Abstract

The objective of these studies was to test the hypothesis that proteins that contain potential polyisoprenyl recognition sequences (PIRSs) in their transmembrane-spanning domain can bind to the polyisoprenyl (PI) glycosyl carrier lipids undecaprenyl phosphate (C55-P) and dolichyl phosphate (C95-P). A number of prokaryotic and eukaryotic glycosyltransferases that utilize PI coenzymes contain a conserved PIRS postulated to be the active PI binding domain. To study this problem, we first determined the 3D structure of a PIRS peptide, NeuE, by homonuclear 2D 1H–nuclear magnetic resonance (NMR) spectroscopy. Experimentally generated distance constraints derived from nuclear Overhauser enhancement and torsion angle constraints derived from coupling constants were used for restrained molecular dynamics and energy minimization calculations. Molecular models of the NeuE peptide were built based on calculations of energy minimization using the DGII program NMRchitect. 3D models of dolichol (C95) and C95-P were built based on our 2D 1H-NMR nuclear Overhauser enhancement spectroscopy (NOESY) results and refined by energy minimization with respect to all atoms using the AMBER (assisted modeling with energy refinements) force field. Our energy minimization studies were carried out on a conformational model of dolichol that was originally derived from small-angle X-ray scattering and molecular mechanics methods. These results revealed that the PIs are conformationally nearly identical tripartite molecules, with their three domains arranged in a coiled, helical structure. Analyses of the intermolecular cross-peaks in the 2D NOESY spectra of PIRS peptides in the presence of PIs confirmed a highly specific interaction and identified key contact amino acids in the NeuE peptide that constituted a binding motif for interacting with the PIs. These studies also showed that subtle conformational changes occurred within both the PIs and the NeuE peptide after binding. 3D structures of the resulting molecular complexes revealed that each PI could bind more than one PIRS peptide. These studies thus represent the first evidence for a direct physical interaction between specific contact amino acids in the PIRS peptides and the PIs and supports the hypothesis of a bifunctional role for the PIs. The central idea is that these superlipids may serve as a structural scaffold to organize and stabilize in functional domains PIRS-containing proteins within multiglycosyltransferase complexes that participate in biosynthetic and translocation processes.

Introduction

A problem of fundamental importance in glycobiology that remains unresolved is how glycoconjugates are translocated across biological membranes and whether their synthesis is linked to translocation. A general unanswered question in eukaryotes, for example, is “How do dolichol (C95)-linked oligosaccharides in the protein N-linked glycosylation pathway traverse the endoplasmic reticulum?” (reviewed in Lennarz, 1987; Schenk et al., 2001). In prokaryotes, a related specific question is “How are the extended capsular polysialic acid chains of neuroinvasive Escherichia coli K1 transported across the plasma (inner) membrane?” (Troy, 1992). Key molecular players implicated in these similar processes are the polyisoprenyl (PI) glycosyl carrier lipids, dolichyl phosphate (C95-P) and undecaprenyl phosphate (C55-P), both of which function as chemical carriers of carbohydrate residues during synthesis in eukaryotes and prokaryotes, respectively (Troy, 1992; Parodi, 2000; Spiro, 2000; Lehrman, 2001). C95-P serves as an intermediate carrier of oligosaccharides during the membrane-directed synthesis of asparagine-linked glycoproteins and glycosylphosphatidylinositol anchors in eukaryotes and for synthesis of O-linked mannosyl chains in yeast (Parodi, 2000; Lehrman, 2001). A comprehensive review on the central role that nine defective genes in the N-linked glycosylation pathway plays in congenital disorders of glycosylation (CDGs) appeared recently (Freeze, 2001a,b).

In prokaryotes, C55-P is the carrier of sugar residues involved in synthesis of cell wall peptidoglycans, lipopolysaccharides, common antigen, and capsular polysaccharides (Osborn, 1971; Troy, 1979). The polysialic acid (polySia) capsule is a neurovirulent determinant in neurotropic E. coli K1 and Neisseria meningitidis as well as an example of a specific type of bacterial capsule whose synthesis involves C55-P as an intermediate carrier of Sia residues (Troy et al., 1975; Masson and Holbein, 1983).

The molecular details of how these structurally unique lipid cofactors function are unknown. It is hypothesized that the sugar residues linked to the lipid by a pyrophosphate bridge at the polar end of the isoprene molecules are translocated from one side of the membrane to the other, and the hydrocarbon chains presumably remain anchored in the nonpolar region of the membrane. This postulate, though widely reported in the literature, has not been well documented by direct experimental evidence (McCloskey and Troy, 1980a,b). Although much is known regarding the biochemistry of PIs, including their synthesis and the transfer reactions they mediate, there is a dearth of biophysical information as to how they may ferry saccharide units across membranes (Lennarz, 1987).

Based on the unusual length and poly-cis geometry of undecaprenyl and dolichyl derivatives (Figure 1), it was proposed earlier that these properties might endow the PIs with some unique physiochemical properties important for their biological function (McCloskey and Troy, 1980a). As a first step in understanding the role of the PIs in transmembrane glycoconjugate processes, previous studies focused on the organization and molecular motions of the PIs in model membranes and the effect these so-called superlipids had on membrane structure. These studies used a variety of biophysical approaches, including electron paramagnetic resonance (McCloskey and Troy, 1980a,b; Troy, 1991), 1H- and 2H-nuclear magnetic resonance (NMR) (de Ropp and Troy, 1984, 1985; de Ropp et al., 1987), 31P-NMR (Vigo et al., 1984; Valtersson et al., 1985; de Ropp et al., 1987; Knudsen and Troy, 1989a), differential scanning calorimetry, and fluorescence depolarization (Vigo et al., 1984; Valtersson et al., 1985). Several key findings emerged from studies using spin-labeled glycosyl carrier lipids, including information on the transbilayer diffusion rates and self-association of the PIs (McCloskey and Troy, 1980a,b). Information on the location and dynamics of the ω-terminus of the polyprenols was subsequently provided by 2H-NMR studies of deuteruim-labeled PIs in host phospholipid vesicles (de Ropp and Troy, 1984, 1985). Importantly, these studies also revealed that C55 and C95 derivatives altered the membrane host packing matrix, thus potentially modulating membrane lipid polymorphism (de Ropp and Troy, 1984, 1985; Knudsen and Troy, 1989a; Troy, 1991). 31P-NMR studies confirmed that the longer-chain PIs induced the formation of nonbilayer or inverted hexagonal organization of phospholipid molecules in phosphatidylcholine (PC) and PC/phosphatidylethanolamine membrane vesicles (Valtersson et al., 1985; de Ropp et al., 1987; Knudsen and Troy, 1989a; Troy, 1991).

Fig. 1.

Structure of the polyisoprenol molecules showing the relevant protons for this study. (a) and (b) represent the poly-CH protons; (c) and (d) the poly-CH2 protons; (e) and (f) the poly-CH3 protons; and (g) and (h) protons in the saturated isoprene units of C95 and C95-P, respectively.

Fig. 1.

Structure of the polyisoprenol molecules showing the relevant protons for this study. (a) and (b) represent the poly-CH protons; (c) and (d) the poly-CH2 protons; (e) and (f) the poly-CH3 protons; and (g) and (h) protons in the saturated isoprene units of C95 and C95-P, respectively.

The unexpected finding that the PIs could change membrane structure from bilayer to nonbilayer led to the hypothesis that alterations in bilayer structure might be important in facilitating the passage of sugar chains across biological membranes (Valtersson et al., 1985; de Ropp et al., 1987; Knudsen and Troy, 1989a; Troy, 1991). Recent studies by Waechter and colleagues have proposed, however, that hypothetical “flippases” may mediate the vectorial movement of Man5GlcNAc2-P-P-C95, Man-P-C95, and Glc-P-C95 from the cytosol across the endoplasmic reticulum membrane into the lumen (Rush et al., 1998; Rush and Waechter, 1998; Schenk et al., 2001). Although it is possible that such enzymes might eliminate the energy barrier for flipping by some unknown mechanism, an unattractive feature of the “flip-flop” model is that it does not adequately account for the considerable energy that would be required to initiate or drive hydrophilic glycosyl chains across a lipid bilayer. It has been estimated, for example, that translocation of one Man5GlcNAc2 chain attached to C95-P would require an energy expenditure of approximately 130–260 kcal/mol, or the equivalent of 13–26 ATP molecules per translocation. This is a minimal estimate, based only on the need to disrupt four hydrogen bonds with water per monomer (Lennarz, 1987). Such a large energy requirement suggests that if transverse diffusion was to occur, then the oligosaccharide chain may be shielded from the hydrophobic interior of the bilayer. Such a possibility would thus favor a nonbilayer or inverted hexagonal (HexII)-mediated model for translocation (Knudsen and Troy, 1989a). Alternatively, a great deal of energy may be required to drive PI-mediated oligosaccharide translocation across membranes.

Albright et al. (1989) first described a 13-amino-acid peptide consensus sequence in the presumed membrane-spanning domain of three yeast glycosyltransferases, Dpm1, Alg1, and Alg7, which was postulated to be a potential dolichol recognition sequence (DRS). A fourth yeast enzyme, Sec59 (a dolichol kinase), also contained a potential DRS (Heller et al., 1992). These findings were extended to mammalian glycosylphosphotransferases (GlcNAc-1- phosphotransferases; GPTs), which catalyzed synthesis of GlcNAc-P-P-C95 (Scocca and Krag, 1990; Zhu and Lehrman, 1990; Lehrman, 1991; Zhu et al., 1992; Datta and Lehrman, 1993). GPT was found to contain two potential DRSs, both of which were shown by mutational analysis to be required for enzyme function (Datta and Lehrman, 1993). In contrast, deletion and mutational analysis of the DRS in yeast dolichylphosphomannose synthase showed that this domain was not essential for enzyme activity (Zimmerman and Robbins, 1993). These conflicting results emphasized the potential limitation of deletion and mutational approaches to determine unambiguously the functional importance of the DRS and highlighted the need for direct structural information. Similar PI-binding sequences were later identified in other proteins of the C95 pathway in yeast and mammalian cells, including Alg2 (Jackson et al., 1993) and ribophorin I and II (Kelleher et al., 1992; Knauer and Lehle, 1999). Ribophorin I and II are two of the nine nonidentical membrane protein subunits that make up the oligosaccharyltransferase (OST) complex (Kelleher et al., 1992; Knauer and Lehle, 1999; Kim et al., 2000).

NeuE and KpsM, two proteins in the multienzyme polysialyltransferase complex in neuroinvasive E. coli K1, were also discovered to contain potential C95 recognition sequences in their membrane-spanning domains (Troy, 1992). Because these prokaryotic proteins would interact with the shorter chain PI, C55-P, we designated the PI binding domain by the more generic descriptor polyisoprenyl recognition sequence (PIRS) to denote that such sequences can interact with PIs other than C95. NeuE was initially postulated to be a sialyltransferase that may begin polySia chain synthesis by catalyzing the transfer of Sia from CMP-Sia to C55-P (Troy, 1992). However, we were surprised to discover a potential PIRS in KpsM because this protein had no known biosynthetic function, having been implicated only in polySia chain translocation (Pavelka et al., 1991; Troy, 1992; Pigeon and Silver, 1994). Table I compares the sequence of the 13-amino-acid PIRS peptides in the eukaryotic glycosyltransferases with the NeuE and KpsM proteins from E. coli K1.

Table I.

Comparison of the conserved 13 amino acid PIRS peptides in eukaryotic glycosyltransferases and in the E. coli K1 NeuE and KpsM proteins

Protein
 
PI substrate
 
PIRS peptides
 
References
 
Sec59 C95-P         10    Albright et al., 1989 
   
Dpm1 C95-P Albright et al., 1989 
Alg7 C95-P Albright et al., 1989 
Alg1 C95-PP-Gn2 Albright et al., 1989 
Alg2 C95-PP-Gn2M2 Jackson et al., 1993 
GPT 1 C95-P Zhu and Lehrman, 1990 
GPT 2 C95-P Zhu and Lehrman, 1990 
Ribophorin I C95-P Kelleher et al., 1992 
NeuE C55-P Troy, 1992 
KpsM C55-P Troy, 1992 
Consensus sequence  Albright et al., 1989 
Protein
 
PI substrate
 
PIRS peptides
 
References
 
Sec59 C95-P         10    Albright et al., 1989 
   
Dpm1 C95-P Albright et al., 1989 
Alg7 C95-P Albright et al., 1989 
Alg1 C95-PP-Gn2 Albright et al., 1989 
Alg2 C95-PP-Gn2M2 Jackson et al., 1993 
GPT 1 C95-P Zhu and Lehrman, 1990 
GPT 2 C95-P Zhu and Lehrman, 1990 
Ribophorin I C95-P Kelleher et al., 1992 
NeuE C55-P Troy, 1992 
KpsM C55-P Troy, 1992 
Consensus sequence  Albright et al., 1989 

The discovery of PIRS in yeast and mammalian glycosyltransferases and ribophorins and in NeuE and KpsM led us to hypothesize that the PIs may serve as a structural scaffold to organize and tether in functional domains PIRS-containing proteins within multiple glycosyltransferase complexes (Troy, 1992). Such a role could possibly function to link biosynthetic and translocation processes. Minimally, this supposition would predict a direct interaction between the PI and PIRS motifs within the membrane-spanning domain of these proteins. No biophysical evidence has been published regarding this possibility. As a first step in testing this hypothesis, we initiated the present studies to address four specific aims. First, we sought to determine the 3D structures of the PIRS peptides by homonuclear 2D NMR methods and molecular modeling. Second, we sought to build 3D structures of C95, C95-P, and C55-P based on 2D 1H-NMR results and refined by energy minimization with respect to all atoms of the PIs using the AMBER (assisted modeling with energy refinements) force field. Our energy minimization studies were carried out on a conformational model of C95 that was originally derived from small-angle X-ray scattering (SAXS) and molecular mechanics methods (Murgolo et al., 1989). Third, using a combination of 2D 1H-NMR, molecular mechanical, and molecular dynamical simulations we sought to determine if PIRS peptides and PIs interacted, and if so, to determine the specificity of the binding, that is, what key contact residues within the PIRS peptides bound to the PIs. Fourth, we sought to estimate the energetics of PIRS:PI binding and to determine how many PIRS peptides could bind a single PI molecule.

Our 2D NMR and molecular modeling results of the PI:PIRS complexes now provide the first biophysical evidence for a direct binding complex between specific contact amino acids in the PIRS peptides and the PIs. We have estimated the energetics of this binding and have determined that a single PI molecule can bind several PIRS peptides. Though the physiological significance of this interaction in vivo remains to be determined, it is anticipated that results from studies using model membranes may eventually help us better understand the importance of key PI:protein interactions in biological membranes, and thus their potential role in glycoconjugates synthesis/translocation processes. They might also lead to a better understanding of the molecular mechanisms of hypoglycosylation associated with some of the CDGs as well as the conformational disorders related to hyperglycosylation (Freeze, 2001a,b).

Results

Determination of the secondary and tertiary structure of the NeuE peptide

A 3D solution structure of the NeuE peptide was determined by high-resolution 1H-NMR. A complete description of all resonance protons in the 13-amino-acid PIRS peptide of NeuE is shown in Table II. The sequential connectivities observed for dNN(i,i+1) in the peptide are shown in Figure 2a. The thickness of the bars indicate the approximate magnitude of the nuclear Overhauser enhancement (NOE) intensity. Because the NeuE peptides contained Pro9, no dNN(i,i+1) connectivities were observed between amino acid residues 8 and 9 and between 8, 9, and 10 (Figure 2a). These results indicated that Pro9 induced a conformational bend in the peptide, which resulted in two distinct segments in the polypeptide backbone. The first segment, in the N-terminal domain, included amino acid residues 1–8, whereas the second segment, making the C-terminal half of the peptide, contained residues 9–13. The secondary and tertiary structural characteristics of the free NeuE peptide are summarized in the following section.

Fig. 2.

Sequential connectivity diagram for the NeuE peptide in the absence (a) and presence of C95 (b) and C95-P (c). The diagram shows the (i,i+1), (i,i+2), (i,i+3), (i,i+5), (i,i+7), and (i,i+8) contacts observed in the NOESY spectra and the spin–spin coupling constants, 3JHNα, measured from the 1D 1H-NMR spectra. The thickness of the bars indicates the approximate magnitude of the NOE intensity. The temperature coefficient of NH protons (ppb/K) is also shown.

Fig. 2.

Sequential connectivity diagram for the NeuE peptide in the absence (a) and presence of C95 (b) and C95-P (c). The diagram shows the (i,i+1), (i,i+2), (i,i+3), (i,i+5), (i,i+7), and (i,i+8) contacts observed in the NOESY spectra and the spin–spin coupling constants, 3JHNα, measured from the 1D 1H-NMR spectra. The thickness of the bars indicates the approximate magnitude of the NOE intensity. The temperature coefficient of NH protons (ppb/K) is also shown.

Table II.

Complete assignment of all resonance protons in the 13-amino-acid residues of the NeuE peptide in the presence (bold) and absence of C95 or C95-P

Amino acid residue no.
 
N-H
 
CαH
 
CβH
 
Other resonance protons
 
C95 
Leu1 NO 3.68 1.58, 1.41 0.99 (CγH) 
 NO 3.77 1.57, 1.45 1.01 
    0.89, 0.66 (CδH) 
    0.84, 0.62 
Ile2 8.38 4.25 1.65 1.05 (CγH) 
 8.43 4.28 1.68 1.03 
    0.75 (CδH) 
    0.78 
Ile3 8.06 4.09 1.62 1.16 (CγH) 
 8.08 4.11 1.65 1.08 
    0.84, 0.71 (CδH) 
    0.81, 0.69 
Leu4 7.94 4.25 1.52, 1.27 1.24 (CγH) 
 7.94 4.26 1.51, 1.28 1.27 
    0.77 (CδH) 
    0.74 
Phe5 7.80 4.50 2.96, 2.72 7.13 (C2,6H) 
 7.80 4.56 2.99, 2.75 7.19 
Leu6 8.05 4.27 1.50, 1.38 1.02 (CγH) 
 8.07 4.31 1.51, 1.40 1.03 
    0.82, 0.71 (CδH) 
    0.86, 0.69 
Ile7 7.61 4.17 1.60 1.00 (CγH) 
 7.61 4.13 1.57 0.97 
    0.71, 0.66 (CδH) 
    0.74, 0.69 
Phe8 8.20 4.62 2.92, 2.75 ND 
 8.20 4.65 2.95, 2.77 ND 
Pro9 NO 4.27 1.81, 1.70 1.50, 1.01 (CγH) 
 NO 4.27 1.85, 1.68 1.64, 1.02 
    3.54, 3.35 (CδH) 
    3.55, 3.33 
Phe10 7.76 4.47 2.91, 2.70 7.15 (C2,6H) 
 7.75 4.44 2.89, 2.74 ND 
Asn11 8.11 4.46 2.49, 2.35 7.37, 6.94 (NH2) 
 8.11 4.52 2.44, 2.39 7.36, 6.95 
Phe12 7.91 4.42 3.00, 2.72 7.16 (C2,6H) 
 7.89 4.43 3.02, 2.74 7.17 
Phe13 8.23 4.35 3.03, 2.94 7.13 (C2,6H) 
 8.27 4.42 3.05, 2.95 7.16 
C95-P     
Leu1 NO 3.68 1.58, 1.41 1.00 (CγH) 
 NO 3.78 1.59, 1.48 0.99 
    0.87, 0.66 (CδH) 
    0.89, 0.67 
Ile2 8.38 4.25 1.65 1.05 (CγH) 
 8.45 4.26 1.61 1.00 
    0.75 (CδH) 
    0.74 
Ile3 8.06 4.09 1.62 1.16 (CγH) 
 8.10 4.08 1.63 1.17 
    0.84, 0.71 (CδH) 
    0.86, 0.79 
Leu4 7.94 4.25 1.52, 1.27 1.24 (CγH) 
 7.96 4.24 1.51, 1.33 1.25 
    0.77 (CδH) 
    0.72 
Phe5 7.80 4.50 2.96, 2.72 7.13 (C2,6H) 
 7.82 4.53 2.95, 2.73 7.15 
Leu6 8.05 4.27 1.50, 1.38 1.02 (CγH) 
 8.08 4.30 1.45, 1.35 0.97 
    0.82, 0.71 (CδH) 
    0.90, 0.69 
Ile7 7.61 4.17 1.60 1.00 (CγH) 
 7.63 4.14 1.58 1.28 
    0.71, 0.66 (CδH) 
    0.69, 0.67 
Phe8 8.20 4.62 2.92, 2.75 ND 
 8.21 4.61 2.91, 2.74 ND 
Pro9 NO 4.27 1.81, 1.70 1.50, 1.01 (CγH) 
 NO 4.26 1.84, 1.72 1.67, 1.02 
    3.54, 3.35 (CδH) 
    3.53, 3.37 
Phe10 7.76 4.47 2.91, 2.70 7.15 (C2,6H) 
 7.75 4.45 2.92, 2.71 ND 
Asn11 8.11 4.46 2.49, 2.35 7.37, 6.94 (NH2
 8.13 4.49 2.49, 2.36 7.40, 6.98 
Phe12 7.91 4.42 3.00, 2.72 7.16 (C2,6H) 
 7.90 4.43 2.99, 2.73 7.18 
Phe13 8.23 4.35 3.03, 2.94 7.13 (C2,6H) 
 8.28 4.36 3.01, 2.92 7.16 
Amino acid residue no.
 
N-H
 
CαH
 
CβH
 
Other resonance protons
 
C95 
Leu1 NO 3.68 1.58, 1.41 0.99 (CγH) 
 NO 3.77 1.57, 1.45 1.01 
    0.89, 0.66 (CδH) 
    0.84, 0.62 
Ile2 8.38 4.25 1.65 1.05 (CγH) 
 8.43 4.28 1.68 1.03 
    0.75 (CδH) 
    0.78 
Ile3 8.06 4.09 1.62 1.16 (CγH) 
 8.08 4.11 1.65 1.08 
    0.84, 0.71 (CδH) 
    0.81, 0.69 
Leu4 7.94 4.25 1.52, 1.27 1.24 (CγH) 
 7.94 4.26 1.51, 1.28 1.27 
    0.77 (CδH) 
    0.74 
Phe5 7.80 4.50 2.96, 2.72 7.13 (C2,6H) 
 7.80 4.56 2.99, 2.75 7.19 
Leu6 8.05 4.27 1.50, 1.38 1.02 (CγH) 
 8.07 4.31 1.51, 1.40 1.03 
    0.82, 0.71 (CδH) 
    0.86, 0.69 
Ile7 7.61 4.17 1.60 1.00 (CγH) 
 7.61 4.13 1.57 0.97 
    0.71, 0.66 (CδH) 
    0.74, 0.69 
Phe8 8.20 4.62 2.92, 2.75 ND 
 8.20 4.65 2.95, 2.77 ND 
Pro9 NO 4.27 1.81, 1.70 1.50, 1.01 (CγH) 
 NO 4.27 1.85, 1.68 1.64, 1.02 
    3.54, 3.35 (CδH) 
    3.55, 3.33 
Phe10 7.76 4.47 2.91, 2.70 7.15 (C2,6H) 
 7.75 4.44 2.89, 2.74 ND 
Asn11 8.11 4.46 2.49, 2.35 7.37, 6.94 (NH2) 
 8.11 4.52 2.44, 2.39 7.36, 6.95 
Phe12 7.91 4.42 3.00, 2.72 7.16 (C2,6H) 
 7.89 4.43 3.02, 2.74 7.17 
Phe13 8.23 4.35 3.03, 2.94 7.13 (C2,6H) 
 8.27 4.42 3.05, 2.95 7.16 
C95-P     
Leu1 NO 3.68 1.58, 1.41 1.00 (CγH) 
 NO 3.78 1.59, 1.48 0.99 
    0.87, 0.66 (CδH) 
    0.89, 0.67 
Ile2 8.38 4.25 1.65 1.05 (CγH) 
 8.45 4.26 1.61 1.00 
    0.75 (CδH) 
    0.74 
Ile3 8.06 4.09 1.62 1.16 (CγH) 
 8.10 4.08 1.63 1.17 
    0.84, 0.71 (CδH) 
    0.86, 0.79 
Leu4 7.94 4.25 1.52, 1.27 1.24 (CγH) 
 7.96 4.24 1.51, 1.33 1.25 
    0.77 (CδH) 
    0.72 
Phe5 7.80 4.50 2.96, 2.72 7.13 (C2,6H) 
 7.82 4.53 2.95, 2.73 7.15 
Leu6 8.05 4.27 1.50, 1.38 1.02 (CγH) 
 8.08 4.30 1.45, 1.35 0.97 
    0.82, 0.71 (CδH) 
    0.90, 0.69 
Ile7 7.61 4.17 1.60 1.00 (CγH) 
 7.63 4.14 1.58 1.28 
    0.71, 0.66 (CδH) 
    0.69, 0.67 
Phe8 8.20 4.62 2.92, 2.75 ND 
 8.21 4.61 2.91, 2.74 ND 
Pro9 NO 4.27 1.81, 1.70 1.50, 1.01 (CγH) 
 NO 4.26 1.84, 1.72 1.67, 1.02 
    3.54, 3.35 (CδH) 
    3.53, 3.37 
Phe10 7.76 4.47 2.91, 2.70 7.15 (C2,6H) 
 7.75 4.45 2.92, 2.71 ND 
Asn11 8.11 4.46 2.49, 2.35 7.37, 6.94 (NH2
 8.13 4.49 2.49, 2.36 7.40, 6.98 
Phe12 7.91 4.42 3.00, 2.72 7.16 (C2,6H) 
 7.90 4.43 2.99, 2.73 7.18 
Phe13 8.23 4.35 3.03, 2.94 7.13 (C2,6H) 
 8.28 4.36 3.01, 2.92 7.16 

All spectra were recorded in DMSO-d6 at 18°C. Chemical shifts in ppm. NO: Not observed. ND: Not determined.

Structure of the NeuE peptide

Strong and medium NOE intensities for dNN(i,i+1) connectivities and strong NOE intensities for dαN(i,i+1) connectivities were observed in both the N-terminal and C-terminal segments of the free NeuE peptide (Figure 2a). The medium connectivities (residues 3–8) are characteristic features of α-helix-like structures, and the strong dαN(i,i+1) connectivities (residues 10–13) are features more in common with β-sheet-like structures (WÏthrich, 1987). In addition, the vicinal coupling constants (3JHNα) were larger than 4 Hz and less than 9 Hz for all residues except Asn11 (Figure 2a). Therefore, the free NeuE peptide did not appear to consist solely of an α-helix or β-sheet structure. Rather, segment 1 (residues 1–8) contained other contacts, notably a weak NOE for dNN(i,i+2) connectivity between residues 4 and 6 and a medium NOE for dαN(i,i+3) connectivity between residues 2 and 5. Further, there was no dαN(i,i+1) connectivity between residues 4 and 5. These data reflected a more α-helix-like structure for the N-terminal half of the peptide (WÏthrich, 1987). In contrast, the C-terminal segment showed both a medium NOE for dNN(i,i+1) connectivity and a strong NOE for dαN(i,i+1) connectivity between residues 9–13 (Figure 2a). Furthermore, the vicinal coupling constant of residue 11 was 13.2 Hz, suggesting that the C-terminal segment of the free NeuE peptide had a more extended β-sheet-like structure.

The structures of the free NeuE peptide derived from the 2D NOE data, energy minimization, and simulated annealing calculations are shown in Figure 3. The top panel shows a stereoview of the backbone conformation after superimposing nine calculated NeuE structures. The lower panel shows a 3D energy minimized structural model of the peptide. The structures were determined using distance constraints and backbone dihedral angle constraints derived from 55 NOE cross-peaks and coupling constants. The average root mean squared deviation (RMSD) of these structures was 2.09 Å for all atoms. This conformation revealed that Leu1, Ile3, Leu6, and Ile7 were located on the same outer surface of the helical domain of the peptide. Superimposition of the terminal residues, 1 and 13, was not as well defined as with the other amino acids, as these residues were more flexible and thus showed fewer NOEs.

Fig. 3.

Structure of the NeuE peptide based on the 2D NOE results, energy minimization, and simulated annealing calculations. (a) Stereoview of the backbone conformation after superimposition of nine calculated NeuE peptide structures; (b) space filling model of the NeuE peptide. The energy minimization and simulated annealing calculations were carried out as described in Materials and methods. The molecular models were built based on calculations of energy minimization by DISCOVER software, as described under Materials and methods.

Fig. 3.

Structure of the NeuE peptide based on the 2D NOE results, energy minimization, and simulated annealing calculations. (a) Stereoview of the backbone conformation after superimposition of nine calculated NeuE peptide structures; (b) space filling model of the NeuE peptide. The energy minimization and simulated annealing calculations were carried out as described in Materials and methods. The molecular models were built based on calculations of energy minimization by DISCOVER software, as described under Materials and methods.

The NMR-derived conformation of the NeuE peptide was compared with the molecular-modeled structures of the PIRS peptides sequences found in the eukaryotic glycosyltransferases Alg1, Alg7, Dpm1, GPT 2, Ribophorin I, and the E. coli K1 Kps M peptide. All of the Pro-containing PIRS peptides had a similar conformational bend as NeuE, thus forming two segments within the peptide backbone. This gave rise in these peptides to a similar conformational structure as described for NeuE. Particularly noteworthy was the finding that amino acids 3, 6, and 7, which are key residues for binding of the NeuE peptide to the PIs, were also located on the external surface of the helical domain. This structural similarity infers that these residues might be involved in binding the PIs. Molecular modeling revealed that Ost4 had a similar helical structure as NeuE, suggesting that this OST subunit might also interact with C95/C95-P.

1H-NMR resonance assignments for C95 and C95-P

Partial resonance assignments for C95 were initially reported in 1976 (Mankowski et al., 1976). Our present 1H-NMR studies include assignments for protons in C95-P and C55-P and extend the preliminary assignments made earlier for C95. No 1H-NMR assignments for C95-P and C55-P have been reported previously. Significantly, we found that the resonance assignments for C55-P, C95-P, and C95 in the membrane mimetic solvent dimethyl sulfoxide (DMSO) and PC vesicles were essentially identical. This finding confirms and extends other studies showing that hydrophobic peptides can adopt the same structure in DMSO as in the native protein, thus providing further evidence that no artifacts are induced in DMSO (Albert and Yeagle, 2000; Yeagle et al., 2000a). The chemical shifts observed in the 1D 1H NMR spectra for the eight different types of relevant protons in C95 and C95-P (Figure 1a–h) are summarized in Table III.

Table III.

Assignment of resonance protons of C95 and C95-P in the presence (bold) and absence of the NeuE peptide

PI
 
Proton designationa
 
(a)
 
(b)
 
(c)
 
(d)
 
(e)
 
(f)
 
(g)
 
(h)
 
C95 5.05 4.96 1.96 1.89 1.61 1.51 1.22 0.81 
 4.94 4.71 ND 1.86 1.60 1.46 1.21 0.81 
C95-P 5.06 4.90 1.96 1.81 1.62 1.52 1.22 0.83 
 5.05 ND ND ND 1.60 1.48 1.20 0.82 
PI
 
Proton designationa
 
(a)
 
(b)
 
(c)
 
(d)
 
(e)
 
(f)
 
(g)
 
(h)
 
C95 5.05 4.96 1.96 1.89 1.61 1.51 1.22 0.81 
 4.94 4.71 ND 1.86 1.60 1.46 1.21 0.81 
C95-P 5.06 4.90 1.96 1.81 1.62 1.52 1.22 0.83 
 5.05 ND ND ND 1.60 1.48 1.20 0.82 

aProtons designated a–h are shown in Figure 1. Protons (a) and (b) represent the poly-CH protons; (c) and (d) the poly-CH2 protons; protons (e) and (f) the poly-CH3 protons; and (g) and (h) the protons in the saturated isoprene units.

The numbers represent chemical shifts in ppm. Spectra were recorded at 500 MHz in DMSO-d6 at 18°C as described in Materials and methods. ND: Not determined.

Conformational models of C95, C95-P, and C55-P

Using molecular mechanics, SAXS, and the MM2 force field program, a solution structure of C95 was first constructed by Murgolo et al. (1989). This model suggested that C95 was composed of a central coiled region flanked by two arms. The length of the molecule was estimated to be about 52 Å, which is about one-half the length of C95 in its extended form (Hanover and Lennarz, 1981). Using our 2D nuclear Overhauser enhancement spectroscopy (NOESY) results and the molecular mechanics/SAXS-derived model of C95 by Murgolo et al. (1989), we sought to determine the 3D structures of C95-P and C55-P and a revised model of C95 by energy minimization and simulated annealing calculations using AMBER force field.

Our 2D 1H NMR studies showed that the NOESY spectra of C95 and C95-P were similar (Figure 4a,b). For example, three cross-peaks were observed in the spectrum of C95-P (Figure 4b), representing NOEs between (1) polyCH-polyCH2, (2) polyCH-polyCH3, and (3) polyCH2-polyCH3, respectively. The same cross-peaks were also observed in the NOESY spectrum of C95 (Figure 4a). The protons responsible for these cross-peaks were all in the unsaturated isoprene units. Because there were no cross-peaks that represented connectivity between the protons of the unsaturated and saturated α-isoprene units, these data further revealed that the unsaturated polyisoprene units, which make up most of the PI backbone, are either in close contact or associated together. Based on these NMR findings, the 3D structures of C95, C95-P, and C55-P were built by energy minimization with respect to all atoms using AMBER force field (Weiner et al., 1984) as described under Materials and methods. The force constants for the phosphate and pyrophosphate groups in AMBER are well defined. These findings revealed that the 3D conformation of the PIs were nearly identical tripartite molecules with their three domains arranged in a coiled, helical structure (Figure 5).

Fig. 4.

2D 1H NMR NOESY spectra of (a) C95 and (b) C95-P. Spectrawere generated as described in Materials and methods. The concentrationof PIs was 1.5 mM, and the mixing time was 500 ms. Cross-peaks A, B, and C represent the following proton pairs: A, polyCH-polyCH2; B, polyCH-polyCH3; C, polyCH2-polyCH3, as described in Figure 1.

Fig. 4.

2D 1H NMR NOESY spectra of (a) C95 and (b) C95-P. Spectrawere generated as described in Materials and methods. The concentrationof PIs was 1.5 mM, and the mixing time was 500 ms. Cross-peaks A, B, and C represent the following proton pairs: A, polyCH-polyCH2; B, polyCH-polyCH3; C, polyCH2-polyCH3, as described in Figure 1.

Fig. 5.

Space filling, 3D structural models of C55-P, left; C95-P, center; and C95, right. The models were determined by a combination of NMR, molecular mechanics, and molecular modeling methodology using the AMBER force field, as described in Materials and methods. The phosphorus atoms of the phosphorylated derivatives are shown in red.

Fig. 5.

Space filling, 3D structural models of C55-P, left; C95-P, center; and C95, right. The models were determined by a combination of NMR, molecular mechanics, and molecular modeling methodology using the AMBER force field, as described in Materials and methods. The phosphorus atoms of the phosphorylated derivatives are shown in red.

In these calculations the lowest steric energy conformations of C95-P and C55-P also gave rise to three geometrical regions, consisting of a central coiled portion and two flanking arms, representing the head and tail domains, respectively. These analyses also revealed that the coil and tail regions of C95-P and C55-P were more compact relative to that of C95. The more compact structures may explain the slightly different conformation induced in the NeuE peptide after binding to C95-P or C55-P, as compared to C95. Similar to free C95, C95-P also contains 1 saturated α-isoprene unit of the R configuration, 15 internal isoprene units of the cis configuration (in the coiled region), and 3 terminal trans-isoprene units (tail region). The sole difference between C95 and C95-P is the phosphate head group that is covalently linked to the saturated α-isoprene unit. C55-P also contains a phosphate head group that, in contrast with C95-P, is covalently linked to an unsaturated α-isoprene unit. C55-P contains eight internal isoprene units of the cis configuration, also in the coiled region, and two terminal trans-isoprene units, located in the tail region. These new 2D NMR-based findings thus allowed the 3D structures of C95, C95-P, and C55-P to be built (Figure 5). The geometrical and size parameters for the lowest energy conformations of these models are summarized in Table IV.

Table IV.

Geometrical and size parameters of the lowest energy conformers of C95, C95-P, and C55-P

PI species
 
Distance between head and tail (Å)
 
Average width (Å)
 
Average radius of coil (Å)
 
Length of coil (Å)
 
Length of head (Å)
 
Length of tail (Å)
 
C95 33 11.56 3.58 28 15 13 
C95-P 32 12.19 4.36 28 15 13 
C55-P 22 7.84 3.17 19 10 
PI species
 
Distance between head and tail (Å)
 
Average width (Å)
 
Average radius of coil (Å)
 
Length of coil (Å)
 
Length of head (Å)
 
Length of tail (Å)
 
C95 33 11.56 3.58 28 15 13 
C95-P 32 12.19 4.36 28 15 13 
C55-P 22 7.84 3.17 19 10 

NMR evidence for a direct binding between the PIRS peptides and the PIs

To determine if the NeuE peptide bound directly to C95 and C95-P, the chemical shift of all protons in the 13-amino-acid NeuE peptide and relevant protons in the PIs were determined (Table II). These data were essential for determining the secondary and tertiary structure of the PIRS peptides and the conformation of the binding complexes with C95, C95-P, and C55-P.

The complete assignment of all resonance protons in the NeuE peptide in the presence and absence of C95 and C95-P is shown in Table II. The sequential connectivities observed for the NeuE peptide in the presence of C95 and C95-P are summarized in Figure 2b and c, respectively. This diagram shows the (i,i+1), (i,i+2), (i,i+3), (i,i+5), (i,i+7), and (i,i+8) contacts observed in the NOESY spectra, and the spin–spin coupling constants, 3JHNα, measured from the 1D 1H-NMR spectra. The temperature coefficient of NH protons (ppb/K) are also shown. In the presence of C95 or C95-P, the resonance and sequential assignments of NeuE were similar to those of the free peptide, except that changes in the chemical shift of some residues were observed, as will be described.

Structure of the NeuE peptide after binding to C95

The dNN(i,i+1) connectivities for residues 3–8 and 10–13; the dαN(i,i+1) connectivities for residues 1–4, 5–8, and 10–13; and the medium NOE for dαN(i,i+3) connectivity between residues 2 and 5 that were observed for the free NeuE peptide (Figure 2a) were also observed when NeuE bound to C95 (Figure 2b). A comparison of the stereoview of the backbone conformation of the NeuE peptide before and after binding to C95 is shown in Figure 6a and b, respectively. As with the free peptide (Figure 6a), the NeuE:C95 docking structure (Figure 6b) is based on both the 2D NOE results and on energy minimization and simulated annealing. The stereoview conformations represent the superimposition of nine NeuE structures calculated using distance and backbone dihedral angle constraints, derived from 85 NOE cross-peaks and coupling constants. The average RMSD for the NeuE:C95 docking structure was 1.87 Å for all atoms.

Fig. 6.

Conformational change induced in the NeuE peptide after binding to C95 and C95-P. The stereoview structures of the NeuE peptide are based on 2D NOE results, energy minimization, and simulated annealing calculations, as described in the text. Superimposition of nine calculated free NeuE structures before docking to PIs (a), and after binding to C95 (b) and C95-P (c). The numbers shown designate the position of the first, third, sixth, and seventh amino acid residue in the PIRS peptide.

Fig. 6.

Conformational change induced in the NeuE peptide after binding to C95 and C95-P. The stereoview structures of the NeuE peptide are based on 2D NOE results, energy minimization, and simulated annealing calculations, as described in the text. Superimposition of nine calculated free NeuE structures before docking to PIs (a), and after binding to C95 (b) and C95-P (c). The numbers shown designate the position of the first, third, sixth, and seventh amino acid residue in the PIRS peptide.

These results revealed that the structure of the NeuE peptide after binding to C95, like the structure of the free peptide, was not singly an α-helix or β-sheet but was also bipartite. For example, the N-terminal half of the molecule contained more α-helix-like character, while the C-terminal segment a more extended β-sheet-like conformation. Several important differences were observed, however, in the NOEs between the free and bound peptide. A comparison of the sequential connectivities of the free peptide with that bound to C95 (Figure 2a,b) revealed the following differences: (1) a medium NOE between residues 2 and 3 for dNN(i,i+1) connectivity, a medium NOE between residues 3 and 5 for dNN(i,i+2) connectivity, and a medium NOE for dαN(i,i+3) connectivity between residues 3 and 6 were found only in the NeuE:C95 bound complex (Figure 2b); (2) the dNN(i,i+1) contacts for residues 3–6 and 11–12, though weak in the NeuE:C95 structure, were absent in the free NeuE peptide; (3) the vicinal coupling constant of residue 11 after binding to C95 was only 3 Hz, suggesting that the C-terminal half of NeuE was less extended after docking; and (4) docking induced a connection between the two NeuE segments, based on the medium dβN(i,i+8) connectivity observed between residues 5 and 13 (Figure 2b).

The backbone conformation of NeuE when bound to C95 (Figure 6b) also revealed that both Leu1 and Leu6 were again positioned on the same outer surface of the helical segment of the molecule, as was observed for the free NeuE peptide (Figure 6a). The superimposition fit for residues 1–13 in the bound state was better than for NeuE alone, because more NOEs were observed. As with the free NeuE peptide, the superimposed conformers of bound NeuE (Figure 6b) showed neither a characteristic α-helix or β-sheet structure. In contrast with the more extended structure seen in the free peptide (Figure 6a), residues 10–13 in the bound structure were in a bent conformation (Figure 6b). This conformational change resulted in the detection of dβN connectivities between residues 5 and 13 (Figure 2b). Therefore, based on the sequential connectivities for NeuE in the presence and absence of C95 (Figure 2a,b) and by comparing the stereoview backbone conformations of free NeuE with the peptide in the bound NeuE:C95 structure (Figure 6a,b), we conclude that docking of NeuE to C95 induced a conformational change in the peptide, resulting in a more compact structure in the carboxyl terminal segment.

Structure of the NeuE peptide after binding to C95-P

Similar to the structure of free NeuE peptide and the structure of NeuE after binding to C95 (Figure 6), strong dαN(i,i+1) connectivities in NeuE were observed in both segments of the peptide after binding to the phosphorylated PI, C95-P (Figure 2c). Segment 1 (residues 1–8), for example, contained more α-helix-like features, as indicated by a strong dαβ(i,i+3) connectivity between residues 2 and 5, a medium dNN(i,i+1) connectivity between residues 4 to 8, a medium dαN(i,i+3) connectivity between residues 3 and 6, and a medium dαN(i,i+1) connectivity between residues 3 and 5 (Figure 2c). Unexpectedly, we found different connection characteristics in both the N- and C-terminal segments of the peptide after binding to C95-P compared with C95. In the interaction between NeuE and C95-P, medium and strong dαβ(i,i+7) connectivities were observed between residues 6 and 13 and between residues 5 and 12 (Figure 2c). In contrast, in the NeuE:C95 complex only a medium dβN(i,i+8) connectivity between residues 5 and 13 was observed (Figure 2b). This suggested that the C-terminal segment of the peptide in the NeuE:C95-P complex was even more compact than in the NeuE:C95 structure. In support of this conclusion, no dNN(i,i+1) connectivity between residues 11 and 12 was observed after NeuE bound C95-P (Figure 2c), as there was in the free peptide (Figure 2a) or in the NeuE:C95 complex (Figure 2b).

A strong NOE was also observed between residues 9 and 10 for dαN(i,i+1), in the NeuE:C95-P structure, as shown in Figure 2c. These sequential connectivity data thus verified that a slightly different conformational change occurred in the NeuE peptide after binding to the phosphorylated PI, C95-P (Figure 6c), compared with the nonphosphorylated alcohol, C95 (Figure 6b). Like the structure of the free NeuE peptide (Figure 6a) and the peptide after binding to C95 (Figure 6b), the conformation of NeuE bound to C95-P (Figure 6c) was determined by superimposing nine peptide structures, calculated by using distance and backbone dihedral angle constraints that were derived from 111 NOE cross-peaks and coupling constants. The average RMSD for the NeuE:C95-P structure was 1.80 Å for all atoms. These structures revealed that Ile3, Leu6, and Ile7 were located on the same outer surface of the helical segment of the peptide molecule (Figure 6c). The conformation of the N-terminal segment of the peptide (residues 2–8) in the NeuE:C95-P docked complex was better defined than the structure of free NeuE or the peptide in the NeuE:C95 complex. In the NeuE:C95-P structure, for example, the C-terminal backbone segment is bent toward the N-terminal segment, such that αβ connectivities between residues 5 and 12 and residues 6 and 13 were detected, as shown in Figure 2c. Energy minimized space-filling models of the NeuE peptide before and after binding to C95 and C95-P are shown in Figure 7a–c, respectively.

Fig. 7.

Energy minimized space-filling models of the NeuE peptide before (a) and after (b) binding C95 and C95-P. These structures show the conformational changes induced in the peptide after binding the PI and were obtained by 2D 1H-NMR, energy minimization, and simulated annealing calculations as described in Materials and methods.

Fig. 7.

Energy minimized space-filling models of the NeuE peptide before (a) and after (b) binding C95 and C95-P. These structures show the conformational changes induced in the peptide after binding the PI and were obtained by 2D 1H-NMR, energy minimization, and simulated annealing calculations as described in Materials and methods.

To determine if any hydrogen bonds were present in the free or bound NeuE peptide, the 1D 1H NMR spectra of all three structures were compared. This comparison revealed that all amide protons in the peptides were shifted up field with increasing temperature. The temperature dependency of the NH protons was found to be high (between 5–8 ppb/K) and uniformly distributed in the free peptide (Figure 2a), suggesting that there were no hydrogen bonds in the free NeuE peptide. However, the temperature dependency of the NH protons was low for residues 5 and 10 when NeuE bound to the PIs (between 1–2 ppb/K) and high for the other residues (3–8 ppb/K, Figure 2b and 4–6 ppb/K, Figure 2c). Accordingly, these data indicated that two hydrogen bonds were formed within NeuE as a result of a conformational change induced in the peptide after binding C95 or C95-P. These hydrogen bonds are thus responsible for the more compact structure seen in the bound peptides (Figures 6 and 7).

The specificity of interaction between PIs and PIRS peptides

A comparison of the 500 MHz 1H-NMR spectra of the NH proton resonances of the NeuE peptide before and after binding C95 or C95-P revealed that most resonances were shifted downfield after binding. A summary of all chemical shifts observed for the amino acid residues in NeuE after binding to C95 and C95-P is shown in Table II. These differences in chemical shift were the first suggestion of a direct interaction between NeuE and the PIs. This binding was further substantiated by chemical shift changes observed for the CH2 and CH3 protons in C55-P after binding NeuE peptide and in the 1H NMR spectra of C55-P in the presence and absence of the eukaryotic Dpm1 peptide in 1,2-dimyristoyl-sn-glycero-3-phosphatidylcholine (DMPC) vesicles (Figure 8). Because of the near identity in these chemical shifts and in the close similarities between the 3D structures of C55-P and C95-P (Figure 5), as well as between NeuE and Dpm1, these findings suggest that the other PIRS peptides (Table I) could also be used to study the specificity of the interaction between these PIRS-containing peptides and the PIs.

Fig. 8.

1H NMR spectra of C55-P in DMPC vesicles containing either the NeuE or Dpm1 peptide. (a) DMPC vesicles alone; (b) DMPC vesicles containing C55-P; (c) or DMPC vesicles containing C55-P and NeuE (A, left panel) or Dpm1 (B, right panel). The molar concentration ratio of NeuE and Dpm1 to C55-P was 0.5. Spectra were recorded at 300 MHz in D2O at 40°C.

Fig. 8.

1H NMR spectra of C55-P in DMPC vesicles containing either the NeuE or Dpm1 peptide. (a) DMPC vesicles alone; (b) DMPC vesicles containing C55-P; (c) or DMPC vesicles containing C55-P and NeuE (A, left panel) or Dpm1 (B, right panel). The molar concentration ratio of NeuE and Dpm1 to C55-P was 0.5. Spectra were recorded at 300 MHz in D2O at 40°C.

Identification of specific protons on NeuE that mediates binding to the PIs

More cross-peaks were observed in the amide proton region of the 2D 1H NOESY spectra of NeuE in the presence of C95 or C95-P compared with that of the peptide alone (Figure 9a–c). These results suggested that the motional state and conformation of NeuE changed after binding to C95 or C95-P. As summarized in Table V, the T1 differences for several key protons of NeuE in the presence and absence of C95 and C95-P validated this conclusion. These results showed that the T1 times for the N-H protons of NeuE were shortest (455–490 ms) in the presence of C95. In contrast, the T1 time for the protons in the γ and δ regions of NeuE were shortest (420 ms) in the presence of C95-P. Therefore, the motional properties of the NeuE peptide were different when bound to C95 than when bound to C95-P, suggesting a different binding interaction between the peptide and the free alcohol and the phosphorylated PI. This conclusion was supported by determining the T1 times for C95 and C95-P protons in the presence and absence of NeuE. As shown in Table VI, the T1 values of the poly-CH2 and CH3 protons adjacent to the head group of C95 were longer in the presence of NeuE than those of C95-P (380 versus 350 ms and 700 versus 420 ms, respectively), whereas the T1 times of the poly-CH3 and poly-CH protons of C95 were shorter than those of C95-P in the presence of the peptide (630 versus 805 ms).

Fig. 9.

Amide proton region in the 2D 1H NOESY spectra of NeuE peptide in the absence of PI (a) and presence of C95 (b) or C95-P (c). The molar concentration of NeuE to C95 or C95-P was 2.3 mM to 0.4 mM (15 mol% PI). The mixing time was 240 ms, and spectra were recorded in DMSO-d6 at 18°C, as described in Materials and methods.

Fig. 9.

Amide proton region in the 2D 1H NOESY spectra of NeuE peptide in the absence of PI (a) and presence of C95 (b) or C95-P (c). The molar concentration of NeuE to C95 or C95-P was 2.3 mM to 0.4 mM (15 mol% PI). The mixing time was 240 ms, and spectra were recorded in DMSO-d6 at 18°C, as described in Materials and methods.

Table V.

T1 Values (ms) for different proton resonances of the NeuE peptide in the presence and absence of C95 or C95-P

Component protons in:
 
N-H
 
Ring proton
 
Protons in γ and δ region
 
NeuE 630–805 540 455 
C95/NeuE 455–490 518 518 
C95-P/NeuE 630–700 546 420 
Component protons in:
 
N-H
 
Ring proton
 
Protons in γ and δ region
 
NeuE 630–805 540 455 
C95/NeuE 455–490 518 518 
C95-P/NeuE 630–700 546 420 
Table VI.

T1 Values (ms) for relevant protons of C95 and C95-P in the presence and absence of the NeuE peptide

Component protons in:
 
Poly-CH2
 
Poly-CH3
 
Poly-CH
 
CH3 protons adjacent to head group
 
C95 350 665 910 950 
C95/NeuE 380 420 630 700 
C95-P 350 630 840 1190 
C95-P/NeuE 350 490 805 420 
Component protons in:
 
Poly-CH2
 
Poly-CH3
 
Poly-CH
 
CH3 protons adjacent to head group
 
C95 350 665 910 950 
C95/NeuE 380 420 630 700 
C95-P 350 630 840 1190 
C95-P/NeuE 350 490 805 420 

Refer to Figure 1 for the position of the relevant protons in C95 and C95-P.

Therefore, based on these chemical shift differences in NeuE, the number of different cross-peaks in the NOESY spectra, and the T1 times for the NeuE protons in the presence and absence of PIs, we conclude that there are specific interactions between the NeuE peptide and C95 and C95-P. These interactions were confirmed by a comparison of the expanded upfield region in the 500 MHz 2D NOESY spectra of the NeuE peptide in the presence and absence of C95 or C95-P (Figure 10). Three cross-peaks revealed an interaction between NeuE and C95, as shown by the boxed areas (A–C) in Figure 10b. These cross-peaks were not present in the 2D NOESY spectrum of NeuE alone (Figure 10a), indicating a specific interaction between contact amino acids in the peptide and C95. The assignment of these three cross-peaks was as follows: (1) cross-peak A represented a δ proton of Leu1 and the CH3 protons of C95; (2) cross-peak B represented another δ proton of Leu1 and other CH3 protons of C95; and (3) cross-peak C represented a δ proton of Leu6 and the CH2 protons of C95. The 2D spectra for the binding of NeuE and C95-P, (Figure 10c), revealed four cross-peaks, as follows: (1) cross-peak A was between the γ proton of Ile3 and the CH3 protons of C95-P; (2) cross-peak B was between the δ proton of Ile3 and the CH3 protons of C95-P; (3) cross-peak C was between the δ proton of Leu6 and the CH2 protons of C95-P; and (4) cross-peak D was between one γ proton of Ile7 and the CH3 protons of C95-P. The presence of these cross-peaks thus provided confirmatory evidence for an interaction between specific protons on NeuE and specific protons on C95 and C95-P (Figure 10b and c, respectively). The docking pairs and energies for the binding of NeuE to C95, C55-P, and C95-P are summarized in Tables VII, VIII, and IX.

Fig. 10.

500 MHz 1H 2D NOESY spectra of the NeuE peptide in the absence of PI (a) and in the presence of C95 (b) and C95-P (c). The cross-peaksin the up-field proton resonance region resulting from the binding of NeuE and C95 (b) or C95-P (c) are shown in the boxed areas denoted A–C (b) and A–D (c). The distance and docking energies for these docking pairs are shown in Table VII, VIII, and IX. The molar concentration of NeuE to C95 or C95-P was 2.3 and 0.4 mM (15 mol% PI), respectively. The mixing time was 400 ms, and the spectra were recorded in DMSO-d6 at 18°C, as described in Materials and methods.

Fig. 10.

500 MHz 1H 2D NOESY spectra of the NeuE peptide in the absence of PI (a) and in the presence of C95 (b) and C95-P (c). The cross-peaksin the up-field proton resonance region resulting from the binding of NeuE and C95 (b) or C95-P (c) are shown in the boxed areas denoted A–C (b) and A–D (c). The distance and docking energies for these docking pairs are shown in Table VII, VIII, and IX. The molar concentration of NeuE to C95 or C95-P was 2.3 and 0.4 mM (15 mol% PI), respectively. The mixing time was 400 ms, and the spectra were recorded in DMSO-d6 at 18°C, as described in Materials and methods.

Table VII.

Docking pairs between the NeuE peptide and C95

Cross-peaksa
 
Proton from NeuE
 
Proton from C95
 
Docking energy (kcal/mol)
 
H-bond
 
 2.9 Å    
L1: Hδ— —CH3 −10 4CO—5HN 
 2.5 Å    
L1: Hδ— —CH2  9CO—10HN 
 2.4 Å    
L6: Hδ— —CH2   
Cross-peaksa
 
Proton from NeuE
 
Proton from C95
 
Docking energy (kcal/mol)
 
H-bond
 
 2.9 Å    
L1: Hδ— —CH3 −10 4CO—5HN 
 2.5 Å    
L1: Hδ— —CH2  9CO—10HN 
 2.4 Å    
L6: Hδ— —CH2   

aDetermined from the NOESY spectrum shown in Figure 10b.

Table VIII.

Predicted docking pairs between the NeuE peptide and C55-P

Predicted cross-peaks
 
Proton from NeuE
 
Proton from C55-P
 
Docking energy (kcal/mol)
 
H-bond
 
 2.3 Å    
I3: Hγ— —CH2 −9 1CO—5HN 
 2.6 Å    
L6: Hδ— —CH3  5CO—10HN 
 2.9 Å    
I7: Hδ— —CH3   
Predicted cross-peaks
 
Proton from NeuE
 
Proton from C55-P
 
Docking energy (kcal/mol)
 
H-bond
 
 2.3 Å    
I3: Hγ— —CH2 −9 1CO—5HN 
 2.6 Å    
L6: Hδ— —CH3  5CO—10HN 
 2.9 Å    
I7: Hδ— —CH3   

The energy minimized NeuE peptide structure (Figure 3) was docked to the 3D structure of C55-P. The docking pairs between peptide and C55-P were then derived by molecular modeling, as described in Materials and methods.

Table IX.

Docking pairs between the NeuE peptide and C95-P

Predicted cross-peaksa
 
Proton from NeuE
 
Proton from C95-P
 
Docking energy (kcal/mol)
 
H-bond
 
 3.4 Å    
I3: Hδ— —CH3 −9 1CO—5HN 
 2.6 Å    
I3: Hγ— —CH3  5CO—10HN 
 3.5 Å    
L6: Hγ— —CH2   
 2.1 Å    
I7: Hγ— —CH3   
Predicted cross-peaksa
 
Proton from NeuE
 
Proton from C95-P
 
Docking energy (kcal/mol)
 
H-bond
 
 3.4 Å    
I3: Hδ— —CH3 −9 1CO—5HN 
 2.6 Å    
I3: Hγ— —CH3  5CO—10HN 
 3.5 Å    
L6: Hγ— —CH2   
 2.1 Å    
I7: Hγ— —CH3   

aDetermined from NOESY spectrum shown in Figure 10c.

3D Structures of the binding complexes between NeuE and the PI derivatives

Figure 11 shows the tertiary structural models of the binding complexes formed between the NeuE peptide and C95 (top right), C55-P (top left), and C95-P (lower panel). Construction of these models was based on the 2D NOE spectral data described and the tertiary structural models for NeuE and the PIs shown in Figures 3 and 5, respectively. The docking complex between NeuE and C95-P (bottom panel, a–e) is shown as it is rotated through 180° to show the stereo views of all of the specific points of interaction. As shown in Figure 6, the Leu1, Ile3, Leu6, and Ile7 residues of NeuE are all located on the same outer surface of the peptide. Because of this accessibility, these amino acids were considered to be the most likely key contact residues involved in the binding to the PIs. This supposition was borne out by the docking structures (Figure 11), which showed that the points of contact between the NeuE peptide and the PIs were specific. For example, three NOE cross-peaks (designated A–C; Figure 10b), which arose from the coupling interaction between two protons that were within van der Waals distance, were involved in the binding of the NeuE peptide to C95 (Table VII; Figure 11 top right). Two of these NOEs (A and B) were between the δ protons of Leu1 and the CH3 and CH2 protons of C95, respectively. The third NOE (C) was between the δ proton of Leu6 and a CH2 proton in C95 (Table VII). Similarly, the binding of NeuE to C55-P (Table VIII; Figure 11 top left) and to C95-P (Table IX; Figure 11 bottom panel) also involved NOEs between specific protons on Ile3, Leu6 and Ile7, and CH3 and CH2 protons from C95-P and C55-P, respectively.

Fig. 11.

3D energy minimized structural models of the NeuE peptide bound to C95, C95-P, and C55-P. These tertiary structural models were builtby a combination of 2D 1H NMR results, energy minimization using the AMBER force field and simulated annealing calculations, as described in Materials and methods. Model of NeuE peptide bound to C55-P (top left): C55-P is color-coded yellow. The Ile3 and Ile7 residues of NeuE are shown in light purple, Leu6 in turquoise, and the remaining amino acids in green. The phosphorus atom is shown in red. Model of NeuE bound to C95(top right): C95 is color-coded yellow, Leu1 is orange, and Leu6 is light purple. The other amino acids in the peptide are shown in green.Model of NeuE bound to C95-P (bottom panels ae): C95-P is shown in yellow. Ile3 and Ile7 are shown in crimison and Leu6 in blue, while theother amino acid residues of the peptide are shown in green. The phosphorus atom is shown in red. The docking structure is shown as it is rotatedthrough 180° from (a) 0°→(b) 45°→(c) 90°→(d) 135°→(e) 180°.

Fig. 11.

3D energy minimized structural models of the NeuE peptide bound to C95, C95-P, and C55-P. These tertiary structural models were builtby a combination of 2D 1H NMR results, energy minimization using the AMBER force field and simulated annealing calculations, as described in Materials and methods. Model of NeuE peptide bound to C55-P (top left): C55-P is color-coded yellow. The Ile3 and Ile7 residues of NeuE are shown in light purple, Leu6 in turquoise, and the remaining amino acids in green. The phosphorus atom is shown in red. Model of NeuE bound to C95(top right): C95 is color-coded yellow, Leu1 is orange, and Leu6 is light purple. The other amino acids in the peptide are shown in green.Model of NeuE bound to C95-P (bottom panels ae): C95-P is shown in yellow. Ile3 and Ile7 are shown in crimison and Leu6 in blue, while theother amino acid residues of the peptide are shown in green. The phosphorus atom is shown in red. The docking structure is shown as it is rotatedthrough 180° from (a) 0°→(b) 45°→(c) 90°→(d) 135°→(e) 180°.

A total of four NOE cross-peaks (designated A–D; Figures 10c) were identified in the interaction between NeuE and C95-P (Table IX). Two of these were between the δ and γ protons of Ile3 and CH3 protons of C95-P, whereas one each occurred between the γ proton on Leu6 and a CH2 proton of the PI and between the γ proton on Ile7 and a methyl proton of C95-P. Also, as shown in Figure 11 (bottom panel) the protons on Ile3 and Ile7 (crimson) and Leu6 (blue) all bound to the coiled region of C95-P. In this model, Ile3, Leu6, and Ile7 were again the most accessible amino acids in the peptide for binding to the phosphorylated PI. A nearly identical docking structure for the interaction between NeuE and C55-P was obtained (Figure 11; top left). In this structure, a total of three NOE cross-peaks involving protons from Ile3, Leu6, and Ile7 and CH2 and CH3 protons in C55-P were predicted (Table VIII). The NOEs, relevant protons, distance between interacting protons, and the docking energies for the NOEs in these NeuE:PI binding complexes are summarized in Tables VII, VIII, and IX.

PI derivatives can bind more than one NeuE peptide

Energy minimization and simulated annealing studies showed that a single C95, C95-P, or C55-P molecule could bind more than one NeuE peptide. Tertiary structural models of these energetically favorable docking structures revealed that C95 and C95-P could each bind at least two NeuE peptides (Figure 12). For C95, the most favorable binding energy for the docking between the three proton pairs on NeuE and the PI was −10 kcal/mol (Table VII). All three of the amino acids involved in this binding are located on the opposite side of the C95 alcohol head group (right side of the PI in Figure 12). When the second NeuE peptide was bound to the opposite or head group side of the same PI (left side of the PI in Figure 12), the binding energy was slightly less favorable (–7 kcal/mol). Similarly, as shown in Figure 12 (top right), one C95-P molecule could bind at least two NeuE peptides via four proton docking pairs, as summarized in Table IX. Again, the most favorable docking energy (–9 kcal/mol) was for the NeuE peptide binding to the helical conformation on the opposite side of the phosphorylated head group (right side of the PI in Figure 12), and the binding energy for the second peptide docking to the head group side of the PI (left side of the PI in Figure 12) was −7 kcal/mol. It was possible to determine by energy minimization and molecular modeling a docking structure for NeuE and C55-P by using the NMR-derived conformation of NeuE, determined after binding to C95-P (Figures 6 and 7c) and the 3D structural similarity between C55-P and C95-P (Figure 5).

Fig. 12.

3D energy minimized structural models of the docking of two NeuE molecules to one molecule each of C95 and C95-P. These tertiary structural models were obtained by a combination of 2D 1H NMR results, energy minimization using the AMBER force field and simulated annealing calculations, as described in Materials and methods. The C95 molecule (left) is shown in yellow, and the two contact amino acid residues in each NeuE peptide, Leu1 and Leu6, are shown in orange and light purple, respectively. The other amino acid residues of the peptide are shown in green. C95-P (right) is also shown in yellow with its phosphorous atom in red. Two of the three contact amino acids, Ile3 and Ile7, are shown in turquoise, and the third, Leu6, is shown in crimison. The other residues of the peptide are in green.

Fig. 12.

3D energy minimized structural models of the docking of two NeuE molecules to one molecule each of C95 and C95-P. These tertiary structural models were obtained by a combination of 2D 1H NMR results, energy minimization using the AMBER force field and simulated annealing calculations, as described in Materials and methods. The C95 molecule (left) is shown in yellow, and the two contact amino acid residues in each NeuE peptide, Leu1 and Leu6, are shown in orange and light purple, respectively. The other amino acid residues of the peptide are shown in green. C95-P (right) is also shown in yellow with its phosphorous atom in red. Two of the three contact amino acids, Ile3 and Ile7, are shown in turquoise, and the third, Leu6, is shown in crimison. The other residues of the peptide are in green.

The interaction of NeuE with the PIs requires a specific binding motif of key contact amino acids in the PIRS peptide

Because the transmembrane spanning domain of most membrane proteins are rich in hydrophobic amino acids, we considered the possibility that the interactions we were studying between NeuE, Dpm1, and the PIs reflected nonspecific interactions expected between the hydrophobic domain of a membrane protein and the PIs, even though none of the other membrane proteins associated with the kps gene complex required for polySia synthesis in E. coli K1 contained PIRS (Troy, 1992). To test this possibility, the specificity of NeuE and PI binding was further substantiated by 2D 1H NOESY experiments using Gramicidin A, a 15-amino-acid hydrophobic peptide, after mixing with various concentrations of C95. The molar ratio of Gramicidin A to C95 was studied at 1:2, 1:1, 2:1, and 4:1, respectively. No cross-peaks between protons of the PI and peptide were observed in the NOESY spectra. Furthermore, mixing Gramicidin A and C95 showed no significant change in the conformation of the PI, thus verifying that there were no direct or specific interactions between this similar hydrophobic control peptide and C95 (spectra not shown). Additionally, the molecular conformation of two mutant NeuE peptides in which the amino acid residues involved in PI binding were mutated was also determined, using the DISCOVER program on a Silicon Graphics Computer. In the first peptide, L1, I3, L6, and I7 residues were replaced with F1, F3, F6, and F7 residues, whereas in the second peptide, F3, F6, and F7 residues replaced the I3, L6, and I7 residues, respectively. These two peptides were then docked to C95 using the INSIGHTII program on a Silicon Graphics Computer, as described under Materials and methods. The docking energies for the mutated peptides were then calculated and found to be considerably larger (−3 kcal/mol) than those of the unmodified NeuE peptide binding to C95 (−10 kcal/mol). This provided additional evidence that L1, I3, L6, and I7 were the key contact amino acid residues in NeuE that appear to constitute a binding motif for the specific interaction between PIRS peptides and the PIs.

On the basis of these findings, we conclude that the physical interaction between the NeuE and Dpm 1 peptides with PIs observed in the present study were specific and not due to nonspecific or irrelevant hydrophobic interactions. Studies to determine more precisely the contribution of each of these amino acids in the binding of PIRS peptides to PIs will require additional NMR experiments in which each of these key residues have been systematically replaced.

Discussion

In these studies we asked whether the potential PIRS in membrane proteins involved in PI-mediated glycoconjugate synthesis could bind to the PI glycosyl carrier lipids, C55-P, C95-P, or C95. The PIRS is postulated to be the active domain of PI binding proteins. To study this problem, we first determined the 3D tertiary structure of the interacting molecules. Accordingly, the assignment of all resonance protons in the PIRS peptide of NeuE and the relevant protons in the PIs were determined using 2D NMR techniques. Using NOESY-derived distant constraints, energy minimization, and simulated annealing calculations, 3D structures of the NeuE peptide and the PIs were determined. These studies revealed that all of the PIs are conformationally nearly identical tripartite molecules, with their three domains arranged in a coiled, helical structure. 1H-NMR studies showed a direct binding of the PIRS peptide to the PIs, and the 3D structures of the resulting molecular complexes were determined. These findings identified the key contact amino acids in the NeuE peptide that constituted a potential binding motif for interacting with the PIs and showed that subtle conformational changes occurred within both the PIs and the NeuE peptide after binding. They further revealed that one PI molecule could bind several NeuE peptides.

Limits of interpretation of experimental results

Our findings raise several relevant points that should be considered with respect to the limits of interpretation of the data. The following discussion addresses these limits and also covers the potential significance of our results in support of the hypothesis that the glycosyl carrier PI lipids may have a bifunctional role in glycosylation/translocation processes. The conformational changes induced on binding and a description of the molecular motions and energetics of binding are also presented.

First, although our biophysical approach was not designed to determine the potential physiological importance of the binding of PIRS peptides to PIs, the high degree of specificity of the interaction suggests a likely biological role. Accordingly, the lack of a demonstrated physiological role of PI–protein interactions should not diminish the potential significance of such a complex, as these interactions may be as important as the multiple protein–protein interactions required for function of the OST complex (Kelleher et al., 1992; Spirig et al., 1997; Knauer and Lehle, 1999; Kim et al., 2000). The biophysical strategy developed here should therefore be viewed as representing the beginning of studies to determine the role of PI–PIRS interactions in glycosylation/translocation events. In this regard, it is noted that neither genetic (deletion or mutational analyses) nor biochemical approaches have been able to unambiguously determine the physiological importance of the PIRS peptides (Scocca and Krag, 1990; Zhu and Lehrman, 1990; Lehrman, 1991; Zhu et al., 1992; Datta and Lehrman, 1993; Jackson et al., 1993; Zimmerman and Robbins, 1993; Rush et al., 1998; Rush and Waechter, 1998).

A second consideration is whether the same structural features present in a 13-amino-acid PIRS peptide exists when the peptide is contained within a full-length protein. This is likely to be true, based on a number of studies showing that small peptides of integral membrane proteins have the same secondary structure as that of the peptide contained within the full-length protein (Dyson et al., 1992; Kahn et al., 1992; Blanco, 1994; Adler et al., 1995; Albert and Yeagle, 2000; Katragadda et al., 2000; Yeagle et al., 2000a,b). For example, one of the more thoroughly studied proteins is rhodopsin, where peptides of both the 6 and 7 transmembrane helical domains have the same 3D structure determined by NMR as the secondary structure of the peptides in the native protein (Kahn et al., 1992; Albert and Yeagle, 2000; Katragadda et al., 2000; Yeagle et al., 2000a). Similarly, studies on the four-helix bundle of myohemerythrin have also shown that small peptides of the protein retain the same secondary structure as the intact protein (Dyson et al., 1992). Based on these findings we conclude that our 3D structure of the transmembrane domain of NeuE is likely similar to the same peptide segment in the intact NeuE protein.

A third consideration to note is that a number of studies have shown that the 1H-NMR-derived solution structure of hydrophobic transmembrane peptide helicies, for example, those of bacteriorhodopsin, are similar in DMSO as the corresponding peptide region in the crystal structure of the protein (reviewed in Yeagle et al., 2000a; Albert and Yeagle, 2000; Katragadda et al., 2000). This indicates that this membrane mimetic solvent can be used for high-resolution NMR studies of hydrophobic peptides.

Conformational features of PIRS peptides

Calculations of structural parameters from spectroscopic data using NMR theory can be limited, particularly for smaller peptides that may undergo conformational changes faster than can be observed on the NMR time scale (Jardetzky, 1981). Because of the flexibility in the terminal residues of the NeuE peptide, the structure shown in Figure 3 may time average several different conformations and thus approximate only one of several potential minimum energy structures. However, application of Chou and Fasman rules (Chou and Fasman, 1974) predicted that the C-terminal segment of the peptide would be in an extended conformation, which is what we found experimentally for the nine calculated backbone conformations of the NeuE peptide. A more rigid C-terminal region was also found in the peptide after docking to C95 or C95-P, showing that binding induced a conformational change in the peptide (Figure 6). Thus the derived structure of the NeuE peptide, particularly after binding to the PI, is likely to be a reasonably accurate conformation.

The Pro9 residue in NeuE induced a distortion or bend after the N-terminal segment (residues 1–8) in the peptide (Figure 3). This is in accord with previous studies on the effect that Pro residues have on the conformation of polypeptide chains (Brandl and Deber, 1986; Deber et al., 1986; Gennis, 1989). Sequence analyses of the other PIRS peptides revealed that they too contained Pro in position 9 or 10, with the exception of Dpm1 and ribophorin I (Table I). All of the Pro-containing PIRS peptides had a similar conformational bend, thus forming two segments within the peptide backbone. A second characteristic common to most of the PIRS peptides is that the N-terminal segment usually contained Leu, Met, Ala, or Cys residues, amino acids that have a higher probability of forming α helices. The key Leu and Ile residues in the NeuE peptide that are involved in PI binding are all within this N-terminal segment, which, due to its distorted helix-like structure, is more rigid than the C-terminal segment. This structural perturbation within the peptide may aid and stabilize its interaction with the PIs. Based on the structural similarities between NeuE and the other PIRS peptides, we predict that the PI binding domain for these peptides will likely be located in their N-terminal segment.

Most of the C-terminal segment (residues 9–13) of the PIRS peptides contained Phe, Tyr, or Asp, residues, which have a higher probability of forming β-sheet or β-turn structures (Gennis, 1989). Thus, this segment of NeuE has a predicted higher probability of forming an extended β-like structure, as do the other PIRS-peptides shown in Table I. Notably, each C-terminal segment also contained Pro, Asn, or Tyr, which have been implicated as the most likely functional residues in transport of hydrophilic substitutes through the nonpolar domain of the membrane (Cantor and Schimmel, 1980). Our finding that there are subtle conformational changes induced in the NeuE peptide after binding C95 or C95-P—particularly in the C-terminal segment, which becomes more compact—suggests that these changes may induce neighboring residues to adopt a more α-helix-like structure. This observation is in accord with the finding that the α-helix content in some serum apolipoproteins increases after binding lipid (Cantor and Schimmel, 1980; Israelachvili and Pashley, 1982; Kahn et al., 1992).

Conformational features of the PIs: a revised C95 model

The 3D structure of C95 determined in this study (Figure 5) has some features in common with an earlier model (Murgolo et al., 1989), yet there are significant differences. The major difference is the length of the PI; our revised structure is considerably shorter, 33 Å in length, compared with 52 Å, as originally proposed. The principal reason for this difference is that our molecular modeling calculations yielded a more energy minimized structure. This resulted from two factors. First, advances in computational speed made since 1989 allowed us to carry out over 1000 iteration steps to calculate our energy minimized structures. Second, the MM2 force field method used by Murgolo et al. (1989) was an older-generation program that lacked many parameters, specifically force constants for the phosphate and pryrophosphate head groups. In contrast, our 3D structures of C95, C95-P, and C55-P were built by energy minimization with respect to all atoms using the AMBER force field (Weiner et al., 1984; Chou et al., 1998; Chou et al., 2000). In AMBER, the force constants for the phosphate and pyrophosphate groups are well defined.

Given that 33 Å is considerably shorter than the thickness of a biological membrane (40–60 Å), this means that if the PIs were oriented perpendicular to the plane of the bilayer, as previously shown for the phosphorylated derivatives (McCloskey and Troy, 1980a,b; de Ropp and Troy 1984; Valtersson et al., 1985; de Ropp et al., 1987; Murgolo et al., 1989), they would not span the bilayer. Rather, such a molecule would penetrate to only about the midbilayer region. This important structural feature of the PIs should be considered when proposing models to explain the function of PIs in glyconconjugate synthesis and translocation processes.

Molecular motions of the PIs: why does NeuE bind to the central coil region of the PIs?

The extensive poly-cis double bond geometry in the PIs likely endows them with the flexibility to fold into coils and to adopt the compact, tripartite structures shown in Figure 5. Because the central coil region (28 Å for C95 and C95-P and 19 Å for C55) is longer than both the head (15 Å for C95 and C95-P and 10 Å for C55 ) and tail (13 Å for C95 and C95-P and 8 Å for C55) regions (Table IV), the molecular motions of the coil region for all three PIs was predicted to be slower than the head and tail region. de Ropp et al. (1987) measured the T1 time of the acetyl ester head group and the tail group for 2H-labeled PIs using 2H NMR. These studies showed that the head and tail groups of the nonphos-phorylated PIs had the same T1 and correlation times. In the present study, we measured the T1 times of the poly-CH, poly-CH2, and poly-CH3 protons in the coil region of C95 and C95-P by 1H NMR and found that they were shorter than the protons near the head groups of C95 (Table VI). These results showed that the correlation time of the coil region of C95 was longer than that of the head and tail regions. Thus, the slower motion in the coil region is what likely facilitates the specific binding of NeuE to this domain of the PI molecules.

It has been shown that free C95 can alter membrane fluidity (Valtersson et al., 1985; de Ropp et al., 1987; Knudsen and Troy 1989a,b; Troy, 1991). Further, the results obtained from differential scanning calorimetry (Valtersson et al., 1985), fluorescence depolarization (Vigo et al., 1984), and 31P-NMR (Valtersson et al., 1985; de Ropp et al., 1987; Knudsen and Troy, 1989a; Troy, 1991) also demonstrated that C95-P can modulate fluidity of a PC bilayer and induce a bilayer to nonbilayer (HexII) structure, even at concentrations as low as 1–5 mol% (Knudsen and Troy, 1989a). Our present findings confirm these earlier reports and show further that the PI-induced fluidity in phospholipid membranes is less pronounced after the PIRS peptides bind the PIs. The T1 time for both the coil (poly-CH2 and poly-CH3) and head group region of C95 or C95-P, for example, are significantly lowered after binding NeuE peptide (Tables V and VI). Binding thus decreases the overall motional rates of the PIs. These findings support our hypothesis that the C95-P-induced alteration in bilayer structure may be modulated or at least partially stabilized when PIRS-containing proteins bind to polyisoprenols.

The results in Tables V and VI further show that the T1 values of protons in the head group of C95-P (Figure 1) were significantly longer than those of C95, indicating greater motion within this region of the phosphorylated PI in comparison with the free alcohol. This is in accord with our previous conclusion that C95 and C95-P have different orientations in membranes, possibly because of the strong charge repulsion between the phosphate head group of C95-P and phospholipids (McCloskey and Troy, 1980a,b). The exclusion interaction caused by these negatively charged groups likely increases the motional rate of the head group and protons in the α-isoprene unit. In contrast, the free hydroxyl head group on C95 is too distantly removed from the phosphate group of the phospholipids to have much effect. As a consequence, the head group and α-isoprene unit in C95 have slower motional rates than that of C95-P.

Conformational differences between the NeuE peptide when bound to C95 and C95-P

Both the 1D 1H NMR and 2D NOESY spectra of NeuE changed significantly after binding C95 or C95-P. No cross-peaks were observed, for example, in the NH-NH, NH-βH, NH-γH, or NH-δ regions of the NOESY spectrum for the free NeuE peptide, using 240 ms mixing time, and only 19 cross-peaks appeared in the NH-α region. In contrast, 46 cross-peaks were observed in the NOESY spectrum when NeuE bound C95, and 70 cross-peaks were observed when the peptide bound C95-P. This suggested that the NeuE peptide had different motional properties when bound to C95 or C95-P. This supposition was verified by determining the T1 values for different protons in the peptide in the presence and absence of C95 or C95-P (Tables V and VI). On the basis of these results, we conclude that the different motional rates reflect a different conformation and/or orientation of the NeuE peptide when bound to C95 or C95-P.

Conformational difference between C95 and C95-P after binding NeuE

Our earlier studies showed that free C95 in model membranes did not distribute homogeneously in a PC bilayer but rather underwent pronounced reversible self-association in which little interaction with the phospholipid acyl chains occurred (McCloskey and Troy, 1980a,b). C95-P, in contrast, remained monomolecularly dispersed. On this basis, we concluded that C95 and C95-P would have different conformations and/or orientations in the membrane, which we now recognize could be modulated by binding a PIRS peptide. Accordingly, differences in orientation and motional rates between C95 and C95-P could be an important factor in affecting the position or extent of NeuE binding. Thus, our energy minimization and simulated annealing calculations, which led to the construction of the structural models for the C55-P–NeuE, C95–NeuE, and C95-P–NeuE complexes (Figures 11 and 12), were significant for two reasons. First, they revealed different conformations for the free and bound PIs; second, they showed that conformational changes occurred in NeuE after binding PI.

Energetics of PIs–PIRS binding: van der Waalsinteractions are the major contributions to PI:PIRS docking

Because of the extreme hydrophobicity of both NeuE and the PIs, strong hydrophobic interactions would be predicted to dominate the interactive forces. Such hydrophobic interactions can influence the binding between two molecules over distances extending up to 100 Å (Israelachvili and Pashley, 1982). The docking energy between PIs:NeuE would thus be expected to differ with various binding sites on a PI molecule. Our results show that the minimum docking energy was −10 kcal/mol for a single NeuE peptide when bound to the opposite side of the head group of the central coil of C95 and −7 kcal/mol when bound to the coiled region on the head group side of the PI (Figure 12). Both of these binding interactions involved protons on the L1 and L6 residues of the NeuE peptide (Table VII). In addition, we found that minimum docking energies were obtained when L1 and L6 (for binding to C95) or I3, L6, and I7 (for binding to C95-P) interacted with the coiled region of the PIs (Table VII). Therefore, the specificity of binding between PIs and the PIRS peptide was further supported by measurements of the docking energies. These results indicated that van der Waals interaction were likely the major contribution to the specific binding between PIs and PIRS peptides because the van der Waals distances for these docking pairs is generally <3.0 Å (Tables VII, VIII, IX).

Possible bifunctional role of the PI carrier lipids in glycosylation/translocation reactions

Our molecular modeling showed that one PI molecule could bind more than one NeuE peptide, with differences in the docking energy between the two sides of the PI molecule (Figure 12). This suggests that it may be possible for a single PI molecule to bind several PIRS-containing glycosyltransferase/translocator proteins within a multienzyme complex. For example, a single C55-P molecule within the polysialyltransferase complex of E. coli K1 could bind to the PIRS motif in several NeuE and/or KpsM molecules to tether the transmembrane domain of these proteins in a complex that may link biosynthetic and translocation events (Troy, 1992).

We found that one C95-P or (C55-P) molecule could bind at least three NeuE peptides. The preferred binding site for the first peptide (N1) on C55-P (−9 kcal/mol) was also in the coiled region on the opposite side of the PI head group. When the second NeuE peptide (N2) was docked to the head group side of the PI, and the third peptide (N3) to the middle of the PI between N1 and N2, the docking energies were −6 and −9 kcal/mol, respectively. However, the position of docking of the three peptides made a significant difference in the docking energies. Although the energy minimized docking energy for binding of the first NeuE to the PIs was always 9–10 kcal/mol, if N2 was docked in close apposition to N1, and N3 close to N2, then the calculated docking energies for the N2 and N3 peptides were −12 and −17 kcal/mol, respectively. This suggested a positive cooperation in binding between the different NeuE peptides. Such cooperation may result from the fact that when the first NeuE is bound to the PI, the docking energy contains contributions from only van der Waals interactions between the peptide and PI. However, when the second and third peptides were moved around the initial NeuE–PI binding complex to determine the minimum docking energy, the binding energy for these peptides also contained contributions from electrostatic interactions between the two peptides, in addition to the van der Waals interactions. The significantly more negative docking energies resulting from each sequential NeuE binding further supported a positive cooperativity between NeuE peptides, a conclusion in accord with our finding that binding induced a conformational change in both NeuE (Figures 7, 8) and the PI (Figure 12).

The tethering of several NeuE and/or KpsM molecules could serve to form a channel or to alter bilayer structure such that nascent or newly synthesized polysialic acid chains could transit from the inner to the outer leaflet of the plasma membrane. The different orientations of PIs in the membrane (McCloskey and Troy, 1980b) could also affect the number of PIRS-containing polypeptides that bind to the PI, thus mediating or stabilizing cooperative interactions to influence biosynthetic and translocation processes. Although it is recognized that the transmembrane movement of lipid-linked oligosaccharides is of fundamental importance in N-linked glycosylation, identification of the putative “flippases” has proven elusive (Helenius et al., 2002). A potential candidate for this reaction in yeast was recently identified as the Rft 1 protein. Rft 1 is an evolutionarily conserved protein that is postulated to be required for the translocation of Man5- GlcNAc2-P-P-Dol across the endoplasmic reticulum membrane (Helenius and Aebi, 2001). Because no information is available regarding how this protein may be involved in this process, it seems premature to describe it as being involved in the flipping reaction. It would also seem important to determine if Rft 1 interacts with C95/C95-P before concluding that no additional factors are required for Rft 1 function (Helenius and Aebi, 2001). The idea that “flippases” and the “flipping reaction” is how PI-linked sugar chains actually traverse membranes appears to be an assumption, so good an assumption that some regard it as fact, even though it is an idea that excludes other possibilities.

It is conceivable that a C95-P–protein complex could also play a regulatory role in the N-linked glycosylation pathway in eukaryotic cells. For example, the OST complex involves nine nonidentical protein subunits, several of which contain potential PI binding sequences. OST catalyzes transfer of C95-P-linked Glc3Man9GlcNAc2-oligosaccharides to Asn-X-Ser/Thr sequons in nascent polypeptide chains as they are translocated across the endoplasmic membrane (Kelleher et al., 1992; Knauer and Lehle, 1999; Kim et al., 2000). The transmembrane domain of some OST proteins have been postulated to stabilize their interactions with other OST proteins, presumably via protein–protein interactions (Kim et al., 2000). Our finding that a PIRS peptide forms a binding complex with C95-P suggests that this binding may be another way to stabilize these multiprotein subcomplexes to facilitate synthesis and/or transport of lipid-linked glycans across membranes. The detailed molecular mechanism(s) of how these synthetic and translocation processes occur, however, is not understood in any biological system and thus awaits further study.

Materials and methods

Source of the PIs and DMPC

The polyisoprenols were purchased from Larodan Fine Chemical Laboratory (Malmö, Sweden). Deuterated DMPC was purchased from Avanti Polar Lipids (Alabaster, AL). The PIs and DMPC were >98% pure when analyzed by thin-layer chromatography (TLC) on silica gel F254–precoated plates (Merck, Darmastadt, FRG). The following solvent systems were employed: (A) CHCl3: CH3OH:H2O (65:25:4 v/v); (B) CHCl3:CH3OH:NH4OH (14.8 M):H2O (70:30:4:1); (C) CHCl3:CH3OH (5:1); and (D) CHCl3:CH3OH:H2O (10:10:3). All PIs gave single spots on TLC in several solvent systems. C95, C95-P, C55-P, and DMPC were dissolved in deuterated chloroform (99.8 atom % D, CDCl3, Aldrich, Milwaukee, WI) and stored as stock solutions at −20°C.

Synthesis of the PIRS peptides

Peptides containing the conserved 13 amino acids of the NeuE and Dpm1 PIRS were synthesized by California Peptide Research (Napa, CA) using standard t-BOC chemistry. The resin containing the peptide was dried under high vacuum, and the peptide cleaved from the resin by 90% HF at 0°C. The released peptides were purified by reverse-phase high-performance liquid chromatography and were estimated to be at least 80% pure. The extremely hydrophobic peptides were not soluble in aqueous solutions but were soluble in deuterated DMSO (DMSO-d6) where they were stored as a stock solution at −20°C. The sequence of the NeuE and Dpm1 peptides used in these studies was as follows:

  • NeuE: NH2-L-I-I-L-F-L-I-F-P-F-N-F-F-COOH

  • Dpm1: NH2-L-F-I-T-F-W-S-I-L-F-F-Y-V-COOH

Preparation of PIRS peptides, PIs, and mixtures of PIRS and PIs in DMSO-d6

To study the interaction between PIRS peptides and PIs, aliquots of the stock solutions of NeuE and C95 and NeuE and C95-P in DMSO-d6 and CDCl3 were mixed in a molar ratio of 2:1. The solvent was removed under a stream of N2. Trace amounts of DMSO-d6 and CDCl3 were removed by overnight lyophilization under high vacuum. Lyophilized samples were dissolved in 0.5 ml of DMSO-d6 by vortexing for 10 min. The final concentration of peptide was 2.3 mM and that of the PIs 0.5 mM. Similar methods were used to prepare the free NeuE and Dpm 1 peptide and PIs alone in DMSO. The viscosity of DMSO (10 M) at 18°C is 8 cP, which approximates that of biological membranes at 37°C (Eto and Rubinsky, 1992). The tendency of hydrophobic peptides to aggregate is also diminished in this membrane mimetic solvent (Wakamatsu et al., 1990). Several studies have shown that the 2D 1H-NMR derived solution structure of hydrophobic transmembrane peptide helicies, for example, those of bacteriorhodopsin, are similar in DMSO as the corresponding peptide region in the crystal structure of the protein (Albert and Yeagle, 2000; Katragadda et al., 2000; Yeagle et al., 2000). Thus, hydrophobic peptides can adopt the same structure in DMSO as in the native protein.

Preparation of deuterated DMPC vesicles

The solvent from an aliquot of the stock solution of deuterated DMPC was removed under a stream of N2. Traces of solvent were removed under high vacuum overnight. One half milliliter of D2O was added to the dried sample, which was vortexed for 10 min to achieve complete suspension of the multilamellar vesicles (MLVs). Small unilamellar vesicles (SUVs) were prepared by sonication of the MLVs on ice for 5 min under an inert atmosphere of argon, as previously described (McCloskey and Troy, 1980a,b). Sonication was carried out using an XL2020 Sonicator (Heat Systems-Ultrasonics Incorporated, Farmingdale, NY) fitted with a microprobe. After sonication, the resulting clear dispersion was centrifuged at 35,000×g for 30 min, and the supernatant containing the SUVs was used for the NMR experiment. The resulting SUVs were relatively homogeneous in size, with an average diameter of approximately 280 Å, as observed in the electron microscope (Zeiss EM 109) after negative staining with 1% uranyl acetate, as previously described (McCloskey and Troy, 1980a).

Incorporation of the PIRS peptide (Dpm1; NeuE) and PIs in DMPC vesicles

The following deuterated DMPC membrane vesicles containing either Dpm1 or NeuE peptide were prepared in the presence and absence of the PIs: (1) Dpm1:DMPC and NeuE:DMPC; (2) DMPC and the PIs, C95:DMPC and C55-P:DMPC; and (3) DMPC and the mixture of either PIRS peptides and PIs, C55-P:Dpm1:DMPC and C55-P:NeuE:DMPC. Vesicles were prepared by adding each component to a small flask to give the desired molar ratios, as described in the text. The bulk of the solvent was removed under a stream of N2 gas. The samples were lyophilized overnight, after which 0.5 ml D2O were added. After incubation at room temperature for 2 h, each sample was vortexed for 10 min to achieve complete suspension of the MLVs. Similarly, SUVs were prepared by sonication of the MLVs, as already described. After sonication, the samples were centrifuged at 13,000×g for 30 min, and the SUVs in the supernatant fraction were transferred to a 5-mm NMR tube.

Determination of the 3D structure of C95, C95-P, and C55-P and the Neu-E peptide: 1H NMR masurements

1H NMR spectroscopy was carried out at 18°C and 25°C on a 500 MHz Nicolet (General Electric) spectrometer. 1H shifts were measured relative to the methyl proton resonance of internal sodium 4,4-dimethyl-4-silapentane-1-sulfonate (DSS). 1D proton spectra were recorded with a sweep width of 8000 Hz and 16 k data points. A total of 1024 scans were accumulated, and an acquisition time of 2.05 s was used for the peptide and the mixture of peptide and PI in both DMSO-d6 and the DMPC membranes. A presaturated pulse sequence was used to inhibit the water resonance peak.

Correlated spectroscopy (COSY), NOESY (Macura and Ernest, 1980; Wuthrich, 1986), total correlation spectroscopy (TOCSY) (Bax and Davis, 1985), and double-quantum filtered COSY (DQF-COSY) (Rance et al., 1983) experiments were carried out on the 500 MHz Nicolet (General Electric) spectrometer. For NOESY experiments, mixing times of 120, 240, 400, and 480 ms were used. For TOCSY experiments, spin-locking fields of 8 kHz and 65 and 80 ms mixing time were used. 3N coupling constants were measured using DQF-COSY and incorporating 1D NMR spectra. All 2D spectral widths were 8000 Hz. The data size in the time domain was 512 points in t1 and 2048 points in t2. For each t1 value, 32 transients were accumulated. Data were processed with a combination of exponential and shifted sine-bell window functions for each dimension. The observed 512×2k complex data matrices were zero-filled to 2k×2k (COSY, NOESY, and TOCSY) or 1K×4K (DQF-COSY). For the DQF-COSY data, the final digital resolution after zero-filling was 1.8 Hz/point. Because of the relatively low solubility of the NeuE peptide in DMPC, it was difficult to obtain a complete 1D and 2D NMR spectra of the peptide in these vesicles alone. Because the peptide was soluble in DMSO, the complete 1D and 2D NMR spectra were determined in this membrane-mimetic solvent at 18°C.

NMR studies of the interaction of PIRS peptides and polyisoprenols

All 1H NMR spectroscopic studies of the binding of PIRS peptides to the PIs were carried out on the 500 MHz Nicolet spectrometer.

Structure refinement and calculations

Experimentally generated distance constraints derived from NOE intensities and torsion angle constraints derived from coupling constant information were used for restrained molecular dynamics and energy minimization calculations. The NOE distance constraints were classified as strong (1.8–2.5 Å), medium (1.8–3.3 Å), and weak (1.8–5.0 Å). The lower bounds for interproton distance were set to the sum of the van der Waals radii of the two protons. Molecular modeling calculations for the NeuE peptide were carried out on a Silicon Graphic Krebs computer using DGII (NMRchitect, Biosym/MSI, San Diego, CA). The 3D structures of C95, C95-P, and C55-P were built by energy minimization using the AMBER force field, where the force constants for the phosphate and pyrophosphate groups are well defined (Weiner et al., 1984; Chou et al., 1998; Chou et al., 1991; Chou et al., 2000). Due to the limited cross-peaks in the NOESY spectra of the PIs, it would have been difficult to obtain a more accurate molecular model using the NOE data and DGII calculations alone. However, Murgolo et al. (1989) had approximated the structure of C95 based on molecular mechanics and SAXS. Their energy minimization was carried out using the MM2 force field program. Therefore, we were able to use their SAXS results and energy minimization using AMBER with our NMR results to determine an energy minimized structure for C95, C95-P, and C55-P.

Calculations for the docking energies between PIRS peptides and PIs were carried out on the Silicon Graphic Krebs Computer using the Insight II program (Biosym). The objective of the binding calculation (nonbond energy) was to evaluate the interaction energy of many orientations of a PI molecule relative to the PIRS peptide, while searching for orientations that resulted in low interaction energies. Insight II provided information for calculating the interaction energy between two molecules using explicit electrostatic or van der Waals energies, or a combination of the two energies. In our calculations, 8 Å was used as the specified intermolecular cutoff parameter.

1
Present address: Harvard Medical School, Beth Israel Deaconess Medical Center, 41 Avenue Louis Pasteur, Research East Room 301, Boston, MA 02115, USA.

We wish to acknowledge the critical advice and help of Drs. Tom Jue, Jeffrey de Ropp, and Gary Smith during the course of these studies. We also wish to extend our appreciation to Dr. K.C. Chou for help in molecular modeling of the 3D structures of the PIs using AMBER, and to Dr. John Voss for his patience and expert help in finalizing the figures. This work was funded in part by National Institutes of Health Research Grants AI09352 and GM55703 (F.A.T.) and by a Hibbard E. Williams Research Grant from the U.C. Davis School of Medicine (F.A.T.). During a portion of these studies, G.-P. Zhou was supported by an NIH Training Grant in Biophysics.

References

Adler, M., Seto, M.H., Nitecki, D.E., Lin, J.H., Light, D.R., and Morser, J. (
1995
) The structure of a 19-residue fragment from the C-loop of the fourth epidermal growth factor-like domain of thrombomodulin.
J. Biol. Chem.
 ,
270
(40),
23366
–23372.
Albert, A.D. and Yeagle, P.L. (
2000
) Domain approach to three-dimensional structure of rhodopsin using high-resolution nuclear magnetic resonance.
Methods Enzymol.
 ,
315
,
107
–115.
Albright, C. F., Orlean, P., and Robbins, P.W. (
1989
) A 13-amino acid peptide in three yeast glycosyltransferases may be involved in dolichol recognition.
Proc. Natl Acad. Sci. USA
 ,
86
(19),
7366
–7369.
Bax, A. and Davis, D.G. (
1985
) MLEV-17-based two-dimensional homonuclear magnetization transfer spectroscopy.
J. Magn. Reson.
 ,
65
,
355
–360.
Blanco, F.J. and Serrano, L. (
1994
) Folding of protein GBI domain studied by the conformational characterization of fragments comprising its secondary structure elements.
Eur. J. Biochem.
 ,
230
,
634
–649.
Brandl, C. and Deber, C. (
1986
) Hypothesis about the function of membrane-buried proline residues in transport proteins.
Proc. Natl Acad. Sci. USA
 ,
83
(4),
917
–921.
Cantor, C.R. and Schimmel, P.R. (
1980
) Introduction to magnetic resonance. In Cantor, C.R. and Schimmel, P.R. (eds), Biophysical Chemistry, Part II: Techniques for the Study of Biological Structure and Function. W.H. Freeman and Co., San Francisco, pp. 409–433.
Chou, P.Y. and Fasman, G.D. (
1974
) Conformational parameters for amino acids in helical, beta-sheet, and random coil regions calculated from proteins.
Biochemistry
 ,
13
(2),
211
–222.
Chou, J.J., Matsuo, H., Duan, H., and Wagner, G. (
1998
) Solution structure of the RAIDD CARD and model for CARD/CARD interaction in caspase-2 and caspase-9 recruitment.
Cell
 ,
94
,
171
–180.
Chou, K.C., Watenpaugh, K.D., and Heinrikson, R.L. (
1999
) A model of the complex between cyclin-dependent kinase 5(Cdk5) and the activation domain of neuronal Cdk5 activator.
Biochem. Biophys. Res. Commun.
 ,
259
,
420
–428.
Chou, K.C., Tomasselli, A.G., and Heinrikson, R.L. (
2000
) Prediction of the tertiary structure of a capase-9/inhibitor complex.
FEBS Lett.
 ,
470
,
249
–256.
Datta, A.K. and Lehrman, M.A. (
1993
) Both potential dolichol recognition sequences of hamster GlcNAc-1-phosphate transferase are necessary for normal enzyme function.
J. Biol. Chem.
 ,
268
(17),
12663
–12668.
de Ropp, J.S. and Troy, F.A. (
1984
) Chemical synthesis and 2H NMR investigations of polyisoprenols: dynamics in model membranes.
Biochemistry
 ,
23
(12),
2691
–2695.
de Ropp, J.S. and Troy, F.A. (
1985
) 2H NMR investigation of the organization and dynamics of polyisoprenols in membranes.
J. Biol. Chem.
 ,
260
(29),
15669
–15674.
de Ropp, J.S., Knudsen, M.J., and Troy, F.A. (
1987
) 2H NMR Investigation of the dynamics and conformation of polyisoprenols in model membranes.
Chem. Scripta
 ,
27
,
101
–108.
Deber, C.M., Brandl, C.J., Deber, R.B., Hsu, L.C., and Young, X.K. (
1986
) Amino acid composition of the membrane and aqueous domains of integral membrane proteins.
Arch. Biochem. Biophys
 .
51
(1),
68
–76.
Dyson, H.J., Merutka, G., Waltho, J.P., Lerner, R.A., and Wright, P.E. (
1992
) Folding of peptide fragments comprising the complete sequence of proteins. Models for initiation of protein folding. I. Myohemerythrin.
J. Mol. Biol.
 ,
226
(3),
795
–817.
Eto, T.K. and Rubinsky, B. (
1992
) Topics in Heat Transfer.
Am. Soc. Mech. Eng.
 ,
206
(2),
47
–53.
Freeze, H.H. (
2001
) The pathology of N-glycosylation-stay the middle, avoid the risks.
Glycobiology
 ,
11
,
37G
–38G.
Freeze, H.H. (
2001
) Update and perspectives on congenital disorders of glycosylation.
Glycobiology
 ,
11
,
129R
–143R.
Gennis, R.B. (
1989
) Biomembranes: Molecular Structure and Function. Springer-Verlag, New York.
Hanover, J.A. and Lennarz, W.J. (
1981
) Transmembrane assembly of membrane and secretory glycoproteins.
Arch. Biochem. Biophys.
 ,
211
(1),
1
–19.
Helenius, A. and Aebi, M. (
2001
) Intracellular functions of N-linked glycans.
Science
 ,
291
(5512),
2364
–2369.
Helenius, J., Ng, D.T., Marolda, C.L., Walter, P., Valvano, M.A., and Aebi, M. (
2002
) Translocation of lipid-linked oligosaccharides across the ER membrane requires Rft1 protein.
Nature
 ,
415
(6870),
447
–450.
Heller, L., Orlean, P., and Adair, W.L., Jr. (
1992
) Saccharomyces cerevisiae sec59 cells are deficient in dolichol kinase activity.
Proc. Natl Acad. Sci. USA
 ,
89
(15),
7013
–7016.
Israelachvili, J. and Pashley, R. (
1982
) The hydrophobic interaction is long range, decaying exponentially with distance.
Nature
 ,
300
(5890),
341
–342.
Jackson, B. J., Kukuruzinska, M.A., and Robbins, P. (
1993
) Biosynthesis of asparagine-linked oligosaccharides in Saccharomyces cerevisiae: the alg2 mutation.
Glycobiology
 ,
3
(4),
357
–364.
Jardetzky, O. and Roberts, G.C.K. (
1981
) NMR in molecular biology. Academic Press, New York, p. 681.
Kahn, T.W., Sturtevant, J.M., and Engelman, D.M. (
1992
) Thermodynamic measurements of the contributions of helix-connecting loops and of retinal to the stability of bacteriorhodopsin.
Biochemistry
 ,
31
(37),
8829
–8839.
Katragadda, M., Alderfer, J.L., and Yeagle, P.L. (
2000
) Solution structure of the loops of bacteriorhodopsin closely resembles the crystal structure.
Biochem. Biophys. Res. Commun.
 ,
1466
,
1
–6.
Kelleher, D.J., Kreibich, G., and Gilmore, R. (
1992
) Oligosaccharyltransferase activity is associated with a protein complex composed of ribophorins I and II and a 48 kd protein.
Cell
 ,
69
(1),
55
–65.
Kim, H., Park, H., and Lennarz, W.J. (
2000
) Studies on the role of the hydrophobic domain of Ost4p in interactions with other subunits of yeast oligosaccharyl transferase.
Proc. Natl Acad. Sci. USA
 ,
97
(4),
1516
–1520.
Knauer, R. and Lehle, L. (
1999
) The oligosaccharyltransferase complex from yeast.
Biochim. Biophys. Acta
 ,
1426
(2),
259
–273.
Knudsen, M.J. and Troy, F.A. (
1989
) Nuclear magnetic resonance studies of polyisoprenols in model membranes.
Chem. Phys. Lipids
 ,
51
(3–4),
205
–212.
Knudsen, M.J. and Troy, F.A. (
1989
) 2H- and 31P NMR studies of glycosyl carrier lipids in membranes. Effect of undecaprenyl and dolichyl phosphate on bilayer structure. Proc. Xth International Symp. on Glycoconjugates, Jerusalem, Israel, Organizing Committee Press.
Lehrman, M.A. (
1991
) Biosynthesis of N-acetylglucosamine-P-P-dolichol, the committed step of asparagine-linked oligosaccharide assembly.
Glycobiology
 ,
1
(6),
553
–562.
Lehrman, M.A. (
2001
) Oligosaccharide-based information in endoplasmic reticulum quality control and other biological systems.
J. Biol. Chem.
 ,
276
,
8623
–8626.
Lennarz, W.J. (
1987
) Protein glycosylation in the endoplasmic reticulum: current topological issues.
Biochemistry
 ,
26
(23),
7205
–7210.
Macura, S. and Ernest, R.R. (
1980
) Elucidation of cross relaxation in liquids by two-dimensional NMR spectroscopy.
Mol. Phys.
 ,
41
,
95
–117.
Mankowski, T., Jankowski, W., Chojnacki, T. and Franke, P. (
1976
) C55-dolichol: occurrence in pig liver and preparation by hydrogenation of plant undecaprenol.
Biochemistry
 ,
15
(10),
2125
–2130.
Masson, L. and Holbein, B.E. (
1983
) Physiology of sialic acid capsular polysaccharide synthesis in serogroup B Neisseria meningitidis.
J. Bacteriol.
 ,
154
,
728
–736.
McCloskey, M.A. and Troy, F.A. (
1980
) Paramagnetic isoprenoid carrier lipids. 1.Chemical synthesis and incorporation into model membranes.
Biochemistry
 ,
19
,
2056
–2060.
McCloskey, M.A. and Troy, F.A. (
1980
) Paramagnetic isoprenoid carrier lipids. 2. Dispersion and dynamics in lipid membranes.
Biochemistry
 ,
19
,
2061
–2066.
Murgolo, N.J., A. Patel, Stivala, S.S. and Wong, T.K. (
1989
) The conformation of dolichol.
Biochemistry
 ,
28
,
253
–260.
Osborn, M.J. (1971) The role of membranes in the synthesis of macromolecules. In Rothfield, L.I. (ed.), Structure and Function of Biological Membranes. Academic Press, New York, pp. 343–400.
Parodi, A.J. (
2000
) Protein glucosylation and its role in protein folding.
Annu. Rev. Biochem.
 ,
69
(1),
69
–93.
Pavelka, M.J., Wright, L.F., and Silver, R.P. (
1991
) Identification of two genes, kpsM and kpsT, in region 3 of the polysialic acid gene cluster of Escherichia coli K1.
J. Bacteriol.
 ,
173
(15),
4603
–4610.
Pigeon, R.P. and Silver, R.P. (
1994
) Topological and mutational analysis of KpsM, the hydrophobic component of the ABC-transporter involved in the export of polysialic acid in Escherichia coli K1.
Mol. Microbiol.
 ,
14
(5),
871
–881.
Rance, M., O.W. Sorensen, Bogenhousen, G., Wagner, G., Ernst, R.R. and Wuthrich, K. (
1983
) Improved spectral resolution in COSY 1H NMR spectra of proteins via double qantrum filtering.
Biochem. Biophys. Res. Commun.
 ,
117
,
979
–987.
Rush, J.S. and Waechter, C.J. (
1998
) Topological studies on the enzymes catalyzing the biosynthesis of Glc-P-dolichol and the triglucosyl cap of Glc3Man9GlcNAc2-P-P-dolichol in microsomal vesicles from pig brain: use of the processing glucosidases I/II as latency markers.
Glycobiology
 ,
8
(12),
1207
–1213.
Rush, J. S., van Leyen, K., et al. (
1998
) Transbilayer movement of Glc-P-dolichol and its function as a glucosyl donor: protein-mediated transport of a water-soluble analog into sealed ER vesicles from pig brain.
Glycobiology
 ,
8
(12),
1195
–1205.
Schenk, B., Fernandez, F., and Waechter, C.J. (
2001
) The in(side) and out(side) of dolichyl phosphate biosynthesis and recycling in the endoplasmic reticulum.
Glycobiology
 ,
11
(5),
61R
–70R.
Scocca, J.R. and S.S. Krag (
1990
) Sequence of a cDNA that specifies the uridine diphosphate N-acetyl-D-glucosamine:dolichol phosphate N-acetylglucosamine-1-phosphate transferase from Chinese hamster ovary cells.
J. Biol. Chem.
 ,
265
(33),
20621
–20626.
Spirig, U., Glavas, M., Bodmer, D., Reiss, G., Burda, P., Lippuner, V., te Heesen, S. and Aebi, M. (
1997
) The STT3 protein is a component of the yeast oligosaccharyltransferase complex.
Mol. Gen. Genet.
 ,
256
(6),
628
–637.
Spiro, R.G. (
2000
) Glucose residues as key determinants in the biosynthesis and quality control of glycoproteins with N-linked oligosaccharides.
J. Biol. Chem.
 ,
275
(46),
35657
–35660.
Troy, F.A. (
1979
) The chemistry and biosynthesis of selected bacterial capsular polymers.
Annu. Rev. Microbiol.
 ,
33
,
519
–560.
Troy, F.A. (
1991
) NMR studies of polyisoprenyl derivatives on bilayer structure and the translocation of polysialic acid chains across the E. coli K1 inner membrane. Nineteenth Nobel Conference on Biosynthesis, Regulation and Products of the Mevalonate Pathway, Carolinska Medico Chirurgiska Institute Press.
Troy, F.A. (
1992
) Polysialylation: from bacteria to brains.
Glycobiology
 ,
2
(1),
5
–23.
Troy, F.A., Vijay, I.K., and Tesche, N. (
1975
) Role of undecaprenyl phosphate in synthesis of polymers containing sialic acid in Escherichia coli.
J. Biol. Chem.
 ,
250
,
156
–163.
Valtersson, C., van Duÿn, G., Verkleij, A.J., Chojnacki, T., de Kruijff, B. and Dallner, G. (
1985
) The influence of dolichol, dolichol esters, and dolichyl phosphate on phospholipid polymorphism and fluidity in model membranes.
J. Biol. Chem.
 ,
260
(5),
2742
–2751.
Vigo, C., Grossman, S.H., and Drost-Hansen, W. (
1984
) Interaction of dolichol and dolichyl phosphate with phospholipid bilayers.
Biochim. Biophys. Acta
 ,
774
(2),
221
–226.
Wakamatsu, K., Choung, S.Y., Kobayashi, T., Inoue, K., Higashijima, T. and Miyazawa, T. (
1990
) Complex formation of peptide antibiotic Ro09-0198 with lysophosphatidylethanolamine: 1H NMR analyses in dimethyl sulfoxide solution.
Biochemistry
 ,
29
(1),
113
–118.
Weiner, S.J., Kollman, P.A., Case, D.A., Singh, C., Ghio, C., Alagona, G., Profeta, S. and Weiner, P. (
1984
) A new force field for molecular mechanical simulation of nucleic acids and proteins.
J. Am. Chem. Soc.
 ,
106
,
765
–784.
Wuthrich, K. (
1987
) NMR techniques for studies of protein-DNA interactions. In Thompson, E.B. and Papaconstantinou, J. (eds.), DNA: Protein interactions and gene regulation. University of Texas Press, Austin, pp. 87–94.
Yeagle, P.L., Danis, C., Choi, G., Alderfer, J.L. and Albert, A.D. (
2000
) Three dimensional structure of the seventh transmembrane helical domain of the G-protein receptor, rhodopsin.
Mol. Vision
 ,
6
(1),
125
–31.
Yeagle, P.L., Salloum, A., Chopra, A., Bhawsar, N., Ali, L., Kuzmanovski, G., Alderfer, J.L. and Albert, A.D. (
2000
) Structures of the intradiskal loops and amino terminus of the G-protein receptor, rhodopsin.
J. Peptide Res.
 ,
55
(6),
455
–465.
Zhu, X.Y. and Lehrman, M.A. (
1990
) Cloning, sequence, and expression of a cDNA encoding hamster UDP-GlcNAc:dolichol phosphate N-acetylglucosamine-1-phosphate transferase.
J. Biol. Chem.
 ,
265
(24),
14250
–14255.
Zhu, X., Zeng, Y., and Lehrman, M.A. (
1992
) Evidence that the hamster tunicamycin resistance gene encodes UDP-GlcNAc:dolichol phosphate N-acetylglucosamine-1-phosphate transferase.
J. Biol. Chem.
 ,
267
(13),
8895
–8902.
Zimmerman, J.W. and Robbins, P.W. (
1993
) The hydrophobic domain of dolichyl-phosphate-mannose synthase is not essential for enzyme activity or growth in Saccharomyces cerevisiae.
J. Biol. Chem.
 ,
268
(22),
16746
–16753.