Biochimie 112 (2015) 96e110

Contents lists available at ScienceDirect

Biochimie journal homepage: www.elsevier.com/locate/biochi

Research paper

Conformational dynamics of bacterial and human cytoplasmic models of the ribosomal A-site c, d, **  Joanna Panecka a, b, Jirí Sponer , Joanna Trylska e, * _ Division of Biophysics, Institute of Experimental Physics, University of Warsaw, Zwirki i Wigury 93, 02-089 Warsaw, Poland  skiego 5a, 02-106 Warsaw, Poland Interdisciplinary Centre for Mathematical and Computational Modelling, University of Warsaw, Pawin c CEITEC e Central European Institute of Technology, Masaryk University, Campus Bohunice, Kamenice 5, 625 00 Brno, Czech Republic d lovopolska  135, 612 65 Brno, Czech Republic Institute of Biophysics, Academy of Sciences of the Czech Republic, Kra e Centre of New Technologies, University of Warsaw, Banacha 2c, 02-097 Warsaw, Poland a

b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 13 November 2014 Accepted 23 February 2015 Available online 3 March 2015

The aminoacyl-tRNA binding site (A-site) is located in helix 44 of small ribosomal subunit. The mobile adenines 1492 and 1493 (Escherichia coli numbering), forming the A-site bulge, act as a functional switch that ensures mRNA decoding accuracy. Structural data on the oligonucleotide models mimicking the ribosomal A-site with sequences corresponding to bacterial and human cytoplasmic sites confirm that this RNA motif forms also without the ribosome context. We performed all-atom molecular dynamics simulations of these crystallographic A-site models to compare their conformational properties. We found that the human A-site bulge is more internally flexible than the bacterial one and has different base pairing preferences, which result in the overall different shapes of these bulges and cation density distributions. Also, in the human A-site model we observed repetitive destacking of A1492, while A1493 was more stably paired than in the bacterial variant. Based on the dynamics of the A-sites we suggest why aminoglycoside antibiotics, which target the bacterial A-site, have lower binding affinities and antitranslational activities toward the human variant.  te  Française de Biochimie et Biologie Mole culaire (SFBBM). All rights © 2015 Elsevier B.V. and Socie reserved.

Keywords: Ribosomal RNA Bacterial A-site Human cytoplasmic A-site Molecular dynamics simulations Aminoglycoside antibiotics

1. Introduction Ribosomes, huge RNA-protein assemblies present in every living cell, catalyze the key and universal process of life e translation of DNA sequence encoded by messenger RNA (mRNA) to amino-acid sequence of proteins [1]. Their mechanism of action is complex and spans different spatial and temporal scales [2e4]. Ribosomes are built of two subunits e small and large e and the primary function of the small subunit is maintaining the fidelity of translation [3]. In particular, ribosomal aminoacylated-tRNA binding site (A-site, decoding site), is responsible for correct recognition of mRNA codon and tRNA anticodon [5,6].

* Corresponding author. CeNT, University of Warsaw, Banacha 2c, 02-097 Warsaw, Poland. Tel.: þ48 (22) 5543 683; fax: þ48 (22) 5540 801. ** Corresponding author. Institute of Biophysics, Academy of Sciences of the Czech lovopolska  135, 612 65 Brno, Czech Republic. Tel.: þ420 5415 17133; Republic, Kra fax: þ420 5412 12179.  E-mail addresses: [email protected] (J. Sponer), [email protected] (J. Trylska).

The adenines 1492 and 1493, according to Escherichia coli sequence numbering of nucleotides, comprise a molecular switch in the A-site that controls fidelity of mRNA decoding [6]. As proved by crystal structures of bacterial ribosomes, during the process they can flip-in and out from the A-site bulge (e.g., see conformations in Fig. 1). When flipped-out, the adenines form a complex with the cognate tRNA anticodon [7] (‘active’ state), otherwise they are ‘inactive’ and tRNA cannot be accepted in the A-site. In the bare Asite the intra-helical states of the adenines were shown to be energetically preferred [8,9]. The proof-reading function of adenines' dynamics has been recently confirmed in the combined theoretical and experimental study by Zeng et al. [10]. The delicate A-site conformational balance has to ensure both sufficient speed and accuracy of translation. Therefore, maintaining this delicate dynamical equilibrium is crucial for preventing mRNA decoding errors [3,11]. Bacterial A-site in the small ribosomal subunit is a target for 2deoxystreptamine aminoglycoside antibiotics [12]. They are anchored in the RNA bulge with the so-called ‘neamine core’ composed of two pseudo-sugar rings, which is a common feature of

http://dx.doi.org/10.1016/j.biochi.2015.02.021  te  Française de Biochimie et Biologie Mole culaire (SFBBM). All rights reserved. 0300-9084/© 2015 Elsevier B.V. and Socie

J. Panecka et al. / Biochimie 112 (2015) 96e110

97

Fig. 1. Simulated crystallographic double A-site models and their secondary structures: (left) bacterial (PDB code: 3BNL [21]) and (right) human (PDB code: 2FQN [22]). Gray labels in the crystal structures describe the geometry of base pairs enclosed in the grey ellipses. The secondary structure of the full simulated bacterial RNA model containing two Asites (site 1 is enclosed in blue frame and site 2 is in grey frame). The secondary structure of the human cytoplasmic A-site is also shown with differences in the sequence with respect to bacterial A-site highlighted in red.

these A-site binding antibiotics (see Fig. S1 and Ref. [13]). They perturb decoding by altering the mobility of the adenines 1492 and 1493, and by promoting their flipped-out states [14,15], which favor near-cognate tRNA codon binding. Binding of these antibiotics is primarily of electrostatic nature, i.e. positively charged amine substituents interact with negatively charged RNA [16,17,18]. Therefore, aminoglycosides are not sufficiently selective and, in particular, they have non-negligible affinities to the A-site of eukaryotic ribosomes [19]. Another reason for the low selectivity of these antibiotics is significant conservation of the A-site sequence, and thus of secondary and tertiary structures, among various species [20]. In particular, the secondary structures of E. coli and cytoplasmic Homo sapiens decoding sites are overall quite similar, with sequence differences present only on one side of the rRNA bulge (compare the minimal A-site sequences shown in Fig. 1). The most notable difference between prokaryotic and eukaryotic organisms is the nucleobase 1408, opposite to the functional adenines 1492 and 1493 (Fig. 1), which is adenine in > 98% of prokaryotic and guanine in > 98% of eukaryotic rRNA sequences [20]. The nucleotide A-site sequence is also variable within the eukaryota taxon. For instance, nucleobases C1409, U1410 and A1490 (of human A-site) feature 90e98% sequence conservation, while A1491 is conserved in only 80e90% of species [20]. Despite A-site rRNA sequence variability among eukaryotic organisms, the minimal A-site rRNA sequence is the same for human and yeast Saccharomyces cerevisiae, which is a much simpler eukaryotic organism [20]. Therefore, considering the A-site only as an isolated fragment extracted from the ribosome, its properties should be similar for the human and simpler eukaryotic organisms, such as yeast. The discussed above sequence differences contribute to the different activities of aminoglycosides towards different A-site variants. In particular, in vivo studies on genetically engineered bacteria with hybrid ribosomes including eukaryotic A-site show that minimal inhibitory concentrations (MIC) of aminoglycosides necessary to inhibit bacterial growth are 2e3 orders of magnitude higher for bacteria containing ribosomes with human A-site sequence than for bacteria with bacterial A-site sequence [23]. A similar study showed that aminoglycosides feature significantly higher MIC for the prokaryotic A-site with only a single A1408G ‘eukaryotic-like’ mutation [24]. However, also the changes in the sequence ‘below’ the A-site bulge may affect paromomycin

translational activities. An example may be Leishmania that has only single nucleotide C1409/ U change in the minimal A-site with respect to the human variant (Fig. 1). Experiments on bacterial cells carrying hybrid ribosomes have shown that for the cells with ribosomes containing the Leishmania A-site, paromomycin-induced growth inhibition was more effective (with MIC of 128 mg/ml) than for cells with the human A-site (MIC > 1024 mg/ml). In summary, the above experiments prove that the changes in nucleotide sequence of A-site affect paromomycin-induced in vivo translation inhibition. Despite the significant influence of the A-site sequence differences on the in vivo biological effect of aminoglycosides, these antibiotics do not efficiently distinguish between the human cytoplasmic and bacterial A-site mimicking oligonucleotide models. Fluorescence experiments on these A-site models [19,25] have shown that the equilibrium association constant Ka determined for paromomycin association with A-site models is only about 10 times lower for the human cytoplasmic A-site than for the bacterial one (([2.4 ± 1.9]  106 M1 versus [3.7 ± 1.5]  107 M1, respectively). Therefore, apart from binding affinities, which are relatively similar for the bacterial and human A-site, there must be other factors that contribute to explaining differences in aminoglycoside activities. For instance, sufficient accessibility of the major groove was suggested to be a critical factor that enables binding of these antibiotics [26]. On the other hand, fluorescence experiments imply that the level of A1492 destacking may determine the level of aminoglycoside miscoding activity [19]. Also, the universally conserved pair of uracils U1406:U1495 positioned just above the A-site bulge is known to create stabilizing contacts with the bound aminoglycoside [14]. Studies on the bacterial A-site have shown that some mutations in these uracils decrease susceptibility to aminoglycoside paromomycin [27,28], although the mutation of this U1406:U1495 base pair to the canonical cWW C1406:G1495 base pair almost does not affect the binding of 4,5-linked aminoglycosides [24]. Therefore, to elucidate the influence of the above structural factors on aminoglycoside binding, it is important to understand the sequence-dependent conformational preferences of this ribosomal mRNA decoding site. Furthermore, the conformational freedom and dynamics of the ribosomal A-site, especially of the crucial nucleotides A1492 and A1493, is likely directly related to the accuracy of mRNA decoding.

98

J. Panecka et al. / Biochimie 112 (2015) 96e110

Experimental studies have shown that this process may have different accuracy in bacterial and eukaryotic cells. In particular, in vitro studies on translation fidelity of ribosomes from various mammal tissues show that decoding errors upon different perturbing factors (such as low temperature or high magnesium ion concentration) that were known to decrease translational accuracy in bacteria, only weakly affected the mammalian cells [29e31]. Also, in a simpler eukaryote, yeast Saccharomyces cerevisiae (which shares cytoplasmic A-site sequence with human) missense translation was shown to occur about three times less frequently than in bacteria (Escherichia coli) [32]. We are aware that multitude of factors (including tRNA competition [33] or influence of elongation factors [34]) may affect the translation accuracy. However, the intrinsic properties of the decoding site forming a part of the helix 44 are important in modulating the miscoding rates in bacteria and eukaryotes. Therefore, we believe that analyzing the sequence and dynamical properties of the minimal A-site is a useful attempt to contribute to understanding of the differences between bacterial and human A-sites. Crystallographic yeast ribosome structures [35,36] and cryoelectron microscopy (cryo-EM) models of yeast and mammalian cytoplasmic ribosomes have been recently solved [4,37e39]. Also, there exists one cryo-EM model of human ribosome [37]. However, the above eukaryotic ribosomal structures have relatively low resolutions (below 3.5 Å [38]). Therefore, we focused on structural data for oligoribonucleotide constructs including only the A-site bulge and a few neighboring base pairs at each side. In particular, the bacterial X-ray A-site model designed by Vicens and Westhof [14], consisting of two symmetrical sites, was frequently crystallized with and without antibiotics. It has been successfully used in previous molecular dynamics (MD) and replica-exchange MD (REMD) computations of the bacterial ribosomal A-site [8,9,13,40e42]. Other computational techniques, such as PoissonBoltzmann model [16], docking [43] and combined virtual screening with 3D-QSAR [44] were also applied to the bacterial Asite. Also, we have recently compared the dynamics and structural properties of the model A-sites in bacterial and human mitochondrial systems [45]. We have found a specific S-turn (S-turn type 2) RNA topology motif that is present in the human mitochondrial structures and is not seen in bacterial and human cytoplasmic ribosome models. The human cytoplasmic A-site model that we analyze in this work has different nucleotide sequence and adopts quite different configurations; in particular it does not contain the S-turn motif. So far, to the best of our knowledge, the internal mobility of the human cytoplasmic A-site has not been computationally studied. Therefore, herein we aimed to describe and compare physicochemical properties of the non-complexed bacterial and human cytoplasmic A-site. We limit our study to the antibiotic-free A-site models, because bacterial A-site complexes with antibiotics have been already extensively studied [13,40e42] and, there are no human A-site model structures with 4,5-linked aminoglycosides specifically bound in the RNA bulge. The aim was to identify structural-dynamics features of the A-site that account for non-negligible affinity of aminoglycosides toward the human target and for the differences between mRNA decoding in bacteria and human. To compare the dynamics of the ribosomal decoding site we performed MD simulations in explicit solvent of the bacterial and human cytoplasmic A-site models. 2. Methods 2.1. Nomenclature The simulated crystallographic models contain two A-sites that were denoted ‘site 1’ and ‘site 2’, as illustrated in Fig. 1. For

consistency in both bacterial and human ribosomal A-site variants we used the E. coli numbering of ribosomal RNA nucleobases. 2.2. System selection, construction and starting conformations As starting configurations we used crystallographic doublestranded oligoribonucleotide models containing two mirror, equivalent A-sites: (i) as a reference e bacterial model (PDB code: 3BNL [21]) resolved at 2.6 Å and (ii) human cytoplasmic model (PDB code: 2FQN [22]) resolved at 2.3 Å (see Fig. 1). The identical binding mode of aminoglycosides in the crystal structures of full bacterial ribosomes and in bacterial A-site constructs confirms the validity of the latter [14,46]. Also, note that superposition of the A-site canonical base pairs (G1405:C1496, C1407:G1494 and U1410:A1490) in the crystal structure of human A-site model (PDB:3BNL, site 2) and the arbitrarily chosen ‘full-ribosomal’ A-site of the same sequence (yeast ribosome, PDB:3O2Z) results in the heavy atom RMSD as low as 1.52 Å, which confirms validity of the human A-site model. The anti-parallel sequences of each strand in the crystal structures are identical (Fig. 1). The sites have identical sequence but their three dimensional conformations differ. In the bacterial variant we observe two inactive (see Introduction) adenine configurations: while in site 1 both A1492 and A1493 are flipped-in, in site 2 A1493 is flipped-in and A1492 is flipped-out. In the human cytoplasmic variant in site 2: A1492 and A1493 acquire an active extra-helical position while in site 1: A1492 and A1493 are flippedin and in addition the neighboring A1491 is flipped-out. Regarding the latter case, even though the crystal packing interactions were reported [22], a similar flipped-out conformation of A1491 also occurs in the crystals of a human A-site model in the complex with apramycin in different crystal context [47,48]. The simulated models consisted of 21 nucleotides in each strand (see Fig. 1). Terminal nucleotide pairs, crystal waters and ions were removed. The hydrogen atoms were added with tleap (AmberTools1.5 [49]) and their positions were optimized in vacuo with the sander program (of Amber package [50]). The energy optimization protocol consisted of 8000 steps of steepest descent method, followed by 2000 steps of conjugate gradient energy algorithm. Next, the model was placed in a truncated octahedral cell of explicit water with solvent layer of at least 10 Å from the solute surface. Final systems consisted of about 40 000 atoms, including about 13 000 waters. In the main series of the MD simulations neutralizing Kþ ions were added at positions of electrostatic potential minima using tleap. Also, in most of the MD simulations weak positional restraints (0.35 kcal/mol/Å2) were applied to the atoms of the termini at each end of the RNA duplex. This was done in order to stabilize the terminal base pairs and to mimic the constraints imposed by the h44 helix of the ribosomal surrounding, of which Asite is a part. 2.3. Molecular dynamics conditions and protocol MD simulations were performed at constant temperature of 300 K (ensured by Berendsen weak coupling algorithm with a time constant of 1.0 ps) and constant pressure of 1 atm (relaxation time of 1.0 ps) using sander and pmemd modules of the AMBER10 package. The systems were simulated in periodic boundary conditions using particle mesh Ewald method [51] with 9 Å cutoff for long range interactions. To allow for longer 2efs time step the SHAKE algorithm [52] was applied to all hydrogens. Center-of-mass motion was removed every 5 ps. This simulation protocol was tested and successfully applied in previous nucleic acid simulations [53e55]. In the first stage solvent atoms were minimized (1000 steps) then temperature was linearly increased from 100 to 300 K

J. Panecka et al. / Biochimie 112 (2015) 96e110

(100 ps) with 25 kcal/mol/Å2 restraints applied to the solute. Next, 5 rounds of alternate stages of minimization and 50-ps MD runs were performed with restraints on the solute gradually reduced (5, 4, 3, 2, and 1 kcal/mol/Å2) and finally e 50-ps MD with 0.5 kcal/mol/Å2 restraints followed by 50-ps MD without the restraints on the solute. The MD production phase of each double A-site model was at least 100 ns and the total aggregate length of the simulations was about 3.8 ms, which means twice as much sampling for single A-site. MD trajectories were saved each 10 ps. 2.4. Force field parameters and simulation conditions MD simulations were performed in various conditions upon changing: (i) RNA force field (Amber parm99 or bsc0cOL3), (ii) type of cations (Kþ or Naþ) and (iii) parameterization of water and ions. In the main set of simulations the RNA models were parameterized with the Amber bsc0cOL3 force field. The latter force field stems from parm99 [56] but includes two recent modifications: bsc0 [57], correcting the a/g behavior of nucleic acid backbone, and cOL3 that prevents the formation of the RNA ladder high-anti c structures [58]. The latter correction is considered to be essential for stability of long RNA simulations. The bsc0cOL3 has been the default RNA force field in the latest (ff10eff14) versions of the AMBER set of force fields. However, for comparison, we have also analyzed the dynamics of the RNA models simulated with the older parm99 force field, since it was used in many previous MD simulations that we refer to. The following solvent parameter combinations were compared: (i) by Joung et al. [59], i.e. for Naþ: radius r ¼ 1.369 Å and well depth E ¼ 0.0874 kcal/mol, for Cl: r ¼ 2.513 Å and E ¼ 0.0356 kcal/mol, with TIP3P water model [60]; (ii) Naþ by Åqvist et al. [61]: r ¼ 1.868 Å and E ¼ 0.00277 kcal/mol, with TIP3P water or (iii) Kþ: r ¼ 1.87 Å and E ¼ 0.1000 kcal/mol, Cl: r ¼ 2.47 Å and E ¼ 0.1000 kcal/mol (cf. Smith & Dang et al. [62,63]) together with SPC/E water model [64]. Finally, the influence of ionic strength was also tested. A summary of MD simulations and their conditions is shown in Table 1. 2.5. Data analysis The MD derived nucleotide conformations were clustered with k-means algorithm using MMTSB Toolset (http://mmtsb.org/). The method divides the set of conformations into separate clusters according to a chosen clustering criterion e we used the cartesian coordinate root mean square deviation (RMSD) with a cluster Table 1 Full list of MD simulations of the A-site models. No. Variant

PDB code Force field Ions

Main set of MD simulations 1 Bacterial 3BNL bsc0cOL3 Kþ (D,0.0 M) 2 Bacterial 3BNL bsc0cOL3 Kþ (D,0.0 M) 3 Human 2FQN bsc0cOL3 Kþ (D,0.0 M) 4 Human 2FQN bsc0cOL3 Kþ (D,0.0 M) Tests of MD conditions and force field 5 Bacterial 3BNL parm99 KCl (D,0.23 M) 6 Bacterial 3BNL parm99 NaCl (C,0.23 M) 7 Bacterial 3BNL bsc0cOL3 NaCl (C,0.23 M) 8 Bacterial 3BNL bsc0cOL3 KCl (D,0.23 M) 9 Human 2FQN bsc0cOL3 KCl (D,0.23 M) 10 Bacterial 3BNL parm99 Naþ (C,0.0 M) 11 Bacterial 3BNL parm99 NaCl (C,0.1 M) 12 Bacterial 3BNL bsc0cOL3 Naþ (C,0.0 M) 13 Bacterial 3BNL bsc0cOL3 NaCl (C,0.1 M) 14 Bacterial 3BNL parm99 Naþ (A,0.0 M) 15 Bacterial 3BNL parm99 NaCl (C,0.2 M)

Water Time (ns) Restraints SPC/E SPC/E SPC/E SPC/E

500 500 500 500

þ e þ e

SPC/E TIP3P TIP3P SPC/E SPC/E TIP3P TIP3P TIP3P TIP3P TIP3P TIP3P

280 150 300 250 230 100 100 100 100 100 100

þ þ þ þ þ þ þ þ þ þ e

Notation for ion parameters: C e by Joung & Cheatham [59], D e Smith & Dang [62,63] and A e Åqvist [61]. Ion concentration of 0.0 M means that only neutralizing cations were added.

99

radius of 2.5 Å. The number of clusters rises if RMSD criterion decreases. While using a specific cluster radius, the number of obtained clusters roughly corresponds to the conformational variability of the system; this interpretation was used throughout the text. The trajectories were analyzed with the ptraj module from AmberTools1.5 [49] and in-house Python, Perl and MATLAB scripts. Gnuplot (http://www.gnuplot.info/) was used to produce plots and VMD (http://www.ks.uiuc.edu/Research/vmd/) to visualize trajectories. Flipping of A1492 and A1493 was characterized with the pseudo-dihedral q angle (defined in Fig. 2a) similar to the one introduced by Song et al. [65] and previously successfully applied in the 2D umbrella sampling study investigating bacterial A-site dynamics [10]. As an example, for A1492 in the bacterial A-site the angle is defined by 4 points (see Fig. 2a): CM1 e the center of mass of the two flanking base pairs (1409:1491 and 1407:1494), CM2 e the center of mass of the phosphate of the flipping nucleotide (of A1492), CM3 e the center of mass of the phosphate of the next nucleotide (of A1493), and CM4 e the center of mass of the flipping nucleobase (A1492). The definition of q for A1493 and in the human A-site is analogical. Only heavy atoms are taken into consideration in the calculations of the centers of mass. If the nucleobase is flipped-in, the q angle is roughly in the range (50+,þ50+), and if the nucleobase is flipped-out q is close to ±180+. However, we need to emphasize that the above criteria are adapted from the study on DNA [65], and they do not necessarily exactly correspond to the flipped-in/out states of nucleobaes in such irregular motif as the RNA bulge. For instance, in the analysis of co-occurrence of the flipped-in/out states of A1492 and base-phosphate interactions we took more relaxed criteria for the A1492 flipped-in state, i.e. with the lower margin of 75+. The choice of this cutoff was based on visual analysis of the trajectory vs. q values. Also, note that the Asite adenines can adopt various intermediate states, as shown in Fig. 2b. In these intermediate states the nucleotide can be shifted towards one of the grooves, either stacked with the neighboring nucleobases or destacked. With the assumed q definition, if the adenines are located from the minor groove side, they acquire negative q angle values, and if they are located from the major groove side they roughly acquire positive q values. The distance between the centers of mass of the nucleobases was used as the descriptor of intra-strand nucleobase stacking, with the distance criterion of 5.0 Å. Hydrogen bonds (H-bonds) were detected with ptraj using 3.5-Å distance cutoff and 120+ cutoff for the dihedral angle (H-donor)dhydrogen,,, (H-acceptor). The observed base pairs were characterized with the Leontis classification [66]. The following abbreviations describing the configurations of base pairs were used: (i) sugar edge e ‘S’, (ii) Hoogsteen edge e ‘H’ and (iii) Watson-Crick edge e ‘W’; ‘c’ for cis and ‘t’ for the trans configuration. The nomenclature for base-phosphate interactions was based on the work of Zirbel et al. [67]. The MINT software (https://bionano.cent.uw.edu.pl/Software/MINT) was used to automatically calculate the occupancies of different basepair types in the MD trajectories. Water densities near the U1406:U1495 base pair were calculated with ptraj after heavy-atom superposition of G1405:C1496 and C1407:G1494 base pairs, and Kþ ion densities e after superposition of nucleotides 1407e1410 and 1490e1494. 3. Results 3.1. General properties under different simulation conditions The simulated oligonucleotide duplexes were globally structurally stable in any studied conditions, yet internally quite flexible.

100

J. Panecka et al. / Biochimie 112 (2015) 96e110

Fig. 2. Mobility of A1492 and A1493 versus simulation time. (a) Definition of the q pseudo-dihedral angle used to characterize nucleobase flipping [65]. For A1492 in the bacterial A-site, the angle is defined by 4 centers of masses (CMi) of certain atom subsets listed in Methods. For each adenine the flipped-in region is defined as (50+,50+) and is marked by grey guides in the plot. (b) The q pseudo-dihedral angle versus simulation time for different MD variants (simulations #1e#4 in Table 1). The typical direction of flipping (in the translation process) through the minor groove is indicated by a continuous arrow, and the opposite direction e by a dashed arrow. For definition of sites see Fig. 1. The positions of adenines in the initial X-ray structures are marked on the y-axis by colored semicircles. The major MD-observed flips of A1492 are indicated by red dashed lines.

The average heavy-atom RMSD values in the range 3.0e5.0 Å show that the systems relaxed from their initial crystal structures (Table S1). However, the RMSD plots (for the simulations #1e#4, in Fig. S2) confirm overall conformational stability of the simulations. As expected, the nucleotides forming the A-site bulge exhibited higher variability of inter-nucleobase interactions than the canonically paired ones (base pair types are listed in Tables S2 and S3). Superposed cluster representative structures shown in Fig. S3 also prove that in most cases the positions of nucleobases in the A-site bulge varied distinctly. The clustering data shown in Fig. S3 confirm that the weak restraints applied to the terminal heavy atoms of the RNA models reduced their flexibility, i.e. the restrained systems yielded fewer conformational clusters. Also, such positional restraints, impose some permanent asymmetry to the system, which may lead to a different dynamics of the two sites in the same RNA model (besides the differences due to the starting structures and sampling). This is, for instance, visible from large differences between the numbers of clusters for the two sites of the same model in the same simulation (Fig. S3). Before selecting final MD conditions we investigated the dynamical behavior of the systems with different RNA and solvent force-fields, and with different types and concentrations of ions. As noted in Table 1, we performed 15 simulations of different lengths, mostly for the bacterial variant. Based on these data, we decided to focus primarily on four simulations performed with the bsc0cOL3 RNA force field [56e58] in presence of Kþ with parameters by Dang

et al. [63] and using the SPC/E solvent model [64]. These simulations were extended to 500 ns. This choice was based on the literature data and the following observations. We noticed that the bsc0cOL3 force field made the RNA nucleotides less flexible than the parm99 (see Fig. S4 for the influence of force field on the residual fluctuations). Additionally, in some simulations with parm99 we clearly observed an increase of the glycosidic c angle. For instance, in a 100-ns simulation with neutralizing Naþ ions, in RNA strand 1 of site 1 the c angle increased by over 20 degrees (Table S1). This suggests the formation of a ‘ladder’-like structure, a known irreversible force field artifact in the RNA simulations carried out without the c-correction [58,68,69]. In contrast, no significant rise of the c angle was observed in the simulations with the bsc0cOL3 correction, consistent with previous MD studies [58,69]. We also investigated the a/g backbone inter-conversions. As expected, these backbone flips occurred more frequently with the parm99 than with the bsc0cOL3 force field (Fig. S5). In both force fields the flips were usually reversible, with the exception of the highly mobile A1492 and A1493 (Fig. S5). In the X-ray structures of bacterial ribosomes, these adenines indeed adopt a multitude of different a/g dihedral angle values (Fig. S6). However, we have to keep in mind the limited resolution of these X-ray structures (typically over 3.0 Å); hence some of the dihedral angles may have been incorrectly determined from the electron densities. On the other hand, even if both of these a/g backbone conformations are

J. Panecka et al. / Biochimie 112 (2015) 96e110

likely to occur for A1492 and A1493, we expect this dynamical behavior to be reversible, which was not observed with the parm99 RNA force field. Therefore, we suggest that the bsc0cOL3 force field may better reflect the properties of the RNA A-site bulges, while the parm99 version may result in unsatisfactory trajectories, in line with ref. [70] and [71]. For an overview of the nucleic acids force fields see Ref. [72]. Although the previously used parm99 provides higher flexibility of the decoding site, this appears to be a consequence of its lower accuracy, with the c angle shift being considered a serious problem, since it is irreversible and leads to spurious structures. Our tests showed that solvent affects some dynamical aspects of the A-site systems. Interestingly, with the bsc0cOL3 force field the fluctuations (Fig. S4) were rather insensitive to the ion concentration increase, with the average RMSF at similar level for the two tested ion concentrations (even lower by about 0.2 Å for the lower ionic strength), in contrast to parm99, with average RMSF higher by 0.4 Å in lower (vs. higher) ionic strength. Additionally, in the simulations of the bacterial A-site with 0.23 M Naþ concentration (ion parameters by Joung et al. [59], MD simulations #6 and #7 in Table 1), the initially flipped-out A1492 remained in the extrahelical position during the entire trajectory, in contrast to all the other simulations. This behavior did not depend on the RNA force field (bsc0cOL3 or parm99). What probably prevented larger scale rearrangements of this nucleotide was a high sodium ion concentration region close to the RNA phosphate of A1492 (see Fig. S7). A similar effect was previously observed by Noy et al. [73] in MD simulations of DNA with the equivalent ion parameters for the SPC/ E model and with higher, 0.5 M, concentrations of Naþ. As suggested by the authors, the ion parameters of Joung and Cheatham may have an increased tendency to directly interact with anionic phosphates [73]. The persistence of A1492 in the flipped-out conformation is unexpected in the light of the current knowledge since, e.g., the flipped-in conformations of both A1492 and A1493 were shown to be energetically preferred [8]. In our previous studies of the models of uncomplexed bacterial A-sites we also observed that A1492 prefers intra-helical states [45,74]. Similar result was suggested by experimental structures without bound cofactors (e.g. Refs. [75] and [76]). These observations made us focus on the simulations with Kþ ions and parameters by Dang et al. [63], in which A1492 flipped in, similarly to the previous simulations of the bacterial A-site [40,45,53,74]. Another reason for using potassium ions (instead of sodium) is that they are dominant in living cells [77]. Obviously, when assessing our tests, we need to keep in mind that our study is also affected by sampling limits. Therefore, our tests of the solvent/ion conditions should not be taken as converged results. Nevertheless, since our tests are consistent also with the cited literature, we believe that our choice of simulation conditions for the main simulations is justified. 3.2. Conformational mobility of A1492 and A1493 in the bacterial A-site

101

restraints, the MD-derived conformations resembled the NMR structure determined by Fourmy et al. (PDB code 1A3M [75]), i.e. A1492 and A1493 were relatively well stacked with each other and partly destacked from the neighboring nucleobases (Fig. 3b and Fig. S8) Full flipping-out was probably disfavored by the interactions of A1492 and A1493 WC edges mostly with A1408 and C1409 (for base pair occupancies see Table S3). In contrast to the configurations in the NMR structure [75], A1492 and A1493 were initially in a syn conformation (Fig. 3b) but then converged to anti. Due to stable base pairing with A1408, the nucleobase A1493 did not flip, i.e. remained in the initial intra-helical configuration. In contrast, in most MD simulations, the initially extra-helical A1492 in site 2 (Fig. 1, bottom) flipped in within the first few nanoseconds. Moreover, for A1492 we also observed a single full flipping-out event at about 180 ns of the unrestrained MD (see flipping angle plotted in Fig. 2b) and a few destacking events of this adenine in both simulations. This behavior is consistent with previous MD simulations of the entire bacterial ribosomal h44 helix, including Asite [53], in which repetitive flipping of A1492 and no flipping of A1493 was observed in 30-ns long simulations (with parm99 force field). On the whole, the experimental and computational data do not clearly resolve the adenine flipping time scale. Fluorescence anisotropy experiments performed on the bacterial A-site model suggested sub-nanosecond time scale of adenine movements [15,78]. In contrast, NMR data for the small A-site models and the recent 2D umbrella sampling simulations indicated that full adenine flipping occurs on the micro-millisecond time scale [10,79]. The authors associated the nanosecond time scale transitions, observed previously by fluorescence methods, with destacking events. This is consistent with the fact that we observed a few adenine destacking events and a single flip on the 100-ns time scale. Our current findings and our previous work on the mobility of A1492 and A1493 in the ribosomal surrounding [74] imply that A1492 may be more flexible than A1493. Further, A1492 flipping in our trajectories is consistent with the observation of two A1492 states (flipped-in and -out) in the crystallographic work of Shandrick et al. [80]. The latter study suggests that these two configurations may be similar in terms of free energy. However, replicaexchange and umbrella sampling MD suggest a lower free energy barrier for flipping of A1493 than A1492 [8,10]. At this moment, we cannot fully resolve the origin of the differences between various simulation studies. Obviously, our standard MD simulations may still provide insufficient sampling, which would be necessary to derive free energies and barriers from the populations. However, also the earlier studies may have their limitations, since the balance of the adenine flipping may depend on the simulation conditions such as the model used, type and strength of the restraints, on the enhanced sampling methods, and also on the force field version [40,74]. 3.3. Overall internal mobility of the human A-site model

Enhanced sampling simulations of the uncomplexed bacterial A-site have been performed [8,10] but constant temperature MD simulation studies [40,53] reached only 20e30 ns time scales. The current study of the bacterial A-site conformational dynamics extends this scale to 100e500 ns, and in addition, for the first time we use the AMBER force field with the corrected ceprofile. In the bacterial A-site A1492 and A1493 adopted intra-helical conformations most of the simulation time (Figs. 2 and 3). In all trajectories A1493 interacted with A1408 via WC edges for about 45e60% of simulation time (Table S3). A1492 also paired with A1408 but less frequently (7e30% of time), consistently with the previous MD study [40]. Moreover, in site 2 (Fig. 1) simulated with

The loss of one canonical G:C base pair in the bulge of the human A-site in comparison with the bacterial one (Fig. 1) enables larger tertiary structure freedom of the human variant. This may be the reason for the trapping of the two human A-sites in distinctly different conformations in the crystal structure (see Fig. 1, right and ref. [22]). Therefore, the simulations performed for these different starting states allowed exploring a larger part of the complex A-site conformational landscape. Indeed, the two initially differing nucleotide arrangements, simulated either with or without restraints, gave different dynamical behavior of the A-site nucleobases (see trajectory clustering in Fig. S3). Specifically, site 2 of the

102

J. Panecka et al. / Biochimie 112 (2015) 96e110

Fig. 3. Typical nucleotide conformations and occurrences of A1492:A1408 versus A1493:A1408 cWW base pairs in the bacterial A-site in MD simulation with restraints (#1 in Table 1). In panel (b) (site 2) syn configurations of glycosydic angle for both A1492 and A1493 are indicated in red. Corresponding plots for the simulation without restraints (#2 in Table 1) are shown in Fig. S17. For clarity only polar hydrogens are shown. For the nucleotide sequence and definitions of site 1 and site 2 see Fig. 1.

human variant (with A1492 and A1493 initially flipped-out) was more conformationally variable than site 1 (with A1492 and A1493 initially intra-helical). Furthermore in the first 300 ns of MD trajectory, the human A-site system yielded on average more clusters than the bacterial one: 6 versus 3, respectively (Fig. S3). However, this conformational diversity was reduced in the remaining part of the trajectories, between 300 and 500 ns. Clustering for each bacterial site gave 2 or 3 conformational clusters (either with or without restraints). The restrained MD simulation of the human variant yielded 4 (site 1) and 3 clusters (site 2) and unrestrained MD e only one cluster for each A-site. Furthermore, the A1492/A1493 flipping dynamics in the human A-site differed from the bacterial one. In site 1 of the human variant (simulations #3 and #4 in Table 1), A1492/93 did not flip at all (Fig. S9). The lack of flipping in site 1 may be partly explained by a specific atypical conformation of site 1 resulting from crystal packing (discussed further). On the other hand, in site 2, the A1492/ A1493 flipping-in movement toward the final intra-helical configuration was more complex than in the bacterial variant (see q angle plots in Fig. 2).

3.4. Flipping movement of A1492 and A1493 in the human A-site may also occur through the major groove In site 2 of the human A-site crystal structure both A1492 and A1493 are in an extra-helical position (see Fig. 1). This site 2 was very flexible in all trajectories and the adenines finally flipped inside the bulge from their initial external configuration (cf. the q flipping angles in Fig. 2). However, the flipping movement occurred through a different route depending on the application of restraints. In site 2, in MD simulation of the human variant without restraints, the adenines flipped through the minor groove (Fig. 2). We expected this path from the analysis of the conformations of A1492/93 in the experimental structures of the bacterial and eukaryotic ribosomes, and of model A-sites (e.g. Refs. [74], [81] or [82]). Since the ribosome structure and mRNA decoding process are conserved to significant extent across all species, we anticipated a similar A1492/93 movement in the studied human A-site. However, in one simulation of the human A-site with restraints (#3 in Table 1) a major-groove adenine flipping pathway occurred

Fig. 4. Flipping-in of A1492 and A1493 via major groove in site 2 of the human A-site model simulated with restraints (MD trajectory #3 in Table 1). For clarity, hydrogens are not shown. For nucleotide sequence see Fig. 1.

J. Panecka et al. / Biochimie 112 (2015) 96e110

103

Fig. 5. Typical conformations and interactions in the human A-site model (site 2), simulated with restraints (MD trajectory #3 in Table 1). (a) representative conformation of the most occupied cluster with the interactions of A1492:C1409:A1491 triple in grey circle; (b) the conformation with destabilized A1493:G1408 base pair (grey circle) and with A1493(N7)eG1408(N2) H-bond characteristic of tHS base pairing; for clarity only polar hydrogens are shown. (c) Distances A1493(N6)eG1408(O6) (in magenta) and A1493(N7)e G1408(N2) (in blue) characteristic for H-bonds formed in cWW and tHS base pairs, respectively (marked also in (a) and (b)). For the nucleotide sequence and the definition of sites see Fig. 1.

(Figs. 2 and 4) resulting in a stable flipped-in conformation (see Fig. 5a and further discussion). Obviously, the movement through the major groove could not happen in the ribosome during decoding because the pathway would be obstructed by the bound mRNA. The observed A1492/A1493 flipping-in through the major groove may be specifically facilitated by the interactions of the adenines with termini that occurred in the simulation #3 of the human A-site model. We did not observe the major groove flipping pathway in any other simulation from the present set. However, we have previously observed such flipping of A1492 and A1493 via the major groove in another MD study for the bacterial A-site simulated in the context of the neighboring ribosome components [74]. In that work the fragment of the h44 helix with the A-site was restrained at the termini, and was simulated without mRNA and Asite-tRNA. The above observations indicate that the type of the simulated A-site model is probably not the sole reason for A1492 and A1493 flipping through the major groove. Therefore, the observed ‘major groove’ pathway may be accessible in the uncomplexed A-site, albeit from our simulation data it seems to be less probable. 3.5. Atypical conformation of the human A-site model with the flipped-out A1491 The human variant site 1 conformation (Fig. 1) is distinctly different than the human site 2 one and also the studied bacterial

one: A1491 is flipped-out, while A1492 and A1493 reside inside the bulge. A direct reason for the flipped-out position of A1491 may be the crystal packing interactions reported in this structure [22]. Additionally, secondary structure of the human A-site comprises more non-canonically paired nucleobases than the bacterial A-site (Fig. 1), which may facilitate this atypical conformation of A1491. Finally, the MD-derived average PeP distances of the base-paired nucleotides are distinctly lower in the human site 1 than in the bacterial A-sites (Fig. S10). Therefore, the base pairing observed in the human A-site renders this RNA bulge more compact than the bacterial one, possibly favoring the flipped-out A1491 conformation. Here, we analyzed if the flipped-out position of A1491 may be energetically preferred in the isolated human A-site, free from the crystal contacts. During 500 ns with positional restraints on termini, the initial flipped-out conformation of A1491 did not change significantly. On the contrary, in the simulation without any positional restraints, A1491 flipped in irreversibly at about 80 ns (compare the initial conformation in Fig. 1 and the final one in Fig. 6d). The different behavior of A1491 in these two simulations may be explained by the fact that unwinding of the RNA helix is necessary for the A1491 flip to occur. In the unrestrained MD we observed the decrease of the helical twist for the base pairs flanking A1491 (A1490:U1410 and G1494:C1407) from about 120 to 100 degrees (Fig. S11). Additionally, the major groove between A1491 and C1404 irreversibly widened upon the A1491 flipping event (Fig. S11).

104

J. Panecka et al. / Biochimie 112 (2015) 96e110

Fig. 6. Conformations of the human A-site, simulated without restraints (MD trajectory #4 in Table 1). (a) Flipping movement of A1492 with frequent interactions marked. (b) A plot and diagram showing co-occurrence of the A1492(O2P)eG1408(N1) 4BPh interaction and flipped-in/out states of A1492 in the time range when the A1492 reversible destacking occurred. The grey shaded area marks the q angle range for which A1492 is flipped-in, consistently with the computed percentages on the right (see Methods). Similar conformations in site 1 and site 2: (c) A1492 fully flipped-out in site 2 and (d): similar interaction network in the final configuration of site 1, but with A1492 flipped-in. For clarity only polar hydrogens are shown. For the nucleotide sequence see Fig. 1.

The above observations are consistent with the fact that in site 1 of the initial crystal structure the backbone and nucleobase oxygens in both RNA strands interact with a positively charged cobalt hexammine(III) ion captured in the major groove (Fig. S12a). Interactions with this ligand may result in a too narrow major groove shape to accommodate A1491 in the initial structure. This observation further agrees with the occurrences of the flipped-out position of A1491 in other crystal structures of the human A-site but complexed with apramycin [47,48] shown in Fig. S13. Apramycin is composed of four linearly connected rings, and is quite long and narrow. In the complexes with apramycin [47,48], due to crystal packing, A1491 stacks with the nucleotides from the neighboring RNA duplex. This external A1491 conformation is observed for two different binding modes of apramycin: either in the vicinity of A1492 and A1493 (Fig. S13a) or distinctly below these adenines (Fig. S13b and c). Further, the space groups of these crystals [47,48] are different, so the crystal contacts with the surrounding are different. Therefore, we suggest that the flipped-out position of A1491 is not due to specific interactions with ligands or specific crystal packing contacts. It might rather be one of the several competing intrinsically stable conformations, which may be selected as the global minimum, for instance, by binding of the charged small ligands (like cobalt hexammine(III) or apramycin) or by non-specific contacts with the RNA surrounding.

3.6. Two A1493:G1408 base pairing patterns in the human A-site model: cWW versus tHS In site 2 of the human variant once A1493 flips in (MD #4 without restraints in Table 1), its q angle levels and no further conformational changes of this nucleotide occur (Fig. 2). After this flipping event, A1493 creates a stable AG tHS base pair (known also as sheared GA base pair) with G1408 (Fig. 6a). This pair is stable in three analyzed sites (with one exception of site 2 in MD #3 in Table 1, discussed further). In site 1 (with A1492 and A1493 initially flipped-in) the geometry of A1493:G1408, close to tHS, was already observed in the crystal structure. Also, in our previous MD simulations of the A-site model, we noted the tHS configuration of A1493:G1408 despite a different simulated system setup [40], i.e. a bacterial A-site crystal structure was used (PDB code:1J7T [14]), the initially bound antibiotic was removed, and the ‘eukaryotic-like’ mutation A1408G was introduced in silico. In site 2 of the human simulation with restraints, in contrast to all other sites of the simulated human models, A1493 and G1408 formed a cWW base pair up to about 370 ns (Fig. 5a and c). This type of base-pairing was established probably because A1492 and A1493 flipped through the major, not the minor, groove (Fig. 4). From this side the sugar edge of G1408, necessary for the formation of the tHS contact, is poorly accessible for A1493 because of steric

J. Panecka et al. / Biochimie 112 (2015) 96e110

obstacles. Still, at certain trajectory intervals (at about 100, 150 and 180 ns, Fig. 5b and c), the cWW base pair broke and tHS pair was almost formed. In the end at about 370 ns the transient cWW base pair finally broke and A:G converged to the more typical tHS pair, which corresponds to the plateau in the plot of the two distances shown in Fig. 5c. Interestingly, this base pair rearrangement also triggered a substantial shift in the position of A1492 towards the major groove in both sites of the simulated model (visible in the plot of q flipping angles in Fig. 2). Overall, our simulations suggest that for A1493:G1408 the tHS base pair formation is more preferred than cWW, which was observed only transiently and only in site 2 of the human system simulated with restraints. Consistently, the quantum chemical calculations showed that gas-phase interaction energies for the tHS A:G base pair are the most favorable ones from the whole tHS base pair family [83]. For instance, the MP2-calculated interaction energy for tHS A:G is about 17 kcal/mol (with inclusion of the critically important sugar-base H-bond), which is comparable to the canonical A:U base pair [83] and to the cWW A:G base pair [84] (both about 15 kcal/mol). Further, the tHS A:G geometry is distinctly more frequent in the experimental RNA structures than cWW A:G [83,85]. The cWW A:G pair is quite flexible and crystal data show that the unpaired NH2 group of guanine, with partial sp3 pyramidalization, tends to get involved in tertiary interactions [85]. Moreover, the cWW base pair seems to require the supporting RNA context to be conformationally stable [85]. Therefore, in the light of the above findings, the transition from the cWW to a tHS configuration in our flexible MD system is expected. Frequent occurrence of A1493:G1408 base pair in the human Asite (higher than in the bacterial variant, Tables S2 and S3) suggests that it would be more difficult for A1493 in the human system to achieve the flipped-out (i.e. active) configuration necessary for the binding of a tRNA anticodon. Again, this could suggest that the higher decoding accuracy in mammals and yeast [29,32] may be connected to the conformational properties of the minimal A-site (apart from other larger scale ribosome factors). Additionally, the discussed base pairing differences between the bacterial and human A-sites may partly explain different binding affinities and translation activities of aminoglycosides. The tHS A1493:G1408 configuration, makes the human A-site narrower than the bacterial A-site with the cWW A1493:A1408 base pair scheme. In Fig. S10 we compared the average PeP distances of opposite backbones in these two A-site models. Since the shape of the rRNA bulge is also an important factor for binding of these antibiotics [26], the narrower bulge may contribute to the experimentally measured lower binding affinities of aminoglycosides in the human A-site in comparison with the bacterial one [19,86]. Moreover, the more stable base pairing of A1493 implicating its higher conformational stability, might be connected to lower destacking ability of A1492 in the human A-site in comparison with the bacterial one, observed in fluorescence experiments on the models with bound antibiotics [19,25]. 3.7. Repetitive destacking of A1492 in the human A-site In the simulations of the bacterial variant a few A1492 destacking events occurred, while A1493 was relatively stably bulged-in (Fig. 2). Similarly, in site 2 of the human variant, simulated without restraints, we observed repetitive destacking of A1492 from A1493. Such conformational change can be considered as an initial phase of full nucleotide flipping-out movement. Destacked A1492 often interacted with the sugar edge of U1410 (Fig. S14), which probably was the reason for the fact that A1492 fully flipped out only once (see Figs. 2b and 6c). A1492 destacking occurred together with other conformational changes like

105

rearrangements in base pairing patterns of A1491 and C1409 (Fig. S15) or formation of the base-phosphate (BPh) interaction between the A1492 backbone and nucleobase of G1408 (Fig. 6a). These interactions were formed in the human A-site systems in which the RNA bulge is more floppy than in the bacterial variant. If A1492 was fully flipped-in (stacked), then the 4BPh interaction (classification by Zirbel et al. [67]) between A1492 and G1408 was rarely detected: both these structural features occurred together only about 10% of trajectory time (Fig. 6b). On the contrary, A1492 in the destacked conformation was often found together with the 4BPh interaction (47% of simulation time). BPh interactions are often found to stabilize nucleobases in crystallographic RNA structures [67], so it is possible that also in the human A-site system they may be functionally important. It is also interesting that, potentially proximal to 4BPh interaction of guanine, a similar 6BPh interaction of adenine would be intrinsically weaker, by 7.0 kcal/ mol when using QM potential energy surface interaction data [67]. Thus, with A at position 1408 (like in the bacterial A-site) instead of G, such BPh interactions would be less favorable. A question arises as to the reasons for the discussed bi-stable dynamical behavior of A1492. First, the human A-site may be too small to accommodate both A1491 and A1492, if a stable tHS base pair (more compact than the cWW arrangement) is formed by A1493 and G1408. For the tHS base pair configuration the average P(A1493)eP(G1408) distance is about 16 Å, while in the case of the cWW pairing this distance reaches about 19.5 Å (Fig. S10). Therefore, due to the limited space, C1409 interacted either with A1491 or A1492 (in total 60 and 54% of time, respectively, Table S2). Furthermore, the mobility of A1492 in site 2 of the human model may be also enhanced due to the relatively weak intra-strand A1491A1492 stacking, if compared to G1491A1492 in the bacterial sites (Fig. S8). 3.8. Conformational preferences for the U1406:U1495 base pair The U1406:U1495 cWW base pair (Fig. 1) creates stabilizing interactions with aminoglycosides that bind in the A-site [13,40,41]. For instance, an earlier MD study [41] showed that the double U1406C/U1495A mutation destabilizes bound paromomycin. The complex of the bacterial A-site with an aminoglycoside paromomycin neighboring these two uracils is shown in Fig. S1. The U1406:U1495 base pair adopts two main conformations c1 and c2 shown in Fig. 7a. Both conformations are observed in the crystals of the A-site models, but only conformation c1 appears in the initial structures used for our simulations [21,22]. On the other hand, conformation c2 occurs almost exclusively in bacterial whole ribosome structures, even without any A-site bound ligands. Vaiana et al. [13] suggested that the U:U conformation c2, more prevalent in the X-ray data, may be the native one to which aminoglycosides bind in the bacterial ribosomal A-site. They speculated that the U:U conformation c1 was a crystallographic artifact. In our MD simulations of free A-site models both U1406:U1495 conformations appeared. The dominant one depended on the simulated site but conformation c2, typical for full ribosomal structures, was overall more frequent in the human A-sites, while c1 in the bacterial variants (Fig. 7b and Fig. S16). Furthermore, the dynamics of the U:U conformational c1 to c2 exchanges differed between the variants and sites. In particular, in the bacterial A-sites the exchanges were less frequent than in the human A-sites. In the latter they occurred on a 10 ns time scale. The only exception from the mentioned above ‘fast’ U:U exchange dynamics was the human site 2 in the simulation with restraints (#3 in Table 1). In this site the cWW type of A1493:G1408 base pairing predominantly formed. In all other systems the more compact tHS was present (compare conformations in Fig. 5a with

106

J. Panecka et al. / Biochimie 112 (2015) 96e110

Fig. 7. The conformational dynamics and hydration of U1406 and U1495. (a) Two typical base pair configurations of U1495:U1406 with schematic representation of paromomycin (PAR) binding mode in the A-site (‘w1’ and ‘w2’ denote water molecules); (b) U,,,U H-bond distances (characteristic of conformation c1) from MD simulations with restraints (#1 and #3 in Table 1; see also the remaining data, for MD trajectories #2 and #4, in Fig. S16). Red frame marks site 2 of the human model in which the uracils behave more similarly to the bacterial A-site. Occupancies of the two configurations are shown in each plot. In (c), (d) and (e) MD-derived water high density regions are shown for site 1 of the human A-site variant. Density isosurfaces are drawn at value 0.19 waters/Å3 in (c) and (d), and at value 0.1 waters/Å3 in (e).

Fig. 6a). In previous MD studies lower conformational stability of this U:U pair was also observed in the ‘eukaryotic-like’ A1408G Asite mutant which resembles the human model studied by us [40]. Notably, in these simulations A1493 and G1408 paired in the tHS configuration, which again agrees with our hypothesis that the frequent U:U exchanges are more preferred in the systems with tHS A1493:G1408 base pairing. Therefore, it seems that the conformational dynamics of these uracils depend on the geometry of the neighboring 1493:1408 base pair. In summary, the frequent conformational exchanges between the two U:U configurations in the human A-site may destabilize the contacts crucial for binding of the so-called neamine core which is an anchor for many A-site binding aminoglycosides (Fig. 7a). In contrast, in the bacterial A-site, possibly due to its overall lower flexibility, this uracil pair can be locked in one conformation for longer time intervals. This dynamical difference, due to the entropic contribution, may partly explain the lower binding affinity of aminoglycosides towards the human A-site target than to the bacterial target.

3.9. Hydration of the U1406:U1495 pair Solvation plays a critical role in the dynamics and conformational stability of RNA systems, including ribosomal A-site [87]. In many crystal structures of the aminoglycoside complexes with model A-sites, crystal waters mediating the interactions with RNA were observed (e.g., ref. [14]). Also, an earlier MD study showed

that some of the aminoglycosideeRNA H-bonds are stably water mediated [13]. The hydration patterns around the U1495:U1406 pair are presented in Fig. 7. Previous MD study [13] showed that in the c1 conformation of U1406:U1495 (Fig. 7a) the O4 carbonyl oxygen of U1406 directly hydrogen bonds to the O6 hydroxyl group of neamine ring II, whereas in the c2 conformation this O4 interacts with the paromomycin O6 oxygen through a water molecule (termed w2). This water molecule in the RNA major groove is characteristic of the crystal A-site complexes with 4,5-linked aminoglycosides such as paromomycin or neomycin [12]. Also, a water molecule at a similar position was captured in one A-site of the human crystal studied in this work. This water was probably stabilized by the cobalt hexammine(III) ion bound near the U1495:U1406 pair (Fig. S12a). However, in the ligand-free simulations of the bacterial and human A-site models, we did not observe high water density at the w2 location (Fig. 7a and d), i.e. in between the O4 oxygens. This observation confirms the results of Romanowska et al. [40], but on a longer 100-ns time scale, that this water molecule may be trapped at the w2 location only upon ligand binding. On the other hand, in the human and bacterial A-site simulations we observed high water densities in the minor groove close to O2 oxygens of U1495 and U1406 (the w1 water molecule in Fig. 7a and d). This structural water molecule bridging between the uracils was identified in previous computational studies of the bacterial Asite [13,40]. It was also captured in one of the sites in the crystal structure complexed with paromomycin [14]. This site adopts the c2 conformation (defined in Fig. 7a, bottom). On the other hand,

J. Panecka et al. / Biochimie 112 (2015) 96e110

this w1 water is not present in any of the crystal models studied in the current work [21,22], which have uracils in the c1 conformation (Fig. 7a, top). Although these experimental structures may suggest that the w1 water is stable only in the c2 conformation, we observed this water molecule for both U:U conformations (Fig. 7e), which is expected because these U:U conformations are symmetric. The reason for this discrepancy is not clear and may stem from some genuine differences between the X-ray crystallography and atomistic MD methods when detecting ordered hydration sites. Finally, we noted that the MD-derived hydration patterns do not overlap well with the positions of crystallographic waters resolved in both human and bacterial A-sites (Fig. 7d); this happens probably due to relatively high mobility of the simulated A-site bulges. Note, however, that positions of ordered water molecules in individual experimental structures of biomolecular systems are often variable, rendering the exact comparison with simulations difficult [88].

3.10. Positions of cations in the context of aminoglycoside binding For the RNA bulges that bind positively charged aminoglycosides, electrostatic interactions are particularly important [16]. The spatial distribution of the well-sampled monovalent cations in MD simulations, may correctly identify the pockets of negative electrostatic potentials favoring ion binding [89]. The locations of cation high density regions in free human and bacterial A-sites are shown in Fig. 8. These ion density maps show that in all sites, both bacterial and human, ions gather near the U:U pair and near the phosphates of A1493 and G1494 (see Fig. S1). These similarities between bacterial and human A-sites are expected because the sequence of A-site near the uracil pair is the

107

same for bacteria and human, as shown in Fig. 1, and in general e it is universally conserved [20]. The above high cation density regions overlap with ring II of paromomycin, a part of the neamine core (Fig. S1), which was shown to serve as an anchor for the A-site aminoglycosides [13], thereby confirming the electrostatic compatibility of the A-site binding cleft and neamine-containing aminoglycosides. Since we observed high ion densities in the free A-site, these electrostatic interactions may help accommodate ring II in the first stage of aminoglycosidee(A-site) recognition, in accord with previous Brownian dynamics studies [90]. On the other hand, the well-localized ions in either bacterial or human A-site must be displaced by an aminoglycoside upon binding and this, in contrast, may disfavor aminoglycoside complexation with both Asite variants. Therefore, the discussed above well-localized high ion densities oppositely contribute to aminoglycoside binding free energies, and based on our results it is impossible to estimate which one would be dominant. In summary, similar electrostatic properties of the human and bacterial A-sites agree with the experimentally derived modest differences in binding free energies for aminoglycosides between the human and bacterial A-sites (DD G of only about 2 kcal/mol [19,25]). Despite the above overall ion-distribution similarities between the human and bacterial models, some particular ion densities do differ between the bacterial and human A-sites. First, as in the work of Romanowska et al. [40], we detected high cation density region near G1408 (in grey circles in Fig. 8b and Fig. S1) that appears only in the human A-site, due to the A1408G sequence variation. One could think of a modified aminoglycoside with a hydrophobic substituent close to this position that could destabilize this antibiotic binding toward the human A-site. Additionally, ion distributions in the bacterial A-site are similar for site 1 and site 2 of the simulated model and independent of the application of restraints in the simulations (Fig. 8a). In contrast, in the human A-site model, due to the overall higher mobility of nucleotides discussed earlier in this paper, ion densities depend on the simulation and site (Fig. 8b). These less precisely localized cation distributions may be one explanation for the, previously mentioned, slightly less energetically favorable binding of aminoglycosides toward the human than to bacterial A-site [19,25]. 4. Conclusions

Fig. 8. High density regions of the Kþ ions observed in MD simulations #1e#4 in Table 1. Isosurfaces are shown for ion concentrations of 0.00625 ion/Å3. The starting structures are shown as a reference. Ion density surfaces were calculated after superposition of heavy atoms; separately for site 1 and for site 2. Paromomycin is shown in a surface representation and was superposed from the PDB crystal structure 1J7T. High cation density regions near G1408, observed only in the human A-site, are enclosed in grey circles.

We performed explicit-solvent MD simulations of the bacterial and human cytoplasmic ribosomal double A-site models, totaling to about 3.8-ms. We tested various simulation conditions such as ionic strength, water models and nucleic acid force fields (parm99 and bsc0cOL3). In particular, with the earlier widely used parm99, we observed a tendency to form a spurious high-anti c ladder-like structure and irreversible a/g flips for certain nucleotides. The parm99 force field also gave higher atomic fluctuations upon decreasing ionic strength. In the end we applied the state-of-theart RNA force field bsc0cOL3. In the human A-site model the flipping-in movement of A1492 and A1493 from the initially flipped-out configuration was more complex and less ordered than in the bacterial A-site. However, this different dynamics may be partly attributed to the differing initial MD conformations of A1492 and A1493. After flipping-in of the adenines, the stability of A1493:G1408 base pair was higher in the simulations of the human A-site than of the corresponding A1493:A1408 in the bacterial variant. The more stable base pairing in the human model may partly explain lower aminoglycoside activities towards the human A-site and overall higher decoding accuracy in mammals (including human). Also, in both human and bacterial variants, typically A1492 was more mobile than A1493. This is consistent with some experimental data [80] and some

108

J. Panecka et al. / Biochimie 112 (2015) 96e110

simulations [40,53], though other studies suggest the opposite dynamical preferences [8,10]. We suppose that the exact relative adenine flipping frequency depends on the type of the studied model, and considering sampling limitations, on the initial coordinates. It may also be affected by the force field. The conformations of these adenines in the crystal structures of bacterial ribosomes and models are indeed variable [74]. Moreover, in the human A-site, we observed an interesting conformational change; a relatively frequent, repetitive destacking of A1492, associated with forming of the 4BPh base-phosphate interaction between the A1492 backbone and the G1408 nucleobase. We also investigated the flipped-out A1491 conformation that occurs in one site of the human model and is most probably enforced by crystal packing. In one MD simulation A1491 flipped in, which was associated with irreversible widening of the major groove and distinct decrease of helical twist. Other crystal structures [47,48] also suggest that this flipped-out A1491 conformation may be correlated with narrowing of the major groove caused by binding of certain ligands (such as apramycin and cobalt hexammine(III)). We found that the U1406:U1495 base pair reversibly changes its conformation both in the human and bacterial variant. More frequent U:U conformational exchanges were observed in the human variants, but only when tHS and not cWW A1493:G1408 base pair was formed. We hypothesize that these frequent U:U conformational changes in the human variant may contribute to weaker aminoglycoside binding to the human A-site than to the bacterial one [19] because this U:U pair forms crucial interactions with aminoglycosides. On the other hand, the water distributions near U1495:U1406 did not differ significantly between the human and bacterial variants, despite different mobility of uracils in these systems. A water molecule positioned in the minor groove was observed for both conformations of uracils. However, the other water molecule present in the complexes with aminoglycosides was absent in our antibiotic-free systems, confirming that this hydration site is occupied only upon ligand binding. Cation distributions were found to be similar in both studied variants, albeit it is more disordered in the human A-site due to its overall larger flexibility. Our study suggests sequence-dependent, structural, hydration, electrostatic, and dynamical reasons for different propensities of aminoglycosides toward human and bacterial ribosomal A-sites. While studying the intrinsic dynamics of these small RNA bulges, we have seen that even small sequence differences influence their volume, shape and mobility. In the future, the simulation methods increasing sampling might be used to explore more fully the conformational space of these RNA bulge systems. Conflict of interest The authors declare no conflict of interest. Acknowledgments blova  for help and advice in JP would like to thank Kamila Re preparing and running preliminary MD simulations and for valuable discussions. JT and JP acknowledge support from the National Science Centre (DEC-2012/05/B/NZ1/00035 and DEC-2011/01/N/ NZ1/01558) and Foundation for Polish Science TEAM/2009-3/8 project co-financed by European Regional Development Fund operated within Innovative Economy Operational Programme. This publication was also co-financed with the European Union funds by the European Social Fund (to JP). JS acknowledges support from the Grant Agency of the Czech Republic (P305/12/G034). JS institutional support was obtained by ”CEITEC e Central European Institute of

Technology” (CZ.1.05/1.1.00/02.0068) from European Regional Development Fund. Calculations were performed using the resources at the University of Warsaw (KDM/ICM grant no G31-4) and the Academy of Sciences of the Czech Republic. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.biochi.2015.02.021. References [1] K.H. Nierhaus, D.N. Wilson (Eds.), Protein Synthesis and Ribosome Structure: Translating the Genome, John Wiley & Sons, Inc, 2006. [2] P.C. Whitford, P. Geggier, R.B. Altman, S.C. Blanchard, J.N. Onuchic, K.Y. Sanbonmatsu, Accommodation of aminoacyl-tRNA into the ribosome involves reversible excursions along multiple pathways, RNA 16 (2010) 1196e1204. [3] J.M. Ogle, A.P. Carter, V. Ramakrishnan, Insights into the decoding mechanism from recent ribosome structures, Trends Biochem. Sci. 28 (2003) 259e266. [4] T.V. Budkevich, J. Giesebrecht, E. Behrmann, J. Loerke, D.J.F. Ramrath, T. Mielke, J. Ismer, P.W. Hildebrand, C.-S. Tung, K.H. Nierhaus, K.Y. Sanbonmatsu, C.M.T. Spahn, Regulation of the mammalian elongation cycle by subunit rolling: a eukaryotic-specific ribosome rearrangement, Cell 158 (2014) 121e131. [5] H.S. Zaher, R. Green, Fidelity at the molecular level: lessons from protein synthesis, Cell 136 (2009) 746e762. [6] N. Demeshkina, L. Jenner, E. Westhof, M. Yusupov, G. Yusupova, A new understanding of the decoding principle on the ribosome, Nature 484 (2012) 256e259. [7] W. Zhang, J.A. Dunkle, J.H.D. Cate, Structures of the ribosome in intermediate states of ratcheting, Science 325 (2009) 1014e1017. [8] K.Y. Sanbonmatsu, Energy landscape of the ribosomal decoding center, Biochimie 88 (2006) 1053e1059. [9] A.C. Vaiana, K.Y. Sanbonmatsu, Stochastic gating and drug-ribosome interactions, J. Mol. Biol. 386 (2009) 648e661. [10] X. Zeng, J. Chugh, A. Casiano-Negroni, H.M. Al-Hashimi, C.L. Brooks 3rd, Flipping of the ribosomal A-site adenines provides a basis for tRNA selection, J. Mol. Biol. 426 (2014) 3201e3213. [11] K.Y. Sanbonmatsu, Flipping through the genetic code: new developments in discrimination between cognate and near-cognate tRNAs and the effect of antibiotics, J. Mol. Biol. 426 (2014) 3197e3200. [12] B. Francois, R.J.M. Russell, J.B. Murray, F. Aboul-ela, B. Masquida, Q. Vicens, E. Westhof, Crystal structures of complexes between aminoglycosides and decoding A site oligonucleotides: role of the number of rings and positive charges in the specific binding leading to miscoding, Nucleic Acids Res. 33 (2005) 5677e5690. [13] A.C. Vaiana, E. Westhof, P. Auffinger, A molecular dynamics simulation study of an aminoglycoside/A-site RNA complex: conformational and hydration patterns, Biochimie 88 (2006) 1061e1073. [14] Q. Vicens, E. Westhof, Crystal structure of paromomycin docked into the eubacterial ribosomal decoding A site, Structure 9 (2001) 647e658. [15] M. Kaul, C.M. Barbieri, D.S. Pilch, Aminoglycoside-induced reduction in nucleotide mobility at the ribosomal RNA A-site as a potentially key determinant of antibacterial activity, J. Am. Chem. Soc. 128 (2006) 1261e1271. [16] G. Yang, J. Trylska, Y. Tor, J.A. McCammon, Binding of aminoglycosidic antibiotics to the oligonucleotide A-site model and 30S ribosomal subunit: Poisson-Boltzmann model, thermal denaturation, and fluorescence studies, J. Med. Chem. 49 (2006) 5478e5490. [17] M. Kaul, C.M. Barbieri, J.E. Kerrigan, D.S. Pilch, Coupling of drug protonation to the specific binding of aminoglycosides to the A site of 16 S rRNA: elucidation of the number of drug amino groups involved and their identities, J. Mol. Biol. 326 (2003) 1373e1387.  ski, P.M. Dominiak, J. Trylska, Electrostatic in[18] M. Kulik, A.M. Goral, M. Jasin teractions in aminoglycoside-RNA complexes, Biophys. J. 108 (2015) 655e665. [19] M. Kaul, C.M. Barbieri, D.S. Pilch, Defining the basis for the specificity of aminoglycoside-rRNA recognition: a comparative study of drug binding to the A sites of Escherichia coli and human rRNA, J. Mol. Biol. 346 (2005) 119e134. [20] J.J. Cannone, S. Subramanian, M.N. Schnare, J.R. Collett, L.M. D'Souza, Y. Du, B. Feng, N. Lin, L.V. Madabusi, K.M. Muller, N. Pande, Z. Shang, N. Yu, R.R. Gutell, The comparative RNA web (CRW) site: an online database of comparative sequence and structure information for ribosomal, intron, and other RNAs.,, BMC Bioinforma. 3 (2002) 2. [21] J. Kondo, E. Westhof, The bacterial and mitochondrial ribosomal A-site molecular switches possess different conformational substates, Nucleic Acids Res. 36 (2008) 2654e2666. [22] J. Kondo, A. Urzhumtsev, E. Westhof, Two conformational states in the crystal structure of the Homo sapiens cytoplasmic ribosomal decoding A site, Nucleic Acids Res. 34 (2006) 676e685.

J. Panecka et al. / Biochimie 112 (2015) 96e110 [23] S.N. Hobbie, S.K. Kalapala, S. Akshay, C. Bruell, S. Schmidt, S. Dabow, A. Vasella, P. Sander, E.C. Bottger, Engineering the rRNA decoding site of eukaryotic cytosolic ribosomes in bacteria, Nucleic Acids Res. 35 (2007) 6086e6093. [24] S.N. Hobbie, P. Pfister, C. Bruell, P. Sander, B. Francois, E. Westhof, E.C. Bottger, Binding of neomycin-class aminoglycoside antibiotics to mutant ribosomes with alterations in the A site of 16S rRNA, Antimicrob. Agents Chemother. 50 (2006) 1489e1496. [25] M. Kaul, C.M. Barbieri, D.S. Pilch, Fluorescence-based approach for detecting and characterizing antibiotic-induced conformational changes in ribosomal RNA: comparing aminoglycoside binding to prokaryotic and eukaryotic ribosomal RNA sequences, J. Am. Chem. Soc. 126 (2004) 3447e3453. [26] D.H. Ryu, R.R. Rando, Decoding region bubble size and aminoglycoside antibiotic binding, Bioorg. Med. Chem. Lett. 12 (2002) 2241e2244. [27] S.N. Hobbie, C. Bruell, S. Kalapala, S. Akshay, S. Schmidt, P. Pfister, E.C. Bottger, A genetic model to investigate drug-target interactions at the ribosomal decoding site, Biochimie 88 (2006) 1033e1043. [28] S.T. Gregory, J.F. Carr, D. Rodriguez-Correa, A.E. Dahlberg, Mutational analysis of 16S and 23S rRNA genes of thermus thermophilus, J. Bacteriol. 187 (2005) 4804e4812. [29] I.B. Weinstein, M. Ochoa, S.M. Friedman, Fidelity in the translation of messenger ribonucleic acids in mammalian subcellular systems, Biochemistry (Mosc.) 5 (1966) 3332e3339. [30] S.M. Friedman, R. Berezney, I.B. Weinstein, Fidelity in protein synthesis. The role of the ribosome, J. Biol. Chem. 243 (1968) 5044e5048. [31] L. Stavy, Miscoding in a cell-free system from spleen, Proc. Natl. Acad. Sci. U. S. A. 61 (1968) 347e353. [32] E.B. Kramer, H. Vallabhaneni, L.M. Mayer, P.J. Farabaugh, A comprehensive analysis of translational missense errors in the yeast Saccharomyces cerevisiae, RNA 16 (2010) 1797e1808. [33] E.B. Kramer, P.J. Farabaugh, The frequency of translational misreading errors in E. coli is largely determined by tRNA competition, RNA 13 (2007) 87e96. [34] E.P. Plant, P. Nguyen, J.R. Russ, Y.R. Pittman, T. Nguyen, J.T. Quesinberry, T.G. Kinzy, J.D. Dinman, Differentiating between near- and non-cognate codons in Saccharomyces cerevisiae, PLoS One 2 (2007) e517. [35] A. Ben-Shem, L. Jenner, G. Yusupova, M. Yusupov, Crystal structure of the eukaryotic ribosome, Science 330 (2010) 1203e1209. [36] A. Ben-Shem, N. Garreau de Loubresse, S. Melnikov, L. Jenner, G. Yusupova, M. Yusupov, The structure of the eukaryotic ribosome at 3.0 Å resolution, Science 334 (2011) 1524e1529. [37] A.M. Anger, J.-P. Armache, O. Berninghausen, M. Habeck, M. Subklewe, D.N. Wilson, R. Beckmann, Structures of the human and Drosophila 80S ribosome, Nature 497 (2013) 80e85. [38] R.M. Voorhees, I.S. Fernandez, S.H.W. Scheres, R.S. Hegde, Structure of the mammalian ribosome-Sec61 complex to 3.4 Å resolution, Cell 157 (2014) 1632e1643. [39] E. Svidritskiy, A.F. Brilot, C.S. Koh, N. Grigorieff, A.A. Korostelev, Structures of yeast 80S ribosome-tRNA complexes in the rotated and nonrotated conformations, Structure 22 (2014) 1210e1218. [40] J. Romanowska, P. Setny, J. Trylska, Molecular dynamics study of the ribosomal A-site, J. Phys. Chem. B 112 (2008) 15227e15243. [41] J. Romanowska, J.A. McCammon, J. Trylska, Understanding the origins of bacterial resistance to aminoglycosides through molecular dynamics mutational study of the ribosomal A-site, PLoS Comput. Biol. 7 (2011) e1002099. [42] M. Dudek, J. Romanowska, T. Wituła, J. Trylska, Interactions of amikacin with the RNA model of the ribosomal A-site: computational, spectroscopic and calorimetric studies, Biochimie 102 (2014) 188e202. [43] F. Barbault, B. Ren, J. Rebehmed, C. Teixeira, Y. Luo, O. Smila-Castro, F. Maurel, B. Fan, L. Zhang, L. Zhang, Flexible computational docking studies of new aminoglycosides targeting RNA 16S bacterial ribosome site, Eur. J. Med. Chem. 43 (2008) 1648e1656. [44] P. Setny, J. Trylska, Search for novel aminoglycosides by combining fragmentbased virtual screening and 3D-QSAR scoring. J. Chem. Inf. Model 49 (2009) 390e400. [45] J. Panecka, M. Havrila, K. Reblova, J. Sponer, J. Trylska, Role of S-turn2 in the structure, dynamics, and function of mitochondrial ribosomal A-site. A bioinformatics and molecular dynamics simulation study, J. Phys. Chem. B 118 (2014) 6687e6701. [46] A.P. Carter, W.M. Clemons, D.E. Brodersen, R.J. Morgan-Warren, B.T. Wimberly, V. Ramakrishnan, Functional insights from the structure of the 30S ribosomal subunit and its interactions with antibiotics, Nature 407 (2000) 340e348. [47] T. Hermann, V. Tereshko, E. Skripkin, D.J. Patel, Apramycin recognition by the human ribosomal decoding site. Blood Cells. Mol. Dis. 38 (2007) 193e198. [48] J. Kondo, B. Francois, A. Urzhumtsev, E. Westhof, Crystal structure of the Homo sapiens cytoplasmic ribosomal decoding site complexed with apramycin. Angew. Chem. Int. Ed. Engl. 45 (2006) 3310e3314. [49] D. Case, T. Cheatham, T. Darden, H. Gohlke, R. Luo, K.M. Merz Jr., A. Onufriev, C. Simmerling, B. Wang, W. R, The Amber biomolecular simulation programs.,, J. Comput. Chem. 26 (2005) 1668e1688. [50] D.A. Case, T.A. Darden, T.E. Cheatham III, C.L. Simmerling, J. Wang, R.E. Duke, R. Luo, M. Crowley, R.C. Walker, W. Zhang, K. Merz, B. Wang, S. Hayik, A. Roitberg, G. Seabra, I. Kolossv ary, K.F. Wong, F. Paesani, J. Vanicek, X. Wu, S.R. Brozell, T. Steinbrecher, H. Gohlke, L. Yang, C. Tan, J. Mongan, V. Hornak, G. Cui, D.H. Mathews, M.G. Seetin, C. Sagui, V. Babin, K.P. A, AMBER 10, University of California, San Francisco, 2008. URL, http://ambermd.org/.

109

[51] T. Darden, L. Perera, L. Li, L. Pedersen, New tricks for modelers from the crystallography toolkit: the particle mesh Ewald algorithm and its use in nucleic acid simulations. Structure 7 (1999) R55eR60. [52] J. Ryckaert, G. Ciccotti, H. Berendsen, Numerical integration of the cartesian equations of motion of a system with constraints: molecular dynamics of nalkanes, J. Comput. Phys. 23 (1977) 327e341. [53] K. Reblova, F. Lankas, F. Razga, M.V. Krasovska, J. Koca, J. Sponer, Structure, dynamics, and elasticity of free 16s rRNA helix 44 studied by molecular dynamics simulations. Biopolymers 82 (2006) 504e520. [54] K. Reblova, E. Fadrna, J. Sarzynska, T. Kulinski, P. Kulhanek, E. Ennifar, J. Koca, J. Sponer, Conformations of flanking bases in HIV-1 RNA DIS kissing complexes studied by molecular dynamics. Biophys. J. 93 (2007) 3932e3949. [55] K. Reblova, Z. Strelcova, P. Kulhanek, I. Besseova, D.H. Mathews, K.V. Nostrand, I. Yildirim, D.H. Turner, J. Sponer, An RNA molecular switch: Intrinsic flexibility of 23S rRNA Helices 40 and 68 5’-UAA/5’-GAN internal loops studied by molecular dynamics methods. J. Chem. Theory Comput. 2010 (2010) 910e929. [56] W.D. Cornell, P. Cieplak, C.I. Bayly, I.R. Gould, J. Merz, K.M.,D.M. Ferguson, D. Spellmeyer, T. Fox, J. Caldwell, P.A. Kollman, A second generation force field for the simulation of proteins, nucleic acids, and organic molecules, J. Am. Chem. Soc. 117 (1995) 5179e5197. [57] A. Perez, I. Marchan, D. Svozil, J. Sponer, T.E. Cheatham, C.A. Laughton, M. Orozco, Refinement of the AMBER force field for nucleic acids: improving the description of alpha/gamma conformers. Biophys. J. 92 (2007) 3817e3829. [58] P. Banas, D. Hollas, M. Zgarbova, P. Jurecka, M. Orozco, T.E. Cheatham, J.S. Sponer, M. Otyepka, Performance of molecular mechanics force fields for RNA simulations: stability of UUCG and GNRA hairpins, J. Chem. Theory Comput. 6 (2010) 3836e3849. [59] I.S. Joung, T.E. Cheatham, Determination of alkali and halide monovalent ion parameters for use in explicitly solvated biomolecular simulations, J. Phys. Chem. B 112 (2008) 9020e9041. [60] W.L. Jorgensen, J. Chandrasekhar, J.D. Madura, R.W. Impey, M.L. Klein, Comparison of simple potential functions for simulating liquid water, J. Chem. Phys. 79 (1983) 926e935. [61] J. Åqvist, Ion-water interaction potentials derived from free energy perturbation simulations, J. Phys. Chem. 94 (1990) 8021e8024. [62] D.E. Smith, L.X. Dang, Computer simulations of NaCl association in polarizable water, J. Chem. Phys. 100 (1994) 3757e3766. [63] L.X. Dang, P.A. Kollman, Free energy of association of the Kþ:18-crown-6 complex in water: a new molecular dynamics study, J. Phys. Chem. 99 (1995) 55e58. [64] H.J.C. Berendsen, J.R. Grigera, T.P. Straatsma, The missing term in effective pair potentials,, J. Phys. Chem. 91 (1987) 6269e6271. [65] K. Song, A.J. Campbell, C. Bergonzo, C. de los Santos, A.P. Grollman, C. Simmerling, An improved reaction coordinate for nucleic acid base flipping studies, J. Chem. Theory Comput. 5 (2009) 3105e3113. [66] N.B. Leontis, J. Stombaugh, E. Westhof, The non-Watson-Crick base pairs and their associated isostericity matrices, Nucleic Acids Res. 30 (2002) 3497e3531. [67] C.L. Zirbel, J.E. Sponer, J. Sponer, J. Stombaugh, N.B. Leontis, Classification and energetics of the base-phosphate interactions in RNA. Nucleic Acids Res. 37 (2009) 4898e4918. [68] P. Banas, A. Mladek, M. Otyepka, M. Zgarbova, P. Jurecka, D. Svozil, F. Lankas, J. Sponer, Can we accurately describe the structure of adenine tracts in BDNA? reference quantum-chemical computations reveal overstabilization of stacking by molecular mechanics, J. Chem. Theory Comput. 8 (2012) 2448e2460. [69] I. Besseova, M. Otyepka, K. Reblova, J. Sponer, Dependence of A-RNA simulations on the choice of the force field and salt strength. Phys. Chem. Chem. Phys. 11 (2009) 10701e10711. [70] M. Zgarbova, M. Otyepka, J. Sponer, A. Mladek, P. Banas, T.E. Cheatham 3rd, P. Jurecka, Refinement of the Cornell et al, nucleic acids force field based on reference quantum chemical calculations of glycosidic torsion profiles. J. Chem. Theory Comput. 7 (2011) 2886e2902. [71] P. Banas, P. Sklenovsky, J.E. Wedekind, J. Sponer, M. Otyepka, Molecular mechanism of preQ1 riboswitch action: a molecular dynamics study, J. Phys. Chem. B 116 (2012) 12721e12734. [72] J. Sponer, P. Banas, P. Jurecka, M. Zgarbova, P. Kuhrova, M. Havrila, M. Krepl, P. Stadlbauer, M. Otyepka, Molecular dynamics simulations of nucleic acids. From tetranucleotides to the ribosome, J. Phys. Chem. Lett. 5 (2014) 1771e1782. [73] A. Noy, I. Soteras, F.J. Luque, M. Orozco, The impact of monovalent ion force field model in nucleic acids simulations.,, Phys. Chem. Chem. Phys. 11 (2009) 10596e10607. [74] J. Panecka, C. Mura, J. Trylska, Interplay of the bacterial ribosomal A-site, S12 protein mutations and paromomycin binding: a molecular dynamics study, PLoS One 9 (2014) e111811. [75] D. Fourmy, S. Yoshizawa, J.D. Puglisi, Paromomycin binding induces a local conformational change in the A-site of 16 S rRNA. J. Mol. Biol. 277 (1998) 333e345. [76] B.T. Wimberly, D.E. Brodersen, W. Clemons Jr., R.J. Morgan-Warren, A.P. Carter, C. Vonrhein, T. Hartsch, V. Ramakrishnan, Structure of the 30S ribosomal subunit. Nature 407 (2000) 327e339. [77] R.H. Kretsinger, V.N. Uversky, E.A. Permyakov (Eds.), ‘Cellular Electrolyte Metabolism’ in Encyclopedia of Metalloproteins, Springer, 2013.

110

J. Panecka et al. / Biochimie 112 (2015) 96e110

[78] C.M. Barbieri, M. Kaul, D.S. Pilch, Use of 2-aminopurine as a fluorescent tool for characterizing antibiotic recognition of the bacterial rRNA A-site. Tetrahedron 63 (2007) 3567e6574. [79] A. Casiano-Negroni, NMR Studies of RNA Conformational Dynamics Induced by Metal Cations and Paromomycin (Ph.D. thesis), The University of Michigan, 2010. [80] S. Shandrick, Q. Zhao, Q. Han, B.K. Ayida, M. Takahashi, G.C. Winters, K.B. Simonsen, D. Vourloumis, T. Hermann, Monitoring molecular recognition of the ribosomal decoding site. Angew. Chem. Int. Ed. Engl. 43 (2004) 3177e3182. [81] D. Fourmy, M.I. Recht, S.C. Blanchard, J.D. Puglisi, Structure of the A site of Escherichia coli 16S ribosomal RNA complexed with an aminoglycoside antibiotic. Science 274 (1996) 1367e1371. [82] J.M. Ogle, D.E. Brodersen, W.M. Clemons, M.J. Tarry, A.P. Carter, V. Ramakrishnan, Recognition of cognate transfer RNA by the 30S ribosomal subunit, Science 292 (2001) 897e902. [83] A. Mladek, P. Sharma, A. Mitra, D. Bhattacharyya, J. Sponer, J.E. Sponer, Trans Hoogsteen/sugar edge base pairing in RNA. Structures, energies, and stabilities from quantum chemical calculations, J. Phys. Chem. B 113 (2009) 1743e1755.

[84] J. Sponer, P. Jurecka, P. Hobza, Accurate interaction energies of hydrogenbonded nucleic acid base pairs. J. Am. Chem. Soc. 126 (2004) 10142e10151. [85] J. Sponer, A. Mokdad, J.E. Sponer, N. Spackova, J. Leszczynski, N.B. Leontis, Unique tertiary and neighbor interactions determine conservation patterns of Cis Watson-Crick A/G base-pairs. J. Mol. Biol. 330 (2003) 967e978. [86] J. Kondo, M. Hainrichson, I. Nudelman, D. Shallom-Shezifi, C.M. Barbieri, D.S. Pilch, E. Westhof, T. Baasov, Differential selectivity of natural and synthetic aminoglycosides towards the eukaryotic and prokaryotic decoding A sites, ChemBioChem 8 (2007) 1700e1709. [87] P. Auffinger, N. Grover, E. Westhof, Metal ion binding to RNA. Met. Ions Life Sci. 9 (2011) 1e35. [88] G.M. Daubner, A. Clery, F.H.-T. Allain, RRM-RNA recognition: NMR or crystallography… and new findings. Curr. Opin. Struct. Biol. 23 (2013) 100e108. [89] S.E. McDowell, N. Spackova, J. Sponer, N.G. Walter, Molecular dynamics simulations of RNA: an in silico single molecule approach. Biopolymers 85 (2007) 169e184. [90] M. Długosz, J.M. Antosiewicz, J. Trylska, Association of aminoglycosidic antibiotics with the ribosomal A-site studied with Brownian dynamics. J. Chem. Theory Comput. 4 (2008) 549e559.

Conformational dynamics of bacterial and human cytoplasmic models of the ribosomal A-site.

The aminoacyl-tRNA binding site (A-site) is located in helix 44 of small ribosomal subunit. The mobile adenines 1492 and 1493 (Escherichia coli number...
3MB Sizes 4 Downloads 7 Views