Probing the Lipid-Protein Interface Using Model Transmembrane Peptides with a Covalently Linked Acyl Chain

Abstract

The aim of this study was to gain insight into how interactions between proteins and lipids in membranes are sensed at the protein-lipid interface. As a probe to analyze this interface, we used deuterium-labeled acyl chains that were covalently linked to a model transmembrane peptide. First, a perdeuterated palmitoyl chain was coupled to the Trp-flanked peptide WALP23 (Ac-CGWW(LA)₈LWWA-NH₂), and the deuterium NMR spectrum was analyzed in di-C18:1-phosphatidylcholine (PC) bilayers. We found that the chain order of this peptide-linked chain is rather similar to that of a noncovalently coupled perdeuterated palmitoyl chain, except that it exhibits a slightly lower order. Similar results were obtained when site-specific deuterium labels were used and when the palmitoyl chain was attached to the more-hydrophobic model peptide WLP23 (Ac-CGWWL₁₇WWA-NH₂) or to the Lys-flanked peptide KALP23 (Ac-CGKK(LA)₈LKKA-NH₂). The experiments showed that the order of both the peptide-linked chains and the noncovalently coupled palmitoyl chains in the phospholipid bilayer increases in the order KALP23 < WALP23 < WLP23. Furthermore, changes in the bulk lipid bilayer thickness caused by varying the lipid composition from di-C14:1-PC to di-C18:1-PC or by including cholesterol were sensed rather similarly by the covalently coupled chain and the noncovalently coupled palmitoyl chains. The results indicate that the properties of lipids adjacent to transmembrane peptides mostly reflect the properties of the surrounding lipid bilayer, and hence that (at least for the single-span model peptides used in this study) annular lipids do not play a highly specific role in protein-lipid interactions.

Introduction

Interactions between proteins and lipids play a key role in the structure and organization of proteins and lipids in membranes (1). Lipids can influence the structure and function of membrane proteins by interacting with their transmembrane segments, and transmembrane segments in turn can affect the organization and structure of lipids.

The hydrophobic nature of the lipid bilayer plays a central role in the interactions between lipids and protein transmembrane segments. For example, the presence of polar or charged residues in this hydrophobic environment is highly unfavorable, and either such residues will be driven to the interior of proteins or other molecular rearrangements will occur to minimize the unfavorable interactions (2). As another example, the extent of hydrophobic matching between the bilayer and the transmembrane segments of integral membrane proteins can modulate the interplay between lipid and proteins. The bilayer may, through its thickness, influence the helical tilt against the bilayer normal, the extent of oligomerization, and/or the secondary structure of transmembrane to minimize the exposure of hydrophobic surfaces to water (3). In turn, phospholipids can adapt to compensate for hydrophobic mismatch between bilayer thickness and protein transmembrane segments (4) by stretching or compressing acyl chains (5,6).

Another factor that may influence the structure of integral membrane proteins is the lateral packing of lipid bilayers. Cholesterol, a common component of mammalian cell membranes, modulates both the packing density and the thickness of phospholipid bilayers (7–10). Hence, cholesterol could have an important role in controlling the activity of membrane proteins. For example, it has been reported that the stability and activity of rhodopsin are influenced by the cholesterol content in the surrounding membranes (11).

All consequences of interactions between lipids and proteins must somehow be transmitted through the lipid-protein interface. However, it is poorly understood how this process occurs. A convenient way to obtain specific information about the protein-lipid interface is to covalently link a specifically labeled acyl chain to the protein as a reporter (12,13), such as by linking palmitoyl chains to cysteine residues at the termini of the protein transmembrane segments. Using this approach, investigators have studied the boundary lipids around, e.g., rhodopsin and gramicidin, by labeling the protein with spin-labeled fatty acids or deuterium-labeled acyl chains, respectively (13,14). Palmitoylation is also interesting from another, more biological point of view, because it is a frequently occurring form of protein modification in membrane proteins (15–18). For example, some results suggest that processes such as protein sorting, ion channel and receptor function are regulated by palmitoylation (19–21). Therefore, studying the behavior of covalently coupled fatty acids may also provide insight into the influence of such modifications.

We chose to use deuterium because it is a convenient and nonperturbing label that allows analysis of (local) acyl chain order in lipid bilayers by deuterium NMR (22). Studies with deuterated hydrocarbon chains attached to the water-soluble peptides N-Ras or GCAP-2 (23,24), or with deuterium-labeled free fatty acids in a lipid bilayer (13) have shown that such chains report bilayer thickness and their length adapts to the thickness of the host membrane. With respect to transmembrane peptides, the method thus far has been applied to palmitoyl chains that were covalently bound to gramicidin in dimyristoylphosphatidylcholine (DMPC) bilayers (13) and to lipid modifications of the synthetic transmembrane peptide LV16, which was used as a model of a fusion peptide (23,25). Surprisingly, very different results were obtained. The studies of gramicidin suggested that the covalently bound chain and the bulk lipids showed similar behaviors, except at the attachment site to the peptide, where the chain was immobilized (13). With the LV16 peptides, on the other hand, the peptide-linked chains did not adapt to the surrounding lipid (23). A possible explanation for this different behavior is that gramicidin, due to the sequence of alternating L and D amino acids, forms a rather unique β-helix structure in which the peptide side chains are packed relatively tightly, thereby inhibiting shielding of the hydrocarbon chain and optimizing interaction with the surrounding lipids. However, it is equally possible that special properties of the fusion model peptide LV16, such as its conformational flexibility (26), are responsible for the lack of response to the host bilayer when hydrocarbon chains are covalently attached to this peptide.

In this study we used synthetic α-helical peptides, which are often used as general mimics of protein transmembrane segments, and also form stable transmembrane helices (27). The palmitoyl chain, perdeuterated or specifically deuterium-labeled at position 9 or 13, was covalently attached to the peptide via a cysteine residue at the N-terminus. To gain insight into whether (and how) annular lipids are influenced by their closeness to α-helical transmembrane segments, we varied the composition of both the lipid environment and the transmembrane peptides. The results show that peptide-linked palmitoyl chains and the free fatty acid sense the lipid environment similarly, and that the peptide composition does not appear to influence the covalently attached lipid in a significantly different manner than it does the bulk lipids. The results will be discussed in the light of the general role of annular lipids in protein-lipid interactions.

Material and Methods

For details regarding the materials and methods used in this work, see the Supporting Material.

NMR measurements

NMR experiments were performed on a Bruker Avance 500 MHz NMR spectrometer. All measurements were performed at 30°C, except for those obtained with perdeuterated di-14:0-phosphatidylcholine (PC), which were performed at 40°C. The ²H-NMR measurements were performed at 76.78 MHz using a quadrupolar echo sequence as described previously (22), with a 6 μs 90° pulse, an echo delay of 40 μs, and a recycling delay of 100 ms. For the samples with perdeuterated acyl chains, a 600 ms recycling delay time was used. An exponential multiplication, corresponding to a line-broadening of 100 Hz was applied to the free induction decay before Fourier transformation. The number of accumulated scans varied from 500,000 to 2,000,000. In general, more scans were required to obtain a good signal/noise ratio for samples with a covalently bound chain, possibly due to increased transverse relaxation rates, as previously observed for lipids in bilayers with high peptide concentrations (28). Peaks at or near the isotropic position were ascribed to residual deuterons in the water with contributions from deuterons in the silicon stopper used to seal the samples. ³¹P NMR was used to confirm that all samples were bilayers and that no nonbilayer components were present. All ³¹P NMR experiments were performed as described previously (29).

Deuterium spectra recorded with di-14:0-PC-d27 were numerically deconvoluted to yield the θ = 0° spectra (22,30), resulting in so-called de-Paked spectra. The de-Paked spectra were then deconvoluted to assign the peaks to carbon atoms along the sn-2 acyl chain. The quadrupolar splittings $\left(\Delta v_Q\right)$ determined from the de-Paked spectra were used to calculate the order parameter, S_CD(i), according to:

$$\Delta v_{Q\left(i\right)}=\frac{3}{4}\left(\frac{e^{2}qQ}{h}\right)\left(3{\text{cos}}^2\theta -1\right)S_{CD\left(i\right)},$$ (1)

where e²qQ/h = 167 kHz (22), and θ is the angle between the bilayer normal and the static magnetic field direction. For θ = 0°, (3 cos² θ − 1) equals unity (26). From the determined order parameters, we estimated the hydrocarbon thickness (D_C) of one monolayer as described previously (31,32).

Results

Chain order of a peptide-linked palmitoyl chain

Our first aim in this study was to obtain information about how palmitoyl chains linked to a transmembrane peptide insert into phospholipid bilayers. To accomplish this, we covalently coupled a perdeuterated palmitoyl chain to a cysteine residue at the N-terminus of a WALP23 peptide, and incorporated the peptide into di-18:1-PC bilayers. Several lines of evidence suggest that this modification does not affect the peptide-lipid interaction. First, circular dichroism experiments on such samples, performed as described previously (33), showed that the modification does not affect the conformation of the peptide (Fig. S1). Second, as measured by ²H NMR on DMPC with perdeuterated chains, the palmitoylated WALP23 peptide was observed to influence the surrounding lipid bilayer in similarity to unpalmitoylated WALP23 (Fig. S2). Third, the orientation of the helical peptide in the bilayer was not altered significantly by addition of the palmitoyl chain, based on deuterium spectra recorded with palmitoylated peptide with a deuterium-labeled alanine side chain (Fig. S3).

Next, ²H NMR spectra of the palmitoylated peptide were recorded and compared with those of free perdeuterated palmitic acid (PA) and di-16:0-PC with a perdeuterated sn-2 chain in di-18:1-PC bilayers (Fig. 1). Fig. 1 A shows the ²H NMR spectrum of 2 mol % di-16:0-PC-d31 in di-18:1-PC. The spectrum shows a number of peak doublets originating from the different ²H-label positions. The difference in frequency between these peaks, or the quadrupolar splitting (Δv_q), is a measure of the order of the palmitoyl chain at that position. The central pair of peaks with a Δv_q of ∼2 kHz corresponds to the terminal methyl group positioned in the disordered center of the bilayer, whereas the outer peaks with a Δv_q of ∼25 kHz correspond to positions in the ordered region close to the carbonyl group. This spectrum closely resembles those reported for samples of pure deuterium-labeled di-16:0-PC in the liquid-crystalline state (22,31,34). From the order parameter profile (shown in Fig. S4), the hydrophobic length of the perdeuterated chain was calculated as described previously (31) to be 14.8 Å. This length is similar to what was previously reported for di-16:0-PC in the liquid-crystalline state (31), and slightly longer than the 13.6 Å reported for di-18:1-PC acyl chains at the same temperature (35).

Figure 1.

Deuterium NMR spectra of 2 mol % di-16:0-PC-d31 (A), 2 mol % perdeuterated free d31-PA (B), and 2 mol % perdeuterated peptide-linked PA (C) in di-18:1-PC bilayers at 30°C.

Overall, rather similar ²H NMR spectra were recorded with free perdeuterated PA in di-18:1-PC bilayers (Fig. 1 B), suggesting that free palmitoyl chains incorporate into di-18:1-PC bilayers much like the lipid-bound chains. The hydrophobic length of the palmitoyl chain was calculated to 14.2 Å. This was less than observed for the lipid-bound chain, suggesting that the free palmitoyl chain is more affected by the bulk lipid than the lipid-bound one. Fig. 1 C shows that the peptide-linked perdeuterated palmitoyl chains also incorporate similarly. However, small but distinct differences are observed in the patterns of the ²H NMR spectra. For example, the spectrum of the peptide-linked chain has a lower intensity of the outer peaks relative to the peaks corresponding to positions closer to the terminal methyl group as compared with the spectra of the lipid-bound and free chains. In addition, Δv_q of the outer peaks in the spectra of peptide-linked palmitoyl chains is ∼23 kHz, i.e., slightly smaller than those found for lipid-bound and free chains (∼24 kHz). The calculated hydrophobic length of the peptide-linked palmitoyl chain is 13.6 Å. Thus, the palmitoyl chain seems to be only slightly perturbed by covalent attachment to the peptide.

Influence of acyl chain length on lipid chain order

To obtain information about how the peptide-linked palmitoyl chains sense the properties of the phospholipid bilayer, we incorporated the peptides into bilayers of different thickness. Unfortunately, although deuterium NMR measurements of d₃₁-palmitoyl-WALP23 in di-18:1-PC bilayers gave well-resolved spectra, the spectra became less well-resolved in the thinner bilayers, showing a different intensity distribution especially near the region attributed to the membrane-water interface (Fig. S5). Therefore, to allow straightforward comparison, palmitoyl chains with deuterium labels in specific positions farther away from the membrane-water interface were used. Unsaturated lipids were used to allow comparison of effects at one temperature, well above the gel-liquid-crystalline phase transition temperature of all lipids. Control experiments were performed with a freely diffusing palmitoyl chain in the same bilayer in the presence of WALP23.

Deuterium NMR spectra of 2 mol % free and peptide-linked 13,13-d₂-palmitoyl chains in PC bilayers are shown in Fig. 2 (the values of the quadrupolar splittings are given in Table S1). Linking the palmitoyl chains to WALP23 significantly decreased the signal/noise level, most likely due to increased transverse relaxation rates, as previously observed for lipids in bilayers with high peptide concentrations (28). In addition, isotropic peaks appeared that were ascribed to residual deuterons in the water, with contributions from deuterons in the silicon stopper used to seal the samples. Nevertheless, well-resolved splittings were obtained. For the free 13,13-d₂-palmitoyl chains in di-14:1-PC bilayers in the presence of equimolar amounts of peptide, the main splitting was 6.8 kHz. As shown in Fig. 2, the value of this splitting increases with increasing bilayer thickness. Similar effects are observed for 2 mol % of the peptide-bound 13,13-d₂-palmitoyl-chain, except that the quadrupolar splittings are slightly smaller, as quantified in Fig. 2 A. In both cases, the quadrupolar splittings depend nearly linearly on the bilayer thickness in the tested range.

Figure 2.

Effects of bilayer thickness on palmitoyl chain order. (A) Deuterium NMR spectra of free and WALP23-linked deuterium-labeled PA in di-14:1-PC, di-16:1-PC, and di-18:1-PC bilayers at 30°C. Palmitoyl chains labeled at carbon 13 were used, and all of the samples contained 2 mol % of the probe (free or peptide-linked). Spectra of the free palmitoyl chains were recorded in the presence of 2 mol % WALP23. (B) Quadrupolar splittings of the main peak doublet of deuterium-labeled palmitoyl chains in di-14:1-PC, di-16:1-PC, and di-18:1-PC bilayers at 30°C. Spectra were recorded with 2 mol % free or WALP23-linked 13,13-d₂-PA.

When the label is located closer to the interface at position 9, similar trends are observed for both the free acyl chain and the peptide-bound chains (Table 1). Again, the quadrupolar splitting increases with bilayer thickness and also here it is smaller for the peptide-bound chain than for the free PA. The most notable difference in the results for the 13,13-d₂-palmitoylchain is the much larger quadrupolar splitting obtained for the 9,9 position, which is in good agreement with previously published results regarding the order parameter profile of perdeuterated di-16:0-PC bilayers in the liquid crystalline state (31,34).

Table 1.

Quadrupolar splittings in kHz meassured with 2 mol % free or peptide-bound 9,9-d₂-PA in PC bilayers without and with cholesterol at 30°C

Lipids	PA	PA+WALP23	PA+WLP23	PA+KALP23	PA −WALP23	PA −WLP23	PA −KALP23
Di-14:1-PC	15.8	16.3	17.6	16.0	14.2	15.0	13.9
Di-16:1-PC	19.0	19.6	20.6	18.7	17.9	18.7	16.1
Di-18:1-PC	20.9	21.6	22.0	20.8	19.6	20.1	17.8
Di-14:1-PC + cholesterol
5 mol %	17.5	17.9	19.1	18.1	15.5	16.2	15.1
10 mol %	18.9	19.0	20.8	18.6	17.2	17.4	15.5
15 mol %	20.8	21.0	22.2	21.2	18.3	19.1	16.7
20 mol %	22.0	21.6	23.3	22.7	19.5	19.8	17.6

Standard deviations of all of the reported quadrupolar splitting were ≤0.2 kHz.

Influence of cholesterol on lipid chain order

Inclusion of cholesterol is known to increase the acyl chain order in PC bilayers in a concentration-dependent manner. To obtain information about how the peptide-lipid interface senses the presence of cholesterol, we recorded ²H NMR spectra of free and WALP23-linked 9,9-d₂-palmitoyl chains incorporated into di-14:1-PC, with an increasing concentration of cholesterol. Representative spectra are shown in Fig. 3 A. All spectra contain only one peak doublet, suggesting that all PA (free or peptide-linked) is present in the same lipid environment independently of the cholesterol concentration. As the cholesterol concentration in the bilayers increases, the quadrupolar splittings of the peak doublets increase for both the free and peptide-bound palmitoyl chains. Fig. 3 B shows that the increase in order is linear with cholesterol concentration up to 20 mol % sterol, which was the highest concentration used in this study. Because the slopes of these lines are the same for the peptide-linked chains as for the free 9,9-d₂-palmitoyl chains, it furthermore can be concluded that cholesterol has a similar ordering effect on both types of acyl chains.

Figure 3.

Effects of cholesterol on palmitoyl chain order, as revealed by the spectra of free and WALP23-linked 9,9-d₂-PA in di-14:1-PC with different concentrations of cholesterol. Deuterium NMR spectra of membranes containing 0, 10 and 20 mol % sterol (A) and the observed quadrupolar splittings (B) as a function of sterol concentration are shown. All spectra were recorded at 30°C with 2 mol % of the deuterium-labeled palmitoyl chains (free or peptide-linked). Spectra of the free 9,9-d₂-palmitoyl chains were recorded in the presence of 2 mol % WALP23.

Influence of the peptide structure on lipid chain order

From the results presented above, it can be concluded that palmitoyl chains that are covalently coupled to transmembrane peptides sense the lipid environment similarly to bulk lipids, although the covalently bound chains are systematically less ordered. Next, we investigated the extent to which the acyl chains sense the properties of the peptides to which they are covalently linked. For this, we chose to include two other model peptides: WLP23 and KALP23. In WLP23, the Leu-Ala sequence in WALP23 has been replaced by Leu-Leu, leading to a poly-Leu sequence, which smoothes the outer contour and increases the hydrophobicity of the transmembrane helix. In KALP23, the flanking Trp residues in WALP23 are replaced by Lys, which is expected to affect interfacial anchoring interactions.

Deuterium NMR spectra of free and WLP23-linked 9,9-d₂-palmitoyl chains in PC bilayers of different thickness and in the presence of cholesterol are shown in Fig. 4. The spectrum of 2 mol % free 9,9-d₂-palmitoyl chains in di-14:1-PC bilayers with 2 mol % WLP23 shows a quadrupolar splitting of 15.0 kHz, which increases with increasing acyl chain length and when cholesterol is present. The peptide-linked chain shows a very similar trend. The quadrupolar splittings from these experiments are given in Table 1. The results show that, in similarity to WALP23, for the WLP peptides the covalently coupled chain has a clearly reduced value of the quadrupolar splitting as compared with the free fatty acid, and the chain order for both the free and linked palmitoyl chains increases with the lipid acyl chain length and the cholesterol content of the host bilayer in a manner very similar to that found with WALP23. However, a striking difference in the results for WALP23 is that for WLP23, systematically higher values are found for the quadrupolar splittings.

Figure 4.

Deuterium NMR spectra of free and WLP23-linked 9,9-d₂-palmitoyl chains in different lipids bilayers at 30°C. Spectra of 2 mol % free or WLP23-linked chains in different PC bilayers, and in a di-14:1-PC bilayer with 20 mol % cholesterol are shown. The spectra of free 9,9-d₂-PA were recorded in the presence of 2 mol % WLP23.

Fig. 5 shows deuterium NMR spectra of free 9,9-d₂-palmitoyl chains in the presence of 2 mol % Lys-flanked KALP23 and of 9,9-d₂-palmitoyl-KALP23 in PC bilayers of different thickness. The measured quadrupolar splitting are listed in Table 1. For the free fatty acid, clearly defined quadrupolar splittings are again observed, which increase with lipid chain length and upon incorporation of cholesterol. However, the spectra of 9,9-d₂-palmitoyl-KALP23 seem to contain a second, broader component with slightly larger splittings that is most clear in the di-14:1-PC bilayer. Characterization of the 9,9-d₂-palmitoylated KALP23 by mass spectrometry showed that some of the peptides were doubly labeled, having a palmitoyl chain also linked to one of the lysines (results not shown). Because the Lys side chains have different, closely spaced attachment sites, the intensities from different labeling positions are expected to result in broader peaks. Therefore, we concluded that the second component probably derives from this population, and hence focused our analysis on the major component. As shown in Table 1, the order of this KALP23-bound chain increases with the lipid chain length and inclusion of cholesterol, again in similarity to the free fatty acid. Also, in similarity to the WALP and WLP peptides, for KALP23 the covalently coupled chain has a clearly reduced value of quadrupolar splitting compared with the free fatty acid, and in all cases the quadrupolar splitting increases with the acyl chain length and cholesterol content of the host bilayer. In addition, a striking difference as compared with the results for WALP23 and WLP23 is that for KALP23, all of the quadrupolar splitting values are systematically smaller.

Figure 5.

Deuterium NMR spectra of free and KALP23-linked 9,9-d₂-palmitoyl chains in lipids bilayers at 30°C. Spectra of 2 mol % free or KALP23-linked chains in different PC bilayers, and in di-14:1-PC bilayer with 20 mol % cholesterol are shown. The spectra of the free 9,9-d₂-PA were recorded in the presence of 2 mol % KALP23.

To test whether the systematic differences among WALP23, WLP23, and KALP23 with regard to quadrupolar splittings, as shown in Table 1, are directly due to differences in the influence of these peptides on the lipids, we tested their effect on lipid chain order in a simple lipid system, DMPC. Fig. 6 A shows the deuterium NMR spectra of a di-14:0-PC-d27 bilayer without and with peptide (peptide/lipid ratio = 1:30). From the de-Paked spectra (not shown), we were able to separate the peaks arising from deuterium atoms at different positions along the acyl chains and to prepare order parameter profiles for di-14:0-PC sn-2 chains without and with peptides (Fig. 6 B). We calculated the hydrophobic thickness (2×D_C) of the bilayers from the order parameters as described by Petrache and co-workers (31). The hydrophobic thickness was calculated to be 24.9 Å for pure di-14:0-PC. In the presence of peptides, the hydrophobic thickness of the bilayers was 25.1 Å with KALP23, 26.4 Å with WALP23, and 27.7 Å with WLP23. These results show that the systematic differences among the effects of the peptides, as shown in Table 1, directly correlate with the differences in their effect on lipid chain order.

Figure 6.

Effects of peptides on phospholipid chain order. (A) Deuterium NMR spectra of di-14:0-PC-d27 were recorded without and with peptide (peptide/lipid ratio = 1:30) at 40°C. (B) Order parameter profiles were prepared from de-Paked spectra.

Discussion

It is generally believed that the interplay between proteins and lipids is important for both the organization of lipids in biological membranes and the structure and function of membrane proteins. In this study, we focus on the effect of proteins on the lipid environment. Clear evidence of such effects has been observed in studies of the interactions between lipids and model peptides that mimic protein transmembrane helices, such as the WALP peptides (reviewed by de Planque and Killian (4)). In previous studies on the effects of WALP-like peptides on lipids, the measured effects were averaged over the bulk lipid in the membrane (5,36). In this work, we specifically investigated the lipid-protein interface by studying the properties of deuterium-labeled palmitoyl chains covalently linked to α-helical transmembrane peptides. This allowed us to monitor the chain order of the palmitoyl chains at the interface and at concentrations where the probability of peptide-peptide interactions is low. As a control, the behavior of free fatty acids was monitored in a bilayer containing equimolar amounts of the corresponding peptides.

Free fatty acids as probe for chain order of bulk lipid

Deuterium-labeled fatty acids incorporated into phospholipids are commonly used as probes to analyze lipid chain order (37). However, the use of free deuterium-labeled fatty acids is not as well documented. Because our approach in this study was to use such fatty acids as probes for the effects of WALP-like peptides on the surrounding lipid environment, we first performed tests of how the fatty acids incorporate in the lipid bilayer. From the similar appearance of the recorded ²H NMR spectra of PA-d31 and di-16:0-PC-d31 in di-18:1-PC bilayers, it was clear that free palmitoyl chains insert into PC bilayers and sense the chain order in much the same way as do PC incorporated palmitoyl chains (Fig. 1), indicating that they may be used as probes to monitor bulk properties of the lipids. Similar observations have been made for acyl chains attached to LV16 peptides and Ras peptides (25,38). The recorded spectra (Fig. S5) in bilayers of different thickness suggest that the hydrophobic length of the peptide-linked palmitoyl chain was adapting to match that of the bulk lipids. This is in line with what previously has been observed for palmitoyl chains that were used to link a water-soluble Ras peptide to the membrane (24). Of interest, with the transmembrane peptide LV16, no such adaptation of the linked acyl chains to the surrounding lipid bilayer has been observed (25). Most likely, this inflexibility of the LV16-linked chains derives from structural features of the peptide and its reported tendency to form β-sheets (26). For example, it is possible that in these structures the chains somehow become shielded from the lipids.

With this approach of using deuterated reporter chains, it is important to keep in mind that in principle the free fatty acids may not accurately reflect the properties of the bulk lipids, because, for example, their lateral distribution and/or vertical positioning in the membrane could deviate from those of the bulk lipids. However, our results suggest that the free fatty acids can accurately monitor changes in lipid chain order, based on 1), measurements of the effects of changes in bilayer thickness and/or cholesterol inclusion on the behavior of the free fatty acid probe (discussed below); and 2), a comparison of how free fatty acid and phospholipid acyl chains respond to WALP23 addition to the bilayers (Fig. S6). In addition, the fact that only one spectral component was detected in all of the experiments with free fatty acids indicates that in all circumstances the fatty acids were homogeneously distributed in the bilayer.

How changes in the lipid environment are sensed at the peptide-lipid interface

It is known that transmembrane peptides can influence the surrounding lipid environment (5), for example, as a response to a hydrophobic mismatch. Because these effects occur at the peptide-lipid interface, one could assume that the annular lipids that are in direct contact with the protein/peptides are affected more than the bulk lipids. Does this mean that the lipids at the surface of the peptide have different properties than a distant lipid? Molecular-dynamics and coarse-grained simulations predict precisely that (39–41). In this study, we addressed the question by experimentally comparing how free and peptide-linked palmitoyl chains sense different bulk lipids.

In a first approach, we varied the bilayer thickness by using three different PCs: di-14:1-PC, di-16:1-PC, and di-18:1-PC. In di-18:1-PC bilayers, the hydrophobic length of WALP23 should match well with the hydrophobic thickness of the bilayer (42), and the degree of positive hydrophobic mismatch should increase with decreasing acyl chain length. Hence, one would expect that upon incorporation into relatively thin bilayers, the covalently linked chain would be stretched more than a free chain, and a much smaller difference would be observed in thicker bilayers. However, comparative measurements of free and WALP23-linked palmitoyl chains showed that the differences in chain order of both chains (free and bound) were similar in all bilayers. Thus, our results show no indication of preferential stretching of the lipids next to the peptide under conditions of positive hydrophobic mismatch.

As a second approach to study how the chain order at the surface of a transmembrane helix is affected by properties of the surrounding lipid bilayer, we examined the effects of cholesterol inclusion on the palmitoyl chains. Cholesterol is known to have an ordering effect on unsaturated acyl chains in PC bilayers (8); hence, the bilayer thickness should increase with increasing cholesterol concentration. Here, the free deuterium-labeled palmitoyl chains also reported a cholesterol-dependent linear increase in chain order that was mimicked by the peptide-linked chains. Therefore, we can conclude that changes in the bilayer thickness affect the free and peptide-linked palmitoyl chains in the same way, leading to the surprising conclusion that changes in bulk lipid properties are sensed similarly at the peptide-lipid interface and two or three lipids away from the transmembrane helices.

The conclusion that bulk lipids and lipids at the peptide surface respond similarly to hydrophobic mismatch is unexpected, particularly if one considers lipid stretching to be the major response to hydrophobic mismatch. A possible explanation could be that in our systems, peptide tilt in response to positive hydrophobic mismatch (43,44) is a more dominant response than lipid chain adaptation. Alternatively, it cannot be excluded that some lipid stretching does occur, but it does so preferentially at the plateau region close to the headgroup and outside the region that is sensed by the reporters at positions 9 and 13 (as used in this study).

Another striking observation is that the quadrupolar splittings were consistently found to be smaller when the palmitoyl chains were coupled to peptides than when they were free in the bilayer in the presence of the same peptide. Similar observations were made in previous studies when deuterium-labeled palmitoyl chains were covalently linked to gramicidin or LV16 peptides (13,25). One possible explanation is that the peptide-linked chains are positioned differently in the bilayer compared with the free fatty acid and the phospholipid-incorporated chains. Also, the difference between peptide- and phospholipid-linked chains may be influenced by the nature of the chemical linkage. Alternatively, it is possible that the peptide has a direct influence on the orientation and/or motion of the acyl chain, resulting in a slightly increased disorder. This explanation seems likely, because recent molecular-dynamics simulations have indicated that WALP-like peptides may be quite dynamic (44–47).

Importance of peptide composition

The effect of transmembrane helices on lipid chain order in bilayers composed of saturated PC has been shown to depend on the hydrophobic length of the transmembrane segments (5). Here, we observed that WLP23, with a hydrophobic length similar to that of WALP23, has a larger ordering effect on the surrounding lipids than WALP23, suggesting that factors other than hydrophobic matching also influence the extent to which transmembrane segments perturb lipid bilayers. These findings fit well with the observations that the antibacterial peptide gramicidin A stretches the acyl chains in adjacent PCs more than WALP peptides of similar hydrophobic length (5), and that although the interactions between different Lys-flanked transmembrane peptides and the surrounding lipids are influenced by the amino acid composition of the transmembrane segments (48,49), the hydrophobic length remains the same.

In line with this, the WLP23-linked palmitoyl chains reported a higher chain order than was observed at the surface of WALP23 in all bilayers. Importantly, as observed with WALP23, the WLP23-linked chains and the free palmitoyl chains responded similarly to increases in the bilayer thickness, caused either by increasing the acyl chain length or including cholesterol, suggesting that the nature of the hydrophobic transmembrane segment does not influence the effect of the bulk lipids on the lipid chain order at the peptide surface.

The chain order at the surface was lower for KALP23 than for WALP23, in agreement with the observation that KALP23 had no ordering effect on the surrounding phospholipids (Fig. 6). When linked to KALP23, the acyl chains were also less sensitive to the hydrophobic thickness of the lipid bilayers than when linked to WALP23 or WLP23. Both the lower chain order and the decreased sensitivity to the hydrophobic thickness of the surrounding bilayers may be related to the anchoring properties of the peptide at the lipid-water interface, which result in a smaller effective hydrophobic length of KALP23 similar to that of a shorter WALP peptide (36,50). This would mean that in the thicker bilayers, there is a negative hydrophobic mismatch between KALP23 and the lipid bilayer. Most likely, the relatively small sensitivity of the KALP23-linked fatty acid to the changes in the hydrophobic thickness of the bilayers is a consequence of this relatively short length, because chain stretching and subsequent adaptation to the thicker bulk lipid environment would be more difficult than they would be for acyl chains that are linked to the effectively longer Trp-flanked peptides. Consistent with this, the free fatty acid was affected similarly by the hydrophobic thickness of the bilayers in the presence and absence of KALP23, suggesting that the bilayer properties dominated over those of the peptides except at the peptide surface.

Concluding Remarks

In this work, we studied the chain order of deuterium-labeled palmitoyl chains covalently linked to transmembrane peptides. By comparing the spectra of deuterated peptide-linked palmitoyl chains with those of free fatty acyl chains incorporated into PC bilayers, we were able to conclude that free fatty acids and chains that are in direct contact with the peptide surface react more or less similarly to changes in bulk bilayer properties. Thus, changes in bulk lipid properties are sensed similarly at the peptide-lipid interface and in bulk lipid, and hence, at least for these single-span model peptides, annular lipids do not appear to play a specific role in protein-lipid interaction. However, the situation may be different for larger proteins, which are less dynamic and have a larger cross-sectional diameter, thereby presenting a less curved surface toward the lipids and possibly allowing more stable interactions. In addition, in such proteins other adaptations to hydrophobic mismatch, such as helical tilt, may be less prominent, as has been predicted from computational studies (40). Further, the presence of helical bundles offers a more structured surface, possibly with pockets that can bind lipids relatively strongly. Hence, it is likely that the role of annular lipids is dependent on the cross-sectional size and three-dimensional structure of the bilayer-spanning part of membrane proteins.