Drosophila melanogaster casein kinase II interacts with and phosphorylates the basic helix-loop-helix proteins m5, m7, and m8 derived from the Enhancer of split complex.

Drosophila melanogaster casein kinase II (DmCKII) is composed of catalytic (alpha) and regulatory (beta) subunits associated as an alpha2beta2 heterotetramer. Using the two-hybrid system, we have screened a D. melanogaster embryo cDNA library for proteins that interact with DmCKIIalpha. One of the cDNAs isolated in this screen encodes m7, a basic helix-loop-helix (bHLH)-type transcription factor encoded by the Enhancer of split complex (E(spl)C), which regulates neurogenesis. m7 interacts with DmCKIIalpha but not with DmCKIIbeta, suggesting that this interaction is specific for the catalytic subunit of DmCKII. In addition to m7, we demonstrate that DmCKIIalpha also interacts with two other E(spl)C-derived bHLH proteins, m5 and m8, but not with other members, such as m3 and mC. Consistent with the specificity observed for the interaction of DmCKIIalpha with these bHLH proteins, sequence alignment suggests that only m5, m7, and m8 contain a consensus site for phosphorylation by CKII within a subdomain unique to these three proteins. Accordingly, these three proteins are phosphorylated by DmCKIIalpha, as well as by the alpha2beta2 holoenzyme purified from Drosophila embryos. In line with the prediction of a single consensus site for CKII, replacement of Ser(159) of m8 with either Ala or Asp abolishes phosphorylation, identifying this residue as the site of phosphorylation. We also demonstrate that m8 forms a direct physical complex with purified DmCKII, corroborating the observed two-hybrid interaction between these proteins. Finally, substitution of Ser(159) of m8 with Ala attenuates interaction with DmCKIIalpha, whereas substitution with Asp abolishes the interaction. These studies constitute the first demonstration that DmCKII interacts with and phosphorylates m5, m7, and m8 and suggest a biochemical and/or structural basis for the functional equivalency of these bHLH proteins that is observed in the context of neurogenesis.

Casein kinase II (CKII) 1 is a ubiquitous protein kinase that is highly conserved among eukaryotes (1,2) and is capable of functioning as an oncogene in mammals (3). CKII is composed of catalytic (␣) and regulatory (␤) subunits that combine to form an ␣ 2 ␤ 2 holoenzyme. With the exceptions of Drosophila melanogaster (4), Caenorhabditis elegans (5), and Schizosaccharomyces pombe (6), CKII from most eukaryotic organisms contains two ␣ subunits, ␣ and ␣Ј, that are encoded by distinct genes. In contrast, ␤ subunit heterogeneity has been documented via protein microchemical approaches in Saccharomyces cerevisiae (7) and via molecular/genetic approaches in Arabidopsis thaliana (8) and D. melanogaster (9).
CKII preferentially phosphorylates Ser/Thr residues in an hyperacidic context (10), although phosphorylation of Tyr has been documented in at least one case, i.e. yeast Fpr3 (11). Analysis of the phosphorylation of synthetic peptides suggests that the consensus site for phosphorylation by CKII can best be described as (S/T)(D/E)X(D/E) (10). Consistent with this, a number of proteins critical for transcription, cell cycle regulation, and signal transduction contain such a site(s) and are known to be phosphorylated in vitro and in vivo (12). Although CKII activity is inhibited in vitro by polyacidic compounds, such as polyaspartate and polyglutamate (13), and stimulated by polybasic compounds, such as polylysine and protamine (14), the in vivo relevance of these observations is currently unknown. Comparisons between recombinant monomeric ␣ subunit and native or reconstituted ␣ 2 ␤ 2 holoenzyme have revealed that the ␤ subunit plays a complex role in regulating the basal activity of the ␣ subunit (15)(16)(17). Although the ␤ subunit stimulates the activity of the monomeric ␣ subunit ϳ5-fold against most substrates, it down-regulates phosphorylation of a select few proteins, notably calmodulin (14,18), the actin bundling protein, Sac6p, 2 and a novel Drosophila zinc finger protein, ZFP35. 3 The ␤ subunit is also subject to phosphorylation by the ␣ subunit (4), but the biological role of this reaction remains undefined.
Genetic analyses in budding and fission yeast have demonstrated that the enzyme is essential for viability (19,20). Studies utilizing temperature-sensitive alleles of the ␣ subunits of yeast CKII indicate a requirement of the enzyme for cell cycle progression in G 1 and G 2 /M (21), in the maintenance of cytoskeletal architecture (22), and for cytokinesis (20). In contrast to the two yeast models, analysis of DmCKII has been stymied, principally due to the absence of mutations, even though the cDNAs encoding DmCKII were the first to be isolated (23). However, two complementary approaches that have recently been applied to DmCKII have proven to be exceptionally useful for analyzing functions of this kinase in a metazoan context. The first of these is the two-hybrid system (24) that has been used to identify CKII interacting proteins, many of which appear to be potential substrates of this kinase. The second is genetic analysis on these interacting proteins via targeted misexpression of transgenes encoding nonphosphorylatable and constitutively phosphorylated variants using the Gal4-UAS system (25), followed by phenotypic analysis. Studies along these lines have identified a novel regulatory (␤Ј) subunit of DmCKII (9), suggested the presence of five alternative transcription start sites in the CKII␤ gene (26), and demonstrated an interaction of the homeobox protein Antennapedia, ANTP, with DmCKII␣ (27). In the case of ANTP, phosphorylation by DmCKII appears necessary for restricting its activity during embryogenesis. A somewhat similar situation exists for another homeobox protein, Engrailed, which also appears to be regulated by CKII-mediated phosphorylation (28). In addition, the segment polarity protein Dishevelled, a component of the wingless/ Wnt signaling pathway (29), is a target for CKII, and both proteins exist as a complex in vivo (30). Collectively, these results suggest that CKII plays crucial roles in embryonic development as well as in cellular differentiation.
In an attempt to better define the physiological role of CKII, we have used the two-hybrid approach (24) to identify and characterize physiological partners of the ␣ subunit of DmCKII (DmCKII␣). One of the proteins identified in this screen is m7, a bHLH-type transcription factor derived from the neurogenic locus E(spl)C (31). The E(spl)C encodes the structurally and functionally similar bHLH proteins mC (also known as m␦), mB (also known as m␥), mA (also known as m␤), m3, m5, m7, and m8 (32,33) and is epistatic to other neurogenic loci, such as Notch, Delta, etc. (34 -36). The segregation of neural and epidermal lineages during development is determined by cell-cell communications that involve two interacting sets of genes: the neurogenic genes mediate signals between adjacent cells, and the proneural genes promote neural development. This nomenclature is explained by the fact that the neurogenic genes are named for their loss-of-function phenotype, whereas the proneural genes are named for their normal function. In a process termed "lateral inhibition," Delta provides the inhibitory signal that is received by Notch. The strength of this signal in the target cell determines functional predominance of the products of either the achaete-scute complex or the E(spl)C. Predominance of the former leads to neurogenesis, whereas that of the latter leads to inhibition of neurogenesis, i.e. epidermogenesis (37). Proteins of the E(spl)C form heterodimers with Groucho (38, 39), a nuclear protein that contains WD40 repeats (40). This interaction occurs via a C-terminal tetrapeptide, WRPW, that is invariant among all E(spl)C members, as well as in Hairy and Deadpan, which regulate segmentation and sex determination, respectively (38). E(spl)C-derived bHLH proteins have been proposed to inhibit neurogenesis via transcriptional repression of proneural genes by binding the N-box sequence, CACNAG (41), as well as by sequestering the proneural proteins themselves (42,43).
The studies described in this report demonstrate that in addition to m7, DmCKII␣ also interacts with two other E(spl) proteins, m5 and m8, and that all three proteins are phosphorylated by DmCKII in vitro. Protein-protein interaction analysis demonstrates a direct physical association between m8 and Drosophila embryo CKII. Using site-directed mutagenesis in combination with phosphorylation and two-hybrid analyses, we have mapped the site of phosphorylation, and we demonstrate that replacement of the phosphoacceptor Ser with Ala (a nonphosphorylatable residue) attenuates interaction of m8 with CKII, whereas substitution with Asp (which mimics the constitutively phosphorylated protein) abolishes the interaction. These studies demonstrate that the bHLH proteins m5, m7, and m8 are new physiological partners and substrates of DmCKII.

EXPERIMENTAL PROCEDURES
Construction of Two-hybrid Plasmids-DNA corresponding to amino acids 1-336 of DmCKII␣ and amino acids 1-215 of DmCKII␤ was amplified by polymerase chain reaction using primers containing two terminal 5Ј bases, a restriction site, and 20 bases of exact homology to the start and stop codon regions. The polymerase chain reaction products were subcloned into the plasmids pGBT9 and pGAD424 (gift of S. Fields, University of Washington) and completely sequenced using the Prism Dye Terminator Cycle sequencing kit (Applied Biosystems). The resulting plasmids express DmCKII␣ and DmCKII␤ as C-terminal fusions with the DNA binding (DB) domain, and activation domain (AD) of S. cerevisiae Gal4 (44), respectively.
Gal4-based Yeast Two-hybrid Screening-Screens were conducted in the Gal4-based version of the two-hybrid system (henceforth referred to as the Fields system) using the yeast strain HF7C (MATa,112, gal4-542, gal80-538, LYS2::GAL1 UAS -GAL1 TATA -HIS3, URA3::GAL4 17mers(x3) -CyC1 TATA -LacZ) (45). HF7C expressing Gal4DB-DmCKII␣ (the bait) was used to screen a 3-18-h D. melanogaster embryo two-hybrid cDNA library (gift of S. J. Elledge, Baylor College of Medicine). This library is contained in the plasmid pACT, which expresses cDNA-derived proteins as C-terminal fusions with Gal4AD (46). A total of 2 ϫ 10 6 transformants were plated on glucose dropout medium lacking tryptophan, leucine, and histidine (47), and colonies exhibiting rapid growth were counterscreened for expression of LacZ (48). Of the 45 His ϩ colonies, 15 tested positive for LacZ and were therefore chosen for further analysis. The library plasmids containing the yeast LEU2 gene were selectively recovered via complementation of the leucine auxotrophy of Escherichia coli HB101 (47). The isolated plasmids were subsequently used to retransform HF7C expressing Gal4DB-alone, GAL4DB-DmCKII␣ and GAL4DB-DmCKII␤. Those cDNAs that induced expression of HIS3 and LacZ only in response to GAL4DB-DmCKII␣, i.e. a bait-specific manner, were identified by sequencing their 5Ј-and 3Ј-ends using the primers 5Ј-ATACCACTACAATGGATGATG-3Ј and 5Ј-ACAGTTGAAGT-GAACTTGCG-3Ј, respectively. All novel cDNAs were completely sequenced using custom primers as described above. One of these cDNAs, DmA51, which encodes the bHLH protein m7, is the subject of this study, whereas others will be reported elsewhere.
LexA-based Two-hybrid Interactions-Explicit interactions between DmCKII␣ and E(spl) proteins were studied in the LexA-based version of the two-hybrid system (49) that was developed in the laboratory of Roger Brent (henceforth referred to as the Brent system). In the Brent system, proteins to be tested for interaction are expressed as fusions with the DNA binding domain of the bacterial repressor, LexA, and the activation domain of protein B42. The yeast strain used for these studies was EGY048 (MATa, trp1, his3, ura3, leu2), which harbors a single chromosomally integrated copy of the yeast LEU2 gene under the control of six LexA operators, and a high copy plasmid, pSH18-34, which expresses E. coli LacZ under the control of eight LexA operators (50). Therefore, expression of the two reporter genes, LEU2 and LacZ, is induced when the interacting complex is tethered to the LexA-operators. Additionally, expression of the AD fusion protein is under control of a GAL-promoter. As a result, reporter gene expression, in an AD fusion protein-dependent manner, is only observed when cells are grown in media containing galactose, but not glucose, as the sole carbon source. Yeast EGY048 containing plasmid pSH18-34 was transformed with a plasmid expressing the B42-derived AD-alone (49) or AD-Dm-CKII␣ fusion protein using lithium acetate (47). A single transformant was selected and subsequently retransformed with plasmids expressing LexA-m7, LexA-m8, LexA-m5, LexA-m3, and LexA-mC (38). Three independent transformants were tested for induction of the LEU2 gene on glucose-and galactose-dropout medium lacking leucine (47) at 29°C for 4 days. In parallel, cultures were analyzed in triplicate for ␤-galactosidase (LacZ) activity using a filter, as well as a solution-based assay (48).
Purification of Glutathione S-Transferase (GST) Fusion Proteins-The construction of plasmids expressing E(spl) proteins as C-terminal fusions with Schistosoma japonicum GST has been described previously (38). Plasmids expressing GST-alone, GST-m7, GST-m8, GST-m5, and GST-mC were transformed into E. coli BL21(DE3) harboring the plas-mid pT-TRX (gift of S. Ishii, Laboratory of Molecular Genetics, Ibaraki, Japan). pT-TRX drives expression of thioredoxin, which increases the solubility and functionality of eukaryotic proteins expressed in E. coli (51). Cultures (100 ml) were grown in 2ϫ YTA (47) containing 150 g/ml ampicillin and 15 g/ml chloramphenicol to an A 600 of 0.7 and induced with 1 mM isopropyl-␤-D-thiogalactoside for 3 h at 30°C with vigorous shaking. All subsequent steps were conducted at 4°C. Cells were harvested, resuspended in 8 ml of phosphate-buffered saline containing 0.1 mM phenylmethylsulfonyl fluoride, 1 mM EDTA, and 0.2% 2-mercaptoethanol and lysed by sonication. Phase contrast microscopy was used to ensure greater than 95% cell lysis. Triton X-100 was added to a final concentration of 1% and mixed for 3 h at 4°C. Insoluble material was removed by centrifugation, and the supernatant was passed twice through a column containing 1 ml of glutathione-Sepharose 4B (Amersham Pharmacia Biotech). The column was washed with 10 bed volumes of phosphate-buffered saline, and bound protein was eluted with 5 ml of 100 mM reduced glutathione in 50 mM Tris, pH 8.0. The eluted protein was concentrated and exchanged into storage buffer (10 mM Tris, pH 8.0, 0.5 mM EDTA, 5% glycerol, 150 mM NaCl, and 0.1 mM phenylmethylsulfonyl fluoride) using a Biomax-10K centrifugal filter device (Millipore). The purity of the fusion proteins was determined by SDS-polyacrylamide gel electrophoresis essentially as described (52), and their concentration was estimated from Coomassie Blue-stained gels relative to known standards.
Purification of DmCKII as a Monomeric Catalytic Subunit and the ␣ 2 ␤ 2 Holoenzyme-The monomeric ␣ subunit of DmCKII was purified to homogeneity from an S. cerevisiae expression system as described (15), and the tetrameric holoenzyme was purified from embryos according to Glover et al. (4), with modifications that will be described elsewhere. The V max of the ␣ subunit monomer is 0.4 mol/min/mg, whereas that of the holoenzyme is 1.6 mol/min/mg, using partially hydrolyzed and dephosphorylated casein (Sigma) as a substrate. These values are similar to those reported earlier (14).
Phosphorylation of m5, m7, and m8 by CKII-The reaction was carried out at 25°C in 50 mM Tris, pH 8.5, 100 mM NaCl, 10 mM MgCl 2 , 10 M ATP, 5 Ci of [␥-32 P]ATP, and ϳ2 g of various GST-E(spl) fusion proteins in a total volume of 40 l. The reaction was initiated with 5 l of the enzyme (either as the ␣ subunit monomer or the holoenzyme) at a concentration of 8 g/ml in 20 mM Tris, pH 8.0, 0.5 mM EDTA, 200 mM NaCl, 10% glycerol, 0.5 mM dithiothreitol, and 0.05% Triton X-100. To study the effects of poly(DL)lysine on phosphorylation, reactions were supplemented to a final concentration of 100 g/ml. The reactions were terminated with 10 l of 5ϫ sample buffer (312 mM Tris-Cl, pH 6.8, 10% SDS, 25% 2-mercaptoethanol, and 40% glycerol), and boiled for 5 min. Samples were separated by SDS-polyacrylamide gel electrophoresis and stained with Coomassie Blue, and the destained gels were exposed to Kodak XAR-5 film at room temperature.
Deletion Mapping and Site-directed Mutagenesis-The construction of variants of m7 lacking either the bHLH domain (m7⌬bHLH), or the WRPW motif (m7⌬WRPW) has been described previously (38). Two variants for mapping of the phosphorylation site were made in the cDNA encoding m8 using the Quick-Change site-directed mutagenesis kit (Stratagene). These are m8S159A (which replaces Ser 159 with Ala), and m8S159D (which replaces Ser 159 with Asp). The two complementary primer sets used for the former variant were 5Ј-CCGGATAT-CACGCCGACTGCGACAGC-3Ј and 5Ј-GCTGTCGCAGTCGGCGTGAT-ATCCGG-3Ј, and those for the latter variant were 5Ј-CCGGATATCA-CGACGACTGCGACAGC-3Ј and 5Ј-GCTGTCGCAGTCGTCGTGATAT-CCGG-3Ј, respectively. The underlined bases correspond to those substituting Ser 159 with either Ala or Asp. A plasmid containing the complete open reading frame encoding m8 was subjected to 17 cycles of polymerase chain reaction using the primer sets described above, and the polymerase chain reaction product was digested with the enzyme DpnI to eliminate the nonmutant plasmid that was used as a template. The reaction mixture was used to transform E. coli DH5␣, and the cDNA from a representative transformant was completely sequenced on both strands using custom primers. Subsequently, the cDNAs encoding m8S159A and m8S159D were subcloned into the EcoRI-BamHI sites of the vectors, pZEX, for expression and purification of GST fusion proteins (38), and pEG202, for expression as LexA fusions for two-hybrid analysis (49). Phosphorylation of wild-type m8 and the two variants, and their interactions with DmCKII␣ in the Brent system were conducted as described above.
In Vitro Interaction and Immunoblotting-Two g of purified GSTalone or GST-m8 were mixed with 25 l of glutathione-Sepharose 4B and incubated overnight at 4°C. The Sepharose was pelleted by centrifugation for 1 min at 2000 ϫ g, and the beads were washed once with 1.5 ml of wash buffer (50 mM Tris, pH 7.5, 5 mM EDTA, 150 mM NaCl, 5% glycerol, 1 mM phenylmethylsulfonyl fluoride, and 0.1% Triton X-100) to remove unbound GST fusion proteins. The washed Sepharose, containing the immobilized GST fusion proteins, was then incubated with 100 ng of purified Drosophila embryo CKII and incubated for 3 h at 4°C. The Sepharose was pelleted by centrifugation for 1 min at 2000 ϫ g, and the supernatant was recovered as unbound material. The pellets were washed two times for 5 min each time with 500 l of wash buffer. Sepharose-bound (pellet) and unbound (supernatant) fractions were resolved by SDS-polyacrylamide gel electrophoresis and electrophoretically transferred to nitrocellulose as described (53). DmCKII subunits were detected by Western blot analysis using primary antibody against DmCKII (54) at a dilution of 1:1000 and secondary antibody (goat-anti-rabbit IgG coupled to alkaline phosphatase, Bio-Rad) at a dilution of 1:3000. Immunoblots were visualized using nitro blue tetrazolium and 5-bromo-4-chloro-3-indoyl phosphate (47).

RESULTS AND DISCUSSION
Isolation of cDNAs Encoding m7-The yeast strain HF7C expressing Gal4DB-DmCKII␣ as a bait was used to screen a D. melanogaster embryo two-hybrid cDNA library. From ϳ2 ϫ 10 6 transformants, 15 clones that activated transcription of HIS3 and LacZ were recovered. All 15 clones induced the two reporter genes only when cotransformed with DmCKII␣ (data not shown). Sequencing of the cDNAs revealed that seven of the clones encode DmCKII␤ (26), one encodes DmCKII␣ (9), two encode DmCKII␤Ј (a novel isoform of the ␤ subunit (9)), one (DmA51) encodes m7, and the rest encode novel proteins that will be described elsewhere. The library plasmid was recovered from yeast clone DmA51 and retested for interaction against various bait constructs. As shown in Fig. 1A, induction of HIS3 and LacZ was observed only when yeast HF7C coexpressed Gal4AD-m7 with Gal4DB-DmCKII␣. On the other hand, neither reporter gene was induced when HF7C was transformed with Gal4AD-m7 by itself or in combination with a plasmid encoding either Gal4DB-alone or Gal4DB-DmCKII␤, suggesting that m7 interacts specifically with the catalytic subunit of DmCKII. The inability of DmCKII␤ to interact with m7 is not due to a lack of expression of the former protein because this construct is expressed in yeast and displays a strong interaction with DmCKII␣ (9).
Sequencing revealed that the DmA51 cDNA encodes fulllength m7 and contains 78 and 228 base pairs of sequence, 5Ј to the initiation codon (ATG) and 3Ј to the termination codon (TAA), respectively. This cDNA, which does not contain any in-frame stop codons 5Ј to the initiation codon, is identical to base pairs 172-1068 of a 4.4-kilobase genomic clone (Fig. 1B) that encodes the m7 and m8 transcription units, each on a single uninterrupted exon (55). The absence of a poly(A) tail in clone DmA51, combined with the presence of a single poly(A) addition signal in the corresponding gene at position 1250 (Fig.  1B), suggests that the isolated cDNA is not full-length with respect to its 3Ј untranslated region. In this regard, we have recently rescreened the Drosophila cDNA library for Dm-CKII␣-interacting proteins and have isolated two additional clones, DmA002 and DmA130, that also encode m7. Our isolation of multiple cDNAs encoding m7 from two independent two-hybrid screens strengthens the likelihood of the relevance of its interaction with DmCKII␣. Apart from length heterogeneity with respect to the DmA51 cDNA, the DmA002 and DmA130 sequences are identical to the corresponding region of the m7 transcription unit and display no polymorphisms (data not shown).
Interaction of DmCKII␣ with E(spl)C-derived bHLH Proteins-The observed interaction between DmCKII␣ and m7 was surprising, as there was no previous indication that m7 is regulated by phosphorylation or that CKII is involved in neurogenesis. Given the structural similarity of all E(spl) proteins (32), we were interested in determining whether DmCKII␣ also interacts with other members derived from this locus. For this purpose, we made use of the Brent system (see under "Experimental Procedures") to remain consistent with the analysis of Paroush et al. (38), who have convincingly demonstrated the interaction of E(spl) proteins with Groucho. We therefore transformed yeast EGY048 with plasmids expressing the ADalone or the AD-DmCKII␣ fusion protein, and the resulting strains were retransformed with plasmids expressing various LexA-E(spl) fusion proteins (38). As shown in Fig. 2, induction of LacZ was specifically observed when cells coexpressed Dm-CKII␣ with m5, m7, or m8. The levels of LacZ induced with m3 or mC were similar to those obtained with LexA-alone, suggesting that the DmCKII␣-m5/7/8 interactions are not mediated via the LexA domain. Furthermore, no significant reporter gene expression was observed when m5/7/8 were tested against the AD-alone, indicating specificity with regard to DmCKII␣. The inability of m3 and mC to interact with DmCKII␣ is not due to attenuated/lack of expression, as these constructs display robust interactions with Groucho that are equivalent to those observed with m5, m7, and m8 (38). The higher levels of LacZ activity observed for mC in combination with the AD-alone and its "silencing" upon expression of AD-DmCKII␣ are consistent with our observation that the unfused AD, in a limited number of cases, confers basal transcription of the reporter genes that is abolished upon expression as a fusion protein. 4 That Dm-CKII␣-m5, -m7, and -m8 interactions display a higher LacZ activity (Fig. 2) than does DmCKII␣-DmCKII␤ (Fig. 1A) does not imply that the former protein pairs interact with a higher affinity. As outlined under "Experimental Procedures," this is a reflection of a high copy/affinity LacZ reporter used in the Brent system. In addition, no induction of LacZ activity was observed when transformants were grown in rich glucose medium (data not shown), suggesting that reporter gene expression was dependent on presence of the AD-DmCKII␣ fusion protein. Identical results were observed for the second reporter gene, LEU2 (data not shown). It should be noted that the two-hybrid interaction between DmCKII␣ and m7 was originally detected using the former protein as a Gal4DB fusion and the latter as a Gal4AD fusion (Fig. 1A), whereas the explicit testing involved the inverse orientation, i.e. m7 as a LexA fusion and DmCKII␣ as an AD fusion (Fig. 2). These results demonstrate that the interactions are neither orientation-specific, as is the case of the interaction of DmCKII␣ with Dm-CKII␤ (9), nor dependent on a specific version of the two-hybrid system.
The apparent specificity of DmCKII␣ for m5, m7, and m8 but not for other members tested (m3 and mC) was surprising given that these proteins are structurally conserved (32). Of particular note are two motifs, the highly conserved HLH domain, which mediates heterodimerization with proneural proteins such as Ac, Da, and l'sc (42), and the invariant WRPW motif (32), which mediates interaction with Groucho (38). We found, however, that variants of m7 lacking either of these motifs interact as effectively with DmCKII␣, as does the wildtype protein (data not shown). These results are consistent with our inability to detect interactions of DmCKII␣ with m3 and mC (both of which contain the HLH and WRPW motifs) and suggest that the interaction domain lies elsewhere in m5/7/8.
Conservation of a CKII Site in m5, m7, and m8 -As men-4 R. L. Trott and A. P. Bidwai, unpublished data. tioned above, all E(spl)C-derived proteins are structurally conserved (32). However, the sequence alignment presented by Delidakis and Artavanis-Tsakonas (32) emphasized conservation of the HLH domain, helices III and IV (also known as Orange domain), a motif in the vicinity of the C terminus with a high PEST score, and the WRPW motif (Fig. 3A). We therefore aligned the seven E(spl) proteins with emphasis on residues N-terminal to the basic domain and those comprising the region from helix IV to the C terminus (C-domain) to determine whether some structural features were unique only to m5, m7, and m8. No sequences in the N terminus were found that were conserved among and/or unique to these three proteins (data not shown). On the other hand, analysis of the C-domain indicates that only these three proteins contain a consensus site for phosphorylation by CKII, 156 SDNE in m5, 168 SDNE in m7, and 159 SDCD in m8, immediately following the highly conserved sequence, (I/L)SP(V/A)SSGY (Fig. 3B), in a region that is characterized by a high PEST score (32). Although PEST-rich sequences act as cis-acting signals that regulate protein turnover (56) and have been suggested to be activated via phosphorylation (see below), the role of this motif in m5/7/8 is currently unknown. This conserved Ser in m5/7/8 conforms to the requirement that it must contain an acidic residue at the nϩ1 and nϩ3 positions to be a target for CKII (10,57,58). It should be noted, that although mB also contains a site for phosphorylation by CKII ( 195 SEDE), it is neither preceded by the (I/ L)SP(V/A)SSGY sequence nor contained within its PEST motif. Interestingly, the cytology and spatial organization of the E(spl)C locus of Drosophila hydei exhibits an extraordinary level of conservation relative to that of D. melanogaster (59). Because the DNA sequence of only D. hydei m8 is currently available, we have compared this protein with m5, m7, and m8 from D. melanogaster. Remarkably, D. hydei m8 also contains the CKII site following the (I/L)SP(V/A)SSGY sequence, both of which are contained within a region with a high PEST score (see Fig. 3B). Given that the two species, D. melanogaster and D. hydei, diverged ϳ60 million years ago (60), evolutionary principles would argue against the notion that the conservation of the aforementioned motifs is merely incidental. A more compelling case will become apparent once the sequences encoding other E(spl) proteins of D. hydei become available.
Phosphorylation of m5, m7 and m8 by Drosophila CKII-One question raised by the sequence alignment was whether the presence of the consensus CKII site in m5, m7, and m8 correlates with their phosphorylation. We have therefore subjected GST-alone, GST-m5, GST-m7, GST-m8, and GST-mC, a noninteracting member, to phosphorylation using two isoforms of CKII, i.e. the monomeric ␣ subunit purified from a yeast expression system (15), and the ␣ 2 ␤ 2 holoenzyme purified from embryos (4). The former isoform mimics the two-hybrid analysis (Fig. 2), whereas the latter mimics the in vivo environment. The results demonstrate that m5, m7, and m8 are phosphorylated by both isoforms of CKII (Fig. 4, B and E, lanes 2-4) and corroborate their observed two-hybrid interaction with Dm-CKII␣. No phosphorylation of either GST or GST-mC (Fig. 4, B and E, lanes 1 and 5) was observed with either enzyme isoform, demonstrating the absence of phosphorylation of the affinity tag used for purification and suggesting that phosphorylation is specific only to those E(spl) proteins that also exhibit a two-hybrid interaction with DmCKII␣. At a quantitative level, however, the rates of phosphorylation of the three E(spl) proteins are different for both enzyme isoforms, such that m5 Ͼ m7 ϭ m8 (compare lanes 2, 3, and 4 in Fig. 4, B and E). What mechanism can account for the observed differences? Detailed kinetic analysis of CKII suggests that whereas DmCKII␣ and the holoenzyme display virtually identical k m values for the protein substrate, the K cat can differ 5-50-fold in a substratedependent manner (14). Furthermore, studies with peptides suggest that whereas the acidic residues at nϩ1 and nϩ3 are absolutely required for phosphorylation, additional acidic residues C-terminal to the nϩ3 position further increase the K cat with marginal effects on the k m (57,61). These criteria, therefore, make it possible to predict the relative rates of phosphorylation of m5/7/8. In this regard, although m7 and m8 fit the consensus, m5 is probably the best because it contains an additional Asp at the nϩ4 position (Fig. 3B). The rank order for phosphorylation is, therefore, predicted to be m5 Ͼ m7 ϭ m8. The analysis presented here essentially reflects this prediction. Because the gel analysis described here inherently reflects a semiquantitative assessment of phosphorylation, kinetic analysis will be necessary to determine whether the observed differences in phosphorylation of m5 versus m7/8 are due to differing catalytic efficiencies (K cat /k m ). That CKII interacts with and phosphorylates these proteins is consistent with the observation that this kinase has been found to exist in a complex with some of its in vivo substrates, such as Topoisomerase II (62), HSP90 (63), ANTP (27), and Dishevelled (30), to name a few.
We and others have previously demonstrated that polybasic compounds, e.g. polylysine, overcome a down-regulation of the holoenzyme that can be conferred by the ␤ subunit of CKII for some substrates (14,64). We were therefore interested in determining whether the limited phosphorylation of m7 and m8, relative to that of m5, might be sensitive to polylysine addition. The results suggest that phosphorylation of m5 is virtually unaffected by this compound when tested with either DmCKII␣ (compare Fig. 4B, lane 2, to Fig. 4C, lane 2) or DmCKII holoenzyme (compare Fig. 4E, lane 2, to Fig. 4F, lane 2). On the other hand, phosphorylation of both m7 and m8 is stimulated by polylysine addition (compare lanes 3 and 4 in Fig. 4B versus 4C  and 4E versus 4F). These results are consistent with previous analysis indicating that phosphorylation of substrates with optimal sites (such as the RII subunit of cAMP-dependent protein kinase) responds modestly, if at all, to polylysine, whereas those that satisfy the minimal requirements of CKII (such as calmodulin) are more responsive (14). This stimulation by polylysine is not a reflection of promiscuous phosphorylation, because DmCKII␣ or the holoenzyme do not phosphorylate GST or GST-mC (lanes 1 and 5 in Fig. 4, C and F) in the presence of polylysine. However, no cellular protein that can mimic the "polylysine effect" with respect to CKII has so far been identified, unlike the case with Ras, which mediates the polylysine-dependent phosphorylation of calmodulin by the insulin-receptor kinase (65).
Is there any evidence, apart from our two-hybrid and phosphorylation analysis, to suggest that m5/7/8 are more closely related to each other than are other E(spl) proteins? We believe that molecular/genetic analyses do, in fact, support this proposal. Using a bacteriophage -based system to detect proteinprotein interactions, Gigliani et al. (43) suggest that m5, m7, and m8 are the most closely related. Although yeast two-hybrid analysis conducted by Alifragis et al. (42) essentially reiterates the closest similarity between m5 and m8, they suggest, however, that m7 be clustered along with mA and mB, a proposal at odds with their own genetic analysis (see below). Because E(spl) proteins homo/heterodimerize and interact with proneural proteins as well, in vivo associations between these proteins is needed to clarify the differences, if any, with regards to m7. Furthermore, and perhaps the most persuasive, is genetic analysis demonstrating that the severity of suppression of bristle development, i.e. neurogenesis, closely correlates with ectopic expression of only m7 and m8 (66). A similar analysis with m5 was, however, precluded even with two copies of the transgene, leading the authors to conclude that, of the seven E(spl) proteins, m5 is probably most inactive/unstable (66). Northern or Western analysis on the m5 transgenics will be necessary to clarify whether this is indeed the case. Given the functional equivalency of m5/7/8 in neurogenesis, we were specifically interested in determining the mechanism by which these proteins interact with DmCKII. We have deferred conducting parallel, and thus redundant, analysis on all three proteins and have selected m8 for these additional studies. This choice was based on our eventual goal of analyzing the significance of this interaction in transgenic flies and to remain consistent with genetic analysis on this protein (see below and Ref. 67) via the GAL4-UAS system (25).
Interaction of m8 with DmCKII-Although the available data demonstrate that m5/7/8 interact with DmCKII and are phosphorylated by it, they do not indicate whether these proteins are capable of direct physical association. Analysis of complex formation between these proteins in the developing Drosophila embryo is currently precluded by the absence of antibodies that specifically recognize the m8 protein (68), coupled with its restricted expression domains within the neuroectoderm (33). As an alternative, we have assessed the ability of recombinant bacterially expressed GST-m8 to form a physical complex with CKII purified from Drosophila embryos. To this end, GST-alone and GST-m8 were purified, immobilized on glutathione-Sepharose beads, and tested for their ability to form a complex with Drosophila embryo CKII. The presence of FIG. 4. Phosphorylation of m5, m7, and m8 by DmCKII. The indicated GST fusion proteins were purified and subjected to phosphorylation using the monomeric ␣ subunit (DmCKII␣) and the ␣ 2 ␤ 2 holoenzyme from Drosophila embryos. A representative gel stained with Coomassie Blue shows the amount and purity of the various GST fusion proteins that were phosphorylated with either DmCKII␣ (A) or the holoenzyme (D). Proteins were phosphorylated in either the absence (B and E) or in the presence (C and F) of 100 g/ml poly(DL)lysine. Samples were electrophoresed in 12% SDS-polyacrylamide gels, stained with Coomassie Blue, and autoradiographed (B, C, E, and F). Arrows in A and D indicate the mobilities of the full-length fusion proteins.
FIG. 5. Interaction of DmCKII with m8. Bacterially expressed GST fusion proteins immobilized on glutathione-Sepharose beads were incubated with Drosophila embryo holoenzyme. The beads were separated from the unbound material as described under "Experimental Procedures," and the bead-bound (P, pellet) and the unbound (S, supernatant) samples were examined by Western blotting using antisera raised against Drosophila embryo CKII. The arrows indicate the immunoreactive bands corresponding to the ␣ and ␤ subunits of DmCKII.
DmCKII in the bead-bound (pellet) and unbound (supernatant) fractions was assessed by Western blotting using an antisera that recognizes both subunits (␣ and ␤) of DmCKII (54). As expected, incubation of the GST beads with DmCKII did not result in any immunoreactive material in the pellet fraction, indicating the absence of an interaction (Fig. 5, compare lanes P and S). On the other hand, incubation of GST-m8 beads with DmCKII resulted in the presence of immunoreactive material in the pellet fraction (Fig. 5, compare lanes P and S), demonstrating that these two proteins form a physical complex. These results suggest that the two-hybrid interaction of DmCKII with m8 is direct and is unlikely to be mediated by the recruitment of yeast proteins.
Mapping the Site of Phosphorylation on m8 -We next sought to define the site of phosphorylation. We therefore generated two variants of m8 with substitutions of the conserved Ser in the CKII site, i.e. m8S159A and m8S159D. The former is a nonphosphorylatable variant, whereas the latter should mimic the constitutively phosphorylated protein, in line with studies on ANTP (27), HP1 (69), etc. GST-m8, GST-m8S159A, and GST-m8S159D fusion proteins were purified and subjected to phosphorylation using DmCKII␣ and the holoenzyme. The results demonstrate that GST-m8 is phosphorylated by the holoenzyme and the ␣ subunit (Fig. 6B, lanes 1 and 4). On the other hand, neither GST-m8S159A (Fig. 6B, lanes 2 and 5) nor GST-m8S159D (Fig. 6B, lanes 3 and 6) are substrates of the two enzyme isoforms. This result strongly suggests that CKII phosphorylation of m8 occurs at Ser 159 , and by corollary the site of phosphorylation on m7 and m5 is most likely to be Ser 168 and Ser 156 , respectively (see Fig. 3B). We consider it unlikely that GST-m8S159A and GST-m8S159D are partially clipped leading to abolished phosphorylation, because, relative to GST-m8, neither protein exhibits altered mobility in SDS-polyacrylamide gels (Fig. 6A, compare lane 1 with lanes 2 and 3). In addition, the inability of the two variants to be phosphorylated by CKII suggest that m8 contains a single site for phosphorylation by CKII, thus corroborating our sequence-based prediction (see Fig. 3B). The ability of m8 to be phosphorylated by DmCKII␣ and the holoenzyme at the identical residue is also consistent with our contention that the substrate specificity of this enzyme is intrinsic to the ␣ subunit (14).
Interaction of m8S159A and m8S159D with DmCKII␣-We were interested in determining whether phosphorylation of m8 affects its interaction with CKII. We, therefore, determined the interaction of DmCKII␣ with m8S159A and m8S159D, relative to wild-type m8. As shown in Fig. 7, replacement of Ser 159 with Ala decreased interaction by ϳ50%, whereas replacement with Asp abolished the interaction. These results suggest that the interaction of DmCKII␣ with m8 appears analogous to that of an enzyme with its substrate and is in line with the interactions of the protein kinase, Snf1, with its substrate, Snf4 (70). Our interpretation of the results is, however, complicated by the fact that phosphorylation of m8 appears to disrupt the complex and may affect protein stability as well (see below). The former possibility is likely, given that DmCKII␣ is catalytically active when expressed in yeast (15), and suggests that the strength of the two-hybrid interaction observed between DmCKII␣ and m5/7/8 may, in fact, represent an underestimate. The likelihood of the latter possibility is difficult to predict, at least in the context of the yeast system used in this study, given that two-hybrid interaction of m8S159D with Groucho appears identical to that observed for wild-type m8. 4 These results suggest that accessory proteins, perhaps lacking in yeast, may be necessary for affecting stability of m8S159D in Drosophila. In addition, although D. melanogaster m5/7/8 and D. hydei m8 contain the conserved sequence, (L/I)SP(V/ A)SSGY, flanking the phosphorylation site (see Fig. 3B), it is presently unknown whether these residues contribute binding energy in addition to that attributable to the CKII site. Studies with Snf1/Snf4 have, in fact, demonstrated the involvement of flanking residues in mediating interactions (71). Further analysis will be needed to determine the intrinsic affinities (K a ) of these bHLH proteins for DmCKII, rather than those (k m ) inferred by kinetic analysis.
Implications of Phosphorylation of m5, m7, and m8 -The results presented above raise the likely prospect that DmCKII interacts with m5/7/8 when these proteins are in the nonphosphorylated state and that the complexes dissociate upon phosphorylation. We obviously cannot extrapolate the two-hybrid and biochemical results to the situation in the epidermal precursors in the developing Drosophila embryo with certainty. However, given the requirements of CKII for cell cycle progression (21) and for checkpoint control (72), it is likely that epidermal progenitors, which are expressing E(spl) proteins, also contain CKII. A direct test of this proposal in the developing embryo still remains a difficult task due to restricted expression of m5/7/8 and the absence of isoform-specific antibodies (see above). At a functional level, our data indicate that interaction and/or phosphorylation of m5/7/8 is unlikely to affect their DNA binding properties (which require the basic region), their ability to heterodimerize with proneural proteins (which requires the HLH domain), or their ability to interact with Groucho (which requires the WRPW motif). What function could then be ascribed to interaction and/or phosphorylation? FIG. 6. Mapping of the CKII phosphorylation site on m8. A and B, the indicated GST fusion proteins were purified and subjected to phosphorylation using the ␣ 2 ␤ 2 holoenzyme from Drosophila embryos (lanes 1-3) and the monomeric ␣ subunit, DmCKII␣ (lanes 4 -6). Lanes 1 and 4, wild-type m8; lanes 2 and 5, m8S159A; lanes 3 and 6, m8S159D. Samples were electrophoresed in 12% SDSpolyacrylamide gels, stained with Coomassie Blue (A), and autoradiographed (B). The position of the GST fusion proteins is indicated by the arrow.
The structural and functional properties common to m5/7/8, and by extension those in D. hydei m8, provide the basis for a likely possibility. As mentioned above, all three proteins contain a PEST-rich motif that harbors an invariant Ser residue that is phosphorylated by CKII. In this regard, a mutation that removes sequences encompassing the PEST-rich region and the resident CKII site acts as a dominant-negative allele with regard to suppression of bristle development (67). That this variant of m8 behaves as a dominant-negative, rather than a loss-of-function (as one would have predicted), suggests that the mutant protein might sequester endogenous wild-type m8, and possibly m5 and m7 as well, thus leading to enhanced neurogenesis. Thus, Giebel and Campos-Ortega (67) propose that this region negatively regulates the activity of m8, a suggestion in line with its ability to homodimerize or heterodimerize with m5 and m7 (42,43). These results and their interpretations are consistent with our proposal that this region of m5/7/8 may influence the stability of these proteins in vivo. In this regard, an interesting parallel has been identified, i.e. activation of the morphogenic protein Dorsal in Drosophila and that of NF-B in humans. Dorsoventral patterning in the Drosophila embryo involves the activities of a transcription factor, Dorsal, and its inhibitor, Cactus (73). Upon receiving the inductive signal, Cactus appears to be phosphorylated by CKII within a motif with a high PEST score (74) and undergoes degradation, thus allowing Dorsal to translocate to the nucleus. A mechanistically similar situation appears to regulate the NF-B/C-Rel family of transcription factors in humans (75,76). A collective theme that emerges from these studies is that phosphorylation, in at least a restricted class of proteins, regulates protein stability via activation of PEST motifs. Our studies further implicate the PEST motif in m5, m7, and m8 as a target for regulation via CKII-mediated phosphorylation. Future studies employing expression of epitope-tagged m8 and its nonphosphorylatable and/or constitutively phosphorylated variants in transgenic flies will be needed to clarify the role of this motif. If these studies indicate this to be the case, we predict that the m8S159A variant would exhibit a longer halflife in vivo, thus leading to its predominance over the proneural proteins and therefore to an inhibition of neurogenesis. If so, the m8S159D variant may exhibit a shorter half-life in vivo, thus preventing antagonism of the proneural proteins. Experiments to address the role of phosphorylation on protein turnover and effects on neurogenesis, via the transgenic route (67), are currently under way. In an interesting twist, Alifragis et al. (42) report that a mutation in the proneural protein, Sc, that replaces Ser 340 with Asp, abolishes its interaction with m3. It should be noted that the sequences flanking Ser 340 (DYIS 340 LWQEQ) do not conform to the consensus for CKII or to that of other Ser/Thr protein kinases with defined substrate specificities. Although their data (42) are conjectural, when they are taken together with our results, it appears that neurogenic as well as proneural proteins may be regulated by phosphorylation.
In summary, the data presented herein demonstrate that select members of the E(spl)C, i.e. m5, m7, and m8, physically interact with DmCKII and are phosphorylated by this enzyme at an invariant Ser residue that is contained within a motif unique to these three isoforms. The suggestion that these three proteins are more functionally related (42,43,59,68) and that the C-terminal domain of m8 acts to negatively regulate function in vivo (67) implicates the PEST motif and its resident CKII phosphorylation site. We believe that the data presented strengthen our contention for the presence of a new functional motif in these transcriptional repressors and raise the possibility that CKII may regulate neurogenesis via posttranslational modification of these proteins.