Ionization Properties of Histidine Residues in the Lipid Bilayer Membrane Environment*

We address the critically important ionization properties of histidine side chains of membrane proteins, when exposed directly to lipid acyl chains within lipid bilayer membranes. The problem is important for addressing general principles that may underlie membrane protein function. To this end, we have employed a favorable host peptide framework provided by GWALP23 (acetyl-GGALW5LALALALALALALW19LAGA-amide). We inserted His residues into position 12 or 14 of GWALP23 (replacing either Leu12 or Leu14) and incorporated specific [2H]Ala labels within the helical core sequence. Solid-state 2H NMR spectra report the folding and orientation of the core sequence, revealing marked differences in the histidine-containing transmembrane helix behavior between acidic and neutral pH conditions. At neutral pH, the GWALP23-H12 and GWALP23-H14 helices exhibit well defined tilted transmembrane orientations in dioleoylphosphatidylcholine (DOPC)and dilauroylphosphatidylcholine (DLPC) bilayer membranes. Under acidic conditions, when His12 is protonated and charged, the GWALP23-H12 helix exhibits a major population that moves to the DOPC bilayer surface and a minor population that occupies multiple transmembrane states. The response to protonation of His14 is an increase in helix tilt, but GWALP23-H14 remains in a transmembrane orientation. The results suggest pKa values of less than 3 for His12 and about 3–5 for His14 in DOPC membranes. In the thinner DLPC bilayers, with increased water access, the helices are less responsive to changes in pH. The combined results enable us to compare the ionization properties of lipid-exposed His, Lys, and Arg side chains in lipid bilayer membranes.

Since the identification of histidine in the active site of chymotrypsin (1,2) and the formulation of the "charge-relay" (or catalytic triad) hypothesis (3), the importance of the imidazole side chain for acid-base catalysis has been widely recognized. Imidazole is the only group on a protein side chain with a pK a around neutrality in aqueous physiological buffer (2). It is therefore vitally important to also know the ionization properties of the His side chain at various locations within lipid bilayer membranes. Indeed, as with soluble proteins, many membrane proteins contain functionally important His residues within their transmembrane domains: for example, those of the photosynthetic reaction center (4), cytochrome c oxidase (5,6), influenza A M2 channel (7)(8)(9)(10), and other transporters and channels (11)(12)(13)(14). Given that proton conduction may require clusters of multiple residues (5,6), it is important to know the limits for the titration properties of candidate residues such as histidines and carboxyl groups, among others, in bilayer membranes.
Polar residues within the membrane core of ␣-helical proteins, although sparse, are highly conserved. In many cases, this conservation is due to the direct involvement of these residues to the function of the protein (15). Numerous membrane proteins display pH-dependent behavior, and often a histidine residue within the membrane core acts as a pH "sensor" for the protein. For example, the protonation of a single His residue in the diphtheria toxin T domain results in a major conformational change leading eventually to the translocation of the catalytic C domain across the endosomal membrane. Interestingly, replacement of this His residue with Arg results in significant protein unfolding at neutral pH (16). Histidine is also a key residue in the pH-dependent M2 channel of the influenza A virus, where a single residue, His 37 , is responsible for channel activation (10) and the proton-selective conductance of the channel (17).
Some of the pK a values for crucial residues in membrane proteins are likely to differ substantially from the canonical values that are observed in aqueous solution (5,6). A key advance toward understanding ionization behavior in lipid membranes was made by MacCallum et al. (18), who employed molecular dynamics simulations to calculate the partitioning of the neutral and charged forms of the side chains of Glu, Asp, Lys, and Arg between bulk water and a lipid bilayer of dioleoylphosphatidylcholine (DOPC). 2 Interestingly, histidine was not modeled because "His is difficult to treat accurately due to its multiple possible protonation states" (18), namely the possibility of deprotonation from either of the two NH groups of charged imidazolium to give two different neutral isomers. With the dearth of modeling results, it becomes ever more important to have experimental results about the pK a of the His imidazole side chain in bilayer membranes.
An avenue for measuring side chain pK a values in the lipid membrane environment is provided by the framework of GWALP23, acetyl-GGALW(LA) 6 LWLAGA- [ethanol]amide (19,20), which possesses two interfacial Trp residues and maintains a preferred and well defined tilted transmembrane orientation with low dynamic averaging (20 -23). Indeed, GWALP23 (or its cousin having Tyr 5 instead of Trp 5 ) has been employed to determine a pK a of 6.5 at 37°C for a lipid-exposed lysine side chain located about 8 Å from the center of DOPC bilayer membranes (24). Within the same framework, arginine is observed not to titrate below pH 9 (24). In agreement with molecular simulations (18), the Arg guanidinium group would prefer to exit the lipid bilayer rather than to deprotonate (25). It is furthermore striking that the pK a of Lys is altered more by the bilayer environment than is the pK a of Arg (24), when compared with a bulk water environment.
With the results for Lys and Arg in hand, in this study we employ the GWALP23 framework to examine the pK a values for His side chains at two locations in bilayer membranes of differing lipid thickness, namely DOPC and DLPC. In parallel with observations for lysine, the lipid bilayer environment tends to favor the neutral unprotonated state of histidine. Specifically, we find that the pK a of a lipid-exposed His imidazole side chain depends upon its bilayer location and its accessibility to the bulk aqueous solution. The results then assume a more general significance because they enable comparisons of the relative influence of the bilayer membrane environment on the pK a values for the side chains of arginine, lysine, and histidine.

Results
To probe the influence of the hydrophobic environment of the lipid bilayer on the ionization behavior of the histidine imidazole side chain, single His residues were introduced into the GWALP23 sequence (Table 1) at position 12, located directly between the two anchoring Trp residues, on the same helix face, separated by seven residues on both sides, or at position 14, located on the opposite face of the helix (Fig. 1). Using peptides with specifically labeled [ 2 H]alanine residues, we employed solid-state NMR spectroscopy as a means to characterize peptide behavior in aligned lipid bilayer membranes. The repeating Leu-Ala core sequence of GWALP23 favors peptide folding into ␣-helical secondary structure within the hydrophobic interior of the lipid bilayer. To verify that the peptides retain their ␣-helical character with the addition of His residues, circular dichroism (CD) spectra were recorded. Indeed, the CD spectra for both GWALP23-H12 and GWALP23-H14 show a minimum near 208 nm and a broad shoulder at 222 nm, indicative of ␣-helical secondary structure. The bilayers also were well aligned, as indicated by the 31 P NMR spectra of the head groups.
To assess the helix orientations, the 2 H NMR spectra of Alad 4 -labeled GWALP23-H12 and GWALP23-H14 peptides were recorded in DLPC and DOPC bilayer-incorporated samples, hydrated with 10 mM buffer at a variety of pH conditions. The 2 H NMR spectra for the aligned samples reveal significant differences in peptide behavior between the GWALP23-H12 and GWALP23-H14 isomers. Under neutral pH conditions, the spectra of the -H12 peptide in bilayers of DOPC or DLPC exhibit distinct signals for each CD 3 methyl side chain of the six core alanine residues, consistent with dynamic averaging about a predominant and well defined tilted transmembrane orientation in each of the lipid bilayer membranes (Fig. 2). Spectra of GWALP23-H14 in DLPC bilayers, hydrated with pH 6 buffer, also display sharp, well defined quadrupolar splittings. In FIGURE 1. Models to illustrate the experimental tilted orientations of charged GWALP23-H14 (left) or neutral GWALP23-H12 (right), with respect to a vertical bilayer normal (see arrow) of DLPC or DOPC lipid bilayer membranes. The peptide with the charged His 14 side chain is more tilted than is GWALP23-H12 in both membranes. In addition to rotational differences, as illustrated, each peptide helix is more tilted in the thinner DLPC bilayer (upper models) than in the thicker DOPC bilayer. The numbers indicate tryptophans 5 and 19, and histidines 12 and 14, in the peptide sequences. The six deuterated alanine methyl groups that underlie the tilt analysis are shown as space filling. Carbons are either green or black, nitrogens are blue, and hydrogens are white. The experimental helix tilt magnitudes and the pK a values of the histidine side chains are explained under "Results." DOPC bilayers, however, spectra of the -H14 peptide appear to have somewhat broader signals with lower signal to noise ratio. When the pH is lowered, the spectral quality improves for GWALP23-H14 (see below), whereas GWALP23-H12 displays multi-state behavior at low pH. As indicated (Fig. 2), above pH 4, the 2 H NMR spectra of the core alanine methyl groups of GWALP23-H12 show well defined signals in DLPC and DOPC bilayers, with quadrupolar splitting magnitudes nearly identical to those of the host peptide, GWALP23 (Table 2). Importantly, the peptide orientation does not change when Leu is replaced with a polar (yet neutral) His residue at position 12, located between Trp 5 and Trp 19 . When the pH is lowered to 2, the His residue becomes charged, leading to spectra that exhibit multiple weak resonances, similar to those of Y 5 GWALP23-K ϩ 12 and GWALP23-R ϩ 12. Furthermore, we also observe the emergence of an additional set of strong resonances, as evident from Fig. 3. These signals likely correspond to a new predominant orientation of the -H ϩ 12 peptide helix, occupied by ϳ70% of the population, in addition to multiple minor states. It is possible, therefore, to monitor the disappearance of the sharp signals observed at pH 4 and the appearance of a new signal at low pH to estimate the filling of this new state as well as the disappearance of the original state as a function of pH. The major tilted transmembrane orientation of GWALP23-H12 is 50% populated at pH 2.6, and the new orientation is 50% populated at pH 2.3 (Fig. 4). These observations are consistent with the fact that the ionization of the His residue of the -H 0 12 peptide does not follow a simple two-state equilibrium. Because the -H 0 12 population interconverts with the new major orientation and the multiple minor states at pH 2.6, the single-state, tilted transmembrane population of the -H 0 12 peptide will be greater than that of the new major orientation or any of the minor states. Also apparent from Fig. 3 is the emergence at low pH of new sets of 2 H resonances with large ⌬ q magnitudes, which likely arise from backbone C␣ deuterons. We note that our d 4 -Ala residues always contain a C␣ deuteron, and resonances from backbone deuterons have often been observed in arginine-containing peptides (25,26). We do not yet understand the molecular properties or environmental conditions that govern the emergence of these signals within the context of the quadrupolar echo pulse sequence. In DLPC bilayers, the spectra of GWALP23-H12 retain strong, well defined signals under low pH conditions.
For the case of GWALP23-H14, when the pH is lowered from 6 to 2 in DOPC bilayers, the 2 H NMR spectral quality improves. Furthermore, changes in ͉⌬ q ͉ for methyl groups of the core alanines, including Ala 7 and Ala 17 , are also observed (Fig. 5). To confirm the spectral assignments of alanine residues when Ala 15 and Ala 17 are labeled, measurements were also recorded using a peptide with a single 2 H-labeled Ala 17 . The collective changes in the ͉⌬ q ͉ magnitudes reveal a single well defined transmembrane orientation of GWALP23-H ϩ 14, with a different helix tilt than that of the neutral -H 0 14 peptide. These results are analogous to those observed for GWALP23-K14 (24). By observing the 2 H quadrupolar splittings of selected alanine CD 3 groups as a function of pH, we obtained sets of titration curves for the histidine imidazole ring of GWALP23-H14 in DOPC bilayers (Fig. 6). It is important to note that the peptide contains no other ionizable groups. The titration curves for the pH dependence of the quadrupolar splittings of deuterated Ala 7 and Ala 17 both reveal a pK a of 4.1 for the buried His residue in GWALP23-H14 at the NMR experimental temperature of 50°C (Fig. 6). Based on the reported temperature dependence of the ionization of the histidine imidazole side chain (27), we

groups in GWALP23 peptides with single residue replacements
Sample orientation is ␤ ϭ 0°. Each value (in kHz) is the average of the magnitude observed at ␤ ϭ 0°and twice the magnitude observed for a ␤ ϭ 90°sample orientation. The position of each labeled alanine is identified. The positions of the variable amino acid residues are indicated as His 12 , His 14 , Arg 12 , Arg 14 , Lys 12 , or Lys 14 ; with the side-chain charge indicated, if known. The label L12,14 refers to host peptide, GWALP23. Dashes indicate that the pH was not buffered, or that particular data points were not measured.

Lipid and peptide pH
Alanine CD 3 position b Based on repeated measurements from multiple duplicated samples, the uncertainty in measuring ͉⌬ q ͉ is within a range of Ϯ 0.6 kHz. c Also observed quadrupolar splittings corresponding to the C␣-deuterons of Ala 7 and Ala 9 with magnitudes of 92 and 102 kHz, respectively, at ␤ ϭ 0°sample orientation.
deduce pK a values of ϳ4.4 at 25°C and ϳ4.3 at 37°C for His 14 . Importantly, no titration is observed for the parent peptide with Leu 14 , as no ionizable groups are present in the peptide. In DOPC bilayer membranes at 37°C, a pK a of 4.3 for His 14 should be compared with values of 6.5 for Lys 14 and Ͼ9 for Arg 14 (24). It therefore appears that the pK a of bilayer-incorporated lysine is lowered (relative to the value in aqueous solution) to a greater extent than that of bilayer-incorporated arginine or histidine residue (see "Discussion").
Interestingly, similar titration experiments in DLPC bilayers show no changes in the quadrupolar splittings of the His 14 or His 12 peptides over a pH range of 2.5-12. Apparently, the histidines in DLPC are sufficiently exposed to the aqueous solvent that their titration does not influence the helix tilt of GWALP23-H14 or GWALP23-H12 in the thinner bilayers of DLPC. Nevertheless, the peptide behavior in DLPC is sensitive to the presence of the polar His 14 . Furthermore, some backbone C␣ deuteron resonances are observed for GWALP23-H14 in DLPC (results not shown).
On the basis of the 2 H quadrupolar splittings of the six core alanine methyl side chains for each of the transmembrane peptides, the helix tilt and rotation were analyzed using the "geometric analysis of labeled alanines" (GALA) (28, 29) approach ( Table 3). The GALA method uses an ␣-helical geometry and a principal "semi-static" order parameter S zz to give an approximate treatment for describing the relative extent of overall motion of the helix. Based upon the 2 H NMR quadrupolar splittings, this method uses the helix tilt (), the azimuthal rotation (), and S zz as variables to find the lowest RMSD value. Alternatively, we also treated the helix dynamics using a modified Gaussian analysis to treat the widths of distributions of tilt () and azimuthal rotation () (22,23). Importantly, the different methods of analysis lead to the same conclusions about the transmembrane helix tilt of GWALP23-H12 and GWALP23-H14.
From the semi-static method, we observe distinctly different orientations in DLPC bilayers for the GWALP23-H12 and GWALP23-H14 peptides (Fig. 7), reflecting different ionization states of the His side chain. The tilt and rotation values of GWALP23-H12 are nearly identical to those of the host peptide, GWALP23 (Table 3; see also Ref. 20). Remarkably, there is The difference in spectra from sharp, well resolved signals above pH 4, to spectra with multiple signals below pH 2.5, indicates that the His 12 residue is charged only under strongly acidic conditions. Red stars in B indicate signals with large ͉⌬ q ͉, which likely correspond to backbone C␣-D nuclei of labeled alanine residues. The peptide/lipid ratio is 1/60 at a temperature of 50°C. . pH dependence of the signal intensities, ؎5% estimated error, for the CD 3 resonances of Ala 7 (red circles, ͦ⌬ q ͦ ‫؍‬ 8 kHz) and Ala 9 (black circles, ͦ⌬ q ͦ ‫؍‬ 28 kHz) of GWALP23-H12 in DOPC bilayers, at ␤ ‫؍‬ 90°s ample orientation. The midpoints, shown by the dashed lines, correspond to 50% maximal peak intensity, observed at pH 2.3 for Ala 9 or pH 2.6 for Ala 7 .
Error bars indicate means Ϯ 5% estimated error. no difference in peptide orientation when neutral (yet polar) histidine is substituted for neutral leucine at position 12, directly between the two interfacial Trp residues, on the same face of the helix. Interestingly, this is not the case for His at position 14, located on the opposite face of the helix. GALA analysis of GWALP23-H14 in DLPC bilayers indicates that the orientation of the His 14 peptide differs from the host peptide, GWALP23, with Leu 14 , with a change of tilt ⌬ of about 5°and a change of rotation ⌬ of about 50°( Table 3). As can be seen in Fig. 7, the orientation for the helix with -H ϩ 14 is nearly identical to that of the -R ϩ 14 or -K ϩ 14 peptide (24).
Similar to the case of DLPC, in DOPC, the orientation of GWALP23-H 0 12 once again is nearly identical to that of GWALP23 itself, as seen in Fig. 8. Interestingly, in addition to the multiple weak resonances, an additional set of signals is observed at low pH. The new peaks at low pH are due to changes in the population of helix orientations, not to changes in the lipid bilayer. As confirmation, no changes are observed in the NMR spectra for the peptide or the lipid for GWALP23 in DOPC at pH 2.0 (data not shown). GALA analysis of GWALP23-H12 at pH 2 reveals a new orientation, in addition to the multi-state behavior, for the charged GWALP23-H ϩ 12. This new orientation likely corresponds to a surface-bound state in which the charged -H ϩ 12 peptide comes out of the membrane, as the new tilt of the core helix is 81°. Nevertheless, in DOPC, -H 0 14 serves to confer a different tilt from that of either the -R ϩ 14 peptide or the GWALP23 parent, as has been observed similarly when -K 0 14 is present (24). GALA analysis of the -H ϩ 14 peptide in DOPC bilayers at pH 2 shows an increased tilt for the charged -H ϩ 14 peptide, when compared with the neutral -H 0 14 peptide. This new orientation is similar to that observed for both the -R ϩ 14 and -K ϩ 14 peptides. By means of quadrupolar wave plots, Fig. 8 summarizes three different preferred tilted states for transmembrane helices that differ only with respect to whether the position 14 side chain is charged (cationic), neutral polar, or neutral hydrophobic.
The accumulated results (Table 3) indicate that the transmembrane orientation of the core helix of GWALP23 depends upon not only the bilayer thickness, but also the identities and charge states of the side chains of residues 12 and 14. Importantly, the helix orientations do not depend upon the method for estimating the helix dynamics, whether semi-static or modified Gaussian. Indeed, the deduced values for the helix tilt ( 0 ) and azimuthal rotation ( 0 ) agree for the two methods of analysis (Table 3). Other than the case of GWALP23-H 0 12 in DOPC, for which the polar imidazole ring is located between two flanking Trp indole rings, the values of from the modified Gaussian analysis (Table 3) are uniformly low, indicating only low to moderate dynamic averaging for each of the target transmembrane helices considered here. The somewhat higher value of for GWALP23-H 0 12 in thicker DOPC bilayers is not surprising, given that the protonation to yield GWALP23-H ϩ 12 leads to a major population exiting to the membrane surface along with multi-state behavior for the minor population of helices that remain in DOPC at low pH (see also Fig. 3).
The accumulated results furthermore illustrate several core principles. A neutral residue at position 12, between the flanking tryptophan indole rings 5 and 19 of GWALP23, whether  leucine or neutral histidine or lysine, confers the same global orientation for the tilted transmembrane core helix. By contrast, a charged residue at position 12, whether arginine or charged histidine or lysine, confers multi-state behavior to the helix, in which a significant fraction of the population exits the membrane (25). Notably, residue 14 is located on the opposite face of the core helix from the flanking tryptophans 5 and 19. When the side chain of position 14 is altered, single-state behavior is always observed, but the favored state for the core helix takes on one of three choices (Fig. 8), depending upon whether the side chain of residue 14 is charged or neutral/polar or neutral/nonpolar. For the position 14 side chains as well as the position 12 side chains, the results are consistent among lysine, histidine, and arginine.

Semi-static GALA and modified Gaussian analysis of transmembrane orientations of related GWALP23 peptides
The parent GWALP23 sequence is acetyl-GGALWLALALAL 12 AL 14 ALALWLAGA-amide. In the noted examples, either residue Leu 12 or Leu 14 (but not both) was changed to His or Lys or Arg, as indicated, and the other residue remained a leucine. a The modified Gaussian analysis followed Sparks et al. (23), with assigned a fixed finite value of 10°. and are related to the widths of distributions of helix orientations, effectively indicating the uncertainties in 0 and 0 . b In DLPC, the results with His 12 and His 14 do not depend on pH. c Because 0 and 0 are derived values from the GALA fits, we report Ϯ numbers that correspond to those ranges where RMSD Ͻ 1.5 kHz. d The results with -R14 ϩ do not depend on pH (24,25). For the L12,14 peptide, the pH was not buffered. e Analysis of -H ϩ 12 orientation is based on quadrupolar splittings of labeled Ala 7 and Ala 9 C␣Ϫdeuterons along with labeled CD 3 groups of Ala 7 , Ala 9 , Ala 11 , and Ala 13 .

Lipid and
About 70% of the -H ϩ 12 population exists in this surface-bound state and the other portion remains transmembrane with multi-state behavior.

Discussion
We have examined the ionization properties of a lipid-exposed histidine residue incorporated into a transmembrane helix at two different positions, as well as the influence of histidine residue titration on the peptide orientations in the hydrophobic environment of the lipid bilayer. The results establish important limits for understanding membrane protein functions that involve proton-transfer reactions or other protonmediated events. We will discuss the dependence of transmembrane helix orientations on His residue location and protonation state, the titration behavior of His 12 and His 14 in bilayer-incorporated GWALP23, the transmembrane to surface transition when His 12 is protonated, and comparisons with the titration behavior of other histidines in soluble proteins or membrane environments. Finally, we will address the current status for understanding and comparing experimental or predicted differences in titration behavior when His or Lys as opposed to Arg is incorporated into a lipid bilayer membrane.
GWALP23-H 0 12, with a neutral imidazole ring, is shown to have a transmembrane orientation identical to that of unmodified GWALP23 in both DLPC and DOPC bilayers. Notably, the helix orientation of GWALP23 is determined by the tryptophan residues, along with unwinding of the helix terminals (30), and the presence of neutral, polar His instead of Leu at position 12 does not alter this orientation. By contrast, GWALP23-H 0 14 adopts a tilted transmembrane orientation distinct from that of the host peptide, similar to that observed (24) for Y 5 GWALP23-K 0 14. Interestingly, the presence of neutral, polar His or Lys at position 14 results in a 40°-50°change in peptide rotation about the helix axis, suggesting that a polar residue instead of Leu at position 14 modulates the orientation of the peptide, along with the interfacial Trp residues and possibly the terminal fraying. Although there is little change in the overall helicity when Leu 14 is substituted by -H 0 14 or -K 0 14, as confirmed by circular dichroism spectra (see also Ref. 23) and by the GALA curves (Figs. 7 and 8), admittedly there could be minor changes in the local helicity or terminal fraying that accompany the non-conservative replacement of Leu with His or Lys.
In DOPC bilayers, we observe major changes in peptide behavior as a function of pH. In the case of His 12 , under strongly acidic pH conditions, the His residue becomes charged and a significant population of GWALP23-H ϩ 12 abandons the transmembrane orientation and exits to the membrane surface. Conversely, protonation of His at position 14 results in a small increase in the peptide tilt and additional azimuthal rotation about the helix axis, with the peptide remaining in a transmembrane orientation. This new orientation most likely provides greater access to the interfacial region of the membrane for the charged His imidazolium ring, which allows the peptide helix to remain in a stable transmembrane orientation. The situation is similar for GWALP23-R ϩ 14 and GWALP23-K ϩ 14 (24,25). Interestingly, the deuterium NMR spectra for GWALP23-H 0 14 are slightly broadened when compared with those of GWALP23-H ϩ 14. The spectral differences suggest differences in peptide dynamics between the neutral and charged forms of the -H14 peptide. That is, the signal broadening observed for the GWALP23-H 0 14 suggests that the neutral form may slow the motion of the peptide helix.
Our results are consistent with a low pK a for His 12 . Because of the multiple states associated with the -H ϩ 12 peptide, in equilibrium with a single transmembrane state for GWALP23-H 0 12, it is not possible to directly assign a pK a value for bilayerincorporated His 12 . Despite these limitations, one observes that the major tilted transmembrane orientation of GWALP23-H 0 12 is 50% populated at pH 2.6, and the perpendicular interfacial orientation of GWALP23-H ϩ 12 is 50% populated at pH 2.3 (Fig. 4). Although not a two-state equilibrium, we suggest that these midpoints constitute upper and lower limits of the pK a for His 12 in DOPC bilayers.
What is the surface orientation of GWALP23-H ϩ 12 at low pH? We have addressed this question using a semi-static GALA analysis with principal S zz order parameter, as well as a modified Gaussian analysis, employing C␣-deuteron quadrupolar splittings from Ala 7 and Ala 9 in addition to the available alanine methyl 2 H quadrupolar splittings. Both methods yield the surface orientation depicted in Fig. 9, with tilt ϭ 81 0 and azimuthal rotation ϭ 296 0 . (The extent of dynamic averaging is low, as is 5°when is fixed at 5°in the modified Gaussian; or S zz is 0.85 in the semi-static treatment.) The azimuthal rotation, which fits the surface orientation of GWALP23-H ϩ 12 (Fig. 9), places the His and Trp side chains, on the same face of the helix, approximately parallel to the membrane surface, such that these side chains point neither into nor out of the membrane. The rotational preference is perhaps a compromise among the aromatic Trp and charged His side chains. Indeed, the rotational preference contrasts with that observed for several amphipathic surface-active antimicrobial peptides. For example, the amphipathic PGLa peptide binds to a bilayer surface so as to orient four lysine residues away from the bilayer, toward the aqueous solution (31). To explore the issue further, we searched for other possible minima in the rotation space for the surface-bound GWALP23-H ϩ 12. In this process, we found no solutions other than the helix orientation shown in Fig. 9A. For comparison, the dependence of RMSD of the fit to the 2 H NMR data upon azimuthal rotation of the GWALP23-H ϩ 12 helix on the surface of a DOPC bilayer is illustrated in Fig. 9B.
In thinner DLPC bilayers, we observe different behavior in response to ionization of His 12 in GWALP23. Interestingly, even at pH 2.1, the majority of the His 12 population remains in the same tilted transmembrane orientation as observed at pH 8. We do, however, observe some signal broadening and the presence of weak multiple signals at pH 2.1. This could indicate the presence of some multi-state behavior for the -H ϩ 12 peptide; however, no signals corresponding to a perpendicular interfacial orientation are observed. It is possible that the pK a of His 12 in DLPC bilayers is much lower than 2.1; however, another likely scenario is that water is able more easily to penetrate the thinner DLPC membrane and thereby satisfy the charge of His 12 , so that no major change in peptide helix orientation is observed when the imidazole ring titrates.
In soluble proteins, the depth of burial of a His imidazole ring and the polarity of the microenvironment influence the His side chain pK a (32). That is, buried histidine residues in soluble proteins are found in mixed polar/apolar environments, and there-fore, the pK a values can vary tremendously depending on a range of factors. For example, His 149 is completely buried within the hydrophobic core of xylanase and has a pK a Ͻ 2.3 (33), whereas His 72 of bovine protein tyrosine phosphatase has a pK a of 9.2, due to electrostatic interactions between the buried imidazole ring and nearby negatively charged side chains (34). In staphylococcal nuclease, the pK a of His 121 has been extensively studied. The observed pK a for this partially buried residue is 5.3; however, when nearby glutamic acid residue Glu 75 is mutated to Ala, resulting in a more hydrophobic microenvironment for the His imidazole ring, the pK a of His 121 drops to 4.0 (35).
A large range of pK a values is also observed for histidine residues of transmembrane domains. His 37 is responsible for pH activation and selectivity of the tetrameric M2 ion channel of influenza A protein. A titration curve generated from 15 N spectra of His 37 yielded doubly degenerate pK a values of 7.6 and 4.5 (36). Single His residues engineered along the helical lining of the nicotinic acetylcholine receptor transmembrane pore have pK a values higher than 6.4, illustrating their accessibility to the aqueous phase (37). In a similar fashion, His residues in model helical peptides LAH2 and LAH4 incorporated in dodecylphosphocholine micelles have pK a values in a range from 4.9 to 6.6 (38), also suggesting aqueous access for the His side chains in the small micelles. These results are seemingly analogous to our observations for His 12 of GWALP23 in the thin bilayers of DLPC. The extent of aqueous access, therefore, is likely to dictate the histidine pK a value in the membrane environment, which, in turn, will influence the pH dependence of membrane protein folding and function.
As an example of the folding/function paradigm, the human chloride intracellular channel 1 converts between cytosolic and membrane forms with changes in pH, with His 74 and His 185 providing major contributions to the pH-dependent conformational stability (39). Our observation of the helix repositioning as a function of the histidine protonation state (Fig. 9) is therefore emblematic of the potential for large-scale conformational rearrangements of membrane proteins in response to pH or FIGURE 9. GWALP23-H ؉ 12 exits the bilayer. A, model to illustrate the transition of GWALP23-H ϩ 12 from its tilted transmembrane orientation (above pH 3) to a primary orientation at the surface of DOPC bilayer membranes (as the pH is lowered below 3). Carbons are green and nitrogens are blue. The midpoint for this transition is between pH 2.3 and 2.6 (see "Results"). The observed azimuthal rotation from 2 H NMR experiments is illustrated for each of the helix orientations. B, dependence of RMSD of the fit to the 2 H NMR data upon azimuthal rotation of the GWALP23-H ϩ 12 helix on the surface of a DOPC bilayer at low pH. The best fit occurs when the azimuthal rotation is 296°, and the corresponding surface-oriented helix in A has a tilt of 81°.
The pH-dependent behavior that we observe for His 12 and His 14 in DOPC bilayers can be compared with results for buried Lys and Arg in analogous positions (25). For arginine, we are unable to observe loss of the positive charge due to deprotonation, up to at least pH 9 (24), and perhaps higher (41). Instead, the charged guanidinium group would rather exit the DOPC bilayer (25) than release a proton. Numerous molecular dynamics simulations are in agreement with this observation for arginine (18,(42)(43)(44)(45). Arginine also remains cationic when buried in the hydrophobic interior of a folded protein (46). We note furthermore that the experimental value for the aqueous pK a of the Arg guanidinium side chain recently has been revised upward to 13.8 (47).
As noted, comparisons among histidine, lysine, and arginine in DOPC bilayer membranes are of interest. The lysine residue Lys 14 has a pK a of 6.5 in DOPC (24), about 4 pH units below that of Lys in aqueous buffer. Residue Lys 12 in the GWALP23 framework also experiences a pK a shift of at least 4 units, a lower limit for the extent of the shift (24). Although buried slightly more deeply than Lys 14 , and at least partially occluded from solvent access by Trp indole rings (26), one is unable to distinguish whether the pK a of Lys 12 is actually lower than that of Lys 14 (24). Nevertheless, the present results resolutely show that the pK a values for His 12 and His 14 on the GWALP23 framework are different from one another in DOPC bilayers. We observe at 50°C a pK a value of 4.1 for His 14 (Fig. 6; corresponding to 4.4 at 25°C) and an even lower pK a between 2.3 and 2.6 for His 12 (Fig.  4). The pK a shift is therefore position-dependent for the histidine side chain on the GWALP23 helix in DOPC. Notably, moreover, the extent of the pK a shift, relative to the aqueous value, follows a trend from arginine to histidine to lysine. For transmembrane GWALP23 in DOPC, a pK a shift is undetected for Arg 14 , and is about 2 pH units for His 14 and 4 pH units for Lys 14 .
We likewise compare our results with recent simulations utilizing a constant pH molecular dynamics technique (48). The agreement between experiments and the computational predictions is quite good, as Panahi and Brooks (48) have calculated pK a values of 4.5 Ϯ 0.3 for His 14 and 4.0 Ϯ 0.1 for His 12 on the GWALP23 helix in DOPC bilayers. As a caveat, one notes that GWALP23-H ϩ 12 moves to primarily a surface orientation at low pH (Fig. 9), although the time scale for the exit of the helix from the bilayer should be slow relative to the time scale of the calculations (48).

Concluding Remarks
Histidine residues are significant for not only soluble protein function but also membrane protein function. Our results fill gaps in the knowledge of His imidazole ionization properties in lipid bilayer membranes. When attached to the transmembrane GWALP23 helix and buried within a bilayer membrane of DOPC, the His imidazole side chain displays position-dependent titration behavior. When located in the central position of the peptide sequence, His 12 displays a pK a of about 2.3-2.6, perhaps influenced also by a tryptophan "cage" (26). The somewhat off-center His 14 displays a pK a of about 4.1, in excellent agreement with computational predictions (48). The transmembrane helix responds to the histidine titration by changing its tilt when His 14 is protonated or by exiting the bilayer when His 12 is protonated. By contrast, in the thinner DLPC bilayer membranes, the more highly tilted transmembrane helices show little response to changes in pH.
The present results allow us to compare the ionization behavior of Arg, Lys, and His side chains in DOPC bilayer membranes. The summary findings are that the Arg guanidinium seeks hydration or exits the bilayer but does not release its proton, the Lys ammonium pK a is at least 4 pH units lower than its aqueous value, and the His imidazolium pK a is about 2-4 pH units lower, depending on its location in the bilayer. These results are of interest and importance for experiments as well as molecular dynamics simulations that address the properties of membrane proteins.

Experimental Procedures
Peptides were synthesized using solid-phase methods on a 0.1-mmol scale using an Applied Biosystems 433A synthesizer from Life Technologies. Protected amino acids were purchased from Novabiochem (San Diego, CA). Histidine and tryptophan side chains were protected with trityl and t-butoxycarbonyl protecting groups, respectively. Peptide cleavage from Rink amide resin was accomplished by treatment at 22°C with a solution of trifluoroacetic acid/triisopropyl silane/water/phenol (85/5/5/5, v/v/v/w) over a 2-h period. The cleavage mixture was then filtered to separate the free peptide from the resin support. The crude peptide was precipitated using a 50/50 mixture of methyl-t-butyl ether and hexane and lyophilized from a 50/50 mixture of acetonitrile and water. Peptides were purified by reversed-phase HPLC on an octyl-silica column (Zorbax Rx-C8, 9.4 ϫ 250 mm, 5-m particle size; Agilent Technologies, Santa Clara, CA) using a gradient of 92-96% methanol, with 0.1% trifluoroacetic acid (v/v), over 24 min. Analytical HPLC and MALDI-TOF analyses were used to verify the peptide purity and identity.
Mechanically aligned samples (1:60, peptide:lipid) for solidstate 2 H NMR experiments were prepared using DOPC or DLPC (Avanti Polar Lipids, Alabaster, AL) and hydrated (45%, w/w) with 10 mM glycine, acetate, or citrate buffer in deuteriumdepleted water at specified pH values between pH 2 and pH 8. Bilayer alignment of each sample was confirmed by 31 P NMR spectroscopy using a Bruker Avance 300 spectrometer (Billerica, MA). Deuterium NMR spectra were recorded using a Bruker Avance 300 spectrometer at 50°C, at ␤ ϭ 90°or ␤ ϭ 0°H macroscopic sample orientations, using a quadrupolar echo pulse sequence (49) with full phase cycling, a 90-ms recycle delay, 3.2-s pulse length, and 115-s echo delay. Between 0.7 and 1 million free induction decays were collected for each 2 H experiment. Fourier transformation was accomplished using an exponential weighting function with 100-Hz line broadening.
The sealed, hydrated samples with ester lipids at pH 2.0 were found to be stable for several days. Samples above pH 4.0 are stable for several months. The intrinsic pK a of the DOPC lipid phosphate group has been measured to be about 0.8 -1.0 (50). Furthermore, our control experiments with GWALP23 in ester lipids showed no changes in the 2 H or 31 P NMR spectra at pH 2.0 when compared with pH 6.0, indicating that the changes we observe at pH 2 for the -H12 and -H14 peptides are solely a result of the changes in ionization state of the His residue.
Helix orientations were analyzed by means of the semi-static GALA method, using the average tilt of the helix axis, the azimuthal rotation , and the principal order parameter S zz as variables (28,29). Additionally, we employed a modified Gaussian approach based on , , a distribution width , and a fixed , as described previously (23).
Author Contributions-R. E. K. conceived and coordinated the study. A. N. M. and D. V. G. designed and performed the experiments. The manuscript was written through contributions of all authors. All authors have reviewed results and given approval to the final version of the manuscript.