1. A genome-wide perspective on translation of proteins Dec 2012 Regulatory Genomics Lecturer: Prof. Yitzhak Pilpel

2. Multiple codons for the same amino acid: opportunities for sophisticated control C1 C2 C3 C4 C5 C6 Serine: UCU UCC UCA UCG AGC AGU Cysteine: UGU UGC Methionine: UGG STOP: UAA, UAG UGA 2

3. W i /W max if W i  0 w i = w mean else { dos Reis et al. NAR 2004 The tRNA Adaptation Index (tAI) ATCCCAAAATCGAAT … … … A simple model for translation efficiency Wobble Interaction 3

4. Codon usage bias is correlated with translation efficiency r=-0.79 (p<0.001) Mutation pattern (neutral) Selection Codon bias

5. Selection of codons might affect: Accuracy Throughput Costs Folding RNA-structure

6. Kinetic proofreading – the problem c + CAAc d dC AAd + cC C – codon c – Cognate tRNA d – None-cognate tRNA cC, dC – tRNA codon pair AAc, AAd, correct and wrong amino acid Assumptions: kc=kd, k’d=100*k’c vd=vc Error rate, Fo= AAd/AAc = dC/cC = k’c/k;d= 0.01 Yet in reality error rate is 10^-4… How can we explain 100 times more accurate translation kc k’c vc vd kdk’d

7. Kinetic proofreading – the solution c + CAAc d dC AAd + cC C – codon C* – Modified Cognate tRNA d* – Modified None-cognate tRNA C*C, d*C – tRNA codon pair m- rate of tRNA modification l’c, l’d – rate of unidirectional reaction in which the modified tRNAs leave the site Assumption: l’c*100=l’d Corrected Error rate, F= (k’c/k’d) (l’c/l’d)= 10^-4 kc k’c vc vd kdk’d c*C m d*C m l’c l’d

8. C c d dC cC Fo d*C c*C Free energy l’c l’d k’c k’d C cC d The energy landscape of kinetic proofreading

10. Analogy: from mating in yeast A protease might help yeast find the right mate by degrading the signal and the noise Barkai et al. Nature 1998 No protease: large errorWith protease: small error

11. Selection of codons might affect: Accuracy Throughput Costs Folding RNA-structure

12. Programmed “errors” in amino acid loading on tRNA in stress Oxygen radicals – toxic !!! A non-Met tRNA As a result, in times of oxidative stress the cell is more protected since it has more Met residues in its proteins !! (Netzer et al Nature 2009)

13. Open questions Does select act to tune the “desired” error rate? How can we find places in genes where high error rate is selected for/against? Does controlling tRNA availability serve as a means to control error rate? Are there additional factors (e.g. location of codon within a gene)?

14. Selection of codons might affect: Accuracy Throughput Costs Folding RNA-structure

15. W i /W max if W i  0 w i = w mean else { dos Reis et al. NAR 2004 The tRNA Adaptation Index (tAI) ATCCCAAAATCGAAT … … … A simple model for translation efficiency Wobble Interaction 15

16. Correlation does not imply causality!! r=0.63 Predicted translation efficiency Measured protein abundance (Ghaemmaghami et al. Nature 2003) Physiological Evolutionary Physiological Z

17. Yet, mRNA also correlates with tAI… and with protein levels… * * * * * * * tAI mRNA

18. tAI and protein levels correlate even among gene populations with same mRNA levels

19. Deletion of a gene with a duplicate has smaller effect on phenotype (Gu et al Nature 2002) Single-copy genes have a tendency to be essential Genes with duplicates tend to be more dispensable

20. As always, correlation doesn’t guarantee causality DuplicationDispensability Z(??)

21. A synthetic library of GFP variants Kudla et al. Science 2009

22. No correlation between CAI and protein expression Protein abundance

23. Correlation does not imply causality!! r=0.63 Predicted translation efficiency Measured protein abundance (Ghaemmaghami et al. Nature 2003) Physiological Evolutionary Physiological

24. Tight RNA structure reduce translation Protein abundance

25. The tightness at the 5’ matters

26. So if codon usage doesn’t affect protein level, what does effect such levels? – It’s the RNA structure and its tightness!

27. Is tightness important throughout, or just at particular locations? ?

28. Natural sequences too show relaxed structure at 5’ (Tuller PNAS 2010) Structural tightness Structural tightness

29. Yet, mRNA structure doesn’t predict expression at all Structural Tightness Protein/mRNA

30. Bioinformatics vs. synthetic biology Bioinformatics Hundreds of thousands of genes All passed through natural selection Synthetic biology Variability is controlled (few confounding factors)

31. No correlation between CAI and protein expression Protein abundance

32. Towards more sophisticated translation efficiency models