Buried lysine, but not arginine, titrates and alters transmembrane helix tilt

Abstract

The ionization states of individual amino acid residues of membrane proteins are difficult to decipher or assign directly in the lipid–bilayer membrane environment. We address this issue for lysines and arginines in designed transmembrane helices. For lysines (but not arginines) at two locations within dioleoyl-phosphatidylcholine bilayer membranes, we measure pK_a values below 7.0. We find that buried charged lysine, in fashion similar to arginine, will modulate helix orientation to maximize its own access to the aqueous interface or, if occluded by aromatic rings, may cause a transmembrane helix to exit the lipid bilayer. Interestingly, the influence of neutral lysine (vis-à-vis leucine) upon helix orientation also depends upon its aqueous access. Our results suggest that changes in the ionization states of particular residues will regulate membrane protein function and furthermore illustrate the subtle complexity of ionization behavior with respect to the detailed lipid and protein environment.

The basic residues lysine (Lys) and arginine (Arg) in membrane proteins are enriched at the interfacial region and are believed to assist in anchoring the transmembrane orientations (1–3). In the hydrophobic core of transmembrane proteins, these polar amino acids are significantly depleted because of the energetic cost of burying them in the nonpolar lipid environment. Nevertheless, although their membrane occupancy is diminished, in many proteins, Lys and Arg can still be found within the bilayer core where they can perform important functional roles. Within this context, the ionization states of basic Lys and Arg side chains may regulate the biological functions of membrane proteins, including sodium channels (4), acetylcholine receptors (5), and integrins (6), among others. For example, charge–charge interactions are implicated for regulating the voltage gating of cation channels (7), and mutations that alter the charge state of an ionizable residue influence the ion selectivity of nicotinic-type receptors (5). Despite their noted importance, the actual ionization states of particular residues in a lipid-bilayer membrane environment are difficult to distinguish directly.

For example, the transmembrane domains of integrin proteins possess conserved basic residues that are essential for signaling (6). The four-helix voltage sensor of voltage-gated potassium channels has multiple Arg and Lys residues (8) that are believed to be charged and to move in response to changes in membrane potential (ie, to open or close the channel) (4, 8). Debate is ongoing as to how these charges may shift to facilitate conformational changes, with the principal discussions involving the probability for a charged side chain, specifically Arg, to reside within the lipid bilayer (9). Although some molecular dynamics simulations (10, 11) have estimated larger energy barriers than translocon-mediated (12) or water-to-bilayer membrane protein–folding experiments (13) for moving Arg into the bilayer, careful consideration of experimental conditions and the underlying molecular context is leading to a convergence of findings from different methods (14–16). In particular, Arg side-chain snorkeling, helix off-center translocation, and membrane deformation all can contribute to maintaining (partial) hydration and thereby lowering the free energy for accommodating a charged residue within a lipid bilayer (14–18). Additionally, for cases in which the aqueous access to Arg may be impeded, structural manipulations to remove the obstruction can restore the accommodation of Arg within a bilayer (19). In relatively thin dilauroyl-phosphatidylcholine bilayer membranes, the cost of placing Arg within a transmembrane helix of outer membrane phospholipase is about 3.7 kcal/mol at pH 3.8 (13, 16, 18), whereas a somewhat larger cost was found for placing Lys at the same position at pH 3.8 (13). Within the context of the existing experiments and simulations, the pH dependence of side-chain ionization merits consideration.

Indeed Coulombic interactions are intensified in the lipid bilayer compared with aqueous solution because of the low dielectric medium. A favorable and sensitive host framework, with measurable pH-dependent properties, is therefore necessary for assessing the ionization states of individual residues. The peptides GWALP23 (3) and Y⁵GWALP23 (20) form particularly favorable 23-residue transmembrane helices, of sequence acetyl-GGAL[W/Y](LA)₆LW¹⁹LAGA-amide, for which the helix tilt is especially sensitive to single-residue substitutions. The host framework is remarkably favorable because of the low extent of dynamic averaging when only one aromatic Trp or Tyr residue flanks each end of a core transmembrane helix (20, 21). Within this framework, it was discovered that Arg confers markedly different properties when present at position 12 (R12) compared with position 14 (R14) (17). Indeed, GWALP23-R14 simply adjusts its tilt and helix azimuthal rotation to provide Arg access to the aqueous interface, whereas GWALP23-R12, in which the Arg is sandwiched between opposing Trp indole rings, adopts radically different multiple orientations that interconvert only slowly in dioleoyl-phosphatidylcholine (DOPC) bilayer membranes (17). In these experiments, the helix orientations and dynamics are assessed by means of solid-state ²H NMR spectroscopy using specifically labeled alanine residues (21).

Here we report the consequences of introducing Lys at position 12 or 14 in the membrane-spanning GWALP23 and Y⁵GWALP23 host peptides. Models for several of the peptide helices are shown in Fig. 1. At low pH, when Lys bears a positive charge, the K12 and K14 peptides behave similarly to the corresponding R12 and R14 peptides in DOPC bilayer membranes. Unlike Arg, which in our framework remains charged up to the experimentally accessible pH 9 in DOPC, we observe titration of the Lys in Y⁵GWALP23. (Above the pH 9 limit, the DOPC lipids begin to hydrolyze.) We present pH-dependent ²H NMR spectra for labeled alanines in Y⁵GWALP23-K12 and -K14, analysis of changes in transmembrane helix orientation when each single Lys is titrated, titration curves, a pK_a value for Y⁵GWALP23-K14, and a pK_a upper limit for Y⁵GWALP23–K12.

Fig. 1.

Models to show peptide helix transmembrane orientations in DOPC bilayers as function of lysine residue position and ionization state. (A) Y⁵GWALP23-K⁺14. (B) Y⁵GWALP23-K⁰14. (C) Y⁵GWALP23-K⁰12. (D) Y⁵GWALP23 itself, with no ionizable residue. See text for the peptide sequences. The side chain of residue 12 is shown as sticks for either leucine (A, B, and D) or lysine (C). [Note that Y⁵GWALP23-K⁺12 exhibits multistate behavior (as does GWALP23-R⁺12 (17)) and would not have a single orientation.] The peptide GWALP23-R⁺14 (17) orients similarly to Y⁵GWALP23-K⁺14, shown in (A). Note that Y⁵GWALP23-K⁰12 (C) orients similarly to Y⁵GWALP23 itself (D). An alternate perspective (top view) is shown in (A2–D2). See also the precession of tilted peptides about the membrane normal, shown in Movie S1.

Results

In each of the Y⁵GWALP23-K12 and -K14 peptides, pairs of alanines within the core (LA)₆ helix were ²H-labeled to different extents (60% and 100% deuteration, to facilitate assignment of NMR resonances). Fig. 2 shows example ²H NMR spectra for labeled alanines 15 and 17 in Y⁵GWALP23-K14 and Y⁵GWALP23-K12 at pH 5.2 and 8.2. Similar to the case of Arg (17), at low pH, distinct major ²H resonances are observed for the alanines only in Y⁵GWALP23-K14, but not in Y⁵GWALP23-K12 (Fig. 2). Indeed, the K12 peptide exhibits multiple weak resonances that suggest, again, multistate behavior, as was observed also with GWALP23-R12 (17). When the pH is raised to 8.2, each Lys residue releases a proton, yielding new ²H NMR spectra for Y⁵GWALP23-K12 and -K14 (Fig. 2). The newly observed quadrupolar splittings for the alanine methyl groups reveal a single major transmembrane orientation for Y⁵GWALP23-K⁰12 and a different helix tilt for Y⁵GWALP23-K⁰14 at pH 8.2, when the Lys (K⁰) are deprotonated (in contrast to the multistate behavior of Y⁵GWALP23-K⁺12 and the orientation of Y⁵GWALP23–K⁺14 at lower pH). We have not observed any such titration behavior for Arg in the corresponding -R12 or -R14 peptides over the pH range accessible for our experiments (up to ambient pH 9.0, corresponding to pH 8.2 at the NMR acquisition temperature of 50 °C). The corresponding Args have remained charged in our experiments, a result that has been suggested by some simulations (10, 11, 22).

Fig. 2.

²H NMR spectra for labeled alanines 15 (50–70% deuterated) and 17 (100% deuterated) in peptides incorporated into hydrated, oriented bilayers of DOPC. (A) Y⁵GWALP23-K14 at pH 5.2. (B) Y⁵GWALP23-K14 at pH 8.2. (C) Y⁵GWALP23-K12 at pH 5.2 (showing multistate behavior). (D) Y⁵GWALP23-K12 at pH 8.2.

Based upon the ²H quadrupolar splittings for all six core alanine methyl groups (Fig. S1 and Table S1), we analyzed the helix tilt for each of the transmembrane peptides (21, 23). (Although excessive peptide dynamics does complicate the determination of helix tilt for some peptides (24–26), the problem is mitigated and the dynamics become tractable when only two aromatic residues are present (3, 21) and additionally when a polar residue is present (17, 21). We therefore estimated the rather minimal dynamics of the K12 and K14 peptides using a principal order parameter (21).)

The tilted transmembrane orientation of Y⁵GWALP23-K⁰12 is identical to that of Y⁵GWALP23 itself (Fig. 3). Remarkably, it makes no difference whether neutral Lys or leucine occupies position 12, which resides between the aromatic residues Y5 and W19, on the same helix face, with seven residues from each of them. By contrast, residue K⁰14, on the opposite helix face, confers a somewhat different helix orientation from that of native Y⁵GWALP23 with L14 (Fig. 3), a change of tilt Δτ of about 3° and a rotation Δρ of about 80° (Fig. 3). At lower pH, the positively charged K⁺14 causes a further Δτ of about 6° and a further Δρ of about 15°, very similar to the influence of R⁺14 in GWALP23 (17). The respective peptide orientations are represented in Fig. 1 and the major dynamic feature, helix precession about the bilayer membrane normal, is illustrated in Movie S1.

Fig. 3.

Quadrupolar wave analysis of tilted transmembrane peptides. (A) Y⁵GWALP23-K14 at pH 5.2 (red; tilt τ = 15°, rotation ρ = 228°) and pH 8.2 (blue; tilt τ = 9°, rotation ρ = 244°). (B) Y⁵GWALP23-K12 at pH 8.2 (blue; tilt τ = 5°, rotation ρ = 308°). The black curves in (A) and (B) represent Y⁵GWALP23 (tilt τ = 5°, rotation ρ = 311°, irrespective of pH). (C) rmsd plots for tilt and rotation of Y⁵GWALP23 (black), Y⁵GWALP23-R14 (orange), Y⁵GWALP23-K14 (red, pH 5.2), and Y⁵GWALP23-K14 (blue, pH 8.2). (D) rmsd plot for tilt and rotation of Y⁵GWALP23-K12 at pH 8.2. Contours are drawn at 1, 2, and 3 kHz.

By monitoring selected alanine methyl group ²H quadrupolar splittings, we obtained a titration curve for Lys in Y⁵GWALP23–K14 (Fig. 4A). Notably, the Lys side chain is the only ionizable group in this peptide. Furthermore, the curves for the pH dependence of the quadrupolar splittings of ²H-A15 and ²H-A17 in the peptide agree on a pK_a of 6.2 (Fig. 4A) at the NMR recording temperature of 50 °C. (Buffer pH values were temperature corrected; see Materials and Methods.) We apply the known temperature dependence of the Lys side chain pK_a (27) to deduce values of ∼6.5 at 37 °C and ∼6.8 at 25 °C for K14 in Y⁵GWALP23-K14. We note also in Fig. 4A that the NMR observables for peptides having the R14 and L14 side chains do not titrate. When L14 and L12 are present, in GWALP23 itself, there exist no ionizable groups, such that titration is not expected, and indeed is not observed. The further observation that Arg itself remains charged agrees with results for buried Arg in soluble proteins (28).

Fig. 4.

(A) Titration curve for Y⁵GWALP23-K14 in oriented DOPC bilayer membranes at 50 °C, based on the ²H quadrupolar splittings for deuterated A15 (black squares) and A17 (red circles). Both datasets agree on a pK_a of 6.2 at 50 °C (dashed vertical segment). When residue 14 is arginine (dashed red segment; A17 data) or leucine (dashed blue segment), the titration behavior is not observed. (B) pH dependence of the main peak intensities for the ²H NMR resonances of labeled A15 and A17 of Y⁵GWALP23-K12 in DOPC. A 50% maximal peak intensity is achieved for both alanines at pH 7.0; 50 °C. The solid-state ²H NMR spectra are included in Figs. S2 and S3.

Discussion

Where is the titrating Lys (K14) residue located in the lipid bilayer? We estimate that the alpha-carbon (C_α) of K14 is about 6 Å from the bilayer center, based upon a 3 Å whole-helix displacement (17) and a 3 Å distance along the helix axis from C_α12 (the peptide center) to C_α14. Depending upon the extent of side-chain “snorkeling,” we estimate that the K14 ɛ-NH₃⁺ group is about 7–9 Å from the bilayer center. For such a location, a pK_a of ∼6.5 (four units below the Lys pK_a in aqueous solution) shows approximate agreement with the (force field dependent) estimates from molecular dynamics simulations of the relative partitioning of charged and neutral Lys analogs into lipid bilayers (11).

The pH-dependent behavior of K14 and K12 (see the following section) can be compared with results for buried Lys in the hydrophobic interior of soluble proteins. Among 25 single-site Lys mutations introduced into staphylococcal nuclease (29), 19 of the Lys residues substituted at internal positions exhibit depressed pK_a values, ranging from 5.3 to 9.3. The larger shifts in pK_a act to decrease protein stability and promote partial unfolding of nuclease in the lower pH range where the internal Lys becomes charged. Indeed, internal Lys-92 with the lowest pK_a (5.3) causes global unfolding of the nuclease when protonated (29). The observed pK_a values have been correlated with a lower dielectric environment in the immediate vicinity of the Lys, such that “apparent” dielectric constants of about 10–11 are estimated when the Lys pK_a is in the range of 6–7 (29). Also in the DOPC bilayer, at the intermediate location of K14 (about halfway between the bilayer center and the interfacial region, namely ∼7–9 Å from each; see the previous section), the dielectric constant will assume an intermediate value. The expected dielectric constant in the vicinity of K14 in GWALP23 is therefore consistent with the measured pK_a value of about 6.5 at 37 °C. We note also that a pK_a of about 8.7 has been suggested for the Asp residue in the transmembrane helix formed by the peptide K₂GL₇DLWL₉K₂A (30). Although the dielectric gradient through the bilayer is very steep, such that pK_a values will vary substantially with distance from the bilayer center (11), it is conspicuous that the reported values for Asp (30) and Lys (this work) are both shifted by about 4 pH units in favor of the neutral form in the bilayer compared with aqueous solution. Because K12 in GWALP23 is located closer to the bilayer center in a lower dielectric environment than K14, one expects an even lower pK_a for K12 than for K14.

Our observations are consistent with a low pK_a for K12. Because of the multistate properties when the Lys side chain is charged, K12 in Y⁵GWALP23-K12 does not exhibit a true titration behavior. At low pH, for the charged K12 residue, where there are multiple resonances arising from each labeled alanine methyl group, it is not possible to group particular sets of ²H-Ala resonances or to assign them to particular helix orientations. The same principle applies for the charged R12 residue (17). As a corollary, it also is not possible to extract pK_a values for K12 in the different helix states. Nevertheless, despite these limitations, a single peptide population emerges for the neutral K12 at high pH in DOPC (Figs. 2 and 3), and we can estimate the filling of this population as a function of pH. The major tilted transmembrane orientation of Y⁵GWALP23-K⁰12 is 50% populated at pH 7.0 at 50 °C (Fig. 4B). Because the K⁰12 population interconverts with multiple and incompletely defined states for the K⁺12 peptide at pH 7, the situation differs from a two-state equilibrium. As a consequence, the single-state population of the K⁰12 peptide will exceed the occupancy of any individual state of the K⁺12 peptide at pH 7. The estimate based on a midpoint for the predominant NMR intensity when the pH is 7.0 (Fig. 4B) hence constitutes an upper limit for the pK_a of K12 in Y⁵GWALP23-K12. Therefore, the experimental pK_a values for K12 and K14 are both below 7.0 for the respective peptides in DOPC bilayer membranes. Indeed, based on the position of Lys in the sequence, for the DOPC bilayer-incorporated peptides, one would expect that the pK_a of K12 in Y⁵GWALP23-K12 should be lower than the value of 6.5 that we observe for K14 in Y⁵GWALP23-K14. However, the multistate behavior, including an interfacial population, indicates that not all Y⁵GWALP23-K12 molecules span DOPC bilayer membranes at low pH.

The preferred orientations for the Y⁵GWALP23-K⁰12, -K⁰14, and -K⁺14 helices in DOPC bilayer membranes are summarized in Fig. 1 and Movie S1. Dynamic averaging around the helix axis is not of major importance for these peptides, even though such averaging is very significant for WALP23 (24, 25, 31), W^2,22W^5,19ALP23 (21), Y^4,5GWALP23 (20), and other similar peptides that have more than one Trp or Tyr residue at one or both ends of the core transmembrane helix (32). For GWALP23 and Y⁵GWALP23, therefore, the helix dynamics are dominated by precession of the helical axis about the membrane normal. Our major findings when a basic residue is substituted into this membrane-spanning α-helical framework include that: (i) the charged forms of Arg and Lys behave similarly near the bilayer center; (ii) the K14 and R14 peptides assume transmembrane topology regardless of the charge status of the single basic side chain; (iii) when the basic side chain at position 12 is charged, the K⁺12 and R⁺12 peptides both exhibit multistate behavior; (iv) neutral Lys at position 12 exerts no influence on the transmembrane helix tilt or rotation, but neutral Lys at position 14 changes the transmembrane helix azimuthal rotation by about 80°; and (v) in these DOPC-incorporated peptides, the Lys residues titrate to surrender a proton between pH 5 and pH 9, but the Arg residues remain charged. Rather than yield a proton, an uncompensated Arg⁺ residue will seek to exit the bilayer. In agreement with results for soluble proteins (28), and with predictions from simulations (10, 11, 22), we have not observed the presence of a neutral Arg side chain in these lipid–peptide systems between pH 4 and pH 9. We believe that these results will have important functional implications for the roles of basic side chains with regard to the mechanisms that govern the voltage gating of cation channels.

Materials and Methods

Labeled peptides were synthesized and oriented; hydrated lipid-peptide samples prepared on thin glass plates as described (20). Each synthetic peptide incorporated two deuterated alanines with differing isotope abundance, namely 100% and 60% deuteration. The oriented plate samples were hydrated using 50 mM acetate, Tris(hydroxymethyl)aminomethane, or bis-Tris buffer in deuterium-depleted water at pH values between 4 and 9. The temperature dependence of buffer pH was measured between 20 °C and 60 °C, and temperature coefficients were found to be −0.031 and −0.019 per degree Centigrade for the Tris and bis⋅Tris buffers, respectively. The enthalpies of ionization are, respectively, 47 kJ⋅mol⁻¹ and 28 kJ⋅mol⁻¹ for Tris and bis⋅Tris (33). The temperature coefficient for the Lys ɛ-amino group is −0.024 per degree Centigrade (27), based on ΔH_ionization of 43 kJ⋅mol⁻¹ (27). Deuterium NMR spectra were recorded as described (20). Helix tilt as a function of the sets of alanine CD₃ quadrupolar splittings was analyzed by established methods (20, 21, 23).

Additional data are included in the Supporting Information.