An Environment-dependent Structural Switch Underlies the Regulation of Carnitine Palmitoyltransferase 1A*♦

Background: The enzyme carnitine palmitoyltransferase 1 regulates the rate of fatty acid oxidation. Results: The regulatory domain of the enzyme constitutes an amphiphilic structural switch. Conclusion: The switch can integrate enzyme inhibitor concentration and membrane characteristics into one regulatory signal. Significance: This represents a highly sophisticated regulatory mechanism for enzymes residing at the membrane-water interface. The enzyme carnitine palmitoyltransferase 1 (CPT1), which is anchored in the outer mitochondrial membrane (OMM), controls the rate-limiting step in fatty acid β-oxidation in mammalian tissues. It is inhibited by malonyl-CoA, the first intermediate of fatty acid synthesis, and it responds to OMM curvature and lipid characteristics, which reflect long term nutrient/hormone availability. Here, we show that the N-terminal regulatory domain (N) of CPT1A can adopt two complex amphiphilic structural states, termed Nα and Nβ, that interchange in a switch-like manner in response to offered binding surface curvature. Structure-based site-directed mutageneses of native CPT1A suggest Nα to be inhibitory and Nβ to be noninhibitory, with the relative Nα/Nβ ratio setting the prevalent malonyl-CoA sensitivity of the enzyme. Based on the amphiphilic nature of N and molecular modeling, we propose malonyl-CoA sensitivity to be coupled to the properties of the OMM by Nα-OMM associations that alter the Nα/Nβ ratio. For enzymes residing at the membrane-water interface, this constitutes an integrative regulatory mechanism of exceptional sophistication.

Mitochondrial oxidative phosphorylation is the primary source of ATP formation in eukaryotes. In most mammalian tissues, this process utilizes either pyruvate or long chain fatty acids (LCFA) 4 as substrates. Carnitine palmitoyltransferase 1 (CPT1), which is an integral protein of the outer mitochondrial membrane (OMM), converts LCFA-CoA esters to carnitine esters. This permits access of the acyl moieties into the mitochondrial matrix, where CPT2 mediates their conversion back to CoA esters followed by their ␤-oxidation (1,2). The inhibition of CPT1 by malonyl-CoA (MCoA), the first intermediate of fatty acid synthesis, enables the enzyme to exert major control over the rate of fatty acid oxidation in all tissues. MCoA acts as a stringent physiological inhibitor of the liver and heart/muscle isoforms CPT1A and -B, respectively (3,4). Moreover, the sensitivity of CPT1A to MCoA inhibition is markedly affected by the lipid composition and fluidity of the OMM, both when they change physiologically, for example under starvation or diabetes (5,6) or when the fluidity of the membrane is altered experimentally in vitro (7). This correlation to the longer term physiological state endows CPT1A with a pivotal metabolic role in fuel selection, particularly in pathological conditions such as type 2 diabetes and obesity (8,9) for the treatment of which CPT1A is a drug target (10). In addition, CPT1A shows very different enzyme kinetics depending whether it resides within the bulk outer membrane or in the contact sites that exist between this membrane and the inner membrane of mitochondria (11).
CPT1A consists of a short, well conserved N-terminal regulatory domain (N) and a large C terminal catalytic domain (CD), which are linked by two transmembrane segments that span the OMM (Fig. 1, A and C). As demonstrated by cross-linking experiments (6), complementary mutagenesis (12), and protein engineering (13), MCoA sensitivity is modulated by a poorly understood interaction between N and CD. Specifically, mutagenesis studies have identified two classes of mutations within N (12, 14 -16). Substitutions or deletions involving approximately the first 18 residues decrease MCoA sensitivity, whereas substitutions or deletions within residues 19 -40 increase MCoA sensitivity (16). The ability to increase MCoA sensitivity is noteworthy because essentially random mutations are unlikely to increase the affinity between N and MCoA or CD, pointing to the evolution of an intricate regulatory mechanism at the membrane-water interface. Also noteworthy are the common structural features of the MCoA inhibitor and PCoA substrate (Fig. 1B), which raises the question as to how the enzyme differentiates between both compounds in the context of N-CD interactions when only the CoA moiety protrudes from the active site (Fig. 1A).
Although it is possible to express full-length CPT1A heterologously in Pichia pastoris (17), structural studies of CPT1 that could provide a rationale for its complex regulatory mechanism have not been achieved yet perhaps because of its membrane association. However, the catalytic domain of CPT1A shares an ϳ30% sequence identity with other MCoA-insensitive acyltransferases of known structure, including carnitine acetyltransferase, choline acetyltransferase, and CPT2 (18 -21). This level of sequence homology permits the construction of meaningful CD structural models (e.g. Fig. 1A), which can explain the effects of certain point mutations within the CD (22)(23)(24). To provide structural insight into the CPT1 regulatory mechanism, it therefore appears most relevant to characterize the structure of the N-terminal domain. A structural model for N has been discussed (12), but without significant sequence identity to known structures, modeling appears challenging in this case. In addition, the absence of an explanation as to how CPT1A senses the molecular properties of the membrane and integrates them into the MCoA concentration-dependent signal warrants structural studies of N in the presence of membrane mimics.
Here, we used multidimensional, heteronuclear NMR spectroscopy to reveal two alternative structures for the N-terminal regulatory domain of CPT1A. Structure-based site-directed mutageneses in conjunction with functional assays of the fulllength protein ascribe physiological relevance to both structures and suggest that their environment-dependent population ratio is used to integrate MCoA concentration, membrane composition, and curvature into one regulatory signal. In combination with molecular modeling of the N⅐MCoA⅐CD complex, the structural basis of CPT1A regulation is proposed. These results provide fundamental insight into the regulation of enzymes residing at the membrane-water interface.

EXPERIMENTAL PROCEDURES
Expression and Purification of N-terminal CPT1A Peptide-Using the cDNA of human liver carnitine palmitoyltransferase 1 (CPT1A), isoform 2 (Mammalian Gene Collection Image 3352642; UniProtKB entry P50416) as template, an insert encompassing Met 1 -Lys 42 was generated by PCR and subcloned into the pET-44 expression vector (Novagen, Inc.) with the third IgG-binding domain of protein G (GB3) as N-terminal fusion protein and an intervening tobacco etch virus protease cleavage site. A truncation variant, N(⌬Met 1 -Val 8 ), and G18A point mutation were subsequently constructed using standard techniques.
NMR Sample Preparation-N peptide concentration was measured in acetonitrile/water solution by UV spectroscopy (⑀ 280 nm ϭ 6,990 M Ϫ1 cm Ϫ1 ), and defined amounts of peptide were freeze-dried. Peptide was taken up in a volume of 280 l to yield the following solutions: 0. The N peptide⅐DDAC micelle complex was aligned relative to the magnetic field by stretched, negatively charged polyacrylamide gel of 320-l volume, which was polymerized from a 4.6% w/v solution of acrylamide, 2-acrylamido-2-methyl-1propanesulfonate, and bisacrylamide with a monomer-tocross-linker ratio of 39:1 (w/w) and a molar ratio of 98:2 of acrylamide to 2-acrylamido-2-methyl-1-propanesulfonate (26). Upon transferring the gel into an open-ended NMR tube, a 2 H splitting of 2.1 Hz was observed following overnight equilibration.
NMR Spectroscopy-NMR experiments were carried out on a cryoprobe-equipped Bruker Avance 700 spectrometer at 35°C. Data were processed and analyzed with the nmrPipe package and CARA. 2 H/ 13 C/ 15 N-labeled N peptide and TROSY-type H-N detection (27) were used for HNCA-, HNCACB-, and HNCO-based backbone assignments, the measurement of 3 J CЈC␥ and 3 J NC␥ couplings (28), and the detection of 1 J NH , 1 J C␣CЈ , and 1 J CЈN , as well as 1 J NH ϩ 1 D NH , 1 J C␣CЈ ϩ 1 D C␣CЈ , and 1 J CЈN ϩ 1 D CЈN couplings (29 -31) using isotropic and aligned samples, respectively. { 1 H}-15 N NOE measurements were carried out also using deuterated peptide with 5 s of presaturation preceded by a recycling delay of 4 s for the NOE experiment and a 9-s recycle delay for the reference experiment. Aliphatic side chain assignments of a protonated sample were achieved using HAHB(CBCACO)NH, (H)CCC(CO)NH, and H(C)CC-(CO)NH experiments (32). Interproton NOEs were measured by 15 N-edited NOESY and HSQC-NOESY-TROSY experiments (120-ms mixing time) in H 2 O using protonated and deu-terated peptide, respectively. A 13 C-edited NOESY spectrum, optimized for the detection of aliphatic nuclei, was recorded with a mixing time of 120 ms in D 2 O solution.
Structure Calculations-The DDAC-bound N peptide structure of the well folded Ala 9 -Trp 39 residues (Fig. 3B) was calculated by simulated annealing, starting at 3000 K using the program XPLOR-NIH (33). The peptide termini were represented by random-coil conformations. Except for Gly 18 , , backbone dihedral angle constraints were derived from N, H ␣ C ␣ , C ␤ , and CЈ chemical shifts using the program TALOSϩ (34). 1 side chain angle restraints were derived from the 3 J CЈC␥ and 3 J NC␥ coupling constants (supplemental Table S1). NOE interproton distance restraints were referenced to the linear ␣-helical structure of His 25 -Trp 39 . In addition to standard force field terms for covalent geometry (bonds, angles, and improper dihedrals) and nonbonded contacts (van der Waals repulsion), dihedral angle and interproton distance restraints were implemented using quadratic square well potentials, and a backbone-backbone hydrogen-bonding potential and torsion angle potential of mean force were employed (35,36). The difference between predicted and experimental residual dipolar couplings (⌬ 1 D) was described by a quadratic harmonic potential. The final values for the force constants of the different terms in the simulated annealing target function are as follows: 1,000 kcal⅐mol Ϫ1 ⅐Å Ϫ2 for bond lengths; 500 kcal⅐mol Ϫ1 ⅐rad Ϫ2 for angles and improper dihedrals, which serve to maintain planarity and chirality; 4 kcal⅐mol Ϫ1 ⅐Å Ϫ4 for the quartic van der Waals repulsion term; 30 kcal⅐mol Ϫ1 ⅐Å Ϫ2 for interproton distance restraints; 500 kcal⅐mol Ϫ1 ⅐rad Ϫ2 for dihedral angle restraints; 0.3 kcal⅐mol Ϫ1 ⅐Hz Ϫ2 for 1 D NH residual dipolar couplings restraints, and 1 D CЈN and 1 D C␣CЈ scaled relative to 1 D NH according to their dipolar interaction constants; 1.0 for the torsion angle potential; and a directional force of 0.20 and a linearity force of 0.05 for the hydrogen-bonding potential. A total of 20 structures were calculated (Table 1). Coordinates, chemical shifts, and structural constraints for N␤ have been deposited in the Protein Data Bank as code 2le3 and BMRB with accession number 17690.
CPT1A Homology Modeling and N Docking-An atomic resolution model of the CD of human CPT1A (residues 168 -766) was derived from rat choline acetyltransferase (PDB code 1T1U; 29% sequence identity) by homology modeling (37). Similar results were obtained when using mouse carnitine acetyltransferase (PDB code 1NDB) or rat choline acetyltransferase (PDB code 1Q6X) templates. Docking calculations with N␤ were performed using the program ClusPro 2.0 (38). MCoA was placed in the CoA and carnitine binding pockets of the N␤⅐CD models using the complexes of the CD of mouse carnitine acetyltransferase with CoA and carnitine, respectively (PDB codes 1ndi and 1ndf, respectively) (20), as templates.
Assay of CPT1 Activity-The construction of the pGAPZ plasmid expressing wild type rat CPT1A has been described previously (16). The point mutations in CPT1A (Fig. 6) were prepared by QuikChange PCR. Transformation into and expression in P. pastoris strain X-33 were performed according to the expression system distributor Invitrogen. Mitochondrially enriched cell-free extracts from P. pastoris expressing CPT1A were prepared and assayed for CPT activity with L-[N-methyl-3 H]carnitine hydrochloride as described previously (17). Briefly, the standard assay buffer contained 150 mM KCl, 1 mM EGTA, 20 mM HEPES, pH 7.2, 1% (w/v) fatty acid-free BSA, and 1 mM dithiothreitol. Reactions were prepared in duplicate with palmitoyl-CoA added to a final concentration of 35 M and the addition of varying amounts of malonyl-CoA in a final volume of 1.1 ml. Yeast extract was added, and reactions were heated to 30°C for 2 min before the addition of L-[ 3 H]carnitine to a final concentration of 500 M and then incubated for a further 4 min before termination of the reaction by addition of 0.3 ml of 6 M HCl and transfer to ice. [ 3 H]Palmitoylcarnitine was extracted into butan-1-ol at low pH and determined by liquid scintillation counting. Nonlinear curve fitting and calculation of enzyme kinetic parameters were performed as described previously (16,39). Intrinsic enzyme activities were unchanged relative to the wild type when normalized for the level of protein expression as quantified on Western blots by anti-C antibodies (40).

N-terminal CPT1A Domain Can Adopt Two Structural
States-Based on the functional properties of CPT1, we hypothesized that N might associate with the CD, with the OMM or exist "free" within the aqueous environment (Fig. 1A). To address this hypothesis, a peptide, also termed N, corresponding to the first 42 residues of human CPT1A (Fig. 1C) was prepared recombinantly. N was found to be insoluble in aqueous solution, suggesting that, in native CPT1A, it cannot exist free in the cytosol. If N were to associate with the OMM, an amphiphilic structure would be required. To detect and stabilize such a structure, we investigated its interaction with phospholipid bicelles and detergent micelles, which constitute well established model systems for the study of protein-membrane interactions (41,42). Upon screening an extensive array of different bicelles and micelles, N was found to be amenable to structural characterization by NMR spectroscopy in the presence of several micelle systems ( Fig. 2A).
Depending on the micelle system employed, one of two characteristic H N -N peak patterns was observed (Fig. 2, B and C). One of these states, termed N␣, was observed in the presence of cationic tetra-or hexadecyltrimethylammonium chloride (TDAC or HDAC) micelles (Fig. 2, A and B). The second state, termed N␤, was obtained not only in the presence of neutral n-octyl-␤-D-maltopyranoside (OM) and zwitterionic DDMG micelles but, unexpectedly, also in the presence of decyl-and dodecyltrimethylammonium chloride (DAC and DDAC) micelles (Fig. 2, A and C). DDAC and TDAC differ in the length of their hydrocarbon tails by two methylene groups ( Fig. 2A), giving rise to different micelle sizes. Like DAC and DDAC, OM and DDMG also exhibit relatively short hydrocarbon tails ( Fig.  2A), which are anticipated to result in small micelles (low aggregation numbers). Thus, among the detergents studied, micelle size rather than headgroup identity determined the structural state of N. This invariance to a wide range of headgroup types clearly indicates that micelles are suitable membrane mimics for N.
To examine the influence of the micelle system on N␣ and N␤ formation in more detail, 13 C ␣ chemical shift assignments were performed for N bound to OM and the four trimethylammonium-based micelles studied ( Fig. 2A). With increasing tail lengths, these latter micelles exhibit aggregation numbers of 30, 44, 62, and 84, respectively (43,44), and thus offer a continuous increase in micelle surface area paired with a decrease in scaffold curvature. In contrast to 1 H N -15 N chemical shifts, 13 C ␣ chemical shifts correlate directly with backbone torsion angles (45,46), rendering them powerful probes of protein conformation. For the five micelle systems studied, only two sets of 13 C ␣ chemical shifts were observed (Fig. 3A). Consequently, N exhibited virtually identical backbone conformations when bound to OM, DAC, or DDAC micelles, as well as when associated with TDAC or HDAC micelles. This observation demonstrated that the conformation of N does not change gradually and continuously according to the size of the micelles provided but interconverts between N␣ and N␤ in a switch-like manner when a discrete micellar size threshold is reached.
In conclusion, micellar membrane mimics support an association of N with the OMM. Because there are two functionally opposing types of mutations within the N-terminal domain of CPT1A with respect to the effects on MCoA sensitivity (12, 14 -16), the detection of two structural states may be significant. N is relatively small (Fig. 1, A and C) and an amphiphilic association with the OMM may require the N-CD association to be also of amphiphilic nature, i.e. micelles may act as folding scaffolds for the OMM-as well as the CD-associated state of the N-terminal regulatory domain.
Structure of N␤ State-To elucidate the structural basis of the switch-like behavior of N and to assign functionality to the respective structural states, the structure of N␤ was determined by multidimensional, heteronuclear NMR spectroscopy. Simulated annealing calculations based on residual dipolar coupling, close interproton distance, and backbone and side chain torsion angle restraints were carried out, resulting in an ensemble of a total of 20 structures for which a coordinate precision of 1.03 Å was obtained for heavy atoms (Fig. 4A). A summary of structural statistics is provided in Table 1. Despite the relatively small size of N, a complex structure, encompassing ␣-helix, ␤-sheet, ␤-turn, as well as an unstructured stretch, was found (Fig. 4, A and B). The eight N-terminal residues of N␤ (Met 1 -Val 8 ) were unstructured and were followed by an anti-parallel ␤-sheet, comprising sheets ␤1 (Phe 10 -Val 14 ) and ␤2 (Ile 19 -Leu 23 ). At the C terminal end of the peptide, an ␣-helix, termed ␣2, was present encompassing His 25 -Trp 39 (Fig. 4A). The ␤1-␤2 connection was formed by Thr-Pro-Asp-Gly, constituting a hairpin of canonical sequence (47). Ser 24 mediated the transition between ␤2 and ␣2. It entertained backbone-tobackbone hydrogen bonds with both the CϭO of the first structured residue (Ala 9 ) and the N-H group of Leu 28 . Moreover, close contacts between Ser 24 (O ␥ ) and Glu 26 (H N ) as well as Ala 27 (H N ) were detected, providing further capping of helix ␣2 (Fig. 4C). Another notable feature of N␤ was the hydrophobic packing interaction between the methyl groups of Ala 27 and Ala 9 (Fig. 4C). With the exception of Phe 12 and Tyr 32 , side chain rotamers were averaged (supplemental Table S1). The regular secondary structure elements of N␤ showed backbone dynamics characteristic of well ordered structural elements as evidenced by close to maximal { 1 H}-15 N NOE values (Fig. 3B). Slight decreases in backbone dynamics were noticeable between secondary structure transitions (Thr 15 -Gly 18 and Ser 24 ).
The N␤ structure was found to be distinctly amphiphilic, which would have been difficult to predict from its primary sequence (Fig. 1C). The residues lining each side of the ␤1-␤2 interface are either hydrophobic or charged/polar, and helix ␣2 exhibits distinctly amphiphilic periodicity (Fig. 4, B and D). N therefore exhibits tertiary amphiphilicity, which is rare (48). Between the longitudinal axes of ␤1-␤2 and ␣2, an angle of ϳ105°was obtained, providing a rationalization for the preference of N␤ to associate with relatively small, highly curved micelle scaffolds. The nonplanar structure of N␤ makes it suitable to associate with the CD (Fig. 1A).
Description of N␣ State and N␣ 7 N␤ Switching Determinant-From the difference of 13 C ␣ chemical shifts of N␣ and N␤ (Fig. 3A), it was evident that both states share helix ␣2 but diverge otherwise. Specifically, for N␣, the structure comprising sheets ␤1-␤2 became destabilized concomitant to the establishment of an N-terminal helix comprising Ala 2 -Val 8 , termed ␣1 (Fig. 3A). The stability of ␣1 was lower than that of ␣2 (Fig. 3B), and therefore, helix ␣1 must be considered to be only partially folded under the current experimental conditions. However, we note that, due to post-translational processing, CPT1 commences with acetylated Ala 2 (49), which is an efficient helix-capping motif (50). Under in vivo conditions, the helical propensity of Ala 2 -Ala 8 in N␣ may therefore be promoted further. In the context of only partially ordered ␣1 and ␤1-␤2 secondary structures, the structural description of N␣ must remain illustrative for these regions at present. Fig. 5 depicts a model that was based on the interpretation of chemical shifts (34) and the overall tertiary arrangement of N␤. Whereas helix ␣2 is well structured, any global N␣ structure merely represents one possible ␣1/␤1-␤2 conformation out of an ensemble of partially folded structural states. We propose that the N␣ structure, which is formed when the ␤1-␤2 conformation is destabilized, will have a propensity to bind to the planar OMM or, alternatively, to the OMM and CD simultaneously (Fig. 1A).
The contrast between the N␣ and N␤ structures suggests that the formation of ␣1 and ␤1-␤2 are mutually exclusive. In the N␤ state, Ala 9 packs against Ala 27 , which directs the unfolded Met 1 -Val 8 residues away from the micelle surface (Fig. 4, B and C). Similarly, because of steric clashes, helix ␣1 cannot be formed without compromising the structural integrity of either ␣2 or ␤1-␤2 (Fig. 5). The mutual exclusivity of ␣1 and ␤1-␤2 appears to be an intrinsic design feature of N, contributing to the switch-like structural transition between N␣ and N␤. To understand the conformational properties of ␣1 in more detail, we next investigated whether residual ␤1-␤2 struc-  DECEMBER 9, 2011 • VOLUME 286 • NUMBER 49

JOURNAL OF BIOLOGICAL CHEMISTRY 42549
tural propensity (Fig. 3) precludes the formation of a stable ␣1 helix. A G18A substitution in N was generated to abrogate the ␤-hairpin, as Ala cannot sustain a backbone torsion angle range compatible with the hairpin. Compared with the wild type N␤ state, G18A led to a loss of NMR chemical shift dispersion (supplemental Fig. 1A), indicating that N␤ could no longer be formed. By contrast, in N␣, the substitution did not increase the helical content of ␣1, as evidenced by virtually identical 13 C ␣ chemical shifts between N␣(G18A) and N␣ (supplemental Fig. 1B). This suggests that, within the context of the micelle system employed, residues Ala 2 -Ala 8 had reached their maximum helical content.
The switch from N␤ to N␣ was observed upon increasing the micelle size (see above). We tested whether this occurs because of an increase in micelle curvature or an increase in micelle surface area that permitted a greater degree of ␣1 binding so as to compete with ␤1-␤2 sheet formation. To distinguish between the two possibilities, a peptide lacking the first eight amino acid residues, N(⌬1-8), was studied. This deletion did not rescue N␤ when binding to TDAC micelles (supplemental Fig. 2), showing that curvature per se was the sole N␣ 7 N␤ switching determinant and that the micelle systems employed did not restrict the structure adopted by N␣. Therefore, all our data indicate that, as the binding surface curvature is increased, the well defined ␤1-␤2/␣2 structure of N␤ cannot be sustained, giving way to the adoption of the ␣1 helix as the default conformation. The switching curvature threshold would appear to be close to the micelle radius of DDAC of 17.1 Å (43,44), at which solution conditions can be chosen to obtain N␣ and N␤ simultaneously (supplemental Fig. 3). In conclusion, the peptide equivalent to the regulatory domain of CPT1A can undergo an amphiphilic environment-dependent structural switch in which a disturbance in target N␤ binding surface curvature flips N␤ to the default N␣ state.
MCoA-dependent Regulation of CPT1A-To assess the functional significance of N␣ and N␤ within the native CPT1A environment, we characterized enzyme kinetics, most notably the concentration of MCoA at which half-maximal inhibition (IC 50 value) is observed, as a function of structure-based point mutations. Following well established protocols (16,17), rat CPT1A mutants were expressed in the yeast P. pastoris where the enzyme resides in its native eukaryotic OMM lipid environment. The N-terminal regulatory domain is fully conserved between human and rat CPT1A except for Y32C (Fig. 1C). First, we studied point mutations aimed at modifying key determinants of the N␤ 3 N␣ transition without affecting side chain contacts external to N to avoid altering any N-CD interactions. The G18A and A9G substitutions fall in this category and were predicted to favor the N␤ 3 N␣ transition ( Fig. 4 and supplemental Fig. S1). When incorporated into CPT1A, either mutation led to a significant increase in the MCoA sensitivity of the enzyme (Fig. 6) without any change to the intrinsic activity of the expressed proteins (data not shown). This result identified N␣ as the conformation that promotes MCoA inhibition, whereas N␤ represents the noninhibitory state. By contrast, disturbing N␣ should make CPT1A less sensitive to MCoA inhibition. Because the N switching machinery resides exclusively within the N␤ structure (Ala 9 /Ala 27 , Thr 15 -Gly 18 , and Ser 24 ), the reverse N␣ 3 N␤ switch can only be achieved by disturbing putative side chain contacts of N␣ with the MCoA⅐CD complex. We predicted that altering a key side chain within helix ␣1 (Fig. 5) would decrease MCoA sensitivity. The charge reversal mutation E3R was employed and resulted in a decrease in enzyme sensitivity to MCoA (Fig. 6), in agreement with our expectation of noninhibitory N␤ benefiting from dysfunctional N␣, i.e. an increased likelihood of N␣ 3 N␤ transitions. In conclusion, we propose that an N␤ 3 N␣ structural switch of the here-determined structures underlies the MCoAdependent inhibition of CPT1A.
In the light of our present observations, it is instructive to revisit the mutagenesis data that have established the existence of positive and negative determinants of MCoA sensitivity in the N-terminal domain of CPT1A. Mutations that make CPT1A more sensitive to MCoA localize within ␤1-␤2 and ␣2 (14) and are anticipated to destabilize the ␤1-␤2/␣2 arrangement of N␤ (e.g. S24A) or to diminish N␤ affinity to its target surface as a result of losing its amphiphilic nature and/or side chain contacts (e.g. G30A and R22A/H25A/K29A). In addition, the ⌬(19 -30) and ⌬(19 -46) deletion mutants, which enhance MCoA sensitivity (16), would prevent the ␤1-␤2/␣2 arrangement of N␤ altogether (Fig. 4) and may be regarded as achieving a permanent N␤ 3 N␣ switch. By contrast, substitutions or deletions that make CPT1A less sensitive to MCoA localize within helix ␣1 except for E26K mutations (12, 14 -16), which resides in helix ␣2 and may make contacts with the CD required for inhibition in accordance with the conservation of helix ␣2 between N␣ and N␤ (Figs. 4 and 5). We regard those "negative" mutations as negating the formation of functional N␣ by disturbing the formation of key side chain contacts between N and MCoA⅐CD and/or by interfering with the mildly amphiphilic

TABLE 1 Structural statistics for N␤ (DDAC-bound N peptide)
Statistics for the 20 calculated simulated annealing structures encompassing structured residues Ala 9 -Trp3 9 are given. Numbers in parentheses indicate number of restraints.

Root mean square deviations from experimental dihedral restraints
All ( nature of ␣1 (Ala 2 , His 5 , and Val 8 line the apolar face, and Asp 3 , Gln 6 , and Ala 7 populate the polar helix face; Fig. 5). In summary, an extensive and consistent body of evidence leads us to propose that the prevalent MCoA sensitivity of CPT1A is determined by the relative population ratio of inhibitory N␣ to noninhibitory N␤ structural states. OMM-dependent Regulation of CPT1A-The N␣ 7 N␤ switch may also explain the well established dependence of MCoA sensitivity on OMM lipid composition and lipid molecular order/fluidity (5-7). As alluded to earlier, based on the requirement of a strongly curved micelle folding scaffold for the formation of N␤, this state appears best suited to interact solely with the CD as illustrated in Fig. 7A. By contrast, N␣ formation requires a less curved surface, and this conformation could be promoted when N is able to bind to the near planar OMM surface or to the OMM and CD simultaneously (Fig. 7C). Thus, a high N␣-OMM affinity and, consequently, the facilitation of the N␤ 3 N␣ transition will make CPT1A more prone to inhibition by MCoA. Conversely, in case of a lowered N␣-OMM affinity, the reverse N␣ 3 N␤ transition becomes more probable, thereby decreasing MCoA sensitivity. We therefore propose that the N␣:N␤ balance not only determines the prevalent MCoA sensitivity of CPT1A but simultaneously couples MCoA sensitivity to the properties of the OMM. Structurally, this translates into an inhibitory OMM⅐N␣⅐MCoA⅐CD and a catalytically active N␤⅐CD⅐LCFA-CoA⅐carnitine complex (Fig. 7, A and C).
The switch mechanism between the two states of the N-terminal domain of CPT1A provides a regulatory mechanism that enables the enzyme to respond to different metabolic states at multiple levels of organization. Whereas changes in MCoA concentration reflect short term substrate availability and lipogenic enzyme activity, changes in membrane properties derive from the longer term metabolic state (51). This includes changes in lipid composition that occur in response to changes in hormonal balance such as the degree of unsaturation of acyl group composition of phospholipids, the identity of phospholipid headgroup, and the cholesterol/phospholipid ratio of the membrane (5,51). Finally, we note that contact sites between the outer and inner membrane of mitochondria, which exhibit a high degree of membrane curvature (52,53), represent an environment where the kinetics of MCoA inhibition with respect to LCFA-CoA shifts from a noncompetitive to a competitive mode (11). Membrane curvature in unilamellar vesicles is limited to a radius of ϳ105 Å (54), whereas N␣ 7 N␤ switching takes place at a radius of ϳ17 Å (see above). It follows that the high degree of membrane curvature at contact sites may disturb optimal OMM-N␣-MCoA-CD associations (Fig.  7C). Therefore, this may represent a further element to the CPT1A regulatory mechanism, as it will make the relative N␣/N␤ ratio also dependent on the degree of membrane curvature of its microenvironment.
Model of the N-CD Interaction-To provide further insight into the structural basis of CPT1A regulation and to illustrate our proposed CPT1A regulatory mechanism, we performed computations that model the docking of N␤ and N␣ to a homology model of the CD. Two pertinent N␤-CD docking results placed N␤ near the carnitine-and CoA-binding site within the CD, respectively (Fig. 7, A and B). Upon inserting MCoA into the CD structure, according to the mode of CoA binding observed in carnitine acetyltransferase (20), we noted that a displacement of carnitine must take place to accommodate the malonyl carboxylate (Figs. 7A and 1B). This binding mode is supported by the observation that when the sulfhydryl group of CoA becomes oxidized to sulfonic acid, it occupies the carnitine carboxylate-binding pocket (20). It also explains the competitive relationship between MCoA and carnitine during catalysis, e.g. MCoA increases K m for carnitine (12,55). It is therefore evident that the binding of MCoA and carnitine/ PCoA is structurally distinct at the carnitine-binding pocket. This difference may give rise to an allosteric conformational  change in the CD that increases the N␤ binding surface curvature, thus triggering an N␤ 3 N␣ switch. Binding of N␤ near the carnitine-binding pocket (Fig. 7A) may facilitate such a mechanism. Alternative N␤-CD associations, for example near the CoA-binding site (Fig. 7B), are also possible. However, because MCoA inhibitor and PCoA substrate share the CoA moiety and differ chemically only near the active site (Fig. 1B), binding of N␤ next to CoA appears to be less suitable to differentiate between inhibitor and substrate binding. An alternative to an allosteric CD conformational change would be a second MCoA-binding site (12) that could also trigger an N␤ 3 N␣ switch if it were to overlap with the N␤-binding site.
Docking of N␣ to the CD domain was complicated by the absence of the OMM in our model but facilitated by a structural constraint derived from complementary mutagenesis. The E26K point mutation decreases MCoA sensitivity of CPT1A, but K561E can compensate for this mutation (12) indicative of the presence of a Glu 26 -Lys 561 electrostatic interaction in the N␣⅐CD complex. In all of our constructed CD models, Lys 561 pointed away from the enzyme surface and into bulk solvent, disfavoring a docking of Lys 561 with N embedded in the CD surface. However, in the context of an OMM-bound N␣, such an interaction was readily obtainable (Fig. 7C). Because of the lowered dielectric constant of the membrane-water interface compared with bulk water, a membrane-proximal salt bridge would also be stronger than a corresponding solvent-exposed interaction (56). When enforcing an electrostatic Glu 26 -Lys 561 interaction in our models, it was found that N␣, and specifically Glu 3 , has access to positive side chain charges on the CD or to the NH 2 group of the adenosine moiety of CoA (Figs. 7C and 1B). When evaluated within the context of the micelles used in this study, the NMR spectral properties of both N␣ and N␤ were essentially unperturbed by the presence of MCoA (supplemental Fig. 4). However, in the absence of a defined CDbound MCoA conformation, this result has limited significance. We also note that the commencement of native CPT1A with acetylated Ala 2 (49) avoids competition for Glu 3 with an N-terminal positive charge. In summary, a destabilization of the N␤ state effected by MCoA may promote CD inhibition through either OMM-N␣-CD or OMM-N␣-MCoA-CD interactions. Based on our structural OMM⅐N␣⅐MCoA⅐CD model, MCoA-dependent inhibitory actions are explained by a blockage of substrate access to the CD by N␣. In support of this model, kinetics of MCoA inhibition with respect to LCFA-CoA are indeed found to be noncompetitive for the protein within the bulk OMM fraction (11).
Conclusions-The pivotal metabolic role of CPT1A is reflected in the intricate structural properties of its relatively FIGURE 7. Models of N␤-CD and N␣-CD associations. A and B, models obtained by docking N␤ to a homology model of the CD using the program ClusPro 2.0 (38). A places the apex of N␤ over the helix formed by Asp 654 -Leu 667 . Arg 655 of this helix contributes to carnitine binding and is also positioned to interact with MCoA as depicted by dashed lines. B docks N␤ near the CoA-binding site. This model localizes N␤ closer to CD mutations that affect MCoA sensitivity (e.g. Met 593 (63)), albeit this may merely reflect their proximity to the CoA-binding site to which MCoA can bind as well. C, model obtained by manually docking N␣ to a homology model of the CD. A Glu 26 -Lys 561 salt bridge was enforced. N␣ was oriented to permit Glu 3 -Lys 560 , Glu 3 -MCoA(NH 2 ), and Lys 41 -MCoA (3Ј-phosphate) interactions. The outer mitochondrial membrane is expected at the bottom of the panel, providing a binding surface for N␣ and parts of the CD. N␣ is shown in red, the N and C domains, making up the catalytic domain (CD), are shown in blue and gray, respectively. Bound MCoA and selected amino acid side chains are illustrated in ball-and-stick representation. DECEMBER 9, 2011 • VOLUME 286 • NUMBER 49 short N-terminal regulatory domain detected in the presence of micellar folding scaffolds. The properties of N enable an amphiphilic, environment-dependent switch that permits interactions with the CD as well as the OMM. This allows N to integrate information concerning location (membrane curvature), long term metabolic state (membrane fluidity and composition), as well as short term metabolic state (MCoA concentration), providing the structural basis for the molecular "memory" of CPT1A (57). Based on the regulated interconversion of N between different environments, its proposed role is etiologically similar to amphiphilic helices in amphotropic proteins (58). Aside from prominent roles in membrane curvature, sensing, and induction (58,59), amphiphilic helices can also regulate enzyme activity by responding to ligands or to covalent modifications (58). In comparison, however, the complexity and sophistication of the integrative regulatory mechanism evolved for CPT1A appears unprecedented.