Energetic principles of stop codon reeading on the ribosome · Principles of stop codon reading on the ribosome Johan Sund1, Martin Andér1 and Johan Åqvist1 1Department of Cell

Principles of stop codon

reading on the ribosome

Johan Sund1, Martin Andér1 and Johan Åqvist1

1Department of Cell and Molecular Biology, Uppsala University,

Biomedical Center, Box 596, SE-751 24 Uppsala, Sweden

Summary word count = 235

Main text word count = 1885

In termination of protein synthesis the bacterial release factors RF1 and RF2 bind

to the ribosome by specific recognition of mRNA stop codons and trigger hydrolysis

of the bond between the nascent polypeptide and P-site tRNA, thereby releasing the

newly synthesized protein. The release factors (RFs) are highly specific for a U in

the first stop codon position and recognize different combinations of purines in the

second and third positions, with RF1 reading UAA and UAG and RF2 reading UAA

and UGA. With recently determined crystal structures of termination complexes it

has become possible to decipher the energetics of stop codon reading by

computational analysis and to clarify the origin of the high RF binding accuracy.

Here, we report molecular dynamics free energy calculations on fourteen different

cognate and non-cognate termination complexes. The simulations quantitatively

explain the basic principles of decoding in all three codon positions and reveal the

key elements responsible for specificity of the RFs. The overall reading mechanism

involves hitherto unidentified interactions and recognition switches that cannot be

described in terms of a tripeptide anticodon model. Further simulations of

complexes with tRNATrp explain the observation of a “leaky” stop codon and

highlight the fundamentally different third position reading by RF2, that leads to a

high stop codon specificity with strong discrimination against the tryptophan codon.

The simulations clearly illustrate the versatility of codon reading by protein which

goes far beyond tRNA mimicry.

Correct termination of protein synthesis is essential for life and must be achieved with

high fidelity. Premature termination of translation at sense codons, leading to

dysfunctional proteins, is rare and occurs much less frequently in bacteria than

misreading of sense codons by near-cognate tRNAs1. Unlike peptide elongation2,3 high

termination fidelity is achieved without the help of a proofreading mechanism4. The

accuracy of RF induced termination originates from discrimination against sense codons

both in terms of RF binding to the ribosome (KM) and the catalytic rate of peptidyl-tRNA

hydrolysis (kcat), where binding normally is the most important factor4. The action of

several of the translation factors has been interpreted in terms of tRNA mimicry5,6,

2

including the class I RFs7, and a key question is whether such mimicry is relevant for

mechanistic details or merely limited to shape similarity.

The class I RFs are multidomain proteins where domain 2 recognizes the stop codon at

the ribosomal decoding site (Fig. 1a). The universally conserved GGQ motif in domain 3

is thereby positioned so that its glutamine residue enters the peptidyl transferase center

(PTC) some 80Å away, triggering hydrolysis of the peptidyl-tRNA bond8. Structural data

for termination complexes at successively higher resolution have been reported9-11 with

medium-resolution (3-3.5 Å) structures recently obtained12-14. These complexes showed

the protein-RNA interactions both at the decoding site and in the PTC, confirming

computational predictions of the role of the GGQ motif in termination15. Further, the

three “monitoring” 16S rRNA bases that recognize correct base pairing in cognate tRNA

complexes16, thereby enhancing fidelity17, do apparently not play such a role in

termination12-14. Hence, compared to codon reading by tRNA, it seems that RF

interactions themselves provide the required discriminatory power.

These new crystallographic complexes provide the data required for a quantitative

understanding of stop codon reading accuracy in termination. This, however, necessitates

evaluation of the interaction energetics involved. Pioneering mutational studies identified

tripeptide motifs (PxT in RF1 and SPF in RF2) as determinants of RF specificity and

indicated that substitutions of other amino acids did not affect codon selectivity7. This led

to the “tripeptide anticodon” model for stop codon reading. The structures12-14 lend some

support to this model as the tripeptides are in contact with the first and second bases of

the stop codons. However, the only seemingly specific interactions with the two motifs

involve the polar Thr and Ser residues (Fig. 1b) and, as shown herein, there is much more

to stop codon reading by RFs.

We have addressed the origin of RF specificity by carrying out molecular dynamics

(MD) free energy simulations of fourteen different codon complexes with RF1 and RF2

on the ribosome (Supplementary Fig. 1), utilizing the crystallographic structures12,14 as

starting points. The energetics associated with first position U/C and second and third

3

position A/G stop codon mutations are summarized in Fig. 1c and Supplementary Table

1. It can be seen that both release factors behave as expected in terms of specificity by

discriminating strongly against CAA. RF1 reads UAG with no difference in binding free

energy compared to UAA, while RF2 reads UAA and UGA with similar affinities.

Further, RF1 discriminates strongly against UGA while RF2 does not tolerate a third

position G and discriminates against both the UAG (stop) and UGG (Trp) codons.

Energetically, the specificities are generally characterized by differences in binding free

energies to the non-cognate codons of about 4 kcal/mol corresponding to a factor of

~1000 in terms of affinity. Such strong discrimination is particularly remarkable for the

third position A/G mutations in the case of RF2, where regular codon-anticodon

interactions with tRNA molecules usually permit both codons to be read. The free energy

differences in Fig. 1 are also very similar to those derived from the experimental KM

values4. Hence, the contribution to RF accuracy from codon binding is well reproduced

by the simulations and allow us to dissect the mechanisms behind the high specificities.

The requirement for a U in the first position is mainly enforced by backbone H-bonds to

Gly116/125 and Glu119/128 (Thermus thermophilus numbering is used throughout with

the notation RF1/RF2), as hypothesized earlier12-14. We can now see quantitatively from

the energy diagram in Fig. 2 that it is essentially these two interactions, together with the

sidechains of Thr186/Phe195, that provide the key discriminatory elements for the first

position. It should also be noted that the polarity of the base increases significantly for

both the U/C and A/G alterations, so that C and G require stronger interactions in the RF-

complexes to compensate for desolvation. Experimentally, RF1 showed an approximately

4-fold higher KM value for CAA than did RF26 and this higher accuracy, also seen in Fig.

1c, seems to be indirectly caused by the Thr186/Phe195 substitution that yields slightly

different conformations of the first position nucleotide (Fig. 2). It is further interesting

that the first position U/C alteration is predicted to yield somewhat larger backbone

displacements (up to ~1.5 Å and localized to the loop containing the tripeptide motif) in

RF1 than RF2 (Supplementary Fig. 2). It thus seems possible that this could have a

structural influence farther away affecting kcat, as observed experimentally4, by adverse

positioning effects on the GGQ loop.

4

Earlier hypotheses have ascribed the dual second position specificity of RF2 to the ability

of Ser193 to form bifurcated H-bonds to both A and G13,14. However, already from the

crystal structures it is clear that such H-bonds are unlikely for geometrical reasons and

the MD simulations show that Ser193 does not donate an H-bond to either A or G.

Instead we find that the RF2 specificity in the second codon position mainly resides in a

fine-tuned recognition switch constituted by Glu128, Asp131, Arg191 and Ser193 that,

by rotation of the Glu128 carboxylate group, allows distinct reading of both A and G (the

Glu119/128 sidechain electron density appears ill-defined in the crystal structures). Here,

Ser193 of the SPF motif reads an A by accepting one H-bond that is lost with a G, but it

instead then positions Glu128 for a bifurcated H-bond to the second nucleotide (Fig. 3).

In RF1, Asp131 is replaced by Leu122 and Ser193 by Pro184 and the switch is lost,

causing Glu119 to engage in a more stable ion pair with Arg182, yielding strong

discrimination against G. The critical role of Glu128 in RF2 explains the deleterious

effect of its mutation to lysine18 and the mechanics of this hitherto unidentified switch is

also reflected by strong conservation of the residues involved (Supplementary Fig. 3).

The proline and phenylalanine of the SPF motif in RF2 effectively constitute a

hydrophobic shield for Glu128 thereby occluding water molecules from its vicinity. In

contrast, in RF1 Val185 (backbone) and Thr186 of the PxT motif provide a hydrophilic

environment allowing water to interact with Glu119, which also affects its

conformational behaviour. The backbone carbonyl of Thr186 in RF1 further contributes

to discrimination against a second position G as its interaction becomes repulsive (Fig.

3b). This effect is not present in RF2 due to different peptide plane orientations of

Thr186/Phe195, that in turn appear coupled to the difference in sidechain length at the

conserved Glu187/Asp196 position. These findings essentially explain the influence of

the tripeptide motifs on second position discrimination7. However, the effects of these

motifs are largely indirect and the key feature is thus the fine-tuned control of the

Glu119/128 environment. In both RFs the intercalation or stacking of His193/202 onto

the second codon base provides an attractive interaction13, that would be less favourable

for pyrimidines (Supplementary Fig. 4).

5

Third position reading is particularly interesting since RF2 must strongly discriminate

against the Trp codon (UGG) to avoid premature termination of translation. There is

indeed a large energetic penalty for binding to UGG (Fig. 1) that is further boosted by a

~10-fold decrease in kcat4. Both RFs recognize A3 by bifurcated H-bonding to the

Thr194/203 sidechain (Fig. 4). In RF1, the simulations predict a key water molecule

bridging between the phosphate group of U531 and the third stop codon base (no waters

are discernible in the experimental electron density). This water has the ability to flip its

orientation so that it can hydrogen bond to both A and G. In RF2, Arg201 blocks this

space and causes repulsion with a third position G, but accepts an A (Fig. 4).

Furthermore, Gln181 in RF1 is a key element in reading a G as its NH2 group replaces

the H-bond lost by Thr194 when A is substituted for G. In RF2, Gln181 is replaced by

Val190 resulting in an unfavourable hydrophobic contact with the partially negative G

oxygen. Hence, the dual specificity of RF1 originates in two recognition switches (a

water and Gln181) that explain the different third position sensitivities. The key roles of

Gln181/Val190 and Ile192/Arg201 are again reflected by conservation (Supplementary

Fig. 3) and consistent with mutation experiments18.

Third position reading apparently has little to do with the tripeptide motifs. The

experimental identification of some omnipotent RF sequences could be explained by the

fact that the Q-VXXXD- - - -I motif was used as a universal cassette for selection7, where

precisely the Q (Gln181) and I (Ile192) positions constitute the RF1 mechanism for

reading both A3 and G3. Therefore such a tripeptide screen would appear to mainly

probe second position specificity and why a phenylalanine in the tripeptide motif was

identified as a “third base discriminator”7 remains obscure. It is further unclear why the

“primary screening” did not give any hits outside the tripeptide motifs7, since several

such sites would be expected from the crystal structures and present results.

The UGG codon is decoded by tRNATrp and both our calculations and experiments4 show

that RF2 does not misread UGG due to a large energetic penalty. Together with the

additional effect on kcat4 this provides a safeguard against premature termination. The

question then arises whether an “opposite” mechanism exists, by which tRNATrp avoids

6

reading UGA, which would insert tryptophan into the growing chain rather than

terminating it. The simulation results (Fig. 1) predict a small discrimination against UGA

by tRNATrp, in line with experimental results19. The UGA stop codon has indeed been

described as a “leaky” one with a reported readthrough frequency of about 102.20 An

example of readthrough is also the Hirsch suppressor, a tRNATrp variant differing by the

single G24A mutation21. Hence, an AC mismatch with the tRNATrp wobble position

apparently does not make codon recognition impossible, but has pronounced effects on

the rate constants of subsequent GTP hydrolysis, peptidyl transfer and rejection steps,

rendering the initial selection complex non-productive19. The reason for why non-

productive binding of tRNATrp to the UGA stop codon does not inhibit termination can

probably be explained by the ~100-fold higher affinity of the RFs19, 22 together with their

higher copy numbers.23,24

Despite the functional similarity of tRNAs and RFs, genetic code reading by protein

factors follow different principles resulting in a higher accuracy without the need for

monitoring rRNA bases. The recognition mechanism of the RFs cannot be described as

tRNA mimicry with a tripeptide anticodon, but involves a range of interactions (polar,

ionic, hydrophobic and stacking) with both the RF backbone and sidechains as well as

water molecules and key groups of the rRNA. The ability of both RF1 and RF2 to each

read two stop codons, with different individual specificities, largely resides in fine-tuned

recognition switches that illustrate the functional diversity of protein-RNA interactions.

METHODS SUMMARY

MD simulations utilizing the free energy perturbation technique and the Bennett

acceptance ratio method25 were use to calculate relative binding free energies for

different cognate and near-cognate mRNA-ribosome complexes with RF1, RF2 and the

tRNATrp anticodon. Free energy differences were calculated for the transformations UAA

CAA, UAA UGA, UAA UAG and UGA UGG using the thermodynamic

cycle in Supplementary Fig.1. Initial coordinates were taken from crystal structures of

RF1 and RF2 bound to 70S ribosomes12,14 and of the small ribosomal subunit loaded with

7

the UUU codon in complex with tRNAPhe, 16 which was changed into a UGG codon and a

CCA anticodon. All MD simulations were performed on spherical systems using the MD

program Q26 utilizing the CHARMM22 force field27 as described earlier17,28. 10-20

replicate simulations of ~5ns were performed for each system with different initial

velocities in order to assess statistical convergence.

Full Methods and associated references accompanies this paper.

8

References

1. Jorgensen, F., Adamski, F. M., Tate, W. P., & Kurland, C. G. Release factor-dependent false stops are infrequent in escherichia coli. J. Mol. Biol. 230, 41-50 (1993).

2. Thompson, R. C., & Stone, P. J. Proofreading of codon-anticodon interaction on ribosomes. Proc. Natl. Acad. Sci. U.S.A. 74, 198-202 (1977).

3. Ruusala, T., Ehrenberg, M., & Kurland, C. G. Is there proofreading during polypeptide-synthesis. EMBO J. 1, 741-745 (1982).

4. Freistroffer, D. V., Kwiatkowski, M., Buckingham, R. H., & Ehrenberg, M. The accuracy of codon recognition by polypeptide release factors. Proc. Natl. Acad. Sci. U.S.A. 97, 2046-2051 (2000).

5. Nissen, P. et al. Crystal Structure of the ternary complex of Phe-tRNA, EF-Tu, and a GTP Analog. Science 270, 1464-1472 (1995).

6. Kristensen, O., Laurberg, M., Liljas, A. & Selmer, M. Is tRNA Binding or tRNA mimicry mandatory for rtanslation factors? Curr. Protein Peptide Sci. 3, 133-141 (2002).

7. Ito, K., Uno, M., & Nakamura, Y. A tripeptide 'anticodon' deciphers stop codons in messenger RNA. Nature 403, 680-684 (2000).

8. Frolova, L. Y. et al. Mutations in the highly conserved GGQ motif of class 1 polypeptide release factors abolish ability of human eRF1 to trigger peptidyl-tRNA hydrolysis. RNA 5, 1014-1020 (1999).

9. Rawat, U. B. S. et al. A cryo-electron microscopic study of ribosome-bound termination factor RF2. Nature 421, 87-90 (2003).

10. Klaholz, B. P. et al. Structure of the Escherichia coli ribosomal termination complex with release factor 2. Nature 421, 90-94 (2003).

11. Petry, S. et al. Crystal structures of the ribosome in complex with release factors RF1 and RF2 bound to a cognate stop codon. Cell 123, 1255-1266 (2005).

12. Laurberg, M. et al. Structural basis for translation termination on the 70S ribosome. Nature 454, 852-857 (2008).

13. Weixlbaumer, A. et al. Insights into translational termination from the structure of RF2 bound to the ribosome. Science 322, 953-956 (2008).

14. Korostelev, A. et al. Crystal structure of a translation termination complex formed with release factor RF2. Proc. Natl. Acad. Sci. U.S.A. 105, 19684-19689 (2008).

15. Trobro, S., & Åqvist, J. A model for how ribosomal release factors induce Peptidyl-tRNA cleavage in termination of protein synthesis. Mol. Cell 27, 758-766 (2007).

16. Ogle, J. M. et al. Recognition of cognate transfer RNA by the 30S ribosomal subunit. Science 292, 897-902 (2001).

9

17. Almlöf, M., Andér, M., & Åqvist, J. Energetics of codon-anticodon recognition on the small ribosomal subunit. Biochemistry 46, 200-209 (2007) .

18. Uno, M., Ito, K. & Nakamura, Y. Polypeptide release at sense and noncognate stop codons by localized charge-exchange alterations in translational release factors. Proc. Natl. Acad. Sci. U.S.A. 99, 1819-1824 (2002).

19. Cochella, L., & Green, R. An active role for tRNA in decoding beyond codon: anticodon pairing. Science 308, 1178-1180 (2005).

20. Parker, J. Errors and alternatives in reading the universal genetic-code. Microbiol. Rev. 53, 273-298 (1989).

21. Hirsh, D., & Gold, L. Translation of UGA triplet in-vitro by tryptophan transfer RNAs. J. Mol. Biol. 58, 459-468 (1971).

22. Zavialov, A. V., Mora, L., Buckingham, R. H., & Ehrenberg, M. Release of peptide promoted by the GGQ motif of class 1 release factors regulates the GTPase activity of RF3. Mol. Cell 10, 789-798 (2002).

23. Dong, H. J., Nilsson, L., & Kurland, C. G. Co-variation of tRNA abundance and codon usage in Escherichia coli at different growth rates. J. Mol. Biol. 260, 649-663 (1996).

24. Adamski, F. M., Mccaughan, K. K., Jorgensen, F., Kurland, C. G., & Tate, W. P. The Concentration of Polypeptide-Chain Release Factor-1 and Factor-2 at Different Growth-Rates of Escherichia-Coli. J. Mol. Biol. 238, 302-308 (1994).

25. Bennett, C.H. Efficient estimation of free energy differences from Monte Carlo data. J. Comput. Phys. 22, 245-268 (1976).

26. Marelius, J., Kolmodin, K., Feierberg, I., & Åqvist, J. Q: A molecular dynamics program for free energy calculations and empirical valence bond simulations in biomolecular systems. J. Mol. Graph. Model. 16, 213-225 & 261 (1998).

27. MacKerell, A. D. et al. All-atom empirical potential for molecular modeling and dynamics studies of proteins. J. Phys. Chem. B 102, 3586-3616 (1998).

28. Trobro, S. & Åqvist, J. Mechanism of the translation termination reaction on the ribosome. Biochemistry 48, 11296-11303 (2009).

10

Supplementary Information accompanies this paper.

Acknowledgements Support from the Swedish Research Council (VR) is gratefully

acknowledged. We thank Dr. Måns Ehrenberg for useful discussions and Drs. Martin

Laurberg, Harry Noller, Albert Weixlbaumer and Venki Ramakrishnan for sending us

their coordinates prior to release and for providing electron density maps.

Author Contributions J.S. and M.A. performed the simulations. All authors analyzed

the data and prepared the manuscript.

Author Information The authors declare no competing financial interests. Correspondence and requests for materials should be addressed to J.Å. ([email protected]).

11

METHODS

Free energies were calculated from molecular dynamics simulations using the single

topology free energy perturbation (FEP) technique together with the Bennett acceptance

ratio method25. Relative binding free energies between different termination complexes

were calculated from the thermodynamic cycle in Supplementary Fig. 1 utilizing the

following mRNA codon transformations, RF1: UAA CAA, UAA UGA, UAA

UAG and RF2: UAA CAA, UAA UGA, UAA UAG, UGA UGG.

Simulations were also carried out for the UGG UGA transformation on the ribosome

in complex with a CCA anticodon in order to evaluate the possible discrimination of

tRNATrp against the UGA stop codon. Each transformation was carried out with and

without the RF or anticodon bound to the solvated ribosome-mRNA complex to evalute

the change in binding free energy.

Initial coordinates were taken from the crystal structures of RF1 and RF2 bound to 70S

ribosomes loaded with the UAA stop codon in the A-site and the crystal structure of the

small ribosomal subunit loaded with the UUU codon in complex with tRNAPhe (PDB

accession numbers 3D5A-3D5D, 3F1E-3F1H, and 1IBM, respectively)12,14,16. UAA was

changed to UGA in order to obtain an initial structure for the UGA UGG

transformation with RF2 (since the mRNA in the 70S RF2 complex with UGA13 ends at

the third codon position that structure was only used as a guide in modifying the above

UAA complex14. The two different crystallographic RF2 complexes13,14 are otherwise

essentially identical). The initial structure with the tRNATrp anticodon was generated

from the UUU-tRNAPhe complex by mutation into a UGG codon and a CCA anticodon.

All MD simulations were performed essentially as reported earlier17,28 with the MD

package Q26 utilizing the CHARMM22 force field27,29. Spherical simulations systems

(40 Å in diameter) centered on the relevant codon position were used and water

molecules close to the sphere boundary were subjected to radial and polarization

restraints according to the SCAAS model26,30. The net charge state of the system was kept

12

neutral with additional Mg2+ counter-ions placed at electrostatic potential minima and

charged groups closer than 5 Å from the simulation sphere boundary neutralized as

described earlier31. A 10 Å cutoff was used for non-bonded interactions, with electrostatic

interactions beyond the 10 Å cutoff treated by the local reaction field multipole

expansion method32. No cutoff was applied to non-bonded interactions involving the base

that was mutated. Simulations of the free and RF1/RF2/tRNATrp bound systems were

performed using the same simulation protocol. The equilibration phase consisted of step-

wise heating from 10 K to 300 K while gradually releasing restraints on heavy solute

atoms, followed by 450 ps of unrestrained equilibration at 300 K. Each system was then

subjected to 5.1 ns of production phase MD for free energy calculations comprising 51

intermediate FEP states, where the first 25 ps of sampling at each state was discarded for

equilibration. A time step of 1 fs was used, in combination with the SHAKE33 procedure

for solvent bonds and angles. Ten replicate simulations were performed for each system

with initial velocities randomized according to the Maxwell distribution. For the UAA

UGA transformations with and without RF2 another ten replicate simulations were

performed to improve the statistics. Error bars in Figs. 1-4 and Supplementary Table 1

are given as standard errors of the mean for the replicate simulations.

References

29. MacKerell, A. D., Wiorkiewicz-Kuczera, J., & Karplus, M. An all-atom empirical energy function for the simulation of nucleic-acids. J. Am. Chem. Soc. 117, 11946-11975 (1995).

30. King, G., & Warshel, A. A surface constrained all-atom solvent model for effective simulations of polar solutions. J. Chem. Phys. 91, 3647-3661 (1989).

31. Trobro, S. & Åqvist, J. Mechanism of peptide bond synthesis on the ribosome. Proc. Natl. Acad. Sci. U.S.A. 102, 12395-12400 (2005).

32. Lee, F. S., & Warshel, A. A local reaction field method for fast evaluation of long-range electrostatic interactions in molecular simulations. J. Chem. Phys. 97, 3100-3107 (1992).

33. Ryckaert, J.-P., Ciccotti, G., & Berendsen, H. J. C. Numerical integration of the cartesian equations of motion of a system with constraints: Molecular dynamics of n-alkanes. J. Comput. Phys. 23, 327-341 (1977).

13

Figure Legends

Figure 1. Stop codon specificity of RF1 and RF2 in termination. a, Schematic

view of the binding of release factors RF1 and RF2 and tRNATrp to the ribosomal A-site

with their cognate codons indicated. b, Interactions of the characteristic PxT (RF1 –

yellow) and SPF (RF2 – cyan) release factor tripeptide motifs with the UAA stop

codon12,14 (mRNA – green), where the polar Thr and Ser sidechains are explicitly shown.

c, Calculated binding free energy changes for RF1 (yellow), RF2 (cyan) and tRNATrp

(blue) caused by different base alterations in the stop codons.

Figure 2. First position reading. Recognition of U and discrimination against C

observed from the average MD structures of termination complexes of RF1 (top, yellow)

and RF2 (bottom, cyan) with the UAA (a) and CAA (b) codons (green). Also shown are

diagrams of the average interaction energies between RF residues and the first codon

position base, where key sidechain (sc) and backbone (bb) contributions are depicted

(RF1 – yellow, RF2 – cyan). The discrimination against C is reflected by more positive

interaction energies than U and this effect is somewhat more pronounced for RF1 than

RF2.

Figure 3. Second position reading. Average MD structures of termination

complexes with the UAA (a) and UGA (b) codons (colour coding as in Fig. 2). While

RF1 (top) discriminates against a second position G, the dual specificity of RF2 (bottom)

arises from a recognition switch involving Glu128 as reflected by the more favourable

interactions with G in RF2 than RF1 in the average energy plot. Note, that the more

negative absolute interaction energies for G than A with Glu119/128 are simply a

consequence of the higher polarity of G that also has stronger interactions with water in

absence of bound RF.

14

15

Figure 4. Third position reading. Average MD structures of termination complexes

with the UAA (a) and UAG (b) codons (colour coding as in Fig. 2 and with G530 of the

rRNA in light blue). RF1 (top) reads both A and G in the third position with a recognition

switch involving Thr194, Gln181 and a water molecule bridging to the rRNA backbone,

while this is prevented in RF2 by the Gln181/Val190 and Ile192/Arg201 mutations as is

also evident from the average interaction energy diagrams. The sidechain rotamer of

Gln181 that is essential for the recognition switch is dictated by an H-bond network

invoving several nearby groups and differs from that proposed in Ref. 12.

a

b c

-3-2-10123456

Transformation

∆∆

G (k

cal/m

ol)

RF1RF2tRNA-Trp

UAA→CAA UAA→UGA UAA→UAG UGA→UGG

UAAUAG

UAAUGA

UGG

E P A E P A E P A

RF1 RF2 tRNATrp

UAA Stop UGA StopUAG Stop UGG Trp

U1

Gly116

Gly125

a b

Glu119

Glu128

Thr186

Phe195

-7-6-5-4-3-2-1012

G116/125 (bb)

E119/128 (bb)

P184/S193 (sc)

V185(sc)/P194

T186/F195 (sc)

Residue

Int.

Ener

gy (k

cal/m

ol)

RF1RF2

-7-6-5-4-3-2-1012

G116/125 (bb)

E119/128 (bb)

P184/S193 (sc)

V185(sc)/P194

T186/F195 (sc)

Residue

Int.

Ener

gy (k

cal/m

ol)

RF1RF2

A2

C1

A2

U1

A2

C1

A2

A2

Pro184

Thr186

Val185

Glu119

Glu187

Leu122 Asp126

Arg182

Glu128

a b

U1

G2

His193

U1

U1

A2

U1

G2

-25-20-15-10

-505

E119/128 (sc)

P184/S193 (sc)

V185(sc)/P194

T186/F195 (bb)

T186/F195 (sc)

Residue

Int.

Ener

gy (k

cal/m

ol)

RF1RF2

-25-20-15-10

-505

E119/128 (sc)

P184/S193 (sc)

V185(sc)/P194

T186/F195 (bb)

T186/F195 (sc)

Residue

Int.

Ener

gy (k

cal/m

ol)

RF1RF2

Ser193

Pro194

Phe195Arg191

Asp196

His202

Asp131Met135

-6-5-4-3-2-101

Q181/V190P184/S193

V185/P194T186/P195

I192/R201

Residue

Int.

Ener

gy (k

cal/m

ol)

RF1RF2 -6

-5-4-3-2-101

Q181/V190P184/S193

V185/P194T186/P195

I192/R201

Residue

Int.

Ener

gy (k

cal/m

ol)

RF1RF2

A3G3

G3A3

G530

G530 U531

U531

Thr203

Thr194

Ile192

Arg201

Gln181

Val190

a b

Supplementary Information

Supplementary Table 1.

Binding free energy differences between with different mRNA codons for ribosome

complexes with RF1, RF2, and tRNATrp, calculated using the FEP method.a

Codons (5’–3’) Complex FEPbindG

UAA→CAA RF1 4.6 ± 0.5

UAA→CAA RF2 3.9 ± 0.6

UAA→UGA RF1 3.2 ± 0.3

UAA→UGA RF2 -1.1 ± 1.2b

UAA→UAG RF1 0.0 ± 0.6

UAA→UAG RF2 4.8 ± 0.6

UGA→UGG RF2 4.0 ± 0.8

UGG→UGA tRNATrp 0.2 ± 0.8 a All values are in kcal/mol. Errors are presented as the standard error of the mean from

ten simulations with different randomized starting velocities. bTwenty independent

simulations were used for the UAA→UGA transformation with RF2.

Supplementary Figure 1. Thermodynamic cycle used for binding free

energy calculations. The relative binding free energy for an RF (or tRNA) betweeen

ribosome-mRNA complexes differing in a single codon mutation is obtained in terms of

the free energies associated with transformation of the codon base with and without the

RF (or tRNA) bound. In the given example the third codon position is altered from an A

to a G and the change in binding free energy ( ) is obtained as

.

UAG UAAbind bind bindG G G

bound freebind mut mutG G G

Supplementary Figure 2. Structural effects of first position U/C alterations.

Comparison of average MD structures illustrating the different reponses of the RF

tripeptide motif backbone segment towards changing the first stop codon position U to a

C. Although the backbone displacements are generally small it is evident that the local

change in RF1 is larger than in RF2 which could possibly be signalled to remote part of

the RF structure, thereby affecting kcat for peptidyl-tRNA hydrolysis. Note, however, that

the size of our current simulation system does not allow such long-range effects to be

explored.

Supplementary Figure 3. Alignment of RF sequences comprising the key

stop codon recognition elements. Prokaryotic RFs show a strong conservation of

the residues constituting the recognition swtiches and reading mechanisms discussed in

the main text. RF1 sequences are shown at the top and RF2 sequences at the bottom with

T. thermophilus numbering. Residues involved in the first, second and third position

reading mechanisms are coloured in green, blue and red, respectively (the backbone of

Glu119/141 is also involved in first position reading). The PxT and SPF tripeptide motifs

are in yellow. Species abbreviations are: Ec, Escherichia coli; Vc, Vibrio cholerae; Hi,

Haemophilus influenzae; Pa, Pseudomonas aeruginosa; Hp, Helicobacter pylori; Bs,

Bacillus subtilis; Rr, Rickettsia rickettsii, Bb, Borrelia burdorferi, Tt, Thermus

thermophilus; Mt, Mycobacterium tuberculosis. An interesting observation is the Glu141

to Asp mutation in Mt RF2 that is found mainly in the Actinobacteria branch. Juding

from both the MD simulations and crystallographic structures it seems likely, however,

that such a change would still allow reading of a G with the same mechanism as with the

consensus Glu residue.

Supplementary Figure 4. Stacking and H-bonding network involving

His193/215. A conserved water molecule (in all simulations) is found to form an H-

bond network between His193/215, the backbone carbonyl of Thr194/216 and the O2’ of

the second position base. These interactions appear to stabilize the conformation of

Thr194/216 in an optimal position for reading a third position A, while simultaneously

orienting His193/215 in such a way that stacking occurs only with the six-membered ring

of a second position purine, but not with a pyrimidine.

Figure S1 (Sund et al.)

Figure S2 (Sund et al.)

U/C

A

U/C

AP

VT

S

PF

a b

Figure S3 (Sund et al.) EcRF1 GTGGDEAALFAGDLFRMYSRYAEARRWRVEIMSASEGEHG…………ESGGHRVQRVPATESQGRIHTSAC VcRF1 GAGGDEAGIFAGDLFRMYSRFAEKKGWRIEVMSSSEAEHG…………ESGGHRVQRVPATEAQGRIHTSAC HiRF1 GTGGDEAGIFAGDLFRMYSRYAESKRWRVEMLSANESEQG…………ESGGHRVQRVPKTESQGRIHTSAC PaRF1 GTGGDEAAIFSGDLFRMYSRYAERQGWRVETLSENEGEHG…………ESGAHRVQRVPETESQGRIHTSAC HpRF1 GTGGDEAGIFVGDLFKAYCRYADLKKWKVEIVSSSENSVG…………EAGTHRVQRVPETESQGRIHTSAI BsRF1 AAGGEEAALFAGNLYRMYSRYAELQGWKTEVMEANVTGTG…………ENGAHRVQRVPETESGGRIHTSTA RrRF1 GSGGEEAALFAAVLFNMYQRYAELKGWRFEILAISDTGIG…………ESGVHRVQRVPETESQGRIHTSAA BbRF1 GTGGEEAALFANNLYSMYIKYSEKKKWKTEIINFNETELG…………ESGVHRVQRIPITESNGRLQTSAA TtRF1 GTGGEEAALFARDLFNMYLRFAEEMGFETEVLDSHPTDLG…………ESGVHRVQRVPVTETQGRIHTSTA MtRF1 GEGGEESALFAADLARMYIRYAERHGWAVTVLDETTSDLG…………EGGVHRVQRVPVTESQGRVHTSAA EcRF2 GSGGTEAQDWASMLERMYLRWAESRGFKTEIIEESEGEVA…………ETGVHRLVRKSPFDSGGRRHTSFS VcRF2 GSGGTEAQDWTSMLLRMYLRWADAKGFKTEVIEVSEGDVA…………ETGVHRLVRKSPFDSGGRRHTSFA HiRF2 GSGGTEAQDWTEMLLRMYLRWAESKGFKTELMEVSDGDVA…………ETGIHRLVRKSPFDSNNRRHTSFS PaRF2 GSGGTEAQDWANMLLRMYLRWADKHGFDATIIELSEGEVA…………EIGVHRLVRKSPFDSGNRRHTSFT HpRF2 GAGGTESQDWASILYRMYLRWAERRGFKSEILDYQDGEEA…………ENGVHRLVRISPFDANAKRHTSFA BsRF2 GAGGTESQDWGSMLLRMYTRWGERRGFKVETLDYLPGDEA…………EKGVHRLVRISPFDSSGRRHTSFV RrRF2 GAGGTESHDWASIMMRMYLRFAERLGFKTEIINMINGEEA…………EAGVHRLVRISPFNAAGKRMTSFA BbRF2 GAGGTEACDWVAMLYRMYSRYAERKKYKTELIDLLEAE-G…………EVGIHRLIRISPFDAAKKRHTSFA TtRF2 GAGGTEACDWAEMLLRMYTRFAERQGFQVEVVDLTPGPEA…………EAGVHRLVRPSPFDASGRRHTSFA MtRF2 GAGGVDAADWAEMLMRMYIRWAEQHKYPVEVFDTSYAEEA…………EQGTHRLVRISPFDNQSRRQTSFA

214 203144

192 122 181119

141

Figure S4 (Sund et al.)