Deregulation of microtubule organization and RNA metabolism in Arx models for lissencephaly and developmental epileptic encephalopathy.

X-linked lissencephaly with abnormal genitalia (XLAG) and developmental epileptic encephalopathy-1 (DEE1) are caused by mutations in the Aristaless-related homeobox (ARX) gene, which encodes a transcription factor responsible for brain development. It has been unknown whether the phenotypically diverse XLAG and DEE1 phenotypes may converge on shared pathways. To address this question, a label-free quantitative proteomic approach was applied to the neonatal brain of Arx knockout (ArxKO/Y) and knock-in polyalanine (Arx(GCG)7/Y) mice that are respectively models for XLAG and DEE1. Gene ontology and protein-protein interaction analysis revealed that cytoskeleton, protein synthesis and splicing control are deregulated in an allelic-dependent manner. Decreased α-tubulin content was observed both in Arx mice and Arx/alr-1(KO) Caenorhabditis elegans ,and a disorganized neurite network in murine primary neurons was consistent with an allelic-dependent secondary tubulinopathy. As distinct features of Arx(GCG)7/Y mice, we detected eIF4A2 overexpression and translational suppression in cortex and primary neurons. Allelic-dependent differences were also established in alternative splicing (AS) regulated by PUF60 and SAM68. Abnormal AS repertoires in Neurexin-1, a gene encoding multiple pre-synaptic organizers implicated in synaptic remodelling, were detected in Arx/alr-1(KO) animals and in Arx(GCG)7/Y epileptogenic brain areas and depolarized cortical neurons. Consistent with a conserved role of ARX in modulating AS, we propose that the allelic-dependent secondary synaptopathy results from an aberrant Neurexin-1 repertoire. Overall, our data reveal alterations mirroring the overlapping and variant effects caused by null and polyalanine expanded mutations in ARX. The identification of these effects can aid in the design of pathway-guided therapy for ARX endophenotypes and NDDs with overlapping comorbidities.

Introduction

The correct development of the mammalian brain requires fine-tuned orchestration of several homeotic transcription factors (TFs), in which deregulation may cause neurodevelopmental diseases (NDDs) (1). Aristaless-related homeobox (ARX; MIM 300382) is an X-chromosome gene that encodes a bifunctional TF absolutely required for proper brain morphogenesis. Mutations in ARX have been found in a wide range of incurable NDDs affecting male children, with severity dependent on the type of alteration (2–4). Point missense mutations affecting the homeobox domain (HD) in ARX lead to loss of function (LoF) causing X-linked lissencephaly with agenesis of the corpus callosum and ambiguous genitalia (XLAG; MIM 300215). In addition to profound structural brain anomalies, XLAG patients have early-onset intractable seizures, and severe psychomotor retardation, often dying within the first months of life (5–7). In contrast, partial LoF alterations by elongations of the first polyalanine (polyAla) tract in the protein cause developmental and epileptic encephalopathy 1 (DEE1; MIM 308350; also named early infantile epileptic encephalopathy 1), a less severe disease compared to XLAG (2,3,7–10). DEE1 patients rather develop pharmaco-resistant paediatric epilepsy characterized by tonic spasms, arrest of psychomotor development, hand dystonia and severe intellectual disability (ID) but with an apparently normal brain structure (11). Previous studies conducted by us and others have shown that XLAG mutations disrupt the transcriptional program controlled by ARX; whereas, expansions of the polyAla tracts are hypomorphic mutations, with reduced transcriptional activity and binding properties (3,4,10,12,13). This partially faulty activity of the elongated ARX protein is associated with the aggregation of mutant proteins, a phenomenon also detected in other instances of pathogenic expanded repeats that might add a possible gain of function (GoF) increasing the phenotype severity (3,14,15).

These findings support the hypothesis that differences in the clinical manifestations of XLAG and DEE1 may correlate with a varying degree of ARX function: the complete loss of transcriptional activity exhibited by LoF mutations produces the severe cortical malformation whereas the partial impairment of transcriptional activity with a potential GoF activity shown by polyAla elongations (partial LoF/GoF) induce the refractory epilepsy (16). Whether, where and when the functional features of polyAla elongations synergize each other and how they impact disease heterogeneity are completely unknown. However, the identification of the molecular basis of Arx(GCG)7/Y pathophysiology remains a big task of ARX studies whose investigation could be fundamental in the identification of relevant therapeutics.

A considerable advance in the analysis of XLAG and DEE1 pathogenesis was afforded by the generation of two Arx transgenic mice: the knockout ArxKO/Y, which harbours a null mutation that abolishes the TF activity of ARX and recapitulates several XLAG signs including a small brain, defective neuronal migration and neonatal death (Table 1) (17), and the knock-in Arx(GCG)7/Y (also referred as Arx polyalanine), harbouring a trinucleotide repeat mutation that expands the first polyAla tract and thus impairs but does not abolish ARX activity. These mice exhibit an apparently normal brain structure but develop severe seizures in young animals (at postnatal days 35 or 40) in a phenotype resembling DEE1 (Table 1) (6,18,19). In ongoing efforts to understand the function of ARX in health and disease conditions, transcriptomic analyses were carried out in embryonic brains of Arx transgenic mice (12,20–22). These studies have shown the crucial role of the cell–cell adhesion pathway and synaptic transmission at the early stage of embryonic development (12,20–22). Very importantly, ARX plays multiple and distinct cell-autonomous roles during corticogenesis in the regulation of cell cycle of neuronal progenitor cells, morphology and radial migration of pyramidal neurons and tangential migration of interneurons (16). Besides, it regulates dorso-ventral identity in the forebrain patterning and growth (23). It is noteworthy that very few pathway biomarkers of the disease at birth have been reported making it difficult to explain at molecular level the basis of XLAG and DEE1—both in patients and mouse models—associated with alterations in ARX with LoF (caused by null mutation) and partial LoF/GoF (caused by polyAla elongation) activities.

Table 1

Summary of the key features of Arx-disease models used in LC–MS/MS study

Name	Mutation	Viability	Brain abnormalities	Epilepsy	Human disease	References
Arx KO/Y	Deletion of exons 1 and 2 of Arx gene	Perinatal death (at postnatal day 1 or 2)	Impaired radial and tangential migration of ventral forebrain neurons;Structural abnormalities in the striatum and pallidum;Decrease in cholinergic interneurons of the striatum and basal telencephalon.	No data	XLAG	(17,18,99)
Arx  (GCG)7/Y	Seven GCG triplets inserted in polyalanine tract 1 of exon 2	Up to 5 months	No gross malformations;Impaired tangential migration of interneurons to the striatum;Loss of cholinergic, striatal and ventral forebrain interneurons.	Spontaneous tonic–clonic seizures (at P30);Sudden death during epileptic attack	DEE1	(19)

Extending further, we have performed for the first time a comprehensive analysis by mass spectrometry (MS)-based label-free quantitative proteomics of neonatal brains of the XLAG-Arx mice (ArxKO/Y) and the DEE1-Arx mice (Arx(GCG)7/Y). Computational examination of the results reveals both overlaps and genotype-dependent differences between the two allelic conditions. Integration with publicly available Arx datasets (RNA-Seq transcriptome and ChIP-on-chip data) implicates several differentially expressed proteins (DEPs) in the neurological disorders. Further functional analysis focusing on top hits validated deregulated molecular pathways in both Arx mutant mice. Furthermore, taking advantage of the evolutionary conserved ARX activity between mammals and nematodes (4), we extended our analysis to the C. elegans alr-1(KO) mutant, ablated for the orthologue of Arx, that shows axon guidance and dendritic defects, touch insensitivity and faulty GABAergic maturation (4,24). Altogether, these findings highlight the pleiotropic consequences of LoF and partial LoF/GoF activities of Arx mutations thus identifying secondary alterations that differentially affect XLAG- and DEE1-Arx brain pups.

Results

Quantitative proteomics uncovers both profound similarities and exclusive features in neonatal brains of ArxKO/Y and Arx(GCG)7/Y mice

The alterations in neonatal brain expression profiles associated with the null mutation in ArxKO/Y and the elongated polyalanine mutation (GCG)7 in Arx(GCG)7/Y were quantitatively determined by label-free quantitative proteomics. Both brain proteomes of ArxKO/Y and Arx(GCG)7/Y pups were compared with XY wild-type (WT) ones. In particular, comparative analysis was performed using five males from each genotype and four matched XY WT mice. Thus, liquid chromatography–tandem mass spectrometry (LC–MS/MS) analysis produced two distinct proteomic datasets, ArxKO/Y and Arx(GCG)7/Y, including 176 and 169 DEPs respectively (Supplementary Material, Tables S1 and S2). The DEPs are graphically represented in volcano plots (Fig. 1A) highlighting interesting proteins that are discussed here. The NSAF protein abundances were able to discriminate between the compared groups for both conditions (Supplementary Material, Fig. S1). Furthermore, the regulated proteomes of the two different genotypes were intersected in order to select common and exclusive features (Fig. 1B). Accordingly, 48 proteins were downregulated in ArxKO/Y datasets and 36 in Arx(GCG)7/Y, exclusively; 13 proteins were commonly down-regulated in ArxKO/Y and Arx(GCG)7/Y. Conversely, 95 proteins were upregulated in ArxKO/Y datasets and 100 in Arx(GCG)7/Y, exclusively; 20 proteins were commonly upregulated in ArxKO/Y and Arx(GCG)7/Y.

Figure 1

Comparative analysis of ArxKO/Y and Arx(GCG)/Y brain proteomes versus XY WT. (A) The differential protein expression profile in brain ArxKO/Y and Arx(GCG)/Y was obtained plotting log2 FC, reported on x-axis, against the statistical significance, reported as –log P-value on y-axis. FC and P-value were calculated using protein quantitative datasets obtained by five independent experiments for ArxKO/Y and Arx(GCG)/Y proteomes and four independent experiments for XY WT. Some selected DEPs are highlighted. (B) The Venn diagram reports common and exclusive deregulated proteins in ArxKO/Y and Arx(GCG)/Y. The common downregulated and upregulated proteins are listed on the right of the figure. (C) Top IPA canonical pathways (left panels) and DAVID Gene ontology terms (right panels) are enriched for both whole neonatal brains of Arx  (KO)/Y and Arx(GCG)/Y. BP, biological processes; CC, cellular component and MF, molecular function.

Top dysregulated canonical pathways and GO terms in ArxKO/Y and Arx(GCG)7/Y

The lists of significant DEPs in the whole brains of ArxKO/Y and Arx(GCG)7/Y were clustered according to their gene ontology (GO) categories, using DAVID. The most interesting altered GO biological process, cellular component and molecular function were reported in Figure 1C (Supplementary Material, Tables S3 and S4). Based on the most significantly enriched IPA canonical pathways (P-value < 1.0E-03), 12 top-ranked terms were related to cytoskeleton, migration and neuronal plasticity in ArxKO/Y (Fig 1C; Supplementary Material, Table S5). Among them, the most significant term results in the remodelling of the epithelial adherens junctions pathway (P-value 9.30E-11), whose main role is to preserve cortical architecture and buffer the mechanical forces exerted by the cortical expansion (25).

In contrast, differential Arx(GCG)7/Y proteome clustered into 10 pathways (P-value < 1.0E-03) (Fig. 1C; Supplementary Material, Table S6) including eukaryotic translation initiation factor-2 (eIF2) signalling (P-value 1.4E-22; Supplementary Material, Table S6), regulation of eIF4 and p70S6K (P-value 8.31E-08), mTOR signalling (P-value 3.56E-06), protein ubiquitination (P-value 1.15E-07) and actin cytoskeleton signalling (P-value 7.85E-05). In addition, splicing of mRNA (P-value 4.26E-11) and processing of RNA (P-value 5.82E-11) were highlighted as significant molecular functions (Supplementary Material, Table S6) in the Arx(GCG)7/Y neonatal brain. These results directly indicate new functions controlled by ARX as protein formation and metabolism, as well as cytoskeleton signalling, are possibly related to the gradual process by which the Arx(GCG)7/Y brain develops seizures. We were surprised to not find specific alterations in the development and functioning of GABAergic neurons or cholinergic neurons that—as previously described—are affected in Arx mutant mice (26–28). Probably due to the analytical sensitivity of untargeted proteomic analysis performed on the whole brain of the Arx mutant mice, we were not able to detect defects affecting a specific brain region or small neuronal populations. Very interestingly, cross-matching the IPA annotation disease tool and OMIM database revealed that 30.1% of DEPs in ArxKO/Y and 27.2% of DEPs in Arx(GCG)7/Y datasets are associated with multiple neurological disorders, including epileptic encephalopathies (EEs), basal ganglia diseases (BGDs) and movement disorders (MDs), all of which present comorbidities overlapping with ARX comorbidities and phenotypes (Supplementary Material, Fig. S2 and Table S7) (29,30). Related to BGDs and MDs, ARX expression is known to be present in neurons of the basal ganglia, a brain area involved in sensorimotor processing and control of precision gripping (29).

Functional enrichment analysis of ArxKO/Y and Arx(GCG)7/Y neonatal brain proteins identified shared and distinct disease-related pathways

The STRING PPI network analysis of the differential ArxKO/Y dataset generated microtubule cytoskeleton (GO_CC:0015630), actin cytoskeleton (GO_CC:0015629) and spliceosome (KEGG:mmu03040) (Fig. 2A) as the main clusters. The relative abundance values of proteins involved in these hits are represented in the heat maps (Fig. 2B). In particular, DEPs as tubulin alpha 1a and 4a (TUBA1A and TUBA4A), tubulin beta 2A class IIa and IIb (TUBB2A and TUBB2B), tubulin beta 3 Class III (TUBB3), tubulin beta 6 Class V (TUBB6), microtubule-associated protein RP/EB family member 1 (MAPRE1), actin-related protein 2/3 complex subunit 5 (ARPC5), actin-related protein 2/3 complex subunit 1A (ARPC1A) and the member RAS oncogene family 5A (RAB5A) are biomarkers for microtubule (MT) and actin–cytoskeleton regulation (Supplementary Material, Table S5 and Fig. S3A) (31–33). In addition, nine DEPs, including the splicing factors poly(U) binding splicing factor 60 (PUF60) and non-POU domain-containing octamer-binding protein (NONO), belong to the spliceosome (Supplementary Material, Table S5; Fig. 2B).

Figure 2

Functional enrichment analysis in ArxKO/Y brain proteome. (A) 176 deregulated proteins in ArxKO/Y were enriched for molecular processes in STRING (Search Tool for the Retrieval of Interacting Genes software) using 0.9 as the interaction score setting. Significantly enriched GO:BP, GO:CC and KEGG pathways related to cytoskeleton, locomotion, synapse structure and spliceosomes are highlighted. Each node represents a deregulated protein. (B) Relative abundance of splicing related proteins (left panel) and cytoskeleton proteins (right panel) in ArxKO/Y and WT XY proteomes is depicted in the heatmap where red and blue colours mean higher and lower protein abundance, respectively.

The STRING PPI network analysis of differential Arx(GCG)7/Y dataset revealed significant enrichment in proteasomes (KEGG:mmu03050), RNA binding (GO_MF:0003723)/spliceosomes (KEGG:mmu03040), translation (GO_BP:0006412)/ribosome (KEGG:mmu03010) and cytoskeleton (GO_CC:0005856)/intracellular transport (GO_BP: 0046907) (Fig. 3A). The relative abundance values of proteins involved in these hits are represented in the heat maps (Fig. 3B). Interestingly, in Arx(GCG)/7Y, 11 DEPs belong to the proteasome complex, including the regulatory particle and α - and β -subunits of 26S (Supplementary Material, Fig. S3B). Conversely, 17 DEPs, including the splicing factors PUF60 and KH RNA-binding domain containing, signal transduction-associated 1 (KHDRBS1; also named SAM68), belong to the spliceosomes (Fig. 3B; Supplementary Material, Table S2). Furthermore, 24 DEPs, including eukaryotic translation initiation factor 4A2 (eIF4A2), are involved in ribosomal activity and translation initiation (Fig. 3B; Supplementary Material, Fig. S3C) (34). Finally, 22 DEPs are involved in cytoskeleton organization/intracellular transport (Fig. 3B). The proteomic data suggest common alterations in pathways, such as cytoskeleton remodelling and spliceosomes, affecting both the analysed Arx-related genotypes, as a consequence of lost or impaired ARX activity. Of note, in Arx(GCG)7/Y, there were exclusive alterations in proteasomes and translation/ribosomes that presumably result from the action of the Arx polyalanine-elongated protein.

Figure 3

Functional enrichment analysis in Arx(GCG)7/Y brain proteome. (A) 169 deregulated proteins in Arx(GCG)7/Y were enriched for molecular processes in STRING (Search Tool for the Retrieval of Interacting Genes software) using 0.9 as the interaction score setting. Significantly enriched GO:BP, GO:CC and KEGG pathways related to proteasome, spliceosomes, ribosomes and cytoskeleton are highlighted. Each node represents a protein. (B) Relative abundance of DEPs belonging to proteasomes, RNA binding/spliceosomes, ribosome/translation and cytoskeleton/intracellular transport in Arx(GCG)7/Y and WT XY proteomes is depicted in heatmaps where red and blue colours mean higher and lower protein abundance, respectively.

Arx KO/Y and Arx(GCG)7/Y DEPs encoded by ARX-bound genes and expressed in the embryonic brain

As ARX is a central TF of mammalian brain development, we investigated which DEPs in ArxKO/Y and Arx(GCG)7/Y are encoded by potential ARX-target genes (35). Coupling publicly available data from chromatin immunoprecipitation followed by hybridization on mouse promoter arrays (ChIP-on-chip) and proteomic results, we identified nine and seven ARX-bound genes, whose encoded proteins are respectively DEPs in ArxKO/Y and Arx(GCG)7/Y (Fig. 4A; Supplementary Material, Table S8). Remarkably, among the nine proteins altered in ArxKO/Y mice (Fig. 4A; Supplementary Material, Table S8), three (i.e. TUBA1A; vesicle amine transport 1, VAT1; and vesicle transport through interaction with T-SNAREs 1B, VTI1B) are involved in cytoskeleton remodelling and synaptic vesicle activities (36–38). Similarly, in Arx(GCG)7/Y, three (i.e. Profilin-2, PFN2; synaptic vesicle glycoprotein 2A, SV2A; and sodium/calcium exchanger 1, SLC8A1) are involved in cytoskeleton and vesicle trafficking (Supplementary Material, Table S8) (39–41). Moreover, a very intriguing ARX target is IF4A2 (alias of eIF4A2), an ATP-dependent RNA helicase in the eIF4 complex that has been identified as critically involved in other trinucleotide repeats disorders (34,42). Finally, RAB5A, involved in cytoskeleton and vesicle trafficking (32), and superoxide dismutase (SOD1), involved in superoxide radical metabolism (43,44), were identified as deregulated in both proteomic datasets (Supplementary Material, Table S8).

Figure 4

Defective levels of tubulin isoforms in Arx-disease models. (A) Venn diagram showing intersections of our proteome datasets with ChIP-on-chip ARX-bound genes and transcriptome data of E14.5 Arx-ablated cortices. Common and distinct DEPs are highlighted including those involved in cytoskeleton and vesicle trafficking (35,45). ∩ represents the intersection of gene sets. (B) Detection of α-tubulin levels in Arx(GCG)7/Y and ArxKO/Y whole brains. Western blot (upper panel) and band quantification (bottom panel) analysis. (C-D) Detection of acetyl α-tubulin expression levels in Arx(GCG)7/Y and ArxKO/Y whole brains and cortex. Western blot (upper panel) and band quantification (bottom panel) analysis. The western blotting experiments were repeated with five WT XY and four Arx(GCG)7/Y and four ArxKO/Y brain samples. As a loading control, β-actin was used. The band quantification of α-tubulin and acetyl α-tubulin was performed with the ImageJ 1.50i software. Columns represent mean ± standard deviation of three independent experiments per genotypes. Student’s t-test was applied with *P < 0.05 **P < 0.01.

As the pathophysiology of XLAG and DEE1 begins early in embryogenesis, we analysed publicly available transcriptome data of E14.5 XY WT and E14.5 Arx-ablated cortices (GSE12956) (20,45) to detect DEPs encoded by early expressed genes (Supplementary Material, Fig. S4A). Relevant to ArxKO/Y, we found that most of the genes in epithelial adherens junction signalling, cytoskeleton remodelling, synaptogenesis and axonal guidance signalling were expressed early in embryogenesis, including the ARX target genes Tuba1a, Rab5a and Vit1b. Moreover, we found that a few of them were significantly deregulated in Arx-ablated cortices (Supplementary Material, Table S8 and Fig. S4A and B). Similarly, in Arx(GCG)7/Y, we noted that genes encoding several DEPs involved in proteasome, RNA binding/splicing and translation/ribosome pathways were already expressed at the E14.5 neocortex (Supplementary Material, Table S8 and Fig. S4C and D). However, consistent with partial LoF activity of polyAla elongations, we did not observe any ARX target gene within the main pathways and cellular processes perturbed in Arx(GCG)7/Y brains. Based on these findings, we conclude that although most of the genes encoding the DEPs identified in our neonatal brain datasets are expressed at an early stage of corticogenesis, only a few are deregulated in the Arx-KO neocortex. These data might mirror distinct selective activity times of ARX in the developing brain.

Altered tubulin isotypes levels in Arx- mice and C. elegans alr-1(KO) animals

Having established by proteomic investigation a marked decrease of several tubulin isotypes in both neonatal brains, we performed western blotting to assess the total amount of α-tubulin in the ArxKO/Y brains with microcephaly and the Arx(GCG)7/Y brain with apparently normal size. In line with the proteomic datasets, a robust decrease of α-tubulin was observed in ArxKO/Y and Arx(GCG)7/Y compared to WT animals (Fig. 4B). In addition, by testing the level of the acetyl α-tubulin fraction, a severe decrease was observed in whole neonatal brains and cerebral cortices in both types of Arx mutant mice (Fig. 4C and D). In further work, we took advantage of the conservation of the ARX pathway between mammals and nematodes (4), and we extended our analysis to the C. elegans alr-1(KO) mutant. Ablated for the orthologue of Arx, alr-1(KO) animals show abnormal dendritic structures, axon guidance defects, touch insensitivity and faulty GABAergic maturation (4,24). There was a pronounced decrease in the total content of α-tubulin in alr-1(KO) compared to WT animals (Supplementary Material, Fig. S5A) suggesting a conserved role in the evolution of ARX/ALR-1 in regulating tubulin expression. Notably, C. elegans expresses nine α-tubulin isotypes (MEC-12, TBA-1, TBA-2, TBA-4 through TBA-9) involved in neurite growth, mechano- and chemo-sensation (46). In the modENCODE database, we found that two of them, mec-12 and tba-9, are annotated as validated ALR-1-binding genes (Supplementary Material, Fig. S5B) (47). Very suggestively, mec-12, an essential gene for the generation and maintenance of the synaptic branch, is the orthologue of ARX-target gene Tuba1a (46,48); while tba-9, a gene involved in sensory cilia, is the orthologue of Tuba4a (49). Collectively, these findings strongly support the inference that microtubule cytoskeleton is a conserved structural network deeply damaged in Arx mutants. This could—at least partially—explain the neuronal abnormalities detected in ArxKO/Y and Arx(GCG)7/Y, as well as those observed in C. elegans alr-1(KO) mutants.

Disorganization of neurite network in ArxKO/Y and Arx(GCG)7/Y cortical neurons

As α-tubulin is central to microtubule structure and functions, we investigated the physiological consequence of the defective α-tubulin content. One of the conditions in which microtubule dynamics is functionally deregulated due to α-tubulin loss is the disorganization of the neurite network. To test this, we plated 1000 cells/mm2 of primary cortical cells for WT, ArxKO/Y and Arx(GCG)7/Y genotypes. Confocal microscopy on Arx(GCG)7/Y and ArxKO/Y cortical neurons (DIV10) labelled with β3-tubulin revealed a drastic alteration of the neurite network in both groups of mice although less severe in Arx(GCG)7/Y animals (Fig. 5A; Supplementary Material, Fig. S5C). Global disorganization of the network is obvious, with neurites entangled and randomly orientated in different directions (Fig. 5A; Supplementary Material, Fig. S5C) suggesting potential defects in axon guidance. Moreover, neurites are shorter and show aberrant variable thickness (some much thinner and others thicker in the region of cone growth) compared to the XY WT counterpart (Fig. 5A; Supplementary Material, Fig. S5C). All these findings indicate that neurite growth and arborization are damaged in both mice and, again, in a more severe form in ArxKO/Y mice. Consistent alteration of tubulin distribution/levels is clearly evident in Arx(GCG)7/Y and ArxKO/Y cortical neurons compared to XY WT in the intensity map (Fig. 5A; lower panels). Comparable results were obtained in immunofluorescence experiments carried out for the microtubule-associated protein 2 (MAP2): ArxKO/Y neurons have fewer neurites per nucleus and some are longer than the control XY cells (Fig. 5B and C; Supplementary Material, Fig. S6). Conversely, a less severe decrease in the number of neurites per nucleus is seen for Arx(GCG)7/Y neurons (Fig. 5B and C, Supplementary Material, Fig. S6). Collectively, these data indicate that MT dynamics and neurite network are affected in both Arx allelic disorders, with an effect that is more pronounced in Arx-XLAG mice that present microcephaly than in Arx-DEE1 mice with apparently normal brain size.

Figure 5

Neurite network in Arx(GCG)7/Y and ArxKO/Y primary cortical neurons. (A, B) Representative immunofluorescence of Arx(GCG)7/Y and ArxKO/Y primary cortical neurons cultured in NB. DIV 10 cells were fixed and stained with β-III tubulin (red) and DAPI (blue) (A) and with MAP2 (green) and DAPI (blue) (B). (C) Quantification of the number of primary neurites in XY WT (N = 57 cells), ArxKO/Y (N = 53 cells) and Arx(GCG)7/Y (N = 62 cells) primary cortical neurons from two independent experiments. Whisker plots represent median with min to max values. One-way ANOVA with Dunn’s multiple comparisons was applied with *P < 0.05. In (A), images were acquired with a confocal microscope taking Z-slices from the top to the bottom of cells; the 3D reconstruction and the corresponding intensity maps are shown. Scale bars represent 5 μm in (A) (original magnification ×20); 50 μm in (B) (original magnification ×40); and 10 μm in (C) (original magnification ×100). Arrows indicate primary neurites.

Arx (GCG)7/Y mice show overexpression of eIF4A2 and evidence of impaired translation capacity

Translation is a specific cellular function that is predicted by IPA and GO analysis to be disrupted in Arx(GCG)7/Y mice brains (Figs 1 and 3). This association, reminiscent of the deficit of translation in other nucleotide repeat expansion diseases (34,50), suggests a convergent feature among them. Noteworthily, eIF4A2, which is a critical component of the CAP-dependent translation machinery particularly important in neurons (Fig. 6A) (42) was found upregulated in the Arx(GCG)7/Y proteome dataset (+ 0.4 fold; P-value 1.17E-02; Supplementary Material, Table S2). Thus, by western blotting we assayed the relative protein content of eIF4A2 in the whole brain and in three neonatal regions of the brain—cortex (CX), hippocampus (HP) and striatum (STR) reported as epileptogenic areas in Arx(GCG)7/Y (19). We also analysed the eIF4A2 content in the cerebellum (CB), which is a region that does not express Arx at any developmental stage and here used as a negative control of ARX activity (51). We found a marked eIF4a2 increase in whole brain and cortex whereas no apparent changes were observed in other brain areas analysed (Fig. 6B and C). We then asked whether the eIF4a2 increase was associated with an effect on protein translation. The surface sensing of translation (SUnSET) assay of global protein synthesis revealed a clear decrease in the total level of de novo proteins in Arx(GCG)7/Y primary cortical neurons compared to XY WT control neurons (Fig. 6D). Consistent with previous findings on expanded GC repeats in C9orf72 and FMR1 genes (50), expanded GC repeats in the Arx gene could induce translational suppression although further studies are needed to deeply define the molecular basis of these phenomena.

Figure 6

Analysis of eIF4A2 levels in Arx(GCG)7/Y brain and of global protein synthesis in Arx(GCG)7/Y primary neurons. (A) Schematic representation of translation initiation. eIF4A2 is an RNA helicase ATP-dependent protein encoded by the Arx target gene Eif4a2 involved in the first step of translation. (B, C) Detection of eIF4A2 expression levels in whole brain and dissected areas of Arx(GCG)7/Y. Western blot (upper panel) and band quantification (bottom panel) analysis. As loading control, Hsp90 antibody was used. Student’s t-test was applied with *P < 0.05. Schematic depiction of mouse brain regions used for eIF4A2 expression analysis is shown. CX, cortex; HP, hippocampus; STR, stratum; CB, cerebellum. (D) Western blotting-SUnSET analysis of primary cortical neuron lysates from XY WT and Arx(GCG)7/Y untreated and puromicyn treated cells. Equal loading was confirmed with coomassie blue staining. Quantitative representation of puromycin incorporation is expressed as the fold change relative to XY WT cells. n = 3 neuronal cell cultures per genotype. Data are presented as mean ± SD and Student’s t-test was applied with *P < 0.05.

Alterations of splicing switches in Arx- mice and C. elegans alr-1(KO) animals

Going into the details of IPA analysis, with reference to RNA splicing in ArxKO/Y and Arx(GCG)7/Y datasets, 8 and 15 DEPs were identified respectively (Fig. 7A). Very interestingly, previous studies carried out in C. elegans alr-1(KO) mutants revealed a role of ARX/ALR-1 in single-neuron gene splicing (52). Based on this evidence, we asked whether alternative splicing (AS) regulation might be a conserved ARX function. Initially, by quantitative western blotting analysis, we validated the decrease of PUF60, a common downregulated transactivating splicing factor in Arx(GCG)7/Y (− 4.4 fold; P-value 1.61E-02; Supplementary Material, Table S1) and ArxKO/Y (−2.2 fold; P-value 5.53E-02; Supplementary Material, Table S2) neonatal whole brain (Supplementary Material, Fig. S7A). Next, to investigate the influence of PUF60 decrease on AS, we assayed the abundance of PUF60-activated AS for three genes—hnRNPD, hnRNPC and hnRNPK—which in turn control alternative and constitutive pre-mRNA splicing (53). By semiquantitative PCR, we tested the two AS isoforms e6(+) and e6(−) of hnRNPD, generated by the skipping or inclusion of exon 6 respectively; the two AS isoforms e14b(+) and e14b(−) of hnRNPK, generated by the skipping or inclusion of exon 14b respectively; and the two AS isoforms e4(+) and e4(−) of hnRNPC, generated by the skipping or inclusion of exon 4 respectively (Fig. 7B). There were no differences in the AS ratio in the whole brain of DEE1 mice compared to the XY mice (Supplementary Material, Fig. S7B), but we detected an increase of hnRNPD e6(+)/e6(−) and of hnRNPC e4b(+)/e4b(−) AS ratio in HP whereas no apparent changes in the hnRNPK e14b(+)/e14b(−) AS ratio were observed (Fig. 7C and D). In addition, no changes were observed in CB samples used as negative controls of ARX activity.

Figure 7

Network of mRNA splicing in Arx mutant mice and analysis of splicing events in hnRNP genes. (A) Network of mRNA splicing in ArxKO/Y and Arx(GCG)7/Y datasets were generated through IPA (QIAGEN Inc., https://www.qiagenbioinformatics.com/products/ingenuity-pathway-analysis/). (B) Schematic diagrams outlining organization of splicing variants of hnRNPD (exon 6), hnRNPK (exon 14b) and hnRNPC (exon 4b) are shown. Arrows indicate the positions of oligonucleotides used in transcript analysis. (C) Semiquantitative PCR analysis with primers for hnRNPD AS exon 6 (e6+), hnRNPK AS exon 14b (e14b+) and hnRNPC AS exon 4b (e4b+) in dissected brain areas of Arx(GCG)7/Y. (D) Ratio of band quantification of AS e(+)/AS e(−) is plotted. Columns represent mean ± SD of three independent tissues samples per genotype. Student’s t-test was applied with *P < 0.05. CX, cortex; HP, hippocampus; STR, stratum; CB, cerebellum.

As alternative splicing governs mammalian brain development, we further corroborated experimentally our proteomic data analysis by verifying the impact of the increase of SAM68 found upregulated in the Arx(GCG)7/Y dataset. For this purpose, we took advantage of the fact that SAM68 directly binds RNA recognition motifs in the introns flanking the highly conserved alternative exon at splice site 4 (AS4) of pre-mRNAs for Neurexin-1 (Nrxn1) (54). Remarkably, Nrxn1—together with Nrxn2 and Nrxn3 genes—belongs to the Neurexin gene family involved in epilepsy, autism and schizophrenia (54). A genetic feature of the Neurexin family of genes is the extensive alternative splicing they undergo generating distinct transcription isoforms encoding a wide range of presynaptic receptors involved in synapse remodelling. In particular, incorporation or skipping of exon 22 at AS4 generates multiple NRXN AS4(+) and NRXN AS4(−) variant proteins (55,56). These isoforms exhibit differential interactions with several ligands mediating synaptogenesis including neuroligins (NLGS) and leucine-rich repeat proteins (Fig. 8A) (55,56). Although SAM68 specifically regulates AS4 of Nrxn1, we also analysed AS4 of Nrxn2 and 3 to obtain more information about the splicing control of this family given its involvement in neurological disorders. By semiquantitative PCR assay, we tested the abundance of AS4(+) and AS4(−) of Nrxn1, 2 and 3 in whole brain and dissected areas of Arx(GCG)7/Y mice. We detected a significant increase in the Nrxn2 AS4(+)/Nrxn2 AS4(−) ratio but no differences in the Nrxn1 AS4(+)/Nrxn1 AS4(−) and Nrxn3 AS4(+)/Nrxn3 AS4(−) ratio were found in the whole brain of DEE1 mice compared to the XY mice (Supplementary Material, Fig. S7C). More importantly, in the dissected Arx(GCG)7/Y areas there was a significant higher level of Nrxn1 AS4(+)/Nrxn1 AS4(−) ratio in CX and HP but decreased in STR (Fig. 8B and C). In line with the whole brain data, a significant higher level of Nrxn2 AS4(+)/AS4(−) ratio was found in HP; and finally, Nrxn3 AS4(+)/AS4(−) ratio was significantly higher in CX (Fig. 8B and C).

Figure 8

Analysis of Neurexins alternative splicing in Arx(GCG)7/Y and ArxKO/Y mice and in C. elegans alr-1(KO). (A) Schematic representation of exon 22 splicing in murine Nrxn1/2/3 genes. Arrows indicate the positions of oligonucleotides used in transcript analysis. (B) Semiquantitative PCR analysis with primers to detect neurexin1/2/3 AS4 in Arx(GCG)7/Y dissected areas. (C) Ratio of band quantification of AS 4(+)/AS 4(−) is plotted. Columns represent mean ± SD of three independent tissues samples per genotype. CX, cortex; HP, hippocampus; STR, stratum; CB, cerebellum. Student’s t-test was applied with *P < 0.05. (D) Schematic representations of exons 25 and 10 splicing in the C. elegans nrx-1 gene. Arrows are primers used for semiquantitative PCR analysis. (E) Semiquantitative PCR analysis with ratio of bands to quantify the alternative isoforms of nrx-1 e25(+), e25(−) (left panel), e10(+) and e10(−) (right panel) in alr-1(KO) L4 larvae mutants. PCR amplicons were run on 2.5% agarose gel, densitometric analysis was performed with the ImageJ software and quantification of the ratio of nrx-1 e(+)/(−) abundance was plotted. Student’s t-test was applied with *P < 0.05.

Given the crucial role of Neurexins in synaptogenesis and that several DEPs of ArxKO/Y dataset are involved in RNA splicing control, we asked whether CX, HP and STR of XLAG mice might present an altered AS profile of Neurexins. Surprisingly, we noted only a significant decrease of Nrxn3 AS4(+)/AS4(−) ratio in the control brain area of CB (Supplementary Material, Fig. S7D). Altogether, these findings suggest that changes of Neurexins repertoire in the epileptogenic areas of Arx polyalanine mice are strictly genotype dependent and can be considered a secondary effect of the polyalanine-elongated variant. Nevertheless, why we detected the aberrant ratio of Nrxn3 isoforms only in CB of ArxKO/Y remains to be explored. One possible explanation is that the null allele of Arx could have a direct or indirect impact on downstream diffusible long-range effectors—secreted outside the ARX-expressed areas—with a key role in CB plasticity. However, a proper dynamic splicing program of Neurexins is required to control synaptogenesis in cerebellar neurons (57).

Likewise in mammals, the single C. elegans orthologue of mammalian Neurexins, nrx-1, controls synaptic remodelling (58) and encodes several isoforms that are temporally controlled (54,59). Although C. elegans lacks a SAM68 orthologue, a sequence homologous to Nrxn1 splice acceptor sequence AS4, which is recognized by SAM68, has been identified in nrx-1 intron 24 (Supplementary Material, Fig. S8) (60). We assayed the two nrx-1 transcript isoforms nrx-1 e25(−) and nrx-1 e25(+) generated by the skipping or inclusion of exon 25 of nrx-1, respectively (Fig. 8D). There was a high level of nrx-1 e25(−) coupled with a drastic decrease of nrx-1 e25(+) in alr-1(KO) mutants and a significantly decreased ratio of the band quantification of nrx-1 e25(+)/nrx-1 e25(−) (Fig. 8E left panel). We furthered the analysis by testing two nrx-1 transcript isoforms nrx-1 e10(−) and nrx-1 e10(+), generated by the skipping or inclusion of exon 10 that is not surrounded by splicing sequences homologous to the SAM68 binding site (Fig. 8D). We observed no change in the nrx-1 e10(+)/nrx-1 e10(−) ratio in the alr-1(KO) animals compared to WT (Fig. 8E; right panel). We therefore suggest that alr-1 has an exon-specific role in regulating nrx-1 e25 isoforms. In contrast, a different molecule or mechanism is the basis of generation of the nrx-1 e10 isoforms. Collectively, our findings point to a new and unanticipated evolutionary conserved Arx-dependent mechanism accounting for alternative splice regulation choice of the Nrxn gene. As NRXN1 AS4(+) and AS4(−) isoforms exhibit differential adhesion/binding with NLGs (e.g. neuroligin 1B postsynaptic receptor weakly binds AS4(−) but AS4(+) strongly) (59), we infer that the expression of an altered Neurexin-1 repertoire might lead to a synaptopathy due to improper axo-dendritic interactions.

Neurexin AS4 splicing in Arx(GCG)7/Y depolarizing cortical neurons

Next, we asked whether an altered Neurexin repertoire might impair the neuronal activity in Arx(GCG)7/Y, the transgenic mouse with spontaneous recurrent seizures. As a high K+-induced depolarization of the neuronal membrane induces a change of synaptic plasticity through an increase of Nrxn1 AS4(+)/AS4(−) ratio (Fig. 9A) (54), we analysed the physiological consequences of this phenomenon in the Arx(GCG)7/Y primary cortical neurons (Fig. 9B). First, primary neuron cultures from Arx(GCG)7/Y and XY WT pups were depolarized with elevated KCl. Successively, we carried out transcript analysis of Nrxn-1 AS4 isoforms and of c-fos whose increase is generally used to validate induction of synaptic plasticity in response to a stimulus (Fig. 9C) (54). As expected, KCl-treated XY WT neurons exhibited a significant upregulation of c-fos and of Nrxn1 AS4(+)/AS4(−) ratio compared to resting (not-treated) XY WT cells (Fig. 9C–E). Conversely, there were reduced levels of c-fos and of Nrxn1 AS4(+)/AS4(−) ratio both in resting and KCl-treated Arx(GCG)7/Y cortical neurons compared to the control cells (Fig. 9C–E). Expanding our analysis to AS4 of Nrxn 2 and 3, there was only a modest increase of Nrxn2 in Arx(GCG)7/Y KCl-treated cortical neurons compared to the resting cells (Fig. 9D and E). In sum, these findings depict a pathogenetic feature that might affect the electrical activity of Arx(GCG)7/Y neurons and thus contribute to the DEE1 phenotype.

Figure 9

Changes of neurexin 1/2/3_AS4 splicing induced by depolarization in Arx(GCG)7/Y primary cortical neurons. (A) Model of the neurexin AS4(+) and AS4(−) isoform splicing stimulated by neuronal activity via SAM68. High K+-induced depolarization triggers a shift in the alternative splicing within cultured neuronal cells via Ca2+/calmodulin-dependent kinase IV (CaMKIV) (54) that is thought to activate the splicing factor SAM68. Specifically, this RBP activates a shift in the alternative splicing of neurexin pre-mRNA that contains the SAM68-consensus recognition motif (in red). Constitutive levels of Nrxn AS4(+) are low in primary cortical neurons and increase upon depolarization. (B) Schematic depiction of isolation of primary cortical neurons from Arx(GCG)7/Y and XY WT pups. (C) Quantitative real-time PCR of c-fos in resting (not-treated) and depolarized neurons treated with NaCl or KCl for 10 min. Two-way ANOVA with Tukey’s multiple comparisons test was applied with *P < 0.05, **P < 0.001. (D) Semiquantitative RT-PCR of alternative isoforms of Nrxn 1/2/3_AS4 in resting and NaCl- or KCl-depolarized cortical neurons. (E) Ratio of band quantification of AS 4(+)/AS 4(−) is plotted. Columns represent mean ± SD of three independent tissues samples. Two-way ANOVA with Tukey’s multiple comparisons test was applied with *P < 0.05.

Discussion

We report an extensive proteomics analysis to reach a consensus view of changes in brains of XLAG and DEE1 mouse models compared to wild-type. We find that most of DEPs and enriched pathways are consistent with pleiotropic defects reflecting overlaps between the two ARX-endophenotypes as well as genotype-dependent differences (Fig. 10). This permitted us to identify functional alterations affecting microtubule network and RNA metabolism, some of them related to defects previously described in ARX patients and mouse models and also involved in other neurological disorders (29,30,33,54).

Figure 10

Secondary ARX-related alterations highlighted in Arx-allelic disorders and involved in macro- and microstructures of the developing brain.

Microtubule network deregulation One of the most interesting observations of our study concerns several DEPs related to microtubule (MT) structure and functioning found in Arx-XLAG and DEE1 pup brains. MTs are built of α and β-tubulin heterodimers that self-assemble head-to-tail into protofilaments and are required to establish correct brain development and network connectivity (30,33,61–63). In both Arx- mice models we detected a robust decrease of α-tubulin global content and α-tubulin acetylation associated with defective neurite growth and arborisation (37,64–66). Furthermore, the clear reduction of α-tubulin in the C. elegans Arx/alr-KO mutant model and the evidence that two α-tubulin isotype genes, mec-12 and tba-9, are ALR-1 targets suggest a conserved role of ARX/ALR-1 in regulating tubulin expression. These data allow us to discover a new disease phenologue pathway in addition to the one we identified earlier (4). Notably, MTs provide a mechanical force for the remodelling of adherens junctions (AJs) and cell–cell adhesions: the former is the most perturbed pathway in ArxKO/Y and the latter is one of main biological process damaged in Arx(GCG)7/Y neonatal brain. As AJs and cell–cell adhesions ensure the fidelity of neural network development, failures in their assembly and disassembly could underlie the defective neurogenesis and synaptogenesis detected in Arx disease mice (25). Moreover, cytoskeleton, migration and neuronal plasticity—which are among the top-ranked terms identified in our proteomic datasets—have been functionally linked to the anatomic defects and neuronal migration alterations previously described in ARX patients (29,67–71).

Nevertheless, these functions were also linked to the aberrant corticogenesis and defective neuronal maturation detected in Arx mutant brains (17,69,72,73). Related to these evidences, the decreased amount of α-tubulin observed by us in ArxKO/Y and Arx(GCG)7/Y brain could explain the defective connectivity and axonogenesis detected—even if at different level of severity—in Arx mutant mice and ARX patients (17,69,72).

Furthermore, this hypothesis is reinforced by our findings in C. elegans mutants. Indeed, the abnormal dendritic structures previously observed in three different classes of alr-1 KO neurons, which are involved in olfaction, thermo-sensation and chemo-sensation (24), could be the consequence of an altered dendrite growth caused by an insufficient amount of α-tubulin. Remarkably, mutations in the human counterpart of several tubulin isotypes found deregulated in Arx mice (i.e. TUBA1A, TUBA4A, TUBB2A and TUBB2B) lead to a wide spectrum of severe brain malformations known as ‘brain tubulinopathies’ presenting lissencephaly and basal ganglia defects (33,62,63). These clinical features overlap with similar anatomic defects described in ARX patients (10,29,68) suggesting that these NDDs share common comorbidities and thus disease hallmarkers.

No less important is the defective level of acetylated α-tubulin detected in both conditions that, as already proved, alter axon branching and neuronal microcircuitry (74–77). We have no evidence that a faulty process common to both mouse models directly compromised acetylation levels, but one can envisage a negative regulatory feedback loop that could control the levels of acetylated-α-tubulin. About the morphological changes of neurite network observed in the primary cortical cultures of both Arx mutants, we infer that they could be the resultant of multiple defects affecting microtubule assembling. These results are in line with the previous findings on the cell-autonomous activity of ARX into regulating the morphology of pyramidal cells (16) that are the main component of primary cortical cells (78). We therefore conclude that the two Arx endophenotypes analysed are two distinct secondary brain tubulinopathies whose pathogenicity depends on the extent of MT impairment: more extensive in ArxKO/Y and relatively constricted in Arx(GCG)7/Y. This conclusion correlates with the poorly developed perinatal ArxKO/Y brain (17,18) and the abnormal axonal arborization and hyperexcitability of Arx(GCG)7/Y mice (27).

RNA metabolism deregulation Understanding the cellular and molecular pathophysiology of Arx polyalanine mice remains a challenge. In particular, to establish whether protein aggregation might increase the severity of the phenotype because it adds a possible gain-of-function to a partial loss of function mechanisms could be relevant to identify appropriate therapeutics. Previous studies carried out by us and others have shown that expansions of polyAla stretches found in DEE1 patients lead to intranuclear protein inclusions consequent of protein misfolding and aggregation (2,3,15). Our brain proteome data for Arx-DEE1 mice provide the first evidence that pathways strongly interconnected with the integrated stress response (ISR), a cellular process that controls protein folding (79–83) and involved in trinucleotide repeat expansion disorders (84,85), are highly affected. Indeed, eIF2 signalling is the most enriched pathway in Arx(GCG)7/Y neonatal brain, along with the regulation of eIF4 and p70S6K and mTOR signalling (79–83). Suggestively, one of the DEPs involved in these processes is eIF4A2, a subunit of the eIF4F complex required for mRNA binding to ribosomes (42,82,86,87). It is encoded by an ARX target gene that is highly expressed at the early stage of embryonic brain development. In addition, the eIF4A2 overexpression coupled with the translational suppression detected in Arx(GCG)7/Y overlaps with similar findings reported in other neurological repeat expansion disorders (34,50). Finally, excessive levels of eIF4A2 have been reported acting as a memory repressor blocking long-lasting forms of synaptic plasticity (82,83,88), a manifestation that correlates well with the behavioural profile of Arx(GCG)7/Y mice (19). Moreover, translation and proteasome function as well as protein ubiquitination—which are among the top-ranked terms identified in Arx polyalanine brains—have been previously related to the protein aggregation caused by ARX polyalanine elongations (15). These results reinforce our hypothesis on the gain-of-function effect of Arx (GCG)7 mutation. Although further studies focused on GCG repeats in ARX are needed, we therefore conclude that misfolded Arx polyalanine elongations might activate a pathogenic mechanism affecting the translation of mRNA.

Another essential mechanism of RNA metabolism affected in Arx polyalanine brains is the control of alternative splicing (AS), a complex phenomenon widely observed at high frequency during brain development, whose misregulation affects synapse plasticity and promotes neuronal hyperexcitability (89–92). This is a current active area of research. The enrichment of DEPs involved in AS detected in Arx-XLAG and Arx-DEE1 pup brains reveals an unexpected molecular scenario that links ARX pathology to AS mis-regulation. As proof-of-concept, we detected defective splicing patterns in a subset of pre-mRNAs that are specific targets of the two DEPs PUF60 and SAM68. Those two splicing factors are both involved in human NDDs. PUF60, a regulatory player of the 3-prime splice site recognition, is encoded by a gene found mutated in patients presenting developmental delay, microcephaly, seizure and skeletal abnormalities (Verheij syndrome, VRJS; MIM 615583) (89). SAM68 is a regulator of activity-dependent AS of Neurexin-1, a gene encoding multiple isoforms of the presynaptic receptor Neurexin-1 and whose mutations have been repeatedly associated with ID, epilepsy, autism and schizophrenia (44,93,94). In Arx(GCG)7/Y hippocampus we detected an aberrant splicing ratio of two PUF60 targets hnRNPD and hnRNPC encoding variant isoforms of these two hnRNPs with many effects on macromolecular dynamics (i.e. DNA replication, transcription and chromatin remodelling); RNA (i.e. mRNA stability and RNA splicing); and protein levels (translation and degradation) (95,96). How these AS abnormalities might impact physiologically on ARX-dependent physiology and contribute to XLAG and DEE1 disease phenotypes is not clear. Instead, even though progress has been made in the understanding of hnRNP action, its functions in brain deserve further investigation. More interestingly, aberrant Nrxn AS4(+)/Nrxn AS4(−) ratios of the SAM68 target Nrxn1 were observed in the epileptogenic niches of Arx(GCG)7/Y mice (cortex, hippocampus and striatum), where seizures jointly start in young animals (19). As these brain areas differ both in ARX expression and neuronal subtype composition (51,97), further studies are required to establish whether the defective NRXN1 repertoire might alter the cell fate in the developing cerebral areas. We therefore infer that ARX is involved in the AS programs of Nrxn1 in important brain areas of DEE1 mice.

A further confirmation of the crucial role of ARX in AS was obtained in alr-1 (KO) worms, in which an altered AS ratio of the Nrxn 1 orthologue was found at the predicted SAM68-binding site. These data, together with the recent findings on the regulatory role of ARX/ALR-1 in splicing in a single C. elegans neuron (52), suggest that ARX is an evolutionarily conserved regulator of AS.

Very important, in addition to participating in the specification of cellular identity, the Nrxn splice repertoire—through dynamic regulation—changes in response to neuron signalling (54,98). Specifically, strong pharmacological or electrical stimulation can induce shifts in the Nrxn AS4(+)/Nrxn AS4(−) ratio (54,98). In the present study, we have discovered a defective increase of AS4(+)/AS4(−) ratio of Nrxn1 associated with a marked defective plasticity of Arx(GCG)7/Y cortical neurons, as evidenced by faulty expression of the firing activity marker c-fos. Considering that neurexins are central regulators of synapse properties, it is reasonable to link the faulty neuronal response and defective splicing. This adds to a substantial body of literature findings that deficiency in the neurexin repertoires can induce behavioural alterations and hyperactivity in rodents (55). Overall, our data explain much of the catastrophic effect of faulty ARX action and shed light on a new function of ARX in controlling AS. This interesting finding opens up further studies aimed at defining the temporal regulation of AS in brain areas of Arx mutant mice. To this end, single-cell RNA sequencing analysis could provide new insights into the global spliceosome activity of ARX, for example, whether ARX controls splice isoform switching during neural development and how it contributes to defining neuronal maturation stages and cellular heterogeneity (92). Remarkably, we discovered a new aspect of ARX pathophysiology involving defects in the combinatorial codes of cell adhesion molecules Neurexin 1 and 2 (55–57), a class of presynaptic proteins that are required for efficient neurotransmission and formation of synaptic contacts (98). As their corresponding human genes were found involved in neurodevelopmental disorders (54,93), it will be interesting to establish how, when and where these processes contribute to the triggering and evolution of epileptogenesis.

In conclusion, we infer that depending on the pleiotropic impact of the null or the polyalanine expansion mutations on the genetic program controlled by ARX, defects in microtubule organization, translation efficiency and splicing patterns converge to generate macro or micro defects in Arx-XLAG and Arx-DEE1 pup brains. Of significance, in this study we highlight the GoF activity of the expanded repeat mutation Arx (GCG)7 by shedding light on an unclear aspect of the pathophysiological consequences of this polyalanine expansion. However, it remains to be determined whether all the expansion mutations found in ARX patients (2,8,10) and modelled in mice (7,13,15,21,27) have a GoF effect or whether there is a relationship dependent on the affected polyAla tract (first or second) or on the number of added alanine codons.

Taken together, these findings led to the identification of novel ARX-related pathway biomarkers opening up to new studies in attempts to find molecular-guided therapy. This could have important applications for both ARX-endophenotypes and NDD with overlapping comorbidities such as lissencephaly and epilepsy caused by mutations in tubulin isoform genes (brain tubulinopathies) (63), basal ganglia disorders (29) and autism spectrum disorders associated with NRXN1 variants (93).

Materials and Methods

Mouse welfare and ethical statements

All experiments in mice were conducted in conformity with the European Community Directive 2010/63/EU and were approved by the Italian Ministry of Health (D.L.gs n. 26/2014) in accordance with the institutional animal care guidelines of the Institute of Genetics and Biophysics ‘Adriano Buzzati-Traverso’, under the accreditation n°307/2018-PR and n° 0009895-P.

Animal models

Arx KO/Y knockout colony was kindly provided by Dr Collombat (99). Arx(GCG)7/Y knock-in colony was purchased from the RIKEN BioResource Center in Japan (strain name: Arx(GCG)7–1 KI (B6), RBRC03654). Mice were maintained by crossing heterozygous females Arx(GCG)7/X with WT males C57BL/6 J and ArxKO/X with WT males C57BL/6 J (Arx is on the X-chromosome). Arx(GCG)7/Y genotyping was performed according to the protocol of the RIKEN BioResource Center (19) and ArxKO/Y genotyping was made according to the protocol described in Collombat et al. (99) using oligonucleotide primer pairs listed in Supplementary Material, Table S9. Sex assessment of male pups was performed by PCR amplification of the Sry gene using the pair of oligos reported in Supplementary Material, Table S9. ArxKO/X and Arx(GCG)7/X heterozygous female pregnancy was assessed after overnight mating and embryonic age was calculated as E0.5 when the vaginal plug was detected. Mice were sacrificed at birth upon deep anaesthesia with isoflurane. Whole brain and brain sections were dissected from XY WT, ArxKO/Y and Arx(GCG)7/Y animals under a stereomicroscope and then snap-frozen in liquid nitrogen (100). C. elegans animals were grown and handled following standard procedures at 20°C, on nematode growth medium (NGM) agar plates seeded with Escherichia coli strain OP50 (101). For alr-1(oy42), a null mutant hence referred as alr-1(KO), the genotype was determined by PCR as described in Poeta et al. (4).

Proteome extraction and sample preparation

Whole frozen brain tissues were mechanically disrupted following the procedure elsewhere described (64,102). Brain homogenates were lysed in ice-cold lysis buffer (50 mM Tris–HCl pH = 7.5, 150 mM NaCl, 1% Triton X-100, 1 mM EDTA) supplemented with protease inhibitor cocktail (Roche, Indianapolis, IN). Supernatants were collected and protein concentrations determined using Bio-Rad Protein Assay Dye Reagent Concentrate (Hercules, CA). For each sample, the volume corresponding to 50 μg of protein was reduced with TCEP (Sigma-Aldrich, St Louis, MO) for 10 min and subsequently carbamidomethylated with iodoacetamide (Sigma-Aldrich, St Louis, MO) for 30 min in the dark. Protein digestion was performed using Sequencing Grade Modified Trypsin (Promega, Madison, WI) on S-Trap (ProtiFi, USA) microspin columns as elsewhere reported (103,104). Digested peptides were eluted from the S-Trap columns, vacuum dried and kept at −80°C until analysis.

Liquid chromatography–mass spectrometry/mass spectrometry and quantitative proteomic analysis

Peptide mixtures were suspended in 0.2% formic acid and 500 ng of each sample were loaded on an EASY-nLC II chromatographic system coupled with an LTQ-Orbitrap XL mass spectrometer (Thermo Scientific, Bremen, Germany). Liquid chromatography—tandem mass spectrometry (LC–MS/MS) analysis was performed as elsewhere reported (105–107). Proteome Discoverer™ version 1.4.1.14 (Thermo Scientific, Bremen, Germany) was employed for proteomic analysis, set as described (106). Mus musculus was selected as taxonomy. ArxKO/Y, Arx(GCG)7/Y and XY WT proteomic datasets resulted constituted by about 1000 proteins identified by a minimum of two peptides (Supplementary Material, Tables S1 and S2). The normalized spectral abundance factor (NSAF) was calculated and used as abundance index for each protein in the global dataset (102,108). Protein relative abundance in ArxKO/Y and Arx(GCG)7/Y samples was calculated as ratio of the NSAFs (both with respect to WT samples), and finally expressed in the log2 scale, obtaining the log2 fold change (FC). In particular, the proteins characterized by a FC > 0.4 or < −0.4 (corresponding to 1.3 ratio) were included in final proteomic datasets as differentially expressed proteins (DEPs). Hierarchical clustering dendrogram using Euclidean distances were carried out. DEPs were further filtered retaining only statistically significant FCs, whereas the Student’s t-test between the two analysed groups was ≤ 0.05 (107,109).

Computational analysis and gene ontology enrichment

The hierarchical clustering analysis and heatmap of the identified proteins were performed using the MetaboAnalyst 4.0 software (http://www.metaboanalyst.ca) (64,110); Data were log-transformed and Pareto-scaled. Normal distribution of NSAF values in ArxKO/Y, Arx(GCG)7/Y and XY WT proteomes was verified (Supplementary Material, Fig. S9). To annotate molecular functions of DEPs, Gene ontology (GO) mapping and Kyoto encyclopaedia of genes and genomes (KEGG) pathways analysis were performed using DAVID gene ontology analysis (http://david.abcc. ncifcrf.gov/). Bioinformatic analysis of altered biological processes was carried out using Ingenuity pathway analysis (IPA; https://digitalinsights.qiagen.com/). Protein–protein interaction (PPI) network analysis was obtained using STRING (Search Tool for the Retrieval of Interacting Genes/Proteins; https://string-db.org) online database (111). The highest confidence (>0.9) of the argument of interactions was set. Public microarray datasets (GSE12956) (20) were downloaded from G.E.O. repository using ad hoc modified GEO2R pipeline. For microarray expression the analysis was performed in R (version 3.6.1) using the following packages: Biobase (version 2.46.0), GEOquery (version 2.54.1), limma (version 3.42.2), gprofiler2 (version 0.1.9). Briefly, extracted data were normalized, divided in two groups (ArxKO/Y and XY WT mice at E14.5) and log2 transformed. Differential expression was measured using limma package. Multiple test correction was performed applying the false discovery rate (FDR) method. The FDR threshold to identify differentially expressed genes was set to ≤ 0.05. Heat maps reporting log2 transformed signal intensities and Z-scores (Supplementary Material, Fig. S1) were generated using custom scripts in R language using the following packages: heatmap.2 function in ggplot2 (version 3.3.2) and RColorBrewer (version 1.1.2). ARX-bound genes were retrieved from ChIP-on-chip data in Supplementary material of the paper from Quillé and colleagues (31). Gene/protein ID conversion were performed using the online tool g:Profiler and the Uniprot database (https://www.uniprot.org) where needed. Intersection between gene/protein lists were first performed in R and then the graphical output (Venn diagram) was generated using Venny online tool (https://bioinfogp.cnb.csic.es/tools/venny/index.html).

Primary cortical neurons, depolarization experiments and immunocytochemistry

Cortical neurons were prepared from Arx(GCG)7/Y and XY WT neonatal brains (3 mice per culture experiment). Cortices were dissected and dissociated to single cells suspension with 0.125% trypsin. Cells were maintained in Neurobasal medium (NB, Gibco, Carlsbad, CA), 1% B-27 plus supplement (Gibco, Carlsbad, CA), 10% Horse serum (Gibco, Carlsbad, CA), 2 mM L-glutamine and penicillin/streptomycin (Gibco, Carlsbad, CA). At DIV 4, cells were treated with 10 mM Cytosine arabinoside (Sigma-Aldrich, St Louis, MO) for 48 h. Membrane depolarization was induced at culture days in vitro 11 (DIV11) by addition of depolarization buffer with high potassium concentration (170 mM KCl, 2 mM CaCl2, 1 mM MgCl2, 10 mM Hepes) for 10 min. NaCl buffer solution was used as a control (112). After 24 h, cortical neurons were collected to perform expression transcript analysis. Quantitative PCR was performed with SYBR-Green-based reagents (Bio-Rad, Hercules, CA) using a CFX96 real-time PCR Detection system (Bio-Rad, Hercules, CA) to evaluate c-fos transcript abundance (Supplementary Material, Table S9).

For immunocytochemistry, primary cortical neurons were cultured on coverslips. At DIV10 were fixed with 4% paraformaldehyde (PFA, Gibco, Carlsbad, CA) in phosphate saline buffer (PBS) (Gibco, Carlsbad, CA), permeabilized in PBS with 0.03% Triton-X and blocked in PBS/bovine serum albumin (BSA) 3%. The following primary antibodies were used: mouse anti-β III tubulin (dilution 1:750, T8578, Sigma-Aldrich, St Louis, MO); rabbit anti-Map2 (dilution 1:400, AB5622, Merck Millipore, Darmstadt, Germany); while as secondary antibodies we used: Texas Red antimouse (T862, 1:400, Life Technologies, Carlsbad, CA) and Alexa Fluor 488 antirabbit (dilution 1:400, A11034, Life Technologies, Carlsbad, CA). Nuclei were counterstained with DAPI (1 μg/mL, D1306, Life Technologies, Carlsbad, CA). Images were taken using a laser-scanning microscope (LSM 700, Carl Zeiss Microimaging, Inc.) equipped with a plan neofluor ×20 (NA 0.5) or plan neofluor ×40 oil immersion (NA 1.3) objective lens or diode lasers (at 405 and 555 nm) were used as light source. Z-slices from the top to the bottom of the cells were collected and 3D reconstructions were carried out using the LSM 700 software. Alternatively, images were acquired using a Nikon Eclipse inverted microscope equipped with Nikon DMS-Ri2 camera.

SUnSET assay

Culture days in vitro 7 (DIV7) of murine primary cortical neurons were incubated with puromycin (Sigma-Aldrich, St Louis, MO) for 30 min. Cells were washed with ice-cold PBS and lysed directly in RIPA buffer (Sigma-Aldrich, St Louis, MO). Samples were analysed by western blotting and puromycin incorporation was detected using the mouse monoclonal antibody MABE343 (dilution 1:10 000, 12D10 clone, Sigma-Aldrich, St Louis, MO). Coomassie staining of total proteins were done and immunolabelling of the housekeeping protein HSP90 was used as loading control.

RNA extraction and alternative splicing analysis

Total RNA extraction and DNAse treatment were performed according to the manufacturer’s protocols (Life Technologies, Carlsbad, CA). Then, reverse transcription was performed with Superscript III Reverse Transcription kit (Life Technologies, Carlsbad, CA). Alternative splicing abundances were determined using the AmpliTaq DNA Polymerase (Applied Biosystems, Thermo Scientific, Bremen, Germany) on the Biorad PCR System (BioRad Laboratories Inc., Hercules, USA). C. elegans samples were similarly treated, except for the flash-freezing of worm pellets before proceeding with RNA extraction in TRI Reagent® (Merck, Darmstadt, Germany) and FastStart™ Taq DNA Polymerase (Roche, Basel, Switzerland) used for determination of alternative splicing abundance. The oligonucleotide sequences used in mice and C. elegans studies are reported in Supplementary Material, Table S9. For Neurexin analysis, alternative transcripts in mice were analysed by using oligonucleotides reported in Iijima et al. (54) whereas in C. elegans oligonucleotides were as reported in Kuroyanagi et al. (113). In mouse, the measures of transcript analysis were normalized to Hprt RNA level, while in C. elegans transcript analysis were normalized to act-1 RNA level. Each experiment assay was performed in triplicate in three independent experiments. DNA fragment intensities were quantified by using ImageJ software.

Immunoblotting and antibodies

Protein extraction in murine brain tissues and in C. elegans were prepared by standard methods described elsewhere (4,100). Membranes were blocked with 5% non-fat dry milk and the following antibodies were used: antitubulin (dilution 1:100 000, #t6074, Sigma-Aldrich, St Louis, MO), antiacetyl tubulin (dilution 1:100 000, #t7451, Sigma-Aldrich, St Louis, MO), anti-Puf60 (dilution 1:1000, ab22819, Abcam, Cambridge, UK) and anti-eiF4A2 (dilution 1:40 000, ab31218, Abcam, Cambridge, UK). As secondary antibodies, we used antimouse (dilution 1:10 000, sc-2005 Santa Cruz Biotechnology, Dallas, TX), antigoat (dilution 1:5000, sc-2020 Santa Cruz Biotechnology, Dallas, TX) and antirabbit (1:10 000, sc-2004 Santa Cruz Biotechnology, Dallas, TX). As loading controls anti-β-actin (dilution 1:5000, sc8432; Santa Cruz Biotechnology, Dallas, TX), anti-GAPDH/anti-Gpd2/3 (dilution 1:5000, ab181602, Abcam Cambridge, UK) and Hsp90 (dilution 1:50 000, E-AB-22072, Elabscience, Houston, TX) were used. The signals were detected with an enhanced chemiluminescence kit (Advansta, San Jose, CA). Films were processed for densitometry scanning using the ImageJ software.

Statistical analysis

One-way ANOVA with Dunn multiple comparisons test, Two-way ANOVA with Tukey’s multiple comparisons test and Student’s t-test were applied using the GraphPad Prism 7 software. P-values < 0.05 were considered significant.

Data Availability

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (available at http://www.proteomexchange.org) via the Proteomics Identification database (PRIDE; http://ebi.ac.uk/pride) partner repository with the dataset identifier PXD021358.

Web Resources

DAVID, https://david.ncifcrf.gov/

GenBank, https://www.ncbi.nlm.nih.gov/genbank/

g:profiler, https://biit.cs.ut.ee/gprofiler/gost/

modENCODE, http://www.modencode.org/

IPA, https://digitalinsights.qiagen.com/

NCBI, https://www.ncbi.nlm.nih.gov/

OMIM, https://www.omim.org/

STRING, https://string-db.org

UCSC, https://genome.ucsc.edu/

Uniprot, https://www.uniprot.org/

Venny, https://bioinfogp.cnb.csic.es/tools/venny/index.html

Wormbase, https://wormbase.org/

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.