Pathogenic missense variants of CSNK2B associated with Poirier-Bienvenu neurodevelopmental disorder impact differently on CK2 holoenzyme formation

2.1 Expression and subcellular localization

POBINDS is mainly caused by nonsense and frameshift variants, but a large number of missense variants spanning the entire protein sequence have also been associated with the disease. We selected four disease-associated variants based on the frequency of the affected amino acid in the literature and the ClinVar database but also based on their localization along the CK2β protein sequence (Figure 1A). The p.Asp32Asn variant is localized to the N-terminal domain (AA 1–54) containing the autophosphorylation site; the p.Arg86Cys mutant is C-terminal to the acidic loop and is one of the highly conserved residues exposed on the surface, facing away from the rest of the protein. p.Arg111Pro is within the Zinc-finger domain, and finally, p.Cys137Phe variant itself provides the sulfhydryl group for zinc coordination that stabilizes the β–β homodimer (Figure 1B and C). In all figures and legends, for simplicity reasons, the one-letter amino acid code is used (D32N, R86C, R111P, C137F).

We transfected the mutant CK2β proteins into Cos7 and HEK293T cells and observed that while WT (wild type) and p.Asp32Asn expressed at similar levels, the p.Arg111Pro and p.Cys137Phe variants are much less expressed, while p.Arg86Cys variant expression is slightly but not significantly reduced (Figure 2A–C). Furthermore, WT, p.Asp32Asn and p.Arg86Cys mutants were predominantly localized to the nucleus, while for p.Arg111Pro and p.Cys137Phe we observed a retention in the cytosol and a more punctate staining (Figure 2D and E). In cells expressing WT, p.Asp32Asn and p.Arg86Cys, the percentage of cells with predominantly nuclear HA signal was 78.5 %, 77.5 %, and 81.9 %, respectively, while for p.Arg111Pro and p.Cys137Phe it was 44.8 % and 24.4 %, respectively (Figure 2D and E). Consistently, for the p.Arg111Pro and p.Cys137Phe variants, we counted 44.2 % and 62.0 % of cells with equal HA signal in cytosol and nucleus, respectively, and also a small number of cells with predominantly cytosolic staining.

Figure 2:

Expression and subcellular localization of POBINDS linked mutant proteins. Immunoblotting of Cos7 and HEK293T lysates overexpressing WT and mutant HA-tagged CK2β mutants and β-actin (A) and quantification thereof (B) (Cos7), (C) (Hek293T). Cos7 cells, stained with anti-HA antibody and DAPI (D). White bar represents 30 and 10 μm in the zoomed image. Quantification of CK2α subcellular localization. For all conditions, more than 200 cells were counted, and three independent experiments performed (E).

Using the translational inhibitor cycloheximide (CHX), we directly addressed the question of reduced stability of the p.Arg111Pro and p.Cys137Phe variants. We analysed lysates after 3 and 6 h of CHX treatment by Western blotting, using anti-HA and anti-CK2β antibodies, in order to compare the stabilities of overexpressed and endogenous proteins (Figure 3A). All overexpressed CK2β proteins were more labile than endogenous CK2β (Figure 3A and B). For example, within 3 h WT HA-tagged CK2β was reduced to 50 % of the level at time = 0. The slopes of the degradation curves were significantly steeper for the p.Arg111Pro and p.Cys137Phe variants, especially after 3 h of incubation, and were not significantly reduced for the p.Arg86Cys variant (Figure 3B), reflecting the results of the steady-state readout without CHX treatment (Figure 2B and C).

Figure 3:

Stability of mutant proteins. Cycloheximide (100 μg/ml) was added to HEK293T cells expressing WT and mutant CK2β proteins in a chase experiment over 6 h. 40 µg of protein lysate was analysed by Western blotting using anti-HA antibody. Results from three independent experiments were quantified, with basal expression levels normalized to 1. Graphs show mean value ± SEM (B). Statistical analysis was performed using two-way ANOVA test with Dunnett post-test.

We wanted to determine whether the proteasomal activity was involved in the degradation of the p.Arg111Pro and p.Cys137Phe variants, since the role of the proteasome in the degradation of misfolded proteins in the unfolded protein response (UPR) pathway is well established (Hetz et al. 2020). To this end, the proteasome inhibitor MG132 was applied to HEK293T cells expressing HA-tagged WT and mutant CK2β. MG132 stabilized all over-expressed proteins while, again, similar to the CHX experiment, it had little effect on endogenous CK2β, as previously shown (Canton et al. 2001) (Figure 4A and B). The fold changes of HA expression, i.e. the ratios of HA-signal in the presence versus in the absence of MG132, clearly indicated that overexpressed proteins were susceptible to proteasomal degradation, with highest ratios of 4.1 and 4.6 observed for the p.Arg111Pro and p.Cys137Phe variants, although the results were not statistically different. Furthermore, the p.Asp32Asn variant showed a trend towards a reduced ratio, corresponding to a reduced susceptibility to proteasomal degradation (Figure 4A and B).

Figure 4:

Degradation of mutant proteins. MG132 (10 µM) was added to HEK293T cells expressing WT and mutant CK2β proteins overnight (16 h). 60 µg of protein lysate was analysed by Western blotting using anti-HA antibody (A). Results from three independent experiments were quantified, fold changes calculated as the ratio of HA signal with/without MG132 and plotted. Graphs show mean value ± SEM (B). Statistical analysis was performed using two-way ANOVA test with Dunnett post-test. Immunofluorescence imaging of Cos7 cells, expressing WT and mutant CK2β, stained with anti-HA, anti-Lamp1 antibodies and DAPI (C). White bar represents 30 μm.

Proteasomal degradation canonically involves fusion of the proteosome with lysosomes, but lysosomal degradation can occur independently of the proteasome (Kim et al. 2024). Thus, we wanted to probe co-localization of the unstable variants p.Arg111Pro and p.Cys137Phe with the lysosomal marker Lamp1, based on the punctate staining observed in the cytosol. We observed partial co-labelling; however, the overlap was not complete (Figure 4C). We therefore suggest that the degradation of the mutant proteins must involve other pathways.

2.2 CK2β homodimerization is affected in POBINDS-associated variants

The CK2 holoenzyme formation comprises a first step of CK2β-β dimerization, followed by CK2α recruitment (Chantalat 1999; Raaf et al. 2008). Based on the location of the CK2β p.Arg111 and p.Cys137 amino acids within the β–β interaction region, where four Zinc-finger-forming cysteines are either present in close proximity, or where p.Cys137 itself provides the sulfhydryl group for bridging, we hypothesized that β–β homodimerization might be impaired. We performed co-immunoprecipitation assays to address this question. We expressed GFP-WT CK2β and HA-tagged wildtype and CK2β mutants in HEK293T cells, we precipitated GFP-WT CK2β and blotted for the presence of HA-tagged versions of CK2β (Figure 5A). As expected, both the p.Cys137Phe and p.Arg111Pro variants co-precipitated less with WT-CK2β mutants than HA-tagged CK2β wildtype, whereas no difference was observed for the p.Asp32Asn and p.Arg86Cys variants (Figure 5A and B).

Figure 5:

CK2β homodimer formation of mutant proteins. GFP-tagged CK2β (WT) and the series of HA-tagged CK2β mutants were expressed in HEK293T cells for 24 h. After lysis, 400 µg of lysate were used for immunoprecipitation using anti-GFP magnetic beads and washed precipitates probed with anti-HA and anti-GFP antibodies (A). Quantification of three independent experiments (B). GFP-tagged CK2β (WT and mutants) and the series of HA-tagged CK2β WT and mutants were expressed in Hek293 cells for 24 h. After lysis, 400 µg were used for immunoprecipitation using anti-GFP magnetic beads and washed precipitates probed with anti-HA and anti-GFP antibodies (C). Quantification of three independent experiments (D). BRET saturation curves of homodimerization of CK2β-β in HEK293T cells transiently transfected with a constant DNA amount of CK2β-Nluc and increasing quantities of CK2β R111P-YFP mutant. The curves were fitted using a non-linear regression equation assuming a single binding site (GrapPadPrism). Results were expressed in milliBRET units (mBU, with 1 mBU corresponding to the BRET values multiplied by 1,000) (E), and for the combination of CK2β-Nluc CK2β R137F-YFP (G). BretMax values for the interaction CK2β WT with R111P (F) and C137F (H). Data were obtained with at least three separate experiments with triplicate determinations. Statistical analysis was performed using two-way ANOVA test with Dunnett post-test (B, D) or with Newman Keuls post-test (F, H).

In POBINDS, about 50 % of the CK2β expressed is expected to be mutant, so in cells two mutant CK2β molecules will also try to form dimers. We wondered whether the dimer formation of two mutant proteins would be even more impaired.

To this end, we expressed both GFP- and HA-tagged mutant pairs, precipitated the GFP-tagged CK2β and blotted for the presence of HA-CK2β (Figure 5C and D). As a result, dimerization of only the p.Arg111Pro and p.Cys137Phe proteins was again significantly impaired, although the reduction compared to WT-WT interaction was less than when testing WT pairing with a mutant molecule. However, these two experiments cannot be directly compared due to differences in the expression of the various constructs.

To validate our findings by another approach, bioluminescence resonance energy transfer (BRET) saturation assays were developed to monitor, quantify, and compare the formation of CK2β-CK2β dimers. BRET saturation experiments have previously been successfully conducted to investigate the relative affinities of various interacting protein pairs (El Khamlichi et al. 2019; Guillemain et al. 2020; Mercier et al. 2002; Wang et al. 2005) and to estimate the proportions of dimeric and monomeric fractions of the two BRET partners (Mercier et al. 2002; Couturier and Jockers 2003).

The BRET signal is expressed as a function of the fluorescence/luminescence measured, which corresponds to the expression of the acceptor (YFP) and the donor (NLuc). Two measures are extracted from saturation curves: BRET_max which represents the maximal BRET signal reached at donor saturation and informs about the relative amounts of dimers, and BRET₅₀, which corresponds to the [YFP]/[luc] ratio at 50 % of BRET_max and has been suggested as an indicator of the relative affinities between donor and acceptor fusion proteins.

We transiently transfected HEK293T with expression vectors for CK2β WT-NanoLuciferase and YFP-tagged CK2β p.Arg111Pro or p.Cys137Phe mutants, by applying escalating doses of YFP-CK2β plasmids and keeping the amount of the donor Luciferase plasmid constant. As a result, BRET assays confirmed the loss of β–β dimerization for both mutants. Compared to the CK2β/CK2β dimerization (BRET_max 221.6 mBu), BRET_max was reduced by factors of 7.4 to 32.4 mBu and by 11.9x to 20.5 mBu, for p.Arg111Pro and p.Cys137Phe, respectively (Figure 5E–H). The strongly reduced BRET is indicative of a lack of dimerization of mutants. We were unable to determine BRET₅₀ values for the mutants due to the very low BRET signal at all Fluo/Luc ratios.

We wanted to determine if CK2α recruitment into the holoenzyme was affected by POBINDS variants. To do this, we again used the BRET technology (CK2α-Luc, CK2β mutants-YFP). We found that none of the variants showed a difference compared to WT, in BRET_max and BRET₅₀ (Figure 6A–E). This was unexpected since it is known that CK2β-CK2β dimerization is a prerequisite for holoenzyme formation, and therefore we expected to see a clear reduction in CK2α recruitment to p.Arg111Pro and p.Cys137Phe mutants.

Figure 6:

CK2α-β formation is not affected by CK2β mutants. BRET saturation curves of heterodimerization of CK2α and CK2β in HEK293T cells. BRET signals were performed on HEK293T cells transiently transfected with a constant DNA amount of CK2α-Nluc and increasing quantities of CK2β-YFP. The curves were fitted using a non-linear regression equation assuming a single binding site (GraphPadPrism). Results were expressed in milliBRET units (mBU, with 1 mBU corresponding to the BRET values multiplied by 1,000). Comparison CK2α-CK2β WT with CK2α-CK2βD32N (A), with R86C (B), with R111P (C) and with C137F (D). Data were obtained with at least three separate experiments with triplicate determinations. Parameters derived from BRET saturation curves of the heterodimerization of CK2. Table listing Bmax and Bret50 values for individual combinations (E). BRET saturation curves comparing CK2α-β and CK2β-β interaction between wildtype CK2 subunits (F), table listing Bmax and Bret50 values (G).

Comparing the affinity of the CK2α-CK2β versus CK2β-CK2β interaction in our cellular BRET system, the saturation curves are qualitatively highly distinct: the CK2β/CK2β curve shows a clear response to expression levels with high maximal BRET (222 mBu) and a high BRET₅₀, whereas CK2α/CK2β curve has reduced BRET_max of 37 mBU, suggesting that there are more CK2β/CK2β dimers compared to CK2α/CK2β dimers (Figure 6F and G). The BRET₅₀ values were slightly lower for CK2α/CK2β (BRET₅₀ = 6.5) compared to CK2β/CK2β (BRET₅₀ = 16.6). The variation in BRET₅₀ values suggests that CK2β may in vivo have slightly higher affinity for binding to CK2α compared to CK2β.

2.3 CK2α activity when complexed with POBINDS-variant proteins

Finally, we wanted to determine the activity of CK2 holoenzyme, composed of at least one mutant CK2β molecule, to phosphorylate a substrate peptide in vitro. We expressed HA-tagged wildtype and CK2β mutants, precipitated them using anti-HA-agarose and performed an in vitro kinase assay on the washed immunocomplexes. For the p.Arg111Pro and p.Cys137Phe mutants, we observed a significant reduction in ADP generation, correlating with ATP consumption, a surrogate for CK2 activity in the ADP-Glo system (Figure 7A). This is not surprising given the low expression of these two mutants, as shown in Figure 1A. Quantification of the presence of CK2α in the precipitates, shows no significant changes in co-precipitated CK2α when normalized to CK2β (Figure 7B), mirroring the findings of the BRET assay in Figure 6. The normalization of ADP generation by the amount of co-precipitated CK2α, restored kinase activity for all mutants, indicating that the activity towards the specific substrate peptide is unaffected by the presence of one or more mutant CK2β molecules in the holoenzyme (Figure 7C). As control, CK2α is not co-precipitated in the absence of HA-CK2β, testifying to the specificity of the precipitation and indicative that the activity measured is due to CK2α-β interaction (Figure 7D).

Figure 7:

In vitro kinase activity of mutant immunoprecipitates. HA-CK2β WT and mutants were expressed in Cos7 cells. 400 μg of cell lysates were used for immunoprecipitation using anti-HA magnetic beads, followed by a kinase assay performed with washed immunoprecipitates. ADP generation was plotted (A). The immunoprecipitates were then separated on a SDS-PAGE gel and amounts of HA-CK2β and CK2α in precipitates were quantified (B). ADP generation was normalized by amount of co-precipitated CK2α (C). Results from three independent experiments were used for quantification. Representative western blot showing CK2α and HA-CK2β after α-HA immunoprecipitation (D). Graphs show mean value ± SEM. Statistical analysis was performed using one-way ANOVA test with Dunnett post-test.

3.1 Stability of variants

We observed a generally reduced stability of exogenously expressed compared to endogenuous CK2β, at steady state and when chased with translational inhibitor cycloheximide. One difference between the two types of protein that may explain this, is the HA-tag on the exogenous CK2β. Importantly, we detected increased instability specifically for p.Arg111Pro and p.Cys137Phe variants, a finding which is corroborated by previous results showing that a Zinc-finger-deficient mutant (Myc-CK2β Cys109Ser/Cys114Ser) is less stable than its WT counterpart (Canton et al. 2001). In this publication, the authors additionally observed reduced phosphorylation of this mutant, rendering it more labile. We further show that in our cell system, CK2β is majorly localized to the nucleus, confirming prior work (Lorenz et al. 1993), but the unstable mutants p.Arg111Pro and p.Cys137Phe are retained in the cytosol.

The mechanism of degradation involves the proteasome which has a stronger effect in the p.Arg111Pro and p.Cys137Phe variants, as well as lysosomal degradation, where the lysosomal marker Lamp1 partially co-localizes with these two mutants. However, we believe that other degradation mechanisms may also be involved. Interestingly, the most stable mutant is the p.Asp32Asn variant, which has been associated to a new intellectual disability-craniodigital syndrome (IDCS) distinct from POBINDS (Asif et al. 2022). In patient cells, the authors observed increased CK2β levels in the nucleus and cytoplasm, together with altered protein-protein interaction and dysregulation of canonical Wnt signalling, leading the authors to postulate a dominant negative effect (Asif et al. 2022). In our cellular system, the p.Asp32Asn mutant does not express to a higher extent than WT protein, but it appears to be the most stable compared to the other mutants tested. Another variant located adjacently in the KEN box, p.Leu39Arg was shown to be less stable, leading to reduced net CK2 activity (Di Stazio et al. 2023). These findings indicate that the precise location of a variant, even within a small motif, is clearly important for exerting biological effects.

3.2 CK2β dimerization and CK2α-β dimerization

Using co-immunoprecipitation and BRET assays, we found that the p.Arg111Pro and p.Cys137Phe variants have a significantly reduced dimerization capacity. This is not surprising, given that p.Cys137 is itself involved in Zn²⁺ binding, while the p.Arg111 is closely adjacent to p.Cys109, another of the four Zinc-finger competent cysteines in the β–β interaction domain. The β–β dimerization is mediated by many hydrophobic interactions between zinc ribbon motifs (composed of three-stranded β sheets, coordinating one Zn²⁺) for each of the two β monomers (Chantalat 1999; Filhol et al. 2005). It is therefore noteworthy to state that mutation of only one residue within this domain almost completely abolishes the dimerization. This finding indeed goes beyond Canton et al. (2001), who observed the impairment of CK2β homodimerization only for a double mutant myc-CK2β p.Cys109Ser/p.Cys114Ser (directly concerning two zinc-binding cysteine side chains) and not an interconnecting residue like Arg111.

The finding that none of the CK2β variants had a direct effect on the recruitment of CK2α into the holoenzyme, was initially unexpected since it has been established that β–β dimerization is needed to form a shared interface where two β subunits interact with two distinct CK2α subunits. This interface is provided by two different C-terminal segments (AA 164–170/interface region 1 and AA 186–198/interface region 2, as shown in Figure 1A) (Niefind et al. 2001).

We can speculate that endogenous CK2β may interact, albeit weakly, with the CK2β mutant, which could be sufficient to recruit CK2α and induce a BRET signal. However, the affinity of wildtype CK2β to itself outweighs binding affinity to zinc-binding motif mutant CK2β, and therefore such dimers are unlikely to occur.

Alternatively, we suggest that the interaction with CK2α may directly occur via monomeric CK2β, which would indicate that α-β dimers exist in our cellular system.

Besides the BRET results, we first observed a reduced co-immunoprecipitation of CK2α protein and activity with CK2β p.Arg111Pro and p.Cys137Phe variants, but this difference was rescued upon normalization, suggesting that at the molecular level, mutant CK2β p.Arg111Pro and p.Cys137Phe are able to interact with CK2α in our cellular overexpression system.

Previous studies had shown that the C-terminal region of CK2β (AA 171–215) stabilizes the CK2α-CK2β interaction (Bidwai et al. 1995; Reed et al. 1994; Sarno et al. 1999). In particular, the region encompassing AA186-193 in interface region 2, bound to CK2α, and a peptide containing this segment was efficient to antagonize holo-complex formation (Laudet et al. 2007; Raaf et al. 2009). Our variants are N-terminal to the crucial interface region 1 and 2, yet the lack of dimerization suggests these regions will fold differently. Structural analysis of CK2β monomer has, to our knowledge, only been performed on C-terminally truncated CK2β (Chantalat 1999).

The question of whether monomeric full-length CK2β can interact with CK2α was addressed previously: In vitro, in the presence of reducing sulfhydryl-modifying reagents, α-β complexes were detected, which, however, did not display activity (Meggio et al. 2000). The difference in catalytic activity could be explained by the in vitro system and the presence of reducing agent, which is not the case in our precipitation experiments.

In contrast, in a Cos7 cell overexpression system, a doubly mutated p.Cys109Ser/Cys114Ser CK2β which does not dimerize, did not co-precipitate with overexpressed HA-CK2α′ (Canton et al. 2001). Further, pulldowns between GST-CK2 α′ and this in vitro transcribed/translated CK2β mutant failed to detect an interaction. In these experiments, the precipitates were not normalized to input, and the input was not shown. Thus, the detection limit may not have been reached due to the low expression/high instability of the CK2β p.Cys109Ser/p.Cys114Ser mutant. Further, as binding partner, CK2α′ was used, which is known to have a one magnitude fold lower affinity for CK2β, compared to CK2α (Bischoff et al. 2011). However, in a yeast-two hybrid system, affinity of the same p.Cys109Ser/p.Cys114Ser non-dimerizing human CK2β mutant to human CK2α was demonstrated (Canton et al. 2001). Thus, we hypothesize that besides instability of the mutant protein, the difference in expression systems (yeast and in vitro translation) could underlie the contradictory results: the yeast proteome is highly similar compared to the mammalian proteome and may alter protein folding through chaperones and post-translational modifications. From this perspective, our results align with the results in yeast.

Regarding the BRET curves of CK2β/CK2β and CK2α/CKβ, the highest BRET signal for CK2β/CK2β (BRET_max = 222 mBu) suggests that there are more CK2β/CK2β dimers compared to CK2α/CKβ (BRET_max = 37 mBu). But these differences could also reflect variations in the orientation of the YFP/RLuc moieties within the proteins. BRET₅₀ values were slightly lower for CK2α/CKβ (BRET₅₀ = 6.5) compared to CK2β/CK2β (BRET₅₀ = 16.6), suggesting that CK2β may have a slightly higher affinity for binding to CK2α compared to CK2β. This contrasts with the knowledge about CK2 subunit interactions showing that CK2β has a stronger affinity to itself, compared to CK2α, which already presents very high affinity to CK2β, K_d being in the nanomolar range (Bischoff et al. 2011). However, affinity experiments have thus far been performed in in vitro systems. Here, we evaluated the physical association of CK2 subunits in intact living cells which differ from in vitro methods that rely on purified proteins. In order to define the affinities in a more quantitative manner, one would have to perform future BRET experiments that allow better curve fitting at even very low fluorescence/luminescence ratios. Finally, the differences observed between BRET and in vitro experiments could be due to the tags, Nanoluciferase and YFP, which may alter the affinity between subunits. Thus, this question needs to be further studied in the future.

3.3 Activity

We did not detect any changes in CK2 activity as a consequence of the variants studied. We used a well-characterized acidic substrate peptide in the in vitro kinase assay. However, one peptide may not be sufficient to detect subtle changes in activity. For example, disease-associated variants of the CK2α subunit, which cause a neurodevelopmental disorder known as OCNDS, have highlighted the importance of expanding the peptide array to capture changes in substrate binding (Werner et al. 2022).

In addition, CK2β is able to recruit substrates (Appel et al. 1995; Bojanowski et al. 1993; Leroy et al. 1999) and bind, as a homodimer, to other proteins and kinases independently of CK2α (Chen et al. 1997; Guerra et al. 2003; Hagemann et al. 1997). Therefore, it is of great importance to study different missense variants in protein-protein interaction assays, such as proximity-labelling, Bio-ID and others, which could also provide insights into the effect of POBINDS missense variants on the interactome and phosphorylation fingerprint of holoCK2 in POBINDS.

One consideration is that our experiments were performed in immortalized cells that already produce a steady state amount of CK2β. It is possible that overexpression per se disrupts the balance between dimerization and holoenzyme formation. CK2β is a highly conserved regulatory subunit and excessive synthesis of CK2β will form dimers that disrupt the overall structure and function of CK2 (Graham and Litchfield 2000; Niefind et al. 2001). However, when comparing the amount of overexpressed HA-CK2β to endogenous CK2β in our experimental system, the exogenous protein levels are always lower and therefore overexpression is not expected to significantly alter homeostatic processes.

We have shown that the precipitation of CK2α is specific and dependent on the presence of HA-CK2β, therefore the ATP hydrolysis activity does not stem from monomeric CK2α, devoid of CK2β. For mutant proteins p.Arg111Pro and p.Cys137Phe this finding argues again that CK2α-β heterodimers can form and be catalytically active. One factor to be taken into consideration is potential co-precipitation of CK2α′, however, we fail to detect this isoform with the pan-CK2α specific antibody, suggesting that if it were precipitated, it would be below our detection limit.

What are the implications of this finding for cells from POBINDS patients carrying these missense variants? If the Zinc-finger mutant proteins are less stable, then even with residual α-β binding, one would expect a reduced net activity of CK2 holoenzyme per amount of tissue. This would suggest that missense variants of the Zinc-finger domain may, at least in part, have a similar effect on CK2 activity as nonsense or splice variants, namely a reduction in the pool of functional holoCK2. Thus, these mutants could be considered to act in a haplo-insufficient manner. Further work is needed to confirm this hypothesis.

Our findings indicate that disruption of β–β dimerization does not necessarily prevent CK2α binding or catalytic activity. While it has previously been shown that preformation of the homodimer is necessary for holoenzyme formation, in the case of CK2β p.Arg111Pro and p.Cys137Phe variants, we still observed α–β interaction by Co-IP and BRET assays.

Based on our results, we further hypothesize that the p.Asp32Asn and p.Arg86Cys, both exposed on the surface and suggested to be important for ligand/substrate binding (Chantalat 1999) may act in a qualitatively different manner, since these variants are as stable as the WT protein, do not affect β–β dimerization, holoenzyme formation or activity towards a peptide, but may alter the profile of protein-protein interaction and in substrate binding.

Our work highlights the need to individually investigate missense variants in order to allow for genotype-phenotype correlations to better understand and treat this neurodevelopmental disorder.

4.1 Plasmids and site-directed mutagenesis

Rat CK2β coding sequence (BC078807) was cloned into PIRESneo3 vector in frame with a N-terminal HA-tag (Rebholz et al. 2009). For BRET assays, coding sequences for rat CK2α (BC091130) was cloned into pEYFP-N1, with EYFP at the C-terminus, and CK2β into pRLuc-N1 and pEYFP-N1 vectors, with RLuc or EYFP at the N-terminus.

Site directed mutagenesis was performed by the Biochemistry and Biophysics (B&B) core facility of the Institute of Psychiatry and Neuroscience of Paris (IPNP). Mutagenesis targeting CK2β was carried out using the Q5^® Site-Directed Mutagenesis Kit (NEB, France) in conjunction with NEBaseChanger V1 software (NEB, France). To construct the pEYFP-Ck2β WT and mutant plasmids, the Ck2β genes and the pEYFP plasmid were amplified with Q5^® polymerase (NEB, France). The resulting amplicons were assembled into the linearized plasmid using the NEBuilder^® HiFi DNA Assembly Kit (NEB, France).

4.2 Cell culture and protein extraction

Human Embryonic Kidney (HEK293T) cells were obtained from Evelyne Bloch-Gallego, Institut Cochin, Paris, France. Cos-7 cells (ATCC #CRL-1651) were obtained from Thierry Galli, IPNP, Paris, France. Cells were grown in DMEM 10 % FBS and transfected with Lipofectamine 2000 (Invitrogen, 52887) according to manufacturer’s instructions. 24 h post transfection, cells were washed with phosphate buffered saline (PBS 0.1 M) and lysis buffer (50 mM Tris–HCl, pH 7.5, 150 mM NaCl, 1 % NP-40, 5 mM EDTA, 1 mM EGTA) containing protease and phosphatase inhibitor cocktails (Roche, 66373700 and 04906837001) was added in each plate/well. Cells were scraped, pooled, and collected. Lysates were centrifuged at 13,000 rpm for 20 min at 4 °C and the supernatant was collected for western blot or immunoprecipitation experiments.

For CHX experiments, cycloheximide (100 μg/ml) was added for 3 or 6 h, cells were immediately lysed thereafter. Untreated cells were lysed at the start of drug incubation.

For MG132 experiments, MG312 (10 µM) was added to cells for 16 h before lysis.

4.3 Western blot analysis

Proteins were separated on a 4–12 % Bis–Tris protein gels by SDS-PAGE and blotted onto either nitrocellulose or PVDF membranes. Non-specific binding sites were blocked either with 5 % non-fat dry milk in PBS/0.1 % Tween20 or with 3 % BSA in PBS/0.1 % Tween20. Membranes were incubated with primary antibodies for 2 h or overnight. The following antibodies were used: anti-HA (Cell Signalling, 3724S), anti-GFP (Abcam, ab1218), anti-β-actin (Santa Cruz, sc-81178), anti-CK2α (Cell Signalling, 2656S) and anti-CK2β (Bethyl, A301-983A). For BRET test, anti-GFP antibody (Cell Signalling, 2956) was used to recognize YFP-tagged proteins. Bands were visualized either by incubation in ECL (Amersham) for chemiluminescence detection by ChemiDoc system (BioRad) or with the Odyssey CLx imaging system (LI-COR) for fluorescence detection. Relative protein amounts were quantified with ImageJ software.

4.4 Immunocytofluorescence

Cells were fixed in 4 % paraformaldehyde for 10 min, rinsed with PBS and then permeabilized with 0.5 % TritonX-100 for 5 min. Cells were blocked with 3 % BSA in PBS for 1 h at room temperature, then incubated for 2 h at room temperature or overnight at 4 °C with the primary antibody diluted in 3 % BSA/PBS. The following antibodies were used: anti-HA (Cell Signalling, 3724S) and anti-Lamp1 (DB Biosciences, 555798). After three washes with PBS, cells were incubated with anti-mouse Alexa Fluor 568, anti-rabbit Alexa Fluor 568 and anti-mouse Alexa Fluor 488 antibody (Invitrogen), diluted in 3 % BSA/PBS during 2 h at room temperature in the dark. After three washes with PBS, cells were mounted using DAPI mounting medium (Prolong, Invitrogen). Immunofluorescent staining was imaged with a Zeiss Axioplan fluorescence microscope and analysed with ImageJ software.

4.5 Immunoprecipitation and immunoblot analysis

HEK293T and Cos7 cells were lysed in lysis buffer (50 mM Tris-HCl pH 7.6, 150 mM NaCl, 5 mM EDTA, 1 mM EGTA, 1 % Triton X-100, protease and phosphatase inhibitor cocktails (Roche, 66373700 and 04906837001). Whole cell extracts were cleared by centrifugation. Protein content was determined using the bicinchoninic acid (BCA) protein assay kit (Thermo Fisher Scientific, 23225), with a bovine serum albumin (BSA) standard series. 400 µg of lysate were used for precipitation: GFP-CK2β was immunoprecipitated using anti-GFP-magnetic antibody for 16 h (ABclonal, AE079). Immune complexes were washed on a magnetic rack and subjected to analysis by SDS/PAGE and immunoblotting. The antigen-antibody complexes were incubated with anti-HA (Cell Signalling, 3724S), anti-GFP (Abcam, ab1218) or pan-CK2α antibody (Cell Signaling #2656), followed by incubation with goat anti-Rabbit IgG (licorbio IRDye 680RD, 926-68071) or goat anti-Mouse IgG (Licorbio IRDye 800CW, 926-32210). Fluorescence was detected by the Odyssey infrared imaging system (LI-COR), and quantified using ImageJ software.

4.6 In vitro kinase assay (IVK)

Wildtype and mutant proteins were expressed in Cos7 cells and lysed in lysis buffer, LB (50 mM Tris-HCl pH 7.6, 150 mM NaCl, 5 mM EDTA, 1 mM EGTA, 1 % Triton X-100, protease and phosphatase inhibitor cocktails (Roche, 66373700 and 04906837001). Immunoprecipitation using anti-HA antibody (0.5 μg per assay point, Cell Signalling, 3724S) was performed at 4 °C for 3 h, followed by two washes in LB and two washes in kinase 1x buffer (MOPS pH 7.4 (50 mM), β-glycerophosphate (10 mM), EGTA (2 mM) and MgCl2 (12 mM)). The immunoprecipitates were divided into two equal parts. Incubation took place at 30 °C for 20 min in kinase buffer containing ATP (150 μM), and either no peptide (control) or CK2 peptide RRRADDSDDDDD (150 uM, Abcam, 204855). Then the ADP-Glo™ (Promega) and detection reagents were added, according to manufacturer’s instructions (Promega ADP glow V9101) and luminescence was measured using the SpectraMax plate reader (Molecular Devices). The value obtained from the control (no peptide) was considered background phosphorylation (autophosphorylation plus eventual phosphorylation of protein(s) co-precipitated with CK2) and was subtracted from the value obtained in the sample incubated with peptide. Precipitates were then analysed by western blotting as described above.

4.7 BRET

HEK293T cells were transfected with the calcium phosphate precipitation method. 48 h after transfection, cells were washed twice with ice-cold PBS and resuspended in HBSS buffer. Intact cells were distributed in 96-well microplates (Optiplate, Perkin Elmer). Furimazine substrate (Interchim, B2W800) was added at a final concentration of 5 μM, and readings were performed with a Mithras LB 943 Multireader (Berthold), which allows the sequential integration of luminescence signals detected with two filter settings (NLuc filter, 480 ± 10 nm; YFP filter, 540 ± 20 nm). Emission signals at 540 nm were divided by emission signals at 480 nm. The BRET ratio was defined as the difference between the emission ratio obtained with co-transfected NLuc and YFP fusion proteins and that obtained with the NLuc fusion protein alone. The results were expressed in milliBRET units (mBU, with 1 mBU corresponding to the BRET ratio values multiplied by 1,000). BRETmax is the maximal BRET signal obtained in milliBRET units and BRET50 represents the ratio of acceptor and donor receptors (acceptor/donor) yielding 50 % of maximum BRET signal. All BRET, luminescence, and fluorescence measurements were performed at 21 °C using a Mithras LB 943 microplate analyser (Berthold). All data analyses were carried out with the GraphPad Prism 9.5.1 software for Windows (GraphPad Software Inc, San Diego, CA, USA). Concentration response curves were fitted by nonlinear regression and saturation curves by a hyperbolic one-binding site equation. The method provided estimates for BRETmax and BRET50 values and corresponding SEM. Each experiment was repeated at least three times.

4.8 Statistics

Data are presented as mean ± standard error of mean (SEM). Statistical differences were assessed using a parametric unpaired Student’s t-tests or one-way or two-away ANOVA with multiple comparisons in Dunnett post-tests or one-way ANOVA followed by the Newman Keuls multiple comparisons test for BRET analysis. A p-value below 0.05 was considered to be statistically significant, with *p < 0.05, **p < 0.01 and ***p < 0.001.