Genetic evidence for splicing-dependent structural and functional plasticity in CASK protein

Abstract

Background:

Pontocerebellar hypoplasia (PCH) may present with supratentorial phenotypes and is often accompanied by microcephaly. Damaging mutations in the X-linked gene CASK produce self-limiting microcephaly with PCH in females but are often lethal in males. CASK deficiency leads to early degeneration of cerebellar granule cells, but its role in other regions of the brain remains uncertain.

Method:

We generated a conditional Cask knockout mice and deleted Cask ubiquitously after birth at different times. We examined the clinical features in several subjects with damaging mutations clustered in the central part of the CASK protein. We have performed phylogenetic analysis and rt-PCR to assess the splicing pattern within the same protein region and performed in silico structural analysis to examine the effect of splicing on CASK’s structure.

Result:

We demonstrate that deletion of murine Cask after adulthood does not affect survival but leads to cerebellar degeneration and ataxia over time. Intriguingly, damaging hemizygous CASK mutations in boys who display microcephaly and cerebral dysfunction but without PCH are known. These mutations are present in two vertebrate-specific CASK exons. These exons are subject to alternative splicing both in forebrain and hindbrain. Inclusion of these exons differentially affect the molecular structure and hence possibly function/s of the CASK C-terminus.

Conclusion:

Loss of CASK function disproportionately affects the cerebellum. Clinical data, however, suggest that CASK may have additional vertebrate-specific function/s that play a role in the mammalian forebrain. Thus, CASK has an ancient function shared between invertebrates and vertebrates as well as novel vertebrate-specific function/s.

Introduction

Pontocerebellar hypoplasias (PCH) are a heterogenous group of degenerative conditions that predominantly affect the volume of the cerebellum and pons [1]. Based on clinical criteria, there are many different types of PCH which are usually progressive with poor prognosis [2]. Multiple studies have identified variants within the X-linked CASK (calcium/calmodulin dependent serine protein kinase) gene as one of the common causes of PCH [3]. In fact, variants in CASK have been observed in studies of PCH types 2, 3 and 4 in the literature [4 5], however CASK-linked PCH remains a unique disorder with non-overlapping etiopathogenesis and symptomatology. Unlike other types of PCH, CASK-linked PCH in girls is thought to be non-progressive. In boys however, CASK mutations produce profound epileptic encephalopathy, supratentorial atrophy with cortical smoothening, and lethality due to dysfunction of the severely hypoplastic brainstem [6–8]. This difference in the presentation and course of the disorder is likely due to X-linkage and random inactivation of the CASK gene in girls [9]. Unlike spino-cerebellar ataxias, PCH usually presents with many cortical (higher) dysfunctions, and severe PCH is associated with overall microcephaly and impaired cortical functions such as intellectual/learning disabilities and seizures, indicating that microcephaly with PCH may just be a more serious manifestation of the same pathological process [2]. It is, however, not known if the molecular pathways producing PCH and microcephaly are identical.

CASK mutation phenotypes are often described as microcephaly with pontine and cerebellar hypoplasia (MICPCH). MICPCH is associated with overall growth retardation, central hypotonia, locomotor incoordination, global developmental delay with intellectual disability, language disability, sensorineural deafness, optic nerve hypoplasia, and seizures in at least 40% of subjects [10 11]. In previous studies we have demonstrated that, similar to human subjects, heterozygous deletion of Cask in mice is sufficient to generate both postnatal microcephaly and cerebellar hypoplasia, indicating that they both result from loss of CASK gene function [12]. However, the overall phenotype in mice is markedly milder than in humans. The heterozygous knockout female mice are reproductive, capable of taking care of their pups, and do not display degenerative changes as measured at 2 years of age. Critically, no specific histological abnormalities have been noted in the forebrain of these mice ([12]) nor any gross electrographic changes in the cortex or hippocampus [13]. Optic nerves and other tissue from these mice, however, display reactive astrogliosis [14]. When CASK is deleted from the cerebellum after neuronal migration is completed, we have documented that cerebellar atrophy results from granule cell loss due to CASK deficiency [9]. Thus, by changing the timing of Cask deletion, we are able to convert cerebellar hypoplasia to cerebellar atrophy with reactive astrogliosis in the Bergmann glia [9]. Strikingly, the neuropathologies observed in Cask mutant mice indicate that the loss of neurons may not depend on a cell-intrinsic CASK function, since cells without CASK remain alive throughout the pathological progression, and furthermore, heterozygous deletion of CASK in mice does not generate any secondary skewing of the X-chromosome[12 15]. Thus, an important remaining question is whether microcephaly and cerebellar hypoplasia have dissociable etiologies owing to unique regulation of either the CASK gene or the molecular function of the CASK protein.

CASK is a multi-domain protein belonging to a family of proteins called MAGUKs (membrane-associated guanylate kinase). CASK has been described as a scaffolding molecule due to its ability to bind several proteins, including membrane proteins such as neurexins [13 16–19]. CASK has also been implicated in transcription activation via interactions with T-box protein Tbr1 and CINAP (CASK interacting nucleosome assembly protein) [20 21]. Although several studies have indicated that MICPCH due to CASK mutation may result from disruption of the CASK-Tbr1 interaction [22 23], experiments from our and other laboratories have indicated that disruption of CASK-neurexin interactions is responsible for the phenotype [24 25]. Still other studies have implicated loss of the CASK-liprins-α interaction as the cause of MICPCH [26–28]. Finally, our group has documented metabolic and mitochondrial functional disruptions in the brain associated with CASK deficiency [12 13]. Despite numerous studies, the exact molecular role of CASK and the mechanism of CASK-linked disorders remains unclear.

Here we develop an inducible knockout model of CASK to demonstrate that, unlike constitutive CASK deficiency, loss of CASK in adulthood does not produce lethality. Further, we demonstrate that cerebellar granule cells are uniquely sensitive to ubiquitous CASK deletion at all ages. Our study thus provides the basis for the disproportionate effect CASK deficiency has on cerebellar volume [29]. Strikingly, several male subjects with damaging CASK mutations fail to display cerebellar hypoplasia, despite presenting with global developmental delay and microcephaly. We demonstrate that these mutations are present on CASK exons that are evolutionary new and vertebrate-specific. The evolutionarily new exons do not code for the canonical domains of CASK, however in silico analysis suggest that the inclusion or exclusion of these exons is likely to affect the structure of the C-terminus of the CASK protein in a manner that may impact homomeric and heteromeric CASK interactions. Thus, we propose that Cask gene function includes both an ancient function that is common between invertebrates and vertebrates crucial for the hindbrain and a vertebrate-specific function that plays a more dominant role only in the forebrain.

Materials and Methods

Statement of ethics

All animal procedures were performed in accordance with the guidelines of the Institutional Animal Care and Use Committee (IACUC) of Virginia Tech. The collection and deidentification of clinical data were performed under guidance of the Virginia Tech Institutional Review Board (IRB).

Development of inducible CASK knockout mice

Cask^floxed mice (JAX Strain #:006382) mice were a gift from Prof. Thomas Südhof and have been published previously [30]. The mice were generated by inserting a loxP site 5’ of exon 1; a neomycin resistance (neo) cassette flanked by two more loxP sites was incorporated in the intron 3’ of exon1 (Cask^floxed). Resultant mice were backcrossed with C57BL/6J background mice for at least 25 generations.

These Cask^floxed mice were then crossed to mice expressing tamoxifen-dependent ER2 (estrogen receptor)-Cre under the CAG promoter (JAX Strain #:004682). Upon intraperitoneal tamoxifen administration in adulthood, Cre-mediated recombination allowed generation of Cask adult knockout mice.

In separate experiments, the ER2-Cre expressing mice were crossed to Ai14-LSL-tdTomato-positive mice (JAX strain 007914) where expression of the fluorescent reporter tdTomato is dependent on functional Cre expression within the nucleus. Degree of observed fluorescence was used to determine the efficacy of Cre-mediated recombination in the presence and absence of tamoxifen.

Brain sectioning and histochemistry.

Adult mice were deeply anesthetized by intraperitoneal injection of tribromoethanol (Avertin) until toe-pinch reflex was absent. They were subsequently euthanized by bilateral thoracotomy, exsanguination, and trans-cardiac perfusion of phosphate buffered saline (PBS) followed by 4% paraformaldehyde. Extracted brains were post-fixed in 4% paraformaldehyde for at least 24 hours before subsequent processing. For staining, brains were sectioned either by Vibratome or cryostat. For Vibratome-prepared sections, post-fixed brains were mounted without embedding on a Vibratome (Leica) and 50–100μm thick sagittal sections were cut from the midline free-floating in PBS. For cryosectioning, brains were cryopreserved by incubation in 30% sucrose solution for 48 hours. They were then embedded in Cryo-Tek and 20 μm thick brain sections were generated using a cryostat (IEC). Both Vibratome and cryostat-prepared floating brain sections were processed similarly in 24-well plates after permeabilization and blocking with 0.1% Triton X-100 and 0.5% fetal bovine serum (FBS) in PBS. Sections were mounted on slides using Vectashield hardset with DAPI mounting medium (Vector laboratories) and coverslips were sealed using nail polish. For optic nerves, tissues were fixed in 2% glutaraldehyde/2% paraformaldehyde (caodylate buffer). Semithin sections were obtained and stained with toluidine blue as previously described[14].

Reverse-Transcription PCR

Mice were anesthetized with intraperitoneal injection of Avertin until toe-pinch reflex was absent. Brain were immediately extracted and divided into forebrain (neocortex and hippocampus) and hindbrain (cerebellum and brainstem). Total RNA was extracted from forebrain and hindbrain separately and subject to reverse transcription PCR using the primers (forward GGGAGTATTACCTTCAAGATTGTG; reverse CTGGATGATGTCACCAACTCG). The generated Cask cDNA samples were then subjected to electrophoresis on a 2.5% w/v agarose gel to separate transcript variants.

Molecular dynamics

Molecular dynamics (MD) simulations using the AMBER99SB-ILDN force field [31] were performed with GROMACS 20.4 [32] on a model derived from a previously described CASK supramodule (PDZ-SH3-GuK; PSG) structure [24]. Exon 19 and 20 were each individually deleted from the PSG structure, and using the Chimera interface for Modeller [33], initial homology models for each splice variant, chosen based on lowest zDOPE score, were created for MD simulations. Structures were solvated (spce water model) in an explicit rhombic dodecahedron water box (solute box distance of 1.0 nm) under periodic boundary conditions, with charges neutralized by chloride ions. Starting structures were energy minimized until convergence at Fmax < 1000 kJ/mol/nm. A 100 ps position-restrained NVT equilibration simulation was run at 300K using a modified Berendsen thermostat, then a 100 ps NPT equilibration simulation using the Parrinello-Rahman barostat was run. After equilibration, an unrestrained 100 ns NPT molecular dynamics simulation was run. Three trajectories, each initiated with different random seeds starting from the minimized structure, were generated for each splice variant.

After post-processing to correct for periodicity, trajectories were analyzed for: 1) root mean square deviation (RMSD) of the protein backbone from the starting structure at each time point (GROMACS rms command), 2) radius of gyration at each trajectory time point (GROMACS gyrate command), and 3) root mean square fluctuation (RMSF) of all alpha carbons (GROMACS rmsf command). Using the equation (8π²/3) × (RMSF)², B-factors were calculated for each residue and averaged across all three trajectories. Cluster analysis of the three concatenated molecular dynamics trajectories (excluding the first 25 ns of each trajectory) for each structure was performed (GROMACS cluster command with a 3.5Å cutoff employing the gromos algorithm [34].

Results

CASK gene function is crucial for the cerebellum throughout the lifespan

MICPCH involves a postnatal progressive microcephaly with a smaller forebrain which is typically detected 3–6 months after birth in children. Deleting CASK constitutively does not produce any immediately obvious gross structural changes in the murine forebrain, but results in early lethality. Heterozygous knockout mice display a smaller forebrain, but no specific histological changes were observed in contrast to cerebellum where a thinning of internal granular layer was observed [12 30]. In order to study the effect of chronic CASK deficiency (~6 months) on forebrain, it is therefore crucial to develop a model where mice survive without CASK expression for a significant period of time, we therefore planned to develop an inducible Cask knockout mouse. To this end, a mouse line expressing Cre-recombinase fused with the estrogen receptor under a pan-mammalian promoter (CAG-ER2-Cre) was crossed with an indicator line carrying the loxP-stop-loxP-tdTomato (LSL-tdTomato) transgene to verify recombination throughout the brain (Figure 1A). Subsequently, we crossed the CAG-ER2-Cre mice with mice carrying hemizygous or homozygous Cask^floxed alleles (Figure 1B). Results indicate that tamoxifen administration rapidly allows recombination in the entire brain, which turns red (Figure 1C). Imaging experiments indicate that while ER2-Cre is slightly leaky without tamoxifen (allowing recombination in 5–12% cells in different nervous tissues), administration of tamoxifen leads to recombination in >90% cells. Recombination is also readily observed in non-nervous tissues like kidney, liver, and intestine (Figure 1D). Consistent with these data, tamoxifen-induced recombination leads to abolition of ~90% CASK expression in brain, leaving other proteins intact (Figure 1E, F). The Cask^floxed mice show reduced expression of CASK due to selection cassette interference as has been previously reported [12 30]. Deletion of CASK does not produce any immediate structural defects in the brain [12 30 35]. We therefore first identified the time point after birth from when CASK was not essential for survival. We deleted CASK at different time points in mice ranging from postnatal day 8 to postnatal day 47. Seven mice were younger than P30 (1 mouse each at P8, P21, P25 and P27 and 3 at P30) None of these mice survived beyond 10 days of the first dose of tamoxifen, indicating that much like constitutive knockout mice, deletion of Cask in young mice is rapidly lethal. The mice also developed tremors and ataxia before death (Suppl. Video 1). We then deleted Cask after 47 days (5 mice), approximately the median of adult myelination status in mice [36]. Cask deletion after 47 days, however, was compatible with survival. For 2–4 months after Cask deletion in >P47 mice, no obvious phenotype was observed. In fact, while Cask deletion before P30 was accompanied by rapid loss of weight, mice over P47 gained weight during the same 10 days period (Figure 1G). Beginning about 5 months after receiving tamoxifen, mice showed obvious signs of ataxia and incoordination (Suppl. Video 2). Rotarod treadmill balancing experiments indicate that these mice have impaired locomotor coordination (Figure 1H). Upon examination of brains from mice who received tamoxifen after P47, we observed that the cerebellum displayed striking atrophy, while the gross structure of the rest of the brain remained unremarkable, (Figure 1I).

Figure 1. Development of adult CASK knockout mice.

A) Cross between ER-Cre and lsl-tdTomato mouse lines results in offspring carrying both transgenes. Intraperitoneal injections of tamoxifen leads to translocation of Cre-recombinase to nucleus and excision of the stop cassette leading to expression of tdTomato. B) Cross between ER-Cre and CASK^floxed mouse lines results in offspring carrying both transgenes. Intraperitoneal injections of tamoxifen leads to translocation of Cre-recombinase to nucleus, excision of exon1, and consequent deletion of CASK. C) Brains from mice of indicated groups. Tamoxifen injection results in widespread expression of tdTomato. D) (Left panel) Representative confocal images from cortex and retina of ER-Cre; lsl-tdTomato mice. Green is DAPI and red is tdTomato; scale bar in brain and in retina is 50μm. (Right panel) Quantitation of cells with tdTomato in ER-Cre; lsl-tdTomato mice with (+TMX) or without (−TMX) Tamoxifen injection, and representative images of liver, kidney and intestine indicates widespread tdTomato expression following tamoxifen injection in ER-Cre-lsl-tdTomato mice. Data plotted as mean±SEM (n=3). E) Immunoblotting against indicated antigen in whole brain lysates of indicated mouse group. F) Western blot quantitation. Data plotted as mean±SEM (n=3), * indicates p < 0.05. G) Weight of mice of indicated ages 8 days after first tamoxifen injection. Data plotted as mean±SEM (n=3); KO= knockout. H) Time spent on an accelerating rotarod in seconds by genotype from left to right: Cask^floxed; ER-Cre mice with and without tamoxifen trafficking (n=3); * indicates p < 0.05 using a two-tailed Student’s t-test. Results are plotted as mean±SEM. I) Brains from mice of indicated group. Note the thin hypoplastic cerebellum (arrow).

Histological examination generally confirmed gross observations. Using DAPI and Nissl staining, we find that while the organization and cellularity of structures like the cortex and hippocampus remain unaltered between mice treated with tamoxifen or vehicle; however the cerebellum specifically undergoes severe degeneration in adult Cask knockout mice (Figure 2 A,B,C). Our data overall indicate that in the entire body, the cerebellum is most vulnerable to CASK loss throughout the lifespan. A closer look at the cerebellum demonstrates that granule cells specifically, but not Purkinje cells, degenerate in the absence of CASK. Our mice may thus be an ideal model for Norman type cerebellar atrophy. Quantification of the inner granular layer shows that mice receiving tamoxifen have less than 40% of the cells compared to mice of the same genotype receiving vehicle (Figure 2E). Finally, we also examined the optic nerve because it is known to undergo hypoplasia in CASK deficiency [14 15]. Surprisingly, no optic nerve atrophy was observed in mice where Cask is deleted after P47, even 6 months after administration when other phenotypes present. These results suggest that certain regions that are vulnerable to CASK loss early in development, such as the brain stem and optic nerve, become more resistant to CASK loss after adulthood. CASK does, however, play a vital and specific role in the maintenance of the cerebellum throughout the life; loss of CASK thus affects the cerebellum disproportionately. These data confirm and extend findings showing that death of granule cells after development underlies diminished cerebellar volume [9].

Figure 2. Specific degeneration of cerebellar granule cells after CASK deletion in adulthood.

Sections of cortex (scale bar=300μm), hippocampus (scale bar=300μm) and cerebellum (scale bar=500μm) stained with Neurotrace^™ in ER-Cre, Cask^floxed mice without tamoxifen injections (A) and seven months after tamoxifen injection (B). Layers of cortex; DG, dentate gyrus, CA3, cornu ammonis 3, and IGL, inner granular layer. Nissl granule staining in cerebellum of ER-Cre; Cask^floxed mice without tamoxifen injections (C) and seven months after tamoxifen injection (D), red asterisk indicates the inner granular layer which is depleted in D (scale bar=40μm). E) Quantitation of Nissl-stained cells in IGL of cerebellum of indicated mice group, asterisk is the IGL. Data plotted as mean±SEM (n=3), * indicates p < 0.05. F) Representative images of semi-thin cross-sections of optic nerves from ER-cre,lsl-tdTomato mice, seven months after either vehicle or tamoxifen injection (scale bar=150μm). G) Quantitation of optic nerve diameter. Data plotted as mean±SEM (n=3).

Deleterious mutations in exon 19 and 20 are associated with milder phenotypes and do not affect the hindbrain including cerebellum

In human subjects, missense mutations throughout the CASK protein can affect the volume of the hindbrain, including the cerebellum[15 24 25 37 38] (Figure 3A). Despite the crucial role for CASK in the cerebellum, it is surprising that there are mutations in the CASK gene that, in boys, produce microcephaly and global developmental delay but no PCH (Figure 3A). These mutations are nearly aphenotypic in females. Notable are two brothers (~8 and ~10 years old) who display global developmental delay but show no evidence of PCH on MRI (Figure 3B)[39]. One of the brothers also displayed unilateral optic nerve hypoplasia. Sequencing indicates that both brothers have a nonsense mutation in the CASK gene (p.L604X). The mother is heterozygous for the same mutation and has no CASK-related manifestations. She displayed a skewing of X-chromosome inactivation (80:20). The mutation (leucine 604) falls between the PDZ and SH3 domain of CASK and is encoded by exon 20.

Figure 3. Damaging mutations in exons 19 and 20 do not affect cerebellum.

A) Domain structure of CASK showing the location of the variants with arrows. Black arrowheads represents examples of missense mutations throughout the protein producing diminished volume of cerebellum. From right to left they are: R106P[37], L209P[38], L354P[37], M519T[24], G659D[15] and V894A[25]. CaMK (calcium/calmodulin-dependent kinase domain), L27 (lin2/7 domain), PDZ (PSD95, dlg/ZO1), SH3 (Src homology 3) and GuK (guanylate kinase). B) Exons, nature of mutation, sex and clinical phenotype. C) Sagittal magnetic resonance image scans of the subjects with the described variants. Arrow indicates cerebellar hypoplasia.

Here we also report a 16-year-old boy with ADHD and mild cognitive deficits. He has graduated from high school and has both receptive and verbal language. No seizures have been noted. He is able to run and care for himself. MRI does not indicate presence of PCH (Figure 3C). Exome sequencing identified a mutation in the CASK gene (c.1783delA. P. S595AfsX23). This frameshift mutation in exon 19 would lead to a CASK protein truncated at the PDZ domain. A sister, although asymptomatic, is also a carrier of the same mutation, indicating that the mutation may have been inherited from the mother who has not been sequenced and is aphenotypic.

We also present here a typical case of a 2-year-old girl with global developmental delay, failure to thrive, and microcephaly. She had an MRI at 3 months and again at 12 months which demonstrated microcephaly, diminished brainstem volume, a dysmorphic and partially absent cerebellum (Figure 3 D), and a posterior fossa cyst. She had an electroencephalograph at 6 months of age which was reportedly normal. Sequencing results indicated a heterozygous de novo mutation in exon 21 (p. Arg639Ter) of the CASK gene.

Overall, the evidence presented here indicates that the mutation in exon 21, which allows formation of a larger protein, is associated with a typical MICPCH case even in the presence of a normal allele of the CASK gene, whereas mutations in exon 19 and 20 are nearly inconsequential in the heterozygous condition and present with much milder phenotypes even in the hemizygous condition in boys. Most dramatically, the mutations in exon 19 do not affect the volume of the cerebellum and pons at all, while mutations in exon 20 may affect the cerebellar volume to some extent without affecting the pons despite presenting with microcephaly [40].

Exons 19 and 20 encode an interdomain loop that is evolutionarily recent and subject to alternate splicing

The CASK protein, encoded by the CASK gene, is a multidomain MAGUK protein which includes a CaM-kinase (calcium/calmodulin dependent protein kinase) domain at its N-terminus and the canonical triad of PDZ (PSD95, dlg, ZO1), SH3 (src homology 3) and GuK (guanylate kinase) domains at its C-terminus. Compared to other MAGUKs, the invertebrate CASK ortholog has a much shorter loop between the PDZ and the SH3 domain. In vertebrates, this region has undergone expansion and is encoded by exon 19 and 20. This region of the protein is thus present only in vertebrates and is missing in CASK orthologs of invertebrates. (Figures 4A). Within vertebrates, the loop shows a high degree of conservation, which may have functional implications. A recent study classified and quantified splicing of CASK in fetal human brain tissue and found extensive splicing in exon 19 and 20 [41]. We also analyzed mRNA obtained from forebrain and hindbrain of mice using RT-PCR with primers targeted across these regions. Multiple bands were observed in all the mice, indicating that these exons are likely to undergo splicing in all vertebrates. We sequenced the bands and found that each of the possible permutations is present both in the hindbrain and forebrain (Figures 4B, C). Such splicing events indicate that transcripts without the damaging mutations in these exons likely exist in cells. The fact that functional protein is likely expressed, even in the presence of deleterious mutations in exons 19 and 20, may explain the milder phenotype seen with these mutations; in other words, although mutations in exons 19 and 20 produce either dysfunctional or nonfunctional protein, it is likely that alternate splice variants unimpacted by mutations in these exons compensate functional loss. Although ratios of transcript types present in the forebrain and hindbrain vary in a statistically significant manner (Figure 4D), these differences do not seem sufficient to explain the complete lack of hindbrain hypoplasia when mutations occur in exons 19 and 20.

Figure 4. Exons 19 and 20 are evolutionarily novel and undergo alternative splicing.

A) Alignment of schematic exon structure of CASK gene in various animal model species compared to human. Exons 19 and 20 (red) are only found in vertebrates. B) Schematic of polymerase chain reaction (PCR) in the area spanning exon 19 and 20 and permutations of possible transcripts are color-coded. C) PCR results from the forebrain and hindbrain of 3 independent C57Bl6 mice. Arrows in gel are color-coded as in B. D) Quantification of band intensity from PCR; bar colors are consistent with B and C. Data plotted as mean±SEM (n=3), * indicates p < 0.05.

Structural implications of alternative splicing in CASK’s C-terminus

To gain insight into the role that the evolutionarily recent exons 19 and 20 play in CASK’s structure, we performed molecular dynamics on three structural variants of CASK (with both exons, with only exon 19, and with only exon 20). Molecular dynamics simulations and accompanying analysis of protein dynamics during each trajectory can help predict protein structural flexibility. B-factors, a parameter that provides an estimate of localized changes in flexibility [42], were derived from root mean square fluctuation (RMSF) calculations performed on the alpha-carbons of each splice variant and revealed that the splice variant with exon 19 exhibited significantly more flexibility in a region of the CASK protein referred to as the hinge region between the SH3 and GuK domains [24] than in the variant with exon 20 (Fig 5A; tan loop). In addition to local changes in flexibility, differences in the radius of gyration of each structural variant provide some insight into the likelihood of a structure being more extended and “open” (Fig. 5B) or more compact and “closed” (Fig. 5C). The CASK structural variant that only contains exon 20 (does not contain the longer exon 19) has an average radius of gyration that is notably smaller (Fig. 5D), with a tighter distribution, than either variant that contains exon 19. Under the conditions of our molecular dynamics simulation, when exon 19 is not present, CASK exists as a much more compact structure, with limited sampling of an extended conformation. This has implications for the likelihood that this particular splice variant (without exon 19) adopts a conformation that is considered to be necessary for interacting with other binding partners.

Figure 5. Inclusion of exons 19 or 20 affects the mobility of the PSG supradomain.

A) Molecular dynamics simulations predict increased mobility in a section of CASK’s hinge region (residues 700–720) with inclusion of exon 19 when compared to a model with exon 20 as indicated by B-factors (Å2) calculated from RMS fluctuations of α-carbons in a homology model of CASK’s PDZ-SH3–GuK supradomain during 100 ns of molecular dynamics simulations at each residue. B) Average structure of CASK containing exon 19 (yellow) predicted from molecular dynamics trajectories. Arrow highlights more open conformation. PDZ domain, purple; SH3 domain, cyan; hinge/hook region, tan; GUK domain, gray. C) Average structure of CASK containing exon 20 (red) predicted from molecular dynamics trajectories. Domain colors as in B. D) Radius of gyration (nm) for three structures of CASK containing both exons 19 and 20 (blue), only exon 19 (yellow) and only exon 20.

Discussion

The CASK gene evolved before emergence of nervous systems in Trichoplax adhaerens [43 44]. We have previously reported that although the CASK protein is highly conserved throughout the animal kingdom, the loops between the domains display a higher degree of evolutionary divergence [43]. Whether these differences support divergence of function remains uninvestigated. CASK is composed of two distinct components, a CaM-kinase (calcium/calmodulin dependent protein kinase) domain which transfers phosphate in the absence of divalent ions and a C-terminus encoding a supra-domain (PSG supradomain) that includes a PDZ domain, an SH3 domain and a GuK domain [18 45]. The PSG domain has evolved in toto and acts as a single biochemical unit [45 46]. Multiple studies have shown that for interactions with membrane proteins such as neurexin, the entire PSG supradomain, not simply the PDZ motif, is involved [17 45]. Furthermore, the PSG supradomain may adopt a more closed or open conformation around a hinge region which may regulate homodimerization of CASK molecules[45 47]. Thus, structural studies of CASK point to a multifaceted function of the CASK protein based on the structural conformation of the PSG supradomain.

Clinical and experimental observations in the mammalian model indicate that the hindbrain, including the cerebellum, is disproportionately sensitive to loss of CASK function. Lethality in the absence of CASK function in humans as well as mice arises from dysfunction of a thinned brain stem leading to respiratory failure and aspirational pneumonia [9 30]. Previously we had published that CASK-linked disorder is degenerative in nature, and the non-progressive nature of MICPCH in girls is simply due to the mosaic nature of CASK expression with heterozygous mutation [9]. Recently, however, regression has been noted in girls beginning in adolescence ([48] and personal communications), indicating the possibility of ongoing and latent degeneration even in girls.

Our results here show that the brainstem and the optic nerve become more resistant to loss of CASK with age and either do not degenerate or may take a much longer time to degenerate clinically than is observable in a murine model, allowing adult knockout mice to survive without developing optic atrophy. The cerebellum, however, remains highly sensitive throughout the lifespan. It is therefore of extreme scientific interest that damaging hemizygous mutations in exon 19 and 20 do not affect the hindbrain, allowing survival of boys with no cerebellar hypoplasia. Interestingly, under heterozygous conditions, these mutations are without notable phenotype. Exons 19 and 20 are evolutionarily novel, vertebrate-specific exons and code for a loop (hinge region) between the PDZ and the SH3 domain of CASK. Due to the presence of these exons, this loop is longer in vertebrate CASK. In a recent study using fetal human brain, it was noted that these exons are subject to splicing and are likely to affect some well-known protein-protein interactions [41]. Our study here indicates that these exons are extensively spliced in the mouse brain. Damaging mutations like nonsense or frame-shift mutations in these exons are unlikely to affect the functions of proteins that are derived from transcripts that lack these exons. Therefore, although the mutations are damaging in nature, the resultant phenotypes are recognizably milder. The major question that needs answering is why these mutations produce global developmental delay and microcephaly while sparing the hindbrain, including the cerebellum. This is especially surprising given that the cerebellar granule cells remain the most sensitive cells to CASK deficiency throughout life. Neither the brain region-specific splicing in humans[41] nor the differences in splicing seen in forebrain and hindbrain of mice, can adequately explain this phenomenon. Strikingly the specific protein interactions that may be affected by these splicing events display significant allostery [41], so we explored the possibility that the inclusion of these exons alters the structure of the protein. We conducted in silico molecular dynamics experiments on a previously described model of the PSG supradomain [24] with inclusion of either exon 19 or 20. We found that presence or absence of these two exons affects the predicted structural flexibility of the PSG domain (Fig 5D), particularly when the larger exon (exon 19) is absent. As described above, these exons contribute to a large, flexible loop that is evolutionarily recent. Modeling demonstrates that this loop confers substantial flexibility on the entire supradomain, allowing this region of CASK to exist in both a compact, closed structure (Fig. 5C) and a more extended structure (Fig 5B) that likely allows CASK (and other MAGUK proteins such as PSD-95) to homodimerize and heterodimerize[15 24 45 47 49] [50]. The ability of CASK to explore a more extended structure when exon 19 is present thus implies that functions (protein interactions) that rely on this extended structure are more evolutionarily recent. Our in-silico analysis therefore also points to different functions that could be attributable specifically to exon 19 and 20. The cases presented here shows that mutation in exon 19 produces a much milder phenotype with lesser locomotor problems than that seen with mutations in exon 20. In fact, a boy with a mutation towards the end of exon 20 (R613X) has been reported to have a smaller cerebellum but a spared brainstem [40]. Thus, based on the loss of specific transcripts, different parts of the brain are variably affected.

The results we present here, in which mutations in these evolutionarily recent exons are surprisingly mild, confirm the hypothesis that CASK has a suite of functions that have evolved along with its structure. Presumably the more evolutionarily ancient functions are preserved in individuals with mutations in either of these exons because splice variants without these exons are still expressed and functions performed by the compact form of CASK are not disrupted. The preservation of the hindbrain and absence of cerebellar hypoplasia seen with mutations in exons 19 and 20 suggest that CASK’s function in cells such as cerebellar granule cells derives from the more compact (and evolutionarily older) form and does not involve interactions with proteins via an extended PSG supradomain. MAGUK genes have shown capacity for functional diversification using multiple mechanisms. CASK orthologs in invertebrates use a second promoter to code for a protein that is distinct from the canonical CASK protein [51]. Here we find that in vertebrates, alternative splicing may diversify the molecular function of CASK. Most multi-exon genes undergo alternative splicing events which play a crucial function in diversifying the cellular proteome. This splicing is regulated both by cis elements (e.g., the specific sequence) and trans elements (e.g., enhancers and repressors of splicing) [52]. Multiple other factors, ranging from the rate of transcription to epigenetics to very late phase elements of the spliceosome may all regulate alternative splicing [52]. In general, current evidence points to a highly dynamic mechanism by which alternative splicing is regulated. Structural and hence likely functional differences in CASK spliced isoforms suggests that its alternative splicing may require functional feedback based on cellular metabolic states. Future studies on the regulation of alternative splicing of the CASK gene will be crucial to understand this process. It is pertinent to mention here that we previously found CASK protein itself capable of interacting with multiple components of the spliceosome [13]. The idea that the CASK protein has an autoregulatory role in its splicing would be an interesting avenue to pursue. Specifically, such a mechanism could potentially alter splicing directly in the presence of exon 19 and 20 mutations. It will also be important to determine if CASK splicing differs in different organs of the body. Overall, the evidence presented here and in the literature suggests that loss of the CASK gene disproportionately affects the cerebellum, but loss of only certain transcripts leaves the cerebellum virtually intact. This provides genetic evidence that CASK has at least two functions regulated by alternative splicing: a primitive function crucial for the hindbrain that may not include exon 19 and 20; and more vertebrate lineage-specific function/s having an augmented role in the forebrain. Our study thus presents a model where, in PCH, cerebral changes may result from a different molecular mechanism even with a monogenic etiology.

Key message.

What is already known on this topic – summarise the state of scientific knowledge on this subject before you did your study and why this study needed to be done

CASK deficiency induces cerebellar hypoplasia along with a smaller forebrain. Absence of CASK in the cerebellum leads to granule cell loss, the role of CASK in forebrain growth remains unclear.

What this study adds – summarise what we now know as a result of this study that we did not know before

The present study extricates the role of CASK in cerebellar maintenance from its role in forebrain. Based on multiple lines of evidence we suggest that CASK is likely to have an alternative splicing-dependent evolutionarily novel role in forebrain growth which is distinct from its role in cerebellum.

How this study might affect research, practice or policy – summarise the implications of this study

Not only does this study highlight how alternative splicing may diversify the molecular function of CASK and promote research in that direction, but it also expands our understanding on how different CASK variants could produce distinct phenotypes.