COUP-TFI promotes radial migration and proper morphology of callosal projection neurons by repressing Rnd2 expression

The mammalian neocortex has undergone pronounced expansion in size and complexity during evolution and is responsible for high cognitive function, sensory perception and consciousness (Northcutt and Kaas, 1995). Newborn neurons migrate from the ventricular zone (VZ) to the pial surface and settle in the cortical plate (CP) in a sequential manner with deeper layers (DLs) formed first, followed by upper layers (ULs) (Rakic, 1978; Super and Uylings, 2001). The latter are developmentally and phylogenetically the last structure to appear in the mammalian cortex (Aboitiz et al., 2003; Molnar et al., 2006). Novel specialized neuronal subpopulations with a high degree of complexity of dendritic and axonal arborization have evolved in the uppermost layers. Callosal projection neurons (CPNs), for example, represent a heterogeneous population of UL pyramidal neurons prevalently located in layers II and III of the neocortex (Molyneaux et al., 2009; Ramos et al., 2008) and projecting either ipsilaterally or to homotypic regions of the contralateral hemisphere, giving rise to the corpus callosum (CC) (Mitchell and Macklis, 2005; Richards et al., 2004).

It has been hypothesized that the inside-out pattern of corticogenesis leading to the birth of the six-layered neocortex was made possible by the appearance of glia-guided neuronal migration (Aboitiz et al., 2003; Molnar et al., 2006). Early-born neurons give rise to layers VI and V (phylogenetically older structures), essentially by somal translocation (Miyata et al., 2004; Nadarajah et al., 2002), whereas late-born neurons follow a complex pattern of migration. During early phases, late-born neurons adopt a multipolar morphology and migrate through the subventricular zone (SVZ) independently of the radial glia scaffold (LoTurco and Bai, 2006; Tabata and Nakajima, 2003). In the upper intermediate zone (IZ), migrating neurons assume a bipolar shape, attach to radial glia and move to the CP by glia-guided locomotion (LoTurco and Bai, 2006; Tabata and Nakajima, 2003). Finally, during late phases of the migratory process, bipolar neurons contact the pial surface detaching from the glia scaffold and reach their target by somal translocation, thus allowing glia-guided upcoming neurons to reach upper positions (Nadarajah et al., 2001).

In the last 20 years, many molecules have been reported to be required for different steps of this process. For example, Lis1 (Pafah1b1 – Mouse Genome Informatics), Dcx, FlnA (LoTurco and Bai, 2006), Cdk5 (Ohshima et al., 2007) and RhoA (Ge et al., 2006; Hand et al., 2005) control multipolar-shaped cell (MSC) migration and the transition from MSCs to bipolar-shaped cells (BSCs) in the IZ. Other molecules, like Mdga1 (Takeuchi and O’Leary, 2006), N-cofilin (Bellenchi et al., 2007), α3β1 and α5β1 integrins (Dulabon et al., 2000; Marchetti et al., 2010) and Dcx are implicated in the maintenance of bipolar cell polarity and interaction of migrating neurons with the radial glia. Finally, the secreted molecule Reelin and its downstream effector Dab1 guide the detachment of migrating neurons from the radial glia, allowing the final somal translocation step beneath the pial surface (Dulabon et al., 2000; Olson et al., 2006; Sanada et al., 2004). Remarkably, disruption of many of these molecules not only leads to impairment in radial migration and lamination, but also causes a striking alteration in dendritic and axonal morphology of UL neurons. However, the correlation between cell shape transition from MSC to BSC and the final morphology and connectivity of UL neurons is still elusive.

Although the number of molecules found to be involved in the regulation of glia-guided radial migration has increased remarkably in recent years, very little is known about their transcriptional regulation. The proneural gene Ngn2 (Neurog2 – Mouse Genome Informatics) was the first factor discovered to activate directly the expression of one of these factors, the small Rho-GTPase Rnd2, which is involved in the regulation of MSC-to-BSC transition and BSC migration during mid-late stages of corticogenesis (Heng et al., 2008; Nakamura et al., 2006). Although Rnd2 transcription is activated by Ngn2 primarily in MSCs along the SVZ and the lower IZ (Heng et al., 2008; Nakamura et al., 2006), its expression was also shown to be enhanced by NeuroD2, a transcription factor that is strongly expressed in the IZ and CP during mid-late stages of corticogenesis (Heng et al., 2008). However, Rnd2 expression is sharply downregulated in the upper IZ, where BSCs attach to the glia and migrate to the CP, suggesting that an independent pathway might negatively regulate Rnd2 expression in this compartment and that such a step might be relevant for correct BSC migration to the CP.

In this study, we found that in the absence of COUP-TFI (Nr2f1 – Mouse Genome Informatics), a transcription factor crucial for corticogenesis and arealization (Armentano et al., 2006; Armentano et al., 2007; Tomassy et al., 2010), Rnd2 expression is abnormally high in the SVZ and IZ during glia-guided radial migration. As a consequence, a greater number of MSCs are stalled in the IZ and BSC migration to the CP is impaired. Although normally specified, CPNs show defects in their postnatal dendritic morphology and in the elongation and arborization of their axons. Unexpectedly, lowering Rnd2 levels in COUP-TFI-deficient neurons rescues migratory defects of COUP-TFI^–/– CPNs, which remarkably recover their morphology in a cell-autonomous manner. In conclusion, we demonstrate that Rnd2 expression levels are precisely fine-tuned by COUP-TFI and are essential for CPN radial migration and maturation.

COUP-TFI null (COUP-TFI^–/–) and COUP-TFI fl/fl mice were generated and genotyped as previously shown (Armentano et al., 2006; Armentano et al., 2007). COUP-TFI fl/fl mice were either crossed to Emx1-Cre for inactivating COUP-TFI in progenitors (COUP-TFI CKO-Emx1) (Armentano et al., 2007) or to Nex-Cre for COUP-TFI inactivation in post-mitotic cells (Goebbels et al., 2006). COUP-TFI^–/– were crossed to the Ngn2^KIGFP mouse line (Seibt et al., 2003) to generate double mutants, COUP-TFI/Ngn2 dKO. Midday of the day of the vaginal plug was embryonic day (E) 0.5. All experiments were conducted following guidelines of the Institutional Animal Care and Use Committee, Cardarelli Hospital, Naples, Italy.

Timed pregnant females received a single intraperitoneal injection of 5-bromo-2′-deoxyuridine (BrdU; 50 mg/kg) at 14.5 days post-coitum (dpc) and three wild-type (wt) and three COUP-TFI^–/– embryos were collected at E16.5. Five regions (three sections per region) were arbitrarily chosen along the rostrocaudal axis and BrdU+ cells were counted with the NIH ImageJ software in a 250 μm-wide portion of the IZ and the CP compartments of the lateral cortex. The number of CP BrdU+ cells was expressed as a percentage of total CP/IZ BrdU+ cells.

Vibratome and cryostat sections were processed for immunocytochemistry and/or in situ hybridization as previously described (Armentano et al., 2007). The following primary antibodies were used: rabbit anti-GFP antibody (1:1000, Chemicon); goat anti-Cux1 (1:50, Santa Cruz); mouse anti-BrdU (1:300, Sigma); rabbit anti-activated caspase 3 (1:100, Cell Signaling Technology); mouse anti-human COUP-TFI (1:100, Abcam); rabbit anti-COUP-TFI (1:500) (Tripodi et al., 2004); rat anti-CTIP2 (1:500, Abcam); mouse anti-Nestin (1:100, Chemicon); rabbit anti-Pax6 (1:100, Chemicon); mouse anti-Satb2 (1:20, Abcam); rabbit anti-Tbr1 (1:1000, gift from R. Hevner, University of Washington, Seattle, USA) and rabbit anti-Cre (1:100, Covance). The following secondary antibodies were used: Alexa Fluor 488-conjugated anti-rabbit; Alexa Fluor 594-conjugated anti-rabbit; Alexa Fluor 594-conjugated anti-mouse; Alexa Fluor 488-conjugated anti-mouse (1:400, Molecular Probes). Antisense RNA Probes were labelled using a DIG-RNA Labelling Kit (Roche).

The pCIG2-IRES-GFP, pCIG2-Cre-IRES-GFP, Rnd2-specific (SH1, SH2) and scrambled (SCR) shRNA-IRES-GFP (Heng et al., 2008) plasmids were injected into E14.5 embryonic vesicles with a PV820 Pneumatic PicoPump (WPI). The electroporations were performed on whole heads using a Tweezertrode electrode (diameter 7 mm; BTX) connected to a CUY 21 EDIT electroporator (NEPA GENE) with the following parameters: five 50 V pulses, P(on) 50 mseconds, P(off) 1 second. Brains were then embedded in 3% agarose and cut using a vibratome (Leica VT1000S) at 300 μm thickness. Slices were cultured for 4 days in Neurobasal Medium (GIBCO) plus N12 and B27 supplements (GIBCO), as previously described (Tripodi et al., 2004). In utero electroporations were performed as previously described (Nguyen et al., 2006) with the following parameters: four 40 V pulses, P(on) 50 mseconds, P(off) 1 second. Analysis of GFP+ cell distribution was performed on the lateral cortex along the rostrocaudal axis. The cortex was subdivided into VZ/SVZ, IZ and CP compartments, and the IZ and CP were further subdivided in lower, medial and upper subcompartments (Heng et al., 2008; Nguyen et al., 2006). Cell counts in the different subcompartments were performed on at least three 10 μm thick planes scanned at different depths of each single slide. The number of GFP+ cells for each compartment was expressed as a percentage of the total number of GFP+ cells. The mean value of GFP+ cells counted in each compartment of the three planes was used to calculate the average value among the same neocortical compartments of rostrocaudal slices from at least three littermates for each genotype.

The neocortex of three wt and three COUP-TFI^–/– brains was dissected and immediately transferred to 500 μl of Trizol (Invitrogen) and processed for total RNA isolation according to the manufacturer’s protocol. RNA (1 μg) was reverse transcribed using a QuantiTect Reverse Transcription Kit (Qiagen). Real-time PCR (QPCR) analysis was performed for all the littermates in triplicate on a 7900HT Fast Real Time PCR System (Applied Biosystems). The PCR reaction was performed using Power SYBR Green PCR Master Mix (Applied Biosystems) according to the manufacturer’s protocol. All the assays were normalized with respect to Gapdh values. Fold-change variations in the levels of mRNA of interest were expressed as a percentage and normalized against wt levels (set as 100%). For primers see Table S2 in the supplementary material.

For each experiment (n=3), neocortices were dissected from 24 mice at E14.5 and diced in ice cold Hanks Buffered Saline Solution (GIBCO). Nuclei were processed and ChIP was performed as described in Kuo and Allis, 1999 (Kuo and Allis, 1999). Antisera directed against COUP-TFI (15 μl) or 3 μg Dicer (Santa Cruz sc30226 as Rabbit IgG control) were used for immunoprecipitation. Putative COUP-TFI binding sites were tested by QPCR analysis of ChIP samples using the LightCycler 480 Real-Time PCR System (Roche). Reactions were performed in triplicate on three independently prepared ChIP samples. Amplification was expressed as fold enrichment compared with non-specific IgG control and relative quantities of immunoprecipitated DNA were calculated as previously described (Heng et al., 2008). For primers see Table S2 in the supplementary material.

All the data were statistically analyzed and graphically represented using Microsoft Office Excel software. The error bars represent s.e.m. Two-tailed Student’s t-test was used for the analysis of statistical significance (^*P≤0.05, ^**P≤0.01, ^***P≤0.005). One-way analysis of variance (ANOVA) was used to compare the means among three or more groups of samples and post-hoc analysis was performed by two-tailed Student’s t-test.

Images were acquired with an Upright microscope Leica DM5000B equipped with Leica IM image management software (Leica Microsystems, Wetzlar, Germany). A Zeiss LSM 710 Laser Scanning Microscope (Carl Zeiss MicroImaging, Jena, Germany) and a Fluorescence Inverted Confocal Microscope LEICA SP2 AOBS (Leica Microsystems, Heidelberg, Germany) were used for image acquisition of ex vivo and in utero electroporated brain slices. A total of 23 z-line scans with a step distance of 0.45 μm were collected at different depths along the z-axis of each lateral neocortex (one to four different regions along the z-axis were scanned depending on the thickness of the electroporated area) and maximum intensity projections were generated either using Zeiss ZEN Confocal Software (Carl Zeiss MicroImaging) or using Leica Confocal Software (Leica Microsystems, Wetzlar, Germany).

The mouse embryocarcinoma cell line P19 was transfected at 60% confluence with pCIG2-[Flag-Rnd2]-IRES-GFP and shRNA-IRES-GFP vectors (Heng et al., 2008) using Lipofectamine 2000 Transfection Reagent (Invitrogen) according to the manufacturer’s protocol. Total protein extraction and western blotting were performed as previously described (Armentano et al., 2006). The following primary antibodies were used: mouse monoclonal anti-Flag (1:1000, Sigma), mouse anti-β3 tubulin (1:1000, Sigma); and appropriate secondary antibodies: goat anti-mouse IgG (H+L) horseradish peroxidase conjugate (1:5000, Bio-Rad). The signal was detected with ECL reagent (Amersham Biosciences).

Brains of three wt and five COUP-TFI CKO-Emx1 P60 mice were vibratome-sectioned coronally at a thickness of 50 μm and processed for Nissl staining, as previously described (Tomassy et al., 2010). Pictures were taken on a LEICA MZ 16 FA Stereomicroscope and converted from RGB to BW by Adobe Photoshop software. Section images were identified with serial numbers and analyzed in both hemispheres, spanning the entire tract. Using a graphic pen, all the different callosal components were labelled with different colours, and images were aligned and integrated by Amira 4.1.2-1 software to obtain 3D pictures.

To investigate the mechanisms underlying alteration of ULs in the absence of COUP-TFI function (Armentano et al., 2007; Faedo et al., 2008; Tomassy et al., 2010), we first assessed whether late-born neurons were normally specified in COUP-TFI^–/– mice. The expression of Satb2, required for late specification of CPNs (Alcamo et al., 2008; Britanova et al., 2008), as well as the expression of Cux1 and Cux2, involved in early and late differentiation of late-born neurons (Cubelos et al., 2010; Nieto et al., 2004; Zimmer et al., 2004), were affected in mutant ULs at P0 (Fig. 1A). The dense Satb2- and Cux1-positive upper cell layer was reduced in thickness and scattered Cux1- and Cux2-positive cells were observed along the radial extent of the mutant cortex. At E16.5, when late-born neurons are predominantly migrating, increased expression of Satb2 and Tbr1, a marker of post-mitotic projection neurons (Hevner et al., 2001) was observed in the upper SVZ and IZ of mutant brains (Fig. 1B,C). No double positive cells for Satb2 and Tbr2 (Eomes – Mouse Genome Informatics), an intermediate progenitor marker (Englund et al., 2005; Kowalczyk et al., 2009; Sessa et al., 2008) were found in COUP-TFI^–/– brains, similarly to control brains (see Fig. S1A in the supplementary material), excluding precocious expression of Satb2 in early post-mitotic neurons. In addition, Satb2 expression was decreased in the CP and Cux1 and Cux2 showed weak or absent expression in the CP (asterisks in Fig. 1D), suggesting overall abnormal cell migration rather than abnormal specification of late-born neurons in the absence of COUP-TFI function.

To investigate further whether absence of COUP-TFI has an effect on radial cell migration during mid-late corticogenesis, we injected BrdU into pregnant females at 14.5 dpc and evaluated the percentage of BrdU+ cells in the CP of E16.5 wt and COUP-TFI^–/– brains along their rostrocaudal axis (Fig. 2A,B; see Table S1 in the supplementary material). The number of BrdU+ cells in the CP was expressed as a percentage of the BrdU+ cells in the IZ/CP regions (Fig. 2B) to avoid any interference derived from increased proliferation of mutant progenitor cells in the VZ/SVZ (Faedo et al., 2008). Interestingly, we observed a significantly lower percentage of cells reaching the CP in mediolateral regions of mutant neocortices (presumptive somatosensory area), whereas rostral- and caudal-most regions were unaffected (Fig. 2B; see Table S1 in the supplementary material).

Next, we assessed directly the behaviour of abnormally migrating cells by ex vivo electroporation (Hand et al., 2005) of a green fluorescent protein (GFP)-expressing vector in the VZ of wt and COUP-TFI^–/– E14.5 brains (Fig. 2C). Slices from rostral to caudal regions of electroporated brains were analyzed after 4 days of culture in vitro (DIV). We observed a striking reduction of neurons reaching the CP principally from rostro- to caudomedial regions of mutant cortices (Fig. 2C-E; see Table S1 in the supplementary material). This defective migration was neither due to impairment of the glial scaffolding nor to abnormal specification of migrating neurons, because almost all GFP-positive cells failing to leave the IZ were Satb2-positive (UL) and Ctip2 (Bcl11b – Mouse Genome Informatics)-negative (deep layers) (see Fig. S2A-D in the supplementary material).

Notably, cell counting revealed a gradual increase in the percentage of GFP+ cells reaching the CP from rostral to caudal regions of wt cortices, with a maximum rate of migration in the caudomedial cortex (Fig. 2F). This migration gradient was lost in COUP-TFI^–/– brains, although no differences were found between rostral- and caudal-most neocortical regions of wt and COUP-TFI^–/– brains (Fig. 2F; see Table S1 in the supplementary material). To exclude any interference from the neurogenic gradient, we limited our statistical analysis of GFP+ cell distribution to the IZ and CP as well as to the lower, middle and upper regions within the CP (Fig. 2E; see Fig. S3A,B in the supplementary material). This analysis confirmed defective cell transition from the IZ to the CP in rostro- to caudomedial regions and showed a significant decrease in cells reaching the upper CP also in caudal-most slices. Analysis of cell migration within the CP also revealed impaired transition from lower to upper CP in rostromedial to caudal regions of COUP-TFI^–/– cortices (see Fig. S3B in the supplementary material). Together, these data indicate that COUP-TFI regulates cell migration from IZ to CP in presumptive somatosensory and occipital cortex during mid-late corticogenesis.

We found that the small Rho-GTPase Rnd2, which we reported to be altered in COUP-TFI^–/– brains (Armentano et al., 2006), is expressed in a high-rostral to low-caudal gradient, complementary to that of COUP-TFI (Fig. 3A) and to the migratory gradient (Fig. 2). Because Rnd2 normally modulates cell migration (Heng et al., 2008), we hypothesized that COUP-TFI represses Rnd2 expression along the rostrocaudal axis. Indeed, we found that at E13.5, COUP-TFI inactivation abolished the Rnd2 gradient and increased Rnd2 expression in caudomedial cortical regions (Fig. 3B). Double labelling of E13.5 medial coronal sections with Tbr2 and Rnd2 confirmed higher levels of Rnd2 expression in the IZ, and more specifically in the CP, despite a normal distribution of basal progenitors (Fig. 3C).

At E15.5, when the majority of UL neurons are generated and start their migration to the CP, high levels of Rnd2 expression were maintained in COUP-TFI^–/– SVZ and IZ (Fig. 3D,E). Abnormal expression levels of Svet1, a marker of SVZ progenitors and multipolar-shaped cells (MSCs) (Sasaki et al., 2008; Tarabykin et al., 2001), and accumulation of Satb2+ cells were also observed in the SVZ and IZ of mutant brain adjacent sections, confirming that UL neurons, and more particularly CPNs, were properly specified but stalled as MSCs in the upper SVZ and IZ (Fig. 3E). Finally, Rnd2 expression levels were still slightly higher in the deep cortex of COUP-TFI^–/– brains at P0 (Fig. 3F). Taken together, our data indicate that the Rnd2 gradient is abolished in COUP-TFI^–/– brains and that Rnd2 expression is increased in abnormally migrating cells.

In light of the role for Ngn2 in promoting Rnd2 expression (Heng et al., 2008), we hypothesized that COUP-TFI and Ngn2 either regulate each other or act through independent pathways to fine-tune Rnd2 expression. To distinguish between these two possibilities, we assessed Rnd2 transcript levels relative to Ngn2 and other factors shown to promote Rnd2 expression (Heng et al., 2008). Despite the remarkable increase in Rnd2 expression in E15.5 COUP-TFI^–/– neocortices (57.9±7.8%, n=3, P=0.002), we found no significant changes in Ngn2, NeuroD2 and Tbr2 levels, and a mild decrease in NeuroD1 expression (12.1±3.1%, n=3, P=0.02) (Fig. 4A; see Table S1 in the supplementary material). This suggests independent control of Rnd2 by COUP-TFI. To confirm this hypothesis further, we generated double COUP-TFI Ngn2 compound mutants (dKO). Interestingly, we found that in dKO brains Rnd2 levels were increased in comparison with Ngn2^–/– brains (Fig. 4B), indicating that repression of Rnd2 expression by COUP-TFI occurs independently of Ngn2. This is confirmed further by the abnormally high levels of Rnd2 in COUP-TFIfl/fl^Nex-Cre brains (Fig. 2C), in which COUP-TFI is specifically inactivated in early post-mitotic cells (see Fig. S4 in the supplementary material).

Finally, by using MatInspector (Quandt et al., 1995) (http://www.genomatix.de/en/produkte/genomatix-software-suite.html), ECR browser (Loots and Ovcharenko, 2007) (http://ecrbase.dcode.org) and ChromAnalyzer (Montemayor et al., 2010), we identified five putative COUP-TFI binding sites at the Rnd2 locus with variable evolutionary conservation in mammals (Fig. 4D). Two binding sites were very close and mapped in the Rnd2 last intron (sites A, A1), whereas the other three mapped in the downstream genomic region of the Rnd2 locus (sites B, C and D). Strikingly, site B mapped between the two previously identified E-boxes enhancing Rnd2 expression in the developing cortex (Heng et al., 2008). To assess whether COUP-TFI binds to these sites in vivo, we performed chromatin immunoprecipitation (ChIP) on E14.5 telencephalic tissue with a COUP-TFI antibody (Tripodi et al., 2004). QPCR analysis of immunoprecipitated material showed enrichment of all five sites, although to different extents (Fig. 4D; see Table S1 in the supplementary material), suggesting that COUP-TFI binds to these sequences with different affinities. Together, our data strongly suggest a direct repression of Rnd2 by COUP-TFI in post-mitotic late-born cortical neurons.

Next, we carefully assessed the morphology of migrating GFP+ cells in wt and COUP-TFI^–/– medial brain slices at the sites where Rnd2 expression is highly increased (Fig. 5A) and observed a significant increase of MSCs in mutant IZ (Fig. 5B; wt: 39.4±2.1%, n=3; KO: 59.1±1.8%, n=4; P<0.005), although their radial distribution along the neocortex was unchanged compared with wt brains (Fig. 5C; see Table S1 in the supplementary material). Accordingly, the BSC percentage decreased from 59.3±1.7% in wt to 38.8±2.1% in COUP-TFI^–/– brains (P<0.005) (Fig. 5B), whereas no alterations were observed in the percentage of round-shaped cells (RSCs, another morphology adopted by migrating neurons) (see Table S1 in the supplementary material). We observed an abnormal BCS distribution along the radial extent of the cortex. Normally, the highest percentage of BSCs is found in the upper IZ and the CP; however, mutant BSCs were mainly located in the IZ (Fig. 5C), indicating that although they adopt a bipolar morphology, COUP-TFI^–/– neurons fail to reach the CP. Thus, our data indicate that COUP-TFI is required for both the morphological transition from multipolar to bipolar shape and the intrinsic migratory property of BSCs within the IZ and CP during UL neuron radial migration.

To test whether abnormally high levels of Rnd2 expression causes the migratory and morphological defects observed in COUP-TFI^–/– brains, we reduced Rnd2 expression levels by electroporation of two Rnd2-specific shRNA–IRES-GFP expressing vectors (noted SH1 and SH2 below) (Fig. 6; see Fig. S5A in the supplementary material) (Heng et al., 2008). As a control we used a scrambled shRNA (SCR). In P19 cells, a strong silencing effect on Rnd2 protein expression was observed with both sh-RNAs, although they exhibited different efficacies (see Fig. S5A in the supplementary material). Electroporation of SH2 in mutant cortices resulted in a remarkable rescue of the number of GFP+ mutant cells reaching the CP (Fig. 6A,B; NULL-SCR: 4.1±0.2%, n=3; NULL-SH2: 19.8±1.6%; n=4; P<0.001) and, within the CP, in the upper cortical region (NULL-SCR: 2.3±1.2%; NULL-SH2: 14.0±0.9%; P=0.003) (Fig. 6A,B). SH1 electroporation instead resulted in a mild decrease in the number of mutant GFP+ cells in the VZ/SVZ compared with mutant SCR-expressing cells. However, mutant SH1+ cells accumulated in the IZ and failed to reach the CP (Fig. 6A,B; see Table S1 in the supplementary material). SH2 was, thus, more effective in rescuing mutant cell transition from VZ/SVZ to CP (Fig. 6B; see Table S1 in the supplementary material), possibly because Rnd2 levels were approximately restored to normal. Conversely, SH1 reduced Rnd2 expression, both in wt and mutant cells, below a critical threshold leading to poor migratory rescue (see Fig. S5A-C and Table S1 in the supplementary material).

Next, we assessed whether rescued GFP+ cells re-acquired their correct shape during migration. The number of MSCs significantly decreased in SH2-electroporated COUP-TFI^–/– cortices, whereas there was a significant and concomitant increase of BSCs (Fig. 6C; NULL-SCR, MSC: 66.5±1.8%, BSC: 31.4±1.2%, n=3; NULL-SH2, MSC: 54.3±1.7%, BSC: 43.8±1.7%, n=4; P<0.05). Furthermore, mutant BSC radial distribution was, remarkably, rescued by SH2 electroporation, whereas no changes were observed in the localization of SH2+ mutant MSCs (Fig. 6D; see Fig. S5D and Table S1 in the supplementary material). Our findings strongly support a role for COUP-TFI, via the regulation of Rnd2, in promoting transition from MSC to BSC and in controlling correct distribution of BSCs in the IZ and CP during migration of late-born neurons.

Next, to assess whether migratory impairment would affect CPN maturation, we inactivated COUP-TFI in single migrating neurons via in utero electroporation of a Cre-IRES-GFP-expressing vector (referred to as CRE) in E14.5 COUP-TFI^flox/flox (floxed) mice. To rescue any eventual morphological defect in mutant neurons we co-electroporated SCR- or SH2 together with CRE. Analysis of Cre expression in electroporated brains at E18.5 confirmed that COUP-TFI was inactivated in single Cre-positive cells of floxed mice and that CRE electroporation did not cause cell death (see Fig. S6A-C in the supplementary material). Thus, GFP-expressing cells could be followed as individual migrating cells extending their axons in a wild-type environment (Fig. 7).

Acute inactivation of COUP-TFI resulted in a significant decrease of GFP+ neurons reaching the upper CP at E18.5 in floxed brains (Fig. 7A,B; wt_CRE/SCR: 54.8±2.3%; fl/fl_CRE/SCR: 35.4±0.8%; n=4; P<0.005). The migratory defect of mutant cells was partially rescued by co-electroporation of SH2 and CRE in floxed brains (fl/fl_CRE/SCR: 35.4±0.8%; fl/fl_CRE/SH2: 43.7% ± 2.4%; n=4, P<0.05) (Fig. 7A,B; see Table S1 in the supplementary material), indicating that COUP-TFI acts cell-autonomously through Rnd2 to control migratory properties. Moreover, CRE/SCR-electroporated floxed brains showed a strong impairment in the elongation of callosal axons, which failed to reach the midline as a compact fibre tract (Fig. 7C; n=4). Remarkably, co-electroporation of CRE and SH2 in floxed brains rescued the midline crossing by GFP+ callosal axons (Fig. 7C), strongly indicating an important role for Rnd2 in axonal extension.

To assess whether mutant callosal axon elongation was permanently arrested or just delayed, we repeated the above experiment and collected brains at P8. We observed that both wt and mutant GFP+ callosal projections were able to reach the contralateral cortex (Fig. 7D). As it is not possible to quantify the percentage of crossing fibres in electroporated brains owing to the variability in the number of cells incorporating GFP, we analyzed the extension and volume of the callosal commissure in COUP-TFIflox/flox^Emx1Cre adult brains (Armentano et al., 2007). Interestingly, we found that the rostrocaudal extension and total volume of the callosal commissure were reduced to 70.7±3.8% and 63.0±7.1%, respectively, in COUP-TFI conditional adult brains (see Fig. S7 in the supplementary material), suggesting that most callosal axons are just delayed in crossing the midline after inactivation of COUP-TFI. Whereas callosal axons normally arborize in layers II/III of contralateral somatosensory regions (S1 and S2) at P8 (Wang et al., 2007) (Fig. 7Di,ii,vii), mutant callosal axons showed poor arborization in these regions (Fig. 7Diii,iv,viii). Strikingly, in SH2-electroporated floxed brains, GFP+ axon innervation of contralateral parietal areas was partially rescued, particularly in S2 (Fig. 7Dv,vi,ix).

In wt control brains, Satb2-positive callosal neurons had a branched apical dendrite and complex apical tufts in layer I at P8 (Fig. 8A). Although the cell-autonomous inactivation of COUP-TFI strongly affected the morphology of these tufts, which barely reached the marginal zone (MZ) (Fig. 8A,B; see Fig. S8A,B and Movies 1 and 2 in the supplementary material), the floxed CRE/SH2-electroporated neurons restored proper contact with the pial surface and apical tuft complexity in layer I (Fig. 8A,B; see Fig. S8A,B and Movie 3 in the supplementary material). However, either mutant or SH2-rescued mutant neurons abnormally aggregated in small clusters in deep positions, which were never observed for wt neurons (Fig. 8A; see Fig. S8B,C in the supplementary material). Finally, dendrite branching, strongly altered in SCR-expressing mutant neurons, was also rescued in SH2-electroporated mutant neurons (Fig. 8B; see Fig. S8A in the supplementary material), although the dendrite length was increased in COUP-TFI-deficient neurons (wt_CRE/SCR: 193.9±12.1 μm; fl/fl_CRE/SCR: 287.3±31.2 μm; n=3; P<0.05) but not restored in CRE/SH2-electroporated brains (wt_CRE/SCR: 193.9±12.1 μm; fl/fl_CRE/SH2: 285.9±5.4 μm; n=3; P<0.005) (see Fig. S8D in the supplementary material). Taken together, our analyses demonstrate a cell-autonomous role for COUP-TFI in migration, targeting, and axonal and dendritic arborization of CPNs and indicate that fine-tuning of Rnd2 levels by COUP-TFI during radial migration favours CPN maturation.

Our data contribute to the delineation of a genetic circuitry underlying Rnd2 expression during glia-guided radial migration and confirms that the fine control of its expression during different phases of this process is crucial for correct radial distribution of UL newborn neurons. We demonstrate that the sharp decrease of Rnd2 in the upper IZ favours glia-guided BSC migration to the CP. We demonstrate that COUP-TFI represses Rnd2 expression levels in the IZ independently of its main activator (Ngn2) in post-mitotic migrating neurons, indicating a direct role of COUP-TFI during early phases of glia-guided radial migration. Moreover, we show that COUP-TFI binds directly to various Rnd2 downstream sequences, probably recruiting chromatin remodelling factors or other co-repressors (Montemayor et al., 2010; Nishihara et al., 2004; Zhang et al., 2009). Strikingly, one of the COUP-TFI binding sites is positioned between two Ngn2-bound sequences in a transcriptional enhancer previously identified in the downstream region of the Rnd2 locus (Heng et al., 2008). COUP-TFI might thus prevent or destabilize the binding of Ngn2 and other basic helix-loop-helix (bHLH) factors to the Rnd2 enhancer in the IZ. This finding correlates with a previous study showing that Ngn2 overexpression does not alter radial distribution of migrating cells in this compartment (Hand et al., 2005). Overall, our data demonstrate that Ngn2 and COUP-TFI are independently involved in the fine-tuning of Rnd2 expression and indicates that COUP-TFI normally promotes glia-guided radial migration by progressively repressing Rnd2 expression along medial and caudal regions of the neocortex during mid- to late corticogenesis.

In this report, we show that the expression of the atypical Rho-GTPase Rnd2 follows a high-rostral to low-caudal gradient, which is complementary to the one described for COUP-TFI (Armentano et al., 2007; Liu et al., 2000; Tomassy et al., 2010; Zhou et al., 2001). This gradient is also complementary to the UL neurons migratory gradient detailed in this study. Thus, high expression levels of Rnd2 in rostral cortex might restrict migrating cells from reaching the CP more than in caudal regions. In the absence of COUP-TFI function, the Rnd2 gradient is abolished and higher Rnd2 expression levels are detected caudally. Accordingly, a lower percentage of cells reach the cortical plate. Thus, our data, together with previous reports, demonstrate that Rnd2 expression is downregulated by increasing levels of COUP-TFI along the rostrocaudal axis, leading to differential rates of migration from rostral to caudal regions of the cortex.

The biological meaning of this migratory gradient is still unclear and deserves further investigation. However, other factors involved in cortical arealization, besides COUP-TFI, seem to modulate neuron radial migration (O’Leary et al., 2007). Furthermore, neurons in ULs (and particularly in layer IV) represent a source of signals for thalamocortical innervation (Shimogori and Grove, 2005), an important process refining cortical arealization. It is thus tempting to hypothesize that COUP-TFI might, at least in part, control the topography of thalamocortical innervation by modulating the timing of neuronal migration to layer IV along the rostrocaudal axis of the neocortex.

Here, we also provide evidence that COUP-TFI is involved in the development of correct dendritic morphology of UL neurons. The precise morphology of these neurons could be controlled during early phases of pyramidal neuron radial migration. A previous report suggested that initial apical dendrite formation occurs during early stages of glia-guided radial migration and that RhoA activity inhibits this process by promoting MSC morphology (Hand et al., 2005). As Rnd2 modulates RhoA activity via pragmin interaction (Pacary et al., 2011; Tanaka et al., 2006), higher Rnd2 expression levels observed in COUP-TFI mutants might alter RhoA activity and thus disrupt the formation of the initial dendrite, ultimately leading to an increased number of MSCs. Our data suggest that altered initial dendritic formation might also affect late dendritic arborization of pyramidal neurons. However, although lowering Rnd2 levels in mutant neurons restores late dendritic complexity, it only partially rescues the morphological transition from MSC to BSC, indicating that most rescued cells were defective BSCs stalled in the IZ and lower CP regions. We cannot exclude that our shRNA strategy failed to re-establish precise levels of Rnd2; however, our results are consistent with previous data showing that artificially enhanced Rnd2 levels only lead to a slight increase in MSC (Heng et al., 2008).

Another possibility is that the development of an abnormal apical dendrite with no or few apical tufts, as observed in COUP-TFI-deficient neurons, might be due to the failure of their leading process to reach the MZ and activate reelin-mediated dendritic branching (Hunter-Schaedle, 1997; Niu et al., 2004; Ohshima et al., 2007; Sanada et al., 2004). It has been shown previously that reelin in the MZ stabilizes the leading process by cofilin phosphorylation (Chai et al., 2009). Lack of proper stabilization of the leading process might not only affect final phases of radial migration but also dendritic maturation. Thus, the remarkable rescue of dendritic complexity observed in SH2-rescued COUP-TFI-deficient CPNs might be due to restored contact of the MZ by their leading processes. Nevertheless, although rescued COUP-TFI-deficient dendrites successfully target the MZ, both mutant and SH2-rescued mutant cells remain beneath the upper CP with abnormally long apical dendrites and form abnormal cell clusters in the ULs. We suggest that these processes might be independent from Rnd2 as it was already reported that the expression of two microtubule-associated proteins, MAP1B (Mtap1b – Mouse Genome Informatics) and MAP2 (Map2 – Mouse Genome Informatics), strongly decreases in COUP-TFI-deficient dendrites (Armentano et al., 2006), whereas cell adhesion increases in the absence of COUP-TFI function (Adam et al., 2000). Furthermore, Rnd2 expression in the CP seems to be very low during late stages of corticogenesis.

Thus, correct positioning of UL neurons does not seem to be a prerequisite for the acquisition of proper dendritic morphology. However, correct timing in reaching the CP, and more particularly the MZ, might be crucial for late-born migrating neurons to acquire a fully mature morphology.

Beside dendritic abnormalities, COUP-TFI-deficient callosal pyramidal neurons show delayed axonal elongation and impaired contralateral innervation. Although previous studies suggested a link between neuron radial migration and CC formation, a direct relationship between proper cell migration and callosal neuron differentiation has not been fully established (Donahoo and Richards, 2009; Richards et al., 2004). Inactivation of various molecules involved in radial migration, such as Dcx and Dclk, leads to affected migration and defective commissural outgrowth (Deuel et al., 2006; Koizumi et al., 2006); however, their expression is maintained at high levels in the CP during late events of differentiation. In this study, we show that altered callosal neuron migratory properties in COUP-TFI null mice correlate with a strong delay in callosal axon midline crossing at perinatal stages as well as reduced arborization at P8. In accordance with previous observations suggesting that axon elongation starts during BSC migration (Hatanaka and Murakami, 2002; Noctor et al., 2004; Norris and Kalil, 1991), we observed altered distribution of mutant BSCs along the IZ and CP and delayed callosal axon elongation, which were both restored by lowering Rnd2 levels in mutant neurons. Accordingly, a previous study demonstrated that the interaction of Rnd2 with plexin D1 inhibits neurite outgrowth in cortical neurons (Uesugi et al., 2009). Finally, because Rnd2 modulates actin dynamics through interaction with the WASP proteins (Kakimoto et al., 2004), it is highly possible that increased Rnd2 levels might alter BSC neurite outgrowth, a process that is at the base of both glia-guided migration and CC formation.

Finally, in light of its expression gradient, we propose that Rnd2 might play an important role in CC formation by modulating the axonal elongation rate along the rostrocaudal axis of the neocortex. Previous studies demonstrated that all components of the CC (rostral, medial and caudal) cross the midline at similar developmental stages. As the midline allows callosal axon crossing mainly in the anterior half of the medial cortex and only for a restricted time, caudal callosal axons should cover greater distances to reach the crossing site and their homotypic regions (Richards et al., 2004). One can thus speculate that caudomedial and caudal CPNs that express lower levels of Rnd2 during migration, elongate their axons at a higher rate to reach the crossing regions at the correct time.

In conclusion, our study clearly demonstrates that COUP-TFI repression of Rnd2 in the IZ is crucial for proper migration of late-born BSC neurons. This suggests that progressive lowering of Rnd2 levels along the rostrocaudal cortical axis modulates the timing of CPN settlement as well as the rate of callosal axon elongation in different regions of the neocortex, ultimately favouring CC formation.

We would like to thank V. Tarabykin for the Svet1 plasmid and Satb2 polyclonal antibody; M. Nieto for the Cux1 and Cux2 plasmids; R. Hevner for the Tbr1 antibody; G. Andolfi for genotyping; M. Giordano for animal husbandry; D. Di Nucci and G. Esposito for surgical assistance; and S. Arbucci, and F. A. Martín for confocal imaging. We are very thankful to R. Rispoli and the TIGEM Bioinformatics Core; and particularly to F. A. Pereira and C. Montemayor for the ChromAnalyzer-driven bioinformatic analysis of the Rnd2 locus; to J. Hazan for critical reading; and to members of the Studer laboratory for fruitful discussions.