CONTACTINS: STRUCTURAL ASPECTS IN RELATION TO DEVELOPMENTAL FUNCTIONS IN BRAIN DISEASE

Abstract

The contactins are members of a protein subfamily of neural immunoglobulin (Ig) domain-containing cell adhesion molecules. Their architecture is based on six N-terminal Ig domains, four fibronectin type III domains, and a C-terminal glycophosphatidylinositol (GPI)-anchor to the extracellular part of the cell membrane. Genetics of neuropsychiatric disorders, particularly autism spectrum disorders, have pinpointed contactin-4, −5, and −6 (CNTN4, −5, and −6) as potential disease genes in neurodevelopmental disorders and suggested that they participate in pathways important for appropriate brain development. These contactins have distinct but overlapping patterns of brain expression, and null-mutation causes subtle morphological and functional defects in the brain. The molecular basis of their neurodevelopmental functions is likely conferred by heterophilic protein interactions. Cntn4, −5, and −6 interact with protein tyrosine phosphatase receptor gamma (Ptptg) using a shared binding site that spans their second and third Ig repeats. Interactions with amyloid precursor protein (APP), Notch, and other IgCAMs have also been indicated. The present data indicate that Cntn4, −5, and −6 proteins may be part of heteromeric receptor complexes as well as serve as ligands themselves.

I. Introduction

The nervous system functions by virtue of neural networks that inter-connect large varieties of nerve cells in a highly organized and controlled manner. These networks are assembled during development and are under constant adaptation during the whole lifespan. Cell adhesion molecules allow nerve cells, neurons, as well as glial cells to interact. Their key role is evident in neurodevelopmental processes such as migration, axon guidance, axon fasciculation, and synaptogenesis and in plastic processes of the mature brain such as synaptic rearrangements, dendritic dynamics, and regeneration. The repertoire of neural cell adhesion molecules is dominated by several large protein families, one of which is the immunoglobulin (Ig) superfamily of cell adhesion molecules, IgCAMs. These proteins are type I transmembrane proteins, share an architecture built on Ig domains, and are subdivided by the presence of additional conserved protein domains. The best known members of neural IgCAMs are the NCAM and L1-CAM families.

A peculiar family of neural IgCAMs is constituted by a six-member group of IgCAMs that are linked to the cell surface by a glycophosphatidylinositol (GPI)-anchor, the contactins (Shimoda and Watanabe, 2009; Fig. 1). Prototypic for the contactins are contactin-1 (Cntn1, aka F3/contactin) and contactin-2 (Cntn2, aka TAG-1). These two proteins as well as their biological functions in neuron–glia interactions and formation of the nodes of Ranvier have been scrutinized in pivotal studies for over two decades (Salzer et al., 2008). These studies revealed principles of structure and function that directed research into the other members of this family.

Fig. 1.

Primary structure of contactin family members. (A) Phylogenetic analysis of human CNTN proteins. Amino acid sequences were aligned using CLUSTALW as implemented in MEGA5, and the tree was generated using MEGA5 (Tamura et al., 2007). (B) Cartoon representing the domain architecture of CNTN family members along with amino acid identity between individual domains of human CNTN2, −3, −4, −5, and −6 with CNTN1. (C) Same as (B) but showing the amino acid identity of individual domains of CNTN3, −5, and −6 with CNTN4.

However, contactin-3 (Cntn3 aka BIG-1), contactin-4 (Cntn4, aka BIG-2), contactin-5 (Cntn5, aka NB-2), and contactin-6 (Cntn6, aka NB-3) have remained underexposed despite multiple in-depth studies by the groups of Watanabe and Yoshihara. Recently, the genetics of neuropsychiatric neurodevelopmental disorders have encountered several of these members and raised the question how they participate in the pathogenesis of disorders such as autism. To understand their role in developmental disorders of the brain, it will be essential to determine the biological and molecular pathways in which these contactins participate. In this chapter, we provide an overview on biological and structural properties that are required to answer these questions.

II. Genetic Implication of CNTN₄, CNTN₅, and CNTN6 in Neurodevelopment

The technological revolution in genetics has opened areas that were almost inaccessible so far. Particularly, the discovery of copy number variations (CNVs) in the human genome has accelerated the assignment of potential disease genes. In the field of neuropsychiatric genetics, not only have mental retardation and autism spectrum disorder (ASD) been the first neurodevelopmental disorders to pinpoint potential disease genes using CNVs (Christian et al., 2008; Kumar and Christian, 2009; Tarpey et al., 2009; Abrahams and Geschwind, 2010; Pinto et al., 2010), but also in schizophrenia and ADHD disease, genes have been indicated with overlap to autism and mental retardation (Burbach and van der Zwaag, 2009; Burbach, 2010). A shared characteristic of the genetics of these disorders is the overwhelming number of potential disease genes. The most recurrent CNVs in autism encompass at best 1–2% of cases, while many cases contain unique CNVs or “private mutations” (Morrow et al., 2008; Pinto et al., 2010). A converging principle arising from recent genetic data is that potential disease genes are part of a limited set of biological pathways (Levitt and Campbell, 2009; Van der Zwaag et al., 2009). The majority of these pathways can be delineated to cell adhesion and related processes (van der Zwaag et al., 2009; Corvin, 2010; Pinto et al., 2010). Amongst the genes that are regularly encountered in ASD-specific CNVs are CNTN4, CNTN5, and CNTN6.

Disruptions of the CNTN4 gene, located at chromosome (chr.) 3p26.2-p26.3, have been identified in several unrelated cases of ASDs by micro-array-based CNV analysis and resequencing of the gene-locus (Bakkaloglu et al., 2008; Glessner et al., 2009; Roohi et al., 2009; Cottrell et al., 2011). The CNTN4 gene is relatively large (spanning almost 1 Mb of genomic sequence) and is situated on the tip of chromosome 3 in close proximity to close homologue of L1 (CHL1), an L1-CAM member, and a second member of the CNTN gene-family, CNTN6 (chr. 3p26.3). Terminal and interstitial deletions of the tip of chromosome 3 (chr. 3p25-p26) are the genetic cause of 3p deletion syndrome, a serious neurodevelopmental disorder in which, in addition to other genes, these three CAMs have been lost (Fernandez et al., 2008; Gunnarsson and Foyn Bruun, 2010; Pohjola et al., 2010). Deletions overlapping the CNTN6 gene alone have also been identified in several unrelated patients with ASDs (Morrow et al., 2008; van der Zwaag et al., 2009). CNTN5, located at chr.11q22.1 and covering over 1.3 Mb of genomic sequence, also features in candidate-gene lists of CNV studies on ASDs (Burbach and van der Zwaag, 2009), and recently, CNTN3 has been found in a case of autism. CNTN5 and CNTN6 have also been implicated in other neuropsychiatric disorders, like recently in anorexia nervosa (Nakabayashi et al., 2009; Wang et al., 2010).

An intriguing finding has been the relatively frequent disruption of the CNTNAP2 gene in forms of schizophrenia, ASDs, mental retardation, and epilepsy (Alarcon et al., 2008; Arking et al., 2008; Friedman et al., 2008; Poot et al., 2010). The gene is located at chr. 7q35-q36 and is one of the largest genes in the human genome, spanning over 2.3 Mb of genomic sequence. The protein product is also known as Caspr2. It is an important interactor of Cntn2 and forms a functional heterodimeric complex with Cntn2 (Poliak et al., 1999). Cntnap2 forms with four structurally related members of a small protein family (Peles et al., 1997a). Of this family, Cntnap1 (aka Caspr) interacts with Cntn1 (Peles et al., 1997b). However, interactions of the other Cntnaps with Cntn3, −4, −5, or −6 have not been established. The role of CNTNAP2 as an ASD-susceptibility gene together with the disruptions of CNTN4, −5, and −6 in autism has emphasized the possibility that there is a “contactin pathway” underlying neurodevelopmental defects leading to autism and other mental disorders. Although tempting in concept, such a pathway has not been established. It would require resolution of the protein networks in which contactins and associated proteins operate in the developing brain. To start understanding such networks, we need to know more details about the spatiotemporal expression patterns of the genes, about the interactions of the proteins during brain development, and consequences of null-mutation. These are subject of the next sections.

III. Expression of Contactins During Brain Development

The spatial and temporal properties of brain expression of contactins may provide clues to potential functions and pathogenesis. The expression patterns of CNTN4, CNTN5, and CNTN6 in the adult human brain have been examined by Northern blot analysis but are limited in scope (Kamei et al., 2000). It was found that CNTN4 mRNA is most prominently present in the cerebellum, occipital lobe, frontal lobe followed by thalamus, cerebral cortex, and the substantia nigra. CNTN5 mRNA was mostly expressed in the occipital lobe and amygdala, followed by the cerebral cortex, frontal lobe, thalamus, and the temporal lobe. Finally, CNTN6 mRNA follows the expression pattern of the CNTN4 gene at a lower expression level (Kamei et al., 2000). In view of the cognitive defects present in neuropsychiatric disorders, the expression in cortical and other limbic areas may be of importance. Human brain expression is of high relevance to disease, particularly as differences in cortical architecture and gene expression exist between humans and rodents (Abrahams et al., 2007; Hansen et al., 2010; Ip et al., 2010). Such differences exist in gene expression of cell adhesion molecules and CNTNAP2 (Abrahams et al., 2007; Ip et al., 2010). Comparison of mouse expression data in the cortical–hippocampal complex present in the Allen brain atlas (Lein et al., 2007) shows similar patterns for Cntn3 to 6 in cortical layers and hippocampal cell groups with slight differences in staining intensities (Fig. 2). A more detailed examination of spatiotemporal expression of these contactins has been obtained in mice and rat and is described below.

Fig. 2.

Expression of Cntn3, Cntn4, Cntn5, and Cntn6 in the cortical–hippocampal complex of the adult mouse brain. In situ hybridization sections were taken from the Allen brain atlas (Lein et al., 2007).

A. Cntn4 Expression in the Mouse Brain

Cntn4 protein expression is detectable during embryonic development of the mouse brain and into adulthood in the axons of a subpopulation of mouse olfactory sensory neurons (OSNs) in the olfactory epithelium (Saito et al., 1998; Kaneko-Goto et al., 2008). From embryonic stage E14 onward, Cntn4 mRNA was found in olfactory cells which peaked between postnatal day P0 and P7 (Saito et al., 1998). In addition, Cntn4-positive cells also peaked in the olfactory epithelium and in the vomeronasal organ at P0 and at P7 declining thereafter as the epithelium and the vomeronasal organ mature. Cntn4 was found to be critical for projection of the respective axons of OSNs to the few topographically fixed glomeruli on the surface of the olfactory bulb (OB; Kaneko-Goto et al., 2008). Therefore, Cntn4 arranges formation and maintenance of a functional odor map, suggesting a function in axonal guidance.

Several axon guidance molecules are involved in the process of olfactory axon pathfinding from the olfactory epithelium to the OB. The manner of axonal guidance for each individual OSN depends on the single olfactory receptor gene choice and the distinct molecule guiding the axon in the particular area. The glomeruli in the OB showed distinct mosaic protein expression of Cntn4, Obcam, and neuropilin-1, which did not appear to overlap. Cntn4 was expressed in a subset of glomeruli at all anteroposterior levels of the OB. In addition, individual glomeruli expressed different levels of Kirrel2 and ephrin-A5 which partially overlapped with Cntn4 in glomeruli. However, when quantified, only ephrin-A5 seemed to show a correlation with that of Cntn4 (Kaneko-Goto et al., 2008).

It was found that Cntn4 protein expression correlates with several olfactory receptor gene choices by individual OSNs. MOR28 is expressed in two large glomeruli in the ventroposterior region of the OB on the medial and the lateral side. Duo antibody labeling for Cntn4 and MOR28 resulted in specific overlap in the glomeruli, and further quantification and investigation in MOR28-transgenic mice showed that the MOR28-positive OSNs consistently express Cntn4 protein at a high level. For the other three olfactory receptors, it was found that Cntn4 was moderate in glomeruli expressing mOR-EG and low in glomeruli positive for mOR256–17 and OR-17 (Kaneko-Goto et al., 2008).

In addition to the olfactory tract, Cntn4 mRNA was also expressed in the CA1 region of the hippocampus, most nuclei of the thalamus, layers II–V of the cerebral neocortex, layer II of the piriform cortex, most layers of the limbic cortices, several nuclei of the hypothalamus, substantia nigra, inferior colliculus (IC), dorsal motor nucleus of the vagus nerve, and hypoglossal nucleus (Yoshihara et al., 1995; Table I; Fig. 2). Very strong Cntn4 mRNA expression was observed in the granular layer of the anterior folia of the cerebellum but only a weak signal in the posterior folia. In addition, a subset of Purkinje cells in lobules 9 and 10 strongly expressed Cntn4 transcript (Yoshihara et al., 1995). There was partial overlap with the expression of the Cntn5 and Cntn6 genes in rat and mouse (Tables I).

Table I.

Comparison of Expression Patterns of Cntn4, Cntn5, and Cntn6

Brain regions	Specific areas/cells	Cntn4	Cntn5	Cntn6
Cerebral cortex	Layers II–V (especially layer V neurons)	X (Yoshihara et al., 1995)
	Layers II–IV; cingulum		X (Ogawa et al., 2001; Li et al., 2003; Toyoshima et al., 2009a,b)
	Layers II/III and V; deep layer pyramidal neurons			X (Lee et al., 2000; Takeda et al., 2003; Sakurai et al., 2010)
Piriform cortex	Layer II	X (Yoshihara et al., 1995)
Limbic cortex		X (Yoshihara et al., 1995)
Auditory cortex	Neuropil		X (Ogawa et al., 2001; Toyoshima et al., 2009a,b)
Piriform cortex			X (Li et al., 2003)
Visual cortex	Layer V; deep layer pyramidal neurons			X (Ye et al., 2008)
Cerebellum	Granular layer of anterior folia (lobules 1–6); Purkinje cells in lobules 9 and 10	X (Yoshihara et al., 1995)
	External granule cells; Purkinje cells		X (Ogawa et al., 2001; Toyoshima et al., 2009a,b)
	Granule cells and molecular layer of lobule 1 to rostral half of lobule 9; Purkinje cells of the caudal half of lobules 9–10			X (Lee et al., 2000; Takeda et al., 2003)
	Deep cerebellar nuclei			X (Lee et al., 2000)
Olfactory epithelium	Olfactory sensory neurons	X (Yoshihara et al., 1995; Saito et al., 1998)
Olfactory bulb	Glomeruli	X (Yoshihara et al., 1995)
	Olfactory nerve layer	X (Yoshihara et al., 1995)
	Glomerular and mitral cell layers		X (Ogawa et al., 2001)
Accessory olfactory bulb			X (Ogawa et al., 2001; Li et al., 2003)	X (Takeda et al., 2003)
Vomeronasal organ	Vomeronasal neuronal precursors	X (Saito et al., 1998)
Hippocampus	CA1 region	X (Yoshihara et al., 1995)		X (Lee et al., 2000; Sakurai et al., 2010)
	Dentate gyrus		X (Ogawa et al., 2001)	X (Lee et al., 2000; Sakurai et al., 2010)
	Subiculum			X (Lee et al., 2000; Sakurai et al., 2010)
Thalamus	Most nuclei	X (Yoshihara et al., 1995)
	Dorsomedial region (AVDM); ventrolateral region (AAVL)		X (Ogawa et al., 2001; Li et al., 2003; Toyoshima et al., 2009a,b)
	Anterodorsal, ventrolateral, medial, and lateral geniculate nuclei			X (Takeda et al., 2003; Sakurai et al., 2010)
Hypothalamus	Supraoptic nucleus	X (Yoshihara et al., 1995)
	Paraventricular nucleus	X (Yoshihara et al., 1995)		X (Lee et al., 2000)
	Mamillary nuclei	X (Yoshihara et al., 1995)
	Ventromedial nuclei			X (Lee et al., 2000)
Inferior colliculus		X (Yoshihara et al., 1995)		X (Lee et al., 2000; Takeda et al., 2003)
	Neuropil; central nucleus (CIC)		X (Ogawa et al., 2001; Toyoshima et al., 2009a,b)
Medial geniculate nuclei	Neuropil		X (Ogawa et al., 2001; Toyoshima et al., 2009a,b)
Cochlear nuclei	Bushy neurons; glutamatergic synapses; posteroventral (PVCN); anteroventral (avcn)		X (Ogawa et al., 2001; Toyoshima et al., 2009a,b)
Ventral acoustic stria	Axons projecting from bushy neurons		X (Toyoshima et al., 2009a,b)
Superior olivary complex	Glutamatergic synapses; neuropil of medial region (MSO); and lateral region (LSO)		X (Toyoshima et al., 2009a,b)
Medial nucleus of the trapezoid body	Calyces of held; glutamatergic synapses		X (Toyoshima et al., 2009a,b)
Inferior olive ponte nuclei			X (Ogawa et al., 2001; Li et al., 2003)
Lateral lemniscus	Ventral nucleus (VNLL)		X (Toyoshima et al., 2009a,b)
Vagus nerve	Dorsal motor nucleus	X (Yoshihara et al., 1995)		X (Lee et al., 2000)
Hypoglossal nucleus		X (Yoshihara et al.,1995)		X (Lee et al., 2000)
Amygdaloid nucleus			X (Li et al., 2003)
	Basolateral region			X (Lee et al., 2000)
Mesencephalic trigeminal nucleus				X (Lee et al., 2000)
Red nucleus				X (Lee et al., 2000)
Substantia nigra		X (Yoshihara et al., 1995)
Caudate putamen			X (Li et al., 2003)
Locus coeruleus			X (Li et al., 2003)	X (Lee et al., 2000)
Corpus callosum				X (Sakurai et al., 2010)

B. Cntn5 Expression in the Rodent Brain

Cntn5 is expressed transiently during the first postnatal week in glutamatergic neurons of the central auditory system of the rat brain. Cntn5 expression reaches maximum levels at postnatal day 14 in the cerebrum and at postnatal day 3 in the cerebellum and declines thereafter (Ogawa et al., 2001). Simultaneously, during this increase of Cntn5 expression, synapse formation and myelination in the central nervous system are upregulated. In situ hybridization demonstrated that Cntn5 mRNA was highly expressed in regions implicated in the central auditory pathways, such as the cochlear nuclei, superior olivary complex (SOC), ICs, medial geniculate nuclei, and auditory cortex (Ogawa et al., 2001). In addition, Cntn5 immunoreactivity is present in bushy neurons of the ventral cochlear nucleus (VCN) and in the ventral acoustic stria, the glutamatergic presynaptic terminals at the lateral superior olive (LSO), and the calyces of Held in the medial nucleus of the trapezoid body (MNTB) at the finalization of auditory brainstem development (Toyoshima et al., 2009a). Between P1 and P7 Cntn5 was transiently expressed in glutamatergic synapses of the VCN and SOC, during the period of completion of young calyces (Toyoshima et al., 2009a). In addition, Cntn5 protein expression was followed by expression of the vesicular glutamate transporter 1 (Vglut1) in the SOC and thereafter is no longer detectable (Toyoshima et al., 2009b), indicating a possible role for Cntn5 in the initial stage of calyx maturation. Cntn5 immunoreactivity was also high in the dorsal posteroventral and anteroventral VCN regions of the cochlear nuclei and gradually decreased toward the ventral posteroventral and anteroventral VCN (Toyoshima et al., 2009a). These regions are known to be high-frequency tonotopic regions indicating that Cntn5 might be involved in the activity-dependent refinement of auditory neural circuits for tonotopic organization.

Mice in which the Cntn5 gene has been substituted by LacZ, coding for β-galactosidase, have been used to determine Cntn5 gene expression in the brain by X-gal staining (Li et al., 2003). At P7, β-galactosidase was detected in the inferior olive pontine nuclei, thalamic nucleus, accessory OB, piriform cortex, amygdaloid nucleus, caudate putamen, locus coeruleus, and cingulum (Li et al., 2003). However, the expression in these regions decline as the mice reach adulthood.

Expression of Cntn5 protein is particularly associated with nuclei of the auditory system. In the central auditory pathway, acoustic stimuli are generated in the inner ear and are transferred from the spiral ganglion to the cochlear nucleus in the brain stem. Excitatory glutamatergic inputs from the bushy neurons of the VCN assemble onto the ipsilateral and contralateral LSO of the SOC (Kil et al., 1995). The VCN also projects excitatory glutamatergic inputs to the contralateral MNTB, which in turn projects inhibitory inputs to the ipsilateral LSO (Kuwabara et al., 1991). Interaural time delays and differences in sound intensity are evaluated in the LSO by the balance between the excitatory inputs from the VCN and inhibitory inputs from the MNTB (Sanes, 1990; Caspary et al., 2008). Major ascending axons from the VCN innervate both sides of the SOC, which projects to the IC through the lateral lemniscus. The IC innervates the medial geniculate nuclei of the thalamus which sends axons to the auditory cortex.

The rodent auditory system is refined, including synaptic maturation, during the first 3 weeks after birth (Kandler and Friauf, 1993; Kil et al., 1995). Around postnatal day 10 in rodents, the terminals of the globular bushy neurons from VCN that innervate cell bodies of principal neurons in the MNTB transform into the young calyxes of Held (Smith et al., 1998) and mature around P14–16 (Kandler and Friauf, 1993; Kil et al., 1995). These structures have been described as the largest synapses in the brain and function to transfer action potentials from globular bushy cells in the VCN to the MNTB principle cells (Kil et al., 1995).

C. Cntn6 Expression in the Mouse Brain

Cntn6 immunoreactivity is only present in neurons of the nervous system and mainly upregulated at the early postnatal stage during mouse brain development (Cui et al., 2004). The expression in the cerebellum increases until P7, where after it dramatically decreases (Lee et al., 2000; Takeda et al., 2003). Cntn6 protein expression reached a maximum at P15 and declined to a constant level in adulthood (Sakurai et al., 2009), suggesting that Cntn6 plays a role in postnatal cerebellar development.

Using X-gal staining on a LacZ knock-in in the Cntn6 gene revealed the brain regions that highly express Cntn6 transcripts: the accessory OB, anterodorsal thalamic nuclei, ICs, layer 5 of the cerebral cortex, and cerebellum (Takeda et al., 2003). The latter was also observed by in situ hybridization for Cntn6 (21). In addition, by in situ hybridization, Cntn6 mRNA expression was also found at the piriform cortex, the hippocampus, hypothalamus, the amygdala, the red nucleus, the pons, inferior olive, and several other nuclei (Lee et al., 2000). At around E17, deep layer pyramidal neurons in the neocortex begin to extend axonal and dendritic processes in which X-gal signal is observed in the deeper layer of the caudal cortex. High-caudal to low-rostral pattern of X-gal signal in the cortex was maintained at P7 (Ye et al., 2008). Immunofluorescent staining of Cntn6 in wild-type mice at P7 shows the strongest signal in the visual cortex, with many deep layer pyramidal neurons stained (Ye et al., 2008).

Cntn6 protein was highly expressed in subpopulations of granule cells and in the molecular layer of lobule 1 to the rostral half of lobule 9 in the cerebellum, whereas the expression in the molecular layer was weak in the rostral region of lobules 9 and 10 (Takeda et al., 2003). However, in the caudal region of lubules 9 and 10, Cntn6 protein is present in the dendrites and somata of Purkinje cells, in contrast to lobules 1–8 (Takeda et al., 2003). During the development of the cerebellum, the Cntn6 gene is first expressed in the Purkinje cells of lobules 9 and 10 and was followed by expression in the internal granule cells of all lobules during cerebellar development, thereby presenting differential expression of Cntn6 over the lobules of the cerebellum.

At P5 and thereafter, in lobule 3 of the cerebellum Cntn6 immunoreactivity was observed in the developing molecular layer and granule cell layer but not in Purkinje cells. During postnatal development, Cntn6 was found along the dendritic branches in the molecular layer and the roots of the stem dendrites of Purkinje cells at P15 (Sakurai et al., 2009). Cntn6 colocalizes with mGluR1a and Vglut1-immunoreactive puncta in the molecular layer of P10 mice, indicating glutamatergic synapses between parallel fibers and Purkinje cells. Cntn6 did not overlap with vesicular glutamate transporter 2 (Vglut2)-immunoreactive puncta in the presynaptic terminals of parallel fibers in the deep molecular layer of P21 mice (Sakurai et al., 2009), which is normally expressed in immature synapses between parallel and climbing fibers and Purkinje cells (Miyazaki et al., 2003). In addition, with Western blotting, Cntn6 protein was only found in the synaptosome fraction. In short, Cntn6 has an important function at the presynaptic termini of parallel fibers, but not in those of climbing fibers, that form synapses with Purkinje cells.

However, Cntn6 immunoreactivity is not only present in the cerebellum and implicated in synapse formation between parallel fibers and Purkinje cells but is also observed in the synapses of the parallel fibers in the hippocampal formation (Sakurai et al., 2010). At P5, Cntn6 immunoreactivity was detected in the subiculum, the stratum lacunosum–moleculare of the CA1 region, in the hilus of the dentate gyrus, and very weakly in the other structures of the hippocampal formation (Sakurai et al., 2010). Cntn6 immunoreactive puncta overlapped with those of Vglut1 and Vglut2 in the subiculum and in the stratum lacunosum–moleculare of CA1 but not with the inhibitory presynaptic marker Vgat (vesicular GABA transporter). In Cntn6-deficient mice, the density of Vglut1- and Vglut2-positive puncta was reduced by 20–30% in the regions where Cntn6 protein is strongly expressed in wild-type mice. However, the Vgat puncta were not affected by Cntn6 deficiency (Sakurai et al., 2010), indicating Cntn6 in the role of glutamatergic but not GABAergic, synapse formation during postnatal development in the hippocampus as well as the cerebellum.

IV. Phenotypes in Contactin Null-Mutants

A. Cntn4 Knockout Mouse Phenotypes

Cntn4-deficient mice have been generated by disrupting exon 2 that contains the translation initiation site and encodes the N-terminal signal peptide of the Cntn4 protein (Kaneko-Goto et al., 2008). In these mice, the gross anatomy and layer organization of the olfactory epithelium appear normal. Expression levels and patterns of other cell surface molecules are not changed, including CAMs present in individual OSN axons. In wild-type mice, MOR28-positive OSNs innervate mostly two large glomeruli on the medial and lateral side of the OB and not so much ectopic glomeruli. In contrast, in the Cntn4-deficient mice, the innervations of ectopic glomeruli were significantly larger (Kaneko-Goto et al., 2008). Other olfactory receptors such as OR-17 and mOR-EG also show highly increased amount of innervations of ectopic glomeruli in Cntn4-deficient mice compared to wild-type mice. The most reasonable assumption of Cntn4 function at this point is an attractive role in convergence and glomerular targeting of OSN axons expressing individual olfactory receptors (Kaneko-Goto et al., 2008).

B. Cntn5 Knockout Mouse Phenotypes

Cntn5-deficient mice have been developed by substituting the translation initiation codon-containing exon of Cntn5 gene by a tau-LacZ gene cassette to produce tau-β-galactosidase protein in place of the Cntn5 protein (Li et al., 2003). Mice deficient for Cntn5 showed little abnormality in the gross brain architecture. Cntn5 protein expression in the auditory pathways in mice was confirmed by the expression pattern of tau-LacZ reporter gene detected by X-gal staining in mice heterozygous for Cntn5. Specifically, Cntn5 protein expression was most distinct in the VCN, SOC, lateral lemniscus, and IC, which is maintained through adulthood. These data correspond to earlier findings in the rat (Ogawa et al., 2001). In wild-type mice at P6, all principal neurons were encircled by calyces of Held in the MNTB, and conversely, in the Cntn5-deficient mice, these principal neurons were scattered without the support of calyces. The calyces that do encircle principal neurons were immature in the absence of Cntn5 (Toyoshima et al., 2009a, b), indicating that a subset of principal neurons may not receive the inputs in Cntn5-deficient mice. Apoptotic activity in the bushy neurons of VCN and in the principal neurons without mature innervations of the MNTB was detected in the Cntn5-deficient mice from P10 to P15 (Toyoshima et al., 2009a,b). In addition, at the same developmental stage, a significant decrease of glutamatergic signal in the LSO was detected.

Wild-type mice demonstrated typical audiogenic seizures in response to acoustic stimuli, but Cntn5-deficient mice were significantly less sensitive to this stimulus (Li et al., 2003). Audiogenic seizures are reflexed by site-specific c-Fos expression which marks enhanced neural excitability (Ishida et al., 2002). Audiogenic seizure induction in wild-type mice showed increased c-Fos expression specifically in the IC, predominantly in the external cortical regions of the IC and dorsal regions of the IC but not in the central nucleus of the IC. In contrast, Cntn5-deficient mice showed profound decrease of c-Fos expression in the external and dorsal cortical regions of the IC. The central nucleus of the IC along with the cochlear nucleus displayed no significant difference in c-Fos expression between the wild-type and Cntn5-deficient animals. The lower induction of neural excitability in the IC of Cntn5-deficient mice was in concordance with the altered behavioral response of these animals to sound stimuli (Li et al., 2003). In addition, in response to pure-tone stimulation after priming, Cntn5-deficient mice demonstrated a scattered and decreased c-Fos expression in the central nucleus of the IC, which is in contrast to the band-like c-Fos expression in the central nucleus of the IC of the wild-type littermates (Li et al., 2003). In addition, Cntn5-deficient mice exhibited a significant increase in the interpeak latencies of auditory brainstem response waves II–III and III–IV (Toyoshima et al., 2009a,b). Auditory brain response waves II–III are responses obtained from spherical and globular busy neurons of the VCN and subsequent targets (Melcher et al., 1996; Melcher and Kiang, 1996), and waves III–IV are combined responses from other regions of the auditory pathway.

In conclusion, the analyses of audiogenic processing in Cntn5 knockout mice suggest that in the absence of Cntn5 globular bushy neurons fail to form or maintain glutamatergic synapses in the LSO and MNTB and subsequently undergo apoptosis. This causes an increase of the interpeak latencies for the auditory brainstem waves. Subsequently, the failure of synapse formation in the MNTB and LSO is able to cause imbalance of integration of binaural sensory information at the LSO. This suggests that Cntn5 is necessary in the final stage of development for tuning neuronal activity in the auditory system. These data suggest that patients that are heterozygous for CNTN5 due to a CNV or coding mutation may be affected in hearing.

C. Cntn6 Knockout Phenotypes

Cntn6-mutant mice were generated by homologous recombination in embryonic stem cells by replacing the exon 2 having the translation initiation site of Cntn6 with LacZ and Neo genes (Takeda et al., 2003). The cytoarchitecture of the brain was not significantly different between Cntn6-deficient mice and wild-type mice at the lightmicroscopic level. Wild-type mice demonstrated high expression of the Cntn6 protein in the Purkinje cells of lobules 9 and 10 (Takeda et al., 2003), which are termed the vestibulocerebellum, which projects to the vestibular nucleus and is important for the control of axial and proximal limb muscles to maintain balance, the control of eye movements, and coordination of the movements of the head and eyes.

Several motor coordination deficits were shown in Cntn6-deficient mice. By testing the ability of the mice to traverse a stationary horizontal rod, psychomotor coordination and the integrity of the vestibular system were evaluated. In the rotorod and horizontal rod test, the Cntn6-deficient mice were inferior to the wild-type mice in their ability to walk on the rods (Takeda et al., 2003) indicating that Cntn6-deficient mice have poor ability to improve their sensorimotor skills upon repeated trials. This demonstrated the relevance of Cntn6 in lobules 9 and 10 of the cerebellum. However, muscle strength was normal in the Cntn6-deficient mice compared to control.

There were no significant differences in the short-term plasticity of either climbing fiber or parallel fiber synapses between Cntn6-deficient and wild-type mice indicating that Cntn6 is not involved in the excitatory synaptic transmission to Purkinje cells. However, output from Purkinje cells to the vestibular nucleus may be impaired without influencing climbing or parallel fiber EPSCs (Takeda et al., 2003). Another reason for the motor impairments could be mossy fiber innervation, which posses a significant amount of Cntn6 transcripts (Lee et al., 2000), so that lack of Cntn6 might affect the excitatory synaptic transmission to either granule cells or Golgi cells.

In the cerebellum, Cntn6 immunoreactivity was observed in a zone underneath the deep EGL and did not overlap with the Cntn2-immunoreactive zone in the deep EGL, in contrast to L1 immunoreactivity which overlapped with both Cntn6 and Cntn2 zones simultaneously (Sakurai et al., 2009). L1 protein is expressed by postmitotic premigratory granule cells migrating radially and extending axons horizontally and on parallel fibers which disappear at the late stage of cerebellar development (Persohn and Schachner, 1987). In Cntn6-deficient mice, there was an increase in the total area of L1 immunoreactivity in the IGL compared to wild-type mice at P5, and there was a significant decrease of cells, indicating delayed development or increased migration of granule cells in the IGL.

In Cntn6-deficient mice at P5, Vglut1 puncta were scattered around the somata of Purkinje cells, and during development until P15, they did not reach the outer edges of dendrites of the Purkinje cells, in contrast to the wild-type mice which displayed Vglut1 puncta from the cell bodies to the outermost edges of dendrites of Purkinje cells (Sakurai et al., 2009). In Cntn6-deficient mice, the Vglut1-positive zone was thinner at the tips of the dendrites, and the density of the puncta was reduced. However, this difference in Vglut1 expression was largely annihilated in adult mice. The mGluR1a-positive zone was similar to the Vglut1-positive zone, reaching from the upper limit of Purkinje cell bodies to the outer edge of dendrites. The thickness of Vglut2 and mGluR1a-positive zones did not differ between Cntn6-deficient and wild-type mice (Sakurai et al., 2009). However, the density of the mGluR1a-positive puncta in Cntn6-deficient mice was reduced by 18% compared to wild-type mice.

In addition, the number of caspase-3 positive cells in the IGL was increased by 60% in the Cntn6-deficient cerebellum when compared to wild type (Sakurai et al., 2009). Thus, during cerebellar development, Cntn6 contributes to synapse formation between parallel fibers and Purkinje cells. Cntn6 deficiency causes reduction in synapse density between parallel fibers and Purkinje cells and also increases granule cell death during cerebellar development. The reduction of synapse formation might be related to the increase of the immature granule cells in the IGL, detected by L1 expression in Cntn6-deficient mice.

The Cntn6 gene is predominantly expressed in layer V of the cortex in the soma and processes of cortical neurons including dendrites and axons (Lee et al., 2000). The distribution of layer V pyramidal neurons in the Cntn6-deficient mice was indistinguishable from that of wild-type littermates, except for layer V in the visual cortex. In this area, the apical dendrites were misoriented, although they did reach layer I and formed apical tufts (Ye et al., 2008). Mice heterozygous for both Cntn6 and Chl1 showed reduced Cntn6 and CHL1 protein levels compared with those of wild-type littermates. Layer V pyramidal neurons in the visual cortex of the compound heterozygous mice showed a more severe misoriented dendrite phenotype than the single-heterozygous mice and their wild-type littermates (Ye et al., 2008). These data indicate that Cntn6 and CHL1 may cooperate in generating this morphological phenotype in the cortex and suggest that functional molecular interactions may exist between these two related proteins.

V. Structural Architecture of Contactins

A. General Domain Structure, Comparison of Domains Between Members

The six members of the contactin family share a common domain organization reminiscent of that of members of the L1 family of neural cell adhesion molecules (Maness and Schachner, 2007). Indeed, contactins include six N-terminal Ig repeats followed by four fibronectin type III (FNIII) repeats (Fig. 1; Gennarini et al., 1989). In contrast to L1, contactins do not possess transmembrane and intracellular regions and are instead tethered to the cell membrane with a GPI anchor. Overall, contactins share 40–60% identity at the amino acid level and cluster essentially in two groups, with CNTN1 and CNTN2 in one group and CNTN3 to −6 in the other group (Fig. 1A). Comparison of individual domains of CNTN3, −5, and −6 to CNTN4 shows high amino acid sequence identity (~70% or better) for the second and third Ig domains where the binding site for a common binding partner is located (see below), whereas the fourth and fifth Ig domains are the most divergent repeats (Fig. 1C).

Transcript variants of the CNTN4 and CNTN6 genes have been described that are generated by alternative splicing. They predict the existence of isoforms of these proteins (Fig. 3). In mouse, these Cntn4 isoforms are Cntn4–1 and Cntn4–2. The latter is over 300 amino acids shorter and truncated at the C-terminus (Fig. 3A). This isoform lacks the three C-terminal FNIII domains as well as the lipid attachment site. Transcripts of this isoform are detectable in the embryonic mouse brain (A. Zuko and J.P.H. Burbach, unpublished) and have been detected in the human brain (Zeng et al., 2002). This isoform may represent a secreted form of Cntn4 with fully functional Ig domains possibly acting as ligand. The Cntn5–2 isoform encodes a protein lacking the N-terminal Ig domain 1 but having a potential signal peptide (Fig. 3B). The transcript is expressed in the embryonic mouse brain (A. Zuko and J.P.H. Burbach, unpublished), but its biological potential is uncertain in view of the role of the Ig domains 1–4 in protein–protein interactions (see below).

Fig. 3.

Isoforms of Cntn4 and Cntn5. (A) Mouse Cntn4 variants 1 and 2 (Cntn4–1 and Cntn4–2) differ in the C-terminal portion due to alternative splicing of the Cntn4 gene. Variant Cntn4–2 is translated from an alternative transcript with an early stop codon. It lacks the three C-terminal FNIII domains and has no probable GPI-modification site as predicted by big-PI-predictor (http://mendel.imp.ac.at/gpi/). Cntn4–2 may be a secreted form. (B) Cntn5–2 is translated from an initiation site on an alternative transcript. The predicted protein has a different N-terminal sequence that includes a signal peptide and lacks the first Ig domain.

B. Properties of Individual Subdomains

Insights into the properties of individual subdomains of contactins have mostly been focused on their four N-terminal Ig repeats. Structural analyses of these domains in human CNTN2 and mouse Cntn4 have shown that they adopt a U-shaped conformation often referred to as a “horseshoe” (Fig. 4; Mortl et al., 2007; Bouyain and Watkins, 2010). In these crystal structures, domains Ig1–Ig2 on one hand and Ig3–Ig4 on the other hand are arranged in an antiparallel fashion. This conformation is made possible because of the presence of an eight amino acid linker region between Ig2 and Ig3, whereas the polypeptide regions between Ig1 and Ig2 and between Ig3 and Ig4 are fairly short. Extensive contacts (>2000Ų) between domains Ig1 and Ig4 and between domains Ig2 and Ig3 stabilize this structure, and it is unlikely that a more linear, open conformation is observed in solution. In addition, analysis of residue conservation at the horseshoe interface indicates that all contactin family members are likely to include this structural motif (Bouyain and Watkins, 2010).

Fig. 4.

Crystal structure of the horseshoe-like motif in the N-terminal domain of mouse Cntn4. A cartoon representing the domain organization of Cntn4 is shown on the left, along with a ribbon diagram in the middle and a surface representation on the right. The letters N and C indicate the N- and C-termini, respectively. Disulfide bonds are shown as orange ball-and-stick models. Asparagine-linked N-acetylglucosamine residues are depicted as gray ball-and-stick models along with the asparagine side chain. Ig domains 1, 2, 3, and 4 are colored cyan, green, gold, and red, respectively. The horseshoe-like structure of Ig domains 1–4 of mouse Cntn4 is closely related to the one adopted by the first four Ig domains of chicken and human Cntn2 and superimpose with rmsd values of 1.6–2.3Å (Freigang et al., 2000; Mortl et al., 2007; Bouyain and Watkins, 2010). Structural images were generated using PyMOL (www.pymol.org).

The horseshoe conformation of Ig domains 1–4 is an important feature of the contactin family. Indeed, this motif mediates heterophilic interactions with several of the known contactin-binding partners: the receptor protein tyrosine phosphatases (RPTPs), Ptprz, and Ptprg for Cntn1, −3, −4, −5, and −6 (see below), and the neural receptors L1 and NrCAM for Cntn2 (Buchstaller et al., 1996; Fitzli et al., 2000). In addition, the horseshoe regions of Cntn2 molecules expressed on opposing cells associate to mediate homophilic cell adhesion (Felsenfeld et al., 1994). However, the structural basis for these interactions remains unclear, as two different homodimerization interfaces have been identified in the chicken and human proteins, respectively (Freigang et al., 2000; Mortl et al., 2007; He et al., 2009). The presence of the horseshoe motif further underlines the resemblance between the contactin and L1 families of neural cell adhesion molecules. Among the four proteins in the L1 family (L1, Chl1, NrCAM, and neurofascin), structural analyses have shown that neurofascin and presumably L1 adopt the antiparallel arrangement observed for CNTN2 and CNTN4, and analysis of residue conservation indicates that Chl1 and NrCAM most likely fold in a similar fashion (He et al., 2009; Liu et al., 2011). Interestingly, L1 family members mediate homophilic cell adhesion using binding interfaces found in the horseshoe motifs (Liu et al., 2011) but have also been shown to bind to several contactin family members (Felsenfeld et al., 1994; Buchstaller et al., 1996; Ye et al., 2008).

In contrast to the Ig domains, little is known about the contribution of the FNIII repeats to the biological functions of contactins. The picture that emerges from work conducted on the chicken Cntn2 homologue axonin is that the FNIII repeats may contribute to the organization or clustering of contactins on the cell surface. Indeed, Kunz et al. (2002) have shown that monoclonal antibodies directed to the fourth FNIII repeat of axonin impaired its homophilic binding properties. The authors concluded that axonin may form cis-oligomers at the cell surface, and that these oligomers would promote the formation of homophilic contacts between axonin-expressing cells. Interestingly, these findings would mirror some of the recent results obtained about classical cadherins (Wu et al., 2010). It is unclear whether the formation of cis-oligomers would also take place for the other members of the family, especially when considering the fact that the sequence identity between the fourth FNIII repeat of Cntn2 and the other contactins is 45% for Cntn1 but falls to 31– 36% for Cntn3 to −6. Nevertheless, it is important to note that contactins may still cluster on the cell surface by mechanisms that do not involve the FNIII repeats. Indeed, GPI-anchored receptors are often found clustered in lipid rafts, which would suggest that contactins could be overrepresented in these microdomains at the cell surface (Harris and Siu, 2002).

VI. Protein–Protein Interactions Mediated by Contactins

A. Interactions with Protein Tyrosine Phosphatases

Physiological ligands for contactins were sought soon after their discovery, a search that has only intensified as CNTN genes is now linked to the pathology of developmental disorders such as schizophrenia and autism (Cottrell et al., 2010; Corvin, 2010). The interactions mediated by Cntn1 and Cntn2 have been described in detail in another review (Shimoda and Watanabe, 2009), and we focus here on the binding partners for Cntn4, −5, and −6. Early on, it was determined that Cntn2 is a homophilic binding molecule in contrast to Cntn1, which has multiple heterophilic binding partners and in particular associates with the tyrosine phosphatase Ptprz (Peles et al., 1995; Sakurai et al., 1997). These findings were of particular interest because the identities of heterophilic binding partners for RPTPs had remained, and still remain, elusive (Johnson and Van Vactor, 2003; Stoker, 2005). Ptprz is a type I transmembrane protein expressed almost exclusively in glial cells that includes a catalytically inactive carbonic anhydrase-like (CA-like) domain that is solely responsible for the interactions with Cntn1 (Peles et al., 1995), a single FNIII domain, a heavily glycosylated spacer region, and two intracellular tyrosine phosphatase domains (Krueger and Saito, 1992). Based on the known interaction between Ptprz and Cntn1, the resemblance between contactin family members and the existence of the Ptprz homologue, called Ptprg, expressed on neurons (Barnea et al., 1993; Lamprianou et al., 2006), it was speculated that contactins could associate with Prprz and Ptprg (Bouyain and Watkins, 2010). In vitro binding assays were used to demonstrate that although mouse Ptprz binds only to Cntn1, Cntn3, −4, −5, and −6, all associate with mouse Ptprg. Further, these interactions are mediated solely by the CA-like domain of Ptprg as is the case for the interactions between Ptprg and Cntn1 (Peles et al., 1995; Bouyain and Watkins, 2010). Although interesting, these findings fall short of demonstrating that Ptprg is a bona fide physiological ligand for Cntn3, −4, −5, and −6 and do not provide any indication as to what the biological functions of interactions between Ptprg and contactin family members may be. In fact, the exact role that interactions between Ptprz and Cntn1 may play in neurogenesis remains unclear.

The recent crystal structure of a complex between the four N-terminal Ig repeats of mouse Cntn4 and the CA-like domain of PTPRG has provided the first structural insights for the interactions between contactin family members and RPTPs (Bouyain and Watkins, 2010). The antiparallel arrangement of Ig domains 2 and 3 in Cntn4 (Fig. 5) creates a contiguous, flat surface in which the PTPRG-binding site is included. This configuration of the binding interface indicates that the horseshoe-like conformation adopted by Cntn4 is critical to its binding interactions with PTPRG. This binding site comes “preformed” because the structure of Ig domains 1–4 of Cntn4 determined in the absence of PTPRG matches closely the structure of the same region when in complex with PTPRG. The complex interface is extensive (~1700Ų) and highly complementary. In broad terms, the PTPRG-binding site in Cntn4 spans both Ig domains 2 and 3. It involves two segments that are found in these repeats: residues 129–142 in Ig2 and residues 220–228 in Ig3 (Fig. 5). Residues in these two segments make van der Waals, hydrogen bond, and electrostatic interactions with residues found in a 14-amino acids beta hairpin loop and a 4-amino acids loop in the CA-like domain of PTPRG with residues in the beta hairpin loop contacting both Cntn4 Ig domains 2 and 3, whereas residues in the short loop interact solely with Ig domain 3.

Fig. 5.

Structural insights into Ptprg–Cntn4 interactions. (A) The Ptprg–Cntn4 complex in two view related by a 30° rotation along a vertical axis. Mouse Cntn4 is shown in surface representation according to the color coding introduced in Fig. 4. The CA domain of mouse Ptprg is shown using a ribbontprg diagram and is colored magenta. In the second view, only the binding regions of Ptprg are shown and they consist of two loop regions. (B) The Ptprg-binding site on Cntn4 spans Ig domains 2 and 3. Ig domains 1, 2, 3, and 4 are colored cyan, green, gold, and red, respectively. Amino acids in Cntn4 that are in contact with amino acids in Ptprg are colored magenta. (C) Conservation of amino acids at the interface between Ptprg and Cntn4. Black lines denote van der Waals interactions, whereas red lines indicate potential hydrogen bonds and salt bridges. Amino acids in the green and gold boxes are located in Ig2 and Ig3 of Cntn4, respectively.

Analysis of the binding site and the conservation of residues have made it possible to rationalize the shared binding interactions observed between Cntn3, −4, −5, and −6 and PTPRG (Bouyain and Watkins, 2010). Indeed, all the amino acids found at the interface between Cntn4 and PTPRG are conserved in Cntn3, −5, and −6 (Fig. 5). Given the fact that it is likely that Cntn3, −5, and −6 adopt a horseshoe-like conformation similar, if not identical, to the one adopted by Cntn4, it is plausible that PTPRG binds to Cntn3, −5, and −6 in much the same way it binds to Cntn4. However, in this case, the biological roles that these four distinct proteins with seemingly identical binding sites for PTPRG would play in neurogenesis are unclear. One could speculate that distinct contactin family members associate with distinct cell surface molecules so that a complex between PTPRG and a contactin molecule in fact involves the formation of higher-order molecular complexes to form active signaling units. Candidate cell surface receptors that associate with contactins might include (a) other contactin molecules to form homo- or heterodimers; (b) members of the L1 family of cell adhesion molecules (see below for insights into Cntn6 and Chl1); or (c) one of the contactin-associated proteins. One clear advantage that would be conferred by the formation of these complexes at the cell surface is the possibility that engagement of a contactin by PTPRG could lead to bidirectional signaling mediated by the tyrosine phosphatase domains of PTPRG in one cell and signaling proteins that would associate with the intracellular region of the putative contactin coreceptor, thus alleviating the absence of an intracellular region in contactins.

B. Interactions of Cntn4 with APP

Kaneko-Goto et al. (2008) have demonstrated the presence of a heterophilic Cntn4 binding partner on olfactory axons. Cntn4-alkaline phosphatase (AP) recombinant fusion protein consisting of mouse Cntn4 extracellular region and human placental AP was used to perform an overlay assay on OB sections from adult wild type and Cntn4-deficient mice. Cntn4-AP strongly bound to the nerve layer and glomerular layer in OB of Cntn4-deficient mice in a mosaic fashion, in contrast to the very weak binding in wild-type mice (Kaneko-Goto et al., 2008). This suggests that the Cntn4 binding partner is occupied in wild type by endogenous Cntn4 protein, which can be unveiled in Cntn4-deficient mice by exogenous Cntn4.

By investigating potential binding partners for the transmembrane protein amyloid precursor protein (APP), Osterfield et al. (2008) found Cntn3 and Cntn4 to bind to APP. APP plays a key role in Alzheimer’s disease, which is characterized by intraneuronal tangles and extracellular plaques consisting of precipitates of amyloid beta-peptides (Aβ), derived from APP after cleavage by β- and γ-secretases. The fusion protein AP–APPsα, the cleaved version of APP and a possible ligand, in chicken was found to be expressed in the OB and to RGC axons of the tectum. The AP–APPsα protein existing only of a middle domain (amino acids 199–345) gave strong RGC binding but little binding in the OB. By contrast, the N-terminal domain of AP–APPsα (amino acids 18–205) showed very strong binding to the OB indicating more than one binding partner for AP–APPsα. In addition, treating the brains with PI-PLC, which cleaves GPI-links, greatly reduced AP–APPsα binding to the OB, whereas binding to tecta appeared less affected (Osterfield et al., 2008).

Coimmunoprecipitation analysis showed that APP has a high affinity interaction with Cntn3 and Cntn4 (Osterfield et al., 2008). After testing of all contactins for affinity to amyloid precursor-like protein 1, it was found that amyloid precursor-like protein 1 binds Cntn3, Cntn4, and Cntn5. In analyzing deletion constructs of Cntn3 and Cntn4, it was found that the FNIII domains are sufficient for binding to the 18–205 amino acid region of APP (Osterfield et al., 2008). The non-GPI-anchored binding partner is NgCAM which coprecipitates with the 199–345 amino acid region of APP. NgCAM is the chick homologue of L1CAM, which has been widely studied for functions in axon growth, guidance, and fasciculation (Maness and Schachner, 2007). Coexpression of Cntn4 or NgCAM with APP in transfected cells demonstrated expression of full-length APP and subsequently an increase of CTFα (the C-terminal fragment α of APP), which may influence further downstream pathways. These results suggest Cntn4 to be implicated with regulation of the APP α-cleavage site.

APP, Cntn3, Cntn4, and NgCAM are all expressed in the RGC layer of the retina and in the tectum in multiple layers, showing that these molecules are suitably placed for interactions that participate in retinal axon development. Retinal explants cultured on NgCAM substrate showed increased RGC axonal outgrowth enhanced by AP–APPsα or by the 18–205 amino acid region of AP–APP but not on laminin or Cntn4 as substrate. As the N-terminal domain of APP, which binds to Cntn3 and Cntn4, is sufficient to increase NgCAM-dependent outgrowth, the role that Cntn4 plays in this process was analyzed further. Soluble Cntn4 reduced levels of axonal outgrowth on AP–APPsα plus NgCAM and on NgCAM alone but showed no effect on laminin-dependent growth of RGC axons. In addition, explants infected with virus expressing shRNA that targets Cntn4 exhibited less axon growth than control explants when grown on AP–APPsα plus NgCAM, but not on infected explants grown on laminin, demonstrating specific functional interactions (Osterfield et al., 2008).

C. Interactions of Cntn6

Both the substrate-bound and soluble forms of Cntn6 have a neurite outgrowth promoting effect on cortical neurons (Lee et al., 2000) as well as regulating the apical dendrite orientation in the neocortex (Ye et al., 2008). Chl1 is a cell surface molecule that is able to regulate apical dendrite orientation of pyramidal neurons in the neocortex. In the developing mouse cortex, Chl1 is expressed in deep layer pyramidal neurons in a low-rostral to high-caudal gradient. Chl1-deficient mice exhibit misoriented apical dendrites of pyramidal neurons (Demyanenko et al., 2004). As Cntn6 demonstrates a similar expression in the cortex and is located next to CHL1 on the human genome, it was postulated that Cntn6 might also contribute to this defect. In addition, animals with protein tyrosine phosphatase α (Ptpra) deficiency showed abnormal apical dendrites oriented sideways or inverted in the caudal (visual), motor, somatosensory cortices (Ye et al., 2008). Ptpra most likely acts downstream of Cntn6 and Chl1.

Ptpra is a receptor-like protein phosphatase and mediates signaling to the intracellular tyrosine kinase p59fyn. The kinase activity of p59fyn is inhibited through intramolecular interaction between phosphorylated Tyr-531 and its SH2 domain, which stabilizes a noncatalytic conformation. Ptpra activates p59fyn via dephosphorylation of the Tyr-531 site (Bhandari et al., 1998). Mice lacking p59fyn demonstrated inversion of apical dendrites of cortical pyramidal neurons (Sasaki et al., 2002), suggesting a possible interaction of Ptpra with Chl1 and Cntn6 in dendrite development in neocortex.

P7 mouse membrane fractions immunoprecipitated were probed for the presence of Chl1 or L1. Chl1 was detected in the precipitates using a Cntn6 antibody but not L1 (Ye et al., 2008). Colocalization of Cntn6 and Chl1 was also observed in the soma and neurites of cultured cortical neurons. Cntn6 microspheres do not bind to Chl1 expressing cells, suggesting that Cntn6 and Chl1 do not function as a ligand for each other (Ye et al., 2008) but may engage in a cis-interaction.

It has been observed that cells expressing only Cntn6 have most Cntn6 protein localized inside the cells and display a very faint Cntn6 cell surface expression (Ye et al., 2008). When cotransfecting cells both with Cntn6 and Chl1 cDNAs, there was an increase in the amount of Cntn6, which moved to the cell periphery and colocalized with cell surface CHL1. In contrast, cotransfection of Cntn6 and Chl1 cDNAs did not change the cell surface level of Chl1 in comparison to single-transfected cells (Ye et al., 2008), possibly explaining why Chl1-deficient mice show a more severe phenotype than the Cntn6-deficient mice.

Coimmunoprecipitation of Prpra together with Chli was observed transfected HEK293T cells, as well as coimmunoprecipitation of Ptpra together with Cntn6, suggesting that Chl1 and Cntn6 are able to signal independently to Ptpra (Ye et al., 2008). Clustering of either Chl1 or Cntn6 independently leads to Ptpra activation and p59fyn dephosphorylation (Ye et al., 2008). These results suggest that the impaired dephosphorylation of p59fyn in the Chl1-and Cntn6-deficient brains is due to the lack of upstream stimulation by Chl1 and Cntn6 of Prpra-signaling to p59fyn, indicating Ptpra as a signal mediator for Chl1 and Cntn6 in the apical dendrite development.

Cntn6 protein is expressed at maximal levels between P7 and P21, corresponding to the time window of oligodendrogliogenesis from progenitor cells and oligodendrocyte maturation (Cui et al., 2004). Cntn6 and Notch1 can be reciprocally coimmunoprecipitated from rat brain membrane extracts, indicating Cntn6 to be a binding ligand of Notch. More specifically, Cntn6 binds to the Notch1 EGF 22–34 repeat region and induces the generation and nuclear translocation of the Notch intracellular domain (NICD) by Notch proteolysis of γ-secretase at the S3 site (Cui et al., 2004). This subsequently initiates promotion of oligodendrogliogenesis and differentiation of neural progenitor cells and oligodendrocyte progenitor cells into oligodendrocytes upon activation of the NICD/Deltex1 signaling pathway (Cui et al., 2004).

In Ptpra-deficient mice, hypermyelination was observed after Ptpra depletion. In Cntn6-deficient mice, neuronal precursor cells gave rise to significantly less oligodendrocytes but more neurons as compared to wild-type littermates (Hu et al., 2006), confirming that Cntn6 promotes oligodendrogliogenesis and in this case via Notch/Deltex1. Interactions have been summarized in Table II.

Table II.

Proteins with Stablized Interactions With Cntn4, Cntn5, and Cntn6

Contactin gene	Interactor	Specifics
Cntn4	Amyloid precursor protein (Osterfield et al., 2008)	Transmembrane protein
	APPsα (Osterfield et al., 2008)	Cleaved ectodomain of APP
	APLP1 (Osterfield et al., 2008)	Amyloid precursor-like protein 1
	PTPRγ (Bouyain and Watkins, 2010)	Protein tyrosine phosphatase γ
Cntn5	APLP1 (Osterfield et al., 2008)	Amyloid precursor-like protein 1
	PTPRγ (Bouyain and Watkins, 2010)	Protein tyrosine phosphatase γ
Cntn6	CHL1 (Ye et al., 2008)	Close homologue of L1; cis-interaction
	PTPRα (Ye et al., 2008)	Protein tyrosine phosphatase α
	Notch (Cui et al., 2004; Hu et al., 2006)	Triggers translocation of NICD
	PTPRγ (Bouyain and Watkins, 2010)	Protein tyrosine phosphatase γ

VII. Final Remarks

The emergence of CNTN3 to −6 in genetic screens for genes causing neuropsychiatric disorders raises the question of how these genes affect processes of brain development and what the molecular mechanisms are. The present data indicate that each contactin has its own expression pattern with specificity in time and space in the developing brain and that partial overlap exists. The cerebral cortex seems an area that expresses the Cntn-4, −5, and −6 proteins, albeit for none of these genes, the cortex is the area with highest expression. It is supposed that the cortex is a critical region involved in cognitive disease symptoms of all neurodevelopmental disorders, including autism. There is also overlap in molecular properties of the Cntn3 to −6 proteins with regard to interaction with Ptprg, which is mediated by conserved motifs in the four N-terminal Ig domains. This suggests that their may be functional redundancy and overlap between these contactin members. However, genetic variation in association with disease has as yet been found in separate contactin genes, and gene knockouts for separate contactins produce distinct phenotypes. Further, combined deletion of CNTN4 and CNTN6, together with CHL1, in the 3p deletion syndrome results in a more severe phenotype. These data suggest that each contactin can independently contribute to appropriate development of the brain, perhaps in a regionally specific manner. Despite these separate functionalities, the molecular mechanisms that underlie the biological functions of each contactin may be related. The known protein interactions suggest that Cntn3 to −6 act in complexes with other proteins in which the contactins can either serve as receptor component with a heterophilic transmembrane protein or as a ligand acting on receptors on other cells. Identification of the protein networks of contactins is highly needed to provide further clues to their biological and pathogenetic roles during brain development and plasticity.