Definition of the Chemical Mechanism of the Photosynthetic Enzyme Rubisco

Virtually all of the carbon in the biosphere is the result of carbon dioxide (CO2) fixation by the photosynthetic enzyme, D-ribulose-1,5-bisphosphate (RuBP) carboxylase/oxygenase (Rubisco). Rubisco also has the dubious honour of being the most abundant protein on earth, a consequence of its woeful catalytic efficiency. Rubisco is both very slow and poorly selective for its substrate CO2, binding also oxygen (O2) and catalysing a competing wasteful oxygenation reaction. As one would expect Rubisco to have been subjected to extreme evolutionary pressures, its incompetency is a great puzzle. Understanding the reasons offers huge potential as a basis for reengineering Rubisco; even modest improvements in efficiency have major implications for improving light, water and nutrient utilization by plants, and, hence, applications for better agricultural crops, greening deserts and degraded land, and for soaking up green-house gases. As Rubisco catalyzes a multi-step reaction involving as many as four enzyme-bound intermediates, whose instabilities give rise to multiple side reactions which further compromise its efficiency, it is possible that the enzyme’s complex structure and function is a compromise solution to effecting quite difficult chemistry. Not surprisingly, much of this chemistry is not approachable by experiment. This provides a (very challenging!) opportunity for computer simulation to try to define intractable issues, particularly the states and roles of the "invisibles" – protons and water molecules. Thus: protonation states of the forest of ionizable residue sidechains in the active site at different stages of the reactions (both carboxylation and oxygenation), as well as the protonation states for the reactant, intermediates and transition states; identities of the proton donors for the various steps; networks for channelling protons produced in the reaction away from the reaction centre; and origin of the water molecule consumed and produced (oxygenation only). We are addressing these issues using a range of computational methods, particularly ab initio QM studies of active-site fragment complexes and MD simulations with hybrid QM/MM potentials.

Principal Investigator

Dr Jill E. Gready
Computational Proteomics Group
John Curtin School of Medical Research


u53, d55

Facilities Used



Mr Lars Andrees
Dr Peter L. Cummins
Dr Bill King
Dr Harald Mauser
Computational Proteomics Group
John Curtin School of Medical Research

Prof T. John Andrews
Research School of Biological Science

RFCD Codes

250106, 250601, 250699, 270108, 270402, 291102, 300203

Significant Achievements, Anticipated Outcomes and Future Work

Results. Because of the complexity of the active site and Rubisco chemistry, as well as difficulty in applying QM/MM methods to an enzyme as large as Rubisco, our initial computational approach has been to map the complete carboxylation reaction using an ab initio QM supermolecule fragment model, as a starting point for these other investigations and simulation methods. In the first study [King et al. Biochemistry 37, 1998, 15414] we established a role as a base for the lysyl carbamate group in the first enolization step, a unique role for a carbamate not known in any other enzyme. This was the first theoretical study that took account of the likely roles of critical features within the active site. A prediction of this study has been tested and supported experimentally [ [Harpel et al. Biochemistry 41, 2002, 1390]. We continued the computational dissection for the subsequent steps – the carboxylation, hydration and C–C bond cleavage reactions – using similar fragment complexes (29 atoms) at the DFT level [1]. The 29-atom model was defined by an Mg2+ ion coordinated to a 2-carbon atom fragment of RuBP (ethen-1,2-diol-2-ate), 3 water molecules (2 representing Asp203 and Glu204) and methylcarbamate representing the carbamylated Lys201. These calculations showed that CO2 is added directly without assistance of a Michaelis complex, in agreement with experiment, and that hydration of the resultant b-keto acid occurs in a separate step with a discrete transition step. The carbamate appears very versatile, acting as a general base not only in the enolization reaction, but also in the hydration and cleavage reactions. While the step with the highest activation energy is the C–C bond cleavage, special arrangements of the metal coordination make it remarkably facile for such a bond-breaking reaction (28-37 kcal/mol depending on the gem-diol hydrate conformer).

QM/MM calculations to mirror the steps of the fragment calculations were then undertaken, to study the effect of the rest of the enzyme environment on the reaction and to calculate free energies [2]. A larger QM region of 165 atoms, rather than the 29-atom fragment-complex model, also allowed investigations of the roles of other active-site groups during the reaction, particularly in possible acid-base catalysis and proton channelling. The QM region included all residues of the first and second coordination sphere of the Mg2+ ion: Lys-175, Lys-177, carbamylated Lys-201, Asp-203, Glu-204, His-294, Lys-334, the Mg2+ ion and complete RuBP. Of interest was whether the b-ketoacid might be stabilized by protonation of the carboxylate group (from CO2); experimentally there is no evidence for reversibility of CO2 addition although the fragment-model calculations suggest it would be energetically accessible. The results showed that carboxylate protonation by Lys-334 was likely and did stabilize the b ketoacid, and that it also reduced the activation free energy for the next, hydration, step: the free energy difference of 16 kcal/mol suggests an almost irreversible CO2 addition. The results also showed that the C–C cleavage reaction is stabilised by the enzyme environment. To fully understand the catalysis of Rubisco, it is essential to know how the two protons formed during each catalytic cycle are transported away from the active site. Despite many experimental efforts, the detailed mechanism is still unknown. We have found that the dianionic phosphate groups of RuBP are acting as strong proton acceptors in the active-site environment. In the first steps of the catalysis, we found strong H-bond coupling between the carbamylated Lys-201 and the P2-phosphate via the substrate water and RuBP-O4. This interaction could be responsible for the recycling of the intermediary protonated carbamate. In the same way, another path for proton transport from the carbamate to the P1-phosphate group via RuBP-O2 and Thr-173 is indicated. Thus, the QM/MM simulations suggest that these phosphates may be responsible for the recycling of the carbamate and the transport of the two resulting protons away from the active site.

In addition to these studies we have undertaken a pilot study to compare the energetics of CO2 or O2 gas addition to the enediolate [King, Gready & Andrews, unpublished, 2000]. The oxygenation reaction, i.e. addition of triplet O2 to a carbanion, is a puzzling reaction requiring a spin inversion, which is inherently improbable in the absence of spin-delocalising mechanisms. QM representations of the O2 addition which would be adequate for treatment of the spin inversion require large correlated calculations, preferably MP2 or higher, which were not possible for our 29-atom fragment model until recently. Initial DFT calculations with a smaller fragment indicated a novel mechanism for the spin inversion for the oxygenation reaction, with much greater stabilization of the oxygenation product compared with the carboxylation product (-15.5 and -4.5 kcal/mol, respectively).

Future work. As we can now begin to attribute specific roles to active-site features and suggest the sources and fates of key protons in the acid/base chemistry that pervades the mechanism, we will expand the QM/MM studies to even larger representations of the active site. Such studies will also include comparisons of mechanistic details of Rubiscos from other species, and of mutants. We will also readdress the carboxylation versus oxygenation issues with DFT and MP2, and maybe higher methods, with a fragment model of comparable size (at least 30 atoms) to those used previously.


Computational Techniques Used

Transition-state, intermediate and product steps on the reaction pathway were performed with GAUSSIAN98 at DFT (B3LYP) level; stationary points were first optimised with the 6-31G* basis set, and single-point calculations with the 6-311G+** basis set were then used to calculate reliable energies. Starting coordinates for a minimal 29-atom active-site fragment model for the enediolate form of RuBP were generated from six sample sets of a 1 ns MD simulation of a Rubisco model. This model was based on a 1.6 Å x-ray structure of activated spinach Rubisco L8S8, and contained two large (L) and two small (S) subunits plus ~500 water molecules; ~1200 residues and 13,500 atoms.

QM/MM calculations were performed with our locally-developed program MOPS, using the semiempirical PM3 method for the QM region and Amber94 forcefield for the MM region. Starting structures were constructed for each stationary point of the carboxylation reaction defined by the fragment calculations, by docking the DFT-optimized structures into the active-site of an L2 dimer model of Rubisco, the minimal functional unit of Rubisco. The L2 dimer was extracted from the 1.6 Å crystal structure and solvated; 16,530 atoms, including 3040 water molecules. The co-crystallized CABP ligand and the Mg-coordination sphere of one active site were replaced by the analogous atoms of the DFT-optimised b ketoacid intermediate, and subjected to a 2 ns MD simulation with the DFT fragment coordinates frozen. For all other stationary points, the coordinates were constructed by replacing the b ketoacid active-site fragment with the analogous atoms of the respective DFT-optimised structures. QM/MM MD simulations of 200 ps were then performed for an expanded QM region of 165 atoms to equilibrate the system, using a 16 Å sphere for the flexible region containing the QM active site, and free energies for each complex on the reaction coordinate calculated by averaging over 800 runs (corresponding to 160 ps) using the linear response approximation.

For status of optimization of GAUSSIAN98 and MOPS for the VPP300, PC and APAC National Facility Compaq SC see reports for u51/d52 and w05.


Publications, Awards and External Funding

1. H Mauser, WA King, JE Gready, TJ Andrews. CO2 fixation by Rubisco: computational dissection of the key steps of carboxylation, hydration and C-C bond cleavage. J Am Chem Soc 123, 2001, 10821-10829.

2. H Mauser, L Andrees, JE Gready, TJ Andrews. QM/MM simulation of the steps of CO2 fixation by Rubisco. M/S in preparation, 2002.


DAAD (Deutscher Akademischer Austauschdienst) Postdoctoral Fellowship (2000-2001) to Dr Harald Mauser.

DAAD (Deutscher Akademischer Austauschdienst) Visiting Studentship (2001) to Mr Lars Andrees.