The hydration of amides in helices; a comprehensive picture from molecular dynamics, IR, and NMR

Abstract

We examined the hydration of amides of α₃D, a simple, designed three-helix bundle protein. Molecular dynamics calculations show that the amide carbonyls on the surface of the protein tilt away from the helical axis to interact with solvent water, resulting in a lengthening of the hydrogen bonds on this face of the helix. Water molecules are bonded to these carbonyl groups with partial occupancy (∼50%–70%), and their interaction geometries show a large variation in their hydrogen bond lengths and angles on the nsec time scale. This heterogeneity is reflected in the carbonyl stretching vibration (amide I′ band) of a group of surface Ala residues. The surface-exposed amides are broad, and shift to lower frequency (reflecting strengthening of the hydrogen bonds) as the temperature is decreased. By contrast, the amide I′ bands of the buried ¹³C-labeled Leu residues are significantly sharper and their frequencies are consistent with the formation of strong hydrogen bonds, independent of temperature. The rates of hydrogen-deuterium exchange and the proton NMR chemical shifts of the helical amide groups also depend on environment. The partial occupancy of the hydration sites on the surface of helices suggests that the interaction is relatively weak, on the order of thermal energy at room temperature. One unexpected feature that emerged from the dynamics calculations was that a Thr side chain subtly disrupted the helical geometry 4–7 residues N-terminal in sequence, which was reflected in the proton chemical shifts and the rates of amide proton exchange for several amides that engage in a mixed 3₁₀/α/π-helical conformation.

The interaction of proteins with water plays a large and essential role in almost all aspects of protein folding and function (Mattos 2002). For example, when helices are formed during the folding of a protein, their amides are partially to fully dehydrated; the degree of residual solvent depends on their location in the protein (Baker and Hubbard 1984; Thanki et al. 1988). Carbonyl oxygen groups on the surface of proteins often retain the ability to form a hydrogen bond to water, while simultaneously forming the normal helical hydrogen bond to an amide proton at position i + 4 in the sequence (Baker and Hubbard 1984; Thanki et al. 1988). Such interactions have been proposed to be important to the stability of the α-helix, as assessed from molecular dynamics simulations (Daggett and Levitt 1992), and in conjunction with measurements of the stability of variants of Ala-rich peptides that form monomeric helices (Luo and Baldwin 1999; Avbelj et al. 2000). However, there are few methods available to directly assess the extent and strength of such hydrogen bonds to water molecules within the helices of folded proteins. X-ray and neutron diffraction have provided the primary methods to examine the presence of water bound to specific sites within proteins (Savage and Wlodawer 1986; Schoenborn et al. 1995). However, the density associated with water molecules is often not well defined, because of disorder or partial occupancy. Also, the crystal structures of proteins are generally solved near liquid nitrogen temperature, which has a dramatic effect on the hydration of the amides as demonstrated in this manuscript. NMR has also been used to explore waters of hydration (Otting et al. 1991; Belton 1994; Otting 1997; Halle and Denisov 2001; Mattos 2002), although only a subset of the crystallographically defined water molecules are typically observed. Here, we use temperature-dependent IR and NMR spectroscopy together with molecular dynamics simulations to examine the extent of solvation of specifically labeled carbonyl groups within a small helical protein. The results advance our understanding of this problem while simultaneously providing simple methods to assess the degree of interaction between water and helical carbonyl groups.

In α-helices of proteins, the amide proton typically forms only a single hydrogen bond (Baker and Hubbard 1984; Thanki et al. 1988). The larger size of the carbonyl oxygen and the presence of lone-pair electrons allow the formation of an additional hydrogen bond to water or other polar side chains (often a Ser or Thr located at positions i + 4 or i + 3 in the helix) (Baker and Hubbard 1984; Thanki et al. 1988). To accommodate this additional hydrogen bond, the amide carbonyl groups on the solvent-accessible side of a helix tend to tilt away from the helical axis by about 20° (Blundell et al. 1983; Baker and Hubbard 1984; Chakrabarti et al. 1986; Barlow and Thornton 1988; Thanki et al. 1988); this tilt increases the length of the intrahelical hydrogen bond and decreases its linearity (Fig. 1 ▶). These geometric differences between solvent-accessible and solvent-inaccessible amide carbonyl groups (Chakrabarti et al. 1986; Olivella et al. 2002) have major affects on their spectroscopic properties. For example, there is a general trend of proton chemical shifts of the amide protons on the solvent-accessible side of the helices tend to be shifted upfield (Goodman and Kim 1991; Wishart et al. 1991; Milne et al. 1998; Cordier and Grzesiek 2002), as compared to the more strongly hydrogen-bonded amide protons on the solvent-inaccessible sides of helices. The location of the amides within helices also influences their rates of exchange with protons from solvent (Goodman and Kim 1991; Milne et al. 1998). Further, the solvent-inaccessible amides have lower temperature coefficients (Δδ/ΔT) (Anderson et al. 1997; Baxter and Williamson 1997), although the correlation is not as great as that with chemical shift and hydrogen-deuterium exchange. Therefore, in this manuscript, we focused on the former two measurements.

Figure 1.

Exposed (left) and buried (right) amide in a helix of α₃D. The schematics provide an idealized description of the geometries of the two groups, as well as their expected spectroscopic properties.

The degree of exposure of amides also has a major effect on the amide I′ band, a mode that primarily reflects the carbonyl stretch (the prime indicates that the amides are deuterated, which is required to minimize the absorbance of the solvent). Recently, we have used isotopic labeling to determine how solvent-accessibility and temperature affect the amide I′ mode, ν_c=o, in the α-helical coiled coil from GCN4 (GCN4-P1′) (Manas et al. 2000). The ¹³C-labeled amide I′ band occurred at lower energy for solvent exposed (1585 cm⁻¹) versus buried positions (1606 cm⁻¹). These data agree with previous studies on unlabeled monomeric helices and coiled coils, which suggested that the solvent accessibility of a helical amide might decrease the frequency of ν_c=o. As the strength of the hydrogen bonds to a carbonyl group increases, ν_c=o generally decreases, in part reflecting a weakening of the C=O bond. Thus, the decrease in the energy of the C=O stretch for exposed amides in helices appears to be a result of their receiving an additional hydrogen bond from solvent, which more than compensates for the decrease in the hydrogen-bond strength associated with the intramolecular helical hydrogen bonds. The temperature dependence of the amide I′ band also provides important environmental information; in our previous study, ν_c=o of the solvent-exposed amides decreased markedly with lowering temperature, presumably reflecting an increase in the strength of the intermolecular hydrogen bond to solvent. By contrast, the amide I′ band of the buried amides were nearly independent of temperature. Finally, the ν_c=o for the solvent-exposed amide showed a dramatic sharpening at lower T, possibly reflecting an increase in the homogeneity and the occupancy of the hydrogen bonds to solvent.

In this manuscript, we confirm and extend these findings by examining a designed globular protein, α₃D, which has numerous advantages compared to other natural or designed systems. The globular structure of α₃D is more typical of small globular proteins than the elongated coiled coils that were studied previously. Further, the protein is stabilized by a relatively simple hydrophobic core, with no contributions from long-range hydrogen bonds that might otherwise complicate the analysis. The solution structure of α₃D has been determined (Walsh et al. 1999), and the dynamics of its side-chain and main-chain atoms have been extensively characterized by NMR (Walsh et al. 2001a), which indicates that the S² order parameters of the main-chain amides are similar to those of natural globular proteins. Finally, there is extensive sequence homology between the three helices, which allows one to average the values of various spectroscopic parameters. Thus, it should be possible to minimize the contributions from variations in local conformation and chemical composition, allowing one to discern more general trends that are associated with solvent exposure of the helices.

Results and discussion

Design and preparation of ¹³C-labeled samples for IR spectroscopy

For IR spectroscopy, it is desirable to label a subset of the amide groups, allowing one to differentiate solvent-exposed from solvent-inaccessible carbonyls. Figure 2 ▶ illustrates the structure of α₃D, with the exposure of the carbonyl groups of the Leu and Ala side chains coded according to the extent to which they form hydrogen bonds to solvent in the molecular dynamics calculations described below. All but two of the seven leucyl residues in α₃D are essentially fully buried from solvent. Conversely, 13 of the 15 Ala residues are exposed to solvent. Thus, we prepared samples in which the amides of either the Ala or Leu were biosynthetically labeled at the carbonyl group with ¹³C. To minimize the contribution from Ala60, which is buried in the core of the protein, we used a previously well-characterized mutant of α₃D, Ala60Ile (Walsh et al. 2001b). For simplicity, we will refer to this derivative as α₃D′ in the sections below.

Figure 2.

Structure of α₃D, with the percent exposure of the carbonyl oxygen atoms coded according to color. On the left are the Ala amides, while the Leu amides are at right.

Molecular dynamics

The motion of α₃D was simulated in a periodic box of water (∼3000 water molecules) using periodic boundary conditions to minimize edge effects. The simulation was performed at 300 K for 10 nsec. A number of parameters, including various torsions, angles, and bond lengths, were monitored to determine which were most appropriate for comparison with the spectroscopic observations. Ultimately, three parameters were found to correlate well with the experimental observations: (1) d_NH-O, the α-helical hydrogen bond length (the distance between a carbonyl at position i and an amide hydrogen at i + 4); (2) ˚_NH-O, the α-helical hydrogen bond angle (the angle formed by the carbonyl oxygen at i, and the amide proton and nitrogen at i + 4); and (3) the percentage of structures in which a given carbonyl is hydrogen-bonded to one or more solvent molecules, n_HOH-O, or the water occupancy about a given carbonyl group.

These three parameters were measured for each helix in α₃D, and are illustrated in Figure 3 ▶. The parameters generally show a periodic distribution, with a repeat matching that of the α-helix. However, helix 1 deviates from this pattern, particularly with respect to the hydrogen bond length and distance. A less substantial deviation from helical periodicity is also apparent at approximately the same position in helix 2. Examination of the structures indicated that the perturbation of the helical periodicity in helix 1 was a result of a persistent (98% occupancy) and strong (average distance 1.63Å) hydrogen bond from the Thr16 side chain to a carbonyl group four residue N-terminal in sequence (Fig. 4 ▶). This hydrogen bond caused a disruption of the helical geometry of residues 9–12, which show mixed 3₁₀/α-helical character. Two waters form hydrogen bonds with the carbonyl groups of residues 8 and 9, which stabilize the 3₁₀–α-helical junction. Significantly, these sites on the protein have the highest water occupancies (Fig. 3 ▶). The structural reorganization N-terminal to Thr16 can also be seen in the φ and ξ angles of the amides (data not presented). Thornton and coworkers (Blundell et al. 1983) have noted an anticorrelation between the φ and ξ torsional angles at individual positions in α-helices. Indeed, these correlated changes from the ideal helical values of φ and ξ contribute to bending of helices on the surface of proteins. A similar negative correlation (correlation coefficient −0.75) was observed for the helical positions of α₃D, except for residues 9–12, just N-terminal from Thr16 (data not shown).

Figure 3.

The parameters n_HOH-O, d_NH-O, and ˚_NH-O as a function of the position in the sequence of α₃D. The values are averages over 2–10 nsec involving 40,000 structures.

Figure 4.

(A) An intramolecular hydrogen bond involving Thr16 disrupts the local helical geometry in helix 1 of α₃D. (B) A ribbon representation of α₃D with the Thr16 side chain in stick representation; the core residues and Ala’s are shown in green and magenta, respectively. (C) Stereo-view of the intramolecular hydrogen bond involving Thr16 in helix 1; the side chain of the core residues and Ala’s are shown in green and magenta, respectively. (D) Stereo-view of α₃D with waters hydrogen-bonded to the backbone shown; side chains are omitted for clarity. The Cα of the core residues and Ala’s are shown in green and magenta, respectively.

Ser40 occurs at a position analogous to Thr16, and similarly engages in a hydrogen bond to a carbonyl at position i-4 in sequence. However, the mean occupancy of this intramolecular hydrogen bond was 34% versus 98% for Thr16. Blaber and coworkers (Blaber et al. 1994) have shown that the β-methyl group of Thr specifically favors this side-chain to main-chain hydrogen bond by destabilizing alternate conformations available to Ser, thus, the mean structural perturbation associated with the side-chain to backbone hydrogen bond at Ser40 appeared to be less severe.

The static and dynamic variations in d_NH-O, ˚_NH-O, and n_HOH-O at a given position of α₃D will affect the IR and NMR spectra differently. The hydrogen bonds to water generally show partial rather than full occupancy (Fig. 3 ▶), and the geometry of the water-carbonyl hydrogen bonds at a given site vary significantly with respect to time (not shown). Also, d_NH-O and ˚_NH-O vary significantly for solvent-exposed residues, while they are essentially constant for the most buried residues. Of particular relevance to the NMR measurements, all three parameters equilibrate rapidly on the 1-nsec time scale. Thus, we would expect the amides to be rapidly exchanging with respect to the NMR measurements discussed below; thus, one should observe single sharp peaks whose positions reflect the population of the various configurations. In contrast, the IR measurements occur on the time scale of a molecular vibration, such that we expect to see a broad spectral distribution that reflects the full ensemble of molecular configurations about each amide bond. More specifically, the dynamics calculations predict that, at room temperature, the amide I′ band should be significantly broader for the exposed carbonyl groups as compared with the buried carbonyls.

Hydrogen bonds between the terminal guanidinium and ammonium groups of Arg and Lys, respectively, have been proposed to stabilize helices by interacting with the carbonyl groups four residues N-terminal to these residues in the sequence of α-helices (Vila et al. 2000, 2001; Garcia and Sanbonmatsu 2002). We therefore measured the frequency with which these residues of α₃D interacted with the appropriate carbonyls in the helix (data not presented). No significant hydrogen bonding was observed, indicating that if this interaction actually stabilizes the structures of helices, it may be limited to the monomeric, Ala-rich helices in which it has been predicted to occur. This conclusion also agrees with surveys of protein structure, which failed to identify this interaction as a feature that occurred frequently in α-helices (Baker and Hubbard 1984).

IR spectra of α₃D′

The IR spectra of unlabeled α₃D’ at temperatures ranging from 300 to 13 K are shown in Figure 5A ▶. The bands in the range from 1610 to 1660 cm⁻¹ arise primarily from the amide I′ band, while the band at ∼1670 cm⁻¹ is from trifluoracetate, which is present as a counterion to the Lys and Arg side chains following purification by reverse-phase HPLC. At room temperature, the amide I′ region is broader than observed for N-methylacetamide (Manas et al. 2000), suggesting the presence of more than one overlapping bands. Indeed, as the temperature is lowered, the broad peak in the α₃D′ spectrum splits into two distinct components, with maxima at ∼1630 and 1645 cm⁻¹, respectively. By analogy to earlier studies (Manas et al. 2000), we assign these peaks to exposed and buried helical amides, respectively (Fig. 3 ▶). The position of the lower-energy band of α₃D′, which is associated with solvent-accessible amides, depends markedly on temperature, while the band associated with solvent-inaccessible amides is largely independent of solvent.

Figure 5.

(A) Temperature dependence of the IR spectra of unlabeled; (B) ¹³C-Ala-labeled; and (C) ¹³C-Leu-labeled α₃D′. Spectra were collected from 290 K and 13 K (direction indicated by arrow) in D₂O-glycerol-d₈ (50:50 v/v) at pD 7.0.

Figure 5B ▶ illustrates that ¹³C-labeling of the Ala residues on the solvent-exposed face of the helices results in a new feature in the amide I′ region of the spectrum. As expected from the molecular dynamics calculations, this peak is quite broad at room temperature, but resolves at low temperatures into a well-resolved peak at ∼1590 cm⁻¹. A second feature evident in the spectra is a diminution in the intensity of the peak associated with the unlabeled exposed amides (at 1630 cm⁻¹ in the low-temperature spectra). Thus, the amide I′ frequencies of the ¹³C-labeled and unlabeled Ala amides are 1590 and 1630 cm⁻¹, respectively. The isotope shift is ∼40 cm⁻¹, which is close to the value of 37 cm⁻¹, expected for an unperturbed oscillator (Decatur and Antonic 1999).

Different behavior is observed when the Leu amides within the solvent-inaccessible core of the protein are labeled (Fig. 5C ▶). As shown in Table 1, the molecular dynamics calculations showed that the values of d_NH-O, ˚_NH-O, and n_HOH-O are more widely distributed (larger standard deviation) for the exposed versus the buried amides. Indeed, for the side-chain–buried ¹³C-labeled Leu amides, a sharp band is observed at ∼1605 cm⁻¹, whose position is essentially independent of temperature. Also, the intensity of the band associated with the unlabeled exposed amides near 1645 cm⁻¹ is markedly decreased. The position of the ¹³C band, together with its temperature independence indicates that it is associated with the ¹³C-labeled amides of the Leu side chains in the solvent-inaccessible interior of the protein.

Table 1.

IR and geometric parameters associated with the exposed Ala and buried Leu residues in α3D′

Sample, temperature	νC=Ob (cm−1)	nHOH−O	νC=O width (cm−1)	σ of dNH−Oc (Å)	σ of ˚NH−Od (°)
13C-Ala α3D′; r.t.a	1593.8	0.58	29.2	0.33	15.0
13C-Ala α3D′; 13 K	1587.9	n.a.	23.4	n.a.	n.a.
13C-Leu α3D′; r.t.a	1606.8	0.12	17.9	0.20	11.8
13C-Leu α3D′; 13 K	1605.0	n.a.	11.8	n.a.	n.a.

^aThe IR data are at 290 K, the simulation temperature was 300 K.

^bThe Gaussian peak position.

^cThe average standard deviation of d_NH−O for the exposed Ala residues (13, 26, 29, 30, 36, 37, 55, 61, 62) and the buried Leu residues (11, 28, 42, 56, 67) in the MD simulation. Averages were calculated over 2–10 nsec.

^dThe average standard deviation of ˚_NH−O for the exposed Ala residues (13, 26, 29, 30, 36, 37, 55, 61, 62) and the buried Leu residues (11, 28, 42, 56, 67) in the MD simulation.

Analysis of the IR spectra

The amide I′ regions of the spectra were fit to a series of Gaussian curves, to allow a more precise determination of the positions, intensities, and widths of the various bands. The amide I′ region of the spectrum of the unlabeled protein is dominated by peaks with Gaussian frequencies and widths as tabulated in Table 1. The protein labeled with ¹³C-Leu shows an additional strong absorption in the 13 K spectrum with a Gaussian frequency of 1605 cm⁻¹ associated with the ¹³C-labeled buried amide carbonyl groups (Fig. 5C ▶). Spectral deconvolution additionally reveals a less intense feature at 1588 cm⁻¹, close to the position expected for exposed, ¹³C-labeled amides. This finding is consistent with the structure, which shows that Leu-21 lies in a solvent-accessible loop between helices 1 and 2 (Fig. 2 ▶). Conversely, the sample labeled with ¹³C-Ala shows a strong absorbance at 1588, with a secondary absorbance at 1605 cm⁻¹, indicating that most of the amides are accessible to solvent.

The peak widths provide information concerning the homogeneity of the environment experienced by the amide groups. At all temperatures, the Gaussian widths of the amide I′ group of exposed amides are approximately twofold broader than for the buried amides (Table 1), presumably reflecting the heterogeneity of the interactions with solvent within the ¹³C-Ala labeled sample. Further, as the temperature is decreased from 290 to 13 K, the width of both bands decreases by ∼6 cm⁻¹. This behavior is consistent with an overall decrease in the disorder of the protein and the surrounding solvent at low temperatures.

The Gaussian frequencies of the solvent-exposed amides depend strongly on temperature, while the frequencies associated with the solvent-inaccessible amides are nearly independent of temperature (Fig. 6A ▶). A plot of the shift in Gaussian frequency relative to that observed at 10 K, shows a strong dependence on temperature from 280 K to 160 K; below this temperature a very small change in δν_C=O is observed. Because the carbonyl groups of the Ala residues are partially exposed to water, one might expect the hydrogen bonding to the backbone to follow the glass transition of the solvent. The absorbance of the OH band, which arises from the HOD in the solvent, is an excellent reporter of this transition (Fig. 6B ▶, filled diamonds). The glass transition occurs at ∼160 to 170 K, and coincides with the point at which the Ala amide I′ band becomes temperature-independent. This finding provides strong evidence for the interaction of water with the solvent-exposed amides.

Figure 6.

Temperature dependence of the peak position (A) and relative absorbance (B) of the peak attributed to the buried ¹³C-Leu amides (filled squares) and the exposed ¹³C-labeled Ala amides (filled circles). The diamonds in B show the relative absorbance of the OH band at 3500 cm⁻¹.

NMR of α₃D

Two parameters were measured by NMR that report on different aspects of the dynamics, hydrogen bonding, and solvent-accessibility of the amide groups of α₃D. The deviations of the amide proton chemical shifts from the value expected for a given amino-acid type in a random coil, Δδ(NH), provide information concerning the strengths of amide hydrogen bonds, and their interactions with water. Figure 7 ▶ illustrates this parameter as a function of position in sequence for the amides in the three helices of α₃D. Within the helical regions, the amide proton chemical shifts are generally shifted upfield relative to the loops, which is consistent with the general trend for helices to experience a net negative Δδ(NH) (Wishart et al. 1991, 1995). Superimposed on this general trend is a sinusoidal variation in this parameter, which matches the period of the α-helix (Kuntz et al. 1991; Zhou et al. 1992). However, there is a significant deviation from this trend in helix 1, at approximately the position of the disruption induced by Thr16.

Figure 7.

Amide proton chemical shifts relative to the values expected for a random coil (A) and the protection factors for hydrogen deuterium exchange (B) for helix 1 (top), helix 2 (middle), and helix 3 (bottom) of α₃D. The bars are drawn with a 3.6-residue helical repeat.

Because a number of different effects can give rise to changes in chemical shift, we take advantage of the strong sequence and structural similarity between the individual helices and compute the average value of Δδ(NH) for all three of the helices (Fig. 8A ▶). Sites within two residues of the ends of the helix were excluded to minimize end effects, and regions that deviated from ideal helical geometries in helix 1 were also excluded. The resulting curve is well described by a sine wave with a period of 3.61 ± 0.08 residues, typical of the α-helical repeat. The mean value of Δδ(NH) for all amides is −0.23 ± 0.04 ppm, which is in good agreement with the mean upfield shift of −0.2 ppm reported by Wishart et al. (1991). The values of Δδ(NH) may be converted to a predicted hydrogen bond length (Cierpicki and Otlewski 2001) using either a d⁻³ (Pardi et al. 1983) or a d⁻¹ (Wishart et al. 1991) distance dependence according to the equations:

Figure 8.

Mean chemical shift indices (A) and protection factors (B) as a function of the position within the regions of helices 1–3. The parameters from a total of 14 residues, corresponding to the central 4 turns of the helices were averaged. The end residues were not included to avoid complications associated with end effects. The residues considered span 5 –18 (helix 1), 29–42 (helix 2), and 54–67 (helix 3). Additionally, the region N-terminal to Thr16 (8–11) was also excluded.

(1)

(2)

where d is the hydrogen bond distance in Å from the amide proton donor to the oxygen acceptor, and Δδ is expressed in parts per minute. Analysis of the data in Figure 8A ▶ indicates that the most solvent-exposed amide would be expected to have a hydrogen bond length of ∼2.23 Å, whereas the most solvent inaccessible amides would have a predicted hydrogen bond length of 1.95 Å, as determined using equation 2. A similar conclusion was reached by examining the relationship between Δδ(NH) and d_NH-O for helices as described in Figure 5 ▶ of Cierpicki and Otlewski (2001). The corresponding values computed using equation 2 are 2.08 and 1.91 Å, respectively. Equation 1 agrees with the range of distances predicted by the molecular dynamics calculations for helix 1, but it underestimates some of the longer distances observed for helix 2 (Fig. 2 ▶). Equation 2 underestimates the longest hydrogen bond distances for both helices.

Hydrogen-deuterium protection factors (Fig. 7B ▶) provide a second parameter defining the environment of the amide protons. These parameters describe the extent to which hydrogen-deuterium exchange is slowed relative to the value expected for a fully exposed amino acid residue of the same type. Amides with protection factors matching the overall stability of the protein are able to exchange only when the entire protein unfolds, whereas amides with very low protection factors generally correspond to disordered or mobile loops in the protein. Intermediate behavior corresponds to structured regions of the protein that are able to locally distort to allow for exchange with solvent. The amides within the helical regions of α₃D all showed significant protection from exchange. Figure 7 ▶ compares the protection factors for the three helices. It is immediately apparent that the structural distortions observed in helix 1 have significantly decreased its stability toward hydrogen-deuterium exchange. The remaining two helices tend to show a sinusoidal variation in the protection factors, although the pattern is less distinct than for the chemical shifts. The deviation from sinusoidal behavior is most significant near the C terminus of helix 3, which we attribute to fraying. The buried amides have protection factors that are well predicted from the global stability of the protein (shaded region in Fig. 7B ▶), while the more exposed amides exchange more rapidly by one to two orders of magnitude. The protection factors were averaged in the same manner as the Δδ(NH), resulting in a very good correlation with the corresponding plot of the averaged chemical shift indices (Fig. 8 ▶).

Conclusions

The hydration of amide-carbonyl groups on the exposed side of helices has been proposed to play an important role in defining the stability of helices, and it also exerts a large effect on their spectroscopic properties. Here, we use a three-pronged approach to examine the hydration of helical amides. The dynamics calculations proved to be pivotal to our interpretation of the spectroscopic parameters, and they have provided a wealth of molecular detail that should be of general relevance to understanding the formation and hydration of helices. These calculations indicate that water molecules form hydrogen bonds with the carbonyl groups of most exposed amides with fractional occupancies of 0.5 to 0.8. The less-than-full occupancy indicates that the free energy of this interaction is relatively weak, on the order of thermal energy at room temperature. This finding is consistent with the relatively small energetic spread in the helical propensities, which span ∼1 Kcal/mole. Temperature-dependent, isotope-edited IR spectroscopy provides a particularly convenient method to probe the hydration of amides in helices. The amide I′ band of buried amides is relatively sharp and its position shows little temperature dependence, indicating that the strength of the hydrogen bond remains roughly constant from room temperature to 13 K. On the other hand, the hydrogen bonds on the surface of the protein depend significantly on temperature. Their partial occupancy and the heterogeneity of their interactions with water result in a broader amide I′ vibration at all temperatures. As the temperature is decreased, one would expect the occupancy and homogeneity of the geometries of the hydrogen-bonded water molecules to increase; this is indeed reflected in the amide I′ band, which sharpens and shifts to lower energy, reflecting a strengthening of the hydrogen bonds.

Proton NMR methods provide a complementary method that focuses on the amide proton rather than the amide carbonyl group. Chemical shifts are particularly sensitive to hydrogen bonding and provide a very rapid sequence-specific readout of this interaction. Our molecular dynamics studies of hydrogen bonding in α₃D complement earlier work on this subject (Kuntz et al. 1991; Zhou et al. 1992; Cordier and Grzesiek 2002) by explicitly considering the role of water in weakening the hydrogen bonds of amides on the surface of the protein. Thus, chemical shifts can now be better interpreted in terms of the occupancy of surface-accessible sites of hydration; as the mean occupancy increases the chemical shifts move progressively more upfield. These findings also provide a framework for the understanding of more modern methods of detecting hydrogen bonds that rely on the weak J-coupling between the donor amide and the acceptor carbonyl carbon, which provides a direct readout of particularly strongly hydrogen-bonded interactions (Cordier and Grzesiek 1999,Cordier and Grzesiek 2002; Cornilescu et al. 1999a,b).

An unexpected finding of these studies was the relatively large structural and thermodynamic effect of a hydrogen bond between a Thr hydroxyl and a main-chain amide four residues prior to it in sequence. This hydrogen bond distorted the angle of the N-terminal carbonyl group, which in turn perturbed the helical parameters of the amide preceding it in sequence. Thus, a single Thr hydroxyl had a major impact on the structure of 4–7 residues distant in sequence. This disruption was evident in the NMR chemical shifts and most markedly in the rates of proton exchange. Indeed, the overall exchange of this helix was significantly more rapid than the other two helices. This helical disruption was not considered in the design of α₃D, which suggests that its stability may be enhanced by either replacing it with another polar residue or by repacking the core to conform to the altered conformation occasioned by the presence of Thr16.

Materials and methods

Sample preparation

Specifically labeled ¹³C(1) Leu- and Ala-α₃D′ samples were produced by growing Escherchia coli strain BL21(DE3) in M9 minimal media supplemented with all the amino acids (as the sole carbon source) and ¹³C(1)-leucine (50 mg/L, Isotec) or ¹³C(1)-alanine (50 mg/L, Isotec), respectively. Uniformly ¹⁵N labeled α₃D was produced by growing in M9 minimal media supplemented with uniformly labeled ¹⁵NH₄Cl (1 g/L, Isotec) as the sole nitrogen source. All α₃D variants were purified as described previously (Walsh et al. 1999).

IR spectroscopy

α₃D variants for IR studies were resuspended in 5 mM potassium phosphate buffer in D₂0 (99.99%, Isotec). The pD was raised to 10 with dilute sodium deuteroxide (Isotec) and incubated at 310 K to ensure complete exchange of amide protons for deuterons. After 3 days at 310 K, the samples’ pD was lowered to 7.0 and lyophilized. IR spectra were acquired with 10 mg/mL of protein in 10 mM potassium phosphate, D₂O-glycerol-d₈ (50/50, v/v).

Infrared spectra were obtained with a Bruker IFS 66 Fourier transform IR instrument. The sample compartment was purged with nitrogen to reduce the contribution from water vapor. The light levels were monitored using an HgCdTe (MCY) detector. The spectral resolution was 2 cm⁻¹. The spectra were smoothed using a nine Savitzky-Golay smoothing algorithm. The sample holder was obtained from Graseby Specac (Smyrna). A 15 mm spacer was used between two CaF₂ windows in the transmission cell. The temperature was regulated using an APD closed cycle Helitran cryostat (Advanced Research Systems). The cryostat sample was filled with He gas at atmospheric pressure, which aids in the transfer of heat from the sample. The outer cryostat windows were made of CaF₂. The inner cryostat windows, which experience the temperature gradient, were 2 mm thick and made of ZnSe (Janos Technology). A special holder for these windows was constructed to minimize strain on the windows from contraction at low temperature (Research Instrumentation Shop, University of Pennsylvania School of Medicine). The temperature was measured with a silicon diode near the sample and the temperature was controlled using a Model 9650 temperature controller (Scientific Instruments). Cryogenic temperature profiles were carried out from high to low. The temperature was measured every 10° and the time for equilibration was 5 to 10 min. The spectra were analyzed for their components using the PeakFit software package (Jandel Scientific Software).

Molecular dynamics calculations

All minimization and dynamics calculations were performed using an in-house version of the program ENCAD (Levitt 1990). All atoms in both the protein and solvent were explicitly present and the force-field parameters have been described previously (Levitt et al. 1995, 1997). The NMR structure (PDB code 2a3d) (Walsh et al. 1999) was used as the starting structure for the simulations. The simulation was performed at neutral pH (Asp, Glu, His, Lys, and Arg were all charged). It was initially minimized 500 steps in vacuo. Then, it was solvated in a box of water molecules extending at least 8 Å from any protein atom, resulting in the addition of 2893 water molecules and total system size of 9819 total atoms. The box dimensions were then increased uniformly to yield the experimental density of water at 298 K (0.997 g/mL). Following that, the system was equilibrated with 5000 steps of water-only minimization, 5000 steps of water-only molecular dynamics at 278 K, another 1000 steps of water-only minimization, 500 steps of protein-only minimization, and finally 500 steps of conjugate gradient minimization of the entire system. Periodic boundary conditions were used throughout to minimize edge effects. This was the starting point for the 10 ns simulation at 298 K using a microcanonical ensemble (constant NVE). An 8 Å force-shifted nonbonded cutoff was employed, and the nonbonded list was updated every five steps. The protocols have been discussed in depth by Levitt and coworkers (Levitt et al. 1995, 1997). Structures were saved every 0.2 ps for analysis, resulting in 50,000 structures. The averages reported in the text were computed from 2–10 ns, or over 40,000 structures.

NMR spectroscopy

A 2-mM ¹⁵N-α₃D sample in 50 mM d₃-acetate and 0.05% sodium azide was used throughout the temperature study. ¹H and ¹⁵N chemical shifts (BMRB 4126) of the backbone amide groups were taken from the 303 K spectrum (Walsh et al. 1999) and extrapolated to the other temperatures (278–323 K). Random coil ¹H chemical shifts were taken from (Wishart et al. 1995). Amide proton chemical shift differences were calculated as Δδ (NH) = δ(NH observed) - δ(NH random coil).

Experiments were done on a Varian INOVA 600 spectrometer equipped with a triple resonance, triple axis gradient probe. Ten temperature points were collected between 278 K and 323 K. The temperature was calibrated using a 100% methanol sample from 278 to 303 K (Raiford et al. 1979). Above 303 K, the temperature was calibrated using a 100% ethylene glycol sample (Raiford et al. 1979). ¹H,¹⁵N-HSQC experiments were carried out as described (Zhang et al. 1994). The ¹H and ¹⁵N carriers were placed at 4.70 and 117 ppm, respectively. The spectral widths were 8000 Hz (¹H) and 1900 Hz (¹⁵N). 1024 and 100 complex points were collected in the ¹H and ¹⁵N dimensions, respectively. The data were processed and analyzed using the program Felix (Biosym Inc.).

Abbreviations

• NMR, nuclear magnetic resonance

• IR, infrared

• δ, chemical shift

• HPLC, high performance liquid chromatography

• nsec, nanosecond

• MD, molecular dynamics