Background
Autism spectrum disorder (ASD) is a chronic neurodevelopmental disorder characterized by repetitive behavior and deficits in social interaction and communication skills. Epilepsy, intellectual disabilities, language delay, anxiety, and hyperactivity are highly comorbid with ASD [
1]. An increased ratio of synaptic excitation/inhibition (E/I) affecting neuroplasticity has been proposed as a common pathway for ASD [
2]. This has been linked to altered functional and structural connectivity. Additional evidence from post-mortem neuropathology also showed reduced parvalbumin and altered density/abundance of glutamatergic receptors including
GRM5 and
GRIA1 in ASD [
3‐
5]. On the other hand, overproduction of GABAergic neurons with
FOXG1 overexpression and accelerated cell cycle were also reported in induced pluripotent stem cells (iPSCs) of sporadic ASD with macrocephaly [
6].
For synaptic excitation, rare mutations in
NRXN,
NLGN, and
SHANK are reported in individuals with ASD and intellectual disability, further supporting the E/I imbalance hypothesis [
7].
NRXN1 and
SHANK2 are in fact the commonest rare genetic factors identified by a meta-analysis of multiple genetic studies [
8,
9]. Notably,
NRXN1 deletions are shared by ASD [
10‐
15], schizophrenia [
16‐
20], intellectual disability [
21], ADHD [
22], and epilepsy [
14,
23‐
26]. Whereas most human deletions involve in 5′ of
NRXN1α+/− with diverse clinical phenotypes, mouse
Nrxn1α−/− mutants display only mild behavioral deficit in nest building but are otherwise viable, fertile, and indistinguishable from wild-type littermates [
27]. This suggests that mouse and human may have different sensitivity to
NRXN1α gene lesions.
Three
NRXN1 family members (
NRXN1-3) exist in the genome, and
Nrxn1α−/−/Nrxn2α−/−/Nrxn3α−/− triple knockout mice are impaired in Ca
2+-triggered neurotransmitter release with altered expression of synaptic Ca
2+ channels and die of lung dysfunction [
28]. Ca
2+ concentration in neurons is tightly controlled by distinct influx/efflux mechanisms. Ca
2+ influx occurs commonly through voltage-gated calcium channels (VGCCs) on membrane [
29,
30], which facilitate a Ca
2+ rise during neuronal firing. The influx of Ca
2+ triggers vesicle exocytosis and neurotransmitter release. The long form of Nrnx1α has been shown to couple release-ready vesicles with metabotropic receptors, facilitating Ca
2+-triggered exocytosis of neurons [
31].
In addition to the long NRXN1α isoforms, which interact with post-synaptic Neuroligins and influence both excitation and inhibition through coupling to GABAergic or NMDA/AMPA receptors [
31] and VGCCs [
32],
NRXN1 also encodes short NRXN1β isoforms by an alternative promoter, which is largely associated with creation of the scaffolding for excitation [
33‐
35].
NRXN1α is therefore proposed to influence E/I balance in both directions, whereas
NRXN1β primarily mediates excitation. Indeed, conditional knockdown of
NRXN1β severely impaired the neurotransmitter release at excitatory synapses [
36]. It is likely that NRXN1
α deletion may display increased neuronal excitability, as a result of reduced ratio of
NRXN1α to
NRXN1β isoforms, and/or a compensatory increase of
NRXN1β expression if it happens. Pak et al. showed a reduced mEPSC frequency in human ESC-derived neurons after disrupting shared exon 19 or 24 of
NRXN1 gene, which knocked out an entire
NRXN1 allele with all
NRXN1α/β isoforms [
37]. However, this is different from the genetics in the majority of patients who carry heterozygous deletion at 5′ of
NRXN1 gene which affect
NRXN1α only, and to date, there have been no patient models to investigate the effects of isoform deletion and/or genetic background. Moreover, it has been shown that common pathophysiological social and cognitive deficits in autism can be linked to gain of function of synaptic proteins and ion channels [
7]. These include hyperactivity in frontal brain regions, high-frequency oscillation in cortical regions, and the presence of clinically apparent seizures in 30% of autistic individuals [
38‐
42]. In addition, mutation in neuronal adhesion molecule
CNTN5 has also shown hyper-excitability and increased excitation in iPSC-derived neurons of ASD individuals [
43]. These studies show the presence of hyper-excitability and hyperactivity in some of the ASD patients.
The iPSC technology now offers significant benefits for disease modeling [
44‐
46], which can be derived from patient somatic tissues. They resemble embryonic stem (ES) cells and can be differentiated into disease cell types, so to provide human models for investigating disease progression and testing therapeutic drugs, in particular for organs such as the brain and heart, which are impossible to culture by conventional methods. We therefore derived iPSCs from controls and ASD patients carrying
NRXN1α
+/− and differentiated them into cortical excitatory neurons, as altered cortical regions, thickness, folding, surface, columnar lamination, and the number excitatory neurons have been reported in ASD [
1,
47‐
51]. We investigated Ca
2+ signaling and the transcriptome in day 100 neurons and provided novel phenotype with increased Ca
2+ transients and upregulated VGCCs in ASD
NRXN1α+/− neurons.
Methods
Participants
Ethical approval for the study was obtained from the St. James’s/Tallaght University Hospital and the Galway University Hospital Clinical Research Ethics Committee. Seven control iPSC lines were derived from five healthy donors (Additional file
1: Table S1). The sample 1C was donated by a healthy sibling of patient ND1, the 4C (male), the 2V (female), and the 3V (male) by healthy volunteers. The NCRM1 control line was derived by NIH from a newborn boy.
All patients had confirmed research diagnoses of ASD with the Autism Diagnostic Interview-Revised and the Autism Diagnostic Observational Schedule (Additional file
1: Table S1) [
52,
53]. Six
NRXN1α+/− iPSC lines were generated from three ASD patients (Additional file
1: Figure S1A). The ND1 was donated by a non-verbal male with severe intellectual disability, autism, infant seizures, developmental delay, self-injurious and aggressive behavior, and carrying de novo
NRXN1α +/− deletion on exons 6–15 (chr2:50711687-51044633, Hg19). The ND2 was a male patient carrying
NRXN1α +/− deletion in exons 1–5 (Chr2:51120335-51360666, Hg19), with autism, language delay, IQ of 78 at age 11, but attended mainstream education. One of ND2’s parents had language delay, and one grandfather and one cousin had ASD. The ND4-1 female was diagnosed with Asperger’s syndrome, social anxiety, psychosis, and mild intellectual disability, with an IQ of 69, a history of seizures, and a paternal
NRXN1α+/− lesion (chr2:50983186-51471321). Her paternal grandmother was institutionalized, and her father and paternal aunt had seizures.
Genomic DNA from parental fibroblasts and iPSC lines was extracted with DNeasy kit (69504, Qiagen). An Illumina 1M SNP array was performed at UCD. All samples passed quality control with call rates > 99%. CNV analysis was carried out using PennCNV. False-positive CNVs were excluded using SNP < 10 or kb < 100. The
NRXN1α deletions were confirmed (Additional file
1: Figure S1A), and additional putative CNVs detected are listed in Additional file
1: Table S7.
iPSC derivation
Skin punches were obtained with consent in the Clinical Research Facility. Biopsy was cut, dragged along the rough surface of culture dishes for adherent culture at 37 °C with 5% CO2 in high glucose DMEM supplemented with 10% FCS, 1% NEAA, and 1% penicillin/streptomycin. The medium was renewed every 2–3 days. Low passage fibroblasts were reprogrammed to iPSCs (Merck-Millipore, SCR510; Thermo Fisher Scientific, or Epi5™ Episomal iPSC Reprogramming Kit; Invitrogen, A15960) and characterized for expression of alkaline phosphatase, NANOG, OCT4, SOX2, SSEA4, TRA-1-60, TRA-1-81, TUJ1, ASM, and AFP.
Neuronal differentiation
The iPSCs were seeded at 45,000–50,000 cells/cm
2, grown to ~ 85% confluency in E8 (Thermo Fisher Scientific, A1517001), and differentiated into neural rosettes for 10–12 days in N2B27 (Thermo Fisher Scientific) with 100 nM LDN193189 (Stem Cell technologies, #72102) and 10 nM SB431542 (Sigma, S4317) [
52,
53]. Neural rosettes were passaged, cultured for another 10 days, and then plated onto poly-D-Lysine/laminin-coated 12-well plates, 15-mm coverslips, or ibidi 8-well chambers for terminal differentiation. Cells were maintained in N2B27 (w/o vitamin A) for 6 days and then in N2B27 plus vitamin A until analyses by immunocytochemistry, immunoblotting, calcium imaging, or RNA sequencing, respectively. All phenotypic analyses were performed at day 100 of differentiation according to previous published protocol [
53].
Immunocytochemistry
Cells were fixed in 4% paraformaldehyde, blocked with 0.2% BSA, and incubated with primary antibodies (Additional file
1: Table S8) at 4 °C overnight. They were washed, incubated for 1 h at room temperature with appropriate secondary antibody (Additional file
1: Table S8), and mounted with DAPI. Images were taken under a fluorescence microscope and quantified by ImageJ.
Calcium imaging
Cultures were washed with artificial cerebrospinal fluid (ACSF), incubated with 2 μM Fluo-4 AM (Thermo Fisher scientific, F14201) in ACSF for 20 min at 37 °C, cultured in normal medium at 37 °C for 20 min, and imaged in warm ACSF in an imaging chamber (Warner Instruments, RC-26GLP) on a Zeiss Axiovert 200 microscope (× 10). Videos were captured with a Hamamatsu ORCA284 at 1 Hz frame rate for 3–5 min and stored as uncompressed image sequences.
Chemicals were added to the ACSF as required, i.e., Na+ channel blocker TTX (Alomone Labs T-550), AMPA/Kainate receptor blocker CNQX (Alomone Labs C-140), NMDA receptor blocker DL-AP5 (Alomone Labs D-140), L-type VGCC blocker Nifedipine (Alomone Labs N-120), P/Q-type VGCC blocker agatoxin (Alomone Labs STA-500), glutamate (Sigma, G8415), ionomycin (Sigma I0634), or γ-aminobutyric acid (Sigma A2129). Videos were recorded continuously.
FluoroSNNAP in MATLAB (MathWorks, Inc.) was used to analyze calcium image sequences [
52,
53]. Neurons with > 5% fluorescence variations during recording were identified by time-lapse analysis and cell soma defined using batch segmentation. A time-varying fluorescence trace was calculated, transient onset identified, and background noise (ΔF/F < 0.05) determined. The frequency, amplitude, duration, and network synchronicity of spontaneous and evoked calcium transients were analyzed by a coding script in R software.
Quantitative RT-PCR
RNA was extracted (Qiagen, 74104) and reversely transcribed (Qiagen, 205311). RT-PCR was executed in triplicate with primers listed in Additional file
1: Table S9. The average cycle threshold (Ct) values were calculated in both control and
NRXN1α+/− lines from three technical replicates. All Ct values were normalized to expression of a house-keeping gene (
GAPDH) as dCt. Relative expression was expressed as 2
–dCt over
GAPDH expression or 2
–ddCt over the target gene expression in control fibroblasts for iPSC characterization.
Transcriptomic analysis
RNASeq was performed by BGI as described previously [
54‐
57] on day 100 cortical neurons from six control iPSC lines of four donors and four
NRXN1α+/− lines of three patients. Transcripts were aligned to GRCH37/hg19, and abundance quantified from the FASTQ in Kallisto (v0.43.1) and presented as transcripts per million (TPM). The two groups were analyzed with false discovery rate (FDR) and adjusted multiple
p value using the DESeq2 in R. PLS discriminant analysis (PLS-DA) was carried out for supervised clustering, confirming the close clustering among controls and patients. PLS-DA is a supervised method for pattern recognition of unsupervised PCA data and uses the partial least squares (PLS) algorithm to explain and predict the membership of observations to several classes using quantitative or qualitative explanatory variables or parameters [
58]. Differentially expressed genes (DEGs) were identified using FDR < 0.05, TPM > 2, > 50% reduction, or > 1.7-fold increase based on TPM ratio and analyzed by STRING and Gene Set Enrichment Analysis (GSEA).
Statistics
All data were expressed as mean ± SEM. All data were tested for normality using the Shapiro-Wilk normality test. Statistical analysis was performed using the Student t test or Mann-Whitney U test with a p < 0.05.
Discussion
NRXN1+/− deletions are the most frequent single-gene disruptions associated with ASD [
10,
12,
14,
15,
59,
60], schizophrenia [
16‐
20], intellectual disability [
21], ADHD [
22], and epilepsy [
14,
23‐
26]. Little is known about the consequences of
NRXN1+/− lesions in patients’ neurons or why the same heterozygous
NRXN1+/− deletions lead to diverse clinical phenotypes. We are the first to report derivation of human iPSCs from ASD patients carrying
NRXN1α+/−. The cortical excitatory neurons from
NRXN1α+/− iPSCs displayed a novel phenotype of increased frequency, duration, and amplitude of Ca
2+ transients. This is supported by transcriptome analyses, which revealed an upregulation of VGCCs (
CACNA1A,
CACNA2D1,
CACNG2, and
CACNG3) and Ca
2+ pathways in
NRXN1α+/− neurons.
Typically in neurons, calcium influx is facilitated by the opening of the α1 subunit in the tetrameric VGCCs in response to membrane depolarizations. The α1 subunit is encoded by
CACNA1A,
CACNA1B,
CACNA1C,
CACNA1D,
CACNA1E, and
CACNA1S genes. Consistent with ASD
NRXN1α+/− phenotype, gain-of-function of VGCCs are implicated in neurodevelopmental disorders (Additional file
1: Table S6). For example, Cav1.2 G406R (
CACNA1C) causes Timothy syndrome with ASD by delayed inactivation and prolonged opening [
61,
62]. Knock-in of the G406R to mice results in autistic phenotype [
61,
62]. Exome sequencing has identified various
CACNA1D mutations (encoding Cav1.3) in ASD [
63‐
66], epilepsy [
67], and developmental delay [
67]. A
CACNA1D paralog,
CACNA1F (Cav1.4), also is linked to New Zealand autistic males with excessive Ca
2+ influx [
61,
62].
We have identified
CACNA1A encoding P/Q-type and
CACNA2D1 encoding L-type VGCC as the most interactive
NRXN1α+/− targets.
CACNA1A is predominantly expressed in neurons and involved in NRXN1α signaling which triggers the release of fusion-ready vesicles [
68].
CACNA1A polymorphisms are associated with Chinese ASD [
68], and
CACNA1A mutations with epileptic encephalopathy [
68]. Additionally, mutations in other VGCCs are also identified as a major pathway in schizophrenia [
68,
69], the common risks across seven brain diseases [
70,
71], and in ASD (Additional file
1: Table S6) [
72,
73]. In addition, loss-of-function mutations in some VGCCs are also reported, i.e.,
CACNA1H R212C, R902W, W962C, and A1874V reduce their activity in ASD [
74];
CACNA2D1 is deleted in epilepsy and intellectual disability [
74];
CACNG2 V143L decreases its binding to GLUR1 or GLUR2 [
75]; and
Cacng2 hypomorph results in epileptic phenotype [
74]. This evidence supports altered VGCCs as a mechanism in ASD
NRXN1α+/− neurons.
The human
NRXN1α+/− phenotype reported here differs from some of the data reported previously. Pak et al. created a mutant human H1 ES cell line with disruption of exon 19 or 24, which are shared by all
NRXN1 isoforms (Additional file
1: Figure S1A), and showed reduced frequency of mEPSCs [
37].
NRXN1 consists of 2 promoters and 11 differentially spliced exons which may result in 2048
NRXN1α and 4
NRXN1β isoforms. The human H1 ES cells (
NRXN1+/−) from Pak et al. are genetically different from the ASD patients here, who carry 1 copy of
NRXN1α+/−. However, qRT-PCR using primer pairs from exons 9–10 or 15–16 demonstrate 24 or 26% (not 50%) reduction. This is likely due to the complex exon usage of differential NRXN1 splicing. Meanwhile, we observe 262% compensational increase in
NRXN1β expression; therefore, the phenotype in this study is likely to result from combinational effects of reduced
NRXN1α and overexpression of
NRXN1β. This may also re-enforce the concept that
NRXN1α and
NRXN1β isoforms play differential roles in neuronal E/I.
Sudhoff et al. propose that Neurexin variants from alternative splicing may perform the same canonical functions but may have different patterns of redundancy [
76‐
78].
Nrxn1α homozygous knockout presented no apparent phenotype, and Pak et al. also showed that mouse
Nrxn1 knockout cells differed from H1 ES cells and displayed no phenotype [
79]. Mice with triple knockout of
Nrxn1α,
Nrxn2α and
Nrxn3α genes were shown to produce different phenotypes in different neurons or synapses [
76,
77]. In hippocampal presynaptic cells, the Ca
2+ influx was reduced in conjunction with lower Cav2.1-mediated transients and elevated axonal mobility of α2δ1 [
80]. Although overexpression of
Nrxn1α and α2δ1 is shown to rescue Ca
2+ currents in
Nrxn1α−/−Nrxn2α−/−Nrxn3α−/− triple knockout mouse neurons, this is yet to be investigated in human cells [
80]. In addition, species differences also exist: i.e., Nrxn1 at
Caenorhabditis elegans acetylcholine neuromuscular synapse is located postsynaptically, not presynaptically [
32], and approximately > 20% of human essential genes are nonessential in mice [
37].
The penetrance of human
NRXN1a+/− is not 100%, and clinical conditions of
NRXN1a+/− are diverse. Therefore, co-factors in the genetic background may play a part in clinical phenotype. Investigations of patient-derived samples are essential for understanding roles of
NRXN1a+/− in different human conditions. The ASD
NRXN1a+/− phenotype here is consistent with the proposal that NRXN1β triggers excitation, and NRXN1α regulates both excitation and inhibition [
33‐
35].
NRXN1α deletions are therefore anticipated to weaken neuronal inhibition and increase excitation. A recent publication has shown that ASD neurons derived from autism CNTN5
+/− or EHMT2
+/− human iPSCs develop hyperactive neuronal networks [
43]. This suggests indirect effects of NRXN1α on Ca
2+ transients. The upregulated CACNA1A, CACNA2D2, and CACNG2 are linked to “the presynaptic depolarization and calcium channel opening” by STRING (Additional file
1: Table S5). Direct interactions of NRXNs with VGCCs are reported but limited. Mouse Nrxn1α is shown to positively modulate Ca
2+ influx through Cav2.1-α2δ1 interaction [
80]. On the other hand, human NRXN1α may also form NRXN1α-Cav2.2-αδ3 complex and negatively regulate Cav2.2 currents in transfected cells [
32]. Furthermore, Neuroligins contain Ca
2+-binding EF-hand domains, and Neuroligin-NRXN1β interaction is dependent on Ca
2+ [
81]. Elevated Ca
2+ transients in human
NRXN1α+/− neurons may therefore also enhance excitation through increased Neuroligin-NRXN1β interactions. Furthermore, we have observed an increase in the expression of few members of SNARE complexes, i.e., synaptotagmins, suggesting an interaction of the cytoplasmic membrane of neurexins with synaptotagmins [
82]. It seems likely that NRXN1α may regulate the level of synaptotagmins or other members of SNARE proteins, which might be critical for neurotransmitter and vesicle release [
83]. Interestingly, two of our ASD patients had a history of seizures. While the patient numbers were small, it appeared that the increase of the frequency was more prominent in two ASD probands with seizure (ND1, ND4) than the ASD without seizure (ND2, Additional file
1: Figure S11). This concurs with disrupted Ca
2+ signaling implicated in a range of neurodevelopmental disorders including ASD and epilepsy [
79,
84‐
88].
The DEGs in
NRXN1α+/− neurons may arise from Ca
2+ influx and voltage-dependent conformational changes of VGCCs. For example, Cav1.2 may interact with αCaMKII, and βCaMKII is then recruited by Ca
2+ mobilization. Voltage-dependent conformational changes can lead to α/βCaMKII activation, CREB phosphorylation and nuclear accumulation [
89], and activation of transcription factors NFAT and MEF2 [
90‐
94]. Therefore, the transcriptomic changes may reflect both the activity-driven alterations and functional features of ASD
NRXN1α+/− neurons.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.