Homology modelling of an antimicrobial protein, Ace-AMP1 ...

Homology modelling of an antimicrobial protein, Ace-AMP1,from lipid transfer protein structuresJérôme Gomar, Patrick Sodano, Marius Ptak and Françoise Vovelle

Background: Plant nonspecific lipid transfer proteins (ns-LTPs) are small basicproteins that facilitate lipid shuttling between membranes in vitro. The function ofns-LTPs in vivo is still unknown. It has been suggested, in relation to their lipidbinding ability, that they may be involved in cutin formation. Alternatively, theymay act in the plant defence system against pathogenic agents. Ace-AMP1 is anantimicrobial protein extracted from onion seed that shows sequence homologywith ns-LTPs but that is unable to tranfer lipids. We have recently determined thethree-dimensional structure of wheat and maize ns-LTPs. In order to compare thestructural features of Ace-AMP1 and ns-LTPs, we have used the comparativemodelling software MODELLER to predict the structure of Ace-AMP1.

Results: The global fold of Ace-AMP1 is very similar to those of ns-LTPs,involving four helices and a C-terminal tail without secondary structure elements.The structure of maize and wheat ns-LTP is characterized by the existence of atunnel-like hydrophobic cavity in which a lipid molecule can be inserted. In theAce-AMP1 structure, this cavity is blocked by a number of bulky residues.Similarly, the electrostatic potential contours of ns-LTPs show some commonfeatures that were not observed in Ace-AMP1.

Conclusions: Although Ace-AMP1 displays a similar global fold to ns-LTPs, itdoes not present a hydrophobic cavity, which may explain why Ace-AMP1cannot shuttle lipids between membranes in vitro. The large differences in theelectrostatic properties of Ace-AMP1 and ns-LTPs suggest a different mode ofinteraction with membranes.

IntroductionAce-AMP1 is an antibiotic protein extracted from onionseed (Allium cepa L.) whose activity, unlike that of otherplant seed antimicrobial proteins, shows little or no reduc-tion at physiological concentrations of inorganic cations[1]. Ace-AMP1 shows sequence homology to plant nonspe-cific lipid transfer proteins (ns-LTPs), which facilitate theinter-membrane transfer of a variety of lipids in vitro [2].They form a group of low molecular weight (9 kDa)homologous proteins with a high isoelectric point. Com-parison of their amino acid sequences reveals a commonpattern of eight cysteines C--C--CC--CXC--C--C formingfour disulphide bridges [3]. The plant ns-LTPs exhibitaffinity for a broad range of phospholipids, fatty acids [4,5]and acyl-coenzyme A esters [6].

Recently, we determined the solution structure of wheat[7] and maize ns-LTPs [8] (NLTA_WHEAT andNLTP_MAIZE, respectively). At the same time, thecrystal structure of the maize ns-LTP was published [9]and detailed comparisons of NMR and X-ray structuresshowed only minor differences (J Gomar, unpublisheddata). More recently, the solution structure of a third plantns-LTP from barley seeds [10] has been reported

(NLT1_HORVU). The structures of these three plant ns-LTPs exhibit very similar global folds; they all includefour helical fragments connected by three loops and a C-terminal tail without regular secondary structure. Ahydrophobic pocket, in which a lipid can be inserted, runsthrough the whole molecule [7–9].

So far, the biological function of plant ns-LTPs in vivohas not been clearly established. On the basis of theirability to transfer lipids in vitro, they were assumed toplay a role in membrane biogenesis by mediating thetransport of lipids from their site of biosynthesis to othermembranes [2]. However, analysis of different cDNAsencoding plant ns-LTPs revealed the presence of a signalpeptide, suggesting the entry of ns-LTPs into the plantsecretory pathway [11–13]. Moreover, ns-LTPs havebeen found predominantly in the cell wall of epidermalcells [14–17]. This localization appears to be inconsistentwith a role in intracellular lipid trafficking. On this basis,ns-LTPs were proposed to be involved in the formationof cutin by transferring cutin monomers from their site ofsynthesis through the cell wall of epidermal cells to theirsite of polymerization [15,17]. It has also been shown thatseveral ns-LTPs are potent growth inhibitors of bacterial

Address: Centre de Biophysique Moleculaire etUniversité d’Orléans, rue Charles sadron, 45071Orléans, Cedex 2, France.

Correspondence: Françoise VovelleE-mail: [email protected]

Key words: antimicrobial protein, homology, lipidtransfer protein, molecular modelling

Received: 05 Feb 1997Revisions requested: 04 Mar 1997Revisions received: 24 Mar 1997Accepted: 11 Apr 1997

Published: 13 May 1997Electronic identifier: 1359-0278-002-00183

Folding & Design 13 May 1997, 2:183–192

© Current Biology Ltd ISSN 1359-0278

Research Paper 183

and fungal pathogens in vitro [1,18,19]. Hence, plant ns-LTPs could play a role in the plant defence system.However, not all ns-LTPs appear to possess antibioticactivities [1]. Wheat ns-LTP, for example, has no antimi-crobial activity, and maize ns-LTP shows only very weakantimicrobial or antifungal activity. ns-LTPs are notfound in all cell types, and in some plant species, differ-ent isoforms are expressed differently [13–16,20]. Allthese results are in agreement with the hypothesis thatdifferent types of ns-LTPs with different tissue speci-ficity (and presumably different function) may coexist ina given plant [19,21].

In this work, we are concerned with the determination ofthe 3D structure of Ace-AMP1. Although, Ace-AMP1 hasbeen classified as a member of the ns-LTP family on thebasis of sequence homologies, it is unable to transfereither phosphatidylcholine or phosphatidylinositol fromliposomes to mitochondrial membranes [1]. In thiscontext, it was important to compare the 3D structure ofAce-AMP1 to the structures of plant ns-LTPs of known invitro activity. Since these proteins have rather different

isoelectric points (9 for maize, 10 for wheat, and 8 forbarley ns-LTPs; 11.8 for Ace-AMP1), the comparison wasextended to molecular properties, such as 3D electrostaticpotential or hydrophobicity potential, that can be derivedfrom the structure and can be related to their function.Wheat and maize ns-LTPs and Ace-AMP1 are good candi-dates for such a study because their lipid transfer andantimicrobial activities in vitro were assessed in compara-tive tests [1]. The aim of this work was to complement ourrecent 3D structure determination of wheat and maize ns-LTPs [7,8] by building the structure of Ace-AMP1. Theprogram MODELLER [22] was used with the experimentallydetermined ns-LTP structures as template.

Results and discussion Sequence alignmentThe alignment of 13 ns-LTP proteins, including Ace-AMP1, is represented in Figure 1. A compilation of themost frequently found residues allows us to propose aconsensus sequence. This consensus sequence is some-what different from that previously proposed byDésormeaux et al. [3], since the number of protein

184 Folding & Design Vol 2 No 3

Figure 1

Amino acid sequence alignment of plant ns-LTPs and Ace-AMP1 (using the one-letter amino acid code). The last line indicates the consensussequence.

sequences considered in the present work is much largerthan in the previous study and the criteria used to estab-lish the consensus are not strictly identical. A comparisonof the Ace-AMP1 sequence with the ns-LTP sequencesshows that all cysteine and proline residues are conserved.Two prolines are structurally very important: Pro13 in Ace-AMP1 numbering induces a kink in the first helix andPro70 curves or stops (in wheat ns-LTP) the fourth helix,thus influencing the position of the C terminus withrespect to the rest of the molecule. In the first half of theAce-AMP1 sequence, most substitutions are conservativerelative to the ns-LTP sequences. In contrast, the asparticacid residue at position 42 (numbering according to theAce-AMP1 sequence) is replaced by a leucine. Consider-ing the structure of maize ns-LTP [8], Asp42 and Arg43,which are conserved among all ns-LTPs, are probablyinvolved in lipid binding. In the second half of the proteinsequence, many nonconservative substitutions areobserved, often involving the substitution of hydrophobicresidues by arginines. A tyrosine at position 81 that is con-served among the ns-LTP family and is probably alsoinvolved in the lipid binding mechanism is replaced by aphenylalanine in the Ace-AMP1 sequence.

When pairwise homologies are considered, high percent-ages of identities and similarities are found between thevarious ns-LTPs (Table 1). The scores are much lowerwhen comparisons are made between ns-LTPs and Ace-AMP1. The highest percentage of identity and similarityis found with maize ns-LTP (32 and 46%, respectively).

Structure calculations10 models of Ace-AMP1 were calculated by optimizing thetarget function from 10 different initial random conforma-tions. Nine of them converged to low values of the targetfunction F and have a good covalent geometry. Thesemodels are not significantly different, since the averageroot mean square deviation (rmsd) for the C� coordinatesis 1.95 Å ± 0.09. Among these structures, the model havingthe maximum number of residues falling in the mostfavoured regions of the Ramachandran plot was chosen tobe the representative Ace-AMP1 structure.

Ace-AMP1 model: comparison with the templatesPROMOTIF [23] and PROCHECK [24] analyses show that thepredicted structure of Ace-AMP1 (Fig. 2) includes four �-helices: H1 (6–18), H2 (25–37), H3 (42–56) and H4 (65–72).Due to the presence of the conservative residue Pro13, theH1 helix can be divided into two subhelices (6–12 and14–18), as the hydrogen bond network is broken betweenresidues 12 and 14. This feature has already been observedon the template structures, especially on the maize struc-ture [8]. In contrast, there is no marked effect of the prolineresidue at position 70. It induces a slight curvature of theH4 helix, as in maize and barley ns-LTPs [8,10]. In wheatns-LTP, two consecutive prolines at positions 70 and 71stop the H4 helix, which is therefore shorter than in theother known ns-LTP structures. Globally, one can observeslight fluctuations of the helix lengths over the different ns-LTPs, the largest differences occuring in H4. In Ace-AMP1,H1 is two residues shorter than in maize or barley ns-LTPs,

Research Paper Homology modelling of an antimicrobial protein Gomar et al. 185

Table 1

Pairwise homologies between ns-LTPs.

Ace- NLTP_ NLTP_ NLTP_ NLT1_ NLTA_ NLTP_ NLTP_ NLTP_ NLTP_ NLTA_ NLTB_ NLTC_AMP1 ORYSA MAIZE ELECO HORVU WHEAT LYCES TOBAC SPIOL DAUCA RICCO RICCO RICCO ⟨similarity⟩

Ace-AMP1 50% 46% 49% 49% 54% 50% 52% 45% 43% 52% 48% 50% 49%

NLTP_ORYSA 33% 92% 90% 80% 81% 80% 83% 84% 76% 76% 64% 62% 79%

NLTP_MAIZE 32% 79% 92% 78% 79% 84% 85% 83% 83% 76% 66% 68% 79%

NLTP_ELECO 33% 73% 76% 74% 74% 80% 81% 82% 75% 78% 68% 64% 75%

NLT1_HORVU 31% 65% 57% 48% 91% 70% 70% 67% 69% 66% 55% 61% 69%

NLTA_WHEAT 29% 63% 59% 53% 73% 70% 71% 69% 67% 67% 59% 58% 66%

NLTP_LYCES 29% 56% 56% 51% 53% 49% 93% 86% 77% 67% 58% 64% 74%

NLTP_TOBAC 26% 54% 55% 54% 51% 51% 84% 85% 77% 66% 61% 63% 70%

NLTP_SPIOL 26% 56% 56% 54% 45% 46% 67% 69% 73% 72% 62% 64% 68%

NLTP_DAUCA 27% 52% 57% 51% 48% 43% 58% 56% 55% 65% 55% 58% 59%

NLTA_RICCO 27% 47% 50% 45% 39% 41% 44% 45% 44% 41% 87% 63% 75%

NLTB_RICCO 26% 43% 44% 39% 38% 39% 40% 41% 39% 41% 70% 60% 60%

NLTC_RICCO 27% 31% 35% 29% 31% 28% 34% 32% 33% 28% 39% 38%

⟨identity⟩ 29% 56% 55% 47% 47% 42% 55% 49% 43% 37% 55% 38%

while H2 and H3 have approximately the same length andH4 is intermediate between maize and wheat structures.The angles between the helices, as calculated with PROMO-TIF, are compatible (within 20°) with the values obtainedfor ns-LTPs [8]. As observed in ns-LTP structures, the Cterminus of Ace-AMP1 is formed by a series of turns. Forexample, a type IV �-turn is formed between residues 87and 90 and is stabilized by a O(87)–N(90) hydrogen bond.Other turns stabilized by hydrogen bonds are found in theloops connecting the helices, such as a �-turn betweenresidues 40 and 42.

The superposition of the C� traces of the template struc-tures and the predicted Ace-AMP1 structure (Fig. 3a–c)clearly shows that there is a better agreement between thesecondary structure elements than between the variableloops and the C-terminal regions. Table 2 gives the pair-wise rmsds of the C� coordinates along the sequence forthe whole backbone (residues 3–92 in the consensussequence numbering) and thus quantifies the visualimpression given by Figure 3. The smallest rmsds arefound for the residues located in the helical regions. Thelargest deviations concern the C terminus, for which fewspatial restraints were generated by MODELLER since thestructural sequence alignment is not efficient in thisregion, which also lacks secondary structures. For compari-son, a superposition of the C� trace of maize, wheat andbarley ns-LTPs with Ace-AMP1 is represented inFigure 3d. This figure and Table 2 show clearly that thedifferences between Ace-AMP1 and the three ns-LTPs arelarger than the differences between the ns-LTP structures.

EvaluationThe stereochemical quality of the predicted model asevaluated by PROCHECK compares favourably with thetemplates: about 72% of the residues fall in the most

favourable region of the Ramachandran plot and 24% inthe additional regions. The covalent geometry of the Ace-AMP1 model is good — there are no deviations from theideal values for valence angles exceeding 10° and none ofthe planar groups deviates for more than 0.04 Å from pla-narity. Only three bond lengths deviate by more than0.05 Å from the ideal values.

A simple way to test the quality of a structure is to inspectthe structure visually on a graphic display. As mentionedin the Materials and methods section, in the earliermodels generated with MODELLER, two arginines wereorientated toward the protein core and so were not acces-sible to the solvent. Examination of the alignmentshowed that Arg6 (in Ace-AMP1) corresponds to a buriedglutamine (Gln6) in maize ns-LTP. Ace-AMP1 shows ahigher percentage of homology with maize ns-LTP thanwith barley and wheat ns-LTPs. Therefore, constraintsissuing from maize ns-LTP had an extra weight com-pared to others, which may explain the ill position of theArg6 sidechain. With respect to Arg35, the difficultiesarise from alignment uncertainties. Compared to Ace-AMP1, the three template sequences have an additionalresidue in the H2 helix. Therefore, the alignment ofArg35 is presumably not optimal. This leads us to recon-sider the alignment in this region (from Arg35 to Arg43).The effect on the structure modelling was a lowering ofthe weight allocated to the template structures, leading tolarger conformational freedom of the end of helix H2 andloop L2. In the same way, the alignment in the C-termi-nal region has been reconsidered, except for a fewresidues (cysteines), since many nonconservative substi-tutions appear in this region.

The accuracy of the model was tested using 3D profile pro-grams (PROSA and 3D1D profiles). PROSA [25,26] calculates


Figure 2

A Molscript [49] stereoview of the global foldof Ace-AMP1.

the conformational energy of each residue of the proteinfrom a compilation of potentials of mean force for all aminoacid pairs. The energy profiles of Ace-AMP1 are negative,corresponding to a correctly folded structure. The com-bined pair interactions–surface energy z-score [27] calcu-lated for Ace-AMP1 (z-score = – 4) is in good agreementwith the expected value for a protein of about 100 residues[27]. This value is, however, much higher than the valuesobtained for the X-ray or NMR structures of the maizeprotein (around –7.5).

The 3D1D profile method [28] is based on empirical statis-tic scores. It calculates the local compatibility between aresidue and its structural environment. The global scorefor the sequence of Ace-AMP1 (S = 28) is below the valuescalculated for the X-ray and NMR structures of maize ns-LTP (S = 45 and 42, respectively), but is in agreementwith a compilation of values obtained for correctly foldedproteins of the same length [29]. The 3D profile plots ofthe Ace-AMP1 model (data not shown) are always largerthan zero, which indicates that most residues lie in acorrect environment.

Hydrophobic cavityThe main feature of the maize and wheat ns-LTP struc-tures is the existence of an elongated cavity runningthrough the whole molecule [7,8] (Fig. 4a,b). This cavityextends from loop L2 to loop L1 and is globally parallelto the H3 helix. In a previous study, we showed that alysolecithine can be inserted in the pocket without majorrearrangements of the protein structure [8]. In order todetermine whether such a hydrophibic cavity is a generalstructural feature related to the special fold of these pro-teins, we looked for cavities in the structures of barley


Figure 3

Superposition of the Ace-AMP1 global foldwith ns-LTP folds from (a) maize, (b) wheat,(c) barley and (d) all three. Ace-AMP1 is incyan, maize ns-LTP in green, wheat ns-LTP inyellow and barley ns-LTP in magenta.

Table 2

Pairwise rmsds between ns-LTPs and Ace-AMP1.

Maize ns-LTP Wheat ns-LTP Barley ns-LTP

Ace-AMP1 3.3 (3.1) 3.1 (2.8) 3.1 (2.9)

Maize ns-LTP 2.2 (2.0) 1.8 (1.6)

Wheat ns-LTP 1.8 (1.6)

Pairwise rmsd values calculated on the C� atoms for the whole backbone(and, in brackets, rmsds calculated on the C� atoms of the four helices).

ns-LTP and Ace-AMP1 using a home-made program(J Gomar, unpublished data). Since the size and shape ofthe cavity depend on the coordinate data set used, we ana-lyzed the NMR structure previously selected for molecu-lar modelling. The calculations confirm our previousresults, in that there is an elongated cavity in the core ofmaize and wheat ns-LTP which is closed at one extremity

by a conserved tyrosine. The estimated volumes are280 Å3 and 380 Å3 for maize and wheat ns-LTPs, respec-tively. In barley ns-LTP, only a very narrow cavity can bedetected (Fig. 4c). Its shape is also elongated and it islocked at one extremity by Tyr81 (in Ace-AMP1 number-ing). Its volume is only 180 Å3, which is not sufficientlylarge to accommodate a lipid. In the Ace-AMP1 structure,we were unable to locate a continuous cavity. The bulkyresidues Trp82, Phe66 and Phe81, whose sidechainsplunge into the protein core, prevent the formation of alarge and continuous cavity (Fig. 4d).

These four proteins, although sharing a similar global 3Dfold, clearly exhibit cavities of different shapes and sizes.Maize and wheat ns-LTPs, which show lipid transfer andlipid binding activities in vitro, display a large cavity inwhich a lipid can be inserted. It is now essential to system-atically study the relationships that may exist between thebinding ability of ns-LTPs and their lipid transfer activity.

Electrostatic potentialIn order to transfer lipids from a donor membrane to anacceptor membrane, ns-LTPs have to approach the mem-brane, to interact with its hydrophilic surface, and presum-ably to insert into the bilayer accessing the hydrophobicenvironment of the lipid acyl chains. Until now, nothingwas known about the ns-LTP–membrane interaction.Electrostatic forces may play an important role both ininducing the approach of the membrane and in stabilizingthe complex formed by the protein and the lipids. In thisstudy, the electrostatic contours of the three ns-LTPs andAce-AMP1 were calculated with the DELPHI program [30]using Amber 4.0 charges. This set of charges was cali-brated to fit the electrostatic potentials from ab initioquantum mechanical calculations [31]. Since electrostaticforces are acting at a distance, we found it to be moreinformative to look at the isopotential contours than at thepotential distribution at the molecular surface. Isocontoursat the level +1.25 and –1.25 kJ mol–1 i.e. ± 0.5 kT at roomtemperature were drawn (Fig. 5) with the GRASP program[32] on a face of the protein including the N terminus,loop L2 and the C terminus. Globally, the four proteins,except for barley ns-LTP, are surrounded by a positivepotential. The shape of the positive contours and the dis-tribution of the negative surface contours are differentamong the four proteins under study and are more or lessrelated to the global charge of the proteins: +17e, +6e, +5eand +2e for Ace-AMP1, maize, wheat and barley ns-LTPs,respectively. As expected, numerous negative patches arefound in the barley protein in the vicinity of the nega-tively charged residues (Asp and Glu), but also in thedirection of the two histidine residues His35 and His58. Inthe PDB file (PDB code 1lip), these two histidines are notprotonated, which is surprising under the NMR experi-mental conditions (pH 4.0 [10]). Introduction of positivecharges on the two histidines modifies the electrostatic


Figure 4

A stereoview of the hydrophobic cavities of (a) maize ns-LTP, (b) wheat ns-LTP, (c) barley ns-LTP and (d) Ace-AMP1.

surface contours (Fig. 5) and allows us to find commonfeatures between the three ns-LTPs. The similaritiesbetween the electrostatic contours are related to con-served residues even though the orientation of theseresidues in the three proteins is different. One canobserve one or two positive areas between loop L2 and theC-terminal tail in the neighbourhood of conservedarginines (Arg44 and Arg89 for wheat and barley proteinsand Arg46 and Arg91 for maize). This region is surroundedby three negative patches. The first one is due to the C-terminal COO- (Val90, Tyr91 and Asn93 for wheat, barleyand maize ns-LTPs, respectively) and the two others toconserved aspartic acid residues (Asp43 and Asp86 forwheat and barley and Asp45 and Asp88 for maize).Depending on the protein considered, the patches aremore or less extended. In maize ns-LTP, for example, thenegative area due to the C-terminal residue is very smalland there is only one large positive contour due to Arg46and Arg91. On the opposite side of the proteins, the nega-tive areas correspond to nonconserved residues and havedifferent positions in the three ns-LTPs.

In contrast to the ns-LTPs, Ace-AMP1 (Fig. 5f) is sur-rounded only by a positive electrostatic potential, as thetwo aspartic acid residues Asp33 and Asp72 are not able tocompensate the effect of the 19 arginines. One can findonly two negative spots on an overall positive surface onone side of the molecule (the opposite side to L2).

For comparison, we have also calculated (Fig. 5c) the elec-trostatic potential for the maize ns-LTP complexed with apalmitate molecule using the crystallographic coordinates[9]. The contours are very similar to the contours of the

uncomplexed form except for a region situated betweenL2 and the C terminus, where a large negative band isfound. This negative band is exactly situated in the regionwhere similarities are found in the electrostatic potentialof the three ns-LTPs.

Since we know very little about the lipid transfer or antibi-otic activities of barley ns-LTP, it is premature to drawstructure/function relashionships. However, it is clear thatthe electrostatic potentials of the three ns-LTPs showcommon features that do not appear on the Ace-AMP1 con-tours. The similarities are restricted to the region situatedbetween loop L2 and the C terminus which is affectedupon lipid binding. This polar region is located at oneextremity of the hydrophobic cavity and one can assumethat it is the interaction site of ns-LTPs with the lipids.

Hydrophobicity potential calculations with the MOLCAD

option of the SYBYL software package [33] show that themolecular surface of the three proteins is hydrophilic. Theonly hydrophobic patches found at the surface of the ns-LTP molecules are located at both extremities of thehydrophobic cavity. As expected, the internal surface ofthe cavity is hydrophobic.

ConclusionsEven though Ace-AMP1 and ns-LTPs display relativelyfew sequence identities and have different functionalproperties in vitro, they show a similar global fold.However, the packing of the sidechains of both classes ofproteins is dissimilar, leading to different organizations ofthe protein core. Both maize and wheat ns-LTPs, whichshow lipid transfer activity in vitro, display a hydrophobic


Figure 5

Electrostatic potential contours. Themolecular surface is shown in white. Theisopotential surfaces for values of 0.5 kT e–1

and –0.5 kT e–1 are coloured in blue and red,respectively. (a) Orientation of the molecules.Compared to Figure 3, the orientationcorresponds to a rotation of 90° of themolecule with respect to a horizontal axis.Loop L3 is in the middle of the picture. (b) Maize ns-LTP. (c) Maize ns-LTPcomplexed with a palmitate molecule.(d) Wheat ns-LTP. (e) Barley ns-LTP. (f) Ace-AMP1.

cavity in the core of their structure. In Ace-AMP1, whichdoes not transfer lipids, the cavity is strongly reduced anddiscontinuous and its size is too small to accommodate alipid. Our results provide evidence for a relationshipbetween the lipid transfer properties of ns-LTPs and theexistence of a hydrophobic cavity. This conclusion should,however, be confirmed by the study of other ns-LTPs.

The electrostatic properties of Ace-AMP1 and ns-LTPs areclearly different. The electrostatic potential is mainly posi-tive around the Ace-AMP1 structure, while negative zonesappear in ns-LTPs. The distribution of electrostatic poten-tials on one face of the ns-LTPs shows similar trends in awell-defined region, which may constitute the interactionsite with the lipid layer. Such similarities are not found forAce-AMP1 and suggest, therefore, that Ace-AMP1 and thens-LTPs interact differently with membranes.

Such considerations may help in the design of mutagene-sis experiments. Before this, the efficiency of MODELLER

should be evaluated on Ace-AMP1 by an NMR study. Inthis work, our model was based on only 32% identitieswith the templates, assuming a similar disulphide networkbetween Ace-AMP1 and ns-LTPs. Usually, a high identityis required to get a precise 3D model of a protein withunknown structure. Regarding the limits of alignmentalgorithms, our manual intervention in optimizing thisalignment, and the assumption that Ace-AMP1 structure isclosely related to ns-LTPs, we have to confirm that ourmodel corresponds to the actual fold of this protein.However, MODELLER [34] has been shown previously topredict correctly 90% of mainchain atoms of the humaneosinophil neurotoxin on the base of low sequence identi-ties to the template (33%).

A new question arises from this study: is Ace-AMP1 amember of the ns-LTP family? As long as the true physio-logical role of these proteins remains unclear, it is difficultto answer this question. Ace-AMP1 and ns-LTPs couldhave issued from a common ancestral gene and theirsequences diverged during evolution. Alternatively, theycould have no functional relationships, and one couldassume that the global fold of these small proteins with ahigh content of cysteines is under the control of the char-acteristic network of disulphide bridges.

Materials and methodsSequence alignmentIn a first step, a multiple sequence alignment was generated with MUL-TALIGN included in the ANTHEPROT package [35] in order to establish aconsensus sequence. 13 different plant ns-LTPs were considered:Ace-AMP1; rice ns-LTP from Oryza sativa seed [36], Swiss-Prot codeNLTP_ORYSA; barley ns-LTP from Hordeum vulgare seedlings [37],NLT1_HORVU; tomato ns-LTP from Lycopersicon esculentum seed[38], NLTP_LYCES; maize ns-LTP from Zea maize seedlings [11],NLTP_MAIZE; Indian finger millet ns-LTP from Eleusine coracana seed[39], NTLP_ELECO; wheat ns-LTP from Tricium aestivum seed [3],

NLTA_WHEAT; tobacco ns-LTP from Nicotiana tabacum flowers [40],NLTP_TOBAC; the isoforms of castor bean ns-LTPs from Ricinuscommunis seed [41], referred to as NLTA_RICCO and NLTB_RICCOand the isoform from Ricinus communis seedlings [42],NLTC_RICCO; carrot ns-LTP from Daucus carota [15],NLTP_DAUCA; and spinach ns-LTP from Spinacia oleraceae leaves[13], NLTP_SPIOL.

Model buildingAce-AMP1 models were built on the basis of three template proteins,namely wheat and barley ns-LTP solution structures and the crystalstructure of maize ns-LTP, which does not differ very much from theNMR model. The various solution structures of wheat and barleyprotein being not significantly different, we selected the minimumenergy conformations of each protein as templates for model building.

Comparative molecular modelling is based on a sequence alignmentbetween the target molecule and the templates. In this work, an itera-tive procedure was used to align Ace-AMP1 with the three templates.In a first step, a structural alignment of the three ns-LTPs was per-formed [22] using SUPERIMPOSE [43]. Two residues were considered tobe aligned when their C�–C� distance was less than a given cut-off(5 Å in this work). Ace-AMP1 and the template proteins were thenaligned using MULTALIGN [35]. Preliminary Ace-AMP1 3D structureswere derived from these alignments. The sequences were subse-quently aligned by a least-square superposition of a Ace-AMP1selected structure and of the template structures using the dynamicprogramming method implemented in MODELLER [22]. The cycle of re-alignment, modelling and structure evaluation was repeated until nofurther improvement on the structure was observed. In the course ofthe modelling study, the alignments were manually edited to optimizethe quality of the derived models.

MODELLER is a modelling program based on the satisfaction of spatialrestraints. Distance and dihedral angle restraints are generated on thetarget sequence from its alignment with the template 3D structures andalso from the statistical analysis of a database including 416 knownproteins structures [22,34]. These restraints in combination withenergy terms enforcing the proper covalent geometry are expressedinto a variable objective function [44] which is optimized using simu-lated annealing and conjugate gradient methods. Disulphide bridgerestraints are automatically generated by the program once the disul-phide bond pairs in the sequence are specified. From sequencehomologies, four S–S bridges, 4–49, 14–27, 28–73 and 47–89, maybe expected in Ace-AMP1. In the course of the modelling process,after evaluation (see below) of the models, several automaticallyderived restraints were removed and the structures were manuallyedited. The sidechains of several arginines located in helices (espe-cially Arg6 and Arg35) were often found in an unfavourable position, onthe hydrophobic side of the two amphiphilic helices H1 and H2. In thiscase, the sidechain atoms were removed and Nilges protocol [45,46]used to reposition the atoms. The arginine sidechain atoms were posi-tioned by a simulated annealing at 1000 K using XPLOR software [47].High harmonic constraints were applied on all other heavy atoms. Thisstep was followed by 5 ps of constrained standard molecular dynamicsat 300 K. The average structure issuing from the molecular dynamicswas then minimized using CHARMM force field.

Additional restraints, such as NMR restraints, can be specified by theuser. In the course of the modelling study, hydrogen bond constraintswere added on the �-helices detected by PROMOTIF analysis. Thisallows us to improve the helix geometry.

Evaluation of the model10 models were generated by varying the initial structures, randomlygenerated in the cartesian coordinates space. A good model corre-sponds to a low value of the objective function. The PROCHECK program[24] was used to assess the stereochemical quality of the models.Additional evaluations were done by various 3D profile programs,


PROSA [25] and 3D1D [29], and also by examination of the structures ona graphic display. A representative model was selected and submittedto further structural analysis and comparisons with other plant ns-LTPs.This was performed using PROCHECK [24] and PROMOTIF [23] and home-made programs.

Electrostatic potential calculations and representationsThe electrostatic potential was calculated by the DELPHI program [30].The linear Poisson–Boltzmann equations were solved using on eachprotein atom Amber fractional charges [48]. The dielectric constantwas set to 2 inside the protein and to 80 outside. The electrostaticpotential contour surfaces were displayed using GRASP [32].

AcknowledgementsWe are grateful to Willem Broekaert for his helpful remarks. This work wassupported by the Centre National de la Recherche Scientifique (CNRS), theRégion Pays de Loire and the Université d’Orléans. It forms part of JérômeGomar’s thesis, supported by a BDI grant.

References1. Cammue, B.P.A., et al., & Broekaert, W.F. (1996). A potent antimicro-

bial protein from onion (Allium cepa L.) seeds showing sequencehomology to plant lipid tranfer proteins. Plant Physiol. 109, 445–455.

2. Kader, J.C. (1990). Intracellular transfer of phospholipids, galactolipidsand fatty acid in plant cells. In Subcellular Biochemistry. (Hilderson,H.J. ed). pp. 69–111, Plenum Publishing Corporation.

3. Désormeaux, A., Blochet, J.E., Pezolet, M. & Marion, D. (1992). Aminoacid sequence of a non-specific wheat phospholipid transfer proteinand its conformation as revealed by infrared and raman spectroscopy.Role of disulfide bridges and phospholipids in the stabilization of the�-helix structure. Biochim. Biophys. Acta. 1121, 137–152.

4. Rickers, I., Tober, I. & Spener, F. (1984). Purification and binding char-acteristics of a basic fatty acid binding protein from Avena sativaseedlings. Biochim. Biophys. Acta 794, 313–319.

5. Yamada, M., Tsuboi, S., Osafune, T., Suga, T. & Takishima, K., (1990).Multifunctional properties of non specific lipid transfer proteins fromhigher plants. In Plant Lipid Biochemistry, Structure, Function andUtilisation. (Quinn, P.J. & Harwoods, J.L., eds), pp 278–280, PortlandPress, London.

6. Ostergaard, J., Vergnolle, C., Schoentgen, F. & Kader, J.C. (1993).Acyl-binding/lipid transfer protein from rape seedlings, a novel cate-gory of proteins interacting with lipids. Biochim. Biophys. Acta 1170,109–117.

7. Gincel, E., Simorre, J.P., Caille, A., Marion, D., Ptak, M. & Vovelle, F.(1994). Three-dimensional structure in solution of a wheat lipid-tranferprotein from multidimensional 1H-NMR data. Eur. J. Biochem. 226,413–422.

8. Gomar, J., et al., & Ptak, M. (1996). Solution structure and lipid bindingof a non specific lipid transfer protein extracted from maize seedling.Protein Sci. 5, 565–577.

9. Shin, D.H., Lee, J.Y., Hwang, K.Y., Kim, K.K. & Suh, S.W. (1995). High-resolution crystal structure of the non-specific lipid-transfer proteinfrom maize seedlings. Structure 3, 189–199.

10. Heinemann, B., Andersen, K.V., Nielsen, P.R., Bech, L.M. & Poulsen,F.M. (1996). Structure in solution of a four helix lipid binding protein.Protein Sci. 5, 13–23.

11. Tchang, F., et al., & Kader, J.C. (1988). Phospholipid transfer protein:full length cDNA an amino acid sequence in maize. J. Biol. Chem. 263,16849–16855.

12. Vergnolle, C., et al., & Kader, J.C. (1988). Synthesis of phospholipidtransfer protein from maize seedlings. Biochem. Biophys. Res.Commun. 1, 37–41.

13. Bernhard, W.R., Thoma, S., Botella, J. & Somerville, C.R. (1991). Isola-tion of a cDNA clone for spinach lipid transfer protein and evidencethat the protein is synthesized by the secretory pathway. Plant Physiol.95, 164–170.

14. Sossountzov, L., et al., & Kader, J.C. (1991). Spatial and temporalexpression of a maize lipid transfer protein gene. The Plant Cell 3,923–933.

15. Sterk, P., Booij, H., Schellekens, G.A., Van Kammen, A. & De Vries,S.C. (1991). Cell specific expression of the carrot EP2 lipid transferprotein gene. The Plant Cell 3, 907–921.

16. Thoma, S., Hecht, U., Kippers, A., Botella, J., De Vries, S. & Som-merville, C. (1994). Tissue-specific expression of a gene encoding a

cell wall-localized lipid transfer from Arabidopsis. Plant Physiol. 105,35–45.

17. Meijer, E.A., De Vries, S.C., Sterk, P., Gadella, D.W., Wirtz, K. & Hen-driks, T. (1993). Characterization of the non-specific lipid transferprotein EP2 from carrot (Daucus carota). Mol. Cell Biochem. 123,159–166.

18. Terras, F.R.G., et al., & Broeckaert, W.F. (1992). Analysis of two novelclasses of plant antifungal proteins from radish (Raphanus sativus L.)seeds. J. Biol. Chem. 267, 150301–15309.

19. Molina, A., Segura, A. & Garcia-Olmedo, F. (1993). Lipid transfer pro-teins (nsLTPs) from barley and maize leaves are potent inhibitors ofbacterial and fungal pathogens. FEBS Lett. 316, 119–122.

20. Fleming, A.J., Mandel, T., Hofman, S., Sterk, P., De Vries, S.C. & Kuhle-meier, C. (1992). Expression pattern of a tobacco lipid transfer proteingene within the shoot apex. Plant J. 2, 855–862.

21. Takishima, K., Watanabe, S., Yamada, M., Suga, T. & Mamiya, G.(1988). Amino acid sequences of two non-specific lipid transfer pro-teins from germinated castor bean. Eur. J. Biochem. 177, 241–249.

22. Šali, A. & Blundell, T.M. (1993). Comparative protein modelling by sat-isfaction of spatial restraints. J. Mol. Biol. 234, 779–815.

23. Hutchinson, E.G. & Thornton, J.M. (1996). PROMOTIF. A program toidentify and analyse structural motifs in proteins. Protein Sci. 5,212–220.

24. Laskowski, R.A., MacArthur, M.W., Moss, D.S. & Thornton, J.M.(1993). PROCHECK: a program to check the stereochemical qualityof protein structures. J. Appl. Crystallogr. 26, 283–291.

25. Hendlich, M., et al., & Sippl M.J. (1990). Identification of native proteinfold amongst a large number of incorrect folds. J. Mol. Biol. 216,167–180.

26. Sippl, M.J. (1993). Recognition of errors in three-dimensional struc-tures of proteins. Proteins 17, 355–362.

27. Sippl, M.J. (1993). Boltzmann’s principle, knowledge-based meanfields and protein folding. An approach to the computational determi-nation of protein structures. J. Comput. Aided Mol. Des. 7, 473–501.

28. Bowie, J.U., Lüthy, R. & Eisenberg, D. (1991). A method to identifyprotein sequences that fold into a known three-dimensional structure.Science 253, 164–170.

29. Lüthy, R., Bowie, J.U. & Eisenberg, D. (1992). Assessment of proteinmodels with 3D-profiles. Nature 356, 83–85.

30. Gilson, K., Sharp, K.A. & Honig, B. (1987). Calculating electrostaticinteractions in biomolecules: method and error assessment. J. Comp.Chem. 9, 327–335.

31. Weiner S.J., Kollman, P.A., Nguyen, D.T. & Case, D.A. (1986). Anapproach to computing electrostatic charges for molecules. J. Comp.Chem. 7, 230–252.

32. Nicholls, A., Sharp, K.A. & Honig, B. (1991). Protein folding and asso-ciation: insight from the interfacial and thermodynamic properties ofhydrocarbons. Proteins 11, 281–296.

33. Heiden, W., Goetze, T. & Brickman, J. (1993). Fast generation of mole-cular surfaces from 3D data fields with enhance marching cube algo-rithm. J. Comp. Chem. 14, 246–251.

34. Šali, A. (1995). Modelling mutations and homologous proteins. Curr.Opin. Biotechnol. 6, 437–451.

35. Geourjon, G., Deleage, G. & Roux, B. (1991). ANTHEPROT: an inter-active graphic software for analysing protein structures fromsequences. J. Mol. Graph. 9, 188–190.

36. Yu, Y.G., Chung., C.H., Fowler, A. & Suh, S.W. (1988). Amino acidsequence of a probable amylase/protease inhibitor from rice seeds.Arch. Biochem. Biophys. 265, 466–475.

37. Mundy, J. & Rogers, J.C. (1986). Selective expression of a probableamylase protein inhibitor in barley aleurone cells: comparison to thebarley amylase/ subtilisin inhibitor. Planta 169, 51–63.

38. Torres-Schuman, S., Gogoy, J.A. & Pintor-Toro, J.A. (1992). A proba-ble lipid transfer gene is included by NaCl in stems of tomato plants.Plant Mol. Biol. 18, 749–757.

39. Campos, F.A.P. & Richardson, M. (1984). The complete amino acidsequence of the �-amylase inhibitor I-2 from seeds of ragi (Indianfinger millet, Eleusine corona Gaert.). FEBS Lett. 167, 221–225.

40. Masuta, C., Furuno, M., Tanaka, H., Yamada, M. & Koiwai, A. (1992).Molecular cloning of a cDNA clone for tobacco lipid transfer proteinand expression of the functional protein in Escherichia coli. FEBS Lett.311, 119–123.

41. Takishima, K., Watanabe, S., Yamada, T. & Mamiya, G. (1986). Theamino acid sequences of the non specific lipid transfer protein fromgerminated castor bean endosperm. Biochem. Biophys. Acta 870,248–255.

42. Tsuboi, S., et al., & Yamada, M. (1991). Organ-specific occurrence,


expression of the isoforms of non-specific lipid transfer protein incastor bean seedlings and molecular cloning of a full length cDNA fora cotyledon specific isoform. J. Biochem. 110, 823–831.

43. Diederich, K. (1995). Structural superimposition of proteins withunknown alignment and detection of topological similarity using a sixdimensional search algorithm. Proteins 23, 187–195.

44. Braun, W. & Go, N. (1985). Calculation of protein conformations byproton–proton distances constraints. A new efficient algorithm. J. Mol.Biol. 186, 611–626.

45. Nilges, M. & Brünger, A.T. (1991). Automated modelling of coiledcoils: application to the GCN4 dimerization region. Protein Eng. 4,649–659.

46. Nilges, M. & Brünger, A.T. (1993). Successful prediction of the coiledcoil geometry of the GCN4 leucine zipper domaine by simulatedannealing, comparison to the X-ray structure. Proteins 15, 133–146.

47. Brünger, A.T. (1992). The X-PLOR Software Manual Version 3.1. YaleUniversity, New Haven C.T.

48. Singh U.C. & Kollman P.A. (1984). An approach to computing electro-static charges for molecules. J. Comp. Chem. 5, 129–145.

49. Kraulis, P.J. (1991). MOLSCRIPT: a program to produce both detailedand schematic plots of proteins structures. J. Applied Crystallogr. 26,283–291.


Because Folding & Design operates a ‘Continuous PublicationSystem’ for Research Papers, this paper has been publishedvia the internet before being printed. The paper can beaccessed from http://biomednet.com/cbiology/fad.htm — forfurther information, see the explanation on the contents pages.

Homology modelling of an antimicrobial protein, Ace-AMP1 ...

Documents

Transcript of Homology modelling of an antimicrobial protein, Ace-AMP1 ...