Unpublished data:
Ab initio calculations of the binding site
Weber et al.
(Weber, P.C., Ohlendorf, D.H., Wendoloski, J.J., Salemme, F.R. Structural
origins of high-affinity biotin binding to streptavidin. Science 243:85-8,
1989)
proposed that stabilization of ionic resonance forms of
biotin by the protein environment contributes to the strong binding of
biotin to streptavidin. To test this hypothesis, and to test the ability of
our classical, nonpolarizable force field to reproduce biotin-protein
interactions, we conducted ab initio calculations for two model systems
simulating biotin/streptavidin and biotin/avidin complexes. The Gaussian
98 suite of programs 70 was used in all calculations. Two levels of theory
were considered: semiempirical with the AM1 Hamiltonian, and ab-initio at
the HF/6-31G* level. Standard counterpoise correction
(Boys, S.F., Bernardi, F. Calculation of small molecular interactions by
differences of separate total energies. Some procedures with reduced errors.
Mol. Phys. 19:553-, 1970)
was used at the ab-initio level.
The model for the biotin/streptavidin complex (system I) was built as
follows. The polar groups of the side-chains from 5 residues (Asn11, Ser15,
Tyr31, Ser 33, Asp116), forming hydrogen bonds to the biotin molecule in the
crystal structure were identified and truncated to the nearest carbon atom
(except for Tyr31, where the whole phenyl ring was included). The charged
side-chain of biotin was replaced by a methyl group. The hydrogen atoms were
placed geometrically according to the hybridization of the heavy atoms. In
the case of OH groups this procedure was ambiguous and several variants were
considered, all consistent with the formation of a hydrogen bonding network.
As some contacts in the crystal structure were too short, the total
interaction was found repulsive at the AM1 level. Four systems different in
the placement of H atoms were fully optimized at the AM1 level, but only one
of them retained the hydrogen bonds between the biotin and the protein
environment observed in the crystal structure. This system was fully
optimized at the HF/6-31G* level, and conserved the above hydrogen bonds at
each optimization step.
To construct the model for the biotin/avidin complex (system II), we
selected 5 side-chain polar groups (Asn10, Ser14, Tyr31, Thr33, Asn116)
forming hydrogen bonds to biotin in the crystal structure. They were
truncated analogously to system I, so that the only difference in
stoichiometry was a neutral NH2 group replacing an O- atom of COO-. Hydrogen
atoms were placed to reproduce the topology of hydrogen bonding network in
system I. The geometry was completely optimized at HF/6-31G* level. While
initial optimization steps led mostly to relaxation in hydrogen bonding
distances, after 20 steps the two Asn residues formed a new hydrogen bond
and one of them consequently broke the hydrogen bond to biotin. One possible
reason for this deviation from the crystal structure is the absence of the
geometrical constraints exerted by the rest of the protein upon the binding
site. For this reason we report the interaction energy (defined as the total
energy of the complex less the sum of total energy for the biotin and the
rest of the model system in the geometry of the complex) obtained in partial
optimization (16 steps) along with complete optimization results in Table 3.
Complete optimization of each system converged in about 160 steps, and took
over 9 days of CPU time on a Compaq DS20 Alpha machine. As one can see from
the Table 3, the intermolecular interaction in system I after the
counterpoise correction is 7 kcal/mol stronger than in system II for the
fully optimized structures and 10.6 kcal/mol for the partially optimized
structures.
To compare the DE values to the MM results we used the standard CHARMM 19
force field (fully charged ionic residues, e=1, and no solvation term) and
calculated the interaction energy between the biotin bicyclic ring and the 5
residues included in the QM calculations. We did this calculation on the
PDB structures minimized for 300 ABNR steps. The values obtained are -37
kcal/mol for streptavidin and either -30 or -35 kcal/mol for avidin
depending on which monomer is used (there are two distinct subunits in the
crystal structure). These values are quite close to the QM results, despite
differences in system preparation and the lack of protein connectivity
constraints in the QM optimizations. The interaction between streptavidin's
Asp 116 and biotin is -10.7 kcal/mol, whereas the interaction of avidin's
Asn 116 and biotin is -6.7 kcal/mol. Thus, the DDE between avidin and
streptavidin tends to be a little underestimated by the CHARMM19 force field
(2-7 kcal/mol vs 7-10.7 kcal/mol from quantum mechanics). The polarization
of biotin by the charged Asp 116 may contribute a couple of kcal/mol extra
to the binding free energy in streptavidin.
It is noted that in EEF1 the Asp residue, like all other ionizable residues,
is neutralized and its solvation free energy is reduced accordingly. In
addition, e=r is used in the calculations. The effective energy between Asp
116 and biotin in EEF1 is -4.3 kcal/mol (van der Waals -2.3, electrostatic
-4, desolvation +2). If we use a charged Asp 116 and assign to its
carboxylate the appropriate solvation free energy (~ -80 kcal/mol) and use
e=1, the effective energy is -4 (van der Waals -2.3, electrostatic -8.4,
desolvation +6.7). Thus, neutralizing the Asp and reducing its reference
solvation free energy produces compensating changes so that the effective
interaction remains about the same.