Protein kinase CK2 phosphorylates a conserved motif in the Notch effector E(spl)-Mγ

Across metazoan animals, the effects of Notch signaling are mediated via the Enhancer of Split (E(spl)/HES) basic Helix–Loop–Helix–Orange (bHLH-O) repressors. Although these repressors are generally conserved, their sequence diversity is, in large part, restricted to the C-terminal domain (CtD), which separates the Orange (O) domain from the penultimate WRPW tetrapeptide motif that binds the obligate co-repressor Groucho. While the kinases CK2 and MAPK target the CtD and regulate Drosophila E(spl)-M8 and mammalian HES6, the generality of this regulation to other E(spl)/HES repressors has remained unknown. To determine the broader impact of phosphorylation on this large family of repressors, we conducted bioinformatics, evolutionary, and biochemical analyses. Our studies identify E(spl)-Mγ as a new target of native CK2 purified from Drosophila embryos, reveal that phosphorylation is specific to CK2 and independent of the regulatory CK2-β subunit, and identify that the site of phosphorylation is juxtaposed to the WRPW motif, a feature unique to and conserved in the Mγ homologues over 50 × 106 years of Drosophila evolution. Thus, a preponderance of E(spl) homologues (four out of seven total) in Drosophila are targets for CK2, and the distinct positioning of the CK2 and MAPK sites raises the prospect that phosphorylation underlies functional diversity of bHLH-O proteins.


Introduction
The bHLH proteins constitute a large group of transcription factors that regulate diverse aspects of cellular and developmental biology [1]. Among these are the diverse bHLH-O repressors [2] that mediate Notch signaling. In addition to the canonical bHLH domains, these proteins contain a second HLH domain (called Orange), which is followed by a C-terminal domain (CtD) of variable length, and terminate with an invariant WRPW motif through which they recruit the essential co-repressor Groucho (Gro, [3]). Since the identification of the founding members, the Drosophila E(spl)-repressors, bHLH-O proteins have been found in all metazoans [4], in which they serve as terminal effectors of Notch signaling. In addition to neurogenesis, these proteins are vital throughout animal development and regulate vital processes such as embryogenesis, myogenesis, somitogenesis, etc. [5,6]. Despite their centrality to Notch biology and an extensive body of knowledge on the importance of post translational modifications (PTM) that impact Notch signaling (for review see [7]), our understanding of the regulation of E(spl)/HES proteins by PTMs such as phosphorylation remains incomplete.
Notch signaling is acknowledged to be a universal arbiter of cell fate choice [5], and is often affected in diseases such as cancers. Although Notch plays distinct tissue-specific roles, the underlying mechanism is largely similar. Signaling is triggered by the Delta/Serrate/Lag (DSL) family of ligands, involves cleavage and nuclear translocation of the Notch intracellular domain (NICD), and culminates with the expression of the E(spl)/HES repressors [8]. The accumulation of E(spl)/HES proteins in the signal receiving cell represses target genes, thereby altering cell fate. How such a deceptively simple signaling pathway acts across development remains to be fully understood. Complexity of this pathway, in part, reflects diversity of E(spl)/HES repressors, and studies in Drosophila have been at the forefront in 1 3 understanding Notch biology and have informed studies in mammalian model organisms.
The E(spl) proteins in Drosophila are encoded by the E(spl)-Complex [2], which harbors seven repressors (Mδ, Mγ, Mβ, M3, M5, M7, and M8), their co-repressor Gro, and four Bearded-family bHLH genes that inhibit Notch signaling. This genetically dense locus (~ 60 Kb) is unique in that the organization and orientation of all genes is conserved in Drosophila species spanning 50 × 10 6 years (MYR) of evolution. E(spl)-repressors are expressed in complex partly overlapping patterns that are tissue/compartment-specific [9], and consequently, the functions of individual members have been gleaned from overexpression-based phenotypes [2]. These studies led to the view that bHLH-O members serve redundant functions [6]. Arguments against this view are the evolutionary conservation and molecular synteny of the E(spl)C [10], tissue-specific expression patterns, homo/ hetero-dimerization specificities, and target gene selectivity [11]. Adding to this complexity is accumulating evidence that these proteins are regulated by phosphorylation [12].
We had previously reported that E(spl)-M8/M5/M7 are phosphorylated by protein kinase CK2 at a highly conserved site in their CtD [13]. This modification of M8 was essential for early steps of eye development, but had a more modest impact during genesis of the sensory bristles [14], two widely used readouts of Notch signaling during Drosophila neurogenesis. This regulation reflects conversion of M8 from a CtD-dependent 'cis'-inhibited state to an active repressor [15]. Activation by phosphorylation also requires MAPK [16], whose recognition site resides in a Ser-rich region of the CtD, which we have termed the P-domain [14,[16][17][18][19]. In addition, the phosphatase PP2A targets the P-domain of E(spl) proteins and opposes the activation of M8, suggesting that dynamic phosphorylation and dephosphorylation may fine tune the activities of E(spl) repressors during Notchmediated tissue patterning [18]. It is noteworthy that a similar P-domain in human HES6 is modified by CK2 + MAPK [20,21], suggesting that PTM by phosphorylation is an important and conserved, yet underappreciated, aspect of Notch signaling. Understanding the generality of PTM of E(spl)/HES repressors via kinases and/or phosphatases would, therefore, enable a better understanding of Notch biology.
The studies we report here uncover novel and direct interactions between CK2 and E(spl)-Mγ. Our studies reveal that Mγ is a bona fide target of this kinase, identify that the primary site for phosphorylation is uniquely positioned adjacent to the WRPW motif, and that this site is invariant in the Mγ protein from 12 Drosophila species. Our studies reveal that a preponderance of E(spl) proteins are regulated by PTM, and provide new insights into the conservation of the kinase sites. Analysis of kinase-site specificity suggests that members of this class of bHLH-O proteins are functionally distinct, and differential regulation by PTM may fine tune Notch signaling.

Identification of CK2 sites in E(spl)-Mγ
The 'sequence motif' search tool at prosite.expasy.org was used to analyze all seven E(spl) proteins for consensus (S-D/ E-X-D/E) sites for CK2. The primary sequences of the Drosophila orthologues were downloaded from 'flybase.org' and aligned using the online software 'MUSCLE' using default parameter settings.

Isolation of E(spl)-Mγ and generation of mutant versions of E(spl)-Mγ
The E(spl)-Mγ open reading frame was PCR amplified from wild-type D. melanogaster (w 1118 flies) using primers complementary to the 5ʹ and 3ʹ ends of the intron-less Mγ gene. The primers included EcoR1 and Xho1 restriction sites at the 5ʹ and 3ʹ ends, respectively. The PCR product was cloned into the vector pBluescript-II (Stratagene/Agilent) and sequenced to verify identity to the reported gene/genome sequence (Flybase ID, FBgn0002735). This entry clone was then used to PCR amplify the N-terminal and C-terminal domains of E(spl)-Mγ, i.e., Mγ-NTD and Mγ-CTD, respectively. Subsequently, site-directed mutagenesis was used to introduce Ala substitutions at predicted CK2 sites. All PCRbased constructs were verified by Sanger-sequencing.

3
Azide) using an Amicon Ultra-15 device (10,000 MW cutoff). The concentration and purity were determined by SDS-PAGE compared to protein standards (Fermentas).

E(spl)-Mγ harbors multiple consensus sites for CK2
We first aligned E(spl)-Mγ homologues from 12 species that represent ~ 50 × 10 6 years (MYR) of Drosophila evolution [29]. All E(spl)-Mγ homologues harbor three conserved consensus sites for CK2 (Fig. 1A), fully meeting the strict recognition requirements for this kinase [30]. Unique among these are Thr 45 and Ser 48 , which reside within the loop of the HLH domain (Fig. 1A). Modification of Ser 48 is predicted to elicit hierarchical phosphorylation of Thr 45 , because Asp/ Glu or pSer/pThr promotes recognition by CK2 with equal potency [31]. Interestingly, of the seven E(spl) bHLH-O proteins in Drosophila, E(spl)-Mγ is the sole member harboring potential site(s) for CK2 within the HLH domain. The third CK2 site (Ser 195 ) is juxtaposed to the WRPW motif ( Fig. 1A), a feature also unique to E(spl)-Mγ. Notably, numerous Asp/Glu, a hallmark of 'high likelihood' targets of CK2, flank Ser 195 . Although E(spl)-Mγ homologues from several Drosophila species harbor the insertion of one/two amino acids immediately N-terminal to Ser 195 , in no case is the DS 195 EDEE motif perturbed (Fig. 1A, numbering reflects the D. melanogaster sequence). We next determined if the insertion/deletion of two amino acids correlates to specific branches of the Drosophila family. Given the known evolutionary relationships between 12 species representing six major groups of fruit flies [29], it appears likely that E(spl)-Mγ from the Hawaiian, Virilis, Repleta, and Obscura groups (harboring an insertion of two residues) are the ancestral form (Fig. 1B). A one residue deletion occurred ~ 37 × 10 6 MYR ago with the emergence of the Willistoni group, and both residues were lost ~ 13 × 10 6 MYR ago with the emergence of the Melanogaster group. The strong conservation of the C-terminal CK2 site despite the insertion/deletion suggests that the one/two residue insertion (in the ancestral species) is unlikely to alter protein structure and/or function.

Phosphorylation of E(spl)-Mγ by CK2
We next determined if E(spl)-Mγ is phosphorylated by CK2. This kinase exits as an α2β2 holoenzyme in all eukaryotes [32], and phosphorylates Ser/Thr (and in restricted cases Tyr) in a messenger-independent manner. Although phosphorylation is catalyzed by the CK2α subunit, the regulatory CK2β subunit promotes or inhibits recognition in a target-specific manner [25,33]. We thus used native α2β2 holoenzyme (CK2-HoloE) purified from Drosophila embryos and monomeric (recombinant) CK2α purified from yeast rescued by a cDNA encoding Drosophila CK2α [26]. Both enzyme forms are pure based on Coomassie staining (Fig. 2, Gel). In the case of CK2-HoloE, the weak staining of CK2β relative to CK2α (Fig. 2) does not reflect sub-stoichiometric levels, as this enzyme preparation has a sedimentation coefficient of 6.4S (see Inset in Fig. 2), consistent with the α2β2 conformation, and reflects weak staining of CK2β from flies and mammals [27,34]. Both forms display kinetic properties and ATP/GTP-dependency identical to those reported by others and us [26,35]. Importantly, both enzyme preparations require ~ 100 mM NaCl for optimal activity and display substrate-specificity identical to that revealed by the Pinna and Krebs laboratories [30,36,37]. Together, these parameters and properties are highly CK2-specific.
The two isoforms of CK2 were used to phosphorylate E(spl)-Mγ. In these assays, 20 ng of CK2α or CK2-HoloE was used and thus proteins with the mobility of CK2-α/β are not seen in SDS-PAGE gels stained with Coomassie (Fig. 2). E(spl)-Mγ was tested as a GST fusion, while GST-alone and GST-E(spl)-M8 served as negative and positive controls, respectively. As shown in Fig. 2, E(spl)-Mγ was robustly phosphorylated by both CK2α and CK2-HoloE. Phosphorylation is specific to E(spl)-Mγ, as GST was not phosphorylated. The band intensities suggest that CK2-HoloE more efficiently phosphorylates E(spl)-Mγ than does CK2α. This difference likely reflects (~ fourfold) greater activity of the holoenzyme [26], rather than a requirement for CK2β. Phosphorylation of E(spl)-M8 (positive control) was observed with both isoforms, although CK2β appeared to modestly diminish phosphorylation, even though this modification has been evinced in in vivo readout assays for Notch signaling. Thus, E(spl)-Mγ is targeted by native Drosophila CK2 independent of CK2β.

Identify the CK2 site(s) in E(spl)-Mγ
We next sought to identify CK2 target sites. Because Mγ harbors three predicted CK2 sites, it was divided into two parts; Mγ-NTD contains the bHLH-O domains, whereas Mγ-CtD contains residues of the CtD including WRPW (Fig. 4A). These constructs separate the CK2 sites within the HLH domain (T 45 LES 48 EGE) from that (S 195 EDE) proximal to WRPW. Wild-type and Ala mutants of Mγ-NTD and Mγ-CTD were purified as GST fusions. For reasons that are unclear, Mγ-NTD was poorly expressed, unstable, and sensitive to degradation/proteolysis despite the use of protease deficient E. coli cells and the inclusion of protease inhibitor cocktails. As an alternative, full-length Mγ and its Ala derivative mutants were used for mapping sites of phosphorylation.
Wild-type Mγ and its Ala mutants were tested for phosphorylation by CK2-HoloE. Analysis with monomeric CK2α was deemed unnecessary given CK2β-independent targeting of Mγ (Fig. 2). In brief, 20 ng of CK2-HoloE was used to phosphorylate 0.5-1.0 μg of Mγ variants under conditions (10 min) where enzyme activity is linear. As shown in Fig. 4B, Mγ-CtD was efficiently phosphorylated by CK2, and this was abrogated by replacing Ser 195 with Ala (Fig. 4B). Under similar conditions CK2 did not phosphorylate GST-alone. We next tested Mγ-NTD and its Ala mutants. While our studies with Mγ-NTD did not reveal any discernible phosphorylation (data not shown), problems with yields and quality remained. We therefore generated Mγ-FL (fulllength) with an Ala substitution at Ser 195 . We reasoned that if this variant was refractory to phosphorylation, it would demonstrate that the two closely juxtaposed and proximal sites Thr 45 and Ser 48 in the T 45 LES 48 EGE motif (see schematic in Fig. 4A) are not targeted by CK2. Indeed, we find that unlike wild-type Mγ-FL, its Ser 195 A mutant was completely refractory to phosphorylation (Fig. 4B), establishing that the CK2 modifies Mγ solely at Ser 195 in its CTD. While the reasons why Mγ-NTD is refractory to CK2 are unclear, but we speculate that localized folding may render Thr 45 / Ser 48 (in the HLH domain) inaccessible to CK2 and hence refractory to phosphorylation.

The importance of PTM of E(spl)/HES proteins via phosphorylation
The phosphorylation of E(spl)-Mγ reveals that a preponderance (four out of seven) of E(spl)C repressors are targeted by CK2. The physiological consequences of phosphorylation are best understood for E(spl)-M8, whose regulation by phosphorylation has been studied during eye and bristle development [14,16]. In either context, the CK2 and MAPK sites (the P-domain) auto-inhibit the O-domain via protein-protein interactions [15]. Phosphorylation overrides auto-inhibition to uncover the O-domain, thereby permitting E(spl)-M8 repressor activity to manifest. Accordingly, removal of the CtD or just the P-domain elicits constitutive (phosphorylation-independent) repression by E(spl)-M8 --  [16]. More recent studies ( [12], and Majot and Bidwai, unpublished) are revealing that M8 is likely to be regulated through additional PTMs and phosphorylation-dependent degradation (see below). The P-domain of M8 harbors two additional Ser residues (Ser 154 and Ser 155 ), which meet the consensus for CK1, a kinase that phosphorylates Ser with pSer/Asp/Glu at the n +4 -to-n +9 positions [39,40]. CK2 targeting of M8 (at Ser 159 ) may thereby favor (hierarchical) CK1-dependent modification of Ser 154 and Ser 155 . This, in turn, generates the pSpSGYHpSDCD 'phospho-degron' (Fig. 4C), which meets the strict consensus for degradation via Slimb/βTrCP [12]. The M8-Slimb/βTrCP interaction has been reported in an in vitro (Drosophila S2) cell-based assay [41], and our studies (unpublished) reveal that mutations in CK1 or Slimb/βTrCP genes strongly diminish in vivo repressor activity of M8. Together, these results suggest that controlled activation of E(spl) proteins (by CK2 and MAPK) and degradation via CK1 and Slimb/βTrCP impact these terminal effectors of Notch signaling during neurodevelopment. A similar sequence is conserved in M5 and M7 Fig. 4 Identify CK2 sites in E(spl)-Mγ. A) Location of the CK2 sites relative to functional domains. Constructs include full-length Mγ, the region from the N-terminus-through-Orange (Mγ-NtD) and the C-terminal Domain (Mγ-CtD). B) The indicated proteins with an intact CK2 site or one mutated with an Ala substitution (Mγ-CtD-S195A or Mγ-FL-S195A) were phosphorylated using the α2β2 holoenzyme (CK2-HoloE). Studies with Mγ-NtD were not conducted due to issues of excessive instability and degradation.  (Fig. 4C). The spacing of the kinase motifs in the P-domain is likely to be germane to understanding the function and regulation of these bHLH repressors. HES6, the mammalian homologue of E(spl)-M8 also conserves similarly positioned CK2, MAPK, and CK1 sites in its P-domain (not shown). We interpret our findings on Mγ in light of those with M8.

CK2-HoloE
While the E(spl)/HES proteins are conserved across the bHLH-O region, sequence divergence is largely restricted to the CtD, which is of variable length and composition. Might CtD divergence underlie functional diversity among this family of repressors? Several lines of evidence support this possibility. (1) The CtDs of E(spl)-M8/M5/M7/Mγ are notable in that they are rich in Ser/Thr residues. Of these the CK2 site is present in all four members, whereas that for MAPK has been lost from only M7 (Fig. 4C). This pattern of phosphorylation-site maintenance/loss is invariant in homologues across 12 Drosophila species (gray box in Fig. 4C), a strong evolutionary argument that this correlation is not merely incidental. (2) It is increasingly being realized that phosphorylation often targets 'intrinsically disordered' (ID) regions, which are rich in Ser/Thr/Asp/Glu residues [42,43]. Accordingly, the CtDs of M8/M5/M7/Mγ (and HES6) are predicted to be disordered [12], but not for those members lacking phosphorylation sites such as E(spl)-Mδ [15]. It is of interest to note that targeted knockdown of CK2 (by RNAi) or overexpression of the phosphatase PP2A strongly antagonizes repression by M8 [18], but neither has an effect on Mδ, which is not phosphorylated by CK2 [17]. While the CK2 and/or MAPK sites of M8/M5/M7 are contained within the ID-region, Mγ is unique in that the CK2 and MAPK sites flank the ID-region [12]. The differential placement of phosphorylation sites relative to regions of ID raises the prospect that phosphorylation effects on Mγ are unlikely to mimic those with M5, M7, M8, and HES6. Indeed, in vivo studies during eye/bristle development reveal that Asp mutations at the CK2 and MAPK sites on M8 potently enhance repressor activity, whereas they have no discernible effects on M5 [16], Kim and Bidwai, unpublished). (3) Unlike the P-domain of M8, which directly mediates auto-inhibition, the CK2 and MAPK sites of Mγ are separated by 31 residues, making it likely that CtD-Orange interactions would be different from those in M8. Evidence in favor of this will require knowledge of residues of the O-domain that directly contact the CtD and participate in auto-inhibition. (4) As mentioned above, M8/M5/M7 conserve a CK2 + CK1dependent phospho-degron, pSpSGYHpS (Fig. 4C). It is difficult to reconcile a similar mechanism for Mγ-turnover, because its SSYAGS sequence does not meet the consensus for Slimb-binding (Fig. 4C), and the absence of Asp/Glu residues C-terminal to this sequence and displacement of the CK2 site preclude the formation of a 'phospho-degron' through hierarchical phosphorylation. Degradation may therefore be of lesser importance for Mγ. We do, however, note numerous conserved Ser/Thr in the linker separating the CK1 and CK2 sites (see Fig. 4C), raising the prospect that distinct kinases may impact Mγ. (5) The juxtaposition of the CK2 site to WRPW is unique to Mγ (Fig. 4C), and it is unknown if CK2 influences Groucho binding. If this were the case, CK2 would modulate formation of an Mγ-Groucho complex, thereby imposing regulation distinct from that for M8, whose interaction with Groucho is phosphorylationindependent [14]. (6) The ability of E(spl) proteins to repress other bHLH factors such as Atonal hinges upon many factors, one of which is the formation of homo/heterodimers [11]. The possibility that phosphorylation of Mγ may bias partner preference for dimerization remains unexplored. Systematic biochemical and genetic analyses are required to identify which E(spl) proteins hetero-dimerize with Mγ, and if phosphorylation influences their binding with bHLH factors other than Atonal.

An expanding role for PTM regulation of Notch signaling
Among the several highly conserved signaling pathways that regulate animal development, the Notch pathway is deceptively simple. Specifically, activation involves cleavage of the ligand-bound Notch, thereby releasing the NICD, which translocates to the nucleus and regulates gene expression in concert with effectors such as Su(H). It has been somewhat of a paradox how such a simple pathway can orchestrate diverse developmental programs. Since the discovery of the first Notch mutations, named to reflect a 'notched wing' phenotype in Drosophila, it has become clear that this pathway is vital for embryogenesis, neurogenesis, organogenesis, and the morphogenesis of adult structures. Furthermore, studies in mammalian model organisms have revealed additional roles in myogenesis, sprouting angiogenesis, kidney, liver, and heart development, stem cell maintenance, and deregulated signaling has been implicated in diverse cancers such as lung, breast, colorectal, prostate, and pancreatic cancers, as well as osteogenic sarcomas [44][45][46]. These clinical findings led to significant efforts to develop inhibitors of Notch signaling as a therapeutic [45,47], but many of these have not borne fruit, perhaps reflecting the broad pleiotropic roles of this signaling pathway and its regulation by an array of PTMs (reviewed in [7]), and its intersection (crosstalk) with other signaling pathways.
Emerging evidence has revealed that Notch signaling involves significant crosstalk with other developmentally important pathways such as EGFR and Wingless/WNT, which are themselves involved with diverse cancers. In addition, an array of PTMs lie at the heart of Notch signaling likely enabling diversity of signaling in both temporal and spatial contexts (reviewed in [7,48]). This diversity of PTMs targets not only the Notch receptor, but many of the 1 3 pathway components. These include PTMs affecting the receptor such as glycosylation, O-glucosylation, O-fucosylation, hydroxylation, acetylation, methylation, and phosphorylation and dephosphorylation, to name a few. Some of these PTMs regulate receptor trafficking, localization, maturation, turnover, and signal strength and/or duration. Another aspect is signal specificity, first identified during early Drosophila eye development, when Notch first signals in a Su(H)-and E(spl)/HES-independent manner, and shortly thereafter (≤ 30 min later) switches into a Su(H)-and E(spl)/HES-dependent mode [49][50][51]. It remains unknown how Notch switches between these two modalities, and the possibility remains that this involves spatial and/or temporal differences in PTMs outlined by Antfolk and coworkers [7]. Similarly, PTMs also target the Notch ligands, and transcriptional complexes (Su(H)/ CSL and MAML). It has accordingly been suggested that regulation of protein stability, protein-protein interactions, localization, and activity may confer fine-tuning of Notch, without which it may be unable to fulfill its diverse roles [7,46,48,52]. Given our findings on CK2 regulation of E(spl)/HES proteins, it will be of interest to determine if mutations in CK2 genes may be linked to aberrant Notch pathologies. For example, several mutations have now been identified in both human CSNK2A1 (encoding CK2α) and CSNK2B (encoding CK2β) genes, and these are involved with a constellation of neurodevelopmental disorders including delayed development, intellectual disability, etc., that are collectively called 'Okur-Chung's Neuropathy' [53][54][55][56][57]. It remains unknown if these pathologies involve aberrant Notch signaling dynamics or specificity.
In summary, our studies reveal that a preponderance of E(spl) proteins are targets of CK2 and contain highly conserved sites for several additional kinases such as MAPK and CK1, thereby implicating multi-site phosphorylation in regulating Notch signaling pathway output. These sites of phosphorylation co-localize with regions predicted to be intrinsically disordered, but are positioned differently in each isoform, raising the prospect that these bHLH repressors and their PTMs are unlikely to serve redundant functions. Although the roles of multi-site phosphorylation of M7, M5, Mγ, and HES6 remain unresolved, our studies on Mγ suggest that PTM by phosphorylation is more central to these Notch effector proteins than has been recognized. Future efforts are needed to clarify the roles of phosphorylation in dimerization of E(spl) proteins, their ability to bind and repress proneural bHLH factors, and determine if these modifications underlie tissue-specific roles of Notch signaling during development of the nervous system and elsewhere. Given that these bHLH-O effectors are expressed in a tissuespecific manner, a better and more detailed understanding of their regulation by PTMs may provide a unique opportunity to interrogate and interfere with the functions of specific members through kinase-specific inhibitors [58], many of which have had success in the clinic.