Semaphorin 6A-Plexin A2/A4 Interactions with Radial Glia Regulate Migration Termination of Superficial Layer Cortical Neurons.

Precise regulation of neuronal migration termination is crucial for the establishment of brain cytoarchitectures. However, little is known about how neurons terminate migration. Here we focused on interactions between migrating cortical neurons and their substrates, radial glial (RG) cells, and analyzed the role of Plexin A2 and A4 (PlxnA2/A4) receptors and their repulsive ligand, Semaphorin 6A (Sema6A), for this process. In both PlxnA2/A4 double-knockout and Sema6A mutant mice, the outermost cortical plate neurons ectopically invade layer 1 at a stage when they should reach their destinations. PlxnA2/A4 proteins are abundantly expressed on their leading processes, whereas Sema6A mRNA is enriched in RG cell somata. Cell-targeted gene expression and conditional knockouts indicate critical roles for these molecules. We hypothesize that the timely appearance of repulsive signaling mediated by Sema6A-PlxnA2/A4 weakens migrating neuron-RG cell interactions, leading to migration termination.

Introduction

Neuronal migration is crucial for establishing a neuronal architecture with functional integrity. During development, newborn neurons migrate from progenitor zones toward their final destinations, where they form specific neuronal architectures and functional circuits. Disturbance of these migrations alters neuronal positioning, causing cytoarchitectonic abnormalities and consequent functional deficiencies (Barkovich et al., 2012, Barkovich et al., 2015, Moffat et al., 2015).

Neuronal migration has been extensively studied using the cerebral cortex. Cortical excitatory neurons, which are born in the cortical ventricular zone (VZ)/sub-VZ, migrate radially through the cortical plate (CP) along radial glial (RG) fibers. Although the RG fibers reach the pial surface spanning the marginal zone (MZ), the future layer 1 (L1), the migrating neurons terminate migration just beneath the MZ and establish a sharply demarcated laminar structure.

The process of neuronal migration can be divided into three steps: initiation, continuation, and termination. Although the cellular and molecular mechanisms for the first two steps have been relatively well studied (for reviews, see Evsyukova et al., 2013, Ohtaka-Maruyama and Okado, 2015, Sekine et al., 2013), our knowledge about the termination of migration is still limited.

Termination may be caused by the detachment of migrating cortical neurons from the RG fibers due to a weakening of adhesion between them at the top of the CP (Goffinet, 1984). Furthermore, increased neuron–neuron adhesion may override neuron–RG fiber adhesion at the top of the CP, thereby contributing to the detachment of neurons from the RG fibers (Goffinet, 1984). The observation that cortical neurons whose migration is terminating are more closely packed supports this notion (Goffinet, 1984, Sekine et al., 2014). However, whether active changes in the interaction between neurons and RG fibers are indeed required for terminal migration is unknown.

To date, several genes are known to affect migration termination. These include genes that disrupt the integrity of the pial basement membrane (BM) (Costa et al., 2001, Herms et al., 2004, Jeong et al., 2013, Moers et al., 2008, Nakagawa et al., 2015, Niewmierzycka et al., 2005), reelin (D'Arcangelo, 2005), reelin-related genes (Ha et al., 2017, Hack et al., 2007, Herrick and Cooper, 2002, Hirota et al., 2018, Kohno et al., 2015), and SPARC-like 1 (Gongidi et al., 2004).

Here we focused on Semaphorins (Semas) and their receptors. Semas are a large and diverse family of signaling proteins composed of secreted as well as transmembrane proteins and were originally identified as repulsive axonal guidance molecules in the nervous system (Goodman et al., 1999). Their receptors are mainly the Plexin (Plxn) family (Battistini and Tamagnone, 2016). Among Semas, the transmembrane-type Sema ligands Sema6A and 6B bind to the PlxnA subfamily members PlxnA2 and PlxnA4 (Suto et al., 2005, Toyofuku et al., 2008, Tawarayama et al., 2010). Their interactions regulate cell adhesion positively or negatively, depending on the situation, and contribute to the establishment of neural circuits such as corticospinal (Faulkner et al., 2008, Rünker et al., 2008) and thalamocortical projections (Leighton et al., 2001, Little et al., 2009, Mitsogiannis et al., 2017), laminar-restricted projections of hippocampal mossy fibers (Suto et al., 2005, Tawarayama et al., 2010), and laminar targeting of retinal neurons (Matsuoka et al., 2011). They also regulate cerebellar granule cell migration (Kerjan et al., 2005, Renaud et al., 2008) and interkinetic nuclear migration of retinal progenitor cells (Belle et al., 2016).

In this study, we show that a temporally regulated change in Sema6A–PlxnA2/A4 interaction between migrating neurons and RG cells is required for correct termination of neuronal migration. Superficial layer neurons (SLNs) located at the outermost part of layers 2 and 3 (L2/3) ectopically invade L1 in PlxnA2/A4 double-knockout (KO) (dKO) mice as well as Sema6A KO mice. Anatomical and genetic analyses suggest that PlxnA2/A4 proteins on the migrating SLNs and Sema6A protein on the RG processes interact to suppress cell invasion into L1, probably by regulating the adhesiveness between SLNs and RG fibers.

Results

Irregular Arrangement of SLNs in PlxnA2/A4 dKO Cortex

PlxnA2 and A4 are widely expressed in mouse cerebral cortex during embryonic and perinatal stages (Murakami et al., 2001, Suto et al., 2003); although their roles have been studied in other systems, little is known about their contribution to cortical development. To explore their role, we first analyzed the cortical phenotype of PlxnA2 and A4 single- and double-mutant mice at P15.

In wild-type (PlxnA2A4), double-heterozygous (PlxnA2A4), or compound mutant mice carrying only a single wild-type allele for PlxnA2 or A4 (PlxnA2A4 and PlxnA2A4), the boundary between L1 and L2/3 was clearly demarcated by a difference in cell density (Figures 1A and S1A–S1C, wild-type, n = 2; PlxnA2A4, n = 3; PlxnA2A4, n = 3; PlxnA2A4, n = 2). In contrast, when both PlxnA2 and A4 were knocked out (PlxnA2A4; PlxnA2/A4 dKO), the boundary became blurred and rippled and SLNs appeared to invade L1 (Figure 1B, n = 3). The mislocation of SLNs across the rostro-caudal axis tended to occur with a medial-low/lateral-high gradient. For example, the mislocation appeared to be more prominent in the lateral regions including the primary somatosensory (S1), auditory, and visual areas (not shown) than in the medial cortical regions that include the primary motor area (M1). In the lateral regions, neurons that invaded L1 sometimes formed clusters, giving rise to a ruffled L1–L2/3 border (Figure 1B, b2, n = 3). These observations indicate that PlxnA2 and A4 redundantly regulate correct neuronal positioning, leading to the formation of a sharp boundary between L1 and L2/3.

Figure 1

Cortical Cytoarchitecture of PlxnA2/A4 dKO Mice

(A and B) Nissl staining of coronal sections from a PlxnA2/A4 double heterozygote, as a control (A), and a dKO mouse (B) at P15. In the dKO mouse, SLNs, which are normally located in the outermost regions of L2/3, were dispersed into L1. M1, primary motor area; S1, primary somatosensory area; S2, secondary somatosensory area. Lower panels show higher-magnification views of dorsal (a1 and b1), dorsolateral (a2 and b2), and lateral (a3 and b3) regions of the sections in (A) and (B). In lateral regions of dKO mice, cells sometimes formed clusters as shown in b2; arrows and an arrowhead indicate vertical and horizontal gaps around clusters, respectively. The frequency of appearance of clusters varied among animals, but they tended to appear in a rostro-medial-low/caudo-lateral-high gradient manner. See also Figure S1. Scale bars: 1 mm in (A and B) and 200 μm in (a1–b3).

Excitatory SLNs Are Mislocated in L1 of PlxnA2/A4 dKO Mice

We next attempted to determine the identity of malpositioned cells in the PlxnA2/A4 dKO mice. At P15, cellular arrangements in the cortex appeared normal (Figures 2A and 2F: hereafter, double-heterozygous mice were used as a control unless otherwise noted). For example, small- and medium-sized neurons were located in L2–4, large-sized neurons in L5 at a low density, and medium-sized neurons in L6, in a manner similar to the control (n = 3). Expression of layer-specific neuronal markers also showed normal neuronal positioning in each layer (Figures 2B–2D, 2G–2I, and S2A–S2C). However, Wolframin protein (Wfs-1)-positive cells, which are normally located in the outer part of L2/3 (Luuk et al., 2008) (Figure 2E), were mislocated in L1 in the dKO mice (Figure 2J). Thus the vast majority of mislocated neurons were SLNs, but not other layer neurons.

Figure 2

Mislocation of SLNs in PlxnA2/A4 dKO Mice and Its Developmental Onset

(A–J) Immunostaining using cortical layer markers. Coronal sections from P15 control (A–E) and PlxnA2/A4 dKO mouse (F–J). NeuN, pan-neuronal marker; Cux1, marker for L2–4 neurons; ER81, marker for L5 neurons; FoxP2, marker for L6 neurons; and Wfs1, marker for upper L2/3 and L5 neurons. Insets at the bottom of (E) and (J) are higher-magnification views of the L1–L2/3 regions indicated by rectangles. Arrowheads indicate Wfs1-positive cells.

(K) Bar histograms showing the number of marker-positive cells within a 100-μm-wide region of S1. Four 100-μm-wide regions within S1 were sampled for each genotype. Data are represented as mean ± SEM. No significant statistical difference was found for any combination of the data in the histogram (one-way ANOVA followed by Tukey's post-hoc test).

(L–M″) Immunostaining using anti-NeuN and Cux1. Mislocated neurons were mostly Cux1-positive.

(N–O′) Immunostaining using anti-NeuN and GABA. Distribution of GABA+ cells appeared similar between control (A2A4) and dKO mice in the locations where ectopic neuronal distribution was evident.

(P–U′) Nissl staining of presumptive primary somatosensory area of control and dKO mice at P1, P3, and P7. (P′–U′) Immunostaining using anti-Satb2 or Cux1. Satb2 marks L2–6 callosal neurons that are excitatory (Britanova et al., 2008, Huang et al., 2013) and start to be expressed earlier than Cux1. By P7, ectopic cells sometimes formed clusters (arrows).

See also Figures S2 and S5. Scale bars: 200 μm in (A–J), 50 μm in insets of (E and J) and in (P′–U′), and 100 μm in (L–O′) and (P–U).

In the dKO cortex, the numbers of lamina-specific neuronal marker-positive cells (Figure 2K) and cleaved caspase 3-positive cells (Figure S2F) were comparable with those in control, suggesting that neither cell type conversion nor cell death is responsible for the mislocalization phenotype. The mislocated SLNs may be mostly excitatory, because (1) they were positive for Cux1, which is expressed in pyramidal neurons, but not parvalbumin-positive inhibitory neurons (Nieto et al., 2004) (Figures 2L–2M″), and (2) the density of GABAergic inhibitory neurons in L1 and nearby appeared unchanged (Figures 2N–2O′). Together, these results indicated that excitatory SLNs showed aberrant distribution in PlxnA2/A4 dKO mice.

Mislocation of SLNs Manifests during the Early Postnatal Stage

To determine the initial event that is primarily responsible for SLN mislocation, we examined the time at which the mislocation first appeared in the PlxnA2/A4 dKO mice. We focused on the prospective S1 area, which at a later stage displayed a typical mislocation phenotype. Nissl staining patterns of dKO mice were indistinguishable from those of control mice at postnatal day (P)1, when cell-sparse L1 became identifiable (Figures 2P and 2Q); however, a few ectopic cells, which were positive for Satb2 (Satb2+), a marker of upper layer neurons, had already appeared in L1 of the dKO mice at this stage (Figures 2P′ and 2Q′). At P3, the L1–L2/3 boundary became less clear (Figures 2R and 2S), largely due to spreading of Satb2+ cells into L1 (Figures 2R′ and 2S′). The malpositioning of Satb2+ cells tended to be more prominent in the lateral than in the medial cortex but these cells did not form clear clusters at this stage. By P7, the disorganized superficial layer structures in the dKO mice were readily recognizable by Nissl staining (Figures 2T–2U′). Thus the first indication of neuronal mislocation occurred at P1, and the mislocation became apparent at P3.

SLNs Appear to Overmigrate beyond the Normal Settling Position

The period of P1–P3 when the mislocated phenotype of neurons becomes apparent corresponds to the period when wild-type SLNs reach the top of the CP to terminate migration (Hatanaka et al., 2004). This suggests that the observed phenotype is a result of aberrant termination of neuronal migration. To explore this possibility, we carried out a birthdating study by injection of bromodeoxyuridine (BrdU) at embryonic day (E)15.5 or E16.5, stages when L2/3 neurons are born. Overall, the distribution of BrdU+ cells was similar between control and dKO mice, indicating their normal migration from the VZ toward L2/3 (Figures 3A–3D). However, the mean cell position (Dm, see legend to Figure 3) was slightly higher in the dKO mice, suggesting the occurrence of overmigration (Figure 3A′–3D′). To further explore this possibility, we labeled SLNs by in utero electroporation (IUE) at E15.5 and examined their morphology at P3. We found that a majority of GFP+ cells had reached the top of the CP in both control and dKO mice (Figures 3E, 3E′, 3H, and 3H′). A close examination, however, revealed that whereas most GFP+ cells were confined to the CP in control mice (Figures 3E and 3E′), a substantial proportion was dispersed into L1 of the dKO mice (Figures 3H and 3H′). In both control and dKO mice, GFP+ cells typically extended their thick apical processes toward the pial surface (Figures 3F and 3G). However, apical processes in dKO mice tended to be shorter than those in the control mice (Figures 3I and 3J). Notably, some GFP+ cells in close proximity to the pial surface extended a thick process pointing to the ventricle (Figure 3J). Together, these observations are consistent with the view that SLNs migrate beyond their normal settling position in PlxnA2/A4 dKO mice.

Figure 3

SLNs Are Ectopically Positioned in L1 of PlxnA2/A4 dKO Mice

(A–D) Distribution of BrdU+ cells. BrdU was injected at either E15.5 (A and B) or E16.5 (C and D), stages of neurogenesis for SLNs in mice (Smart and Smart, 1982). BrdU+ cells were examined at P7. (A and C) Control and (B and D) PlxnA2/A4 dKO mice. (A′–D′) Bar histograms showing laminar distribution of BrdU+ cells. The cerebral wall was divided into 10 equal bins (bin 1–10) from the bottom of L6 to the pial surface, which is delineated by dotted lines in (A–D). Bins 10 and 9–7 approximately correspond to L1 and L2–4, respectively. The mean cell position (Dm) in each sample was calculated as Σ(NiDi)/Σ(Ni), where Ni is the number of BrdU+ cells in bin Di. Data are represented as mean ± SEM. The Dm value of the dKO mouse was slightly higher, showing a significant difference for E15.5 (p = 0.01), but not E16.5, injection (p = 0.22); in both cases, the proportion of cells in bin 1 was significantly higher in the dKO mouse. *p < 0.05 (Student's t test).

(E–J) Location and morphology of SLNs at the stage of migration termination. GFP+ SLNs of control mice were localized to the uppermost regions of L2/3 (E and E′), whereas those of PlxnA2/A4 dkO mice were additionally found in L1 (H and H′). The dotted line in (E′) and (H′) indicates the position of the L1–L2/3 boundary. (F and I) Higher-magnification views, and (G and J) 3D reconstruction images of GFP+ SLNs. The dotted line indicates the position of the pial surface.

See also Figure S5. Scale bars: 100 µm in (A–D), (E and H) and (E’ and H’), and 50 µm in (F and I) and (G and J).

Although breaches in the pial BM result in neuronal ectopias (Costa et al., 2001, Herms et al., 2004, Jeong et al., 2013, Moers et al., 2008, Nakagawa et al., 2015, Niewmierzycka et al., 2005), we found unaltered laminin expression in the dKO mice (Figure S2G), indicating that BM defects were not the cause of the phenotype observed here.

Overlapping Expression of PlxnA2 and A4 in Neurons at the Top of the CP

Both PlxnA2 and A4 mRNA showed developmental stage-dependent expression profiles from E15.5 through to P7 (Figures 4A–4H, Murakami et al., 2001, Suto et al., 2003). Importantly, at P0 and P3, both transcripts were abundant at the top of the CP (arrowheads), where SLNs should settle. We also found that both PlxnA2 and A4 proteins were enriched in L1 (Figures 4M–4N″). Within this layer, the site of high PlxnA2 and A4 immunoreactivities was found in close proximity to SLN leading processes (Figures 4O and 4P). These findings are consistent with the idea that the leading processes of SLNs express both PlxnA2 and A4 proteins.

Figure 4

PlxnA2, PlxnA4, and Sema6A Expression at Perinatal Stages

(A–L) In situ hybridization analysis of PlxnA2, PlxnA4, and Sema6A mRNA expression at E15.5, P0, P3, and P7 in presumptive S1 area. Both PlxnA2 and A4 mRNAs were enriched in the upper part of the CP at P0 and P3 (arrowheads). PlxnA2 mRNA was also enriched in the deeper part. Sema6A mRNA expression was generally scattered throughout the cerebral wall, but was high in the cortical VZ at P0 and P3 (arrows in J and K). Open arrows in (L) point to the location of presumptive barrels in L4. IZ, intermediate zone.

(M–N″) PlxnA2 and PlxnA4 immunoreactivities (green) with nuclear counterstain (blue) at P3 (M and N), nuclear stain alone (M′ and N′), and an enlarged view (M″ and N″). Weak whitish cell nuclear staining in the L5 region in (N) is non-specific, as it was observed after immunostaining without anti-PlxnA2 or anti-A4 antibodies.

(O and P) PlxnA2 and PlxnA4 immunoreactivities (red) with GFP+ SNLs (green).

Scale bars: 200 µm in (A, E and I) and (B, C, F, G, J, and K) and (D, H and L), and 50 µm in (M–N’), (M” and N”) and (O and P).

Cell-Autonomous Function of PlxnA2

PlxnA2/A4 expression on SLNs raises the possibility that these molecules function cell autonomously in SLNs. We tested this by introducing PlxnA2 (Figure S3A) into the PlxnA2/A4 dKO mice at E15.5 by IUE. In wild-type (Figures S3B and S3C) and control mice (Figure 5A, see also Figure 3E), expression of GFP alone (GFP) or a mixture of GFP and PlxnA2 (GFP/Plxn2), respectively, yielded virtually no GFP+ or GFP/PlxnA2+ cells in L1 (0% of total GFP/PlxnA2+ cells above the white matter [WM], n = 3 brains, total 1,336 cells, control, Figure 5H; wild-type, data not shown). In contrast, in dKO mice transfected with GFP, a substantial fraction of GFP+ cells were observed in L1 (32.6% ± 8.4%, n = 3 brains, total 1,923 cells, Figures 5B and 5H). Expression of Sab2 indicates that these were SLNs (Figure 5E). However, forced expression of GFP/PlxnA2 rescued this phenotype, greatly reducing the number of GFP/PlxnA2+ cells in L1 (8.1% ± 0.7%, n = 3 brains, total 2,012 cells, Figures 5C and 5H). Importantly, untransfected SLNs were still abundantly mislocated in L1 and there was no difference in the number of GFP−/Satb2+ cells between GFP- and GFP/PlxnA2-transfected animals (Figure 5G). Thus, these results indicate that PlxnA2 functions cell autonomously for proper SLN positioning.

Figure 5

Transfection of SLN Progenitors with PlxnA2 Rescues SLN Mislocation in PlxnA2/A4 dKO Mice

(A–C‴) Transfection of SLN progenitors in the VZ with GFP alone or a mixture of GFP and PlxnA2 plasmids. IUE was performed on control (A) or dKO mice (B and C) at E15.5 and brains were examined at P3. (A–C) Left and right panels indicate the distribution of GFP+ cells and nuclear counterstain, respectively. (A′–C′) Higher-magnification merged views of GFP fluorescence and Satb2 staining. (A″–C″) GFP and (A‴–C‴) Satb2 views. Dotted lines indicate the L1–L2/3 boundary. The region located within 70 μm from the pial surface was regarded as L1. Note that some PlxnA2/GFP-high+ cells in dKO mice aggregated in the CP (arrowheads in C) and the WM (see Figure S3).

(D) Transfection of GFP plasmid into SLN progenitors of Sema6A/ mice, by IUE at E15.5; brain was examined at P3. Left and right panels indicate the distribution of GFP+ cells and nuclear counterstain, respectively. (D′) Higher-magnification merged views of GFP fluorescence and Satb2 staining. (D″) GFP and (D‴) Satb2 views.

(E) Higher-magnification views of the region indicated by a rectangle in (B′), indicating that ectopically positioned GFP+ cells were Satb2-positive (arrowheads).

(F) Higher-magnification views of the regions indicated by dotted rectangles in A″ (left panel) and C″ (right panel).

(G) Bar histogram showing the number of GFP−/Satb2+ cells in L1. Cell numbers counted in an individual section are indicated by crosses.

(H) Bar histogram showing the percentage of GFP+ cells in L1 among total GFP+ cells above the WM. In (G and H) three or four sections each from three brains were examined, although two brains were used for Sema6A/. The numbers in parentheses indicate the numbers of sections in each group. Data are represented as mean ± SEM. *p < 0.05 (one-way ANOVA followed by Tukey's post-hoc test). n.s., not significant. The results of statistical analyses are indicated only on the focused groups.

Scale bars: 100 µm in (A–D) and (A’–D’”), and 20 µm in (E) and (F).

Some GFP/PlxnA2+ cells of the dKO mice appeared to prematurely halt migration beneath L1, exhibiting elaborated apical processes (Figure 5F). This may be because PlxnA2 started to be expressed in SLNs earlier than its endogenous expression under a strong constitutive promoter (Niwa et al., 1991). On closer examination, the elaborated apical processes were somewhat tangled, implying an additional effect of PlxnA2 expression. In extreme cases, GFP/PlxnA2-high+ cells in the WM of the dKO mice showed prominent aggregation (Figure S3F).

Mutation of Sema6A Phenocopies the PlxnA2/A4 dKO Phenotype

PlxnA2 and A4 are the receptors for the transmembrane Semas, Sema6A and 6B. Although both Sema6s bind to PlxnA2 and A4 (Suto et al., 2005, Toyofuku et al., 2008, Tawarayama et al., 2010), the binding affinity of Sema6B to PlxnA2 is relatively low (Tawarayama et al., 2010). To identify the Semas involved in proper SLN positioning, we first examined two different mutant alleles for Sema6A, Sema6A gene trap (Sema6A, Rünker et al., 2011) and Sema6A exon 3 deletion (Sema6A, Figure S4), and found that both homozygotes showed a disrupted L1–L2/3 boundary (Sema6A at P15, n = 2 brains, Figure 6A and Rünker et al., 2011; Sema6A at P3, n = 6 brains, Figure 6C), similar to the PlxnA2/A4 dKO cortex. Quantitatively, we found no significant changes in the number of cortical neurons between control and Sema6A mice (Figures S2D and S2E), as was the case for control and PlxnA2/A4 dKO mice (Figures S2C and S2E). Moreover, SLN mislocation became manifest around P3 (Sema6A, Figure S5A) and was most severe in the lateral region, showing a medial-low/lateral-high gradient. Immunohistochemical examination indicated that their BM structure, similar to PlxnA2/A4 dKO mice, was normal (data not shown). To corroborate these findings, we used the same IUE protocol that was applied for PlxnA2/A4 dKO mice. We labeled SLNs by IUE with GFP at E15.5 and examined GFP+ neurons at P3. Of the labeled cells in the CP and L1, the proportion of those in L1 in Sema6A mice was comparable to that in PlxnA2/A4 dKO mice (Figures 5D and 5H, 33.2 ± 3.11%, n = 2 brains). Furthermore, the morphologies of these cells resembled those in PlxnA2/A4 dKO mice (Figure S5B, see also Figure 3). Unlike in Sema6A mice, we found no ectopic SLNs in L1 of the Sema6A+/Sema6B compound mutant mice (n = 2 brains; Figure 6B), indicating the importance of Sema6A. In addition, when PlxnA2 was electroporated into Sema6A mice, SLN mislocation was hardly rescued (Figure S5C). Taking these observations together, it is highly likely that Sema6A is the PlxnA2/A4 ligand involved in proper positioning of SLNs.

Figure 6

Cortical Cytoarchitecture of Sema6A Mutant Mice, Identity of Sema6A-Expressing Cells, and Sema6A Protein Localization

(A and B) NeuN staining of a coronal section from a P15 mouse homozygous for Sema6A gene trap allele (A) and a compound mutant mouse heterozygous for Sema6A gene trap allele and homozygous for Sema6B knockout allele (B). The lower panels of (A and B) show higher-magnification views of the dorsal (a1 and b1), dorsolateral (a2 and b2), and lateral (a3 and b3) regions indicated by rectangles in (A and B).

(C) Immunostaining of coronal sections from P3 wild-type and Sema6A mice using cortical layer markers. Tbr1 is a marker for L6 neurons. For Ctip2, Satb2, and Cux1, see legend of Figure 2. Dotted lines indicate the L1–L2/3 boundary. Compare with Figure S2B, and see also Figures S4 and S5.

(D–H) Sema6A-expressing cells in the cerebral cortex at P1. These were identified by X-gal staining (blue) in sections taken from heterozygotes of Sema6A gene trap mice. As the trapped gene is fused to β-geo (a fusion between β-galactosidase and neomycin phosphotransferase, Leighton et al., 2001), cells expressing endogenous Sema6A can be identified by β-geo activity, which is retained within the endoplasmic reticulum. (D–F″) X-gal staining in L1 (D–F), anti-Ctip2 (D′) or anti-reelin antibody staining (E′ and F′), and merged views (D″–F″). Note that because Ctip2 expression marks a subset of GABAergic interneurons and L5 excitatory neurons (Arlotta et al., 2005), it is likely that Ctip2+ cells in L1 (yellow arrowheads) are GABAergic cells. (G) X-gal staining with an anti-Satb2 antibody staining (brown). (H) X-gal staining in the VZ.

(I) Binding profiles of AP-PlxnA2 on wild-type and Sema6A cortical slices, and of AP-Sema6A on a wild-type cortical slice, at P1. No specific binding of AP-PlxnA2 was observed on Sema6A slices, verifying a Sema6A–PlxnA2 molecular interaction.

Scale bars: 1 mm in (A and B), 200 μm in (a1–b3), 100 μm in (C) and (I), 25 μm in (D–F″), and 50 μm in (G and H).

The Source of Sema6A for Interaction with PlxnA2/A4 in L1

We next examined the distribution of Sema6A transcript during the perinatal period to identify the source of Sema6A that interacts with PlxnA2/A4. Sema6A mRNA was sparsely expressed throughout the cerebral cortex, but was prominent in the VZ from P0 through to P3 (Figures 4I–4L). To further examine Sema6A expression, we used mice heterozygous for the Sema6A gene trap allele, which can report expression of the trapped gene as ß-geo gene expression (Leighton et al., 2001), and found that at least two types of cells expressed Sema6A at P1: GABAergic interneurons in L1, as judged by Ctip2 expression (Figure 6D), and in the CP, as judged by the lack of Satb2 expression (Figure 6G), and VZ cells, especially those facing the lateral ventricle (Figure 6H). Neither Cajal-Retzius cells in L1 nor Satb2-positive excitatory cortical neurons expressed Sema6A (Figures 6E–6G).

Owing to the lack of specific antibodies, we could not directly examine the distribution of Sema6A protein responsible for SLN mislocation. However, as an alternative method, we employed a ligand–receptor-binding assay using alkaline phosphatase (AP)-conjugated recombinant protein for the sema domain of PlxnA2 (AP-PlxnA2, Renaud et al., 2008). When AP-PlxnA2 protein was applied to P1 live cortical slices, it bound most strongly to L1 (Figure 6I). Likewise, AP-Sema6A, consisting of the ectodomain of Sema6A and AP (Suto et al., 2005), bound most strongly to L1 (Figure 6I). These results led us to the idea that Sema6A protein is mainly distributed in L1 and functions via binding to PlxnA2/A4 on SLN leading processes extending into L1.

Both GABAergic interneurons and cortical VZ cells can synthesize Sema6A protein in L1. Although the somata of VZ cells are located far from L1, these cells could be a source of Sema6A proteins observed in L1. This idea gains support from the fact that VZ cells transfected with myc-tagged Sema6A displayed expression of exogenous Sema6A protein throughout the RG cell processes, including their endfeet, in L1 (Figures S3G–S3K′).

Sema6A Expression in VZ Cells Is Responsible for the Proper Termination of SLN Migration

The above observations indicate that both GABAergic and VZ cells could be sources of Sema6A protein. To identify the Sema6A source responsible for termination of SLN migration, we took a conditional (c) KO approach, using a Sema6A-floxed mouse (Sema6A), whose exon 3 is flanked by the loxP sequence (Figure S4). As described above, mice lacking exon 3 (Sema6A) phenocopied the PlxnA2/A4 dKO mice in terms of SLN mislocation (Figures 7A, 7E, and 7G, n = 3 brains for each genotype, see also Figures 5D, 5H, S2D, and S2E, and S5), indicating that Cre-mediated recombination of this floxed allele indeed disrupted Sema6A functions. We first crossed Sema6A mice with the Emx1-Cre line (Iwasato et al., 2008), in which Cre is expressed in VZ cells and excitatory neurons of the pallium, but not in GABAergic interneurons, which are derived from the subpallium (Gorski et al., 2002). The resulting number of Satb2+ cells in L1 of Emx1-Cre;Sema6A mice was almost identical to that of Sema6A mice, indicating that the major cell source of Sema6A regulating SLN positioning is the pallium (Figure 7B, n = 3 brains).

Figure 7

Sema6A Mice Reveal a Requirement for Sema6A in VZ Cells for Proper Termination of SLN Migration

(A–F) Satb2 staining of the presumptive S1 area of (A) Sema6A, (B) Emx1-Cre; Sema6A, (C) Sema6A, (D) Nestin-CreER; Sema6A (administered tamoxifen at E17.5), (E) PlxnA2A4, and (F) PlxnA2A4 mice at P3. Dotted lines indicate the L1–L2/3 boundary. Note that SLNs of Sema6A, Emx1-Cre; Sema6A, Nestin-CreER; Sema6A, and PlxnA2A4 dKO mice were less tightly packed than those of the control groups.

(G) Bar histogram showing the number of Satb2+ cells in L1 in a 450-μm-wide strip. Three sections each from three brains were examined. Data are represented as mean ± SEM. *p < 0.05 (one-way ANOVA followed by Tukey's post-hoc test). n.s., not significant. In Nestin-CreER; Sema6A cKO mice, the phenotype was milder than that in Sema6A KO mice. This may be due to insufficient recombination, because tamoxifen-induced Cre recombination is not very effective in general (Vooijs et al., 2001). See also Figure S6. Scale bar, 100 μm in (A–F).

(H) Model for mislocation of SLNs in Sema6A KO/cKO and PlxnA2/A4 dKO mice. The illustration indicates the proposed mechanism for termination of neuronal migration and proper positioning of SLNs through Sema6A (blue)–PlxnA2/A4 (red) interaction. (Left) SLNs migrate toward the pial surface along fibers of RG cells, whose cell bodies are located in the VZ. Sema6A protein synthesized in VZ cells travel distally along the fibers at the stage when SLNs reach the top of the CP. Meanwhile, PlxnA2/A4 expression in SLNs is up-regulated and concentrated on apical processes extending into L1. (Middle and right) In both Sema6A KO/cKO and PlxnA2/A4 dKO mice, SLNs appear to migrate beyond their final destination, thus invading L1. We propose that L1 Sema6A on RG cell fibers acts with PlxnA2/A4 on leading processes of SLNs to detach them from the fibers, which is a key step for proper termination of migration.

As described above, the Emx1-Cre line inactivates Sema6A in pallial VZ cells and their descendants in both embryonic and perinatal stages. To remove Sema6A specifically in perinatal VZ cells (Figures 4J, 4K, and 6H), we crossed Sema6A mice with the Nestin-CreER line (Imayoshi et al., 2006), in which CreER, a tamoxifen-inducible Cre recombinase, is continuously expressed in VZ cells from the embryonic through to the perinatal period. We first confirmed that tamoxifen administration at E17.5 induced recombination almost exclusively in VZ cells (Figures S6A–S6D). We then administered tamoxifen to Nestin-CreER; Sema6A mice at E17.5 and found that Sema6A was indeed inactivated in VZ cells (Figures S6E and S6F). In these mice, the SLN mislocation phenotype observed in Sema6A was recapitulated (Figures 7D and 7G, n = 3 brains). As a control, SLN-targeted Sema6A inactivation was carried out by IUE with pDCX-Cre plasmid, which drives Cre expression in postmitotic premigratory neurons (Franco et al., 2011). As expected, this did not affect SLN positioning (Figure S6G, n = 3 brains). Taken together, our findings indicate that the source of Sema6A responsible for termination of SLN migration is cortical VZ cells.

Reelin+ Cajal-Retzius Cells and Reelin Signaling Pathway Are Unchanged

Reelin is essential for proper laminar positioning of cortical neurons (D'Arcangelo et al., 1995). Several reelin-related mutant mice exhibit a mis1location phenotype of SLNs resembling Sema6A KO (Sema6A and Sema6AΔ/Δ) and PlxnA2/A4 dKO mice (Ha et al., 2017, Hack et al., 2007, Herrick and Cooper, 2002, Hirota et al., 2018, Kohno et al., 2015). This raises the possibility that reelin signaling is altered in Sema6A KO and PlxnA2/A4 dKO mice. We found that Cajal-Retzius cells, a major source of reelin in L1 of developing mouse (Ogawa et al., 1995), were distributed normally in PlxnA2/A4 dKO mice (Figures S7A–S7D″). There was no difference in their number between control (11.3 ± 0.9, in a 400-μm-wide MZ at P1, n = 6 strips, 2 brains) and PlxnA2/A4 dKO mice (10.8 ± 1.0, n = 6 strips, 2 brains). Furthermore, there were no notable changes in reelin-dependent cells or reelin signaling in these mice: (1) the morphology of RG cell processes, which is affected by reelin signaling (Chai et al., 2015, Hartfuss et al., 2003), appeared normal (Figures S7E and S7F) in PlxnA2/A4 dKO mice and (2) the expression profiles of Dab1, which is an essential component of the canonical reelin signaling pathway and whose expression level is regulated by reelin signaling (Arnaud et al., 2003, Howell et al., 1999, Howell et al., 2000), were unaffected both in Sema6A and in PlxnA2/A4 dKO mice (Figures S7G–S7K). These findings argue against the possibility that mislocation of SLNs is mediated by attenuation of reelin signaling.

Discussion

In this study, we have shown that many SLNs were mislocated in L1 of PlxnA2/A4 dKO and Sema6A KO/cKO mice. This is most likely to be a result of overmigration past their final destinations. Given that PlxnA2/A4 and Sema6A were expressed in migrating cortical neurons and RG cells, respectively, the most likely explanation for the phenotype is that the normal detachment of migrating SLNs from RG endfeet was abrogated in these mice due to a lack of repulsive interactions caused by Sema6A–PlxnA2/A4 signaling (Figure 7H). We hypothesize that this detachment is critical for SLNs for their migration termination at their final destinations.

Functional Requirement for PlxnA2/A4 in Migrating SLNs and for Sema6A in the RG Scaffold

Sema6A can interact with PlxnA2/A4 in either cis (on the same cell) or trans (on different cells) to transduce their signals (Battistini and Tamagnone, 2016). These two types of interactions seem to differently regulate the strength or functionality of the signals (Battistini and Tamagnone, 2016, Haklai-Topper et al., 2010, Suto et al., 2007). Typically, Sema6A–PlxnA2/A4 trans-interaction transduces predominantly repulsive signaling (Battistini and Tamagnone, 2016). In this context, it is important to know whether or not Sema6A and PlxnA2/A4 are expressed on the same cells. We found that PlxnA2, and probably A4 as well, act cell autonomously in SLNs. On the other hand, these cells do not express Sema6A. Moreover, SLN-specific inactivation of Sema6A caused no notable phenotype, supporting a non-cell-autonomous function of Sema6A. Thus, Sema6A most likely functions in trans with PlxnA2/A4 of the SLNs, which probably elicits repulsive signaling.

Several lines of evidence indicate that the major source of Sema6A is cortical VZ cells. This is because conditional removal of Sema6A from cells of pallial origin using Emx1-Cre driver mice caused a mislocation phenotype resembling that in Sema6A KO mice. Moreover, selective removal of Sema6A from VZ cells at a late embryonic stage induced neuronal invasion into L1 (Figure 7). Although a minor contribution of Sema6A from other cells, such as GABAergic cells and immature glial cells, cannot be ruled out, our results collectively support the conclusion that cortical VZ cells are the major source of Sema6A required for termination of SLN migration.

The next question that arises is where the Sema6A–PlexinA2/A4 interaction occurs. Given that cortical VZ cells extend RG fibers reaching the pial surface during the embryonic and early postnatal periods, and that Sema6A mRNA expression in VZ cells was high at the time when SLNs reach their target, it is likely that this interaction occurs at L1 and nearby. Indeed, AP-PlxnA2 preferentially bound to the region of L1 and nearby where binding of AP-Sema6A also occurred.

Independence of SLN Mislocation from Thalamocortical Projection Error

Abnormal thalamocortical projection to L1, which can be observed at P2, has been reported in Sema6A and PlxnA2/A4 dKO mice (Little et al., 2009, Mitsogiannis et al., 2017, Rünker et al., 2011), raising the possibility that the phenotypes observed here occurred as a consequence this projection error. However, we observed almost no such aberrant thalamocortical projection in Sema6A cKO mice crossed with Emx1-Cre mice (data not shown).

Critical Role of Sema6A–PlxnA2/A4 Interaction in Termination of Neuronal Migration

How does Sema6A–PlxnA2/A4 interaction contribute to establishing correct positioning of SLNs? We consider that the mislocation of SLNs observed in KO mice is a consequence of a failure in termination of migration. Goffinet (1984) proposed that when neurons terminate migration, the affinity between the neurons and the RG fibers is attenuated, thus leading to detachment of migrating neurons from their glial substrate. The present study is consistent with this hypothesis: first, PlxnA2/A4 expression in SLNs appeared to be up-regulated at the terminal phase of migration, whereas Sema6A was concurrently supplied from VZ cells to their endfeet, and second, both Sema6A and PlxnA2/4 appeared to be localized to the site where the detachment probably takes place. Given that Sema6A–PlxnA2/A4 trans-interactions generally mediate a repulsive response (Battistini and Tamagnone, 2016), the most likely scenario is that SLNs terminate migration as a result of Sema6A–PlxnA2/A4 interactions, which may reduce the affinity between migrating SLNs and their RG substrates.

Arguably, the fact that PlxnA2/A4 mRNAs are broadly expressed in the upper layers from P0 raises the possibility that SLN mislocation occurred as a secondary effect. We cannot completely exclude this possibility. However, highly localized distribution of Sema6A in the L1 region indicates that Sema6A can directly interact with PlxnA2/A4 expressed by SLNs. Further studies are needed to validate this view.

An additional prediction by Goffinet (1984) is that an enhancement of neuron–neuron adhesion would contribute to neuronal detachment from the substrate. Our findings are consistent with this idea: SLNs were less densely packed in both Sema6A and PlxnA2/A4 dKO mice at P3 (e.g., Figure 7), whereas forced PlxnA2 expression in SLNs of the dKO mice induced aggregation of these neurons (e.g., GFP/PlxnA2-high cells, Figure S3F). This apparent up-regulation of neuronal adhesion may be a direct consequence of Sema6A–PlxnA2/A4 signaling, because Plxn signaling also mediates axon fasciculation (Rünker et al., 2008, Rünker et al., 2011, Suto et al., 2005), probably by enhancing axon-axon adhesiveness via homophilic binding (Ohta et al., 1995). Alternatively, the neuronal aggregation may have been a secondary result of a reduction of neuron–glia adhesion. In any case, the relative strength of adhesiveness among different cell types is integral to understanding the termination of migration and the consequent correct positioning of SLNs.

Possible Downstream Signaling Pathway of Sema6A–PlxnA2/A4

Plexin family members are known to work via small GTPases, which elicit repulsive effects on axons by regulating cytoskeletal and adhesive dynamics (Alto and Terman, 2017). In addition, in Drosophila, Mical, an oxidoreductase enzyme that disassembles F-actin, has been shown to directly interact with the cytoplasmic region of PlxnA and mediate a repulsive activity (Yang and Terman, 2013). Interestingly, in mice and chickens, Sema-mediated repellent interactions between boundary cap cells and immature spinal motor neurons act as a stop signal to regulate positioning of motor neurons (Bron et al., 2007, Mauti et al., 2007). This repulsive signal is mediated by MICAL3 (Bron et al., 2007). It would be interesting to examine whether the same effector also functions in Sema6A–PlxnA2/A4 signaling for proper SLN positioning.

Independence of Sema6A–PlxnA2/A4 Signaling from Reelin Signaling

In the present study, we found no obvious interactions between Sema6A–PlxnA2/A4 and reelin signaling. Although cortical phenotypes of mice in which either signaling pathway is mutated are somewhat similar, we noticed on closer examination that their phenotypes are distinguishable. For example, in Reln mice (Kohno et al., 2015), SLNs rarely reach beneath the pial surface, whereas they do so in Sema6A KO/cKO and PlxnA2/A4 dKO mice. It is conceivable that these two individual signaling pathways regulate termination of neuronal migration independently or additively to ensure proper development of the laminar architecture of the cortex.

Implied Existence of a Specific Mechanism for Termination of SLN Migration

Sema6A–PlxnA2/A4 signaling is required for superficial but not deeper layer neurons. We consider two alternative scenarios for this differential effect: (1) neurons in different layers use different signaling pathways for detachment from RG fibers and (2) termination of SLN migration requires a specific mechanism in addition to the termination mechanisms required for neurons in other layers. For example, the last-born SLNs are special in terms of lacking a following migratory neuron. Further studies are needed to uncover whether a temporally regulated change in the interaction between neurons and their substrate is a general mechanism for termination of neuronal migration in the cortex.

Limitations of the Study

We found evidence that an interaction between Sema6A and PlxnA2/A4, which are expressed on RG cells and SLNs, respectively, regulates laminar positioning of SLNs. We speculate that migrating SLNs become detached from their substrate, RG, on arrival at their final positions, due to the repulsive interaction between Sema6A and PlxnA2/A4. However, technical difficulty prevented us from confirming this by directly observing the process of detachment itself. Analysis of downstream effector(s) of Sema6A–PlxnA2/A4 signaling may also be needed for clarifying the process in mechanistic terms in the future.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.