For bacterial expression, a plasmid containing the human CSNK2A1 gene in an Invitrogen Gateway donor vector (pDONR223) was provided by The Broad Institute and was transferred to the pDEST17 backbone following the standard Gateway Protocol (Life Technologies). Generation of the CK2^K198R mutant was done using the Q5^® Site-Directed Mutagenesis Kit (New England BioLabs).

The Proteomic Peptide Library (ProPeL) approach is a mass spectrometry-based method to experimentally determine high-resolution kinase specificity motifs. In this method, an exogenous kinase of interest is expressed in Escherichia coli and phosphorylates the endogenous proteome according to its native substrate preferences. These phosphorylated proteins can then be identified by LC-MS/MS to create a list of phosphorylation sites to be used to visualize substrate specificity. ProPeL experiments were carried out as previously described (Chou et al., 2012; Lubner et al., 2018) with the following conditions for in vivo proteome phosphorylation: all CK2 constructs were expressed in E. coli OverExpress C43 (DE3) cells (Lucigen) by IPTG induction. Optimal expression conditions were determined to be mid-log induction followed by expression for 24 h at 37°C in TB media (data not shown).

Comparable expression of CK2^WT and CK2^K198R was confirmed by western blotting using the primary antibody Anti-CSNK2A1 antibody (Abcam ab10466) at a 1/5,000 dilution and the secondary antibody IRDye^® 680RD Donkey anti-Rabbit IgG at a 1/5,000 dilution.

Peptide samples were subjected to mass analysis using a Thermo Scientific Ultimate 3000 RSLCnano ultra-high performance liquid chromatography (UPLC) system coupled to a high-resolution Thermo Scientific Q Exactive HF mass spectrometer. An aliquot of each peptide preparation in Solvent A (0.1% formic acid in H₂O) was injected onto a Waters nanoEase M/Z Peptide BEH C18 analytical column (130Å, 1.7 μm, 75 μm × 250 mm) and separated by reversed-phase UPLC using a gradient of 4–30% Solvent B (0.1% formic acid in acetonitrile) over a 100- min gradient at 300 nL/min flow. Peptides were eluted directly into the Q Exactive HF using positive mode nanoflow electrospray ionization and 1.5 kV capillary voltage. MS scan acquisition parameters included 60,000 resolution, 1e6 AGC target, maximum inject time of 60 ms, and a 300–1800 m/z mass range. Data-dependent MS/MS scan acquisition parameters included 15,000 resolution, 1e5 AGC target, maximum ion time of 40 ms, loop count of 15, isolation window of 2.0 m/z, dynamic exclusion window of 30 s, normalized collision energy of 27, and charge exclusion “on” for all unassigned, +1, and >+8 charged species.

Peptides were identified using MaxQuant (v1.6.10.43) and the embedded Andromeda search engine (Cox and Mann, 2008). The raw data was searched against three databases: an in-house-generated protein database consisting of 6xHis-tagged CK2 wildtype and mutant sequences, the complete UniProt E. coli reference proteome (identifier UP0000068040, accessed 11 September 2020), and the MaxQuant contaminants database. Variable modifications were oxidation of Met, acetylation of protein N- termini, deamidation of Asn/Gln, and for enriched samples, phosphorylation of Ser/Thr/Tyr. Carbamidomethylation of Cys was set as a fixed modification. Protease specificity was set to trypsin, allowing a maximum of two missed cleavages. LFQ quantification was enabled. All results were filtered to a 1% false discovery rate at the peptide spectrum match and protein levels; all other parameters were kept at default values. MaxQuant-derived output was further analyzed in accordance with the ProPeL method (Chou et al., 2012).

Given that the ProPeL methodology for kinase motif determination is carried out in the context of a closed system (the E. coli expressed proteome) the scoring of phosphorylated and unphosphorylated serine, threonine, and tyrosine residues in E. coli could be used to convert motif scores to probabilities of modification. Specifically, phosphorylated 15mers (for either the wild type or K198R data set) were mapped back onto the UniProt reference K12 E. coli proteome to generate a list of expressed proteins phosphorylated by CK2 in E. coli. These proteins were then subjected to a full trypsin digestion in silico and resultant peptides with a length up to 50 residues were retained to generate a list of all potential phosphorylation sites that could have been hypothetically observed in our experiments. Seven residues upstream and downstream from each serine, threonine, and tyrosine residue in these peptides was extracted to create a complete list of serine, threonine, and tyrosine-centered 15mers. Each 15mer was labeled as either positive or negative depending on whether it was observed to be phosphorylated in our ProPeL experiments, resulting in 1,350 positive (phosphorylated) and 26,331 negative (not observed to be phosphorylated) 15mers in the CK2^WT data set, and 1,024 positive and 20,083 negative 15mers in the CK2^K198R data set. Every peptide was scored using the appropriate wild type or K198R serine, threonine, or tyrosine-centered position weight matrix derived from the log-odds binomial probabilities observed in the corresponding serine, threonine, or tyrosine-centered pLogo, as we have shown previously (Schwartz et al., 2009). Note, the value for the central phosphorylated residue was derived from pLogo containing phosphorylated serine, threonine, and tyrosine-centered 15mers relative to presumed unphosphorylated serine, threonine, and tyrosine-centered 15mers (i.e., the pLogo with the unfixed central residue, see Figure 1). The complete set of scored 15mers was sorted in descending order and was used as a reference table to determine probabilities of modification based on score - specifically by taking the number of positive peptides above a score threshold divided by the total number of positive and negative peptides above a score threshold. Complete reference tables for CK2^WT and CK2^K198R are available in Supplementary Tables S5, S6. This methodology was utilized to score and assign probabilities of modification to all serine, threonine, and tyrosine residues in human axonally localized ion channels (see Table 1, Supplementary Table S7).

To obtain a higher resolution motif than we previously published in 2012 (Chou et al., 2012) for CK2^WT, several new ProPeL experiments were carried out utilizing more sensitive instrumentation that allowed confident identification of low abundance phosphorylated peptides that escaped identification previously. Experiments with CK2^WT identified 3,894 tryptic phosphopeptides (Supplementary Table S1) with 2,845 phosphopeptides classified as high confidence (i.e., containing at least one phosphorylation site with greater than 0.9 probability of localization determined by MaxQuant/Andromeda). Parallel ProPeL experiments with CK2^K198R identified 3,322 (Supplementary Table S2) tryptic phosphopeptides with 2,439 phosphopeptides classified as high confidence. Each high confidence phosphorylation site was mapped back to the UniProt reference K12 E. coli proteome and extended to create lists of unique 15mers as previously described (Lubner et al., 2018). Finally, 312 phosphorylation sites detected by our group and others as endogenous to E. coli (Macek et al., 2008; Chou et al., 2012; Soares et al., 2013; Lubner et al., 2016) were removed from each list prior to motif analysis. This process yielded a data set comprised of 1,350 unique phosphorylation sites (818 pS, 422 pT, 110 pY) for CK2^WT (Supplementary Table S3) and 1,024 unique phosphorylation sites (619 pS, 251 pT, 154 pY) for CK2^K198R (Supplementary Table S4). Taken together these results indicate that CK2^K198R is an active kinase able to phosphorylate substrates under in vivo conditions. Furthermore, given that the CK2^WT and CK2^K198R ProPeL experiments were carried out under identical conditions, and western blotting revealed comparable levels of protein expression (Supplementary Figure S1), the number of non-unique phosphorylation sites identified could be used as a proxy for kinase activity. We identified 5,362 non-unique phosphorylation sites attributed to CK2^WT and 4,146 non-unique phosphorylation sites attributed to CK2^K198R. When considering the total amount of material analyzed by LC-MS/MS for each kinase (1570 ng for CK2^WT, 1430 ng for CK2^K198R) the overall kinase activity of CK2^K198R can be approximated at 85% of CK2^WT, thus suggesting that the OCNDS CK2^K198R mutation results in only a slight decrease in kinase activity relative to wild type CK2. This observed activity of CK2^K198R stands in contrast to previous results indicating that the mutation results in a loss of activity (Dominguez et al., 2021).

The generation of pLogos (O’Shea et al., 2013) provides a visual representation of the relative statistical significance of amino acids (represented by single letter abbreviations) at positions within a 15-residue window of a central phosphoacceptor residue. Amino acid preferences at each position are stacked and sized according to significance relative to the background data set. The most significant residues are displayed with the largest size and are positioned closest to the x-axis. Background data sets representing unphosphorylated residues were created independently for CK2^WT and CK2^K198R. Each background was generated through the in silico tryptic digestion of proteins identified by the presence of at least one phosphorylated peptide in either the CK2^WT or CK2^K198R experiments. Background lists included serine, threonine, and tyrosine sites in proteins known to be expressed that could potentially have been observed via MS/MS but were never detected in a phosphorylated state. These sites were subsequently extended seven amino acids upstream and downstream of each phosphorylatable residue to generate two unique background lists for use in pLogo generation.

The overall pLogo generated with the current CK2^WT data set (Figure 1A) recapitulated our previously published CK2^WT pLogo (Chou et al., 2012); namely a strong preference for acidic residues (D/E) at the +1 and +3 positions with the +1 position being the most statistically significant position in the pLogo overall. Consistent with our prior hypothesis that the K198R mutation would impact the specificity of the +1 position, the pLogo generated for CK2^K198R indicated a massively decreased preference for acidic (D/E) residues in the +1 position, as well as an increased preference for both glycine and alanine, with glycine superseding aspartate to become the most statistically significant residue at the +1 position (Figure 1B). Interestingly, the reduction in significance was limited to the +1 position, as the preference for acidic residues at the +3 position was maintained at nearly identical levels (compare Figures 1A,B).

Further comparison of the CK2^WT and CK2^K198R derived pLogos indicated an unexpected reordering of phosphoacceptor preference. Though both mutant and wild type CK2 strongly favored serine phosphorylation over threonine or tyrosine phosphorylation, CK2^WT strongly disfavored tyrosine as a phosphoacceptor and was relatively neutral towards threonine as a phosphoacceptor (Figure 1A), while CK2^K198R strongly disfavored threonine as a phosphoacceptor and was relatively neutral towards tyrosine as a phosphoacceptor. It is important to note that although CK2^K198R did not display a preference for tyrosine phosphorylation (i.e., it was not phosphorylated by CK2^K198R at a level greater than its relative frequency among phosphorylatable residues), its shift is significant as it represents a near doubling of the proportion of phosphorylated tyrosine residues in the K198R data set (15%) relative to wild type (8.1%).

We next investigated whether the decreased preference for acidic residues at the +1 position observed in the overall CK2^K198R pLogo was consistent among each individual phosphoacceptor. To investigate this question, we fixed each central residue (serine, threonine, and tyrosine) to generate phosphoacceptor specific pLogos as shown in Figure 2. As was observed in the overall pLogo, the serine and threonine centered CK2^WT pLogos indicated a preference for acidic residues at the +1 and +3 positions (Figures 2A,B). The preference for acidic residues was only significant at the +1 position of the tyrosine centered pLogo (Figure 2C). In the CK2^K198R serine, threonine, and tyrosine centered pLogos (Figures 2D–F), the decreased preference for acidic residues at the +1 position was preserved. Interestingly however, the tyrosine centered pLogo suggested a modest increased preference for acidic residues at the –1 and –2 positions (Figure 2F). Importantly, the pLogos shown in Figures 2C,F represent the first-ever documentation of a tyrosine motif specificity for CK2.

The ProPeL methodology, carried out in living E. coli, provided a unique opportunity to explore the gain/loss of CK2^WT and CK2^K198R substrates at the site level. Specifically, we were interested in investigating whether our observed decreased preference for acidic residues at the +1 position in CK2^K198R resulted from the loss of existing E. coli substrates, the addition of completely new substrates, or a combination of the two. Figure 3A shows a Venn diagram representing phosphorylation site overlap between CK2^WT and CK2^K198R indicating the latter; namely, approximately half of the CK2^WT sites were detected in the CK2^K198R mutant, yet over a third of the sites detected in the K198R mutant were not observed in the wild type kinase data set.

To further explore motif differences between the three subsets noted in the Venn diagram, pLogos were generated for each subset (Figures 3B–D). The pLogo generated from the set of phosphorylation sites unique to CK2^WT included a highly pronounced +1 aspartate specificity that became more attenuated in the overlapping data set and was entirely replaced by glycine (and to a lesser extent, alanine and leucine) in the sites unique to CK2^K198R. The +3 acidic residues were similarly attenuated in the mutant kinase data set. Finally, with regard to the phosphoacceptor residue, the phosphorylation sites unique to CK2^WT showed a highly significant disfavoring of tyrosine residues relative to serine residues that was lost in the CK2^K198R mutant (which displayed a general neutrality to the phosphoacceptor residue).

Taken together these results point to a significant shift at the substrate level for the K198R mutant kinase, whereby the most canonical wild type CK2 phosphorylation sites are either lost or phosphorylated with decreased efficiency, coupled with a large increase in previously unphosphorylated sites.

CK2^K198R is the most frequently occurring mutation in patients with OCNDS (Chiu et al., 2018; Owen et al., 2018). Substrate rewiring in this mutant, indicated by our specificity and E. coli substrate analyses, is likely to contribute to the etiology of OCNDS. In an effort to offer a potential mechanism for the role of CK2^K198R in OCNDS pathology we developed a predictive strategy for determining CK2^WT and CK2^K198R phosphorylation sites on human proteins based on our prior scan-x post-translational modification predictor (Schwartz et al., 2009), but with the advantage of calculated probabilities of modification (see Methods). Due to the neurodevelopmental nature of OCNDS, in the present study we focused our predictions of potential CK2^WT and CK2^K198R phosphorylation sites to ion channels localized to the axon. The top ten predicted sites (from a list of over 3,000 total sites) with the greatest change in probability between CK2^WT and CK2^K198R are listed in Table 1 (a complete list of predicted sites is available in Supplementary Table S7). Of the sites listed, two have an increased probability of modification by CK2^K198R and the remaining eight are less likely to be phosphorylated by CK2^K198R. Additionally, four of the ten sites included in Table 1 have been previously shown to be phosphorylated (with two additional sites being paralogs of phosphorylation sites). We have created a website Located at https://okurchung.com that provides users with an interactive view of our predictive results.