Alpha protocadherins and Pyk2 kinase regulate cortical neuron migration and cytoskeletal dynamics via Rac1 GTPase and WAVE complex in mice

Diverse clustered protocadherins are thought to function in neurite morphogenesis and neuronal connectivity in the brain. Here, we report that the protocadherin alpha (Pcdha) gene cluster regulates neuronal migration during cortical development and cytoskeletal dynamics in primary cortical culture through the WAVE (Wiskott-Aldrich syndrome family verprolin homologous protein, also known as Wasf) complex. In addition, overexpression of proline-rich tyrosine kinase 2 (Pyk2, also known as Ptk2b, Cakβ, Raftk, Fak2, and Cadtk), a non-receptor cell-adhesion kinase and scaffold protein downstream of Pcdhα, impairs cortical neuron migration via inactivation of the small GTPase Rac1. Thus, we define a molecular Pcdhα/WAVE/Pyk2/Rac1 axis from protocadherin cell-surface receptors to actin cytoskeletal dynamics in cortical neuron migration and dendrite morphogenesis in mouse brain.


Introduction
The human brain contains approximately 86 billion neurons, and each neuron engages in several thousand specific synaptic connections, resulting in complex neural networks with over 10 15 specific connections. These complex neural circuits are required for normal brain function, and inappropriate assemblies of neural circuits underlie neurodevelopmental and neuropsychiatric disorders (Hyman, 2008). A remarkable feature of neurodevelopment is the long-distance neuronal migration from the site of origin to the final destination (Angevine and Sidman, 1961;Ayala et al., 2007). For example, cortical immature neurons generated from the proliferative ventricular and subventricular zones (VZ/SVZ) migrate radially through specific phases to appropriate laminar positions in an 'inside-out' manner and then differentiate into distinct subtypes of cortical neurons (Angevine and Sidman, 1961;LoTurco and Bai, 2006;Rakic, 1974). The cortical migration phases include somal translocation, multipolar migration, and glial-guided locomotion (Ayala et al., 2007;Cooper, 2014;Noctor et al., 2004). Newly born bipolar neurons in SVZ assume multipolar or stellate morphology and migrate randomly in the intermediate zone (IZ), moving tangentially, up, or down (Ayala et al., 2007;Cooper, 2014;Jossin and Cooper, 2011;Nadarajah et al., 2003;Noctor et al., 2004;Tabata and Nakajima, 2003). They then transit into bipolar again near the border of IZ/CP (cortical plate) and resume final radial migration to settle in appropriate cortical layers (Ayala et al., 2007;Cooper, 2014;Jossin and Cooper, 2011;Nadarajah et al., 2003;Noctor et al., 2004;Tabata and Nakajima, 2003). Abnormal neuronal migration results in various neurodevelopmental and psychiatric diseases (Ayala et al., 2007;LoTurco and Bai, 2006;Valiente and Marín, 2010); however, the underlying molecular mechanisms for the abnormal neuronal migration is largely unknown.
Human genetics studies have implicated mutations of the clustered protocadherin (Pcdh) cell adhesion genes in the 5q31 region for various developmental and psychiatric disorders (Anitha et al., 2013;Iossifov et al., 2012;Pedrosa et al., 2008;Shimojima et al., 2011). Similar to Dscam1 in Drosophila (Zipursky and Sanes, 2010), diverse clustered Pcdh genes play an important role in establishing neuronal identity and connectivity in the vertebrate brain (Garrett et al., 2012;Lefebvre et al., 2012;Molumby et al., 2016;Nicoludis et al., 2016;Rubinstein et al., 2015;Schreiner and Weiner, 2010;Suo et al., 2012;Thu et al., 2014;Wu and Maniatis, 1999). In mice, 58 clustered Pcdh genes are organized into three closely linked Pcdh a, b, and g clusters (Pcdha, Pcdhb, and Pcdhg) (Wu et al., 2001). The Pcdh a and g clusters are each consisted of variable and constant regions, similar to that of the Ig, Tcr, and Ugt1 gene clusters (Wu, 2005;Wu and Maniatis, 1999;Wu et al., 2001;Zhang et al., 2004). In particular, the variable region of the mouse Pcdha cluster contains 12 highly similar 'alternate exons', a1-a12, whose promoters are stochastically activated by distal enhancers, and two divergent c-type 'ubiquitous exons', ac1 and ac2, whose promoters are constitutively activated by distal enhancers ( Figure 1A) (Guo et al., 2012). Each variable exon is separately spliced to the common set of downstream constant exons, generating diverse mRNAs and proteins. CCCTC-binding factor (CTCF)/Cohesin-mediated topological chromatin-looping domains are crucial for proper expression of Pcdha proteins (Guo et al., 2015;Huang and Wu, 2016). Each variable exon encodes an extracellular domain (ectodomain EC1-6), a transmembrane eLife digest There are hundreds of billions of neurons in a human brain, and each one can form several thousand connections with other neurons. This complex network determines our thoughts, memories, personality, and behavior, but how does it form? During brain development, specific areas give rise to new neurons, which then migrate long distances to other parts of the brain. Upon arrival, they generate several structures, called dendrites, which connect with other neurons.
To distribute themselves correctly, the migrating immature neurons must be able to travel long distances and steer clear of one another. The dendrites from a single mature neuron must also avoid each other, a phenomenon known as self-avoidance. Certain membrane-spanning proteins, called clustered protocadherins, may help neurons achieve this. The portion of the protocadherins that sits on the cell surface is highly variable, and acts as a zipcode that helps cells to recognize one another. However, the section of the protein inside the cell varies little and is shared by all members of a protocadherin family. When the clustered protocadherin is 'switched on', this internal segment can trigger a cascade of reactions that create changes in the cell. Yet, little was known about the nature of this signaling cascade.
Using gene editing in mice, Fan, Lu et al. focus on the signaling cascade of the clustered protocadherin alpha family. The experiments show that the internal portion of these proteins interacts with a protein complex called WAVE. It also inhibits an enzyme known as Pyk2, which increases the activity of another enzyme called Rac1 GTPase, that then further activates WAVE. This results in the WAVE complex also interacting with the internal skeleton inside the neurons and dendrites, which regulates the ability of these cells to migrate and of the dendrites to avoid each other.
Many brain conditions, such as autism spectrum disorders or depression, result from abnormal neuronal migration and connectivity. Mutations in the genes of clustered protocadherins increase the risk of these disorders. By showing how these proteins help to regulate the migration and connectivity of neurons, Fan, Lu et al. add to our understanding of brain development in health and disease. segment, and a juxtamembrane variable cytoplasmic domain (VCD) (Shonubi et al., 2015;Wu and Maniatis, 1999), whereas the three constant exons encode a common membrane-distal constant domain (CD) of all Pcdha proteins ( Figure 1A). This suggests that diverse extracellular cues converge on a single intracellular signaling pathway. However, the functional significance of this intriguing arrangement remains obscure.
Here, we report that Pcdha proteins play a critical role in neuronal migration and cytoskeletal dynamics. Specifically, we define an actin cytoskeleton remodeling pathway by which Pcdha regulates lamellipodial and filopodial dynamics and neuronal migration as well as dendrite morphogenesis through interaction with WAVE complex via the WIRS (WAVE interacting receptor sequence) motif of Pcdha constant domain (CD). In addition, Pyk2 regulates cortical neuron migration by inactivating the small GTPase Rac1. Given that actin cytoskeletal dynamics are central for neurite morphogenesis and neuronal migration, our findings have interesting implications for mechanisms of Pcdha functions in dendrite self-avoidance and neuronal self/nonself recognition in normal brain development as well as aberrant neuron migration and dendrite morphogenesis underlying complex neurodevelopmental diseases.

Defective cortical neuron migration with Pcdha knockdown
We mapped the embryonic expression pattern of Pcdha by using a GFP knockin mouse line (Pcdha GFP ) (Wu et al., 2007) and found that Pcdha proteins are expressed throughout the developing forebrain ( Figure 1B). Immunostaining with an antibody against alpha constant domain (aCD) revealed that Pcdha proteins are expressed in all cortical regions and most prominently in the intermediate zone and marginal zone (IZ and MZ) of the developing neocortex ( Figure 1C). RT-PCR with isoform-specific primers showed that, starting at E10, every member of the Pcdha cluster is expressed in the developing brain ( Figure 1-figure supplement 1A). Pcdha knockdown (aKD) with two independent shRNAs, each targeting a distinct subdomain of the constant region by in utero electroporation (IUE), revealed a significant decrease of migrating neurons in the cortical plate (CP) and a concomitant increase within the lower intermediate zone, suggesting defects in multipolar migration ( Figure 1D and Figure 1-figure supplement 1B). The aKD multipolar neurons in the intermediate zone display stunted processes, as shown by lucida drawings ( Figure 1E). Live cell imaging of brain organotypic slice culture demonstrated the slower velocity of multipolar migration of aKD neurons compared to controls ( Figure 1F-H and Video 1). In addition, early born aKD neurons also have migration defects, suggesting that Pcdha is also required for glia-independent somal translocation ( Figure 1I and J). This suggests that Pcdha is required for migration of immature cortical neurons.
To rule out the possibility of altered progenitor proliferation, we labeled aKD mouse brain with BrdU and analyzed cell proliferation. Compared with controls, aKD results in no significant difference  (Wu et al., 2007). The phenotypic discrepancy may be due to known genetic compensation mechanisms induced by deletion but not knockdown (Rossi et al., 2015).

Rescuing cortical neuron migration by single Pcdha isoforms and constant domain
To rescue the migration defect, we constructed shRNA-resistant forms of a6 (a6*), which represents members of the alternate a1-a12, and of the two divergent c-types (ac1* and ac2*) ( . Indeed the single a6* isoform rescues the aKD migration defect. The Pcdh ac1* also rescues the migration defect; however, ac2* does not ( Figure 2A and Figure 2-figure supplement 1B). This suggests that ac2 has distinct functions other than cortical neuron migration, consistent with very recent findings that ac2 endows serotonergic neurons with a single cell-type identity and specifically mediates the axonal tiling and assembly of serotonergic neural circuitries .
To investigate whether the extracellular domain and transmembrane segment play a role in cortical neuron migration, we replaced them with a myristoylation signal to attach the shRNAresistant intracellular domain (ICD) to the plasma membrane (Myr-a6ICD*, Myr-ac1ICD*, Myr-a c2ICD*) (Figure 2-figure supplement 1A). We found that Myr-a6ICD* and Myr-ac1ICD* rescue the migration defect, and Myr-ac2ICD* does not (Figure 2-figure supplement 1C and D). This suggests that the intracellular domain of Pcdha plays an important role in cortical neuron migration. To investigate why Myr-ac2ICD* cannot rescue the migration defect, we constructed an ac2 VCD-deleted form, which is, by definition, a with DAPI. (D) Cortical coronal sections of E19.5 mouse brain which were electroporated at E15.5 with control (SCR: scrambled) or aKD (a shRNA1 or a shRNA2) plasmids. Nuclei were counterstained with DAPI. Quantification of GFP + cell distribution across the cortex (divided into ten equal bins) is shown on the right. n = 6 brains for each group. Statistical significance was assessed using one-way ANOVA, followed by a post-hoc Tukey's multiple comparisons test. (E) Representative multipolar neurons and their lucida drawings in the red boxes shown in (D). Asterisks indicate multipolar cells. (F) Embryonic brains were electroporated at E15.5 and organotypic slices were prepared from brains at E17.5. Representative frames from a 10 hr timelapse imaging experiment are shown. Asterisks indicate one migrating cell. See also Video 1. (G and H) Typical migration traces (G) and migration velocity (H) of control and aKD neurons in a time-lapse experiment shown in (F). n = 15 cells for each group. Student's t test. (I) Cortical coronal sections of E15.5 embryos electroporated at E12.5 with control or aKD plasmids. Nuclei were counterstained with DAPI. (J) Quantification of E15.5 control and aKD GFP + cells in CP, IZ, and VZ. n = 5 brains for SCR, n = 4 brains for aKD. Student's t test. Data as mean ± SEM. ****p<0.0001. *p<0.05. See

Rescuing cortical neuron migration by the WAVE complex
Recent studies linked Pcdha6 to the WAVE complex through the WIRS (WAVE interacting receptor sequence) motif within the Pcdha constant domain . We thus investigated whether Pcdha regulates cortical neuron migration through WAVE. Remarkably, we found that overexpression of either WAVE2 (Wasf2) or Abi2 in vivo rescues the cortical neuron migration defect of aKD neurons ( Figure 2B) although they themselves have no apparent influence on cortical neuron migration ( Figure 2C). Consistently, endogenous Pcdha and WAVE2 co-localize in primary cultured cortical neurons ( Figure 2D and E). In addition, mutating the WIRS motif (from FITFGK to FIAAGK) of a6*, ac1*, and Myr-aCD* (a6*-AA, ac1*-AA, and Myr-aCD*-AA) abolishes the rescue effect ( Figure 2F and G, in comparison to Figure 2A and A role of Pyk2 in cortical neuron migration Pcdha physically interacts with and negatively regulates the Pyk2 kinase (Chen et al., 2009). In addition, we previously found that Pcdha regulates dendritic and spine morphogenesis through inhibiting Pyk2 activity (Suo et al., 2012). To this end, we investigated whether knockdown of Pyk2 could rescue cortical neuron migration defects of aKD. Although Pyk2 (Ptk2b) knockdown (Pyk2KD) per se or CRISPR knockout of Pyk2 (Pyk2KO) does not affect cortical neuron migration ( Figure 3A  We next asked whether overexpression of Pyk2 (Pyk2OE) could recapitulate aKD cortical neuron migration defects. We found that the majority of Pyk2OE cells are stalled in the middle intermediate zone (mIZ) ( Figure 3B), a stage little later than the stalling of aKD cells ( Figure 3A). In addition, these mIZ cells have aberrant multipolar morphology with supernumerary primary processes in comparison to single leading processes of control cells ( Figure 3C-E). For the very few Pyk2OE cells in the lower cortical plate (CP), they harbor elaborated leading processes ( Figure 3F and G); by contrast, control cells displayed typical bipolar morphology with a single or bifurcated thick leading process ( Figure 3F and G). Pyk2OE leads to the inhibition of Rac1 activity (Suo et al., 2012). As Rac1 is thought to provide the spatial information for actin polymerization (Tahirovic et al., 2010), loss of Schematics of Pcdha protein structure with the AA mutation of the WIRS motif. (G) Cortical coronal sections of E19.5 embryonic brains electroporated at E15.5. Nuclei were counterstained with DAPI. Quantification of GFP + cell distribution is shown on the right. n = 6 brains for each group. Data as mean ± SEM. Statistical significance was assessed using one-way ANOVA, followed by a post hoc Tukey's multiple comparisons test. ****p<0.0001. See Rac1 activity leads to aberrant actin polymerization at many sites with no controlled spatial information, resulting in supernumerary primary processes ( Figure 3C-E) and more branchy morphology ( Figure 3F and G). Finally, time-lapse imaging showed that there is a significant difference of velocity of cortical neuron migration between Pyk2OE and control cells ( Figure 3H and I, and Video 2). These data suggest that Pyk2OE partially recapitulates cortical neuron migration defects.
We next examined the orientation of the Golgi apparatus of cells in mIZ, which is essential for transporting vesicles for oriented motility (Jossin and Cooper, 2011), by immunostaining with a Golgi marker GM130 ( Figure 3J). Most Golgi complexes are normally localized in front of the cell nucleus and are oriented toward the cortical plate (Jossin and Cooper, 2011). However, the polarity of most Pyk2OE cells is disrupted, showing oblique or inverted orientation of the Golgi apparatus ( Figure 3J  To rule out the potential nonspecific effect of the CAG promoter, which is active in both progenitors and postmitotic neurons, we ectopically overexpressed Pyk2 at E15.5 only in postmitotic neurons using the NeuroD promoter (Jossin and Cooper, 2011). We found that Pyk2OE under the NeuroD promoter also significantly impairs cortical neuron migration in postmitotic neurons (Figure 3-figure supplement 1E-G). Taken together, this suggests that Pcdha regulates cortical neuron migration, at least in part, through inhibiting Pyk2 kinase activity.

Regulation of cortical neuron migration by Pyk2 via Rac1
We previously found that Rac1 is epistatic downstream of Pyk2 in dendrite development and spine morphogenesis (Suo et al., 2012). To investigate whether Pyk2-Rac1 pathway also functions in cortical neuron migration, we overexpressed a constitutive active form Rac1 (Rac1 Q61L ) in Pyk2OE neurons. We found that Rac1 Q61L rescues defects of multipolar migration and morphology of Pyk2OE neurons ( Figure 4A-C), although Rac1 Q61L itself has no apparent effect on cortical neuron migration ( Figure 4D). However, overexpression of another constitutively active form of Rac1 (Rac1 G12V ) impairs cortical neuron migration ( Figure 4D) (Konno et al., 2005) and cannot be used to rescue, likely because it has a lower affinity for GTP and thus lower constitutive activity than Rac1 Q61L . Thus, the two constitutively active forms of Rac1 have distinct roles in cortical neuron migration ( Figure 4A and D). Together, we conclude that Pyk2OE inhibits multipolar-bipolar transition and leads to aberrant branchy morphology in the intermediate zone by inactivating the small GTPase Rac1.

Dissection of Pyk2 domain in cortical neuron migration
Pyk2 functions as an enzyme through its middle kinase domain and as a molecular scaffold through its N-terminal FERM (four-point-one, ezrin, radixin, moesin) domain (Figure 4-figure supplement 1A) (Chen et al., 2009;Lev et al., 1995;Suo et al., 2012). We systematically engineered Pyk2 by mutating a series of key residues of its enzymatic kinase cascade. We found that overexpression of Pyk2 Y402F , an autophosphorylation mutant that still can be activated by endogenous Pyk2, as well as Pyk2 Y579F , Pyk2 Y580F , and Pyk2 Y881F , still recapitulate the migration defects of aKD (Figure 4-figure supplement 1A and B). However, overexpression of Pyk2 K457A , which has a mutation at the catalytic center and is completely kinase-dead (Suo et al., 2012), cannot recapitulate the migration defects of aKD (Figure 4-figure supplement 1A and B). This suggests that the catalytic activity of overexpressed Pyk2 is essential for recapitulating the migration defects of aKD.
Remarkably, overexpression of the Pyk2 FERM domain alone recapitulates the blocking activity of Pyk2OE ( . This is consistent with that Pyk2 has important kinase-independent functions in contextual fear memory (Suo et al., 2017). Together, we conclude that both Pyk2 kinase cascade and FERM scaffold are crucial for blocking cortical neuron migration.
As stated above, constitutive active Rac1 Q61L rescues the blocking effect of Pyk2OE ( Figure 4A). However, we found that Rac1 Q61L cannot rescue the blocking activity of FERM domain (Figure 4figure supplement 1D). This suggests that constitutive active form of Rac1 only functions downstream of the kinase cascade but not the FERM scaffold of Pyk2.

Pcdha in lamellipodial formation and cytoskeletal dynamics
We next investigated actin dynamics underlying neuronal migration in primary cultured cortical neurons. The early development of primary cultured neurons can be divided into two stages: stage 1, in which the cell body is surrounded by flattened lamellipodia and stage 2, in which the lamellipodia transform into definitive processes with growth cones (Dotti et al., 1988). At stage 1, we found that the size of lamellipodia around cell cortex in aKD neurons decreases significantly compared with controls ( Figure 5A and B). In addition, a6*, ac1*, or Myr-aCD* rescues the aKD lamellipodial shown on the right. n = 6 brains for each group. Data as mean ± SEM. Statistical significance was assessed using one-way ANOVA, followed by a post hoc Tukey's multiple comparisons test. ns, not significant; ***p<0.001; ****p<0.0001. See aKD, aKD+ac1*, aKD+ac1*-AA, and aKD+ac2*; n = 15 cells for aKD+a6*, aKD+a6*-AA; n = 12 cells for aKD + Myr-aCD*-AA, aKD + Abi2; n = 11 Figure 5 continued on next page defect. By contrast, ac2* does not rescue ( Figure 5C and D), which is consistent with that ac2* cannot rescue the defects of cortical neuron migration ( Figure 2A). Moreover, mutating the WIRS motif (from FITFGK to FIAAGK) in either a6*, ac1*, or Myr-aCD* abolishes their rescue effects ( Figure 5E and F), similar to the situation in cortical neuron migration ( Figure 2G). Finally, both WAVE2 and Abi2 rescue the lamellipodial defect ( Figure 5G and H). In addition, consistent with stage 1, both WAVE2 and Abi2 rescue the lamellipodial defect of stage 2 aKD neurons ( Figure 5-figure supplement 1G and H).
Finally, aKD lamellipodial dynamics are significantly impaired in comparison with control neurons, whose veil-like lamellipodia are motile and are constantly extending and retracting in both stage 1 and stage 2 neurons ( Figure 5I, Figure 5-figure supplement 1I, Video 3 and Video 4). These data demonstrated that Pcdha is indispensable for lamellipodial dynamics. Because lamellipodial dynamics are essential for cell migration (Krause and Gautreau, 2014), this suggests that cortical neuron migration defects of aKD are a consequence of impairment of lamellipodial formation and cytoskeletal dynamics.

A comparison between PcdhaKD and Pyk2OE in cytoskeletal dynamics
Consistent with that Pyk2KD rescues cortical neuron migration defects of PcdhaKD ( Figure 3A), we found that knockdown of Pyk2 in aKD cells Filopodia are thin membrane protrusion pushed by underlying actin bundles and filopodial formation is also dependent on Arp2/3 complex (Mattila and Lappalainen, 2008), we found that Pyk2OE results in a significant increase of filopodial number per stage 1 neuron as well as of primary neurite number per stage 2 neuron despite no alternation in aKD cells ( Figure 6D-G). Finally, similar to the rescue of cortical neuron migration defects of PykOE ( Figure 4A), we found Rac1 Q61L rescues both lamellipodial and filopodial defects of Pyk2OE ( Figure 6D-G). In summary, although both aKD and Pyk2OE impact cytoskeletal dynamics, they have subtle differences on both lamellipodia and filopodia.
To see whether growth cones with lamellipodia and filopodia are affected in vivo, we co-electroporated Lifeact, an actin marker, with either aKD or Pyk2OE plasmids into the developing mouse cortex. In the lower intermediate zone, aKD neurons exhibit abnormal enrichment of Lifeact-labeled actin structures in stunted processes and cell bodies, while the control neurons extend long processes with growth cones (Figure 6-figure supplement 1A). In the upper intermediate zone, Pyk2OE neurons exhibit branchy morphology with multiple aberrant processes; however, the control neurons have normal bipolar morphology with single leading processes and growth cones (Figure 6-figure supplement 1B).

Discussion
Recent studies revealed that a zipper-like ribbon structure assembles from combinatorial cis-and trans-interactions between like-sets of the clustered Pcdhs located in apposed plasma membranes of neighboring cells (Nicoludis et al., 2016;Rubinstein et al., 2015;Schreiner and Weiner, 2010;Thu et al., 2014;Wu, 2005). These protocadherin interactions could provide enormous diversity and exquisite specificity for neuronal connectivity and neurite self-avoidance required for mammalian brain development. While exquisite specificity is determined by strict homophilic trans-interactions of highly diversified EC2/3 (Goodman et al., 2017;Molumby et al., 2016;Nicoludis et al., 2016;Rubinstein et al., 2015;Schreiner and Weiner, 2010;Thu et al., 2014;Wu, 2005); enormous diversity is mainly generated by promiscuous cis-interactions of highly conserved EC5/6 (Nicoludis et al., 2016;Rubinstein et al., 2015;Schreiner and Weiner, 2010;Thu et al., 2014;Wu, 2005). One intriguing genomic architecture of the Pcdha cluster is multiple tandem variable exons followed by a single set of three constant exons, encoding a common cytoplasmic constant domain, which is shared by all members of the Pcdha family ( Figure 1A) (Huang and Wu, 2016;Wu and Maniatis, 1999). The extracellular domains of Pcdha provide enormous diversity and exquisite specificity for cell recognition and adhesion (Nicoludis et al., 2016;Rubinstein et al., 2015;Schreiner and Figure 6 continued stage 2 neuron. Student's t test; n = 12 cells for both groups. (D) Primary cultured cortical neurons, derived from E17.5 embryonic cortices which were electroporated at E15.5 with indicated plasmids, were in-vitro cultured for 24 hr and immunostained by a Tuj1 antibody for tubulin, counterstained with phalloidin for F-actin. Arrowheads, lamellipodia; Arrows, filopodia. (E) Quantification of lamellipodial size per stage1 neuron shown in (D). Statistical significance was assessed using one-way ANOVA, followed by a post hoc Tukey's multiple comparisons test. n = 10 cells for each group. (F) Quantification of filopodial number per stage 1 neuron shown in (D). Statistical significance was assessed using one-way ANOVA, followed by a post hoc Tukey's multiple comparisons test. n = 10 cells for each group. (G) Quantification of primary neurite number per stage 2 neuron shown in (D). Statistical significance was assessed using one-way ANOVA, followed by a post hoc Tukey's multiple comparisons test. n = 13 cells for each group. All data are presented as a scatter-dot plot. The median is shown as a line with the interquartile range. ****p<0.0001; ***p<0.001; ns, not significant. See  Thu et al., 2014;Wu, 2005). However, the intracellular Pcdha signaling pathway is largely unknown.
We propose a Pcdha-based WAVE clustering model for cortical neuron migration (Figure 7). Distinct Pcdha isoforms on the cell surface recruit WAVE complex to the cell cortex under the plasma membrane. This is strongly supported by (1) the specific interaction between members of the Pcdha family and the WAVE complex through the WIRS motif in Pcdha constant domain ; (2) the rescue of migration and lamellipodial defects of aKD neurons by WAVE complex subunits WAVE2 and Abi2; and (3) the abolishment of the rescue effect by WIRS mutations. The WIRS motif of members of the Pcdha family binds to a composite surface formed by Abi2 and Sra1 of the WAVE complex . In addition, the Pcdha proteins may also recruit WAVE complex through the direct binding of Abi2 C-terminal SH3 domain to the four protocadherin PXXP motifs, which are specific to the constant domain of the Pcdha but not Pcdhg family (Wu and Maniatis, 1999). Consistently, WAVE2 and Abi2 are required for growth cone activity during cortical neuron migration (Xie et al., 2013).
Our finding that aKD blocks lamellipodial and filopodial formation and cytoskeletal dynamics is also consistent with the WAVE clustering model. Taken together, we suggest that Pcdha regulates the formation and dynamics of lamellipodial and filopodial protrusions underlying cortical neuron migration through the WAVE/Pyk2/Rac1 axis (Figure 7). We noted that aKD neurons stall in the lower intermediate zone and Pyk2OE neurons stall in the middle intermediate zone. In other words, aKD phenotype is more severe than that of Pyk2OE. In addition, aKD neurons display stunted processes while Pyk2OE neurons have branchy morphology. Consistently, the WAVE clustering model suggests that, in addition to disinhibition of Pyk2 and consequently inhibition of Rac1, aKD also impairs the membrane recruiting of the WAVE complex directly (Figure 7).
It is puzzling why Pcdhac2 is different from other members of the Pcdha family (Figures 2A,  5C and D, and However, a recent study revealed an intriguing role of ac2 in serotonergic axonal local tiling and global assembly . Given the known role of variable cytoplasmic domain of clustered Pcdh proteins in their cytoplasmic association (Shonubi et al., 2015), the unique sequences of the ac2 variable cytoplasmic domain may restrict its role to axonal tiling of serotonergic neurons but not cortical neuron migration.

Generation of CRISPR mice
Mouse lines of Pyk2KO and Pyk2 Y402F were generated by using CRISPR/Cas9. Briefly, sgRNA scaffold sequences were constructed in the pLKO.1 plasmid. The construct was then used as template for amplifying a PCR product containing T7 promoter and sgRNA target sequence. The PCR product was gel-purified and used as templates for in vitro transcription of sgRNA (T7-Transcription Kit, Invitrogen). Cas9 mRNA was transcribed in vitro from linearized pcDNA3.1-Cas9 plasmid (T7-ULTRA-Transcription Kit, Ambion). Both Cas9 mRNA and sgRNAs were purified (Transcription Clean-Up Kit, Ambion), mixed in M2 (Millipore) at the concentration of 100 ng/ml, and then injected into the cytoplasm of fertilized eggs of C57BL/6 mice. For Pyk2 Y402F mice, single-stranded oligo-donor nucleotides (ssODN) with mutation at Y402 residue and nonsense mutation at PAM sequence were coinjected together with the Cas9 mRNA and sgRNA. After equilibration for 30 min, 15-25 injected fertilized eggs were transferred into fallopian tube of pseudopregnant ICR mouse females. Offspring of these mice were genotyped by PCR, restriction endonuclease digestion, and Sanger sequencing. All oligos used are listed in Supplementary file 1.

In utero electroporation (IUE)
IUE was performed as previously described with modifications (Saito and Nakatsuji, 2001). Briefly, dams were anesthetized with pentobarbital sodium. pLKO.1-shRNAs (2 mg/ml) for knockdown or pCAG-Myc (2 mg/ml) constructs for overexpression were mixed with GFP-expressing plasmid pCAG-eGFP (0.5 mg/ml) and 0.05% fast green. Laparotomy was performed to expose the uteri. The plasmid mixture was injected into the lateral ventricle of the embryonic brain. Five electrical pulses were applied at 40 Volts for a duration of 50 ms at 900 ms intervals using a tweezertrode (3 mm, BTX) with an electroporator (Gene Pulser System, Bio-Rad). The uterine horns were placed back into the abdominal cavity to allow the embryos to continue normal development.
Cortical neuron primary culture, organotypic slice culture, and timelapse imaging For cortical neuron primary culture, electroporated cortices were collected from E17.5 embryos in Hanks' Balanced Salt Solution (HBSS) with 0.5% glucose, 10 mM Hepes, 100 mg/ml penicillin/streptomycin. The cortices were then digested with 0.25% trypsin for 10 min at 37˚C. The reaction was terminated with 0.5 mg/ml trypsin inhibitor for 3 min at room temperature (RT). The cortical tissues were gently triturated in the plating medium (MEM with 10% FBS, 1 mM glutamine, 10 mM Hepes, 50 mg/ml penicillin/streptomycin) until fully dissociated. Cell viability and density were determined using 0.4% trypan blue and a hemocytometer. The dissociated cells (1 Â 10 5 ) were plated into fourwell chamber or 35-mm glass-bottom Petri dish precoated with 100 mg/ml poly-L-lysine (Sigma) and 5 mg/ml laminin (Invitrogen). The cells were incubated with 5% CO 2 at 37˚C for 4 hr. The plating medium was then replaced with a serum-free culture medium (Neurobasal medium, 2% B27, 0.5 mM glutamine, 50 mg/ml penicillin/streptomycin supplemented with 25 mM glutamate). For immunocytochemistry, cells were cultured for additional 20 hr in vitro (hiv).
For cortical organotypic slice culture, the head of E17.5 embryos were briefly placed in 70% ethanol and the brains were carefully dissected. The brains were embedded in 3% low-melting agarose and glued to the chuck of a water-cooled vibratome (Leica). The 250-mm-thick whole-brain coronal sections were cut and collected in the sterile medium. The organotypic slices were carefully placed in a 0.4 mm membrane cell culture insert (Millipore) in a six-well plate. Slices were cultured in slice culture medium: 67% Basal Medium Eagle (BME), 25% HBSS, 5% FBS, 1% N2, 1% penicillin/streptomycin/glutamine (Invitrogen) and 0.66% glucose (Sigma). Slices (three per well) were cultured in sixwell plates at 37˚C and 5% CO 2 , incubated for 6-8 hr. The membrane insert with slices was then transferred on to a glass-bottom Petri dish (MatTek). Images were taken at 3 mm steps with 10-15 optical sections and were captured every 15 min for up to 16 hr with the Nikon A1 confocal laser microscope system.
For single-cell time-lapse imaging, cortical neurons were plated into a 35-mm glass-bottom Petri dish. Images were taken at 1 mm steps with 10-15 optical sections and were captured every 5 min for up to 10 hr with Nikon A1 confocal laser microscope system.

Immunocytochemistry, immunohistochemistry, and imaging
Primary cultured cortical neurons were washed once with PBS, fixed in 4% PFA for 20 min at RT, washed and permeabilized with 0.2% Triton X-100 for 10 min. After blocking with 5% BSA, cells were incubated with primary antibodies at 4˚C overnight followed by incubation of secondary antibodies for 1-2 hr at RT. F-actin was labeled by Alexa-546 phalloidin (Sigma). For immunohistochemistry, the dams were sacrificed, and embryonic brains were fixed in 4% PFA overnight at 4˚C. The brains were then sectioned at 50 mm with a vibratome (Leica). Sections were washed three times in PBS, blocked in 3% BSA, 0.1% Triton X-100 in PBS for 1 hr at RT, and then incubated with primary antibodies at 4˚C overnight and secondary antibodies at RT for 1-2 hr. Cell nuclei were visualized with DAPI. Images were collected with a confocal microscope (Leica) under a 10x objective for brain sections. High-resolution images were collected under a 60x oil objective with a 3x digital zooming factor for primary cultured neurons.

Cell culture and western blot
HEK293T cells were maintained in DMEM with 10% FBS and 100 mg/ml penicillin/streptomycin. Cultured cells were transfected using Lipofectamine 2000 (Invitrogen). Total protein of HEK293T cells was extracted by lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% NP40, 0.5% sodium deoxycholate, 0.1% SDS) with protease inhibitors and then centrifuged at 12,000 Â g at 4˚C for 30 min. The lysates were subjected to Western blot analyses.

Reverse transcriptase PCR (RT-PCR)
Total RNA was extracted from embryonic mouse brain tissues with TRIzol (Ambion). The reversetranscription reaction was performed with 1 mg total RNA preparations. All oligos used are listed in Supplementary file 1.

Statistical analysis/image analysis and quantification
For each group, the IUE experiments were performed using at least three pregnant female mice, by which we usually harvested at least six embryonic brains. We obtained 15~20 sections from each electroporated brain, and quantified one typical section per brain. Nearly identical areas in the presumptive somatosensory cortices of anatomically matched brain sections were chosen for imaging and quantification. For bin analysis, the cortices were divided into ten equal bins and all GFP + neurons in each bin were counted. In total, about 150~300 cells were counted per section. Statistical significance was assessed using one-way ANOVA, followed by a post hoc Tukey's multiple comparisons test.
In primary culture experiments, the development stage of cultured neurons were defined as in Dotti's paper: at stage 1, the cell body was surrounded by flattened lamellipodia; at stage 2, the lamellipodia transformed into neural processes with growth cones (Dotti et al., 1988). We immunostained the cultured cells with Tuj1 (Neuron-specific class III beta-tubulin) antibody, a neuron-specific marker, to exclude differentiated glia or radial glia. For quantification, we selected neurons with typical stage 1 or stage 2 morphology based on GFP and phalloidin signals. For stage 1 neurons, we selected the lamellipodia region by the wand tool in the ImageJ software (NIH) and measured the area size. For stage 2 neurons, the neurite tips with F-actin-enriched protrusions two folds larger than its width were defined as 'neurite with lamellipodia'. Sholl analysis was performed as previously described (Suo et al., 2012).
The significance of differences between two groups was analyzed using unpaired Student's t tests. One-way ANOVA was used for multiple comparisons by the GraphPad software. Shen, Hong Shao, Lun Suo, Data curation, Formal analysis, Investigation; Qiang Wu, Conceptualization, Supervision, Funding acquisition, Project administration, Writing-review and editing Author ORCIDs Qiang Wu http://orcid.org/0000-0003-3841-3591

Ethics
Animal experimentation: Animal experimentation: All procedures involving animals were in accordance with the Shanghai Municipal Guide for the care and use of Laboratory Animals, and approved by the Shanghai Jiao Tong University Animal Care and Use Committee (protocol #: 1602029). Data availability All data generated or analysed during this study are included in the manuscript and supporting files. Source data files have been provided for all figures.