Quantitative analysis of amphiphilic properties of membrane interactive proteins

Abstract

Integral membrane proteins often possess lipophilic a-helical regions of approximately 21 amino acid residues that are able to traverse the bilayer. These may either be amphiphilic or predominantly hydrophobic (such as those found in the photoreaction centres). Such helices are distinguished by low charge densities and large mean hydrophobicities. Several hydrophobicity measures (scales) and algorithms for identification of these segments have been developed.
Here, a survey of currently available hydrophobicity scales and prediction techniques has been conducted and a brief description of each of the techniques is reported. The hydrophobic moment methodology, even after twenty years of existence, appears to be the most widely used technique for amphiphilic helical structure prediction. The reliability of this methodology at predicting structure and function relationship in membrane proteins was tested using proteins and peptides from different but known classes. In predicting structure, the hydrophobic moment appears to make predictions consistent with those from other researchers using other prediction techniques based on other alternative properties. However, the predictions for function were not consistent with the role of proteins in some cases.
Scatterplots of mean hydrophobicity and mean hydrophobic moment for different window sizes revealed that there is a negative association between the two variates and, that mean hydrophobic moment decreases as window size increases. It was also deduced from 99% bootstrap confidence intervals for the correlation coefficients, that short window sizes (7-15) are
more discriminative than tong ones (16-20). Angular frequencies between 95° and 107 0 were also investigated for all window sizes and it was observed that different window lengths have different optimum angular frequencies. Variation in both window size and angular frequency were seen to affect prediction with window sizes 7-15 residues giving better prediction.
A data set of 403 transmembrane segments was assembled. The compositionat and distributional properties of constituent amino acids were investigated. It was observed that transmembrane segments tend to posses high occurrence of charged amino acid residues at the interface, large planar molecules just within the membrane interior and smalt hydrophobic
residues in the central region. When the amino acid residue compositions of the boundaries of transmembrane a-helices were compared statisticatly, it was observed that at the 5% significant level there is no difference between the boundaries within, or between classes of transmembrane spans. It was also observed that the assumed length of twenty-one residues is, on average, reasonable for uncleaved sequences but that twenty-two may be appropriate for stop transfer sequences.
When the presence of a hydrophobicity gradient in transmembrane a-hetices was examined, 25% of the sequences in the data set showed the distribution of hydrophobicity which is observed in typical transmembrane spans (tow-high-low). However, 13% showed the distribution of hydrophobicity which is similar to that seen in tilted peptides. It would appear
that some transmembrane a-helices do have a hydrophobicity gradient, but this has not been related to any biological role. A new measure of amphiphilicity, which takes into account the third dimension of a helix, was also developed and compared to the conventional hydrophobic moment using bootstrap and regression modelling with a categoricat predictor. From the 95% bootstrap confidence intervals for estimates of model parameters, it was concluded though that there is no difference between the two measures, implying that the loss of spatial information in the conventional hydrophobic moment model, is not significant.