1Laboratory of High Performance Computing,
Institute of Physical Biology USB and Institute of Landscape Ecology AS CR,
University of South Bohemia, Zámek 136, CZ-373 33 Nové Hrady, Czech
Republic, email: ettrich@greentech.cz
2Institute
of Physics, Charles University, Ke Karlovu 5, CZ-12116 Prague 2, Czech
Republic
3Department
of Biochemistry, Faculty of Science, Charles University, Hlavova 2030, CZ-12840
Prague 2, Czech Republic
4Institute
of Physiology of AS CR, Vídeňská 1083, CZ-14220 Prague 4, Czech Republic
5Institute of
Biophysics, Faculty of Medicine, Charles University, V Úvalu 84, CZ-15000
Prague 5, Czech Republic
The NMR solved structure of Na+/K+-ATPase
N-domain [1] represents the first breakthrough in solving the structure of the
ATP-binding domain with a high resolution method. However, Hilge et al. [1] marginalized the role of protein molecular modeling done prior to
their NMR-study. Nearly three years ago we published a paper describing a
computational model of the H4–H5 loop [2] of Na+/K+-ATPase.
This model, and especially its improvement connected with usage of vibrational
spectroscopy [3], leads to a root mean square deviation of Ca (RMSD)
~1.6 Å with respect to the NMR average structure [1] and
1.4 Å with respect to the recently published crystal structure at
2.6 Å resolution [4]. RMSD values of NMR structures [1]
and the crystal structure [4] are all in the range of 1.4 Å and so, apart from several wrongly modeled
loop regions, the computational model describes the structure correctly. Our
model allowed computer docking of ATP to the N-domain and thus already at that
time to identify the amino acids forming the binding pocket. Experiments based
on the structural model followed and we were able to describe the binding
interaction in more detail [5-7]. E446, F475, K480, Q482, K501, G502 were
identified as amino acids forming the binding pocket correctly. Moreover, the
aromatic stacking interaction between ATP and F475 together with one hydrogen
bond of the NH2 of the adenosine moiety was described as most
important for binding [6]. However, F548 was described incorrectly to be close
to the binding site and one hydrogen bond was incorrectly modeled with E446 and
not with Q482. Molecular modeling proposed the closest distance 3.8
Å between Q482 and ATP and thus the mechanism of participation of this
residue in ATP recognition is rather indirect. Nevertheless, on basis of our
experiments we could correctly state that F548 does not influence binding of
ATP and that the replacement of Q482 with leucine results in a strong inhibition
of both ATP binding [7]. The structurally important hydrogen bond between R423
and E472 was proposed at first on basis of our model [7]. Finally, our model
had to be corrected after the publication of the NMR-structure only to a small
extent and thus for several years it was the most correct structure for
designing structural experiments. In spite of the fact that protein modeling is
often underestimated, our work shows that molecular modeling in combination
with vibrational spectroscopy and other low resolution techniques is a powerful
tool for fast and relatively correct determination of protein structures in
general.
This
research was supported by the Grant Agency of the Czech Republic (Grants No.
309/02/1479, 206/03/D082), and by the Ministry of Education of the Czech
Republic (Grants no. MSM113100001, LN00A141).
References
1. M. Hilge et
al., Nat. Struct. Biol. 10 (2003)
468–474.
2. R. Ettrich
et al., J. Mol. Model. 7 (2001) 184–192.
3. K. Hofbauerová
et al., Biopolymers 67 (2002) 242–246.
4. K.O. Hakansson, J. Mol. Biol. 332 (2003)
1175–1182.
5. M. Kubala
et al., Biochem. Biophys. Res. Commun.
297 (2002) 154–159.
6. K. Hofbauerová
et al., Biochem. Biophys. Res. Commun. 306 (2003) 416–420.
7.
M. Kubala et al., Biochemistry 42 (2003) 6446–6452.