Skip to main content
NCBI home page
Protein Science : A Publication of the Protein Society logoLink to Protein Science : A Publication of the Protein Society
. 2001 Aug;10(8):1529–1538. doi: 10.1110/ps.6301

Modeling of the structural features of integral-membrane proteins reverse-environment prediction of integral membrane protein structure (REPIMPS)

Siavoush Dastmalchi 1, Michael B Morris 1, W Bret Church 2,3
PMCID: PMC2374085  PMID: 11468350

Abstract

The Profiles-3D application, an inverse-folding methodology appropriate for water-soluble proteins, has been modified to allow the determination of structural properties of integral-membrane proteins (IMPs) and for testing the validity of solved and model structures of IMPs. The modification, known as reverse-environment prediction of integral membrane protein structure (REPIMPS), takes into account the fact that exposed areas of side chains for many residues in IMPs are in contact with lipid and not the aqueous phase. This (1) allows lipid-exposed residues to be classified into the correct physicochemical environment class, (2) significantly improves compatibility scores for IMPs whose structures have been solved, and (3) reduces the possibility of rejecting a three-dimensional structure for an IMP because the presence of lipid was not included. Validation tests of REPIMPS showed that it (1) can locate the transmembrane domain of IMPs with single transmembrane helices more frequently than a range of other methodologies, (2) can rotationally orient transmembrane helices with respect to the lipid environment and surrounding helices in IMPs with multiple transmembrane helices, and (3) has the potential to accurately locate transmembrane domains in IMPs with multiple transmembrane helices. We conclude that correcting for the presence of the lipid environment surrounding the transmembrane segments of IMPs is an essential step for reasonable modeling and verification of the three-dimensional structures of these proteins.

Keywords: Membrane proteins, transmembrane helices, fold recognition, molecular modeling, protein structure, structure prediction, GPCRs

Integral-membrane proteins (IMPs) fulfill a variety of important cellular functions and make up a large fraction of all proteins. It is estimated that 35–40% of all yeast genes and 20% of all human genes code for IMPs (Goffeau et al. 1993; Jones 1998), reflecting the fact that they play critical roles in maintaining the homeostasis and responsiveness of cells, organs, and organisms. Given the rapid advance of large-scale gene sequencing projects, complete protein sequences for many key organisms are known or will be known in the near future (Kyrpides 1999). In consequence, the gap between the number of sequenced proteins and those for which experimentally determined three-dimensional structures are known is expanding despite continued improvements in the power and speed of biophysical methods such as electron microscopy (Stowell et al. 1998), X-ray crystallography (Abrahams and De Graaff 1998; Brunger et al. 1998; Stoddard 1998), multidimensional nuclear magnetic resonance (NMR) (Case 1998; Dotsch and Wagner 1998; Marassi and Opella 1998), and site-directed spin labeling (Hubbell et al. 1998).

Comparative modeling of the three-dimensional structures of proteins based on sequence similarity can now be applied with reasonable accuracy to ten times more protein sequences than the number of experimentally determined protein structures (Sanchez and Sali 1997). Although this approach becomes unreliable when the sequence identity drops below 20–30% (Sander and Schneider 1991), it has been estimated that more than 25% of all sequences in the SWISS-PROT sequence database can be modeled by homology (Rost et al. 1995). The success of such modeling aside, knowledge about the structure of IMPs remains sparse because so few high-resolution three-dimensional structures are available (Table 1). This lack of experimental data means that it is not usually possible to use sequence homology methods to build accurate three-dimensional models of IMPs.

Table 1.

Representative IMPs whose three-dimensional structures have been solved to better than 3.6 Å

Proteins PDB code (resolution, Å) Reference
Bacteriorhodopsin 1AP9 (2.5) Pebay-peyroula et al. 1997
1AT9 (3.0) Kimura et al. 1997
1BM1 (3.5) Takeda et al. 1998
1BRD (3.5) Henderson et al. 1990
1BRR (2.9) Schertler et al. 1993
1BRX (2.3) Luecke et al. 1998
2AT9 (3.0) Mitsuoka et al. 1999
2BRD (3.5) Ceska and Henderson 1990
Rhodopsin (bovine) 1F88 (2.8) Palczewski et al. 2000
Cytochrome c oxidase 1OCC (2.8) Tsukihara et al. 1996
2OCC (2.3) Yoshikawa et al. 1998
1AR1 (2.8) Iwata et al. 1995
Cytochrome bc1 complex 1BCC (3.16) Zhang et al. 1998
Photosynthetic reaction centre 1PRC (2.3) Deisenhofer et al. 1995
1PSS (3.0) Chirino et al. 1994
1AIG (2.6) Stowell et al. 1997
1AIJ (2.2) Stowell et al. 1997
1PCR (2.6) Ermler et al. 1994
Light-harvesting complex 1KZU (2.5) McDermott et al. 1995
1LGH (2.4) Koepke et al. 1996
Porin 1PRN (1.96) Weiss and Schulz 1992; Kreusch et al. 1994
Outer membrane phospholipase A 1QD6 (2.1) Snijder et al. 1999
H+ gated potassium channel 1BL8 (3.2) Doyle et al. 1998
Mechanosensitive ion channel 1MSL (3.5) Chang et al. 1998
Open in a new tab

A more comprehensive and continually updated table is available at http:// blanco.biomol.uci.edu/Membrane_Proteins_xtal.html (S. White, pers. comm.).

A number of inverse-folding methodologies have been developed for water-soluble proteins which, rather than relying on sequence alignments, attempt to predict three-dimensional structures of proteins on the basis of physicochemical parameters (Sippl 1990; Bowie et al. 1991; Jones et al. 1992; Bowie and Eisenberg 1993; Bryant and Lawrence 1993; Godzik et al. 1993; Godzik 1995; Sippl 1995; Hu et al. 1997). The methodologies use potential functions frequently involving terms for pairwise amino acid interaction and solvent exposure. The forms of the potential functions and the weighting of the terms involved differ from method to method. Recently incorporated aspects in fold recognition include multiple sequence alignment and secondary structure predictions (Jones et al. 1999; Koretke et al. 1999; Murzin 1999).

In the Profiles-3D inverse-folding method (Biosym/Molecular Simulations, Profiles-3D 95.0), the physicochemical environments of residues in the three-dimensional structures of proteins are quantified in terms of (1) the area of a residue's side chain buried away from the aqueous phase, (2) the fraction of side-chain area in contact with a polar environment (polar atoms from the aqueous phase and from other residues), and (3) local secondary structure (Bowie et al. 1991; Luthy et al. 1992). Based on these criteria, the probability of finding specific residues in a particular class of environment can be estimated by analyzing the well-defined structures of a set of water-soluble proteins. These probabilities are used to generate a 3D→1D scoring table which can be used to link sequences and three-dimensional structures together; that is, possible structures for a protein can be generated following the conversion of the string of residues to a string of environment classes. In addition to finding compatible structures for a given sequence, the alignment of an amino acid sequence with a 3D profile can be used for other purposes such as testing the validity of a preliminary or model structure, and finding sequences compatible with a defined structure.

We considered that such a general quantitative description of environment preference could be adapted to predict structural features of IMPs. Despite the difference in polarity of their surrounding environments, it has been suggested that IMPs and water-soluble proteins are variations on common structural themes, differing primarily in the polarity of the residues on the protein surface (Rees et al. 1989). Our aim was to modify the Profiles-3D approach so that it would take into account the fact that certain residues in IMPs are exposed to a lipid environment rather than an aqueous environment. By making this correction, it should be possible to predict structural features of IMPs and to assess the validity of both experimentally determined and modeled structures of IMPs.

Results

Compatibility scores of IMPs improve when the lipid environment is included

Table 2 shows the total of the compatibility scores for ten representative IMPs whose structures are known to high resolution. When using the Profiles-3D program alone, the assumption is made that all residues on the surface of the protein are exposed to an aqueous environment. This gives rise to totals of the compatibility scores which are less than the scores expected for these proteins based on their sequence lengths. In fact, for bacteriorhodopsin, the value of the total score is close to the lowest acceptable score. A score below this indicates that the structure is incorrect. In this case, however, the low Profiles-3D score probably does not indicate an incorrect structure but rather indicates that 56% of the surface area of bacteriorhodopsin is exposed to a lipid rather than an aqueous environment.

Table 2.

Total of the compatibility scores calculated for ten structurally known IMPs using the Profiles-3D program and correcting the scores on the basis that particular residues are exposed to a lipid-based rather than an aqueous environment

Profiles-3D scorea
Calculation method Self compatibility score Expected scoreb Lowest acceptable scorec Lipid- corrected scored
Cytochrome c oxidase (1OCC)e 799 822 370 901
Cytochrome bc1 complex (1BCC)e 775 924 415 875
Bacteriorhodopsin (2BRD)e 47 101 45 116
Halorhodopsin (1E12)e 67 108 48 128
Bovine rhodopsin (1F88)e 105 154 69 168
Potassium, H+ gated channel (1BL8)e 137 177 80 201
Mechanosensitive ion channel (1MSL)e 63 249 112 143
Photosynthetic reaction center (1PRC)e 415 547 246 526
Porin (1PRN)e 97 132 59 165
Outer membrane phospholipase A (1QD6)e 168 231 104 247
Open in a new tab

The Profiles-3D program assumes that the protein is exposed only to an aqueous environment and generates aa total compatibility score for that protein based on its three-dimensional structure, ba score expected for a correct structure of sequence length L based on a set of structures solved to better than 2 Å resolution (Luthy et al. 1992), and ca lowest acceptable score which is 0.45 times the expected score. If the Profiles-3D score is lower than the lowest acceptable score, this indicates that the three-dimensional structure of the protein is incorrect or, alternatively, that it has not been placed in the correct environment. dThis score has been generated on the basis that some of the residues in these IMPs are exposed to a lipid rather than an aqueous environment and the compatibility score for each of these residues has been corrected based on the approach outlined in Materials and Methods. ePDB accession number.

We have recalculated the totals of the compatibility scores for these proteins after first estimating which residues are in contact with lipid in the membrane and assigning the correct environmental class for lipid-exposed residues using Equation 1 and A*b (see Materials and Methods for details). On this basis, the totals of the compatibility scores for these proteins are now similar to or exceed the expected scores in all cases (Table 2). As expected, the biggest percentage correction was for bacteriorhodopsin, whose score improved ∼150%. Other than for halorhodopsin and mechanosensitive ion channel, the relative improvements are smaller for the other proteins such as cytochrome c oxidase and the photosynthetic reaction center because a smaller percentage of residues are in contact with lipid.

Grigorieff et al. (1996) reported the presence of ten lipid molecules (phosphoric acid 2,3-bis-(3,7,11,15-tetramethyl-hexadecyloxy)-propyl ester 2-hydroxo-3-phosphonoxy-propyl ester) associated with bacteriorhodopsin. The lipid molecules are mainly in contact with helices I, II, IV, VI, and VII. In contrast, helices II and V are not covered or are only partially covered by the lipid molecules. The compatibility scores for the residues in contact with the lipid molecules decreased after eliminating the lipid molecules, whereas the scores for residues which were not covered by the lipid molecules remained the same (Fig. 1 ▶). This result also highlights the need to consider the presence of a lipid environment when calculating compatibility scores.

Fig. 1.

Fig. 1.

Open in a new tab

Compatibility scores calculated by Profiles-3D plotted against the residue number for bacteriorhodopsin in the presence (solid line) and absence (dashed line) of ten lipid molecules (phosphoric acid 2,3-bis-(3,7,11,15-tetramethyl-hexadecyloxy)-propyl ester 2-hydroxo-3-phosphonoxy-propyl ester) associated with the protein (Grigorieff et al. 1996). The lipid molecules are mainly in contact with helices I, III, IV, VI, and VII, whereas helices II and V are not covered or are only partially covered. The lines represent scores smoothed by a sliding 21-point window. The seven solid bars represent the locations of the helices as described in the Materials and Methods section.

For each transmembrane portion of the α-helices of bacteriorhodopsin, photosynthetic reaction center, and cytochrome c oxidase, we calculated the areas of the side chains exposed to lipid and averaged this area per residue. We then calculated the compatibility score for each of these transmembrane domains assuming that the protein had been inserted into a membrane, subtracted the score calculated assuming an aqueous environment, and divided the difference by the number of residues involved. The final value is a measure of the improvement in the compatibility score for the transmembrane domain compared to what would normally be calculated using Profiles-3D. Figure 2 ▶ shows the correlation between this improved score and the average area of side chain exposed to lipid. As expected, those transmembrane domains which had the greatest area per residue exposed to the lipid had the largest improvement in their compatibility score per residue.

Fig. 2.

Fig. 2.

Open in a new tab

Correlation between the average improvement of the compatibility scores per residue after correcting for the presence of a lipid environment and the average area of lipid exposure per residue for helical transmembrane segments of bacteriorhodopsin, photosynthetic reaction center, and cytochrome c oxidase. The transmembrane portion of each helix was determined as outlined in Materials and Methods. For each helix, the areas of the side chains exposed to lipid were calculated, and this area was averaged per residue. The compatibility score for each of these helices was then calculated assuming a membrane environment, and from this was subtracted the value calculated assuming an aqueous environment. The final value was then divided by the number of residues involved and is a measure of the improvement in the compatibility score for the transmembrane domain.

Rotating individual helices in IMPs

We tested whether our method can discriminate between different three-dimensional structures for the same IMP. This was done by making changes to the native structure and then recalculating the total of the compatibility scores. We used two different approaches: in the first, we used the structure for bacteriorhodopsin that contains a retinal molecule bound to helix seven at residue Lys216. Six of the seven transmembrane helices were fixed, and the remaining helix was rotated around its long axis at 36 10° intervals. At each interval, the total of the compatibility scores of the rotated helix was calculated and corrected for the fact that many of the residues were exposed to a lipid rather than an aqueous environment. Figure 3A ▶ shows that the total of the compatibility scores for helix VII of bacteriorhodopsin was maximal near the beginning and end of the rotation process (i.e., near 0° and 360°), which corresponds to the native position of the helix in the 3D structure. The total generally falls in the middle of the rotational range, indicating that residues were being placed in incompatible environments. Similar trends were observed for helices I–V of this protein. Helix VI, however, was an exception with the plot showing a broad maximum centered at 130° (data not shown) and the value at 0° being among the lowest calculated. The results for all helices were similar in both the presence and absence of retinal.

Fig. 3.

Fig. 3.

Open in a new tab

(A) Values of helix compatibility scores versus rotation for helix VII of bacteriorhodopsin. The long axis of the helix was located and the amide bonds at the boundaries of the helix cleaved. Fixed rotations of 10° around the long axis were then performed through the InsightII command line. After each rotation, the amide bonds were reformed and the total of the compatibility scores for the helix was recalculated. Positions at 0° and 360° represent the native position of the helix. (B) The helix compatibility scores for Helix II of bacteriorhodopsin versus the helix rotational shift into and out of the membrane. Each rotation step represents the helix being screwed into or out of the membrane one residue at a time; that is, the translation of the helix by a distance corresponding to the vertical distance between two consecutive Cα in the helix and a 100° rotation around the helix axis (see Materials and Methods for details). Positive rotation steps represent shifts into the membrane and negative steps represent shifts out of the membrane as viewed from the extracellular side of the membrane.

In the second test, individual helices of bacteriorhodopsin were advanced into or out of the membrane one residue at a time by screwing the long axis of the helix. The process used effectively achieved rotation steps of ∼100°, resulting in the loss of one residue from one end of the transmembrane segment, and the addition of a new residue to the other end of the segment. Figure 3B ▶ shows that the total of the compatibility scores for helix II of bacteriorhodopsin drops progressively from its maximum value obtained with the helix in the native position. This was also the case for helices I–V, and VII, both in the absence and presence of retinal. Again, helix VI did not exhibit the same behavior as the other helices; it showed no obvious trend with rotation.

Validating the location of transmembrane domains using REPIMPS

We tested whether our method could accurately locate the transmembrane domain of IMPs containing a single helix crossing the membrane. First, a preliminary set of 15 proteins was selected from the SWISS-PROT and Tr EMBL protein sequence databanks, each protein believed to contain a single α-helical TM domain. The polypeptide for each selected protein was folded into a single ideal helix. Compatibility scores for each residue based on a lipid environment were calculated, and the resulting values were plotted against the residue number for each of the 15 proteins. Figure 4 ▶ shows the result for HLA class I histocompatibility antigen (SWISS-PROT code 30443). The compatibility scores for residues 308–332 are all strongly positive, indicating that this region defines the transmembrane domain. The boundaries of the domain are sharply defined, because the compatibility scores of consecutive residues switched from being negative immediately outside the transmembrane domain to positive within the domain. Other methods (Table 3) of estimating the location of transmembrane domains also place the transmembrane domain of HLA class I histocompatibility antigen in this region, and the same region is reported in SWISS-PROT (Bairoch and Apweiler 2000).

Fig. 4.

Fig. 4.

Open in a new tab

Predicting the location of the α-helical transmembrane segment of HLA class I histocompatibility antigen. The polypeptide was folded into a single ideal helix as outlined in Materials and Methods. A compatibility score for each residue was then calculated based on those residues being in a lipid environment. The bar covers those residues (308–332) predicted to form the transmembrane domain.

Table 3.

Calculation of the compatibility scores (CS) for residues 8–20 (helix I) of bacteriorhodopsin. Scores were calculated using Profiles-3D (Water CS) and assuming the entire protein was placed in a lipid environment (Lipid CS)

Residue number and name SSa Ab (Å)b Fb Envb Water CSb Ab* (Å)c F*c Envc Lipid CSc Final CSd
8 PRO C 72.22 0.67 P2 0.44 123 0.26 B1 0.59 0.44
9 GLU H 132.21 0.44 B2 −0.58 151 0.31 B1 −2.15 −0.58
10 TRP H 71.18 0.79 P2 −1.09 234 0.10 B1 1.11 1.11
11 ILE H 63.99 0.66 P2 −0.59 157 0.07 B1 1.11 1.11
12 TRP H 174.66 0.37 B2 1.01 234 0.12 B1 1.11 1.11
13 LEU H 149.72 0.26 B1 1.3 154 0.24 B1 1.30 1.30
14 ALA H 36.25 0.60 E 0.44 71 0.11 P1 0.76 0.76
15 LEU H 69.00 0.58 P2 −0.46 154 0.03 B1 1.30 1.30
16 GLY H 40.00 0.11 E 0.63 40 0.00 E 0.63 0.63
17 THR H 96.96 0.35 P1 0.39 113 0.21 P1 0.39 0.39
18 ALA H 10.06 0.89 E 0.44 71 0.03 P1 0.76 0.76
19 LEU H 83.97 0.49 P1 −0.3 154 0.03 B1 1.30 1.30
20 MET H 169.86 0.31 B1 1.26 172 0.30 B1 1.26 1.26
Open in a new tab

a SS = Secondary structure, C = coil, H = helix. Classification of secondary structure was determined by the Kabsch-Sander method from the X-ray structure.

b Values calculated using Profiles-3D assuming the protein was exposed to an aqueous environment. Ab = area of side chain buried away from water; F = fractional area of side chain in contact with polar atoms; Env = environment (see Materials and Methods).

c Parameters are analogous to b but calculated assuming the protein was exposed to a lipid environment. Ab* is equivalent to the total accessible area of the side chain; F* has been calculated using Equation 1.

d Water CS values are used for residues 8 and 9. Lipid CS values are used for residues 10–20.

The predictive power of our method was examined by locating the transmembrane segments for the set of 15 proteins and was compared with that of five other prediction methods (Table 4). For 14 of the 15 proteins, REPIMPS selected the SWISS-PROT deposited location of the transmembrane domain and was superior to the other five methods. For one of the proteins (neurogenic locus delta protein precursor, SWISS-PROT accession number P10041), REPIMPS selected a sequence for the transmembrane domain (residues 595–619) different from that listed in the SWISS-PROT database (residues 654–677). However, all of the other five methods used selected a region similar to that selected by REPIMPS as being transmembrane, and none predicted the location stated in the database. For two other proteins, the Ser/Thr-protein kinase IRE1 precursor (SWISS-PROT accession number P32361) and a 40.1 kD protein encoded by the HMC operon of Desulfovibrio vulgaris (SWISS-PROT accession number P33389), REPIMPS selected the correct position of the transmembrane domain plus one and two additional regions, respectively, as transmembrane domains.

Table 4.

Prediction of the location of the single α-helical transmembrane region of a set of 15 IMPs using a suite of methodsa,b

REPIMPS SOSUI SPLIT TMHMM TMpred TopPred
Total number of predicted TM segments (T) 18 23 24 15 37 31
Number of correctly located TM segments (C) 14 12 14 12 14 14
Number of the proteins predicted as water-soluble protein (S) 0 1 0 1 0 0
Prediction index: C/(T + S) 0.78 0.50 0.58 0.75 0.38 0.45
Open in a new tab

a SWISS-PROT accession numbers are O42204, O75503, P04195, P07359, P08195, P10041, P12555, P14585, P23654, P30443, P31789, P32361, P33389, P33767, and P34891.

b Web locations of these programs are listed in Materials and Methods.

We used REPIMPS to predict the location of the transmembrane segments of the 7-transmembrane protein, bacteriorhodopsin. A single ideal helix was built from the sequence of bacteriorhodopsin, and the compatibility scores were calculated based on the presence of a lipid environment. Fourier analysis was performed on the resulting string of compatibility scores to detect the existence of a periodicity of 3.6 residues. The analysis was carried out using a sliding window of different sizes (14–18) scanning through the sequence of compatibility values. The magnitude of the Fourier power spectrum at 100° (corresponding to the periodicity of an α-helix) at each starting position of the sliding window was calculated and plotted against the starting position (Fig. 5 ▶). A line of moving average per eight data points gave rise to seven peaks corresponding to the positions of the seven transmembrane segments of bacteriorhodopsin (Fig. 5 ▶).

Fig. 5.

Fig. 5.

Open in a new tab

Predicting the location of the seven transmembrane domains in bacteriorhodopsin. The polypeptide was folded into a single ideal helix (as outlined in Materials and Methods), and the compatibility score for each residue was then calculated based on those residues being in a lipid environment. A Fourier analysis was performed on the string of compatibility scores using a Fortran program modified from the original code of Cornette (Cornette et al. 1987). The program calculated the power spectrum of the Fourier analysis at 100° (corresponding to the periodicity of α-helix) using a sliding window of 17 residues. The power spectrum values were smoothed using the eight-point moving average function in Excel (solid curve). The thick portion of the solid bars represents the location and size of the transmembrane domain portions of the seven α-helices, as determined by the process outlined in Materials and Methods. The thick portion of the bars and the thin lines extending from them represent the full length of each of the seven α-helical regions as calculated based on Kabsch-Sander's algorithm.

Discussion

We have adapted the Profiles-3D application, an inverse-folding methodology appropriate for water-soluble proteins, to develop and test the idea that it might be useful for determining structural properties of IMPs. The results using REPIMPS show that it is possible to use the same scoring system and set of environment classes for both IMPs and water-soluble proteins. The only correction required is to ensure that residues in contact with lipid are assigned to the correct class through the use of Equation 1. By taking into consideration the fact that certain residues in IMPs are exposed to a lipid environment, the total of the compatibility scores for a series of IMPs whose three-dimensional structures are known universally improved such that they were now close to or exceeded the expected scores (Table 2; Fig. 1 ▶). In addition, the greater the average area of exposure of transmembrane residues to the lipid environment, the greater the improvement in the compatibility score (Fig. 2 ▶). Thus, correcting for the presence of the lipid environment surrounding transmembrane segments is an essential step for reasonable modeling and verification of the three-dimensional structures of IMPs. It should be possible to refine this approach by producing a scoring table specifically for the transmembrane segments based on the currently available three-dimensional structures of IMPs. The table itself could be refined as further high-resolution structures of IMPs become available.

The relative orientation and depth of the helical transmembrane segments are important structural features of IMPs. Usually the decision on determining the interior-facing side or lipid-facing side of helical transmembrane segments is made by the analysis of hydrophobicity moments or conservation patterns in these regions. In general, the hydrophobic and less-conserved side is more likely to face toward the lipid bilayer (Baldwin et al. 1997). However, sometimes a clearly more hydrophobic side of a transmembrane helix is not apparent or there are no homologous proteins to determine the conserved face of the helix.

The suitability of our method for determining the relative orientation of TM segments can be inferred by the results shown in Figure 3 ▶. Although the biologically relevant form of bacteriorhodopsin is the trimer (Grigorieff et al. 1996; Muller et al. 1997; Moller et al. 2000), we have performed calculations on monomeric bacteriorhodopsin on the basis that contacts between monomeric subunits in the trimer are predominantly hydrophobic (e.g., see PDB accession number 1AP9). Our method predicted the correct orientation of six out of seven helical transmembrane segments of bacteriorhodopsin; that is, the highest total of the compatibility scores belonged to the helix in its native orientation, while deviation from the native position reduced the total. Helix VI proved to be the exception, indicating either that our methodology needs to be improved or that some functional significance is associated with this helix. Subramaniam et al. (1999) suggested that the movements of four helices (I, II, VI, and VII) are involved in bacteriorhodopsin's proton pumping. For example, helix VII moves in the transition between states BR and M resulting from an isomerization of retinal (Luecke et al. 1999), and it has been suggested that at least two residues, Val177 and Trp182, move in helix VI. Luecke et al. (1999) also suggested a refinement of the description of a tilt in helix VI, with Tyr185 and Pro186 acting as a hinge during proton pumping. In addition, helix VI may play a role in other states of the transport cycle, including state O. Our method cannot describe precise conformational changes, but it may be useful in identifying regions in IMPs that have functional consequences.

We also used REPIMPS to predict the TM segments of IMPs which span the bilayer just once. Figure 4 ▶ shows the TM region of the HLA histocompatibility antigen predicted by our method, which is in agreement with the reported results (Bairoch and Apweiler 2000). In a further test, the accuracy of predicting TM segments by our method for a set of 15 such proteins was compared to that of other methods. The results in Table 4 strongly support the capability of the method to predict the location of helical TM segments with greater accuracy than a series of other methods. The only `incorrect' predictions made by REPIMPS were as follows: For the neurogenic locus delta protein precursor (SWISS-PROT accession number P10041), the single transmembrane domain selected was shifted ∼60 residues to the N-terminal side of the location of the domain, as deposited in SWISS-PROT. However, all of the other five prediction methods used also located the transmembrane domain to the same position as REPIMPS, and none predicted the location listed in SWISS-PROT. For the 40.1 kD protein encoded by the HMC operon of D. vulgaris (SWISS-PROT accession number P33389), REPIMPS selected two additional transmembrane segments in addition to the correct segment. The location of one of the `incorrect' segments (residues 229–247) was also predicted by TopPred, whereas the location of the second `incorrect' segment (residues 6–33) was predicted by three of the five other methods used. For the Ser/Thr-protein kinase IRE1 precursor (SWISS-PROT accession number P32361), REPIMPS selected one additional transmembrane segment (residues 8–26) in addition to the correct segment (residues 531–555). However, all of the other methods predicted the presence of the additional transmembrane segment.

In a final validation of the REPIMPS method, we tested its ability to successfully predict the location of the seven transmembrane segments for bacteriorhodopsin (Fig. 5 ▶). This involved transforming the structure of bacteriorhodopsin into an ideal helix and calculating the compatibility scores for the residues assuming the whole structure was placed in a lipid environment. Under these circumstances, those residues that would normally be extramembrane and therefore exposed to the aqueous environment should receive low compatibility scores. Within the α-helical transmembrane domains of the native structure, there should be a periodicity in the residues facing the lipid environment based on the natural periodicity of the α-helix of 3.6 residues/turn, and those residues would be expected to receive a high compatibility score. Using this approach (Fig. 5 ▶), we observed seven maxima, with each maximum located close to the center of a transmembrane domain of bacteriorhodopsin.

In summary, we have modified an environment-based inverse-folding method originally developed for water-soluble proteins. This approach has the potential to assess the validity of experimentally determined IMP structures and model structures and to predict structural features of IMPs of unknown structure. We are in the process of optimizing the parameters for our method and extending the work presented here to test other structural parameters such as packing. For example, it may be possible to develop other potential terms based on the success of other inverse-folding methods used for water-soluble proteins. In addition, we are in the process of refining our method for assessing the reliability of models of IMP structures by encompassing a description of the membrane environment. Methodologies based on inverse-folding potentials promise advantages over sequence-alignment methods. Because only a small number of existing structural templates for IMPs exist, we hope to examine the generality of the templates themselves, with the goal of reducing reliance on multiple sequence alignment and secondary structure in predicting the three-dimensional structures of IMPs.

Materials and methods

Databases and programs

All structure coordinates for the proteins used in this study were retrieved from the Protein Data Bank (Berman et al. 2000). Sequences and their annotations were from the SWISS-PROT and Tr EMBL protein sequence database (Bairoch and Apweiler 2000). Computation was carried out on a Silicon Graphics workstation (Power Indigo2, R10000, 195 MHz processor). The Profiles-3D application was used within the InsightII molecular modeling package (v95.5, Molecular Simulations, San Diego).

The Profiles-3D software

For each residue in a protein of known three-dimensional structure, the Profiles-3D software will calculate (1) the area of each residue's side chain that is buried away from the aqueous phase (Bowie et al. 1991), (2) the fraction of the area of each residue's side chain that is in contact with polar atoms (either from the solvent or other atoms in the protein), and (3) the secondary structure (α-helix, β-strand, and other) for the residue based on the KABSCH-SANDER algorithm (Kabsch and Sander 1983). Based on this information, each residue is assigned to one of 18 environment classes. A profile table can then be constructed in which each position in the sequence is assigned an environment class and a compatibility score for the residue filling the position. The compatibility scores are derived from a basic set of 16 water-soluble proteins whose three-dimensional structures are known to high resolution and which represent a variety of protein folds. The total of the compatibility scores, taking into account the size of the protein, can be used to assess the validity of the structure.

Correcting compatibility scores for residue side-chains in a lipid environment

For IMPs whose three-dimensional structures are known to high resolution (Table 1), a significant proportion of the residues would normally be in contact with lipids of the membrane rather than the aqueous environment. Thus, a correction to values of the areas of a side chain buried away from the aqueous phase and in contact with polar atoms is required for those residues within the membrane. The equation used to correct the fractional area of side chain in contact with polar atoms, F*, is:

graphic file with name M1.gif 1

where F is the uncorrected fractional area of side chain in contact with polar atoms (i.e., calculated by the Profiles-3D assuming that any exposed area of side chain faces an aqueous environment). Ab (Å2), also calculated by the Profiles-3D program, is the area of side chain buried away from the aqueous environment and is defined as the solvent-accessible area of the side chain (At) in a Gly-X-Gly tripeptide minus the solvent-accessible area of the side chain in the protein. Values of At include the α carbon atom (Bowie et al. 1991). For those residues within the membrane, the total area of side chain buried away from the aqueous environment (A*b) was considered to be equal to At.

By using F*, A*b, and the local secondary structure of each residue located within the membrane of an IMP, it is possible to assign the appropriate environment class for each residue from the set of 18 environment classes. In this way, a residue exposed to lipid will be assigned differently and will have a new compatibility score, whereas a residue exposed to the aqueous phase will retain the original class and compatibility score normally assigned by Profiles-3D.

Table 2 shows those proteins for which corrected compatibility scores were obtained. Each of these proteins (except for porin and outer membrane phospholipase A) contains helical transmembrane domains. An empirical approach was used to identify the boundary between those residues in the polypeptide sequences exposed to the aqueous phase and those buried in the membrane. This was done by first calculating (using Profiles-3D) the compatibility scores for each residue in an IMP assuming that the IMP was exposed only to water, and then recalculating (using an in-house Fortran program) the values using Equation 1 assuming the protein was entirely within the lipid membrane. For the proteins with helical transmembrane domains, the point within each helix at which a marked improvement in the compatibility score was observed for the lipid-based value over the water-based value was taken to be the boundary (Table 3). This approach accommodates the situation where a helical region may extend beyond the membrane into the aqueous environment. In ∼75% of cases, this boundary was clear. In those cases where the boundary was not as clearly defined by this approach, one turn of helix was considered to lie outside of the lipid membrane. A similar approach was used to locate the boundaries of porin. In those cases where the boundary was not clear, it was assumed that charged or polar residues (Ser, Thr, Tyr, Asn, Gln) represented the location of the boundary and that residues at the top of the β turns were located outside of the membrane. It was assumed that water molecules filled the pore of the protein.

Once the boundaries were located, the total of the compatibility scores for an IMP structure was calculated using the lipid-based values for those residues located in the lipid bilayer and water-based values for those residues lying outside the bilayer (Table 3).

Rotation of individual helices within bacteriorhodopsin

Bacteriorhodopsin was selected because it contains seven close-to-ideal helices approximately perpendicular to the membrane. The crystallized protein appears to retain both native conformation and activity (Portmann et al. 1991; Landau and Luisi 1993; Hochkoeppler et al. 1995). The rotation of individual helices was performed in two ways: (1) An individual helix was selected, the long axis of the helix located, and the amide bonds at the boundaries of the helix cleaved. Fixed rotations of 10° around the long axis were then performed through the InsightII command line. After each rotation, the amide bonds were reformed, and the total of the compatibility scores for the chosen helix was recalculated. (2) An individual helix was selected and effectively screwed into or out of the membrane one residue at a time. In order to simulate this type of rotation, we used the mutation tool of the Swiss-PdbViewer program (v3.5b1). The backbone coordinates of the selected helix remained the same but the first residue of the helix within the lipid environment was replaced by the previous residue. In the same way, the second residue was replaced by the first residue. This residue replacement procedure was continued up to the end of the membrane region of the helix and left the last residue, previously lying within the membrane, now lying outside of it. This in fact simulates the translation of the helix into the membrane by a distance corresponding to the vertical distance between two consecutive Cα in helix and a 100° rotation around the helix axis. The mutation tool in Swiss-PdbViewer automatically selects the most favorable rotamer to minimize steric clashes while increasing the number of hydrogen and disulfide bonds (Ponder and Richards 1987). Any remaining steric clashes were eliminated later either manually or using the "Fix selected side chains" option from the Tools menu.

Detecting the transmembrane region of IMPs with a single transmembrane domain

The sequences of a random selection of 15 IMPs each believed to contain a single α-helical transmembrane domain were downloaded from SWISS-PROT and Tr EMBL. For this set, the highest pairwise alignment score was 19%. The sequence for each selected protein was folded into a single ideal helix using the Swiss-PdbViewer program (v3.5b1). The α-helical structures were built from the sequences in FASTA format (Guex and Peitsch 1997). Side chains were placed automatically by the program in most favorable rotamer (Ponder and Richards 1987). A compatibility score for each residue in each protein was then calculated based on all residues being in a lipid environment.

Several other prediction methods were used to detect the location of transmembrane regions of the set of 15 single α-helix-containing IMPs listed in Table 4. These programs were run from their web sites with the various options set at default values. The names and the web addresses of these programs are as follows: SOSUI (Hirokawa et al. 1998), http://sosui.proteome.bio.tuat.ac.jp/sosuiframe0E.html.; SPLIT35 (Juretic and Lucin 1998), http://pref.etfos.hr/split/; Tmpred, http://www.ch.embnet.org/software/TMPRED_form.html; TMHMM (v. 0.1) (Sonnhammer et al. 1998), http://www.cbs.dtu.dk/services/TMHMM-1.0/; and TopPred2 (von Heijne 1992), http://www.sbc.su.se/∼erikw/toppred2/.

Predicting the location of the seven transmembrane domains in bacteriorhodopsin

The sequence of bacteriorhodopsin was folded into a single ideal helix using the Swiss-PdbViewer program, as described above, and the compatibility score for each residue was then calculated based on all residues being in a lipid environment. A Fourier analysis was performed on the string of compatibility scores using a Fortran program modified from the original code of Cornette (Cornette et al. 1987). The program calculated the power spectrum of the Fourier analysis using a sliding window of 14–18 residues.

Acknowledgments

This work was supported by a scholarship from the Iranian Ministry of Health and Medical Education to S. Dastmalchi and Australian Research Council and Sydney University Sesqui R and D grants to M.B. Morris.

The publication costs of this article were defrayed in part by payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 USC section 1734 solely to indicate this fact.

Article and publication are at http://www.proteinscience.org/cgi/doi/10.1101/ps.6301

References

  1. Abrahams, J.P. and De Graaff, R. 1998. New developments in phase refinement. Curr. Opin. Struct. Biol. 8 601–605. [DOI] [PubMed] [Google Scholar]
  2. Bairoch, A. and Apweiler, R. 2000. The SWISS-PROT protein sequence database and its supplement TrEMBL in 2000. Nucleic Acids Res. 28 45–48. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Baldwin, J.M., Schertler, G.F.X., and Unger, V.M. 1997. An alpha-carbon template for the transmembrane helices in the rhodopsin family of G-protein-coupled receptors. J. Mol. Biol. 272 144–164. [DOI] [PubMed] [Google Scholar]
  4. Berman, H.M., Westbrook, J., Feng, Z., Gililand, G., Bahat, T.N., Weissig, H., and Shindyalov, P.E. 2000. The protein data bank. Nucleic Acids Res. 28 235–242. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Bowie, J.U. and Eisenberg, D. 1993. Inverted protein structure prediction. Curr. Opin. Struct. Biol. 3 437–444. [Google Scholar]
  6. Bowie, J.U., Luthy, R., and Eisenberg, D. 1991. A method to identify protein sequences that fold into a known three-dimensional structure. Science 253 164–170. [DOI] [PubMed] [Google Scholar]
  7. Brunger, A.T., Adams, P.D., and Rice, M.L. 1998. Recent developments for the efficient crystallographic refinement of macromolecular structures. Curr. Opin. Struct. Biol. 8 606–611. [DOI] [PubMed] [Google Scholar]
  8. Bryant, S.H. and Lawrence, C.E. 1993. An empirical energy function for threading protein sequence through the folding motif. Proteins 16 92–112. [DOI] [PubMed] [Google Scholar]
  9. Case, D.A. 1998. The use of chemical shifts and their anisotropies in bimolecular structure determination. Curr. Opin. Struct. Biol. 8 624–630. [DOI] [PubMed] [Google Scholar]
  10. Ceska, T.A. and Henderson, R. 1990. Analysis of high-resolution electron diffraction patterns from purple membrane labeled with heavy-atoms. J. Mol. Biol. 213 539–560. [DOI] [PubMed] [Google Scholar]
  11. Chang, G., Spencer, R.H., Lee, A.T., Barclay, M.T., and Rees, D.C. 1998. Structure of the MscL homolog from Mycobacterium tuberculosis: A gated mechanosensitive ion channel. Science 282 2220–2226. [DOI] [PubMed] [Google Scholar]
  12. Chirino, A.J., Lous, E.J., Huber, M., Allen, J.P., Schenck, C.C., Paddock, M.L., Feher, G., and Rees, D.C. 1994. Crystallographic analyses of site-directed mutants of the photosynthetic reaction center from Rhodobacter sphaeroides. Biochemistry (Mosc). 33 4584–4593. [DOI] [PubMed] [Google Scholar]
  13. Cornette, J.L., Cease, K.B., Margalit, H., Spouge, J.L., Berzofsky, J.A., and DeLisi, C. 1987. Hydrophobicity scales and computational techniques for detecting amphipathic structures in proteins. J. Mol. Biol. 195 659–685. [DOI] [PubMed] [Google Scholar]
  14. Deisenhofer, J., Epp, O., Sinning, I., and Michel, H. 1995. Crystallographic refinement at 2.3 A resolution and refined model of the photosynthetic reaction centre from Rhodopseudomonas viridis. J. Mol. Biol. 246 429–457. [DOI] [PubMed] [Google Scholar]
  15. Dotsch, V. and Wagner, G. 1998. New approaches to structure determination by NMR spectroscopy. Curr. Opin. Struct. Biol. 8 619–623. [DOI] [PubMed] [Google Scholar]
  16. Doyle, D.A., Cabral, J.M., Pfuetzner, R., and Michel, H. 1998. The structure of potassium channel: Molecular basis of K+ conduction and selectivity. Science 280 69–77. [DOI] [PubMed] [Google Scholar]
  17. Ermler, U., Fritzsch, G., Buchanan, S.K., and Michel, H. 1994. Structure of the photosynthetic reaction centre from rhodobacter sphaeroides at 2.65 angstroms resolution: Cofactors and protein-cofactor interactions. Structure 2 925–936. [DOI] [PubMed] [Google Scholar]
  18. Godzik, A. 1995. In search of the ideal protein sequence. Protein Eng. 8 409–416. [DOI] [PubMed] [Google Scholar]
  19. Godzik, A., Kolinski, A., and Skolnick, J. 1993. De novo and inverse folding predictions of protein structure and dynamics. J. Comput. Aided Mol. Des. 7 397–438. [DOI] [PubMed] [Google Scholar]
  20. Goffeau, A., Nakai, K., Slonimski, P., and Risler, J.L. 1993. The membrane proteins encoded by yeast chromosome III genes. FEBS Lett. 325 112–117. [DOI] [PubMed] [Google Scholar]
  21. Grigorieff, N., Ceska, T.A., Downing, K.H., Baldwin, J.M., and Henderson, R. 1996. Electron-crystallographic refinement of the structure of bacteriorhodopsin. J. Mol. Biol. 259 393–421. [DOI] [PubMed] [Google Scholar]
  22. Guex, N. and Peitsch, M.C. 1997. SWISS-MODEL and the Swiss-PdbViewer: An environment for comparative protein modeling. Electrophoresis 18 2714–2723. [DOI] [PubMed] [Google Scholar]
  23. Henderson, R., Baldwin, J.M., Ceska, T.A., Zemlin, F., Beckmann, E., and Downing, K.H. 1990. Model for the structure of bacteriorhodopsin based on high-resolution electron cryo-microscopy. J. Mol. Biol. 213 899–929. [DOI] [PubMed] [Google Scholar]
  24. Hirokawa, T., Boon-Chieng, S., and Mitaku, S. 1998. SOSUI: classification and secondary structure prediction system for membrane proteins. Bioinformatics 14 378–379. [DOI] [PubMed] [Google Scholar]
  25. Hochkoeppler, A., Landau, E.M., Venturoli, G., Zannoni, D., Feick, R., and Luisi, P.L. 1995. Photochemistry of a photosynthetic reaction centre immobilized in lipidic cubic phases. Biotechnol. Bioeng. 46 93–98. [DOI] [PubMed] [Google Scholar]
  26. Hu, W.P., Godzik, A., and Skolnick, J. 1997. Sequence-structure specificity—how does an inverse folding approach work? Protein Eng. 10 317–331. [DOI] [PubMed] [Google Scholar]
  27. Hubbell, W.L., Gross, A., Langen, R., and Lietzow, M.A. 1998. Recent advances in site-directed spin labeling of proteins. Curr. Opin. Struct. Biol. 8 649–656. [DOI] [PubMed] [Google Scholar]
  28. Iwata, S., Ostermeier, C., Ludwig, B., and Michel, H. 1995. Structure at 2.8 Å resolution of cytochrome c oxidase from Paracoccus denitrificans. Nature 376 660–669. [DOI] [PubMed] [Google Scholar]
  29. Jones, D.T. 1998. Do transmembrane protein superfolds exist. FEBS Lett. 423 281–285. [DOI] [PubMed] [Google Scholar]
  30. Jones, D.T., Bryson, K., Tress, M.L., and Hadley, C. 1999. Successful protein fold recognition using sequence and secondary structure constrained threading methods. Proteins S3 104–111. [DOI] [PubMed] [Google Scholar]
  31. Jones, D.T., Taylor, W.R., and Thornton, J.M. 1992. A new approach to protein fold recognition. Nature 358 86–89. [DOI] [PubMed] [Google Scholar]
  32. Juretic, D. and Lucin, A. 1998. The preference functions method for predicting protein helical turns with membrane propensity. J. Chem. Inf. Comput. Sci. 38 575–585. [Google Scholar]
  33. Kabsch, W. and Sander, C. 1983. Dictionary of protein secondary structure: Pattern recognition of hydrogen-bonded and geometrical features. Biopolymers 22 2577–2637. [DOI] [PubMed] [Google Scholar]
  34. Kimura, Y., Vassylyev, D.G., Miyazawa, A., Kidera, A., Matsushima, M., Mitsuoka, K., Murata, K., Hirai, T., and Fujiyoshi, Y. 1997. Surface of bacteriorhodopsin revealed by high-resolution electron crystallography. Nature 389 206–211. [DOI] [PubMed] [Google Scholar]
  35. Koepke, J., Hu, X., Muenke, C., Schulten, K., and Michel, H. 1996. The crystal structure of the light-harvesting complex II (B800-850) from Rhodospirillum molischianum. Structure 4 581–597. [DOI] [PubMed] [Google Scholar]
  36. Koretke, K.K., Russell, R.B., Copley, R.R., and Lupas, A.N. 1999. Fold recognition using sequence and secondary structure information. Proteins S3 141–148. [DOI] [PubMed] [Google Scholar]
  37. Kreusch, A., Neubueser, A., Schiltz, E., Weckesser, J., and Schulz, G. 1994. The structure of the membrane channel porin from Rhodopseudomonas blastica at 2.0 angstroms resolution. Protein Sci. 3 58–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Kyrpides, N.C. 1999. Genomes OnLine Database (GOLD 1.0): A monitor ofcomplete and ongoing genome projects world-wide. Bioinformatics 15 773–774. [DOI] [PubMed] [Google Scholar]
  39. Landau, E.M. and Luisi, P.L. 1993. Lipidic cubic phases as transparent, rigid matrices for the direct spectroscopic study of immobilized membrane proteins. J. Am. Chem. Soc. 115 2102–2106. [Google Scholar]
  40. Luecke, H., Richter, H.T., and Lanyi, J.K. 1998. Proton transfer pathway in bacteriorhodopsin at 2.3 angstrom resolution. Science 280 1934–1937. [DOI] [PubMed] [Google Scholar]
  41. Luecke, H., Schobert, B., Richter, H.T., Cartailler, J.P., and Lanyi, J.K. 1999. Structural changes in bacteriorhodopsin during ion transport at 2 angstrom resolution. Science 286 255–260. [DOI] [PubMed] [Google Scholar]
  42. Luthy, R., Bowie, J.U., and Eisenberg, D. 1992. Assessment of protein models with three-dimensional profiles. Nature 356 83–85. [DOI] [PubMed] [Google Scholar]
  43. Marassi, F.M. and Opella, S.J. 1998. NMR structural studies of membrane proteins. Curr. Opin. Struct. Biol. 8 640–648. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. McDermott, G., Prince, S.M., Freer, A.A., Hawthornthwaite-Lawless, A.M., Papiz, M.Z., Cogdell, R.J., and Isaacs, N.W. 1995. Crystal structure of an integral membrane light-harvesting complex from photosynthetic bacteria. Nature 374 517–521. [Google Scholar]
  45. Mitsuoka, K., Hirai, T., Murata, K., Miyazawa, A., Kidera, A., Kimura, Y., and Fujioshi, Y. 1999. The structure of bacteriorhodopsin at 3.0 Å resolution based on electron crystallography: Implication of the charge distribution. J. Mol. Biol. 286 861–882. [DOI] [PubMed] [Google Scholar]
  46. Moller, C., Buldt, G., Dencher, N.A., Engel, A., and Muller, D.J. 2000. Reversible loss of crystallinity on photobleaching purple membrane in the presence of hydroxylamine. J. Mol. Biol. 301 869–879. [DOI] [PubMed] [Google Scholar]
  47. Muller, D.J., Schoenenberger, C.A., Schabert, F., and Engel, A. 1997. Structural changes in native membrane proteins monitored at subnanometer resolution with the atomic force microscope: A review. J. Struct. Biol. 119 149–157. [DOI] [PubMed] [Google Scholar]
  48. Murzin, A.G. 1999. Structure classification-based assessment of CASP3 predictions for the fold recognition targets. Proteins S3 88–103. [DOI] [PubMed] [Google Scholar]
  49. Okada, T., Trong, I.L., Fox, B.A., Behnke, C.A., Stenkamp, R.E., and Palczewski, K. 2000. X-Ray diffraction analysis of three-dimensional crystals of bovine rhodopsin obtained from mixed micelles. J. Struct. Biol. 130 73–80. [DOI] [PubMed] [Google Scholar]
  50. Palczewski, K., Kumasaka, T., Hori, T., Behnke, C.A., Motoshima, H., Fox, B.A., Le Trong, I., Teller, D.C., Okada, T., Stenkamp, R.E., et al. 2000. Crystal structure of rhodopsin: A G protein-coupled receptor. Science 289 739–745. [DOI] [PubMed] [Google Scholar]
  51. Pebay-peyroula, E., Rummel, G., Rosenbusch, J.P., and Landau, E.M. 1997. X-ray structure of bacteriorhodopsin at 2.5 angstroms from microcrystals grown in lipidic cubic phases. Science 277 1676–1681. [DOI] [PubMed] [Google Scholar]
  52. Ponder, J.W. and Richards, F.M. 1987. Tertiary templates for proteins. Use of packing criteria in the enumeration of allowed sequences for different structural classes. J. Mol. Biol. 193 775–791. [DOI] [PubMed] [Google Scholar]
  53. Portmann, M., Landau, E.M., and Luisi, P.L. 1991. Spectroscopic and rheological studies of enzymes in rigid lipidic matrices: The case of a-chymotripsin in lysolectin/water phase. J. Phys. Chem. 95 8437–8440. [Google Scholar]
  54. Rees, D.C., DeAntonio, L., and Eisenberg, D. 1989. Hydrophobic organization of membrane proteins. Science 245 510–513. [DOI] [PubMed] [Google Scholar]
  55. Rost, B., Casadio, R., Fariselli, P., and Sander, C. 1995. Transmembrane helices predicted at 95% accuracy. Protein Sci. 4 521–533. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Sanchez, R. and Sali, A. 1997. Advances in comparative protein-structure modelling. Curr. Opin. Struct. Biol. 7 206–214. [DOI] [PubMed] [Google Scholar]
  57. Sander, C. and Schneider, R. 1991. Database of homology-derived protein structures and the structural meaning of sequence alignment. Proteins 9 56–68. [DOI] [PubMed] [Google Scholar]
  58. Schertler, G.F., Villa, C., and Henderson, R. 1993. Projection structure of rhodopsin. Nature 362 770–772. [DOI] [PubMed] [Google Scholar]
  59. Sippl, M.J. 1990. The calculation of conformational ensembles from potentials of mean force, an approach to the prediction of local structures in globular proteins. J. Mol. Biol. 213 659–683. [DOI] [PubMed] [Google Scholar]
  60. Sippl, M.J. 1995. Knowledge-based potentials for proteins. Curr. Opin. Struct. Biol. 5 229–235. [DOI] [PubMed] [Google Scholar]
  61. Snijder, H.J., Ubarretxena-Belandia, I., Blaauw, M., Kalk, K.H., Verheij, H.M., Egmond, M.R., Dekker, N. and Dijkstra, B.W. 1999. Structural evidence for dimerization- regulated activation of an integral membrane phospholipase. Nature 401 717–721. [DOI] [PubMed] [Google Scholar]
  62. Sonnhammer, E.L.L., von Heijne, G., and Krogh, A. 1998. A hidden Markov model for predicting transmembrane helices in protein sequences. Proc. Int. Conf. Intell. Syst. Mol. Biol. 6 175–182. [PubMed] [Google Scholar]
  63. Stoddard, B.L. 1998. New results using laue diffraction and time-resolved crystallography. Curr. Opin. Struct. Biol. 8 612–618. [DOI] [PubMed] [Google Scholar]
  64. Stowell, M.H., McPhillips, T.M., Rees, D.C., Soltis, S.M., Abresch, E., and Feher, G. 1997. Light-induced structural changes in photosynthetic reaction centre: Implications for mechanism of electron-proton transfer. Science 276 812–816. [DOI] [PubMed] [Google Scholar]
  65. Stowell, M.H.B., Miazawa, A., and Unwin, N. 1998. Macromolecular structure determination by electron microscopy: New advances and recent results. Curr. Opin. Struct. Biol. 8 595–600. [DOI] [PubMed] [Google Scholar]
  66. Subramaniam, S., Lindahl, M., Bullough, P., Faruq, A.R., Tittor, J., Oesterhelt, D., Brown, L., Lanyi, J.K., and Henderson, R. 1999. Protein conformational changes in the bacteriorhodopsin photocycle. J. Mol. Biol. 287 145–161. [DOI] [PubMed] [Google Scholar]
  67. Takeda, K., Sato, H., Hino, T., Kono, M., Fukuda, K., Sakurai, I., Okada, T., and Kouyama, T. 1998. A novel three-dimensional crystal of bacteriorhodopsin obtained by successive fusion of the vesicular assemblies. J. Mol. Biol. 283 463–474. [DOI] [PubMed] [Google Scholar]
  68. Tsukihara, T., Aoyama, H., Yamashita, E., Tomizaki, T., Yamaguchi, H., Shinzawa-Itoh, K., Nakashima, R., Yaono, R., and Yoshikawa, S. 1996. The whole structure of the 13-subunit oxidized cytochrome c oxidase at 2.8 Å. Science 272 1136–1144. [DOI] [PubMed] [Google Scholar]
  69. von Heijne, G. 1992. Membrane protein structure prediction. Hydrophobicity analysis and the positive-inside rule. J. Mol. Biol. 225 487–494. [DOI] [PubMed] [Google Scholar]
  70. Weiss, M.S. and Schulz, G.E. 1992. Structure of porin refined at 1.8 Å resolution. J. Mol. Biol. 227 493–509. [DOI] [PubMed] [Google Scholar]
  71. Yoshikawa, S., Shinzawa-Itoh, K., Nakashima, R., Yaono, R., Yamashita, E., Inue, N., Yao, M., Fei, M.J., Libeu, C.P., Mizushima, T., et al. 1998. Redox-coupled crystal structural changes in bovine heart cytochrome c oxidase. Science 280 1723–1729. [DOI] [PubMed] [Google Scholar]
  72. Zhang, Z.L., Huang, L.S., Shulmeister, V.M., Chi, Y.-I., Kim, K.K., Hung, L.-W., Crofts, A.R., Berry, E.A., and Kim, S.-H. 1998. Electron transfer by domain movement in cytochrome bc1. Nature 392 677–684. [DOI] [PubMed] [Google Scholar]

Articles from Protein Science : A Publication of the Protein Society are provided here courtesy of The Protein Society

RESOURCES