Exascale computing provides access to unprecedented computational resources and it is worth considering the beneficial contributions of this computing to structural biology. One area to benefit from these advancements is molecular dynamics (MD) simulations of biomolecular complexes as they have been known to push computational resources, scaling beyond 100,000 processing cores [1,2]. MD simulations are a powerful set of tools that provide access to biomolecular structural dynamics and conformational rearrangements, given enough sampling. Expanded computational resources provide two fundamental benefits to MD simulations: 1. Increased size of systems to be simulated and 2. Increased length of the simulations achieved. The latter provides access to simulations of slower conformational dynamics while the former grants access to simulations of larger systems that were previously unattainable. An area to benefit tremendously from increased MD simulation capabilities is the studies of ribonucleoprotein complexes (RNPs) and their necessary conformational rearrangements.

Advancements to Exascale computing coincide with the advancements in cryo-EM, resulting in a rapid increase in the structural determination of RNP structures (Figure 1a). New cryo-EM and X-ray crystallography structural determinations are providing a wealth of stable or semi-stable models of RNP conformations. This offers a vast opportunity for MD simulations to be used to study the local structural dynamics and conformational rearrangements required by RNPs. For medicine, performing these studies is imperative, as inhibiting conformational rearrangements, particularly for rate-limiting steps, represent attractive drug-targetable states. However, due to their relatively large size, MD simulations of RNPs are computationally expensive [3], yet, with advancements in computational resources, the number of RNP simulations and their timescales are steadily increasing (Figure 1a, b).

Studies of RNP structural dynamics have included the use of molecular mechanics (MM), coarse-grained (CG), or quantum mechanics (QM) simulation methods. Explicit solvent MM simulations encompass all-atom RNP complexes solvated in a buffer-like system, often mimicking in vitro conditions. This approach employs Newtonian mechanics and the use of empirical forcefields, allowing the protein to sample configurations feasible based on features such as van der Waals forces, electrostatic interactions, or steric hindrance. A drawback to this approach is that conformational rearrangements that occur on timescales in the millisecond to second timescale are not achievable, currently. To investigate longer timescale sampling of conformational landscapes, CG simulations have been employed, where the number of atoms in the simulation or the forcefield complexity is reduced. However, CG simulations lack finer details, making specific interactions often unquantifiable. In understanding the behavior of electrons during catalytic events in biomolecular complexes, QM simulations are preferred. Often, QM simulations are coupled with MM simulation (QM/MM) providing access to conformational sampling of RNP complexes and characterization of the catalytic event performed by the RNP. Nevertheless, QM simulations are computationally expensive and do not achieve long timescales. To address specific hypotheses, one must choose the correct type of simulation to employ.

The RNP system that has become a hallmark of studying RNP structural dynamics is the ribosome. In this mini-review we will discuss recent advancements in how MD simulations have studied conformational rearrangements of the ribosome, focusing on recent insights into the elongation phase of protein synthesis.

Ribosomes simulations marked a significant milestone as they were the first all-atom simulation to exceed 1 million atoms [5]. Since this achievement, ribosome simulations have increased in number, describing the structural dynamics of this molecular machine [3,6, 7, 8, 9]. Ribosomes are the cellular RNP responsible for protein synthesis and are composed of both large and small subunits (SSU and LSU), the latter is separated into a head and body domain. During active protein synthesis the SSU undergoes an intersubunit rotation relative to the LSU and the SSU undergoes intrasubunit rotation of the head relative to the body.

Protein synthesis can be decomposed into 4 essential phases in prokaryotes including initiation, elongation, termination, and recycling (Figure 2). Translation initiation involves the formation of an initiation complex when the SSU binds to messenger RNA (mRNA), fMet-tRNAfMet, and initiation factors (IF) 1, 2, and 3. Release of IF3 allows for binding of the LSU through the process of subunit joining (Figure 2). Guanosine triphosphate (GTP) hydrolysis by IF2 and disengagement of both IF1 and IF2 triggers formation of a ribosome complex ready to enter the elongation phase of translation [10]. Elongation follows two GTP hydrolysis-dependent processes in a cyclical manner: Delivery of aminoacyl-(aa)tRNA to the ribosomal A site by Elongation Factor Thermo unstable (EF-Tu) and peptide bond formation, followed by translocation of the mRNA-tRNA2 substrate from the A and P sites to the respective P and E sites by Elongation Factor G (EF-G) (Figure 2). After the polypeptide has achieved its final length and the ribosome encounters a stop codon on the bound mRNA, Release Factor 1 or 2 (RF1 or RF2) binds to the A site and catalyzes cleavage of the nascent polypeptide chain from the tRNA. RF3 disengages RF1 or RF2 from the A site in a GTP hydrolysis-dependent mechanism, allowing for EF-G and Ribosome Release Factor (RRF) to stimulate subunit separation and recycling (Figure 2). Although other phases of translation have been studied by MD simulations [7], here we will focus on recent studies of elongation.

tRNA selection commences by an initial selection of aa-tRNA involving formation of Watson-Crick codon-anticodon interactions between the codon of the mRNA and anticodon of the tRNA. To achieve high fidelity in translation, tRNA selection proceeds via a kinetic proofreading mechanism, providing a second opportunity to discriminate for the correct tRNA [11, 12, 13, 14]. Both initial selection and proofreading involve several distinct conformations of the ribosome, tRNA, and EF-Tu and are separated by EF-Tu-catalyzed GTP hydrolysis [15,16]. The first of a series of conformational rearrangements involves initial binding of the ternary complex composed of EF-Tu (GTP) and aa-tRNA to the SSU, in the absence of codon-anticodon interactions. The aa-tRNA samples configurations until codon-anticodon interactions are formed adopting the Codon Recognition (CR) state. Domain closure of the SSU at the ribosomal A site docks EF-Tu on the Sarcin-Ricin Loop of the LSU, stimulating GTP hydrolysis. Upon Pi release EF-Tu undergoes a conformational change, releasing and allowing aa-tRNA to approach the fully accommodated A/A position (where the letter denotes the position of the tRNA within the SSU and LSU, respectively) (Figure 3) [16,17].

Initial MD simulations were implemented to study the interactions between the ribosome and the aa-tRNA as it traverses the accommodation corridor into the A site [18]. Thereafter, studies focused on the energetics of the interactions between the mRNA codon and aa-tRNA anticodon, including the contributions of tRNA modifications and the decoding 16S rRNA nucleotides of G530, A1492, and A1493 [19, 20, 21, 22, 23]. These investigations have continued towards the interaction of antibiotics with ribosomes and how they impact the structural dynamics of regions such as the A-site or the nascent peptide exit tunnel, contributing to ribosome stalling and inhibition of protein synthesis [24, 25, 26, 27∗∗, 28, 29, 30, 31, 32, 33, 34]. Simulations have been used to characterize how aminoglycosides alter the conformational landscape of the decoding center to impact the fidelity of translation [24, 25, 26, 27∗∗]. Extending beyond initial selection, simulations have been used to characterize how evernimicin or hygromycin A impacts the movement of the aa-tRNA into the ribosome during the proofreading stage of tRNA selection [27,28]. Beyond tRNA selection, simulations have also been critical for our understanding of macrolide interactions with the nascent peptide exit tunnel [29,30] and for describing how these antibiotics remodel the peptidyl transferase center [33], as well as how they interact with stalling peptide sequences [31,32]. More recently, high-resolution cryo-EM coupled with MM simulations characterized how rapid freezing impacts direct and water-mediated ribosome interactions with lincomycin [34]. Collectively, these simulations were critical to our understanding as to how ribosome–antibiotic interactions are mediated and how they inhibit protein synthesis.

To study the conformational rearrangements of the ribosome and aa-tRNA during tRNA selection structure-based simulations were utilized. These types of simulations allow for sampling between two ribosomal conformations that are set as energetic minima on a minimally frustrated landscape. Thermal fluctuations facilitate sampling between two biomolecular conformations and native contacts, derived from a pre-defined model, energetically stabilize the system. The shape of the overall free energy landscape is governed by the pre-defined models, the configurational entropy, and the steric contributions of the system [35]. These simulations have explored the reversible fluctuations of aa-tRNA accommodation [36], as well as the steric effects contributed by uL11 [37] and the A-site finger [38] to this process. Furthermore, they have been implemented to quantify the diffusion of tRNA within the ribosome, facilitating more accurate estimations of the timescales of these types of simulations [39]. Recent advancements to the structure-based models have included electrostatic considerations allowing for characterization of the contribution of ions, particularly Mg2+, to tRNA selection. These results indicated that fluctuations of tRNA during accommodation are highly dependent on ion concentration, where diffuse Mg2+ concentrations can increase or decrease barriers for tRNA accommodation [40].

More recently structure-based simulations described structural distinctions between selection of cognate and near-cognate tRNA during initial selection and proofreading [27]. These simulations commenced from the conformation of initial binding of ternary complex to the SSU and proceeded to the fully accommodated state (Figure 3). It was observed that near-cognate aa-tRNA follows divergent pathways into the ribosomal A site in comparison to cognate aa-tRNA, resulting in misaligned tRNA positioning. This misalignment can be observed prior to release from EF-Tu, indicating that the divergent behavior originates from codon-anticodon interactions. The alternate trajectory of near-cognate aa-tRNA promotes interactions between the acceptor stem of the tRNA and helix 71 of the 23S rRNA, resulting in increased flexibility of this region of the tRNA. Misaligned tRNA positioning restricts progression of the tRNA into the ribosomal A site, resulting in the incoming amino acid being distal from the nascent peptide chain, limiting access to peptide bond formation. These findings provide a structural framework for proofreading and support earlier hypothesis that tRNA selection involves alignment of cognate tRNA and misalignment of near-cognate tRNA [53].

Late-stage events of aa-tRNA accommodation into the A site, involving movement of the 3′-CCA end of the tRNA were also studied with structure-based simulations, revealing two distinct pathways [54]. Both pathways include movement of the 3′-CCA end of the tRNA to a pocket formed by helix 89/90/92 (H89/90/92) of the 23S rRNA. The first and more sampled pathway involves the 3′-CCA proceeding along H89 to the A site. The second pathway includes the 3′-CCA end moving away from H89 after entering the H89/90/92 pocket by rotating ∼180° before moving along H93 and the P-site tRNA into the A site. By modulating steric contributions of ribosomal elements, it was identified that C2573 of the 23S rRNA influences the kinetics of 3′-CCA accommodation. These in silico studies are comparable to those identified in vivo where C2573A mutation reduced the error frequency of translation, reported by β-galactosidase assays [55]. The findings that near-cognate aa-tRNA is misaligned relative to cognate aa-tRNA and the contributions of C2573 as a steric barrier for 3′-CCA end accommodation indicates that the precise geometry of the accommodating tRNA relative to the accommodation corridor is an important contributor to tRNA selection.

Early simulations of aa-tRNA selection were often performed in the absence of EF-Tu as the focus was placed on the interactions between tRNA and mRNA or tRNA and ribosome. Recently, investigations have been utilized to study how EF-Tu contributes to the overall aa-tRNA selection process [56] and how EF-Tu undergoes conformational changes during release of aa-tRNA [57, 58, 59, 60∗]. Structure-based simulations revealed that EF-Tu, when bound to ribosomes after aa-tRNA release, reduces the barrier for aa-tRNA to enter an elbow-accommodated conformation [56]. This is achieved by the steric contribution of EF-Tu and the kinetics of EF-Tu release from the ribosome. Steric contributions and EF-Tu dissociation from the ribosome have evolved to be well balanced so that EF-Tu can contribute to reducing the barrier of aa-tRNA accommodation while still allowing for dissociation of incorrect aa-tRNA during proofreading. Similar studies simulated EF-Tu release from the ribosome, indicating that EF-Tu undergoes domain separation and enters an extended intermediate conformation where domain I likely disengages the ribosome first [57]. This separation of domain I was found to be energetically more favorable compared to domain I rotation followed by domain separation [58]. Switch I (Gly 40 – Ile 62) of EF-Tu increases mobility prior to domain separation whereas switch II (Asp 80 – Gly 100) conformational rearrangements occur late in EF-Tu's conformational change and lock it in the post-hydrolysis conformation [58]. Correspondingly explicit solvent MM simulations reveal that EF-Tu post-GTP hydrolysis undergoes rotation of domain I about the switch II helix and a separation of domain I relative to domain II [59]. These simulations reveal that increased switch I flexibility after GTP hydrolysis is due to a loss in contact between the γ-phosphate of GTP and switch I [59].

To characterize the conformational dynamics of switch I during tRNA selection, structure-based simulations were performed of aa-tRNA release during EF-Tu conformational change on the ribosome. These structure-based simulations revealed that switch I movements are sterically restricted by aa-tRNA progression towards the A site. For tRNA selection to proceed, switch I was observed to traverse the minor and major grooves of the aa-tRNA before compacting against domain III of EF-Tu. The restriction of aa-tRNA progression into the A site indicates a regulatory role of switch I in aa-tRNA selection and correlates to smFRET data indicating that EF-Tu dissociates during aa-tRNA accommodation [28,60]. The dynamics of switch I behavior were supported later by time-resolved cryo-EM ensembles showing similar configurational rearrangements of EF-Tu on the ribosome during aa-tRNA selection [61]. These recent insights into tRNA selection indicate that not only does EF-Tu deliver aa-tRNA to the A site but also that it actively contributes to the overall tRNA selection process.

After peptide bond formation tRNAs are repositioned from the classical A/A and P/P pre-translocation (PRE-C) positions to the hybrid A/P or A/P∗ (PRE-H1 or PRE-H2) (tRNA elbow G19-C56 pair remains fixed with A-site finger) and P/E positioning through spontaneous counterclockwise intersubunit rotation (Figure 3) [42]. This process requires several intermediates that have been resolved by a variety of biochemical and structural techniques [41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52,62, 63, 64, 65, 66, 67, 68]. In brief, EF-G binding to PRE-H1 coincides with transition to the first intermediate conformational basin (INT1), where the SSU has rotated by ∼11° and the head has swiveled ∼6° [42,45,63]. Head swivel continues until ∼16° and there is partial reverse intersubunit rotation achieving the second intermediate basin (INT2) [42,45]. This conformation leads to a 2/3 translocation of the codon-anticodon with respect to the SSU and the tRNA achieves its fully translocated position with respect to the LSU. Upon return of the SSU to its unrotated position a third intermediate conformation (INT3) is observed by smFRET, corresponding to hyper-swivel positioning or tilting motions of the SSU head domain [45,69,70]. Reverse head swivel to 0° transitions the ribosome to the POST conformation whereby the A site is available for another round of elongation or termination (Figure 3).

The energetics and structural fluctuations involved in the movement of the tRNA during intersubunit rotation have been investigated through use of both explicit solvent and structure-based simulations [71, 72, 73, 74, 75, 76, 77∗, 78]. Given that tRNA hybrid state formation is dependent on counterclockwise intersubunit rotation of the SSU and mRNA-tRNA2 translocation is dependent on head-swivel motions of the SSU, simulations have been implemented to decrypt these motions. Explicit solvent simulations were implemented to describe body rotation, head swivel, and tRNA displacement diffusion and kinetics [71]. Further simulations of ribosomal conformations in the PRE and POST basins described the energetic contributions of intersubunit contacts between the LSU and SSU stabilizing different intersubunit rotation conformations [72]. More recent investigations have utilized structure-based simulations to investigate the energetic barriers of intersubunit rotation [73] and structural dynamics of eukaryotic intersubunit rotation [74]. These latter simulations represent a new frontier for structure-based simulations which include solving the conformational changes of eukaryotic ribosomes.

To describe tRNA fluctuations during intersubunit rotation explicit solvent simulations were used to quantify the transition rates of tRNA within the PRE and POST basins. In so doing, these simulations described the fluctuations of tRNA during intersubunit rotation and the contributions of the L1 stalk in assisting the formation of the P/E tRNA positioning during SSU rotation [75]. For a more global view of translation targeted MD was used to simulate the hierarchical order of subunit rotation, tRNA elbow movements, and 3′-CCA end movement during transitions of the deacyl-tRNA from the P/P to the P/E position [76]. These simulations described an intermediate ensemble which forms due to steric interactions between helix 74 of the 23S rRNA and the 3′-CCA end of the tRNA. The intermediate conformation of hybrid P/E tRNA formation was investigated further with structure-based simulations identifying two additional intermediates and an additional free energy minimum [77]. In the first intermediate the tRNA elbow and 3′-CCA end are dissociated from but remain near the P site, in the second intermediate the elbow domain of the tRNA is in the E site but not the 3′-CCA end. These intermediates represent similar conformations to those solved by cryo-EM [50]. The energetic minimum conformation positions the tRNA elbow in E site but the 3′-CCA end extends beyond the P or E site. Separation of the two intermediate conformations was identified to involve steric contributions from residues U2067, A2077, C2078, U2423, A2435 of the 23S rRNA and Asn29 of bL33. Similar steric contributions of the A-site finger were observed to contribute to the barrier of A/P tRNA positioning [42,78]. These findings indicate that the fluctuations of tRNA to hybrid positionings are governed by the geometric shape of the ribosome itself.

Late-stage events of translocation involving translocation of the mRNA-tRNA2 substrate with respect to the SSU involving head-domain swivel motions have more recently been investigated by MD simulations [69,70,79]. These simulations suggest transitions between the hybrid configuration to the POST configuration can occur by a combined head swivel and head tilt mechanism [70,79] or through hyper swivel motions of the SSU head [69]. In the head tilt model the head swivel achieves ∼16°, positioning the tRNA in a chimeric position (ap/P and pe/E position, where the tRNA is in the A and P site or P and E site with respect to the SSU head and body), consistent with translocation intermediate conformations [41]. For the head to return to the ∼0° swivel position in the POST state it requires a head-tilt motion of ∼10° orthogonal to the head-swivel axis. Steric barriers provided by interactions between the deacylated-tRNA and the PE loop (G1338-U1341 of the 16S rRNA) and uS13 necessitate the head tilting motions in this model. The hyper-swivel model requires the head-swivel motions to adopt an angle of ∼24°, extending beyond the 16° translocation intermediate position [69]. This model is consistent with smFRET intermediate (INT3) [45] and represents the maximal swivel angle before steric interactions occur between the SSU head and body. The benefit to this model is that it describes how the head swivel can position the tRNA to their final translocated position (P/P and E/E) through a continuous forward swivel trajectory. Reverse swivel to the POST configuration is then achieved by dissociation of the tRNA in the E-site at the hyper-swivel position, whereby, the steric barrier between the PE loop and deacyl-tRNA would be alleviated not requiring head tilt. Both these models describe how ribosomes can return the 0° head-swivel position, with hyper-swivel describing positioning of the mRNA-tRNA2 substrate to its final translocated position.