Biochemical Journal

Research article

A CD study of uncoupling protein-1 and its transmembrane and matrix-loop domains

Masoud Jelokhani-Niaraki, Marina V. Ivanova, Bonnie L. McIntyre, Cheryl L. Newman, Fern R. McSorley, Elizabeth K. Young, Matthew D. Smith


Conformations of the prototypic UCP-1 (uncoupling protein-1) and its TM (transmembrane) and ML (matrix-loop) domains were studied by CD spectroscopy. Recombinant, untagged mouse UCP-1 and a hexahistidine-tagged version of the protein were obtained in high purity following their overexpression in Escherichia coli. The TM and ML domains of hamster UCP-1 were chemically synthesized. Conformations of both recombinant UCP-1 proteins were dominantly helical (40–50%) in digitonin micelles. Binding of the purine nucleotides GDP and GTP to UCP-1, detected in the near-UV CD region, supported the existence of the functional form of the protein in digitonin micelles. All individual TM and ML peptides, except the third ML domain, adopted helical structures in aqueous trifluoroethanol, which implies that, in addition to six TM segments, at least two of the ML domains of the UCP-1 can form helical structures in membrane interface regions. TM and ML domains interacted with vesicles composed of the main phospholipids of the inner membrane of mitochondria, phosphatidylcholine, phosphatidylethanolamine and cardiolipin, to adopt dominantly β- and/or unordered conformations. Mixtures of UCP-1 peptide domains spontaneously associated in aqueous, phospholipid vesicles and digitonin micelle environments to form ordered conformations, which exhibited common features with the conformations of the full-length proteins. Thermal denaturations of UCP-1 and its nine-peptide-domain assembly in digitonin were co-operative but not reversible. Assembly of six TM domains in lipid bilayers formed ion-conducting units with possible helical bundle conformations. Consequently, covalent connection between peptide domains, tight domain interactions and TM potential are essential for the formation of the functional conformation of UCP-1.

  • CD spectroscopy
  • conformational analysis
  • overexpression in Escherichia coli
  • pipette-dipping patch-clamp
  • thermal denaturation
  • transmembrane and matrix-loop peptide domains
  • uncoupling protein-1


Thermogenin or UCP-1 (uncoupling protein-1) is the prototypic member of the mitochondrial inner-membrane UCPs, a subfamily of the large family of mitochondrial integral membrane carrier proteins. UCP-1 which is specifically found in mammalian brown adipocytes, causes thermogenesis in these cells and non-shivering thermogenesis in the whole organism [1,2]. It is widely accepted that thermogenesis is the result of dissipation of heat energy through proton leak across the inner mitochondrial membrane. Proton leak from the intermembrane space to the matrix uncouples the electron transport chain processes from ATP synthesis. Release of heat as the protonophoric function of UCP-1 prevents energy storage as fat and could be considered as an anti-obesity activity [3]. Recently, two UCP homologues, UCP-2 and UCP-3, have been discovered in the mitochondria of various mammalian tissues [4]. Other UCPs, UCP-4 and UCP-5, have been traced in brain tissues and are much less homologous with any of the other UCPs [5]. In contrast with UCP-1, despite possible structural similarities, none of the other four UCPs are believed to cause thermogenesis as their principle physiological role. It seems that the proton transport activity and reduction of the TM (transmembrane) potential are universal features of the biological activity of UCPs [6].

Two general mechanisms have been proposed for proton transfer by UCP-1. One proposes that a proton pathway exists within the UCP-1 molecule [7], while the other suggests that a flip-flop of the carboxylate group of the fatty acids adjacent to the protein molecule facilitates proton flux across the membrane [8]. Purine nucleotides (e.g. ADP, ATP, GDP and GTP) all bind tightly to UCP-1 and inhibit proton transport, and are considered as both inhibitors and regulators of UCP-1 activity [6,7]. In addition to its proton-transport function, it has been found that, similar to other members of the family of mitochondrial anion carriers, UCP-1 can also transport anions such as Cl [9].

Despite the existence of extensive data on the biological function of UCP-1, little data has been documented on its structure. According to hydropathy plots, the UCP-1 monomer is composed of six TM α-helices, and the conformations of its extramembrane segments are not clearly known. Studies using micellar systems suggest that UCP-1 may be able to form dimers, which might be the aggregation state of the protein in membranes [10]. However, the monomeric compared with dimeric state of functional UCP-1 is still debated. There are even less structural data available for other UCPs. To study the conformation and ion-transport properties of uncoupling proteins, we have previously synthesized the six TM domains of human UCP-2 [11]. Among these TM domains, TM2 was able to selectively transport anions such as chloride and could possibly participate in the anion-transport function of the protein.

In the present study, we have expressed recombinant mUCP-1 (mouse UCP-1) in Escherichia coli, and chemically synthesized the six TM and three ML (matrix-loop) segments of the prototypic golden hamster UCP-1 to study the conformation of UCP-1 and its peptide domains in membrane-like environments. For this purpose, we have used CD spectroscopy for conformational analysis of the proteins, individual peptides and peptide assemblies. We have also employed pipette-dipping patch-clamping to examine the ion-conducting properties of peptide assemblies.



Phospholipids, POPC (1-palmitoyl-2-oleoyol-sn-glycero-3-phosphocholine), POPE (1-palmitoyl-2-oleoyol-sn-glycero-3-phosphoethanolamine), POPG [1-palmitoyl-2-oleoyol-sn-glycero-3-(phosphor-rac-1-glycerol); sodium salt], TOCL (1,1′2,2′-tetraoleoyl cardiolipin; sodium salt) and DPhPC (1,2-diphytanoyl-sn-glycero-3-phosphocholine) were from Avanti Polar. Detergents N-lauroylsarcosine (sodium salt) and digitonin were from Sigma and Calbiochem-EMD Biosciences respectively. Fmoc (fluoren-9-ylmethoxycarbonyl)-protected amino acids, coupling reagents and solvents used for peptide synthesis were from either Advanced ChemTech or Novabiochem-EMD Biosciences. Proofstart DNA polymerase (Qiagen) was used for all PCR amplifications. All other chemicals were of high-purity reagent grade and used as received.

UCP-1 constructs

The mUCP-1 cDNA clone (pET-mUCP-1) was a gift of Dr Martin Brand (MRC Dunn Human Nutrition Unit, Cambridge, U.K.). For the purpose of producing a His6 (hexahistidine) affinity-tagged version of mUCP-1, the cDNA was first sub-cloned into pQE30 (Qiagen), and ultimately into pET21B (Novagen). Primer-adapters were used to incorporate BamHI and SacI restriction sites into the ends of the mUCP1 cDNA by PCR, which facilitated cloning into pQE30. The new mUCP-1 construct was again amplified using primer-adapters, and ultimately cloned into the NdeI and SacI sites of pET21B. The final construct, pET21B-TEV-mUCP1His, encodes a recombinant version of UCP-1, denoted UCP-1His6, that includes an N-terminal His6 tag separated from the endogenous mUCP-1 start codon by a TEV (tobacco etch virus) protease recognition site (ENLYFQG) and also includes the endogenous stop codon. The pET-mUCP-1 and pET21B-TEV-mUCP1His6 constructs were introduced into E. coli BL21(DE3) to facilitate overexpression of the corresponding recombinant proteins.

Recombinant UCP-1 and UCP-1His6 expression and purification

Overexpression of recombinant UCP-1 and UCP-1His6 was achieved by induction with 0.4 mM IPTG (isopropyl β-D-thiogalactoside) for 3 h at 37 °C with aeration. Cell pellets containing recombinant UCP-1 were resuspended in TGE (Tris-glucose-EDTA) buffer [25 mM Tris/HCl (pH 8.0), 50 mM D-glucose and 10 mM EDTA], incubated for 20 min with lysozyme (200 μg/ml) and sonicated. The IBs (inclusion bodies) were then collected by centrifugation at 28500 rev./min for 15 min at 4 °C using a JA 30.50 Ti rotor (Beckman Coulter), and washed four times using TGE buffer containing 1% (w/v) Triton X-100. The final IB pellet was resuspended in 1 ml of TGE buffer containing 2 mM DTT (dithiothreitol).

Cell pellets containing recombinant UCP-1His6 were resuspended in extraction buffer [20 mM Tris/HCl (pH 8.0), 500 mM NaCl and 20 mM imidazole], incubated for 20 min with lysozyme (200 μg/ml) and sonicated using a probe-tip sonicator. IB proteins were collected by centrifugation at 40000 rev./min for 15 min at 4 °C in a TLA100.3 rotor (Beckman Coulter). The IBs were washed three times with extraction buffer containing 1% Triton X-100, once with extraction buffer, and finally dissolved in extraction buffer containing 8 M urea and 20 mM imidazole. Protein suspensions were centrifuged [40000 rev./min (rotor TLA100.3; Beckman Coulter) for 15 min at 4 °C] to remove insoluble aggregates, and the supernatant was diluted with extraction buffer such that the final urea concentration was 6 M. UCP-1His6 was then purified using Ni-NTA (Ni2+-nitrilotriacetate) chromatography under denaturing conditions according to the manufacturer's protocol (Novagen). Briefly, urea-solubilized IBs were applied to a column containing Ni-resin. The resin was washed with extraction buffer containing 6 M urea and 40 mM imidazole, and UCP-1His6 was eluted using extraction buffer containing 6 M urea and 500 mM imidazole.

Protein determination and SDS/PAGE analysis

Protein concentrations were determined either using a dot blot technique as described by Ghosh et al. [12] or with the DC RC protein assay kit (Bio-Rad Laboratories) for detergent-containing samples. Proteins were separated using SDS/PAGE in gels containing a 4% stacking gel and a 12% resolving gel, and were visualized using Coomassie Brilliant Blue R-250.

UCP-1 peptide domain constructs

Linear sequences of the TM and ML peptides of the golden hamster UCP-1 (92% sequence homology with mUCP-1) (Table 1) were manually synthesized by solid-phase procedures using Fmoc-chemistry [13]. Briefly, Fmoc-AAc-Wang resin (Advanced ChemTech) was used to initiate the synthesis from the C-terminal amino acid, termed AAc, as described previously [11,13]. Each amino-acid-coupling used HOBt (1-hydroxybenzotriazole), HBTU {N-[(1H-benzotriazol-1yl)(dimethylamino)methylene]-N-methylmethanaminium hexafluorophosphate N-oxide} and DIPEA (N,N-diisopropylethylamine) as coupling reagents in NMP (N-methylpyrrolidone). Fmoc was removed from peptide sequences on resin with 20% piperidine in NMP. Synthesized peptides were simultaneously cleaved and deprotected from the resin, then analysed and purified by RP-HPLC (reversed-phase HPLC) on a Waters 600E HPLC system using a Waters C4 Delta Pak column (300 mm×7.8 mm internal diameter, 15 μm particle size, 10 nm pore size) for purification and Waters C4 Symmetry column (150 mm×4.6 mm internal diameter, 5 μm particle size, 30 nm pore size) for analysis. A ZQ-4000 Micromass/Waters mass spectrometer with an ESI (electrospray ionization) probe was used to measure peptide masses. Peptide concentrations in aqueous stock solutions were determined by amino acid analysis.

View this table:
Table 1 Sequences of TM and ML peptide segments of the golden hamster UCP-1

The underlined sequences represent putative TM segments from the analysis of the hamster UCP-1 sequence (accession number P04575) using the Swiss-Prot database. Golden hamster UCP-1 has a total of 306 amino acids. The total number of amino acids in peptide segments is 276, which is 90% of the number of amino acids in the native protein. All cysteine residues were replaced by a serine residue, with their positions marked with ‘*’. The biological activity of UCP-1 was not affected by replacing cysteine residues with serine [11,27]. The free energy of transfer of putative TM segments (underlined) from water to octanol (ΔGWO), water to POPC bilayer interface (ΔGWB), and the difference between the two (ΔΔGWBO) were calculated on the basis of whole-residue hydrophobicity scales [28]. The free energy of transfer of the peptide bonds to the octanol phase (representing the hydrophobic core of bilayers) used for calculations was −0.4 kcal/mol. The hydrophobic moment (HM) values of the putative segments are shown in parenthesis after the free energy values. All calculations were performed by the Membrane Protein Explorer software (MPEx), provided by Stephen White's group (University of California, Irvine, CA, U.S.A.) at: AA, amino acid; HM, hydrophobic moment; IF-Oct, interfacial to octanol; MW, molecular mass.

Preparation of proteins for CD experiments

Solubilization of UCP-1 was achieved using a modified version of a protocol described previously [14]. Specifically, UCP-1 IBs (∼5 mg) were dissolved in 400 μl of buffer 1 [2% (w/v) sarcosyl, 2 mM DTE (dithioerythritol), 2 mM PMSF and 50 mM ammonium bicarbonate (pH 8.0)], incubated at room temperature (22 °C) for 10 min, and centrifuged at 17000 g for 30 min at 4 °C. The supernatant was diluted 4-fold with buffer 2 [2 mM DTE, 2 mM PMSF and 50 mM ammonium bicarbonate (pH 8.0)], then five times further with buffer 3 [1 mM DTE, 50 mM 2-mercaptoethanol, 0.2% (w/v) digitonin and 100 mM potassium phosphate (pH 8.0)] containing 1% (w/v) sarcosyl. Buffer exchange and protein concentration was achieved using diafiltration. Specifically, the solution was concentrated 3-fold by centrifugation at 5000 g at 4 °C in a centrifugal filter device with a 10 kDa molecular-mass limit (Amicon Ultra 10K; Millipore), then diluted 3-fold with buffer 3 and concentrated 3-fold again. This was repeated three times using buffer 3 then twice using buffer 4 [1 mM DTE, 1 mM EDTA and 10 mM potassium phosphate (pH 8.0)]. To remove any remaining sarcosyl, 2 ml of the UCP-1 solution was incubated overnight with 1.64 g of Dowex mesh 16–50 (1.6 g of the Cl form and 0.04 g of the OH form; Dow Chemical Company) on a rotating mixer at 4 °C. The protein-containing solution was removed and the ion-exchange beads were washed twice with 2 ml of buffer 4, which was combined with the UCP-1 solution.

Purified UCP-1His6 was precipitated using TCA (trichloroacetic acid), washed twice with 0.5% (w/v) TCA and once with water. The precipitated protein was resuspended in 400 μl of buffer A [2% (w/v) sarcosyl, 2 mM DTE and 50 mM ammonium bicarbonate (pH 8.0)], and incubated for 1 h at room temperature. The solubilized protein was diluted 4-fold with buffer B [2 mM DTE and 50 mM ammonium bicarbonate (pH 8.0)], then five times further with buffer 3 containing 1% (w/v) sarcosyl. Buffer exchange and protein concentration was achieved as described for UCP-1, except that buffer D [1 mM DTE, 1 mM EDTA and 100 mM potassium phosphate (pH 8.0)] was used in place of buffer 4. UCP-1His6 protein solutions were diluted before CD analysis such that the final buffer composition was 1 mM DTE, 5 mM 2-mercaptoethanol, 0.02% (w/v) digitonin, 0.9 mM EDTA and either 10, 50 or 100 mM potassium phosphate (pH 8.0).

Preparation of unilamellar vesicles and peptides for CD experiments

Two phospholipid vesicle systems were used for spectroscopic studies. The first three-lipid system was composed of POPC, POPE and POPG phospholipids to model negatively charged cytoplasmic membranes. The second four-lipid system contained PC (phosphatidylcholine), PE (phosphatidylethanolamine), PG (phosphatidylglycerol) and CL (cardiolipin) to model the inner mitochondrial membranes.

Chloroform solutions of the lipids, POPC/POPE/POPG (5:4:1 molar ratio) and POPC/POPE/TOCL/POPG (4:3.5:2:0.5 molar ratio) were dried under a mild nitrogen flush in a round-bottomed flask to form a thin film. The film was dried overnight under vacuum, and then hydrated with buffer [10 mM Tris and 100 mM NaF (pH 7.4)]. The resultant multilamellar vesicles were then freeze–thawed a few times and extruded through a 100 nm filter in a LiposoFast apparatus (Avestin) [15]. The prepared unilamellar vesicles were less than 100 nm in diameter and stable in the dark for several days at 4 °C.

Peptide samples in lipid vesicles and micelles were prepared by the addition of appropriate amounts of peptide stock solutions in water (100 μM) to vesicle or micelle dispersions in buffer, and incubated at room temperature before CD measurements.

CD measurements and analysis

CD spectra were measured on an Aviv 215 spectropolarimeter (Aviv Biomedical). Ellipticities are reported as mean residue ellipticity. The measurements were carried out in rectangular quartz cells with 0.1 cm (far-UV) and 0.5 cm (near-UV) pathlengths. The reported spectra were measured at 0.5 nm/s scanning speed and are an average of four scans. Sodium fluoride was used in place of sodium chloride in buffer (Tris buffer as described above) to reduce the high noise levels of chloride ions below 200 nm. CD spectra were analysed further for estimation of secondary structure content by the program CDSSTR, which uses a singular value deconvolution algorithm [16]. The analysis was based on a set of 48 reference proteins and performed on the Dichroweb website [17].

Patch-clamp sample preparation and measurements

Pipette-dipping patch-clamp experiments were performed as described previously [18]. Patch pipettes were of thin hard-glass borosilicate type (1.16 mm internal diameter) (Warner Instruments). The pipettes were pulled through a two-pull method to give approx. inner diameters of 1 μm, and used without being heat-polished or siliconized. The pipette-tip resistance in electrolyte solutions was approx. 20 MΩ. Peptide mixture stock solutions were prepared as 60 μM aqueous solutions (10 μM for each peptide). The peptide mixture final concentration in patch pipettes was 500 nM.

To form the monolayer, chloroform solutions of DPhPC lipids were mildly evaporated and the residual lipid was redissolved in pentane before being spread on aqueous surfaces (1–2 μl of 10 mg/ml pentane solutions) in 3.5 cm diameter Petri dishes. Seals of 3–10 GΩ were formed at the tip of pipettes. Experiments were also performed in the absence of peptides to rule out artefacts due to membrane breakdown or non-specific leakage through the seals.

Single-channel currents were amplified by an Axopatch 200B (Axon Instruments) patch/whole cell microelectrode amplifier. Digidata 1322A (Axon Instruments), a low-noise digitizer, was used for data acquisition. Data were analysed by the pCLAMP 9 software (Axon Instruments) using a 1 kHz low-pass filter.


UCP-1 and UCP-1His6 expression and purification

Recombinant UCPs have been overexpressed as either bacterial IB proteins [14,19] or as His6-tagged proteins in yeast mitochondria [20]. In the present study, recombinant versions of mUCP-1 were prepared for conformational analysis by CD. Both versions of the UCP-1 were produced in E. coli as IB proteins, a characteristic which enabled a high degree of purity to be achieved, even in the absence of an affinity tag (Figure 1A). Indeed, after washing with Triton X-100, the UCP-1 IB protein (∼33 kDa) was of adequately high purity for conformational analysis (Figure 1A, lanes 1 and 2). The UCP-1His6 protein (∼33 kDa) was also expressed exclusively in IBs; however, following solubilization in urea, IMAC (immobilized metal-ion-affinity chromatography) purification under denaturing conditions yielded protein that was markedly more pure than that in the washed IBs (Figure 1B, compare lanes 1 and 2).

Figure 1 Recombinant mUCP1 expression and purification

Recombinant mUCP1 proteins expressed in E. coli were resolved using SDS/PAGE (12% gels) and stained with Coomassie Brilliant Blue. (A) Triton X-100-washed mUCP1 IBs (IB, lane 1, 10 μg) and digitonin-solubilized mUCP1 (Dig, lane 2, 10 μg). (B) Triton X-100-washed mUCP1His6 inclusion bodies (IB, lane 1, 10 μg), urea-solubilized mUCP1His6 purified by IMAC (IMAC, lane 2, 5 μg), digitonin-solubilized mUCP1His6 (Dig, lane 3, 5 μg) and dodecyl maltoside-solubilized mUCP1His6 (DDM, lane 4, 5 μg). Molecular-mass markers (kD) are shown (lane labelled M).

Both versions of the protein were solubilized using 2% sarcosyl, which was subsequently substituted with digitonin using diafiltration and ion-exchange resin. SDS/PAGE analysis revealed that both versions of recombinant UCP-1 remained soluble following this treatment (Figure 1A, lane 2 and Figure 1B, lanes 2 and 3), and were of adequately high purity for analysis using CD.

It is noteworthy that, to our knowledge, the present study represents the first example of the successful expression, purification and structural analysis of a His-tagged version of a UCP produced in E. coli.

Conformation of UCP-1 and UCP-1His6 in micelles

Information on the conformation of UCP-1 has been limited to IR and CD studies in micelles [20,21] and a modelling study [22]. A high-resolution crystal structure of bovine mitochondrial AAC (ADP/ATP carrier) in complex with its inhibitor carboxyatractyloside has provided a structural basis for other mitochondrial transporters [23]. In the AAC protein structure, in addition to six TM helices, three shorter helices were observed in the ML regions. Further crystallographic data supported a lipid-mediated interaction between the AAC monomeric units, which suggests the existence of a dimerization interface between monomers [24]. Previous biophysical and biochemical data suggested a dimeric functional unit for AACs [25]. However, a recent study of the yeast AAC protein suggests that the functional unit of the protein is monomeric [26]. Comparably, the association state (monomeric compared with dimeric) of the functional unit of UCP-1 has not yet been fully established. Mitochondrial AAC proteins and UCP-1 monomers have comparable molecular masses (∼33 kDa), both bind to nucleotides, and experimental data supports similarities in the structure of their functional form. The structural similarity between the two proteins has been supported by attempts at sequence-dependent computational homology modelling of UCP-1 on the basis of the known AAC structure (M.V. Ivanova and M. Jelokhani-Niaraki, unpublished work, and [20]).

In the AAC crystal structure, 67% of residues form helical structures, of which 55% are in TM and 12% are in ML domains. Hydrophobic scales applied to the UCP-1 sequences estimate that at least 43% of residues can be considered as TM segments. In an IR study on the secondary structure of hamster UCP-1 in RTX-100 (a cyclohexane-containing derivative of Triton X-100) micelles the helix and β-structure were estimated to be 50% and 28–30% respectively [21]. In a more recent CD study, the helical content of rat UCP-1, expressed in yeast and carrying a His6-affinity tag on its C-terminus, in DDM (n-dodecyl β-D-maltoside) micelles was estimated to be 68% [20]. The high helical content of the CD spectrum prompted the same investigators to mention the possibility of the existence of non-TM helices, similar to those observed in the ML sequences of AAC.

In the present study, we have measured the CD spectra of UCP-1 in digitonin, a mild detergent used to reconstitute membrane proteins. The far-UV CD spectra of UCP-1 and UCP-1His6 depicted in Figure 2 imply comparable conformations. The CD spectrum of UCP-1His6 was superimposable on the UCP-1 spectrum at wavelengths above 220 nm. For these measurements we used reducing agents DTE and 2-mercaptoethanol to minimize the possibility of oxidation of cysteine side chains that could lead to aggregation of UCP-1 monomers. The CD spectra of proteins in Figure 2 have a minimum at ∼220 nm, a shoulder between 209–210 nm and a maximum at ∼194 nm. The overall spectra of the proteins indicate the presence of both helical and β-structure components. Estimates of the secondary structure composition of both versions of UCP-1 (Table 2) suggest that UCP and UCP-1His6 are approx. 41% and 50% helical respectively. The β-structure content is higher in UCP-1 and the secondary structures of the proteins have comparable shares of turn and random structures. In addition to differences in primary structures, the degree of purity could also affect the overall helical content of each protein.

Figure 2 CD spectra of mUCP-1 and mUCP-1His6 in digitonin micelles

The inset shows the near-UV spectrum of UCP-1 at 10× concentration. The proteins were in 0.02% digitonin micelles in a buffer composed of 1 mM DTE, 5 mM 2-mercaptoethanol, 0.9 mM EDTA and 10 mM potassium phosphate (pH 8.0) at 25 °C.

The far-UV CD spectra of UCP-1 and UCP-1His6 in digitonin micelles reveal dominantly helical profiles. On the other hand, the CD spectra of these proteins are complex and indicate the presence of 18–25% β-structure. The CD spectra of UCP-1 and UCP-1His6 in digitonin micelles are distinctly different from those of the same proteins (results not shown) and a C-teminus His6-tagged rat UCP-1 (96% sequence homology with mUCP-1) in Tris buffer and 0.03% DDM [20]. In addition to the overall difference in conformation possibly caused by different micellar environments, the differences between the CD spectra of UCP-1 proteins can be also attributed to dissimilar packing of structural domains in protein monomers and different degrees of association. The conformation of UCP-1 and UCP-1His6 in 0.02% digitonin may represent a fully or partially associated form of the protein. In support of formation of an associated state, an increase in the phosphate salt or protein concentration did not significantly change the overall conformation of UCP-1His6 protein, but increased the contribution of the β-structure in proteins (Figure 3 and Table 2). Near-UV CD spectra of proteins give useful information about the changes in the microenvironment of aromatic amino acid and cystine residues. mUCP-1 (GenBank® accession number P12242) contains two tryptophan, ten tyrosine and 13 phenylalanine residues. The influence of these aromatic residues was apparent in the near-UV spectrum of UCP-1 shown in the inset in Figure 2. Tyrosine ellipticity was dominant with a maximum at 272 nm, but tryptophan residues also made a major contribution to the spectrum at higher wavelengths. The spectrum at a high concentration of the protein (32.5 μM) could also represent the local environment of aromatic residues in the associated and/or more densely packed form of the protein.

Figure 3 CD spectra of mUCP-1His6 and its binding to GDP and GTP nucleotides in digitonin micelles

The inset shows the near-UV spectra of UCP-1His6 in the absence and presence of GDP and GTP. The proteins were in 0.02% digitonin micelles in a buffer composed of 1 mM DTE, 5 mM 2-mercaptoethanol, 0.9 mM EDTA and 50 mM potassium phosphate (pH 8.0), at 25 °C.

View this table:
Table 2 Secondary structure composition of UCP-1 proteins and peptide assemblies

Deconvolution of the CD spectra was performed using the CDSSTR program on the DICHROWEB website at: See also the Experimental section. NRMSD, normalized root mean square deviation (indicates the best fit between calculated and experimental CD spectra).

Purine nucleotides strongly bind to UCP-1 to inhibit its ion-transport function. Basic amino acids Arg83 (TM2), Arg182 (TM4) and Arg276 (TM6) are important for binding of these nucleotides to UCP-1 [7,10]. Aromatic amino acids, especially tyrosine and tryptophan, are located close to these arginine residues, for example Tyr68, Tyr194 or Trp173, Trp280 (Table 1). In Figure 3, we show that binding of purine nucleotides GDP and GTP to UCP-1His6 can be detected by CD. The overall conformation of the protein (far-UV CD) was not strongly influenced by nucleotide binding. In contrast, the CD spectra of the nucleotide-bound proteins were significantly different from the free protein in the near-UV region (Figure 3, inset). Binding of nucleotides drastically changed the spectrum of the protein causing a 3-fold increase in intensity and a blue-shift of maximum from 295 nm to 280 nm. The shifted spectra of nucleotide-bound UCP-1His6 are reminiscent of the spectrum of UCP-1 in the near-UV region (Figure 2), and in addition to conformational changes in the monomeric protein, may indicate an increase in formation of associated proteins.

Overall, the results of the CD spectroscopic studies of the UCP-1 and UCP-1His6 proteins in digitonin micelles indicated comparable protein conformations and provided sufficient evidence for the direct binding of purine nucleotides to UCP-1, which might be an indication that UCP-1His6 is present in a functional form in digitonin micelles.

Design and synthesis of UCP-1 peptide domains

Six TM and three ML domains of golden hamster UCP-1 were chemically synthesized on the basis of the relative hydrophobicity of the amino acid sequence of the protein (Table 1). On average, the peptide sequences had 94% sequence homology with the comparable sequences in mUCP-1. To avoid further complications in spectroscopic measurements due to intermolecular disulfide bond formation, all cysteine residues in the peptide sequences were replaced by serine residues. This Cys→Ser exchange does not influence peptide conformations [11]. Moreover, it has been suggested that cysteine residues are not significant for UCP-1 function [27]. The total number of amino acids in these segments is 276, which represents 90% of the total number of amino acids in UCP-1. Only the short terminal sequences and loops in the mitochondrial intermembrane space (total of 30 amino acids) are excluded. A few flanking amino acid residues were added to both ends of the putative TM sequences to improve the solubility of otherwise insoluble peptides in aqueous environments. The whole-residue hydrophobicity scales [28] were used to assess the affinity of TM segments for insertion into the membrane interior (Table 1). The ΔΔGWBO in Table 1 is the difference between the free energy of partition of potentially helical peptides in the membrane interface region and octanol (representing the membrane interior). Negative ΔΔGWBO values and small hydrophobic moments (low amphipathicity) [29] of all putative TM segments were in agreement with the tendency of these hydrophobic segments to partition into the lipid bilayer interior.

Conformation of individual UCP-1 peptide domains

CD spectra of six TM and three ML peptides in Tris buffer exhibited unordered conformations (results not shown). Conformation of TM helices in helix-promoting organic solvents such as TFE (trifluoroethanol) can be compared with their conformation in micelles or mixed micelle-phospholipid milieus [30,31]. CD spectra of the TM segments of UCP-1 in 50% TFE confirmed the helix-forming tendencies of these peptide domains (Figure 4A). Transmembrane peptides TM1–TM6 exhibited α-helical conformations (double minima at ∼208 and ∼222 nm, maximum between 192–194 nm) with different degrees of helicity. Interestingly, two of the three ML peptide segments, ML12 and ML34, formed α-helical conformations in 50% TFE, whereas ML56 was less ordered (Figure 4B). For comparison, the CD spectra of the ML domains in Tris buffer are shown as an inset in Figure 4(B). These results imply that in addition to six TM segments at least two of the ML domains of the UCP-1 proteins have the potential to form helical structures in hydrophobic environments. The helical portions of the ML segments of UCP-1 are reminiscent of the ML short helices in the crystal structure of AAC, which could lie on the membrane interface [23].

Figure 4 Far-UV CD spectra of UCP-1 TM domains (A) and ML domains (B) in 50% TFE

The inset in (B) shows the CD spectra of ML peptides in Tris buffer. The peptides were either in 50% (v/v) TFE in a buffer composed of 10 mM Tris and 100 mM NaF (pH 7.4) or only in Tris buffer with the same composition, at 25 °C.

The inner membrane of mitochondria is mainly composed of PC, PE and CL phospholipids. Unilamellar vesicles with comparable composition to the inner membranes of mammalian mitochondria in the absence (POPC/POPE/POPG; 5:4:1 molar ratio; LS1, lipid system 1) and presence (POPC/POPE/TOCL/POPG; 4:3.5:2:0.5 molar ratio; LS2, lipid system 2) of CL were used to study the conformation of all nine peptide segments of UCP-1 in lipid bilayers. CL is a specific mitochondrial lipid that interacts with membrane proteins and influences their structure [24,32]. Despite the absence of PG in the mitochondrial membranes, these lipids were used to preserve the overall negative charge balance of the vesicles at neutral pH. The peptides were rapidly mixed with vesicles directly from aqueous solutions without the interference of organic solvents. The spontaneous interaction of peptides with bilayer surfaces and their conformation in vesicular milieus were detected by CD [Figures 5A (LS1) and 5B (LS2)]. The conformation of TM segments in both lipid systems were different from each other and had high β-structure and random coil but low helical content. Spontaneous insertion of helical peptides into the lipid bilayer hydrophobic interior could be prevented both by the added flanking charged/polar residues and lateral pressure of bilayer surfaces. The CD spectra of TM segments in both lipid systems therefore implied that, in general, peptides could associate to form β-structures, only partially penetrate in bilayer interiors and partly lie on the lipid bilayer interface with water. Moreover, conformations of individual TM segments were different in the two lipid systems. For example, the conformation of TM4 was radically different in the two vesicle systems (16% helix, 39% β-strand and 24% random in the LS1; 1% helix, 39% β-strand and 38% random in the LS2).

Figure 5 Far-UV CD spectra of TM and ML domains of UCP-1 in PC/PE/PG (A) and PC/PE/CL/PG (B) vesicles

The insets in both (A) and (B) show the CD spectra of ML peptides in vesicles. The vesicles were in Tris buffer at 25 °C as described in the legend of Figure 4.

The ML peptide domains of UCP-1 in the two lipid vesicle systems adopted mainly random conformations with negligible helical and some β-structure content (insets of Figures 5A and 5B). The most significant conformational change in the two lipid systems was observed in the ML34 domain, which had low β-structure content (11%), but high contents of turn (37%) and random (40%) conformations in LS1; ML34 had more β-structure (28%) and less turn (31%) and random (32%) conformations in LS2. It has been reported that this central ML could be involved in fatty-acid binding, resulting in activation of the proton transport function of UCP-1 [33].

Overall, TM and ML peptides, except for ML56, adopted helical structures in aqueous organic solvents. TM domains associated in lipid vesicles, had a high content of β-structure and distinctly different conformations, whereas ML domains were dominantly random coil in lipid systems. CL seemed to interact specifically with some of the peptide domains (Figure 5B).

Conformation of associated UCP-1 peptide domains

Dissected peptide domains of UCP-1 interacted in aqueous environments and spontaneously formed associated structures (Figure 6A). Formation of distinct associated structures was not observed in the CD spectra of individual peptides in buffer (results not shown). The six TM peptides in buffer associate to form a dominantly β-sheet conformation (38% β-structure, 38% random with almost no helix). The conformation of the mixture of three ML domains was dominantly unordered, but these peptides interacted with the six TM domains to form an assembly in buffer with less β-sheet content (34% β-structure, 43% random and 3% helix). A further proof of self-association of peptides is exhibited in Figure 6(B), where the increase in temperature broke apart the associated β-sheet conformation to form less ordered conformations at high temperatures. The association of the nine UCP-1 domains was concentration-dependent and occurred even when each peptide concentration was at 1 μM (Figure 6B, inset).

Figure 6 Far-UV CD spectra of assemblies of TM and ML peptide domains in buffer (A); temperature-dependence and concentration-dependence (inset) of the CD spectra of assembly of nine peptide domains in buffer (B); and CD spectra of assemblies of TM and ML peptide domains in PC/PE/PG vesicles (C)

The inset in (B) shows the concentration-dependent CD spectra of the peptide assembly. The Tris buffer composition at 25 °C is described in the legend of Figure 4.

The association of TM domains also occurred in both LS1 and LS2 systems (results are only shown for LS1). The conformation of mixed peptide domains in LS1 in Figure 6(C) indicates high β-sheet content and an increase in helical content compared with the conformation in buffer (Table 2). The conformation of the mixture of nine peptides in LS1 was close to 50% random coil, implying that addition of the three ML segments did not stabilize ordered conformations. The CD spectra in Figure 6(C) have an isodichroic point, which suggests the existence of at least two associated forms at equilibrium. It should be noted that in both aqueous and lipid vesicle environments, the CD spectra upon addition of three ML to six TM peptides were not additive. The CD spectra therefore represent the conformation of assemblies formed by an interacting collection of peptides.

Temperature-dependent unfolding of UCP-1 and associated UCP-1 peptide domains

Thermal denaturation of UCP-1 and its nine associated peptide domains in digitonin micelles was examined by CD (Figures 7A and 7B). Although the backbone conformations of the protein and peptide assembly differed (Figures 7A and 7B, and Table 2), their near-UV CD spectra were comparable (Figure 7C). The backbone conformation of the protein was relatively stable in the 10–80 °C range; however, the near-UV CD spectra revealed a changing side-chain conformation (Figure 7A and inset). Deconvolution of the far-UV CD spectra showed a decrease in the helical content from 42% (10 °C) to 24% (80 °C), and an increase in the random conformation from 20 to ∼30%. From this observation it can be concluded that UCP-1 was partially denatured in the applied temperature range. This partial thermal denaturation had a sigmoidal path and could therefore be explained by a two-state model with a transition temperature, Tm, of 54.6±3.1 °C.

Figure 7 Temperature-dependent CD spectra of UCP-1 (A); temperature-dependent CD spectra of the assembly of TM and ML domains (B); and near-UV CD spectra of UCP-1 and the peptide assembly (C)

The inset in (A) shows the temperature-dependent near-UV spectra of UCP-1. The inset in (B) is the transition curve for thermal denaturation of the peptide assembly monitored at 220 nm. The protein and peptide assembly were in 0.02% digitonin. The phosphate buffer composition is described in the legend of Figure 2. Near-UV CD was measured at 25 °C.

Compared with the whole protein, the nine peptide domain assembly melted at much lower temperatures (37.7±0.7 °C) and transformed from one conformation (∼30% β-structure, 17% turn, 50% random coil at 10 °C) to another conformation (∼38% β-structure, 25% turn, 30% random coil at 80 °C) (Figure 7B). Thermal denaturation of the associated peptide assembly also followed a sigmoidal curve (inset of Figure 7B) that could be explained by a two-state denaturation model. The isodichroic range of the CD spectra around 205 nm could further support the existence of two main conformations of the peptide assembly at equilibrium.

Assuming a two-state transition at equilibrium (folded ↔ unfolded) and by adopting a co-operative unfolding (or partial unfolding) model for UCP-1 and its nine peptide segments assembly, the van't Hoff equation: Embedded Image(1) could be applied to calculate the effective enthalpies of unfolding or transformation of the ‘melting unit’ of the protein, which can be different from the protein monomer [34]. It was assumed that the effective enthalpy did not change in the temperature range used for experiments. The van't Hoff plots for UCP-1 and peptide assembly in digitonin micelles are shown in Figure 8. The ΔHv values for UCP-1 and the peptide assembly were 18.6 and 26.5 kcal/mol respectively (1 kcal≈4.184 kJ). The higher van't Hoff enthalpy values for peptide assembly implied a higher degree of disorder (higher entropies) of unfolding, which was expected for an assembly of nine peptides compared with the full-length protein. The entropies of unfolding (ΔSv) for UCP-1 and the peptide assembly were 56.5 and 85.1 cal·mol−1·K−1 respectively. The thermal denaturation of UCP-1 and the peptide assembly in digitonin were not reversible and the Tm of the reversal cooling process were at lower temperatures (e.g. ∼ 40 °C for UCP-1).

Figure 8 van't Hoff plots for thermal denaturation of UCP-1 (●) and the assembly of TM and ML domains (○) in digitonin micelles

The folded and unfolded states were at equilibrium and the Kunfolding (Keq) values in the plots were calculated from the ellipticity ratios of unfolded and folded states at each temperature. The ΔHv values were calculated from the slopes of the lines, −ΔHv/R. Entropies of unfolding were calculated from the equilibrium conditions at Tm, where ΔSvHv/Tm.

Comparison between the conformations of UCP-1 and associated UCP-1 peptide domains

UCP-1 and UCP-1His6 in 0.02% digitonin micelles adopted dominantly helical conformations (40–50%). The conformation of the assembly of nine peptide domains in 0.02% digitonin micelles was largely a combination of β-sheet and random coil. An increase in the digitonin concentration to 0.2% increased the content of ordered structures in the conformation of the peptide assembly (Table 2). A comparison between the conformational features of proteins and peptide assemblies is exhibited in Figure 9(A). Conformation of peptide assemblies had a low helical (less than 10%) and high random coil (35–50%) content, whereas UCP-1 and UCP-1His6 had lower β-structure and turn content in comparison with peptide assemblies (Table 2). An interesting feature of the CD spectra in Figure 9(A) is the partial overlap of the six TM peptide assembly and the UCP-1 spectra with comparable ellipticities above 207 nm. Covalent connection of TM and ML assemblies seems to be essential to induce helical structures in UCP-1; however, these peptide segments could also spontaneously assemble to form distinct ordered conformations.

Figure 9 Comparative far-UV CD spectra of UCP-1 and its TM and ML peptide assemblies (A) and the ion-channel formation in the assembly of six TM domains in lipid bilayers (B)

CD measurements were at 25 °C. The patch-clamp experiments were performed at ambient temperatures in 5 mM Hepes and 500 mM KCl buffer at pH 7.4, using DPhPC as the lipid bilayer. Electrolyte composition was symmetrical in the pipette and bath. The open–closed probability histogram of the ion channel is shown on the right-hand side in (B). The ion current for the single-state channel at +150 mV TM potential was 3.4 pA (23 pS conductance).

We also employed the pipette-dipping patch-clamp technique to examine the possible ion-conducting function of an assembly of the six TM domains in zwitterionic PC phospholipid (DPhPC) bilayers (Figure 9B). The peptide mixture formed stable single-state low-conductance (23 pS) ion-conducting units. Conductance and ion specificity of ion channels can be related to their pore size in lipid bilayers. The ion specificity of this relatively narrow ion channel remains to be explored. Considering the high tendency of the UCP-1 TM segments to form helical structures in hydrophobic environments and membrane interiors (Figure 4A and Table 1) it is plausible that the pore structure of the TM peptides was composed of a tightly packed helical bundle. If this assumption is correct, then the application of potential difference across the lipid bilayer had transformed the dominantly β-structure/random coil conformation of the peptide assembly to helical conformations that could transverse the bilayer interiors to form ion channels.

From the conformational studies of the TM and ML segments of UCP-1, it can be concluded that, in addition to the amino acid sequences and spontaneous conformations of these domains, structural constraints such as head-to-tail covalent connection between domains, tight packing of TM domains, membrane-lipid interactions with domains and electrical potential across membranes are essential factors that determine the functional structure of UCP-1.


The present study was supported by grants from the Canada Foundation for Innovation and the Natural Sciences and Engineering Research Council of Canada to M. J.-N. (CFI: 6786; NSERC: 250119) and M. D. S. (CFI: 11292; NSERC: 312143). B. L. M. was the recipient of an NSERC Undergraduate Summer Research Assistant scholarship.

Abbreviations: AAC, ADP/ATP carrier; CL, cardiolipin; DDM, n-dodecyl β-D-maltoside, DPhPC, 1,2-diphytanoyl-sn-glycero-3-phosphocholine; DTE, dithioerythritol; Fmoc, fluoren-9-ylmethoxycarbonyl; IB, inclusion body; ML, matrix-loop; NMP, N-methylpyrrolidone; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PG, phosphatidylglycerol; POPC, 1-palmitoyl-2-oleoyol-sn-glycero-3-phosphocholine; POPE, 1-palmitoyl-2-oleoyol-sn-glycero-3-phosphoethanolamine; POPG, 1-palmitoyl-2-oleoyol-sn-glycero-3-[phosphor-rac-1-glycerol] (sodium salt); TCA, trichloroacetic acid; TFE, trifluoroethanol; TGE, Tris-glucose-EDTA; TM, transmembrane; TOCL, 1,1′2,2′-tetraoleoyl cardiolipin (sodium salt); UCP, uncoupling protein


View Abstract