ACTAUNIVERSITATIS

UPSALIENSISUPPSALA

2017

Digital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Science and Technology 1549

Computational Studies of ProteinSynthesis on the Ribosome andLigand Binding to Riboswitches

CHRISTOFFER LIND

ISSN 1651-6214ISBN 978-91-513-0051-1urn:nbn:se:uu:diva-328583

Dissertation presented at Uppsala University to be publicly examined in B41 BMC,Husargatan 3, Uppsala, Friday, 13 October 2017 at 13:15 for the degree of Doctor ofPhilosophy. The examination will be conducted in English. Faculty examiner: ProfessorAlexander MacKerell Jr (University of Maryland School of Pharmacy).

AbstractLind, C. 2017. Computational Studies of Protein Synthesis on the Ribosome and LigandBinding to Riboswitches. Digital Comprehensive Summaries of Uppsala Dissertations fromthe Faculty of Science and Technology 1549. 64 pp. Uppsala: Acta Universitatis Upsaliensis.ISBN 978-91-513-0051-1.

The ribosome is a macromolecular machine that produces proteins in all kingdoms of life.The proteins, in turn, control the biochemical processes within the cell. It is thus of extremeimportance that the machine that makes the proteins works with high precision. By using threedimensional structures of the ribosome and homology modelling, we have applied moleculardynamics simulations and free-energy calculations to study the codon specificity of proteinsynthesis in initiation and termination on an atomistic level. In addition, we have examined thebinding of small molecules to riboswitches, which can change the expression of an mRNA.

The relative affinities on the ribosome between the eukaryotic initiator tRNA to the AUGstart codon and six near-cognate codons were determined. The free-energy calculations showthat the initiator tRNA has a strong preference for the start codon, but requires assistance frominitiation factors 1 and 1A to uphold discrimination against near-cognate codons.

When instead a stop codon (UAA, UGA or UAG) is positioned in the ribosomal A-site, arelease factor binds and terminates protein synthesis by hydrolyzing the nascent peptide chain.However, vertebrate mitochondria have been thought to have four stop codons, namely AGAand AGG in addition to the standard UAA and UAG codons. Furthermore, two release factorshave been identified, mtRF1 and mtRF1a. Free-energy calculations were used to determine ifany of these two factors could bind to the two non-standard stop codons, and thereby terminateprotein synthesis. Our calculations showed that the mtRF’s have similar stop codon specificityas bacterial RF1 and that it is highly unlikely that the mtRF’s are responsible for terminatingat the AGA and AGG stop codons.

The eukaryotic release factor 1, eRF1, on the other hand, can read all three stop codonssinglehandedly. We show that eRF1 exerts a high discrimination against near-cognate codons,while having little preference for the different cognate stop codons. We also found an energeticmechanism for avoiding misreading of the UGG codon and could identify a conserved clusterof hydrophobic amino acids which prevents excessive solvent molecules to enter the codonbinding site.

The linear interaction energy method was used to examine binding of small molecules tothe purine riboswitch and the FEP method was employed to explicitly calculate the LIE b-parameters. We show that the purine riboswitches have a remarkably high degree of electrostaticpreorganization for their cognate ligands which is fundamental for discriminating againstdifferent purine analogs.

Keywords: Binding free energy, Ribosome, Codon reading, Translation initiation, Translationtermination, Mitochondrial translation, Release factor, Purine riboswitch, MolecularDynamics, Free Energy Perturbation

Christoffer Lind, Department of Cell and Molecular Biology, Computational Biology andBioinformatics, Box 596, Uppsala University, SE-751 24 Uppsala, Sweden.

© Christoffer Lind 2017

ISSN 1651-6214ISBN 978-91-513-0051-1urn:nbn:se:uu:diva-328583 (http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-328583)

“You miss 100% of the shots you don’t take”

- Wayne Gretzky

List of Papers

This thesis is based on the following papers, which are referred to in the text by their Roman numerals.

I Lind, C., Sund, J., Åqvist, J. (2013) Codon-reading specificities

of mitochondrial release factors and translation termination at non-standard stop codons. Nature Communications, 4:2940

II Sund, J., Lind, C., Åqvist, J. (2014) Binding site preorganization and ligand discrimination in the purine riboswitch. Journal of Physical Chemistry B, 119:773—782.

III Lind, C., Åqvist, J. (2016) Principles of start codon recognition in eukaryotic translation initiation. Nucleic acids research, 44(17):8425—8432

IV Lind, C., Oliveira, A., Åqvist, J. (2017) Origin of the omnipo-tence of eukaryotic release factor 1 (submitted)

Reprints were made with permission from the respective publishers. Additional publications

i. Åqvist, J., Lind, C., Sund, J., Wallin, G. (2012) Bridging the gap between ribosome structure and biochemistry by mechanistic computations. Current Opinion in Structural Biology. 22:815—823.

ii. Lind, C., Esguerra, M., Åqvist, J. (2017) A close-up view of co-don selection in eukaryotic initiation. RNA Biology, 14(7):815—819

Contents

Introduction ................................................................................................. 9Outline ............................................................................................. 10

The genetic code and RNA ....................................................................... 11The structure of DNA ...................................................................... 11What about RNA? ........................................................................... 14Cracking the code ............................................................................ 15

The ribosome ............................................................................................ 18Initiation of protein synthesis .......................................................... 20Elongation of the peptide chain ....................................................... 23Termination of protein translation ................................................... 27

Riboswitches ............................................................................................. 30

The present work ....................................................................................... 32Start codon selection in eukaryotic initiation (Paper III) ................ 32Stop codon recognition in two different eukaryotic compartments (Papers I & IV) ................................................................................ 33Ligand binding in purine riboswitches (Paper II) ............................ 40

Computational Methods ............................................................................ 42Molecular mechanics ....................................................................... 42Molecular dynamics ........................................................................ 43Free-energy perturbation ................................................................. 45Linear interaction energy ................................................................. 48Homology modeling ........................................................................ 49

Sammanfattning på svenska ...................................................................... 50

Acknowledgements ................................................................................... 53

References ................................................................................................. 54

Abbreviations

DNA RNA mRNA tRNA rRNA (e)RF mtRF (e)IF EF-Tu GTP GDP ATP ORF MD FEP LIE T R kB

Deoxyribonucleic acid Ribonucleic acid Messenger-RNA transfer-RNA ribosomal-RNA (eukaryotic) Release factor Mitochondrial release factor (eukaryotic) Initiation factor Elongation Factor thermo unstable Guanosine-5’-triphosphate Guanosine diphosphate Adenosine-5’-triphosphate Open Reading Frame Molecular dynamics Free-energy perturbation Linear interaction energy Absolute temperature Gas constant (1.987 x 10-3 kcal/K/mol) Boltzmann’s constant (1.380 x 10-16 J/K)

9

Introduction

This thesis, and the work presented within, is primarily focused on various binding events on the ribosome investigated through computational ap-proaches. The concept of molecular binding, and especially, reversible bind-ing, is that the molecules should also be able to dissociate after such events. This ability to dissociate is fundamental for a cell to continue growing. But, interacting molecules should not bind too weakly so that the biological rele-vance of the binding is lost. Let us think of the ribosome for instance, whose function is to produce new proteins for the cell by translating a specific mes-senger-RNA (mRNA). The fully functional ribosomal machine, in other words, determines if a cell will continue to grow or not. In order to do so, the ribosome needs to rapidly decide what to take in, and what to discard, and in which order. During protein synthesis this is evident, when a large number of tRNA’s are practically barraging the ribosome, and the accuracy of protein synthesis relies on the basis that a large number of different molecules and proteins associate and dissociate from each other in an extremely precise man-ner; first, the ribosome must start translation at the correct codon, second, it must select the correct transfer-RNA (tRNA) to bind the correct codon and third, stop accepting tRNA and instead allow binding of a peptide release fac-tor (RF) protein to one of three stop codons to terminate translation, and break the nascent chain. Before any of these events can take place, the ribosome itself must be formed. The ribosome is assembled by a handful of ribosomal RNA (rRNA) molecules and a large number of ribosomal proteins that must all come together, perfectly and in sequence.

In computational chemistry and biology, molecular binding of different molecules can be calculated and compared to experimentally obtained data via the concept of thermodynamic stability, or free-energy. Free-energy calcula-tions have played a large role in connecting biochemical experiments and structural interpretations of biological processes (see Åqvist et al., 2012; Lind et al., 2017 and references therein). Thus, it is of great importance to be able to calculate, with high precision, free-energies at the atomistic level in order to understand the macroscopic behavior of, for instance, the ribosome.

10

Outline The research about the ribosome presented here, involves the first and last steps of protein synthesis. Here the focus was on the eukaryotic initiation pro-cess and how such a complex process can be extremely precise (Paper III). Its direct opposite, which describes the termination of mRNA translation is pre-sented in Papers I and IV. Paper I describes the human mitochondrial termi-nation system and the usage of non-standard stop codons, followed by describ-ing the omnipotence in eukaryotic cytosolic termination by eRF1 in Paper IV. Paper II, considers a specific element of the mRNA molecule called a ri-boswitch. Via binding of small molecules, the riboswitch exerts a wide range of regulatory control comprising transcription, translation and splicing. In our study, we focused on the purine riboswitches found in bacteria.

To evaluate the various binding events, whether it is a small ligand binding to a riboswitch, a large tRNA searching for a specific three-letter codon or a release factor binding a stop codon with highly specific protein-RNA interac-tions, all projects have been approached through computational methods. Cal-culating free-energies is a great addition to traditional biochemical methods of measuring thermodynamic parameters, as free-energy calculations often are directly comparable to experimental data. It also provides the missing link be-tween structure and function.

In the next sections I will give a brief historical overview of how determin-ing the structure of the DNA helix and how early work in protein synthesis started a whole era of ribosomal research, and how the ribosome has been studied since then.

Finally, I will present my own contributions to the ribosomal field, and the methods I have been using throughout this thesis. For the interested reader, I recommend reading this comprehensive summary and the attached papers to get a more detailed view of my contributions in the field of ribosomal research.

11

The genetic code and RNA

Every single process inside of a living cell is governed by precise interactions of small building blocks, atoms. These atoms in turn make up molecules. Mol-ecules can range from a few interacting atoms to enormous macromolecular machines made up of millions of atoms. Proteins are made up by amino acids, which are coded for in genes and transcribed from long nucleic acid sequences in our DNA. But not all genes code for proteins made up by amino acids. One such molecular machine is the ribosome. Here, both RNA and proteins are assembled as different molecules of various sizes, to function as one unit. Non-covalent binding events on the ribosome are of critical importance for its function. A detailed insight into this large complex can help us understand the fundamental driving forces as to why a biological reaction happens one way and not the other.

But, the for us today with modern technologies, obvious structure, form and shape of nucleic acids has not always been so clear. Therefore, I would like to start from the beginning and take a moment to introduce the structural point-of-view starting from the determination of the DNA structure and slowly moving on to RNA and protein synthesis.

The structure of DNA In 1953, Pauling and Corey proposed an initial model of the DNA molecule (Pauling and Corey, 1953). Their model comprised a three-chained cylindrical molecule with the phosphorous groups pointing inwards, thus making up the core of the helix and the bases pointing away from each other. Each chain would resemble the way in which a protein a-helix is structured. However, the duo argues the different possible compositions of the core, but rule out both options of purine-pyrimidine groups as well as sugar groups. They fur-ther argue that their proposed model, with the bases on the periphery would allow nucleic acids to “interact vigorously with other molecules”. The authors, however, seemingly did not include the biological aspects of DNA and the genetic code in their model, with, for instance, how the genetic code is main-tained and protected against changes, gene expression via transcription and translation nor did they model the phosphate backbone correctly.

12

Experimental work by Chargaff (Chargaff et al., 1950) investigated the dis-tribution of the four DNA units (purines: adenine (A) and guanine (G); pyrim-idines, thymine (T) and cytosine (C). In RNA thymine is replaced by Uracil (U)) within different organisms and concluded the existence of a pattern cor-related to its distribution. They had found that, in principle, the ratio of pyrim-idines and purines were equal and later (Zamenhof et al., 1952) that one could differentiate this even further stating that the ratio of adenine to thymine, and guanine to cytosine, often were very close. This study must have been a mag-nificent piece of information needed to finally solve the shape and structure of DNA.

The double helical structure of DNA was first proposed by James D. Wat-son and Francis H. C. Crick in their classical Nature paper in 1953 (Watson and Crick, 1953a). Here they describe both their own model and why the model of Pauling and Corey was unsatisfactory. In my opinion, by reading and analyzing much of the, at the time, ongoing research about deoxyribonu-cleic acid and its structure, not much of this research concerned its stability. However, Watson and Crick bring attention to this matter by pointing out that the model proposed by Pauling and Corey was lacking stability – “Without the acidic hydrogen atoms it is not clear what forces would hold the structure together, especially since the negatively charged phosphates near the axis will repel each other” (Watson and Crick, 1953a). They also managed to incorporate the findings of Chargaff to their model in order to solve which nucleotides would interact (base-pair). Their model (Figure 1) turned out as two antiparallel helical chains, meaning that the chains have opposite directions, and intertwine around the same axis. Hence if strand 1 contained an A at a given position, the corresponding position on strand 2 would have to be a T. Equally with a G, the complement would have to be a C. Hence the idea of the genetic code was made up by its complementariness. They further found that these specific pair-ings were only allowed if the bases are in their most probable tautomeric keto form rather than their enol form (a topic that recently caused some quarrel in another context (Pavlov et al., 2017)). In their view, as an A-T bond, in the keto form, only allows for two hydrogen bonds, it was thus assumed that also a G-C bond would consist of two hydrogen bonds. This error was, however, addressed by Pauling and Corey three years later in a study on specific hydro-gen-bond formation on DNA (Pauling and Corey, 1956). The DNA model and how it interacted also satisfied earlier work by Astbury and Bell that had found that each nucleotide in their stacked plane was separated by 3.4 Å (Astbury and Bell, 1938), which is true for standard B-form DNA. While Watson and Crick were quite confident that their model was correct, they could not be a hundred percent sure.

13

Figure 1. Representation of a current 3D reconstruction of the DNA helix first pro-posed by Watson and Crick. The arrows show the directionality of each DNA strand. This image was created from a solution NMR structure of DNA duplex, PDB: 2L8Q (Julien et al., 2011)

Along with Watson and Crick’s proposed model, analysis of X-ray crystallog-raphy diffraction by helices was done, suggesting that the two-chain nucleic acid helical structure do exist in biological systems (Wilkins et al., 1953). Later it was further proved by Franklin and Gosling that the proposed DNA model indeed seemed to be correct, and that the DNA molecule can be made in two conformers (A and B form) (Franklin and Gosling, 1953a). They later showed that the two-chain helix could be determined in both forms and that both exist in double helical form under different conditions (Franklin and Gos-ling, 1953b).

Some years before Watson and Crick proposed the DNA model, it had been suggested that deoxyribonucleic acid indeed was the material of genes, and “the available evidence strongly suggests, then nucleic acids of this type must be re-garded not merely as structurally important but as functionally active in determining the biochemical activities and specific characteristics” (Avery et al., 1944). DNA replication was later discussed and the idea of complementarity and how the DNA helix could duplicate itself (Watson and Crick, 1953b).

It should not go unnoticed that Bruce Fraser proposed an earlier model sim-ilar to the one of Watson and Crick, with striking similarities, with external phosphate and the bases making internal hydrogen bond interaction. However, this model was also made up by three strands of DNA and not two. Although this paper is stated to be ‘in press’ when Watson and Crick published their model, this work was never published. Additionally, Sven Furberg also touched on the possibility that DNA could have a rod-like structure (Furberg, 1952), but he never managed to piece his data together.

14

What about RNA? So far, only the structure of DNA had been determined, yet, how the genetic code and its message turned into proteins was still a mystery. Although there was an existing knowledge about RNA and that it was often found inside the cytosol where protein synthesis occurred (Caspersson, 1947; Caspersson and Schultz, 1939) its function had yet to be determined.

During the 1950s, large molecules containing a significant part of the cells total RNA had been discovered. These molecules were initially termed ‘mi-crosomal particles’ (it was agreed to rename these particles to ribosomes in 1958 (Roberts, 1958) and I will use the term ribosome). In a paper from 1958, Crick argues for the essence of protein synthesis and the potential role of these particles in protein synthesis (Crick, 1958). Since it had already been stated that DNA makes up genes that, somehow, need to be converted into proteins, he argues the protein synthesis is a flow, a flow of information. Zamecnik and co-workers had studied protein synthesis in rats using radioactively labeled amino acids that constantly would attach to the ribosomes. Soon after, more evidence of ribosomes being an active part in protein synthesis appeared, nonetheless, how it was connected was still not clear. One of the theories, of a more speculative nature, suggested that each ribosome contained specific cavities in which the amino acids would bind, however, all attempts to re-create such cavities had failed. This was the so-called one gene, one ribosome one protein hypothesis.

Crick now proposed two complementary theories of his view on protein synthesis:

a. the sequence hypothesis, where he explains that the specificity of a piece of RNA (or DNA) is solely expressed on the sequence made up by the bases of a given molecule, which in turn, determines the se-quence of the protein product.

b. the central dogma, “the transfer of information from nucleic acid to nucleic acid, or from nucleic acid to protein may be possible, but transfer from protein to protein, or from protein to nucleic acid is im-possible” (Figure 2) (Crick, 1958)

Essentially, this is exactly how protein synthesis occurs – the flow of infor-mation – but Crick did not manage to sort the puzzle completely. He argued that there existed two kinds of RNA and that it was possible that the ribosome, which in major part consists of RNA, was one of the them. His idea was that each ribosome carries a template RNA, meaning that each ribosome contained the message coding for each protein. The other type of RNA, the adaptor or

15

Figure 2. An unpublished drawing of the central dogma by Crick on ‘Ideas on Pro-tein Synthesis’ October 1956 (Image: Wellcome Library, London.)

the soluble RNA, would carry each amino acid to its corresponding cavity on the ribosome. In this view, which goes a little in line with the one-gene, one-ribosome theory, the ribosome would contain the template RNA sequence car-rying the information from the gene. Hoagland, Zamecnik and Stephenson had shown that it was possible to attach amino acids to these soluble RNA mole-cules and by diffusion they could bind the ribosome (Hoagland et al., 1958; 1957). The soluble RNA, however, was later shown to be a different part of protein synthesis, namely what today is referred to as tRNA. Although, one could argue that Crick was spot on with the template located inside the ribo-some, he did not see that one RNA molecule was still missing, the messenger.

Although it had been mentioned earlier (Dounce, 1953; Jacob and Monod, 1961) that there should be some sort of message carrying the information from the gene to an intermediate instead of the ribosome being this intermediate, it was not until two independent groups accomplished the isolation of the spe-cific molecule, announcing the isolation of the mysterious messenger RNA molecule as an unstable short-lived intermediate (Brenner et al., 1961; Gros et al., 1961). It had now been showed that after phage infection in E. coli, bacterial protein synthesis would stop, but not phage protein synthesis and ribosome biogenesis. Hence the previously proposed theory that each ribo-some was gene specific had now been proved false.

Cracking the code While it would take some years before the entire codon table (Figure 3) was completed in 1966, pioneering work by Nirenberg and Matthaei (Nirenberg and Matthaei, 1961) showed that polyuridylic acid contained the necessary

16

information to form poly-L-phenylalanine. But if the code was made from a single uracil or more they could not tell.

Figure 3. The standard genetic amino acid codon table depicted as a wheel. Starting at the middle (first position) and read to its outer boundary (second and third posi-tion, respectively) show which codon representing which amino acid. For instance, starting at the middle A and next letter U followed by G, reads out as AUG, which is the start codon for almost every ribosomal translated protein in our genome.

Although there were speculations on a potential signature for the coding se-quence of mRNA (Crick, 1958; Gamow, 1954), Crick and co-workers deter-mined that the coding sequence had to be made from a three nucleotide tem-plate, and the amino acids were incorporated in a sequential matter, and no overlapping sequences were allowed (Crick et al., 1961). Within a few years, Nirenberg and Leder developed a sophisticated method using 14C-labeled ami-noacyl-tRNA (they neatly labeled one amino acid at a time) and let random triplets of RNA pass through a filter covered with ribosomes. By washing off the unbound tRNA they could easily detect which amino acid was bound to both tRNA and ribosome (Nirenberg and Leder, 1964). In a series of articles Nirenberg and colleagues managed to determine an almost complete codon table (Nirenberg et al., 1966) and a new era of protein synthesis had begun.

17

Hopefully this short historical summary of how determining the DNA struc-ture and how early understanding of protein synthesis was somewhat coupled, has given the reader a better understanding of what fundamental and difficult challenges were solved in the early history of protein synthesis research. The following sections will shift focus to the present research.

18

The ribosome

The ribosome is a large ribonucleoprotein complex with enzymatic activity (Nissen et al., 2000) found in all living cells. It is universally comprised of two subunits, one small and one large. Both subunits play specific roles during protein synthesis. The small subunit holds the mRNA and binds the anticodon stem loop (ASL) of tRNA’s passing through the ribosome, while the large subunit interacts with the body of the tRNA and the growing nascent peptide chain in the exit tunnel (Figure 4). Even though these events are highly con-served through all kingdoms of life, the ribosomes themselves come with dif-ferent compositions regarding RNA-protein ratio. Table 1 shows the main dif-ferences between ribosome compositions for ribosomes of different biological kingdoms.

Table 1. Comparison between bacterial, mammalian mitochondria, chloroplasts in plants and human ribosomes.

Species Property

Bacteria (E. coli)

Mitochondria (Mammalian)

Chloroplast

Eukaryote (Human)

Molecular mass 2.3 MDa 2.7 MDa 2.6 MDa ~4.3 MDa Subunits 30S + 50S 28S + 39S 30S + 50S 40S + 60S Sedimentation coefficient 70S 55S 70S 80S Small subunit composition

16S rRNA 1534 nt

21 proteins

12S rRNA 962 nt

30 proteins

16S rRNA 1491 nt

25 proteins

18S rRNA ~1800 nt

33 proteins Large subunit composition

23S rRNA 5S rRNA

3024 nt

33 proteins

16S rRNA

1560 nt 52 proteins

23S rRNA 5S rRNA

4.8S rRNA 3033 nt

33 proteins

28S rRNA 5.8S rRNA 5S rRNA ~4000 nt a 47 proteins

a The cytosolic eukaryotic large subunit varies between 26S-28S and with it the number of nucleotides range from ~3400 to ~5035

As seen in Table 1, the size of the ribosomes is not conserved throughout kingdoms, both regarding mass, number of nucleotides, and ribosomal pro-teins. But, if one looks closely at the peptidyl transferase center or the mRNA decoding site, one finds high conservation of both structure and key residues (Woese and Fox, 1977).

19

As one may have understood from the Introduction, early detailed under-standing of protein synthesis was rather limited. About two decades ago, the ribosomal field had a major breakthrough regarding structural information. Now the first high-resolution x-ray crystallographic structures of the prokary-ote Thermus thermophilus small subunit (Schluenzen et al., 2000; Wimberly et al., 2000) and the large subunit of Haloarcula marismortui (Ban et al., 2000) had been solved. Since then, a large variety of ribosome structures have been solved in various stages of protein synthesis. The development in the field of structural determination has been a tremendous help in fine-tuning understanding of the ribosomes function and recent advances of Cryo-EM methods have successfully been able to further broaden our understanding of the ribosome.

Figure 4. Structural view of the bacterial 70S ribosome in complex with mRNA (black), A-site tRNA (yellow), P-site tRNA (red) and E-site tRNA (magenta) and the nascent peptide chain (dark blue) shown in surface representation. The 30S small subunit and 50S large subunit are shown as transparent green cartoons, respectively. Ribosomal proteins are shown in blue as cartoon representation (Voorhees et al., 2009). The peptide chain and the extension of mRNA was modeled from other struc-tures (Hussain et al., 2016; Su et al., 2017).

20

Initiation of protein synthesis Bacteria Each round of protein synthesis is initiated by an mRNA binding to the initi-ation factor (IF) 3 bound 30S subunit. The 5’ UTR of the mRNA contains a specific purine rich segment called Shine-Dalgarno (SD) sequence (Shine and Dalgarno, 1974) by which it base-pairs to its complementary sequence, anti-SD, on the 16S rRNA. These interactions properly position the start codon in the ribosomal P-site. Subsequently, IF1 and IF2•GTP bind to the 30S subunit and enhance the recruitment of fMet-tRNAfMet. IF2 recognizes N-formylated methionine on the initiator-tRNA (Antoun et al., 2006a) which is not present on the tRNAMet used during elongation. IF1 binds to the ribosomal A-site and blocks premature binding of elongator tRNA’s. IF1 has further been shown to bind and protect A-site rRNA residues known to be important for correct tRNA selection during elongation (Dahlquist and Puglisi, 2000; Moazed et al., 1995). IF3 has been suggested to indirectly control the accuracy of initiation by destabilizing all incoming P-site tRNA’s (Antoun et al., 2006a) and will spontaneously dissociate upon initiator tRNA association (Antoun et al., 2006b). When the correct tRNA has bound, the 50S subunit will associate forming the translation competent 70S ribosome. This leads to GTP hydrolysis on IF2 and dissociation of IF1 and IF2 (Figure 5).

Figure 5. Simplified overview of initiation of protein synthesis in bacteria.

Eukaryotes While the main purpose of eukaryotic initiation remains the same as in pro-karyotes, the mechanism by which it is achieved is far more complex. Again, to initiate protein synthesis the start codon must be positioned in the ribosomal P-site on the small subunit. However, both the mRNA and the eukaryotic 40S subunit lack the SD sequence, meaning that mRNA does not have a specific anchoring point on the ribosome. Instead, upstream of the, usually, first in-frame AUG codon, the mRNA contains a specific pattern known as the Kozak sequence (Kozak, 1987) consisting of predominantly an adenine at the -3 po-sition, relative to AUG. Here, the initiation complex is formed via a scanning

21

mechanism (Kozak, 1978) using at least 12 different initiation factors (prokar-yotes use only 3). The small ribosomal subunit first forms a complex with initiation factors eIF1, eIF1A, eIF3 and eIF5 (Asano et al., 2000). This com-plex then recruits the initiator tRNA (note that the eukaryotic initiator tRNA is not formylated as it is in prokaryotes) in a ternary complex with GTP bound eIF2 forming the 43S pre-initiation complex (Figure 6).

Simultaneously, initiation factors 4A, B, E and G bind to the mRNA and together with poly-adenine binding protein (PABP) encircle mRNA. Now the 43S complex and the mRNA complex join in a 48S scanning competent com-plex. The 43S complex binds close to the m7G 5’-cap of the mRNA and begin an ATP driven scan along mRNA (Hinnebusch, 2014). During scanning, the ribosome complex switches between two conformations, PIN when the tRNA senses the mRNA triplet (scanning) and POUT when the tRNA moves away from the mRNA in order for a new triplet to enter the P-site. Once the start codon is located, the IFs dissociate from the 43S complex and the large subunit joins.

Since the start codon is not pre-positioned in the P-site, as in prokaryotes, this process must be highly accurate not to miss, or wrongly initiate, at an incorrect codon. In Paper III we calculated binding free-energies of different near-cognate codons relative to the AUG start codon.

So far, only one start codon has been mentioned, namely the standard AUG. But, it should not go unnoticed that in some cases there are alternative start codons. Based on E.coli usage, one could say that the AUG codon is universal as it is used in 83% of all translated mRNA, followed by GUG (14%) and UUG (3%) (Blattner et al., 1997). However, these are all initiated with the standard initiator fmet-tRNAfmet. Although alternative start codons may be ex-tremely rare in eukaryotes, recent data shows that in mammalian translation, the mRNA coding for the major histocompatibility complex (MHC) class I can utilize the leucine-tRNA to initiate at the CUG codon (Starck et al., 2012).

22

Figure 6. Simplistic view of the scanning mechanism leading to the AUG start co-don in eukaryotic protein synthesis. The pre-initiation complex (PIC) binds close to the m7G-cap of the mRNA. The cap and the poly-A tail makes the mRNA connected via the multifactor complex (MFC) made up from different initiation factors and other proteins. The mRNA is then threaded through the PIC in the search for the co-don. Initiator tRNA (red), eIF1 (orange), eIF1a (light orange) and eIF2 with a, b and g subunits (olive) are shown in surface representation. For details, see page ZZZ.

23

Elongation of the peptide chain Not much is known about the elongation step in eukaryotes, since the majority of studies have focused on the bacterial translational system. However this process is believed to be very conserved between different kingdoms (Rodnina and Wintermeyer, 2009).

The elongation process can be divided into two phases; decoding and trans-location (Schmeing and Ramakrishnan, 2009). I will mainly discuss the de-coding phase since it is more closely related to the topics and methods studied within this thesis.

Decoding Following translation initiation, the mRNA is ready to be translated by se-quentially adding one amino acid at a time to the growing peptide chain into the final product of a functional protein. Now, the 70S ribosome holds the initiator tRNA in the P-site and the A-site is empty, free to accept incoming aminoacylated tRNA’s. This tRNA, loaded with its amino acid and in ternary complex with EF-Tu and GTP, binds the ribosome. EF-Tu binds the highly conserved sarcin-ricin loop (SRL) on the 50S subunit which coordinates the catalytic center for GTP hydrolysis (Schmeing et al., 2009; Voorhees and Ra-makrishnan, 2013; Voorhees et al., 2010). Upon binding, the anticodon stem-loop of the tRNA base-pairs with the A-site mRNA codon, while the body of the tRNA is still bound to EF-Tu. This conformation is referred to as the A/T-state (Figure 7).

Figure 7. Schematic overview of the peptide elongation cycle. (Left to right) The cy-cle begins with a tRNA in the P-site (red) holding the nascent chain an empty A-site. tRNA (yellow) binds the A-site in ternary complex with EF-Tu(GTP) (orange) in the A/T-state. Here, the tRNA can either be accepted or rejected. If accepted, GTP hy-drolysis on EF-TU leads to a conformational change on the tRNA. EF-Tu(GDP) dis-sociate and the tRNA gets fully accommodated and the peptide chain is transferred to the A-site tRNA.

24

Initial selection: At this point, an initial tRNA selection occurs based on sequence complementarity of the tRNA and the mRNA which forms a mini-helix in the A-site. Here, the key rRNA residues A1492 and A1493 on helix 44 of the 30S subunit flip out from the helix from an OFF to ON state, simi-larly G530 switches its conformation from syn to anti (Ogle et al., 2001). The concept of ‘monitoring bases’ arose from comparison of the anticodon stem loop (ASL) bound 30S ribosomes with both paramomycin (Fourmy et al., 1996), which binds in the pocket created when the bases pass from OFF to ON, and without (Ogle et al., 2001) and the 30S ribosome with an empty A-site (Wimberly et al., 2000). It was suggested that the bases would actively ‘monitor’ correct Watson-Crick base-pairing geometries between the codon-anticodon, hence the ON and OFF notation (see Satpati et al., 2014).

Recent cryo-electron microscopy (Cryo-EM) structures of pre-accommo-dated intermediates (preceding the A/T conformation) show tRNA binding to the open conformation A-site without base-pairing to the codon and before any contact between EF-Tu and the SRL (Loveland et al., 2017). These new structures suggest that the G530 residue plays a central role prior to the con-formational change of residues A1492 and A1493. The authors suggest, that for a cognate tRNA, before any contact of the codon-anticodon, the confor-mations of the monitoring bases are: G530, A1492 and A1913 (of the LSU) OFF, while A1493 is SEMI-ON as it is prearranged to bind the codon-antico-don. Followed by a conformational change in the tRNA, which shifts it closer to the codon and G530 interacts with the sugar base of the second nucleotide on the anticodon and third nucleotide on the mRNA, G530 SEMI-ON, while A1492 and A1493 can alternate between ON/OFF (Loveland et al., 2017). Once the tRNA adopts the A/T state, G530, A1492, A1493 and A1913 are all ON as seen previously for cognate tRNA in A/T (Schmeing et al., 2009; Voor-hees et al., 2010) and A/A states (Demeshkina et al., 2012; Selmer et al., 2006).

Earlier studies have, indirectly, suggested that the monitoring bases only “monitor” or flip out on correct codon-anticodon interactions when a cognate tRNA enters the ribosomal A-site. The correct interactions or geometry would send a signal for EF-Tu (some 75Å away) to hydrolyze GTP (Gromadski and Rodnina, 2004; Pape et al., 1998). Kinetic experiments show that GTP hydrol-ysis on EF-Tu is dramatically decreased for near- or non-cognate tRNA’s, which would be an effect of slow codon recognition. This was presented as the ‘induced fit’ model and suggested that the ribosome reacts uniformly to different kinds of mismatches (Gromadski et al., 2006). Although the induced fit model may be correct it has been shown that the E. coli ribosome has an error frequency of about 10-4 – 10-3 but varies depending on the codon (Kramer and Farabaugh, 2007; Parker, 1989) which could argue against a uniform re-sponse to near-cognate tRNA’s.

Computational approaches have also been used to elucidate the atomistic mechanisms of discrimination of near-cognate codons (Allnér and Nilsson,

25

2011; Almlöf et al., 2007; Sanbonmatsu and Joseph, 2003) and the function of the monitoring bases (Satpati et al., 2014; Zeng et al., 2014), however, here the authors seem to draw different conclusions on the matter. The first model (Satpati et al., 2014) suggests that all tRNA, regardless of whether they are cognate or not, must (or can) pass from the A/T state to the A/A state with the monitoring bases ON. The calculated energies show that it would be more difficult for a near-cognate codon to pass from one state to the other due to large free-energy barriers associated with turning the monitoring bases from OFF to ON. Earlier molecular dynamics simulations on codon-anticodon in-teractions (Almlöf et al., 2007) suggested an alternative function for the mon-itoring bases A1492 and A1493. Here it was seen that the monitoring bases could function as a water shield preventing unwanted solvent molecules to interrupt the codon-anticodon interactions. This idea was also supported in the later study (Satpati et al., 2014).

The second study suggests that the energy barrier for GTP hydrolysis is associated with the flipping of the bases from OFF to ON. In the case of a cognate tRNA the energy of flipping would thus be negative (favourable) while reaching the same state in the near-cognate case would be positive (un-favourable). Hence, their conclusion states that, for a near-cognate tRNA the monitoring bases A1492 and A1493 would, with high probability, remain in their OFF conformation (Zeng et al., 2014). A recent study on mammalian ribosomes (Shao et al., 2016) argues an in-between model. The eukaryotic monitoring bases A1824, A1825 and G626 (E. coli: A1492, A1493 and G530, respectively) would act very much like the bacterial ones, but depending on the context of the A-site tRNA, the probability for the monitoring bases to stay ON is much greater for a cognate tRNA than a missense tRNA. This indicates that, if a near-cognate tRNA binds, the bases switches from OFF to ON, but can flip back to their native conformation and induce a destabilizing effect which ultimately leads to tRNA rejection.

However, structures of both the A/T and A/A states of near-cognate tRNA in the A-site exist, showing that the G530 conformation goes from OFF to ON as well as A1492, A1493 and A1913 have identical conformations to those in cognate structures (Demeshkina et al., 2012; Loveland et al., 2017; Rozov et al., 2015) hinting that the energetic barrier passing from one state to the other may not be so high.

Proofreading: Before a tRNA is fully accommodated on the ribosome (A/A state), proofreading occurs (Johansson et al., 2008; Ruusala et al., 1982; Thompson and Stone, 1977), that contributes to the total accuracy of tRNA selection. The proofreading was initially believed to occur in one step, but recently, this view was challenged, suggesting that there is actually two proof-reading steps (Ieong et al., 2016), rather than one. This can lead to either spon-taneous dissociation after dissociation of EF-TU or together with EF-Tu and GDP of an incorrect tRNA (Ieong et al., 2016).

26

The total accuracy for tRNA selection can be summarized as a function of these steps depending on the rate constants in the different steps:

𝐴 = 𝐼 ∙ 𝐹&'& =

()*+)

,-)

()*+.)

,-.) =

(0)

(0.)

10)

10.)

12)

12.) (1)

where

𝐹&'& = 𝐹6or

𝐹&'& = 𝐹6 ∙ 𝐹9

where A is the total accuracy, I initial selection, Ftot is the proofreading factor, kcat/Km is the kinetic efficiency for cognate (c) and near- or non-cognate (nc). kI is the rate of ternary complex association to the A-site and PI is the proba-bility that this leads to GTP hydrolysis on EF-Tu and PF the probability of peptide bond formation (Ieong et al., 2016; Johansson et al., 2011).

Translocation Once the tRNA has passed all selection steps, the amino acid in the PTC makes a nucleophilic attack on the carbonyl carbon in the ester bond linking the nas-cent peptide chain to the P-site tRNA which transfers the peptide chain to the tRNA of the A-site (Hansen et al., 2002; Trobro and Åqvist, 2005). The P-site tRNA is now deacylated. This triggers ribosome ratcheting, which causes the two ribosomal subunits to fluctuate in a counterclockwise rotation relative to each other (Moazed and Noller, 1989; Valle et al., 2003). During the fluctua-tion, the ASL of the A- and P-site remains in their respective positions on the small subunit, the rotation, however, puts the A-site and the P-site acceptor ends in the P-site and E-site, respectively, on the large subunit. These states are denoted as A/P and P/E hybrid states. At some point during the rotational fluctuation, EF-G binds on the A-site side of the ribosome. The ratcheted hy-brid conformation is stabilized by binding of EF-G•GTP (Spiegel et al., 2007). Although EF-G is known to catalyze the movement of the tRNA and mRNA on the small subunit, the mechanism by which this is achieved has not yet fully been determined.

The above described elongation processes are continuous and repeated until the ribosome reaches a stop codon, for which there are no cognate tRNA’s.

27

Termination of protein translation Once the translational apparatus reaches the end of an open reading frame (ORF), one of the three stop codons UAA, UGA and UAG (Brenner et al., 1967; 1965) is presented in the ribosomal A-site. For these codons, there are no cognate tRNA. Instead, two class-I release factors in bacteria, RF1 and RF2 bind these codons and terminate peptide translation by releasing the nascent chain from the tRNA (Capecchi, 1967; Caskey et al., 1971; Vogel et al., 1969) (Figure 8). Since there are three codons and only two RFs, they share the same specificity for UAA. Moreover, in addition to the UAA codon, RF1 binds to UAG and RF2 to UGA (Scolnick et al., 1968).

Figure 8. Termination of protein synthesis in bacteria. Once a stop codon is located in the A-site, the RF (orange) binds the ribosome and cleave the nascent peptide chain held by the P-site tRNA. RF3(GDP) (blue) binds the complex to help dissocia-tion of the RF. The ribosome is now done with one round of protein synthesis and can be recycled by subunit separation and can initiate on a new mRNA.

Early theories speculated that the RF’s would interact with the stop codons in a similar fashion as tRNA does (Ito et al., 1996; Moffat and Tate, 1994) , this is known as the mimicry theory. An impressive genetics and mutational study of RF1 and RF2 later identified universally conserved motifs in domain 2, which is the codon reading domain (Ito et al., 2000). The more variable, yet conserved, PxT (x denotes any amino acid) was found in RF1 and the SPF motif in RF2. Although RF1 and RF2 are structural homologues, they only have a sequence homology of around 35%, the sequence similarity is over 50% (simple sequence alignment E. coli K12 strain). However, a substantial sequence homology is found in the key domains interacting with the codon and around the GGQ motif (Craigen and Caskey, 1987).

Comparable to tRNA, the RFs do not only bind to specific codons, but are also capable of catalyzing the final cleavage of the peptide bond by hydrolysis

28

of the peptide-tRNA in the P-site (Buckingham et al., 1997; Nakamura and Ito, 1998; Tate and C. M. Brown, 2002) and releasing it from the ribosome. This is done by the universally conserved GGQ motif (Frolova et al., 1999). The theory based on tRNA mimicry suggested that the stop codon would be read similarly as the standard sense codons with the bases stacked next to each other. This was also proposed in low-resolution structures (Petry et al., 2005). High-resolution reconstructions of the prokaryotic termination complex later revealed the mimicry not to be valid. Instead, the mRNA presented an unex-pected conformation (Korostelev et al., 2008; Laurberg et al., 2008; Weixlbaumer et al., 2008). Now only the first two bases were stacked, sepa-rated by a histidine. This conformation positions the third nucleotide perpen-dicular to the first and second bases (Figure 9). The structures show that the RF makes direct contacts with the stop codons, not only via Watson-Crick interactions but also through the Hogsteen edge of the nucleotides. Conse-quently, the interactions are distinctly different from Watson-Crick tRNA reading.

Figure 9. Structural view of RF2 (blue) binding to the UAA stop codon (black). RF2 is shown in cartoon representation, and the RF2 specific SPF motif (E. coli number-ing) is shown as sticks, as well as the histidine separating the second and third nucle-otide of the stop codon in both RF2 and RF1.

Since the release factors are proteins, the mechanism for reading the stop co-dons differ significantly from the interactions made during the elongation pro-cess. For instance, here, the need to boost the accuracy by the monitoring bases is no longer needed (Youngman et al., 2007) still the RF’s maintain an error frequency of 10-3 – 10-6 (Freistroffer et al., 2000). In the bound structures of bacterial RF’s, the monitoring bases are in a somewhat SEMI-ON confor-mation, with A1492 switched OFF (Laurberg et al., 2008). It was shown that upon binding the RF’s induce a conformational change in the decoding site and activation of both A1492 and A1493 could impair RF’s binding as it would produce a steric hindrance for RF to bind the A-site (Youngman et al.,

29

2007). Hence the discriminatory power seems to originate directly from the protein-mRNA interactions. Computer simulations later revealed that the co-don specificity did not only originate from the PxT or SPF motifs, but was made up from a large range of interactions involving both the RFs and water molecules (Sund et al., 2010)

The eukaryotic release factor, eRF1, on the other hand, is distinctly differ-ent from the bacterial RFs, but shares a larger homology to the RF found in Archaea (aRF1) (Atkinson et al., 2008; Bult et al., 1996). A major difference between the eukaryotic RFs and its bacterial analogs regards the codon speci-ficity. As described above, the bacterial RFs have dual specificity while e/aRF1 recognize all three stop codons (reviewed in Ehrenberg and Tenson, 2002). Within the stop codon binding domain, several conserved amino acid motifs have been found; TASNIKS, YxCxxxF (Bertram et al., 2000; Frolova et al., 2002; Song et al., 2000) and GTS (Conard et al., 2012; Kolosov et al., 2005; Wong et al., 2012) (human eRF1 numbering 58-64, 125-131 and 31-33, respectively). These conserved motifs have been subjected to a large number of mutagenesis and cross-linking studies aiming to decode how the omnipo-tence has been retained for the e/aRF (Blanchet et al., 2015; Chavatte et al., 2002; Conard et al., 2012; Frolova et al., 2002; Kryuchkova et al., 2013; Seit-Nebi et al., 2002). Even though the human eRF1 structure was solved in 2000 (Song et al., 2000), it took over a decade until high-resolution structures of eRF1 bound on the ribosome were available (A. Brown et al., 2015; Matheisl et al., 2015). Again, the conformation of the stop codon revealed an unex-pected twist. Unlike the conformations seen for both tRNA and prokaryotic RFs bound in the A-site, the mRNA now showed a much more compact shape (Figure 10).

The principles of stop codon recognition in eukaryotes will be covered in more detail in the coming sections.

Figure 10. Structural view of eRF1 (blue) binding to the UAA stop codon (black). eRF1 is shown as cartoon representation with the NIKS loop motif, Arg68 and Glu55 shown as sticks (Homo sapiens numbering).

30

Riboswitches

The above sections have to a large degree contained the detailed function of RNA in the form of rRNA, tRNA and mRNA, and their interplay. But, RNA can act in different ways. This is evident in the case of riboswitches. These molecules are parts of mRNA and can sense levels of a diverse range of me-tabolites throughout the cell and thereby serve as RNA genetic controllers. Via binding to the appropriate metabolite, the riboswitches change their con-formations resulting in modulating the expression of the gene, containing the riboswitch (Winkler and Breaker, 2003). Riboswitches not only have the ca-pability to bind small molecules, such as amino acids or nucleotide deriva-tives, they can also bind ions (Cromie et al., 2006; Ren et al., 2012) and large molecules such as tRNA (Henkin, 2008; Saad et al., 2013). These ligand sens-ing RNAs are commonly found in the 5’ untranslated region (UTR) of mRNA’s.

Riboswitches are typically made up from two distinct functional domains (Figure 11): an aptamer domain (Gold et al., 1995; Hermann and Patel, 2000) that plays a role as ligand binding domain, and an expression platform domain that interacts with the element involved in gene expression, for example, the ribosome.

Figure 11. General overview of a riboswitch being part of bacterial mRNA. When binding of the metabolite occur on the aptamer domain it induces a conformational change of the expression platform resulting in a change of expression in the down-stream gene

31

Riboswitches were first discovered in bacteria (Mironov et al., 2002; Nahvi et al., 2002; Winkler et al., 2002) and since then they have been believed to be functional in all three kingdoms (Serganov and Nudler, 2013; Wachter, 2010). Their ability to regulate specific gene expression is not only a critical feature for the living cell, but it is also of great interest from a pharmaceutical viewpoint. Riboswitches have, since their discovery, been found in 36 human bacterial pathogens and the riboswitch can therefore be a suitable target for new antibiotics (Breaker, 2012). One such riboswitch is the purine riboswitch which are usually found in genes regulating transport and metabolism of pu-rines in which, upon binding, can either silence or activate gene expression. These riboswitches bind either adenine or guanine via Watson-Crick interac-tions and a single pyrimidine residue (uracil or cytosine) at position 74 defines the specificity of the riboswitch (Mandal et al., 2003; Mandal and Breaker, 2004).

32

The present work

Start codon selection in eukaryotic initiation (Paper III) As has been described above, how protein synthesis is initiated in eukaryotes is quite different when compared to bacteria. One clear difference is that the eukaryotic ribosome does not contain the SD sequence. This means that, the mRNA cannot bind at a pre-determined spot, and position the start codon at the P-site. Instead, it relies on a scanning mechanism (Kozak, 1989). The scan-ning is initiated once the 43S PIC has been formed and uses the anticodon of the initiator tRNA Met-tRNAi to identify the correct AUG start codon (Figure 6). Since each mRNA is unique, the sequence before the start codon, the 5’ UTR, varies immensely, not only by sequence but also in length. With this in mind, to find the correct AUG triplet among a large variety of possible triplets then seems a bit of challenge.

In this project, we investigated how the initiator tRNA manages to distin-guish the specific interactions made up from AUG compared to six out of nine possible single nucleotide mismatches. To examine the relative start codon specificity, MD simulations and FEP calculations were performed for the near cognate codons. During the search for the correct codon the tRNA switches between the scanning conformation Pin and Pout. The Pin conformation is in which the tRNA scan the nucleotide content of the P-site. If the codon is in-correct, the tRNA switches conformation to Pout which is about 7Å away from the codon, letting a new nucleotide thread into the P-site. During the FEP cal-culations, the POUT was approximated by removing the tRNA from the simu-lated structure.

The FEP calculations show that the specificity for AUG compared to the six near cognate codons is very strong, 5-8 kcal/mol, corresponding to a factor > 4000 in affinity (Table 2). But as it turns out, the full discriminatory control does not sit entirely on the specific codon-anticodon made from the anticodon-mRNA interactions. During start codon scanning, the initiator tRNA is accom-panied by two initiation factors, eIF1 and eIF1A, that binds in the E- and A-site, respectively. The proximity of the IF’s to the codon-anticodon mini-helix provides an additional dimension for high fidelity as is evident when either one or both IFs are removed from the simulations. Here, a drastic decrease of the energies was shown. Notably, the effect of the IF closest to the first nucle-otide position was largest for transformation of adenine to guanine and equally

33

at the third position where guanine to adenine transformation had the largest effect.

Table 2. Calculated relative free-energies (kcal/mol) of initiator tRNA binding dif-ferent mRNA codons in presence and absence of eIF1 and eIF1A. Mutation

AUG®GUG AUG®CUG AUG®ACG AUG®AAG AUG®AUA AUG®AUC

+eIFs 7.77 ± 0.44 5.21 ± 0.47 6.25 ± 0.20 8.40 ± 0.65 6.48 ± 0.50 6.86 ± 0.54 –eIF1 2.76 ± 0.48 5.50 ± 0.47 7.17 ± 0.45 4.40 ± 0.80 5.76 ± 0.44 6.54 ± 1.00 –eIFs 2.80 ± 0.41 3.75 ± 0.44 6.11 ± 0.31 3.24 ± 0.36 3.10 ± 0.41 5.30 ± 1.01

It was not too surprising that the two proteins flanking the codon would affect codon selection. However, that the major contribution for discrimination against GUG (first position mismatch) and AUA (third position mismatch) is almost entirely made up from the presence of the IFs making the intrinsic tRNA discrimination less than 3 kcal/mol, was not expected. This corresponds to a 30-fold decrease in accuracy. The energies of the A-C mismatch at the middle position of the triplet was not affected by the presence of the IFs but the bulkier A-A mismatch, which had the largest calculated relative free-en-ergy, showed that half of the discrimination originates from eIF1. The calcu-lations performed in Paper III covered two thirds of the possible combinations of AUG and different compositions of pyrimidines and purines at different positions. It showed a rather uniform discrimination towards these combina-tions. It could thus be assumed that, with a scanning rate of ~8 nucleotides per second, the UUG, AGG and AUU codons would show similar large discrim-inations. Further, other triplets, for instance GUC, would follow the same trend of discrimination as the ones calculated. It is then clear that the IF’s eIF1 and eIF1A are crucial for codon selection and ensure a rapid scanning.

Stop codon recognition in two different eukaryotic compartments (Papers I & IV) The direct opposite of translation initiation, as discussed above, is the termi-nation in protein synthesis and here I will present our findings on translation termination. Papers I & IV describe peptide release factors and stop codon recognition in human mitochondria and cytosolic translation termination.

Mitochondria (Paper I) Most eukaryotic cells contain an organelle called mitochondria. It is required for production of the high-energy compound adenosine triphosphate (ATP) through oxidative phosphorylation. ATP is used as an energy source in the majority of the vital functions of the cell. In humans this involves muscle con-traction, cell signalling, DNA and RNA synthesis etc.

34

The mitochondrion has a DNA and translational system of its own, alt-hough it is rather small. The human mitochondrial genome contains only 37 genes (Anderson et al., 1981) coding for 13 electron transport proteins, two rRNA ribosomal subunits and 22 tRNA’s. All other proteins needed for its function are expressed, translated, and imported from the cytosol, this includes the mitochondrial release factors.

Two proteins identified inside the mitochondrion were classified as mito-chondrial release factors, mtRF1 and mtRF1a (Soleimanpour-Lichaei et al., 2007; Zhang and Spremulli, 1998) based on sequence similarity to bacterial RF1. Biochemical experiments have shown mtRF1a to have peptide release activity on both, stop codon programed bacterial ribosomes, as well as in vivo (Akabane et al., 2014; Soleimanpour-Lichaei et al., 2007). In mitochondria, the standard stop codons are reduced from three to only two with UGA read by tryptophan tRNA. However, the genes coding for two proteins being mem-bers in the human respiratory chain, mtCO1 and mtND6 end with codons AGA and AGG, respectively. These codons usually code for arginine.

Comparing the sequences of mtRF1 and mtRF1a one finds two additional insertions in mtRF1, Arg-Thr (RT) immediately before the tip of a-5 helix reading the uracil in bacteria, and Gly-Leu-Ser (GLS) in the recognition loop (see Figure 1 in Paper I for sequence alignment). It was initially hypothesized that these insertions were needed for AGA and AGG reading (Young et al., 2010), although, it showed low activity on bacterial ribosomes.

At the time, no crystal structures of the mitochondrial ribosome, nor the mtRF’s, were available (there are still no accessible structures of the mtRF’s). Instead, exploring the stop codon reading of the mtRF’s by computational means appeared viable. Homology models of both mtRF1 and mtRF1a were made with RF1 as template. The bacterial RF1 was used due to the mtRF’s having similar PxT motifs and that the RF2 UGA codon is not a stop codon in mitochondria. The models took into account the insertions of mtRF1, hence rRNA residues close to the recognition loop were added to the template struc-ture as well as one residue from the large subunit (A1913C). The models were then subjected to MD and FEP calculations. Although the energies of RF1 and RF2 had been obtained earlier (Sund et al., 2010), new control calculations were done, including the transformation of UAA to AAA. The simulations showed that the mtRF’s follow similar discrimination pattern to those of RF1. The simulations further show that both mtRF1 and mtRF1a can read UAA and UAG, but no sign of reading an adenine as the first nucleotide (Table 3).

35

Table 3. Calculated relative free-energies (kcal/mol) of bacterial and mitochondrial release factors.

Mutation Release factor

UAA®AAA UAA®UGA UAA®UAG UGA®AGA

RF1 4.6 ± 0.9 4.8 ± 0.6 0.1 ± 0.9 mtRF1a 4.9 ± 1.5 3.5 ± 1.1 0.8 ± 0.9 5.8 ± 1.1 mtRF1 5.1 ± 1.3 6.0 ± 0.7 0.5 ± 1.0 7.5 ± 0.9 RF2 -0.6 ± 0.5 5.8 ± 0.7

In conclusion, we could discard the hypothesis that mtRF1 would read the non-standard stop codons in human mitochondria. Instead, we proposed an alternative termination mechanism at these codons. Two other proteins found in mitochondria, homologues to RF1, are the immature colon carcinoma tran-script-1 (ICT1) and C12orf65 (Antonicka et al., 2010; Richter et al., 2010). These proteins however, lack the codon reading domain but contain the char-acteristic GGQ motif. ICT1 is a homolog of the bacterial alterative rescue fac-tor arfB (YaeJ), but contrary to arfB, ICT1 has been shown as an integral part of the mitochondrial ribosome (Greber et al., 2014; Richter et al., 2010). ICT1 was shown to induce peptide release on a large number of codons as well as an empty A-site (Richter et al., 2010). YaeJ rescues stalled ribosomes absent of mRNA or in the presence of repeated rare codon clusters such as AGA and AGG (Handa et al., 2011). However, the structure of YaeJ on an empty A-site shows that the alpha helix sensing the A-site would clash with the codon if one were present. Instead, a simple model of ICT1 based on the YaeJ structure was made on one of the alternative paths of the helix. ICT1 could then sense the non-standard stop codons without clashing with the codon. Our prediction was later supported by biochemical studies (Akabane et al., 2014).

Another hypothesis on the AGA and AGG codon reading involved a -1 frameshift (Temperley et al., 2010). In humans, CO1 and ND6 mRNA’s end with UAGA and UAGG. A -1 frameshift would thus create a situation where a standard stop codon UAG, which we show can be read by mtRF1a, sits in the A-site. However, the occurrence of a uracil before a non-standard stop codon is far from universal (Duarte et al., 2012; Lind et al., 2013). Further, another homology model study of mtRF1 suggested that mtRF1 could be a RF recognizing an empty A-site (Huynen et al., 2012) and rescue stalled ribosome similar to tmRNA in bacteria.

Both mtRF1 and mtRF1a are expressed and found inside the mitochondria. If one of these two have no function, what is the reason for having both? Es-pecially since none of them is coded by the mitochondrial DNA but has to be imported. The mtRF1 gene arose from a duplication of the mtRF1a gene (Young et al., 2010) and it could be that it is actually just a junk protein being expressed along with mtRF1a. Furthermore, does termination of 13 proteins require two RFs? In conclusion, we determined mtRF1a to be similar to bac-terial RF1 both regarding codon reading and discrimination against non UAA

36

and UAG codons, as well as proposing a hypothesis on termination of non-standard stop codons. Structural information of the mitochondrial ribosome in complex with the mtRF would be very helpful to figure out the last details about mitochondrial translation termination.

Eukaryotes (Paper IV) Cytosolic termination in eukaryotes, on the other hand, is carried out by a sin-gle RF. The eukaryotic release factor 1 (eRF1) has evolved to read all three universal stop codons. This means that eRF1, unlike the bacterial RF’s, man-ages to read combinations of both purines in the second and third position of the stop codon, and at the same time distinguish a GA or an AG pair from a GG pair.

eRF1 binds the ribosomal A-site accompanied by the class-II release factor eRF3•GTP in a pre-termination complex. Binding to the ribosome triggers GTP hydrolysis on eRF3 inducing a conformational change in eRF1. The pres-ence of eRF3•GTP greatly increase peptide release efficiency of eRF1 (Al-kalaeva et al., 2006). Thus, for eRF1 to switch between a compact eRF3 bound conformation to an accommodated one requires GTP hydrolysis and dissoci-ation of eRF3 (Alkalaeva et al., 2006; Shao et al., 2016). However, comparing the pre-accommodated and the accommodated structures reveals that the N-domain (eRF1 codon binding domain) and the conformation of the stop codon are very similar (A. Brown et al., 2015; Matheisl et al., 2015; Shao et al., 2016), indicating that the initial recognition of the stop codon has already hap-pened in the pre-accommodated binding event.

The RF’s described above and its eukaryotic counterpart are functional homologs, but they do not overlap structurally nor sequentially, apart from the GGQ motif, essential for peptide release. Just as the RF structures are differ-ent, this also extends to the conformation of the stop codons (Figures 9 & 10). Instead of stacking the two first nucleotides, and the third being flipped out separated by a histidine (Korostelev et al., 2008; Laurberg et al., 2008), eukar-yotes adopt another conformation. Recently determined structures of the ter-mination complex revealed the stop codon in a U-turn motif (A. Brown et al., 2015; Matheisl et al., 2015). The U-turn motif was first discovered in the an-ticodon loop of tRNA (Quigley and Rich, 1976), and has since been identified in several different structures, frequently occurs in rRNA (Gutell et al., 2000), and is a highly stable motif. A U-turn motif consists of three nucleotides and is usually made up of a UNR sequence. Here N stands for any nucleotide and R for purines. The U-turn motif is recognized sequence-wise by the UNR se-quence and structurally by a sharp turn of the phosphate backbone between the first and second nucleotide allowing the uracil to hydrogen bond to the phosphate of the 3’ residue following the UNR motif. Thus, the U-turn is a trinucleotide motif with a four-nucleotide requirement. The U-turn enables the three nucleotides succeeding the uracil to be easily accessed for solvent or

37

other interaction partners such as codon-anticodon (D'Ascenzo et al., 2017) or as in the translation termination case, its cognate receptor.

Prior to the recent structures of the eukaryotic termination complex (A. Brown et al., 2015; Matheisl et al., 2015; Shao et al., 2016), codon reading of eRF1 was biochemically studied and it was found that highly conserved motifs in the N-domain were important for its codon specificity. The TASNIKS se-quence was shown to cross-link to the invariant first position uracil (Chavatte et al., 2002). Reading of the AA, AG or GA combinations was believed to be done by the YxCxxxF and GLS motifs (Bulygin et al., 2011; Conard et al., 2012; Kolosov et al., 2005). Although the structures are now available and yield important information, they do not fully explain how eRF1 can bind the combinations AA, AG or GA but not GG. For instance, different interactions of the YxCxxxF motif where one group argues it interacts with the second nucleotide (Matheisl et al., 2015), while the second show it to interact with rRNA A1825 which would stabilize the codon via stacking of the last two nucleotides (A. Brown et al., 2015). The structures reveal the position of the crucial residue Glu55, whose importance is explained by its avoidance to read a double guanine pair due to electrostatic repulsion between the negatively charged amino acid and the O6 oxygen in guanine (A. Brown et al., 2015).

To further determine which residues are important for eRF1 to retain a high selectivity, we mutated each stop codon to either a near-cognate codon, by changing the first uracil to cytosine, or by changing one stop codon to another. Since the structures of all codons were solved, we could apply the transfor-mation between the two as well, e.g. UAA ® UGA and UGA ® UAA. We also mutated UGA/UAG ® UGG. The mutations were done using the FEP method. The energies reveal that eRF1 has a high selectivity for stop codons compared to the near-cognate codons (Table 4), while a low selectivity is seen within the different stop codons.

As was suggested by both cross-linking experiments, and the structures, Lys63 of the NIKS loop can directly interact with the O4 atom of the uracil. When changing this nucleotide to a cytosine, the positively charged Lys can no longer interact due to the NH2 group on cytosine. Since the stop codon takes a U-turn conformation, where a cytosine is not allowed, a cytosine will prob-ably weaken the stability of the stop codon conformation. As our MD simula-tions suggest, the NIKS loop appears very stable with limited backbone fluc-tuation. We reason that the stability partly comes from the highly conserved Arg68 (in some species, a lysine is present at this position), as it was seen during the MD simulations, where it adds two hydrogen bonds to the backbone of Ile62 and Lys63. Based on experimental work, Arg68Ala mutation com-pletely abolishes termination (Kryuchkova et al., 2013). We simulated this situation by manually changing the arginine to alanine and repeated the first position FEP calculations (Table 5). Interestingly, the discrimination towards the CAA, CGA and CAG codons drop from ~6 kcal/mol to less than 2 kcal/mol.

38

Table 4. Calculated relative free-energies (kcal/mol) of WT eRF1.

Mutation UAA®CAA UAA®UGA UAA®UAG UAA®UGG UGA®CGA UAG®CAG

Energy 5.65 ± 0.60 0.67 ± 0.57 -1.13± 0.53 6.54 ± 0.70 6.93 ± 0.43 6.71± 0.73

Table 5. Calculated relative free-energies (kcal/mol) of the Arg68Ala mutated eRF1.

Mutation UAA®CAA UGA®CGA UAG®CAG

Energy 1.80 ± 0.78 0.71 ± 0.75 1.08 ± 0.93

Likewise, discrimination against the UGG codon follows the predictions made for the first position. However, the position of Glu55 is not only distinguishing a GG pair. Since the stop codon has a U-turn conformation, and U is read in one part of the RF, we looked at the last two nucleotides separately. Here, we calculated the interaction energies between these two nucleotides to the sur-rounding amino acids. The amino acids were selected based on their experi-mental effect on termination efficiency (Kryuchkova et al., 2013). Now we could calculate the total energetics that the interactions of these amino acids contribute to the binding. We further dissected these energies into residue spe-cific ones (Figure 12). It is quite clear that the total interaction energy in read-ing AA, AG or GA is favourable, while the total for GG is close to zero. In the GG situation, a large positive (unfavourable) interaction energy for the GG-Glu55 pair was expected, but when looking at each nucleotide we see that Glu55 has unfavourable interactions to all combinations of guanine (Figure 12). While reading a guanine in the second position, the negative Glu55 inter-actions are countered by Cys127 of the YxCxxxF motif. Similarly, Thr32 of the GTS motif seems to interact with a third position G. Looking at the calcu-lated pair-wise interaction energies, of the YxCxxxF and GLS motifs, not many of the residues seem to contribute in a large extent to codon reading. But instead there is an interplay between Glu55 and Cys127. However, there does not seem to be much solvent aiding codon reading as is seen in bacteria (Sund et al., 2010). The last two nucleotides bind between two a-helices which largely contain hydrophobic amino acids, perhaps to clear the binding site from solvent. We identified a number of conserved hydrophobic residues that are not sequential. These residues are probably there to block the codon from becoming solvent exposed. Even one water molecule at the wrong place could potentially remove the natural discriminatory control within eRF1 and in-crease misreading of near-cognate codons.

39

Figure 12. Average non-bonded interaction energies between the last two codon ba-ses and their surroundings. Green, blue, cyan and pink bars denote the UAA, UGA, UAG and UGG codon, respectively. The inset box show the base specific interaction to Glu55.

In addition to the stop codon selection by eRF1, we support the structural ex-planation to discrimination against UGG, where Glu55 obstructs a double gua-nine pair. Additionally, we identified a cluster of hydrophobic residues sur-rounding the last two nucleotides preventing unwanted solvent molecules to access the binding site and reduce the accuracy. These newly identified im-portant residues for codon selection, and Arg68 providing stabilizing interac-tions of the NIKS loop, may help to broaden the view of how the eukaryotes evolved and ended up with one omnipotent RF rather than having multiple factors terminating one of the most important tasks within a living cell.

It is interesting to speculate as to why the stop codon conformation varies depending on which RF to bind. Considering that elongation codons are read in a similar fashion regardless of the biological system, may this only come from the shape of the codon reading domain of eRF1, and when does the con-formational change occur? Structures of eRF1 in both pre-accommodated and accommodated states show the codon being in the U-turn already in the pre-accommodated state (Shao et al., 2016). Since GTP hydrolysis is required for efficient termination and conformational change of eRF1, could the switch from a stretched and stacked conformation to U-turn conformation be a signal for GTP hydrolysis on the eRF3?

40

Ligand binding in purine riboswitches (Paper II) Riboswitches are small non-coding cis regulatory structures within the 5’ UTR of mRNAs. The riboswitch binds various ligands, such as ions or specific me-tabolites in order to regulate gene expression. Upon cognate ligand binding, the mRNA structure induces a conformational change in the expression plat-form, which gives the riboswitch its name. The conformational switch leads to an altered gene expression of that specific biological expression pathway (Winkler and Breaker, 2005).

With an increasing number of antimicrobial resistance in pathogenic bac-teria, discoveries of novel drugs are very much needed. As RNA research has progressed, so has the interest for drugs targeting RNA. Due to the regulatory properties of riboswitches, as an active part in gene regulation and binding of small molecules, has led to an interest in synthetic ligands to control and downregulate important genes, thus making the riboswitch a putative drug tar-get (Blount and Breaker, 2006; Deigan and Ferré-D'Amaré, 2011). Although it has proven difficult to identify synthetic riboswitch ligands using conven-tional screening methods based on fragment and affinity, or structure based design (Lünse et al., 2014; Matzner and Mayer, 2015; Mulhbacher et al., 2010), a recent study identified ribocil, a highly selective chemical modulator of the riboflavin riboswitch in E. coli after screening nearly 60,000 com-pounds (Howe et al., 2015).

In this project, we studied the purine riboswitch using both the linear inter-action energy (LIE) method and FEP. The first purine riboswitch was identi-fied as part of the xpt-pbuX operon in Bacillus subtilis coding for xanthine phosphoribosyltransferase and a xanthine specific permease. This riboswitch showed great affinity for guanine but discriminated against adenine (Mandal et al., 2003), and was therefore classified as a guanine riboswitch (GR). Later, the authors identified a similar riboswitch in the same bacteria, but in the yhlD gene coding for a purine efflux pump (Mandal and Breaker, 2004). This one bound adenine, but not guanine, hence an adenine riboswitch (AR). While the two riboswitches appeared to be very similar as to one residue. The interacting nucleotide in GR is C74, and the corresponding residue in AR is U74. These two residues make up the interaction surface of the GR and AR. The im-portance of residue 74 was made clear when a C74U (denoted as GRA) mu-tation abolished the guanine specificity in GR and instead showed a preference to adenine (Mandal and Breaker, 2004). These riboswitches, GR and GRA, have been under the scope of both crystallographers and subject to ligand binding studies (Gilbert et al., 2009; 2006). These binding data was used to evaluate the LIE method. The calculations included four and seven binding ligands for GR and GRA, respectively. Additionally, in the GR calculations adenine was used as a non-binder control, while guanine and hypoxanthine were used in GRA since their respective discrimination against guanine and adenine is more that 10,000-fold.

41

As it turned out, the purine riboswitches, especially GRA, have specific methods of discrimination. It showed a large electrostatic preorganization within the riboswitch binding site. As a consequence, the standard LIE param-eterization, b=0.43 (Åqvist and Hansson, 1996), which has successfully been used for affinity predictions of small molecules in protein complexes (Brands-dal et al., 2003) and RNA-RNA interactions (Almlöf et al., 2007), did return low correlation to the experimental results. Instead, we explicitly evaluated each ligand and its specific b-parameter, by calculating the energetic cost of charging the ligands both in water and in the riboswitch. That is, in discrete steps turning off the atomic charges, while preserving the interatomic interac-tions.

Whereas the calculated b-parameter obeyed the standard parameter in wa-ter, bwater ≈ 0.43, the calculations of the riboswitch bound ligands revealed a much higher b. Even larger than the strictly projected b=0.5 derived from the linear response approximation (LRA) method. Since the generalized LIE equation relies on a single b for both ligand bound to a complex and free in solution (bcomplex ≈ bwater), different b parameters thus make the standard equa-tion unjustified. Our calculations now gave bcomplex ≈ 0.63 for all binders while lower for non-binders, e.g., a calculated bcomplex of 0.53 for guanine binding the GRA riboswitch.

The higher value of b originates from a significant contribution to the av-erage electrostatic ligand-surrounding interactions from the off-term (here; the atomistic ligand charges turned off) in the LRA equation and thus need to be considered. Here, the biological meaning of a large b, is that it reflects an electrostatically pre-organized ligand binding site. Examining the purine ri-boswitch binding site, this might not come as a surprise. Since the Y74 residue interacts via Watson-Crick base-pairing and upon binding, the riboswitch vir-tually limits the internal degrees of freedom, the ligand is not allowed to adopt different conformations once bound.

In this project, we evaluated the standard LIE parameterization to predict ligand binding to the purine riboswitches GR and GRA. Our initial calcula-tions did not succeed as the standard LIE parameters could not reproduce the experimentally measured binding energies. However, it revealed a remarkable feature of a pre-organized binding site that gives the riboswitch its specificity. We further discovered that GR and GRA act with different mechanisms, where GR discriminated against adenine due to loss of hydrogen bonds and GRA requires a pre-organized discrimination to avoid G-U wobble interac-tions.

42

Computational Methods

This section will introduce and describe the computational methodologies used in Papers I-IV. Just as in any scientific field, each problem or question requires its own methods and approaches to investigate. The field of compu-tational chemistry was greatly noticed in 2013 when pioneering work by Mi-chael Levitt at Stanford University, Arieh Warshel at University of Southern California and Martin Karplus at Harvard University were awarded the Nobel Prize in Chemistry as they managed to connect the classical Newtonian phys-ics with quantum physics and the development of multiscale models for com-plex chemical systems.

Molecular mechanics Molecular mechanics is a way to describe and model a molecular system based on the laws of classical mechanics. Every atom is described as a sphere with a partial charge, often located at the center of the atom. The interacting forces between atoms are described by a potential energy function termed a force field. The force field is used to calculate the energy of a given system, both regarding the forces between atoms within a molecule, as well as the interac-tion between molecules. Accurate force fields are of fundamental importance for theoretical chemists and biologists to reproduce nature by computational means. The parameters used to describe a molecule, for instance, the force constant and bond length of a simple carbon-carbon bond, were initially de-rived from experiments (Lifson and Warshel, 1968) and new force fields can utilize high-order quantum calculations (Donchev et al., 2005). I will not dis-cuss too much on the details of force fields, but at least show a typical form of the force field equation (2). If one wishes to dig deeper into force fields and potential energy functions, I may suggest reading ‘Computer simulation of liquids’ by Allen and Tildesley or ‘Empirical Force Fields’ by MacKerell in Computational methods for protein structure prediction and modeling I: Basic characterizations.

43

𝑈;'& = 𝑘= 𝑏 − 𝑏@ 9

='ABC

+ 𝑘E 𝜃 − 𝜃@ 9

GAHIJC

+ 𝑘K 𝜓 − 𝜓@ 9

MN;O'P;JOC

+ 𝑘Q 1 + 𝑐𝑜𝑠 𝑛𝜙 − 𝛿BMXJBOGIC

2

+1

4𝜋𝜖@

𝑞M𝑞^𝑟M,^A'AP='ABJB

;GMOC(M,^)

+ 𝐴M,^𝑟M,^69

−𝐵M,^𝑟M,^bA'AP='ABJB

;GMOC(M,^)

Molecular dynamics Given a potential energy function, such as the one given in equation (2) above, one can calculate all possible thermodynamic properties. While static struc-tures usually mean a lot in the structural determination field, we would like to see how they are in motion, just like in nature. To obtain meaningful infor-mation and describe a system as close as possible to how the system behaves, one needs lots of different configurations. These system configurations could be compared to the case where one would take fast (usually 1-2 femtoseconds) snapshots of a dynamic biological system.

One common method to mimic the natural dynamic motion of a system is by using molecular dynamics (MD). By applying Newton’s second law, one can calculate the forces acting on each atom since the potential energy func-tion given by equation (3) is related to the force acting on every atom by:

𝑭M = − defg+

d𝒓i= 𝑚M𝒂 𝑡 M,i = 1, 2, 3, … , N (3)

So, in equation (3) the force F, on atom i, at each instant t, is calculated as the negative positional derivative of the potential energy function. If we can calculate the force on every atom at every instant, solving equation (3) will give us the energy evolution over time and result in a trajectory where we can follow the atoms in position and momentum.

Regardless of what one wants to simulate, molecular dynamics simulations of any system require a structure, usually obtained from experimental work such as X-ray crystallography, cryo-electron microscopy, nuclear magnetic resonance (NMR) or theoretic work such as homology modelling. These dif-ferent methods have one thing in common, they produce static models. If one

44

would employ multiple runs of MD on such a structure, the result, given the same starting condition, would always come out identical. To solve that, each repeated MD is initiated with randomized starting velocities from the Max-well-Boltzmann velocity distribution P(vi) at a given temperature.

𝑃 𝒗M = Ni9s(tu

𝑒P

wi𝒗ix

xytz (4)

A fundamental assumption in statistical thermodynamics, known as the er-godic principle, is that all configurations, of any given system, shall be equally sampled. This complete set of configurations is known as an ensemble. How-ever, most simulations will never sample high-energy configurations. In order to link computational experiments to bulk-like ones performed either in vitro or in vivo, the ergodic postulate comes in handy. It states that the time average of a system is equal to the ensemble average equation (5,6). Meaning that, if you over a long time sample many different configurations, the configuration sampled the most will likely represent the true configuration. Given this, it is possible to connect macroscopic experiments with the microscopic ones per-formed by the computer.

If one takes the average of infinitely many systems in different microscopic configurations, macroscopic properties can be formulated as the ensemble av-erage as described by equation (5,6)

𝐴 &MNJ = lim}→�

1𝜏

𝐴 𝑡 𝑑𝑡�

&�@

5

𝐴 &MNJ = 𝐴 JACJN=IJ 𝐴 𝒑, 𝒓 𝜌 𝒑, 𝒓 𝑑𝒑�𝑑𝒓�(6)

where the property of A is a function of momenta pi(t) and coordinates r(t) of N number of particles at a given time t. ρ(p,r) is the probability density used to calculate the probability to find a given configuration. In a canonical ensemble, the expected outcome of system A can be calculated using the Boltzmann distribution.

45

While the MD method simulates and generates configurations by solving Newton’s law of motion, another popular simulation method is Monte Carlo (MC) dynamics. The MC method instead generates new configurations ran-domly and then evaluates the change in potential energy. If a new configura-tion has a negative change (lower energy than the previous one) the new con-figuration is accepted. If, however, a positive change occurs, the new config-uration is only accepted if the probability of that configuration is within the Boltzmann distribution.

The papers included in this thesis all employed the molecular dynamics method implemented in the Q software (Marelius et al., 1998b).

Free-energy perturbation Molecular dynamics can be used to generate large numbers of configurations of a system, and accordingly describe the large conformational changes of a protein. The majority of the papers presented and discussed above have had another scope of interest. Namely, RNA or protein binding events on the ri-bosome, or small ligands binding to a riboswitch. On a relative timescale, binding of a molecule to its target can seem like a fast process, it is, however, usually not easily achieved computationally, although not impossible (Shan et al., 2011). Instead, relative free-energy differences can be calculated between states, e.g. two ligands or nucleotides. These calculated relative free-energies can be compared to experimentally measured binding constants via equation (7).

∆𝐺 = ∆𝐺@ − 𝑅𝑇𝑙𝑛 ,���

(7)

Where Kd is the dissociation constant, c0 is the standard state concentration, R is the gas constant, T is the absolute temperature and, ∆G0 the Gibbs standard free-energy, meaning the free-energy of a reaction or binding event where 1M of A+B can form AB in equilibrium. With time, the concentrations of free A+B and complex AB in a reversible binding event will reach equilibrium and ∆G=0 and equation (7) can be written as:

∆𝐺@ = 𝑅𝑇𝑙𝑛 ,���

(8)

This relates the dissociation constant directly to the ∆G0 which can be used theoretically to predict the affinity difference between two ligands or mole-cules.

In statistical thermodynamics, the Helmholtz free-energy, A, for a closed system is defined using the canonical ensemble partition function Q where the volume, V, temperature T, and number of particles (N) remain constant.

46

𝐴 = −𝑘�𝑇𝑙𝑛𝑄 𝑁, 𝑉, 𝑇 𝑄 𝑁, 𝑉, 𝑇 = 𝑒P�i(tu

M

(9)

In equation (9) Q is the sum of all quantum states i, with energy Ei representing all degrees of freedom of the system, including contributions from transla-tions, rotations and vibrations. Gibb’s free-energy is often referred to in bio-chemical experiments, as stated in equation (8). Such experiments are per-formed at constant temperature and pressure, justifying the use of the Gibbs free energy instead of the Helmholtz free-energy which is defined for systems at constant temperature and volume. Gibb’s free-energy can be expressed in a number of different ways (McQuarrie and Simon, 1999), such as

𝐺 = 𝐻 − 𝑇𝑆 10

where H = U + PV is the enthalpy, T is the temperature and S is the entropy. Further Gibb’s free-energy can be related to Helmholtz free-energy, A, using the relation

𝐺 = 𝐴 + 𝑃𝑉 11

where A = U – TS. If one differentiates equation (11) while keeping the pres-sure and volume constant, which is the case for biological systems, the PV term can be neglected, and the free-energy difference, ∆G, between two states X and Y, (e.g. a release factor binding the stop codon UAA and CAA) can be obtained with the free-energy perturbation (FEP) formula commonly known as Zwanzig’s formula (Zwanzig, 1954)

∆𝐺�→� = 𝐺� − 𝐺� = −𝑘�𝑇𝑙𝑛 𝑒P∆�fg+ytz � 12

where;

∆G < 0; favourable reaction ∆G = 0; equilibrium ∆G > 0; disfavourable reaction

Here ∆Upot is the potential energy difference between state X and Y calculated on the configuration and potential represented by state X.

Since the energy difference is calculated between two states based on the initial state X, the limitation of equation (12) is that it requires a significant overlap in Upot. While MD is used to generate different configurations, FEP is used to drive the system between two states with different potential energies.

47

To get meaningful and converged sampling, the transformation of X to Y is done by introducing several intermediate states creating a mapping parameter l

𝑈M = 𝜆6M 𝑈� + 𝜆9M 𝑈�𝑎𝑛𝑑𝜆6 + 𝜆9 = 1 13

With the mapping parameter li one can create a linear combination of the two states, where l=(1,0) in the initial state and l=(0,1) in the product state using i number of steps (51 steps are used in the papers presented in this thesis). This way of dividing the transformation thus creates i-1 unphysical states. Since the free-energy is a thermodynamic state function, it is not path dependent. As we now have divided the FEP into i number of steps, equation (12) must be re-written to account for all intermediate steps

∆𝐺�→� = 𝐺� − 𝐺� = −𝑘�𝑇 𝑙𝑛 𝑒P(ei¡¢Pei(tu M

AP6

M�@

14

The above equation can be used to calculate the free-energy difference of a ligand at its binding site. But since molecules do not just appear in a binding site, we must also calculate the free-energy difference in water (or without the receptor) to get a relative free-energy between two molecules. This is achieved using a thermodynamic cycle (Figure 13).

Figure 13. Representing a thermodynamic cycle where the release factor binds to ei-ther a UAA stop codon or a CAA near-cognate codon.

48

While in a test tube, one can measure the binding of molecules X and Y (the horizontal arrows in Figure 13), we can calculate the free-energy difference of molecules being either in bulk solvent (the left vertical arrow) or at its bind-ing site (the right vertical arrow). This allow us to calculate the relative free-energy difference, ∆∆Gbind, of RF binding to either UAA or CAA, as the ex-ample above, via equation 15

∆𝐺='£ABN£&G&M'A − ∆𝐺£A='£ABN£&G&M'A = ∆∆𝐺=MAB = ∆𝐺=MABN£&G&M'A − ∆𝐺=MAB¤& (15)

Linear interaction energy While FEP is generally limited to small changes of potential energy, there are other semi-empirical methods for determining absolute binding free-energies, such as the linear response approximation (LRA), molecular mechanics-Pois-son-Boltzmann surface area (MM-PBSA), and the linear interaction energy (LIE) method. The LIE method was used in Paper II in predicting the binding of small ligands to the purine riboswitch. In LIE, one only simulates the end-states instead of sampling unphysical intermediates. Two separate MD simu-lations are used where one collects the ligand interaction energies to its sur-rounding, allowing the free-energy to be predicted as:

∆𝐺=MAB = 𝛼 𝑈IPC¦B¤

= − 𝑈IPC¦B¤§ + 𝛽 𝑈IPCJI

= − 𝑈IPCJI§ + 𝛾 (16)

Here, one collects the ligand-surrounding (l-s), van der Waals (vdw), and elec-trostatic (el) interactions in the unbound (f) and bound (b) state. a and b in equation (16) are theoretically and/or empirically derived scaling factors for the non-polar and polar interactions of the free-energy of binding, respec-tively. While a has been empirically estimated to be 0.18 (Åqvist et al., 1994), b was assigned 0.5, initially based on the LRA method (Lee et al., 1992). A revised model later divided b based on the property of the ligand, e.g. for di-polar compounds with two or more hydroxyl groups b=0.33 (Marelius et al., 1998a). In standard LIE, b is assumed to be equal in both bound and free states giving equation (16), but should explicitly be written as:

∆𝐺=MAB = 𝛼 𝑈IPC¦B¤= − 𝑈IPC¦B¤

§ + 𝛽= 𝑈IPCJI= − 𝛽§ 𝑈IPCJI

§ + 𝛾 (17) where the sub-indexes b and f denote bound and free ligand, respectively. As described in Paper II, this is the case for ligand binding in the purine ri-boswitches where bb is significantly larger than bf (Sund et al., 2015). The g parameter is usually necessary for calculations of absolute binding free ener-gies, but not essential for calculations of relative binding affinities as the term would cancel out as one computes the ∆∆Gbind. It has been shown that the g

49

parameter correlate with the hydrophobicity of the binding site (Hansson et al., 1998; Mishra et al., 2012).

Homology modeling The only thing one basically needs to run an MD simulations or a free-energy calculation, is an initial starting structure. The structures used in MD, FEP or LIE calculations are usually obtained through experimental methods such as X-ray crystallography, Cryo-EM or NMR. Even though there has been a rapid development in the structural field over the years, and new structures are avail-able every day (Berman et al., 2013), far from all structures are solved. Luck-ily, even though small variations in sequence exists between two related pro-teins, their structural fold is usually more conserved than the sequence (Cho-thia and Lesk, 1986). Homology modeling was used in Paper I to create a structural model of the mitochondrial release factors, since no structures were available. As of today, 1st of September 2017, these structures are yet to be determined.

In homology modeling, one requires at least one known structure, which is used as a template to create a three-dimensional structure of the target mole-cule. The starting point in homology modeling is alignment of the sequences. The alignment is useful to identify conserved regions which helps to optimize the model. Generally, a-helices and b-sheets are conserved between homo-logues proteins, while stretches of loops can be more dynamic and thus more complex to model.

50

Sammanfattning på svenska

Den här avhandlingen sammanfattar hur man med hjälp av datorsimuleringar och fria-energiberäkningar kan koppla samman biokemiska experiment och experimentellt bestämda kristallstrukturer.

Avhandlingen behandlar en av de mest fundamentala processerna i våra celler, nämligen proteinsyntesen. Dessa processer är livsviktiga för alla le-vande celler och här beskrivs dem på en atomistisk nivå, där en inkorrekt vä-tebindning kan ge betydande konsekvenser. Ribosomen är enorm makro-mo-lekylär maskin och brukar kallas för cellens proteinfabrik, där gener översätts till proteiner. Proteinerna gör i princip allt i cellen, från att, till exempel, känna igen och ta upp koffein när vi dricker kaffe, till att kopiera vårt DNA.

Varje protein har sin egen unika sekvens. Gener blir kopierade från DNA med hjälp av ett protein, RNA polymeras, som skapar ett mRNA. mRNA:t binder sedan till ribosomen där det översätts till proteiner. I människor och andra eukaryota organismer måste ribosomen leta sig fram till den korrekta positionen på mRNA:t för att starta proteinsyntesen. Detta görs genom att skanna mRNA:t. I bakterier däremot så finns det ett inbyggt bindningsställe som positionerar mRNA:t korrekt. Ribosomen översätter sedan den sekvens av nukleotider som mRNA:t består av. Detta sker med hjälp av tRNA mole-kyler. tRNA:t binder till tre nukleotider på mRNA:t som bildar ett kodord, kodon, som motsvarar en specifik aminosyra. tRNA har två ändar; en som binder till mRNA:t (den avkodande delen) och en som är laddad med den ami-nosyra som mRNA:t kodar för. Ribosomen kommer tillslut till ett kodon som inte kodar för någon aminosyra, detta betyder stopp. Istället för att ett tRNA binder till ribosomen, så binder då ett protein kallat termineringsfaktor. Dess funktion är att avsluta proteinsyntesen genom att frigöra den aminosyrakedja som bildats på ribosomen.

Arbetet som presenteras i den här avhandlingen redogör för den första och den sista delen av proteinsyntesen, och hur både nukleinsyror och proteiner känner igen rätt från fel.

I papper III undersöktes noggrannheten hos det första tRNA som binder och startar proteinsyntesen i eukaryoter. Detta tRNA binder till, och scannar längs med, mRNA:t för att hitta start kodonet. När tRNA:t scannar mRNA:t kommer det att känna av många kodon som är väldigt likt start kodonet. Vår hypotes var att denna process måste vara väldigt exakt för att proteinsyntesen ska påbörjas på korrekt kodon. Genom fria-energiberäkningar testade vi skill-naden i bindnings fri-energi mellan det universella start kodonet, AUG, och

51

sex andra kodon. Beräkningarna visade en stor noggrannhet för AUG relativt de andra kodonen. Dessutom visade våra beräkningar att noggrannheten inte endast låg hos interaktionerna mellan start kodonet och tRNA:t, utan en stor del av noggrannheten kom från två proteiner.

I nästa del studerades hur proteinsyntesen avslutas. Efter att ribosomen har läst klart ett mRNA, stöter den på en stoppsignal i form av ett kodon som inte kodar för något naturligt tRNA. Dessa kodon är UAA, UAG och UGA, och kallas stopp kodon. I papper I studerades termineringsprocessen i mitokondrier. Mitokondrien är en organell i eukaryota celler och har en viktig funktion i att producera adesintrifosfat (ATP). ATP är en energirik molekyl som används i de flesta biokemiska processer i cellen. När det mänskliga mitokondrie genomet sekvenserades fann man variation från de vanliga stopp kodonen. CO1 och ND6 är två proteiner viktiga i ATP produktionen. Det spe-ciella med dessa proteiner är att dess stopp kodon inte är något av de tre van-liga stoppkodonen. Istället avslutas de humana CO1 och ND6 proteinerna med kodonen AGA och AGG. I mitokondrier finns det inga tRNA som binder dessa kodon. Vi fann även att detta var fallet i andra genom tillhörande rygg-radsdjur, fast inte alltid vid samma proteiner.

Mitokondrien har sitt eget DNA, men innehåller inte alla de proteiner som den behöver. Den större delen av proteinerna kodas för i cellens kärna och importeras sedan till mitokondrien. Det här gäller även termineringsfak-torerna. Tidigare har två proteiner identifierats som mitokondrie-termine-ringsfaktorer (mtRF1 och mtRF1a), baserat på proteinsekvenslikheter (homo-logi) till termineringsfaktorerna i bakterier. Dessvärre, finns det inga kristall-strukturer av dessa proteiner, så vi genererade homologimodeller som sedan användes för beräkningar av fria-energier. Våra beräkningar visade att de fö-reslagna mitokondrie-termineringsfaktorerna inte kan läsa AGA och AGG kodonen. Istället föreslog vi en alternativ teori, där ett annat protein, ITC-1, som också är homologt till de bakteriella termineringsfaktorerna, skulle kunna hoppa in och avsluta proteintillverkningen när ett mRNA har AGA eller AGG som stoppkodon. Denna teori fick senare stöd av biokemiska data.

Papper IV behandlar en annan termineringsfaktor, den eukaryota termine-ringsfaktorn eRF1. Till skillnad mot de bakteriella termineringsfaktorerna, RF1 och RF2, som binder till kodonen UAA, UAG respektive UAA, UGA, kan eRF1 läsa alla tre stoppkodonen. Forskare har länge undersökt hur eRF1 kan binda till både UGA och UAG, vilket betyder att termineringsfaktorn kan binda till en guanin nukleotid i både andre och tredje positionen av stoppkodo-net, men den måste undvika att läsa båda samtidigt.

Nyligen presenterades kristallstrukturer av eRF1 bunden till den eukaryota ribosomen. Teorin om hur eRF1 undviker att läsa UGG kodonet visade sig, strukturellt, bero på positionen av en aminosyra, glutaminsyra-55. För att yt-terligare testa denna teori, har vi beräknat fria-energiskillnader mellan, både olika stoppkodon, samt förändringar i första positionen och för UGG. Våra

52

beräkningar visade att eRF1 har en tydlig noggrannhet för de tre stoppkodo-nen. Vi kunde även fastslå teorin om funktionen angående glutaminsyra-55. Utöver glutaminsyra-55, föreslog vi även att cystein-127 är en viktig kompo-nent i bindning till både UAG och UGA, samt att stoppkodonet är omgärdat av en hydrofob ficka av konserverade aminosyror. Fickan har tidigare inte identifierats då den inte är del av en oavbruten aminosyrasekvens, utan utgör ett strukturellt motiv.

Den sista delen handlar om hur uttryck av gener kan regleras genom bind-ning av småmolekyler till en del av ett mRNA, en så kallad riboswitch. I detta fall studerades en riboswitch med specificitet att binda purin-molekyler, till exempel adenin och guanin. Här användes, utöver FEP, även LIE metoden för att beräkna bindnings-fria-energier för olika molekyler. Här visade det sig att riboswitcharnas bindningssätt var specifikt för respektive riboswitch. Detta specifika bindningssätt lyckades inte återspegla de korrekta energierna med de standardiserade LIE-parametrarna. Istället beräknades nya parametrar ex-plicit via FEP som visade att purin riboswitcharna har ett pre-organiserat sätt att binda småmolekylerna, vilket i sin tur, ger upphov till en specifik nog-grannhet att binda rätt molekyl.

53

Acknowledgements

To everyone that have contributed to this thesis, thank you! I would also like to mention: Johan Åqvist, my supervisor. Thank you for giving me the opportunity to do my doctoral studies in your group, for continuously helping me to improve as a scientist and for teaching me the importance of ‘Blow your horn!’ Suparna Sanyal, my co-supervisor, thank you for always being honest, taking care of me at conferences, and always answer my questions with a smile. All my friends and colleagues at ICM for the past years, and specifically the Åqvist group: Ana, Geir, Henrik, Hugo, Jaka, Jessica, Johan S, Lars, Masoud, Mauricio, Silvana, Sudarsan, Willem, Yao & Yasmin, thank you for all the good times from boat rides, skiing, running, potlucks, musicals, julbord, build-ing walls at conferences, the salame and most importantly, you have made BMC a fun place to work at. Thank you!

A big and special thank you to Johan Sund, who took me in as a project student and helped getting me started when my studies began. I remember one time as a project student, he told me to make some input files for a FEP calculation and showed me an example, and then left on vacation. I spent over two weeks on figuring out how to make it and it taught me hove to survive in the office on my own. To my parents, thank you for always supporting me no matter what I do. Sofie, Schatje, words are not enough to describe how grateful I am that you said ‘Hi’ that one time. Thanks to all my friends outside of BMC and Sweden for great company and good times.

54

References

Akabane, S., Ueda, T., Nierhaus, K.H., Takeuchi, N., 2014. Ribosome rescue and translation termination at non-standard stop codons by ict1 in mammalian mito-chondria. PLoS Genet. 10, e1004616. doi:10.1371/journal.pgen.1004616

Alkalaeva, E.Z., Pisarev, A.V., Frolova, L.Y., Kisselev, L.L., Pestova, T.V., 2006. In vitro reconstitution of eukaryotic translation reveals cooperativity between re-lease factors eRF1 and eRF3. Cell 125, 1125–1136. doi:10.1016/j.cell.2006.04.035

Allnér, O., Nilsson, L., 2011. Nucleotide modifications and tRNA anticodon–mRNA codon interactions on the ribosome. RNA 17, 2177–2188. doi:10.1261/rna.029231.111

Almlöf, M., Andér, M., Åqvist, J., 2007. Energetics of codon−anticodon recognition on the small ribosomal subunit. Biochemistry 46, 200–209. doi:10.1021/bi061713i

Anderson, S., Bankier, A.T., Barrell, B.G., de Bruijn, M.H., Coulson, A.R., Drouin, J., Eperon, I.C., Nierlich, D.P., Roe, B.A., Sanger, F., Schreier, P.H., Smith, A.J., Staden, R., Young, I.G., 1981. Sequence and organization of the human mitochondrial genome. Nature 290, 457–465.

Antonicka, H., Østergaard, E., Sasarman, F., Weraarpachai, W., Wibrand, F., Peder-sen, A.M.B., Rodenburg, R.J., van der Knaap, M.S., Smeitink, J.A.M., Chrzanowska-Lightowlers, Z.M., Shoubridge, E.A., 2010. Mutations in C12orf65 in patients with encephalomyopathy and a mitochondrial translation defect. Am. J. Hum. Genet. 87, 115–122. doi:10.1016/j.ajhg.2010.06.004

Antoun, A., Pavlov, M.Y., Lovmar, M., Ehrenberg, M., 2006a. How initiation fac-tors maximize the accuracy of tRNA selection in initiation of bacterial protein synthesis. Mol. Cell 23, 183–193. doi:10.1016/j.molcel.2006.05.030

Antoun, A., Pavlov, M.Y., Lovmar, M., Ehrenberg, M., 2006b. How initiation fac-tors tune the rate of initiation of protein synthesis in bacteria. EMBO J. 25, 2539–2550. doi:10.1038/sj.emboj.7601140

Asano, K., Clayton, J., Shalev, A., Hinnebusch, A.G., 2000. A multifactor complex of eukaryotic initiation factors, eIF1, eIF2, eIF3, eIF5, and initiator tRNA(Met) is an important translation initiation intermediate in vivo. Genes Dev. 14, 2534–2546. doi:10.1101/gad

Astbury, W.T., Bell, F.O., 1938. Some recent developments in the x-ray study of proteins and related structures. Cold Spring Harbor Symposia on Quantitative Biology 6, 109–121. doi:10.1101/sqb.1938.006.01.013

Atkinson, G.C., Baldauf, S.L., Hauryliuk, V., 2008. Evolution of nonstop, no-go and nonsense-mediated mRNA decay and their termination factor-derived compo-nents. BMC Evolutionary Biology 2008 8:1 8, 290. doi:10.1186/1471-2148-8-290

Avery, O.T., MacLeod, C.M., McCarty, M., 1944. Studies on the chemical nature of the substance inducing transformation of Pneumococcal types. J. Exp. Med. 79, 137–158. doi:10.1084/jem.79.2.137

55

Åqvist, J., Hansson, T., 1996. On the validity of electrostatic linear response in polar solvents. J. Phys. Chem. 100, 9512–9521. doi:10.1021/jp953640a

Åqvist, J., Lind, C., Sund, J., Wallin, G., 2012. Bridging the gap between ribosome structure and biochemistry by mechanistic computations. Curr. Opin. Struct. Biol. 22, 815–823. doi:10.1016/j.sbi.2012.07.008

Åqvist, J., Medina, C., Samuelsson, J.-E., 1994. A new method for predicting bind-ing affinity in computer-aided drug design. Protein Eng. 7, 385–391. doi:10.1093/protein/7.3.385

Ban, N., Nissen, P., Hansen, J., Moore, P.B., Steitz, T.A., 2000. The complete atomic structure of the large ribosomal subunit at 2.4 Å resolution. Science 289, 905–920. doi:10.1126/science.289.5481.905

Berman, H.M., Coimbatore Narayanan, B., Di Costanzo, L., Dutta, S., Ghosh, S., Hudson, B.P., Lawson, C.L., Peisach, E., Prlić, A., Rose, P.W., Shao, C., Yang, H., Young, J., Zardecki, C., 2013. Trendspotting in the Protein Data Bank. FEBS Lett. 587, 1036–1045. doi:10.1016/j.febslet.2012.12.029

Bertram, G., Bell, H.A., Ritchie, D.W., Fullerton, G., Stansfield, I., 2000. Terminat-ing eukaryote translation: domain 1 of release factor eRF1 functions in stop co-don recognition. RNA 6, 1236–1247.

Blanchet, S., Rowe, M., Haar, von der, T., Fabret, C., Demais, S., Howard, M.J., Namy, O., 2015. New insights into stop codon recognition by eRF1. Nucleic Acids Res. 43, 3298–3308. doi:10.1093/nar/gkv154

Blattner, F.R., Plunkett, G., Bloch, C.A., Perna, N.T., Burland, V., Riley, M., Col-lado-Vides, J., Glasner, J.D., Rode, C.K., Mayhew, G.F., Gregor, J., Davis, N.W., Kirkpatrick, H.A., Goeden, M.A., Rose, D.J., Mau, B., Shao, Y., 1997. The complete genome sequence of Escherichia coli K-12. Science 277, 1453–1462. doi:10.1126/science.277.5331.1453

Blount, K.F., Breaker, R.R., 2006. Riboswitches as antibacterial drug targets. Nat. Biotechnol. 24, 1558–1564. doi:10.1038/nbt1268

Brandsdal, B.O., Osterberg, F., Almlöf, M., Feierberg, I., Luzhkov, V.B., Åqvist, J., 2003. Free energy calculations and ligand binding. Adv Protein Chem 66, 123–158.

Breaker, R.R., 2012. Riboswitches and the RNA world. Cold Spring Harb Perspect Biol 4, a003566. doi:10.1101/cshperspect.a003566

Brenner, S., Barnett, L., Katz, E.R., Crick, F.H.C., 1967. UGA: A third nonsense tri-plet in the genetic code. Nature 213, 449–450. doi:10.1038/213449a0

Brenner, S., Jacob, F., Meselson, M., 1961. An unstable intermediate carrying infor-mation from genes to ribosomes for protein synthesis. Nature 190, 576–581.

Brenner, S., Stretton, A.O.W., Kaplan, S., 1965. Genetic Code: The “nonsense” tri-plets for chain termination and their suppression. Nature 206, 994–998. doi:10.1038/206994a0

Brown, A., Shao, S., Murray, J., Hegde, R.S., Ramakrishnan, V., 2015. Structural basis for stop codon recognition in eukaryotes. Nature 524, 493–496. doi:10.1038/nature14896

Buckingham, R.H., Grentzmann, G., Kisselev, L., 1997. Polypeptide chain release factors. Mol. Microbiol. 24, 449–456. doi:10.1046/j.1365-2958.1997.3711734.x

Bult, C.J., White, O., Olsen, G.J., Zhou, L., Fleischmann, R.D., Sutton, G.G., Blake, J.A., FitzGerald, L.M., Clayton, R.A., Gocayne, J.D., Kerlavage, A.R., Dougherty, B.A., Tomb, J.F., Adams, M.D., Reich, C.I., Overbeek, R., Kirk-ness, E.F., Weinstock, K.G., Merrick, J.M., Glodek, A., Scott, J.L., Geoghagen, N.S., Venter, J.C., 1996. Complete genome sequence of the methanogenic ar-chaeon, methanococcus jannaschii. Science 273, 1058–1073.

56

Bulygin, K.N., Khairulina, Y.S., Kolosov, P.M., Ven’yaminova, A.G., Graifer, D.M., Vorobjev, Y.N., Frolova, L.Y., Karpova, G.G., 2011. Adenine and gua-nine recognition of stop codon is mediated by different N domain conformations of translation termination factor eRF1. Nucleic Acids Res. 39, 7134–7146. doi:10.1093/nar/gkr376

Capecchi, M.R., 1967. Polypeptide chain termination in vitro: isolation of a release factor. Proc. Natl. Acad. Sci. U.S.A. 58, 1144–1151.

Caskey, C.T., Beaudet, A.L., Scolnick, E.M., Rosman, M., 1971. Hydrolysis of fMet-tRNA by peptidyl transferase. Proc. Natl. Acad. Sci. U.S.A. 68, 3163–3167.

Caspersson, T., 1947. The relations between nucleic acid and protein synthesis. Symp. Soc. Exp. Biol. 1, 127–151.

Caspersson, T., Schultz, J., 1939. Pentose nucleotides in the cytoplasm of growing tissues. Nature 143, 602–603. doi:10.1038/143602c0

Chargaff, E., Zamenhof, S., Green, C., 1950. Composition of human desoxypentose nucleic acid. Nature 165, 756–757.

Chavatte, L., Seit-Nebi, A., Dubovaya, V., Favre, A., 2002. The invariant uridine of stop codons contacts the conserved NIKSR loop of human eRF1 in the ribo-some. EMBO J. 21, 5302–5311.

Chothia, C., Lesk, A.M., 1986. The relation between the divergence of sequence and structure in proteins. EMBO J. 5, 823–826.

Conard, S.E., Buckley, J., Dang, M., Bedwell, G.J., Carter, R.L., Khass, M., Bed-well, D.M., 2012. Identification of eRF1 residues that play critical and comple-mentary roles in stop codon recognition. RNA 18, 1210–1221. doi:10.1261/rna.031997.111

Craigen, W.J., Caskey, C.T., 1987. The function, structure and regulation of E. coli peptide chain release factors. Biochimie 69, 1031–1041. doi:10.1016/0300-9084(87)90003-4

Crick, F., 1958. On protein synthesis. Symp. Soc. Exp. Biol. 138–163. Crick, F.H.C., Barnett, L., Brenner, S., Watts-Tobin, R.J., 1961. General nature of

the genetic code for proteins. Nature 192, 1227–1232. doi:10.1038/1921227a0 Cromie, M.J., Shi, Y., Latifi, T., Groisman, E.A., 2006. An RNA sensor for intracel-

lular Mg2. Cell 125, 71–84. doi:10.1016/j.cell.2006.01.043 D'Ascenzo, L., Leonarski, F., Vicens, Q., Auffinger, P., 2017. Revisiting GNRA and

UNCG folds: U-turns versus Z-turns in RNA hairpin loops. RNA 23, 259–269. doi:10.1261/rna.059097.116

Dahlquist, K.D., Puglisi, J.D., 2000. Interaction of translation initiation factor IF1 with the E. coli ribosomal A site. J. Mol. Biol. 299, 1–15. doi:10.1006/jmbi.2000.3672

Deigan, K.E., Ferré-D'Amaré, A.R., 2011. Riboswitches: Discovery of drugs that target bacterial gene-regulatory rnas. Acc. Chem. Res. doi:10.1021/ar200039b

Demeshkina, N., Jenner, L., Westhof, E., Yusupov, M., Yusupova, G., 2012. A new understanding of the decoding principle on the ribosome. Nature 484, 256–259. doi:10.1038/nature10913

Donchev, A.G., Ozrin, V.D., Subbotin, M.V., Tarasov, O.V., Tarasov, V.I., 2005. A quantum mechanical polarizable force field for biomolecular interactions. Proc. Natl. Acad. Sci. U.S.A. 102, 7829–7834. doi:10.1073/pnas.0502962102

Dounce, A.L., 1953. Nucleic acid template hypotheses. Nature 172, 541–542. doi:10.1038/172541b0

Duarte, I., Nabuurs, S.B., Magno, R., Huynen, M., 2012. Evolution and diversifica-tion of the organellar release factor family. Mol. Biol. Evol. 29, 3497–3512. doi:10.1093/molbev/mss157

57

Ehrenberg, M., Tenson, T., 2002. A new beginning of the end of translation. Nat Struct Biol 9, 85–87. doi:10.1038/nsb0202-85

Fourmy, D., Recht, M.I., Blanchard, S.C., Puglisi, J.D., 1996. Structure of the A site of Escherichia coli 16S ribosomal RNA complexed with an aminoglycoside an-tibiotic. Science 274, 1367–1371.

Franklin, R.E., Gosling, R.G., 1953a. Molecular configuration in sodium thymonu-cleate. Nature 171, 740–741.

Franklin, R.E., Gosling, R.G., 1953b. Evidence for 2-chain helix in crystalline struc-ture of sodium deoxyribonucleate. Nature 172, 156–157.

Freistroffer, D.V., Kwiatkowski, M., Buckingham, R.H., Ehrenberg, M., 2000. The accuracy of codon recognition by polypeptide release factors. Proc. Natl. Acad. Sci. U.S.A. 97, 2046–2051. doi:10.1073/pnas.030541097

Frolova, L., Seit-Nebi, A., Kisselev, L., 2002. Highly conserved NIKS tetrapeptide is functionally essential in eukaryotic translation termination factor eRF1. RNA 8, 129–136. doi:10.1017/s1355838202013262

Frolova, L.Y., Tsivkovskii, R.Y., Sivolobova, G.F., Oparina, N.Y., Serpinsky, O.I., Blinov, V.M., Tatkov, S.I., Kisselev, L.L., 1999. Mutations in the highly con-served GGQ motif of class 1 polypeptide release factors abolish ability of hu-man eRF1 to trigger peptidyl-tRNA hydrolysis. RNA 5, 1014–1020.

Furberg, S., 1952. On the structure of nucleic acids. Acta Chemica Scandinavica 6, 634–640.

Gamow, G., 1954. Possible relation between deoxyribonucleic acid and protein structures. Nature 173, 318–318. doi:10.1038/173318a0

Gilbert, S.D., Reyes, F.E., Edwards, A.L., Batey, R.T., 2009. Adaptive ligand bind-ing by the purine riboswitch in the recognition of guanine and adenine Analogs. Structure 17, 857–868. doi:10.1016/j.str.2009.04.009

Gilbert, S.D., Stoddard, C.D., Wise, S.J., Batey, R.T., 2006. Thermodynamic and ki-netic characterization of ligand binding to the purine riboswitch aptamer do-main. J. Mol. Biol. 359, 754–768. doi:10.1016/j.jmb.2006.04.003

Gold, L., Polisky, B., Uhlenbeck, O., Yarus, M., 1995. Diversity of oligonucleotide functions. Annu. Rev. Biochem. 64, 763–797. doi:10.1146/an-nurev.bi.64.070195.003555

Greber, B.J., Boehringer, D., Leitner, A., Bieri, P., Voigts-Hoffmann, F., Erzberger, J.P., Leibundgut, M., Aebersold, R., Ban, N., 2014. Architecture of the large subunit of the mammalian mitochondrial ribosome. Nature 505, 515–519. doi:10.1038/nature12890

Gromadski, K.B., Daviter, T., Rodnina, M.V., 2006. A uniform response to mis-matches in codon-anticodon complexes ensures ribosomal fidelity. Mol. Cell 21, 369–377. doi:10.1016/j.molcel.2005.12.018

Gromadski, K.B., Rodnina, M.V., 2004. Kinetic determinants of high-fidelity tRNA discrimination on the ribosome. Mol. Cell 13, 191–200.

Gros, F., Hiatt, H., Gilbert, W., Kurland, C.G., Risebrough, R.W., Watson, J.D., 1961. Unstable ribonucleic acid revealed by pulse labelling of escherichia coli. Nature 190, 581–585. doi:10.1038/190581a0

Gutell, R.R., Cannone, J.J., Konings, D., Gautheret, D., 2000. Predicting U-turns in ribosomal RNA with comparative sequence analysis. J. Mol. Biol. 300, 791–803. doi:10.1006/jmbi.2000.3900

Handa, Y., Inaho, N., Nameki, N., 2011. YaeJ is a novel ribosome-associated pro-tein in Escherichia coli that can hydrolyze peptidyl-tRNA on stalled ribosomes. Nucleic Acids Res. 39, 1739–1748. doi:10.1093/nar/gkq1097

58

Hansen, J.L., Schmeing, T.M., Moore, P.B., Steitz, T.A., 2002. Structural insights into peptide bond formation. Proc. Natl. Acad. Sci. U.S.A. 99, 11670–11675. doi:10.1073/pnas.172404099

Hansson, T., Marelius, J., Åqvist, J., 1998. Ligand binding affinity prediction by lin-ear interaction energy methods. J. Comp. Aid. Mol. Des. 12, 27–35. doi:10.1023/A:1007930623000

Henkin, T.M., 2008. Riboswitch RNAs: using RNA to sense cellular metabolism. Genes Dev. 22, 3383–3390. doi:10.1101/gad.1747308

Hermann, T., Patel, D.J., 2000. Adaptive recognition by nucleic acid aptamers. Sci-ence 287, 820–825. doi:10.1126/science.287.5454.820

Hinnebusch, A.G., 2014. The scanning mechanism of eukaryotic translation initia-tion. Annu. Rev. Biochem. 83, 779–812. doi:10.1146/annurev-biochem-060713-035802

Hoagland, M.B., Stephenson, M.L., Scott, J.F., Hecht, L.I., Zamecnik, P.C., 1958. A soluble ribonucleic acid intermediate in protein synthesis. J. Biol. Chem. 231, 241–257.

Hoagland, M.B., Zamecnik, P.C., Stephenson, M.L., 1957. Intermediate reactions in protein biosynthesis. Biochim. Biophys. Acta 24, 215–216. doi:10.1016/0006-3002(57)90175-0

Howe, J.A., Wang, H., Fischmann, T.O., Balibar, C.J., Xiao, L., Galgoci, A.M., Ma-linverni, J.C., Mayhood, T., Villafania, A., Nahvi, A., Murgolo, N., Barbieri, C.M., Mann, P.A., Carr, D., Xia, E., Zuck, P., Riley, D., Painter, R.E., Walker, S.S., Sherborne, B., de Jesus, R., Pan, W., Plotkin, M.A., Wu, J., Rindgen, D., Cummings, J., Garlisi, C.G., Zhang, R., Sheth, P.R., Gill, C.J., Tang, H., Roe-mer, T., 2015. Selective small-molecule inhibition of an RNA structural ele-ment. Nature 526, 672–677. doi:10.1038/nature15542

Hussain, T., Llácer, J.L., Wimberly, B.T., Kieft, J.S., Ramakrishnan, V., 2016. Large-scale movements of IF3 and tRNA during bacterial translation initiation. Cell 167, 133–144.e13. doi:10.1016/j.cell.2016.08.074

Huynen, M.A., Duarte, I., Chrzanowska Lightowlers, Z.M.A., Nabuurs, S.B., 2012. Structure based hypothesis of a mitochondrial ribosome rescue mechanism. Biol. Direct 7, 14. doi:10.1186/1745-6150-7-14

Ieong, K.-W., Uzun, Ü., Selmer, M., Ehrenberg, M., 2016. Two proofreading steps amplify the accuracy of genetic code translation. Proc. Natl. Acad. Sci. U.S.A. 113, 13744–13749. doi:10.1073/pnas.1610917113

Ito, K., Ebihara, K., Uno, M., Nakamura, Y., 1996. Conserved motifs in prokaryotic and eukaryotic polypeptide release factors: tRNA-protein mimicry hypothesis. Proc. Natl. Acad. Sci. U.S.A. 93, 5443–5448.

Ito, K., Uno, M., Nakamura, Y., 2000. A tripeptide “anticodon” deciphers stop co-dons in messenger RNA. Nature 403, 680–684. doi:10.1038/35001115

Jacob, F., Monod, J., 1961. Genetic regulatory mechanisms in the synthesis of pro-teins. J. Mol. Biol. 3, 318–356. doi:10.1016/s0022-2836(61)80072-7

Johansson, M., Bouakaz, E., Lovmar, M., Ehrenberg, M., 2008. The kinetics of ribo-somal peptidyl transfer revisited. Mol. Cell 30, 589–598. doi:10.1016/j.mol-cel.2008.04.010

Johansson, M., Ieong, K.-W., Åqvist, J., Pavlov, M.Y., Ehrenberg, M., 2011. Rate and accuracy of messenger RNA translation on the ribosome, in: Ribosomes. Springer Vienna, Vienna, pp. 225–235. doi:10.1007/978-3-7091-0215-2_18

Julien, O., Beadle, J.R., Magee, W.C., Chatterjee, S., Hostetler, K.Y., Evans, D.H., Sykes, B.D., 2011. Solution structure of a DNA duplex containing the potent anti-poxvirus agent cidofovir. J. Am. Chem. Soc. 133, 2264–2274. doi:10.1021/ja109823e

59

Kolosov, P., Frolova, L., Seit-Nebi, A., Dubovaya, V., Kononenko, A., Oparina, N., Justesen, J., Efimov, A., Kisselev, L., 2005. Invariant amino acids essential for decoding function of polypeptide release factor eRF1. Nucleic Acids Res. 33, 6418–6425. doi:10.1093/nar/gki927

Korostelev, A., Asahara, H., Lancaster, L., Laurberg, M., Hirschi, A., Zhu, J., Trakhanov, S., Scott, W.G., Noller, H.F., 2008. Crystal structure of a translation termination complex formed with release factor RF2. Proc. Natl. Acad. Sci. U.S.A. 105, 19684–19689. doi:10.1073/pnas.0810953105

Kozak, M., 1989. The scanning model for translation: an update. J. Cell Biol. 108, 229–241.

Kozak, M., 1987. An analysis of 5'-noncoding sequences from 699 vertebrate mes-senger RNAs. Nucleic Acids Res. 15, 8125–8148.

Kozak, M., 1978. How do eucaryotic ribosomes select initiation regions in messen-ger RNA? Cell 15, 1109–1123. doi:10.1016/0092-8674(78)90039-9

Kramer, E.B., Farabaugh, P.J., 2007. The frequency of translational misreading er-rors in E. coli is largely determined by tRNA competition. RNA 13, 87–96. doi:10.1261/rna.294907

Kryuchkova, P., Grishin, A., Eliseev, B., Karyagina, A., Frolova, L., Alkalaeva, E., 2013. Two-step model of stop codon recognition by eukaryotic release factor eRF1. Nucleic Acids Res. 41, 4573–4586. doi:10.1093/nar/gkt113

Laurberg, M., Asahara, H., Korostelev, A., Zhu, J., Trakhanov, S., Noller, H.F., 2008. Structural basis for translation termination on the 70S ribosome. Nature 454, 852–857. doi:10.1038/nature07115

Lee, F.S., Chu, Z.-T., Bolger, M.B., Warshel, A., 1992. Calculations of antibody-an-tigen interactions: microscopic and semi-microscopic evaluation of the free en-ergies of binding of phosphorylcholine analogs to McPC603. Protein Eng. 5, 215–228. doi:10.1093/protein/5.3.215

Lifson, S., Warshel, A., 1968. Consistent force field for calculations of confor-mations, vibrational spectra, and enthalpies of cycloalkane andn alkane mole-cules. J. Chem. Phys. 49, 5116–5129. doi:10.1063/1.1670007

Lind, C., Esguerra, M., Åqvist, J., 2017. A close-up view of codon selection in eu-karyotic initiation. RNA Biol. 1–5. doi:10.1080/15476286.2017.1308998

Lind, C., Sund, J., Åqvist, J., 2013. Codon-reading specificities of mitochondrial re-lease factors and translation termination at non-standard stop codons. Nat. Com-mun. 4, 2940. doi:10.1038/ncomms3940

Loveland, A.B., Demo, G., Grigorieff, N., Korostelev, A.A., 2017. Ensemble cryo-EM elucidates the mechanism of translation fidelity. Nature 546, 113–117. doi:10.1038/nature22397

Lünse, C.E., Schüller, A., Mayer, G., 2014. The promise of riboswitches as potential antibacterial drug targets. Int. J. Med. Microbiol. 304, 79–92. doi:10.1016/j.ijmm.2013.09.002

Mandal, M., Boese, B., Barrick, J.E., Winkler, W.C., Breaker, R.R., 2003. Ri-boswitches control fundamental biochemical pathways in Bacillus subtilis and other bacteria. Cell 113, 577–586.

Mandal, M., Breaker, R.R., 2004. Adenine riboswitches and gene activation by dis-ruption of a transcription terminator. Nat. Struct. Mol. Biol. 11, 29–35. doi:10.1038/nsmb710

Marelius, J., Hansson, T., Åqvist, J., 1998a. Calculation of ligand binding free ener-gies from molecular dynamics simulations. Int. J. Quant. Chem. 69, 77–88. doi:10.1002/(SICI)1097-461X(1998)69:1<77::AID-QUA10>3.0.CO;2-2

60

Marelius, J., Kolmodin, K., Feierberg, I., Åqvist, J., 1998b. Q: a molecular dynamics program for free energy calculations and empirical valence bond simulations in biomolecular systems. J. Mol. Graph. Model. 16, 213–25– 261.

Matheisl, S., Berninghausen, O., Becker, T., Beckmann, R., 2015. Structure of a hu-man translation termination complex. Nucleic Acids Res. 43, 8615–8626. doi:10.1093/nar/gkv909

Matzner, D., Mayer, G., 2015. (Dis)similar analogues of riboswitch metabolites as antibacterial lead compounds. J. Med. Chem. 58, 3275–3286. doi:10.1021/jm500868e

McQuarrie, D.A., Simon, J.D., 1999. Molecular Thermodynamics. Sterling Publish-ing Company.

Mironov, A.S., Gusarov, I., Rafikov, R., Lopez, L.E., Shatalin, K., Kreneva, R.A., Perumov, D.A., Nudler, E., 2002. Sensing small molecules by nascent RNA. Cell 111, 747–756. doi:10.1016/S0092-8674(02)01134-0

Mishra, S.K., Sund, J., Åqvist, J., Koča, J., 2012. Computational prediction of mon-osaccharide binding free energies to lectins with linear interaction energy mod-els. J. Comput. Chem. 33, 2340–2350. doi:10.1002/jcc.23081

Moazed, D., Noller, H.F., 1989. Intermediate states in the movement of transfer RNA in the ribosome. Nature 342, 142–148. doi:10.1038/342142a0

Moazed, D., Samaha, R.R., Gualerzi, C., Noller, H.F., 1995. Specific protection of 16 S rRNA by translationalinitiation factors. J. Mol. Biol. 248, 207–210. doi:10.1016/S0022-2836(95)80042-5

Moffat, J.G., Tate, W.P., 1994. A single proteolytic cleavage in release factor 2 sta-bilizes ribosome binding and abolishes peptidyl-tRNA hydrolysis activity. J. Biol. Chem. 269, 18899–18903.

Mulhbacher, J., Brouillette, E., Allard, M., Fortier, L.-C., Malouin, F., Lafontaine, D.A., 2010. Novel riboswitch ligand analogs as selective inhibitors of guanine-related metabolic pathways. PLOS Pathogens 6, e1000865. doi:10.1371/jour-nal.ppat.1000865

Nahvi, A., Sudarsan, N., Ebert, M.S., Zou, X., Brown, K.L., Breaker, R.R., 2002. Genetic control by a metabolite binding mRNA. Chem. Biol. 9, 1043–1049. doi:10.1016/S1074-5521(02)00224-7

Nakamura, Y., Ito, K., 1998. How protein reads the stop codon and terminates trans-lation. Genes Cells 3, 265–278.

Nirenberg, M., Caskey, T., Marshall, R., Brimacombe, R., Kellogg, D., Doctor, B., Hatfield, D., Levin, J., Rottman, F., Pestka, S., Wilcox, M., Anderson, F., 1966. The RNA code and protein synthesis. Cold Spring Harbor Symposia on Quanti-tative Biology 31, 11–24. doi:10.1101/SQB.1966.031.01.008

Nirenberg, M., Leder, P., 1964. RNA codewords and protein synthesis. Science 145, 1399–1407. doi:10.1126/science.145.3639.1399

Nirenberg, M.W., Matthaei, J.H., 1961. The dependence of cell-free protein synthe-sis in E. coli upon naturally occurring or synthetic polyribonucleotides. Proc. Natl. Acad. Sci. U.S.A. 47, 1588–1602. doi:10.1073/pnas.47.10.1588

Nissen, P., Hansen, J., Ban, N., Moore, P.B., Steitz, T.A., 2000. The structural basis of ribosome activity in peptide bond synthesis. Science 289, 920–930. doi:10.1126/science.289.5481.920

Ogle, J.M., Brodersen, D.E., Clemons, W.M., Tarry, M.J., Carter, A.P., Ramakrish-nan, V., 2001. Recognition of cognate transfer RNA by the 30S ribosomal subu-nit. Science 292, 897–902. doi:10.1126/science.1060612

Pape, T., Wintermeyer, W., Rodnina, M.V., 1998. Complete kinetic mechanism of elongation factor Tu dependent binding of aminoacyl tRNA to the A site of the E.coli ribosome. EMBO J. 17, 7490–7497. doi:10.1093/emboj/17.24.7490

61

Parker, J., 1989. Errors and alternatives in reading the universal genetic code. Micro-biol. Rev. 53, 273.

Pauling, L., Corey, R.B., 1956. Specific hydrogen-bond formation between pyrim-idines and purines in deoxyribonucleic acids. Archives of Biochemistry and Bi-ophysics 65, 164–181. doi:10.1016/0003-9861(56)90185-0

Pauling, L., Corey, R.B., 1953. A proposed structure for the nucleic acids. Proc Natl Acad Sci U.S.A.

Pavlov, M.Y., Liljas, A., Ehrenberg, M., 2017. A recent intermezzo at the Ribosome Club. Philos. Trans. R. Soc. Lond., B, Biol. Sci. 372, 20160185. doi:10.1098/rstb.2016.0185

Petry, S., Brodersen, D.E., Murphy, F.V., Dunham, C.M., Selmer, M., Tarry, M.J., Kelley, A.C., Ramakrishnan, V., 2005. Crystal structures of the ribosome in complex with release factors RF1 and RF2 bound to a cognate stop codon. Cell 123, 1255–1266. doi:10.1016/j.cell.2005.09.039

Quigley, G.J., Rich, A., 1976. Structural domains of transfer RNA molecules. Sci-ence 194, 796–806. doi:10.1126/science.790568

Ren, A., Rajashankar, K.R., Patel, D.J., 2012. Fluoride ion encapsulation by Mg2+ ions and phosphates in a fluoride riboswitch. Nature 486, 85–89. doi:10.1038/nature11152

Richter, R., Rorbach, J., Pajak, A., Smith, P.M., Wessels, H.J., Huynen, M.A., Smeitink, J.A., Lightowlers, R.N., Chrzanowska-Lightowlers, Z.M., 2010. A functional peptidyl-tRNA hydrolase, ICT1, has been recruited into the human mitochondrial ribosome. EMBO J. 29, 1116–1125. doi:10.1038/emboj.2010.14

Roberts, R.B., 1958. Microsomal particles and protein synthesis. London: Per-gamon.

Rodnina, M.V., Wintermeyer, W., 2009. Recent mechanistic insights into eukaryotic ribosomes. Curr. Opin. Cell Biol. 21, 435–443. doi:10.1016/j.ceb.2009.01.023

Rozov, A., Demeshkina, N., Westhof, E., Yusupov, M., Yusupova, G., 2015. Struc-tural insights into the translational infidelity mechanism. Nat. Commun. 6, ncomms8251. doi:10.1038/ncomms8251

Ruusala, T., Ehrenberg, M., Kurland, C.G., 1982. Is there proofreading during poly-peptide synthesis? EMBO J. 1, 741–745.

Saad, N.Y., Stamatopoulou, V., Brayé, M., Drainas, D., Stathopoulos, C., Becker, H.D., 2013. Two-codon T-box riboswitch binding two tRNAs. Proc. Natl. Acad. Sci. U.S.A. 110, 12756–12761. doi:10.1073/pnas.1304307110

Sanbonmatsu, K.Y., Joseph, S., 2003. Understanding discrimination by the ribo-some: stability testing and groove measurement of codon–anticodon pairs. J. Mol. Biol. 328, 33–47. doi:10.1016/S0022-2836(03)00236-5

Satpati, P., Sund, J., Åqvist, J., 2014. Structure-based energetics of mRNA decoding on the ribosome. Biochemistry 53, 1714–1722. doi:10.1021/bi5000355

Schluenzen, F., Tocilj, A., Zarivach, R., Harms, J., Gluehmann, M., Janell, D., Ba-shan, A., Bartels, H., Agmon, I., Franceschi, F., Yonath, A., 2000. Structure of functionally activated small ribosomal subunit at 3.3 Å resolution. Cell 102, 615–623. doi:10.1016/S0092-8674(00)00084-2

Schmeing, T.M., Ramakrishnan, V., 2009. What recent ribosome structures have re-vealed about the mechanism of translation. Nature 461, 1234–1242. doi:10.1038/nature08403

Schmeing, T.M., Voorhees, R.M., Kelley, A.C., Gao, Y.-G., Murphy, F.V., Weir, J.R., Ramakrishnan, V., 2009. The crystal structure of the ribosome bound to EF-Tu and aminoacyl-tRNA. Science 326, 688–694. doi:10.1126/sci-ence.1179700

62

Scolnick, E., Tompkins, R., Caskey, T., Nirenberg, M., 1968. Release factors differ-ing in specificity for terminator codons. Proc. Natl. Acad. Sci. U.S.A. 61, 768–774.

Seit-Nebi, A., Frolova, L., Kisselev, L., 2002. Conversion of omnipotent translation termination factor eRF1 into ciliate-like UGA-only unipotent eRF1. EMBO Rep. 3, 881–886. doi:10.1093/embo-reports/kvf178

Selmer, M., Dunham, C.M., Murphy, F.V., Weixlbaumer, A., Petry, S., Kelley, A.C., Weir, J.R., Ramakrishnan, V., 2006. Structure of the 70S ribosome com-plexed with mRNA and tRNA. Science 313, 1935–1942. doi:10.1126/sci-ence.1131127

Serganov, A., Nudler, E., 2013. A decade of riboswitches. Cell 152, 17–24. doi:10.1016/j.cell.2012.12.024

Shan, Y., Kim, E.T., Eastwood, M.P., Dror, R.O., Seeliger, M.A., Shaw, D.E., 2011. How does a drug molecule find its target binding site? J. Am. Chem. Soc. 133, 9181–9183. doi:10.1021/ja202726y

Shao, S., Murray, J., Brown, A., Taunton, J., Ramakrishnan, V., Hegde, R.S., 2016. Decoding mammalian ribosome-mRNA states by translational GTPase com-plexes. Cell 167, 1229–1240.e15. doi:10.1016/j.cell.2016.10.046

Shine, J., Dalgarno, L., 1974. The 3'-terminal sequence of Escherichia coli 16S ribo-somal RNA: complementarity to nonsense triplets and ribosome binding sites. Proc. Natl. Acad. Sci. U.S.A. 71, 1342–1346.

Soleimanpour-Lichaei, H.R., Kühl, I., Gaisne, M., Passos, J.F., Wydro, M., Rorbach, J., Temperley, R., Bonnefoy, N., Tate, W., Lightowlers, R., Chrzanowska-Lightowlers, Z., 2007. mtRF1a is a human mitochondrial translation release fac-tor decoding the major termination codons UAA and UAG. Mol. Cell 27, 745–757. doi:10.1016/j.molcel.2007.06.031

Song, H., Mugnier, P., Das, A.K., Webb, H.M., Evans, D.R., Tuite, M.F., Hem-mings, B.A., Barford, D., 2000. The crystal structure of human eukaryotic re-lease factor eRF1—mechanism of stop codon recognition and peptidyl-tRNA hydrolysis. Cell 100, 311–321.

Spiegel, P.C., Ermolenko, D.N., Noller, H.F., 2007. Elongation factor G stabilizes the hybrid-state conformation of the 70S ribosome. RNA 13, 1473–1482. doi:10.1261/rna.601507

Starck, S.R., Jiang, V., Pavon-Eternod, M., Prasad, S., McCarthy, B., Pan, T., Shas-tri, N., 2012. Leucine-tRNA initiates at CUG start codons for protein synthesis and presentation by MHC class I. Science 336, 1719–1723. doi:10.1126/sci-ence.1220270

Su, T., Cheng, J., Sohmen, D., Hedman, R., Berninghausen, O., Heijne, von, G., Wilson, D.N., Beckmann, R., 2017. The force-sensing peptide VemP employs extreme compaction and secondary structure formation to induce ribosomal stalling. Elife 6, e25642. doi:10.7554/eLife.25642

Sund, J., Andér, M., Åqvist, J., 2010. Principles of stop-codon reading on the ribo-some. Nature 465, 947–950. doi:10.1038/nature09082

Sund, J., Lind, C., Åqvist, J., 2015. Binding site preorganization and ligand discrimi-nation in the purine riboswitch. J. Phys. Chem. B 119, 773–782. doi:10.1021/jp5052358

Tate, W.P., Brown, C.M., 2002. Translational termination: "stop" for protein synthe-sis or “pause” for regulation of gene expression. Biochemistry 31, 2443–2450. doi:10.1021/bi00124a001

Temperley, R., Richter, R., Dennerlein, S., Lightowlers, R.N., Chrzanowska-Lightowlers, Z.M., 2010. Hungry codons promote frameshifting in human mito-chondrial ribosomes. Science 327, 301–301. doi:10.1126/science.1180674

63

Thompson, R.C., Stone, P.J., 1977. Proofreading of the codon-anticodon interaction on ribosomes. Proc. Natl. Acad. Sci. U.S.A. 74, 198–202. doi:10.1073/pnas.74.1.198

Trobro, S., Åqvist, J., 2005. Mechanism of peptide bond synthesis on the ribosome. Proc. Natl. Acad. Sci. U.S.A. 102, 12395–12400. doi:10.1073/pnas.0504043102

Valle, M., Zavialov, A., Sengupta, J., Rawat, U., Ehrenberg, M., Frank, J., 2003. Locking and unlocking of ribosomal motions. Cell 114, 123–134. doi:10.1016/S0092-8674(03)00476-8

Vogel, Z., Zamir, A., Elson, D., 1969. Possible involvement of peptidyl transferase in the termination step of protein biosynthesis. Biochemistry.

Voorhees, R.M., Ramakrishnan, V., 2013. Structural basis of the translational elon-gation cycle. http://dx.doi.org/10.1146/annurev-biochem-113009-092313. doi:10.1146/annurev-biochem-113009-092313

Voorhees, R.M., Schmeing, T.M., Kelley, A.C., Ramakrishnan, V., 2010. The mech-anism for activation of GTP hydrolysis on the ribosome. Science 330, 835–838. doi:10.1126/science.1194460

Voorhees, R.M., Weixlbaumer, A., Loakes, D., Kelley, A.C., Ramakrishnan, V., 2009. Insights into substrate stabilization from snapshots of the peptidyl trans-ferase center of the intact 70S ribosome. Nat. Struct. Mol. Biol. 16, 528–533. doi:10.1038/nsmb.1577

Wachter, A., 2010. Riboswitch-mediated control of gene expression in eukaryotes. RNA Biol. doi:10.4161/rna.7.1.10489

Watson, J., Crick, F., 1953a. A structure for deoxyribose nucleic acid. Nature 171, 737–738. doi:10.1038/171737a0

Watson, J.D., Crick, F.H., 1953b. Genetical implications of the structure of deoxyri-bonucleic acid. Nature 171, 964–967. doi:10.1001/jama.1993.03500150079031

Weixlbaumer, A., Jin, H., Neubauer, C., Voorhees, R.M., Petry, S., Kelley, A.C., Ramakrishnan, V., 2008. Insights into translational termination from the struc-ture of RF2 bound to the ribosome. Science 322, 953–956. doi:10.1126/sci-ence.1164840

Wilkins, M.H.F., Stokes, A.R., Wilson, H.R., 1953. Molecular structure of deoxy-pentose nucleic acids. Nature 171, 738–740. doi:10.1038/171738a0

Wimberly, B.T., Brodersen, D.E., Clemons, W.M., Morgan-Warren, R.J., Carter, A.P., Vonrhein, C., Hartsch, T., Ramakrishnan, V., 2000. Structure of the 30S ribosomal subunit. Nature 407, 327–339. doi:10.1038/35030006

Winkler, W., Nahvi, A., Breaker, R.R., 2002. Thiamine derivatives bind messenger RNAs directly to regulate bacterial gene expression. Nature 419, 952–956. doi:10.1038/nature01145

Winkler, W.C., Breaker, R.R., 2005. Regulation of bacterial gene expression by ri-boswitches. Annu. Rev. Microbiol. 59, 487–517. doi:10.1146/annurev.mi-cro.59.030804.121336

Winkler, W.C., Breaker, R.R., 2003. Genetic control by metabolite-binding ri-boswitches. Chembiochem 4, 1024–1032. doi:10.1002/cbic.200300685

Woese, C.R., Fox, G.E., 1977. Phylogenetic structure of the prokaryotic domain: The primary kingdoms. Proc. Natl. Acad. Sci. U.S.A. 74, 5088–5090. doi:10.1073/pnas.74.11.5088

Wong, L.E., Li, Y., Pillay, S., Frolova, L., Pervushin, K., 2012. Selectivity of stop codon recognition in translation termination is modulated by multiple confor-mations of GTS loop in eRF1. Nucleic Acids Res. 40, 5751–5765. doi:10.1093/nar/gks192

64

Young, D.J., Edgar, C.D., Murphy, J., Fredebohm, J., Poole, E.S., Tate, W.P., 2010. Bioinformatic, structural, and functional analyses support release factor-like MTRF1 as a protein able to decode nonstandard stop codons beginning with ad-enine in vertebrate mitochondria. RNA 16, 1146–1155. doi:10.1261/rna.1970310

Youngman, E.M., He, S.L., Nikstad, L.J., Green, R., 2007. Stop codon recognition by release factors induces structural rearrangement of the ribosomal decoding center that is productive for peptide release. Mol. Cell 28, 533–543. doi:10.1016/j.molcel.2007.09.015

Zamenhof, S., Brawerman, G., Chargaff, E., 1952. On the desoxypentose nucleic ac-ids from several microorganisms. Biochim. Biophys. Acta 9, 402–405. doi:10.1016/0006-3002(52)90184-4

Zeng, X., Chugh, J., Casiano-Negroni, A., Al-Hashimi, H.M., Brooks, C.L., 2014. Flipping of the ribosomal A-site adenines provides a basis for tRNA selection. J. Mol. Biol. 426, 3201–3213. doi:10.1016/j.jmb.2014.04.029

Zhang, Y., Spremulli, L.L., 1998. Identification and cloning of human mitochondrial translational release factor 1 and the ribosome recycling factor. BBA -Gene Structure and Expression 1443, 245–250.

Zwanzig, R.W., 1954. High temperature equation of state by a perturbation method. I. nonpolar gases. J. Chem. Phys. 22, 1420–1426. doi:10.1063/1.1740409

Acta Universitatis UpsaliensisDigital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Science and Technology 1549

Editor: The Dean of the Faculty of Science and Technology

A doctoral dissertation from the Faculty of Science andTechnology, Uppsala University, is usually a summary of anumber of papers. A few copies of the complete dissertationare kept at major Swedish research libraries, while thesummary alone is distributed internationally throughthe series Digital Comprehensive Summaries of UppsalaDissertations from the Faculty of Science and Technology.(Prior to January, 2005, the series was published under thetitle “Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Science and Technology”.)

Distribution: publications.uu.seurn:nbn:se:uu:diva-328583

ACTAUNIVERSITATIS

UPSALIENSISUPPSALA

2017