Prebiotic Peptide Synthesis on Blue Gene Platforms at "Iron-Sulfur-World" Conditions

The exploration of possible scenarios for prebiotic molecular synthesis, including biopolymers such as peptides, is paramount to understanding how primitive life emerged on the young Earth. Amongst a vast amount of different hypotheses, evidence accumulated that mineral surfaces in conjunction with water at extreme thermodynamic conditions might offer favorable reaction environments. In particular, an unusually precise proposition for biomolecular synthesis on iron/ sulfur minerals in hot-pressurized water as found close to deep-sea hydrothermal vents has been detailed in the framework of the so-called "Iron-Sulfur-World" (ISW) scenario [1]. Certainly, understanding the interplaying fundamental issues of unusual chemical reactivity at extreme conditions, liquid state theory of solvation, and physical chemistry of mineral/water interfaces turn out to be of overriding importance here. In the intricate reaction chain from small molecules to functional proteins the formation of the peptide bond as such is, without any doubt, a key step. Although significant experimental support of the major ingredients has been accumulated by several groups, pertinent experiments lack detailed molecular insight into how small "inorganic" reactants transform into biomacromolecular products. Still, preliminary ab initio molecular dynamics (AIMD) simulations at ISW conditions were carried out only a few years back [2-4].

In the long-term project "Full in Silico Exploration of Possible Routes to Prebiotic Peptide Synthesis by Ab Initio Metadynamics" devoted to fundamental research in Chemistry the importance of high temperature and pressure ISW reaction conditions including mineralsurfaces is being assessed by AIMD techniques. The primary goal here is to provide vital molecular level understanding about the pertinent reactions in the "virtual lab" [5] which is otherwise difficult or even impossible to obtain in real laboratory experiments. Greatly expanding our initial work [2-4], our computations on the IBM Blue Gene systems JUBL and JUGENE at the John von Neumann Institute for Computing (NIC) at Forschungs zentrum Jülich (FZJ) during the last two years have unveiled very important mechanistic and energetic details of peptide synthesis at ISW conditions. It is stressed that the unprecedented computational complexity of our in silico prebiotic peptide synthesis demanded an investment of up until now about five nanoseconds of AIMD simulation time in total, which was only possible due to generous access to these efficient resources at NIC. In 2008 the above-mentioned project has been elected to be the first "NIC Excellence Project of the Year".

The "Virtual Lab" Approach to Chemistry

Recent advances in both computer technologies and simulation methods, in particular Car-Parrinello AIMD [6,7] in conjunction with efficient sampling methods like the powerful "metadynamics" technique developed by Laio and Parrinello (see Ref. [8] for a review), make it possible to study truely complex chemical reaction networks in the "virtual lab" [5]. All calculations presented here were performed using Hohenberg- Kohn-Sham density functional theory in its efficient plane wave pseudopotential implementation [7] within the CPMD software package [9].

Since the beginning of the 1990ies, the CPMD code [9] has been designed by Jürg Hutter from the onset to run efficiently on all kinds of parallel platforms as explained in detail in NIC Lecture Notes [7]. However, low-latency interconnects are required to run this parallel AIMD code efficiently. To get around load balancing problems on platforms of ever growing processor numbers a second level of parallelization named the "task-grouping" of processors has been implemented into CPMD some time ago [7]. Furthermore, the hierarchical multi-level strategies [7] that combine distributed-memory and shared-memory parallelization are highly suited for ultra-dense massively parallel HPC machines such as the Blue Gene architecture [10] in particular.

Figure 1: Relative scaling performance when using N processors with respect to 512 processors (i.e. the ratio of the computer time for one AIMD step per processor using N processors to that for 512 processors multiplied by N) for the CPMD code for one of the systems studied on JUGENE at NIC. Red and green lines are using Open MP threads 1 and 4, respectively. The yellow line is the performance using the multiple walker technique where the number of walkers, n, is reported in parenthesis. In the inset the scaling behavior going from 512 processors to 4,098 processors is magnified for clarity.

However, electronic structure calculations and thus AIMD require nontrivial parallelization strategies as the character of the underlying off-lattice quantum problem is not easily suited for partitioning without making use of further approximations such as done in linear scaling algorithms. This is due to the spatially non-local character of quantum-mechanical wavefunctions. The multiple walker metadynamics algorithm [8], which we have successfully implemented, is a linear scaling algorithm in itself and thus improves the net scaling behavior of CPMD for our given system sizes at hand on a large number of processors. Figure 1 shows the performance of CPMD on the JUGENE Blue Gene/P installation at NIC employing up to one half of the whole machine with and without multiple walkers. Still, due to the inherent scaling limitations in any quantum simulation code for typical system sizes like that in the test case shown, it is much less efficient to go beyond a Blue Gene mid-plane for practical simulations; see for instance in Figure 1 that the scaling dropped down to about 33 % when using a full mid-plane with respect to using only a quarter of a midplane. However, the parallelization among walker replicas is extremely efficient due to the loose coupling of the walkers such that the communication characteristics of Blue Gene systems can be exploited using many racks for a single AIMD simulation. As usage of n walkers will decrease the total length of the simulation by a factor of about n the effective scaling of a multiple walker algorithm can be estimated based on CPU time per AIMD step divided by n. It is evident based on the scaling behavior shown in Figure 1 that use of all processors of JUGENE is now possible for our system of interest without the need to increase the size of the system even when certain processor groups must communicate via highlatency interconnects.

The Key Result: Peptide Synthesis Cycle

The simulations of peptide bond formation [11,12] between two glycine molecules were carried out using three vastly different reaction conditions: ambient bulk water at about 300 K and 0.1 MPa (ABW), hot-pressurized bulk water at about 500 K and 20 MPa (HPW), and hot-pressurized water at the pyrite interface (PIW). The effective free energy barriers estimated along the peptide synthesis cycle leading to diglycine are reported in Figure 2. It is evident from the mechanism depicted in Figure 2 that it is the neutral form of the amino acid glycine 2 that is required and not the zwitterionic form 1 for its reaction with the COS molecule to form thiocarbamate (see step B). The thiocarbamate 3, in turn, leads to an activated amino acid in form of its so-called Leuchs anhydride 5 that easily adds to another amino acid (or peptide) to form a peptide bond (in step E) which finally yields an elongated peptide. As apparent by the computed free energy surface Figure 3 HPW extreme conditions stabilize the neutral form 2, consistent with the lowering of the dielectric constant of HPW, whereas in ABW neutral glycine converts easily to the zwitterion 1 on an ultrafast timescale of ca. 1 ps.

In other words, such extreme thermodynamic conditions are found to increase the concentration of the neutral amino acid by shifting the equilibrium between the neutral and the charged zwitterionic forms of amino acids toward neutral form, thus favoring the formation of peptides. This is an interesting result and immediately reveals the importance of hot-pressurized conditions for this route to peptides. Moreover, it was found that these extreme HPW conditions speed up the production of peptides by accelerating individual steps of the whole peptide synthesis cycle according to the free energy barriers reported in Figure 2.

Figure 2: Full peptide synthesis cycle comprising input of an amino acid (or peptide), here glycine, and its activation followed by elongation using another amino acid (or peptide), here another glycine, as well as termination and hydrolysis as a major reverse reaction. The calculated free energy barriers (given in kBT energy units) for individual steps of the mechanism leading to diglycine formation are color coded. Blue: ambient bulk water (ABW), green: hot-pressurized bulk water (HPW), red: hot-pressurized water at the pyrite interface (PIW). The crossed direct formation path C´ is very unlikely in view of its high activation free energy compared to the indirect path via isocyanate 4. See Ref. [11] for details.

Another discovery from our simulations is the so-called isocyanate pathway leading to the formation of the activated form of amino acid, Leuchs anhydride 5, from thiocarbamate 3. Compared to a direct cyclization of the thiocarbamate 3 to form Leuchs anhydride 5, the indirect isocyanate pathway, i. e. first forming an isocyanate 4 which rapidly cyclizes to Leuchs anhydride, is very much lower in terms of free energy barriers. This result confirms earlier experimental speculations about such a route including an equilibrium between 5 and 4. Our calculations also shed light on the productivity of the cycle including the formation of isocyanate being the ratedetermining step of the whole peptide cycle. Based on a simple estimate, the time scale for forming peptide bonds along this route is in the order of a few minutes at hot-pressurized conditions whereas it would be several years at ambient conditions! It is also shown by the simulations that hydrolysis of the synthesized peptide is slower than the rate-determining formation step and, therefore, a net accumulation of peptide can be expected in agreement with experimental conclusions.

Figure 3: Free energy surface for the conversion between zwitterionic 1 and neutral form 2 of glycine (a) in ambient bulk water (ABW) and (b) in hot-pressurized bulk water (HPW); color bar shows the relative free energy F in kJ/mol energy units. Metadynamics AIMD simulations were performed using two collective variables: coordination numbers of nitrogen and carboxy oxygen to all hydrogen atoms in the system, c(NGly - H) and c(OGly - H), respectively. See Ref. [12] for details.

Possible roles of Fe/S mineral surfaces were also investigated via simulations using an ideal pyrite surface, FeS₂(001), as the simplest model. By decreasing the entropic contribution to the free energy barriers this surface is found to accelerate several reaction steps in the peptide synthesis cycle by lowering the corresponding free energy barriers up to a factor of two, which increases the respective reaction rates exponentially. In addition, more interesting effects like scaffolding of reactant molecules favoring the formation of transition states and thereby speeding up the reaction by several orders of magnitude have been identified. In particular, the pyrite surface is found to favor the preformation of the five-membered ring that is characteristic of Leuchs anhydride 5 as demonstrated in Figure 4.

Based on free energy calculations and careful examination of various reaction mechanisms, our studies underpin the importance of extreme conditions and mineral surfaces for peptide synthesis along a putative route [11,12]. These comprehensive simulations delineate pathways connecting the crucial activation and elongation steps through which peptides can be produced out of amino acids and COS via an indirect isocyanate/Leuchs anhydride route. Importantly, the data provide convincing evidence that all steps along the proposed synthesis cycle are clearly favored in hot-pressurized water when compared to ambient conditions providing in total a productive synthesis cycle [11]. Beyond the specific case, these findings imply that "different chemistry" must be considered when discussing putative prebiotic synthesis scenarios at extreme aqueous conditions.

What Next?

Although a significant step forward, the cycle in Figure 2 is based on a set of disconnected free energy calculations whereas a single, global free energy landscape is necessary to fully explore this rather complex reaction network including the roles of reverse and side reactions. This, again, will be a challenge to both algorithms to sample free energies beyond three or four dimensions and platforms to carry out such a unified, ultramassive simulation. Another key issue that still remains largely unexplored is the role of mineral surfaces since defective surfaces are known to be most active in heterogeneous catalysis whereas an ideal, non-defective pyrite has been used up to now in our "virtual lab" [5]. Including defects at the mineral/water interface, most desirably as "dynamical degrees of freedom" in an AIMD simulation, certainly adds a lot more complexity to the problem but would be another necessary step forward. However, looking back at recents developments it appears to optimists that such dreams might become true much sooner than currently expected.

Figure 4: Mechanism of the formation of Leuchs anhydride 5 from thiocarbamate 3 in hot-pressurized water at the pyrite interface (PIW) based on the free energy surface; color bar shows the relative free energy ΔF in kJ/mol energy units. Metadynamics AIMD simulations were performed using two collective variables: distance between the carbon atom of the COS entity to one of the carboxylate oxygen atoms d[C-O] and coordination number of nitrogen to all hydrogen atoms in the system c[N-H]. Three representative real space confi gurations sampled from these simulations at the pyrite-water interface demonstrate scaffolding due to FeS2(001) by preformation of the cyclic topology of Leuchs anhydride upon bidentate chemisorption. Color code: hydrogen (white), oxygen (red), carbon (green), nitrogen (blue), sulfur (yellow), iron (ocher); labeling is according to Figure 2. See Ref. [11] and upcoming publications for details.

Thanks

We are most grateful to Alessandro Curioni and Jürg Hutter for help in porting CPMD onto the Blue Gene/L and P platforms JUBL and JUGENE at NIC. This research is supported by Deutsche Forschungsgemeinschaft via Normalverfahren MA 1547/7.

References

[1] Wächtershäuser, G.
Groundworks for an Evolutionary Biochemistry: The Iron-Sulfur World, Progress in Biophysics and Molecular Biology, 58, pp. 85-201, 1992

[2] Boehme, C., Marx, D.
Glycine on a Wet Pyrite Surface at Extreme Conditions, The Journal of the American Chemical Society, 125, pp. 13362-13363, 2003

[3] Pollet, R., Boehme, C., Marx, D.
Ab Initio Simulations of Desorption and Reactivity of Glycine at a Water-Pyrite Interface at "Iron-Sulfur World" Prebiotic Conditions, Origins of Life and Evolution of Biospheres, 36, pp. 363-379, 2006

[4] Nair, N. N., Schreiner, E., Marx, D.
Glycine at the Pyrite-Water Interface: The Role of Surface Defects, The Journal of the American Chemical Society, 128, pp. 13815-13826, 2006

[5] Marx, D.
Theoretical Chemistry in the 21st Century: The "Virtual Lab", in: Proceedings of the "Idea-Finding Symposium: Frankfurt Institute for Advanced Studies", (Greiner, W.; Reinhardt, J., Eds.), EP Systema, Debrecen, pp. 139-153, 2004

[6] Car, R., Parrinello, M.
Unifi ed Approach for Molecular Dynamics and Density-Functional Theory, Physical Review Letters, 55, pp. 2471-2474, 1985

[7] Marx, D., Hutter, J.
Ab Initio Molecular Dynamics: Theory and Implementation, in: Modern Methods and Algorithms of Quantum Chemistry, (Grotendorst, J., Ed.), John von Neumann Institute for Computing (NIC), Forschungszentrum Jülich, Germany, Vol. 3, pp. 301-449, 2000

[8] Laio, A., Parrinello, M.
Computing Free Energies and Accelerating Rare Events with Metadynamics, in: Computer Simulations in Condensed Matter: From Materials to Chemical Biology, Ferrario, M.; Ciccotti, G.; Binder, K., Eds., Springer-Verlag, Berlin Heidelberg, Vol. 1, pp. 315-347, 2006

[9] CPMD
Hutter, J. et al., Copyright: IBM Corp 1990-2008 and MPI für Festkörperforschung Stuttgart 1997-2001; see http://www.cpmd.org

[10] Hutter, J., Curioni, A.
Car-Parrinello Molecular Dynamics on Massively Parallel Computers, ChemPhysChem, 6, pp. 1788-1793, 2005

[11] Schreiner, E., Nair, N. N., Marx, D.
Infl uence of Extreme Thermodynamic Conditions and Pyrite Surfaces on Peptide Synthesis in Aqueous Media, The Journal of the American Chemical Society, 130, pp. 2768-2770, 2008

[12] Nair, N. N., Schreiner, E., Marx, D.
Peptide Synthesis in Aqueous Environments: The Role of Extreme Conditions on Amino Acid Activation, The Journal of the American Chemical Society, in press, 2008

• Nisanth N. Nair
• Eduard Schreiner
• Dominik Marx

Ruhr-Universität Bochum, Lehrstuhl für Theoretische Chemie

  top