Interactions between Casein Kinase Iε (CKIε) and Two Substrates from Disparate Signaling Pathways Reveal Mechanisms for Substrate-Kinase Specificity

Background Members of the Casein Kinase I (CKI) family of serine/threonine kinases regulate diverse biological pathways. The seven mammalian CKI isoforms contain a highly conserved kinase domain and divergent amino- and carboxy-termini. Although they share a preferred target recognition sequence and have overlapping expression patterns, individual isoforms often have specific substrates. In an effort to determine how substrates recognize differences between CKI isoforms, we have examined the interaction between CKIε and two substrates from different signaling pathways. Methodology/Principal Findings CKIε, but not CKIα, binds to and phosphorylates two proteins: Period, a transcriptional regulator of the circadian rhythms pathway, and Disheveled, an activator of the planar cell polarity pathway. We use GST-pull-down assays data to show that two key residues in CKIα's kinase domain prevent Disheveled and Period from binding. We also show that the unique C-terminus of CKIε does not determine Dishevelled's and Period's preference for CKIε nor is it essential for binding, but instead plays an auxillary role in stabilizing the interactions of CKIε with its substrates. We demonstrate that autophosphorylation of CKIε's C-terminal tail prevents substrate binding, and use mass spectrometry and chemical crosslinking to reveal how a phosphorylation-dependent interaction between the C-terminal tail and the kinase domain prevents substrate phosphorylation and binding. Conclusions/Significance The biochemical interactions between CKIε and Disheveled, Period, and its own C-terminus lead to models that explain CKIε's specificity and regulation.


Introduction
CKI family members have been implicated in a wide range of signaling activities including involvement in the Wnt, planar cell polarity (PCP) and circadian rhythms pathways [1][2][3]. In mammals, the CKI family of serine/threonine kinases consists of seven distinct genes (a, b, c 1 , c 2 , c 3 , d, e) that code for enzymes with highly conserved kinase domains and divergent amino-(N-) and carboxy-(C-) termini. CKIe in particular has a long (,100 residue) C-terminus that undergoes inhibitory autophosphorylation at conserved serine and threonine residues [4][5][6][7]. Subcellular localization of different isoforms in some cases is determined by splice variants (in the case of CKIa) [8,9] or fatty-acylation (in the case of CKIc) [10], but Western and Northern blotting experiments, and in situ hybridization studies indicate that the a, d, and e isoforms have generally overlapping expression [9,[11][12][13][14]. It is therefore interesting that despite high sequence similarity and common expression patterns, CKI family members have different targets in the cell [10,[15][16][17][18].
Physiological outputs (hormonal, metabolic and cellular) that are entrained to the 24 hour cycle of the earth's rotation have been described in organisms from cyanobacteria through humans, and are based on molecular clockwork, [reviewed in 3,34,35]. The cyclic expression and localization of the transcriptional regulator, Period (Per), is in large part regulated by phosphorylation by CKIe in mammals [36][37][38][39][40]. There are three Per isoforms in mice (mPer1, mPer2, mPer3) that display different expression and regulation patterns [41,42]. Mutations in CKIe disrupt the circadian pathway through aberrant or reduced phosphorylation of Per [39,[43][44][45]. Both CKIe and CKIa isoforms are found in cells that exhibit circadian cycling, including the master circadian oscillator in mice, the suprachiasmatic nucleus [14]. Why one CKI isoform is active in the pathway while the other is not has not been investigated.
Per and Dsh bind specifically to CKIe [19,37,[46][47][48]. To determine how CKIe and CKIa differ with respect to binding by substrates, we performed a series of in vitro experiments to probe the direct interaction between CKIe and two substrates from disparate signaling pathways. We show that, surprisingly, features C-terminal to the kinase domain of CKIe are not responsible for the difference in binding of the substrate proteins to the two CKI kinases. Instead, two residues on the kinase domain of CKIe determine Dsh's and Per's affinity for CKIe, whereas the Cterminus of CKIe stabilizes interactions between CKIe and Dsh and Per. Finally, we show that autophosphorylation of CKIe's Cterminus inhibits substrate binding. Using chemical cross-linking and tandem mass spectrometry, we demonstrate that autophosphoryation of the C-terminal tail changes its interaction with the kinase domain, revealing why the autophosphorylated tail inhibits substrate phosphorylation and binding.

Results
CKIe, but not CKIa, binds to xDsh and mPer1 in vitro Per1 and Dsh bind to CKIe in vivo, but they do not interact with the related isoform CKIa [23,[47][48][49]. In order to determine if these substrates bind to the two kinases with the corresponding affinities in vitro, we used a Glutathione S-Transferase (GST) pulldown system to follow protein-protein interactions. Due to protein instability and toxicity in E. coli, prior in vitro studies of CKIe have used a C-terminally truncated form of the protein called CKIeDC [48,50,51], which eliminates the last 98 residues of the protein.
Since we wanted to include all of the protein in our analysis, we developed an expression system for large quantities of pure fulllength Xenopus CKIe fused to GST (CKIe-GST; see Methods). Using a similar strategy we also expressed full-length Xenopus CKIa fused to GST (CKIa-GST). Both kinases were catalytically active against b-catenin, indicating that they were both well folded (data not shown). Because CKIe autophosphorylation regulates substrate binding (see below), we used dephosphorylated CKIe for our binding studies except where otherwise noted. Using GST-pulldown assays we found that 35 S-labelled Xenopus Dsh (xDsh) and mouse Per1 (mPer1) bound specifically to CKIe compared to CKIa (Fig. 1), recapitulating previously published in vivo results [19,37,[46][47][48]. This provided us with a useful assay system to examine CKIe specificity for these substrates.
CKIe does not require its C-terminus to interact with substrate proteins CKIe and CKIa are 75% identical, and 89% similar in their kinase domains (Fig. 2). This high degree of conservation between the kinases suggested that full binding by mPer1 and xDsh to CKIe might depend on the long C-terminal tail, which is unique to CKIe. We first examined binding of mPer1 and xDsh to CKIeDC, which truncates the kinase at residue 319 (Fig. 2). Figure 3A shows that xDsh and mPer bound to CKIe and CKIeDC to the same extent, demonstrating that CKIe does not require the autophosphorylation region of the C-terminus to bind to substrate proteins.
Because truncation of CKIe at residue 319 did not diminish xDsh or mPer1's affinity for the kinase, another part of the protein must determine CKIe's interaction with substrate proteins. CKIeDC retains 22 residues that are not conserved between CKIe and CKIa ( Fig. 2) and these could contribute to the substrate-kinase interaction. We produced CKIeDDC, which is truncated at residue 295 and therefore even more closely resembles CKIa (Fig. 2). Figure 3B shows that CKIeDC and CKIeDDC each bound to xDsh or mPer1with the same affinity. CKIe must therefore bind to substrate proteins through its kinase domain.

Residues N275 and R295 are responsible for enhanced substrate binding to CKIe
To identify residues within the CKIe kinase domain that could confer specificity, we examined the structure of CKId. CKId is considered a very close relative of CKIe since they share 82% identity over the entire protein and 96% identity in their kinase domains, and since their cellular activities are indistinguishable [11,13,24,48,52]. Residues shown in cyan in Figure 4A are conserved between CKIe and CKId but are not conserved in CKIa. Two residues that are conserved in CKIe and CKId but not CKIa, and have solvent exposed side-chains, Asparagine 275 (N275) and Arginine 279 (R279), are shown in red (Fig. 4A). In CKIa, the corresponding residues have chemical properties that differ substantially from the CKIe residues (AspRIle and ArgRThr), and therefore these residues could contribute specificity to CKIe-substrate interactions.
We initially changed N275 and R279 to alanines in the context of the conventional CKIeDC truncation (CKIeDC N275A/ R279A, Fig. 2). These mutations significantly disrupted the binding of xDsh to CKIe, although they had no effect on mPer1 binding (Fig. 4B, lane 4, and Fig. 4C). However, when we replaced N275 and R279 with the corresponding CKIa residues Ile and Thr, respectively (CKIeDC N275I/R279T, Fig. 2), the binding between CKIe and mPer1 was abolished and the binding between CKIeDC and xDsh was further reduced (Fig. 4B, lane 5, and Fig. 4C). In conclusion, N275 and R279 are required for strong binding between CKIe and xDsh, whereas the corresponding chemically distinct residues in CKIa prevent binding of mPer1 to CKIe.

The C-terminus of CKIe modulates the effect of mutations in the kinase domain
In order to determine what role the C-terminus might play in substrate recognition, we introduced the N275I and R279T substitutions into full-length CKIe (CKIe N275I/R279T, Fig. 2). xDsh's binding to CKIe N275I/R279T is marginally enhanced compared to its binding to CKIeDC N275I/R279T, while mPer1 binding is completely restored (compare Fig. 4 to Fig. 5). We conclude that the C-terminus is not necessary for CKIe to bind to substrates (see Fig. 3), but that it stabilizes otherwise weakened interactions between xDsh or mPer1 and CKIe.
We constructed a chimeric protein that contains the kinase domain of CKIa and the C-terminus of CKIe to determine if the C-terminus of CKIe could confer CKIe-like binding to CKIa  (CKIaRe, Fig. 2). Neither xDsh nor mPer1 showed strong binding to the chimeric protein, compared to CKIe (Fig. 6A). Their weak binding to CKIaRe along with the data presented above indicates that the C-terminus is able to contact and partially stabilize the interaction between CKIe and its substrates, but it is not sufficient for the binding of mPer1 or xDsh to CKI.

Conversion of CKIa to an xDsh-binding protein
We hypothesized that CKIa's residues I283 and T287, which correspond to CKIe's N275 and R279, were responsible for preventing interactions with CKIaRe. We therefore engineered CKIaRe I283N/T287R, which is nearly identical to CKIaRe but contains two point mutations that could potentially enhance binding by xDsh and mPer1 (Fig. 2). Impressively, conversion of CKIa residues I283 and T287 to the CKIe identity enabled xDsh to fully interact with the kinase. In contrast, mPer1 did not bind CKIaRe I283N/T287R. This data clearly demonstrates that xDsh depends on the CKIe residues N275 and R279 for strong binding to CKIe, and reveals that mPer1 requires an additional CKIe-like environment elsewhere on the kinase in order to bind CKIa.

Phosphorylation of the C-terminus of CKIe inhibits the binding of target proteins
Our data shows that the C-terminus of CKIe stabilizes some substrate interactions, specifically in the absence of motifs that are required for binding ( Fig. 6A and B). However, autophosphorylation of the C-terminus has also been shown to inhibit CKIe's activity towards protein targets [6,7,53,54]. We hypothesized that this effect might be partially mediated by a change in affinity for substrates. When CKIe is allowed to fully autophosphorylate, neither mPer1 nor xDsh bind to it (Fig. 7). In addition, hyperphosphorylated CKIe no longer binds to mAxin, a protein that is able to bind to both CKIa and CKIe. Hyperphosphorylation of CKIe's C-terminus thus inhibits the binding of substrate proteins, and may act as a regulatory mechanism to control phosphorylation of targets. Autophosphorylation of the C-terminus of CKIe stabilizes its interaction with the kinase domain Despite numerous attempts, we were unable to crystallize fulllength autophosphorylated CKIe, raising the possibility that the phosphorylated C-terminus does not adopt a single conformation, which is necessary to obtain a crystal. It was possible, however, that the autophosphorylated C-terminus of CKIe adopts a stable set of conformations that blocks the binding site for substrates. As an alternative approach to crystallography, we performed a set of crosslinking and mass spectrometry experiments to establish how autophosphorylation of CKIe's C-terminus interferes with substrate binding.
We first looked for evidence of a conformational change by following differences in electrophoretic mobility of purified, hyperphosphorylated CKIe upon the addition of 1-Ethyl-3-[3dimethylaminopropyl]carbodiimide Hydrochloride (EDC), a zerolength crosslinker. When two or more proteins are crosslinked by EDC, the expected product is a band of approximately their additive molecular weights. In the case of CKIe, no high molecular weight bands appear, demonstrating that CKIe does not form dimers or higher-order oligomers in solution. Instead, the single band corresponding to CKIe broadened considerably, and showed faster mobility after EDC treatment as seen by PAGE analysis (Fig. 8A, lane 4). Interestingly, unphosphorylated CKIe did not show a change in mobility after crosslinking and it continued to migrate as a single band (Fig. 8A, lane 2). The broad band that appears after crosslinking suggests that the autophosphorylated CKIe occupies multiple conformations.
We investigated where CKIe's hyperphosphorylated C-terminus binds to the kinase domain using mass spectrometry on three different trypsin-digested preparations of phosphorylated and then crosslinked protein. In order to alleviate potential problems with phosphorylation inhibiting tryptic cleavage or detection of peptides on the mass spectrometer, we treated one sample with Shrimp Alkaline Phosphatase (SAP) after crosslinking but before trypsin cleavage. SAP treatment reduced the level of phosphorylation of crosslinked CKIe as seen by radioactive labeling on the protein (data not shown).
Mass spectrometry analysis showed that the sequence PED-LDRERREHDREER is crosslinked to the kinase domain ( Fig. S1 and Fig. S2). PEDLDRERREHDREER corresponds to the 20 amino-acid stretch that was removed from CKIeDC to make CKIeDDC (Fig. 2). Two crosslinks along the PEDLDRERREHD-REER peptide position it directly along the basic groove that abuts the active site of CKIe, which was postulated to be a possible phosphate recognition region [50,[55][56][57][58][59][60] (Fig. 8B). Lysines 130 and 232 are part of the kinase domain and were crosslinked to opposite ends of the peptide (PEDLD and EHDREER, respectively). We did not identify any crosslinks in the rest of the Cterminus. This is in agreement with our PAGE results, which indicate that the crosslinked protein sample is highly heterogeneous and would have reduced abundance of unique cross-linked peptides (Fig. 8A).
We also collected spectra for two phosphorylated peptides that were not crosslinked efficiently. These peptides, MGQLRGS (p)AT(p)RALPPGPPAGAAPNR and ISASQAS(p)VPFDHLGK ( Fig. S3 and Fig. S4) are phosphorylated at previously described autoinhibitory autophosphorylation sites [6]. The two phosphorylated peptides were found even in the sample treated with SAP. Presumably dephosphorylation of these tryptic peptides was blocked due to inefficient contacts between SAP and the phosphates, suggesting that the phosphate modifications may be protected through interactions with the rest of CKIe.

Discussion
CKIe and CKIa are closely related kinases with very divergent C-termini. We show here that recombinant CKIe and CKIa interact in vitro with two known CKIe substrates similarly to the previously published in vivo interactions between endogenous proteins [23,[47][48][49]. Using recombinant proteins, we demonstrate that CKIe's C-terminus is not essential for binding to xDsh and mPer1, and show that two residues in the conserved kinase domain, N275 and R279, provide a critical interface for binding by these substrates. Additionally, we show that the C-terminus of CKIe can positively regulate substrates by stabilizing contacts with the kinase domain when dephosphorylated, whereas it can negatively regulate substrate-kinase interactions when autophosphorylated. Finally, we show that a conformational change occurs in the C-terminus of CKIe such that the C-terminus probably binds near the active site when the kinase is autophosphorylated.

Selectivity of CKIe binding for Dishevelled
The binding of xDsh to CKIe and not CKIa primarily requires the residues N275 and R279 (Fig. 9A). Mutation of these residues to either alanines or the CKIa identities, isoleucine and threonine respectively (Fig. 9B), prevented binding. Thus, the identity of these residues is essential, suggesting that Dsh directly contacts these residues. We also noted that the extent to which Dsh was phosphorylated (visualized by a change in electrophoretic mobility, Fig. 4) was different depending on the mutations that were made. The identities of the residues at positions 275 and 279 in CKIe may therefore also be important for correctly positioning Dsh for full phosphorylation. The C-terminal tail of CKIe provided some enhancement of xDsh binding and increased the level of phosphorylation by the mutant protein (Fig. 5), but it was not sufficient for binding. Strikingly, we saw full binding of xDsh to CKIa simply by changing the I283 and T287 to the CKIe identity and adding on the C-terminal tail (Fig. 9C). In the absence of the C-terminal tail, Dsh did not bind to the mutated CKIa (data not shown), which strengthens our hypothesis that the C-terminus of CKIe does contact Dsh, but is not required in for interactions with wild-type CKIe. These results allow us to fully explain the selectivity of CKIe for Dishevelled.

Selectivity of CKIe binding for Period
The binding of mPer1 to CKIe is more complex than for Dishevelled. Changing residues N275 and R279 to alanine in the context of CKIeDC did not disrupt mPer1 binding, whereas mutation of these residues to the CKIa identity in CKIeDC eliminated all binding. This result indicates that the CKIa residues at this position prevent binding by mPer1, but that the residues may not be directly involved in its binding (Fig. 9B). This is supported by the observation that changing these residues in CKIa to the CKIe identity (as well as adding the C-terminal tail) did not rescue mPer1 binding even though it completely rescued xDsh binding.
The C-terminal tail does have an important contribution to mPer1 binding since full-length CKIe N275I/R279T could fully bind mPer1 whereas CKIeDC N275I/R279T could not bind mPer1. Thus the C-terminal tail can override the repulsive effect of N275I and R279T. However, since the tail did not rescue CKIaRe I283N/T287R, it alone is not sufficient for binding mPer1 to CKIa. These results reveal that another site (or sites) on CKI must also contribute to the difference in binding between CKIe and CKIa.
One CKI-binding motif, F-X-X-X-F (where F is phenylalanine and X can be any amino acid residue), is present in mPer1 but not in xDsh [49]. Studies of the NFAT family of proteins suggest that the F-X-X-X-F motif enables stable interactions with both CKIa and CKIe [49]. Mutation of the phenylalanine residues to alanines abolishes binding by NFAT to CKIa and CKIe. mPer1 shows the same loss of binding when its F-X-X-X-F motif is mutated [37,49].
Since the F-X-X-X-F motif binds both CKIa and CKIe, it contributes to the binding affinity of mPer1 to CK1 but it is not involved in the selectivity. Thus, we conclude that another residue (or residues) on CKIe besides N275 and R279 contributes to the selectivity (Fig. 9A). This additional residue (or residues) could be amino acids that in CKIe positively enhance the binding of mPer1, or residue (or residues) that in CKIa repel mPer1. Nonetheless, our results demonstrate that the lack of the tail and changes in two residues in CKIa are sufficient to explain the lack of binding of CKIa to mPer1.

The role and structure of the C-terminal tail
The C-terminus of CKIe has been shown in previous reports to be catalytically autoinhibitory and important for regulation within signaling pathways [61][62][63]. Wnt signaling lowers the level of autophosphorylation on CKIe's C-terminus [63] though only a subset of the sites are phosphorylated in vivo, judging by polyacrylamide gel and isoelectric focusing data. Our data show that phosphorylation of the C-terminus blocks interactions between CKIe and binding partners, and it is therefore likely that some portion of the lowered phosphorylation activity seen in vivo is due to less efficient substrate binding.
The C-terminus of CKIe has been proposed to act as a pseudosubstrate for the kinase during autoinhibition [4,50,57]. The positively charged groove in front of the active site has been postulated to bind phosphopeptides that helps determine the S(p)-X-X-S specificity of the kinase family [1,4,50,[55][56][57][58][59]64]. Thus the phosphates in the tail could occupy that site and prevent substrates from binding. Our data suggest the alternative view that the phosphorylation of the tail causes the unphosphorylated but highly charged peptide, PEDLDRERREHDREER to sit across the front of the kinase's active site when the C-terminal tail is autophosphorylated, thus preventing substrate access to the active site and/or phosphate recognition groove ( Fig. 8B and 9E). Intriguingly, two previous structures of CKIdDC failed to show any density for this peptide [50,58], indicating that phosphorylation of the C-terminal tail is necessary to lock this peptide into place.
In the model illustrated in Figure 9D, after PEDLDRERREHD-REER is directed through the phosphate recognition groove, the rest of the C-terminus is led to the back of the kinase (Fig. 9D). The back face is very basic as shown in an electrostatic view (Fig. 8B), providing many sites for the phosphates in the tail to interact. The heterogeneity of the cross-linked phosphorylated CKIe and the lack of identification of prominent cross-linked tail peptides suggests that the interaction of the tail with the back side is highly variable, which also explains our inability to crystallize full-length CKIe (the same was also found with full-length CKId [50,58]). Once at the back side of the kinase domain, some phosphorylated residues on the Cterminus are buried in the basic face to stabilize the conformation (Fig. 9E). The long C-terminus could then fold against the back side of the kinase domain to occlude the binding sites for Dsh, Per and Axin. Thus, we suggest that phosphorylation of the tail has two roles, to bring the proximal C-terminus in front of the active site to prevent substrate phosphorylation and to prevent substrate binding using the distal part of the tail.
Casein Kinase family members have many targets in the cell. The results of this study designate two amino acid residues that determine the ability of two different CKI isoforms to select and bind their target proteins. Furthermore, our work on auto-and dephosphorylated CKIe introduces a means to regulate CKIe's in the presence of its targets. These two biochemical mechanisms could help maintain the strict spatial and temporal control of CKIe and CKIa that is necessary in both the circadian rhythms and Wnt pathways. Purified GST and GST-fusion proteins were bound to glutathione sepharose and bound GST-CKIe was incubated with SAP or ATP for 1 hour. Resin was incubated with 35 S-Methionine-labeled xDsh, mAxin or mPer1. 10% of the mPer1and mAxin input and 25% of the xDsh input were run. Coomassie stained gel shows the levels of GST fusion protein used for each pull-down. doi:10.1371/journal.pone.0004766.g007 Figure 8. Analysis of the binding of the C-terminal tail. (A) Full-length CKIe was incubated with either SAP or ATP prior to reaction with the EDC crosslinker. Lane 2 shows that there is no change in the apparent molecular weight of dephosphorylated CKIe. In lane 4, there is marked change in the migration of autophosphorylated, crosslinked CKIe (bracketed). Asterisk shows a high molecular weight species that may correspond to SAP-CKIe oligomers (lane 2). (B) Space-filling models of CKId are shown. The APBS plugin for PyMol (DeLano Scientific LLC) was used to establish electrostatic potential of solvent exposed atoms. Positively charged areas are shaded blue and correspond to basic regions of the protein; negatively charged regions are red, and correspond to acidic areas. The highly basic groove that has been postulated to be a phosphate recognition region is conservered across the CKI family. The two identified cross-linked residues are indicated with an X. The cartoon line on the left side diagram shows the position of the first 20 amino acids of the tail based on the crosslinking data. The dotted line on the right side shows the proposed extension of the tail onto the backside of the kinase. doi:10.1371/journal.pone.0004766.g008

Materials and Methods
Expression of GST-CKIe, GST-CKIa and CKI mutant proteins GST-CKIe was expressed in E. coli in Terrific Broth for 18 hours at 16uC. Cells were grown to OD 600 = 2 in baffled flasks and were induced with 0.5 mM IPTG. Cells were pelleted and opened by sonication, in the presence of PMSF, Leupeptin and Pepstatin A. Lysate was cleared by centrifugation, and was applied to Glutathione Sepharose 4B resin (Amersham) or Glutathione Sepharose resin (Clontech). Resin was washed three times with 20 mM Tris, pH 8, 250 mM NaCl, 5 mM DTT and the GST fusion protein was eluted with 15 mM glutathione in the same buffer. Protein was dialyzed overnight against 20 mM Tris, pH 8, 50 mM NaCl, 5 mM DTT at 4uC and was applied to 1 or 5 mL HiTrap Q sepharose column (Amersham, GE Healthcare). Protein was eluted by ion exchange using 20 mM Tris, pH 8, 800 mM NaCl, 1 mM DTT. Fractions containing the cleanest fusion protein were frozen in liquid nitrogen and stored at 280uC. For generation of CKIe without the GST-tag, ion-exchange fractions were pooled and re-bound to glutathione resin. After being washed with 20 mM Tris, pH 8, 250 mM NaCl, 5 mM DTT, the resin was incubated over night with TEV protease at room temperature, in the same buffer, with 1 mM EDTA and 1 mM DTT added. The supernatant from the overnight cleavage contained CKIe, which was either frozen in liquid nitrogen and stored at 280u C, or was further purified by gel filtration on a Sephadex 200SE column. Gel filtration fractions were pooled and concentrated on VivaSpin MWCO 10,000 filtration devices (GE Healthcare) and were frozen in liquid nitrogen and stored at 280uC.
GST-CKIa was expressed in E. coli, in Luria Broth for three hours at 18uC. Cells were grown to OD 600 = 1, and were induced with 0.5 mM IPTG. Mutants of CKIa and CKIe were expressed at 18-24uC, for 3-18 hours, depending on the construct. Cells were grown to OD 600 = 1, and were induced with 0.5 mM IPTG.
GST was expressed in E. coli from the pGex-4T1 plasmid as in [65].

GST pull-down experiments
Pull-down experiments were as performed in [65] with the following differences. For most experiments, cell pellets were thawed, sonicated and the lysate was applied to glutathione resin for 10 minutes at room temperature. Beads were washed three times with wash buffer (20 mM Tris, pH 7.5, 150 mM NaCl, 0.5% Nonidet P-40, 1 mM DTT). Binding reactions were done at 4u C for 40 minutes in wash buffer. After binding, beads were washed three times for 5-10 minutes with wash buffer prior to addition of sample buffer. Except where noted, all full-length CKIe and CKIaRe constructs were dephosphorylated using Shrimp Alkaline Phosphatase (Fermentas) for 1.5 hours at 30uC after binding to glutathione sepharose. The bound protein was then washed three times to remove traces of phosphatase. For experiments requiring hyperphosphorylated CKIe, GST-CKIe was bound to glutathione and allowed to autophosphorylate in the presence of 10 mM MgCl 2 and 2 mM ATP for 1.5 hours. Except where noted, all experiments that required full-length CKIe were run with chromatographically purified protein (including crosslinking and mass spectrometry experiments) and all other experiments were run with partially purified protein from crude E. coli lysate.

Crosslinking and mass spectrometry
Purified CKIe (2 mg/mL) was incubated with 10 mM MgCl 2 and 10 mM ATP for 1.5 hours, at 30uC. CKIe was diluted 4-fold in 100 mM potassium phosphate buffer (pH 7.5) and incubated with 100 mM EDC (Pierce) for three hours at room temperature. The crosslinked protein was dialyzed against two changes of 20 mM Tris, pH 7.5, 50 mM NaCl for 5 hours at 4uC. Protein was concentrated by evaporation to 10 mL and resuspended in 25 mM Ammonium bicarbonate (pH 7.8) in 6 M Urea. CKIe that was dephosphorylated prior to trypsinization was incubated with SAP for 1.5 hours at 30uC after dialysis; tryptic peptides that were dephosphorylated were also incubated with SAP for 1.5 hours at 30uC, immediately after tryptic digestion. Peptides were desalted on C-18 spin columns (The Net Group, Inc) and infused into a hybrid-LTQ-Orbitrap mass spectrometer via electrospray ionization (ESI). Crosslinked peptides were identified by using the open modification search tool POPITAM [67]. Posttranslational modifications that corresponded to peptides were searched to identify intramolecular crosslinks. Figure S1 Tandem mass spectrum of cross-linked peptide from dephosphorylated CKIe with a precursor mass of 2686.3876 daltons. All labeled fragment ions were identified with more than 10 ppm mass accuracy, written in red above the ion peak. Ions that are labeled in green are from peptide DVKPDNILMGLGKK and ions labeled in blue are from peptide NPEDLDRER. Solid horizontal bars mark ions whose relative abundance was greater than the scale at left. Found at: doi:10.1371/journal.pone.0004766.s001 (1.02 MB EPS) Figure S2 Tandem mass spectrum of cross-linked peptide from phosphorylated CKIe with a precursor mass of 3982.0332 daltons. All labeled fragment ions were identified with more than 10 ppm mass accuracy, written in red above the ion peak. Ions labeled in green are from peptide KMSTPIEVLCK and ions labeled in blue are from peptide FGAARNPEDLDRERREHDREER. Solid horizontal bars mark ions whose relative abundance was greater than the scale at left. Found at: doi:10.1371/journal.pone.0004766.s002 (0.73 MB EPS) Figure S3 Tandem mass spectrum of phosphorylated peptide MGQLRGSATRALPPGPPAGAAPNR with a precursor mass of 2502.7794 daltons. All labeled fragments were identified with SEQUEST, and the spectrum shown was taken directly from the SEQUEST results. Charge states are indicated by +, and phosphorylated residues are represented by lower case letters in the peptide sequence. Found at: doi:10.1371/journal.pone.0004766.s003 (0.62 MB EPS) Figure S4 Tandem mass spectrum of phosphorylated peptide ISASQASVPFDHLGK with precursor mass of 1636.5736 daltons. All labeled fragments identified with SEQUEST, and the spectrum shown was taken directly from the SEQUEST results. Charge states are indicated by +, and phosphorylated residues are represented by lower case letters in the peptide sequence. Found at: doi:10.1371/journal.pone.0004766.s004 (0.60 MB EPS)