INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

The ability to predict the effects of individual or multiple amino acid sequence modifications on the structure and function of peptides is a key step toward the de novo design of functional proteins (Koeppe and Andersen, 1996). To be able to predict the folding and conformation of a sequence of amino acids, one must first determine how individual amino acids will respond to the environment in which they are placed. Important interactions may occur with the surrounding environment including neighboring amino acids, and/or with residues that are distant in sequence but come in close proximity during the folding of the peptide. A peptide with a well-defined structure and function can be used as a model to study how changes in the primary sequence may alter its properties.

Gramicidin A channels have a well-defined structure and function (for reviews, see Andersen and Koeppe, 1992; Killian, 1992; Cardew, 1999), and serve as an excellent model to examine the effects of single and multiple sequence modifications (Durkin et al., 1990; Becker et al., 1991, 1992; Koeppe et al., 1994a; Salom et al., 1995, 1998; Jude et al., 1999). Gramicidin channels assemble in membranes or membrane-like environments by the association of two monomers that emanate from opposite sides of a lipid bilayer (O'Connell et al., 1990). Each gA monomer consists of 15 alternating L- and D-amino acids with an N-terminal formyl group and a C-terminal ethanolamine (Table 1; Sarges and Witkop, 1965). Conventional gA channels are formed when opposing RH ^6.3-helical monomers are joined at their formyl-NH termini by six intermolecular hydrogen bonds (Fig. 1). The peptide backbone amide hydrogen atoms of residues 1, 3, and 5 of one monomer hydrogen bond with the carbonyl oxygen atoms of residues 5, 3, and 1, respectively, of the opposing monomer, and vice versa (Koeppe et al., 1994a). Channels formed by gA in membranes have a conformation that is not found outside of lipid bilayer membranes or membrane-like environments (Wallace et al., 1981; Wallace and Ravikumar, 1988; Langs, 1988; Bystrov and Arseniev, 1988; Abdul-Manan and Hinton, 1994; Cotten et al., 1999). The well-defined conformation of gA channels in lipid bilayers provides a way in which to examine the structural consequences of sequence modifications.

View larger version (109K): [in this window] [in a new window]   FIGURE 1   CPK model to illustrate the proximity of the side chains (highlighted in dark, and labeled) of valines 7 and 1 on one subunit, and the newly introduced 5-B on the other subunit, of a [Val5, D-Ala8]gA channel. In this energy-minimized model (not yet fit to NMR data), the side-chain 1 angles are illustrated at 64°, 182°, and 180° for valines 5-B, 1, and 7, respectively, top to bottom.

The preference for the formation of a RH conformation for gA channels has been previously investigated (Koeppe et al., 1992, 1994a; Providence et al., 1995). The contribution of the three central valine residues (Table 1) to the formation of the RH conformation was examined by executing a one-residue shift toward the N-terminus (Koeppe et al., 1994a). The resulting [Val⁵, D-Ala⁸]gA retained the RH SS ^6.3-helical conformation. This conservative modification resulted in the formation of channels with reduced ion conductance and channel duration. The reduced single-channel duration was attributed to a steric conflict between Val¹ in one monomer and Val⁵ in the other monomer created by the one-residue shift. A modification of the orientation and motion of Val¹ may in turn affect Val⁷ of the same monomer due to intramolecular interactions (Fig. 1).

To further investigate this issue, we have incorporated deuterium (²H) labels at Val 1, 5, or 7 in [Val⁵, D-Ala⁸]gA. ²H-NMR spectra of oriented and nonoriented samples in hydrated DMPC have been recorded. This technique allows the orientations and motions of the labeled valines to be estimated and compared to native gA.

Gramicidin A may also be used as a model to study the effects of acylation on integral membrane proteins because it occurs naturally in both free and acylated forms (Koeppe et al., 1985), and the acyl chain interacts sterically with Trp⁹ and D-Leu¹⁰ (Koeppe et al., 1995, 1996). To better understand acylation effects, deuterated [Val⁵, D-Ala⁸]gA analogs were palmitoylated and ²H-NMR spectroscopy was again performed. The results indicate that the covalent attachment of a fatty acid to the ethanolamine of [Val⁵, D-Ala⁸]gA does not significantly influence Val¹, Val⁵, or Val⁷.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Formyl-L-Val, non-deuterated Fmoc amino acids, and Fmoc-L-Trp resin for peptide synthesis were from Bachem (King of Prussia, PA), Peninsula Labs (Belmont, CA), and Advanced Chemtech (Louisville, KY). Deuterated L-Val was purchased from Cambridge Isotope Labs (Andover, MA). Deuterated L-Val was converted to the Fmoc derivative, and peptides specifically labeled at either valine 1, 5, or 7 in [Val⁵, D-Ala⁸]gA were prepared by previously described methods (Greathouse et al., 1999). Fifty milligrams of each specifically labeled [Val⁵, D-Ala⁸]gA was acylated and purified using previous methods (Koeppe et al., 1996).

Oriented and nonoriented samples containing 10:1 DMPC/gramicidin were prepared with ²H-depleted H₂O as described in Koeppe et al. (1996). Cuvettes with oriented samples at 40% hydration were sealed with glass lids and epoxy. Control-oriented lipid samples from which gramicidin was omitted gave spectra with only a small sharp peak at 0 Hz attributable to HOD. Centrifuged, cut glass tubes that contained nonoriented samples with excess H₂O were sealed only with stoppers and parafilm, and so were more susceptible to contamination from atmospheric HOD.

²H-NMR spectra were recorded as previously described (Koeppe et al., 1996), on a Bruker AMX 300 spectrometer modified for wideline operation with a 7.5-mm-diameter solenoid coil. Spectra were recorded using the quadrupolar echo sequence with full phase cycling (Davis et al., 1976), a 3.0-µs 90° pulse, a 30- or 100-ms interpulse time, and 0.9-1.5 million scans. The echo delay time was varied from 30 to 75 µs. A line broadening of 200 Hz for oriented samples, or 3000 Hz for nonoriented samples at 4°C, was applied to increase the signal-to-noise ratio.

To estimate the orientations and dynamics of the valine side chains both oriented and nonoriented samples were analyzed. Powder spectra for nonoriented samples at 4°C were simulated using the Fortran program MXQET (Greenfield et al., 1987), which has kindly been made available by Professor Robert L. Vold (http://nmr.physics.wm.edu). All simulated spectra were calculated on a Silicon Graphics O₂ workstation and were corrected for finite pulse length and echo delay. Single- and double-precision calculations gave identical results for the "hopping" regimes considered here. Angles of 70° (the tetrahedral supplement) between the deuterated (methyl) site and the hopping (C-C) axis, and of 120° between respective jump sites, were used. The site occupancies, exchange rates, and quadrupolar coupling constants were varied in the calculations. As has been done previously, the asymmetry parameter  was approximated as zero (Lee and Cross, 1994; Lee et al., 1995).

For the analysis of spectra from oriented samples at higher temperatures, the quadrupolar splittings (_q values) were converted to C-²H bond orientation angles using the relation (Killian et al., 1992):

in which e²qQ/h is the quadrupolar coupling constant (~168 kHz for C-²H bonds),  is the angle between the C-²H or C-CD₃ bond and the membrane normal,  is the angle between the membrane normal and magnetic field, either parallel ( = 0°) or perpendicular ( = 90°), and  is either 0° for the C and C deuterons or ~109.5° in the case of the tetrahedral geometry for the valine methyls (Killian et al., 1992).

For a starting atomic model, the "residue replace" function of InsightII (Biosym) was used to convert a gA model (Koeppe et al., 1996) to [Val⁵, D-Ala⁸]gA. With ₁ set initially to 180° for Val¹ and Val⁷and to 60°, 180°, or 300° for Val⁵the model was energy-minimized successfully as described by Killian et al. (1992). In all test cases, the initial valine orientations changed little during the minimization, indicating that each starting position was allowable and free of serious steric conflicts. The backbone of one of these structuresthat where ₁ was near 60° for Val⁵was used to generate Fig. 1, and the same backbone was used to fit the deuterium NMR data. On this backbone, the valine side chains were allowed to rotate through 360°. Analysis began with the implementation of previous software, which rotates a given side chain about the C-C bond (₁ torsion angle) in 0.5° increments (Koeppe et al., 1994b). At every interval in ₁, the calculated _q values were compared to experimental results. As determined previously for Val¹ (Killian et al., 1992; Lee and Cross, 1994; Lee et al., 1995), this method resulted in no static solutions for any of the L-Val side chains that were examined.

A second method was then used to allow each valine to hop rapidly between any two ₁ endpoints near 60°, 180°, or 300° (with tolerances of ±10°). This hopping is consistent with the data and simulations for cooled, nonoriented samples; see below. For rapid hopping, the calculated _q is the occupancy-weighted average of the individual values for the different states. Fractional occupancies of the two states were varied from 0 to 100% in 1% intervals, and the rms deviations between the theoretical and experimentally obtained _q values for each valine were minimized. This approach can be considered a subset of a generalized three-state model, in which one of the occupancies is close to zero. Rapid averaging of two states in all cases was sufficient to fit the data from oriented as well as nonoriented samples.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

²H-NMR measurements of nonoriented samples

Spectra were recorded at 4°C for individually deuterated valine 1, 5, or 7 [Val⁵, D-Ala⁸]gA samples in hydrated DMPC bilayers. By lowering the temperature below the gel-to-liquid crystalline phase transition point, global motion about the helix channel axis is eliminated (Cornell, 1987; Nicholson et al., 1987). Due to the "freezing out" of the global motions, local methyl group motions dominate the spectra in the region of 40 to +40 kHz (Lee et al., 1995). Fig. 2 shows the ²H-NMR powder pattern spectra of each individually labeled valine at low temperature, together with spectral simulations. In each case the central sharp peak is an experimental artifact that is not seen with such intensity in the simulations (see also Lee and Cross, 1994). The ²H-NMR powder pattern spectrum for d₈-Val¹ in [Val⁵, D-Ala⁸]gA is broad and flattened with distinct shoulders that appear intermediate between a single hump and a Pake pattern. The result suggests that the local motions of Val¹ have been slowed to an intermediate rate of side-chain hopping, as compared to Val¹ in native gA (Lee et al., 1995). Indeed, the spectrum for Val¹ in [Val⁵, D-Ala⁸]gA (Fig. 2 A) can be simulated nicely using a two-state model with relative occupancies of 0.7 and 0.3, an exchange rate of 5 × 10⁴ s¹, and a QCC of 49-50 kHz (Fig. 2 B).

View larger version (14K): [in this window] [in a new window]   FIGURE 2   Experimental (A) and simulated (B) 2H-NMR spectra of nonoriented samples of labeled Val1 (top), Val5 (middle), or Val7 (bottom) in [Val5, D-Ala8]gA, hydrated in DMPC at 10:1 lipid/peptide; 4°C. The echo delay time is 75 µs and is accounted for in the simulations. The center peak at 0 Hz is attributed to residual HOD and is not simulated. The simulated spectra in (B) were calculated using a two-site model with occupancies of (0.7, 0.3) and jump times of 5 × 104, 3 × 105, and 8 × 104 s1 for valines 1, 5, and 7, respectively.

The results for Val¹ in [Val⁵, D-Ala⁸]gA should be compared with the findings of Lee and Cross (1994), who have fit Val¹ in gA using a two-state (0.7, 0.3) model with QCC of 40 kHz and exchange rate of 10⁶, or a three-state model with QCC of 46 kHz and exchange rate of 6.7 × 10⁵. The QCC of ~168 kHz for a static C-²H bond (Burnett and Muller, 1971) is reduced to 168/3 = 56 kHz by methyl rotation, and is further reduced by ~8% due to backbone librational averaging (Hing et al., 1990; Prosser et al., 1991; Killian et al., 1992). The theoretical maximum QCC for a side-chain methyl in a bilayer-incorporated gramicidin channel is therefore ~51 kHz. This value agrees with data for alanine-d₄ labeled gramicidin (Lee et al., 1993) and could be further reduced for valines by additional side-chain motions (Lee and Cross, 1994).

Compared to gA (at 4°C), we find a higher QCC and a 10-20-fold slower hopping rate for the Val¹ side chain in [Val⁵, D-Ala⁸]gA. A QCC of 49 kHz was sufficient to fit not only Val¹ but also Val⁵ and Val⁷ in [Val⁵, D-Ala⁸]gA (Fig. 2), and the value of QCC was independent of the experimental echo delay (see below). The hopping rates for Val⁷ and Val⁵ are somewhat faster than Val¹ in [Val⁵, D-Ala⁸]gA, being 8 × 10⁴ and at least 3 × 10⁵, respectively, at 4°C (Fig. 2).

The hopping rates for Val¹ and Val⁷ in [Val⁵, D-Ala⁸]gA are in a range that should be sensitive to the experimental echo delay time, _echo. This is indeed the case, as is illustrated for Val¹ in Fig. 3. For echo delays of 30-75 µs, the spectral shapes varied, but each spectrum could be simulated using the same QCC of 49 kHz and exchange rate of 5 × 10⁴ as used in Fig. 2. By contrast, the more rapid hopping of Val¹ in gA (Lee and Cross, 1994) was insensitive to _echo between 30 and 75 µs (simulations not shown).

View larger version (13K): [in this window] [in a new window]   FIGURE 3   Experimental (A) and simulated (B) 2H-NMR spectra of a nonoriented sample of labeled Val1 in [Val5, D-Ala8]gA in DMPC at 4°C, as a function of the echo delay time, echo. The peak at 0 Hz is attributed to residual HOD and is not simulated. The simulated spectra in (B) were calculated using a two-site model with occupancies of (0.7, 0.3) and a jump time of 5 × 104.

In [Val⁵, D-Ala⁸]gA, the side-chain hopping rate (at 4°C) increases in the order Val¹ < Val⁷ < Val⁵. In comparison to native gA, the introduction of Val⁵ slows the hopping rate of Val⁷ by about one order of magnitude and of Val¹ to a somewhat greater extent.

²H-NMR measurements of oriented samples and comparison to native gA

²H-NMR spectra of [Val⁵, D-Ala⁸]gA deuterated at position 1, 5, or 7 in oriented DMPC bilayers were recorded at several temperatures, echo delays, and sample orientations. Fig. 4 shows results for Val¹ at different values of _echo. Fig. 5 shows results for all three valines at 50°C and  = 90°. As for gA (Hing et al., 1990; Prosser et al., 1991; Killian et al., 1992; Ketchem et al., 1993), each of the backbone C-²H quadrupolar splittings (_q) for [Val⁵, D-Ala⁸]gA are found between 100 and 105 kHz at  = 90° (Fig. 5), characteristic of the ^6.3 helix. There is no significant change in the peptide backbone conformation, thus confirming earlier conclusions that were based on circular dichroism, ¹H-NMR, and hybrid channel results (Koeppe et al., 1994a). The backbone C-²H quadrupolar splittings in [Val⁵, D-Ala⁸]gA exhibit no significant change from their counterparts in native gA (Table 2).

View larger version (9K): [in this window] [in a new window]   FIGURE 4   2H-NMR spectra as a function of the echo delay for an oriented sample of labeled Val1 in [Val5, D-Ala8]gA, hydrated in DMPC at 10:1 lipid/peptide; 50°C;  = 90°.

View larger version (15K): [in this window] [in a new window]   FIGURE 5   2H-NMR spectra of oriented samples of labeled Val1 (A), Val5 (B), or Val7 (C) in [Val5, D-Ala8]gA, hydrated in DMPC at 10:1 lipid/peptide; 50°C;  = 90°. Vertical expansions show peaks from the deuterons on the  and  carbons.

View this table: [in this window] [in a new window]   TABLE 2   Magnitudes of quadrupolar interactions for deuterated valines 1, 5, and 7 in [Val5, D-Ala8]gA, palmitoyl-[Val5, D-Ala8]gA, and corresponding sites in gA, at 50°C and  = 90°

The methyl deuterons undergo very fast motional averaging of the CD₃ top, which results in a small _q value and large intensity. The remaining spectral component with low intensity and intermediate _q then in each case can be assigned to the C-²H (Killian et al., 1992; Lee and Cross, 1994). The spectra of Val¹ and Val⁷ reveal a distinct spectral component with _q of 84 and 80 kHz, respectively, values that are significantly increased from those of Val¹ and Val⁷ C-²H in native gA (Table 2). The C-²H peak for Val⁵ in [Val⁵, D-Ala⁸]gA is less clear, but we make a tentative assignment to a _q near 50 kHz for a rather broad feature in Fig. 5 B. This is not an unreasonable assignment for a valine C-²H, since the C-²H of Val⁷ in free and acylated gA has _q of 39 and 35 kHz, respectively (Table 2; Koeppe et al., 1995).

That distinct resonances are observed when  = 90° confirms that the helix is undergoing fast axial reorientation. When samples are turned to  = 0°, the _q values should increase by a factor of 2, and this is indeed observed for the methyl deuterons (Fig. 6). At  = 0°, the spectral lines assigned to C-²H and C-²H (Fig. 6) become broad and weak, as others have observed (Hing et al., 1990; Killian et al., 1992).

View larger version (14K): [in this window] [in a new window]   FIGURE 6   2H-NMR spectra at 50°C of oriented samples of labeled Val1 (A), Val5 (B), or Val7 (C) in [Val5, D-Ala8]gA at  = 0°. The signals from the deuteron on Val7-C are labeled in the expansion of spectrum (C).

Val¹

Val⁷

Val⁵

Spectral changes after acylation

To determine whether or not covalently attached acyl chains influence residues positioned toward the middle of a [Val⁵, D-Ala⁸]gA channel, the effects of acylation on the ²H-NMR spectra were investigated. The spectrum of d₈-Val¹ in either gA (not shown) or [Val⁵, D-Ala⁸]gA (Fig. 7 A) is not greatly affected by acylation. This result is not surprising since Val¹ is far from the acylation site on the ethanolamine (Koeppe et al., 1996).

View larger version (9K): [in this window] [in a new window]   FIGURE 7   2H-NMR spectra at 50°C of oriented samples of labeled Val1 (A), Val5 (B), or Val7 (C) in acyl-[Val5, D-Ala8]gA at  = 90°. Peaks due to the deuterons on the  and  carbons are labeled in (A) and are also evident in (C).

Val⁵ and Val⁷ also are not much influenced by acylation. As before acylation, the spectrum of the acyl-[ d₈-Val⁵, D-Ala⁸]gA lacks spectral components that may be attributed to the C-²H or C-²H (Fig. 7 B). The methyl spectral component is difficult to interpret but appears to be comprised of multiple peaks and quite similar to the spectrum before acylation. The similarity of the Val⁷ spectra before (Fig. 5 C) and after (Fig. 7 C) acylation is of particular interest due to the fact that when gA is acylated there is a small change in the _q of the C-²H and a minor peak disappears; such changes do not occur for [Val⁵, D-Ala⁸]gA.

Evaluation of side-chain rotameric states

The spectra of nonoriented samples at 4°C (Figs. 2 and 3) indicated (at least) two rotameric states with approximate occupancies of (0.7, 0.3) and respective exchange rates of 5 × 10⁴, 3 × 10⁵, and 8 × 10⁴ for valines 1, 5, and 7 in [Val⁵, D-Ala⁸]gA. Consistent with these results, the spectra of oriented samples at 50°C (Figs. 4 and 5) could not be fit by single static solutions for valine 1, 5, or 7. A second method of analysis was employed to allow the side chain to "hop" between two conformations (see Methods) and fit the observed _q values for the oriented samples. One expects that the exchange rates should be faster and the site occupancies may vary somewhat from the values at 4°C.

For Val¹ in [Val⁵, D-Ala⁸]gA, the number, intensity, and arrangement of quadrupolar splittings are different from Val¹ of native gA (Table 2). The C-²H _q (at  = 90°) has increased from 68 kHz in gA to 84 kHz in [Val⁵, D-Ala⁸]gA. The Val¹ methyl deuterons in [Val⁵, D-Ala⁸]gA give a strong peak at 0.0 kHz and less intense peaks near 7.8 and 11.0 kHz that are difficult to interpret. The spectral component at zero is too intense to be a contribution only from residual HOD. Analysis of the remaining non-zero spectral components (7.8 and 11.0 kHz) alone did not lead to any satisfactory two-state or three-state solutions. Separate analyses were then performed in which zero was one _q and the second _q was either 7.8 or 11.0 kHz. Both of these procedures resulted in two-state hopping models that gave good fits to experimental data with similar fractional occupancies. These results, together with the dependence on _echo (Fig. 4), suggest that Val¹ in [Val⁵, D-Ala⁸]gA exists in several substates, each of which exhibits rapid two-state hopping about ₁. The substates seem to be induced by the presence of Val⁵. Transitions among the substates exhibit complicated dynamics, for which we have insufficient information to make an interpretation.

Using quadrupolar splittings of 84 kHz for C-²H and (0.0, 11.0) for the methyls, the Val¹ dynamics fit nicely to two-state hopping about ₁ between 185° (88% occupancy) and 65° (12% occupancy), or between 175° (85% occupancy) and +70° (15% occupancy; Table 3). For both fits, the sum of squared deviations between observed and calculated values is <1%. A graphical representation of these solutions is shown in Fig. 8. Similar results, with slightly different fractional occupancies, were obtained using splittings of (0.0, 7.8) kHz for the Val¹ methyls. These results agree well with the conclusions from the powder patterns at 4°C (above), except that the hopping rates are faster and the occupancy of the major state is increased from ~70% to ~85% at 50°C. The conclusions would not be significantly altered by considering a three-state model.

View this table: [in this window] [in a new window]   TABLE 3   Two-state hopping model, best fit 1 solutions and fractional occupancies of individually deuterated valine 1, 5, or 7 in [Val5, D-Ala8]gA

View larger version (17K): [in this window] [in a new window]   FIGURE 8   Two different fits to the 2H-NMR data based on rapid two-state hopping of the side chain of Val1 in [Val5, D-Ala8]gA: solid line, hopping between 1 of 175° and 70°; dashed line, hopping between 185° and 295°. "Sumsq" denotes the sum of squared deviations between observed and calculated q values (see Methods).

The quadrupolar splittings of Val⁵ and Val⁷ (Table 2) were similarly fit to describe the rotameric states of these valines in [Val⁵, D-Ala⁸]gA. Once again, two-state models were sufficient to reproduce the experimental data with low sums of squared deviations (below 1%). For Val⁷, the fractional occupancy when ₁ is near 180° (83%) is close to that for Val¹. Val⁵ also has a major conformation with ₁ in the vicinity of 180°, albeit with a lower occupancy (Table 3).

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

[Val⁵, D-Ala⁸]gA was synthesized to test the influence of the central valines upon the helix sense of gA channels (Koeppe et al., 1994a). Gramicidin A and [Val⁵, D-Ala⁸]gA are true sequence isomers, in which only a single valine and a single alanine have been interchanged in the sequence. The D-Val⁶-L-Val⁷-D-Val⁸ sequence in gA is shifted to L-Val⁵-D-Val⁶-L-Val⁷ in [Val⁵, D-Ala⁸]gA. This change does not alter the RH helix sense of gramicidin channels in DMPC or DPhPC membranes, or in SDS micelles, as has been shown by CD spectroscopy, hybrid channel analysis, and two-dimensional NMR (Koeppe et al., 1994a). Single channels formed by [Val⁵, D-Ala⁸]gA exhibit ~0.55 of the Na⁺ conductance of gA channels and have an average duration that is ~0.45 that of gA channels.

When a RH gA channel assembles within a phospholipid membrane (O'Connell et al., 1990), the transmembrane dimer is stabilized by six hydrogen bonds that link the backbone carbonyl and NH groups of residues 15, 33, and 51, respectively. The opposing side chains of these residues in the gA dimer are Val¹Ala⁵ and Ala³Ala³ (Fig. 9). A significant packing change in [Val⁵, D-Ala⁸]gA is that Val⁵ in one component monomer will now oppose Val¹ in the other to give the interactions Val¹Val⁵ along with Ala³Ala³ (Fig. 9). It has been suggested that steric interference between the side chains of Val¹ and Val⁵ could account for the altered single-channel properties of [Val⁵, D-Ala⁸]gA (Koeppe et al., 1994a). Furthermore, Val⁷ sits above Val¹ in the next helical turn of the same monomer in the RH channel, and so it is plausible that a Val¹Val⁵ interaction, across the dimer junction, could be propagated into a secondary effect on Val⁷. In this paper we have used specific deuterium labeling and ²H-NMR spectroscopy to examine the average orientations and dynamics of the side chains of valines 1, 5, and 7 in [Val⁵, D-Ala⁸]gA, in order to make comparisons with the properties of valines 1 and 7 in gA.

View larger version (20K): [in this window] [in a new window]   FIGURE 9   Schematic representations of the hydrogen bonding at the subunit junction in channels formed by (A) gA and (B) [Val5, D-Ala8]gA. L-amino acids are enclosed in boxes, whereas D-amino acids are not. The arrows represent the directions for continuing each helical subunit. In RH gA helices, the formyl groups participate in intramolecular hydrogen bonds (C==O- - -). Each dimer is stabilized by six intermolecular hydrogen bonds (- - -) that link the C==O and N-H groups, respectively, of residues 1/5, 3/3, and 5/1 of the component monomers. Single-channel data as well as 2H-NMR spectra indicate that some steric interference between the Val1 and Val5 side chains (* *) modulates the properties of [Val5, D-Ala8]gA channels.

Valine orientations and dynamics

Val⁵

1

Val¹ and Val⁷ in [Val⁵, D-Ala⁸]gA as opposed to gA

Effect of acylation on gA and [Val⁵, D-Ala⁸]gA

Previously, it was shown that Val⁷ in gA has a major conformation in which the side chain undergoes rapid three-state hopping (Table 3; Lee et al., 1995) and a minor conformation that is represented by a minor additional quadrupolar splitting from the methyl groups (Koeppe et al., 1995). The minor peaks are not an artifact because they are observed whether samples are oriented at  = 0° or 90°, and in spectra from independent samples in two different laboratories (Koeppe et al., 1995; Lee et al., 1995). When gA is acylated, there is no change in the major conformation of Val⁷, but the minor conformation disappears (Koeppe et al., 1995). We now report also that the ²H-NMR spectrum of Val¹ in gA after acylation is identical to that before acylation (Table 2; figure not shown).

For acyl [Val⁵, D-Ala⁸]gA, the ²H-NMR spectra (Fig. 7) suggest only small changes in the dynamics of valines 1, 5, and 7, but no large changes in the major conformation of any of the valines following acylation. The change in the pattern of minor peaks for the methyls of Val¹ after acylation (compare Figs. 5 A and 7 A) is reminiscent of Val⁷ in gA. For Val⁷ in [Val⁵, D-Ala⁸]gA, the relative intensity of the peaks with _q of 8 kHz increases after acylation (Fig. 7 C). These effects could be caused by changes in the side-chain dynamics. The major finding, however, is that acylation has very little effect on the properties of valines 1 and 7 in gA or of valine 1, 5, and 7 in [Val⁵, D-Ala⁸]gA. The results are consistent with the earlier finding that the acyl chains are largely on the same side of a gramicidin channel dimer, where they each pass near the side chains of residues 9 and 10 of their respective subunits (see Fig. 7 of Koeppe et al., 1996). In this location, the acyl chains are radially ~140-180° away from the Val⁷-Val¹-Val^5B "triad" of Fig. 1.

	CONCLUSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Deuterium NMR spectra show that for the gramicidin sequence isomer [Val⁵, D-Ala⁸]gA in phospholipid membranes, the backbone folding is unchanged. Interactions with valine 5 slow the motions of the valine 1 and 7 side chains and correlate with altered single-channel properties. Acylation of [Val⁵, D-Ala⁸]gA affects these valines very little.