Loss of the Chr16p11.2 ASD candidate gene QPRT leads to aberrant neuronal differentiation in the SH-SY5Y neuronal cell model

Altered neuronal development is suggested to be one of the major drivers in the etiology of autism spectrum disorders (ASD). Neuropathological studies based on postmortem brains of ASD patients reported aberrant neuronal development including reduced dendritic branching in the hippocampus [1], smaller pyramidal neurons in the language associated Broca’s area [2], abnormal minicolumnar organization in the cerebral cortex leading to decreased inter-areal connectivity [3], and disorganized layers of the cortical areas [4]. The underlying etiology of ASD is mainly based on different genetic findings including de novo copy number variations (CNVs). These CNVs, in particular deletions, have been recurrently shown to alter genic regions in ASD individuals [5], specifically affecting neurodevelopmental genes [6].

One of the most recurrent CNVs in ASD resides within Chr16p11.2 spanning ~ 600 kb. Overall, duplications and deletions of 16p11.2 can be identified in 0.8% of ASD cases [7]. A deletion of this region is associated with a nine times higher likelihood of developing ASD, and the duplication is associated with a nine times higher risk of both ASD and schizophrenia [8]. While developmental delay or intellectual disability can occur in some cases of 16p11.2 duplication carriers, they are more common in deletions [9, 10].

The 16p11.2 CNV region spans 29 genes which showed gene dosage-dependent expression in lymphoblastoid cell lines (LCLs) of CNV carriers, leading to a differential expression of genes implicated in biological processes such as synaptic function or chromatin modification [11].

A study in zebrafish showed that the majority of the human Chr16p11.2 homologous genes are involved in nervous system development: loss of function of these genes led to an altered brain morphology for 21 of 22 tested genes [12]. Double heterozygous knockouts of the Chr16p11.2 homologs double C2 domain alpha (DOC2A) and family with sequence similarity 57, member Ba (FAM57BA) induced hyperactivity, increased seizure susceptibility and increased body length and head size in zebrafish [13]. In mice, CNVs of the homologous 16p11.2 region induced differing phenotypes. Two studies reported the deletion to result in a reduction of the skull [14] or brain size [15], accompanied by gene dosage-sensitive changes of behavior and synaptic plasticity [14] as well as altered cortical cytoarchitecture, and reduction of downstream extracellular signaling-related kinase (ERK/MAPK) effectors [15]. Comparing individual brain regions in mice with deletions to wild-type animals, Horev and associates identified six regions with an increased volume, which were not altered in duplication carriers (see Dataset S04 in the original publication [16]). In another study, mice carrying a heterozygous microduplication of the region showed increased dendritic arborization of cortical pyramidal neurons [17]. Via network analysis of protein-protein-interaction, the authors identified the gene coding for mitogen-activated protein kinase 3 (MAPK3) as hub gene. MAPK3 plays a role in signaling cascades involved in proliferation and differentiation. A recent study focused on different effects of 16p11.2 deletions in male and female mice and reported impairments of reward-directed learning in male mice accompanied by male-specific overexpression of dopamine receptor D2 (DRD2) and adenosine receptor 2a (ADORA2A) in the striatum [18]. Both genes have been discussed in the context of ASD [19, 20].

While the functional validation of the entire CNV models the genomic status of the patients, investigating gene dosage effects of single genes located in Chr16p11.2 is useful to understand their individual contribution to the complex and diverse pathologies of ASD. In zebrafish, the suppression of potassium channel tetramerization domain containing 13 (KCTD13) was associated with macrocephaly whereas overexpression led to microcephaly [21]. In mice, the same study showed a reduction of KCTD13 to result in increased proliferation of neuronal progenitors, which is also suggested to result in macrocephaly. Further, a heterozygous deletion of the gene coding for major vault protein (MVP) induced a reduction of functional synapses in mice [22]. TAO kinase 2 (TAOK2), also located in 16p11.2, was found to be essential for the development of basal dendrites and axonal projections in cortical pyramidal neurons of mice [23]. Chr16p11.2 genes DOC2A, KIF22, and T-box 6 (TBX6) are required for the development of neuronal polarity in mouse hippocampal cultures [24].

In ASD patients, multiple brain measures such as the thalamic or total brain volume were reported to be increased in 16p11.2 deletion carriers and reduced in duplication carriers [25, 26]. Another study integrated physical interactions of 16p11.2 proteins with spatiotemporal gene expression of the human brain. The authors identified the KCTD13-Cul3-RhoA pathway as being crucial for controlling brain size and connectivity [27]. Still, only few genes of the Chr16p11.2 region have been investigated for their specific role in neuronal differentiation in human models.

Here, we investigated the SH-SY5Y neuroblastoma cell model as a well-studied and feasible model for neuronal differentiation in vitro. Previously, we reported that a continuous application of brain-derived neurotrophic factor (BDNF) and retinoic acid (RA) leads to neuronal cells of most likely cortical identity with a transcriptomic signature reminiscent of that of neocortical brain tissue developed for 16–19 weeks post-conception [28]. In addition, we showed expressed genes in the differentiated SH-SY5Y model to be co-regulated within modules, several of which were associated with neurodevelopmental disorders such as the orange module. We then implemented three complementary statistical methods to identify genes that were (i) differentially regulated upon differentiation, (ii) significantly involved in the independent processes active during differentiation and/or (iii) that were significantly changed over time. Finally, we described a list of 299 robustly regulated genes that appeared to be significant in all three analyses [28].

We here report that of the 29 genes located within Chr16p11.2, a total of 10 genes were identified by at least one of the three implemented statistical approaches. However, only the gene coding for quinolinate phosphoribosyltransferase (QPRT) was identified by all three analyses. In addition, QPRT was one of the most highly expressed genes of the Chr16p11.2 region and showed the highest regulatory fold change (FC) after induction of neuronal differentiation. Also, QPRT was co-regulated with an early upregulated gene module (MEorange) which showed significant enrichment for ASD candidate genes [28].

QPRT codes for an enzyme of the kynurenine pathway, the primary route for tryptophan catabolism, which results in the production of nicotinamide adenine dinucleotide (NAD⁺). In addition, it is the only enzyme catabolizing quinolinic acid (QUIN), a potent excitotoxin acting as N-methyl-D-aspartate receptor (NMDA-R) agonist. QUIN is also linked to astroglial activation and cell death as originally identified in the context of Alzheimer’s disease [29]. QPRT-KO mice showed increased QUIN levels in the brain [30] and increased excretion of QUIN in urine [31]. A significant increase of QUIN was observed in blood plasma of children with ASD when compared to their age-matched healthy control siblings [32]. Furthermore, QPRT was identified as an interaction partner of the ASD candidate neuroligin 3 (NLGN3; [33]), suggesting an involvement of QPRT in the formation of the postsynaptic density.

Here, we hypothesized that QPRT is implicated in neuronal differentiation and that reduced QPRT expression following its deletion results in alterations of neuromorphological development. We first tested the gene dosage-dependent expression of QPRT in a patient-specific LCL of one Chr16p11.2 deletion carrier. We then analyzed the expression of QPRT and its co-regulated gene set for correlation with the development of neuronal morphology in SH-SY5Y wild-type (WT) cells. To study the effects on neuronal morphology, we inhibited QPRT function in SH-SY5Y cells using (i) siRNA knockdown (KD), (ii) chemical mimicking of loss of QPRT, and (iii) complete CRISPR/Cas9-mediated knock out (KO). QPRT-KD cells underwent morphological analysis. Chemically inhibited and QPRT-KO cells were characterized using viability assays. To understand the effects of QPRT loss on the kynurenine pathway and QUIN levels, we additionally performed a metabolite analysis of the generated QPRT-KO cells. To explore the systems-wide interaction network of QPRT, we investigated the transcriptomic signature of QPRT-KO cells. Finally, to understand the role of QPRT in neural development, we tested the genes associated with QPRT-KO for enrichment among gene-networks implicated in human brain development [34].

Based on our differential analysis, we report a causal relation between QPRT, located in the ASD-associated CNV region Chr16p11.2 and neuronal differentiation of SH-SY5Y cells. A gene dosage reduction or inhibition of QPRT affects morphological parameters during neuronal differentiation as well as the regulation of genes and gene networks that were previously implicated in ASD. This includes processes like synapse organization or brain development. In summary, our findings suggest a neurodevelopmental role for QPRT in the etiology of ASD in 16p11.2 deletion carriers.

As expected from previous findings [11], in one deletion carrier with ASD, we confirmed that QPRT was expressed in a gene dosage-dependent manner strengthening the role of QPRT as a potential risk gene in the pathology of Chr16p11.2 deletion syndrome. By comparing our previously published transcriptome data of SH-SY5Y wild-type cells [28] to the morphological changes during neuronal differentiation described in the present study, we found QPRT to be correlated with the development of the neuritic complexity. These results suggest a potential regulatory link between QPRT and neuronal maturation. Our findings of the KD and KO cell models further underline this interpretation: the KD of QPRT led to subtle changes of SH-SY5Y neuronal complexity in line with previous reports using mouse models of Chr16p11.2 [17] as well as iPS cells generated from 16p11.2 CNV carriers [59] and postmortem studies of ASD individuals [1]. Inhibition of QPRT activity as well as genetic KO both led to cell death of differentiating but not proliferating cells. These findings suggest that enough protein is left for the survival of differentiating cells upon KD of QPRT while the loss of QPRT is lethal for differentiating SH-SY5Y cells. Interestingly, the administration of the QPRT substrate QUIN, which is a potent excitotoxin, did not show any effect on neither proliferating nor differentiating cells. Previous findings suggested that a reduction of QPRT, the only enzyme catabolizing quinolinic acid (QUIN), leads to an accumulation of QUIN, which in turn may induce neuronal cell death by over-activating NMDA-R and increase nitric oxide (NO) production [46]. Altered QUIN levels could also lead to a change in NAD⁺ production which in turn could change poly (ADP-ribose) polymerase (PARP) activity [60]. In QPRT-KO mice, the striatum showed an accumulation of QUIN leading to neurodegeneration [30] as well as altered expression of enzymes of the kynurenine pathway and of NMDA-receptors. No ASD-like behaviors were studied in these mice, but the animals showed no growth or developmental abnormalities. As the kynurenine pathway was associated with Parkinson’s disease [61], the authors performed a behavioral test measuring the stride length of WT and KO mice. Indeed, they reported shorter stride lengths in aged but not middle-aged QPRT-KO mice as usually seen in mouse models of Parkinson’s disease [30]. In another study, no differences of histological features of the cerebrum of QPRT-KO mice were observed [31]. Although we were not able to mimic QPRT loss in WT cells by application of increased QUIN levels, we tried to rescue differentiating QPRT-KO cells from cell death during differentiation by modulation of QUIN-related metabolism and signaling. Inhibition of NMDA-receptors was expected to protect QPRT-KO cells from the possible QUIN-induced excitotoxicity. However, the application of different concentrations of an NMDA-R inhibitor did not prevent cell death. Similarly, inhibition of NOS1 with the aim to reduce the production of NO did not affect viability. Finally, we tried to increase the viability of differentiating cells via the application of NAD⁺ to rescue PARP activity. Again, QPRT-KO cells could not be rescued from cell death during differentiation. In accordance with these observations, QPRT-KO cells did not show changes in the expression levels of any genes coding for PARP enzymes. In addition, QUIN could not be detected in any of our samples, and none of the metabolites of tryptophan catabolism was significantly changed upon KO of QPRT in the performed metabolite assays. Taken together, we conclude that the detrimental effect of QPRT loss on viability during SH-SY5Y differentiation is independent of QUIN levels as well as of QUIN-induced metabolic changes or the kynurenine pathway in general.

To elucidate the underlying mechanisms of QPRT-KO-induced cell death during SH-SY5Y neuronal differentiation, we performed a hypothesis-free transcriptomic approach. These findings suggest that loss of QPRT may lead to an increased negative regulation of cytoskeleton organization and to an inhibition of neuronal differentiation and dendritic spine development. In our KD model, we observed an alteration of neuritic complexity, which strongly supports the functional role of QPRT in these processes.

The functional association of QPRT with ASD is supported by the following results of our study: QPRT-KO led to inhibition of GABRB3, which has been well established as an ASD risk gene [62]. GABRB3 codes for a subunit of an inhibitory GABA receptor. It is located on Chr15q11-13, a region strongly implicated in ASD [63]. Similarly, the gene SNTG2 also downregulated in QPRT-KO cells, codes for a synaptic scaffolding protein involved in actin and PDZ domain binding. SNTG2 is located in 2p25.3, a region linked to intellectual disability [64] and associated with ASD [65]. Furthermore, SNTG2 protein interacts with neuroligins (NLGN), and this interaction is altered by ASD associated mutations in the NLGN genes [66]. QPRT was also found to physically interact with NLGN3 (neuroligin 3; [33]). Although the function of this interaction is still unclear, it is likely that QPRT is involved in the formation of the postsynaptic density of GABAergic neurons. KCNQ3, another ASD candidate gene, downregulated upon QPRT-KO, encodes a protein regulating neuronal excitability. Variants of this gene were found to be implicated in the development of ASD and epilepsies [67]. Finally, we also found CNTNAP2, a well-replicated risk gene for ASD [68] located in 7q22-q36, to be downregulated in QPRT-KO cells. In a previous study, we reported CNTNAP2 promoter variants reducing transcription to be risk factors for ASD [69].

Besides ASD-associated differentially regulated genes, we report six novel candidates (i.e., COX17, GUCA1A, COX17P1, VSTM2A, LINC01760, and ARHGAP20). At this stage, we cannot find any link to neuronal development for the photoreceptor-associated guanylate cyclase activator 1A GUCA1A, the preadipocyte development implicated V-set and transmembrane domain containing 2A VSTM2A gene, and the long intergenic non-coding RNA LINC01760. However, the cytochrome c oxidase copper chaperone COX17 (and its pseudogene COX17P1) and the Rho GTPase-activating protein 20 ARHGAP20 are functionally related to neuronal development: COX17 is part of the terminal component of the mitochondrial respiratory chain, catalyzing the electron transfer from reduced cytochrome c to oxygen, and might thus be involved in the regulation of oxidative stress and energy metabolism, both processes that have previously been associated with ASD [36]. The ARHGAP20 enzyme is implicated in neurite outgrowth and thus potentially associated with the here observed neuromorphological phenotypes [70].

Translating the in vitro transcriptomic effect of QPRT-KO in SH-SY5Y cells to the gene networks active during human brain development [34], we observed differentially regulated genes to be enriched in modules previously associated with cell cycle regulation (Kang-Module 1), transcription factors regulating progenitor cell fate (Kang Module 20), neuronal development (Kang-Module 10 and 20), morphogenesis (Kang Modules 17 and 2), and synaptic transmission (Kang-Module 2 and 15). In addition, the Kang-Modules identified to be affected by QPRT-KO were also reported to be highly co-expressed (ρ ≥ 0.68) with markers for glutamatergic (Kang-Module 10) and GABAergic neurons (Kang-Modules 2 and 15) or astrocytes (Kang-Module 2). In addition, Kang-Module 1 is implicated in the early development of the hippocampus and amygdala, two regions significantly associated with ASD after meta-analysis [71, 72]. These associations identified in the translational approach suggest that QPRT loss might trigger alterations in the development of excitatory/inhibitory neuronal networks, a pathomechanism postulated to underlie the etiology of ASD [73, 74].

In summary, our findings allow us to conclude that in the neuroblastoma model SH-SY5Y a loss of QPRT impairs neuronal development in vitro by changing genetic networks previously implicated in the etiology of ASD. To confirm these results and the role of QPRT in the etiology of ASD in general, further studies in other neuronal or animal models are needed. In particular, analyzing the above described QPRT-KO mouse model [30, 31] could elucidate the effect of QPRT loss at systems level.

Here, we report QPRT, a gene within the ASD-associated 16p11.2 CNV region, to be essential for and causally related to SH-SY5Y neuronal differentiation in vitro. This corroborates postmortem findings of a disturbed neuronal development in ASD. The functional mechanism is still elusive; however, based on our results, we can exclude alterations of QUIN levels or other products of the kynurenine pathway in the here described SH-SY5Y model. Further, the transcriptomic approach suggests that a reduced availability of QPRT impacts on genes and networks associated with ASD, such as neuron differentiation, synapse organization, or the development of excitatory/inhibitory neurons. Overall this study suggests an alteration of QPRT expression or of QPRT related genes to be underlying the etiology of ASD in 16p11.2 CNV carriers. Further studies are needed to confirm our findings in more mature neuronal systems or animal models.