Interdomain and Membrane Interactions of CTP:Phosphocholine Cytidylyltransferase Revealed via Limited Proteolysis and Mass Spectrometry* □ S

CTP:phosphocholine cytidylyltransferase (CCT) is a multi-domain enzyme that regulates phosphatidylcholine synthesis. It converts to an active form upon binding cell membranes, and interdomain dissociations have been hy-pothesized to accompany this process. To identify these interdomain and membrane interactions, the tertiary structures of three forms of CCT (cid:1) were probed by moni-toring accessibility to proteases. Time-limited digestion with chymotrypsin or arginine C of soluble CCT (cid:1) (CCT sol ), phospholipid vesicle-bound CCT (CCT mem ), and a soluble constitutively active CCT truncated at amino acid 236 generated complex mixtures of peptides that were resolved and identified by gel electrophoresis/im-munoblotting and by matrix-assisted laser desorption/ ionization-mass spectrometry, with or without coupling to capillary liquid chromatography. Identification of cleavage sites enabled assembly of peptide bond accessibility maps for each CCT form. Our results reveal a (cid:1) 80-residue core within the catalytic domain (domain C) as the most inaccessible region in all three forms and the C-terminal phosphorylation domain as the most accessible. Membrane binding has little effect on the protease accessibility of these domains. To map the protease sites onto the catalytic domain, its three-dimensional (cid:4) was at Mass standards (MS-CAL1; were used for external calibration daily. Two mass spectrometers were used in this study. A linear MALDI-time of flight-MS (Perseptive Biosystems Voy-ager-DE, Framingham, MA) was used exclusively to analyze dried droplet preparations. Mass spectra of CCT fragments (cid:2) 10 kDa were collected with an accelerating voltage of 25 kV, a grid voltage of 88.0%, a guide wire voltage of 0.030%, and a delay time of 350 ns, and the low mass gate was set at 10 kDa. Mass spectra of CCT fragments (cid:5) 10 kDa were collected with an accelerating voltage of 20 kV, a grid voltage of 93.7%, a guide wire voltage of 0.010%, and a delay time of 50 ns, and the low mass gate was set at 800 Da. Mass spectra for both the dried droplet and CapLC sample preparations were collected on a MALDI -time of flight-MS equipped with delayed ion extraction and operated in linear or reflectron mode (MALDI-LR; Waters Technologies). For all of the samples the source voltage was 15 kV, the detector voltage was 1.8 kV, and the delay time was 500 ns. To accommodate the wide mass range of peptides from 1 to 42 kDa for full-length CCTs and from 1 to 27 kDa for CCT236, each sample was analyzed using separate pulse and/or reflec- tron voltage conditions to optimize the sensitivity and resolution for peptides detected in each mass range, typically 1–7, 2–15, and 10–50 kDa. The reflectron operation mode offers higher resolution but lower sensitivity; thus it was used only for peptide ions with m / z (cid:5) 7000. In addition each sample was prepared in three modes: dried droplet with CHCA matrix, dried droplet with MALDI matrix sinapinic acid (3,5- dimethoxy-4-hydroxycinnamic acid), and CapLC fractionation depos-ited onto a

region showed that it is a continuous amphipathic ␣-helix in a membrane-mimetic environment (10). By comparing the secondary structure content of soluble and membrane-bound forms of CCT derived from deconvoluted circular dichroism spectra, Taneva et al. (11) showed that this domain is a mixture of conformers in the soluble form of the enzyme and is transformed into an ␣-helix in the presence of membranes containing anionic lipids. The structure of the N-terminal domain (N) is by contrast a black box. It bears no significant sequence homology with any other sequence in the NCBI protein structure database. The C-terminal domain (domain P), which is known to house up to 16 phospho-serine sites (12), is also structurally uncharacterized, as are the linkers between domains C and M and between domains M and P.
CCT␣ is a noncovalently associated homodimer of a 367amino acid subunit. The dimer interface involves both domains N and C (13). Binding to anionic lipid membranes causes a repositioning of the dimer interface (13), but whether this change is instrumental in the activation process is not known.
Previous analysis by limited proteolysis with chymotrypsin and mapping the major SDS-PAGE-resolved fragments by epitope-specific antibodies gave rise to the proposal that CCT consists of a protease-resistant compact head (up to residue 225) and a protease-sensitive, flexible, exposed tail (14). This approach suffered the limitation of inaccurate estimation of the specific cleavage sites. Moreover, the effects of membrane binding on the tertiary structure assessed by protease accessibility had not been examined. In this study we compare the protease accessibility of three forms of CCT: the soluble form, the membrane-bound form, and the form truncated at the end of domain C. The identification by MALDI-MS of Ͼ100 peptide fragments generated by time-limited digestion with chymotrypsin or Arg-C resolved the uncertainty associated with fragment identification based only on electrophoretic mobility and epitope-specific antibody reactivity and enabled the construction of a peptide bond accessibility map for each CCT form.
According to a prominent model for CCT regulation (15), the soluble form of CCT is inhibited by interdomain interactions between domain M and the catalytic domain (Fig. 1B). Membrane binding relieves inhibition by dissociation of this domain interaction. That domain M is autoinhibitory is supported by the findings that deletion of domains M and P, but not P alone, results in a constitutively active enzyme (15,16). Our data support this general model and provide a clue as to the origins of the interdomain regulatory contacts. The accessibility of the interdomain contact points involved in the regulatory confor-mational switch should be low in the soluble form and high in the membrane-bound form. Based on this premise the present work identifies domain N rather than domain C as the best candidate for a site of contact with the C-terminal region of domain M.

EXPERIMENTAL PROCEDURES
Protein Purification and Preparation of CCT mem -Recombinant rat CCT ␣-isoform was purified from a baculovirus expression system using the method of Friesen et al. (15) as modified by Davies et al. (17). CCT236 was constructed, expressed, and purified as described (11). To prepare CCT mem , sonicated egg phosphatidylglycerol (PG) vesicles were prepared (18) and mixed at Ն100-fold molar excess over CCT 5 min prior to the initiation of proteolysis reactions.
Proteolysis and Gel Electrophoresis/Immunoblot Analysis-The chymotrypsin digestion method has been described previously (14,10). Briefly, the reactions were at 37°C and contained ϳ12 M CCT (without or with Ն1.2 mM PG vesicles) and a 1:250 weight ratio of chymotrypsin to CCT in 10 mM Tris, pH 7.4, 2 mM dithiothreitol, 0.15 M NaCl, 1 mM EDTA. Aliquots of the reaction mixture were transferred after various times to vials containing the protease quencher, phenylmethylsulfonyl fluoride (final concentration, 10 mM). Arg-C reactions contained ϳ15 M CCT (without or with Ն1.2 mM PG vesicles) and a 1:175 weight ratio of Arg-C/CCT in 50 mM Tris, pH 7.8, 8 mM CaCl 2 , 0.5 mM EDTA, 5 mM dithiothreitol, 120 mM NaCl. The aliquots were transferred at various times to tubes containing the quencher, EGTA (final concentration, 20 mM). A portion of each quenched sample was aliquoted for MALDI-MS analyses, and the remainder was analyzed by SDS-PAGE. The samples were boiled for 3 min in Laemmli sample buffer containing 4% SDS and 2% ␤-mercaptoethanol (19). Protein fragments were separated on 12% polyacrylamide gels (19). The fragments were visualized by silver staining (20) or by electrophoretic transfer to polyvinylidene difluoride and reaction with antibody against amino acids 1-15 (anti-N), 164 -176 (anti-cat), or 256 -288 (anti-M) as described (13).
Delipidation of CCT mem Samples Prior to MS Analysis-The presence of PG vesicles in the CCT mem samples inhibited the production of ionized peptides by interfering with the requisite MALDI matrix/peptide co-crystallization and desorption processes. To alleviate this interference, we extracted the phospholipids from the CCT mem samples prior to MALDI analysis by the dried droplet method. While vortexing, a 20-l sample was injected into 0.25 ml of ethanol/diethyl ether/H 2 O (74.6/24.9/0.6) (21) at Ϫ20°C. The vortexing continued for 3 min at Ϫ20°C. After sedimentation at 13,000 rpm for 3 min at 0°C, a minute white pellet was visible. The solvent in the supernatant was removed, and the pellet was dried under N 2 and redissolved in 10 mM Tris, pH 7.4, 1 mM EDTA, 2 mM dithiothreitol, 0.15 M NaCl. This protocol removed 98% of the phospholipid as assessed by tracking radiolabeled tracer phospholipid but retained each of the CCT fragments with masses Ͼ6000 Da, as assessed by SDS-PAGE/silver staining of the sample before and after extraction.
Capillary Liquid Chromatography-Capillary liquid chromatographic separation of the peptides was performed using a CapLC System (Waters Technologies, Milford, MA) with a Symmetry 300 C 18 , 5-m packing diameter, 0.32-mm-inner diameter ϫ 150-mm-long column. Proteolytic CCT digests were analyzed by CapLC without prior sample delipidation. For the CapLC separation, we used an injection volume of 5.0 l and a flow rate of 5.0 l/min. A gradient of solution A (0.1% trifluoroacetic acid in acetonitrile) and solution B (H 2 O) was created as follows: 10% A and 90% B at 0 min; 43% A and 57% B at 80 min; 60% A and 40% B at 100 min. No peptides were eluted with additional eluent. The flow exiting the CapLC column was directed into a device (LC-MALDIprep module; Waters Technologies) that automatically concentrates and deposits the eluent onto a MALDI target (22). In this device eluent was directed through a small orifice where it was aerosolized in a 40°C chamber and delivered by a 15-p.s.i. N 2 flow onto the MALDI target plate precoated with the matrix ␣-cyano-4-hydroxycinnamic acid (CHCA). The instrument was synchronized with the CapLC to deposit and concentrate 25 l of the eluent onto discrete ϳ1-mm spots on the MALDI plate at 5-min intervals, creating an archive of the LC-separated peptides onto different positions on the MALDI target.
Mass Spectrometry-Crude mixtures of the peptides were prepared for MALDI analysis without prior CapLC separation, using the dried droplet method. 1.0 l of crude digest was mixed with 1.0 l of sinapinic acid matrix (10 mg/ml in acetonitrile with 0.25% trifluoroacetic acid in distilled deionized water, 60:40) or 1.0 l of CHCA (10 mg/ml in 0.1% trifluoroacetic acid in acetonitrile:methanol, 50:50) in a microcentrifuge tube. An aliquot of 1.0 l was delivered to the target plate and dried at room temperature. Mass standards (MS-CAL1; Sigma) were used for external calibration daily. Two mass spectrometers were used in this study. A linear MALDI-time of flight-MS (Perseptive Biosystems Voyager-DE, Framingham, MA) was used exclusively to analyze dried droplet preparations. Mass spectra of CCT fragments Ͼ10 kDa were collected with an accelerating voltage of 25 kV, a grid voltage of 88.0%, a guide wire voltage of 0.030%, and a delay time of 350 ns, and the low mass gate was set at 10 kDa. Mass spectra of CCT fragments Ͻ10 kDa were collected with an accelerating voltage of 20 kV, a grid voltage of 93.7%, a guide wire voltage of 0.010%, and a delay time of 50 ns, and the low mass gate was set at 800 Da. Mass spectra for both the dried droplet and CapLC sample preparations were collected on a MALDI -time of flight-MS equipped with delayed ion extraction and operated in linear or reflectron mode (MALDI-LR; Waters Technologies). For all of the samples the source voltage was 15 kV, the detector voltage was 1.8 kV, and the delay time was 500 ns. To accommodate the wide mass range of peptides from 1 to 42 kDa for full-length CCTs and from 1 to 27 kDa for CCT236, each sample was analyzed using separate pulse and/or reflectron voltage conditions to optimize the sensitivity and resolution for peptides detected in each mass range, typically 1-7, 2-15, and 10 -50 kDa. The reflectron operation mode offers higher resolution but lower sensitivity; thus it was used only for peptide ions with m/z Ͻ7000. In addition each sample was prepared in three modes: dried droplet with CHCA matrix, dried droplet with MALDI matrix sinapinic acid (3,5dimethoxy-4-hydroxycinnamic acid), and CapLC fractionation deposited onto a target plate precoated in CHCA matrix. For example, 72 mass spectra were acquired for just one CCT sol digest using the CapLC preparation method. Each mass spectrum was the sum of at least 200 laser shots.
Peptides were identified by matching the measured monoisotopic or average mass to theoretical monoisotopic or average masses generated via in silico digest using a web-based program offered by the University of California at San Francisco, MS Digest (prospector.ucsf.edu/ucsf-html4.0/msdigest.html), taking into account the N-terminal acetylation of CCT and the phosphorylation of domain P on serines.
The heterogeneity of CCT phosphorylation presented a challenge to detect all of the phosphopeptides present. We explored the efficacy of four matrices (CHCA, sinapinic acid, 2,5-dihydroxybenzoic acid, and 2,4,5-trihydroxyacetophenone) prepared in different solvent blends and ammonium acetate additives. The highest phosphopeptide ion signalto-noise ratio was obtained using sinapinic acid. To verify phosphopeptide assignment, some of the crude digest mixtures were treated with protein phosphatase 1␣ catalytic subunit (PP1␣). The digested CCT (8 M) was reacted for 5 min at 30°C with 0.2 units PP1␣ (1 unit hydrolyzes 0.1-1 nmol PO 4 /min) in the presence of 0.2 mM MnCl 2 and 5 l of reaction volume. The reaction was stopped with K 2 HPO 4 , pH 7.7 (final concentration, 20 mM). This procedure resulted in a shift of phosphopeptide peaks by units of Ϫ80 Da because of dephosphorylation.
Construction of Peptide Bond Accessibility Maps (PBAMs)-For each form of CCT, we converted the collection of peptides identified by mass into a map that facilitates rapid visualization of peptide bond accessibility to proteases. The data used to build the PBAMs was derived primarily from the low mass fragments because there were a limited number of identified high mass fragments and because of the higher mass accuracy of the MALDI-MS in the low mass region. However, the cleavage sites obtained from low mass peptides were cross-checked for concurrence in high mass fragments. To build each PBAM, the starting and ending residues of all of the peptides were tabulated according to the time they were detected. To generate a chymotrypsin PBAM, we recorded each chymotrypsin site with Ͼ50% theoretical probability of cleavage based on PeptideCutter (www.expasy.org/tools/peptidecutter), an algorithm based on the work of Keil (23). Peptide bonds uncleaved within 5 min were colored black. Other sites were color-coded based on the time at which one or more cuts at that site was first detected. To build an Arg-C PBAM all arginines except the one followed by Pro were color-coded.
Modeling of CCT-The amino acid sequence of rat CCT␣ was submitted to the Polish fold recognition meta-server (24). This process identified GCT (Protein Data Bank code 1COZ) as the closest match. Once the pair-wise alignment between the CCT and GCT sequences was obtained, the alignment was submitted to the MODELLER program (25) to generate homology models of CCT. The default parameters for MODELLER were used, and in addition, the "loop-modeling" option was enabled. The models generated were refined by molecular dynamics with simulated annealing (a functionality in MODELLER) to improve the quality of the predicted structure. The models were also verified for favorable geometrical and stereochemical properties using Verify3D (26) and PROCHECK (27). The root mean square deviation between the CCT models and the GCT template were calculated. The main chain atoms in the best model exhibited an average root mean square deviation of 0.75 Å from the backbone of the template from which they were built. PROCHECK revealed that 94% of the residues occupied the most favored region of a Ramachandran plot, and only one residue, Ser 174 in the extra CCT loop, occupied the disallowed zone of the Ramachandran plot. The positioning of the side chains on this model was further refined using the software scwrl (28). For molecular visualizations we used the graphics software GRASP2 (trantor.bioc. columbia.edu/grasp2/) and PyMol (pymol.sourceforge.net/).

Comparison of the Chymotrypsin Digestion Pattern of CCT sol ,
CCT mem , and CCT236 by Electrophoresis-CCT sol refers to fulllength CCT in buffered saline, pH 7.4. CCT236 refers to a truncated form encompassing amino acids 1-236, also in buffered saline. CCT mem was formed from CCT sol by incubation with sonicated PG vesicles for 5 min, which is in excess of the time required for full engagement of the vesicles (11). CCT bound to PG vesicles is representative of the membrane-activated form. The enzymatic activity and chymotrypsin digestion pattern of CCT bound to PG vesicles is very similar to those obtained with CCT bound to other anionic lipid vesicles such as PC/oleic acid or lysoPC/PG (11). The three forms of CCT were digested in parallel, and the fragments were separated by electrophoresis. Each distinct band was assigned a number in order of increasing migration rate. For CCT sol, shown in the left-hand lanes of Fig. 2 (A-C), bands 1-11 reacted with antibodies against the N terminus and the catalytic domain (anti-N and anti-cat). This indicates that these fragments arise from progressive digestion of CCT from the C terminus. Bands 7-11 did not react with antibody directed against amino acids Val 256 -Ser 288 in the core of domain M (anti-M; not shown). Bands 7-11 were also generated upon digestion of CCT236 (Fig. 2, A-C, right-hand panels) and reacted with both anti-N and anti-cat antibodies. Thus bands 7-11 are produced by cleavage in domain C, proceeding from the C terminus of that domain. Bands 13-17 migrating ahead of the 21-kDa marker were not produced upon digestion of CCT236 and did not react with either the anti-N or anti-cat domain antibody (Fig. 2, B and C), indicating that they derive from the C-terminal domains, M and/or P. In agreement with previous antibody mapping of chymotrypsin fragments (29,30), bands 13-17 reacted with anti-M (Fig. 2D, lanes 2 and 3). Previous work also showed that bands in the 11-17-kDa range react with a lipid photolabel (29). No fragments of CCT236 reacted with anti-M (Fig. 2D, lane 1). Upon dephosphorylation of the digests of both CCT forms with PP1␣, the electrophoretic mobility of bands [13][14][15] increased, indicating that these fragments contain domain P (Fig. 2D, lanes 4 and 5). Dephosphorylation did not affect the mobility of band 17, selective to CCT mem , indicating that it does not contain domain P. Together these data suggest the following assignments: bands 1-4 span the N terminus to cleavage sites in domain P, bands 5-6b span the N terminus to cleavage sites in domain M, and bands 7-11 span the N terminus to sites in domain C; bands 13-16 derive from fragments containing both domains M and P; band 17 derives from domain M.
The electrophoretic profiles of CCT sol and CCT mem differ as follows: bands 5, 6, 6b, and 15 are missing from CCT mem , band 12 in CCT sol is replaced by 12b in CCT mem , band 13 is more prominent in CCT mem , and band 17 is generated only in CCT mem (Fig. 2). Other bands appear to be produced and processed at similar rates in both CCT forms. These data suggest that the rates of proteolysis of domain P and domain C are similar for both the soluble and membrane-bound forms, that cleavage sites in domain M to generate bands 5, 6, 6b, and possibly 15 are bypassed in CCT mem , and that the cleavage pattern of fragments containing domains M and P is altered in membrane-bound CCT.
Comparison of the Chymotrypsin Digestion Pattern of CCT sol , CCT mem , and CCT236 by MALDI-MS-To precisely identify the chymotryptic fragments of the three CCT forms, we analyzed the digests by MALDI-MS. A mass spectrum of peptides in the 19 -42-kDa range is shown in Fig. 3, and the masses of the observed peaks are compared with their theoretical masses in Table I. Because of a limited number of cleavage options and an average mass accuracy of ϳ100 ppm for peptides in this mass range, this analysis readily identified all of the primary cleavages giving rise to gel bands 1-11 (Table I). Peaks 1, 2, and 3 contain all or portions of the domain P, so they appear very broad because of heterogeneous phosphorylation. In agreement with the gel electrophoretic analysis, species corresponding to 30, 27, and 26.5 kDa (gel bands 5, 6, and 6b) were observed only in the CCT sol samples. These species correspond to Met 1 -Phe 263 , Met 1 -Tyr 240 , and Met 1 -Phe 234 . Thus membrane association of CCT converts these sites from highly accessible to very inaccessible.
Interestingly, although CCT contains seven chymotrypsin sites with Ͼ50% propensity for cleavage between Phe 263 and Leu 311 , we did not observe, on gels or in the mass spectra, fragments corresponding to Met 1  high mass spectra where these fragments would appear is shown in Fig. 4 (boxed areas). The lack of cleavage to produce these fragments indicates that the C-terminal end of domain M is inaccessible in both forms of CCT.
The mass spectra in the 10 -28-kDa range of CCT sol and CCT mem digests also show some differences (Fig. 5). Many of these species are heterogeneously phosphorylated giving rise to a set of peaks separated by 80 Da, the additional mass associated with each phosphorylation (Fig. 5, inset). This provides additional proof that these species are derived from the Cterminal portion, because the phosphorylation of CCT occurs exclusively within the last 50 residues (12,30). A set of peaks in the range 11,500 -12,100 Da, corresponding to phosphorylated versions of Val 264 -Asp 367 , is clearly present only in CCT sol . This species probably corresponds to gel band 15, observed only in CCT sol . Other peptides missing from CCT mem digests over a 15-min period but present in CCT sol digests were generated by cleavage at residues Phe 234 , Tyr 240 , and Phe 263 (Fig. 5). Thr 226 -Asp 367 does not accumulate in CCT mem because of accelerated processing of this fragment at Leu 351 , Phe 338 , and Leu 311 compared with CCT sol . In corroboration with the high mass N-terminal fragment data ( Fig. 3 and Table I), the identification of these peptides from the C-terminal domains shows that Phe 234 , Tyr 240 , and Phe 263 are exposed in CCT sol and are buried upon membrane binding.
Peptide Bond Accessibility Maps of the Three CCT Forms-The chymotrypsin-digested samples shown in Fig. 2 were subjected to a complete MALDI-MS analysis of peptides in the 1100 -7000-Da range. MALDI-MS cannot be relied upon to quantitatively detect all peptide species in a crude mixture in proportion to their abundance because of potential signal suppression of minor species by the more predominant ones and differential desorption from the matrix (31,32). To reduce signal suppression, to increase confidence that relative peptide ion detection reflects sample abundance, and to separate peptides with near identical masses, the samples were subjected to CapLC prior to MALDI-MS analysis (22). For example, the CapLC separation enabled isolation of the peptide Ala 217 -Phe 234 from Arg 42 -Tyr 59 , despite a mere 2-Da mass difference, because they had different LC elution times ( Supplementary  Fig. 1A). The Ͼ1000 individual mass spectra collected using these approaches yielded full coverage of the CCT sequence. Importantly, more than 95% of the observed peptides were definitively assigned to CCT sequences with mass accuracies of Ͻ0.01% variance. All of the peptide bond cleavages identified are displayed as PBAMs in Fig. 6. A full description of PBAM construction is provided under "Experimental Procedures." The color coding indicates the digestion times at which each site was first cleaved, identified by detection of a peptide starting or ending at that site. The peptide bond accessibility displayed in Fig. 6 does not account for the differences in numbers or signal intensity of peptides generated by cleavage of a peptide bond. If a 25-s digest of CCT sol generated five new peptides ending at Tyr 216 and only two peptides ending at Phe 234 , both sites are colored yellow, because both sites were cleaved within the same time frame. The PBAMs in Fig. 6 show that region I is less accessible in CCT mem , region II is more accessible in CCT mem and CCT236, and region III is least accessible in all forms. The peptides contributing to Fig. 6 were derived mainly from secondary cleavages but included many primary cleavages, e.g. We were concerned that the secondary peptides would reflect cleavages that occurred after the tertiary structure of the enzyme was compromised by initial cuts in the peptide backbone. However, the peptide bond accessibility derived from small mass fragment identification was compatible with the analysis of the large primary fragments. The agreement between the two approaches can be explained in that we analyzed the fragmentation patterns for Յ5 min, during which time the native structure of the core was likely retained. Retention of native structure is supported by the results of mapping the protease sites onto a structural model of domain C (see "Discussion"). Both analytical approaches concurred on four points. First, the C-terminal end and the hinge between domain C and M are the most accessible regions in CCT sol and CCT mem . Second, amino acids Phe 85 -Phe 165 constitute the most inaccessible core in all three CCT forms. Third, the entire domain M is inaccessible in CCT mem . Fourth, the C-terminal end of domain M between amino acids Leu 274 -Leu 303 (inclusive) is relatively inaccessible even in CCT sol. With regard to last point, we did not observe a primary fragment ending at Phe 285 at any time point. Peptides ending at Phe 285 were not observed until after scission at Tyr 226 .
The low mass peptide data contributing to the PBAMs shown in Fig. 6 revealed enhanced accessibility of amino acids 41 and 59 in domain N of CCT mem and CCT236 (region II). We also observed enhanced chymotrypsin accessibility of these sites in high mass spectra of CCT236 digests, where Arg 42 -Asn 236 was produced immediately and processed within 1 min by cleavage at Tyr 59 and by C-terminal trimming at Tyr 225 , Tyr 216 , and Tyr 182 . However, the analysis of high mass fragments of CCT mem did not identify peptides derived from cuts at these sites. To verify that domain N accessibility is greater in CCT mem , we utilized another protease, Arg-C. CCT has four potential Arg-C cleavage sites between Met 1 -Leu 41 in domain N, but no chymotrypsin sites in this region. Arg-C also has one potential site, Arg 283 near the chymotrypsin site, Phe 285 , within the domain M region that appeared to be relatively buried in both CCT forms.

Comparison of the Arg-C Digestion Pattern of CCT sol and CCT mem -Because
Arg-C digestion proceeded more slowly than chymotrypsin and because CCT has fewer sites for Arg-C, the fragmentation patterns were less complex than those generated by chymotrypsin. We identified by electrophoresis and immunoblotting a few large cleavage fragments (Ն34 kDa) generated by digestion for 1.5 min with Arg-C that did not react with anti-N (data not shown), indicating early cleavages that removed the N-terminal segment. Mass spectra obtained from a dried droplet preparation of the crude digestion mixture revealed three high mass fragments, Lys 13 -Arg 309 , Lys 13 -Arg 353 , and Lys 13 -Asp 367 , that account for the electrophoresis and immunoblotting observations (Fig. 4C). The peptides from a time course of Arg-C digestion were analyzed by CapLC-MALDI, and the peptide bond accessibility map derived from this data is shown in Fig. 6 (E and F). A peptide corresponding to Met 1 -Arg 12 appeared within a few seconds of digestion of CCT sol and CCT mem . Met 1 -Arg 14 was also isolated from a Ͻ10-s digestion of CCT sol . The immediate cleavage at Arg 12 and Arg 14 shows clearly that the N terminus of CCT is highly accessible. Otherwise, the Arg-C cleavage pattern was similar to that of chymotrypsin in that the C-terminal end was rapidly cleaved relative to a protease-resistant core between Arg 78 -Arg 196 (Fig. 6, E and F, region III). There are only two Arg-C  (10). By contrast Arg 283 , was positioned in the nonpolar face of the amphipathic helix (10), and accordingly, Arg 283 is inaccessible to Arg-C in CCT mem (Fig. 4).
Two other important differences between CCT sol and CCT mem , evidenced from chymotrypsin digestion, were also revealed by Arg-C digestion: (i) Sites within domain N at Arg 36 , Arg 42 , Arg 61 , and Arg 69 were cleaved to generate a variety of peptides in CCT mem , whereas none of these sites were cleaved in CCT sol for at least 10 min of digestion ( Fig. 6C and Supplementary Fig. 1C). Thus the limited digestion pattern generated by two different proteases indicate that membrane binding increases the accessibility of the C-terminal side of domain N (Fig. 6, region II). (ii) Arg 283 within the C-terminal side of domain M was cleaved in CCT sol (Fig. 6C and Supplementary  Fig. 1C), but like Phe 285 , this occurred only after cleavage of the hinge between domains C and M. As shown in Fig. 4 (C and D), large fragments corresponding to Met 1 -Arg 283 or Lys 13 -Arg 283 were not observed, although Lys 13 -Arg 223 and Lys 13 -Arg 309 were prominent. Thus Arg-C and chymotrypsin proteolysis revealed the relative inaccessibility of the C-terminal region of domain M in CCT sol .
Domain P Phosphopeptides Reveal the Degree of Phosphorylation of CCT␣-The PBAMs reveal that domain P in both CCT forms was highly accessible to both chymotrypsin and Arg-C. All Arg-C sites in domain P were cleaved immediately in CCT mem . However, only three of the four chymotrypsin sites in domain P were cleaved; Phe 334 was not cleaved. The phosphorylation state of CCT purified by the method of MacDonald and Kent has been explored by proteolysis, high pressure liquid chromatography and sequencing (12). The phosphorylation state of CCT purified by the more recent protocol of Friesen et al. (15), which we used, has not been explored. This information is important for studies on the membrane affinity of CCT purified by this method, because the phosphorylation state affects membrane affinity (33), and for consideration of CCT forms for crystallization. Cuts between Leu 311 and the C terminus generated heterogeneously phosphorylated peptides with molecular masses Ͻ7000 Da. An average mass accuracy of Ͻ50 ppm in the 1000 -7000 Da range (i.e. Ϯ0.3 Da for a 6000-Da peptide) enabled facile identification of peptides sharing the same amino acid sequence but varying in the number of phosphoryl (80 Da) units (Table II). The average degree of phosphorylation for each peptide was calculated from the relative signal intensities within each set of phosphopeptides. For the peptide spanning the entire domain P, Gln 312 -Asp 367 , the masses matched for four, five, six, and seven phosphates. Our analysis did not enable identification of each phosphorylated site; however, the results (Table II) indicate a phosphorylation pattern that is very similar to that obtained by MacDonald and Kent (12). The phosphorylated serines are spread over the entire 50-residue domain.

Implications of the Peptide Bond Accessibility Data on CCT
Structure-The idea that domain M functions as an autoinhibitory domain in CCT␣ is strongly supported by the constitutive activity of CCT236. The present work has begun to unravel the nature of the inhibitory interactions of domain M by mapping its potential interdomain contact sites. The construction of the accessibility maps for CCT (Fig. 6) has generated a model that confirms the bipartite tertiary structure of CCT sol , that of a compact head followed by a loosely folded and perhaps flexible tail. The maps show that the tertiary structure of domains N and C in CCT mem resembles that of CCT236 rather than CCT sol . The maps reveal that the central core of the catalytic domain remains inaccessible upon membrane binding and that  Fig. 2. Red, Ͻ10 s; yellow, 20 s; green, 60 s; blue, 300 s; black, sites with Ͼ50% propensity for chymotrypsin cleavage that were not cut within 300 s (site 221 had Ͻ50% propensity). E and F, peptide bond accessibility to Arg-C of CCT sol (E) and CCT mem (F). Red, Ͻ10 s; yellow, 90 s; green, 240 s; blue, 600 s; black, sites that were not cut within 600 s. Only new cut sites measured at each time point were plotted. PBAMs assembled from independent digests with chymotrypsin and Arg-C yielded similar findings. Region I was more accessible in CCT sol , region II was more accessible in CCT mem and CCT236, and region III was least accessible in all forms. domain P remains the most accessible region. They show clearly that the region buried upon membrane binding is restricted to domain M. A region on the C-terminal flank of domain M is inaccessible in both the soluble and membrane forms of the enzyme, and the C-terminal region of domain N becomes more accessible in CCT mem . Because membrane binding exposes a region of domain N, but not domain C, we hypothesize that in the soluble form domain M of CCT interacts with the region of domain N that becomes exposed upon membrane binding of domain M. The domain M-domain N interaction may be responsible for the autoinhibition of the enzyme.
Protease Accessibility of the Catalytic Domain Remains Unaltered Upon Membrane Binding-Residues Phe 85 -Phe 165 of all three CCT forms are relatively inaccessible to chymotrypsin and residues Arg 78 -Arg 196 are inaccessible to Arg-C in both the soluble and membrane forms of CCT. Chymotrypsin digested domain C progressively from the C terminus at Phe 191 , Tyr 182 , and Tyr 173 , but only after cleavage in the very accessible hinge region (amino acids 216 -225). After these primary cleavages secondary peptides were generated at Leu 158 , Trp 151 , Phe 124 , Phe 121 , Tyr 107 , Leu 96 , and Tyr 80 . We did not observe any consistent differences in the chymotrypsin peptide bond accessibility in domain C between the three CCT forms with the exception of sites Tyr 80 and Phe 85 near the junction with domain N.
To assess the positioning of the accessible and inaccessible sites on the catalytic domain, we took advantage of the solved structure for a homologous cytidylyltransferase, GCT (8). Residues Arg 78 -Arg 211 of CCT␣ share ϳ60% similarity and ϳ30% identity with residues 3-129 of GCT. Residues 78 -211 of CCT␣ were modeled upon the atomic coordinates available for this enzyme (Protein Data Bank code 1COZ) using MODELLER. Fig. 7A shows the close match between peptide backbones (see "Experimental Procedures" for discussion of the modeling). Fig.  7B presents a surface rendering of the CCT domain C dimer, highlighting the chymotrypsin and Arg-C sites. The purpose of this representation is to localize the protease sites with respect to major domain C surfaces and secondary structural elements. In general the protease accessibility data correlate well with surface accessibility of sites on this model of the native domain C fold. The sites most accessible to chymotrypsin (Trp 151 , Leu 158 , Tyr 173 , Tyr 182 , and Phe 191 ; colored orange in Fig. 7B, bottom image) are on the periphery of the catalytic dimer, located in loops, the ends of structural elements, or within the peripheral 3-10 helix. Chymotrypsin sites that were cleaved more slowly, i.e. between 1 and 5 min (colored green in Fig. 7B), were at the start of ␤2 (Tyr 107 ), the end of ␣-B (Phe 121 ), or in loop L2 (Phe 124 ). Of the sites inaccessible to chymotrypsin, Phe 85 , Phe 88 , Leu 96 , and Leu 108 are completely buried on the model. The remaining uncleaved sites had only partially exposed side chains on the model (colored blue in Fig. 7B). Arg-C did not cleave any sites in the entire catalytic domain over a 10-min time-frame. Arg 94 and Arg 147 appear fairly exposed (Fig. 7B, top image), but they are positioned in the middle of ␣-helix A and ␤-3, respectively, impeding access of the protease to the peptide backbone. However, Arg 162 and Arg 140 are in turns and appear very exposed on the upper surface of the model in Fig. 7B. We suggest that this surface might be involved in interactions with domain N (Fig. 8). If domain M were bound tightly to the surface of domain C in CCT sol , one would anticipate that those domain C sites buried by domain M would become more accessible in CCT mem and CCT236. Because the Arg-C and chymotrypsin sites in domain C were relatively inaccessible in all three forms, we have not found data to support such a model. Our results suggest that the catalytic core is tightly folded and remains quite rigid upon membrane binding of domain M. Thus the changes at the active site upon membrane binding or truncation to generate CCT236 do not involve major reorganization of structural elements in domain C that would be detected by protease accessibility.
Domain P Remains Highly Accessible Upon Membrane Binding-In contrast to domain C, domain P is the region of highest protease accessibility in both CCT sol and CCT mem . The rapid decrease in intensity of band 1 (Fig. 2), the decreased relative intensity of peak 1 to peak 2 in high mass MALDI-MS data (Fig. 3, B and C) and the absence of the peptide Tyr 226 -Asp 367 (Fig. 5) suggest that it may even be more accessible in CCT mem . This observation would appear to contradict a previous suggestion that domain P can serve as a second membrane-binding domain acting together with domain M. This was suggested by the lipid responsiveness of the activity of a CCT mutant missing residues 257-309, i.e.  4 , column A indicates the potential number of phosphates, based on the total number of serines in the peptide sequence. Column B indicates the average number of phosphates/peptide based on results of CCT␣ purified by the MacDonald and Kent method (12). Column C indicates the average number of phosphates/peptide calculated from our MALDI-MS data of CCT␣ purified by the Friesen method (15). S was Ͼ90% phosphorylated as determined by MacDonald and Kent (12), and the rest were ϳ50 -60% phosphorylated.  (23), but we did not detect any peptides with N or C termini at any of these positions except for Phe 285 . There is only one site for Arg-C cleavage in this region at Arg 283 . Both Phe 285 and Arg 283 are cut only after cleavage in the hinge region. (Chymotrypsin cleavage at Tyr 225 precedes that at Phe 285 ; Arg-C cleavage at Arg 223 or Arg 245 precedes cleavage at Arg 283 ). We never observed high mass fragments terminating at these residues, even after the longest digestion time. This region of inaccessibility could be a contact site between domain M and another domain (e.g. domain N, as shown in Fig. 8), a potential contact site subject to modulation upon membrane binding.
Membrane Binding Buries Residues Phe 234 -Leu 303 -Analyses of both high mass and low mass peptides were consistent in showing that Tyr 240 , Phe 263 , and Phe 285 as well as Arg 283 , accessible in CCT sol , are uncleaved even at the longest digestion time in CCT mem . Residue Phe 234 is also more proteaseresistant in CCT mem than CCT sol . The residues flanking this region, Tyr 225 and Arg 223 , Leu 311 and Arg 309 are very rapidly cleaved in both forms. That Phe 234 is relatively inaccessible to proteases in the presence of membranes may reflect its burial in lipid or a loss of flexibility of the peptide backbone in the near vicinity of the membrane embedded helix. Peptides spanning residues Asn 236 -Leu 299 fold into amphipathic ␣-helices in the presence of anionic vesicles (35)(36)(37). The lipid interaction is accompanied by fluorescence blue shifts of Tyr 240 and Trp 278 , indicative of transfer to a more nonpolar environment (35)(36)(37). The tryptophan fluorescence at positions 263 or 278 is sensitive to quenching by lipids with brominated acyl chains (35,36). These data argue that the membrane-embedded region encompasses residues 234 -285, at a minimum. The C-terminal flank of domain M between residues 286 and 303 is inaccessible in both forms; thus it could be argued that this region remains locked in interdomain contacts upon membrane binding rather than dissociating to bind lipids. However this scenario does not provide an explanation for the enhanced accessibility of domain N in CCT mem or CCT-236. Thus the membrane binding region is likely to encompass residues 234 -303.
Membrane Binding Increases the Accessibility of Sites in Domain N-Arg-C generated immediate cleavage at Arg 12 and Arg 14 in both CCT sol and CCT mem , showing for the first time that the N-terminal 15 residues of CCT are not a part of a compact fold (Fig. 8). Analysis of low mass peptides after chymotrypsin or Arg-C digestion indicated membrane-enhanced access to the peptide backbone at positions Arg 36 , Leu 41 , Arg 42 , Tyr 59 , Arg 61 , Arg 69 , and Tyr 80 . In the constitutively active CCT236, domain N was also more accessible in this region. There is other evidence for a restructuring of CCT in this portion of domain N upon membrane binding (13). Cysteine 37 is within disulfide bond distance with its counterpart in the opposite subunit in the soluble form of CCT, i.e. at the dimer interface. In CCT mem these cysteines are repositioned to be outside disulfide bond distance (13). We have observed that in CCT sol the C-terminal region of domain M is buried, as is the C-terminal segment of domain N. This same region of domain N increases its accessibility upon membrane binding, whereas sites in domain C remain inaccessible to protease. One explanation to account for these changes is that there is an interaction between residues Val 274 -Leu 303 in domain M with residues Arg 36 -Arg 69 in domain N in CCT sol , that protects these regions from protease cleavage (Fig. 8). Upon membrane binding and dislocation of domain M, the sites in domain N become accessible. The accessibility of this region of domain N in CCT236, which lacks domain M, also supports the notion of a domain N/domain M interaction in CCT sol . Because Arg-C and chymotrypsin cleavage sites do not mark the entire surface of domain C, it is quite possible that domain M contacts a portion of domain C as well as domain N in CCT sol . Upon membrane binding, domain N remains bound to domain C but repositions subtly (Fig. 8). We propose that the interaction of domain M with N contributes to autoinhibition of CCT in its soluble form. Our modified hypothesis for autoinhibition can now be tested using other approaches to specifically probe for interactions between domains M and N.
The domain N Ϫ domain M interaction could interfere with substrate docking or with the movements of helix B and loop L2 within the catalytic domain, which by analogy with GCT (9) may be important for catalysis (38). Comparison of the x-ray-derived structures of GCT in complex with substrate or with product has led to the proposal that during a catalytic cycle a ϳ10-residue segment encompassing the C terminus of helix B and the adjoining loop (L2) moves to engage the substrate, glycerol phosphate (3-Å shift at the point of maximal dislocation) and to stabilize the transition state at the ␣ phosphate of CTP (9). Charge stabilization of the transition state is accomplished by Lys 44 and Lys 46 in GCT (9) and Lys 122 in CCT, which are situated in the L2 loop (38). Mutation of Lys 122 to alanine in CCT reduces k cat /K m by a factor of ϳ200,000 (38). A similar relocation of basic side chains is involved in catalysis of other structurally similar nucleotidyltransferases (9). The contacts between domains N and M in CCT sol may constrain this movement to prevent ideal orientation of Lys 122 . FIG. 8. Domain reorganization upon activation of CCT by membrane binding. Proposed working model of domain organization around the catalytic domain dimer developed using the newly collected data as well as existing data. A, CCT sol ; B, CCT mem ; C, CCT236. To differentiate between the components of the two CCT molecules that comprise each dimer, the domain letters carry subscript numbers denoting CCT molecules 1 and 2. CTP is docked into the approximate region identified from the solved structure of the GCT-CTP complex.