Position dependence of non-polar amino acid intrinsic helical propensities1

J. Mol. Biol. (1998) 278, 279±289

Position Dependence of Non-polar Amino Acid Intrinsic Helical Propensities Michael Petukhov1*, Victor MunÄoz1, Noboru Yumoto2 Susumu Yoshikawa2 and Luis Serrano1 1

European Molecular Biology Laboratory (EMBL) Meyerhofstrasse 1, Heidelberg D-69012, Germany 2

Osaka National Research Institute, AIST, Ikeda, Osaka 563, Japan

Until now and based on the success of the helix/coil transition theory it has been assumed that the a-helical propensities of the amino acids are position independent. This has been critical to derive the set of theoretical parameters for the 20 natural amino acids. Here, we have analyzed the behavior of several non-polar residues, Val, Ile, Leu, Met and Gly at the N-cap, at each position of the ®rst helical turn and at a central helical position of a 16-residue peptide model system that starts with eight consecutive alanine residues. We have interpreted the results from these experiments with the model of the helix/coil transition (AGADIR), that indicates that the intrinsic helical propensity is position dependent. Gly, Val and Ile are more favorable at the ®rst turn than in the middle of the a-helix, while for Leu and Met we observe the opposite behavior. The differences between the observed helical propensities are as large as 1.0 kcal/mol in some cases. Molecular modeling calculations using the ECEPP/2 force-®eld equipped with a hydration potential show that this effect can be explained by the combination of three factors: (a) the sidechains in the ®rst helix turn are more solvent-exposed; (b) they have fewer intramolecular van der Waals' contacts; and (c) they posses higher con®gurational entropy than that in the central position of an a-helix. The position-dependent results of the calculations are in reasonable agreement with the experimental estimates and with the intrinsic propensities of the amino acids derived from the statistical analysis of the protein structure database. # 1998 Academic Press Limited

*Corresponding author

Keywords: stability; hydration; entropy; secondary structure; folding

Introduction Substitution experiments in short polyalaninebased peptides have provided a perfect framework to extensively study the helix/coil transition (Scholtz & Baldwin, 1992). This has permitted a quantitative understanding of the cooperativity of a-helix formation and an in-depth analysis of the physical factors governing a-helix formation in water (Scholtz & Baldwin, 1992; Padmanabhan & Baldwin, 1994; MunÄoz & Serrano, 1995a). The current view is that each amino acid type has a characteristic intrinsic helical propensity to populate the a-helix conformation (Creamer & Rose, 1994). In a-helices there are also side-chain:sidechain interactions (between residues at positions i,i ‡ 3 and i,i ‡ 4), interactions of charged or polar residues with the helix macrodipole and capping interactions between the residues ¯anking the 0022±2836/98/160279±11 $25.00/0/mb981682

a-helix and the free NH and CO groups at the ®rst or last helical turn (for a review, see Padmanabhan & Baldwin, 1994; MunÄoz & Serrano, 1995a). More recently, local motifs involving residues outside the helix that pack against helical residues have been described at both the N terminus (hydrophobic staple; Seale et al., 1994; MunÄoz et al., 1995) and C terminus (Schellman motif; Aurora et al., 1994; Viguera & Serrano, 1995). Moreover, a signi®cant contribution to helix stability has been experimentally demonstrated for the hydrophobic staple (MunÄoz et al., 1995; MunÄoz & Serrano, 1995d). The intrinsic helical propensity of the amino acids is normally assumed to be independent of their position within the a-helix because this structure is very symmetrical (Finkelstein et al., 1991; Stapley et al., 1995; MunÄoz & Serrano, 1995a). The ability shown by helix/coil transition models to # 1998 Academic Press Limited

280

Figure 1. Model of the nine-residue Ala-based helix used for our calculations. All hydrogen atoms except those in the amide groups (shown in yellow) of the backbone have been omitted for clarity. The arrow indicates the possible steric clash between the side-chain of Val at a central position of the helix and carbonyl oxygen atom of a residue at position N2.

describe all the experimental observations obtained so far has further supported this hypothesis. However, the ®rst a-helix turn is not geometrically equivalent to the rest of the helix, as illustrated in Figure 1, which shows that the rotational freedom of the side-chain of Val at this point is restricted as compared to that at position N1. In fact, the conformation gÿ (w1 ‡ 60) is completely prohibited for Val at central positions of protein helices because of steric hindrance between the side-chain and the backbone carbonyl oxygen atom of the i ÿ 3 residue. This is observed also in the distribution of rotamers in the protein structures database (McGregor et al., 1987; and see Figure 2). By contrast, this conformational state is well populated for terminal resides at position N1 (see Figure 2). To study the signi®cance of the positional effect on the stability of an a-helix, we synthesized ®ve series of 16 residue Ala-based peptides substituted with Val, Ile, Leu, Met and Gly at N-terminal positions N1, N2, N3, N4 and the central position, Nc (corresponding to position N7 of the nomenclature used by Richardson & Richardson, 1988), that allowed us to safely measure (without in¯uence of

Amino Acid Intrinsic Helical Propensities

Figure 2. The populations of side-chain rotamers for two typical cases: Val at N1 (®rst N-terminal position; open diamonds) and at N5 (central position; open squares) of the a-helices of 315 protein crystal structures at high resolution. The data were obtained with the PROMOTIF program (Hutchinson & Thornton, 1996). In all case, only a-helices of more than nine residues were used. Each point represents a single case in the database. It can be observed that although the number of cases of Val at position N1 is less than at position N5, the 60 w1 rotamer is signi®cantly populated, while it is absent at position N5.

the side-chain:side-chain and helix macrodipole interactions) the helical propensities of the amino acids in the ®rst four and in the central positions. The notations of amino acid positions in an a-helix used here are in accordance with the nomenclature established by Richardson & Richardson (1988). The peptides were analyzed by far-UV circular dichroism and the results interpreted in free energy terms by using the helix/coil transition algorithm AGADIR1s (MunÄoz & Serrano, 1997). To explain the differences in free energy experimentally observed, we estimated the con®gurational entropy of side-chains, van der Walls', electrostatic and solvation contributions at terminal positions N1, N2, N3, N4 and central position N5 of nine-residue Ala-based model helices, using calculations based on ECEPP/2 potential (Momany et al., 1975; Nemethy et al., 1983), combined with a solvation potential (Oii et al., 1987) and the classical Boltzmann-Gibbs approach. These results are compared with the distribution of amino acids in the protein structure database.

Results and Discussion Design of the polyAla-based peptides The template sequence used in this work is (the ®rst seven Ala are N1 through 7): Fr/Ac-AlaAla-Ala-Ala-Ala-Ala-Ala-Ala-Arg-Ala-Ala-Ala-

281

Amino Acid Intrinsic Helical Propensities

Arg-Gly-Gly-Tyr-Am, which is substituted with Val, Ile, Leu, Met and Gly at positions N1, N2, N3, N4 and N7. Here we use the Richardson & Richardson (1988) nomenclature for a-helices to designate the position in the sequence of the substituted residue. All the peptides were amidated (Am), to prevent charge-end effects and there is a Tyr residue at the C terminus, preceded by Gly, to measure peptide concentration. There are two Arg residues at the C-terminal half of the peptide to favor peptide solubility. In these peptides, non-polar side-chains of the guest residues do not interact with the helix macrodipole, nor do they have side-chain:side-chain interactions at any of the positions tested. This allows direct measurement of the intrinsic helical propensities of the chosen amino acids at terminal and central positions. Position N7 is used as the reference for the helical propensity at the center of the a-helix. N-capping propensities are speci®cally measured in the peptides with a ``free'' N terminus (Fr, non-acetylated). In contrast to the standard acetylated peptides (Ac), in which the acetyl group acts as N-cap, in the free peptides the ®rst residue cannot adopt a helical conformation and the change in helical content is due only to its N-capping propensity. CD analysis and comparison to helix/coil transition algorithms The concentration dependence of the far-UV CD of the different peptides was determined in the 10 mM to 500 mM range (data not shown). No concentration dependence was found, except for the peptides containing Val at position N-cap and N1. These two peptides have been eliminated from our analysis. Figure 3 shows CD spectra, obtained as described in Materials and Methods, for the series of synthetic peptides. Table 1 summarises the results of the experiments, peptide sequences and estimated experimental helical content for all the peptides analyzed here. The measured ellipticities correlate well with two other CD parameters that are independent of peptide concentration: the position of the minimum around 208 nm and the ratio R1 (this value is obtained by dividing the ellipticity at 193 nm by the ellipticity minimum in the range 200 to 210 nm (Bruch et al., 1991; data not shown). The estimated population values do not always adopt a U-shape when plotted versus the position in the peptide (see Table 1), as expected if the intrinsic helical propensities are the same at each position. In AGADIR, weak i,i ‡ 4 Ala-Leu, Ala-Val and Ala-Ile interactions were considered (MunÄoz & Serrano, 1995b). These interactions can, in principle, explain the absence of a U-behavior of the helical content when introducing an unfavorable residue at different positions in the peptide. However, modi®cation of this parameter in AGADIR can explain the experimental data for position N7 for only some amino acids (i.e. Leu and Met), but not for the other positions (data not shown). This indicates that at least

some of the amino acids analyzed here do indeed have different helical propensities depending on the position they occupy in the a-helix.

Determination of the positional helical propensities using AGADIR1s Direct interpretation of the experimental helical content in terms of intrinsic helical propensities is dif®cult because the helix/coil is not an all or non transition. Therefore, in the peptide there is not a single a-helix in equilibrium with the coil state, but a broad ensemble of helical conformations with different lengths and involving different residues of the sequence. This information can be directly derived, however, from the ®tting of the far-UV CD data for the different peptides to a statistical mechanical model of the helix/coil transition. For this purpose we have used the algorithm AGADIR1s that includes a single-sequence version of the helix/coil transition (MunÄoz & Serrano, 1997) and the whole set of free energy parameters previously described (MunÄoz & Serrano, 1995a,c). Table 2 shows the results obtained after ®tting the AGADIR1s set of energy parameters to the CD experimental helical contents (see Materials and Methods). Because the peptides containing Val at positions N1 (acetylated) and N-cap (free) showed a concentration-dependent CD spectrum, it was impossible to obtain the N-cap propensity of this residue from our set of data. In order to derive the propensity of Val at the other positions of the a-helix we assumed that its N-cap propensity was the same as that of Ile. This is probably a very reasonable assumption, as Doig & Baldwin (1995) have reported very similar N-cap propensities for these two residues. In fact, the relative differences between the amino acids analyzed here that we observe at position N-cap (from the free peptides) are similar to those found by Doig & Baldwin (1995). The global analysis with a helix/coil model shows that, to reproduce the experimental data the helical propensity of all the amino acids tested must be position dependent. The greatest differences are generally obtained between the ®rst helical position (N1) and the central position (N7), with a rather smooth transition for the intermediate positions (Table 2). However, although the changes in helical content due to the positiondependent intrinsic propensity are obvious, these changes result from very small differences in free energy (of the order of 0.1 to 0.2 kcal/mol). Gly is an obvious exception, showing changes in free energy of up to 1 kcal/mol. There is also a qualitative difference between the b-branched sidechains (Val and Ile) and Leu and Met. The b-branched residues are less unfavorable at the ®rst turn of the helix, while Leu and Met have higher propensity at central helical positions. It is also of interest that Gly is not only very favorable as N-cap (Serrano & Fersht, 1989) but also at other positions of the ®st helical turn.

282 The energy contributions to the intrinsic helical propensities Determination of the con®gurational entropy for the 20 amino acids obtained using the ECEPP/2 potential (Momany et al., 1975; Nemethy et al., 1983) in combination with a classical Boltzmann-

Amino Acid Intrinsic Helical Propensities

Gibbs approach, has been found to reproduce satisfactorily the conformer distributions of non-polar amino acids in the protein database (Abagyan & Totrov, 1994) and those obtained from theoretical simulations using Monte-Carlo techniques (Creamer & Rose, 1994), and different empirical force-®elds (Lee et al., 1994). On the basis of these

Figure 3. CD spectra of the series of peptides used in this study. The indexes of the peptides are indicated. The sequences, mean residue ellipticities and estimations of helical content are given in Table 1. Other experimental conditions are given in Materials and Methods.

283

Amino Acid Intrinsic Helical Propensities

Table 1. Sequences and results of CD measurements of the helical contents for the series of synthetic peptides used in this study Peptide FA1 FI1 FL1 FM1 FG1

Sequence 

NH2 -AAAAAAAARAAARGGY-NH2 NH2 -IAAAAAAARAAARGGY-NH2 NH2 -LAAAAAAARAAARGGY-NH2 NH2 -MAAAAAAARAAARGGY-NH2 NH2 -GAAAAAAARAAARGGY-NH2

ÿ[y]222 (deg cm2/dmol)

Helical content (%)

7300  220 7040  210 7800  240 6100  180 10,200  300

22.3  1.0 21.5  1.0 23.9  1.0 18.4  1.0 31.2  1.0

A1

Ac-AAAAAAAARAAARGGY-NH2

15,500  470

47.3  1.5

I1 I2 I3 I4 I7

Ac-IAAAAAAARAAARGGY-NH2 Ac-AIAAAAAARAAARGGY-NH2 Ac-AAIAAAAARAAARGGY-NH2 Ac-AAAIAAAARAAARGGY-NH2 Ac-AAAAAAIARAAARGGY-NH2

12,600  370 11,200  330 10,810  330 12,170  360 10,280  310

38.6  1.2 34.3  1.0 33.0  1.0 37.2  1.2 31.4  1.0

L1 L2 L3 L4 L7

Ac-LAAAAAAARAAARGGY-NH2 Ac-ALAAAAAARAAARGGY-NH2 Ac-AALAAAAARAAARGGY-NH2 Ac-AAALAAAARAAARGGY-NH2 Ac-AAAAAALARAAARGGY-NH2

14,080  420 11,770  350 12,160  360 13,560  420 13,550  420

43.0  1.4 36.0  1.1 36.3  1.1 41.5  1.4 41.4  1.0

M1 M2 M3 M4 M7

Ac-MAAAAAAARAAARGGY-NH2 Ac-AMAAAAAARAAARGGY-NH2 Ac-AAMAAAAARAAARGGY-NH2 Ac-AAAMAAAARAAARGGY-NH2 Ac-AAAAAAMARAAARGGY-NH2

15,100  450 13,180  410 11,990  360 13,790  430 13,860  430

45.5  1.5 39.7  1.2 35.8  1.0 42.5  1.3 42.2  1.5

G1 G2 G3 G4 G7

Ac-GAAAAAAARAAARGGY-NH2 Ac-AGAAAAAARAAARGGY-NH2 Ac-AAGAAAAARAAARGGY-NH2 Ac-AAAGAAAARAAARGGY-NH2 Ac-AAAAAAGARAAARGGY-NH2

13,200  400 13,060  400 9670  300 7170  210 4360  130

40.3  1.2 39.4  1.2 29.5  1.0 21.5  1.0 15.2  1.0

V2 V3 V4 V7

Ac-AVAAAAAARAAARGGY-NH2 Ac-AAVAAAAARAAARGGY-NH2 Ac-AAVAAAAARAAARGGY-NH2 Ac-AAAAAAVARAAARGGY-NH2

12,070  360 11,210  330 10,980  330 9980  300

36.9  1.1 34.2  1.0 33.6  1.0 30.5  1.0

Far-UV CD spectra of the peptides were obtained at pH 10 (FX1 series), or pH 7 for the rest of the peptides, in 5 mM sodium phosphate buffer at 5 C. Peptide concentrations were 50 mM. The percentage of a-helix was calculated with the empirical equation ÿ100(y222/(39,500(1-2.57/n))) by Chen et al. (1974); where n is the number of peptide bonds and y222 is an experimentally observed ellipticity of peptide at 222 nm.

observations, we have used the ECEPP/2 potential with the accessible surface hydration potential (Oii et al., 1987), to estimate contributions from the con®gurational entropy, non-bonded intramolecular interactions and solvation, to the position dependence of the intrinsic helical propensities of the non-polar amino acids measured by CD. To perform our calculations we used a model polyalanine helix (f, c and o, of ÿ60 , ÿ 40 and 180 , respectively) with blocked N and C termini (acetyl and amide groups). In Table 2 we present the independent contributions of non-bonded interactions, solvation and side-chain entropy to the intrinsic helical propensities of the amino acids at different positions. The values of: ÿ EECEPP
EECEPP ˆ EECEPP helix coil

…1†

represent the pure change of non-bonded interactions (
EECEPP) due to a-helix formation. Nonbonded interactions of non-polar residues, except Gly, in a-helices are more favorable in the central positions than in the ®rst helical turn. This is mainly due to fewer van der Waals' contacts between side-chains of the guest residue and the rest of the helix molecule. It is, of course, the opposite for Gly, which has no side-chain. In the value

of
EECEPP this effect is partly compensated by the unfolded state of the molecule, which has the same chemical structure, and where the terminal residues have fewer van der Waals' contacts as well. However, it is still pronounced in Leu and Met and, with opposite direction, in Gly. Generally, the solvation of the non-polar residues in the ®st helix turn is less favorable (0.1 to 0.4 kcal/mol) than in the center of the helix. This is mainly due to the non-polar residues burying a signi®cant hydrophobic surface in the center of the helix, while in the ®rst turn they are less buried and they partly screen the ®rst four non-bonded amide groups from the solvent. For Gly, the effects are opposite and it has more favorable energy (ÿ0.1 to ÿ0.6 kcal/mol) in the ®rst helical turn. Less restricted rotational freedom of side-chains and hence higher con®gurational entropy is the next important contributor to the changes of intrinsic helical propensities. As we can see, there are signi®cant differences in the changes in con®gurational entropy between the N-terminal and central position of the a-helix, for the majority of the amino acids. Generally, the side-chains at positions N1 and N2 have similar changes in conformational entropy, because of the absence of steric restric-

284

Amino Acid Intrinsic Helical Propensities

Table 2. Contributions to the intrinsic helical propensities of several non-polar amino acids at central and N-terminal positions of a nine-residue a-helix model

Amino acid/ position


EECEPP (kcal/mol)


EHydr (kcal/mol)

Change of conf. Entropy (ÿT
S) at T-278 K (kcal/mol)



Gtheor relative to position Nc(N5) change of free energy of a-helix formation (kcal/mol)



Eexp relative to position Nc(N7) change in intrinsic helical propensitya (kcal/mol)



Eexp relative to Ala change in intrinsic helical propensitya (kcal/mol)

Ile/N1 Ile/N2 Ile/N3 Ile/N4 Ile/Nc

ÿ15.68 ÿ15.62 ÿ15.59 ÿ15.48 ÿ15.54

9.49 9.32 8.89 8.73 9.07

ÿ0.07 ÿ0.03 0.36 0.44 0.38

ÿ0.17 ÿ0.24 ÿ0.25 ÿ0.22 0.00

ÿ0.13 ÿ0.03 0.00 ÿ0.17 0.00

0.41 0.51 0.54 0.37 0.54

Leu/N1 Leu/N2 Leu/N3 Leu/N4 Leu/Nc

ÿ17.32 ÿ17.46 ÿ17.60 ÿ17.80 ÿ17.83

9.35 9.27 9.27 9.26 9.32

0.29 0.40 0.36 0.46 0.44

0.39 0.28 0.10 ÿ0.01 0.00

0.00 0.25 0.15 0.00 0.00

0.21 0.46 0.36 0.21 0.21

Val/N1 Val/N2 Val/N3 Val/N4 Val/Nc

ÿ16.18 ÿ15.99 ÿ16.20 ÿ15.96 ÿ15.93

9.12 8.99 8.90 9.04 8.99

0.25 0.27 0.46 0.42 0.46

ÿ0.33 ÿ0.25 ÿ0.36 ÿ0.02 0.00

± ÿ0.17 ÿ0.12 ÿ0.12 0.00

± 0.39 0.44 0.44 0.56

Met/N1 Met/N2 Met/N3 Met/N4 Met/Nc

ÿ17.48 ÿ17.83 ÿ17.99 ÿ18.22 ÿ18.32

9.62 9.64 9.46 9.50 9.51

ÿ0.14 ÿ0.04 0.20 0.25 0.22

0.59 0.36 0.26 0.12 0.00

ÿ0.10 0.10 0.20 0.00 0.00

0.06 0.26 0.36 0.16 0.16

Gly/N1 Gly/N2 Gly/N3 Gly/N4 Gly/Nc

ÿ16.04 ÿ16.51 ÿ16.27 ÿ16.07 ÿ16.03

8.78 8.65 9.04 9.11 9.20

± ± ± ± ±

ÿ0.43 ÿ1.03 ÿ0.40 ÿ0.13 0.00

ÿ0.60 ÿ0.85 ÿ0.50 ÿ0.20 0.00

0.66 0.41 0.76 1.06 1.26

The average values of EECEPP and EHydr were calculated separately for folded and unfolded states using the Boltzmann factor for all possible conformers of a particular amino acid side-chain in the grid of 10 step, as described in Materials and Methods. Values of
EECEPP and
EHydr were calculated as the differences between folded and unfolded states. Con®gurational entropy, S was calculated using the classical Boltzmann formula S ˆ ÿ RPiln(Pi), where R is the gas constant and Pi are the probabilities of the conformational states, which were calculated from the canonical Boltzmann-Gibbs distribution. a The changes of intrinsic helical propensities between the central and several N-terminal positions were obtained by ®tting the AGADIR1 s set of energy parameters (MunÄoz & Serrano, 1997) to CD measurements of helical content of the Ala-based peptides listed in Table 1. The error for positions N2 to N4 and N7 is around 0.1 kcal/mol, while at position N1 it is around 0.2 kcal/mol. The error in the estimations was obtained by determining the intrinsic contribution that will give a helical content 3% higher than the experimental one. The difference between the intrinsic contribution obtained by reproducing the experimental value and that described above, is the error in the energy estimation.

tions associated with the previous helical turn. The differences between the changes in conformational entropy at the N-terminal positions and those at central positions depend on the amino acid, varying from 0.1 to 0.2 kcal/mol for Val and Leu, to around 0.4 to 0.5 kcal/mol for Met and Ile. The side-chain of Ala has, of course, the same entropy at terminal and central positions. This is the most probable explanation for the fact that models incorporating a uniform loss of entropy yield rather accurate results for Ala-based peptides (Finkelstein et al., 1991; Stapley et al., 1995; MunÄoz & Serrano, 1995a). The guest residues have a transitional position at N3 where restrictions to rotational freedom become operative. It is of interest that for nonN-blocked peptide helices, as well as protein helices, the transitional position should be shifted to N4, because the blocking groups such as acetyl can effectively play the role of a capping residue and restrict the rotational freedom of the sidechain at N4 in N-blocked helices (see Figure 1). As for several C-terminal positions (the last helix turn)

these seem to have basically the same entropic properties as those of central positions because of the Christmas tree-like arrangement (with the top being the C terminus) of side-chains in an a-helix as shown in Figure 1. The theoretical free energy differences for each amino acid with respect to the central position are presented in Table 2. There is a reasonably good correlation (the Pearson's correlation coef®cient and the mean quadratic deviation are 0.793 and 0.202, respectively), between the experimental and theoretical free energy values (data not shown). This correlation improves signi®cantly if we eliminate the data at position N1 (the Pearson's correlation coef®cient is 0.91). Except for a few cases (mainly positions N1 for Leu and Met) the direction and magnitude of the changes in intrinsic helical propensities are predicted well. The relatively poor correlation between theoretical prediction and experimental data for substitutions at position N1 can be explained by the unique environment of this position, which only has an acetyl group pre-

285

Amino Acid Intrinsic Helical Propensities

Figure 4. Frequencies of several non-polar amino acids at N-terminal and central positions of a-helices from 315 non-related protein structures at high resolution. The data were obtained with the WHATIF program (Vriend, 1990).

ceding it. Therefore, its conformational entropy, van der Waals' interactions and solvation terms in the denatured state could be different from that for positions N2, N3, N4 and Nc. This should not be the case for Gly and, in fact, the prediction for this residue at position N1 is within the experimental error. The possible formation of non-a-helical conformations such as 310-helices that are quite probable at the end of helices (Millhauser et al., 1997) and the interaction of terminal aliphatic side-chains (at position N0 ) with Ala at position N4, akin to that found in a hydrophobic staple motif (MunÄoz & Serrano, 1995a), also could explain the behaviour of the N1 position.

and central positions. This is probably because the position effect for this amino acid is small and therefore, easily overcome by tertiary interactions in proteins. Finally, Gly, the amino acid with the most pronounced effect, does not have a sidechain. Therefore, we should expect to ®nd the best correlation with the database frequencies, as the in¯uence of tertiary interactions should be minimal. Figure 5 shows a clear correlation between the experimental and theoretical changes of intrinsic helical propensities, and the frequencies of Gly at different positions in protein a-helices.

The position dependence of intrinsic propensities and the protein database

There have been a signi®cant number of experimental studies in which the same amino acid (Asp, His, Glu, Gly) has been placed at different positions of a guest peptide (see HuyghuesDespointes et al., 1993, and references therein; Chakrabartty et al., 1991). U-shaped curves for positional dependencies of helical contents were obtained in those experiments, which were previously interpreted in terms of helix fraying

It has been shown that the f, c distribution of the different amino acids correlates well with the intrinsic secondary structure propensities of the 20 amino acids (MunÄoz & Serrano, 1994; Swindells et al., 1995). Moreover, this distribution can be used to reproduce NMR experimental data in random coil peptides (Serrano, 1995; Smith et al., 1996). Therefore, we should expect to ®nd a reasonable correlation between the positional helical propensities described here and the relative preferences of the same amino acids for positions N1 to N4 and central helical positions. Figure 4 shows that, indeed, such an agreement exists. In particular Gly and Ile are more frequent at those positions of protein a-helices (N1, N2, N3 and N4, respectively) where signi®cant favorable changes of its intrinsic helical propensities have been found (see Table 2). Leu and Met at position N2 and N3 are signi®cantly less frequent than their counterparts at central positions, in agreement with the predictions. Val was found to have slightly ( ÿ 0.1 kcal/mol) better helical propensities at the ®rst helix turn. However, it has signi®cantly lower frequencies at positions N1 and N2, and approximately the same frequencies at positions N3, N4

Discussion

Figure 5. Correlation between the experimental and the theoretical changes of intrinsic helical propensities and frequencies of Gly at different positions in protein a-helices. The data were obtained as described in Materials and Methods

286 and/or interactions with the helix macrodipole (Huyghues-Despointes et al., 1993). However, the possibility of a position dependence of the helical propensities of the amino acids was not reported in those studies. Looking at the original design and at the positions mutated in those studies it is easy to understand why no indication of a positional effect was found. In all the cases for the majority of the substitutions, possible i,i ‡ 3 or i,i ‡ 4 interactions with Gln or Lys residues, introduced to make the peptide soluble, are found. Moreover in the case of the Gly series, the ®rst two positions of the peptide after the acetyl group, where we found the strongest position effect, were not mutated. In our peptide design we have avoided all possible i,i ‡ 3 and i,i ‡ 4 interactions, and mutated in a consecutive manner all positions in the ®rst helical turn, which allows detection of the positional effect. It is evident that similar positional effects will be found for the remaining 15 amino acids, although in the case of charged or aromatic residues, the interpretation will be more dif®cult due to added factors like charge-dipole interactions and aromatic effects on the CD or NMR spectra, respectively. Interestingly, we have found a good correlation between the experimental data and the statistic positional preference in the protein database, con®rming what has been found in other instances, that the protein database can be used to derive pseudo-energy potentials that correlate with free energies. However, the most interesting result is the fact that we can reproduce to a signi®cant extent the positional free energy contribution using the ECEPP/2 potential in combination with an accessible surface hydration potential (except for position N1). This opens the possibility of performing similar calculations in the case of other amino acids. It should be expected that the positional effect would not be of similar magnitude for all amino acids, since as we have seen it is due to a combination of at least three different factors, van der Waals' interactions, entropy of side-chain and solvation. In some cases, the three factors could go in the same direction and large positional energy differences will be found, in other cases they could cancel out and no signi®cant difference will be found. However, from a pure thermodynamic point of view, in those cases in which different effects will cancel out, the temperature and solvent dependence will not be the same. We have recently introduced the positional dependence of the helical intrinsic propensities in a new version of the helix/coil transition algorithm, AGADIR1s-2 (L.S., unpublished). Consideration of positional effects improves signi®cantly the predictive power of the algorithm.

Conclusions We have analyzed the intrinsic helical propensity of several non-polar amino acids at several

Amino Acid Intrinsic Helical Propensities

positions of the a-helix. Our results show that this propensity is different at each position of the ®rst turn helix turn and in the middle of a a-helix. The effects are complex and vary in magnitude and sign for different amino acids. Gly shows the strongest dependence on position (ÿ0.85 kcal/mol) and is, like Ile and Val, more favorable at the ®rst turn. Leu and Met are, on the other hand, more favorable in the center of the helix. To rationalize our results in terms of individual free energy contributions we have performed molecular mechanics calculations. The calculations indicate that the position dependence of the intrinsic helical propensity is due to three factors: (a) greater solvent exposure of the side-chain in the ®rst helix turn in a central position of the helix; (b) fewer intramolecular van der Waals' contacts; (c) higher con®gurational entropy. The incorporation of this effect into helix/ coil algorithms should improve their ability to predict the helical tendency of proteins and peptides.

Materials and Methods Experimental procedures The peptides were synthesized by the EMBL peptide synthesis service using Fmoc chemistry and PyBOP activation at a 0.025 mmol scale. Peptide homogeneity and identity were analyzed by analytical HPLC, amino acid analysis and matrix-assisted laser desorption time-of¯ight mass spectroscopy. The concentration of peptide was determined from the UV absorbance of the C-terminal Tyr using the method of Gill & von Hippel (1989). The error in the concentration determination was around 2%. Circular dichroism (CD) spectra were recorded on a Jasco-710 instrument at pH 7 and a temperature of 278 K. To check for concentration dependence of the CD spectra, different dilutions of the peptides (10 mM to 500 mM) were scanned. CD spectra in the range of 190 to 250 nm were obtained using the continuous scan option (100 nm/minute/scan speed), with a one second response time taking points every 0.2 nm; 30 scans were taken for every sample. The helical content of the peptides was estimated using the mean residue ellipticity at 222 nm (Chen et al., 1974). Survey of the Protein Data Bank Population maps of amino acids were derived from 315 protein crystal structures at high resolution with less than 50% homology (Vriend, 1990) with the PROMOTIF program (Hutchinson & Thornton, 1996) and the WHATIF program (Vriend, 1990). The crystal structures of the proteins were taken from the Brookhaven Protein Data Bank (release of April 1997). We searched for the sequence motif STC/H/H/H/H/H/H/H/H/H, where S is strand, T is turn, C is coil and H is helix. Calculations based on statistical mechanics The change in free energy for a-helix formation upon mutation in a helical peptide cannot be precisely calculated using a standard two-state model. A more precise estimation can be obtained by ®tting a helix/coil transition algorithm to the changes in helical content detected by far-UV CD (MunÄoz & Serrano, 1995b,c). The

287

Amino Acid Intrinsic Helical Propensities calculations based on statistical mechanics were made with the AGADIR1s computer program (MunÄoz & Serrano, 1997), modi®ed to include the possibility that the residue immediately following an acetyl group, or preceding an amide group, is helical. The following energy contributions are included in the calculations of the partition functions of all possible helical segments in AGADIR1s: hydrogen bonding in the backbone (ÿ0.875 kcal/mol); loss of entropy required for adoption of helical dihedral angles by amino acids; capping interactions at N and C termini; and interactions of charged groups with themselves and the helix macrodipole. To obtain the energy parameters, we used the approximation described by MunÄoz et al. (1995) and Myers et al. (1997). Essentially, we treated the peptide as a homopolymer in which we varied one parameter at a time. First of all, the intrinsic helical propensity of Ala and its N-capping contribution was modi®ed until we could reproduce within a 1% difference the helical content of the control non-acetylated peptide having Ala at position N-cap. The ®nal value of the intrinsic helical propensity for Ala was found to be ‡0.618 kcal/mol. Since the non-blocked peptides at the N terminus were analyzed at pH 10, we did not expect any charge interactions of the free N-terminal group with the rest of the helix. Once we could reproduce the experimental results of the reference peptide with or without an N-terminal acetyl group, we introduce the N-capping propensities of Ala, Leu, Ile, Val, Met and Gly in order to obtain the experimental numbers of the corresponding N-terminal unprotected peptides. Those were found to be ‡0.5, ‡0.4, ‡0.6, ‡0.6, ‡0.8 and ÿ0.05 kcal/mol, respectively. The N-capping values were ®xed and then we modi®ed the intrinsic propensities of the ®ve amino acids at positions N1 to N4 and Nc, in order to reproduce the experimental values within a 1% difference. Calculations based on molecular mechanics Energy pro®les of side-chains of natural amino acids in the folded state were calculated for terminal positions N1, N2, N3, N4 and central position Nc (corresponding to position N5 of the nomenclature used by Richardson & Richardson, 1988) of nine-residue Ala-based model helices. The N and C termini of the model a-helix were acetylated and amidated, respectively. The energy calculations were made with the BKS molecular modeling program (Mazur & Abagyan, 1989) using the ECEPP/2 force-®eld (Moman et al., 1975; Nemethy et al.,1983). All atoms in the peptides were treated explicitly. Bond lengths and bond angles were ®xed at their standard values during the energy calculations and minimization. The van der Waals', electrostatic, hydrogen bond and torsion potentials were included in the energy calculations. The choice of dielectric constant is always a problem in molecular-mechanical calculations. A low dielectric constant (e 2 to 4) yields good results for internally buried groups of globular proteins, while the effective dielectric constant of bulky water (e 81) is more appropriate for solvent-exposed groups (for a review, see Warshel & Russel, 1984). Several distance-dependent models for e have been successfully employed to mimic the water screening of electrostatic interactions in proteins. In this study, we chose the two extreme cases (e 2 and 81) to examine the dependence of the results of calculations on the dielectric model used. It was found that results of calculations were almost independent of the e value used (data not shown), indicating that electrostatic interactions play a minor role in the intrinsic helical pro-

pensities of non-polar amino acids. Therefore, we used a value of 81 for the data shown in Table 2. To avoid the necessity of renewing interaction lists during the energy calculations, we used the complete list of non-bonded interactions of the peptide under investigation. The solvation energy term was modeled by continuum approximation model for protein: solvent interactions (Oii et al., 1987). Accessible surface area was calculated with the NSC program (Eisenhaber et al., 1995). The source code of the algorithm was kindly provided by the author of program. The van der Waals' radii and atomic solvation parameters were taken from Oii et al. (1987). Energy pro®les were calculated with grid steps of 10 , with 50 subsequent steps of energy minimization by the conjugate gradient method. The dihedral angles f,c and o of the backbone of peptides in the folded state were ®xed at standard values of ÿ60 , ÿ40 and 180 , respectively. In the unfolded state, these angles were initially set to 180 and allowed to vary by a conjugate gradient energy minimization algorithm. The ÿCH3 groups were considered not to contribute to entropy because of their symmetries and the conformations of these groups were ®xed with dihedral angles of 60 . Other dihedral angles of side-chains were varied, as described above. Calculations of configurational entropy The con®gurational entropy of amino side-chains, S was calculated from the energy pro®les at 278 K by a classic Boltzmann-Gibbs approach: X …2† S ˆ ÿR Pi ln…Pi † where i is the index of summation over all selected conformational states (see below), R is the gas constant and Pi is the probability of the state, which can be calculated from the canonical Gibbs distribution: X XX exp…ÿEij =RT†= exp…ÿEij =RT† …3† Pi ˆ Index j indicates the summation over the grid points of the phase space that belong to the selected conformational states and for which the energy of the system, Eij, has been calculated. The conformational states were separated as described (Lee et al., 1994). The linkage between two sp3 carbon atoms had three conformers: gÿ, 0 to 120 : t, 120 to 240 ; and g‡, 240 to 360 . The discrete summation over the grids with steps of 10 was found to provide an adequate approximation of the partition fraction: the number of points is still manageable and differences between estimates of entropy obtained with grids 10 and 1 in all cases were below 1% (data not shown).

Acknowledgements M.P. is the recipient of a short-term fellowship from the EMBO program for Eastern European Countries. The authors are grateful to Dr E. G. Hutchinson and Dr J. M. Thornton (Department of Biochemistry and Molecular Biology, University College, London) for the recent version of the PROMOTIF computer program and Dr F. Eisenhaber (European Molecular Biology Laboratory, Heidelberg, Germany) for NSC computer programs that were used in this study. The new version of AGADIR1s, including a position dependence of the intrinsic propen-

288 sities, is available at http://www.embl-heidelberg.de/ Services/index.html#5.

References Abagyan, R. & Totrov, M. (1994). Biased probability Monte-Carlo conformational searches and electrostatic calculations for peptides and proteins. J. Mol. Biol. 235, 983± 1002. Aurora, R., Srivanasan, R. & Rose, G. D. (1994). Rules for a-helix termination by glycine. Science, 264, 1126± 1130. Blaber, M., Zhang, X.-J. & Matthews, B. W. (1993). Structural basis of amino acids alpha-helical propensity. Science, 260, 1637± 1640. Bruch, M. D., Dhingra, M. M. & Gierasch, L. M. (1991). Side chain backbone hydrogen bonding contributes to helix stability in peptides derived from an a-helical region of carboxypeptidase A. Proteins: Struct. Funct. Genet. 10, 130± 139. Chakrabartty, A., Schellman, J. A. & Baldwin, R. L. (1991). Large differences in the helix propensities of alanine and glycine. Nature, 351, 586 ± 588. Chen, Y. H., Yang, J. T. & Chau, K. H. (1974). Determination of the helix and b-form of proteins in aqueous solution by circular dichroism. Biochemistry, 14, 3350± 3359. Creamer, T. P. & Rose, G. D. (1994). a-Helix-forming propensities in peptides and proteins. Proteins: Struct. Funct. Genet. 19, 85± 97. Doig, A. J. & Baldwin, R. L. (1995). N- and C-capping preferences for all 20 amino acids in a-helical peptides. Protein Sci. 4, 1325± 1336. Eisenhaber, F., Lijnzaad, P., Argos, P., Sander, C. & Scharf, M. (1995). The double cubic lattice method: ef®cient approach to numerical integration of surface area and volume and to dot surface contouring of molecular assemblies. J. Comput. Chem. 16, 273± 284. Finkelstein, A. V., Badretdinov, A. Y. & Ptisyn, O. B. (1991). Physical reasons for secondary structure stability: a-helices in short peptides. Proteins: Struct. Funct. Genet. 10, 287± 299. Gill, S. C. & von Hippel, P. H. (1989). Calculation of protein extinction coef®cients from amino acid sequence data. Anal. Biochem. 182, 319 ± 326. Hutchinson, E. G. & Thornton, J. M. (1996). A program to identify and analyse structural motifs in proteins. Protein Sci. 5, 212 ± 220. Huyghues-Despointes, B. M. P., Scholtz, J. M. & Baldwin, R. L. (1993). Effect of a single aspartate on helix stability at different positions an a neutral alanine-base peptide. Protein Sci. 2, 1604± 1611. Lee, K. H., Xie, D., Freire, E. & Amzel, L. M. (1994). Estimation of changes in side chain con®gurational entropy in binding and folding: general methods and application to helix formation. Proteins: Struct. Funct. Genet. 20, 68 ±84. Mazur, A. K. & Abagyan, R. A. (1989). New methodology for computer-aided modelling of biomolecular structure and dynamics. 1. Non-cyclic structures. J. Biomol. Struct. Dynam. 6, 815± 832. McGregor, M. J., Islam, S. A. & Sternberg, M. J. E. (1987). Analysis of the relationship between sidechain conformation and secondary structure in globular proteins. J. Mol. Biol. 198, 295 ± 310. Millhauser, G. L., Stenland, C. J., Hanson, P., Bolin, K. A. & van de Ven, F. J. (1997). Estimating the

Amino Acid Intrinsic Helical Propensities relative populations of 310-helix and alpha-helix in Ala-rich peptides: a hydrogen exchange and high ®eld NMR study. J. Mol. Biol. 267, 963± 974. Momany, F. A., McGuire, R. F., Burgess, A. W. & Scheraga, H. A. (1975). Energy parameters in polypeptides. VII. Geometric parameters, partial atomic charges, nonbonded interactions and intrinsic torsional potentials for the naturally occurring amino acids. J. Phys. Chem. 79, 2361± 2381. MunÄoz, V. & Serrano, L. (1994). Intrinsic secondary structure propensities of the amino acids, using statistical phi-psi matrices: comparison with experimental scales. Proteins: Struct. Funct. Genet. 20, 310± 311. MunÄoz, V. & Serrano, L. (1995a). Helix design, prediction and stability. Curr. Opin. Biotechnol. 6, 382± 386. MunÄoz, V. & Serrano, L. (1995b). Elucidating the folding problem of helical peptides using empirical parameters. II. Helix macrodipole effects and rotational modi®cation of the helical content of natural peptides. J. Mol. Biol. 245, 275± 296. MunÄoz, V. & Serrano, L. (1995c). Elucidating the folding problem of helical peptides using empirical parameters. III. Temperature and pH dependence. J. Mol. Biol. 245, 297± 308. MunÄoz, V. & Serrano, L. (1995d). Analysis of i,i ‡ 5 and i,i ‡ 8 hydrophobic interactions in a helical model peptide bearing the hydrophobic staple motif. Biochemistry, 34, 15301± 15306. MunÄoz, V. & Serrano, L. (1997). Development of the multiple sequence approximation within the AGADIR model of a-helix formation. Comparison with Zimm-Bragg and Lifson-Roig formalisms. Biopolymers 41, 495± 509. MunÄoz, V., Blanco, F. & Serrano, L. (1995). The ``hydrophobic staple'' motif. A role for loop-residues in a-helix stability and protein folding. Nature Struct. Biol. 2, 380±385. Myers, J. K., Pace, C. N. & Scholtz, J. M. (1997). A direct comparison of helix propensity in proteins and peptides. Proc. Natl Acad. Sci. USA, 94, 2833± 2837. Nemethy, G., Pottle, M. S. & Scheraga, H. A. (1983). Energy parameters in polypeptides. 9. Updating of geometrical parameters, nonbonded interactions and hydrogen bond interactions for the naturally occurring amino acids. J. Phys. Chem. 87, 1883± 1887. Oii, T., Oobatake, M., Nemethy, G. & Scheraga, H. A. (1987). Accessible surface areas as a measure of the thermodynamic parameters of hydration of peptides. Proc. Natl Acad. Sci. USA, 84, 3086± 3090. Padmanbhan, S. & Baldwin, R. L. (1994). Helix-stabilizing interaction between tyrosine and leucine or valine when spacing if i,i ‡ 4. J. Mol. Biol. 241, 706 ± 713. Richardson, J. S. & Richardson, D. C. (1988). Amino acid preferences for speci®c locations at the ends of a-helices. Science, 240, 1648± 1652. Seale, J. W., Rajgopal, S. & Rose, G. D. (1994). Sequence determinants of the capping box, a stabilizing motif at the N-terminal of a-helices. Protein Sci. 3, 1741± 1745. Scholtz, J. M. & Baldwin, R. L. (1992). The mechanism of a-helix formation by peptides. Annu. Rev. Biophys. Biomol. Struct. 21, 95± 118. Serrano, L. (1995). Comparison between the f distribution of the amino acids in the protein database and NMR data indicates that amino acids have various f propensities in the random coil conformation. J. Mol. Biol. 254, 322± 333.

Amino Acid Intrinsic Helical Propensities Serrano, L. & Fersht, A. R. (1989). Capping and a-helix stability. Nature, 342, 296± 299. Smith, L. J., Bolin, K. A., Schwalbe, H., MacArthur, M. W., Thornton, J. M. & Dobson, C. M. (1996). Analysis of main chain torsion angles in proteins. Prediction of NMR coupling constants for native and random-coil conformations. J. Mol. Biol. 255, 494 ± 506. Stapley, B. J., Rohl, C. A. & Doig, A. J. (1995). Addition of side chain interactions to modi®ed Lifson-Roig helix-coil theory: application to energetics of phenyalaninemethionine interactions. Protein Sci. 4, 2383± 2391.

289 Swindells, M. B., MacArthur, M. W. & Thornton, J. M. (1995). Intrinsic f, c propensities of amino acids, derived from the coil regions of known structures. Nature Struct. Biol. 2, 596± 603. Viguera, A. R. & Serrano, L. (1995). Experimental analysis of the Schellman motif. J. Mol. Biol. 251, 150 ± 160. Vriend, G. (1990). WHATIF: a molecular modeling and drug design program. J. Mol. Graph. 8, 52± 56. Warshel, A. & Russel, S. (1984). Calculations of electrostatic interactions in biological systems and in solution. Quart. Rev. Biophys. 17, 283 ± 422.

Edited by A. R. Fersht (Received 18 September 1997; received in revised form 21 January 1998; accepted 30 January 1998)