SARA regulates neuronal migration during neocortical development through L1 trafficking

During embryonic corticogenesis, radial glia (RG) cells divide symmetrically to produce two identical daughter RG at the ventricular zone (VZ). Alternatively, RG divide asymmetrically to produce an RG cell and an intermediate basal progenitor. In turn, basal progenitors predominantly generate neurons (Taverna et al., 2014). Newborn neurons migrate radially to the cortical plate (CP) after temporally halting in the intermediate zone (IZ). These neurons mainly become projection neurons by extending axon processes tangentially across the IZ (Noctor et al., 2004). The final laminar cortical layers are formed ‘inside-out'. Later-born neurons migrate past the earlier born neurons, and hence older neurons are situated deeper in the cerebral cortex than younger neurons (Hatanaka et al., 2004).

Neuron migration requires the concerted establishment of traction force by forming attachments between the leading processes (LPs) and the extracellular matrix (ECM) and/or neighboring cells, as well as detachment at the cell rear (Elias et al., 2007; Jossin and Cooper, 2011; Shieh et al., 2011). Emerging evidence shows that perturbing components crucial for endocytic trafficking (e.g. Rab5, Rab11, dynamin, clathrin) leads to impaired cortical neuron migration by altering the surface distribution of adhesion molecules (e.g. N-cadherin, β1-integrin) (Shieh et al., 2011; Kawauchi et al., 2010).

L1 (L1CAM) is a cell adhesion molecule best known for its importance in axon growth, guidance and fasciculation (Kamiguchi et al., 1998). Recent studies showed that L1 suppression also disrupts the radial locomotion of cortical neurons (Kishimoto et al., 2013). Mutations in the L1 gene have been linked to hydrocephalus in several human congenital brain disorders (Kamiguchi et al., 1998; Weller and Gärtner, 2001). In vitro studies have shown that axonal plasma membrane expression of L1 is tightly regulated by a transcytotic pathway (Wisco et al., 2003; Yap et al., 2008). However, little is known about the mechanisms underlying the surface expression of L1 in vivo and its physiological relevance during cortical development.

Smad anchor for receptor activation (SARA; also known as Zfyve9) is an early endosome (EE) protein that acts as a downstream effector of Rab5-mediated EE fusion (Hu et al., 2002). Like overexpression of constitutively active Rab5, overexpression of SARA also causes the enlargement of EEs (Itoh et al., 2002; Seet and Hong, 2001). Furthermore, SARA has been shown to be involved in the vesicular trafficking of a variety of proteins, including Delta, Notch, uninflatable, rhodopsin, transferrin and Smad (Coumailleau et al., 2009; Loubéry et al., 2014; Chuang et al., 2007; Hu et al., 2002; Tsukazaki et al., 1998). A recent report revealed that SARA is unequally distributed to the two daughter cells of zebrafish spinal cord neural precursor cells that undergo asymmetric division, and that the expression level of SARA plays a determining role in the cell fate of the daughter cells (Kressmann et al., 2015). In the present study, we investigate the expression level and function of SARA during cortical development of the mouse brain.

We employed both biochemical and immunohistochemical methods to investigate SARA expression in the developing mouse neocortex. In velocity gradient density fractions of embryonic day (E) 15 mouse brains, SARA was co-enriched with two other EE markers: EEA1 and Rab5 (Fig. 1A). In E15 mouse cortical slices, SARA immunofluorescence appeared in bright puncta throughout all layers. SARA is particularly enriched in the apical endfeet of nestin-labeled RG at the ventricle borders (Fig. 1B).

We then determined whether SARA distributes equally into the daughters of apically dividing RG. We identified mitotic cell pairs in cortical slices of brains, which had been transfected at E13.5 with GFP by means of in utero electroporation (IUE), that exhibited characteristic condensed chromatin. Cortical slices were subjected to immunolabeling for endogenous SARA 40 h after transfection (Fig. 1C). To determine if the symmetrical or asymmetrical modes of divisions affected SARA distribution among the two daughter cells we measured their cleavage plane angle. As expected (Haydar et al., 2003), most apically diving cells presented a vertical cleavage plane (60-90°) relative to the horizontal ventricle border (Fig. 1D). We assessed SARA fluorescence intensity in each daughter cell and a ratio was calculated between cell pairs. Endosomes positive for SARA expression segregated similarly along all the focal planes (Fig. S1). For the three cleavage plane categories, SARA⁺ EEs were roughly equally distributed among the two cells with a ratio close to 1 (Fig. 1E). A similar analysis for Rab5 also points to a symmetrical distribution in apically dividing cells (Fig. 1F,G).

To investigate the function of SARA in RG, we performed loss-of-function analysis by delivering a plasmid encoding both SARA short-hairpin (sh) RNA (SARAsh) and GFP into E13.5 cortex. Scrambled control shRNA (Ctrolsh) provided a control. The knockdown (KD) effect in SARAsh was previously validated (Chuang et al., 2007; Arias et al., 2015) and confirmed by SARA immunohistochemistry in transfected cortical slices (Fig. S2A).

The distribution patterns of cells transfected with Ctrolsh or SARAsh were comparable in brains harvested 40 h after electroporation (Fig. 2A). To further test whether SARA plays a role in neurogenesis, we performed a cell cycle exit analysis. A single pulse of BrdU was given to mice 24 h after IUE, and the brain slices harvested 24 h later were immunolabeled for Ki67 and BrdU (Fig. 2C,D). The cell cycle exit index was similar between cells transfected with Ctrolsh or with SARAsh (Fig. 2B).

Furthermore, SARAsh transfection did not affect the expression pattern or number of Pax6-labeled apical progenitors and Tbr2 (Eomes)-labeled basal progenitors (Fig. 2E,F,I,J). The fraction of mitotic [phospho-histone H3 (PH3)-positive] cells was indistinguishable between Ctrolsh and SARAsh transfections(Fig. 1G,K). Similar to Ctrolsh-transfected brain slices, the expression of SARAsh did not alter the organization of RG processes as revealed by nestin immunostaining (Fig. S2B). Finally, SARAsh-transfected cells were able to reach the IZ and display the neuron marker Tuj1 (Tubb3) to a similar extent as control cells (Fig. 2H,L; Fig. S2C). These data derived from our KD experiments suggest that SARA is dispensable for the proliferation and differentiation of RG progenitors.

We next examined brains 3 days after transfection and found that the large majority of control neurons (>60%) had migrated into the CP (Fig. 3A,D). By contrast, only ∼14% of SARA-suppressed neurons had migrated into the CP; instead, a significantly higher fraction (∼75%) remained in the IZ (Fig. 3B,D).

Increased apoptosis is unlikely to explain the reduced number of SARA KD neurons at the CP because the number of cells expressing cleaved PARP (Fig. S3A,B), a main target of active caspase 3 (Tewari et al., 1995), was similar in the control and SARAsh-transfected brains.

At this developmental stage, the large majority (∼74%) of control neurons, either in the IZ or CP, had already developed a bipolar morphology with a pia-directed LP. The orientation of the LPs was mostly between 75° and 90° relative to the ventricle border (Fig. 3A,E,H). By contrast, most SARAsh-transfected neurons in the IZ were tilted: ∼30% angled 0°-20° and ∼50% angled 20°-45° (Fig. 3B,H, arrowheads in box 2 of F). Although ∼6% of SARA-suppressed neurons developed a vertical angle, both the LPs and trailing processes of these neurons were abnormally curved and branched (Fig. 3B,I,J, arrows in box 1 of F).

To rule out off-target effects of SARAsh, we showed that cells transfected with the rescue plasmid SARAsh/SARA*, which encodes SARAsh, shRNA-resistant SARA (from cDNA) and GFP, were able to reach the CP and displayed rather normal looking, pia-oriented LPs (Fig. 3C,D,G). These results suggest that the SARA KD-mediated phenotypes are specific.

The migration defect of SARA-suppressed neurons was unlikely to be due to defective polarity. This is supported by the finding that the large majority (∼90%) of SARAsh-transfected cells had their γ-tubulin-labeled centrosome localized normally between the nucleus and the LP, indicating a largely intact internal polarity (Fig. S3C).

Previous studies showed that L1 KD neurons exhibit migration defects (Kishimoto et al., 2013) similar to SARA KD neurons. In E16.5 mouse cortex, L1 was abundantly expressed on the axonal tracks that are tangentially distributed along the IZ (Fig. 4A), as expected (Fushiki and Schachner, 1986; Chung et al., 1991). Weak L1 signal was also detected in the CP and marginal zone (MZ), and little or no L1 was detectable in the VZ and subventricular zone (SVZ).

Furthermore, we show that L1 signals were frequently associated with SARA-labeled endosomes in dissociated cortical neurons (Fig. 4B), indicating that L1 is trafficked through SARA-expressing EEs. Consistent with this notion, the surface distribution of L1 in axons is increased in SARA KD neuronal cultures (Arias et al., 2015). Thus, we hypothesized that the SARA KD-induced neuronal migration defect could be related to changes in the surface expression of L1. To test this, we labeled the endogenous surface L1 of neurons isolated from transfected brains. In these experiments, an antibody that specifically recognizes the extracellular domain of L1 was used for immunostaining under non-permeabilizing conditions. Notably, surface L1 was predominantly expressed in a single process (probably the future axon) of GFP-transfected neurons (arrows in Fig. 4C, top). By contrast, increased L1 was apparent on almost all processes of SARAsh-transfected neurons (arrows in Fig. 4C, middle). Quantification studies showed that SARAsh-transfected neurons had ∼3.5-fold more L1 on the plasma membrane of neuronal processes compared with the control GFP-transfected neurons (Fig. 4D). By contrast, the surface expression of β1-integrin in SARA KD neurons was unchanged compared with controls (Fig. S4A,B).

To evaluate whether the increase in surface L1 affects cell adhesion, we examined cortical cells dissociated from brain cortices 2 days after transfection and cultured for an additional 2 days. We used nestin and doublecortin (DCX) antibodies to label progenitor cells and neurons, respectively. We found that neuronal processes extended from control neurons preferentially contacted nestin⁺ progenitor cells rather than DCX⁺ neurons (Fig. 4E,F; Fig. S4C). By contrast, SARAsh-transfected neurons preferentially contacted other DCX⁺ neurons, rather than nestin⁺ progenitor cells.

The above results prompted us to hypothesize that the impaired IZ exit of migrating neurons upon SARA suppression is due to elevated L1 surface expression. Consistent with this model, in 3-day transfected brains the majority of neurons overexpressing L1-YFP also failed to reach the CP and were retained in the IZ (Fig. 5A,C). These neurons were horizontally or obliquely aligned, similar to SARA-suppressed neurons (Fig. 5B,D).

Resembling SARAsh-transfected neurons, the surface L1 signal was similarly increased in multiple processes of L1-YFP-expressing neurons isolated from transfected brains (Fig. 4C,D). Consistently, the ectopically expressed L1-YFP was also prominently detected in the LPs of migrating neurons (arrows, Fig. 5B). 3D rendering of confocal images showed that both the soma and processes of L1-YFP-overexpressing neurons in the IZ made multiple contacts with the processes extended by other transfected neurons (Fig. 5B,E,F; Movie 1), indicating that increased cell-cell contact in these neurons is the result of increased surface L1.

Next, we addressed whether upregulated L1 expression accounts for SARA KD-mediated phenotypes. We generated an L1 shRNA (L1sh)-expressing plasmid and validated its KD effect by immunoblotting of co-transfected L1-YFP cell lines (L1 levels were reduced to below 40% of control levels; Fig. 5G). Similar to previous reports (Kishimoto et al., 2013), neurons transfected for 3 days with our L1sh construct led to migration arrest around the IZ (Fig. S5A,B). Finally, 3-day transfected neurons coexpressing SARAsh and L1sh displayed a vertical orientation and ∼60% of them reached the CP, similarly to Ctrolsh-transfected neurons (Fig. 5H,I). These results suggest that SARA KD-mediated phenotypes can be effectively rescued by simultaneous KD of L1.

To further delineate the mechanism underlying the SARA KD-mediated neuronal migration defect, we imaged transfected neurons on live brain slices for 12-14 h. As expected (Noctor et al., 2004), multipolar Ctrolsh-transfected neurons exited the IZ, acquired a bipolar morphology with a pia-directed LP and migrated towards the CP (Fig. 6A,C; Movie 2). By contrast, neurons expressing SARAsh exhibited characteristic multipolar dynamics throughout the recording. These neurons constantly extended and retracted their neuronal processes in various directions, and frequently changed orientation (Fig. 6B,C; Movie 3). Almost no SARAsh-transfected cells were able to exit the IZ during the 14 h imaging period. Furthermore, the migration speed of transfected neurons was significantly slower in SARAsh-expressing cells compared with their control counterparts (Fig. 6D).

By examining 5-day transfected brains, we found that both Ctrolsh- and SARAsh-transfected neurons were able to reach the CP with their LPs contacting the MZ (Fig. 7A-C). Immunolabeling of SARAsh-transfected neurons showed that SARA was still knocked down at this time point (Fig. S6A,B). Thus, these results indicate that the delayed IZ exit of SARAsh-transfected neurons was overcome with time. Consistent with retarded migration through the IZ, SARA-suppressed neurons were localized in the more superficial cortical layers relative to Ctrolsh-transfected neurons (Fig. 7B,C). In addition, SARA-suppressed neurons tended to cluster together, presumably due to stronger adhesion.

Notably, the LPs of both control and SARA-suppressed neurons reached the MZ (Fig. 7D,E). Despite their similarity in length (Fig. 7F), the LPs of SARAsh-transfected neurons exhibited a greater curvature than those of controls (Fig. 7D,E), as reflected in a statistically significant difference in the curvilinear index (Fig. 7G). These results suggest that SARA is important for timely migration and correct positioning in the CP and for the LP morphogenesis of cortical projection neurons.

We next evaluated whether the delayed migration of SARA KD neurons affected later corticogenesis. We studied transfected brains at postnatal day (P) 15, by which time neuronal migration is largely completed (Molyneaux et al., 2007). At this time point, SARA KD neurons were distributed to more superficial layers compared with control neurons (Fig. 8A,B). Using layer-specific markers, we found that ∼70% of Ctrolsh-expressing cells localized to the deep (V and VI) layers that expressed Ctip2 (Bcl11b), whereas a minor fraction expressed the upper-layer (II-IV) marker Cux1 (Fig. 8C). By contrast, the large majority of SARA KD neurons mapped to Cux1⁺ layers II-IV. Consistently, relative to the controls, the fraction of SARA KD neurons positive for Ctip2 and Cux1 decreased and increased, respectively (Fig. 8D). At P15 we did not detect any major differences in apical dendrite morphology between Ctrolsh- and SARAsh-transfected cells (not shown).

These data support the idea that SARA KD causes migration defects that affect laminar cortical layer development in the postnatal brain.

Previous studies showed that SARA-expressing EEs play an important role in determining asymmetric cell division in Drosophila sensory organ precursor (SOP) cells by modulating Notch-Delta signaling (Coumailleau et al., 2009). However, SARA itself is dispensable for the asymmetric segregation of Delta and Notch and SOP development. Subsequent studies showed that in dividing cells of the zebrafish spinal cord SARA segregates asymmetrically, whereas Rab5 distributes symmetrically. In this model, the daughter cells that inherited more SARA-expressing EEs remain as proliferative progenitors through a Notch-dependent pathway, whereas the daughter cells containing fewer SARA-expressing EEs tend to leave the cell cycle and differentiate into neurons (Kressmann et al., 2015).

Here, we show that in the E15 mouse cortex SARA, as well as Rab5, is roughly equally distributed between the two daughters of a dividing RG cell. Furthermore, the expression patterns of specific markers for neuron, apical progenitor, basal progenitor, and mitotic cells are comparable between SARAsh- and control-transfected brains. Progenitor cell cycle exit is also unaffected after SARA KD. Thus, our results suggest that SARA is not crucial for mitosis and cell fate determination of RG in the developing mouse neocortex. This finding is unexpected given that SARA is enriched in RG apical endfeet, and that Notch is crucial for the asymmetric cell division of rodent RG (Bultje et al., 2009). We cannot completely exclude the possibility that residual SARA in the SARAsh-transfected apical progenitors obscured our detection of its role in neurogenesis. However, we favor a model whereby a redundant endocytic pathway is used to compensate for Notch-mediated asymmetric division when SARA is suppressed in the mammalian neocortex. SARA is a FYVE (Fab1p, YOTB, Vac1p and EEA1) domain-containing protein, several of which, such as EEA1 and Hrs, are also enriched in EEs and have been shown to be involved in endocytic vesicular trafficking (Stenmark and Aasland, 1999; Panopoulou et al., 2002) (see below).

Both the cell distribution and live imaging studies showed that SARA suppression causes a substantial delay in the migration of postmitotic neurons through the IZ. Increased surface expression of L1 is likely to contribute to this phenotype because: (1) SARA KD neurons are stalled at the IZ, a region in which L1 is highly expressed (Fushiki and Schachner, 1986; Chung et al., 1991); (2) overexpression of L1 phenocopies SARA KD phenotypes; and, most importantly, (3) L1 suppression can rescue SARA KD-mediated defects.

The surface level of L1 appears to be tightly regulated during neuronal development: neurons expressing either too little (Kishimoto et al., 2013; Demyanenko et al., 1999) or too much (this work) L1 exhibit impaired migration and morphology/orientation of their LPs. Weakened adhesions may fail to provide sufficient traction force for forward movement, whereas excessive adhesions cause cells (and/or cell processes) to stick to each other and/or ECM, preventing their detachment (Shikanai et al., 2011). L1 is typically enriched on the axonal surface of wild-type neurons (Fig. 9A), but is increasingly expressed on all processes of SARA KD or L1-YFP-overexpressing neurons (Fig. 9B). We showed that SARA-suppressed neurons preferentially adhere to other neurons at the expense of their adhesion to progenitor cells. In vivo, the increased L1 in LPs of migrating neurons might enable preferred binding with neurites of other IZ-localized neurons and/or axonal processes of CP neurons. L1 is a transmembrane protein with extracellular immunoglobulin-like and fibronectin III domains. It may interact with L1 molecules (transhomophilic interaction) or other adhesion molecules such as axonin-1 (contactin 2), integrins and phosphacan (Ptprz1) (transheterophilic interaction) (Kuhn et al., 1991; Yip and Siu, 2001; Milev et al., 1994). Alternatively, L1 can also bind to the ECM. Enhanced adhesion might explain the abnormal orientation of mutant neurons and their reduced migration towards the CP (Fig. 9B).

The multipolar-to-bipolar transition is an important step before engaging in RG-guided migration. Namely, multipolar cells in the upper SVZ and lower IZ need to acquire a bipolar morphology in which a stable pia-directed LP heads the migration out of the IZ and into the CP (Kriegstein and Noctor, 2004; LoTurco and Bai, 2006). Our time-lapse experiments revealed that the multipolar-to-bipolar transition is substantially impaired by SARA KD. These neurons consistently drifted randomly at the IZ and exhibited numerous highly dynamic neurites. The SARA KD-induced radial migration defect was partially overcome with time. In line with the idea that SARA-suppressed neurons arrive later at the CP, their somata resided in the most upper cortical layers relative to control neurons.

Although the importance of cytoskeletal and motor proteins (e.g. filamin, LIS1, DCX, RhoA, myosin II, dynein) in neocortical development has been extensively established (Nagano et al., 2004; Tsai et al., 2005, 2007; Bai et al., 2003; Cappello et al., 2012; Solecki et al., 2009), only recently has evidence begun to emerge that endosomal trafficking also participates in guided neuron migration and proper positioning (Shieh et al., 2011; Kawauchi et al., 2010; Zhou et al., 2007). It is well established that internalized surface molecules can either be transported back to the plasma membrane from the EEs via a ‘fast recycling pathway' or through Rab11⁺ recycling endosomes via a ‘slow recycling pathway' (Maxfield and McGraw, 2004). It is also well known that transferrin and its receptor are trafficked through the slow recycling pathway. SARA has been shown to regulate transferrin recycling; overexpression of SARA significantly delays the recycling of transferrin as well as of transferrin receptor (Hu et al., 2002). However, the level of SARA is not critical for the internalization rate of transferrin.

In cortical neurons, L1 has been shown to be transcytosed from the somatodendritic domain to the axonal plasma membrane through the slow recycling pathway (Wisco et al., 2003). Our unpublished data (Y.-C. Hsu, J.-Z.C. and C.-H.S.) show that recycling endosome genesis requires SARA. Thus, L1 might be shifted to the fast recycling pathway in SARA-suppressed cells, and this might explain its overall increased surface expression. Since only a trace amount of internalized L1 (∼10%) undergoes lysosomal degradation (Schäfer et al., 2010), it is less likely that dysregulated degradation levels accounts for the increased L1 surface expression in SARA KD cells.

Consistent with the role of SARA in Rab5-mediated EE fusion (Hu et al., 2002), SARA KD neurons exhibit branched LPs and migration defects similar to those observed after Rab5 KD (Kawauchi et al., 2010). Rab5 KD phenotypes rely on elevated surface expression of N-cadherin. Although we do not know whether SARA KD also affects N-cadherin expression, surface β1-integrin expression was unchanged in SARA KD neurons. This indicates that the increase in L1 in SARA KD neurons is selective. Our other data suggest that SARA KD phenotypes can be accounted for by the increase in L1. Importantly, elevated L1 expression also explains the neuronal alignment defects observed in SARA KD but not in Rab5 KD (Kawauchi et al., 2010).

In addition to SARA, several other proteins known to be involved in Rab5-mediated EE fusion (e.g. Rabex-5, EHD1/4, EEA1) have been implied in L1 surface recycling (Aikawa, 2012; Yap et al., 2010; Lasiecka et al., 2014). For example, the surface expression of L1 is increased in dendrites of EHD1-overexpressing neurons (Yap et al., 2010). Similar to our current finding, a compensatory mechanism emerges later to alleviate this defect. This redundancy on trafficking regulation reiterates the importance of tight control of the L1 surface level. Notably, L1 is upregulated in post-stroke cortex and has been implicated in axonal regeneration after spinal cord injury (Carmichael et al., 2005; Roonprapunt et al., 2003; Chen et al., 2007). Thus, our finding that SARA regulates the surface expression level of L1 in the cerebral cortex in vivo might be of clinical relevance.

Finally, although highly controversial (Bakkebø et al., 2012; Runyan et al., 2012), SARA has been reported to be important for activation of the TGFβ signaling pathway (Tsukazaki et al., 1998; Itoh et al., 2002). Whether this functional aspect of SARA plays a role in L1 surface expression remains to be investigated.

Expression and shRNA constructs were generated and transfected as described in the supplementary Materials and Methods.

Velocity density gradient sedimentation was conducted as previously described (Sachdev et al., 2007). E15.5 mouse brains were harvested and homogenized. The high-speed supernatant fraction was fractionated on a 5-20% linear sucrose gradient in 11 ml Tris-KCl buffer (20 mM Tris pH 7.5, 50 mM KCl, 5 mM MgCl₂, 0.5 mM EGTA, 0.5 mM DTT, protease inhibitors) at 32,000 rpm (100,000 g) at 4°C for 16 h. Equal amounts of each fraction were subjected to immunoblotting assays using a standard method and the antibodies described in the supplementary Materials and Methods.

IUE procedures were performed on E13.5 mouse brains as previously described (Li et al., 2011; Artegiani et al., 2012). Briefly, the embryonic ventricles were injected with a mixture of plasmids (2 µg/µl), Fast Green dye and Tris-EDTA buffer. Immediately, three square electric pulses (30 V) were delivered using a BTX electroporator (Harvard Apparatus). Embryos were placed back to their original position and sutured. Female mice were allowed to recover from anesthesia on a warm plate.

For cell cycle exit analysis, pregnant mice (20-30 g) were injected with 1 mg BrdU 24 h after IUE, and fetal brains were harvested 24 h after BrdU treatment. Electroporated brains were harvested at the indicated time points. Embryonic brains were fixed by immersion in 4% paraformaldehyde (PFA) and 0.0125% glutaraldehyde overnight at 4°C, while postnatal animals were transcardially perfused with 4% PFA and brains postfixed by immersion in the same fixative. Brains were embedded in low melting point agarose, and sectioned by vibratome (40 µm). All animal manipulations were performed in accordance with the guidelines for animal experiments at Weill Medical Cornell IACUC, at Instituto M. y M. Ferreyra CICUAL, and Landesdirektion Sachsen.

In some experiments, electroporated brain cortices were dissociated with trypsin, dispersed on poly-L-lysine-coated coverslips (Thermo Scientific) and incubated at 37°C in Neurobasal medium (Gibco) supplemented with B27, N2 and Glutamax (Gibco). For co-culture of primary cortical neurons with nestin⁺ cells, Neurobasal medium supplemented with 10% horse serum was used. At the indicated time points (2 or 5 days after seeding), cells were fixed with 4% PFA and 4% sucrose in PBS followed by immunolabeling using a standard protocol and antibodies described in the supplementary Materials and Methods. Surface L1 or β1-integrin labeling was conducted similarly, except that Triton X-100 was omitted during the incubation with the primary and secondary antibodies. Cells were then permeabilized with 0.2% Triton X-100 before cytoskeleton counterstaining.

Details of quantitative image analyses, including statistical analysis, are provided in the supplementary Materials and Methods.

Brains were harvested 2 days after electroporation (E15.5) and embedded in low melting point agarose at 37°C. The tissue was immediately sliced at 250 µm with a vibratome. Brain sections were transferred to membrane inserts for 6-well culture plates (BD Falcon), preincubated with 1.5 ml Neurobasal medium supplemented with 10% horse serum and antibiotics. The tissue was allowed to recover and to attach to the insert for 2 h before imaging.

Time-lapse imaging was performed with a Zeiss LSM 780 inverted confocal microscope using a Plan-Apochromat 10× air objective (NA 0.45) and a temperature-controlled incubator (Life Imaging Services), with CO₂ set to 5% (Pecon). Stacks were captured every 30 min for 12 h. The migrating neurons were tracked and analyzed with the Manual Tracking plugin in ImageJ (NIH). At least six slices from three independent experiments were analyzed for each condition.

We thank the light microscopy facility of the BIOTEC/CRTD for excellent support; H. Kamiguchi (RIKEN, Japan) and D. Carrer (INIMEC-CONICET, Argentina) for reagents; M. Lorenzatti for assistance with Fig. 9; and A. Caceres for discussion.