Direction des sciences du vivant

Our group conducts research in the field of computational biology and chemistry. In particular, we apply theoretical techniques to the study of physical phenomena closely related to biophysics and biology. Our interests are both in the solution of challenging biophysical problems and in the methods used in such undertakings. We pursue the following research themes: 1) Theoretical investigation of metallo-proteins. 2) Structure and dynamics of hydration around proteins, micelles and inverse micelles. 3) Mesoscopic modeling of protein-protein and protein-solvent interactions

1 Metallo-proteins: Generalities.

One-third of all proteins are "metallo-proteins", chemical combinations of amino acids with ions of metals such as Fe, Ca, Cu, Zn, Mn etc. The metal ions in metallo-proteins are critical to the protein's function, as they are required in numerous essential enzymatic reactions. The capability that transition metals have to exist in a variety of oxidation states is an important asset for proteins, which need easily available electron donors/acceptors to carry out electron transfer reactions in certain enzymes. In the photosynthesis of plants, for instance, a Mn cluster is directly responsible to fixate atmospheric CO2 and produce dioxygen, essential for life on earth. Also, transition metal ions, such as Zn, Cu, Mn, are used by various forms of superoxide dismutase (SOD) to detoxicate the cell from free oxygen radicals, a byproduct of many biological reactions known to cause common pathological conditions in man.  Thus, the ability to understand and, ultimately, to control the binding and activity of protein metal sites is of basic biological and medical importance. In the past, we have been involved in the study of metallo-proteins involved with different aspects of photosynthesis of bacteria and plants. More recently, we have begun to investigate the production of oxygen in photosystem II and the Ca pumping in ATPases.
 

                 ©CEA/M. Marchi

1a Metallo-proteins: Electron Transfer in Bacterial Photosynthesis
An important part of our current research activity is concerned with the elucidation by theoretical means of the molecular mechanism of electron transfer in photosynthetic reaction center (RC) proteins.  In the bacterial RCs, the first step of photosynthesis is a highly directional electron transfer, which shuttles an electron on only one of the two symmetry related branches after photoexcitation.  This directionality is also very robust with respect to mutations. To tackle electron transfer (ET) in complex dielectric media such as proteins, an easy, but not simplistic, approach is to use reduced models of the electronic transitions in combination with molecular dynamics simulation. Once MD determines the important parameters of the model, the quantum mechanics can be done straightforwardly thus computing activation energies and reaction rates. This technique has already been applied in the past by us and others to understand both electron transfer and electronic excitation in proteins. More recently, we have carried out a new and extensive modeling of photosynthesis by developing first force fields for the photosynthetic cofactors (chlorophylls, pheophytins and quinones)1 , 2 , 3, 4 , 5 , 6 and then computing the free energy surfaces for the primary electron transfer in the bacteria Rhodobacter   sphaeroides7 (see Fig. 2).  In the latter investigation, the kinetic parameters for the electron transfer along the active (L) and inactive (M) branches have been determined. With no post--processing parameter fit, our modeling computes, for two different charge distributions, the driving forces for the transfer of an excess electron to the cofactors on the L--branch in good agreement with experiments. A multi-exponential kinetics of the primary charge separation is also predicted, consistent with experimentally observed kinetics3. In a latest development we have investigated the kinetic of electron transfer between ferredoxin and photo-system I8. Extensions to the study of higher plant, oxygen forming, photo-systems are also planned.

Fig. 2 Pictorial view of an instantaneous configuration of one hydrated photosystem II of cyanobacterium Thermosynechococcus elongatus in a POPC membrane after 9 ns of MD simulation. The lipid is shown in space filled, whereas ribbons are used for the protein. The 36,471 water molecules are not shown.

1b Metallo-proteins: Metallic Clusters in Photosystems
In a new project we theoretically investigate the role of the Mn cluster in the production of oxygen by the photosystem II protein (PS II). Many clues on the mechanism of biological oxygen production have been gathered over the years by a variety of spectroscopic measurements and biochemical experiments: At least 4 states of different charge and spin seem to be involved in the oxygen cycle – this is the so-called S-state cycle. Recently, the geometry of this oxygen-evolving center, or OEC, has been obtained from the first X-ray crystal structure of the PS II of the cyanobacterium Thermosynechoccus elongates at atomic resolution. This cluster - a cubane-like Mn3CaO4 aggregate connected to another Mn ion through an oxygen bridge – is very close to a tyrosine residue, Tyrz, which provides a relay between the OEC and the primary electron donor of the photochemical cycle. Although, this new structure provides an essential contribution to our knowledge of PS II activity, the mechanism for oxygen formation is still far from clear. Many proposed mechanisms assign a predominant role to two of the four Mn’s in donating electrons to Tyrz and in binding to the substrate water molecule, before the formation of dioxygen.
In order to provide crucial insights on the role of each of the metal ions and their ligands in the oxygen production, we plan a theoretical investigation of the structure of the electronic states and density on the OEC atoms and their protein environments. It is our opinion that such a piece of information can be gathered directly only through electronic ab initio DFT (Density Functional Theory) calculations on the possible charged and protonated states of the OEC consistent with the S-state cycle. Since DFT calculations are capable of predicting atomic structures with a high degree of precision, our theoretical calculations will also be useful to validate the position of the metals  - in particular Ca - in the cluster. Indeed, given the experimental error in precisely positioning the metal ions in such a large protein complex, the Ca coordinates are still controversial. Moreover, DFT calculations will be capable of providing structural data for the OEC in conditions of pH and environment for which PS II is known to be functional - crystallization conditions inhibits the photosynthesis.
Only in the last few years DFT calculations have been used to investigate molecules of biological interest. In particular, our group has carried out a few studies on the cofactors of bacterial photosynthesis, such as bacteriochlorophyls1, quinones3, 9 and, more recently, Mn catalase (manuscript in preparation). For the OEC of PS II, we will need to include in the calculation the amino acid bound directly to the cluster and some water molecules of the immediate environment. With modern parallel computers state of the art DFT code from Parrinello’s group in Lugano with whom we collaborate, electronic calculation of complex structures such as that of the OEC and its ligands are within our reach. Preliminary molecular dynamic simulations of PSII in a POPC membrane and water have provided initial configurations for our DFT calculations.

As a final remark, we emphasize that the scope of our investigation goes beyond the oxygen formation in PS II. Indeed, numerous active sites of other interesting metallo-proteins could be studied with our approach as well. I think of possible extensions of our study to detoxicating proteins such as SOD or to Fe-S cluster containing proteins involved in interesting catalytic activities such as production of hydrogen in Fe-Ni or Fe-only hydrogenase, both themes having an obvious technological appeal.

                 ©CEA/M. Marchi


 
1c Metallo-proteins: Calcium Transport in Sarcoplasmatic Reticulum ATPase
Cells use different mechanisms to protect against toxic agents or excess metal ions through detoxication pathways, which in turn are controlled by transport proteins. We are interested in one of such proteins, calcium ATPase, which provides activated transport through the sarcoplasmic reticulum for calcium and, possibly, strontium (SERCA1a protein). In order to gain insights in the mechanism of metal ion transfer through the membrane, we have engaged very recently in the molecular modeling of SERCA1a. In particular, we study the calcium ion binding to the protein and investigate its relations to the conformational stability of the protein – two conformationally different structures are know from X-ray for Ca bound and Ca depleted forms of SERCA1a.. Given the complexity of this protein (more than 15400 atoms) and its environment, at the same time hydrophobic and hydrophilic in nature (a part of the protein is bound to the cellular membrane, while the other is hydrated), its simulation requires a considerable calculation effort (see Fig. 3). Through long molecular dynamics runs, we have detected significant concerted movements of trans-membrane helices M1 and M2 perpendicular to the membrane, whereas helices M3-M10 move in the opposite direction. At present, we investigate the relation of this finding vis-à-vis the mechanism of the conformational transformation, which takes place in SERCA1a at the same time as the transfer of calcium. More recently, we have mutated in silico the two calcium ions in strontium (a metal of larger size) to locate the amino acids most involved in the ion affinity of the protein. Our preliminary results shows that in addition to residues bound directly to Ca a few amino acids as far 10 Å from ion play a significant rôle in the binding of Ca.

2 Structure and Dynamics of Solvation

Biological water at the interface of proteins is critical to their equilibrium structures, enzyme functions, and to phenomena such as molecular recognition and protein-protein interactions. In the past, we have characterized the dynamics of biological water around globular proteins10 , 11 by extensive computer modeling. Our simulations have demonstrated that modern force fields and current molecular modeling are able to reproduce the time scales of the biological water dynamics as detected by nuclear magnetic relaxation dispersion and other spectroscopic techniques. In this way, the mechanism of water attachment to the protein and the relation between water rotational and translational diffusions has been elucidated. Our current research is then aimed at identifying how and why water interacts differently with proteins with respect to other amphiphilic aggregates such as micelles12, 13 , 14} and reverse micelles15 , 16 , 17, 18.

3a Implicit Models of Solvent
We develop reduced (implicit) model of solvent to both accelerate the sampling of the conformational space of a biopolymer and compute directly thermodynamics properties such as free energies. In particular, we have introduced a novel method to simulate hydrated macromolecules with a dielectric continuum representation of the surrounding solvent19.  This approach allows the solution of the dielectric continuum problem ``on the fly'', while the molecular coordinates are propagated. It is similar in spirit to techniques used by Parrinello and coworkers in ab initio density functional theory calculations. Indeed, in our approach, the interaction between the solvent and the molecular degrees of freedom is described by means of a polarization density free energy functional, which is minimum at electrostatic equilibrium.   Recently, improved models of dielectrics for proteins and peptides have been developed to reproduce experimental solvation energies20, while electrostatic free energy functionals simpler and quicker to compute are being investigated20 , 21 , 22 , 23.

Fig. 1 Model of electrostatic interaction between two BPTI molecules. To each sphere representing an aminoacid is assigned a charge and a polarizable dipole. In blue is a dielectric continuum description of the solvent.

                ©CEA/M. Marchi
 
3b Mesoscopic Modeling
In addition to implicit models of solvent, we are also interested in devising new reduced models of solute in order to study phenomena of aggregation or crystallization among soluble and membrane bound proteins, and, possibly, drug binding to active sites. This undertaking involves the development of new mesoscopic models: Our idea is to model proteins as rigid geometrical objects whose surface represents a low grade approximation of the interface between the protein itself and the solvent (see for instance Fig. 1). The same approach will be used to model drug molecules, but additional degree of flexibility will be added. These objects will interact with each other with electrostatic, repulsive and excluded volume potentials.  Solvents of different ionic strength and, eventually, a membrane can be included using our dielectric continuum model discussed in the above 14. Possible applications of this approach are the study of the functional mechanism of molecular motors, such as proton and Ca ATPase. Application of this approach to aggregation of biomolecules is also envisaged.