MedicineLakex

Length-dependent translation initiation benefits the functional proteome of human cells

Volume 11 Number 2 February 2015 Pages 321–666 Molecular
BioSystems
Interfacing chemical biology with the -omic sciences and systems biology Tong Wang, Gong Zhang et al.
Length-dependent translation initiation benefi ts the functional proteome of human cells Length-dependent translation initiation benefitsthe functional proteome of human cells† Cite this: Mol. BioSyst., 2015,
Jieming Guo,‡ab Xinlei Lian,‡a Jiayong Zhong,a Tong Wang*a and Gong Zhang*a We previously found that shorter mRNAs are preferably translated in various eukaryotic cells. However, the theoretical basis of this phenomenon is unclear. We hypothesize that shorter mRNA length correlates to the decreased translational error rate to reduce the energy consumption on defective protein degradation. In this study, we established a computational model to explain the length- dependent translation initiation efficiency. We provided mathematical evidence that this translational preference, rather than the protein degradation, is a major factor to shape the genome-wide length- dependent protein abundance. As deducted, we simulated that shorter mRNA length is a determinant of Received 4th August 2014, initiation circularization time. Furthermore, our model unveiled that preferentially translating shorter Accepted 8th October 2014 mRNAs benefits the energy efficiency on the proteome functionality. We proposed that cancer cells DOI: 10.1039/c4mb00462k tend to hijack this evolutionary mechanism by counteracting the higher translational error rate. In conclusion, our model provides insights into the nature of the global length-dependent translational control and its biological significance.
known that the average protein length in eukaryotes significantlyexceeds prokaryotes in all of the functional classes.9,10 In eukaryotes, translation initiation is a rate-limiting process These findings led to an important question in the field for protein synthesis that serves as a primary determinant of regarding why human cells need this length-dependent translation protein productivity1–3 (reviewed in ref. 4 and 5). Although this initiation. According to the following rationale, we hypothesize that complex process involves multiple factors, numerous evidence the translational preference of shorter mRNAs will decrease the obtained from yeast and plant cells has revealed the connection error rate of protein synthesis, thus reducing the energy demand, between mRNA/coding sequence (CDS) length and translation and is beneficial to a functional proteome of human cells.
initiation, namely, mRNAs encoding larger proteins tend to be First, maintaining a functional proteome is a requirement less translated.2,6,7 While in human normal and cancer cells, for a living cell in the steady-state (reviewed in ref. 11). As the we confirmed this length-dependent translation initiation in a counterpart, the ‘‘wrong'' proteome, including misfolded, genome-wide scale; interestingly, we reported that cancer cells unfolded and/or incorrectly translated proteins, should be are more preferentially translating shorter mRNAs.8 In a recent maximally avoided in general. Equally important is that remov- report, Shah et al. simulated the translation initiation in yeasts, ing these proteins from the cell is an energy-consuming task, proposing that ribosome availability is the primary rate- primarily taken by the ubiquitin–proteasome system (UPS), limiting factor of initiation; in addition, this model successfully autophagy, and exocytosis. Otherwise, unfolded protein predicted that the initiation probability is negatively correlated responses (UPRs) are being initiated and amplified.
to the gene length in yeasts (R = 0.56).3 But, it is seemingly Concerning this ‘‘wrong'' proteome formation, the error that contradictory that larger proteins are often considered to have occurred during translation is one of the most important more functions and favoured per evolution.9,10 For example, it is mechanisms; however, it is often overlooked or underestimatedwhen modelling the translation process. Specifically, ribosomepremature drop-off, frameshift and amino acid misincorpora- tion can take place at the probability of 103 to 104 in Institute of Life and Health Engineering, Key Laboratory of Functional ProteinResearch of Guangdong Higher Education Institutes, Jinan University, eukaryotic cells.12,13 Kinetic competition is one of the major Guangzhou 510632, China. E-mail: [email protected], causes of these errors (reviewed in ref. 14). Therefore, the [email protected]; Tel: +86-20-85225960, +86-20-85224031 cellular physiological condition changes, including transla- b Guangdong Experimental High School, Guangzhou, China tional demand, tRNA availability, and intracellular diffusion, † Electronic supplementary information (ESI) available: Fig. S1–S3. See DOI: can significantly alter the error rate.15 As to cancer cells, they ‡ These authors contributed equally to this work.
have remarkably higher tRNA concentrations than normal 370 Mol. BioSyst., 2015, 11, 370--378 This journal is The Royal Society of Chemistry 2015 Molecular BioSystems cells,16–18 with an approximately 6-fold increase of errors mouse ESCs and HeLa cells, the mRNA length minimally corre- during protein synthesis.19 The cell has to invest energy to lates to the mRNA half-life (R o 0.01, Fig. S2, ESI†), suggesting clear these defective proteins with aberrant sequences or struc- that the mRNA degradation rate is not a major contributor to tures that mitigate the fitness.20–23 To avoid this energy waste, the overall protein abundance. Our results emphasized that cells should have evolved a mechanism to balance the protein the mRNA length is a key factor of protein abundance that production quantity and quality.
is associated with translational initiation. This reciprocally To address our hypothesis, we proposed a model explaining confirmed that the length-dependent TR is a major determinant the length-dependent translation initiation efficiency that to shape the proteome abundance.
reflected the major determinants for shaping the length- It is notable that the same trend applies to the yeast (Fig. S3, dependent proteome. This model elucidated the contribution ESI†), suggesting that the translational preference towards of gene length to a functional proteome in terms of circulation smaller proteins may benefit organisms in a conserved way, efficiency in translation initiation and the translational error thus being maintained along with evolution.
rate. We computationally found the strongly negative correla-tion between the protein synthesis and the mRNA length that The mRNA circularization is the determinant process of length- leads to optimized energy consumption.
dependent translation initiation The cap-dependent translation initiation in eukaryotic cellscomprises the following major steps: (a) 43S pre-initiation complex formation (including 40S ribosome subunit, eIF2ternary complex, eIF3, eIF5); (b) cap-binding by the eIF4F Length-dependent translation is the major determinant of complex and poly-A binding by PABP; (c) mRNA circularization; length-dependent protein abundance (d) attachment of the 43S complex to mRNA and scanning of With human normal bronchial epithelial cells (HBE), lung the initiation codon; and (e) recruitment of the 60S ribosome cancer A549 and H1299 cells, and human colon cancer Caco- subunit (Fig. 2A and reviewed in ref. 29). In steady-state cells, 2 cells, we previously reported the direct analysis on the the concentration of all molecules is constant. Therefore, all of translation initiation efficiency, represented by the translation the binding kinetics are constant in a given environment (steps ratio (TR), defined as the abundance ratio of translating mRNA (a), (b), (d) and (e)). The only length-dependent step is the to mRNA for a certain gene.8,24 We further reported the mRNA circularization (step (c)).
translatome and transcriptome profiles of human liver cancer It is known that mRNA molecules can be modelled as worm-like cell lines, such as Hep3B cells.25 chain (WLC) with the persistence length of 0.79 nm.30 The average We then examined whether the length-dependence of trans- distance between the 50-cap and the 30-tail can be described as lation initiation efficiency shapes the proteome. In human 2PL, where P is the persistence length and L is the length genome, mRNA length significantly and positively correlates of the mRNA (Fig. 2B). The circularization is a passive diffusion to the CDS length (Fig. S1, ESI†), suggesting that longer mRNAs process with no energy input. Therefore, it follows the Brownian in general encode larger proteins. In human cells originated motion model.15,31 The tail of a longer mRNA needs more time to from various tissues, mRNA abundance minimally correlates to diffuse to the cap and thus is less efficient in the initiation.
the length with Pearson R values ranging from 0.10 to 0.30 We then simulated the diffusion of the tail complex to the cap (Fig. 1A). The lower TR for longer mRNAs leads to the more complex. The average length of the polyA tail is 250 adenosines,32 significant negative correlation of the RNC–mRNA and the and each PABP protein binds to 27 adenosines on average.33 mRNA length with remarkable greater R values (Fig. 1B). We Therefore, each tail complex consists of 9 PABPs and 250 adeno- previously showed that the RNC–mRNA strongly correlates to sines, with the molecular weight of 723 kDa. As the experimentally proteins in their abundance.8 In addition, it is known that the measured average density of protein is 1.37 g cm3,34,35 the molar protein synthesis is the major determinant of the protein volume of the tail complex can be calculated to be 5.13  abundance.26 Hence, we further posit a significant negative 105 cm2 mol1, and the radius is approximately 5.9 nm. The cap correlation between the protein abundance and the mRNA complex consists of eIF4A, eIF4B, eIF4E and eIF4G, with the total length. This was confirmed by analyzing the label-free mass molecular weight of 315.4 kDa, and the radius is approximately spectrometry data at least in cervical cancer HeLa cells,27 with 4.5 nm. The diffusion coefficient can be calculated as: R = 0.48 (Fig. 1C); in addition, we found that the protein degradation rate k deg has minor contribution to the protein abundance (R = 0.20, Fig. 1D). This result conformed to aspecific study with mouse embryonic stem cells (ECSs), proposing where kB is the Boltzmann constant, T is the temperature, rp is the that the protein degradation rate plays a minor role in shaping radius of the particle and Z is the viscosity. We used the experi- the proteome abundance, although the protein abundance is mentally measured viscosity of Z = 0.88  103 Pa s in HeLa cells.36 determined by both the synthesis rate and the degradation rate.26 Therefore, the diffusion coefficients of the tail complex and the We further examined the length dependency of the mRNA cap complex are 4.37  1011 m2 s1 and 1.72  1010 m2 s1, degradation rate, which also serves as a factor that regulates respectively, which are comparable to the other protein–RNA protein production efficiency.28 In the yeast, Arabidopsis thaliana, complexes, such as the ternary complex of tRNA.37 The Brownian This journal is The Royal Society of Chemistry 2015 Mol. BioSyst., 2015, 11, 370--378 371

Molecular BioSystems Translation initiation efficiency is a primary determinant of the length-dependent proteome abundance. (A) mRNA abundance versus mRNA length correlation in human HBE, Hep3B and HeLa cells. (B) RNC–mRNA abundance versus mRNA length correlations. (C) Protein abundance versusmRNA length correlation in HeLa cells. (D) Protein degradation rate versus mRNA length correlation in HeLa cells.
motion simulation was performed by using these parameters average to loop, comparable to the initiation probability of long and at the time interval of 0.04 ms. Circularization was detected genes in yeasts (233 seconds).3 After mRNA circularization, step when the center of the tail complex and the cap complex are no (d) and step (e) happen with the rate constant k 4 3 min1 and larger than the sum of their radii.
k 4 1 min1 in the rabbit reticulocyte in vitro translation The simulation result showed that the frequency of mRNA system, respectively,38 comparable to k1 = 0.076 s1 and k2 = circularization decreases almost exponentially against mRNA 0.019 s1 measured in Saccharomyces cerevisiae.39 Therefore, for length within the range of 300–10 000 nt (Fig. 2C). The prob- long mRNAs, the rate-limiting step is the circularization.
ability of a successful collision (forming the stable mRNA loop We also performed simulation assuming various viscosities and ready to recruit 43S pre-initiation complex) has not been (Fig. 2D). The results showed that higher diffusion coefficient reported in the literature; yet we refer to the probability of caused by lower viscosity facilitates the circularization for all successful collision of the ternary complex to deliver amino mRNAs. Thus, with this model, we explained the mechanism of acids on the ribosome to be 1.96%.15 Therefore, longer mRNAs the translational preference of smaller mRNAs in terms of need considerably extended time to form the loop for trans- circularization, which further orchestrates the protein abun- lation. For example, a 7000 nt mRNA needs 387 seconds on dance also in a length-dependent manner.
372 Mol. BioSyst., 2015, 11, 370--378 This journal is The Royal Society of Chemistry 2015 Molecular BioSystems Model and simulation of the mRNA circularization during the translation initiation. (A) Major steps of translation initiation. (B) Model of mRNA and the initial setting of the simulation. Cap complex and the tail complex are linked by mRNA modelled as worm-like chain (WLC). (C) Simulated collisionfrequency of the cap complex and the tail complex versus the mRNA length. (D) Successful circularization time versus mRNA length under differentviscosities (Z).
Energy consumption and functionality of a proteome is approximately 96.5% of all of the human proteins according to balanced by the translational preference of shorter mRNAs the NCBI RefSeq database. We then investigated proteins In steady-state eukaryotic cells, the size of most proteins distri- within this range and defined a parameter, amount ratio bute between 100 and 3000 amino acids, which consists of (AR), to describe the ratio between the synthesis quantity of This journal is The Royal Society of Chemistry 2015 Mol. BioSyst., 2015, 11, 370--378 373 Molecular BioSystems the largest and the smallest protein categories: AR = A100/A3000, We thus can calculate the ratio between the total function where An means the synthesis quantity of the proteins with score of the proteome and the energy consumption: length n. The synthesis amount of the proteins between 100 and 3000 amino acids was logarithmically interpolated.
Incorporation of each amino acid into the growing nascent polypeptide chain costs one GTP (one unit of energy). There- fore, the total energy needed for synthesizing the proteins with A higher r value represents higher energy efficiency for a proteome functionality that benefits the organism.
We calculated r values under f p L0, L1/3 and L2/3, respec- tively (Fig. 3A). In all cases, the r values increase with the AR, Errors occur during the synthesis of proteins, including showing that higher translation efficiency of smaller proteins the misincorporation, frameshift, premature drop-off and provides higher energy efficiency to gain the proteome func- misfolding. The probability of these errors is proportional tionality (Fig. 3A). We calculated the AR in steady-state HeLa to the protein length.12,15,40 We sum all of these prob- cells from the actual proteome data27 to be 206.5 (marked in abilities as the error probability per amino acid e. Therefore, Fig. 3A), approaching the plateau. Consistently, an AR that is the total error probability to synthesize proteins with length higher than 206.5 will not remarkably affect the changing rate of r, with the dr/dAR value less than 106 in HeLa cells (Fig. 3B).
Therefore, higher AR can be beneficial within certain dynamic And the amount of functional (error-free) proteins with Upregulation of the translational preference of small proteins is potentially beneficial for cancer cells M = Ai(1  Perr) = Ai(1  e)Li Due to the higher tRNA concentration, the translation elonga-tion speed is remarkably higher in cancer cells.17 This leads to The misfolded or damaged proteins have to be actively higher error rates especially on folding, increasing the e in the removed by UPS and/or other protein quality control mechan- model described above. We therefore calculated the curves isms, which are all energy-consuming bioprocesses. We here of AR versus r at increasing e (Fig. 3). In all cases of f p L0, focused on UPS to computationally elucidate and exemplify L1/3 and L2/3, the r value decreases with the increase in the error such energy consumption. ATP serves as the energy source of rate, given the same AR. An increase in AR can increase the UPS-mediated protein degradation, and energy consumption in r value and thus counteracts the negative effect of larger e, ATP units is approximately 1/3 of the protein length in the which is potentially beneficial to the cancer cells (Fig. 3C). With unfolding phase and equal to the protein length in the degra- the active translation, cancer cells can generate enough amount dation phase.23 On average, the number of ATPs used during of functional large proteins even at higher AR.
the degradation is slightly more than the protein length.23Thus, the energy used for degradation by UPS can be approxi-mately described as: Edeg,i = Li AiPerr = Li Ai[1  (1  e)Li] We here simulated that shorter mRNA length considerablydeceases circularization time, and reduces the translational A well-folded protein will have functional sites to interact error dependent energy consumption. We believe the two with other molecules to carry out bioprocesses. We define a mechanisms should be the outcome of evolution, determining ‘‘function score'', f, to quantitatively reflect the functions of a the length-dependent translation initiation in steady-state protein molecule. The higher f a protein has, the more it can be human cells. Equally important is that they are helpful to involved in bioprocesses. Protein length may correlate to f with understand why human cancer cells often amplify such a one or more of the following types: (a) f is irrelevant to the length-dependent preference,8 namely, cancer cells tend to length, i.e. f p L0, which corresponds to those proteins that do hijack this evolutionary benefit to counteract the energy not interact with any other proteins; (b) f p volume p L1/3. The demand caused by their high translation error rates.
greater volumes of such proteins could accommodate more There are very few models published to explain the global structural domains with substrate binding pockets for catalysis length-dependent feature of translation. Our findings further and regulation, which facilitate the domain cooperation;41,42 demonstrated that the rate-limiting step of circularization is at and (c) f p area p L2/3. The functionality of these proteins least one of the determinants of such a bioprocess in human relies on the surface interaction with other proteins.43 Regard- cells. For certain, we should not overlook numerous gene- less of the type, the total function score of the proteins with specific studies in the field, implicating other regulatory length Li can be calculated as: mechanisms, such as the antisense-RNA (e.g. microRNA),sequence/structure-specific 3'-UTR binding proteins, upstream Fi = Mi fi = Ai(1  e)Lifi microORFs, and the specific sequence around the start codon 374 Mol. BioSyst., 2015, 11, 370--378 This journal is The Royal Society of Chemistry 2015 Molecular BioSystems AR versus r values at different error rates (0.1–0.8%). (A) The calculation was performed assuming f = L0, L1/3 and L2/3, respectively. (B) Derivatives of the curves calculated in (A). (C) Schematic illustration of the benefit of elevating translational preference in cancer cells. In cancer cells, the error rate ishigher, therefore the r value decreases if the AR remains constant (route a). Elevating AR will increase the r value (route b), compensating for the highererror rate.
and upstream regions44,45 (reviewed in ref. 46–48). To be noted, has minor contributions to the mRNA–protein abundance we previously found that the longer mRNA produces less correlation in yeast, mouse and human cells (0.2%, o10% and molecules of proteins before it is degraded.28 Considering o5%, respectively).26,52 Therefore, the mRNA stability was not the mRNA stability, the attached ribosomes protect mRNAs included in our model.
from endonucleic degradation in prokaryotes that is also Indeed, energy efficiency is a selective force during the termed as the ribosome shielding effect.49 While in eukaryotes, evolution. Recent evolutionary studies on genetic codes have mRNAs are largely degraded via decapping and exonucleic considered the translational fidelity, i.e. the translation error digestion.28,50 This feature allows longer genes in eukaryotes caused by the misincorporation of amino acids and its conse- than prokaryotes.28 Although previous studies suggested that quent energy penalty.13,14,53,54 However, this type of error is of inhibition on translation accelerates mRNA decay under stress low probability, ca. 103 to 104 in different types of eukaryotic conditions (reviewed in ref. 50), recent studies have shown that cells (reviewed in ref. 13). In comparison, the misfolded newly- the mRNA half-life correlates weakly to the ribosome density synthesized proteins can be a problem that costs more energy: and the number of ribosomes per mRNA (Spearman r = 0.36), 30% of the newly synthesized proteins are immediately at least in yeasts and under physiological conditions. This degraded due to their improper conformation (reviewed in indicated that the ribosome density has a minor effect on ref. 23). Since the chaperon-dependent folded proteins consist mRNA stability under physiological conditions.51 Further studies of only a lesser fraction of the proteome,55 the majority of from other groups experimentally found that the mRNA stability misfolding happens co-translationally under physiological This journal is The Royal Society of Chemistry 2015 Mol. BioSyst., 2015, 11, 370--378 375 Molecular BioSystems conditions, which is highly relevant to the translation elonga- cellular stiffness, in accordance with their less-organized cyto- tion speed (reviewed in ref. 5). Furthermore, translational skeleton. This has been suggested to accelerate nutrient uptake, pausing is more important to the larger proteins, as the facilitate cellular deformation and thus to increase the meta- number of pausing sites increase with the protein length.22,40 static potential of cancer cells56,57 (reviewed in ref. 58). Second, Therefore, these co-translationally misfolded proteins and viscosity is one of the major determinants of diffusion.59 In the subsequent clearance of these proteins are a primary source the crowding cytoplasmic environment, increased cytoplasmic of energy waste in protein synthesis; however, we would viscosity leads to significantly decreased diffusion coefficients of re-emphasize that it was overlooked in the previously reported molecules, and the amplitude of such diffusion coefficient length-dependent translation models. Our model successfully decreases of large molecules is much higher than that of the supported such an argument in terms of higher energy effi- small molecules.60 The formation of the 43S pre-initiation ciency by translating shorter mRNAs. We believe that the novel complex is also inhibited at higher cellular viscosity,61 leading and essential features of this energy model can be summarized to inhibited translation initiation, while with less effects on as: (1) we considered the errors during the protein synthesis translation elongation.15 These facts implicate that higher and (2) we assessed the benefit in the view of energy efficiency viscosity severely extended the time needed for all steps in the of maintaining a steady-state proteome.
translation initiation except mRNA circularization (Fig. 4A).
Our study suggested that the down-regulation of the cytoplasmic Considering these effects, lower cytoplasmic viscosity amplifies viscosity in cancer cells can considerably augment the translational higher translational preference towards small proteins (Fig. 4B preference of small proteins, which is energetically beneficial to and C). In this regard, our study provided new insights and cancer cells. To explain the mechanism of this phenomenon, we system connections between cellular mechanics, length-dependent propose that the viscosity difference between normal and cancer system regulation of the whole translatome and malignant cells should be an important determinant. First, it is known that phenotypes of cancer, which may give hints to understand cancer cells tend to have lower cytoplasmic viscosity to soften their cancer cells from a new perspective.
Proposed explanation of the possible contribution of viscosity to augmented translational preference on smaller proteins in cancer cells.
(A) Definition of the time of major translation initiation steps. (B, C) The total time for initiation versus mRNA length. The situations at low (B) and high (C)viscosities are shown, respectively. The time needed for other steps except tc are totalized as T = tb + td + te, shown as the black dotted line. T isindependent of length, but dependent on viscosity per diffusion model. Therefore, the ratio of the total translation initiation time for large to smallproteins is greater in the scenario of low viscosity than that in high viscosity, resulting in augmented translational preference towards small proteins.
376 Mol. BioSyst., 2015, 11, 370--378 This journal is The Royal Society of Chemistry 2015 Molecular BioSystems 24 J. Zhong, Y. Cui, J. Guo, Z. Chen, L. Yang, Q. Y. He, G. Zhang and T. Wang, J. Proteome Res., 2014, 13, 50–59.
This work was collectively supported by the National Science 25 C. Chang, L. Li, C. Zhang, S. Wu, K. Guo, J. Zi, Z. Chen, Fund of China (81322028, 81372135), National High-Tech R&D J. Jiang, J. Ma, Q. Yu, F. Fan, P. Qin, M. Han, N. Su, T. Chen, (863) Program of China (2014AA020504), the Key Project of K. Wang, L. Zhai, T. Zhang, W. Ying, Z. Xu, Y. Zhang, Y. Liu, Chinese Ministry of Education (212207), and Research Fund for X. Liu, F. Zhong, H. Shen, Q. Wang, G. Hou, H. Zhao, G. Li, the Doctoral Program of Higher Education (20124401120008).
S. Liu, W. Gu, G. Wang, T. Wang, G. Zhang, X. Qian, N. Li,Q. Y. He, L. Lin, P. Yang, Y. Zhu, F. He and P. Xu, J. ProteomeRes., 2014, 13, 38–49.
J. Schuchhardt, J. Wolf, W. Chen and M. Selbach, Nature, 1 R. Duncan, S. C. Milburn and J. W. Hershey, J. Biol. Chem., 2011, 473, 337–342.
1987, 262, 380–388.
27 S. B. Cambridge, F. Gnad, C. Nguyen, J. L. Bermejo, M. Kruger 2 Y. Arava, Y. Wang, J. D. Storey, C. L. Liu, P. O. Brown and and M. Mann, J. Proteome Res., 2011, 10, 5275–5284.
D. Herschlag, Proc. Natl. Acad. Sci. U. S. A., 2003, 100, 28 A. Valleriani, G. Zhang, A. Nagar, Z. Ignatova and R. Lipowsky, Phys. Rev. E: Stat., Nonlinear, Soft Matter Phys., 3 P. Shah, Y. Ding, M. Niemczyk, G. Kudla and J. B. Plotkin, 2011, 83, 042903.
Cell, 2013, 153, 1589–1601.
29 R. J. Jackson, C. U. Hellen and T. V. Pestova, Nat. Rev. Mol.
4 C. E. Aitken and J. R. Lorsch, Nat. Struct. Mol. Biol., 2012, 19, Cell Biol., 2010, 11, 113–127.
30 F. Vanzi, Y. Takagi, H. Shuman, B. S. Cooperman and 5 G. Zhang and Z. Ignatova, Curr. Opin. Struct. Biol., 2011, 21, Y. E. Goldman, Biophys. J., 2005, 89, 1909–1919.
31 M. Dlugosz and J. Trylska, BMC Biophys., 2011, 4, 3.
6 C. Branco-Price, R. Kawaguchi, R. B. Ferreira and J. Bailey- 32 M. Brook, J. W. Smith and N. K. Gray, Reprod., 2009, 137, Serres, Ann. Bot., 2005, 96, 647–660.
7 L. Ciandrini, I. Stansfield and M. C. Romano, PLoS Comput.
33 D. A. Mangus, M. C. Evans and A. Jacobson, Genome Biol., Biol., 2013, 9, e1002866.
2003, 4, 223.
8 T. Wang, Y. Cui, J. Jin, J. Guo, G. Wang, X. Yin, Q. Y. He and 34 P. G. Squire and M. E. Himmel, Arch. Biochem. Biophys., G. Zhang, Nucleic Acids Res., 2013, 41, 4743–4754.
1979, 196, 165–177.
9 J. Zhang, Trends Genet., 2000, 16, 107–109.
35 K. Gekko and H. Noguchi, J. Phys. Chem., 1979, 83, 2706–2714.
10 L. Brocchieri and S. Karlin, Nucleic Acids Res., 2005, 33, 36 T. Kalwarczyk, N. Ziebacz, A. Bielejewska, E. Zaboklicka, K. Koynov, J. Szymanski, A. Wilk, A. Patkowski, J. Gapinski, 11 J. A. Gerlt, Nat. Biotechnol., 2002, 20, 786–787.
H. J. Butt and R. Holyst, Nano Lett., 2011, 11, 2157–2163.
12 C. G. Kurland, Annu. Rev. Genet., 1992, 26, 29–50.
37 A. Werner, Nucleic Acids Res., 2011, 39, e17.
13 M. Lynch, Trends Genet., 2010, 26, 345–352.
38 J. R. Lorsch and D. Herschlag, EMBO J., 1999, 18, 6705–6717.
14 I. Wohlgemuth, C. Pohl, J. Mittelstaet, A. L. Konevega and 39 M. G. Acker, B. S. Shin, J. S. Nanda, A. K. Saini, T. E. Dever M. V. Rodnina, Philos. Trans. R. Soc. London, Ser. B, 2011, and J. R. Lorsch, J. Mol. Biol., 2009, 385, 491–506.
366, 2979–2986.
40 G. Zhang, M. Hubalewska and Z. Ignatova, Nat. Struct. Mol.
15 G. Zhang, I. Fedyunin, O. Miekley, A. Valleriani, A. Moura Biol., 2009, 16, 274–280.
and Z. Ignatova, Nucleic Acids Res., 2010, 38, 4778–4787.
41 H. Hegyi and M. Gerstein, Genome Res., 2001, 11, 1632–1640.
42 R. S. Wang, Y. Wang, L. Y. Wu, X. S. Zhang and L. Chen, M. R. Rosner and T. Pan, Nucleic Acids Res., 2009, 37, BMC Bioinf., 2007, 8, 391.
43 J. Warringer and A. Blomberg, BMC Evol. Biol., 2006, 6, 61.
17 A. Ujvari, R. Aron, T. Eisenhaure, E. Cheng, H. A. Parag, 44 N. T. Ingolia, S. Ghaemmaghami, J. R. Newman and Y. Smicun, R. Halaban and D. N. Hebert, J. Biol. Chem., J. S. Weissman, Science, 2009, 324, 218–223.
2001, 276, 5924–5931.
45 K. Bentele, P. Saffert, R. Rauscher, Z. Ignatova and 18 Y. Zhou, J. M. Goodenbour, L. A. Godley, A. Wickrema and N. Bluthgen, Mol. Syst. Biol., 2013, 9, 675.
T. Pan, Biochem. Biophys. Res. Commun., 2009, 385, 160–164.
46 J. W. B. Hershey, N. Sonenberg and M. B. Mathews, Cold 19 M. C. Luce, K. D. Tschanz, D. A. Gotto and C. L. Bunn, Spring Harbor Perspect. Biol., 2012, 4, a011528.
Biochim. Biophys. Acta, 1985, 825, 280–288.
47 J. Jia, P. Yao, A. Arif and P. L. Fox, Curr. Opin. Genet. Dev., 20 U. Schubert, L. C. Anton, J. Gibbs, C. C. Norbury, 2013, 23, 29–34.
J. W. Yewdell and J. R. Bennink, Nature, 2000, 404, 770–774.
48 E. Huntzinger and E. Izaurralde, Nat. Rev. Genet., 2011, 12, 21 K. C. Keiler, P. R. Waller and R. T. Sauer, Science, 1996, 271, 49 C. Deneke, R. Lipowsky and A. Valleriani, Phys. Biol., 2013, 22 I. Fedyunin, L. Lehnhardt, N. Bohmer, P. Kaufmann, G. Zhang and Z. Ignatova, FEBS Lett., 2012, 586, 3336–3340.
50 S. Huch and T. Nissan, Wiley Interdiscip. Rev.: RNA, 2014, 5, 23 A. L. Goldberg, Nature, 2003, 426, 895–899.
This journal is The Royal Society of Chemistry 2015 Mol. BioSyst., 2015, 11, 370--378 377 Molecular BioSystems 51 S. Edri and T. Tuller, PLoS One, 2014, 9, e102308.
57 Y. Bishitz, H. Gabai, P. Girshovitz and N. T. Shaked, 52 G. Wu, L. Nie and W. Zhang, Curr. Microbiol., 2008, 57, J. Biophotonics, 2013, 7, 624–630.
58 D. Wirtz, K. Konstantopoulos and P. C. Searson, Nat. Rev.
53 A. S. Novozhilov, Y. I. Wolf and E. V. Koonin, Biol. Direct, Cancer, 2011, 11, 512–522.
2007, 2, 24.
59 M. C. Konopka, I. A. Shkel, S. Cayley, M. T. Record and 54 P. Shah and M. A. Gilchrist, PLoS Genet., 2010, 6, e1001128.
J. C. Weisshaar, J. Bacteriol., 2006, 188, 6115–6123.
55 M. J. Kerner, D. J. Naylor, Y. Ishihama, T. Maier, 60 J. T. Mika, G. van den Bogaart, L. Veenhoff, V. Krasnikov H. C. Chang, A. P. Stines, C. Georgopoulos, D. Frishman, and B. Poolman, Mol. Microbiol., 2010, 77, 200–207.
M. Hayer-Hartl, M. Mann and F. U. Hartl, Cell, 2005, 122, 61 M. Brigotti, P. G. Petronini, D. Carnicelli, R. R. Alfieri, M. A. Bonelli, A. F. Borghetti and K. P. Wheeler, Biochem.
56 J. Farrell, Med. Hypotheses, 1988, 25, 119–123.
J., 2003, 369, 369–374.
378 Mol. BioSyst., 2015, 11, 370--378 This journal is The Royal Society of Chemistry 2015

Source: http://bioinformatics.jnu.edu.cn/translatomics/Publications/MBS_GJM.pdf