Novel missense mutations in PTCHD1 alter its plasma membrane subcellular localization and cause intellectual disability and autism spectrum disorder

Abstract The X‐linked PTCHD1 gene, encoding a synaptic membrane protein, has been involved in neurodevelopmental disorders with the description of deleterious genomic microdeletions or truncating coding mutations. Missense variants were also identified, however, without any functional evidence supporting their pathogenicity level. We investigated 13 missense variants of PTCHD1, including eight previously described (c.152G>A,p.(Ser51Asn); c.217C>T,p.(Leu73Phe); c.517A>G,p.(Ile173Val); c.542A>C,p.(Lys181Thr); c.583G>A,p.(Val195Ile); c.1076A>G,p.(His359Arg); c.1409C>A,p.(Ala470Asp); c.1436A>G,p.(Glu479Gly)), and five novel ones (c.95C>T,p.(Pro32Leu); c.95C>G,p.(Pro32Arg); c.638A>G,p.(Tyr213Cys); c.898G>C,p.(Gly300Arg); c.928G>C,p.(Ala310Pro)) identified in male patients with intellectual disability (ID) and/or autism spectrum disorder (ASD). Interestingly, several of these variants involve amino acids localized in structural domains such as transmembrane segments. To evaluate their potentially deleterious impact on PTCHD1 protein function, we performed in vitro overexpression experiments of the wild‐type and mutated forms of PTCHD1‐GFP in HEK 293T and in Neuro‐2a cell lines as well as in mouse hippocampal primary neuronal cultures. We found that six variants impaired the expression level of the PTCHD1 protein, and were retained in the endoplasmic reticulum suggesting abnormal protein folding. Our functional analyses thus provided evidence of the pathogenic impact of missense variants in PTCHD1, which reinforces the involvement of the PTCHD1 gene in ID and in ASD.


| BACKGROUND
From the last few decades, the molecular study of neurodevelopmental disorders revealed the involvement of numerous genes essential in neuronal developmental processes like morphogenesis or neuronal plasticity and synaptogenesis (Bourgeron, 2015;Gilman et al., 2011;Krishnan et al., 2016;Laumonnier et al., 2007;Moyses-Oliveira et al., 2020;Parenti et al., 2020). Among them, microdeletions and mutations in the X-chromosomal PTCHD1 (patched domain containing 1) gene were described in patients with autism spectrum disorders (ASDs) and/or intellectual disability (ID; Chaudhry et al., 2015;Filges et al., 2011;Marshall et al., 2008;Noor et al., 2010). More precisely, PTCHD1 gene disruption or truncation causes a recessive X-linked nonsyndromic neurodevelopmental disorder characterized by variable features of ID, ASD, global developmental delay in their childhood with also infantile hypotonia (especially hypotonic face features), motor incoordination, and sometimes dysmorphic features . Some patients display behavioral or psychiatric issues, including attention deficit with or without hyperactivity (ADHD), sleep disruptions, aggressive, or impulsive behavior (Chaudhry et al., 2015). No evidence for a recognizable and significant pattern of congenital anomalies or serious medical co-morbidities in association with disruptions or deletions of this gene was highlighted.
PTCHD1 encodes a membrane protein located in GABAergic neurons (Wells et al., 2016) and in glutamatergic neurons (Tora et al., 2017), with 12 predicted transmembrane domains, two large extracellular loops, and intracellular amino and carboxy-terminal tails. Recent studies highlighted that PTCHD1 can bind postsynaptic scaffold proteins PSD95 and DLG3 (SAP102) via a PDZ binding domain motif at the end of the C terminal tail (Tora et al., 2017;Ung et al., 2018), as well as retromer complex proteins SNX27, and also VPS26B and VPS35 but independently from the PDZ-domainbinding motif (Tora et al., 2017). PTCHD1 is also predicted to be a receptor, homologous to PTCH1, which is involved in the Sonic Hedgehog (SHH) pathway, however, PTCHD1 does not include the SHH ligand-binding site motif which is present in PTCH1 and has been found unable to modulate the SHH pathway (Bosanac et al., 2009;Ung et al., 2018).
Single nucleotide variants (SNVs), including missense, nonsense or truncating mutation in PTCHD1 have been highlighted in patients with NDD. Strikingly, little is known about the pathogenic evaluation of PTCHD1 missense variants using biological in vitro experiments, which leads to significant issues in genetic medical diagnosis. We established a functional analysis to investigate 13 missense variants using in vitro non-neuronal (HEK 293T) and neuronal (Neuro-2a and primary neurons) cellular models. Here, we report findings showing differential impacts of PTCHD1 variants on the cell membrane localization pattern, as well as protein expression levels. Specifically, variants located in structural domains such as transmembrane segments abolished the membrane expression of PTCHD1, which was retained in the endoplasmic reticulum. Besides the increasing number of pathogenic microdeletions and truncating mutations of PTCHD1 causing ASD and ID, our data provide further evidence of the role of PTCHD1 in neurodevelopmental disorders, with the growing contribution of missense mutations leading to loss-of-function consequences.

| Patients and identification/follow-up of missense variants in PTCHD1
The genetic investigations regarding the patients with the five novel PTCHD1 variants were approved by the local Institutional Review Boards (Lille, Lyon, Paris, Strasbourg), and written informed consent was obtained from the patients' parents, including explicit permission to share clinical and identifying information.
The molecular analyses were performed through genetic diagnosis protocols using whole-exome strategies on the affected boy and co-segregation processed on parents' genomic DNA when available. The collection of these variants was possible through the  (Kumar et al., 2009; https://sift.bii.a-star.edu.sg/), Polyphen-2 (Adzhubei et al., 2010;http://genetics.bwh.harvard.edu/ pph2/), and UMD Predictor (Salgado et al., 2016;http://umd-predictor. eu/). As SIFT is a tool relying on sequence homology between species, the protein sequence was taken from all species referred to on Uniprot (https://www.uniprot.org/uniprot/), corresponding to 816 complete or partial PTCHD1 orthologues. Protein sequences were aligned with Clustal Omega (Sievers et al., 2011) on Uniprot. In SIFT, the human PTCHD1 protein sequence (Refseq NP_775766.2) was entered as the query sequence, the others as related sequences in SIFT Aligned Sequences tool, with the FASTA file of aligned sequences generated by Uniprot after the alignment. In UMD-Predictor, "single analysis" was performed on the human PTCHD1 transcript ENST00000379361 from Ensembl (https://www.ensembl.org/). The presence of the different variants in the general population was assessed using gnomAD database (http://gnomad.broadinstitute.org/, v2.1.1), including both total and "non-neuro" data sets.

| PTCHD1 expression plasmid and sitedirected mutagenesis
The sequence of full-length coding Ptchd1 mouse cDNA was amplified from the IMAGE cDNA clone 40095445 (GenBank accession number BC116312; Source BioSience) and cloned in pAcGFP1-N vectors using In-Fusion cloning strategy (Catalog no. 632501, Clontech) to generate PTCHD1-GFP with GFP tag at the C-terminal end (previously described in Ung et al. (2018)). Importantly, the murine PTCHD1 protein shares the same amino acids length (888) with 98% of sequence identity between murine and human PTCHD1 proteins.
The mutant PTCHD1-GFP plasmids were generated using the Q5 ® Site-Directed Mutagenesis Kit (New England Biolabs) according to the manufacturer's recommendations. The primer sequences for each mutagenesis are indicated in Table S1. The efficiency of the site-directed mutagenesis was verified by Sanger Sequencing.

| Gene expression study by qPCR
HEK 293T cells were transfected with pAc-PTCHD1-GFP different plasmids using lipofectamine 2000 as described above. Two days after transfection total RNA was extracted from HEK 293T samples with TRIzol and then purified using a DirectZol Kit (Ozyme). About 1 μg of RNA was used to generate cDNA using a PrimeScript RT Reagent Kit (Takara, RR037B). PCR amplification of Ptchd1 cDNA was performed with a forward primer located in exon 1 is 5ʹ-GCCAACATGCTAGACCAACA-3ʹ, and the reverse primer located in exon 2 is 5ʹ-CCCGAGCATTCTTTAGCTCTT-3ʹ. The human GAPDH cDNA was used as a reference gene with a forward primer: (5ʹ-CTGCACCACCAACTGCTTAG-3ʹ) and a reverse primer (5ʹ-GTCTTCTGGGTGGCAGTGAT-3ʹ). PCR analyses were done in duplicates using SYBR Green Takyon Kit (Eurogentec, UF-NSMT-B0701) and a LightCycler 480 (Roche) on total cDNAs obtained from four independent cultures. Data were normalized to the GAPDH gene and to the pAc-PTCHD1-GFP WT control according to the 2 −ΔΔCp method. Data statistical analysis was carried out with GraphPad Prism 8, using a Kruskal-Wallis test with Dunn's multiple comparisons from at least four independent experiments.

| Western blot
Two days posttransfection, cells were washed two times with cold PBS then were lysed using RIPA buffer (Pierce) with protease inhibitors cocktails for 15 min on ice. After incubation, the lysates were centrifugated at 15,000g for 10 min. The supernatants were collected and total protein lysates were quantified. A total of 30 μg of non-denatured proteins were processed for SDS-PAGE at 180V during 1 h using mini-PROTEAN precast gels 4%-20% TGX stain-free (456-8094, Bio-Rad). SDS-PAGE gels were blotted on PVDF membrane using TransBlot Turbo (Bio-Rad) followed by blocking the membrane for 1 h with 5% milk diluted in TBS-Tween (0.05% tween 20). Primary antibodies for PTCHD1-GFP (1/2000, 632592, Takar-aBio) and actin (1:100,000, A3854, Sigma-Aldrich) were incubated overnight at 4°C in 5% milk in TBST buffer. After three washes (10 min) with TBST, membranes were incubated with corresponding HRP-conjugated secondary antibodies (Goat anti-rabbit 1/2500, W4018, Promega) for 45 min at room temperature. Proteins were detected after 1-min incubation for Actin blot and 5-min incubation for PTCHD1-GFP blot with Clarity western ECL Kit (1705061, Bio-Rad), and the membranes were visualized in a Chemidoc Touch imaging system (Bio-Rad). Immunoblots were analyzed by Chemidoc Touch (Bio-Rad) by calculating the volume of each band for the same area. The bands' intensity was normalized to Actin. Data statistical analysis was carried out using GraphPad Prism 8 and using the Kruska-Wallis test with Dunn's multiple comparisons test.

| Image analysis
The cellular-imaging study was performed using a confocal microscope Leica SP8 and the associated software Leica Application Suite X (LAS X). The White Light Laser (WLL; set at 546 nm) maximum intensity was limited at 70%, the 488 nm argon laser at 30%, and the 405 nm diode. Inner settings were performed depending on the emitted fluorescence from cells. Sequential acquisitions were made, and high-resolution z stack images of cells were taken with the optical magnification of 6× 300 with optical section separation (z interval) of 0.6 μm. The Image colocalization Manders' coefficient (Aaron et al., 2018) was used to quantify PTCHD1-GFP overlapping with each used co-marker (calnexin, PSMB5, and Na + /K + APTase) and was calculated via Fiji software (with automatic thresholds), on three consecutive stack image per condition. Data analysis was carried out using GraphPad Prism 8. We used a d'Agostino-Pearson  Table 1). As no functional validation was performed on these variants, the authors did not conclude about their pathogenicity degree. The five remaining, novel variants included in our study were identified in the French National Genetics Network on Intellectual Disability for genetic molecular diagnosis purposes. The p.Pro32Arg mutation was found in a patient with ID and inherited from his mother; the p.(Pro32Leu) was identified in a patient with NDD and was also present in an uncle HALEWA ET AL.  Figure 1a).

| The PTCHD1 protein is relatively intolerant to missense changes
We reviewed all the missense variants present in the general population using the full gnomAD data set (v2. 1.1, 141,456 samples). Almost 30% F I G U R E 1 Schematic representation of the structural organization of PTCHD1 protein and localization of the variants. (a) The predicted secondary structure of PTCHD1 protein (NP_775766.2, 888 amino acids) has been generated using TMHMM server 2.0 (http://www.cbs.dtu. dk/services/TMHMM/), revealing the 12 transmembrane domains (their respective boundaries are indicated by the amino acid number) embedded in the plasma membrane structure, the intracellular loops, N-terminal, and C-terminal ends, as well as the two extracellular loops.  Gambin et al., 2017;Marshall et al., 2008;Noor et al., 2010;Pinto et al., 2010;Whibley et al., 2010). Furthermore, the available functional studies performed so far, in cellular and animal models, were all based on the absence of PTCHD1 expression (Murakami et al., 2019;Tora et al., 2017;Ung et al., 2018;Wells et al., 2016).
The analysis of the coding sequence of the PTCHD1 gene in numerous cohorts of individuals with ID/ASD led to the description of few missense variants, however, without any relevant evidence supporting their pathogenicity.
We found that six missense mutations impair the membrane proteins would not function as for the wild-type protein in the developing human brain, leading to a possible loss-of-function consequence.
In view of these findings, we also found these six particular mutations induce a major decrease in PTCHD1 protein expression level. The associated pathophysiological mechanism is still speculative but would affect either the PTCHD1 sorting and trafficking toward plasma membrane, from the integration of PTCHD1 into ER to the vesicular transport, or the structural conformation of the protein, which will induce the proteasomal degradation process (Guna & Hegde, 2018;Ott & Lingappa, 2002). Gly300Arg. As PTCHD1 does not include a peptide signal sequence at its N-terminal end, the first TM will contain the signal anchor sequence, which binds the signal recognition protein (SRP) during translation and SRP leads the ribosome-nascent-chain to the ER for translocation and membrane integration (Ott & Lingappa, 2002).
The ineffective plasma membrane targeting of the PTCHD1 protein may be a direct clue supporting a pathological phenotype. Although disrupting variants in PTCHD1 are highly penetrant (Chaudhry et al., 2015), we found that the Pro32Leu variant has also been detected in the patient's uncle who has not been diagnosed with NDD, suggesting a variable penetrance in this family, the presence of secondary genetic/nongenetic events that would compensate the impact of the PTCHD1 Pro32Leu variant, or the possibility that it might be unrelated to the disease.
| 857 only two missense variants are referenced in the second extracellular loop (amino acids 521-695).
As classically undertaken in X-linked disorders, the X-inactivation status of the patient's mothers who transmitted their variant (although it is doubtful as to whether methylation status in mothers' lymphocyte DNA would bear any relation to that in the brain), as well as the segregation analyses in unaffected and/or affected relatives from different generations, would provide further arguments on the association of the candidate variant with the genetic disorder. This is however not always possible (i.e., de novo mutation or studies of trios cohorts), and the clinical validation of missense variants without functional evidence is still challenging to conclude about the pathogenicity degree, even with various prediction software. For example, we found that the variant Lys181Thr prevented membrane localization of PTCHD1 and caused ER retention, whereas it was predicted as benign by SIFT and Polyphen-2 (but not with UMD predictor, Table 1). It is thus crucial to establish such experimental strategies to provide formal evidence for a relevant evaluation of missense candidate variants, as recently illustrated for the synaptic NLGN3 and NLGN4X proteins (Nguyen et al., 2020;Quartier et al., 2019), and to gain more knowledge on the structural organization and functional domains of the PTCHD1 protein, which is still an orphan receptor.
F I G U R E 4 Subcellular localization of GFPtagged PTCHD1 proteins in transfected Neuro2a cells. Representative confocal microscopy images of immunofluorescence experiments in Neuro2a cells at 48 h posttransfection and fixation. The anti-GFP antibody revealed the subcellular localization of Ptchd1-GFP WT or variant forms (green fluorescence). The DAPI staining indicates the position of the nuclei. The subcellular compartments targeted in our analysis were: (a) the plasma membrane (Na + /K + ATPase antibody, red), (b) the endoplasmic reticulum (Calnexin antibody, red), or (c) the proteasome 20S (PSMB5 antibody, red). n = 3 independent transfections, with a minimum of 6-8 images for each condition. Scale bar, 20 μm. WT, wild-type