Introduction
Poirier–Bienvenu neurodevelopmental syndrome [Online Mendelian Inheritance in Man (OMIM) #618732, POBINDS] is a recently described rare autosomal dominant neurodevelopmental disorder characterized by early-onset seizures, developmental delay, hypotonia, and dysmorphism (Ernst et al., 2021). POBINDS is caused by a de novo variant in the CSNK2B gene (located at 6p21.33), which encodes the casein kinase 2 enzyme β subunit (CK2β). CK2 is a ubiquitous protein serine/threonine kinase, a heterotetrameric enzyme consisting of two catalytic CK2α or CK2α′ subunits and two regulatory β subunits (Niefind et al., 2001). CK2β is a highly conserved regulatory subunit, and excessive synthesis of CK2β will form dimers that disrupt the complete structure and function of CK2 (Graham and Litchfield, 2000; Niefind et al., 2001). In previous studies, CK2 was found to be widespread in the brain (Guerra et al., 1999), and conditional knockout of CSNK2B mice induced age-dependent reduction of grip strength and impaired neuromuscular transmission and synaptic gene expression (Eiber et al., 2019). Knockdown of CSNK2B in mouse embryonic neural stem cells (NSC) impaired cell differentiation and reduced dendritic length and branching (Yang et al., 2018). These studies suggest a pivotal role for CSNK2B in neurodevelopmental disorders. To date, more than 40 individuals with POBINDS have been reported in the literature (Li et al., 2019; Ernst et al., 2021; Yang et al., 2021), but the genotype–phenotype correlations of POBINDS are still unclear.
In this study, we describe the clinical and genetic characteristics of eight de novo individuals with POBINDS, most of whom suffered developmental delay, generalized epilepsy, and hypotonia. Two missense, two nonsense, one frameshift, and three intron variants were detected, with only c.58G>T being reported before (Ernst et al., 2021). Minigene experiments confirmed that two intron variants (c.367+5G>A and c.367+6T>C) resulted in skipping of exon 5, leading to premature termination of mRNA transcription. Combining our data with the available literature, missense, nonsense, frameshift, and splicing site variants were found to cause POBINDS. The variant types do not reflect the severity of POBINDS. Reduced CK2 holoenzyme activity may represent a unifying pathogenesis. Missense variants in the zinc finger domain result in manageable seizures (p = 0.009) and milder intellectual disability (ID) (p = 0.003) than missense in other domains of CSNK2B. This is the first study of the genotype–phenotype correlations of CSNK2B variants and expands the knowledge of POBINDS.
Discussion
CSNK2B was first reported by Poirier et al. (2017) in two individuals with early-onset epilepsy, developmental delay, hypotonia, and malformations, and was documented as the causative gene for POBINDS by OMIM in 2020.3 A total of 57 individuals with CSNK2B mutations have been reported to date (including eight cases in this article), and the largest sample study to date was published in Epilepsia by Ernst et al. (2021). The clinical phenotype of POBINDS is complex, with great inter-individual variability and uneven severity of phenotype. Although these studies have expanded the phenotype and genotype spectrum of POBINDS, the genotype–phenotype correlations in POBINDS are still unclear.
Two splicing variants (c.367+5G>A and c.367+6T>C) were located in non-canonical splicing sites. The minigene assays confirmed that these two atypical splice variants lead to skipping of exon 5, resulting in the insertion of a premature stop codon in p.Leu98Alafs*11. A total of 11 splice variants (including three in this article) have been reported in the current literature (11/56), of which six (c.175+2T>G, c.367+2T>C, c.367+5G>A, c.367+6T>C, c.292−2A>T, and c.558−3T>C) have been experimentally confirmed to lead to abnormal splicing of CSNK2B (Poirier et al., 2017; Yang et al., 2021). The pathogenicity of atypical splicing site variants is difficult to decipher, and our work confirmed that these variants indeed affect splicing, which is valuable for subsequent research.
Frameshift variants often bring in a premature termination codon, leading to nonsense-mediated decay (NMD) (Shoemaker and Green, 2012). However, among frameshift variants of CSNK2B, six out of nine [p.(Met132Leufs*110), p.(Asp155Alafs*71), p.(Leu167Serfs*60), p.(Pro179Tyrfs*49), p.(Asn181Thrfs*46), and p.(Lys208Glnfs*38)], induced a prolonged termination codon downstream of the wild-type termination codon, not upstream. In eukaryotes, mRNAs with these variants may not be directly identified by NMD mechanisms. NMD escape was found in other protein-coding extending like COQ8A, but enzyme activities were significantly reduced (Liu et al., 2014). The functional effects of these six frameshift variants in CSNK2B need further verification, but extensive amino acid sequence disruption cannot be ignored. Individuals with these six variants did not show any phenotypic pattern. Another 3 frameshift and 11 nonsense variants were predicted to introduce a stop codon, none of which were in the last exon or the last 50 base pairs of the penultimate exon. In this study, these variants were ranked into LOF variants. Individuals with LOF variants did not show any clear correlations in symptoms, except dysmorphisms (p = 0.041). There may be some objective reasons for this phenomenon, such as a small sample, varying phenotype details, and different indicators of developmental assessment in the age strata.
Interestingly, we found some duplicated mutations in the cohort, which may be hotspot mutations or regions. Ernst et al. (2021) and Yang et al. (2021) have reported three start losses with different variants (c.1A>G, c.2T>A, and c.3G>A). The effect of the start loss mutations is still unpredictable. Firstly, the phenotypes of these three individuals are polytropic, c.1A>G showed only mild speech and motor developmental delay. c.2T>A had mild ID, while c.3G>A had profound ID. c.2T>A and c.3G>A were both diagnosed with epilepsy, c.2T>A showed drug-refractory epilepsy, while c.3G>A showed seizure-free. Second, the start loss of CSNK2B was also recorded in the gnomAD database4 with a relative high population frequencies. Considering that the start loss variant may theoretically affect the open reading frame and leads to gene silencing (Chen et al., 2020), or re-initiate translation by another “ATG” triple nucleotide and mainly affects phenotypic penetrance (Benkirane et al., 2021), whether those variants are pathogenic requires further study. There were eight individuals with c.G94 mutation, and detailed phenotypes were available for four individuals. In these four individuals, some mutual characteristics were found, such as moderate to severe ID, and malformations. Three individuals had seizures and two were diagnosed with epilepsy, both of which manifested as absence epilepsy and drug-refractory. These phenotype profiles suggest that mutations at the c.G94 may be involved in signaling a variety of developmental processes, such as organogenesis (Bandyopadhyay et al., 2016). The untranslated region (UTR) of exon 5 (c.292-367) may be a hotspot region for CSNK2B mutations. A total of 11 individuals were reported in this region, 5 individuals with detailed phenotypes available. Four individuals were diagnosed with epilepsy, all with effective antiepileptic treatment. Three with seizures before 6 months of age. The degree of ID varied from mild to severe, except one individual, had normal development (c.367+5G>A), and three cases had malformations (c.292-1G>A, c.367+2T>C, and c.367+5G>A). Although all five splice site mutations have been experimentally confirmed to cause skipping of exon 5 (except for c.292-1G>A), the varied clinical phenotypes suggest that CSNK2B may influence POBINDS onset and progression through some unknown pathways.
Regarding the location of the mutation, the individuals with variants in the zinc finger domain (residues 105–140) had manageable seizures and milder ID, which was inextricably linked to the structure and function of CK2β. Zinc finger structures play a crucial role in the regulation of transcription, adaptation of protein interactions, and post-transcriptional regulation in eukaryotes (Laity et al., 2001; Ecco et al., 2017). Zinc finger mutations (cys109, 114) result in the inability of CK2β to form dimers and loss of the interaction with the catalytic subunit CK2α, which is accompanied by increased degradation of the mutant protein (Canton et al., 2001). The individual in our cohort (c.325T>C) was mildly affected and had manageable epilepsy, mild ID, and mainly language delay. This does not appear to be parallel to the severity of the mutation. However, the situation in vivo is complex, and developmental delay may be a direct cause of seizures or due to genetic pleiotropy. Increased degradation of CK2β degradation may trigger a compensatory mechanism that regulates CK2β protein levels. Our data reveal corresponding associations from both clinical and bioinformatics analysis, suggesting that the mutation in this region may be less harmful, and also worth further functional study.
In addition to the zinc finger region, CK2β includes the N-terminal (residues 1–54) containing the autophosphorylation site, an acidic groove region (residues 55–77), and a C-terminal (residues 171–215) that stabilizes the CK2β-CK2α interaction (Reed et al., 1994; Bidwai et al., 1995). We compared missense variants across domains, and the results were not convincing due to the small number of variants. The C-terminal-truncated CK2β loses its ability to bind stably to the catalytic subunit (Sarno et al., 2000). p.(Pro179Tyrfs*49)-mutated CK2β is unable to bind to the CK2α, although it is still expressed (Nakashima et al., 2019). The frameshift of CSNK2B (c.499del) resulted in decreased protein expression (Yang et al., 2021). These studies suggest that the CSNK2B mutation is likely to affect the function of CK2 in a haploinsufficient manner. However, we cannot exclude a dominant-negative effect of CK2β. It was previously shown that although CSNK2B heterozygous mice did not differ significantly from wild-type mice in CK2β expression levels, some of the heterozygous mice failed to survive (Blond et al., 2005). This may have been due to the mutated CK2β forming non-functional tetramers that inhibited the normal catalytic activity of the wild-type CK2α. Such phenomena are not uncommon in proteins that perform functions in multimeric forms (Ahn et al., 2003). The reduced CK2 holoenzyme activity may represent a uniform pathogenesis.
The severity of the neurological phenotype ultimately depends on the spatiotemporal distribution and catalytic activity of CK2. The CK2 holoenzyme is a heterotetramer composed of two catalytic subunits (α/α′) and two regulatory subunits (β), which can bind as α2β2, αα′β2, or α′2β2 forms. During early development (embryonic and fetal stages), the mRNA levels of CK2β were significantly higher than those in childhood and adulthood (Yang et al., 2018). CK2 activity was higher in cortical pyramidal neurons and large striatal neurons than in other neuronal populations (Castello et al., 2017). Interestingly, in different brain regions of adult rats, the expression of each subunit of CK2 is different at the transcriptional level and the protein level (Castello et al., 2017). This may form distinguishing CK2 heterotetramers in different brain regions, resulting in different enzymatic activity of CK2.
In addition, CK2 is involved in protein phosphorylation, and these affected proteins/signaling pathways may be the main cause of the progression of neurodevelopmental disorders. CK2 catalyzes the phosphorylation of Kv3.1 and calmodulin (Miller, 2000; Macica and Kaczmarek, 2001), the latter of which directly affects the stability of KCNQ2 channels (Benaim and Villalobo, 2002). CK2 is abundantly expressed in the postsynaptic density (Soto et al., 2004), and inhibition of CK2 blocks N-methyl-D-aspartate receptor-dependent synaptic transmission and long-term potentiation (Kimura and Matsuki, 2008). Meanwhile, knockdown of CSNK2B not only impaired cell differentiation of mouse embryonic NSC (Yang et al., 2018) but also disrupted the migration of mouse immortalized neuronal GN11 cells (Lettieri et al., 2019). The deletion of CSNK2B in embryonic NSC impaired the proliferation of forebrain NSCs and inhibited the differentiation of NSCs into oligodendrocyte precursor cells, resulting in defects in brain development (Huillard et al., 2010). Conditional knockout CSNK2B mice showed an age-dependent reduction in grip strength, impaired neuromuscular transmission, and inordinate synaptic gene expression (Eiber et al., 2019). Abnormal emotional behavior, including autistic behavior, irritability, depression, and hyperactivity, may be associated with abnormalities in the CK2-mediated dopamine signaling pathway (Rebholz et al., 2009; DiCarlo and Wallace, 2022). CK2 is also involved in signaling for a variety of developmental processes, such as embryogenesis and organogenesis (Bandyopadhyay et al., 2016). The above CK2 signal involvements underlie the phenotypes of neurodevelopmental disorders such as epilepsy, developmental delay, hypotonia, and dysmorphic disorders.
Limitations of this study include the heterogeneity of the clinical data. When analyzing the reported cases, we relied heavily on the authors’ descriptions of the individuals, and the data lacked a standardized assessment. Especially concerning the judgment of developmental delay, age stratification needs to be introduced to standardize the determination of the degree of developmental delay. The reported age of some patients is too young, and some late-onset neurodevelopmental symptoms need to be determined by long-term follow-up. Additionally, the statistical power of our Chi-square and Fisher’s exact test was generally below 0.8, which may omit some true positive results. This question requires more samples to solve. Moreover, the mechanism by which CSNK2B variants cause CK2 dysfunction remains unclear, especially the effect of missense variants on CK2, which requires more experimental models. With the increase of the functional study and refinement of phenotypic profiles, the clarity of genotype–phenotype correlations is expected to be further improved.
In conclusion, this study further enriched the phenotype–genotype profile of POBINDS. Minigene experiments confirmed that two intron variants resulted in skipping exon 5, leading to a premature termination of mRNA transcription. The pathogenicity of intronic variants requires more experiments to decipher. Taken together, our analysis results showed that missense, nonsense, frameshift, and splicing variants were common forms of CSNK2B variants, and individuals with variants in the zinc finger domain had manageable seizures and milder ID. The variant types do not reflect the severity of the POBINDS. Determining further genotype–phenotype correlations in POBINDS requires more functional research and refinement of phenotypic profiles.