Comparison of Ab Initio, DFT and Semiempirical QM/MM approaches for description of catalytic mechanism of hairpin ribozymes

Vojte Mlynsky, Pavel Banas, Jiri Sponer, MarcW. van der Kamp, Adrian J. Mulholland, and Michal Otyepka, Journal of Computational and Theoretical Chemistry, 2014, DOI:10.1021/ct401015e

Contributed by Esteban Vohringer-Martinez

I enjoyed very much reading this paper where the authors report a very interesting study in which the performance of semiempirical methods in QM/MM calculations is addressed in detail.

The reaction under study is the reversible phosphodiester bond cleavage and ligation in hairpin ribozymes. These ribozymes belong to a small group which catalyze the reaction without any metal ion at comparable rates.

From experimental evidence and previous theoretical calculations it was known that there are two main players in the catalytic reaction: A38 and G8; two neighbouring bases to the phosphodiester bond to be cleaved between the Adenine -1 and Guanine +1 (see scheme below). However which was not clear from the theoretical studies nor the experiment is the protonation state for the A38 and G8 bases. The experimental studies showed a pH dependence in the catalytic activity of the ribozyme implying different catalytic activity as a function of protonation state.

To account for different protonation states the authors studied two reaction mechanisms: the mono anionic one (top) where only the phosphatediester group is deprotonated and the dianionic mechanism where the G8 bases is also deprotonated (bottom). In both mechanism they assumed the A38 to be protonated (see reaction scheme).

The energetics of the reaction are followed along two reaction coordinates represented by a linear combination of the cleaving and forming bonds (d1 and d2) in the proton transfer step and the oxygen-phosphor distance as shown in the scheme. 

The energies were calculated with the SCS(spin component scaled)-MP2  single point energy calculations (CBS limit) on BLYP optimized geometries (6-31G(d,p) in a QM/MM electrostatic embedding with the AMBER99sb force field.  The test methods include the MPW1K hybrid functional, the BLYP GGA DFT functional and the semiempirical methods AM1/dPhoT and SCC-DFTBPR.

For me the main finding of this paper is that Ab-Initio and all DFT methods predict a concerted nucleophilic and proton transfer mechanism whereas both semiempirical methods yield a sequential mechanism where first the proton is transferred and then the nucleophilic attack takes place. Interestingly both semiempirical methods yield a stable intermediate in the dianionic mechanism after the proton transfer step which is not present in the ab-initio and DFT results. 

These result raise doubts about the ability of semiempirical methods to provide correct structures and reaction mechanisms in enzyme catalysis and suggest that a careful calibration of these methods for each system should be performed prior to its usage. 

One additional aspect to consider is that the activation barrier of the semiempirical methods were close to the ab-initio, DFT methods and experiment. However, a match of barrier heights as shown in this study does not guarantee a correct reaction mechanism.

Finally, the authors also compare their results to free energy calculations employing semiempirical methods. But, as the authors also conclude the almost negligible entropy contribution they report should be taken with care due to the very short sampling time of only 50ps.