|Home | About | Journals | Submit | Contact Us | Français|
The hexapeptide acetyl-Trp-Leu5 (AcWL5) has the remarkable ability to assemble reversibly and spontaneously into β-sheets on lipid membranes as a result of monomer partitioning followed by cooperative assembly This system provides a unique opportunity to study the thermodynamics of protein folding in membranes, which we have done using isothermal titration calorimetry (ITC) and differential scanning calorimetry (DSC). The results, which may represent the first example of reversible thermal unfolding of peptides in membranes, help to define the contribution of hydrogen bonding to the extreme thermal stability of membrane proteins. ITC revealed that the enthalpy change for partitioning of monomeric, unstructured AcWL5 from water into membranes was zero within experimental error over the temperature range of 5 °C to 75 °C. DSC showed that the β-sheet aggregates underwent a reversible, endothermic, and very asymmetric thermal transition with a concentration-dependent transition temperature (Tm) in the range of 60 °C to 80 °C. A numerical model of nucleation and growth-dependent assembly of oligomeric β-sheets, proposed earlier to describe β-sheet formation in membranes, recreated remarkably well the unusual shape and concentration-dependence of the transition peaks. The enthalpy for thermal unfolding of AcWL5 β-sheets in the membrane was found to be about 8(±1) kcal mol−1, or about 1.3(±0.2) kcal mol−1 per residue.
The thermodynamic driving forces for membrane protein (MP) folding and assembly are not fully understood quantitatively. One reason for this is that reversible thermal unfolding transitions of membrane-buried secondary structure elements cannot be observed,1 because of the high thermodynamic cost of breaking H-bonds in the non-polar membrane interior. The free energy cost can exceed 80–100 kcal mol−1 for a 20 amino acid (AA) transmembrane (TM) α-helix.2 The few available calorimetric studies of MPs detect only the thermal unfolding of extra-membrane domains and the rearrangement of membrane-spanning segments,1 and even these transitions are mostly irreversible. The goal of the work described here was to use a peptide model system for which reversible unfolding transitions in the membrane are experimentally accessible (Figure 1) in order to obtain quantitative information about the forces that stabilize membrane proteins.
Beginning with the seminal calorimetric measurements of Privalov and colleagues 30 years ago,3–7 the thermodynamics of unfolding single-domain soluble proteins have been described well, both experimentally and theoretically8–10 Soluble proteins are only marginally stable with a net free energy of folding, ΔGfold, ranging from −5 to −15 kcal mol−1, corresponding to transition midpoint temperatures (Tm) in the range of 50–90 °C. The small ΔGfold value for soluble proteins in the aqueous phase apparently arises from strong thermodynamic compensation between the stabilizing hydrophobic effect and destabilizing polar group/peptide bond dehydration, described in detail by Yang et al.11 The free energy balance for MPs is problematic, because of the difficulty of unfolding them inside membranes.
The literature contains only a few accounts of the thermal denaturation of membrane proteins. In one of the earliest references, Jackson & Sturtevant12 reported the irreversible denaturation of the membrane protein bacteriorhodopsin at about 100 °C in its native membrane. Bacteriorhodopsin has since been studied by other laboratories as well, and these studies invariably find that bacteriorhodopsin in membranes undergoes an irreversible structural transition in the temperature range 90 °C to 100 °C. FTIR spectroscopy13,14 showed that the protein retained much of its α-helical structure even above the transition temperature. This is in full agreement with structural studies of the thermal unfolding of cytochrome c oxidase, for example. The multiple transitions of cytochrome c oxidase were interpreted as resulting from unfolding of extrinsic domains and an irreversible structural rearrangement of membrane-spanning segments, which did not unfold.15–17 Similar results have been reported for bovine rhodopsin,18 the GLUT1 glucose transporter,19 a Ca2+ ATPase,16,17,20 and a mechanosensitive channel of Escherichia coli.21 Like bacteriorhodopsin, most of these membrane proteins have high-temperature thermal transitions that have scan-rate-dependent calorimetric properties. This means that the transitions were irreversible, because the high temperature state converted irreversibly to an additional state. This thermally induced, inactive state could have resulted from aggregated protein in the membrane or from a state with non-native topology. As further evidence of incomplete unfolding of these membrane proteins, their calorimetric enthalpies for the thermal transitions have a mean value of about 0.35(±0.05) kcal mol−1 per residue. This is about half of the value of that measured for soluble proteins,22 and less than one-third the value we report here for AcWL5 unfolding in membranes (see below). The image that emerges from this literature is that the thermal denaturation of membrane proteins is an irreversible and kinetically controlled process, rather that an equilibrium process, and that it does not involve the unfolding of the membrane-spanning segments. The membrane-spanning segments are thus very stable toward unfolding. Because of this, published calorimetric studies of membrane protein folding are of little utility in understanding the thermodynamic driving forces for membrane protein folding.
Direct experimental measurements of the energetics of peptide folding and unfolding in membranes to date is limited to our studies of AcWL5 folding23,24 and a few studies of interfacial folding of amphipathic α-helices and their diastereomeric analogs.25–27 Membrane partitioning and secondary structure formation of polypeptides in the membrane are coupled through the cost of partitioning the peptide bond.28 Specifically, the partitioning of an open peptide bond from water into the bilayer interface is unfavorable28 by at least 1.2 kcal mol−1. The reduction of that unfavorable interaction in the membrane through the formation of intramolecular H-bonds is the primary driving force for the formation of secondary structure.23 Thus, relative to folding in an aqueous environment, a peptide partitioned into a membrane will be much more likely to have secondary structure.2 Based on several model peptide studies, the net free energy of secondary structure formation for peptides in membranes, including the contribution of hydrogen bonding, configurational entropy, and bilayer effects,29 is in the range of −0.15 to −0.7 kcal mol−1 per residue.24–27,30 We have estimated that the assembly of AcWL5 β-sheets in membranes is driven by a favorable folding free energy of 0.4–0.6 kcal mol−1 per residue.24 Considering that a TM β-barrel contains a minimum of 80 residues31 and a TM α-helix is about 20 residues long,32 the net H-bond driving force for structure formation in membranes can be very large, consistent with the extreme thermal stability of membrane-spanning proteins.1
AcWL5 is a well-characterized hexapeptide that folds cooperatively in membranes into large oligomeric β-sheets.23,24 Oligomers assemble in membranes via a three-step process of monomer partitioning, β-sheet nucleation, and β-sheet growth.24 This reversible (equilibrium) folding process, illustrated schematically in Figure 1, produces membrane β-sheets that are estimated to contain 10–20 peptides each.24 The reversible unfolding of AcWL5 is clearly revealed by circular dichroism (CD) spectroscopy measurements. Plots of molar ellipticity at 199 nm versus temperature revealed a characteristic secondary-structure melting curve with an inflection at about 60 °C and transition half-width of about 20 °C, consistent with the cooperative melting of 50 or more residues (see Figure 6 of Wimley et al.24). However, the melting was not the result of a simple two-state transition, because the collection of CD spectra gathered at different temperatures did not have an isodichroic point. In the work described here, we present calorimetric studies of folding/unfolding reactions of AcWL5 β-sheets that provide further insights into the melting transition in membranes and hence the energetics of MP stability.
The folding–unfolding equilibria of AcWL5, outlined in Figure 1, include the thermally induced β-sheet-to-coil transition on (or possibly in) the membrane and the membrane-to-aqueous phase partitioning of the unfolded peptides.24 The latter transition makes this a particularly convenient experimental system, because the AcWL5 peptides that unfold in the membrane can also partition into the aqueous phase, and thereby remove the penalty associated with the partitioning of open H-bonds into the membrane. This couples β-sheet unfolding to a shift of peptide molecules from the membrane to the aqueous phase.24 In contrast, the high hydrophobicity of full-length MPs prevents their partitioning into the aqueous phase during thermal unfolding. In our experiments, we have studied both the partitioning and unfolding of AcWL5 by calorimetry. Isothermal titration calorimetry (ITC) was used to characterize the partitioning step and differential scanning calorimetry (DSC) to characterize the thermal unfolding step. Together, they provide a complete image of the thermodynamics of β-sheet unfolding in membranes.
The ITC measurements, designed to measure the enthalpy of water-to-membrane partitioning (ΔHm) of unfolded AcWL5, were made under conditions in which β-sheets do not form:24 high lipid concentration (100 mM lipid) and low peptide concentration (≤10 μM AcWL5). Under these conditions,24 the peptide is a monomeric, interfacially bound, random-coil with a free-energy of transfer ΔGm = −5.2 kcal mol−1. We also measured the ΔHm value for AcWL4, which has nearly identical33 membrane interactions but does not form β-sheets at any concentration, because it has a much higher water solubility. ITC results for the partitioning of monomeric random-coils of both peptides into POPC bilayers are shown in Table 1. Within experimental error, the net enthalpy of partitioning of AcWL5 was zero across the entire temperature range, which implies a small or zero heat capacity change (ΔCp) for partitioning as well. This was surprising because the hydrophobic effect with characteristic negative heat capacity is generally the main driving force for the partitioning of hydrophobic peptides into membranes. The experimental uncertainty for AcWL5, due to its low solubility, was large enough that small, but significant, values of ΔHm and ΔCp could not be excluded. However, measurements performed using the more soluble AcWL4 peptide also yielded a ΔHm value that was zero within ± 0.1 kcal mol−1 at both 25 °C and 50 °C. This observation gives us confidence that the contribution of ΔHm to the thermodynamics of AcWL5 unfolding in membranes is small.
In order to gain insights into the negligible enthalpy and heat capacity, we examined several other peptides and model compounds. Using exactly the same lipid and buffer system as in this work, we measured ΔHm for the peptides AcWLWLL and AcWLGLL and found values of −0.54(±0.2) and −0.2(±0.1) kcal mol−1, respectively, at 25 °C with associated heat capacity changes of −41 and −18 cal mol−1 K−1 (unpublished observations). In contrast, the 13-residue antimicrobial peptide indolicidin34 was found to partition exothermically at 25 °C with a ΔHm a value of + 3.3(±0.5) kcal mol−1 at 25 °C and ΔCp value of −330 cal mol−1 K−1. The bee venom peptide melittin had a ΔHm value of +4.0 kcal mol−1 at 25 °C with a ΔCp value of −300 cal mol−1 K−1 in the same system (unpublished observations). We have also measured ΔHm by ITC for the side-chain analogs of tryptophan (unpublished observations). For N-methyl indole and 3-methyl indole,35 we found values of −2.2(±0.2) and −1.5(±0.2) kcal mol−1, respectively, with corresponding ΔCp values of −17 and −29 cal mol−1 K−1. If AcWL4 and AcWL5 make an exothermic enthalpy contribution to ΔHm due to their tryptophan residues, then there must also be an endothermic contribution from other sources. We speculate that the unfavorable partitioning of the open backbone of these random-coil peptides into the bilayers may be strongly endothermic, possibly due to strong electrostatic interactions with bilayer surfaces.26 But the interpretation of these and other data in the literature26,27,30 is further complicated by the induction of structure that accompanies partitioning.27,30 One can only say at this point that a wide range of partitioning enthalpies and heat capacity changes are possible for peptides interacting with phosphatidylcholine membranes, because of the composite nature of peptide–bilayer interactions2,36 and the many different competing interactions that contribute.
Unlike the ITC measurements, the DSC experiments were done under conditions24 that promote cooperative assembly of AcWL5 into multimeric (β-sheets: low lipid concentration (2 mM POPC) and high AcWL5 concentration (greater than 50 μM). We found earlier24 that the transfer free energy ΔGmA of monomer to β-sheet aggregate was −3.70 kcal mol−1, and that β-sheet formation was thermally reversible. Upon heating, the membrane-bound β-sheets underwent a cooperative β-sheet-to-coil transition centered at about 65 °C. The peptide was monomeric and unstructured above the transition temperature, and some of the bound peptide partitioned back into the aqueous phase. The DSC experiment was thus expected to measure the energetics of β-sheet unfolding in membranes, and subsequent partitioning of unfolded peptides back into the aqueous phase. Because the ΔHm value for AcWL5 is zero (Table 1), the enthalpy measured by DSC should therefore represent mainly unfolding of the β-sheets in membranes, ΔHmA.
Examples of raw excess heat capacity curves for POPC vesicles and POPC vesicles containing AcWL5 β-sheets are shown in Figure 2. Except for a small negative shift in the y-axis position of the curve, the vesicle-only trace is identical to the water versus water instrumental background. This is consistent with the expectation that no thermal transitions occur in POPC over the temperature range37 of 5 °C to 95 °C. In the vesicles + peptide scan, there is an asymmetric, endothermic transition that occurs over the temperature range of 40 °C to 80 °C. This corresponds exactly to the sheet-to-coil transition observed with CD spectroscopy24
In Figure 3(A), the baseline-subtracted transition curves are plotted for samples of 2 mM POPC vesicles containing between 60 and 140 μM AcWL5. Linear baselines were drawn by extrapolation from the regions before and after the transition. We could not detect evidence of a significant shift in baseline heat capacity during the transition. We estimate that we should have been able to measure a baseline heat capacity difference if it were equal to or larger than 0.005 mcal °C−1, which is equivalent to about 100 cal mol−1 K−1 or about 15 cal mol−1 K−1 per residue for 100 μM AcWL5 in the 0.519 ml sample cell. For comparison, a single buried leucine residue in a soluble protein is expected to contribute a heat capacity change of approximately 50 cal mol−1 K−1 upon unfolding.38 In the ITC experiments (above), we estimated that the heat capacity change for the water-to-membrane partitioning was smaller than 20 cal mol−1 K−1. Assuming, then, no baseline heat capacity change, we estimate that the enthalpy of unfolding to be no smaller than 1.3(±0.2) kcal mol−1 per residue (below).
The asymmetric shape of the DSC transition curves is quite unusual for protein or peptide transitions. As shown in Figure 3(A), these peaks have broad and overlapping leading (i.e. low-temperature) edges, and have very steep high-temperature edges that shift to higher temperatures with higher peptides concentrations. Can the unusual characteristics of these curves be explained with the nucleation-growth model that we have used24 to explain the assembly of AcWL5 β-sheets in membranes? To answer this question, we used the model and our previous iterative computational approach to simulate the calorimetry experiment. Briefly, we used the equation24 for total concentration CT of peptide on the membrane:
where Cm is the membrane monomer concentration calculated from the infinite-dilution monomer partition coefficient Kx (7500) and the equilibrium peptide concentration in water (CW) using Cm = KxCW/W, where W is the concentration of water (55.3 M). The parameters σ and s are the nucleation and growth parameters, respectively.39 For each temperature, the molar concentrations of soluble monomeric peptide (Cw), the mole-fraction concentration of bound monomeric peptide (Cm), and the mole-fraction concentration of peptide bound as multimeric β-sheets (Csheet = CT − Cm) were determined from the equations above using values of σ and s calculated from x = exp(−ΔGX/RT) and ΔGX = ΔHX − TΔSX where x = σ or s. For simulations using the experimental lipid and peptide concentrations, we found a good empirical match with the experimental data when ΔHmA was set to − 10 kcal mol−1 and ΔGσ and ΔGS were set to + 1.8 and − 4.2 kcal mol−1, respectively. These values are somewhat different from those obtained from partitioning measurements, but are acceptably close given the exquisite sensitivity of the system.24
The excess heat capacity curves in the simulation were calculated by numerical integration using:
where Csheet is the mole fraction concentration, L is the molar lipid concentration (2 mM), and Vcell is the volume of the calorimeter cell (0.5185 ml). Notice in the simulated curves that all the characteristics of the experimental curves are recreated, including the unusual shape, asymmetry, concentration-dependence, Tm values, and the magnitude of . We conclude that the very unusual shapes of the calorimetric transitions in AcWL5 β-sheets in membranes are well accounted for by the nucleation and growth model of AcWL5 assembly into β-sheets in POPC bilayers.
The goal of our work was to use the experimentally accessible reversible unfolding transitions of the AcWL5 system (Figure 1) to obtain quantitative information about the high thermal stability of membrane proteins. Because β-sheets of AcWL5 in membranes contain 10–20 peptides each,24 or roughly 100 amino acids, they are comparable in size to small, single-domain proteins. At the same time, because the peptides are water-soluble and only six residues long, it is possible to unfold them thermally in order to study their structural transitions. We show in Figure 4 the concentration-dependence of ΔHmA and Tm for AcWL5 β-sheet unfolding in membranes. Due in part to increased fractional peptide binding,33 both values increase with peptide concentration, approaching Tm = 80 °C and ΔHmA = 7 kcal mol−1 at the highest concentrations of AcWL5 studied. Based on our earlier work,24 we estimate that the fractional binding at the highest and lowest peptide concentrations used, 80–90% and 65–80% were bound, respectively. Because the binding/assembly process is extremely cooperative, it is not possible to predict fractional binding more precisely. Dividing the observed enthalpy of unfolding, which increases from 5 to 7 kcal mol−1 over the range of peptide concentrations studied (Figure 4), by our highest (most conservative) estimates of fractional binding, we obtain a value for the enthalpy of unfolding of no less than 8(± 1) kcal mol−1 of peptide, or about 1.3(±0.2) kcal mol−1 per residue.
In order to test the reasonableness of this estimate, we examined the thermal unfolding of AcWL6 in membranes. AcWL6, containing a single additional Leu residue, folds into β-sheets in membranes that are very similar to AcWL5 β-sheets.24 If each additional hydrogen bond contributes up to 1.3 kcal mol−1 to stability as our calorimetry results suggest, then AcWL6 β-sheets will be significantly more stable than those assembled from AcWL5. We have reported previously that AcWL6 β-sheets did not appear to unfold at any experimentally accessible temperature (up to 100 °C) when examined with CD spectroscopy24 DSC measurements performed in conjunction with our measurements for AcWL5 revealed no evidence of unfolding transitions at any temperature up to 100 °C, confirming the extreme stability of AcWL6 membrane β-sheets. The excess heat capacity trace for AcWL6 in POPC (not shown) was essentially identical to the POPC-only trace shown in Figure 2. These results strengthen the idea that hydrogen bonding makes very large net contributions to membrane protein stability.
In order to verify further that our DSC observations for AcWL6 are reasonable, we performed numerical simulations like those in Figure 3. Simulations were run at the lowest measured concentration of ~50 μM peptide in 2 mM lipid using the parameters from the simulation of AcWL5. We found that a net increase in the nucleation and growth free energies (ΔGS and ΔGσ; equation (1)) of 0.5 kcal mol−1 or greater, coupled with the known increase in water-to-membrane partitioning of AcWL633 is sufficient to increase the Tm value to well over 100 °C. This supports our conclusion from the AcWL5 data that each hydrogen bond that forms in a membrane environment makes a substantial net contribution to the stability of membrane proteins.
Because ΔHm is zero (Table 1), the enthalpy measured for AcWL5 unfolding is derived from the unfolding steps in the membrane environment. In contrast, Seelig and colleagues26,27 have studied the formation of interfacial α-helices and reported ΔHmA of about 0.7 kcal mol−1 per residue. How do these numbers compare to the enthalpy of unfolding of soluble proteins and peptides in water? As discussed extensively by Yang et al.,11 the per-residue enthalpy accompanying soluble protein unfolding is generally in the range of −0.1 to 0.6 kcal mol−1. Scholtz et al.,40 however, found that the per residue enthalpy of unfolding of a 50-residue polyalanine helix to be 1.3 kcal mol−1. The difference arises, according to Yang et al.,11 from the electrostatic cost of dehydrating the secondary structure elements when they assemble to form a protein's tertiary structure. That is, there is an unfavorable enthalpy of dehydration that works against the favorable enthalpy of H-bond formation. What are the consequences of applying this reasoning to the unfolding of our β-sheet aggregates on or in membranes? To begin to address that question, structural information about the membrane aggregates is required.
The three-dimensional structure and topology of the β-sheet aggregates is uncertain, but the available evidence does not rule out the possibility that they may form membrane-spanning structures (Figure 1). Polarized attenuated total reflectance FTIR spectroscopy showed that the interstrand hydrogen bonds are parallel to the membrane surface,41 consistent with either an interfacial orientation parallel to the membrane surface or transmembrane orientations normal to the bilayer surface (Figure 1). Fluorescence experiments41,42 are unambiguous in placing the tryptophan residues of the peptide β-sheets in the interfacial region, partially exposed to the aqueous phase. Furthermore, the pKa value of 5.7 for the carboxyl termini of the β-sheets change very little (W.C.W, unpublished observations) compared to interfacially bound, random-coil monomers,43 consistent with an interfacial location of the carboxyl groups. However, in a family of related peptides of the form AcWLLXLL, it was found that only the members with the most non-polar X-residue side-chains assembled readily into multimeric β-sheets. All other side-chains, even those hydrophobic enough to partition strongly into bilayer interfaces, were monomeric random-coils at all concentrations.42 This acute sensitivity of secondary structure to the polarity of the side-chain at the X-position suggested a mid-bilayer (hydrocarbon core) location for the central residues of the hexapeptide β-sheets, because of the energetic cost of placing a polar moiety in the hydrocarbon core.42,43 A transmembrane topology would place the carboxy termini and the Trp residues of the β-sheet at the bilayer interfaces, which is consistent with the known interfacial localization of aromatic residues44 in β-barrel membrane proteins.45 The transmembrane separation of the Trp residues would be ~20 Å, which is rather smaller than the 30 Å thickness46 of the bilayer hydrocarbon core. Nevertheless, it corresponds closely to the average transmembrane distance between the aromatics of β-barrel membrane proteins45 and the tryptophan residues of gramicidin dimers.47,48
The membrane environment complicates any interpretation of ΔHmA because the peptides are in a membrane environment. Changes in lipid–lipid and lipid–protein interactions must surely contribute, but just how is not clear. It may not be surprising then that our value of ΔHmA= 1.3 kcal mol−1 per residue for a membrane environment is virtually identical to the enthalpic cost of breaking H-bonds in the unfolding of polyalanine helices in water determined by Scholtz et al.40 This may mean that the β-sheet aggregates are located on the membrane surface with high water exposure. If, on the other hand, the aggregates have a transmembrane configuration, then the observed value likely arises from a more favorable contribution from H-bonding in the membrane coupled with a very unfavorable enthalpic cost of dehydrating the aggregates upon membrane insertion. In this case, our value of ΔHmA represents a lower limit on the enthalpic cost of breaking H-bonds in the non-polar environment of the bilayer hydrocarbon core. Yang et al.11 estimate that the polar group enthalpy favoring the unfolded state of soluble proteins is in the range of − 0.4 to − 1.4 kcal mol−1 per residue, suggesting that for buried aggregates the enthalpy for breaking H-bonds in the hydrocarbon core is no smaller than 1.7 to 2.7 kcal mol−1 per residue.
This work was supported by grants GM-46823 (to S.H.W) and GM-60000 (to W.C.W) from the National Institute of General Medical Sciences.