If a membrane protein consists of several helices that are clustered to form a pore for the translocation of hydrophilic solutes, the helices will not be hydrophobic over their entire surfaces. On the side facing the lipids they will be hydrophobic, but on the side facing the pore, they are likely to be at least partially hydrophilic. Such helices are called amphipathic and, because of the hydrophilic nature of the substrates, they are envisaged to play a major role in the structure of the sugar transporters. The oldest method of illustrating such helices is to construct a helical wheel (Schiffer and Edmundson, 1967), but a more quantitative plot was proposed by Eisenberg et al. (1982) who introduced the hydrophobic moment. These plots are well suited to demonstrate the amphipathic nature of an individual helix, but not appropriate to identify amphipathic helices on an amino acid sequence of several hundred residues, such as GLUT1. Consequently, an additional method for evaluating the consensus structure of the sugar transporter family was required that possessed the ability to detect the relative accessibility of residues to the surrounding solvent, plus helical periodicity.
One of the features of an amino acid alignment is that information can be gained regarding the constraints that have been placed upon each residue throughout evolution. Such constraints become apparent in the different character that is exhibited between residues buried within the core of a protein and those found at the surface. In essence, buried residues are more conserved than those that are exposed. Moreover, the cores of water soluble proteins tend to be more hydrophobic than their surface positions which are in contact with aqueous solution, whereas the cores of membrane proteins tend to be more polar than their lipid facing exteriors (Rees et a/., 1989). These differences can be used to predict the extent to which each position in a protein is buried by examining the residues present at each position of a sequence alignment.
The periodicity of residues on the face of an a-helical structure that are exposed to a membrane, coupled with the increased sequence variability of exposed residues suggests the possibility of identifying exposed residues by analysing the sequence alignments of homologous proteins. Assuming (a) that the sequence represents a transmembrane helix and (b) that the helix is positioned on the exterior of a helix bundle, the residues in contact with the lipid bilayer may be identified from the pattern of hypervariable positions occurring with a periodicity of about 3.6 residues in a family of sequence alignments. Even a superficial examination of the sugar transporter sequence alignment reveals an apparent periodicity of certain residues, perhaps the most prominent of which is the presence of glycine residues separated by three or four residues in helices 2, 3, 4 and 9.
This type of approach has been used previously with hydrophobicity scales (Cornette et a/., 1987) and variability characteristics (Donnelly at a/., 1989) found within amino acid alignments. In addition, the relative directions of the conserved and variable faces of a membrane-spanning helix can be used to predict whether an exposed face is in contact with a lipid or aqueous environment. Another method of predicting the faces of helices in contact with the lipid bilayer from sequence alignments, is to predict the accessibility of each residue position in the alignment from the substitution pattern at that position. The structural environment of an amino acid residue provides constraints on the evolutionary diversity of that residue. The amino acid substitution patterns are characteristic of their structural environment so that the mutational properties of an exposed residue are different to those of a buried one. Consequently, it is possible to predict the structural environment of residues from a sequence alignment, for which only a substitution pattern is known.
Environment-dependent substitution tables derived from accessible and inaccessible residues in aligned protein structures (Overington ef a/., 1992) are used to predict whether the substitution patterns in sequence alignments are more typical of buried or exposed residues. Substitution tables derived from
residues that are accessible to a lipid environment are used to predict and orientate transmembrane helices (Donnelly et a/., 1993).
Fourier transform methods provide a quantitative approach for characterising the periodicity of conserved and variable residues in a family of aligned sequences (Komiya at a/., 1988). First, the variability profile is constructed from aligned sequences of the helical regions. Next, the residue positions with greatest variability consistent with an a-helical periodicity are determined by fitting a cosine curve to the variability profile. The residue positions for which this Fourier series has the greatest amplitude correspond to the most variable positions. Calculation of these positions for the 11 RC transmembrane helices shows a strong correlation between the most variable positions and the exposed positions (Rees at a/., 1989). The variablity profile may also be used to predict the presence of a-helical segments which are usually identified from hydropathy plots or hydrophobic moment analysis.
The periodicity of hydrophobicity (H), conservation (C) and substitution (S) for the putative membrane spanning helices of the sequence alignment were calculated using PERCON, PERHYD and PERSCAN software (Donnelly atal.,
1989,1993) at University College London. Table 3.3 provides some of the data obtained.
It is clear from the data presented in Table 3.3 that the periodicity of residues throughout the putative transmembrane spanning domains indicates a-helical structure for each of the regions analysed. Although not shown in Table 3.3, the programs also calculate a property of the alignment termed the alpha periodicity (AP). AP is analogous to y used by Komiya at a i (1988) and to the amphipathic index AI used by Cornette at a i (1987), although the precise boundaries of the helical regions of the power spectrum differ in the latter. Komiya at a i (1988) suggest that a value of AP greater than 2 indicates that the helical periodicity is significant. Larger values of AP correspond to a greater fraction of the P(&) curve in the a-helical region. If peripheral helices, in a helix
Table 3.3 Predicted number of residues per turn for each putative transmembrane helix.
Periodicity in the patterns of residue substitution (S), conservation (C) and hydrophobicity (H) were calculated from the alignment of the sugar transporter family.
Family 1 Families 1, II, III and IV
S C H S C H 1 3.33 3.40 3.43 3.00 3.50 3.33 2 3.60 3.36 3.71 3.75 3.46 3.60 3 3.64 3.56 3.46 3.53 3.19 3.50 4 3.50 3.53 3.71 3.19 3.43 3.71 5 4.00 3.33 3.87 3.64 3.33 3.83 6 3.46 3.40 3.43 3.64 3.36 3.40 7 3.03 3.64 3.27 3.21 3.13 3.40 8 3.53 3.43 3.36 3.46 3.13 3.53 9 3.36 3.24 3.30 3.43 3.00 3.71 10 3.64 3.36 3.00* 3.67 3.60 3.00 11 3.71 3.05 3.79 3.71 3.10 3.56 12 3.08 3.08 3.67 3.43 3.24 3.50 AP < 2
AP is the ratio of the extent of the periodicity in the helical region of the spectrum compared with that over the whole spectrum.
bundle, have greater AP values than core helices, then it is consistent with the analysis that membrane exposed residues are more poorly conserved than buried residues. Consequently, the AP provides a good measure of the surface exposure of the a-helix and is helpful, therefore, in deriving information from sequence data about the three-dimensional structure.
All AP values for each putative helix of the sugar transporter family, whether determined for Family I alone or the entire sugar transporter family, were greater than 2, except for the one noted in Table 3.3. Consequently, it is predicted that each of the regions of sequence analysed, that is, the putative transmembrane domains, demonstrate a significant periodicity that is consistent with a-helical structure. The most probable explanation for the lower AP value for the periodicity of hydrophobic residues in helix 10 is the presence of the highly conserved GPGPIPW motif. It seems possible that this region, perhaps, represents a deviation from the normal helical structure or, a systematic shift in exposed residues due to interactions with adjacent helices.
In order to generate information regarding the faces of helices which may be exposed to the surrounding lipid bilayer, or those faces in direct contact with adjacent helices, the periodicity of patterns of substitution, conservation and hydrophobicity were calculated. Vectors for each residue position were calculated from statistical tables derived from sequence alignments and summed for each putative transmembrane helix analysed. The resulting moments are illustrated on Figures 3.6 and 3.7 for the assessment of Family I alone, and the whole sugar transporter family, respectively.
For a membrane protein possessing a bundle of a-helices, it would be expected that the most hydrophobic residues would be on the outside of the helices. These residues would also represent the least conserved. In contrast, most residues on the inner faces of the helices, that is, in contact with other helices or forming a hydrophilic pore, would not be expected to be hydrophobic, but would show the greatest degree of conservation. Further, it would be expected that the mutational
CD K ) G ÿ M L A L V N I N s ^
Figure 3.6 Patterns in periodicity of Family I.
Moments of substitution (S), conserved (C) and hydrophobic (H) patterns superimposed upon helical wheel plots for the putative transmembrane helices of Family I.
CD
W
G V M