Predicting the topology of eukaryotic membrane proteins

Abstract

We show that the so-called 'positive inside' rule, i.e. the observation that positively charged amino acids tend to be more prevalent in cytoplasmic than in extra-cytoplasmic segments in transmembrane proteins [von Heijne, G. (1986) EMBO J. 5, 3021-30271, seems to hold for all polar segments in multi-spanning eukaryotic membrane proteins irrespective of their position in the sequence and hence can be used in conjunction with hydrophobicity analysis to predict their transmembrane topology. Further, as suggested by others, we confirm that the net charge difference across the first transmembrane segment correlates well with its orientation [Hartmann, E., Rapoport, T. A. and Lodish, H. F. (1989) Proc. Natl Acad. Sci. USA 86, 5786-57901, and that the overall amino-acid composition of long polar segments can also be used to predict their cytoplasmic or extra-cytoplasmic location [Nakashima, H. and Nishikawa, K. (1992 We present an approach to the topology prediction problem for eukaryotic membrane proteins based on a combination of these methods.Integral membrane proteins seem to come in two basic varieties: the helical bundle proteins with membrane domains built from a bundle of hydrophobic transmembrane a helices [l, 21, and the P-barrel proteins which form large, anti-parallel / 3 barrels with a hydrophobic outer surface facing the lipids [3, 41. In both cases, the requirement that a continuous outer apolar surface must be formed has necessitated a structurally rather well-defined solution. This greatly facilitates the task of predicting structure from sequence, and the secondary structure of both types of protein can now be fairly well predicted [4, 51. For the helical bundle proteins, two easily identified features of the nascent chain seem to be the major structural determinants: the long, apolar stretches that form the transmembrane a helices and the biased distribution of arginines and lysines in the polar regions exposed on either side of the membrane, the 'positive inside' rule [6, 71. In fact, we have recently shown that transmembrane segments in bacterial inner-membrane proteins can be predicted with very high confidence if this charge bias is taken into account [5].The possibility of similarly improving the prediction of transmembrane helices in eukaryotic membrane proteins has not been explored so far. A priori, there are a number of reason why a charge-bias analysis of the kind shown to work well for prokaryotic proteins may be less informative in this case: in contrast to the situation in prokaryotes, the initial insertion of eukaryotic proteins into the membrane of the endoplasmic reticulum is thought to be largely co-translational, suggesting that only the most N-terminal apolar segment(s) will determine the final topology; the (Arg+Lys) bias between cytoplasmic and extra-cytoplasmic (lumenal) segments is less extreme than in prokaryotes [7] ; it appears that the net charge difference (including negatively charged Asp and Glu residues) across the first transmembrane segment, rather than the differe...

Methods for studying transmembrane peptides in bicelles: consequences of hydrophobic mismatch and peptide sequence

“…It is believed that in this latter case, the tryptophan residue is able to control the association of the peptide with the bicelle. This result is not surprising given previous studies that have shown clustering of aromatic residues at the memebrane-water interface [43][44][45][46]. However, this result does demonstrate the power of one amino acid in dominating lipid-peptide interactions as well as the need for careful consideration when constructing putative transmembrane domains.…”

Topology prediction for helical transmembrane proteins at 86% accuracy–Topology prediction at 86% accuracy

“…Prediction methods were designed to predict the locations of HTMs (von Heijne, 1981(von Heijne, , 1986a(von Heijne, , 1986b(von Heijne, , 1992Argos et al, 1982;Kyte & Doolittle, 1982;Engelman et al, 1986;Cornette et al, 1987;von Heijne & Gavel, 1988;Degli Esposti et al, 1990;von Heijne & Manoil, 1990;Landolt-Marticorena et al, 1992;Donnelly et al, 1993;Edelman, 1993;O'Hara et al, 1993;Sipos & von Heijne, 1993;Jones et al, 1994;Persson & Argos, 1994;Donnelly & Findlay, 1995; and the orientation of HTMs with respect to the cell (dubbed topology, Fig. 1; von Heijne & Gavel, 1988;von Heijne, 1989von Heijne, , 1992Nilsson & von Heijne, 1990;Sipos & von Heijne, 1993;Jones et al, 1994;. If the locations of the HTMs and the topology are known with sufficient accuracy, 3D structure can be predicted successfully for the membrane spanning segments by an exhaustive search of the entire possible structure space .…”

“…In general, current 1D predictions are not accurate enough to provide the demanded precision in locating the helices. (2) A simple and successful technique to predict topology is the positive-inside rule (von Heijne & Gavel, 1988;Hartmann et al, 1989;von Heijne, 1989von Heijne, , 1992Boyd & Beckwith, 1990;Dalbey, 1990;Nilsson & von Heijne, 1990;Sipos & von Heijne, 1993): positively charged residues occur more often in intra-cytoplasmic than in extra-cytoplasmic regions. Applying this rule for the prediction of topology relies crucially on a correct prediction of the nontransmembrane regions.…”

“…We based our analyses on proteins for which experimental information about the locations of HTMs is annotated in the SWISS-PROT database (Manoil & Beckwith, 1986;von Heijne & Gavel, 1988;von Heijne, 1992;Sipos & von Heijne, 1993;Bairoch & Boeckmann, 1994;Jones et al, 1994). The proteins were chosen to meet two criteria: (1) reliability: experimental information should be as reliable as possible (Manoil & Beckwith, 1986;von Heijne, 1992); and (2) comparability: the data set should be similar to those used by others .…”

Role of 26S proteasome and HRD genes in the degradation of 3-hydroxy-3-methylglutaryl-CoA reductase, an integral endoplasmic reticulum membrane protein.

“…Both proteins appeared to have membranespanning domains. The HRD3 gene encoded a protein with four features consistent with Hrd3p being a transmembrane protein oriented with the amino terminus on the lumenal site of a membrane (Sipos and von Heijne, 1993, and references therein, and Doolittle, personal communication). These features included a candidate cleavable hydrophobic N-terminal signal sequence, several asparagine-linked glycosylation consensus sites (NxT/S) in the hydrophilic region following the signal sequence, a hydrophobic region of 22 amino acids with flanking charge density and hydrophobic moment predicted for a transmembrane span, and a relatively small cytoplasmic domain with a stop-transfer sequence immediately adjacent to the transmembrane span.…”