Intellectual disability-associated gain-of-function mutations in CERT1 that encodes the ceramide transport protein CERT

Intellectual disability (ID) is a developmental disorder that includes both intellectual and adaptive functioning deficits in conceptual, social, and practical domains. Although evidence-based interventions for patients have long been desired, their progress has been hindered due to various determinants. One of these determinants is the complexity of the origins of ID. The ceramide transport protein (CERT) encoded by CERT1 mediates inter-organelle trafficking of ceramide for the synthesis of intracellular sphingomyelin. Utilizing whole exome sequencing analysis, we identified a novel CERT variant, which substitutes a serine at position 135 (S135) for a proline in a patient with severe ID. Biochemical analysis showed that S135 is essential for hyperphosphorylation of a serine-repeat motif of CERT, which is required for down-regulation of CERT activity. Amino acid replacements of S135 abnormally activated CERT and induced an intracellular punctate distribution pattern of this protein. These results identified specific ID-associated CERT1 mutations that induced gain-of-function effects on CERT activity. These findings provide a possible molecular basis for not only new diagnostics but also a conceivable pharmaceutical intervention for ID disorders caused by gain-of-function mutations in CERT1.


Introduction
ID is a developmental disorder that includes both intellectual and adaptive functioning deficits in conceptual, social, and practical domains [1]. ID has an overall general population prevalence of~1%, which varies by age [1]. Although evidence-based interventions for patients have long been desired, their progress has been hindered due to various determinants [2,3]. One of these determinants is the complexity of the origins of IDs, which include: inborn genomic mutations, malconditions (e.g., hypoxia) in the fetal period, detrimental stresses (e.g., physical and/or mental abuse) in the infant period, an inferior social environment (e.g., unavailability of education) in childhood, and other factors. Recent advances in human medical genomics have facilitated the identification of causative mutations of various genetic diseases, which may provide rationales for personally tailored interventions for patients with specific mutations. Whole exome sequencing (WES) analysis has been applied to ID studies, and has identified various de novo mutations that are strongly associated with the induction of IDs [4][5][6][7][8][9]. Recently, several independent studies found that de novo mutations in CERT1 can induce IDs [4][5][6][7]. The lipid ceramide is synthesized in the endoplasmic reticulum (ER) and the ceramide transport protein CERT encoded by CERT1 transports ceramide from the ER to the trans-Golgi regions. Once localized to the trans-Golgi regions, ceramide is metabolized to sphingomyelin (SM), which is one of the major phospholipid types in mammalian cells [10,11]. Previous studies demonstrated that a serine-repeat motif (SRM) of CERT undergoes multiple phosphorylations, which down-regulate CERT activity [12]. Intriguingly, various examples of ID-related de novo mutations in CERT1 were mapped to the region encoding the SRM [4,5,7]. These mutations are expected to inhibit SRM hyperphosphorylation, thereby disrupting the post-translational down-regulation of CERT. However, it has not been shown whether these mutations affect the phosphorylation of the CERT SRM and render the mutated CERT abnormally active.
In this study, we identified a novel de novo missense mutation, which substitutes a serine at position 135 for a proline (S135P), in CERT1 (NM_0031361.3:c.403T>C:p.[Ser135Pro]) by trio WES analysis of a patient with a severe ID and systemic symptoms. Moreover, we showed that this novel mutation and other previously reported mutations inhibited the hyperphosphorylation of the CERT SRM, which resulted in abnormally activated CERT.

Study subjects and ethics statement
This study was approved by the Review Board and Ethics Committee of the Kanagawa Children's Medical Center (Approval number: 118-17). Written informed consent (as outlined in PLOS consent form) was obtained from the patient's parents, which included consent for the pictures appearing in the manuscript. The biochemical and cell biological analyses of LCLs were also approved by the Medical Research Ethics Committee of the National Institute of Infectious Diseases (Approval number: 1030).

Lysenin tolerance assay
Lysenin purified from the coelomic fluid of the earthworm Eisenia foetida was a gift from Yoshiyuki Sekizawa (Zenyaku Kogyo Co., Tokyo, Japan) [20]. Cells were seeded at an initial density of 1.0 x 10 4 cells/well in a 24-well plate and were grown under normal conditions overnight. The cells were then incubated in serum-free DMEM containing 250 ng/ml of lysenin for 1 h. The cell viability was measured with LDH cytotoxicity assay kit (Nacalai Tesque, Kyoto, Japan) according to the manufacturer's protocol. In our assay, the cell viability ("survival ratio after lysenin-treatment") was defined as the following formula: 1-(A1-A2)/(A3-A2). Here, A1, A2, and A3 represent LDH activity (measured by absorbance) from lysenin-treated cells, untreated control cells, and detergent-treated cells, respectively. The LDH activity from culture medium that was not used to grow cells was measured and subtracted as a background control. The luminescence was measured with an iMark microplate reader (Bio-Rad Laboratories, Hercules, CA, USA).

Immunofluorescence
Cells grown on coverslips (Matsunami Glass, Osaka, Japan) were washed with PBS and fixed with 4% paraformaldehyde (Mildform 10N; WAKO, Osaka, Japan) for 15 min at room temperature. The fixed cells were then permeabilized with 0.1% TritonX-100 in PBS for 5 min, blocked with 3% bovine serum albumin (Sigma-Aldrich, St. Louis, MO, USA) in PBS, and incubated with specific primary antibodies for 1 h. After washing with PBS, the cells were incubated with either Alexa Fluor 594-conjugated goat anti-mouse IgG or Alexa Fluor 647-conjugated goat anti-rabbit IgG secondary antibodies for 1 h. The cells were viewed using a confocal laser microscope (LSM700; Carl Zeiss, Jena, TH, Germany) equipped with a 100 × oil immersion objective lens (1.46 NA; Carl Zeiss, Jena, TH, Germany), and the view was captured with the ZEN software (Carl Zeiss, Jena, TH, Germany). Photoshop CS6 software (Adobe Systems Corporation, San Jose, CA, USA) was used for the final presentation of the images.

λPPase treatment
Cells were lysed with the lysis buffer lacking phosphatase inhibitor cocktails, and the protein concentration of each lysate was determined with the BCA method. Next, the lysates (50 μl) were incubated with 400 units of λPPase (New England Biolabs, Beverly, MA, USA) at 30˚C for 15 min.

Establishment of CERT1 KO cell lines
The CRISPR guide RNA (gRNA) sequence designed to target the human CERT1 gene was cloned into the pX458 vector. The target sequence is 5'-GCTCTGATTATCCGACATGG-3'.

Western blotting
Cells were lysed with a lysis buffer [50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, complete EDTA-free protease inhibitor cocktail (Roche, Mannheim, Germany), and phosphatase inhibitor cocktails 2 and 3 (Sigma-Aldrich, St. Louis, MO, USA)] and centrifuged at 20,000 × g for 20 min at 4˚C. The supernatants were collected, and the protein concentrations were quantitated with the bicinchoninic acid (BCA) method (Thermo Fisher Scientific, San Jose, CA, USA) using bovine serum albumin as the standard. The lysates were then solubilized with immunoblot sample buffer [46.7 mM Tris-HCl (pH 6.8), 5% glycerol, 1.67% SDS, 1.55% dithiothreitol, and 0.003% bromophenol blue]. Next, the samples were separated by SDS-PAGE, transferred to an Immobilon-P polyvinylidene difluoride membrane (Millipore, Darmstadt, Germany), and blotted with primary and secondary antibodies. Each protein signal was detected with the Immobilon Western Chemiluminescent HRP Substrate (Millipore, Darmstadt, Germany). The signal intensities were captured with the LuminoGraph image analyzer (ATTO, Tokyo, Japan) and the images were processed using Photoshop CS6 software (Adobe Systems Corporation, San Jose, CA, USA).

Sphingolipid labeling with radioactive serine and thin layer chromatography (TLC)
The metabolic labeling of sphingolipids with radioactive serine was performed as described previously [21]. Briefly, LCLs or HCT116 cells (5.0 x 10 5 -1.0 x 10 6 cells) were cultured in a 12or 6-well plate, respectively. After overnight incubation under normal conditions, LCLs or HCT116 cells were incubated with 1% Neutridoma-SP (Roche, Mannheim, Germany) and 18.5 kBq of L-[U-14 C]serine (Moravek Biochemicals, Brea, CA, USA) for 16 h in RPMI medium supplemented with 10% FBS or in serum-free low glucose DMEM, respectively. After the labelled cells were washed with phosphate buffered saline (PBS), they were lysed with a 0.1% sodium dodecyl sulfate (SDS) solution. The lipids were extracted from the lysate and separated by TLC with the solvent methyl acetate/n-propanol/chloroform/methanol/0.25% KCl (50/50/50/20/18, vol/vol) as described previously [22]. The radioactive lipids on TLC plates (Millipore, Darmstadt, Germany) were detected using an image analyzer (Typhoon FLA 7000; GE Healthcare, Madison, WI, USA). The intensities of the radioactive lipid bands visualized on the TLC plates were determined by densitometric scanning using ImageJ software (National Institutes of Health). The intensity levels of SM were normalized to phosphatidylserine in order to compare the relative levels of labelled SM among different cell lines.

Statistics
For data analysis of biochemical assays, a one-way analysis of variance (ANOVA) followed by Tukey's or Dunnett's test was conducted. Descriptive statistics are presented as the mean ± standard error of the mean (SEM) from technical replicates. Results were considered significant at the 95% significance level (p < 0.05). The statistical analysis was performed using R software (R Core Team).

Phenotypic details of a proband
The proband was a 23-year-old girl and the second child of healthy nonconsanguineous parents. She was born after a full-term delivery with normal birth parameters after an uneventful pregnancy (Fig 1A). Her developmental milestones were severely delayed. She attained head control at the age of 6 months. She could not roll over but could sit with support at the age of 4 years. At 5 years of age, she developed epilepsy triggered by an influenza virus infection and began taking an anti-epilepsy drug. Brain magnetic resonance imaging (MRI) at 5 years of age showed delayed myelination of the cerebrum and hypoplasia of the corpus callosum. Her intelligence quotient was less than 35 at 6 years of age and she could not stand alone nor speak any meaningful words. The progression of her leukodystrophy, especially in the frontal lobe, was confirmed by MRI at 23 years of age (Fig 1B and 1C). On physical examination, she exhibited several systemic complications and distinctive facial features (Fig 1A and  1D) (for additional clinical information, see S1 Table).

Mutation screening and assessment
At 20 years of age, we conducted trio WES analysis of the patient and her parents and identified a novel heterozygous missense variant in CERT1 (NM_001130105.1:c.787T>C:p.[Ser263-Pro]) (Fig 2A and 2B). Trio Sanger sequencing confirmed that this was a de novo variant ( Fig  2C) that was absent in the control healthy population found in the genome aggregation database (gnomAD; https://gnomad.broadinstitute.org/) and the Japanese multi omics reference panel (jMorp; https://jmorp.megabank.tohoku.ac.jp/202001/variants). In silico pathogenicity prediction tools indicated strong pathogenicity of this variant (Table 1). According to the American College of Medical Genetics and Genomics (ACMG) guidelines [14], this variant was pathogenic. The CERT1 probability score that predicts the presence of autosomal dominant changes was high (pAD = 0.849) [23], while the CERT1 probability score for the loss-offunction intolerant was not high (pLI = 0.67, observed/expected metric = 0.21) (https:// gnomad.broadinstitute.org/gene/ENSG00000113163?dataset=gnomad_r2_1). These data indicated that CERT1 likely acquires pathogenicity by dominant active or negative mechanisms and is likely not a haploinsufficiency gene. This is consistent with the fact that all the previously reported CERT1 variants that cause ID or autism spectrum disorder (ASD) were missense variants, and not truncation variants ( Fig 2D and Table 1) [4-7, 9, 24, 25].

Predominant form of CERT1 transcripts
CERT1 contains 20 exons and produces at least three alternatively spliced variants, which we have named isoforms 1-3 (S1 Fig). The isoform 3 transcript is the longest CERT1 mRNA isoform that encodes a putative very long variant of CERT (CERT/VL). However, RNA sequencing data from human tissues found in the Genotype-Tissue Expression project (GTEx; https://gtexportal.org/home/) indicated that the expression of exon 1 from CERT1 (that is present in isoform 3) is absent or very rare in all human tissues except for the testis (S2 Fig). In contrast, the CERT1 isoform 1 transcript consisting of exons 2-11 and 13-19 was the most abundant type. Moreover, the isoform 2 transcript consisting of exons 2-19, which encodes a long variant of CERT (CERT/L) is expressed but at lower levels than isoform 1 in most tissue types (S2 Fig). These expression levels are consistent with a previous mRNA hybridization study [26]. Hereafter, we refer to isoform 1 as the predominant CERT1-derived transcript, which is translated into the 598 amino-acid protein CERT (NP_112729.1) (S1 Fig). Our ID patient contained a mutation in isoform 1, which is translated into a variant form of CERT that substituted a serine at position 135 for a proline (S135P) in the SRM.

Characterization of CERT in trio-derived permanent cell lines
Our previous study demonstrated that phosphorylation of multiple S/T residues of the CERT SRM down-regulated the activity of CERT. Moreover, a S132A substitution in CERT renders the protein constitutively active [12]. Recent human WES studies have revealed various examples of missense mutations (e.g., S132L, S135C and S138C) that generated amino acid replacements in the SRM of CERT and induced inherent ID disorders [4,5,7]. However, it has not been determined whether the ID-associated missense mutations in CERT1 affected the phosphorylation status of the CERT SRM. Thus, we examined this phenomenon using trioderived B-cell lines that were established by Epstein-Barr virus infection. A maximum of 10 phosphorylations occur on the SRM of CERT and the SRM-hyperphosphorylated form of CERT can be separated from its dephosphorylated and/or hypophosphorylated (de/hypophosphorylated) forms in SDS-PAGE [12]. The hyperphosphorylated form of CERT is far more abundant than the de/hypophosphorylated forms in CHO cells [10], HeLa cells [12], and human colon cancer-derived HCT116 cells (S3B Fig). Moreover, Western blotting of lymphoblastoid cell (LCL) extracts derived from our ID patient and her parents showed that the hyperphosphorylated form of CERT was more abundant than the de/hypophosphorylated forms in both the mother and father, while the de/hypophosphorylated forms were more abundant than the hyperphosphorylated form in the patient (Fig 3A). These results are in line with the finding that the patient's mutation is heterozygous and suggest that the CERT S135P mutant is incapable of becoming hyper-phosphorylated. A faint band comprising the hyperphosphorylated form of CERT/L was also observed in all trio LCLs (Fig 3A). The de/ hypophosphorylated forms of CERT/L were not clearly observed, presumably due to their size overlap with the hyperphosphorylated form of CERT (Fig 3A). When cell lysates were treated with the lambda protein phosphatase (λPPase), the Western blotting patterns of the trio were converted into two bands. The upper and lower bands correspond to the de-phosphorylated forms of CERT/L and CERT, respectively ( Fig 3B). The phosphatase treatment experiment validated the accuracy of the assignments of the various CERT forms in the untreated samples. It should also be noted that no protein derived from the putative CERT1 isoform 3 transcript was detected in LCLs (Fig 3A and 3B) and HCT116 cells (S3B Fig).
In order to explore whether the ID-associated mutations rendered CERT abnormally active, we conducted metabolic labelling of sphingolipids with radioactive serine. All the types of sphingolipids and several types of glycerophospholipids use serine as their metabolic precursor and CERT activity in living cells can be semi-quantitatively monitored by the level of de novo SM synthesis [10]. Under our experimental conditions, no significant difference was observed in the de novo SM synthesis among the trio LCLs, although the synthesis of SM in patientderived LCLs was slightly higher than that in the parents (Fig 3C).

Characterization of CERT S135 mutants in the absence of the endogenous wild-type CERT
In order to confirm that the CERT S135P mutant protein is incapable of being hyperphosphorylated, we characterized CERT mutants in the absence of the endogenous wild-type (WT) CERT. For this purpose, we disrupted both alleles of CERT1 in HCT116 cells, and various constructs encoding CERT mutants tagged with influenza hemagglutinin (HA) or monomeric Venus (mVenus) were stably expressed in the HCT116 CERT1 knockout (KO) cells. Like the endogenous CERT in HCT116 cells, the ectopically expressed HA-or mVenus-tagged WT CERT displayed a major hyperphosphorylated form (Fig 4A and S3C Fig). In contrast, all the ectopically expressed HA-or mVenus-tagged CERT S135P, S135C, and S135A mutant constructs only displayed de/hypophosphorylated forms, which indicated that these mutants were not hyperphosphorylated on the SRM.
Next, we conducted metabolic labeling of lipids with radioactive serine. Metabolic labelling of SM, but not other sphingolipids (i.e., ceramide and glucosylceramide) nor glycerophospholipids (i.e., phosphatidylserine and phosphatidylethanolamine), was markedly reduced in CERT1 KO cells, compared with parental HCT116 cells (Fig 4B and S3D Fig). This reduction was restored to the parental control levels when either mVenus-tagged WT CERT or CERT/L was expressed in the CERT1 KO cells (S4 Fig). When the mutant CERT S135A, S135C, or S135P constructs were expressed, the levels of labelled SM were significantly higher than the level observed in the WT CERT rescued cells, despite lower expression levels of these CERT mutants compared to WT CERT levels (S3C Fig). The sensitivity to lysenin, an SM-binding cytolysin, was monitored as another measure of SM synthesis. Cells expressing the CERT mutants exhibited higher lysenin sensitivity (Fig 4C), which indicated that the intracellular CERT S135A, S135C, and S135P mutants exhibited higher activity than the WT CERT.
When mVenus-tagged CERT variants were expressed in HCT116 CERT1 KO cells, the WT CERT was diffused throughout the cytoplasm with a partial enrichment in the peri-nuclear regions, which are adjacent to a 130 kDa cis-Golgi matrix protein (GM130) (Fig 4D). Interestingly, all the CERT S135 variants exhibited a punctate distribution pattern, and part of the puncta co-stained with the vesicle-associated membrane protein-associated protein (VAP)-A (Fig 4D).

The abnormal phenotypes of CERT S135 mutants require their Golgi-and ER-binding activities
CERT is recruited to the Golgi apparatus via binding of its pleckstrin homology (PH) domain to phosphatidylinositol 4-phosphate (PtdIns4P) in the trans-Golgi regions [10] while CERT is associated with the ER via binding of its diphenylalanine in an acidic tract (FFAT)-motif to VAP [27]. Replacement of a glycine at position 67 to glutamic acid (G67E) in CERT disrupts the PtdIns4P-binding activity of the PH domain [10]. The punctate distribution of the CERT S135P mutant was abrogated by an additional mutation of either G67E (i.e., G67E/S135P) or deletion of the FFAT-motif (i.e., S135P/ΔFFAT) (Fig 5A and 5B). When the intracellular activity of CERT was assessed by both the lysenin sensitivity assay and de novo SM synthesis, the enhanced activity of CERT conferred by the S135P mutation was partially blocked by the additional mutation of G67E or ΔFFAT (Fig 5C and 5D). Together with our previous studies showing the analogous behavior of other activated CERT mutants (i.e., S132A and S315E) [12,28] that exhibited punctate distributions, these results suggested that the constitutively activated CERT mutants commonly display a subcellular punctate distribution, depending on their ability to associate with both Golgi-embedded PtdIns4P and ER-embedded VAP-A. Although the entity of the CERT-associated puncta remains unclear, the common intracellular punctate distribution pattern of various activated CERT mutants may be applicable as a molecular diagnostic assay to assess whether CERT is abnormally activated by CERT1 mutations.

Discussion
The novel CERT1 missense mutation in our patient is located in the SRM, which is also the site of other CERT1 mutations previously found in patients with ID ( Fig 2D) [4,5,7]. Biochemical characterization revealed that amino acid substitutions at S135 in CERT resulted in the loss of SRM hyperphosphorylation, thereby abnormally activating the CERT mutants. Thus, ID patients with activated CERT might be treated with pharmacological inhibitors of CERT activity. To this end, small chemicals that inhibit the inter-membrane ceramide transfer activity of CERT in cultured cells have been developed [29][30][31].
Our patient exhibited multiple congenital anomalies including cerebral leukodystrophy, skeletal abnormalities and distinctive facial features (Fig 1). It is currently infeasible to determine whether these traits are typical in patients with abnormally activated CERT because clinical features of patients with CERT1 mutations have not been reported in detail [4-7, 9, 24, 25]. One case of a patient with a substitution of a glycine at position 243 for an arginine (G243R) in CERT exhibited unilateral renal aplasia as well as abnormal brain structures, which included dilated ventricles and periventricular leukomalacia [6]. Additional clinical case reports are required to elucidate whether congenital anomalies other than ID are typical features of CERT1 mutation. Considering that CERT mutations have been identified as the cause of ASD outside of SRM [24,25], future case accumulation may indicate that CERT is the causative gene of neurodevelopmental disorders, including ID, developmental delay, and ASD.
Importantly, the ID-inducing missense mutations in CERT1 described previously have been classified as heterozygous dominant mutations. However, we failed to detect a significant difference in SM synthesis among the trio-derived LCLs by metabolic lipid-labeling experiments and lysenin tolerance assay (Fig 3C and 3D). Although we currently lack a conclusive answer to this discrepancy, there are a few possible explanations: (1) both the metabolic lipidlabeling experiment and lysenin tolerance assay might not be sensitive enough to detect an enhancement in intracellular CERT activity and (2) LCLs might not be a good cell model to monitor changes in the activity of CERT with ID-inducing mutations.
Our patient exhibited delayed cerebrum myelination. Galactosylceramide (GalCer) and its sulfonated metabolite are abundant in the brain and testis, and are essential for the electrophysiological function and stability of the myelin sheath [32,33]. Since ceramide is a common precursor for SM, glucosylceramide, and GalCer, the enhanced production of SM due to constitutively activated CERT might induce a shortage of the precursor ceramide for GalCer, thereby inhibiting myelination. In order to test this hypothesis, further studies including the generation of a model ID animal with a heterozygous CERT1 mutation is required. In conclusion, the present study provided a possible molecular basis for not only new diagnostics but also a conceivable pharmaceutical intervention for ID disorders caused by gain-offunction mutations in CERT1. The heatmap summarizes the median read counts per base of each exon across all tissues. This indicated that the transcript containing exon1 is barely detected in all tissues except for testis and Epstein-Barr virus-transformed lymphocytes. In most tissues, RNA-seq short reads from exon12 were detected at lower frequencies than those from exon11 or exon13. (B) Junction expression data of CERT1 (https://gtexportal.org/home/ transcriptPage). The heatmap summarizes the median raw read counts of junctions from individual isoforms. This demonstrated that read counts of junction21 connecting exon1 with exon2 were barely detected, which indicated that isoform 3 of CERT1 is not expressed in most human tissues. The read counts of junction10 connecting exon11 with exon13 were detected at higher frequencies than junction9 or junction11, which indicated that the isoform 1 of CERT1 is the predominant transcript in most human tissues. (PDF)

S3 Fig. Establishment of CERT1 KO HCT116 cells. (A)
Diagram of the location of gRNAtargeted sequences of human CERT1. The 5'-untranslated regions (UTRs) and coding regions are shown as white and black rectangles, respectively. The red lines represent the first and second methionines. In the WT allele, the protospacer adjacent motif (PAM) sequence is shown in bold. The red highlighted sequence (ATG) indicates the second methionine within exon2. The gRNA-targeted sequence is underlined. A 1 bp insertion that generates a frame-shift mutation is shown in blue in both CERT1 alleles from the CERT1 KO cell line. (B) Cell lysates prepared from WT and CERT1 KO HCT116 cells were incubated with or without λPPase and analyzed by Western blotting with the indicated primary antibodies. (C) WT and CERT1 KO HCT116 cells stably expressing various mVenus-CERT constructs were lysed and subjected to Western blotting analysis with the indicated primary antibodies. (D) WT and CERT1 KO HCT116 cells stably expressing the indicated mVenus-CERT variants were cultured with L-[U-14 C]serine for 16 hr. Metabolically labelled lipids were separated by TLC analysis and visualized using an image analyzer. A representative image is shown. PC, putative phosphatidylcholine. Note that serine is directly incorporated into PS by the base exchange reaction with PC, and labelled PS is converted to PE by decarboxylation [34,35]. Metabolic labeling of PC with L-[U-14 C]serine can occur by two different pathways [34,35]. One pathway occurs via Nmethylation of labelled PE, and the other pathway occurs via the conversion of serine to pyruvate, which is then anabolized to fatty acids via acetyl-CoA, followed by the synthesis of PC