Self-association of Transmembrane Domain 2 (TM2), but Not TM1, in Carnitine Palmitoyltransferase 1A

EXPERIMENTAL PROCEDURES

Preparation of Rat Liver Mitochondria—Mitochondria were prepared from livers of fed male Wistar rats. The liver was rapidly excised from the rat and placed in 20 ml of ice-cold homogenization medium containing 300 mm sucrose, 5 mm Tris-HCl, 1 mm EDTA (pH 7.4). All procedures were carried out at 4 °C. The liver was minced using scissors and washed 2–3 times in ice-cold medium. The minced liver was homogenized using a Porter-Elvenheim Teflon glass homogenizer. The homogenates were centrifuged at 500 × g for 10 min to remove debris. The supernatant was then decanted and centrifuged at 12,000 × g for 10 min. The resultant crude mitochondrial fraction was collected and resuspended in a medium containing 300 mm sucrose, 5 mm Tris, 1 mm EDTA (pH 7.4), and gently resuspended. After this, the mitochondria were sedimented by centrifuging at 9000 × g for 10 min to give a mitochondrial fraction. For rapid preparation of the mitochondria, the second centrifugation step was omitted.

Preparation of P. pastoris Mitochondria—Yeast cultures were grown for 24 h at 30 °C with shaking at 200 rpm. All initial steps were carried out at room temperature. Cells were collected by centrifugation at 4000 × g for 5 min, and the resultant pellet was weighed. The appropriate amounts of (2 ml g^-1 yeast) and dithiothreitol (3.084 mg g^-1 yeast) were added. The mixture was incubated at 30 °C for 10–15 min with shaking at 150 rpm. The cells were then sedimented by centrifugation as before and washed in 1.2 m sorbitol/20 mm potassium phosphate buffer (pH 7.4). This pellet was resuspended in 30 ml of 1.2 m sorbitol, 20 mm potassium phosphate, and zymolase was added (3.4 mg g^-1). The cells were incubated at 30 °C with shaking (150 rpm) for 20 min. All subsequent steps were carried out at 4 °C. An equal volume of mannitol buffer (0.6 m mannitol, 1 mm KCl, 10 mm MES, and 0.5 mm EDTA) was added containing a mixture of protease inhibitors (Roche Applied Science) as recommended by the manufacturer. Spheroplasts were collected by sedimentation and washed in the mannitol-containing buffered medium. The pellet was then resuspended in 20 ml of the same medium and homogenized using a Porter-Elvenheim Teflon glass homogenizer. Homogenates were centrifuged at 800 × g for 10 min to remove debris. The supernatant was then decanted and further centrifuged at 12,000 × g for 15 min to obtain the mitochondrial fraction. The pellet was washed and resuspended in a medium containing 0.6 m sorbitol and 20 mm Hepes (pH 7.4).

Blue Native PAGE of CPT1A—Samples were prepared and subjected to Blue Native PAGE using the Novex Bis-Tris gel system (Invitrogen). Briefly 100 μg of mitochondrial fraction were collected by centrifugation for 12 min at 12,000 × g in a cooled centrifuge at 4 °C and resuspended in 100 μl of Native sample buffer (Invitrogen) containing 0.5% digitonin. After solubilization for 15 min on ice, the extracts were centrifuged for 30 min at 50,000 × g, and the supernatants were recovered and supplemented with 0.5 μl of G-250 additive (Invitrogen) for each 10 μl of extract before electrophoresis on 4–16% gradient Novex Bis-Tris gels (Invitrogen). The calibration standard kit used was comprised of eight proteins with a molecular mass covering the range between 20 and 1200 kDa.

Western Blot Analysis—Separated proteins were blotted onto polyvinylidene difluoride membranes using Tris-glycine transfer buffer. After transfer, the membrane was washed for 15 min in 8% acetic acid and allowed to air-dry. The membrane was then washed in methanol, and proteins were detected as described previously using the ECL detection system (Pierce). The antibodies used were antipeptide antibodies raised against the Arg⁴²⁸-Lys⁴⁴¹ sequence within the C-segment of rat CPT1A (19).

TOXCAT Assay and Construction of Chimera—The self-association of rCPT1A transmembrane domains in a natural membrane was studied using the TOXCAT assay, the details of which have been described previously (18). Briefly, TOXCAT employs a chimeric protein in which the α-helical TM domain of interest is inserted between the N-terminal DNA binding domain of ToxR, a dimerization-dependent transcriptional activator, and maltose-binding protein, a monomeric periplasmic anchor protein. The fusion protein is constitutively expressed in E. coli together with a chloramphenicol acetyltransferase (CAT) reporter gene under the control of a ToxR-responsive ctx promoter. Oligomerization of the TM domains within the bacterial inner membrane results in oligomerization of the ToxR domain, transcriptional activation of the ctx promoter, and CAT expression. The amount of CAT expressed in this system is proportional to the strength of oligomerization of the TM domains. The expression vectors (pccKAN, pccGpA-wt, and pccGpA-G83I) and E. coli strain NT326 were kindly provided by Prof. Donald M. Engelman. TOXCAT chimera were constructed according to a reported protocol (18) and expressed in E. coli. Before performing the assay, membrane insertion of all constructs was confirmed using sodium hydroxide washes (20), and correct orientation of the TOXCAT constructs in the membrane was confirmed through protease sensitivity in a spheroplast assay (18). Similar expression levels for all constructs were confirmed via Western analysis against the maltose-binding protein domain before performing CAT assays. The CAT assays were performed using the FAST CAT kit (Molecular Probes and Invitrogen).

Peptide Synthesis and Purification—A peptide corresponding to the second TM domain (TM2) of CPT1A was synthesized at the Keck Facility, Yale University using Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry. The sequence of the TM2 peptide was COCH₃-KKNIVSGVLFGTGLWVAVIMT-MRK-CONH₂, containing residues Lys¹⁰²-Arg¹²³ as well as non-native lysine residues (to aid solubility) and end-caps on the C and N termini (M_r = 2689 Da). The peptide was purified by reversed-phase high performance liquid chromatography using a linear acetonitrile gradient from 30 to 100% containing 0.1% trifluoroacetic acid on a Phenomenex C4 semipreparative column. The purity of pooled peptide fractions was confirmed by electrospray ionization time-of-flight mass spectroscopy (ESI-TOF-MS microTOF, Bruker) before subsequent lyophilization. Peptides were stored as dry powders until use.

Circular Dichroism (CD)—CD spectra were measured using a Jasco J715 spectropolarimeter (Jasco UK, Great Dunmow, UK) and 1.0-mm path-length quartz cuvettes (Starna, Optiglass Ltd, Hainault, UK). All spectra were recorded from 190 to 260 nm (data below 200 nm are not shown due to high noise of light scattering) using a 2.0-nm spectral bandwidth, 0.2-nm step resolution, 100-nm min^-1 scanning speed, and 1-s response time. The TM2 peptide was prepared in 50 mm sodium phosphate buffer (pH 7) containing 10 mm SDS or 46 mm dodecylphosphocholine (DPC) (Avanti Polar Lipids, Alabaster AL) and 100 mm NaCl to a final peptide concentration of 40 μm. CD spectra were collected at room temperature.

Peptide Cross-linking in Detergents—Cross-linking reactions were carried out for 20 μm solutions of the TM2 peptide dissolved in DPC micelles at peptide:micelle ratios ranging from 1 to 20 (corresponding to DPC concentrations ranging from 2.7 to 24 mm). All samples were prepared in 20 mm sodium phosphate buffer and 150 mm NaCl (pH 8). Bis[sulfosuccinimidyl]suberate (BS³, Pierce) was used to cross-link the peptide in solution via primary amine groups according to the manufacturer's protocol. The cross-linking reaction was terminated after 30 min by the addition of 1 m Tris-HCl (pH 8). Uncross-linked and SDS cross-linked samples were prepared as controls. All samples were analyzed by gel electrophoresis, and peptides were visualized on gels by staining either with Coomassie Blue or with silver nitrate.

Analytical Ultracentrifugation—Sedimentation velocity measurements were carried out on a Beckman XL-1/A analytical ultracentrifuge (Beckman Coulter, Fullerton CA) housed in the Department of Biological Sciences, University of Birmingham, Birmingham, UK. TM2 peptide samples were prepared in 50 mm Tris-HCl (pH 7.5), containing 15 mm DPC (Avanti Polar Lipids), 100 mm NaCl, and 52.5% D₂O (Cambridge Isotope Laboratories, Andover MA) to match the buoyant density of the detergent. When the solvent matches the buoyant density of the detergent micelles, the only contribution to the buoyant molecular weight is from the peptide, as described (21).

Data were collected using absorbance optics set to 280 nm at two peptide concentrations (142 and 62 μm) using a double-sector centerpiece at a speed of 40,000 rpm and a temperature of 25 °C. 400 scans were recorded for each sample, with 50 s between each scan. The moving boundary was monitored by repetitive radial scanning at a constant step size of 0.003 cm at 280 nm using a UV absorption optical system. Fitting of the resulting profiles was achieved using SEDFIT (22) to generate a continuous sedimentation coefficient distribution, which was subsequently converted to a molecular mass distribution using a peptide monomeric molecular mass of 2689 Da, a buffer density of 1.05971 g ml^-1, a buffer viscosity of 1.0267 centipoise, and a partial specific volume of 0.7792 ml g^-1 (calculated using SEDNTERP).

Computational Searches Using CHI—Structural calculations were performed using the CNS searching of helix interactions (CHI) package, the details of which have been described previously (23–25), on an 8-node dual 2.66-GHz Xenon processor Linux cluster (Streamline Computing, Warwick). CHI was used to create models of rCPT1A TM2 homotrimers and homohexamers. Using CHI, either three (for trimer) or six (for hexamer) canonical α-helices containing residues Lys¹⁰²-Arg¹²³ of TM2 were built. The starting geometries incorporated both right-handed (-25°) and left-handed (25°) crossing angles and an axis-to-axis distance between the helices of 10.4 Å. In a search of approximately symmetrical interactions, the helices were simultaneously rotated about their central axis in 45° increments from 0° to 360°. After each rotation, molecular dynamics simulations were performed using simulated annealing of atomic coordinates. Four different molecular dynamics simulations were performed for each starting geometry, and energy minimization of structures was carried out both before and after molecular dynamics simulations. Groups of structures with a backbone root-mean-square deviation of ≤1 Å were placed in clusters of 10 or more members followed by a calculation of an average structure for each cluster and energy minimization.

RESULTS

DISCUSSION

Oligomer Formation in Natural Membranes—The involvement of the TM domains of CPT1A in the regulation of its activity and its sensitivity to malonyl-CoA has been strongly inferred for some time through numerous findings on the modulation of the interaction between the N- and C-terminal segments (for review, see Ref 40). This interaction between the N- and C-terminal segments of the molecule has now been modeled using in silico techniques based on homology with other transferases for which x-ray crystal structures have been determined (13). Moreover, we have shown that such a role for TM1-TM2 interactions occurs very much in the context of protein-lipid interactions, making CPT1A sensitive to changes in lipid composition that accompany changes in diet and pathophysiological state (e.g. diabetes, fasting) (15). Previous studies have also raised the possibility of an additional role for the TM domains, namely that of driving the oligomerization of CPT1A (16). In those studies, analysis of the native protein either in rat liver mitochondria or that expressed heterologously in yeast yielded either trimeric or hexameric oligomers in a manner that suggested that the fundamental oligomeric state is the trimer and that the hexamer results from the dimerization of the trimer (16). In general, trimers were observed after gel filtration, whereas hexamers were observed after Blue Native gel electrophoresis (16). The CPT1A protein N-terminally truncated at residue 82 (i.e. lacking TM1 but retaining most of the loop region and TM2 (9)) was similarly found to yield only trimers. Although these data indicated that TM2 was sufficient for oligomerization to occur, they did not exclude the possibility that TM1 plays an additional/complementary role in oligomerization and may be interpreted as indicating that TM1 may determine the order of oligomerization. Therefore, in our studies, in addition to characterization of the oligomeric characteristics of the native rCPT1A and truncation/deletion mutants, we have quantified the oligomerization propensities of both TM1 and TM2 in isolation using the TOXCAT assay and have explored in more detail the biophysical properties of TM2 (which is the only one to give a substantial signal in TOXCAT) with respect to its oligomerization properties.

Importantly, our experiments on CPT1A in rat liver mitochondria and the native protein expressed heterologously in yeast mitochondria showed that both trimers and hexamers can be detected using the same technique (Blue-Native gel electrophoresis). Similar observations were made for all the truncation and deletion mutants tested when expressed in yeast (Fig. 1b). In particular, contrary to previous observations (16), we found that the (Δ1–82) CPT1A truncation mutant also exists in a mixture of the two oligomeric states (Fig. 1b). These observations suggested that the two oligomeric forms may exist in equilibrium in vivo. However, our time-course experiments on rapidly prepared rat liver mitochondria suggest that in vivo CPT1A is likely to occur as a hexamer which de-oligomerizes rapidly during preparation of subcellular fractions and especially with storage of the mitochondria, particularly at 37 °C. The observations obtained on rapidly fractionated rat liver tissue indicate that the trimer (although stabilized by non-covalent interactions that we have modeled; see below) may not be a physiologically relevant entity but may arise as an artifact during the necessarily lengthy preparation of purified mitochondrial fractions by differential centrifugation (this study) or on self-forming gradients (16).

It has also been observed that there was no difference in the order of oligomerization of CPT1A in rat liver mitochondria isolated from animals in different physiological states (fasting, diabetic) known to result in a wide range of sensitivities of the enzyme to malonyl-CoA inhibition (16). Therefore, in this study we have tested this observation directly by studying the oligomerization state of mutants that contained deletion mutations within the N-terminal 47 amino acid residues that have previously been shown (see e.g. Ref. 26) to display a range of several orders of magnitude in their malonyl-CoA sensitivities. The data in Fig. 1b show that we always observed similar proportions of trimer and hexamer irrespective of which N-terminal mutant was studied. This strengthens the evidence that malonyl-CoA sensitivity is independent of the quaternary structure of CPT1A and is instead determined by intramolecular (N-C) interactions (40) to which TM1-TM2 (hetero-oligomeric) interactions are likely to be central.

Self-association of TM2, but Not TM1, Reflects the Oligomerization of the Native Protein—The above interpretations are supported by the biophysical experiments we performed on the peptide corresponding to TM2 in vitro. The experimental demonstration using TOXCAT that only TM2 has homo-oligomerization potential enabled us to focus on the biophysical properties of a peptide corresponding to this TM. When the TM2 peptide was dissolved in DPC micelles and cross-linked with BS³, oligomers corresponding to the hexamer were observed, suggesting that the TM2 sequence has inherent primary and secondary structural properties that favor hexamerization. The fact that the hexamer also appears to be a favored oligomeric state in vivo suggests that these interactions are also important in the context of the full-length protein when this is stabilized within a lipid bilayer. De-oligomerization to the trimeric form occurs rapidly during preparation and storage of mitochondria (when disruption of contacts with other intracellular systems, loss of contact site density (20) and changes in outer membrane lipid composition because of phospholipase activity are known to occur). It is likely that de-oligomerization to the trimeric form of the native rCPT1A protein is also favored during digitonin- or Triton X-100 solubilization of mitochondria, in preparation for BN-PAGE or gel filtration (16). In this context, it is of interest to note that when the size of the molecular entity C-terminal of TM2 was much smaller than that of native CPT1A (e.g. the truncated CPT1A-dihydrofolate reductase chimera, CPT1A-(97–147) dihydrofolate reductase), only the respective hexamer was observed (16). This suggests that whereas TM2-driven oligomerization results in the hexamer as a default quaternary structure for the native protein, the large C-terminal domain of the native protein imposes intermolecular constraints within that oligomeric structure that are overcome only within the mitochondrial outer membrane in vivo and which result in the relatively rapid deoligomerization to the trimer.

Such close correspondence between the oligomerization behavior of rCPT1A in vivo and those of TM2 in vitro provide further evidence that TM2 plays a central role in the oligomerization of the native protein, as suggested by the strong propensity of this TM to oligomerize in E. coli inner membrane (in the TOXCAT assay).

Molecular Modeling of the Role of the GXXXG(A) Motifs in TM2 Oligomerization—As discussed above, the hexamer of rCPT1A is likely to be the most important oligomeric form of rCPT1A in vivo, although the inability to prepare mitochondria sufficiently rapidly from intact rat liver does not allow a definitive exclusion of the co-existence of the trimer. Certainly, the hexameric form of TM2 is the predominant species when the corresponding peptide is studied in DPC-micelle solution; other oligomeric states are observed only at very low or very high M:P ratios. The molecular models shown in Fig. 5 illustrate predicted conformations that may be adopted by the TM2 trimers and hexamers. Importantly, Fig. 5d highlights the possible role of the GXXXG and GXXXA motifs present within TM2 (and conserved in all CPT1A proteins sequenced to date) in stabilizing the hexameric structure. In both of these motifs, Gly or other small residues (e.g. Ala) are found four residues apart (one turn of the helix) and, therefore, appear on the same helical face. The presence of a “patch” of these small residues allows the helices to pack tightly together, and this close-packing stabilizes helix-helix interactions (36, 41, 42).

The angular separation of the GXXXG and GXXXA motifs within the predicted helical structure of TM2 (Fig. 6b) is approximately equal to the apex angles of a hexagon (120°). Therefore, in the hexamer both motifs could theoretically pack against adjacent TM2 helices, hence favoring this oligomeric form. However, in the native protein, intermolecular steric hindrance may destabilize this tertiary structure such that during preparation of mitochondria the trimer is formed (Fig. 1c). This is also evidenced by data showing that when a smaller molecular entity replaced the bulky C-terminal segment of native rCPT1A with the structure of dihydrofolate reductase, only hexamers were observed (16). Although further work is required to confirm the mechanism of association, GXXXG and GXXXA motifs are well known to contribute to TM interactions.

The trimer, although appearing not to be the physiologically relevant quaternary structure of rCPT1A, is predicted to be stabilized by the possible close packing of at least one of the GXXXGorGXXXA motifs and by interhelical hydrogen bonding, as shown in Figs. 5, a and b. The relative disposition of the GXXXGorGXXXA motifs within the predicted α-helical structure of TM2 is incompatible with favorable packing of both of these motifs simultaneously in the formation of homo-trimers, which would require an angle of ∼60° (Fig. 6b).

Functional Significance of CPT1A Oligomerization—Previous data (16) as well as data from the present study raise the question as to the function of CPT1A hexamer formation. It has been suggested that oligomerization (to form trimers or a dimer of trimers) may be a mechanism for the formation of a channel in the mitochondrial outer membrane through which the acylcarnitine product of the reaction catalyzed by CPT1 may gain access to the intermembrane space (16). The molecular models in Fig. 5 suggest that the hexamer (but not the trimer) is capable of forming a pore of significant dimensions; however, validation of its possible function as a channel for acylcarnitines awaits further study. In this context, two considerations are pertinent. First, we have previously shown by performing functional target size analysis on CPT1A in whole rat liver mitochondrial and purified outer membranes that the functional size of rCPT1A (i.e. the size of the molecular entity required for catalytic activity) is the same as that of the monomeric molecular mass (∼88 kDa) of the protein (43). Therefore, in the rat liver mitochondrial outer membrane each monomer within the hexameric form would function as an independent entity in terms of its catalytic function. Second, if as suggested (16) the putative channel formed by the oligomerization of CPT1A were necessitated by the inability of porin (VDAC1) to allow acylcarnitine esters access into the intermembrane space (a prospect raised previously (44)), it would need to be suggested that either this effect of porin is unidirectional or that transport of acylcarnitines out of mitochondria (which is considerable) also occurs necessarily through the putative pore/channel created by CPT1A oligomerization. These are now testable hypotheses, and the present studies suggesting that stabilization of TM2 hexamerization by GXXXG and GXXXA interactions may be central to rCPT1A oligomerization will enable further studies on CPT1A mutants that are unable to oligomerize. These should be useful in obtaining information regarding the function of CPT1A oligomerization with respect to long-chain fatty acyl moiety entry into intact mitochondria.