Transcript
Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology 1306
Accuracy of mRNA Translation in Bacterial Protein Synthesis JINGJI ZHANG
ACTA UNIVERSITATIS UPSALIENSIS UPPSALA 2015
ISSN 1651-6214 ISBN 978-91-554-9383-7 urn:nbn:se:uu:diva-262901
Dissertation presented at Uppsala University to be publicly examined in B10:2, BMC, Husargatan 3, Uppsala, Friday, 4 December 2015 at 13:00 for the degree of Doctor of Philosophy. The examination will be conducted in English. Faculty examiner: Professor Olke C. Uhlenbeck (Northwestern University). Abstract Zhang, J. 2015. Accuracy of mRNA Translation in Bacterial Protein Synthesis. Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology 1306. 49 pp. Uppsala: Acta Universitatis Upsaliensis. ISBN 978-91-554-9383-7. Reading of messenger RNA (mRNA) by aminoacyl-tRNAs (aa-tRNAs) on the ribosomes in the bacterial cell occurs with high accuracy. It follows from the physical chemistry of enzymatic reactions that there must be a trade-off between rate and accuracy of initial tRNA selection in protein synthesis: when the current accuracy, the A-value, approaches its maximal possible value, the d-value, the kinetic efficiency of the reaction approaches zero. We have used an in vitro system for mRNA translation with purified E. coli components to estimate the d- and A-values by which aa-tRNAs discriminate between their cognate and near cognate codons displayed in the ribosomal A site. In the case of tRNALys, we verified the prediction of a linear trade-off between kinetic efficiency of cognate codon reading and the accuracy of codon selection. These experiments have been extended to a larger set of tRNAs, including tRNAPhe, tRNAGlu, tRNAHis, tRNACys, tRNAAsp and tRNATyr, and linear efficiency-accuracy trade-off was observed in all cases. Similar to tRNALys, tRNAPhe discriminated with higher accuracy against a particular mismatch in the second than in the first codon position. Remarkably high d-values were observed for tRNAGlu discrimination against a C-C mismatch in the first codon position (70 000) and for tRNAPhe discrimination against an A-G mismatch in the second codon position (79 000). At the same time, we have found a remarkably small d-value (200) for tRNAGlu misreading G in the middle position of the codon (U-G mismatch). Aminoglycoside antibiotics induce large codon reading errors by tRNAs. We have studied the mechanism of aminoglycoside action and found that the drug stabilized aminoacyl-tRNA in a codon selective in relation to a codon non-selective state. This greatly enhanced the probability of near cognate aminoacyl-tRNAs to successfully transcend the initial selection step of the translating ribosome. We showed that Mg2+ ions, in contrast, favour codon non-selective states and thus induce errors in a principally different way than aminoglycosides. We also designed experiments to estimate the overall accuracy of peptide bond formation with, including initial selection accuracy and proofreading of tRNAs after GTP hydrolysis on EF-Tu. Our experiments have now made it possible to calibrate the accuracy of tRNA selection in the test tube to that in the living cells. We will now also be able to investigate the degree to which the accuracy of tRNA selection has been optimized for maximal fitness. Keywords: protein synthesis, genetic code, misreading, error hot spots, kinetics, aminoglycoside Jingji Zhang, Department of Cell and Molecular Biology, Structure and Molecular Biology, 596, Uppsala University, SE-751 24 Uppsala, Sweden. © Jingji Zhang 2015 ISSN 1651-6214 ISBN 978-91-554-9383-7 urn:nbn:se:uu:diva-262901 (http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-262901)
To my princess
List of Papers
This thesis is based on the following papers, which are referred to in the text by their Roman numerals. I
II
III IV
Johansson, M., Zhang, J., Ehrenberg, M. (2012) Genetic code translation displays a linear trade-off between efficiency and accuracy of tRNA selection. Proc Natl Acad Sci U S A, 109(1):131–136 Zhang, J., Ieong, K.W., Johansson, M., Ehrenberg, M. (2015) Accuracy of initial codon selection by aminoacyl-tRNAs on the mRNA-programmed bacterial ribosome. Proc Natl Acad Sci U S A, 112(31):9602–9607 Zhang, J., Pavlov, M., Ehrenberg, M. (2015) On the mechanism of translation error induction by aminoglycoside antibiotics. (Manuscript) Zhang, J.*, Ieong, K.W.*, Ehrenberg, M. (2015) Enhanced proofreading neutralizes potential error hot spots in genetic code translation by transfer RNAs. (Manuscript) *Co-first author
Reprints were made with permission from the respective publishers.
Contents
Introduction ................................................................................................... 11 Early kinetic models of translation accuracy............................................ 12 Linus Pauling’s accuracy model .......................................................... 12 Bacterial protein synthesis ................................................................... 14 High and tunable intracellular accuracy of mRNA translation ............ 15 Kinetics and single step substrate selection ......................................... 16 Proofreading selection ......................................................................... 18 The present work........................................................................................... 22 Recent two step kinetic modeling of translation accuracy ....................... 22 In vivo accuracy studies............................................................................ 23 Mg2+ dependent efficiency and accuracy trade-off (Paper I and II) ......... 24 Aminoglycoside dependent efficiency-accuracy trade-off (Paper III) ..... 31 Accuracy enhancement by proofreading (Paper I and IV) ....................... 37 Conclusions ................................................................................................... 41 Summary in Swedish .................................................................................... 43 Acknowledgements ....................................................................................... 45 References ..................................................................................................... 47
Abbreviations
30S 50S 70S A A aa aa-tRNA A site ASL ATP C E. coli EF-G EF-Tu fMet G GTP IC IF IleRS Mg mRNA PTC P site PEP rRNA R group RNA T3 tRNA U ValRS
The small subunit of a bacterial ribosome The large subunit of a bacterial ribosome The complete bacterial ribosome Adenosine Accuracy Amino acid Aminoacyl-tRNA Aminoacyl-tRNA site Anticodon stem loop Adenosine 5’-triphosphate Cytidine Escherichia coli Elongation factor G Elongation factor Tu Formylmethionine Guanosine Guanosine 5’-triphospate Initiation complex Initiation factor Isoleucyl-tRNA synthetase Magnesium Messenger RNA Peptidyl transferase center Peptidyl-tRNA site Phosphoenolpyruvate Ribosomal RNA Amino acid side chain Ribonucleic acid Ternary complex (aa-tRNA·EF-Tu·GTP) Transfer RNA Uridine Valyl-tRNA synthetase
Introduction
In the living cell, genetic information stored in molecules of DNA is transcribed into messenger RNA (mRNA) molecules and then translated into protein by the ribosome. The ribosome reads the genetic code, composed of base triplets (codons) in the mRNA sequence. Each codon is decoded by a transfer RNA (tRNA) carrying the corresponding amino acid. There are 61 codons to code for the 20 natural amino acids and three stop codons, which tell the ribosome to end translation of the message. It is vital for life that this process of translating the genetic code in mRNA into proteins is both fast and highly accurate. This thesis concerns the accuracy of bacterial protein synthesis and is split into two sections. First there is a brief history of accuracy research including Linus Pauling’s first accuracy model from 1957, a short overview of bacterial protein synthesis, Robert Loftfield’s in vivo accuracy measurements in 1972, Luigi Gorini’s ribosomal accuracy mutants in 1971, Jaques Ninio’s kinetic interpretation of Gorini’s data and John Hopfield’s proofreading mechanism in 1974, as well as experimental evidence for proofreading from Anne Norris Baldwin and Paul Berg (1966), John Hopfield (1976), Robert Thomson and Pamela Stone (1977) and Alan Fersht’s double sieve mechanism (1979). The second section concerns the present work and includes Marina Rodnina’s kinetic model for tRNA selection by the ribosome in 2004 and Philip Farabaugh’s in vivo measurements of accuracy in 2014. It focuses on our measurements of the Mg2+ concentration (Paper I and II) and aminoglycoside (Paper III) dependent speed and accuracy trade-off in initial codon selection as well as our measurements of the total accuracy (Paper IV).
11
Early kinetic models of translation accuracy Linus Pauling’s accuracy model Already in the 1950s Linus Pauling formulated essential concepts for the later understanding of the accuracy of protein synthesis. He used his experimental estimates of the precision by which antibodies can discriminate between a cognate antigen and its near cognate competitors along with physical chemical modeling of ligand binding to proteins to discuss the frequency of errors in protein synthesis. He postulated the existence of template proteins designed to recognize their cognate amino acids (aac) and reject near cognate amino acids (aanc). Discrimination was modeled as an equilibrium between a template protein, E, and its amino acid ligands (Pauling, 1957): Kc
E aa c nc
nc
c nc C
Cc is the cognate and Cnc the near cognate complex, Kc/nc the dissociation equilibrium constant for cognate and near cognate complex formation, i.e. the ratio between the dissociation rate constant and the association rate constant. Pauling defined the affinity difference between cognate and near cognate substrates as the discrimination ratio Knc/Kc, the d-value. He focused on how well a template protein with a binding pocket for a methyl group could possibly discriminate against a hydrogen atom and, vice versa, how well a template protein cognate to a hydrogen atom (group) could discriminate against a methyl group (Table 1).
12
Table 1. Comparison of binding constant for antibody to cognate and near cognate antigens (d-value).
Experimental data
London theory of electronic dispersion forces
Forces between antibody and hapten
Template
Van der Waals repulsion
Antibody
Cognate substrate
Near cognate dsubstrate value1
Benzoic acid with methyl group Antigen with Antigen with Antibody hydrogen methyl group group Antigen with Van der Acetylcholine Antigen with hydrogen Waals attrac- esterase inhibitor methyl group group tion Protein
Benzoic acid with hydrogen group
Alanine
100
Glycine
3-6
6-7
4.3
1
The ratio of near cognate and cognate dissociation constants (d-value) are from (Pauling, 1957).
Linus Pauling identified van der Waals attraction, van der Waals repulsion, hydrogen-bond formation along with attraction between positively and negatively charged groups as important factors for amino acid recognition by template proteins (Pauling, 1957). He expected discrimination against near cognate amino acids with smaller R group than the cognate amino acid to be more difficult than when the near cognate amino acid had a larger R group. This hypothesis was based on the idea that steric exclusion by van der Waals repulsion would be a better discriminator than the differential binding energy provided by van der Waals attraction to the larger side group. He found that an antibody, evolved for cognate binding to benzoic acid with a hydrogen group, had 100 times larger dissociation constant for benzoic acid with the hydrogen replaced by a methyl group (Table 1). From this he postulated that the machinery for protein synthesis in the living cell would not be able to discriminate against near cognate amino acids with a methyl group replacing the hydrogen of the cognate amino acid better than by a factor of 100, which would lead to amino acid substitution errors in the 1% range in such cases. He exemplified with the amino acids alanine (methyl group) and glycine (hydrogen group). However, in the case where a near cognate amino acid has smaller R group than the cognate one, then even larger error frequencies could be expected. In this case it is the van der Waals attraction energy that is important. Here, he calculated the error to be around 20% for incorpora13
tion of near cognate glycine instead of alanine, meaning that the accuracy of this mis-incorporation would be about 4 (Table 1). A similar example is incorporation of valine as the near cognate amino acid with a hydrogen group instead of isoleucine with a cognate methyl group.
Bacterial protein synthesis Synthesis of peptide chains on the mRNA programmed ribosome includes four stages: initiation of protein synthesis, peptide chain elongation, termination of protein synthesis and recycling of ribosomes. In brief, these stages are as follows. Initiation To initiate protein synthesis, initiation factor 3 (IF3) binds to the small (30S) subunit, preventing it from premature docking with the large (50S) ribosomal subunit (Antoun et al., 2006b). An mRNA binds to the 30S subunit and is positioned with its initiation codon (AUG) in the ribosomal P site by base pairing between its Shine-Dalgarno sequence (SD-sequence) and the anti-SD sequence of the 16S ribosomal RNA (rRNA). With the help of initiator factors 1 (IF1) and 2 (IF2) initiator tRNA (fMet-tRNAfMet) binds to the initiation codon (Antoun et al., 2006a). Then the 50S subunit docks with the 30S subunit, the initiation factors dissociate from the ribosome after GTP hydrolysis on IF2 and the 70S initiation complex with initiator tRNA in P site is ready for entry of the first elongator tRNA into the A site. Peptide Elongation and Translocation Aminoacyl-tRNA in ternary complex with EF-Tu and GTP binds to the A site of the 70S post-translocation complex with peptidyl-tRNA in the P site (Schmeing et al., 2009). Cognate interaction between the anticodon of the incoming aminoacyl-tRNA and the mRNA codon in the A site rapidly leads to GTP hydrolysis on EF-Tu and release of EF-Tu in the GDP form, to be recycled back to EF-Tu·GDP by elongation factor Ts (EF-Ts). Then the nascent peptide chain is transferred from the P site tRNA by peptide bond formation with the amino acid of the A-site accommodated aminoacyl-tRNA. Then EF-G·GTP binds to the pre-translocation ribosome with peptidyltRNA in hybrid A/P state and deacylated tRNA in hybrid P/E state (Ermolenko et al., 2007; Spiegel et al., 2007), hydrolyses GTP and translocates the mRNA by one codon and the tRNAs to P/P and E/E states, respectively (Rodnina et al., 1997). After release of EF-G·GDP the now empty A site is ready for yet another round of peptide elongation.
14
Termination When the A site displays one of the three stop codons (UAA, UAG or UGA), a class one release factor (RF1 or RF2) binds to the A site (Scolnick et al., 1968) and promotes ester bond hydrolysis and release of the completed peptide chain from the P site bound peptidyl-tRNA. The GTP form of the class two release factor (RF3) induces the ribosome conformational change that releases the class one release factor from the post-termination ribosome (Gao et al., 2007). After GTP hydrolysis on RF3 it leaves the posttermination ribosome (Zavialov et al., 2001; Zavialov et al., 2002). Recycling Ribosome recycling factor (RRF) and EF-G·GTP split the post-termination complex into ribosomal subunits in a GTP hydrolysis dependent manner (Karimi et al., 1999; Zavialov et al., 2005). Then IF3 binding to the 30S subunit removes its remaining deacylated tRNA, making the ribosomal subunits ready for yet another round of initiation (Karimi et al., 1999; Peske et al., 2005).
High and tunable intracellular accuracy of mRNA translation Robert Loftfield developed assays to measure the accuracy of protein synthesis in the living cell, which made it possible for him to seriously test Pauling’s prediction of sloppy intracellular messenger RNA translation (Loftfield and Vanderjagt, 1972). He studied isoleucine-containing peptide fragments from rabbit haemoglobin and estimated the frequency of valine substitution for isoleucine as between two and six per 10 000. This frequency is about a thousand times smaller than Pauling’s predictions and Loftfield speculated that cells may have a “scavenging mechanism for eliminating errors, but more probably Nature has found means for accentuating the small differences among amino acids and the similarly small differences among trinucleotide code-words”. These words were remarkably prophetic in that we now know that single step discrimination can be much better than the poor d-values estimated by Pauling and that there indeed exist if not scavenging so at least editing mechanisms that allow for repeated selection of the same “small difference” between cognate and near cognate substrates. A plethora of antibiotic drugs bind to the bacterial ribosome (Yonath, 2005) and some of these induce high amino acid substitution errors in the living cell as well as in the test tube (Ogle and Ramakrishnan, 2005). One of these is streptomycin (Sm), a bacteriocidal drug which has generated a large number of Sm-resistant, Sm-dependent, Sm-pseudo dependent and Sm-hyper sensitive mutants in clever selection schemes pioneered by Luigi Gorini. He found that some Sm-resistant bacteria (strA) appeared to have more accurate mRNA translation than wild type strains due to amino acid substitutions in 15
ribosomal protein S12 of the small ribosomal subunit. He also found that other ribosomal mutants displayed not only increased sensitivity to Sm but also reduced accuracy of mRNA translation. These low accuracy mutations were localized to alterations in proteins S4 and S5, of the small ribosomal subunit (Ozaki et al., 1969). Gorini was fascinated by his finding that a single ribosome alteration in a ribosomal protein could leave cognate codon reading virtually unchanged but greatly affect the propensity of near cognate reading up or down.
Kinetics and single step substrate selection Jacques Ninio not only introduced chemical kinetics and used this tool to tentatively interpret Gorini’s complex data sets (Ninio, 1974) but also forged generally applicable concepts to explain selective enzymatic pathways in general. Here, we describe a modified version of Ninio’s kinetics albeit keeping in line with his general ideas. The mRNA programmed ribosome selects aminoacyl-tRNAs in ternary complex with elongation factor Tu (EFTu) and guanosine-5’-triphosphate (GTP) for binding to the ribosomal A site and subsequent hydrolysis of GTP by EF-Tu. A simple aa-tRNA selection scheme that summarizes the essence of Ninio’s contribution can be written: k1
T3 R
c nc
q1c
c nc
C1
kc
nc
2
nc
Here the same type of ternary complex (T3) binds to a ribosome programmed with its cognate, R c , or a near cognate, R nc , codon with the same association rate constant (k1) to form a cognate, C1c, or a near cognate ribosome complex, C1nc. Products are formed with rate constants k2c/nc. We define a discard parameter a (q1c /k2c) which determines the efficiency of cognate codon reading. The discrimination value (d) is (q1nc /q1c)· (k2c /k2nc). The steady state flows to product formation are given by:
k j T3 R cat Km c
c
c
k k1 T3 R c T3 R c 1 c 1 a q 1 1c k2
k j nc T3 R nc cat Km
16
nc
T3 R nc
k1 k1 T3 R nc 1 d a q1nc 1 nc k2
The ratio between cognate and near cognate flow is given by: c
T3 R c jc j nc T3 R nc
k cat K m T3 R c 1 d a nc T3 R nc 1 a k cat Km
If in both cognate and near cognate cases the T3 and R concentrations are equal, then c
jc j nc
k cat K m 1 d a A nc 1 a k cat Km
The accuracy (A) is defined as the ratio between cognate, (kcat/Km)c, and near cognate efficiency, (kcat/Km)nc, of product formation. The d-value is normally much larger than one. If it is smaller than one, the definition of cognate and near cognate would change. It is interpreted as the maximal possible accuracy of cognate in relation to near cognate product formation (Ehrenberg and Blomberg, 1980; Ehrenberg and Kurland, 1984) and can be written as
d eG
c ,nc
/ RT
Here, ΔΔGc,nc is the standard free energy difference between cognate and near cognate ribosome complex in the transition state of the GTP hydrolysis reaction. R is the gas constant and T the absolute temperature. The a-value can be understood as an accuracy tuning parameter that determines how much of the d-value that the enzyme uses in its current A-value. Now we come back to Ninio’s kinetic explanation of the phenotypes of Gorini’s ribosome mutants (Ninio, 1974). The Ram mutation decreases the q1c/nc parameter by the same factor for cognate and near cognate substrate, meaning that the phenotype of Ram is reduced a-parameter. In the extreme case, when a is close to zero, the A goes to its minimum value of 1. So the Ram mutation is an error prone mutant as Gorini’s data showed. The StrA mutation decreases the k2c/nc parameter which increases the a parameter. In the extreme case, when a is very large, then A approaches its maximum value d. So the phenotype of the StrA mutation is hyper accurate code translation. This means, in other words, that Ninio suggested that these mutations affected the a-
17
parameter up or down but left the d-parameter unaltered. In this case there is a linear relation between (kcat/Km)c and A: kcat K m
c
dA k1 d 1
Plotting (kcat/Km)c versus A leads to a straight line:
By changing the a-value to change (kcat/Km)c and A, the d-value can be obtained from the intercept of the line with the x-axis. The accuracy A is essentially the inverse of the error frequency E: E
j nc j nc j c
1 1
c
j j nc
1 1 ~ 1 A A
Proofreading selection Linus Pauling’s predictions of the accuracy of protein synthesis in living cells had turned out to be severe underestimates. The deviation between prediction and reality could mean that he underestimated the d-values that can be achieved by the aminoacyl-tRNA synthetases that couple the amino acids to their cognate tRNAs (Loftfield et al., 1973). Another option would be that the living cell could, somehow, use the same (small) intrinsic discrimination capacity, the d-value, several times and in this way supersede the physical chemical limitation of selective enzymes.
18
One intriguing experiment suggested that the latter option could be true. It was found by Anne Norris Baldwin and Paul Berg in 1966 (Baldwin and Berg, 1966) that, after activation of Ile to Ile-AMP by isoleucyl-tRNA synthetase (IleRS), addition of tRNAIle led to rapid formation Ile-tRNAIle: IleRS Ile AMP tRNA Ile Ile tRNA Ile IleRS AMP
However, after activation of the near cognate amino acid Val to Val-AMP by IleRS, addition of tRNAIle just leads to hydrolysis of Val-AMP: IleRS Val AMP tRNA Ile Val tRNA Ile IleRS AMP
In the mid-seventies, John Hopfield (Hopfield, 1974) and Jacques Ninio (Ninio, 1975) proposed a kinetic proofreading step after initial selection, that would allow the same difference in standard free energy between near cognate or cognate substrate to be used several times. A generic scheme for proofreading applied to mRNA translation may be written as: k1
T3 R c nc
c nc k 2c nc C1 q1c
c nc
C2
qc
nc
R c nc
kc
nc
3
nc
2 aa tRNA EF Tu GDP
In this scheme, initial selection of ternary complex ends through GTP hydrolysis by EF-Tu in complex C1c/nc leading to complex C2c/nc and release of inorganic phosphate (Pi). From complex C2c/nc the reaction can either lead back to ribosomal complex Rc/nc through release of EF-Tu·GDP and aatRNA with rate constant q2c/nc or, alternatively to A-site accommodation and peptide bond formation with rate constant k3c/nc. We note that in this scheme the influx from ribosome complex Rc/nc over the vertical proofreading path has been neglected. The reason why this can be done and, in fact, why proofreading works at all is that the ternary complex concentration is very far above equilibrium with EF-Tu·GDP and aa-tRNA. In the living cell such out of equilibrium shifts have been estimated as 108 (Hopfield, 1974), which provide yet another upper limit to how much proofreading can enhance the accuracy of enzymatic selection (Ehrenberg and Blomberg, 1980). At equal concentrations of cognate and near cognate ribosomes the accuracy, A, is given by:
19
c
kcat c Km 1 d1 a1 1 d 2 a2 j A nc nc 1 a1 1 a2 j kcat Km Here a1=q1c /k2c, a2=q2c /k3c, d1=(q1nc /q1c)· (k2c /k2nc) and d2=(q2nc /q2c)· (k3c /k3nc). The maximum accuracy is d1d2. This scheme shows in concrete terms that the total accuracy of an enzyme system can be much higher than the d-value of initial selection, here equal to d1. Referring to the above question; the A could be larger than d1. It also follows from this type of scheme that the proofreading contribution to the accuracy is equal to the ratio between the numbers of GTPs (or ATPs in other cases) hydrolyzed per near cognate (fnc) and cognate (fc) product formation (Hopfield, 1974). In 1976, Hopfield and Yamane proved experimentally that on average 1.5 ATPs are hydrolyzed per Ile-tRNAIle formed by IleRS in the ATP driven aminoacylation reaction. In contrast, 270 ATPs were consumed per formed Val-tRNAIle by IleRS. This demonstrated that proofreading exists in this aminoacylation reaction and suggested the proofreading contribution to the overall accuracy of 180 (270/1.5) (Hopfield and Yamane, 1976). We note that the probabilities that an aminoacyl-tRNA “survives” the proofreading step in aminoacylation is 1/fnc=P2nc in the near cognate and 1/fc = P2c in the cognate case, so the F=P2c/P2nc =fnc/fc. Now, the accuracy A is the initial selection, I , multiplied with the proofreading selection F: A=I·F. Thompson and Stone (1977) were the first to demonstrate proofreading of aa-tRNAs in bacterial protein synthesis (Thompson and Stone, 1977). Later, Ruusala et al. (1982) demonstrated proofreading of aa-tRNAs in a more complete in vitro system for bacterial protein also containing elongation factor G (EF-G) (Ruusala et al., 1982). Now, we have seen that proofreading of aa-tRNAs enhances the accuracy of protein synthesis on the ribosome and that proofreading of amino acids by aminoacyl-tRNA synthetases enhance the accuracy of aminoacylation. What about the d-value estimates provided by Pauling? Are they accurate? Alan Fersht measured kcat/Km for activation of Val (cognate) and Ile (near cognate) by valyl-tRNA synthetase (ValRS) (Fersht and Dingwall, 1979), and showed that the cognate initial selection reaction was 60 000 times more efficient than the near cognate one, i.e. that I=60 000 (Fersht and Dingwall, 1979). This means a d-value at least 600 times larger than Linus Pauling’s estimate of d=100 for a template protein we can now identify as ValRS (Pauling, 1957). Fersht also measured the kcat and Km for activation of Val and Ile by isoleucyl-tRNA synthetase (IleRS) (Fersht, 1977). From the kcat/Km values for both cognate Ile and near cognate Val, the initial selection is about 200 20
(Fersht, 1977; Fersht, 1979), i.e. about 50 times larger than Linus Pauling’s prediction, which was 4.3 (Pauling, 1957). These comparisons show that Pauling’s d-value estimates for common amino acids were far below the mark. This means, in conclusion, that the discrepancy between Pauling’s prediction that protein synthesis in living cells is permeated with amino acid substitution errors (Pauling, 1957) and Loftfield’s experimental observations that such errors are extremely rare stems from his underestimated d-values and from mother Nature’s invention of proofreading. One reason why Pauling’s d-values came out so low, could be that they were based on equilibrium constants, while Fersht’s estimates were based on the efficiency of catalytic reactions. In the latter case it is likely that substrates are positioned in precise positions to minimize the entropy of activation of reactions. Fersht found that Ile activation by IleRS to Ile-AMP complex is a fast, initial step and Ile transfer to tRNAIle is a slow, rate limiting step (Fersht and Kaethner, 1976). He measured the efficiency of ValRS activation of Ile (large side group), Val (cognate) and Thr (isosteric side group) (Fersht and Dingwall, 1979). He found that the aa-activation site can discriminate against amino acids with side chains larger than that of Val, but not against those with smaller side chains. Discrimination against substrates larger than the cognate constitutes the first sieve in the double sieve mechanism proposed by Fersht to explain amino acid selection by aminoacyl-tRNA synthetases (Fersht and Dingwall, 1979). The editing site, in contrast, could discriminate against amino acids with side groups smaller than that of the cognate substrate. This constitutes the second sieve of the double sieve mechanism suggested by Fersht.
21
The present work
Recent two step kinetic modeling of translation accuracy A kinetic model for initial codon selection was proposed by Marina Rodnina and collaborators (Gromadski and Rodnina, 2004): k1
T3 R
c nc
q1
k2
C1c nc
k c nc
3 C2c nc
q 2c nc
In the above scheme, R is the ribosome, T3 is the ternary complex, c stands for cognate, nc for near cognate, C1 is the initial binding complex. The first binding step is not codon-anticodon selective, and thus with the same dissociation constant, Kd (q1/k1), for cognate and near cognate ternary complex (Rodnina et al., 1994). When the Mg2+ concentration was varied from 5 mM to 10 mM, Kd decreased ~ 100 fold (Rodnina et al., 1996). At Mg2+ concentrations less than 3 mM, Kd is above 30 µM, leading to rapid dissociation from C1 (rate constant q1) after binding to (rate constant k1) this state. Under these low Mg2+ conditions, there is negligible inhibition of protein synthesis by near cognate ternary complex (Johansson et al., 2008). The forward rate constant, k2, leading to codon recognition in complex C2 is the same for cognate and near cognate reactions. The C2 complex is much more stable for cognate than for near cognate ternary complex: the ratio q2nc /q2c ratio is about 1000 (Gromadski and Rodnina, 2004). GTPase activation and hydrolysis are irreversible steps as indicated in the above scheme. The rate of conformation change of EF-Tu after GTP hydrolysis has been analyzed by fluorescence from mant-GTP replacing GTP in the ternary complex. It was found that k3c is larger than 260 s-1 and that k3nc is 0.4 s-1 (Gromadski and Rodnina, 2004). The much smaller value of k3nc than of k3c has been used to characterize GTPase activation as an induced fit mechanism (Zaher and Green, 2009). To measure the proofreading contribution to accuracy, purified ternary complex was mixed with ribosomes programmed with cognate or near cognate codon for the aminoacyl-tRNA and the relative yield (1/fc or 1/fnc) of
22
peptide bond formation was estimated in each case (Gromadski and Rodnina, 2004). Rodnina and co-workers estimated K2c/nc-values for the equilibrium between complexes C1 and C2, containing Phe-tRNAPhe interacting with either one cognate or six near cognate codons in the ribosomal A site. They found similar K2nc values in all near cognate cases and suggested uniform accuracy in ternary complex recognition (Gromadski et al., 2006). Using pre-steady kinetics, Hani S. Zaher and Rachel Green investigated the above mentioned classical ribosome mutants StrA and Ram found by Gorini (Zaher and Green, 2009). The Ram mutation mainly increased the error of initial ternary complex selection, whereas the StrA mutation mainly decreased the error in proofreading selection of aminoacyl-tRNA.
In vivo accuracy studies The accuracy of the steps in the central dogma displays great variation. In DNA replication, transcription and translation accuracy levels of 108-1010 (Kunkel and Bebenek, 2000), 104 (Rosenberger and Foskett, 1981) and 103104 (Bouadloun et al., 1983; Edelmann and Gallant, 1977; Kramer and Farabaugh, 2007; Laughrea et al., 1987), respectively, have been estimated. Codon reading accuracy measured in vivo using enzyme (firefly luciferase or beta-galactosidase) mutants that need missense errors for residual activity have an error background level between 10-6 and 10-5 (Manickam et al., 2014). With these in vivo assays Farabaugh and co-workers have estimated different types of missense errors when tRNAGlu, tRNATyr or tRNAAsp misread their near cognate codons (Kramer and Farabaugh, 2007; Manickam et al., 2014). Concerning the first codon position, the U36-U1 mismatch is error prone (tRNALys misreading stop codon UAA). U-C, U-G, A-A, A-C, A-G, C-C, C-U and C-A mismatches do not lead to significant errors. The most error frequent codon reading concerns the second and third codon positions. In the second codon position, U35-G2 (tRNATyr misreading codons UGC, UGU; tRNALys misreading codons AGA, AGG; tRNAAsp misreading codons GGU, GGC; tRNAGlu misreading codons GGA, GGG) is a mismatch with high error propensity. In the third codon position, mnm5s2U34-U3/C3 (tRNALys misreading codons AAU, AAC; tRNAGlu misreading codon GAU, GAC) is more error prone than Q34-A3/G3 (tRNAAsp misreading codon GAA, GAG; tRNATyr misreading codon UAA, UAG). In contrary, our biochemical system with E. coli components of high purity measures error levels down to 10-8 (Paper IV).
23
Mg2+ dependent efficiency and accuracy trade-off (Paper I and II) The following generic model for initial selection of codons by tRNAs may serve as the starting point for a discussion about the accuracy of bacterial protein synthesis:
Fig 1. Kinetic scheme of initial codon selection on the mRNA programmed ribosome.
The cognate (c) and near cognate (nc) kcat/Km-values (efficiencies) for GTP hydrolysis are given by:
kcat / Km
c , nc
ka
kcc ,nc ka c , nc c , nc c , nc kc k d 1 kd / kcc , nc
The (normalized) accuracy of initial selection is defined as the ratio between cognate and near cognate kcat/Km-values so that: A kcat / Km / kcat / Km c
nc
1 kdnc / kcnc 1 da 1 kdc / kcc 1 a
It follows that the discrimination parameter, d, and discard parameter, a, are given by: d
kcc kdnc eG / RT ; a kdc / kcc kcnc kdc
In an experiment where a is varying, e.g. by changing the free Mg2+ concentration, at constant ka and d-values, there is a linear trade-off between cognate efficiency and accuracy:
kcat / K m
24
c
ka
dA d 1
We measured the cognate and near cognate efficiency by performing GTP hydrolysis experiments (Fig 2)
Fig 2. Measurements of kcat/Km-parameters for GTP hydrolysis during cognate or near cognate codon reading. Time evolution of the level of [3H]GDP in response to [3H]GTP·EF-Tu·Cys-tRNACys binding to 70S ribosomes programmed with cognate codon or near cognate codons. For the cognate reaction in short timeframe (top left panel), data was fitted to a single exponential function. Each near cognate experiment was performed in parallel with a cognate experiment (in black). The very same ternary complex mixture was here used for both cognate and near cognate reactions, and both curves were jointly fitted with sharing of parameters to increase precision of the measurement. In all experiments ribosomes were in excess over ternary complexes, and kcat/Km-values were calculated from the apparent GTP hydrolysis rate constant divided by the active ribosome concentration (here, 0.7 µM and 1.8 µM ribosomes were used for cognate and near cognate reactions, respectively). The decrease in [3H]GDP level in the long time-frame is due to spontaneous dissociation of [3H]GDP from EF-Tu followed by its rapid regeneration to [3H]GTP by pyruvate kinase. All experiments were performed in polymix buffer with the addition of 2 mM extra Mg(OAc)2.
25
From such experiments the initial selection accuracy can be calculated from ratios between cognate and near cognate kcat/Km-values. With this experimental tool, the GTP hydrolysis efficiency for cognate and near cognate codon reading by tRNAGlu, tRNAPhe, tRNAHis, tRNACys, tRNATyr, tRNAAsp and tRNALys was studied. Increasing kcat/Km-values for cognate and near cognate codon reading at different Mg2+ concentrations are displayed in Figs 3A and 3B, respectively. Selected linear trade-off plots are shown in Fig. 3C and all the seven association rate constants, ka, for cognate codon reading are shown in Fig. 3D.
26
Fig 3. The rate-accuracy trade-off. (A) Efficiency of cognate GTP hydrolysis, (kcat/Km)c, for different tRNAs (see panel B for symbol legend) reading their fully matched codons at varying Mg2+ concentration. (B) Efficiency of near cognate GTP hydrolysis, (kcat/Km)nc, for different tRNAs reading single mismatch codons at varying Mg2+ concentration. (C) Efficiency of cognate GTP hydrolysis, (kcat/Km)c, versus the accuracy (calculated as the ratio (kcat/Km)c/(kcat/Km)nc) for different tRNAs reading single mismatch codons as indicated in panel B. In each tRNA misreading case, cognate and near cognate (kcat/Km) values were measured at different Mg2+ concentrations as shown in panels A and B. The x-axis intercept gives the maximal accuracy, d, for each misreading case, and the y-axis intercept gives the rate constant for association of each cognate tRNA to the ribosome. (D) The rate constant, ka, for association of different aa-tRNAs in ternary complexes to ribosomes, estimated from the linear dependence between cognate GTP hydrolysis efficiency and accuracy. Data in panels A, B and C represent weighted averages from at least two experiments ± propagated standard deviation. Error bars in panel D represent the standard deviation estimates from the parameter fitting procedure, where experimental errors, such as in panel A and B, are used as weights.
27
As seen, all cognate kcat/Km values saturate at high Mg2+ concentration but not the near cognate ones, as explained in paper III. The cognate saturation values, ka , are in the range from 75 µM-1s-1 to 200 µM-1s-1 among the seven studied tRNAs. In all cases, we observe linear trade off plots with y-axis intercepts estimating ka-values and x-axis intercepts estimating d-values:
28
Fig 4. Maximal accuracy variation dependent on codon position, mismatch identity, and aa-tRNA. Maximal accuracy values, d, for single mismatch readings by different tRNAs, summarized with respect to mismatch codon position (columns) and mismatch identities (rows for the anticodon bases and colours for the codon bases as indicated in the figure).
29
The d-values varied from 200 to 84,000 with different types of mismatch, tRNA and codon position. In general, with the same type of tRNA and mismatch, the second codon position had highest and the third codon position the lowest d-value. Pyrimidine-purine mismatches show the lowest d-values in general. There are also error hot spots: tRNAGlu misreading GGA (Gly) with d-value 200, GAU (Asp) with d-value 240, GAC (Asp) with d-value 650; tRNAHis misreading CGC (Arg) with d-value 250, CAA (Gln) with dvalue 200.
30
Aminoglycoside dependent efficiency-accuracy tradeoff (Paper III) Taking advantage of quench-flow techniques and data analysis, previously applied to initial codon selection (Paper II), we estimated the efficiency (kcat/Km) by which Phe-tRNAPhe in ternary complex with EF-Tu and GTP selects a cognate (UUC) and a near cognate codon (CUC) at different Mg2+ concentrations in the presence and absence of paromomycin (Figure 5A).
Fig 5. Mg2+ concentration dependence of the efficiency of ribosome induced GTP hydrolysis in initial selection of ternary complex. Efficiency of GTP hydrolysis for tRNAPhe-containing ternary complex reading cognate (UUC) codon (kcat/Km)c in absence ( ) or presence ( ) of paromomycin or reading near cognate (CUC) codon in absence ( ) or presence ( ) of paromomycin at varying Mg2+ concentration.
The cognate efficiency was identical in the absence and presence of paromomycin and varied about fourfold from its lowest to its plateau value as the free Mg2+ concentration increased from 1 to 22 mM. The near cognate efficiency, in contrast, increased by four orders of magnitude in the absence and by two orders of magnitude in the presence of paromomycin. All efficiencies reached the same plateau value in the limit of high Mg2+ concentration. The observations that paromomycin did not affect the variation of cognate kcat/Km and that all efficiency curves reached the same efficiency plateau can be accounted for by a simple three step model for initial selection:
31
Fig 6. Schematic and efficiency of initial codon selection on the mRNA programmed ribosome. (A) Kinetic scheme of initial codon selection on the mRNA programmed ribosome. (B) Simplified kinetic scheme of panel corresponding to the cartoon in (A). (C) Efficiency of cognate and near cognate initial codon selection.
R2 is a ternary complex-containing ribosome complex without codonanticodon contact, as suggested (Rodnina et al., 1994), while R3 is a ternary complex-containing ribosome complex with codon-anticodon contact in which the monitoring bases A1493, A1492 and G530 are activated and in contact with the minor groove of the codon-anticodon helix (Carter et al., 2000; Lynch et al., 2003). Hence, the only codon specific parameters are a2=q2c/k3c and a2d2=q2nc/k3nc so that c
k cat Km
k1 1 a 1 1 a 2
k cat Km
nc
k1 1 a1 1 a2 d 2
In all cases, parameter a1, unaffected by amino glycoside binding decreased with increasing Mg2+ concentration from about three towards zero. Parameter a2, unaffected by Mg2+ concentration, decreased by a large factor upon paromomycin addition from a value much smaller than one to an even smaller value. Parameter a2d2, unaffected by Mg2+ concentration, was greatly reduced by aminoglycoside binding from a very large to a smaller but still large number. Since in the cognate case a2 was much smaller than one and 32
became even smaller upon drug addition, the cognate curves in Fig. 5 were insensitive to the presence or absence of the drug. Since, in contrast, a2d2 was much larger than one, its large decrease with paromomycin was clearly seen. Cognate and near cognate kcat/Km-values were used to construct efficiency-accuracy trade-off lines for ternary complex-containing Phe-tRNAPhe misreading CUC (Fig. 7A), UCC (Fig. 7B) and UUA (Fig. 7C) in absence of aminoglycosides or presence of paromomycin, gentamicin or neomycin. For each near cognate codon, different aminoglycosides resulted in different slopes of the straight lines, whereas the intercepts with the y-axis, ka, representing the cognate rate constant for association of the specific aa-tRNAcontaining ternary complex to the ribosome, was the same as without durgs. All d-values obtained in this work are summarized in Table 2 along with previously obtained tRNAPhe data (Zhang et al., 2015).
33
Fig 7. Efficiency-accuracy trade-off and rate constants for cognate ternary complex dissociation from the ribosome. Efficiency of cognate GTP hydrolysis, (kcat/Km)c, versus the accuracy (calculated as the ratio (kcat/Km)c/(kcat/Km)nc) for tRNAPhe reading CUC (A), UCC (B) and UUA (C) codons, in each case with no drug, paromomycin, gentamicin or neomycin. In each tRNA misreading case, cognate and near cognate (kcat/Km) values were measured at different Mg2+ concentrations. The x-intercept gives the effective discrimination parameter, de, for each misreading case, and the y-intercept gives the rate constant for association of each cognate tRNA to the ribosome. (D) Time evolution of dipeptide bond formation in experiment, where A-site bound ternary complex EF-Tu(H84A)·GTP·Phe-tRNAPhe with GTPase deficient EF-Tu mutant His84A is chased by ternary complex with wild type EF-Tu.
34
Table 2. Rate constant q2 of cognate reaction compared with d-values for different near cognate codons with and without antibiotics No Paromomycin Gentamicin Neomycin drug chase of ternary 1.04 0.08 0.016 0.008 complex, q2c (s-1) Fold change by 1 14 70 170 drug 1760 ± d-vaue tRNAPhe 75 ± 10 17 ± 4 3.9 ± 0.6 on CUC 180 Fold change by 1 23 104 451 drug 5900 ± d-vaue tRNAPhe 420 ± 76 103 ± 20 30 ± 5 on UCC 970 Fold change by 1 14 57 197 drug 1200 ± d-vaue tRNAPhe 111 ± 22 14,6 ± 0,8 9 ± 1.7 on UUA 190 Fold change by 1 11 82 133 drug Form the Mg2+ dependent steps each a1-value at each Mg2+ concentration could be estimated. The effective d-value (a2d2) decreased 23-, 14-, and 11fold upon paromomycin addition in cases of tRNAPhe reading CUC (Fig. 7A), UCC (Fig. 7B) or UUA (Fig. 7C) codon (Table 1). If we take the average of these three numbers, the mean d-value decrease by paromomycin was 16 fold (Table 3). Upon gentamicin addition the corresponding d-values decreased by factors of 104, 57 and 82 with an average of 80. Upon neomycin addition these d-value reductions were 451, 197 and 133 with an average of 260 (Table 3). The smallest effective d-value in Table 1 is 4, obtained from tRNAPhe reading CUC codon in the presence of neomycin. This result shows that any residual codon selectivity of the parameter ratio q1/k2 in Fig. 6 must be smaller than 4, in line with the hypothesis that complex R2 lacks codon-anticodon contact. From the above section, we know that aminoglycosides profoundly affect the effective d-value a2d2. To find out if aminoglycosides affect a2, d2 or both parameters we performed experiments in which cognate ternary complex with a GTPase deficient EF-Tu mutant was chased from the ribosomal A site with a native ternary complex (complex III, Fig. 6A).
35
For this, we used the GTPase deficient EF-Tu mutant (H84A) (Daviter et al., 2003) to form the mutant ternary complex (EF-Tu(H84A)·GTP·PhetRNAPhe) for pre-incubation with ribosomes with UUC programmed A sites UUC for formation of complex III (Fig. 6A). Addition of the native ternary complex leads to peptide bond formation at a rate determined by the compounded rate constant, qdiss, for dissociation of the mutated ternary complex (Fig. 7D):
1 1 c 1 a1 c qdiss q2 Since we have estimated a1 as 1.0 at 2.3 mM free Mg2+ concentration we could estimate q2c as 1 s-1 in the absence of aminoglycoside (Table 2). Using the same method, estimated q2c values in the presence of paromomycin, gentamicin or neomycin as 0.08, 0.016 and 0.008 s-1, respectively (Table 2). Table 3. Average d-value change compared with ternary complex dissociation rate change -1 -value change q2c (s ) change No drug 1 1 Paromomycin 16 14 Gentamicin 80 70 Neomycin 260 170 Table 3 shows that the fold change in effective d-value in response to addition of each one of the drugs closely corresponds to the fold change in the cognate back reaction parameter q2 (Table 2). This result strongly suggests that the effective d-value change in response to aminoglycoside addition is caused by uniform reduction of the back rate constant from complex C2 (recent two step kinetic modeling translation accuracy section) for cognate and near cognate ternary complex in an aminoglycoside specific manner. This would mean that the reduction in effective d-value observed here does not reduce the intrinsic accuracy of the reaction (d2) but only how much of it is expressed in ternary complex selection for GTP hydrolysis (a2). This conclusion is in sharp contrast to previous suggestions that aminoglycosides greatly enhance the rate constant k3 for near cognate but not cognate ternary complex, which would imply a great reduction in the intrinsic discrimination parameter d2 (Pape et al., 2000).
36
Accuracy enhancement by proofreading (Paper I and IV) We also designed experiments to estimate the overall accuracy of peptide bond formation, which includes the accuracy enhancement provided by proofreading of tRNA following initial selection of ternary complex. We used the following generic model for overall accuracy selection of codons by tRNAs:
Fig 8. Kinetic scheme of peptide bond formation on the mRNA programmed ribosome. aa-tRNAs can be rejected during initial selection or at the proofreading step. The two selection steps are separated by hydrolysis of EF-Tu bound GTP.
In the experiments we used Lys-tRNALys, Glu-tRNAGlu and Phe-tRNAPhe with either fully matched cognate codons or single mismatched near cognate codons:
Fig 9. The overall accuracy of tRNA selection was measured for Lys-tRNALys, Glu-tRNAGlu, and Phe-tRNAPhe reading all possible single-mismatch codons. Compared to their fully matched codons AAA (tRNALys), GAA (tRNAGlu) and UUC (tRNAPhe), mismatch codon positions are underlined.
We estimated the kcat/Km values for dipeptide formation when the same ternary complex reacted with cognate and near cognate programmed ribosomes at varying Mg2+ concentration.
37
Figure 10. Measurements of cognate and near cognate kcat/Km-values for dipeptide formation. (A) Time evolution of f[3H]Met-Glu formation. Ternary complexes EF-Tu·GTP·Glu-tRNAGlu were reacted with 70S initial complexes programmed with f[3H]Met-tRNAfMet in the P site and a cognate codon GAA (black) or near cognate codon GGA (red) in the A site. Reactions were performed at increasing complex concentration as indicated in the figure. Ternary complexes were in excess over ribosomes so that the rate of dipeptide formation kdip was limited by ternary complex concentration. (B) Concentration dependence of the rate of dipeptide formation kdip estimated from (A). Insert: near cognate reaction. Experiments were performed in Polymix buffer with 2.3 mM free Mg2+.
From such experiments the overall accuracy can be calculated from ratios between cognate and near cognate kcat/Km-values. With this experimental tool, the dipeptide bond formation efficiency for cognate and near cognate codon reading by tRNALys, tRNAGlu and tRNAPhe was studied. Increasing kcat/Km-values for cognate and near cognate codon reading at different Mg2+ concentrations cause the overall accuracy decrease as shown (Fig. 11):
38
Figure 11. The rate-accuracy trade-off in overall selection. Rate-accuracy tradeoff plots in log-log scale for overall accuracy for tRNAGlu selection of GAA in relation to the GGA codon.
The trade-off line for overall accuracy of codon reading now makes it possible to calibrate accuracy data obtained in vitro with the translation accuracy in vivo from the previous section (in vivo accuracy studies).
39
Figure 12. Comparison of in vivo and in vitro misreading error frequency. In vivo data (black stars) from the Farabaugh lab (Manickam et al., 2014) are based on induction of bioluminescence by mistranslation by tRNAGlu ternary complex from E. coli strains with β-galactosidase mutants. In vitro measurements (red squares) were performed at 2.3 mM free Mg2+ and calibrated to the in vivo condition according to the abundance of the 2 competing tRNA species in vivo (Dong et al., 1996) (see Materials and Methods) and assuming different ternary complexes as well as release factors have similar efficiencies for binding to ribosomes in the living cell. Mismatch codon positions are underlined.
The best correspondence between our in vitro accuracy measurements and those in vivo is at 2.3 mM free Mg2+ concentration in our in vitro system. Here our measurements and those in vivo agree well at the error levels from 10-5 and higher. It is seen that when our error estimates go down below 10-6, the in vivo estimates does not respond, a result very likely due to an error background in this range. In the case of tRNAGlu reading G in second and U or C in third position (Zhang et al., 2015), the overall accuracy is in a range below 1/3 000. So the proofreading remedies those initial selection errors to a tolerable level. 40
Conclusions
Like the children of moles, which always inherit the skills to dig holes, the children of humans inherit the merits and demerits of their parents. Like parent, like offspring. An implication of these proverbs of Chinese and English origin is that the accuracy of genetic code translation is an essential feature of life. The accuracy of messenger RNA translation is high in the living cell and depends on two consecutive steps: initial reading of the genetic code by transfer RNA followed by transfer RNA guided proofreading of the initial code reading. This thesis is about speed and accuracy of genetic code translation. In brief, we have used biochemical experiments to demonstrate a linear trade-off between the speed and accuracy of messenger RNA translation on the bacterial ribosome (Papers I and II). We have clarified the mechanism by which antibiotic drugs of the aminoglycoside type decrease the accuracy of messenger RNA translation (Paper III). We have shown that the accuracy amplification by proofreading is essential to remedy error hot spots in initial readings of messenger RNA (Paper IV). Over the past six decades scientists have been brooding about how gene expression to protein by messenger RNA transcription and translation can have both high speed and high accuracy. Speculative explanations have been given, but hard experimental evidence on how the speed-accuracy trade-off is resolved in living cells has been hard to come by. Here, quench-flow and other techniques have been used to monitor the speed of correct messenger RNA reading performed at different Mg2+ ion concentrations to tune the level of misreading (Paper I). Plots of correct initial reading speed versus accuracy of initial genetic code reading, defined as the inverse of the misreading frequency, were used to estimate the maximal reading accuracy at zero reading speed and the maximal reading speed in the low accuracy limit. We studied the speed-accuracy trade-off for initial messenger RNA reading by seven transfer RNAs. These experiments involved fourteen correct genetic code letter readings and fifty six of the most common misreadings by these transfer RNAs, and covered about 15% of all correct and incorrect messenger RNA readings. There was a 400-fold variation in the accuracy of initial code reading, in which we identified distinct error hot spots. We suggested that such error hot spots forced Nature to evolve a proofreading step for genetic code translation (Paper II). 41
We subjected the effects of aminoglycoside antibiotics on initial code reading to an in depth biochemical study (Paper III). We identified two fundamental reading steps. First, transfer RNA binds to the ribosome in complex with its protein factor in a state that is blind for the genetic message. Then, transfer RNA moves to a reading activated state. We found that Mg2+ ions shift the equilibrium from free transfer RNA equally to the blind and reading activated states. Aminoglycosides, in contrast, only shift the equilibrium from the reading blind to the reading activated state. These findings have vast implications for the mechanism of initial code reading as well as for the mechanism by which aminoglycosides corrupt the accuracy of genetic code translation to protein (Paper III). We calibrated the total genetic code reading accuracy as obtained in our biochemical assay to that in the living bacterial cell. It follows from our data that the error frequency of genetic code reading varies by five orders of magnitude within a subset of three transfer RNA types. As predicted (Paper II), the potentially deleterious error hot spots in initial codon reading are neutralized by enhanced proofreading (Paper IV).
42
Summary in Swedish
Precis som råttors avkomma som alltid gräver hål, så även med människor, söner ärver sina fäders egenskaper, både bra och dåliga. Sådan far sådan son. Både det kinesiska och det svenska ordspråket antyder att noggrannheten i överföringen av informationen i den genetiska koden är viktig i naturen. I den här avhandlingen undersöks noggrannheten i mRNA translationen på djupet, både initialselektionen och korrekturläsningssteget. Direkta mätningar av den operativa diskrimineringskonstanten i initialselektionen hur hastigheten i kodontranslationen minskar när noggrannheten går mot sitt maximala värde (Artikel I och II). Aminoglykosider kan minska den operationella, men inte den verkliga, diskrimineringskonstanten (Artikel III). Förstärkningen av noggrannheten genom korrekturläsningssteget är speciellt viktig för att hantera de ”hot spots” där noggrannheten i initialselektionen är särskilt låg (Paper IV). Under de senaste 60 åren har forskare försökt förstå hur levande celler kan uppnå både hög noggrannhet och hög hastighet när de uttrycker sina gener. En modell för både hastigheten och noggrannheten i bakteriers proteinsyntes publicerades i tidskriften Current opinion in microbiology år 2008 men experimentellt stöd för denna hypotes saknas fortfarande. Med hjälp av både quench-flow och manuella tekniker upptäckte vi att genom att variera koncentrationen av Mg2+ joner så kunde vi demonstrera det linjära hastighet-noggrannhet sambandet experimentellt (Paper I). Från de här linjära sambanden kan vi få fram både de operationella diskrimineringskonstanterna, noggrannheten när hastigheten med kognat substrat är noll, och associationshastighetskonstanten för kognat substrat när noggrannheten är 1. Dessa samband bidrar även med en visualisering av hur hastigheten minskar när noggrannheten ökar. Genom mätningar av hastigheten och noggrannheten hos 7 kognata, 7 wobble, och 56 när-kognata kodon-antikodon par, motsvarande ungefär 15 % av den genetiska koden, upptäckte vi en stor variation (~400 ggr) i noggrannheten för initialselektionen av aminoacyl-tRNA av den bakteriella ribosomen (Artikel II). Vi identifierade även flera fel ”hot spots” där noggrannheten är särskilt låg. Vi föreslår att korrekturläsningssteget har utvecklats för att korrigera för dessa ”hot spots”. Vi har undersökt denna operationella diskriminering mellan rätt och fel tRNA i initialselektionen med hjälp av aminoglykosid antibiotika för att få vidare förståelse för hur den fungerar i levande celler (Artikel III). Det visar 43
sig att det finns två trippelkomplexbundna ribosomala stadier. Först ett stadie som inte är selektivt med avseende på kodon-antikodon interaktionen. Sen ett stadie som är selektivt med avseende på kodon-antikodon interaktionen. Mg2+ joner skiftar jämvikten mellan fritt trippelkomplex och det första bundna stadiet mot det bundna stadiet men lämnar det andra stadiet ostört. Aminoglykosidantibiotika skiftar jämvikten mellan de två bundna stadier mot det selektiva stadiet men lämnar jämvikten mellan fritt och bundet trippelkomplex ostört. Genom att jämföra våra mätningar in vitro med in vivo data fann vi att ungefär 50% av både det operationella diskrimineringsvärdet och den kognata hastigheten utnyttjas av levande celler (Artikel IV). Korrekturläsningen är viktig för att kompensera för den oanvända delen av den operationella diskrimineringskonstanten för att minimera skada på det bakteriella proteomet från felläsning av den genetiska koden under proteinsyntesen, speciellt vid fel ”hot spots” i initialselektionen.
44
Acknowledgements
During the last six years of my PhD studies, many people were there for me when I studied here, when I worked out scientific problems, when I needed friends and support, and when I laughed and grieved. They brought something wonderful to my life and without them my life in Sweden would be significantly different. I would like to take this opportunity to thank all of them. First and foremost I want to express my very sincere and special appreciation to my main supervisor, Måns Ehrenberg. His professional knowledge, excelsior scientific attitude, inclusive mind and high prestige have created such an inspiring academic environment. Not only did it give me space to learn and grow, but provided me with the chance to communicate with some of the world’s top scientists. All of these make him to be one of the best supervisors I have ever had. I thank him for giving me the opportunity of being his student, for trusting me gradually and for his support at any moment. Without his guidance and constant feedback this Ph.D. would not have been achievable. I’d also like to give a heartfelt, special thanks to my co-supervisor, Suparna Sanyal, who offered help of all sorts, useful tips with experimental design and interesting talks. Great thanks to my co-supervisor, Michael Y. Pavlov, for equation setup and insightful comments on drafts. I would like to thank Professor Anthony C. Forster too, for the useful tips on presentation. My thanks also go out to all the former and present members in the Molecular Biology program. I am especially grateful to Magnus Johansson who shared his experimental knowledge when I was a newcomer to our lab, and gave me lots of useful comments on manuscripts. He witnessed my growth in scientific research. I am also very grateful to Ka-Weng Ieong for collaboration. Anneli Borg, thank you for chipper laughter. Thanks to Harriet Mellenius for dissertation party information. Gabriele Indrisiunaite, we had many scientific discussions. Onur Ercan’s jokes lightened my daily work. Ajith Harish shared his career experience. I can’t forget Raymond Fowler who was behind me every time when I need components and chemical orders. Chandra Sekhar Mandava, a good-tempered person. Ravi Kiran Koripella’s introduction of Indian language helped me to learn about another ancient country. Xueliang Ge, the nicest and most patient roommate 45
I have ever had. Ram Gopal Nitharwal interpreted some knowledge of tRNA synthetase for me. Yanhong Pang, her joy, happiness and laugher kept all throughout our chats. A lot of thanks to Mikael Holm, I got many benefits after our every discussion about projects, and he gave me a big hand in the end of writing my thesis. Petar Kovachev, gave me his jokes. Gürkan Korkmaz, gave me a chance to experience X-ray in akademiska sjukhuset together. I couldn’t forget Matteo Libero Baroni’s very special good food too. And I also remembered it was Kristin Peisker told me how to prepare ATP solution, and I had a very interesting talk with Sunanda Chatterjee. Ranjeet Kumar always welcomed me with nice juice when I went to his apartment. There were many scientific questions in Neelanjan Vishnu’s brain which led me to wonder his cranial capacity. Marek Kwiatkowski’s chemistry knowledge was helpful to understand my work. Josefine Liljeruhm shared her synthetic biology book. Sofia Pytharopoulou, we shared the same bench. I knew more about American from Tyson Shepherd’s talking. Jinfan Wang, our planting and competing on HPLC made me cherish my working time more than before. Thank all of you for coloring my working days. I grew more and learned more every day with all of you working in the same lab. I also give my thanks to the reliable Chinese friends who made my life in Sweden much more convenient and entertaining. Leena Tirkkonen, my Mom in Sweden. Thanks for giving me a room to live when I first arrived in Sweden, for driving me out to feel the country scenes of Sweden, and for everything we experienced together. I would also like to say a heartfelt thank you to my parents, for helping in whatever way they could during this challenging period. My hard-working parents have sacrificed their lives for myself and provided unconditional love and care. I love them so much, and I would not have made it this far without them. A very special thank you to Mr. Xuejun Zhang and Mrs. Xiaojuan Gao for allowing me to take care of their little princess, for always being so supportive of our love. I am thankful for it and appreciate it. There are so many very special people who have touched my life in past six years, but there is one that stands out – my lovely princess, Bo. She saw something in me that I didn’t know was there; she believed in me until I learned to believe in myself; she brought me to sense wonderful parts of the world and explored something which would happen in the future; she accompanied me in laughter and facing my fears; she gave me a colorful life and she brightened my days and soul. There were many possibilities for us to miss each other, while I’m lucky that I could find her crown and she would like to share her life with me. Thank you, my princess, for being in my life. And finally to myself, thankful for me, never give up, ever. 46
References
Antoun, A., Pavlov, M.Y., Lovmar, M., and Ehrenberg, M. (2006a). How initiation factors maximize the accuracy of tRNA selection in initiation of bacterial protein synthesis. Mol Cell 23, 183-193. Antoun, A., Pavlov, M.Y., Lovmar, M., and Ehrenberg, M. (2006b). How initiation factors tune the rate of initiation of protein synthesis in bacteria. Embo J 25, 2539-2550. Baldwin, A.N., and Berg, P. (1966). Transfer Ribonucleic Acid-Induced Hydrolysis of Valyladenylate Bound to Isoleucyl Ribonucleic Acid Synthetase. J Biol Chem 241, 839-&. Bouadloun, F., Donner, D., and Kurland, C.G. (1983). Codon-specific missense errors in vivo. Embo J 2, 1351-1356. Carter, A.P., Clemons, W.M., Brodersen, D.E., Morgan-Warren, R.J., Wimberly, B.T., and Ramakrishnan, V. (2000). Functional insights from the structure of the 30S ribosomal subunit and its interactions with antibiotics. Nature 407, 340-348. Daviter, T., Wieden, H.J., and Rodnina, M.V. (2003). Essential role of histidine 84 in elongation factor Tu for the chemical step of GTP hydrolysis on the ribosome. J Mol Biol 332, 689-699. Dong, H.J., Nilsson, L., and Kurland, C.G. (1996). Co-variation of tRNA abundance and codon usage in Escherichia coli at different growth rates. J Mol Biol 260, 649-663. Edelmann, P., and Gallant, J. (1977). Mistranslation in E. coli. Cell 10, 131-137. Ehrenberg, M., and Blomberg, C. (1980). Thermodynamic Constraints on Kinetic Proofreading in Biosynthetic Pathways. Biophys J 31, 333-358. Ehrenberg, M., and Kurland, C.G. (1984). Costs of Accuracy Determined by a Maximal Growth-Rate Constraint. Quarterly reviews of biophysics 17, 45-82. Ermolenko, D.N., Majumdar, Z.K., Hickerson, R.P., Spiegel, P.C., Clegg, R.M., and Noller, H.F. (2007). Observation of intersubunit movement of the ribosome in solution using FRET. J Mol Biol 370, 530-540. Fersht, A.R. (1977). Editing Mechanisms in Protein-Synthesis - Rejection of Valine by Isoleucyl-Transfer-Rna Synthetase. Biochemistry-Us 16, 1025-1030. Fersht, A.R. (1979). Editing mechanisms in the aminoacylation of tRNA. Cold Spring Harbor Monograph Archive 9, 247-254. Fersht, A.R., and Dingwall, C. (1979). Evidence for the Double-Sieve Editing Mechanism in Protein-Synthesis - Steric Exclusion of Isoleucine by ValylTransfer Rna-Synthetases. Biochemistry-Us 18, 2627-2631. Fersht, A.R., and Kaethner, M.M. (1976). Mechanism of Aminoacylation of Transfer-Rna - Proof of Aminoacyl Adenylate Pathway for Isoleucyl-TransferRna and Tyrosyl-T-Rna Synthetases from Escherichia-Coli-K12. BiochemistryUs 15, 818-823.
47
Gao, H.X., Zhou, Z.H., Rawat, U., Huang, C., Bouakaz, L., Wang, C.H., Cheng, Z.H., Liu, Y.Y., Zavialov, A., Gursky, R., et al. (2007). RF3 induces ribosomal conformational changes responsible for dissociation of class I release factors. Cell 129, 929-941. Gromadski, K.B., Daviter, T., and Rodnina, M.V. (2006). A uniform response to mismatches in codon-anticodon complexes ensures ribosomal fidelity. Mol Cell 21, 369-377. Gromadski, K.B., and Rodnina, M.V. (2004). Kinetic determinants of high-fidelity tRNA discrimination on the ribosome. Mol Cell 13, 191-200. Hopfield, J.J. (1974). Kinetic proofreading: a new mechanism for reducing errors in biosynthetic processes requiring high specificity. Proceedings of the National Academy of Sciences of the United States of America 71, 4135-4139. Hopfield, J.J., and Yamane, T. (1976). Experimental-Evidence for Kinetic Proofreading in Aminoacylation of Transfer-Rna-Ile Biophys J 16, A16-A16. Johansson, M., Bouakaz, E., Lovmar, M., and Ehrenberg, M. (2008). The kinetics of ribosomal peptidyl transfer revisited. Mol Cell 30, 589-598. Karimi, R., Pavlov, M.Y., Buckingham, R.H., and Ehrenberg, M. (1999). Novel roles for classical factors at the interface between translation termination and initiation. Mol Cell 3, 601-609. Kramer, E.B., and Farabaugh, P.J. (2007). The frequency of translational misreading errors in E. coli is largely determined by tRNA competition. Rna 13, 87-96. Kunkel, T.A., and Bebenek, K. (2000). DNA replication fidelity. Annu Rev Biochem 69, 497-529. Laughrea, M., Latulippe, J., Filion, A.M., and Boulet, L. (1987). Mistranslation in 12 Escherichia-Coli Ribosomal-Proteins - Cysteine Misincorporation at Neutral Amino-Acid-Residues Other Than Tryptophan. Eur J Biochem 169, 59-64. Loftfield, R.B., Pastuszy.A, and Lovgren, T. (1973). Mechanism of Aminoacylation of Transfer-Rna. Fed Proc 32, 460-&. Loftfield, R.B., and Vanderjagt, D. (1972). The frequency of errors in protein biosynthesis. The Biochemical journal 128, 1353-1356. Lynch, S.R., Gonzalez, R.L., and Puglisi, J.D. (2003). Comparison of x-ray crystal structure of the 30S subunit-antibiotic complex with NMR structure of decoding site oligonucleotide-paromomycin complex. Structure 11, 43-53. Manickam, N., Nag, N., Abbasi, A., Patel, K., and Farabaugh, P.J. (2014). Studies of translational misreading in vivo show that the ribosome very efficiently discriminates against most potential errors. Rna 20, 9-15. Ninio, J. (1974). Semi-Quantitative Treatment of Missense and Nonsense Suppression in Stra and Ram Ribosomal Mutants of Escherichia-Coli Evaluation of Some Molecular Parameters of Translation Invivo. J Mol Biol 84, 297-313. Ninio, J. (1975). Kinetic Amplification of Enzyme Discrimination. Biochimie 57, 587-595. Ogle, J.M., and Ramakrishnan, V. (2005). Structural insights into translational fidelity. Annu Rev Biochem 74, 129-177. Ozaki, M., Mizushim.S, and Nomura, M. (1969). Identification and Functional Characterization of Protein Controlled by Streptomycin-Resistant Locus in E Coli. Nature 222, 333-&. Pape, T., Wintermeyer, W., and Rodnina, M.V. (2000). Conformational switch in the decoding region of 16S rRNA during aminoacyl-tRNA selection on the ribosome. Nat Struct Biol 7, 104-107.
48
Pauling, L. (1957). The Probability of Errors in the Process of Synthesis of Protein Molecules (Birkhäuse, Basel), pp597-602. Peske, F., Rodnina, M.V., and Wintermeyer, W. (2005). Sequence of steps in ribosome recycling as defined by kinetic analysis. Mol Cell 18, 403-412. Rodnina, M.V., Fricke, R., and Wintermeyer, W. (1994). Transient Conformational States of Aminoacyl-Transfer-Rna during Ribosome Binding Catalyzed by Elongation-Factor Tu. Biochemistry-Us 33, 12267-12275. Rodnina, M.V., Pape, T., Fricke, R., Kuhn, L., and Wintermeyer, W. (1996). Initial binding of the elongation factor Tu•GTP•aminoacyl-tRNA complex preceding codon recognition on the ribosome. J Biol Chem 271, 646-652. Rodnina, M.V., Savelsbergh, A., Katunin, V.I., and Wintermeyer, W. (1997). Hydrolysis of GTP by elongation factor G drives tRNA movement on the ribosome. Nature 385, 37-41. Rosenberger, R.F., and Foskett, G. (1981). An estimate of the frequency of in vivo transcriptional errors at a nonsense codon in Escherichia coli. Mol Gen Genet 183, 561-563. Ruusala, T., Ehrenberg, M., and Kurland, C.G. (1982). Is There Proofreading during Polypeptide-Synthesis. Embo J 1, 741-745. Schmeing, T.M., Voorhees, R.M., Kelley, A.C., Gao, Y.G., Murphy, F.V., Weir, J.R., and Ramakrishnan, V. (2009). The Crystal Structure of the Ribosome Bound to EF-Tu and Aminoacyl-tRNA. Science 326, 688-694. Scolnick, E., Tompkins, R., Caskey, T., and Nirenber.M (1968). Release Factors Differing in Specificity for Terminator Codons. P Natl Acad Sci USA 61, 768&. Spiegel, P.C., Ermolenko, D.N., and Noller, H.F. (2007). Elongation factor G stabilizes the hybrid-state conformation of the 70S ribosome. Rna 13, 14731482. Thompson, R.C., and Stone, P.J. (1977). Proofreading of Codon-Anticodon Interaction on Ribosomes. P Natl Acad Sci USA 74, 198-202. Yonath, A. (2005). Antibiotics targeting ribosomes: Resistance, selectivity, synergism, and cellular regulation. Annu Rev Biochem 74, 649-679. Zaher, H.S., and Green, R. (2009). Fidelity at the molecular level: lessons from protein synthesis. Cell 136, 746-762. Zavialov, A.V., Buckingham, R.H., and Ehrenberg, M. (2001). A posttermination ribosomal complex is the guanine nucleotide exchange factor for peptide release factor RF3. Cell 107, 115-124. Zavialov, A.V., Hauryliuk, V.V., and Ehrenberg, M. (2005). Splitting of the posttermination ribosome into subunits by the concerted action of RRF and EFG. Mol Cell 18, 675-686. Zavialov, A.V., Mora, L., Buckingham, R.H., and Ehrenberg, M. (2002). Release of peptide promoted by the GGQ motif of class 1 release factors regulates the GTPase activity of RF3. Mol Cell 10, 789-798. Zhang, J., Ieong, K.W., Johansson, M., and Ehrenberg, M. (2015). Accuracy of initial codon selection by aminoacyl-tRNAs on the mRNA-programmed bacterial ribosome. Proceedings of the National Academy of Sciences of the United States of America 112, 9602-9607.
49
Acta Universitatis Upsaliensis Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology 1306 Editor: The Dean of the Faculty of Science and Technology A doctoral dissertation from the Faculty of Science and Technology, Uppsala University, is usually a summary of a number of papers. A few copies of the complete dissertation are kept at major Swedish research libraries, while the summary alone is distributed internationally through the series Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology. (Prior to January, 2005, the series was published under the title “Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology”.)
Distribution: publications.uu.se urn:nbn:se:uu:diva-262901
ACTA UNIVERSITATIS UPSALIENSIS UPPSALA 2015
