Hierarchical clustering of gene expression patterns in the Eomes + lineage of excitatory neurons during early neocortical development

Validation of the analytic grouping technique

A method to quantify the optical density of the in situ patterns (QISPs) across the cerebral wall was developed (see Methods). The method of QISP generation is potentially subject to error due to variability of ROI positioning and histological heterogeneity between cortical areas and sections. To assess the precision of the QISP method, multiple measurements along the rostral caudal axis were performed using a set of E14.5 in situ hybridizations for Dab1 (Additional file 6), which at E14.5 is strongly expressed in the VZ and cortical plate (CP) [32], providing two distinct regions of peak signal. The Dab1 in situ profile was measured across the cerebral wall at 5 different points within the rostral cortex on a single in situ image (Additional file 6a). The Pearson’s correlation of these five distinct quantifications ranged from 0.94 to 0.99 with an average value of 0.97 (n = 5). This indicates that values obtained with the ROI approach were consistent and largely insensitive to the precise position of the ROI. Similarly, we compared the Pearson’s correlation of Dab1 signal from three different, sagittal in situ images derived from sections of rostral cortex along the lateral-to-medial axis (Additional file 6b). These correlations ranged from 0.87 to 0.97 with an average correlation of 0.93 (n = 4). Importantly, for both the rostral-caudal and lateral-medial planes the peak signals within the VZ and CP were always restricted to within 10% of the cerebral wall (i.e., ± 2 sROI). These tests validated the QISP analysis by indicating robust concordance of optical density measurements between ROIs from different positions in the rostral cortex and along the lateral-to- medial axis.

To objectively evaluate genes of interest we quantified the in situ pattern for Eomes itself, reasoning that up-regulated genes would display a QISP equivalent to, or localized more superficially than, the QISP derived for Eomes (Tbr2). This assumption is based on the accepted observation that, at this developmental stage, the most mature neurons are located more superficially in the cortex than immature neurons [33]. Eomes displayed little expression in ventricular sROIs_1-2 and higher expression in the more superficial sROIs_3-11. We therefore compared the average in situ expression in sROIs_1-2 versus the average expression in sROIs_3-20. Transcripts where this ratio exceeded 1.0 (i.e., where expression was strongest in the deep VZ) were considered invalid. This procedure identified 11 invalid genes which were then removed from the initial, post-Genepaint list of 328 genes. Thus 97% (n = 317) of the ≥ 3-fold up-regulated, RMA ≥ 7 transcripts displayed a Genepaint-derived in situ hybridization pattern that was spatially consistent with the Eomes + excitatory neuron lineage.

Analytic grouping of the E14.5 excitatory neuron transcriptome

To explore in greater detail groups of spatially co-expressed genes, the 317 QISP-validated genes were imported into the Multiple Experiment Viewer [27, 28], the signal (expression) values normalized by gene, and then hierarchically clustered. This procedure groups together similar QISPs, and can reveal groups of genes that are spatially co-expressed and thus potentially co-regulated [34].

The clustering of similar QISPs revealed several distinct spatial distributions of dynamically regulated genes in the radial dimension of the developing cerebral cortex (Figure 3). Approximately 70% of up-regulated QISPs within the Eomes + lineage display peak expression in the CP and upper intermediate zone (IZ). The majority of these QISPs displayed similar (i.e., tightly clustered) expression patterns, characterized by gradually increasing expression in the IZ and peak expression in the CP. The most heterogeneous expression patterns were observed in those QISPs with peak expression in the VZ and IZ, with a number of QISPs displaying dual expression peaks (e.g., Gdap1, Bcl6) or expression restricted to either the IZ or VZ (e.g., Neurod1 or Sstr2).

Four groups of genes – that is, four groups of clustered QISPs – were examined using a network analysis tool (IPA, Ingenuity Systems). The four groups were: (1) all 317 genes; (2) 44 “deep group” genes that achieve highest in situ expression in the VZ; (3) 49 “middle group” genes that achieve highest expression in the IZ and lower CP; and (4) 224 “superficial group” genes with greatest expression in the CP. This approach was adopted because neurons in the deep, middle, and superficial regions of the cortex have distinct morphologies and modes of migration [1] and thus can be expected to manifest dynamic regulation of genes related to morphology, navigation and migration.

Network analysis of all up-regulated genes

The “deep group” of 44 up-regulated genes (14% of the total), which is spatially co-expressed with Eomes (2.2 fold up), achieved highest expression in the VZ (sROIs _1-10). This group includes clusters 1, 5, 7, 8 and 11-15 (Figure 3). Along with Eomes, the deep group is characterized by the expression of the transcriptional regulators Neurod4, Bcl6, Tanc2 and Myt1. Earlier studies focusing upon this deep area of cortex revealed (a) neurons with a long leading process that translocate out of the VZ [19, 38, 39] and (b) compact intermediate precursor cells in the SVZ [40]. Thus, at this point in development the GFP + cells in the VZ and SVZ are predicted to represent a mixture of intermediate precursors and postmitotic neurons [41]. Consistent with this prediction, IPA analysis identified DNA Replication, Recombination and Repair as a functional category for these group (p = 1.37E-04; Table 1, Additional file 7). Surprisingly, Carbohydrate Metabolism was also identified (p = 3.00E-05): the deep group includes genes involved in the synthesis (C1ql3 and Lpl) as well as level regulation (Chga, Gipr, Pcsk2 and Sstr2) of D-glucose. The top “Canonical Pathways” include Protein Kinase A Signaling and Cardiac β-Adrenergic Signaling. Additionally, a number of genes involved in intracellular calcium signaling (Ryr3, Camk2b, Cacna1h) showed elevated expression in this group. Network analysis of deep group genes thus suggests that the phenotypic transition from radial glial precursor to intermediate precursor/immature cortical neuron coincides with significant changes in mechanisms of glucose metabolism and second messenger signaling.

The “middle group” comprises 49 genes, or 16% of the ≥ 3 fold up data set. The middle group QISPs, comprising clusters 9 and 10, are maximal in the IZ, where developing excitatory neurons display a multipolar morphology. Multipolar neurons (MPNs) possess multiple fine processes that lack definitive orientation, and migrate relatively slowly in the radial direction [1]. The top “Molecular and Cellular Functions” in this group are Cellular Development and Cell Morphology (Table 1, Additional file 7). The middle group is dominated by the basic helix loop helix (bHLH) transcription regulators Neurod6 (71.5 fold), Neurod1 (30.9 fold), Nhlh1 (11.6 fold) and Neurod2 (7.3 fold), which represent some of the most highly up-regulated genes in the entire data set. These bHLH transcription factors have well-described functions in excitatory cortical neuron differentiation and development [6, 42]. The top “Canonical Pathways” include Axonal Guidance Signaling and Semaphorin Signaling in Neurons, as well as Inositol Phosphate Metabolism. Of particular interest within the middle group genes is cluster 9, consisting of 26 genes that includes seven with known chemotropic signaling functions in axon guidance. The expression of this set of axonal guidance genes could promote the generation and extension of axons that are observed to trail the migrating cortical neuron in the upper IZ and CP [39, 43]. In addition to the highly up-regulated transcription factors, the top up-regulated genes include Vglut1/Slc17A6 (19.4 fold) a glutamate transporter that identifies excitatory cortical neurons [44]. Network analysis of “middle group” genes thus indicated that developing excitatory neurons in the IZ are molecularly defined by a high expression of canonical bHLH neuronal transcription factors and axonal guidance molecules, particularly those of the Semaphorin signaling pathway.

The “upper group” of genes (n = 224) were observed in clusters 2, 3, 4 and 6. These genes, comprising 71% of the up-regulated gene population, achieved highest expression in the developing cortical plate (CP), an area that contains migrating neurons with an elongated or bipolar shape, and post-migratory differentiating neurons. Within the upper group the most prominent transcriptional regulators were Tbr1 (43.1 fold) [45], Bhlhb5 (19 fold) [46], and Hivep2/Schnurri2 (18 fold). Additional transcription factors in the upper group included Prdm8 [47], Satb2 [48], Sox5 [11], and Bcl11b/Ctip2 [49]. All of these transcriptions factors, with the exception of Schnurri2, have demonstrated roles in corticogenesis. Schnurri2 is, however, a BMP signaling effector, with a neuronal function suggested by the observations that mice deficient in Schnurri2 are hyperactive and exhibit hypersensitivity to stress [50].

Neurons within the forming CP migrate along radial glia fibers, from which they later detach and extend dendrites. It is therefore not surprising that the top “Molecular and Cellular Functions” in the upper group are Cell-To-Cell Signaling and Interaction, Cellular Assembly and Organization, and Cellular Movement (Table 1, Additional file 7). The top “Canonical Pathways” include Corticotropin Releasing Hormone Signaling and CDK5 Signaling. A role for CDK5 signaling mechanisms is established for many neurodevelopmental functions, including the migration of cortical neurons along radial glial fibers and subsequent cortical layer formation and dendritogenesis [51, 52]. In contrast the role(s) of Corticotropin Releasing Hormone Signaling in these neurons is less clear. The eight genes from the ≥ 3 fold up-regulated category that populate the Corticotropin Releasing Hormone Signaling category include four with known functions in brain development, including the cannabinoid receptor Cnr1 [53], the transcription factor Mef2C [54], and nitric oxide synthase Nos1 [55]. The remaining genes in this group include the protein kinases Prkar1β, Prkcε and Prkcζ Network analysis of “upper group” genes thus suggest maturing excitatory neurons in the developing CP express a number of regulatory and signaling factors, including the transcription factors Tbr1, Bhlhb5, Hivep2/Schnurri2, Prdm8, Satb2, Sox5, and Bcl11b/Ctip2.

Analysis of highly up- and down-regulated transcription factors

To explore the spatial relationships of key transcription factors (TFs) in the cortical lineage, we identified all ≥ 3 fold up- and down-regulated genes associated with the Gene Ontology descriptor “transcription, nuclear” and clustered their respective QISPs (Figure 4). These transcription/nuclear QISPs span the cerebral wall at E14.5, with subsets of QISPs showing peak expression in the VZ, IZ and CP (Figure 4a). We were initially surprised that Eomes itself was upregulated only 2.2 fold, a value below our threshold for analysis. There are several reasons for this low fold up value. The first is that GFP- cells express Eomes mRNA (RMA = 10.7). The presence of Eomes mRNA in the absence of GFP protein could imply post-transcriptional regulation of the 5’ UTR of the Eomes message that is included in the Eomes: :eGFP transgene [26]. Similarly, the presence of Eomes mRNA in GFP- cells might reflect the rapid up regulation of Eomes mRNA in a subpopulation of cells that has yet to express sufficient GFP protein for FAC detection and sorting into the GFP + group. Finally, Eomes mRNA is sharply down regulated as neurons leave the SVZ [3], while GFP protein expression persists and is expressed in all cortical zones superficial to the SVZ, at this developmental time point. Thus the GFP + cell population is likely larger than the Eomes mRNA + cell population, effectively reducing the average Eomes mRNA expression level in the GFP + population. Thus post-transcriptional regulation and/or the difference between sharp Eomes mRNA regulation and the delayed and extended GFP protein expression may account for this relatively low fold up level for this critical TF.

We first compared these genes to an earlier study of bHLH-family TFs denoted as the “Dorsal Telencephalon Network” [4]. Comparison to this independent data set of 13 TFs revealed 7 (54%) to be members of our ≥ 3 fold up-regulated category: Neurod1 (31 fold), Neurod2 (7.3 fold), Neurod6 (72 fold), Tbr1 (43 fold), Id2 (4.1 fold), Nhlh1 (12 fold) and Satb2 (12 fold) (Figure 4). Other dorsal telencephalon TFs outside the ≥ 3-fold up-regulated category include the early genes Pax6 (7.6 fold down-regulated), Neurog 1 (1.3 fold down-regulated), Neurog 2 (1.1 fold down-regulated) and Eomes itself (2.2 fold up-regulated). Although Pax6 expression in the GFP + population was below threshold (RMA = 6.2) Neurog1, Neurog2, and Eomes were not (RMA values of 7.6, 10.1, and 11.9, respectively). This suggests that the GFP + neurons in our study substantially down-regulate Pax6 relative to neuronal precursors, but do express the neuronal specifying TFs Neurog1, Neurog2 and Eomes. Additionally, two genes specific to cortical layer V-VI, Etv1 and Otx1 [56], were down-regulated in our data (5 fold and 2 fold, respectively), an observation supported by strong Genepaint in situ hybridization signals at E14.5 in the ventricular zone for both genes. Because postnatal expression of Etv1 and Otx1 is specific to cortical layer V-VI, this result suggests temporally differential (i.e., developmentally early and late) requirements for these TFs during cortical development. In contrast to the “Dorsal Telencephalon Network” we identified the layer II-IV marker Satb2 within our ≥ 3 fold up-regulated data set (12 fold), an observation confirmed by Genepaint-derived in situ hybridization. Consistent with the dorsal telencephalic identity of the Eomes + population, none of the nine genes associated with the “Ventral Telencephalon Network” [4] were observed in the ≥ 3 fold up-regulated population. By comparison to published studies of excitatory neuron differentiation [2–4, 57], this TF expression analysis developmentally positions the “maturation spectrum” of the isolated E14.5 GFP + neurons to a period after the expression of early dorsal specification genes (e.g., Pax6), coincident with neural specification genes (e.g., Neurog1, Neurog2), and prior to the expression of late layer-specific markers (e.g., Etv1 and Otx1).

We hypothesized that the highly expressed “middle” and “deep” group TFs are critical molecular regulators, perhaps functioning to simultaneously repress precursor-specific TFs and promote the expression of TFs required for proper differentiation of cortical excitatory neurons. To test this idea, a network analysis tool (IPA; see Methods) was used to explore within our dataset possible regulatory interactions among the dynamically expressed TFs (Figure 5). All TFs from the ≥ 3 fold up- and down-regulated populations were imported into the pathway analysis and the network of direct and indirect interactions was identified. All interacting genes that were identified, but absent from the populations of ≥ 3 fold up- and down-regulated TFs, were assigned to a group denoted “IPA Interacting Genes.”

Approximately 430 connections were identified for TFs in the up- and down-regulated groups.

Only 4 direct interactions were reported between the “deep group” or “middle group” genes and either the “upper group” genes or the down-regulated genes (Figure 5). Numerous direct interactions, however (50 of 54), were detected that connected “middle” and “deep” genes to the “IPA Interacting Genes” group. Included in this group are signaling molecules that function non-cell autonomously in neuronal development: Notch signaling [58], Wnt-βcatenin signaling [59], VEGF [60], TGF β [37] and retinoic acid signaling [61]. Each of these signaling pathways has been shown to regulate either the production and/or differentiation of excitatory cortical neurons. Surprisingly, only 14, or ~3%, of all detected interactions were direct between any group of dynamically regulated TF genes (i.e., between the 4 lower boxes in Figure 5). The results of this network analysis support models of neurogenesis that predict complex interactions between specific cell autonomous (e.g. transcription regulatory networks) and non-cell autonomous signaling mechanisms (e.g., Notch, TGFβ signaling) during the maturation of cortical excitatory neurons.

Analysis of highly up- and down-regulated receptor genes

To more fully explore the role(s) of dynamically regulated receptors in the differentiation and migration of cortical neurons, we identified within the ≥ 3 fold up- and down-regulated gene populations those with the Gene Ontology descriptor “receptor activity,” and clustered their respective QISPs (Figure 6). The down-regulated receptors were enriched for signaling systems with established functions in neuronal precursors (Figure 6b, 6b’), including the Wnt-βcatenin signaling family members Sfrp2, Fzd8, Lrp4, and Wnt7a. Wnt-βcatenin signaling controls the production of neurons by neural precursors [62] and regulates intermediate precursor proliferation [59, 63]. The FGF receptors Fgfr2 and Fgfr3, and Notch family members Notchr1, Notchr2 and Notchr3 were also down-regulated. Notch receptors have multiple roles in neurogenesis and differentiation, including the regulation of neuronal precursor differentiation [58, 64], the promotion of glial fate [65], and dendritic growth [66, 67]. Fgfr2 is required for excitatory neuron development [68], with Fgfr3 required for attainment of normal cortical and hippocampal volume and caudal cortex development [69]. These results suggest that, along with a dynamic enhancement of pro-differentiation genes, genes responsible for the maintenance of precursor/progenitor/glial phenotypes are dyna mically suppressed at E14.5 in the lineage of excitatory cortical neurons upon exiting the VZ.

Thirty four receptors were identified in the ≥ 3 fold up category (Figure 6a, 6a’). The top category of up-regulated receptors includes those with established roles in axonogenesis (8.14E-11 p-value): members of the Semaphorin signaling family (Nrp1, Sema6d, Plxna2, Sema7a, Plxna4, Sema4g, Plxnd1); Netrin receptors (Unc5d, Dcc); the Slit signaling receptor Robo1. Because they are observed in clusters 9 and 10 (Figure 3), many of these canonical chemotropic signaling receptors are co-expressed in the IZ. Specifically, Semaphorin signaling molecules are expressed by excitatory neurons in the IZ soon after exit from the VZ, with Netrin receptors and Robo1 expressed in more superficial domains. These chemotropic signaling systems have multiple roles and may be enabled for axonal growth [70–72], neuronal migration [73], dendritic growth [74–76] and cell survival [77].

Also included among the most highly up-regulated receptors were eight GTP-binding protein-coupled receptors (Gprs) (2.87 E-5 p-value), including Mcr4 (melanocortin receptor 4), Cnr1 (cannabinoid receptor1), Sstr2 (somatostatin receptor 2), Gipr (gastric inhibitory polypeptide receptor), and the muscarinic acetylcholine receptor Chrm3. Cnr1 has multiple developmental roles including axonal guidance [78]. Specific developmental functions of Mcr4, Gipr, and Chrm3 are unknown, although Mcr4 interacts with Cnr1 to regulate food intake in adults [79], Gipr regulates insulin secretion after feeding [80], and Chrm3 is required for pancreatic insulin and glucagon release [81] and neuronal endocannabinoid release [82]. The up-regulated Gprs also include the orphan receptor Gpr21, as well as a highly conserved receptor, Gpr85 (also known as Sreb2), known to control brain size [83]. The Flamingo homolog Celsr3, which is required for neuronal projections to the anterior commissure as well as subcerebral targets, is also up-regulated [84].

Analysis of highly up- and down-regulated neurotransmitter receptors and synaptic genes

Neurotransmitters and their receptors have important functions in the developing nervous system [85]. Although our dataset contained no neurotransmitter receptors down-regulated ≥ 3 fold, the up-regulated category contained a variety of neurotransmitter receptor subunits. Cholinergic receptor subunit β2 (Chrnb2) and Chrna7 are up-regulated in the IZ and CP by 4.5 fold and 2.8 fold, respectively. These receptor subunits form heteromeric receptor complexes in the developing brain, suggesting an early role for cholinergic signaling as excitatory neurons migrate out of the VZ. Although deletion of Chrna7 in mice does not cause overt structural abnormalities [86], recent studies indicate that haploinsufficiency in CHRNA7 may cause a range of neurodevelopmental phenotypes associated with the 15q13.3 syndrome, including developmental delay, mental retardation and seizure [87]. Although no glutamate receptors were upregulated ≥ 3 fold, N-methyl-D-aspartate (NMDA) receptor subunits Grina and Grin1 were mildly up-regulated (2.6 fold and 2.0 fold, respectively; data not shown). In addition to mediating adult forms of synaptic plasticity, NMDA receptors may mediate forms of spontaneous activity in the developing cortex [88] as well as structural plasticity in the late embryonic and early postnatal period [89, 90]. mRNA encoding GABA (gamma-amino-butyric-acid) receptor subunits were also up-regulated, including Gabarg2 (4.6 fold) and Gabbrb2 (3.9 fold). Gabarb3 was upregulated 2.8 fold. Deficiency in Gabarg2 may be causal in Dravet syndrome, a form of childhood epilepsy [91]. GABA functions as an excitatory neurotransmitter in early development, xand deficiency in GABAA receptor subunits alter Inhibitory Post Synaptic Potentials (IPSPs) but not synapse number [92]. During early periods GABA can function as a depolarizing neurotransmitter and may relieve Mg²⁺ blockade of NMDA receptors [93, 94]. One glycine receptor subunit, Glra2 (10.7 fold up-regulated) was identified in this group, although mice lacking Glra2 display no overt abnormalities of cortical structure [95].

Synaptogenesis begins quite early in cortical development, at ~ E15 in subplate neurons in the rodent temporal cortex [96], and continues well into the postnatal period [97]. At the developmental point of our analysis, E14.5, few if any synapses are present in the developing CP. Despite this ultrastructural feature, we detected not only multiple mRNAs encoding neurotransmitter receptors, but also message for 22 other synaptic genes identified by the Gene Ontology descriptor “synapse” or “synaptic” (Figure 7). Included in this group are synaptotagmin IV (SytIV), a calcium sensing protein involved in neurotransmitter release [98], and synaptic vesicle glycoproteins Sv2a and Sv2b that may also be involved in regulated secretion [99]. Additionally detected were three members of the Leucine Rich Repeat Transmembrane family (Lrrtm1-3), which function as synaptic organizers [100]. These results admit the possibilities that neurotransmitter receptors and synaptic mRNAs expressed by these immature excitatory neurons either (a) do not encode functional proteins, or (b) encode proteins involved in non-canonical functions. In either case, the results are consistent with important roles for classical neurotransmitter signaling molecules in shaping the developmental trajectory of differentiating excitatory cortical neurons.

Analysis of highly up- and down-regulated voltage-dependent channels

Early patterns of electrical activity and calcium transients are known to shape the developing nervous system [101–104]. For this reason we queried the grouping of QISPs associated with known voltage-dependent channels (Figure 8). One down-regulated, and ten up-regulated, transcripts were identified and grouped. Of the up-regulated transcripts, four encode subunits of voltage-dependent calcium channels (Cacna1e, Cacna1h, Cacnb3, Cacna2d1), four encode voltage-dependent potassium channels (Kcna1, Kcnj3, Kcnq3, Kcna6), and two encode voltage-dependent sodium channels (Scn3b, Scn2a1). Hierarchical clustering of these transcripts’ in situ hybridization profiles revealed peak expression within the upper-most ten sROIs, corresponding to the IZ, CP and MZ. This expression pattern is spatially consistent with the acquisition of voltage-dependent ionic currents in multipolar and migrating neurons [103]. Only one gene (Kcnj10), encoding the weakly inwardly rectifying potassium channel subunit Kir4.1, was down-regulated (3.3 fold), and this observation was consistent with the corresponding in situ hybridization data. Although robustly down-regulated at this developmental time point by differentiating excitatory cortical neurons, we note that deficiency in KCNJ10 produces seizures, sensorineural deafness, ataxia, mental retardation, and electrolyte imbalance (SeSAME syndrome) [105, 106]. This analysis is consistent with an important contribution of electrical activity toward shaping the development of cortical neurons (and presumably their synaptic/functional networks), but also implicates specific channel types and sub-units in the manifestation of these early activity-dependent processes by cortical excitatory neurons.

Analysis of highly up- and down-regulated adhesion molecules

Spatial distributions of adhesion molecules, and changes in such distributions, can in principal convey substantial morphologic and migratory changes upon developing neurons. To group expression patterns of known adhesion molecules within the Eomes + neuronal lineage, we sorted within the ≥ 3 fold up-regulated category for the Gene Ontology descriptor “adhesion” (Figure 9). Twenty one genes were identified that were verified by screening the Genepaint database. These genes include several Immunoglobulin Super Family Members (IGSFMs), such as L1cam, Dscam, Nfasc, Cadm3, Cntn1 and Cntn2. All of these IGSFM genes have confirmed neurodevelopmental functions: L1cam is involved in neuronal migration and neurite outgrowth [107]; Dscam is a co-receptor for Netrin signaling and is required for commissural axon pathfinding [108]; Cadm3, also known as Nectin-like 1 or Syncam3, is a cell adhesion molecule involved in axonal myelination [109]; Nfasc and Cntn1 interact to promote the formation of paranodal axo-glial junctions during axonal myelination [110]; Cntn2, also known as Tag-1, is highly expressed on axons and can serve as a substrate for interneuron migration [111] (although Tag-1 knockout mice display no overt cortical phenotype [112], there is a disruption of paranodal myelin junctions [113]). Several Cadherins (Cadh8, Cadh11 and Cadh13) as well as Protocadherin 20 were observed to be up-regulated, although to date none of these Cadherin genes have identified roles in cortical development. Surprisingly, no members of the integrin family were observed in the ≥ 3 fold, or the ≥ 2 fold, up-regulated groups, suggesting this important family of adhesion proteins is not dynamically regulated at the transcriptional level in the Eomes + lineage at E14.5.

In contrast other adhesion molecules, such as Fibronectin (Fn1), were down-regulated in the developing Eomes + lineage. Two integrins that via heterodimerization form a fibronectin receptor (and an activating integrin for TGFβ signaling), Itgv and Itgβ8, were also down-regulated. Conditional deletion of Itgαv in radial glia and neurons causes cerebral hemorrhage [114] while reduced expression of αvβ8 integrin is associated with arteriorvenous malformations [115]. Other down-regulated genes encoding adhesion molecules were: Jam3, a tight-junction protein essential for maintaining the integrity of the cerebrovascular endothelium [116]; Cdh6, which is expressed in the VZ where it may serve to assign the cortico-striatal boundary [117]; two additional IGSFM genes, Boc and Cdon, that are evident in the VZ, where they function in Sonic Hedgehog signaling and are implicated in forms of holoprosencephaly [118, 119]. Taken together these groupings suggest that expression (and suppressed expression) of numerous adhesion molecules, particularly specific members of the cadherin and IGSF families, are required elements for migration, neurite formation, and subsequent myelination of developing excitatory cortical neurons.