Comprehensive investigation of CASK mutations and other genetic etiologies in 41 patients with intellectual disability and microcephaly with pontine and cerebellar hypoplasia (MICPCH)

The CASK gene (Xp11.4) is highly expressed in the mammalian nervous system and plays several roles in neural development and synaptic function. Loss-of-function mutations of CASK are associated with intellectual disability and microcephaly with pontine and cerebellar hypoplasia (MICPCH), especially in females. Here, we present a comprehensive investigation of 41 MICPCH patients, analyzed by mutational search of CASK and screening of candidate genes using an SNP array, targeted resequencing and whole-exome sequencing (WES). In total, we identified causative or candidate genomic aberrations in 37 of the 41 cases (90.2%). CASK aberrations including a rare mosaic mutation in a male patient, were found in 32 cases, and a mutation in ITPR1, another known gene in which mutations are causative for MICPCH, was found in one case. We also found aberrations involving genes other than CASK, such as HDAC2, MARCKS, and possibly HS3ST5, which may be associated with MICPCH. Moreover, the targeted resequencing screening detected heterozygous variants in RELN in two cases, of uncertain pathogenicity, and WES analysis suggested that concurrent mutations of both DYNC1H1 and DCTN1 in one case could lead to MICPCH. Our results not only identified the etiology of MICPCH in nearly all the investigated patients but also suggest that MICPCH is a genetically heterogeneous condition, in which CASK inactivating mutations appear to account for the majority of cases.


Introduction
The CASK gene (OMIM: Ã 300172), which encodes a member of the MAGUK (membraneassociated guanylate kinase) protein family, is highly expressed in the mammalian nervous system of both adults and fetuses, and plays several roles in neural development and synaptic functions [1,2]. Since 2008, when both a heterozygous deletion and point mutations affecting CASK were reported to cause intellectual disability and microcephaly with pontine and cerebellar hypoplasia (MICPCH, OMIM: #300749) [3,4], various types of CASK aberrations have been reported in more than 50 MICPCH patients [5,6]. Due to its location on the X chromosome, loss-of-function of CASK usually leads to the manifestation of MICPCH in females, whereas a complete loss of this gene is believed to be lethal in males. Indeed, the vast majority of typical MICPCH by CASK aberrations are seen in female patients [4][5][6][7][8].
Since our first report [3], we have recruited 41 MICPCH patients to investigate their etiologies by genetic testing of CASK and other genome-wide approaches such as single nucleotide polymorphism (SNP) array, targeted resequencing and whole-exome sequencing. Besides 32 cases in which the molecular diagnosis could be attributed to CASK loss-of-function mutations, including a rare somatic mosaic mutation in a male patient, we identified a mutation in ITPR1, another gene in which mutations are causative for MICPCH. Moreover, in the analysis we identified aberrations in genes that might be associated with MICPCH: HDAC2, MARCKS, RELN, and possibly HS3ST5. Interestingly, our findings also suggested that concurrent mutations of both DYNC1H1 and DCTN1 might cause MICPCH. Our current study comprehensively clarified the etiologies of MICPCH in the cohort of patients and suggested heterogeneity of MICPCH.

Results
Overall, we identified causative or candidate genomic aberrations in 37 of the 41 patients (90.2%) ( Table 1). Briefly, 23 patients had point mutations of CASK, nine patients had CNVs affecting CASK and five had aberrations involving other genes. Among all, parental samples were available for 13 patients (Table 1). Their samples were also tested to confirm that all the patients' mutations were de novo, except for patient 37.

Subjects
We recruited 41 patients clinically diagnosed with MICPCH, including 16 patients who have previously been reported [3,7,9] (Table 1). All patients were examined and evaluated by clinical dysmorphologists in each hospital and were enrolled in this study if they met the criteria for both microcephaly and pontocerebellar hypoplasia in brain magnetic resonance imaging (MRI) (Fig 1). All the patients were Japanese.
Patient's clinical features are shown in Table 1. The cohort was largely comprised of females, with only six male patients, and their ages varied between five months to 17 years old at the last follow-up examination. Development has been markedly retarded, and some of them also showed seizures. Only six patients could walk at the time of the evaluation, and only three could speak a few words. While the occipital-frontal circumference (OFC) at birth ranged from normal to severe microcephaly in a few cases, at the last examination most of the patients showed a severe microcephaly (<-3.0SD). Although three cases (patients 36, 37, and 41) had OFC larger than -3.0 SD, we decided to include them in this cohort because their pontocerebellar hypoplasia resembled those of the other patients. Although most patients did not show major anomalies besides microcephaly, some of them also presented with ocular abnormalities and/or muscular hypotonia. From the results of the brain MRI examination, the severity of the pontocerebellar hypoplasia was found to be variable, whereas the size of the corpus callosum tended to be normal [10]. DNA and chromosomes were extracted from peripheral blood by standard methods. A lymphoblastoid cell line (LCL) was also established for all patients and available samples of parents (Table 1) by infecting lymphocytes with an Epstein-Barr virus, as previously described [11].
Missense mutations were detected in three cases, and in silico analysis predicted them to be disease-associated variations. The c.1910G>A (p.G637D) variant in patient 20, c.1061T>C (p. L354P) variant in patient 21, and c.317G>C (p.R106P) variant in patient 22 were predicted in silico to be damaging or disease-associated variation by PolyPhen-2, SIFT, or SNPs&GO, respectively, and their REVEL scores were also high (S1 Table).
Among the six male patients, CASK aberrations were identified in three; other than patients 21 and 22 described above, Patient 23 who presented with typical MICPCH (Fig 1) was found to have an indel between exon 15 and intron 15 resulting in a stop codon in a heterozygouslike pattern, despite being male (2A and 2B). This result suggests that the patient had the mutation in a mosaic state. Primer sets specific for the indel mutation amplified a product only in this patient ( Fig 2C). Additionally, real-time quantitative PCR with the same primer sets confirmed a decreased copy number of both intact and indel alleles relative to a male control ( Fig 2D). Thus, we concluded this patient is a rare case of a somatic mosaic indel mutation CNV involving CASK and other candidate genes SNP array led to the discovery of heterozygous deletions involving whole or part of the CASK gene in six cases (patients 24 to 29), and intragenic duplications in three cases (patients 30 to 32) (Table 1). We have previously described patients 30 and 31 in detail [7]. In patient 32, oligonucleotide array determined exons 4 and 5 to be duplicated (S1 Fig). In order to precisely map the rearrangement, genomic PCR was performed with a primer combination specific to a tandem duplication structure (S2 Fig), and a product was detected only in the patient (S3 Fig). Sequencing of this product showed that intron 5 was followed by a duplicated intron 4 (S4 Fig), resulting in a shift in the reading frame and thereby producing an aberrant transcript. As a consequence, all intragenic duplications probably also resulted in loss-of-function of CASK.
Fine mapping of the CNVs using primer sets flanking the breakpoint (BP) junctions was performed in seven cases, patients 24-27 and 30-32. One case, patient 25, had a 126-bp sized Alu repeat at the BP (S5 Fig, upper), indicating it was likely mediated by non-allelic homologous recombination (NAHR) [12,13]. In contrast, one case, patient 26, had no homology at the BP (S5 Fig, middle), therefore the rearrangement might have been induced by non-homologous end joining (NHEJ). The other five cases, patients 24, 27, and 30-32, had a microhomology of 1-3 base pairs at the BP junction (S5 Fig, lower), likely caused by microhomology-mediated break-induced replication (MMBIR). These findings suggest that most of the CNVs were not induced by a specific motif, suggesting CNVs involving CASK are probably non-recurrent ( Table 1).
The SNP array analysis also identified a CNV not encompassing CASK in patient 33, who presented with typical MICPCH (Fig 1): a 13-Mb heterozygous deletion at 6q21-q22.31 (Table 1; Fig 3A and 3B). This result is concordant with previous reports of deletion overlapping with this case also showing MICPCH [14,15]. We confirmed she had no mutations of CASK (data not shown).

Targeted resequencing
Among 231 mutations detected in the six cases, only two mutations were in exons of RELN [OMIM: Ã 600514] (S1 Table). One was c.4918A>G (p.I1640V) in patient 34, who is female and shows typical MICPCH (Fig 1), and the other was c.7093G>A (p.V2365M) in patient 35, who is male and also shows typical MICPCH (Fig 1). Both of the mutations were confirmed by direct sequencing (Fig 3C). The substituted isoleucine and valine are highly conserved at least in primates and placental animals (Fig 3D), suggesting that both amino acids are important for RELN function. In-silico analyses found p.I1640V not to be damaging ('tolerated' by Poly-Phen-2 and SIFT, and 'neutral' by SNPs&GO), and scores of REVEL and CADD also suggested that mutation was not probably damaging (S2 Table). However, the mutated residue was predicted to be smaller than the wild type (WT), leading to loss of interactions, according to prediction by HOPE. The evaluation of p.V2365M may be vague: it was evaluated to be 'damaging' by PolyPhen-2 and SIFT and CADD scores were very high, while SNPs&GO judged it 'neutral' and REVEL score was rather low. HOPE predicted the 3D structure to be bigger than the WT and the mutated residue might prefer to be in another secondary structure, thus the local conformation would be slightly destabilized.

Whole-exome sequencing
In patient 11 eight single-nucleotide variations (SNVs) and four indel mutations were extracted as candidate mutations. Among them, one insertion of a single guanine, c.1677dupG (p.R560Afs Ã 20) in exon 18 of CASK caused a frameshift leading to a stop codon ( Fig 3E). As the insertion was located at continuous guanines and close to the exon-intron junction, it might have been overlooked in the previous steps by the direct sequencing and targeted resequencing.
In patient 36, candidate SNVs were identified in three genes, ITPR1, WNK3, and CYCL1. Among them, c.7753A>C (p.T2585P) in exon 58 of ITPR1 [OMIM: Ã 147265] was functionally the strongest candidate (Fig 3E). Heterozygous mutations of ITPR1 are known to cause . Specifically, SCA29 is characterized by a psychomotor delay with an onset in the infantile period and an atrophic cerebellum, but not severe microcephaly [16], which is consistent with the mild microcephaly shown by patient 36 (Table 1, Fig 1). In total, in-silico analyses suggested this variant is probably diseaseassociated (S2 Table), and HOPE predicted that the mutation introduces a more hydrophobic residue at this position, hence the 3D structure of the protein might disrupt an α-helix, impairing the correct folding. Taken together, these findings indicate that this mutation may account for the patient's etiology.
In patient 37, since the de novo analysis and compound heterozygous variant filtering failed to indicate a functionally significant candidate gene, we performed a protein-protein network analysis in order to assess possible interactions between the mutated genes found in the patient. Thus, all novel SNVs extracted in patient 37 were evaluated by the STRING database (http://string-db.org/), which indicated a close interaction between DYNC1H1 [OMIM: Ã 600112] and DCTN1 [ Ã 601143], both being missense variants (Fig 3F, S6 Fig). While the maternally-inherited c.11824C>T (p.P3942S) variant in DYNC1H1 was predicted in silico as benign to neutral with lower REVEL score, the paternally-inherited c.497C>G (p.S166C) mutation in DCTN1 was predicted to be neutral to damaging (S2 Table). Despite the fact that the in-silico predictions did not show strong evidence of pathogenicity of the variants alone, the possibility remains that the combined occurrence of both variants in the same protein complex might be associated with the patient's phenotype.

Discussion
Since the first report of CASK aberrations associated with MICPCH, more than 50 cases have been reported [4][5][6][7]. Here we summarized our investigation of 41 MICPCH cases, and in total we were able to identify causative or candidate genomic aberrations in 37 out of 41 cases (90.2%) ( Table 1). Among the 37 positive cases, 23 cases (57.6%) had a point mutation of CASK and nine cases (27.3%) had CNV involving CASK (Table 1, Fig 4A and 4B). We also found five cases, two males and three females, with aberrations in other causative and/or candidate genes. For 13 cases whose parental samples were available, analysis of the parental samples showed that all the mutations were de novo, except patient 37 who may have compound mutations from both of the parents (see below).
Mapping of the breakpoint junctions of seven cases with the CNVs including CASK revealed that they might have been originated by several mechanisms of genomic rearrangements, namely NAHR, NHEJ and MMBIR (S2A- S2C Fig), suggesting that CNVs involving CASK are generally non-recurrent. This non-recurrent nature of CASK CNVs is concordant with the fact that most CASK-related point mutations arise de novo.
Atypical cases in this cohort were represented by three male cases found with CASK aberrations. Two cases (patient 21 and 22) had a missense mutation [9], and the other case (patient 23) was a somatic mosaic of WT and indel alleles (Fig 2). Recent studies have shown that CASK mutations in male patients lead to a broad spectrum of phenotypes including MICPCH. While missense mutations corresponding to functional change of the CASK protein causes non-syndromic intellectual disability [17][18][19] or FG syndrome [20], a decrease in the normal expression appears to correlate with the manifestation of MICPCH [8]. The complete loss of CASK expression was associated with a severe MICPCH phenotype and likely caused the reduced viability or in utero lethality [4,8], which is consistent with the neonatal lethality of Cask knockout mice [21]. Thus, CASK-inactivating mutations in mosaic state (like the case of patient 23) might be able to escape early lethality due to incomplete loss of CASK expression, thereby resulting in typical MICPCH.
Aside from CASK, we also found another known gene in which mutations are causative for MICPCH, through the identification of an ITPR1 mutation in patient 36 by WES (Fig 3E). ITPR1 has previously been established as a causative gene of SCA29 associated with psychomotor delay and atrophy of the cerebellum, likely consistent with the brain malformation of patient 36 [16]. However, it may be distinguished from typical MICPCH caused by CASK mutation by a relatively milder microcephaly, thicker corpus callosum and non-atrophic brainstem.
Our investigation found a few genes that might be associated with MICPCH. Our findings in patient 33 support evidence linking the deletion of 6q21 to MICPCH (Fig 3A and 3B), concordant with previous reports of typical MICPCH cases with deletion at 6q21 [14,15]. In particular, Case 7 reported by Rosenfeld showed typical MICPCH and had the smallest deletion, including only three protein-coding genes, MARCKS [OMIM: Ã 602940], HDAC2 [ Ã 605164], and HS3ST5 [ Ã 609407]. MARCKS is reported to bind actin to support cytoskeletons, and HDAC2 is a histone deacetylase that acts as a suppressor of the expression of other genes [22]. While its function remains uncertain, HS3ST5 is highly expressed in the fetal brain [23]. These findings strongly suggest that haploinsufficiency of HDAC2, MARCKS, and possibly HS3ST5 may be relevant to MICPCH, even though Hdac2 or Marcks mutant mice showed conflicting phenotypes. Marcks +/mice showed no brain abnormality, while Marcks -/resulted in perinatal lethality with abnormal brain development [24]. Both heterozygous and homozygous Hdac2deficient mice were viable and had normal brains, and additionally, the spine density was markedly increased in Hdac2 -/mice [25]. On the other hand, mice with conditional knockout of both Hdac2 and Hdac1, a homologous gene of Hdac2, showed severe hippocampal and cerebellar abnormalities [26]. Although these results suggested that either HDAC2 or MARCKS alone might not cause brain malformations, haploinsufficiency of both genes may be relevant to MICPCH due to the collapse of the interaction between them.
Targeted resequencing screening only detected heterozygous mutations in RELN in two cases, of uncertain pathogenicity (Fig 3C). RELN encodes a secreted protein playing an important role in neural migration in the developing brain and spinal cord [27,28]. While homozygous loss-of-function mutations of RELN cause lissencephaly with cerebellar hypoplasia [OMIM: #257320] [29,30], heterozygous missense mutations of RELN cause autosomal-dominant lateral temporal epilepsy (ADLTE) with no brain anomaly [31]. Although the nature of the current mutations remain uncertain, an interesting report that CASK regulates the expression of RELN through interactions with TBR1 may suggest that RELN would be also associated with MICPCH [32].
WES analysis also identified an interesting pair of mutations of DYNC1H1 and DCTN1 in patient 37 (Fig 3F), respectively inherited from the mother and father. DYNC1H1 encodes a member of the cytoplasmic dynein heavy chain [33], while DCTN1 encodes p150(Glued), the largest subunit of dynactin [34,35]. The dynein-dynactin complex plays an important role during mitosis and is necessary for synapse stabilization [36]. Heterozygous mutations in the motor domain of DYNC1H1 have been associated with autosomal dominant mental retardation-13 (MRD13) with neuronal migration defects [OMIM: #614563], in which a subset of patients also present with microcephaly along with a small cerebellum and/or brainstem [37,38]. However, the vast majority of the reported mutations were de novo. Mutations in DCTN1 cause distal hereditary motor neuronopathy type VIIB [OMIM: Ã 607641] and, so far, there is no association with brain malformations. Despite the uncertainty of the clinical significance of both variants, one might speculate the possibility of a 'double-hit' effect similar to a digenic inheritance model, in which variant genotypes at two loci (often involving interacting proteins) explain the phenotypes of some patients more clearly than the genotypes at one locus alone [39]. This tentative etiology may also explain the reason why the phenotype of patient 37 is relatively milder (Table 1). Nevertheless, further studies are needed to assess the combined effect of the two variants in the stability of the dynein-dynactin complex.
Our present study comprehensively explained the etiologies of MICPCH using several genomic analysis techniques. Not only did it identify CASK but also other genes involved in the etiology of MICPCH. Our findings demonstrate that MICPCH is a genetically heterogeneous condition, in which CASK inactivating mutations appear to account for the majority of MICPCH cases and with severer phenotypes, while the non-CASK mutation cases tend to have milder microcephaly. For the remaining cases, WES or WGS analysis may identify the undiscovered causative genes.

Informed consent
Written informed consent was obtained from the parents of all the subjects. The consent form includes an agreement to have their clinical details and genetic analyses published without personal information.

Direct sequencing of CASK
Point mutations within the coding region of CASK were analyzed by exon amplification and direct sequencing as previously described [7].

Microarray analysis
For patients in whom no pathogenic CASK point mutation was found, SNP array (Huma-nOmniExpress-12 v1.0; Illumina, San Diego, CA, USA) was performed in order to detect genomic copy number variant (CNV) involving CASK. Oligonucleotide array (Human CGH Array 2.1Mb; Roche-Nimblegen, Madison, WI, USA) was subsequently applied to the positive cases to better determine the extent of the CNVs. DNA labeling, hybridization and washing were carried out according to the directions provided by the manufacturer. All CNVs were confirmed by fluorescence in situ hybridization (FISH), performed as previously described [40].

Genomic PCR and real-time quantitative PCR
Genomic PCR was performed to amplify and sequence the breakpoint junctions of the CNVs. We also performed genomic PCR to confirm the presence of mosaicism in patient 19 by designing intact-specific and mutation-specific primer sets (S3 Table). The same primer sets were used for real-time quantitative PCR assay in order to determine the proportion of mosaic alleles on the genomic DNA of the patient, using the 7500 Real-Time PCR System (Applied Biosystems, Grand Island, NY, USA) and KAPA SYBR1 FAST qPCR Master Mix (KAPA Biosystems, Wilmington, MA, USA), according to the manufacturers' instructions.

Targeted resequencing
Samples of six patients (patients 11, 34, 35, 38, 39, and 41) in whom neither point mutation of CASK nor CNVs involving CASK had been identified in the previous steps were subjected to targeted resequencing (amplicon sequencing) on the Ion Torrent PGM platform (Life Technologies, Carlsbad, CA, USA). Primer sets targeting the whole CASK region, including all exons including 5' and 3' untranslated regions (UTR), introns, promoter region, and all exons of 16 candidate genes interacting with CASK or corresponding to pontocerebellar hypoplasia in the literature (Table 2), were designed using the manufacturer's AmpliSeq primer design tool (http://www.ampliseq.com). Library preparation and sequencing were performed according to the manufacturer's protocol. Base-calling and alignment to the human reference genome (GRCh37/hg19) were carried out using Torrent Suite 3.6 (Life Technologies). Variant detection was performed with the Variant Caller plug-in software of Torrent Suite.

Whole-exome sequencing
Among the eight negative cases after the targeted resequencing, we applied whole-exome sequencing (WES) for three cases (patients 11, 36, and 37) whose parental samples were available. Exome library was prepared using the SureSelect Human All Exon V5+UTRs Kit (Agilent, Santa Clara, CA, USA) and sequenced using the HiSeq2500 platform with Cluster/ SBS kit v3 (Illumina). We extracted candidate mutations as follows: de novo variations were first extracted by comparing the sequence data between the trio of patient and parents, followed by filtering of variants with low read-depth (RD<4) or those occurring at non-coding regions. Next, variations observed in healthy populations were excluded in reference to a database of known