Emerging evidence implicating a role for neurexins in neurodegenerative and neuropsychiatric disorders

Synaptopathies are brain disorders characterized by dysfunctional synapses, which are specialized junctions between neurons that are essential for the transmission of information. Synaptic dysfunction can occur due to mutations that alter the structure and function of synaptic components or abnormal expression levels of a synaptic protein. One class of synaptic proteins that are essential to their biology are cell adhesion proteins that connect the pre- and post-synaptic compartments. Neurexins are one type of synaptic cell adhesion molecule that have, recently, gained more pathological interest. Variants in both neurexins and their common binding partners, neuroligins, have been associated with several neuropsychiatric disorders. In this review, we summarize some of the key physiological functions of the neurexin protein family and the protein networks they are involved in. Furthermore, examination of published literature has implicated neurexins in both neuropsychiatric and neurodegenerative disorders. There is a clear link between neurexins and neuropsychiatric disorders, such as autism spectrum disorder and schizophrenia. However, multiple expression studies have also shown changes in neurexin expression in several neurodegenerative disorders, including Alzheimer's disease and Parkinson's disease. Therefore, this review highlights the potential importance of neurexins in brain disorders and the importance of doing more targeted studies on these genes and proteins.


Introduction
There is accumulating evidence to suggest that synaptic dysfunction is present in both neuropsychiatric disorders, such as autism spectrum disorders (ASDs), schizophrenia and bipolar disorder (BD), and neurodegenerative disorders, such as Parkinson's disease (PD), Alzheimer's disease (AD) and Huntington's disease (HD) [1]. In fact, involvement of the synapse is such a prominent feature of the pathogenesis of various brain disorders that it has led to the coining of a specific term, 'synaptopathies'. Indeed, in the case of PD, the involvement of synaptopathy as an initial and central event in the disease pathogenesis, which precedes neuronal damage, has been postulated [2]. Synaptic dysfunction can occur due to mutations that alter the structure and function of synaptic components or abnormal expression levels of a synaptic protein.
Synapses are specialized junctions between neurons that transmit information and they connect neurons into millions of 'neural circuits' that underlie all brain functions [3]. The information transmitted allows the nervous system to respond to external stimuli and controls bodily functions, behaviour, emotions and memories [4]. This system is tightly controlled and regulated, and even slight perturbations can lead to synaptic dysfunction. An important aspect of synapse biology is the cell adhesion molecules that connect pre-and post-synaptic compartments [5]. These interactions in the synaptic cleft help to maintain synapse structure by delineating mutual boundaries [6]. These proteins are also important in synapse plasticity as synaptic cell adhesion is able to regulate the remodelling of synapses [7]. Interestingly, they are also involved in trans-synaptic signalling [5]. Thus, these proteins are highly important in the organization of synaptic junctions and overall brain function.
Neurexins are one type of synaptic cell adhesion molecule. They are pre-synaptically localized and bind to neuroligins and other proteins in the post-synapse (figure 1). Neurexins and their common binding partners, neuroligins, have recently gained more pathological interest as variants in both have been associated with several neuropsychiatric disorders, including autism and schizophrenia [8]. This further suggests that synaptic dysfunction plays a role in the development of these disorders. Synaptic dysfunction is also known to occur in neurodegenerative disorders [9]; however, it was considered an endpoint of these disorders, due to the considerably later onset of clinical symptoms and progressive appearance of cognitive deficits. This dichotomy has, recently, been challenged by the creation of 'disease-in-a-dish' models for multiple central nervous system (CNS) pathologies [9]. This research has identified commonalities between developmental and degenerative disorders, at both the cellular and molecular level, with most of these common mechanisms meeting at the synapse level [9]. Indeed, our laboratory has, recently, found a novel variant ( p.G849D) in the NRXN2 gene which may be implicated in PD [10]. Therefore, we believe it is important to investigate the potential role of neurexins in various neuropsychiatric and neurodegenerative disorders.
In this review, we summarize some of the key physiological functions of the neurexin protein family and the protein networks they are involved in. We also examine the available published literature to determine what research has been done on neurexins in neuropsychiatric and neurodegenerative disorders. This analysis provides an overview on what progress has been made in understanding the roles of synaptic functioning in these disorders and reveals the gaps in knowledge in this field.

Structure and biological functions of neurexins
Neurexins were first identified using affinity chromatography when neurexin 1α was found in rat brain extract on a column of α-latrotoxin [11]. α-latrotoxin is a potent neurotoxin from black widow spider venom that stimulates synaptic vesicle exocytosis and induces massive neurotransmitter release [12]. This work has been continued by Südhof and coworkers [13] who have characterized the neurexin proteins and their binding partners, the neuroligins [14]. In mammals, the neurexins are encoded by three NRXN genes (NRXN1-3), each of which has both an upstream promoter that is used to generate the α-neurexins, and a downstream promoter that is used to generate the shorter β-neurexins [13,15]. Neurexins also undergo extensive alternate splicing at five splice sites, generating significant diversity of more than 2000 potential variants [13,16]. The fact that neurexin splice insert sequences and their positions are well conserved among neurexin genes and between species supports the idea that alternative splicing has important functional roles.
The neurexins are transmembrane proteins that consist of an extracellular region responsible for trans-synaptic interactions, a transmembrane domain and a smaller cytoplasmic domain named PSD-95, DLG1, ZO-1 binding domain (PDZ) that is involved in intracellular protein interactions and signalling (figure 2) [13]. α-neurexins are composed of six large extracellular laminin/neurexin/sex   Neurexins are localized pre-synaptically and are distributed to both excitatory and inhibitory synapses [8]. Their functions are mediated by their binding to neuroligins (figure 1). Neuroligins have five known isoforms and are expressed post-synaptically [17]. Consequently, neurexins and neuroligins form synaptic complexes in the synaptic cleft and have been found to control synapse formation, maturation, validation and function [17]. Various combinations of the different neurexins and neuroligin binding partners at synapses may also help determine synapse specificity through differential interactions between multiple splice variants and isoforms of these proteins [8].
Primarily, neurexins function to maintain synaptic organization. Gene ontology (GO) analysis by WebGestalt (http:// www.webgestalt.org) [18] of the three neurexins indicates that they all function in protein binding, ion binding and possess molecular transducer activity (figure 3). They are also involved in cellular component organization, developmental processes, response to stimuli, cell communication and biological regulation. These processes thus demonstrate how neurexins are able to maintain synaptic organization but also show their multi-functional nature. As such, it is conceivable that disruptions in neurexins could be detrimental to their various functions and affect overall neuronal function and integrity.

Biological pathways and interacting partners of neurexins
To understand the broader biological pathways that the neurexins are involved in, protein-protein interaction network analysis and co-expression analysis was performed using the tools, STRING (https://string-db.org) [19] and GeneMania (https://genemania.org) [20]. STRING finds related genes by accessing the STRING database which contains experimental data and computational predictions. Data in STRING are weighted and integrated and a confidence score of 0-1 is calculated for all interacting protein partners. GeneMANIA finds proteins related to neurexins by leveraging functional association data, such as interactions, pathways, co-expression, colocalization and protein domain similarity. All functional data for the proteins observed in these networks were obtained from UniProt (https://www.uniprot.org) [21], unless otherwise stated, while pathway data were obtained from KEGG (https://www.kegg.jp) [22].

String analysis
Weighted string analysis was conducted on neurexin 1, 2 and 3 individually to determine their binding partners (figure 4a-c   royalsocietypublishing.org/journal/rsob Open Biol. 11: 210091 interacting proteins with very high confidence scores since the lowest score across the analyses was 0.969. This means that there is strong experimental evidence that these proteins interact with one or more of the neurexins. Analysis of the identified neurexin binding partners revealed many proteins important in the maintenance and functioning of synapses. Notably, variants in several of these proteins are implicated in neuropsychiatric and developmental disorders. Variants in neuroligin 1 and SHANK2 have been implicated in susceptibility to autism [23,24], while variants in neuroligin 3 have been implicated in Asperger syndrome and autism [25]. Variants in neuroligin 4X have been implicated in X-linked forms of Asperger syndrome, autism susceptibility and mental retardation [25][26][27]. Variants in CASK have been implicated in FG syndrome 4, an X-linked genetic disorder and mental retardation [28][29][30][31][32], variants in DLG4 have been implicated in intellectual developmental disorder 62 [33,34] and variants in synaptotagmin-1 have been implicated in Baker-Gordon syndrome [35]. Furthermore, binding partners of these proteins as well as the pathways they occur in could also give insight into the development of disease. CASK binds to amyloid precursor protein and neuroligin 1 binds to amyloid-β, both of which are important in AD. LRRTM3 is also a known positive regulator of amyloid-β formation. Notably, LRRTM3 may be considered a candidate gene for late-onset AD as it promotes the processing of amyloid precursor protein which leads to toxic amyloid-β accumulation [36]. DLG4 is involved in dopamine receptor binding and synaptotagmin-1 regulates dopamine secretion. The loss of dopamine functioning is crucial in PD. Indeed, DLG4 is involved in several pathways of neurodegeneration (in multiple diseases), the HD pathway as well as cocaine addiction.

GeneMANIA
GeneMANIA analysis was performed on the neurexins to reveal further potential protein-protein interactions (figure 4d; electronic supplementary material, table S1). We performed the analysis by selecting only proteins with stronger evidence of neurexin interactions, such as interactions with physical evidence, and evidence from co-expression and co-localization studies.
All of the binding partners observed by STRING analysis were still present; however, more interacting proteins were also identified. These proteins have more diverse functions but still function in overall synapse maintenance.
This analysis further identified afadin (AFDN), Rho GTPase activating protein 10 (ARHGAP10), cerebellin 1 (CBLN1), dystroglycan (DAG1), microtubule actin cross-linking factor 1 (MACF1), neurexophilin-2, neurexophilin-3, PDZ domain-containing protein 2 (PDZD2), proteolipid protein 1 (PLP1), syndecan binding protein 1 (SDCBP), SDCBP2, SH3 domain-containing GRB2-like protein 2 (SH3GL2), signalinduced proliferation-associated 1-like protein 1 (SIPA1L1), synaptotagmin-13 (SYT13) TAFA chemokine-like family member 1 (TAFA1), TUBB-like protein 1 (TULP1) and XKrelated protein 4 (XKR4) as interactors of one or more neurexin proteins. AFDN, ARHGAP10, MACF1 and SIPA1L1 are all involved in actin filament binding/ organization, while PDZD2, SDCBP and SDCBP2 are involved in cell binding and cytoskeletal organization.  royalsocietypublishing.org/journal/rsob Open Biol. 11: 210091 Dysregulation of any of these proteins could thus affect cell adhesion and binding at the synapse. In addition, variants in MACF1 have been implicated in lissencephaly 9 with complex brainstem malformation [37]. CBLN1 is essential for synapse integrity and plasticity and its disruption could lead to synapse dysfunction. DAG1 has multiple functions, such as laminin and basement membrane assembly, cell survival, peripheral nerve myelination, nodal structure and cell migration. Variants in DAG1 have been implicated in both type A and type C muscular dystrophy-dystroglycanopathy [38][39][40][41]. Muscular dystrophies are genetic disorders characterized by the degeneration of skeletal muscle. Type C muscular dystrophy-dystroglycanopathy affects the limbgirdle area [40], while type C is congenital with brain and eye anomalies [39]. Neurexophilin-2 and neurexophilin-3 are both ligands for α-neurexins and are involved in the neuropeptide signalling pathway. Disruption of these proteins could, therefore, affect neurotransmitter release and the subsequent signalling. PLP1 is the major myelin protein in the CNS and is important for maintaining the structure of myelin. Disruption of this protein could, therefore, negatively affect the downstream myelination of neurons, as is seen in multiple sclerosis (MS). Interestingly, PLP1 is also involved in the development of the substantia nigra, the main brain region affected by PD. Therefore, PLP1 alterations could also lead to disruptions in this brain region. SH3GL2 has been implicated in synaptic vesicle endocytosis, while synaptotagmin-13 may be involved in transport vesicle docking to the plasma membrane. Dysregulation of these proteins could thus affect neurotransmitter functioning. TAFA1 is involved in the modulation of neural stem cell proliferation and differentiation; therefore, dysregulation of this protein could result in developmental disorders. TULP1 is required for normal development of photoreceptor synapses. Variants in TULP1 are associated with Leber congenital amaurosis [42,43] and retinitis pigmentosa [42,[44][45][46][47]. However, this protein is also involved in actin filament binding, therefore, its dysregulation could also affect cell adhesion and binding at the synapse. Not much is known about XKR4 except that it is involved in apoptosis during development. Therefore, its dysregulation could also possibly result in developmental disorders. GO terms and physiological/pathway information on all binding partners identified by STRING and GeneMANIA are available in electronic supplementary material, figure S1 and table S1.

Role of neurexins in neuropsychiatric disorders
Literature-based searches using neurexin as a search term identified several studies that reported an association of neurexins in various neuropsychiatric disorders. The main findings of these studies are reported in table 1 and are summarized below.

Human studies
NRXN1 has been well documented for its association with ASDs [54]. Several genetic analyses of families and populations of people with ASD have shown that copy number variations (CNVs) and de novo mutational events at the NRXN1 locus are enriched in ASD [48,49,51,52,54]. In one study, NRXN1 was sequenced in cases of ASD with mental retardation [50]. Mutations (c.-3G > T in the Kozak region, c.3G > T at the initiation codon ( p.M1), p.R375Q and p.G378S) were found in the NRXN1β coding region thereby potentially implicate synapse dysfunction an important determinant in ASD [50]. The first evidence for a potential role of NRXN2 in ASD was provided by a report of a frameshift mutation within NRXN2 exon 12 (c.2733delT) in a boy with ASD and his father who had severe language delay [57]. This mutation results in a truncated neurexin 2α protein that lacks the binding sites for the established post-synaptic binding partners LRRTM2 and neuroligin 2 [57]. Subsequently, a 21-year-old man with a clinical phenotype including autistic traits, such as speech and language deficits and pathological insistence on routine, was reported to have a 570 kb de novo deletion of 24 genes at chromosome 11q13.1, including NRXN2 [58].
Using microarray analyses on RNA extracted from brain tissue, Mirnics et al. [69] did not observe a difference in neurexin 1 expression between schizophrenia (SCZ) and control samples. However, since then, a link between neurexin 1 and SCZ has been reported in other studies. One study reported that NRXN1 deletions are more common in those with SCZ; however, it also found that there was incomplete penetrance of these deletions in families with SCZ [72]. Kirov et al. [70] observed a deletion in an SCZ patient at 2p16.3 that disrupts NRXN1 and predicted that it was highly likely to be pathogenic. Also, NRXN1 deletions were shown to segregate with several neuropsychiatric disorders in a study of a complex family [78]. The proband had SCZ and other members of his family had mental retardation, schizophreniform disorder and affective disorder [78]. After genotyping the proband and eight family members, they found two rare deletions upstream of the NRXN1 gene (2p16.3) that co-segregate with these disorders [78]. Notably, this shows that deletions in NRXN1 may manifest as multiple neuropsychiatric phenotypes.
Angione et al. [64] implicated NRXN1 in epilepsy. They identified a 2p16.3 deletion in an 8-year-old male patient diagnosed with epilepsy showing symptoms of myoclonicatonic seizures (EMAS) [64]. This deletion included the first five exons of the NRXN1 gene [64].
NRXN genes may also be involved in treatment response. In one study, it was found that variants in NRXN1 may affect the long-term treatment outcome of patients with BD by modulating the effects of antipsychotics [61]. In a study of Levetiracetam resistance, an antiepileptic drug, Grimminger et al. [63] found that neurexin 1 is differentially expressed in non-responder and responder patients with mesial temporal lobe epilepsy (mTLE), whereby lower levels of neurexin 1 were observed in responder patients.

Association studies
A genome-wide association study (GWAS) by Liu et al. [51] specifically examined NRXN1 in an autism cohort of the Chinese Han population and discovered 22 variants that were associated with ASD. In this cohort, one SNP (rs2303298) was also significantly associated with a risk of developing ASD [51]. Furthermore, a GWAS of SCZ in Spain showed that a NRXN1 single nucleotide polymorphism (SNP) (rs3850333) was close to the significance threshold [71], while another GWAS in American patients of European or African ancestry showed that NRXN1 is associated with antipsychotic royalsocietypublishing.org/journal/rsob Open Biol. 11: 210091 'trios' consisting of a mother, father and an ASD-affected child a de novo CNV in NRXN1 was discovered in a large cohort of families with a single ASD-affected child and at least one unaffected sibling [48] performed genetic analyses to identify CNVs ASD NRXN1 GWAS 1174 families from SSC were genotyped rare de novo events/CNVs at NRXN1 are strongly associated with autism [49] identified CNVs and de novo events ASD NRXN1β genetic analysis 86 patients with both ASD and mental retardation four novel mutations in NRXN1β were identified by sequencing the exon of the gene in cases with autism and mental retardation [50] The coding sequence of the NRXN1β gene was analysed by PCR ASD NRXN1 genetic analysis 313 ASD patients and 500 healthy controls from a Chinese autism cohort were recruited 22 variants in the NRXN1 gene were discovered in the Chinese Han population; one SNP (rs2303298) was significantly associated with a risk of autism in this cohort [51] performed genomic DNA sequencing ASD NRXN1 genetic analysis 2478 ASD individuals from SSC and 719 ASD individuals from AGRE recurrent CNVs in NRXN1 are enriched in autism. [52] 580 controls from ClinSeq and NIMH used a custom microarray to analyse CNVs ASD NRXN1, 2 and 3 cell culture iPSCs were produced from probands and unaffected family members neurexin 1, 2 and 3 mRNA is overexpressed in patient-derived iPSCs and differentiated organoids [53] iPSCs underwent neuronal differentiation to organoids RNA sequencing was performed on both iPSCs and differentiated organoids ASD NRXN1 genetic analysis 2591 families from SSC were genotyped NRXN1 is an ASD risk gene [54] identified CNVs, de novo deletions and ASD risk genes ASD NRXN1 and 2 animal study RNA was isolated from the whole brain of age-matched monoamine oxidase A KO mice and wild-type mice neurexin 1 and 2 are downregulated in monoamine oxidase A KO mice [55] Microarrays were used to determine gene expression levels ASD α-NRXNs animal and cell culture study transfected C. elegans strains and HEK-293 cells with plasmids expressing NRXN1α and different α2δ subunits changes in α-neurexin binding to α2δ-3 subunits of N-type calcium channels could be important in some forms of autism spectrum disorders [56] performed co-immunoprecipitation and pull-down assays ASD NRXN2 genetic analysis 142 ASD patients and 94 non-syndromic controls observed a frameshift mutation in NRXN2 exon 12 in a patient with ASD inherited from a father with severe language delay [57] sequenced NRXN 1,2 and 3 genes  [72] genotyped samples and analysed them for CNVs and deletions there is incomplete penetrance of NRXN1 deletions in families with schizophrenia SCZ NRXN1 genetic analysis data from 572 SCZ patients and 551 controls were used to select genes for sequencing missense variants at NRXN1 may be protective against schizophrenia susceptibility [73] 153 SCZ patients and 153 controls were sequenced for 21 chosen genes using NGS SCZ NRXN1 cell culture isolated primary rat neurons from hippocampi overexpressing Caveolin-1, a potential therapeutic for schizophrenia, in neurons increased expression of proteins involved in synaptic plasticity (PSD95, synaptobrevin, synaptophysin, neurexin 1 and syntaxin 1) as well as DISC1 [74] differentiated human neurons derived from human iPSCs overexpressed Caveolin-1 in both cell types western blotting was used to measure the expression of proteins involved in synaptic plasticity as well as DISC1, an SCZ-associated protein (Continued.) royalsocietypublishing.org/journal/rsob Open Biol. 11: 210091 animal study generated iPSCs from 5 childhood-onset SCZ patients and 4 controls neurexin 1 was downregulated in chimeric mice produced from iPSCs derived from patients with childhood-onset schizophrenia [75] differentiated iPSCs into glial cells transplanted glial cells into mice via injection into the corpus collosum performed molecular analyses on both the differentiated glial cells and chimeric mice SCZ NRXN1 GWAS obtained genetic data and treatment response data of 302 SCZ patients treated with lurasidone and 117 SCZ patients treated with a placebo from two clinical SCZ trials NRXN1 is associated with antipsychotic response to lurasidone in schizophrenia patients [76] performed a GWAS SCZ NRXN1 cell culture generated iPSCs from 3 NRXN1 deletion SCZ patients and 3 controls and differentiated them into human iNs heterozygous NRXN1 deletions impair neurotransmitter release and synaptic function, and increase the levels of the synaptic scaffolding protein, CASK in human iNs but not mESCs generated from NRXN1 KO mice [77] generated mESCs from NRXN1 KO mice analysed neuronal development, synapse formation and neurotransmitter release SCZ and other neuropsychiatric disorders NRXN1 genetic analysis recruited a family with multiple neuropsychiatric disorders two rare deletions upstream of the NRXN1 gene (2p16.3) segregate with schizophrenia, schizophreniform disorder, and affective disorder in a family [78] the proband has SCZ, while other family members have mental retardation, schizophreniform disorder and affective disorder genotyped the proband and eight family members royalsocietypublishing.org/journal/rsob Open Biol. 11: 210091 response to lurasidone in SCZ patients [76]. Additionally, an association study on Spanish SCZ patients showed that missense mutations in NRXN1 may actually protect against susceptibility to SCZ [73]. In a Taiwanese GWAS study, a significant association between NRXN3 and BD was found [60]. And finally, an association study on Australian borderline personality disorder (BPD) patients showed that several NRXN3 SNPS were nominally associated with BPD in heroin-dependent cases [62].

In vitro and in vivo models of disease
Functional in vitro and in vivo studies have also found evidence for the roles of neurexins in ASD. Monoamine oxidase A knockout (KO) mice, which are an animal model for autism, exhibited downregulated levels of both neurexin 1 and neurexin 2 [55]. Furthermore, mice colonized with the microbiota of ASD patients showed differential splicing of NRXN2 [59]. Another animal study showed that changes in the binding of α-neurexins to N-type calcium channels could be important for some forms of ASD as it mediates synaptic inhibition [56]. Finally, a study using ASD patientderived induced pluripotent stem cells (iPSCs) and differentiated organoids showed that neurexin 1, 2 and 3 mRNA is overexpressed in these samples [53].
One study examined neurexins in Fragile X syndrome, a genetic disorder with features similar to ASD, and characterized by the silencing of the FMR1 gene [79]. Individuals with Fragile X experience a range of neurodevelopmental problems, such as learning disabilities and cognitive impairment, and males are usually more severely affected. Using FMR1 KO mice, researchers probed brain sections to determine the levels of neurexin 1, 2 and 3 [66]. Interestingly, they found that neurexin 3 mRNA levels are increased in female mice but decreased in male mice and predicted that this may help explain the sex difference observed in this disorder [66].
In an animal study of SCZ, neurexin 1 was found to be downregulated [75]. This study generated iPSCs from patients with childhood-onset SCZ, differentiated them into glial cells and injected the glial cells into mice to form chimeric mice as a model organism [75]. Interestingly, an in vitro study of SCZ showed that overexpressing Calveolin-1, a potential therapeutic for SCZ, actually increased the levels of neurexin 1 as well as other proteins involved in synaptic plasticity [74].
Neurexin 2α has been implicated in epilepsy and, more specifically, in seizures. Researchers observed an increase in neurexin 2α expression in the dentate gyrus of the hippocampus in an induced-seizure mouse model [65]. Finally, in one study, a rat chronic mild stress model of depression was used to determine if neurexin expression was altered in major depressive disorder; however, no change in neurexin 1, 2 or 3 levels was observed [67].
So far, there have been two studies validating the effect of NRXNs in vitro, both by Pak et al. [68,77]. These studies cultured human stem cells as well as mice cells generated from NRXN1 KO mice. The first study introduced two conditional NRXN1 mutations previously seen in a range of neuropsychiatric disorders, including ASD and SCZ, into human embryonic stem cells (hESCs) using adeno-associated virus recombination and differentiated them into human-induced neurons (iNs) [68]. These cells were compared to cortical neurons generated from NRXN1α KO mice [68]. The second study generated iPSCs from three NRXN1 deletion SCZ patients and three controls, and again differentiated them into human iNs [77]. These cells were compared to mouse embryonic stem cells (mESCs) from NRXN1 KO mice [77]. Both studies showed that heterozygous NRXN1 deletions were able to impair neurotransmitter release and synaptic function, and increase the levels of the synaptic scaffolding protein, CASK, in human iNs but not in mice cells [68,77]. Therefore, these studies provide evidence that NRXN1 deletions exhibit a major synaptic transmission phenotype in humans and are thus meaningful at a pathophysiological level.
In summary, these studies demonstrate a link between NRXNs and neuropsychiatric disorders such as ASD and SCZ, especially involving full or partial deletions of these genes. NRXNs have also been associated with BD and BPD. In addition, protein expression studies have shown changes in neurexin expression in animal models of epilepsy/seizures and Fragile X syndrome.

Role of neurexins in neurodegenerative disorders
Additionally, literature-based searches provided proof for the involvement of neurexins in various neurodegenerative disorders, and these studies are listed in table 2 and discussed below.

Human studies
Studies examining cerebrospinal fluid (CSF) from AD patients have observed lowered expression of neurexin 1 [81], as well as neurexin 2α and neurexin 3α [85]. In addition, it was found that these changes precede the neurodegeneration markers as they were observed in the preclinical stage 1 of AD [85]. Moreover, Aβ 42 fibrils in CSF were found to bind to neurexin 1, 2 and 3 as well as proteoglycans and growth factors [83]. Levels of the synaptic proteins neuronal pentraxin 2 (NPTX2), GluA4containing glutamate (AMPA4), neuroligin 1 and neurexin 2α are also declined in plasma neuron-derived exomes and this decline was associated with AD progression [82]. Neurexin 3 protein expression has also been seen to be specifically downregulated in blood samples of AD patients [84]. Another expression analysis on CSF from MS patients identified neurexin 2α levels as a potential biomarker for the disorder [98], while a genetic analysis found that a mutant miRNA, MIR8485, overexpresses neurexin, which leads to a calcium overload in pre-synapses [99]. It was hypothesized that this could induce neurodegeneration in MS [99].
A study examining gene expression in brain tissue samples of patients with PD found that genes related to nerve function, such as protocadherin-8 (PCDH8) and neurexin 3, were downregulated [109].
Two studies on mild cognitive impairment (MCI) found promising results. MCI is a milder form of dementia that is considered the intermediate state of cognitive decline between normal ageing and dementia [114]. Berchtold et al. [95] found that neurexin 1 and neurexin 2 are upregulated in MCI. In addition, neurexin 1 expression was found to be associated with longitudinal phenotypes in MCI, but not in AD [96].
One study examined neurexins in order to identify genes that are differentially regulated by HIV encephalitis [94]. This royalsocietypublishing.org/journal/rsob Open Biol. 11: 210091  GWAS of brain structure obtained neuroimaging and genetic data from 818 subjects as part of ADNI NRXN3 (KIAA0743) is associated with temporal lobe structure in AD patients [80] performed a GWAS with 546,314 SNPs using temporal lobe and hippocampal volume as quantitative phenotypes AD NRXN1 protein expression analysis collected CSF samples from 10 AD patients and 15 healthy controls the concentrations of the synaptic proteins neurexin 1 and neuronal PTX1, as well as neurofascin, were significantly lowered in AD CSF [81] analysis using LC-MS/MS AD NRXN2α protein expression analysis collected blood and CSF samples from 28 AD patients and 28 controls significantly lower levels of the synaptic proteins NPTX2, AMPA4, neuroligin 1 and neurexin 2α in the plasma neuron-derived exomes correlate with AD progression [82] extracted plasma neuron-derived exomes CD81, NPTX2, AMPA4, neuroligin 1 and neurexin 2α proteins were quantified using ELISAs AD NRXN1, 2 and 3 protein expression analysis collected CSF samples from six AD patients and five non-AD patients Aβ 42 fibrils in AD CSF are involved in binding to proteoglycans, growth factors and neuron-associated proteins, such as neurexin 1, 2 and 3 [83] binding assays were performed to determine which proteins in CSF bind to Aβ 42 fibrils and/or protofibrils AD NRXN3 transcriptome and RNA expression analysis selected data from 263 AD patients and 151 non-demented controls sampled from the religious orders study neurexin 3 expression is downregulated in AD [84] performed RNA expression profiling AD NRXN2α and NRXN3α protein expression analysis collected CSF samples from AD patients and cognitively normal controls (three stage study with different n for each stage) levels of neurexin 2α and neurexin 3α, as well as other synaptic proteins are decreased in preclinical AD CSF [85] performed LC-MS/MS and LC-SRM AD and ageing NRXN1, 2 and 3 microarray analysis obtained frozen brain samples from 26 AD cases and 55 non-AD controls from National Institute on Ageing Alzheimer's disease brain banks SYNAPTIC proteins, including neurexin 1, 2 and 3, undergo altered expression in ageing and AD [86] used microarrays to evaluate expression profiles of 340 synaptic genes AD and ageing NRXN3 animal study mice were divided into four groups, with four mice in each group: memory intact AD-transgenic mice, memory impaired AD-transgenic mice, memory intact aged mice and memory impaired aged mice neurexin 3 is downregulated in AD-transgenic mice with impaired memory, but not in normal aged mice with impaired memory [87] performed proteomics on the hippocampus of each mouse (Continued.) royalsocietypublishing.org/journal/rsob Open Biol. 11: 210091 microarray analysis performed a microarray analysis on 47 post-mortem brain samples from cognitively intact aged individuals from the MAP study neurexin 1 and 3 have decreased expression in ageing and AD but have increased expression in association with late-life physical activity [88] identified 48 microarrays from the public GEO: 16 young cases, 18 cognitively intact aged cases and 14 AD cases analysed data to identify genes related to physical activity, ageing and AD ageing NRXN3 animal study cerebella were removed from three adult C57BL/6 J mice and three aged C57BL/6Jnia mice neurexin 3 is downregulated in the cerebellum of aged mice [89] RNA was extracted and sequenced ageing NRXN2 methylation analysis monocytes were purified from PBMCs CpG sites associated with NRP1, NRXN2 and miR-29b-2 are hypomethylated in monocytes during ageing [90] analysis of methylation was performed on genomic DNA from monocytes ageing NRXN1 animal study 28 Swiss albino mice were divided into four groups by age: young, adult, middle age and old neurexin 1 and neuroligin 3 are differentially expressed in cerebral cortex and hippocampus during different stages of ageing, which might be responsible for alterations in synaptic plasticity during ageing [91] molecular techniques were used to analyse neurexin 1 and neuroligin 3 expression HIV encephalitis NRXN1 microarray analysis received cortical brain tissue from 13 HIV patients: eight with HIV encephalitis and five without neurexin 1 is downregulated in HIV encephalitis [94] extracted total RNA performed microarray analysis mild cognitive impairment (MCI) NRXN1 and 2 microarray analysis obtained frozen brain samples from 16 MCI cases, 25 AD cases and 24 aged controls from National Institute on Aging Alzheimer's Disease brain banks neurexin 1 and 2 are upregulated in MCI [95] extracted total RNA performed a microarray analysis (Continued.) royalsocietypublishing.org/journal/rsob Open Biol. 11: 210091 performed a co-immunoprecipitation assay with neurexin 1 and acetylcholinesterase co-transfected HEK-293 cells to express neurexin 1β and neuroligin 1 and cultured these cells in acetylcholinesterase conditioned media performed a co-immunoprecipitation assay with neurexin 1β and neuroligin 1 neurotoxicity NRXN3β animal study groups of 3 Sprague-Dawley rats were treated with sarin via intramuscular injection sarin exposure causes a persistent downregulation of neurexin 1β and breakdown of the blood-brain barrier [105] rats were sacrificed 15 min or 3 months after sarin exposure dissected brains and extracted total RNA performed a microarray analysis neurotoxicity NRXN2α animal study wild-type zebrafish were repeatedly exposed to domoic acid via intracoelomic injection neurexin 2α was upregulated in zebrafish two weeks after exposure to domoic acid, a neurotoxin [106] dissected brains and extracted total RNA performed a microarray analysis cell culture cultured SH-SY5Y cells and primary mouse mesencephalic cells downregulation of neurexin 1 mRNA and protein was observed in the 6-OHDA-induced cell culture models of PD [107] treated cells with 6-OHDA performed RAP-PCR and analysed the PCR products using RT-PCR and qRT-PCR PD NRXN2 animal study transgenic mice were assigned to 4 treatment groups with 20 mice per group transgenic mice overexpressing α-synuclein have increased levels of neurexin 2 [108] cholesterol oximes were administered in food pellets chronic administration of cholesterol oximes to these mice decreased neurexin 2 levels TH+ neurons were isolated from the substantia nigra and subjected to a transcriptome analysis PD NRXN3 genetics analysis obtained genomic data of 29 PD samples and 18 controls from the GEO database genes related to nerve function, such as PCDH8 and neurexin 3, are downregulated in PD brain tissue samples [109] analysed the data to identify disease-related genes and differential gene expression (Continued.) royalsocietypublishing.org/journal/rsob Open Biol. 11: 210091 animal study adult Wistar rats were divided into five treatment groups, with 6-8 rats in each group neurexin 1 is significantly decreased in the striatum of 6-OHDA-induced rats [110] experimental groups had 6-OHDA brain injections with or without different concentrations of allopregnanolone treatment with allopregnanolone attenuates this and other molecular changes western blots were performed to evaluate the levels of the synaptic proteins PSD95 and neurexin 1 in the striatum PD NRXN1 RNA expression analysis MRI data from 149 PD patients and 64 healthy controls were obtained from the PPMI database neurexin 1 does not have an expression pattern that predicts regional atrophy in PD [111] 17 genes of interest implicated in PD were selected for whole-brain expression analysis PD NRXN1 animal study adult Wistar rats were divided into seven treatment groups, with seven rats in each group neurexin 1 expression is decreased in the striatum of 6-OHDA-induced rats [112] experimental groups had 6-OHDA brain injections with or without different concentrations of apelin-13 6-OHDA rats treated with apelin-13 showed increased neurexin 1 expression in the striatum western blots were performed to evaluate the levels of the synaptic proteins PSD95, neurexin 1 and neuroligin in the striatum spinal muscular atrophy (SMA) NRXN2α animal study used HB9:D3cpv/MN-transgenic zebrafish and Smn-/−/SMN2 mice SMN-deficiency downregulates neurexin 2α expression and alters its splicing in zebrafish and mouse models of SMA [113] isolated total RNA from both models performed a microarray analyses and qRT-PCR royalsocietypublishing.org/journal/rsob Open Biol. 11: 210091 microarray study showed that neurexin 1 is downregulated in HIV encephalitis. Finally, González-Velasco et al. [92] showed that neurexin 2 and neurexin 3 mRNA levels are downregulated in ageing. Another study found that neurexin 1, 2 and 3 underwent altered expression in both AD and ageing [86]. A more recent study from the same group confirmed decreased expression of both neurexin 1 and neurexin 3 in AD and ageing [88]. Interestingly, they also found that late-life physical activity is associated with increased expression of these proteins [88].

Association study
A GWAS performed by Stein et al. [80] showed that the SNP rs7155434 within NRXN3 is associated with temporal lobe structure in AD patients. Temporal lobe volume deficits are a known risk factor for AD; therefore, this study potentially implicates NRXN3 with AD risk [80].

In vitro and in vivo models of disease
Several studies involving cell culture and/or rodent disease models have also shown differences in the expression of neurexin proteins. Three studies showed that neurexin 1 is downregulated in PD. One of these measured neurexin 1 mRNA in two 6-OHDA (6-hydroxydopamine)-induced cell culture models; one using human neuroblastoma (SH-SY5Y) cells and the other using primary mouse mesencephalic cells [107]. The other studies used a 6-OHDA-induced rat model of PD and both saw a decrease in neurexin 1 in the striatum [110,112]. In addition, these studies showed that treatment with apelin-13 [112] or allopregnanolone [110] is able to attenuate this change. Apelin-13 is an endogenous ligand for APJ [115] that has been investigated as a potential protective neuropeptide due to the role of the apelin-APJ system in neuronal survival [116], while allopregnanolone is a reduced metabolite of progesterone [117] and has reduced CSF levels in PD patients [118]. Freeze et al. [111], however, noted that the expression pattern of neurexin 1 does not predict regional atrophy in PD. This suggests that neurexin 1 is not a marker for PD; however, it does not exclude it as an important protein in PD pathogenesis. Another study in PD-transgenic mice overexpressing α-synuclein found that neurexin 2 expression was also upregulated [108]. In addition, chronic administration of cholesterol oximes was able to increase the transcription of cytoprotective genes and undo transcriptome alterations, including the alteration of neurexin 2 expression [108].
Two studies using induced models of MS implicated neurexins in this disorder. One study induced experimental autoimmune encephalomyelitis (EAE) in rats and observed downregulation of neurexin 3 [97]. This is a commonly used model that mimics certain aspects of MS. The other study used an in vitro model of MS, cultured human monocytes, and observed an association between ceramide-induced hypermethylation of DNA and neurexin 1 mRNA [100].
An animal study performed by Neuner et al. [87] showed that neurexin 3 is downregulated in AD-transgenic mice, but not in normal aged mice with impaired memory. However, Popesco et al. [89] found that neurexin 3 is downregulated in the cerebellum of aged mice. Another study found that levels of both neurexin 1 and neuroligin 3 are differentially expressed in cerebral cortex and hippocampus of mice and that these expression levels change during different stages of ageing [91]. They predicted that this may be responsible for the changes in synaptic plasticity observed with age [91]. Finally, a DNA methylation study by Tserel et al. [90] showed that CpG sites associated with NRP1, NRXN2 and miR-29b-2 are hypomethylated in monocytes during ageing.
To date, only one study has examined neurexins in amyotrophic lateral sclerosis (ALS) and spinal muscular atrophy (SMA). In a cell culture model of ALS, neurexin 1 and other RNA targets of TDP-43 were dysregulated [93]. TDP-43 is a component of the cytoplasmic inclusion bodies present in ALS patients [93]. Fragments of TDP-43 are ubiquitinated, hyperphosphorylated and then accumulate in neurons and glia [119]. In zebrafish and mouse KO models of SMA, the SMN-deficiency downregulated neurexin 2α expression and altered its splicing [113]. SMA is associated with mutation or deletions in the SMN gene [120] and lack of the SMN protein causes degeneration and results in anterior horn cell dysfunction.

Models of induced neurodegeneration and toxicity
Several studies investigated neurexins in models of neurodegeneration or toxicity instead of studying a specific neurodegenerative disease.
Four studies examined the role of neurexins in models of induced neurodegeneration. Two of these studies hypothesized that neurexin 1 could play a role in neurodegeneration. The first study showed that a novel TDP-43 binding miRNA, miR-NID1 (miR-8485) is able to repress neurexin 1 and predicted that this could play a role in neurodegeneration [103]. Xiang et al. [104] found in vitro that excessive glycosylated acetylcholinesterase could competitively disrupt the neurexin 1β-neuroligin junctions and impair the integrity of glutamatergic synapses, which could lead to neurodegeneration. The other two studies showed that neurexin 3 is downregulated in animal models of neurodegeneration [101,102]. Suh et al. [102] saw that neurexin 3 was downregulated in the hippocampus of mice treated with kainic acid, an inducer of neurodegeneration, while Swanberg et al. [101] found that neurexin 3 is downregulated in rats with axonal damage caused by ventral root avulsion.
Two studies were conducted in animal models of neurotoxicity. One study exposed zebrafish to chronic, low levels of the neurotoxin domoic acid and saw an upregulation of neurexin 2α after two weeks [106]. The other study exposed rats to acute doses of sarin, which caused a persistent downregulation of neurexin 1β and breakdown of the blood-brain barrier [105].
In summary, multiple studies have shown changes in neurexin expression in AD, ALS, MS, PD and SMA. Many of these studies have observed downregulation of protein expression for neurexin 1, 2 and 3 in these disorders. Similarly, downregulation of neurexin 1, 2 and 3 were observed in disorders such as HIV encephalitis and MCI and in studies on ageing, in models of neuronal toxicity, and animal models of MS and ALS.

Concluding remarks
A clear link between synaptic dysfunction and neurodegenerative as well as neuropsychiatric disorders has been established in recent years. Our literature-based searches revealed several royalsocietypublishing.org/journal/rsob Open Biol. 11: 210091 studies that have linked CNVs, deletions or expression changes in neurexins to different disorders. The evidence is most compelling for a role of neurexins in neuropsychiatric disorders, particularly in regard to the involvement of neurexin 1 in ASD and SCZ. Currently, there is comparatively less evidence for the involvement of neurexins in neurodegenerative disorders. Although there have been some studies that have suggested that neurexins may be important in these disorders, at this stage more experimental data are still needed to draw concrete conclusions. Therefore, it is apparent that more targeted studies in various disorders involving these genes as well as the proteins they encode are warranted. In terms of their broader biological and physiological functions, the neurexins function as molecular inducers, are involved in iron and protein binding, and play a role in cell-to-cell communication and response to stimuli, consequently making them critical for normal cell functioning. Furthermore, these proteins interact with various other proteins such as the neuroligins and the LRRTM proteins identified via protein interaction networks. This implicates the neurexins' involvement in synaptic integrity and functioning making them promising candidates as disease genes for a wide range of brain pathologies.
In summary, this review serves to highlight the potential importance of the neurexin genes and proteins in human disease and recommends that more targeted studies on these genes and proteins are warranted. Furthermore, with the wealth of exomic and genomic sequences and genome-wide transcriptomic datasets now available, it has become plausible to interrogate them for their involvement in various human disorders, on a scale not previously possible. In addition, the human neurexin protein structures urgently need to be solved to understand the function and infer accurate proteinprotein interactions as well as to understand the effect of mutations on the protein structure. Ultimately, improved knowledge on synapses and their individual components are necessary to develop novel therapeutic approaches for the emerging and exciting field of synaptopathies.
Data accessibility. This article has no additional data.