3024
Biomacromolecules 2008, 9, 3024–3032
Molecular Dynamics Simulation of the Cooperative Adsorption of Barley Lipid Transfer Protein and cis-Isocohumulone at the Vacuum-Water Interface S. R. Euston,*,†,‡ P. Hughes,†,‡ Md. A. Naser,†,‡,§ and R. E. Westacott§ School of Life Sciences, International Centre for Brewing and Distilling, and School of Engineering and Physical Sciences, Heriot-Watt University, Edinburgh, EH14 4AS, United Kingdom Received April 21, 2008; Revised Manuscript Received July 25, 2008
Molecular dynamic simulations have been carried out on systems containing a mixture of barley lipid transfer protein (LTP) and cis-isocohumulone (a hop derived iso-R-acid) in one of its enol forms, in bulk water and at the vacuum-water interface. In solution, the cis-isocohumulone molecules bind to the surface of the LTP molecule. The mechanism of binding appears to be purely hydrophobic in nature via desolvation of the protein surface. Binding of hop acids to the LTP leads to a small change in the 3-D conformation of the protein, but no change in the proportion of secondary structure present in helices, even though there is a significant degree of hop acid binding to the helical regions. At the vacuum-water interface, cis-isocohumulone shows a high surface activity and adsorbs rapidly at the interface. LTP then shows a preference to bind to the preadsorbed hop acid layer at the interface rather than to the bare water-vacuum interface. The free energy of adsorption of LTP at the hopvacuum-water interface is more favorable than for adsorption at the vacuum-water interface. Our results support the view that hop iso-R-acids promote beer foam stability by forming bridges between separate adsorbed protein molecules, thus strengthening the adsorbed protein layer and reducing foam breakdown by lamellar phase drainage. The results also suggest a second mechanism may also occur, whereby the concentration of protein at the interface is increased via enhanced protein adsorption to adsorbed hop acid layers. This too would increase foam stability through its effect on the stabilizing protein layer around the foam bubbles.
Introduction The primary stabilizing factor of fluid interfaces in foods and beverages is usually an adsorbed layer of protein or peptide.1 Often these are found in combination with other ingredients such as polysaccharides2,3 and low molecular weight surfactants.1 These can interact synergistically with the proteins to stabilize the interface further or, in some circumstances, can reduce interfacial stability via competitive adsorption.1 Of particular interest to the food emulsion formulator are the low molecular weight emulsifiers, such as Tweens, Spans, monoglycerides, phospholipids, and so on. These are used to aid formation of small droplets during homogenization of food emulsions, which are subsequently stabilized by more slowly adsorbing proteins.1 Low molecular weight emulsifiers have been shown to compete for interfacial area with proteins and, if they are present at a high enough concentration, can displace protein from the surface, thus leading to interface destabilization.1 In alcoholic beverages various proteins and peptides derived from malted barley have been found to play a role in the formation and stability of the foam of the head of beer.4,5 The best characterized of these are protein Z, a 40 kDa protein found in barley, and barley lipid transfer protein (LTP), a 9.7 kDa peptide.4-9 It has been found that in beer foam the hydrophobic bittering agents derived from hops, the iso-R-acids, also have a role in controlling foam stability. Hughes and Wilde10 hypothesize that the iso-R-acids interact with hydrophobic proteins and peptides, such as LTP, and form a cross-link between different * To whom correspondence should be addressed. Tel.: +44 131 451 3640. Fax: +44 131 451 3009. E-mail:
[email protected]. † School of Life Sciences. ‡ International Centre for Brewing and Distilling. § School of Engineering and Physical Sciences.
proteins at the interface. This creates a film that is more resistant to foam drainage, thus improving foam stability. They also found that trans-isomers of isocohumulone and isoadhumulone were more surface active than cis-isomers, although they did not determine whether this led to differences in foam stability. In this paper, we report molecular dynamics simulation results for a mixture of barley LTP and cis-iso-cohumulone and for the coadsorption of LTP and cis-iso-cohumulone at the vacuum-water interface. This is an extension of our earlier studies on LTP adsorption at both vacuum- and decane-water interfaces.11,12 Our aim is to throw light on the role of hop isoacids in controlling structure and stability of adsorbed protein layers of the LTP molecule.
Simulation Methodology All simulations were carried out using the GROMACS13 molecular dynamics package. Two sets of simulations were carried out. In the first simulation, a single LTP molecule was simulated in a water box in the presence of 24 cis-isocohumulone molecules. cis-Isocohumulone is one of the most abundant hop acids in beer, although its effect on beer foam stability is less than that of the less abundant transisocohumulone. To generate a molecular topology for the cis-isocohumulone, the molecule was drawn using the Chem3D software package. The cis-isocohumulone was defined as one of its enol tautomeric forms (Figure 1), which represents the tendency for one of the carbonyl groups in the molecule to readily tautomerize into the enolate anion in solution. A number of enol tautomers are possible. We have chosen the one that has the lowest energy and is likely to be the most stable in solution. Hydrogen atoms were added to that structure using the open babel file format conversion package, and the structure was crudely minimized using the semiempirical quantum mechanical method PM3.14 The stability of the different enol tautomers of cis-
10.1021/bm8004325 CCC: $40.75 2008 American Chemical Society Published on Web 10/09/2008
Molecular Dynamics Simulation of LTP Adsorption
Figure 1. Enol tautomer of cis-isocohumulone.
isocohumulone was compared after minimization, and the lowest energy form was taken as the starting structure. This structure was inserted at the center of a box of 1 × 1 × 1 nm3. The structure was further minimized under the GROMOS force field15 using a quasi-Newtonian followed by a steepest descent algorithm. A total of 27 of the minimized hop acid structures were then stacked in a 3 × 3 × 3 arrangement in a 6 × 6 × 6 nm3 water box (with the water density scaled to 1000 g/L). The distance between hop acid in the boxes was 1 nm. A single LTP was inserted into the middle of the box replacing three hop acid molecules. The starting structure was the NMR structure determined by Heinemann et al.6 deposited in the Brookhaven Protein database as 1lip. Counterions were added to the simulation cell to neutralize the charge on the protein. Addition of hop acids required removal of overlapping water molecules. The final number of water molecules in the simulation cell was 6801, and there were 24 cis-isocohumulone molecules. The box was then used to carry out molecular dynamics simulation for 2 ns to check the stability of the system, and having checked the energy conservation and radial distribution function and density of water, a full 20 ns simulation run was carried out. In the second simulation the coadsorption of LTP and cis-isocohumulone at the water-vacuum interface was simulated. To form the water-vacuum interface in this system, a box of size 6 × 6 × 6 nm3 with water and 24 hop acids was expanded 7 nm on each side in the z-direction, while maintaining periodic boundary conditions at the x and y faces. The system was run for 2 ns using MD to stabilize the interface. During this time the hop acids partitioned themselves in the interfacial region. A single LTP molecule was added close to the center of the cell, with the precise position being determined partly by the need to avoid overlap with hop acid molecules, and the simulation was then run for a final 20 ns. Again, counterions were added to neutralize the charge on the protein. The final number of water molecules in the simulation cell was 6799. Interactions were modeled using the OPLS all-atom force field.16 Electrostatic interactions were calculated using the particle mesh Ewald summation method. A switching function was used to smoothly decay the Lennard-Jones potential to zero after a certain cutoff distance. A cutoff of 0.9 nm was used, with the interactions decaying exponentially to zero at 1.2 nm. The system was coupled to a Berendsen thermostat17 to maintain the temperature at 25 °C. Simulations were run for up to 20 ns. No attempt was made to parametrize a force field for the hop acid explicitly. Rather, we used the OPLS force field to model interactions between hop acids, hop acid and water and protein and hop acid. As for the protein, the Ewald summation method was used to model electrostatic interactions, with the same switching parameters for Lennard-Jones interactions as defined above. Assignment of partial charges to the hops acid molecule was carried out according to the parameters defined for each atom in the OPLS force field. Nor was any attempt made to test the validity of the hop acid model by simulating hop acid systems to check density, or heat of solvation since experimental data for comparison is not readily available from the literature. Thus, at present the model for hop acid remains untested.
Biomacromolecules, Vol. 9, No. 11, 2008
3025
The choice of the LTP and hop concentrations for the simulation was complicated by several factors. The concentration of protein and iso-R-acids in beer has been measured as between 20-250 mg/mL for protein18 and 17-55 µg/mL for iso-R-acids.19 However, both protein and hop acid will be found preferentially in the lamella of the foam bubbles in the head of beer, and thus the concentrations are likely to be much higher than this under these conditions. We are not aware of any studies that have measured the elevated concentration of protein and hop acid in the foam. Therefore, any choice of protein and hop acid concentration for our simulations will be an educated guess. We need at least one LTP molecule in the simulation. We are also limited in system size to one that is large enough to give meaningful results, but not so large as to give prohibitively long simulations. We have calculated that given our system size the concentration of LTP in our simulation box is about 75 mg/mL. This is in the middle of the experimental concentration range,18 but may be somewhat low given our expectation that protein will be found preferentially in the foam lamellae close to the air-water interface. Similarly, we have calculated that the concentration of cis-isocohumulone in our simulation is about 70 mg/mL, which is several orders of magnitude higher than the measured total iso-R-acid concentration of beers. However, this may not be an unreasonable concentration given the affinity of the hop acids for the interfacial region. The structures of the simulated adsorbed proteins were characterized in several ways. Changes in the secondary structure were analyzed by determining the number of amino acid residues existing in helices, β-sheet, or in random coils as a function of simulation time, using Deep View/Swiss-PdbViewer.20 Changes in tertiary structure were estimated by following the atom mass weighted root-mean-square deviation (RMSD) and the radius of gyration (Rg). The RMSD for conformations during the simulation was calculated relative to the starting native conformation. Density profiles for protein, water, and hop-acid across the simulation cell were determined. The change in the number of hydrogen bonds (protein-protein and protein-water) was also calculated depending on the system. The free energy of adsorption for the protein was estimated using the probability ratio method.21 The distance between the center of mass of the protein and the interface in the z-direction was calculated. The z-coordinate dimension was divided into a series of intervals (or bins) of width ∆r. The frequency fi of the protein being in a bin with a specific width ∆ri was calculated and the probability was calculated as21
pi )
fi
with
∑f
∑p )1 i
(1)
i
The normalized probability density is
P′i )
pi ∆ri
with
∑ P′ ∆r ) 1 i
i
(2)
From this normalized probability density, the free energy of adsorption was calculated according to the following relation:
∆Gi ) Gi - G0 ) -RTln
[ ] P′i P′0
(3)
where R is the ideal gas constant and T is absolute temperature. The subscript 0 defines a reference state for a given molecular system. In these systems the reference state was taken as the starting position of the protein in the z-direction. A theoretical analysis of cis-isocohumulone binding to LTP was carried out using the PatchDock server.22 This utilizes an algorithm that finds docking sites between two molecules (in our case cisisocohumulone and LTP) in a highly efficient manner.22 PatchDock finds docking sites by searching for molecular shape complementarity between a protein and a small molecule. This is achieved by representing both molecules as a Connelly dot surface, and dividing this into concave, convex and flat patches. Complementary patches are matched and then evaluated and scored according to geometric fit and desolvation of the surface. The final decision on the accuracy of the docking is
3026
Biomacromolecules, Vol. 9, No. 11, 2008
Euston et al.
Figure 3. Conformation of LTP/hop acid complex in a water box showing the position of the lipid binding pocket. The highlighted hop acid molecule is bound close to the entrance to the binding pocket. The letters A-D represent the helices in LTP as identified by Heinemann et al.6
Figure 2. (a) Snapshot conformation for native LTP in a water box in the presence of 24 cis-isocohumulone molecules after 20 ns of an MD run. (b) Composite of 24 PatchDock19 generated conformations of hop acid binding to native LTP. Both LTP conformations have been rotated to give approximately the same orientation so as to highlight the similarity in cis-isocohumulone binding for the two methods. The letters A-D represent the helices in LTP as identified by Heinemann et al.6
assessed by RMSD clustering.22 We have used this program to calculate likely surface binding sites for the hop acid onto the surface of native LTP. The Q-site web server (http://bmbpcu36.leeds.ac.uk/qsitefinder/) was used to identify potential binding sites in the LTP structure and to calculate the volume of the binding pocket. The methodology23 calculates the interaction energy between regions of a target protein and a simple van der Waals probe. This information is used to identify energetically favorable binding sites. Binding pockets are identified by clustering favorable probe sites according to their spatial proximity. The putative binding pockets are then ranked according to the sum of interaction energies for sites within each cluster.
Results and Discussion Figure 2a is a snapshot conformation for a system of one LTP molecule and twenty four cis-isocohumulone molecules in a water box after 20 ns of an MD simulation. We find that
as the simulation progresses the hop acids self-associate with each other and then bind to the surface of the LTP molecule (Figure 2a; a separate simulation of cis-isocohumulone in water has been carried out and a snapshot conformation showing selfassociation of the hop acids is included as Supporting Information). This is not surprising because both LTP and cis isocohumulone is known to be hydrophobic (hops acids are often extracted from hops as an essential oil). From Figure 2a it appears that the cis-isocohumulone molecules interact preferentially with certain regions of the LTP molecule. Detailed inspection of the conformations suggests that the interaction does not occur in the binding pocket of the LTP molecule,9 but occurs at the surface of the protein. We can compare our simulated conformations with those predicted by the PatchDock server.22 Figure 2b is a composite picture of a single LTP molecule with the first 24 most likely (highest score) hop acid interaction conformations. This conformation has been rotated so as to give approximately the same LTP orientation as in Figure 2a. In Figure 2b some of the hop acid molecules overlap with each other, because each binding conformation is generated for an LTP-hop acid pair. It is obvious from Figure 2 that both simulated and predicted hop acid binding sites are in the same region of the molecule. The region of most likely binding of hop acids is a relatively thin band on the surface that extends over some half to two-thirds of the molecule. Large areas of the molecule have a reduced probability for hop acid binding. It has been hypothesized that during hop binding to LTP, water molecules solvating the amino groups of the polypeptide chain are replaced by the carbonyl and enolate groups of iso-R-acids, with binding of the ion-dipole type.24,25 The stability of the complex is increased further by hydrophobic interactions between the iso-R-acids side chains and hydrophobic region of polypeptides. Our simulated conformations in Figure 2 and more detailed analysis of the full MD trajectory for the system do not suggest that there is any interaction of hop acids with the internal hydrophobic binding pocket of LTP. Figure 3 highlights the amino acids that line the binding pocket in LTP, and it is apparent that no hop acid molecules occupy this binding pocket. It is possible that the hop acid is either too large, or has the wrong shape for it to fit the binding pocket. We do notice, however, a single hop acid molecule that is bound to the surface of the LTP molecule close to the binding pocket (highlighted
Molecular Dynamics Simulation of LTP Adsorption
Biomacromolecules, Vol. 9, No. 11, 2008
3027
Table 1. Identification of Amino Acid Interaction Sites for cis-Isocohumulone amino acid within a certain distance (Å) of the hop acid cis-isocohumulone identifier
