1. Codon Bias and Regulation of Translation among Bacteria and Phages
Thesis defense of
Marc BAILLY-BECHET
Advisor: Massimo VERGASSOLA
Institut Pasteur, Dept Genomes & Genetics, Unit « In Silico » Genetics

2. Summary
Introduction to the bacterial translation system and the codon bias
Structuration of the bacterial chromosomes by codon bias domains
Why tRNAs in phages?

3. Translation processes in prokariotuc cells

4. Transfer RNA
tRNAs are the small RNAs that link an amino-acid to the peptide sequence
They have a special palindromic structure
They are amino acid specific AND codon « specific » (wooble)
They differ greatly in number in the cell (from ~100 to ~5000 for a given amino acid)

5. Degeneracy of the genetic code

6. Differential usage of synonymous codons at the genome scale

7. Causes of the codon bias
Non-selective causes of the codon bias
Mutation biases (e. g. towards high/low G+C)
Strand bias on the chromosome (GT bias)
Selective causes of the codon bias:
Translation efficiency
Translation accuracy
Codon-anticodon selection ?
Codon robustness ?

8. tRNA concentration correlates to codon bias
Dong et al. (1996) J. Mol. Biol. 260:649

9. Codon bias domains over bacterial chromosomes

10. Motivations of the project
Aim: clustering the genes of an organism according to their codon bias
Biological interests:
Functional analysis of the groups of genes
Role of codon bias in the chromosome structuration
Comparison of the genome organization between species
Inference of some codon bias causes from the classification

11. Previous results Methods: correspondance analysis
2 main sub-groups of genes identified in multiple organisms:
Highly expressed
Horizontal transfer genes
Methodological difficulties:
Choice of the number of groups
Choice of the distance
Kunst et al. (1997), Nature 390:249

12. Key idea about the method: the optimization criteria
Each group is defined by the probability distribution of codon usage generated by the genes it contains
A good classification is one which maximize the gain of information on these probability distributions, relative to a uniform prior distribution

13. The clustering algorithm
…….
Threshold
C =40
…….

14. Key idea about the method: selection of the number of groups
The good number of groups is the one maximizing the average stability of genes attribution inside the groups, relative to the expected stability in absence of structure (random case)

15. Number of groups and clustering significance

16. Codon usage inside the groups

17. Tests of the algorithm

18. Gene function is correlated with codon bias
Highly expressed genes, translation and ribosomal proteins : COG J (9/22).
Unknown genes, pathogenicity islands and horizontally transfered genes : COG - (17/19).
Metabolism (synthesis & transport) : COG C (4/6), E (7/4) et F (7).
Membrane and carbohydrate metabolism genes : COG G (6) et M (3/3).
B. subtilis only -- Motility genes : COG N (5).

19. Anabolic genes are grouped on the lagging strand

20. Replication and transcription machineries collisions
Anabolic genes are usually transcribed when no replication occurs
=> being on the lagging strand is not counter-selected.
Mirkin & Mirkin (2005) Mol Cell Biol. 25(3): 888

21. Codon bias domains

22. Group by group analysis : influence of the GC%

23. Acknowledgements (I)
Frank Kunst and all the GMP Team

24. Why tRNAs in phages?

25. What’s a phage?

26. Motivations of the project
Understanding the presence of tRNAs inside bacteriophages
Correlation to the host or phage codon bias?
Differences between lytic and temperate phages?
Selection acting on tRNA acquisition and implications for phage evolution?

27. Acquisition of tRNA sequences by bacteriophages
Lysogenic phages are known to insert in microbial genomes in tRNA sequences
=> Imprecise excision could explain the acquisition of tRNA sequences
Lytic phages cause liberation of the host genetic material after cell lysis
=> Acquisition of tRNAs sequences in the surrounding media or neighbour hosts

28. Datas Beginning : 200 DNA phage genomes, 23 hosts, 240 tRNAs
Taken out :
Non sequenced hosts
Phages genomes without tRNAs
tRNAs inserted in prophagic regions
Phages having tRNAs their host do not have
Final dataset :
37 phages, 15 hosts, 169 tRNAs
(6 duplicates, 1 triplet)

29. tRNA distribution in phages

30. Correlation of host and phages codon bias
< R > = 0.77  0.27 real data
< R > = 0.38  0.42 phage-random host
=> Codon usage is correlated between the host and the phage
< R > = 0.83  0.14 real data - Temperate
< R > = 0.61  0.39 real data - Lytic
=> The correlations are higher in temperate phages

31. Phage codon frequency distribution is related to tRNA content

32. First conclusions
Lytic phages have a codon usage less similar to the one of their hosts when compared to temperate phages
Lytic phages have more tRNAs than temperate ones
Codon usage is more biased in lytic phages than in temperate ones
Both seem to have tRNAs corresponding to the codons they use more

33. Random uptake hypothesis
tRNA content of host matches codon bias
Codon bias of phage matches the host one’s
=> No need for the phage to have tRNAs !
Random uptake hypothesis: the tRNA content of a phage should be proportional to its host tRNA content, and so would be indirectly correlated to the codon bias of the phage

34. Statistical tests of the random uptake hypothesis
Significance for high values of <f>: p = 0.68
No specific enrichment in tRNAs for the phage high frequency codons
Significance for high values of <∆f>: p <
Significant enrichment in tRNAs for the codons the phage uses more than its host

35. Modelisation of the acquisition and loss processes
Gain
Loss

36. Inference of the parameters by maximum likelihood
Probability
Likelihood of
the real data,
given the model
Maximum likelihood
Most probable

37. Evolutive processes tested
Selection based on:
Frequency of usage of the corresponding codon in the phage genome (+)
Frequency of usage of the corresponding codon in the host genome (-)
Difference of codon usage frequencies between phage and host genome (+)
Duplication of tRNA on the phage genome

38. Master model equation results
Selection based on the phage frequency of codon usage is non significant (p=0.15)
Selection based on the rarity of the codon in the host genome is slightly significant (p=0.018 before Bonferroni correction)
Selection based on the difference of frequencies of codon usage between phage and host is highly significant (p<2.10-7)
The tRNA duplication hypothesis has to be rejected

39. Adaptative selection of tRNAs?
Selection relative to the phage codon usage only could lead to a static tRNA content, and could be non-optimal after an host change
Selection relative to the host codon usage only does not take into account the quick phage sequence evolution
Selection needs to take both into account to be adaptative and gives rise to a useful tRNA content

40. Conclusions
Translational selection is a strong pressure acting on phage tRNA content
tRNA content among phages is optimized to compensate for differences between host and phage codon usage
This pressure is more important in lytic phages

41. Acknowledgements (II)
Massimo Vergassola
Eduardo Rocha
The committee members
Yves Charon
Guillaume Cambray
Aymeric Fouquier d’Herouel
All the family and friends who came today!

42. Supp. Mat. Part 1

43. Codons probability distributions

44. Tests of the algorithm (II)
High CAI genes share the same codon bias:
32/59 in group 1 of B. subtilis
33/33 in group 1 of E. coli
Genes in the same operon or pathway tend to belong to the same group

45. Transcription and translation
From Miller et al., 1970, Science 169:392

46. Translation regulation and synchronization by tRNA recycling
Gene Gene Gene 3

47. Recycling phenomenon analysis
On average, tRNA recycling should not increase translation speed
Recycling could induce a coupling between close ribosomes, allowing for protein synthesis synchronization
Synthetases are the limiting factor as they prevent in most cases a tRNA used by a ribosome to be re-employed by another close one

48. Supp. Mat. Part 2

49. Phage codon frequency distribution is related to tRNA content

50. Master equation model (I)
Random excision

51. Modelisation of the acquisition and loss processes (II)
Gain
Loss

52. Master equation models (II)
Random excision
Random excision + selective loss
Random excision + selective loss + random copy

53. Selection is significant event relative to random hosts