Nonspecific lipid transfer proteins (ns-LTP1) form a multigenic protein family in plants. In vitro they are able to bind all sort of lipids but their function, in vivo, remains speculative. A ns-LTP1 isolated from wheat seed was crystallized in the presence of lyso-myristoyl-phosphatidylcholine (LMPC). The structure was solved by molecular replacement and refined to 2.1 Å resolution to an R-factor of 16.3% and a free R-factor of 21.3%. It reveals for the first time that the protein binds two LMPC molecules that are inserted head to tail in a hydrophobic cavity. A detailed study of the structure leads to the conclusion that there are two lipid-binding sites, one of which shows a higher affinity for the LMPC than the other. Comparison with other structures of lipid-bound ns-LTP1 suggests that the presence of two binding sites is a general feature of plant ns-LTP1.
Nonspecific lipid transfer proteins (ns-LTPs) are ubiquitous lipid binding proteins in plants. Two main groups of ns-LTPs, ns-LTP1 and ns-LTP2, have been identified with molecular masses of about 9 and 7 kDa, respectively. ns-LTP1 are the best characterized group and have a basic pI and a conserved cysteine pattern. They form a multigenic family whose genes are spatially and temporally regulated . In vivo, ns-LTP1 are synthesized with a N-terminal signal peptide, which indicates that they follow the secretory pathway, in agreement with their location in extracellular layers, i.e. cell walls or cutin, and in vacuolar structures [1, 2]. Such locations are consistent with neither a role in membrane biogenesis nor with a role in the β-oxidation pathway of fatty acids. Sterk et al.  suggested that they are involved in the formation of cutin layers by transporting hydrophobic cutin monomers. Finally, the anti-microbial activities of ns-LTP1 observed in vitro suggest a possible role in the defence of plants against bacterial pathogens. Although these two latter hypotheses are attractive and consistent with most of the biological data, there is no clear evidence for such roles in vivo (for an extensive review, see  and the references cited therein).
Structural biology is a complementary approach to the search for the biological function of proteins. Determination of three-dimensional structures of ns-LTP1 has shown that these proteins share a common fold, stabilized by four disulphide bonds, and are characterized by a four helix bundle covered partly by a long C-terminal arm with several turns. The loose folding of the hydrophobic side chains in the protein core can produce a cavity, even in the absence of lipids. Such cavities have been seen in wheat, barley, rice and maize ns-LTP1 whose structures were determined by either X-ray crystallography or multidimensional NMR or both [5–9]. This cavity is able to bind different types of monoacylated and diacylated lipid molecules, including fatty acids [6, 10, 11], fatty acyl CoA [12, 13], lyso-phosphatidylcholine (LPC) [7, 14] and phosphatidylglycerol . Therefore these proteins can be unambiguously assigned to this group of lipid binding proteins. The presence of this cavity should constitute an important feature for their biological function. It is noteworthy that in Ace-AMP1, an anti-microbial protein from onion seed with a ns-LTP1 fold, no cavity is present and the protein cannot transfer or bind any lipids. However, this protein interacts with lipidic membranes and changes their permeability . The core of Ace-AMP1 has many aromatic side chains including two tryptophan residues which are not found in other ns-LTP1. Previously, the structure of a hydrophobic protein from soybean containing the same disulphide bond pattern revealed a similar four helix fold but did not show a hydrophobic cavity as in the ns-LTPs . These divergences in lipid/protein interactions for proteins sharing the same folding pattern could indicate that ns-LTP1 have different biological functions.
A study of the relationship between the structure of the hydrophobic cavity and the lipid binding properties of the protein should bring useful information on the functionality of ns-LTP1. The volume of the cavity is variable for the lipid-free protein and can increase when large lipids are bound. This increase may or may not be accompanied by an increase in the protein size [9, 13, 15]. Furthermore, a comparison of the structures of barley ns-LTP1 complexed with acyl CoA  and maize ns-LTP1 complexed with palmitic acid  has revealed opposite orientations of the lipid molecules. In this work, the determination of the structure of wheat ns-LTP1 complexed with two lyso-myristoyl-phosphatidylcholine (LMPC) molecules provides new insight into understanding lipid binding and ns-LTP1 function.
Materials and methods
Purification of wheat ns-LTP1
Wheat bran (1 kg) was extracted with 5 L of deionized water. After filtration through a Büchner funnel and centrifugation at 5000 g for 20 min, the extract was loaded on a column (5 × 30 cm) packed with a cation-exchange resin (Streamline, Pharmacia). The fractions were eluted by applying a gradient from 0 to 0.7 m NaCl in 20 mm Mes pH 5.6 buffer. ns-LTP1 was detected by electrophoresis and immunoblotting as described in . ns-LTP1-enriched fractions were pooled, dialysed overnight against deionized water, and freeze-dried. The dry material was solubilized and loaded on a gel filtration column (3 × 100 cm) packed with Sephadex G50 and eluted in 20 mm Mes pH 5.6 buffer. Finally, ns-LTP1 was purified from the enriched fractions by semipreparative C18 reversed-phase HPLC on a 5-µm, 300-Å bonded silica column (25 × 1 cm) using a gradient of water/acetonitrile/0.05% trifluoroacetic acid (1% acetonitrile per min) at 50°C and freeze dried. The purity of the protein was checked by analytical HPLC and by mass spectrometry as described in . The binding of LMPC to the purified protein was checked by fluorescence measurements .
Crystallization of wheat lipid transfer protein was only possible in the presence of a phospholipid. Previous crystals obtained in the presence of ammonium sulphate  were not reproducible constantly and therefore not suitable for structure refinement. New crystal forms were obtained using the hanging drop method from a protein solution at 6 mg·mL−1 with a reservoir containing 30% (w/v) polyethylene glycol 4000 and 0.2 m ammonium acetate in 0.1 m citrate buffer, pH 5.6, in the presence of LMPC (4 lipids per protein monomer) at 20°C. The lyophilized protein was solubilized in the lipid containing solution prior to crystallization. Several lipid to protein ratios were tested, ranging from 1 to 10 lipids per protein monomer; the ratio of 4:1 led to the best crystals. Similarly, cocrystallization with three different lipid species was attempted: LMPC, lyso-palmytoyl-phosphatidylcholine and di-heptanoyl-phosphatidylcholine, the choice of LMPC being dictated by crystal quality. Two crystal forms were obtained in the same drops: monoclinic P21 with cell parameters a = 48.30 Å, b = 41.90 Å, c = 51.10 Å, β = 113.3°, two monomers per asymmetric unit; and orthorhombic P212121 with cell parameters a = 42.29 Å, b = 53.55 Å, c = 71.49 Å, two monomers per asymmetric unit. Only the latter crystal form was used for solving and refining the structure.
Structure determination and refinement at 2.1 Å resolution
A first model was obtained by molecular replacement from the orthorhombic crystal form and X-ray diffraction data to 2.6 Å resolution. It was then refined to 2.1 Å resolution using data collected with a better crystal. Data were collected at room temperature on a FAST detector mounted on a Elliot GX21 X-ray generator using a copper rotating anode (Enraf-Nonius). Data were reduced using the madnes package , the procor program  and the CCP4 package . Data collection statistics are summarized in Table 1. The structure was solved by molecular replacement using a model of maize ns-LTP1 solved by X-ray diffraction at the laboratory (unpublished data) and the amore package . The self-rotation function indicated the presence of two monomers in the asymmetric unit. They were located using the data between 10 and 3.5 Å resolution, with a correlation of 41.0% and an R-factor of 45.2%. The structure was refined up to 2.6 Å with the program xplor 3.8.1  and the model was progressively improved by alternating manual corrections from 2Fo-Fc and Fo-Fc maps (program o), with energy minimization and simulated annealing (up to 3000 K) using the parameters as in Engh and Huber . Noncrystallographic symmetry restraints were maintained between the main chain atoms of the two molecules of the asymmetric unit. The 2.6 Å model was then refined using the 2.1 Å data set with a bulk solvent correction (free R-factor set of reflections same as in the 2.6 Å data set and extended to the high resolution). All restraints were relaxed between the two monomers. Water molecules were gradually added. Individual B-factors (temperature factors) were refined. The phases were improved using the ARP program  and lipid molecules appeared clearly in the Fo-Fc density. Two lipids per monomer were gradually built and refined from the coordinates and geometric constraints of the crystallographic structure of lyso-phosphatidyl-ethanolamine . The final model at 2.1 Å resolution contains 1626 nonhydrogen atoms, including 1356 protein atoms, 124 lipid atoms and 145 water atoms. Refinement statistics are summarized in Table 1 (R-factor = 16.3%, free R-factor = 21.3%). The Ramachandran plot calculated with the program procheck for both monomers shows no residues in the forbidden areas and 94.5% of the residues in the most favoured areas. The mean B-factor calculated for all atoms in the asymmetric unit (22.5 Å2), is comparable to the value obtained from the Wilson plot (20.5 Å2). Figure 1 was drawn with the program molscript and Figs 2, 4, and 6 were drawn with the program grasp.
Table 1. Data processing and refinement statistics.
Resolution limits (Å)
Number of measurements
< I/σ >
Native 2.6 Å
Native 2.1 Å
aRsym and R-factors are defined as follows: Rsym = ΣhΣi|I(h)i<I(h)>|/ΣhΣiI(h)i, where I(h) is the reflection h, Σh is the sum over all reflections and Σi is the sum over the i measurements of reflection h, R-factor = Σ|Fobs–Fcalc|/ΣFobs, where Fobs and Fcalc are the observed and calculated structure factor amplitudes, respectively. The reflection set for the free R-factor contains 5% of the total set of unique reflections.
The asymmetric unit is composed of two monomers of 90 residues. Two molecules of LMPC are bound per monomer. The wheat ns-LTP1 consists of four helices, H1 (residues 3–17), H2 (residues 25–37), H3 (residues 41–55), and H4 (residues 63–73) and a long C-terminal loop LCter (residues 74–90) ( Fig. 1). The helices were defined using the program DSSP . They are linked by three loops, L1-2 (residues 18–24), L2-3 (residues 38–40) and L3-4 (residues 56–62). Helix H1 is significantly bent around Pro13 and helix H4, at Ile69, before the doublet Pro70-Pro71. LCter is a long winding loop made of two successive turns (residues 74–80 and residues 81–90) maintained close to the core of the structure by the disulphide bridge Cys48-Cys87. The structure contains two pairs of disulphide bridges located by pairs at opposite sides of the protein ( Fig. 1). Electron density maps reveal clear unique conformations for the disulphide bridges Cys13-Cys27, Cys28-Cys73 and Cys48-Cys87, while two alternate conformations are observed for the Cys3-Cys50 disulphide bond. In all loops, the electron density is well defined and atomic B-factors for the main chain atoms are below 30 Å2. The protein is best described as a helix bundle with H1/H2 and H3/H4 (including LCter) forming two antiparallel helix sets, surrounding a strikingly large tunnel which runs through the protein ( Fig. 2) covered predominantly with hydrophobic residues: Val10 and Cys13 from helix H1, Cys27, Val31 and Leu34 from helix H2, Ala38 from loop L2–3, Ala47, Cys48, Leu51, Lys52 and Ala55 from helix H3, Leu61 from loop L3–4, Ala66, Arg67, Ile69 and Pro70 from helix H4 and Val75, Leu77, Tyr79, Ile81, Ser82, Leu83 and Ile85 from loop LCter. The two extremities of the tunnel are exposed to the solvent and include both hydrophobic or hydrophilic residues: Asp7, Arg11, Leu14, Ile54 and Ile58 on one side, and His35, Arg44, Pro78 and Val90 on the other side. The two monomers of the asymmetric unit, which were refined independently, have similar conformations with root mean square (rms) deviations of 0.32 Å for the main chain atoms and 0.9 Å for all atoms.
Binding of LMPC
In both monomers (A and B), the lipid molecules correspond to well defined electron densities which show unambiguously the presence of two alkyl chains complexed to each monomer ( Fig. 3). Lipid 1 (A1 and B1 for monomers A and B, respectively) and lipid 2 (A2 and B2) are positioned head to tail, crossing the protein through the hydrophobic tunnel ( Fig. 4). The aliphatic chains are buried inside the tunnel. The polar head groups are exposed to solvent areas, between LCter and the beginning of H3 for lipid 1 and between L3–4 and the end of H1, for lipid 2. We notice that the choline group is only visible in lipid B1 ( Fig. 3). This is consistent with the observed B-factors. In the aliphatic chains, the atomic B-factors are similar to the mean B-factor of the protein ( Fig. 5). In the polar head groups, they are much higher, especially in lipids A2 and B2, increasing from the carbonyl to the choline, with maximum values of 80–100 Å2. Choline groups are likely to be multiconformational. The comparison between the two monomers A and B of the asymmetric unit shows that lipids in site 1 (A1 and B1) superimpose all along the aliphatic chain up to the phosphate group ( Fig. 4). Lipids in site 2 (A2 and B2) superimpose only along the aliphatic chains with a 1-Å translation. There are more lipid/protein interactions in site 1 than in site 2. This is highlighted by several hydrophobic interactions in site 1 with His35, Cys48, Leu51, Ala55, Ala66, Asn76, Leu77, Pro78, Ile81, Ser82, Leu83, Val90, and by a hydrogen bond between the carbonyl of the ester bond in LMPC and the hydroxy group of Tyr79. In site 2, only a few hydrophobic interactions are observed with Val10, Leu14, Val31, Ile54, Ile58 and His59.
Lipid binding in wheat ns-LTP1
Our crystallographic study reveals for the first time that wheat ns-LTP1 binds two molecules of LMPC. In the present structure, the cavity described in the lipid-free protein  becomes a tunnel formed by the displacement of helix H1 and loop LCter. Both lipids are located head to tail inside the hydrophobic core with their aliphatic chains in the highly hydrophobic part of the tunnel and their polar head groups directed towards the solvent areas, at each end of the tunnel. In site 1, LMPC strongly interacts with the protein through hydrophobic interactions and through a hydrogen bond with Tyr79. In site 2, LMPC is involved only in few hydrophobic interactions. The absence of hydrogen bond linking lipid 2 to the protein, as well as the lower number of hydrophobic interactions suggest that lipid 2 has a lower affinity for the protein than lipid 1. These findings are in agreement with fluorescence experiments which show that wheat ns-LTP1 can bind more than one lipid molecule per protein . Fluorescence experiments under resonance energy transfer conditions suggest that the maize LTP is also able to bind two lipids in the cavity .
Lipid binding in other ns-LTP1
Several lipid-free or lipid-bound structures of ns-LTP1 have previously been determined by NMR spectroscopy or X-ray diffraction. The rms differences between these structures and the present one are listed in Table 2 and show high similarities between the structures, although a systematic difference is observed in the loop LCter from residue 76 to the end. This allows the protein structure to adjust to the presence of different lipids by modifying the position of the loop LCter. The size of the cavity increases in order to accommodate the binding of one or two lipid molecules.
Table 2. Comparison with other ns-LTP1 structures. The numbers indicate the rms displacement (Å) calculated with program lsqkab on main chain atoms, and in parentheses on all atoms. References are shown in square brackets.
Comparison with wheat ns-LTP1 bound to di-myristoyl-phosphatidyl-glycerol (DMPG). The solution structure of wheat ns-LTP1 complexed with DMPG has been determined by NMR . Comparison with the present structure shows that the orientation of DMPG corresponds to site 1 of LMPC with the polar head group lying in a similar position ( Fig. 6a). In the NMR structure, the Tyr79 side chain is directed towards the solvent, preventing the formation of a hydrogen bond between this residue and the lipid. This is in contradiction with our structure, as well as with the maize ns-LTP1 structure complexed with palmitate , in which a hydrogen bond links the hydroxyl group of Tyr81 to the carboxyl group of the fatty acid. As in the NMR structures, aliphatic chains of DMPG were modelled without any experimental constraints ; this may explain the discrepancies in the localization of DMPG and LMPC aliphatic chains.
Comparison with maize ns-LTP1 bound to palmitate. In the crystal structure of maize-ns-LTP1 complexed with palmitate , the palmitate has its polar head group in site 1, close to the ester group of LMPC, and its aliphatic chain in site 2 ( Fig. 6b). Similarly to our structure, the polar head group forms a hydrogen bond with Tyr81. This indicates that this tyrosine, which is highly conserved in ns-LTP1 sequences could play a role in lipid binding to site 1.
Comparison with barley ns-LTP1 bound with palmitoyl coenzyme A (PCoA). In the structure of PCoA-bound-barley ns-LTP1 determined by NMR , PCoA is oriented according to site 2 and its palmitoyl aliphatic chain is strongly curved inside the hydrophobic pocket, extending out of site 2 ( Fig. 6c). The polar head of PCoA is directed towards site 2, is highly disordered and shows multiple conformations. Both PCoA in barley ns-LTP1, and LMPC (site 2) in wheat ns-LTP1, only interact through hydrophobic contacts with the protein.
Our present study shows for the first time, that a ns-LTP1 is able to bind two monoacylated lipids inserted head to tail in the hydrophobic cavity. This structural analysis highlights the existence of two lipid binding sites of different affinity for lipids, with more lipid–protein interactions in site 1 than in site 2. Orientations of the lipid molecules which are found to be opposite in barley and maize lipid-bound structures are shown here to be both possible. This explains why only one LPC molecule was observed in previous NMR studies of maize ns-LTP1 . A large excess of lipid would have been necessary to saturate site 2. However, in barley ns-LTP1 complexed with PCoA, steric constraints and additional interactions between PCoA and the protein are more favourable to the binding of PCoA in site 2.
ns-LTP1 function has not been clearly determined yet. In this regard, it is interesting to note that some structural and binding properties are shared between plant ns-LTP1 and serum albumin, a protein involved in the transport of fatty acids in the blood of mammals. Serum albumin is a protein with a higher molecular mass than ns-LTP1 but each helical domain encloses a tunnel which is the binding site for fatty acid. Furthermore as in ns-LTP1, one of these domains can bind two fatty acid molecules and the lipid carboxylate is stabilized by both hydrogen and electrostatic interactions with arginine, lysine and tyrosine side chains . Therefore, these similarities and the data obtained in this work sustain the role of plant ns-LTP as a transporter of fatty acids or fatty acid derivatives (e.g. hydroxy-fatty acids, acyl-CoA) necessary during the formation of cutin layers or during lipid mobilization in germinating seedlings, the current hypothesis for the biological function of plant ns-LTP .
We are grateful to Mogens Lehmann for having initiated this work and for useful discussions during this study, to Richard Kahn and Emile Duée for helpful advice. We thank Christine Saint Pierre and Eric Forest for mass spectrometry analysis. We also thank Denize Sy, Patrick Sodano, Françoise Vovelle and Marius Ptak for providing unpublished coordinates of the solution structures of lipid-free and lipid-bound wheat ns-LTP1 and for stimulating comments.
Note: the novel atomic coordinates and structure factors have been deposited with the Brookhaven Protein Data Bank (Accession numbers: Id 1bwo and 1bwosd).