J. Phys. Chem. B 2007, 111, 10965-10969
10965
Short-Range Structure of a GM3 Ganglioside Membrane: Comparison between Experimental WAXS and Computer Simulation Results Marcello Sega and Renzo Vallauri* INFM-CNR and Department of Physics, UniVersity of Trento, Via SommariVe 14, I-38050 PoVo, Trento, Italy
Paola Brocca and Laura Cantu` Department of Chemistry, Biochemistry and Biotechnologies for Medicine, UniVersity of Milan, Via F.lli CerVi 93, I-20090 Segrate, MI, Italy and INFM
Simone Melchionna SOFT-INFM-CNR and UniVersity of Rome “La Sapienza”, P.le A. Moro 2, I-00185 Rome, Italy ReceiVed: April 11, 2007; In Final Form: June 12, 2007
The local structure of a GM3 ganglioside bilayer, whose wide-angle X-ray spectrum is reconstructed from molecular dynamics simulations, is found to compare quantitatively well with the experimental one. By separating inter- and intramolecular contributions, correlations between distinct head groups are shown to contribute in a substantial way to the total scattering intensity. This finding supports the hypothesis of a strong local head group order as recently formulated on the basis of calorimetry and X-ray experimental data.
1. Introduction During the past decade gangliosides, an interesting class of biological amphiphiles, have been the subject of many investigations, since they are found to play an active role in many fundamental processes taking place in the outer plasma membrane of eukaryotic cells.1-3 The hydrophilic moiety of gangliosides is composed of a variable number of saccharidic rings and carries a net negative charge because of the presence of sialic acid residues, thus presenting a very strong hydrophilic character. The nonsugar moiety, which accounts for the hydrophobic nature of gangliosides, is a ceramide double tail. As is typical for amphiphiles in solution, gangliosides above a critical concentration (of the order of 10-8-10-9 M) organize themselves in a variety of different structures. The properties of the aggregates have been extensively studied by means of various experimental techniques such as NMR, calorimetry, neutron, light, and X-ray scattering (see for example refs 4-7). Three main features distinguish gangliosides from phosphocholines, one of the most studied class of amphiphiles: (a) the presence of ceramide as the hydrophobic portion, (b) a bulky head group, which largely contributes to the whole molecular hindrance, and (c) its ionic nature. In particular, the size of the head group is expected to play a key role in determining the global shape of the aggregate, in sharp contrast to phosphocholines, and in determining nontrivial effects on the short-range spatial correlations. As a matter of fact, gangliosides form aggregates of very different shapes and dimensions, ranging from globular micelles to vesicles, according to the hydrophobichydrophilic balance of the molecule. More complicated bicontinuous structures can also be formed in an appropriate range of concentration.8 Ganglioside GM3 (in Figure 1), the subject of the present investigation, is of physical interest because is * To whom correspondence
[email protected].
should
be
addressed.
E-mail:
able to form spontaneously vesicles in solution at high dilution5,7 and, to increase the surface curvature with concentration, forming ribbons (paper in preparation). On the few-nanometers scale GM3 aggregates are of the bilayer type. Recently, Wide Angle X-Ray Scattering (WAXS) experiments on GM3 aggregates have been carried out with the aim of gathering information on the short range molecular organization.9 On the basis of an unexpected complexity of the measured WAXS spectra, as compared to the case of phosphocholines, and of their sensitivity to temperature variations, it has been proposed that the structuring of the GM3 head groups occurs at the bilayer surface, strongly contributing to the formation of a globally ordered short-range structure.9 As a computational study, we have undertaken for the first time extensive molecular dynamics simulations of a GM3 ganglioside bilayer.10 The good agreement previously obtained between the measured SAXS spectra and the reconstructed scattering pattern from the simulation data has proved the reliability of the implemented force field.10 In the present paper we extend further our study to the investigation of the local structure of the simulated GM3 bilayer by comparing simulated and experimental WAXS data. Scattering pattern reconstruction relative to a bulk hydrocarbon system (pentadecane) has also been used as a reference system to investigate the source of the observed discrepancies. 2. Simulation Details The GM3 bilayer has been simulated at atomistic detail, using a force field derived from the GROMOS87 one.11 Several modifications have been introduced to improve the force field reliability in reproducing various properties, such as the hydrophobic tails hindrance and the conformation of saccharidic rings of the head group. Both these changes and the procedure employed to create the starting configuration and bringing the system to equilibrium have been thoroughly described in a
10.1021/jp072834a CCC: $37.00 © 2007 American Chemical Society Published on Web 08/24/2007
10966 J. Phys. Chem. B, Vol. 111, No. 37, 2007
Sega et al.
Figure 1. Chemical structure of the ganglioside GM3. The subunits of the head group are sialic acid (NeuAc), galactose (Gal), and glucose (Glc). The double tail of ganglioside GM3, ceramide, is composed of a sphingosine and a fatty acid.
previous paper.10 In particular, the Lennard-Jones parameters of the aliphatic chains have been taken from ref 12. The simulations have been performed using the GROMACS molecular dynamics package13,14 at constant temperature (333 K) and pressure (1 atm) via the Berendsen weak-coupling algorithm.15 The Ewald method, in the form of the particle-mesh Ewald algorithm,16 has been employed for the treatment of electrostatic interactions. The relative strength of the Coulomb interaction at the cutoff was set to 10-5, the mesh spacing to 0.12 nm and the spline order to 4. The long-range correction to the pressure for the Lennard-Jones interactions have not been included explicitly, therefore, following ref 12, an appropriate rescaling of all distances has been applied a posteriori on the configurations used for data analysis. The system consisted of 128 GM3 molecules, together with the corresponding Na+ counterions, and 6724 SPC water molecules. After the equilibration run, which took about 35 ns, the production run has been carried on for another 15 ns, sampling the quantities of interest every 100 ps. Analysis of the computed deuterium order parameter, which gives a measure of the orientational order of the C-C bonds of the tails,17 showed clearly that the bilayer is structured in the LR, that is, disordered liquid crystalline, phase. 3. Data Analysis The X-ray scattering intensity is defined as the Fourier components of the electronic density,18
I(q) ) 〈|
∑ j,m
fj(q) exp(iq‚rj,m)|2〉
Figure 2. GM3 Experimental WAXS (solid line) and simulated (dashed line) scattering intensities I(q). The experimental and simulated bulk water spectra have been subtracted from the corresponding curves.
intermolecular contributions to the scattering intensity, which are directly accessible in simulation. By allowing the index m in eq 1 to run over a specific moiety, one can easily calculate its contribution to the total scattering. In our case, this is of particular interest, because it allows a separate investigation of the local structuring of the head and tail groups. The intra- and intermolecular contributions to the scattering intensity are defined by
Iintra(q) ) 〈
(1)
where the index m identifies a particular molecule, and j identifies an atom belonging to that molecule, having indicated the atomic positions by rj,m. The symbol 〈...〉 indicates an ensemble average together with an average over all possible orientations of the sample or, equivalently, of the wavevector q. The symbol fj(q) represents the form factor of atom j and is calculated via the Cromer-Mann formula, which fits the atomic scattering data measured experimentally.18 It is a common practice in presenting WAXS experimental data to appropriately subtract the spectra arising from the vessel and the aqueous solvent, so to obtain a pattern representative of the solute (GM3, in our case) scattering intensity. To compare the experimental and simulation spectra, we have similarly removed the contribution of the solvent from our numerical data by subtracting the one obtained from a separate simulation of bulk water at the same thermodynamic condition. Finally, the resulting spectrum has been scaled by a global factor to match with the height of the experimental peak. Let us remark that the pattern obtained in this way arises from the sum of the scattered radiation of the GM3 molecules and GM3-water cross-contributions. The subtraction of bulk water signal has been performed only for the sake of direct comparison with the experimental data but not for the other analysis. The interpretation of experimental data can be further aided by looking at molecule-separated, as well as intra- and
| ∑ fj(q) exp(iq‚rj,m)|2〉 ∑ m j
(2)
and
Iinter(q) ) I(q) - Iintra(q)
(3)
respectively. As already introduced in ref 19 from the ratio of the interand intramolecular terms one can obtain the structure factor S(q) ) Iinter(q)/Iintra(q) + 1, which allows for an absolute comparison between the patterns of molecules or moieties with different electronic density. The quantity S(q) - 1 acts as an order parameter that quantifies the degree of local structuring for the molecule or its moieties, becoming in fact zero when no spatial correlation exists between distinct molecules or groups. 4. Results and Discussion The computed scattering intensity I(q) and the recently obtained experimental results9 are illustrated in Figure 2. The overall agreement is remarkably good, showing a strong resemblance in shape and intensity over the whole investigated range, thus confirming the reliability of our computational model. The main discrepancies are a slight shift of the main peak and the absence of the structures in the simulated spectrum at high scattering vector (q = 27 and q = 34 nm-1), as compared to the experimental one. Notably, the simulation data reproduce quite well the small wavevectors (q < 12 nm-1) part of the spectrum, whose relatively highsand almost constantsintensity
Short-Range Structure of a GM3 Ganglioside Membrane
Figure 3. Pentadecane experimental (squares) and simulated (circles) scattering intensity.
is peculiar to the GM3 spectrum, as opposed to the case of phosphocholines, for which the WAXS intensity below 10 nm-1 drops considerably. As shown in detail later on, the main contribution to the WAXS peak is due to the hydrophobic chains as in the case of phosphocholines,19,20 and therefore one may be tempted to attribute the shift to the nonoptimal force field associated to the chain carbon groups. However, recent simulations of phosphocholine bilayers in LR phase,19 employing the same parameters for the aliphatic chains as used here, suggested that to eliminate the small shift observed in the peak position, a reparametrization of the complete Ganglioside force field, not just of the tails’ part, would be more appropriate. To investigate this point further, we decided to perform a simulation of bulk liquid pentadecane (CH3(CH2)13CH3) using the same set of force field parameters, thus allowing a separate check of the reliability of the force field relative to the aliphatic tails in reproducing its WAXS spectrum. The results are reported in Figure 3, compared with available experimental data regarding pentadecane at 293 K and 1 atm.21 The system has been simulated at the experimental thermodynamic conditions. The two spectra compare quantitatively very well, especially in the high wavevectors region (q J 16 nm-1), the differences being more pronounced in the lower part of the spectrum. Notably, the position of the experimental peak appears at lower q, with respect to the simulated one, in contrast to the GM3 case. This opposite behavior rules out the possibility of simply correcting the shift seen on gangliosides by reparametrizing the Lennard-Jones potential of the aliphatic carbon groups. As a consequence it appears that the force-field of the whole GM3 molecule may be invoked to explain the differences between experimental and simulation data. In particular, keeping in mind the bulkiness of the GM3 head groups, as opposed to the high flexibility and elongated structure of the tails, a different packing of the heads appears to be a possible explanation of the observed discrepancies. To gain more insight into the differences between the simulated and actual system, one can compare quantities like the area per head and the typical interchain distance. The value of the area per head obtained from the simulation (0.67 ( 0.02 nm2) matches within error bar with the value of 0.70 ( 0.07 nm2 estimated from experimental SAXS measurements.5 The two data are compatible, but because of the large error bar in the experimental value, the comparison is not conclusive. On the contrary, more accurate informations can be obtained from the shift in the WAXS peak position qo itself. This can be used to estimate the discrepancy in the position do of the peak in the tails radial distribution function, directly obtaining an informa-
J. Phys. Chem. B, Vol. 111, No. 37, 2007 10967
Figure 4. GM3 intramolecular scattering intensity (solid line) compared with its contributions from the head groups (dashed line) and tails (dotted line).
tion on the density in the tail region. Using the Ehrenfest relation do ) 1.2285 × 2π/qo (see, e.g., ref 23), one obtains a difference in the typical distance between chains, do, of about 0.02 nm (=4%), corresponding to a less packed arrangement of the simulated system (i.e., bilayer chains are more fluid than the experimental counterpart). Such effect may originate from an underestimation of some effect of mutual interaction among head groups. The other major difference between the GM3 spectra is the lack of specific structure for the simulation data q > 20 nm-1. The analysis of the experimental results, supplemented by density and calorimetric measurements, as reported in ref 9, suggests that the high q intensity arises from local structuring of the GM3 head groups. Consequently one can infer that our simulated system shows a slightly different (looser) spatial arrangement in the head-group region with respect to the real one. To clarify the role of different parts of GM3 we have decomposed the spectrum in contributions arising from the hydrophilic (head group) and the hydrophobic (tail) components. The separated spectra are reported in Figure 4. The spectrum arising from the tails displays a distinct peak in the 13 j q j 15 nm-1 range, while the heads present a shoulder in the same range, thus indicating that the local structuring of both heads and tails occurs at approximately the same typical distances (of the order of 0.4 nm). In the q range beyond 20 nm-1 cross interference effects between heads and tails occur, as evidenced by comparing the total GM3 spectrum with the separate contributions from these groups. In Figure 5a,b, we report the scattering intensities of the heads and tails further separated into inter- and intramolecular contributions. It is apparent that the peak at q = 15 nm-1 stems primarily from the intermolecular correlation for both the intensities relative to head and tail groups. Since the intensity due to intramolecular correlations appears to be mostly structureless, one can gain a deeper understanding of the local structure resulting from the intermolecular interactions by evaluating S(q), defined previously. The calculation has been separated in contributions arising from heads and tails and is reported in Figure 6. The distinct peak found at q = 15 nm-1, present in both spectra, indicates that each component, either a head or a tail, is preferentially surrounded by elements of the same moiety but belonging to different GM3 molecules, all falling at an average distance of 0.4 nm. It is worth pointing out that the intensity of the peak due to the head groups is substantial (about half the one relative to the tails). This feature seems to be peculiar to the GM3 bilayer. In the case of DPPC,
10968 J. Phys. Chem. B, Vol. 111, No. 37, 2007
Sega et al. presence of excluded volume effects.22 From these considerations, one can argue that the head groups interact via a much softer effective potential than for the tails, the spectrum of the former having much weaker oscillations beyond the first peak. This can be ascribed to the branched structure of the sugar rings, that does not allow efficient packing like in the case of a monatomic fluid, or aliphatic chains, as well as to the hydration of the water exposed region. Nevertheless, the presence of a first, defined peak in the head group pattern is doubtless the fingerprint of correlations which have no counterpart in the case of phosphocholines. 5. Conclusions
Figure 5. Scattering intensity from GM3 head groups (left panel) and tails (right panel): total intensity (solid line), intramolecular (dotted line), and intermolecular (dashed line) contributions.
Figure 6. Order parameter of the tails (solid line) and head groups (dashed line). The thin solid line represents the structure factor of liquid pentadecane.
on the contrary, it has been observed that, although a peak in S(q) also develops for the head groups, its height is much smaller and located at a position shifted to the right with respect to the tails (see ref 19, Figure 2). Ganglioside GM3 displays instead a marked correlation for the head groups on the scale of 0.4 nm, which is most probably related to packing forces of comparable strength with that of the tails, thus allowing the head groups of different molecules to be in close contact with each other. Let us remark that in the definition of S(q) the intermolecular scattering intensity is normalized to the scattering power of the molecule, that is, the intramolecular term. Therefore, the higher value of the GM3 head group peak in S(q), with respect to its DPPC counterpart, is not directly related to the higher molecular weight of the GM3 head group. In view of understanding the role played by the tail liquid crystalline order on the observed structural pattern, the same type of analysis has been performed on the results of another simulation of liquid pentadecane at 333 K and 1 atm, whose spectrum is reported in Figure 6. The two profiles appear to be quantitatively very similar, except for the higher peak of GM3 tails with respect to pentadecane, indicating that higher structuring is induced by the liquid crystalline nature of the GM3 bilayer. From a qualitative point of view, however, both the head and tail structure factors appear to have the same features of a disordered system composed of monatomic units interacting via a central potential: namely, a first, well-defined peak, followed by rapidly decaying oscillations. As is well-known, in the case of monatomic liquids the main peak indicates the presence of near neighbors at distances of the order of the Van der Waals diameter, while oscillations at larger q reflect the
The analysis of the WAXS spectra obtained by computer simulation of a GM3 bilayer has shown a good agreement with the available experimental data. By analyzing the different contributions to the overall spectrum, we found that at the position of the characteristic peak a strong tail-tail correlation is present, as seen also in phospholipids, but also that a well marked contribution stems from the head-head correlations. The latter appears to be a distinguishing feature of gangliosides, as opposed to phosphocholines, for which the peak arising from head-head correlations is smaller and located at higher wavevectors. This observation enforces the notion that the packing of the GM3 heads plays an important role in determining the overall arrangement of the bilayer. In fact, the order parameter derived from the structure factor shows that the degree of correlation of the heads is about half the tails one and located at the same position. This is a strong indication that the interpretation of the experimental WAXS structuring as containing an important contribution of the head group is convincing. We found also that while the tails structure factor appears to resemble that of liquid pentadecane, as well as that of simple monatomic fluids, the heads structure factor suggests a much softer effective potential between the head groups, arising from the branched nature and the asphericity of the sugar rings. In summary, the current study underscores the competition of excluded volume and hydrophobic forces in affecting the phase diagram of gangliosides. It would be thus instructive to undertake further studies on different global arrangements, such as micelles or vesicles, to investigate the interplay between these forces on length scales much larger than those characteristic of tail and head groups. Acknowledgment. The computations have been partly performed on the HPC facility of the Department of Physics, University of Trento. References and Notes (1) Hakomori, S.; Young, W. W., Jr. Handbook of Lipid Research. 3. Sphingolipid Biochemistry. Plenum Publishing Corp.: New York, 1983; pp 381-436. (2) Masserini, M.; Ravasi, D.; Sonnino, S. Trends Glycosci. Glycotechnol. 2001, 13 (71), 239-250. (3) Hakomori, S.-I. J. Biol. Chem. 1990, 265 (31), 18713-18716. (4) Siebert, H.-C.; Reuter, G.; Schauer, R.; von der Lieth, C. W.; Dabrowski, J. Biochemistry 1992, 31, 6962-6971. (5) Cantu`, L.; Favero, E. D.; Dubois, M.; Zemb, T. N. J. Phys. Chem. B 1998, 102, 5737-5743. (6) Cantu`, L.; Corti, M.; Favero, E. D.; Muller, E.; Raudino, A.; Sonnino, S. Langmuir 1999, 15 (15), 4975-4980. (7) Cantu`, L.; Corti, M.; Favero, E. D.; Raudino, A. J. Phys. II 1994, 4, 1585. (8) Boretta, M.; Cantu`, L.; Corti, M.; Favero, E. D. Physica A 1997, 236, 162-176. (9) Brocca, P.; Cantu`, L.; Favero, E. D.; Dubois, M.; Motta, S.; Tunesi, S.; Zemb, T. Coll. Surf. 2005, A259, 125-133.
Short-Range Structure of a GM3 Ganglioside Membrane (10) Sega, M.; Brocca, P.; Melchionna, S.; Vallauri, R. J. Phys. Chem. 2004, B108 (52), 20322-20330. (11) van Gunsteren, W.; Berendsen, H. Groningen Molecular Simulation (GROMOS) Library Manual; Biomos: Groningen, Germany,1987. (12) Berger, O.; Edholm, O.; Jahnig, F. Biophys. J. 1997, 72, 20022013. (13) Lindahl, E.; Hess, B.; van der Spoel, D. J. Mol. Model. 2001, 7, 306-317. (14) Berendsen, H. J. C.; van der Spoel, D.; van Drunen, R. Comput. Phys. Commun. 1995, 91, 43-56. (15) Berendsen, H. J. C.; Postma, J. P. M.; DiNola, A.; Haak, J. R. J. Chem. Phys. 1984, 81, 3684-3690. (16) Essman, U.; Perera, L.; Berkowitz, M. L.; Darden, T.; Lee, H.; Pedersen, L. G. J. Chem. Phys. 1995, 103, 8577-8592.
J. Phys. Chem. B, Vol. 111, No. 37, 2007 10969 (17) Seelig, A.; Seelig, J. Biochemistry 1974, 13 (23), 4839-4845. (18) International Tables for Crystallography; Hahn, T., Ed.; Reidel: Dordrecht, The Netherlands, 1991; Vol. C, pp 500-502. (19) Sega, M.; Garberoglio, G.; Brocca, P.; Cantu`, L. J. Phys. Chem. 2007, B111, 2484-2489. (20) Luzzati, V.; Mustacchi, H.; Skoulios, A.; Husson, F. Acta Cryst. 1960, 13, 660-667. (21) Habenschuss, A.; Narten, A. H. J. Chem. Phys. 1990, 92 (9), 56925699. (22) McDonald, I. R.; Hansen, J.-P. Theory of Simple Liquids; Elsevier: London, 2006. (23) Warren, B. E. Phys. ReV. 1933, 44, 969-973.
