Interdomain communication in the phosphatidylcholine regulatory enzyme, CCTα, relies on a modular αE helix

CTP:phosphocholine cytidylyltransferase (CCT), the rate-limiting enzyme in phosphatidylcholine (PC) synthesis, is an amphitropic enzyme that regulates PC homeostasis. Recent work has suggested that CCTα activation by binding to a PC-deficient membrane involves conformational transitions in a helix pair (αE) that, along with a short linker of unknown structure (J segment), bridges the catalytic domains of the CCTα dimer to the membrane-binding (M) domains. In the soluble, inactive form, the αE helices are constrained into unbroken helices by contacts with two auto-inhibitory (AI) helices from domain M. In the active, membrane-bound form, the AI helices are displaced and engage the membrane. Molecular dynamics simulations have suggested that AI displacement is associated with hinge-like bending in the middle of the αE, positioning its C terminus closer to the active site. Here, we show that CCTα activation by membrane binding is sensitive to mutations in the αE and J segments, especially within or proximal to the αE hinge. Substituting Tyr-213 within this hinge with smaller uncharged amino acids that could destabilize interactions between the αE helices increased both constitutive and lipid-dependent activities, supporting a link between αE helix bending and stimulation of CCT activity. The solvent accessibilities of Tyr-213 and Tyr-216 suggested that these tyrosines move to new partially buried environments upon membrane binding of CCT, consistent with a folded αE/J structure. These data suggest that signal transduction through the modular αE helix pair relies on shifts in its conformational ensemble that are controlled by the AI helices and their displacement upon membrane binding.

The lipid composition of cell membranes must be regulated to maintain fluidity, elasticity, curvature, surface charge, impermeability, and the optimal function of both integral and peripheral membrane proteins that depend on these physical properties. The major membrane phospholipid component in most eukaryotic cells is phosphatidylcholine (PC). 4 The geometrical and packing properties of PC molecules impart strong barrier functions and low curvature strain. In addition to its membrane structural role, PC is a major component of lung surfactant, serum lipoproteins, and bile.
PC homeostasis is controlled in part by the activity of the rate-limiting enzyme in the CDP-choline pathway, CTP:phosphocholine cytidylyltransferase (CCT) (1). CCTs catalyze the synthesis of CDP-choline from CTP and phosphocholine. CCT␣ is the dominant isoform, and CCT␤ is tissue-restricted. Whereas their disordered termini are divergent, their catalytic and membrane-binding domains are Ͼ98% similar within a species (1). CCT␣, hereafter referred to as simply CCT, is the most intensely studied form. It is activated by translocation to membrane sites that have reduced PC content, according to the prevailing theory. When unbound to membranes in vitro, the CCT catalytic efficiency is reduced by more than 2 orders of magnitude by effects on both k cat and the K m for CTP (2,3). It is the physical properties of the PC-depleted membrane that attract CCT, properties such as high surface charge and packing voids that accompany a bilayer relatively enriched in acidic and/or nonbilayer-forming lipids (4). Once CCT is membranebound and activated, PC production will ramp up and the membrane PC content will be restored, with ensuing release of CCT. As testament to its importance in the supply of PC, three human inherited diseases are caused by single nucleotide changes in PCYT1A, the gene for the CCT␣ isoform. The mutations destabilize and/or inactivate the enzyme (5) and manifest as dystrophy in retina, bone and cartilage, or adipose tissue (6 -9). The question we are trying to solve is: How does membrane binding activate this enzyme?
CCTs are homodimers with each monomer composed of an N-terminal catalytic domain (domain C) and a C-terminal regulatory domain. The latter contains a membrane-binding domain (M) of ϳ70 amino acids and a C-terminal phosphorylation region (P) of ϳ50 amino acids. The M domain of CCT has two conformations. Engagement with membranes is accompa-nied by the folding of Ͼ60 residues into a long amphipathic helix (Fig. 1A), which binds via a combination of electrostatic and hydrophobic interactions (10 -12). When unbound to membranes, this domain is composed of an auto-inhibitory (AI) helix (residues 272-293) tethered to the catalytic domain via a disordered "leash" (residues ϳ236 -271) (2,13,14). In the crystal structure of an inactive CCT dimer containing the complete catalytic domain and the M domain with a shortened leash, each AI helix docks onto the ␣E helix pair that extends from the base of the active site ( Fig. 1B) (15). The two ␣E helices and two AI helices form a stable four-helix complex that positions the C termini of the AI helices in backbone-to-backbone contact with loop L2 in the active site (Fig. 1B). Loop L2 houses a key catalytic residue, Lys-122 (16,17).
We recently explored potential mechanisms for catalytic silencing by the AI helix. MD simulations of the CCT dimer with and without the AI helices revealed that docking of the AI helices increased the frequency of an electrostatic trap for Lys-122 that captures its amino group, pre-empting contacts with substrates. The AI helices also constrained the dynamics of the ␣E helices (18). But upon removal of the AI helices in silico, the dynamics of the ␣E helices was dramatically increased. The binding of CCT to membrane vesicles in vitro, which displaces the two AI helices, was linked to an increase in the distance between the two ␣E helices, monitored by cross-linking at a site in the middle of the ␣E helix. When remodeling of the ␣E helices was prevented by a constraining disulfide bond, the enzyme, although membrane-bound, was not activated (18). These findings suggest that reorganization of the ␣E helix pair accompanies membrane binding and is required for full activation of CCT. However, in the absence of a solved structure of membrane-bound CCT, the precise conformation and mechanism of modulation of the ␣E helices during activation remain unknown.
All eukaryotic CCTs explored thus far are lipid-regulated, including CCTs from unicellular S. cerevisiae and P. falciparum. These CCTs have a highly conserved sequence corresponding to the 22-residue ␣E helix seen in the solved mammalian CCT structure obtained from crystals grown in the absence of lipid (PDB code 4MVC). The ␣E is followed by a turn at Gly-224. This turn was visible in the low-resolution structure of CCT-312 (15). The turn is followed by a structurally unresolved, conserved 10-residue linker to domain M that we refer to as the J (joining) segment (Figs. 1 and 2). In contrast, nonlipid regulated cytidylyltransferases, such as the CTs that use glycerol phosphate or ethanolamine phosphate as cytidyl acceptors (GCT and ECT) have short ␣E helices that terminate at or before the residue corresponding to Val-210 (Figs. 1E and 2). The differences between regulated and constitutive CTs suggest that the ␣E and J sequence are conserved in CCTs for an important function, such as transducing the membrane-binding event to the activation of the catalytic domain. We refer to the N-terminal ␣E segment in CCTs (residues 202-211) that resembles the GCT and ECT ␣E sequences as the ␣E N and consider the ␣E N to be part of the catalytic domain fold, as it contains residues that contribute to the active site (19 -22). We refer to residues 212-215 as the ␣E hinge (predicted to be helix-breaking) (1,23) and residues 216 -223 as ␣E C (see Fig. 2, diagram). Within the CCT ␣E hinge and ␣E C , two tyrosines (Tyr-213 and Tyr-216; mammalian numbering), two arginines (Arg-219 and  Fig. 1 images. The N-terminal 40 residues, present in the constructs shown in A-D, is disordered and not visible in any of the structures. GCT lacks this disordered region. B, CCT sol ; residues 40 -223 from PDB entry 4MVC. The AI helices (residues 276 -295; magenta) are docked onto the two ␣E helices (cyan) to create a four-helix complex. The red circle highlights the inhibitory interaction between the C terminus of the AI helix and loop L2 housing the catalytic Lys-122. C, CCT sol with AI helices removed. From a representative MD simulation with 4MVC as the starting structure, CTP in the active site, and constraints on the ␣E C to preserve helical structure (18). The image shown corresponds to the configuration shown in Fig. 5E of Ramezanpour et al. (18). The ␣E helix and the aromatic interaction between Tyr-213, Tyr-216, and Phe-124 in loop L2 is highlighted. D, CCT-236; residues 40 -215/216 with bound CDP-choline, from PDB entry 3HL4. E, GCT; residues 1-126 with bound CDP-glycerol from PDB entry 1N1D. GCT lacks the N-cap (dark blue) found in CCT, has a short loop L5, and, relevant to our discussion, has a shortened ␣E helix resolved to Ile-126 that is analogous to Ile-209 in rat CCT␣.

Modular interdomain helix regulates CCT␣
Arg-223), and a glycine (Gly-224) appear to be universal. The J segment (residues 225-234) is less strictly conserved, but the sequence conservation is much higher than the M domain.
The previously reported simulations showed that without the constraining AI helices, the two ␣E helices bent at the predicted break in each of the 10 replicates, and in six of them, the bending led to a new stable conformation featuring splayed helices, each contacting the L2 loops at the base of the active sites, in trans (18) (Fig. 1C). We hypothesize that a similar conformational change in the ␣E helices may occur upon membrane engagement via the M domain and that this is required for activation. In this work, we explored this hypothesis by (i) examining the impacts of mutations at conserved sites in the ␣E on lipid activation and (ii) probing the environment of conserved tyrosines and engineered tryptophans within the ␣E hinge, ␣E C , and J segment in various CCT forms: the soluble form (CCT sol ), the membrane-bound form (CCT mem ), and a fragment encompassing the catalytic domain and allosteric linker (CCT-236). The catalytic domain and the M and P regions have been extensively probed via mutagenesis. However, the ␣E and J segments, which constitute the only physical pathway between the membrane-bound M domain and the active site, have never been explored to determine the contribution of conserved sites to catalysis or regulation of catalysis by lipids.

CCT is nonfunctional when truncated to resemble GCT
The k cat /K m (CTP) value measured for CCT when membrane-bound (9800 s Ϫ1 M Ϫ1 ; Fig. 3A) is similar to that obtained for the constitutively active GCT (13,500 s Ϫ1 M Ϫ1 ) (20) with its short ␣E helix. We queried whether a similarly truncated CCT would be constitutively active. However, when we truncated CCT at residue 212 to create a short ␣E helix like GCT, the enzyme activity was completely obliterated (Fig. 3, A and B). The transition temperature for unfolding (T m ) for CCT-212, as monitored by Sypro Orange fluorescence, was reduced 10°C, from 55°C (full-length) to 45°C (CCT-212) ( Table 1). These data suggest a requirement for some or all of the C-terminal region for CCT catalytic function and catalytic domain stability.

Membrane activation of CCT requires domain M and the J segment
In the absence of membranes the C-terminal region is inhibitory (2,24,25). How much of the enzyme C-terminal to domain C contributes to the activation of catalysis in CCT mem ? Analysis of CCTs progressively truncated from the C terminus showed that deletion of the P region has little impact on activity (Fig. 3A), as noted previously (26 -28). But the activity was progressively compromised with increasing truncation of the M domain, especially (Please note that the JBC is not responsible for the long-term archiving and maintenance of this site or any other third party hosted site.) Middle, sequence alignment that color-codes amino acid conservation. The asterisks above the sequence indicate residues in the rat CCT␣ that were mutated to tryptophan in this study. Bottom, the corresponding sequences of the nonlipid-regulated cytidylyltransferases, GCT and ECT (N-terminal repeat), are provided.

Modular interdomain helix regulates CCT␣
with loss of the hydrophobic AI helix between residues 272 and 295 (Fig. 3A), that contributes strongly to the membrane-binding free energy (2). These decreases are due to impacts on both the k cat and the K m for CTP (Fig. S1).
Whereas CCT-212 had no measurable activity, ϳ5% of WT catalytic function is retained with a truncation at the start of domain M (CCT-236; Fig. 3A). The truncation at residue 236 did not destabilize the catalytic domain fold, as the T m for unfolding was equivalent to the full-length enzyme ( Table 1). The 10°C T m depression measured for CCT-212 indicates that the catalytic domain is adequately stabilized by the ␣E C and J segment. The partial constitutive activity of CCT-236 makes sense as it lacks the AI motif. But the low activity of CCT-236 compared with the full-length enzyme in its membrane-bound form also argues that full activation requires domain M to be present and engaged with the membrane.
Following the turn at Gly-224, the J segment comprises 10 conserved amino acids linking the ␣E C to the start of domain M. To assess the importance of this sequence to CCT activity, we substituted a glycine-rich J segment, replacing residues 227-232 with GGGSGG, or we simply deleted residues 227-232. Neither of these constructs affected the association of domain M with membrane vesicles as monitored by a lipid fluorescence quenching assay (Fig. S2); nor did they affect the unfolding transition temperature (Table 1). However, both were completely devastating to enzyme V max assayed with or without lipid (Fig. 3B). The loss of all activity accompanying the deletion of the J segment is consistent with total inactivation upon truncation at residue 228 (26). Deletion of the 6 residues from the J segment had a milder impact on the activity of CCT-236, lowering it 3-fold (Fig. 3B). Thus, both the ␣E C and the J segment improve the catalytic function of a CCT missing domain M and are essential for any activity in CCTs with domain M. These data argue that the J segment functions as more than a simple tether between domains.

Conservative site mutations in the ␣E hinge reduce membrane activation
We analyzed the impact on activity of site-specific mutations at the conserved polar residues in the ␣E hinge and ␣E C segments in the context of CCT-312. This construct contains all of the catalytic and M domains, missing only the P region. This truncation has no negative impact on binding to anionic membrane vesicles in vitro (15,23,28), as the phosphorylation of this region serves mainly to fine-tune the sensitivity to membrane negative charge (29, 30). A nonconservative replacement of Arg-211 with alanine had no impact on activity with or without lipid vesicles, and conservative substitutions at Tyr-216, Arg-219, Asn-220, and Arg-223 had no effect or a small inhibitory effect (Fig. 4A). But several mutations showed activity reductions of Ͼ2-fold: D212A, Y213F, and D214A (D214N) in the ␣E hinge; Y216A at the start of the ␣E C ; and G224A in the predicted turn at the end of the ␣E C . To determine whether these inhibitory effects were selective for membrane-induced activity, we also created the mutations in the context of CCT-236. D212A and especially Y216A and G224A were strongly inactivating in the context of both CCT-236 and CCT-312, suggesting functional roles that are independent of a membranebound M domain, but the ␣E hinge substitutions at Tyr-213 and Asp-214 had minor or negligible consequences on the activity of CCT-236 (Fig. 4B). That the hinge mutations required a membrane interaction to manifest a defect suggested that the hinge may be a critical region for signal transduction by membrane binding of domain M.

Destabilizing ␣E hinge mutations enhance CCT activity
To probe the importance of the ␣E hinge in membrane activation, we focused on the hinge residue Tyr-213, which shows different interactions in the crystal structures of fully silenced CCT-312 versus the partially active CCT-236. In the silenced CCT structure, Tyr-213 is buried in the hydrophobic interior of the four-helix complex and forms a weak H-bond to Asp-214 in trans across the dimer interface (Fig. 5A). In CCT-236, one of these H-bonding partnerships is intact, but in chain B, a bend at the hinge places the Tyr-213 hydroxyl within H-bonding distance from a backbone carbonyl on loop L2 at Error bars, 95% confidence interval. The dashed line was added to guide the eye. B, the specific activities of the indicated purified CCTs were determined using the standard assay conditions optimal for WT CCT in the absence (gray) or presence (green) of 0.2 mM PC/PG (1:1) SUVs. One unit is 1 nmol of CDPcholine formed per min. Data are means Ϯ range (error bars) of two independent determinations, except n ϭ 5 for CCT-312. GG-J, replacement with 227 GGGSGG 232 ; ⌬J, deletion of residues 227-232.

Modular interdomain helix regulates CCT␣
Phe-124 in chain A (Fig. 5B). This is the same carbonyl that participates in the backbone trap for the catalytic lysine (15). In simulations of the catalytic dimer without the AI helices, Tyr-213 is at the center of the ␣E hinge that samples different configurations on a nanosecond scale (18).
Mutations that replaced Tyr-213 with small aliphatic residues (Thr, Ala, and Gly) destabilized the soluble, silenced form, leading to an increase in the k cat /K m in the absence of lipid vesicles. This effect was seen in the context of CCT-312, which contains the M domain, and, somewhat surprisingly, CCT-236, which does not (Fig. 5C). Thus, the effect of these hinge mutations on activity can be independent of the four-helix complex. These data suggest that weakening of the interactions between the two ␣E helices may facilitate CCT catalysis. In contrast, tyrosine or phenylalanine at position 213 maintains low activity in the lipid-free state, suggesting that these bulky groups stabilize ␣E helix interactions and/or the four-helix complex in CCT-312. In keeping with a destabilizing role of the small aliphatic substitutions at Tyr-213, we found that the unfolding transitions of CCT-236 bearing Ala, Thr, or Gly at residue 213 were reduced 4 -7°C, and the T m for CCT-312(Y213A) was also reduced 7°C (Table 1). A Phe substitution in the AI helix at Trp-278, a residue vital for maintaining a web of interactions at the base of the four-helix complex (15), also lowered the T m by 7°C (Table 1; W278F). Fig. 5D shows the packing gap that is created between ␣E helices when Tyr-213 is changed to alanine.
The activity of CCT-236 is elevated Յ2-fold by lipids, and in the presence of lipid vesicles, the small aliphatic substitutions at Tyr-213 also stimulated activity. For membrane-bound CCT-312, the small aliphatic substitutions were activating compared with a hydrophobic phenylalanine substitution. However, none of the substitutions could mimic the role of tyrosine in the activation of CCT-312 by lipids. Neither the Y213F nor Y213A hinge mutation affected CCT-312 binding to anionic lipid vesicles (Fig. S3); thus, the effects were due to a block in signaling from the membrane-bound M domain. The data of Fig. 5C suggest that one role for membrane binding of domain M is to impart activating properties to Tyr-213. These data suggest that Tyr-213 evolved as the solution to effective silencing and activating functions and thus is tuned for regulation by lipids.

Tyr-213 resides in different chemical environments in the inactive, partially active, and fully active forms
MD simulations of residues 40 -223 revealed that removal of the AI helices lifted the constraints on the ␣E helices, leading to hinge bending and persistent contact between Tyr-213 and Tyr-216 with Phe-124 and Val-126 of loop L2 (18) (Fig. 1C). To probe the ␣E conformational switch in vitro, we investigated the environment of Tyr-213 and Tyr-216 in three forms of CCT: the soluble inactive form, the fully activated membrane-bound form, and partially active soluble CCT-236. Based on the solved structures of membrane-free CCT-312 and CCT-236, we expected that Tyr-213 would be very protected from solvent in the absence of lipid as it is buried deep within the four-helix bundle and that it would be

Modular interdomain helix regulates CCT␣
more exposed in CCT-236 (Fig. 5, A and B). How does its exposure in membrane-bound CCT compare with these constructs? We probed the environment of these tyrosines by iodination potential, using chloramine-T-catalyzed iodination of the native enzyme. The iodination was followed by complete trypsin digestion, separation, identification, and quantification of peptides by HPLC and MS. This provided us with information on the accessibility of all 11 tyrosines in the full-length enzyme. Each tyrosine can have 0, 1, or 2 iodo groups, depending on its solvent exposure. This produced a set for each peptide with increasing mass of 126 Da per iodo group (see Figs. S4 and S5 for some examples). The iodination potential correlated well with the accessibility of tyrosines that are buried in the core of the catalytic fold or exposed on a surface loop (Table 2 and Fig. 6A). Furthermore, iodination of Tyr-240 in the M domain was reduced 35% in the presence of membranes. These results suggested that our method can provide faithful assessment of the environment of CCT's tyrosines. The two tyrosines of the ␣E are housed in a single tryptic peptide corresponding to 212 DYDVYAR 218 ; thus, this peptide can contain 0, 1, 2, 3, or 4 iodo groups. These species were separated by HPLC and identified by their masses. Moreover, using a modified LC separation method, we separated the seven positional isomers of this peptide containing 1, 2, or 3 iodo groups, and each unique species was identified by MS/MS fragmentation (Fig. S5).
Iodination of the ␣E peptide containing Tyr-213 and Tyr-216 was semi-protected compared with tyrosines located on the periphery of the catalytic domain and more exposed than those buried in the hydrophobic core (Table 2). Iodination was enhanced by membrane binding, from 1.44 Ϯ 0.16 iodo groups/ peptide in CCT sol to 1.82 Ϯ 0.01 in CCT mem . The average iodination of the peptide was even higher in our analysis of CCT-236, 2.2 Ϯ 0.04 (Fig. 6B). Fragmentation analysis of the multiply iodinated peptides (1-3 iodo groups/peptide) provided the iodination status of each tyrosine and the Tyr-213/Tyr-216 iodination ratio (Fig. 6C). Tyr-216 iodination was invariant among the three CCT forms (1.06 Ϯ 0.02), whereas Tyr-213 iodination was only 0.4 Ϯ 0.11 iodo groups/peptide in CCT sol and increased to 0.77 Ϯ 0.01 in CCT mem. The ratio of iodination on Tyr-213/Tyr-216 increased from 0.39 Ϯ 0.11 to 0.73 Ϯ 0.01 upon membrane binding, but this ratio is less than that of CCT-236 (1.0 Ϯ 0.04).
These data suggest that displacement of the AI helices that accompanies membrane binding increases the accessibility of Tyr-213, which is not surprising. But the reduced iodination of Tyr-213 of membrane-bound CCT compared with CCT-236 suggests an environment for Tyr-213 in CCT mem that is more sequestered from solvent than in a truncated soluble CCT. The latter is expected to have a disordered ␣E C and J segment. Thus, Tyr-213 resides in a semi-protected environment in CCT mem , which is compatible with a folded population for the ␣E C in membrane-bound, active CCT.

Membrane binding promotes a new, sequestered environment for residue 216 in the ␣E
We further probed the environment of the ␣E hinge and ␣E c compared with other segments of CCT by examining the fluorescence of engineered Trp residues and their accessibility to the quencher acrylamide. We engineered Trp substitutions in the ␣E C helix and J segment in CCT-312 Trp-free backgrounds (sites shown in Fig. 2 above the sequence alignment) and also explored native Trps in the active site (Trp-151) and in the AI segment of domain M (Trp-278). The lipid-stimulated activities of the purified single Trp mutants at Y216W (␣E C ), and

Modular interdomain helix regulates CCT␣
Y225W and F234W (J segment) were similar to WT (Fig. 7A), indicating tolerance to substitution from one aromatic residue to another. However, the activities of L221W, L230W, and especially V232W were impaired, indicating intolerance of an aliphatic to aromatic switch in the ␣E C and J segments. The T m for unfolding was not altered for the Y216W, L221W, L230W, or V232W substitutions compared with the parent enzyme (W278F; T m ϭ 50.5 Ϯ 0.5°C; Table 1), and the chymotrypsin digestion patterns were also undistinguishable from the control, 5 confirming that the mutations do not destabilize the catalytic domain fold. Thus, we opted to analyze these variants, but 5 D. G. Knowles, unpublished observation.

Table 2
Iodination status of tyrosines in CCT CCT mem was prepared by incubating full-length WT CCT with 100-fold molar excess PG SLVs. CCT sol or CCT mem was incubated with oxidized sodium iodide for 2 min using a NaI/Y molar ratio of 3 or a ratio of 2 in the single case of the peptide housing Tyr-133. After complete trypsinolysis, fast-HPLC separation, and MS analysis, the fraction of each peptide sequence with 0, 1, 2, 3, or 4 iodo groups was calculated. The table reports the average number of iodo groups per peptide species. Data are means Ϯ average deviation of two or three independent iodination and MS analyses. Analyses of iodination of these peptides using different iodination ratios of NaI/Y ϭ 1, 2, or 5 provided the same conclusion with regard to whether the tyrosine was exposed or buried. Examples of the distribution of iodinated species are shown in Fig. S4. The values for DYDVYAR peptide in this table are very similar but not identical to the values in Fig. 6 because of the differing separation and analytical methods used here (fast HPLC separation and MS) versus Fig. 6

Modular interdomain helix regulates CCT␣
not the completely inactive V232W. The tryptophans engineered into the ␣E C and J segment did not impair the lipid concentration dependence for activation by small anionic lipid vesicles, suggesting no impact on domain M-membrane interaction (Fig. S6). The fluorescence spectra for these CCTs (Fig. 7B) varied in peak wavelength, reflecting the variable environments for each tryptophan. The spectra for Trp-55 in the N-terminal region of domain C was red-shifted (362 nm), reflecting its solvent exposure, and its emission peak was unaffected by lipid vesicles. Trp-278 in the AI helix was highly blue-shifted (326 -328 nm) in both CCT sol and CCT mem , reflecting its buried disposition within the four-helix complex and the lipid bilayer, respectively (11,15,31,32). The spectra for Trp-151 in the catalytic domain showed a peak at ϳ345 nm that was not blue-shifted by lipid vesicles but was reduced in intensity ϳ20%. The spectra for four

Modular interdomain helix regulates CCT␣
tryptophans in the ␣E C /J segment (Trp-221, Trp-225, Trp-230, Trp-234) were moderately blue-shifted in the soluble form (ϳ345 nm), and lipid vesicles caused additional 4 -5-nm blue shifts without significant changes in intensity. The spectral behavior of Trp-216 was the most striking. It was solventquenched in CCT sol and dramatically dequenched in CCT mem , along with a ϳ4-nm blue shift, indicating distinct environments for the two states.
Changes in solvent exposure of the single Trps were also assessed by the degree of acrylamide quenching of soluble and membrane-bound forms. Fig. 7C shows Stern-Volmer quenching constants (K SV ) for the entire set. For comparison, the K SV for Trp amino acid in solution was 25 M Ϫ1 . All sites showed reduced acrylamide quenching in membrane-bound CCT, with the exception of Trp-230. Membranes offered a high degree of protection for both Trp-234 and Trp-278, lowering the K SV by factors of ϳ2.5, suggesting that Trp-234 marks the N terminus of domain M. Acrylamide quenching of a Trp engineered into the N-terminal region (F55W) was reduced ϳ30%, but given the small sample number, this was not statistically significant. The active site Trp-151, however, showed a significant 2-fold reduction in K SV , suggesting a more secluded environment in CCT mem . In soluble CCT, the AI helices should partially protect the nonpolar ␣E c residues from solvent; thus, one might anticipate an increase in acrylamide quenching upon membrane binding. However, acrylamide quenching was slightly decreased for Trp-221 and Trp-225 and strongly reduced for Trp-216. Trp-216 showed the strongest membrane protection from acrylamide of all sites, with a 3-fold reduction in K SV (Fig. 7C). In CCT mem , the K SV constant for Trp-216 was only 5 Ϯ 0.8, equivalent to those in the M domain, suggesting similar solvent-shielded environments. The combined changes in fluorescence intensity, peak wavelength, and accessibility to solvent indicate a dramatically different environment for Trp-216 in CCT sol versus CCT mem .

Discussion
This work is the first report of the participation of residues in the ␣E and J segments in the regulation of catalysis upon membrane binding via domain M. The ␣E C and J segments may function together as an allosteric switch. Truncation and deletion mutations in ␣E C /J completely obliterated activation of CCT mem by lipids. Even nonconservative single-site mutations in this region caused drastic reductions in enzyme activity, suggesting a role beyond the simple tethering of two domains. The data here also suggest a conformation for the ␣E C /J in CCT mem that differs from previous models that portrayed it as an extended, unstructured tether between the catalytic and membrane-bound domains (e.g. see Fig. 1). The disposition of the ␣E C /J in relation to the catalytic domain and the membrane surface is explored in the companion paper (33). The emerging model presented here (Fig. 8) and bolstered by data in the companion paper proposes that the allosteric function of the ␣E c and J segment rely on their metastability. Membrane binding of the M domain provides an anchor to restrict the ensemble of conformers and thereby direct the folding of the allosteric linker into conformations productive for catalysis.

Tyr-213 has a role in both silencing and activation
The mutation scan through the ␣E directed us to focus on the consequences of mutations at the ␣E hinge. Two hinge mutations we initially tested had impacts only on CCT with an intact M domain (CCT-312 but not CCT-236), implying hinge involvement in signal transduction from the M domain. A set of substitutions for Tyr-213 provided evidence that this tyrosine has evolved as an effective solution to both silencing and activation of catalysis. In support of a silencing role, constitutive activity was elevated by mutations that change bulky Tyr-213 to smaller residues that destabilized the four-helix complex. In the solved structure (PDB code 4MVC (15)), Tyr-213 contributes to the stability of the inactivating four-helix complex by occupying the center of a hydrophobic pocket bounded by Ala-217 of the same ␣E, Val-210 of the partnering ␣E, and Phe-285, Ile-289, and Phe-289 of the AI helix. Tyr-213 can also H-bond in trans with Asp-214. That a Phe substitution for Tyr-213 did not elevate the activity of the soluble form argues that the polar contact is a weak one and that hydrophobic packing optimization is the dominant role for Tyr-213 in silencing.  -headed arrows), restricted by interactions with the AI helices within a 4-helix complex. The k cat /K m Ϸ 50 s Ϫ1 M Ϫ1 . Image adapted from PDB entry 4MVC. State II, e.g. CCT-236. Removal of the AI without membrane binding leads to a multitude of ϳisoenergetic states for the ␣E C , and the configuration most productive for catalysis (e.g. conformer a) is visited infrequently (p Ͻ 0.1). This state with a disassembled four-helix complex can also be envisioned as an intermediate to membrane binding. k cat /K m Ϸ 500 s Ϫ1 M Ϫ1 ; image from PDB entry 3HL4. The ␣E C helices were placed manually. State III, CCT mem -M domain embedding in the membrane promotes a folding pathway for the ␣E C and J segment, with the lowest energy configuration shown speculatively as a helix-turn-helix following the ␣E hinge. The configuration of the ␣E that is most productive for catalysis (conformer a) is stabilized by interactions with the J segment and with the membrane surface. Arrows, small rotations about the ␣E hinge. The k cat /K m Ϸ 10,000 s Ϫ1 M Ϫ1 ; image from PDB entries 4MVC (residues 40 -216 only), 1PEI, and 1PEH. The ␣E C and J segments were placed manually. ␣E residues Tyr-213 and Tyr-216 are shown in stick representation in all three states.

Modular interdomain helix regulates CCT␣
The latter would be disrupted by the small aliphatic substitutions (Fig. 5D).
Destabilization of inactive conformers (e.g. by membrane displacement of the AI helices or a Y213A mutation) provides the opportunity for generating new activating conformers. Without the docked AI helices, there are very few ␣E C protein contacts (see Fig. 5D), so that they appear poised for rearrangement. However, if dynamics alone were sufficient for activation, then the glycine substitution should have had the greatest desilencing effect, but glycine was a poorer substitute than Ala or Thr for both CCT sol and CCT mem activity. Thus, whereas hinge dynamics enables the ␣E C /J to adopt different environments, constrained hinge movements, limited by the membrane-embedded M domain, reduce the number of nonproductive conformers. We had previously hypothesized an ␣E-ordering role for membrane binding in the activation of CCT, based on correlations between crystallographic B factors for the ␣E helix and enzyme activity in various CT structures (19). Our new analysis of mutational effects supports an organizing role for the domain M-membrane interaction.
In support of an activating function for Tyr-213, no substitution was as effective as the native tyrosine for CCT mem . We do not know the precise role of Tyr-213 in catalysis, but the data of Fig. 5C suggest that membrane binding empowers its participation (e.g. by stabilizing a specific orientation of the Tyr-213 hydroxyl for H-bonding). Because both the crystal structure of CCT-236 and the MD simulations featured an interaction in the bent ␣E configuration between the Tyr-213 hydroxyl and a carbonyl oxygen at Phe-124 in loop L2, we had hypothesized that the membrane-promoted bending of the hinge could position Tyr-213 to capture one partner of the trap for Lys-122, the carbonyl at Phe-124 (1,18). In this way, Tyr-213 could participate in the lipid regulation of catalysis. However, that Y213A is partially active is difficult to align with this mechanism. Although its mechanism of action remains to be solved, our data suggest that the aromaticity and/or hydrophobicity of Tyr-213 is important for silencing and activation, and its H-bonding capacity is required for optimal catalytic function.

␣Ec/J mutations can modulate CCT activity independent of domain M
CCT-236 is considered to be in a "desilenced," lipid-independent state, released from the inhibitory action of the AI helix but unable to fully populate the state required for full activation. But surprisingly CCT-236 is weakly activated by lipids (Figs. 3-5), an effect that will be clarified in the companion paper (33). Thus, it is an imperfect model to use compared with CCT-312 to tease out the specific lipid-related effects from nonlipid-related effects on activity. ␣E hinge, ␣E C , and J segment mutations affected the activity of CCT-236, consistent with a role for this region in allosteric control of catalysis. The catalytic enhancement created by mutations at Tyr-213 that destabilize ␣E-␣E packing indicates that the activity inherent in the truncated CCT is also dependent on a malleable ␣E C /J conformational ensemble. However, unlike CCT-312, the activation of CCT-236 by lipids was less effective with the native tyrosine at position 213 than with alanine. One explanation for these results is that only when the M domain is membrane-bound can a hinge conformation be promoted that positions the Tyr-213 hydroxyl to participate (indirectly) in the reaction.

A guided conformational ensemble model elucidates the impacts of ␣E hinge mutations
Collectively, these results suggest a model where the allosteric linker samples multiple conformers when the AI is removed, but not all conformers are productive for catalysis (Fig. 8). In CCT-236, the conformation of the ␣E C /J that is optimal for full activation of the enzyme (hypothetical state a in Fig.  8) is a rare occurrence. Instead, it likely samples many iso-energetic, low-activity conformations with low energy barriers between them, and without a firm domain M anchor onto the membrane, the barrier for acquisition of the fully active conformation is large. In the solved structure of CCT-236, the ␣E C and J segment are present, but not visible, indicating their disorder. Substituting Gly, Thr, or Ala for Tyr-213 slightly increases the fraction of the conformational ensemble with high activity (e.g. state a) at the expense of the unbent ␣E helix conformation (Fig.  8). For full-length CCT (or CCT-312) bound to the membrane via domain M, the conformational ensemble for the ␣E and J segment would be restricted, with a high proportion of fully active conformers (Fig. 8, state III). The folded M domain helix would provide an anchor point at the C terminus of the ␣E C /J. The sequestered, compact environment for the ␣E hinge and the J segment suggested by the data in this paper and the companion paper (33) supports a constrained conformational ensemble in CCT mem .

The ␣E helix residues Tyr-213 and Tyr-216 sample a solvent-protected environment in CCT mem
Upon membrane binding, the AI helices dissociate from the ␣E helices, and thus the tyrosines of the ␣E hinge should become more solvent-accessible. However, iodination and fluorescence analysis showed that the new environment for these tyrosines in CCT mem is not compatible with full solvent exposure. Previous work showing protection of Tyr-213 from chymotrypsin cleavage in CCT mem (13) is in agreement with the protected status observed in the iodination experiments. The different iodination levels for Tyr-213 suggested different environments in the silenced, desilenced, and activated CCT forms, and this supports the modular conformational ensemble model for the allosteric linker. The Y216W substitution we made to probe fluorescence is nondisrupting, as it has no effect on the tertiary structure, membrane binding, or enzyme activity. Trp-216 was much less accessible to acrylamide in CCT mem versus CCT sol . The large increase in fluorescence intensity of Trp-216 upon membrane introduction also supports sequestration from solvent. However, Tyr-216 retained similar exposure to the I ϩ species. Because the vesicle surface in the iodination reactions possessed very high negative charge (100% PG), the I ϩ species would concentrate near the surface (Ͻ20 Å), and this is a potential weakness of the iodination analysis. Thus, the degree of protection indicated by the iodination levels for the ␣E tyrosines in CCT mem could in fact be underestimated.
In CCT sol , Trp-216 may be quenched by surrounding amino acids and is dequenched upon membrane binding. In the solved structure (PDB code 4MVC), Tyr-216 is situated within

Modular interdomain helix regulates CCT␣
quenching distance of Asn-220 and Tyr-213 in the ␣E and Met-292 in the AI. The dequenching and blue shift as well as the strong protection from acrylamide quenching observed for Trp-216 in CCT mem is consistent with remodeling of the ␣E so that Trp-216 is in a chemically different and sequestered environment. That Tyr-216 can be substituted with Trp or Phe without functional consequences, but not with alanine (or cysteine) 6 also argues that the activated conformation places this residue in a nonpolar environment, and this is essential for activity.
An aromatic cluster between the two tyrosines in the ␣E and Phe-124 in L2 was observed as a dominant interaction in silico in all ten 1000-ns simulations lacking the AI helix (18). However, the aromatic clustering may not reflect the dominant conformers in CCT mem in vitro. The simulated starting structures lacked the J segment, domain M, and a bilayer, and in their absence, the intrinsic drive to bury nonpolar residues resulted in the aromatic cluster. In vitro and in vivo, a bent helix may be stabilized by interactions other than or in addition to the aromatic cluster. Possibilities, explored in the companion paper (33), include interactions with the J segment following the ␣E and/or the membrane surface.
Our analysis of this previously unexplored, interdomain linker elucidates the ␣E C /J segment as a signal transducer that has been evolutionarily tuned to elicit a response to membrane binding. The response includes hinge-induced reorientations, changes in solvent exposure, and new interactions forged by highly conserved residues. Data in the subsequent paper (33) suggest that the ␣E C and J segments are close to the membrane surface in CCT mem but are not membrane-penetrant and that they form a compact structure that brings the active site close to the membrane surface.

Construction, expression, and purification of CCTs
The untagged full-length rat CCT␣ was expressed in Trichoplusia ni cells using baculovirus infection and purified as described (12). The untagged catalytic fragment (CCT␣-236) was also expressed and purified from T. ni cells (24). Construction of the N-terminally His-tagged, C-terminal truncations of rat CCT␣ were also described previously (15,34). The truncated variants encoded CCTs terminating at residues 312, 304, 293, 271, 255, and 236. In addition to these, we used QuikChange mutagenesis (Stratagene) and HisCCT-236 as the template to construct HisCCT-212 by engineering a stop codon after codon 212.
We found during the progress of this work that engineering a C-terminal His tag increased the purity of the CCT-312 preparations fractionated on nickel-agarose. Thus, some of the constructs utilized a CCT-312 (His) variant (described previously (18)) as template for further mutagenesis. For spectroscopy in the presence of lipid vesicles, we found that deleting the N-terminal NLS sequence ( 12 RKRRK 16 ) eliminated turbidity of the samples caused by NLS-mediated vesicle aggregation (18,34). Thus, all constructs to be used in fluorescence analysis utilized CCT-312-His ⌬12-16 as template for further mutagenesis. Construct DNA sequences were analyzed by Eurofins Genomics, which confirmed the presence of the engineered mutation as well as the absence of any nonsolicited mutations. His tags for purification at either the N or C termini were provided by pET14b or pET24a plasmids, respectively. Tagged CCTs contained either the N-terminal, non-CCT sequence (MGSSH 6 SSGLVPRGSH) or a C-terminal sequence following residue 312 ( 313 LEHHHHHH 320 ). Neither was cleaved prior to analysis. A list of the constructs, parent plasmids, and cell fractions for purification is provided in Table S1.
All constructs were expressed in E. coli Rosetta DE-3 cells (15). For cells harboring pET14b plasmids, the lysogeny broth growth medium contained 100 g/ml ampicillin and 34 g/ml chloramphenicol. For pET24a plasmids, the medium contained 50 g/ml kanamycin and 34 g/ml chloramphenicol. 400-ml cultures were agitated (225 rpm) at 37°C until the optical density at 600 nm reached 0.45-0.55. CCT expression was induced with isopropyl ␤-D-1-thiogalactopyranoside (1 mM), and cultures were agitated for an additional 3 h at 37°C (for CCT-312-His) or 20 -21°C (for His-CCT-236). After expression, cells were harvested by centrifugation at 4500 ϫ g for 15 min and frozen at Ϫ80°C. A small-scale analysis of CCT distribution between 20,000 ϫ g supernatant and pellet fractions was done using a modification of a previously published procedure (5) to assess the appropriate fraction for purification.
Depending on partitioning behavior, the CCTs were purified from the soluble fraction, the particulate fraction, or the wholecell lysate (see Table S1 for construct-specific information). Cell lysis and purification from the 20,000 ϫ g supernatant by Ni-agarose chromatography was carried out as described (34,35). To purify CCT from the 20,000 ϫ g pellet, we washed the pellet twice using gentle sonication with lysis buffer (20 mM NaP i , 150 mM NaCl, 1% Triton X-100, 2 mM DTT, protease inhibitors, pH 7.4), and the washed pellet was solubilized in 10 ml of denaturation buffer (6 M guanidinium-HCl, 500 mM NaCl, 50 mM NaP i , pH 8, 25 mM imidazole, 1 mM DTT) by intermittent vortexing at 20 -21°C for 1 h. After removing nonsolubilized material by centrifugation for 20 min at 13,000 ϫ g at 20 -21°C, the supernatant was applied to a 3-5-ml Ni-NTAagarose column at 20 -21°C. The CCT was refolded on-col-6 S. G. Taneva, unpublished observation.
Some of the constructs were distributed between the soluble and particulate fractions. We found that CCT purification directly from the whole-cell lysate provided better yield without loss of purity for these CCTs. Cell pellets from 200-ml cultures were resuspended in 5 ml of lysis buffer with sonication on ice. The lysate (5 ml) was added to 50 ml of 50 mM NaP i , pH 8, 6 M guanidinium HCl, 500 mM NaCl, 25 mM imidazole, 1% Triton X-100, 1 mM DTT, gently mixed for 1 h at 20-21°C, and centrifuged at 13,000 ϫ g for 30 min. The supernatant was applied to a Ni-NTA-agarose column at 20 -21°C, and the CCT was isolated as described above. The concentrations of the pure enzymes were determined by the method of Bradford (36) using BSA as a standard.

CCT enzyme activity
The specific activity was determined as described previously (5). For the standard assay, optimal for WT CCT, the concentrations of CTP and phosphocholine in the assay were 8 and 1.6 mM, respectively. Kinetic constants were determined by varying the CTP concentration from 0 to 16 mM using a fixed (2 mM) concentration of phosphocholine. The data were evaluated with GraphPad Prism, and the V max and K m values were obtained using this software. r 2 values for the fits to the data were Ͼ0.96. Sonicated lipid vesicles were prepared as described previously (37), except the suspensions were sonicated for 15 min on 50% output and were centrifuged for 4 min at 13,000 ϫ g to remove titanium debris introduced by the sonicator probe. The vesicles were stored under argon at 4°C and used within ϳ2 days.

Thermal denaturation using SYPRO Orange
The T m for unfolding was measured using the hydrophobic dye SYPRO Orange (Invitrogen, Molecular Probes), as described (5).

Iodination of CCTs
Iodination of untagged CCT-236 and full-length CCT-367 utilized chloramine-T-derivatized Iodination Beads. A single bead washed in 10 mM Tris, pH 6.5, was placed into a microcentrifuge tube together with 100 l of 3.5 mM NaI, 10 mM Tris, pH 6.5, and incubated for 5 min. The Iodination Bead was removed, and aliquots from the reaction solution containing the oxidized I ϩ were added to 0.1-ml samples containing 20 M protein in 10 mM Tris, pH 7.4, 100 mM NaCl, 0.2 mM DTT to give a final molar ratio of NaI to tyrosine of 0, 1, 2, 3, and 5. The resulting mixtures were incubated for 2 min (or 4 min for NaI/ tyrosine ϭ 5) at 20 -21°C, and the reaction was stopped by the addition of DTT to a final concentration of 3 mM. Aliquots were removed to assess the completeness of iodination based on a detectable mass shift via SDS-PAGE analysis. The remainder was digested with trypsin for 1 h at 37°C using a trypsin/CCT mass ratio of 0.2. The digestion was quenched with 2 mM phenylmethylsulfonyl fluoride.
Iodination of membrane-bound CCT was performed as above in the presence of a 100-fold molar excess of PG sucroseloaded vesicles (SLVs). SLVs were prepared by extrusion of multilamellar vesicles at 20 -21°C using polycarbonate membranes with a 100-nm pore diameter (38). The protein-SLV mixtures were incubated for 5 min at 20 -21°C. After iodination and trypsin digestion as described above, the SLVs were sedimented at 100,000 ϫ g for 1 h at 20°C to remove lipid. The supernatant was removed and used for MS analysis.
For MS analysis, we adjusted the pH of the protein digest to pH Յ4.0 with TFA (final concentration of 0.1%). We desalted 25-l samples using 0.6 l of ZipTip C18 resin according to the manufacturer's protocol (Millipore). Peptides were eluted from the ZipTip with 10 l of 50% acetonitrile, 0.1% TFA.

HPLC-MS analysis of iodinated CCT peptides
Iodinated and desalted protein digests were analyzed with a Bruker maXis Impact Quadrupole Time-of-Flight LC/MS System. The system consists of an Agilent 1200 HPLC and a Bruker maXis Impact Ultra-High Resolution tandem TOF mass spectrometer. We used positive electrospray ionization for MS and MS/MS experiments and the following parameters: gas temperature at 180°C, gas flow rate at 9 liters/min, nebulizer at 4 bars, capillary voltage at 3800 or 4200 V, analyte mass range of 50 -1500 Da, and mass-dependent collision energies ranging from 10 to 40 eV. Data acquisition and processing utilized Compass 1.5 and Biotools 3.2 SR3 software provided by the Bruker system.
We used two different HPLC separation methods, which are briefly summarized here. Details of these methods are described in the supporting material. 1) "Fast" LC/MS was used to determine the extent of iodination of all of the tyrosinecontaining peptides in CCT sol and CCT mem . We obtained the peak intensities for the tyrosine-containing peptides that matched the theoretical masses in a list of CCT peptides containing 0, 1, or 2 iodines (or 3 or 4 iodines in the case of peptides containing 2 tyrosines). The extent of tyrosine iodination was determined as the ratio of the peak intensity of the modified peptide to the sum of the intensities of modified and unmodified peptides. The data from this method appear in Table 2 and Fig. S4. 2) "Slow" LC/MS was used to resolve the isomers of the ␣E heptapeptides containing all of the iodinated species of Tyr-213 and Tyr-216, coupled to MS/MS fragmentation to identify the iodinated tyrosines in each resolved species. The relative abundance of the various iodinated species of this ␣E peptide was assessed from extracted ion chromatograms that segregated the m/z species matched to each iodinated and unmodified peptide away from all other CCT peptides (Fig. S5). Quantitative analysis utilized integrated peak areas from these chromatograms. The two mono-iodinated species ( 212 DY(I) DVYAR 218 and 212 DYDVY(I)AR 218 ) did not separate by the first two methods. Their relative abundance was obtained from fragmentation analysis. The data using the latter methods appear in Fig. 6.

Modular interdomain helix regulates CCT␣ Tryptophan fluorescence and acrylamide quenching
For fluorescence acquisition, 3 M CCT was incubated in buffer (10 mM Tris, pH 7.4, 150 mM NaCl, 2 mM DTT) or 450 M vesicles in buffer for 5 min at 20 -21°C. Sonicated POPC/egg PG (1:1) vesicles (2 mM lipid) were prepared as described above. Sample volume was 0.3 ml. Fluorescence scans were measured at 20°C on a Cary Eclipse fluorescence spectrophotometer. The excitation wavelength was 295 nm (to minimize tyrosine excitation) with a 5-nm excitation slit, and the emission was scanned twice from 305 to 420 nm with a 10-nm emission slit and a scan rate of 120 nm/min. Acrylamide (6 mol) was added to the same sample with pipette-mixing to a final concentration of 20 mM, and the fluorescence scan was repeated. This cycle was continued in 20 mM increments to a final acrylamide concentration of 120 mM. Lipid and spent dialysis buffer background spectra were subtracted to generate corrected emission spectra. The fluorescence intensity maximum for each spectrum was recorded at the peak emission wavelength average for each acrylamide quenching set, and these values were used to generate linear plots of unquenched/quenched fluorescence versus concentration of acrylamide for each single-Trp CCT construct. The slopes of these plots were reported as the Stern-Volmer quenching constants (K SV ) in the absence and presence of vesicles, where F o represents unquenched fluorescence, F q is quenched fluorescence, and [Q] is the molar concentration of acrylamide.