 15 s at the annealing temperature (Robo1, glyceraldehydes-3-phosphate dehydrogenase [GAPDH]: 62°C, Robo2: 63°C), and 30 s at 72°C for 25 cycles (Robo1, GAPDH) or 30 cycles (Robo2). The PCR products were separated on a 2% agarose gel and stained with ethidium bromide.
A quantitative real-time PCR analysis was performed using the SYBR green-labeling system (SYBR Premix Ex Taq II; Takara) and the ABI Prism 7700 Sequence Detection System (Applied Biosystems). Amplifications were performed in a 96-well optical plate, and the thermocycle conditions were as follows: 5 s at 95°C, 10 s at 62°C, and 30 s at 72°C for 40 cycles. The primers used were as follows: Robo1-F, 5′-CTCCCGTCTGATGACACACAATACC-3′; Robo1-R, 5′-CATTAAGGGTTAGG-CAATCAATCAGCAACAC-3′; Robo2-F, 5′-GCCCAGCTGCGGATCACCTC-3′; Robo2-R, 5′-CCGGCCCCACCCCTTTTTCC-3′; GAPDH-F, 5′-GTCATCATCTCC-GCCCCTTCTGC-3′; GAPDH-R, 5′-GATGCCTGCTTCACCACCTTCTTG-3′. A quantitative analysis was performed using the delta–delta Ct method with GAPDH as an internal control.
The data were evaluated using the Student t-test and the software package JSTAT. All values were expressed as the mean ± standard error of the mean, and P-values <0.05 were considered significant.
Spatiotemporal Expression Patterns of Robo1 mRNA and Protein in the Developing Neocortex
To investigate the roles of Robo1 in the developing neocortex, we first examined its expression in embryonic and postnatal mouse cortices by in situ hybridization (Fig. 1A–F). A signal for Robo1 was detected in the preplate (PP) as early as E12.5 and, by E16.5, Robo1 mRNA exhibited a layer-specific expression pattern within the CP. In situ hybridization using an ER81 probe (a marker for layer V neurons) in adjacent sections indicated that Robo1 mRNA is highly expressed in layer V neurons and moderately expressed in neurons of layers II/III and VI at P1 (Fig. 1D and data not shown). While the level of Robo1 mRNA expression was maintained in deep-layer neurons (layers V and VI), its expression in the upper part of layers II/III was upregulated by P1, and remained high at P3 (Fig. 1D,E). Its expression gradually decreased after P7, but it was still detectable at P15 (Fig. 1F and data now shown).
The layer-specific expression pattern of Robo1 mRNA in the CP suggested that these cells were pyramidal neurons. To assess the identity of the Robo1 expressing cells, we next examined the stage-dependent localization of Robo1 protein within pyramidal neurons. At E15.5, Robo1 protein was strongly expressed in the intermediate zone (IMZ) and weaker in the CP (Fig. 1G–K). At this stage, the cortical IMZ consists of corticofugal and thalamocortical axons, as well as tangentially migrating interneurons (Andrews et al. 2006, 2008; López-Bendito et al. 2007) and radially migrating pyramidal cells, making it difficult to discern the source of Robo1 protein expression. We, therefore, co-labeled Robo1 protein with a cortical callosal projection neuron marker, Satb2 (Alcamo et al. 2008; Britanova et al. 2008). We found that Robo1 protein was at times localized at the leading process of Satb2-expressing neurons in the IMZ and CP (Fig. 1K,K′), suggesting that it was expressed in migrating pyramidal neurons. We further examined the localization and distribution of Robo1 protein at E18.5 and P1 (Fig. 1L–U′). We found that, although Robo1 mRNA was expressed in cells of layers V, VI, and upper part of layers II/III at these stages (Fig. 1C,D). Robo1 protein was predominantly detected in the MZ, and in a graded manner, in the cell somata of neurons within the CP (Fig. 1N,S). These data suggest that Robo1 protein mainly localizes in the neurites and axons after translation. We further assessed the co-localization of Robo1 protein with Satb2 whose expression is maintained in callosal projection neurons in the CP (Alcamo et al. 2008; Britanova et al. 2008). We found numerous examples of such co-localization (Fig. 1P′,U′). These results indicate that Robo1 is indeed expressed in differentiated pyramidal cells within the neocortex.
Suppression of Robo1 Expression in the Developing Mouse Neocortex
Our analyses revealed that Robo1 protein was expressed in a graded manner within the CP during critical stages of pyramidal cell migration and positioning (Fig. 1K′,N,S), and we wished to explore the requirement of this molecule in these events. We, therefore, analyzed Robo1−/− knockout mice to assess the effect of loss of Robo1 in the distribution of neocortical neurons. We utilized a Robo1 knockout line in which the whole Robo1 gene (exons 1–22) is deleted (Andrews et al. 2008), and examined the cytoarchitecture of the adult neocortex by Nissl staining (Fig. 2A). Notably, in contrast to the organized alignment of neurons in the supragranular layers (layers II/III) of the Robo1 wildtype cortex (Fig. 2A), Robo1−/− cortices exhibited dense cell distribution in layers II/III (Fig. 2A). We further quantified the density of cells in these layers in Robo1−/− cortices by counting the number of DAPI+ nuclei. Cell density was significantly higher in Robo1−/− cortical layers II/III than those in the Robo1+/+ cortex (Robo1+/+: 162.0 ± 5.9; Robo1−/−: 214.1 ± 11.2 × 103 cells/mm3, P< 0.01, Fig. 2B). As we previously reported an increase in the total number of interneurons in the adult Robo1−/− cortex (Andrews et al. 2008), we next examined whether this increase in cell density in Robo1−/− cortical layers II/III attributes to any changes in pyramidal cell distribution, by utilizing a specific marker for layers II/III pyramidal cells, Cux1 (Fig. 2D,E, Nieto et al. 2004). Notably, the distributions of Cux1-expressing neurons in Robo1−/− cortices were significantly dense throughout layers II/III compared with that of control cortices (bins 2–5, n = 12 sections, P< 0.05, Fig. 2E). To determine whether this change is due to altered distribution or increased proliferation, we quantified the total number of Cux1-expressing cells within layers II/III in Robo1−/− cortices. We observed no changes in the total Cux1-positive cell number in Robo1−/− cortices (Fig. 2F), indicating that the observed increase in cell density in Robo1−/− attributes to changes in layers II/III pyramidal cell distribution, rather than increased proliferation and/or survival of these cells upon loss of Robo1. Consequently, the average thickness of layers II/III in the Robo1−/− cortex was significantly thinner as compared with Robo1 wildtype controls (Robo1+/+: 299.1 ± 3.9; Robo1−/−: 270.2 ± 4.1 μm, P< 0.05, Fig. 2C), resulting in a superficial shift of layer IV toward the pial surface (Fig. 2D,E; note that bin 6 in Fig. 2E includes Cux1-positive layer IV cells in the Robo1−/− cortex but not the Robo1+/+ cortex). Together, these observations further supported the idea that Robo1 is required for normal laminar formation by upper-layer neurons, in particular layers II/III pyramidal cells.
Thus, to clarify the cell-autonomous requirement of Robo1 in the laminar distribution of layers II/III neurons, we directly suppressed its expression in a discrete population of cortical pyramidal cells using RNA interference. We designed 4 Robo1-shRNA-expression vectors (Robo1-shRNA1–4) to target distinct regions in the Robo1 coding sequence (Supplementary Fig. S1A). These vectors were co-transfected with Robo1-expression vector into COS7 cells and the expression of recombinant Robo1 protein were examined. Robo1-shRNA1 and Robo1-shRNA4 greatly reduced the amount of Robo1 protein expression (Supplementary Fig. S1B), whereas Robo1-shRNA2 and Robo1-shRNA3 had weaker effects (data not shown). The introduction of a mutation within the Robo1 target region in Robo1-shRNA1 (Mt-Robo1-shRNA1) abolished the knockdown effect of Robo1-shRNA1 (Supplementary Fig. S1C). We further confirmed that Robo1-shRNA1 had a little effect on Robo2 expression (Supplementary Fig. S1D); consequently, Robo1-shRNA1 was used in subsequent experiments.
To confirm the effect of Robo1-shRNA1 in vivo, we introduced Robo1-shRNA1 together with EGFP-expression vector (pCAGGS-EGFP) into the lateral ventricle of E15.5 embryos using in utero electroporation. At P1, reduced Robo1 protein expression was observed in EGFP-positive cells transfected with Robo1-shRNA compared with control-shRNA in the mouse cortex (Supplementary Fig. S1E). We further isolated EGFP-positive cells from the P1 mouse neocortex using FACS, and examined Robo1 mRNA expression by quantitative RT-PCR analysis (Supplementary Fig. S1F). Robo1 mRNA expression level in the Robo1-shRNA-transfected cells was significantly reduced than that in the control cells (Supplementary Fig. S1F), whereas Robo2 mRNA expression was unaffected (Supplementary Fig. S1F). We further examined the amount of total Robo1 protein by dissecting EGFP-expressing neocortical tissues at P1. The amount of Robo1 protein was also significantly decreased by immunoblot (Supplementary Fig. S1G). Together, these results indicate that Robo1-shRNA1 does effectively reduce the expression of endogenous Robo1 in vivo.
Robo1-Suppressed Neurons Exhibit a Delay in their Migration to the Cortical Plate
To investigate the role of Robo1 in the development of neocortical layers II/III pyramidal neurons, we introduced Robo1-shRNA1 or control-shRNA together with pCAGGS-EGFP into the lateral ventricle of E15.5 embryos using in utero electroporation as described above. Previous studies demonstrated that newborn neurons generated in the VZ at E15.5 differentiate into projection neurons of layers II/III (Hatanaka et al. 2004). To validate the specificity of shRNA transfection in cortical pyramidal cells, we examined the expression of cell type markers, Satb2 (cortical callosal projection neuron marker; Alcamo et al. 2008; Britanova et al. 2008), calbindin (cortical interneuron marker; Anderson et al. 1997), and GFAP (astrocyte marker; DeArmond et al. 1980) in EGFP-positive cells by immunohistochemistry (Supplementary Fig. S2). We found the co-expression of EGFP with Satb2-positive (Supplementary Fig. S2C,F), but not calbindin-positive (Supplementary Fig. S2I,L), or GFAP-positive (Supplementary Fig. S2O,R) cells, demonstrating that Robo1-shRNA electroporation into the lateral ventricle at E15.5 specifically targets layers II/III pyramidal cells without affecting their neuronal identity.
Since the laminar organization was disrupted in the Robo1−/− cortex (Fig. 2), we selectively suppressed Robo1 in layers II/III neurons and examined their positions at the early postnatal period. Specifically, when Robo1-shRNA constructs were electroporated at E15.5 and examined at P1, >35% of EGFP-positive cells reached the superficial portion of the CP (bin 1) in control cortices (Fig. 3A,B). By contrast, in Robo1-shRNA1-transfected cortices, about 25% of EGFP-positive cells reached the superficial part of the CP (bin 1) at P1, but many cells were diffusely distributed throughout the CP (Fig. 3A,B). A similar result was obtained using Robo1-shRNA4 (Fig. 3A). These results suggest that suppression of Robo1 may cause a delay in neuronal migration toward the pial surface.
To address this, we assessed the localization of EGFP-positive cells at earlier time point post-electroporation. We labeled neurons born at E15.5 by injecting EdU into pregnant dams 30 min prior to shRNA electroporation. The positions of both EdU single-labeled and EdU/EGFP co-labeled neurons were assessed at E18.5 (Fig. 4). Consistent with a previous report (Hatanaka et al. 2004), many E15.5 EGFP-transfected neurons that co-labeled with EdU have entered the CP in control cortices at E18.5 (Fig. 4A,A′, 50.7% in bins 1–8 in Fig. 4C). In contrast, we observed only few EdU+/EGFP+ cells entering the CP in Robo1-shRNA1-transfected cortices at this stage (Fig. 4B,B′, 25.5% in bins 1–8 in Fig. 4C), whereas EdU+/EGFP-cells were unaffected (Fig. 4A′,B′,C). Together, these data indicate that Robo1 suppression causes a delay in upper-layer pyramidal neuron migration, by affecting the timing of entry into the CP.
Robo1 is Required for Inside-Out Segregation of Layers II/III Pyramidal Cells
Our results demonstrate that loss of Robo1 affects the early phase of radial migration of layers II/III pyramidal neurons. To further explore the positioning defect caused by Robo1 suppression at later stages, we assessed the distribution of Robo1-shRNA-transfected neurons at P4, P8, P12, and P15 (Fig. 5). At P4, most EGFP-positive cells of Robo1-shRNA-transfected cortices have finally reached the superficial portion of the CP; however, a small number of EGFP-positive cells were still present in middle, and deeper positions (Fig. 5A and data not shown). These results indicate that although Robo1-shRNA-transfected neurons exhibit a delay in entering the CP, these neurons are still capable of migrating to the outermost position of the CP, regardless of suppression of Robo1. However, after P8, many EGFP-positive cells were located in the upper portion of layers II/III beneath the MZ (Fig. 5A,B), showing a distribution substantially different from control neurons. Whereas control EGFP-positive neurons were distributed throughout the mid- to upper portion of layers II/III (bins 1–3 in Fig. 5B), Robo1-shRNA1 transfected were predominantly located in the most superficial part (bin 1) of these layers (P8; control: 40.0 ± 3.6% in bin 1 and 55.6 ± 1.7% in bin 2, Robo1-shRNA1: 83.3 ± 3.4% in bin 1 and 13.5 ± 1.6% in bin 2, n = 12 sections, P< 0.01, P15; control: 49.5 ± 5.8% in bin 1 and 41.4 ± 1.3% in bin 2, Robo1-shRNA1: 84.7 ± 2.7% in bin 1 and 13.6 ± 2.1% in bin 2, n = 12 sections, P< 0.01). Such differences were not observed in the Mt-Robo1-shRNA1-transfected cells (data not shown). To further assess whether the positional changes in Robo1-shRNA1-transfected cells were caused by specific suppression of Robo1 expression, we transfected Robo1 cDNA (pCAGGS-Robo1) together with Robo1-shRNA1 and pCAGGS-EGFP into the lateral ventricle of the E15.5 cortex. Radial distribution patterns of Robo1-shRNA1 plus Robo1 cDNA-transfected EGFP-positive cells were identical to those of control-shRNA-transfected cells in this case (data not shown). These results indicate that the distinct localization observed in Robo1-shRNA1-transfected cells was caused by a specific suppression of Robo1 expression.
Our results have so far demonstrated that within layers II/III pyramidal cells, suppression of Robo1 expression causes: 1) delay in neurons to enter the CP; 2) relatively normal radial migration, where Robo1-suppressed neurons eventually reach the upper CP by P4; and 3) abnormal distribution of neurons after the first postnatal week, where neurons take up the outermost positions in the CP immediately underneath the MZ. Together, these results indicate that loss of Robo1 has a significant impact on the positioning of layers II/III neurons even after they have reached the pial surface. It is known that pyramidal neurons of the neocortex follow an inside-out neuronal distribution, where later-born neurons migrate past earlier-born cells to dominate a more superficial position within the CP. However, the cell and molecular mechanism by which the relative positioning of these neurons is established remains unclear (Magdaleno et al. 2002; Yoshida et al. 2006; Cooper 2008; Kubo et al. 2010). Since Robo1-suppressed cells appear to localize to the most superficial region within the neocortex, even after the first postnatal week (Fig. 5A,B), when the layer organization has been well-defined (Auladell et al. 1995), we next assessed the distribution pattern of temporal cohorts of neurons by performing sequential in utero electroporation analysis. When control-shRNA was transfected with pCAGGS-EGFP at E16.5, the majority of EGFP-positive neurons gave rise to the uppermost layers II/III neurons in the neocortex (bin 1 in Supplementary Fig. S3), whereas electroporation at E17.5 resulted in mainly gliogenesis (data not shown). This indicated that electroporation at E16.5 targets the latest cohorts of layers II/III pyramidal cells in the neocortex. We, therefore, co-electroporated control-shRNA or Robo1-shRNA1 together with pCAGGS-EGFP vector at E15.5, and the same cortices were subsequently electroporated with same shRNA constructs together with pCAGGS-DsRed at E16.5 (Fig. 6). As expected, control-shRNA-transfected cortices exhibited a typical inside-out distribution of temporal neuronal cohorts at P8, where E16.5 control-shRNA-electroporated DsRed-positive cells were localized superficial to E15.5 control-shRNA-electroporated EGFP-positive cells (Fig. 6C). Notably, Robo1-suppressed layers II/III neurons did not exhibit a characteristic inside-out distribution and, instead, later-born neurons were intermingled within the residual space that was unoccupied by E15.5-transfected EGFP-positive neurons (Fig. 6F). Furthermore, both E15.5 and E16.5 Robo1-shRNA-transfected neurons exhibited a dense distribution in the radial extent of the CP, resulting in a more compact lamina of electroporated cells, similar to that of upper CP neurons observed in the Robo1 null cortex (Fig. 2). Together, these results indicate that upper-layer pyramidal neurons require Robo1 for their correct positioning and inside-out segregation within the CP.
We further examined whether suppression of Robo1 expression has any effect in the dendritic development of cortical pyramidal neurons, by assessing EGFP-positive cells located in layers II/III of E15.5 Robo1-shRNA-transfected cortices (Fig. 7). Although differences in morphology between Robo1-shRNA- and control-shRNA-transfected EGFP-positive cells were not apparent at P4 (Fig. 7A), marked differences in the growth and branching of neurites were observed at a subsequent stage. At P8, Robo1-shRNA1-transfected cells exhibited a more complex neurite branching at the apical side than control-shRNA-transfected cells (Fig. 7A). Quantitative analysis revealed that the number of neurites at the apical side in Robo1-suppressed cells was significantly larger than that in control cells (control: 1.4 ± 0.2, Robo1-shRNA1: 5.0 ± 0.4, n = 25 cells, P< 0.01), whereas no difference was observed at the basal side (control: 2.5 ± 0.2, Robo1-shRNA1: 2.7 ± 0.4, n = 25 cells, P< 0.01) (Fig. 7B). To further assess whether distinctive positioning and neuronal morphology of Robo1-suppressed cells were secondary effect to changes in their identity after migrating to their final position, we examined the expression of cell type specific markers in EGFP-positive cells at P8. Immunohistochemical analysis revealed that these cells express Brn2 (layers II/III pyramidal neuron marker) and NeuN (neuronal marker), but not GFAP, similar to control EGFP-positive cells (data not shown). Together with the analysis of short-term fate mapping of Robo1-shRNA1-transfected cells (Supplementary Fig. S2), these results demonstrate that Robo1-suppressed neurons show distinctive positioning without an overt change in their identity.
To further validate the specific requirement for Robo1 in the distribution of layers II/III pyramidal cells, we next manipulated Robo1-expression levels within cortical upper-layer neurons, and assessed their effect on neuronal positioning. Our earlier expression studies indicated that Robo1 mRNA was normally expressed in layer V neurons at a higher level than those observed in layers II/III neurons (Fig. 1D). Therefore, one possibility is that the level of Robo1 expression may determine the final positions of pyramidal cells in a layer subtype-specific manner. To test this possibility, we electroporated pCAGGS-Robo1 into the E15.5 embryo neocortex, and examined the positions of EGFP-positive cells at P8 (Supplementary Fig. S4). Upon introduction of pCAGGS-Robo1, we detected higher levels of Robo1 protein expression in the VZ, IMZ, and CP when compared with control cortices transfected with pCAGGS (Supplementary Fig. S4A and data not shown), whereas the position of EGFP-positive cell showed no significant changes in their distribution patterns (Supplementary Fig. S4B,C). These results showed that Robo1-overexpressing layers II/III neurons do not exhibit a shift to a deeper position as that of layer V neurons within the CP. Notably, however, we found an increase in the numbers of EGFP-positive axons projecting into the striatum (Supplementary Fig. S5D), which were normally not observed in control EGFP-transfected layers II/III neurons (Supplementary Fig. S5B). These results implied that Robo1-expression levels may be important for refining the axonal projection patterns within distinct layer subtypes. Finally, we transfected Robo1-shRNA in mouse neocortices at E14.5, where the majority of neurons consists of layer IV neurons that do not normally express Robo1 at perinatal stages (Fig. 1), and compared the positions of electroporated cells at P8. In these cases, the laminar distribution patterns and morphology of both Robo1- and control-shRNA-transfected cells were identical (Fig. 8C,F). These results indicate that layers II/III neurons have a specific requirement for Robo1 in their normal distribution, consistent with the restricted expression of Robo1 within the upper-layer neocortex.
To date, growing numbers of genes have been identified that control the movement of neocortical neurons during the early steps of their radial migration (Caviness and Rakic 1978; Walsh and Goffinet 2000; Tsai and Gleeson 2005; Huang 2009; Honda et al. 2011). In contrast, the mechanisms by which neuronal distribution of neocortical pyramidal cells is controlled during terminal stages of migration have remained largely elusive. Here, using an in vivo loss-of-function approach with RNA interference and Robo1 knockout mice, we directly demonstrate that Robo1 expressed in upper-layer pyramidal neurons plays a significant role in determining their positions within the neocortex.
Robo1 Regulates the Laminar Distribution of Layers II/III Pyramidal Cells
Our studies show that Robo1 is expressed in layers II/III pyramidal neurons during the critical period of their migration and final positioning. Notably, while the level of Robo1 mRNA expression remains constant in deep-layer neurons (layer V), its expression in the upper part of layers II/III is upregulated during the early postnatal period (Fig. 1). Moreover, Robo1 protein is predominantly distributed in the apical neurites of these neurons and primarily near the MZ (Fig. 1P′,U′). In agreement with its expression pattern, we found that Robo1 plays a critical role in the positioning of these neurons during postnatal stages, where the major changes in the neuronal distribution of Robo1-suppressed layers II/III pyramidal cells occur between P4 and P8 (Fig. 5A). This implies that possible interactions between Robo1 and other molecules during the first postnatal week may be crucial in determining the final positions of these cells, with those that lack Robo1 failing to descend to deeper positions in the CP.
What, then, is the cellular mechanism by which Robo1 regulates the correct positioning of layers II/III pyramidal neurons? In principle, the precise inside-out distribution of temporal cohorts of cortical neurons can be achieved through a number of different mechanisms. Most simply, Robo1 may regulate correct positioning of layers II/III neurons by promoting their early migration (Fig. 4), as delay in entering the CP would result in their more superficial positioning. However, although our results demonstrate that loss of Robo1 affects the time of entry of layers II/III neurons into the CP, hence early migration is delayed, we found that most E15.5 Robo1-suppressed cells are able to reach the pia by P4, when positional defects are least prominent. Furthermore, if all Robo1-suppressed cells were delayed in their migration in a similar manner, one would expect such cells born at E15.5 to take deeper positions within the CP than cells born at E16.5. However, this appears not to be the case in the sequential electroporation experiments (Fig. 6). As such, these results imply that neuronal positioning of layers II/III pyramidal neurons requires later action of Robo1 after they reach the pial surface, rather than simply carrying over the migration delay from earlier development.
It has been proposed that the interaction between migrating neurons and the molecules expressed in the MZ has a prominent influence in facilitating or ceasing neuronal migration at the appropriate time. By far the most studied, extracellular matrix glycoprotein Reelin and its signaling pathway play critical roles in radial migration and influences the outcome of neocortical layer formation (Caviness and Sidman 1973; D'Arcangelo et al. 1995; Ogawa et al. 1995; Howell et al. 1997; Sheldon et al. 1997). Notably, the phenotype caused by Robo1 suppression is clearly distinct from that of known Reelin pathway interference analysis (Trommsdorff et al. 1999; Olson et al. 2006; Feng et al. 2007; Hack et al. 2007; Hashimoto-Torii et al. 2008; Yano et al. 2010; Franco et al. 2011; Sekine et al. 2011). Whereas reelin and Dab1 mutant mice exhibit an inverted laminar neuron distribution, cells of layers II/III that lack Robo1 expression appear to be anchored to the most superficial part of the CP, beneath the MZ. One possibility is that the detachment of migrating pyramidal neurons from radial glial fibers may be affected in the absence of Robo1 through misregulation of cell–cell and cell–substrate adhesion molecules. Although we did not observe obvious clusters of radial glia-associated Robo1-suppressed cells in our analysis, it has been demonstrated that Slit/Robo signaling inhibits cell adhesion during neurite outgrowth and axon extension in chick neural retina (Rhee et al. 2002, 2007). Thus, Robo signaling may also play a role in the adhesive interaction between cortical pyramidal cells and radial glia through similar mechanisms.
A second and intriguing hypothesis is that Robo1 may regulate correct positioning of layers II/III neurons through their dendritic development. Following detachment from radial glia, pyramidal cells extend apical dendrites and elaborate branching toward their target layers depending on the position of their cell body (Barnes and Polleux 2009). In our studies, Robo1-suppressed layers II/III pyramidal cells exhibited dense distribution in the radial extent of CP when compared with that of control cells (Figs 2 and 6). These observations suggest that the failure of Robo1-suppressed neurons to interact with local cues may prohibit their normal dendritic extension, which leads to abnormally dense neuronal distribution along the radial axis (Fig. 6F). In Robo1-suppressed layers II/III cells, their superficial positioning may not be the primary cause of the changes in their dendritic morphology, as both E16.5 control- and Robo1-shRNA-transfected cells showed similar neuronal distributions (Supplementary Fig. S3), but the morphological differences were still apparent (Fig. 6E). Indeed, it has been suggested that defect in dendritogenesis may be instructive, rather than a result of perturbed layer formation (Teng et al. 2001; Olson et al. 2006; Nichols and Olson 2010; Franco et al. 2011). Thus, further investigations on how distinctive dendritic patterning is achieved in Robo1-suppressed neurons (Fig. 7), may provide novel insights into understanding the relationship between dendritogenesis and fine-tuning of layer construction in the neocortex.
Requirement for Robo1 in Cortical Neuron Subtypes
Earlier studies on the role of Robo1 in neocortical development have provided compelling evidence for its involvement in interneuron development and axonal path formation (Andrews et al. 2006, 2008; López-Bendito et al. 2007; Hernandez-Miranda et al. 2011). In contrast, the specific roles for Robo1 in upper-layer pyramidal neurons have remained elusive. Our RNA interference approach has enabled us to dissociate the requirement for Robo1 in the development of layers II/III pyramidal neurons, as they terminate migration and assume their final positions during the first postnatal week. Interestingly, although it has been suggested that later-born neurons may utilize a common migration strategy (Nadarajah et al. 2001; Hatanaka et al. 2004), our temporal electroporation analysis has revealed that E14.5 Robo1-shRNA-transfected cells, unlike that of E15.5 or E16.5 neurons, do not exhibit changes in their positioning (Fig. 8F). Our observation is consistent with a previous study that cortical neurons born at E14 have a distinct character from that of layers II/III neurons (Ajioka and Nakajima 2005). Since the majority of E14 born neurons adopt a layer IV fate (Takahashi et al. 1999) and do not express detectable levels of Robo1 (Fig. 1), these results imply that Robo signaling may act in a temporally and spatially restricted manner, where layer IV neurons are refractory to loss of Robo1 expression. Indeed, a new study examining the role of Robo4, a different Robo family receptor expressed in the neocortex reported that the laminar organization of the neocortex develop normally in cortex-specific Robo4 knockout mice, but the radial migration of E14.5 Robo4-suppressed mouse cortical neurons was delayed (Zheng et al. 2011). These results highlight the importance of distinct Robo receptors in regulating neuronal migration and positioning in a layer-specific manner.
In summary, we have presented evidence that Robo1 regulates the positioning of layers II/III pyramidal neurons in the developing neocortex. Collectively with previous reports (Hatanaka et al. 2004; Hack et al. 2007), our current study suggests that although laminar identities of neurons may be largely determined at birth (McConnell and Kaznowski 1991), the mechanisms by which layer-specific neurons navigate to their final positions may involve cell type- and context-dependent combinatorial codes that further refine the formation of neocortical laminae.
This work was supported by grants from the Ministry of Health, Labor and Welfare of Japan and the Ministry of Education, Culture, Sports, and Science and Technology of Japan, Grant-in-Aid for Scientific Research on Innovative Areas “Neural Diversity and Neocortical Organization,” from the Wellcome Trust (Program Grant 089775 to J.G.P. and W.D.A.) and from the Strategic Research Program for Brain Sciences “Understanding of molecular and environmental bases for brain health.”
We thank Dr Fujio Murakami for generous gift of anti-Robo1 antibody, Dr Satoru Masuda for FACS experiment, Dr Shigeki Yuasa, and Dr Toshio Terashima for valuable advice and all members of the Hanashima Laboratory for valuable discussions. Conflict of interest: none declared.