Solution NMR determination of hydrogen bonding and base pairing between the glyQS T box riboswitch Specifier domain and the anticodon loop of tRNAGly

(1)

Solution NMR determination of hydrogen bonding and base pairing

between the glyQS T box riboswitch Speciﬁer domain and the anticodon

loop of tRNA

Gly q

Andrew T. Chang, Edward P. Nikonowicz

⇑

Department of Biochemistry and Cell Biology, Rice University, Houston, TX 77005-1892, United States

a r t i c l e

i n f o

Article history: Received 11 July 2013 Revised 30 August 2013 Accepted 2 September 2013 Available online 11 September 2013 Edited by Michael Ibba

Keywords: Heteronuclear U-turn tRNA Transcription regulation Riboswitch Antitermination Difference spectrum

a b s t r a c t

In Gram-positive bacteria the tRNA-dependent T box riboswitch regulates the expression of many amino acid biosynthetic and aminoacyl-tRNA synthetase genes through a transcription attenuation mechanism. The Specifier domain of the T box riboswitch contains the Specifier sequence that is complementary to the tRNA anticodon and is flanked by a highly conserved purine nucleotide that could result in a fourth base pair involving the invariant U33of tRNA. We show that the interaction

between the T box Speciﬁer domain and tRNA consists of three Watson–Crick base pairs and that U33

confers stability to the complex through intramolecular hydrogen bonding. Enhanced packing within the Speciﬁer domain loop E motif may stabilize the complex and contribute to cognate tRNA selection.

1. Introduction

In Gram-positive bacteria, the transcription of many tRNA syn-thetase genes and genes involved in amino acid metabolism is reg-ulated in a tRNA-dependent manner by the T-box riboswitch[1,2]. The T-box riboswitch includes a 200–300 nt region of mRNA lo-cated 50 _{to the translation start codon (also known as the mRNA}

leader region) that can form multiple conserved secondary struc-ture elements and selectively binds gene-speciﬁc tRNA species (Fig. 1)[3]. The binding of uncharged tRNA stabilizes a 30_-proximal

RNA hairpin designated the antiterminator helix and prevents pre-mature transcription termination[4].

The Speciﬁer domain (SD) is a structural element in the 50

re-gion of the mRNA leader that is variable in size and contains the Speciﬁer sequence, three nucleotides that are complementary to the anticodon nucleotides of the cognate tRNA. The speciﬁcity of

the riboswitch for tRNA is primarily achieved through pairing of the Speciﬁer sequence nucleotides with the anticodon of tRNA

[4,5] and changes in this sequence can switch the speciﬁcity of the T-box riboswitch to allow recognition of other tRNA species

[4–6]. In addition to the Speciﬁer sequence, the SD contains a loop E structural motif that is necessary for proper regulatory function and structure maintenance [7–9]. The SD also contains a highly conserved purine residue immediately 30_{to the Speciﬁer sequence}

that is positioned to pair with the invariant U33of tRNA[6,10],

cre-ating the potential for a fourth base pair between the anticodon loop and the Speciﬁer loop. This residue is protected from Mg2+ cleavage in the tRNA–mRNA leader complex[11], supporting the possibility that the tRNA–SD interaction involves four base pairs

[12]. Recent SHAPE analysis of tRNA in complex with stem I of the Geobacillus kaustophilus glyQS T-box riboswitch (which in-cludes the Speciﬁer domain) indicated protection of U33 but not

of the conserved adenine[13]. However, in the recently reported co-crystal structure of the Oceanobacillus iheyensis glyQ riboswitch Stem I with tRNAGly_{, the Speciﬁer–tRNA interaction involves three}

base pairs and the U33of tRNA loops out[14].

We have used NMR spectroscopy and isothermal calorimetry (ITC) to examine the interaction between the anticodon arm of tRNAGly,GCC _(ASLGly_{) and the SD of the Bacillus subtilis tyrS mRNA}

http://dx.doi.org/10.1016/j.febslet.2013.09.003

Abbreviations: NOESY, nuclear Overhauser effect spectroscopy; 2D, two dimen-sional; HMQC, heteronuclear multiple quantum coherence; NH, imino; CW, continuous wave; SD, speciﬁer domain

q

This work was supported by National Institutes of Health grant GM73969 to E.P.N.

⇑ Corresponding author. Fax: +1 713 348 5154. E-mail address:[email protected](E.P. Nikonowicz).

(2)

leader containing the glycyl Speciﬁer sequence, GGC. The SD– ASLGly_{complex is formed by three stacked intermolecular}

Wat-son–Crick G–C base pairs. The conformation of the ASLGly _loop

transitions from dynamic and disordered to a moderately stable U-turn structural motif in the complex. A U33A mutant of ASLGly

that is unable to form the canonical U-turn motif retains the ability to bind SD but with reduced affinity. Our data are consistent with a configuration where the conserved purine 30 _{to the Specifier}

se-quence and the conserved purine 30 _{to the anticodon (residue 37}

of tRNA) stack against the ends of the intermolecular helix and may confer additional stability to the complex. These results are consistent with the co-crystal structure of the glyQ T box–tRNA complex and observations of solution SHAPE experiments[13,14].

2. Methods

The RNA sequences shown inFig. 1were prepared by in vitro transcription with T7 RNA polymerase using synthetic DNA tem-plates and either unlabeled or13_C/15_{N -labeled 5}0_-NTPs_[15]_{. The}

RNA molecules were puriﬁed using 20% (w/v) preparative poly-acrylamide gels, electroeluted, and precipitated with ethanol. The RNA molecules were suspended an extensively dialyzed against 10 mM KCl and 5 mM potassium phosphate, pH 6.8. The samples were then heated to 90 °C for 60 s and snap cooled on ice before addition of MgCl2to 2.0 mM and 10% D2O. All RNA samples were

concentrated to a volume of 330 mL. The sample concentrations varied between 0.5–1.0 mM and were checked for RNA integrity using denaturing PAGE.

All NMR spectra were acquired on Varian Inova 600 and 800 MHz spectrometers equipped with cryogenically cooled1_H–

[13_C,15_{N] probes and solvent suppression was achieved using}

bino-mial read pulses. For selectively decoupled 1D difference experi-ments, pairs of1_{H spectra were recorded with application of}

on-or off-resonance low power (833 Hz) continuous wave (CW)15_N

decoupling during acquisition. 2D15_N–1_{H HMQC spectra were}

col-lected to identify 15_N–1_{H chemical shift correlations. 2D NOESY}

and NOESY–HMQC spectra (tm= 180 ms) and were acquired at

16 °C to obtain sequence speciﬁc NH1_{H resonance assignments.}

Typically, the data points were extended by 25% using linear pre-diction for the indirectly detected dimensions. NMR spectra were

processed and analyzed using Felix 2007 (Felix NMR Inc., San Die-go, CA).

A VP-ITC calorimeter (MicroCal, Inc.) was used for the ITC experiments. The concentrations of RNA in the injection syringe and sample cell were 260–350

l

M and 20–30

l

M, respectively. Thirty 10

l

L injections into 1.8 mL sample cell volume were per-formed at 10 °C with 5 min between injections. Control titrations (forward and reverse) were performed and yielded similar results

[16]. The ITC data was analyzed using the vendor-supplied soft-ware (ORIGIN v7.0) and plots ofDH versus mole ratio were gener-ated from the raw thermograms. The ﬁnal 4–6 points from each experiment were extrapolated to obtain a straight line that was subtracted from all the data before determining Ka (association

constant) and n (reaction stoichiometry) by ﬁtting the points using a non-liner least squares model for a single binding site.

3. Results and discussion

The SD sequence (Fig. 1) corresponds to that of the B. subtilis tyrS mRNA leader with the tyrosyl UAC Speciﬁer sequence replaced by the glycyl GGC[5]. The imino (NH) resonances of the SD and ASLGly _{molecules were monitored using} 15_N–1_{H HMQC spectra}

(Fig. 2) and assigned using NOESY-based experiments. Preliminary studies to identify and optimize conditions that stabilize the SD– ASLGly_{complex demonstrated that Mg}2+_{was necessary. Addition}

of Mg2+ _{to ASL}Gly _{disrupts the C}

32-A+38 base pair of the hairpin

[17,18], but has little effect on the NH spectrum (Fig. 2B). Addition of Mg2+ _{to SD reinforces the loop E motif as evidenced by the}

appearance of NH resonances for nucleotides U12and G26that

par-ticipate in reverse Hoogsteen U-A and sheared A–G base pairs, respectively (Fig. 2A). The NH resonance of the bulged G11

nucleo-tide shifts 1.0 ppm upﬁeld in response to Mg2+_{binding. At the base}

of the Speciﬁer loop, the NH resonances of U35, U5, and G6are

dou-bled and the splitting is most pronounced for residues proximal to the loop–helix junction. Exchange cross peaks between the split resonances in the NOESY spectrum (not shown) are indicative of two conformations at the base of the Speciﬁer loop.

The SD–ASLGly_{complexes were prepared such that one of the}

RNA molecules was13_C/15_{N-labeled and the other was unlabeled.}

In the HMQC spectrum of SD, complex formation leads to the appearance of two additional guanine NH resonances at 11.85 ppm (G29) and 13.14 ppm (G30) (Fig. 2A). These resonances

give rise to a cross peak in the NOESY spectrum (Fig. 2C). The res-onance at 13.14 ppm also has a NOE cross peak with a NH reso-nance at 12.78 ppm. A15_{N-edited NOESY spectrum indicated this}

later NOE cross peak is inter-molecular, establishing the identities of the NH resonances at 13.14 and 12.78 ppm as G30of SD and⁄G34

of ASLGly_{, respectively. Also in the complex, split SD resonances of}

U35, U5, and G6collapse into single peaks with chemical shifts

un-ique to the complex (Fig. 2A). In the loop E motif, the U12NH

res-onance intensiﬁes and the G26 15N NH resonance shifts 4.0 ppm

upﬁeld. The shift of the G26resonance is consistent with formation

of the extensive hydrogen bond network involving the base O6, N1, and N2 atoms of this residue in the context of a loop E motif[19]. Together, these results indicate that binding of ASLGly_imposes

con-formational ordering to the base of the Speciﬁer loop and within the loop E motif.

The free form of ASLGly_{has a ﬁve base pair stem, but the seven}

nucleotide loop does not adopt the archetypal U-turn motif[17]. The 15_N–1_{H HMQC spectrum contains four major peaks}

sponding to the stem nucleotides and a ﬁfth weak peak that corre-sponds to the terminal G27(Fig. 2B). Three additional peaks in the

NH1_{H spectrum are exchange broadened and do not give rise to}

peaks in the HMQC spectrum. Because there is no evidence for multiple conformations of the stem nucleotides in other ASLGly

spectra, these resonances were tentatively assigned to U33 and Fig. 1. (Left) Generalized secondary structure of the T box riboswitch in complex

with tRNA. The domain of Stem I containing the Specifier sequence and the anticodon loop of tRNA are highlighted. (Right) The nucleotide sequences corresponding to the Specifier domain of the tyrS T box riboswitch with glycyl Specifier sequence GGC (tyrSGGC_{) and the anticodon arm of tRNA}Gly,GCC_(ASLGly_{) from} Bacillus subtilis (tRNA numbering used for ASLGly_{). The Specifier (red) and anticodon} (green) nucleotides confer specificity to the interaction. The highly conserved A (magenta) in the Specifier domain and the invariant U (orange) of the anticodon loop also have been proposed to pair[11]. Nucleotides of the SD that form the loop E motif are shown in blue.

(3)

G34. The exchange-broadened peaks were characterized using1H

1D difference experiments (Fig. 3). The15_{N carrier was positioned}

at different chemical shifts in the NH nitrogen region and pairs of

1_{H spectra were recorded with and without low power CW} 15_N

decoupling. The NH1_{H resonances display incomplete cancellation}

when the15N carrier is positioned at or near the frequencies of the corresponding NH nitrogen resonances.15_{N decoupling centered at}

161 ppm (uridine NH region) leads to the appearance of two peaks, 10.69 ppm and 11.45 ppm (Fig. 3), suggesting two conformations of U33. Decoupling centered at 145 ppm (guanine NH region) leads

to the appearance of a peak at 10.37 ppm that was assigned G34.

Titration of unlabeled SD with labeled ASLGly causes 0.1 ppm downﬁeld shifts of the loop-proximal G39 and G40 NH 1H

reso-nances (Fig. 2B). Although no new cross peaks become apparent, the G34 (assigned in the NOESY) and G39resonances overlap and

appear as a single peak at 12.72 ppm (Fig. 2B). To locate the U33

NH resonance in the complex, 1D difference spectra were again

acquired.15_{N decoupling centered at 161 ppm shows the U}

33NH

resonance is a single peak at 11.64 ppm (Fig. 3). Although the U33

NH1_{H resonance in the complex is moderately broad, the chemical}

shift of the resonance is consistent with the spectroscopic signa-ture of a uridine nucleotide that participates in a U-turn motif

[20,21]. Hydrogen bonding involving a 20_{-OH also can lead to NH} 1_{H chemical shifts over this range, but a moderately stable U-turn}

that limits U33 dynamics is consistent with SHAPE experiments

that show protection of this residue after complex formation[13]. To further explore the importance of the U33 residue for SD

binding, a U33A mutant of ASLGlywas examined (Fig. 1). The

invari-ant U33 of tRNA molecules is the basis of the structural motif

known as the U-turn[22]. The U-turn motif reverses the direction of the phosphate backbone and orients the anticodon bases for pre-sentation to the mRNA codon. The motif is stabilized by cross-strand hydrogen bonds involving the U33 NH and 20-OH groups

and by stacking of the anticodon bases. The A33 mutant is not Fig. 2. (A)15

N–1

H HMQC imino spectra of SD (green), SD with Mg2+

(red), and SD in complex with ASLGly

(black). The G29and G30resonances appear only in the spectrum of the complex and have chemical shifts consistent with G–C base pairs. The appearance of the U12and G26indicates Mg2+increases the stability of loop E motif whereas the doubling of G6and U5indicates that Mg2+induces structural heterogeneity at the base of the Speciﬁer loop. The in vitro transcription reaction was primed with unlabeled 50 -GMP, therefore the G1NH resonance does not appear in this spectrum. (B)15N–1H HMQC imino spectra of ASLGly(green), ASLGlywith Mg2+(red), and ASLGlyin complex with SD (black). The NH resonances from the stem nucleotides are minimally altered by complex formation. (C) Imino proton connectivities between adjacent base pairs in the 180 ms mixing time NOESY spectrum of the SD–ASLGly

complex. The dashed lines trace the connectivities of the ASLGly

molecule in the complex. For clarity, the (⁄

) was added to indicate ASLGly

nucleotides. The ASLGly

is in 50% molar excess and the dotted lines trace the connectivities among the NH resonances of excess ASLGly

molecules. The labels identify cross peaks between NH protons of neighboring base pairs. Connectivity through the Speciﬁer codon triplet, G29–G30–C31, is highlighted in blue (G29and G30of SD and ⁄

G34of the ASL). The inter-base pair connectivity is broken between G6–C33and C31– ⁄

G34and between G29– ⁄

C10and the reverse Hoogsteen U12–A27. The absence of a G6– ⁄

G34 cross peak indicates that the lower stem and the Speciﬁer–anticodon helix are not coaxially stacked and that A32may provide a platform to stabilize the C31–⁄G34base pair. Chemical shifts are provided inSupplementary Table S1.

(4)

isosteric with U33 and cannot form the characteristic hydrogen

bond network, but A33also is not expected to sequester the

antico-don bases and prevent intermolecular pairing. Indeed, the NMR spectrum of the U33A mutant of ASLGlyindicates that the loop is

disordered (not shown). Native PAGE reveals an incomplete band shift of U33A ASLGlyby SD and NMR spectra indicate U33A

substitu-tion weakens the afﬁnity of ASLGlyfor SD. The HMQC spectrum of SD in the presence of two mole equivalents of U33A ASLGly

(Fig. 3B) displays features of free SD (resonances U5, U35, and G6)

and bound SD (resonances U5, U35, G6, and U12) which indicate a

lifetime for the complex >0.01 s. The NH resonances G29and G30

re-main absent from the spectrum and the G26resonance is exchange

broadened beyond detection, the later indicating a lifetime for the complex of 0.003 s. The different lifetimes suggest that binding involves more complex structural changes in the loop E motif than at the base of the Speciﬁer loop. The addition of one equivalent of ASLGlyto the same sample displaces the A33mutant and eliminates

the SD spectral heterogeneity.

ITC measurements reveal that the complex between SD and A33–ASLGlyis at least ten-fold weaker than the complex between

SD and U33–ASLGly (Fig. S1) (350 ± 68 nM vs 3.2 ± 0.8

l

M). The

afﬁnity of ASLGly

for this SD sequence is slightly stronger than the ASLGly_{afﬁnity for the native glyQS Speciﬁer domain sequence}

of 1.2

l

M[9]. Also, no binding could be detected using a non-cog-nate ASL corresponding to the unmodiﬁed form of the glycyl iso-type tRNAGly,UCC _[17]_{. Two compensatory mutants of SD, A}

32U

and A7U/A32U, were tested for their ability to associate with

U33A–ASLGlyand restore complex stability. However, complex

for-mation could not be detected by NMR or by native PAGE analysis. In the SD molecule, U32within the context of A7extends the lower

helix into the Speciﬁer loop by one Watson–Crick base pair and potentially limiting the conformational freedom of the Speciﬁer se-quence. The A7U/A32U double mutant does not appear to form an

additional base pair, but the absence of A32in both mutants also

could result in loss of important stacking interactions that are ob-served in the co-crystal structure[14].

Fig. 4 depicts a model of the SD–ASL complex based on the NMR-informed secondary structure of the complex and the solu-tion structure of the tyrS SD[8,9]. As in our previous model[8], rotation of the Speciﬁer bases toward the minor groove allows pairing with the anticodon bases and requires minimal rearrange-ment of the SD loop. Pairing via the minor groove face of SD also positions the 30 _{ﬂanking unpaired adenine bases, A}

37 of ASLGly

and A32of SD, for stacking against each end of the

anticodon–Spec-iﬁer helix. These features are exhibited in the recently reported tRNA and T-box Stem I co-crystal structure[14]. The crystal struc-ture also shows the loop E motif in the Speciﬁer domain collapses to a more compact state when the tRNA is bound, consistent with changes in the NH spectrum of SD (Fig. 2A).

Fig. 4. Model of the interaction between the Speciﬁer Sequence (green) and the ASLGly

loop (red). The loop E motif nucleotides are highlighted in blue. The solution structure of the SD (PDB 2kzl)[9]was docked with an ASLGly

sequence modeled on the anticodon arm of tRNALys,3

(PDB 1xmo) [24]. Base pair and loop E motif constraints (derived from NOE connectivities and comparison of15

N–1 H HMQC spectra) were applied and the complex subjected to energy minimization using Xplor-NIH. Rotation of the Specifier bases towards the minor groove side of the Specifier domain allows anticodon binding with minimal adjustment of the SD structure[9]. The A37and A32bases of ASLGlyand SD, respectively, are positioned to stack on the ends of the anticodon–Specifier helix as seen in the co-crystal structure of T-box stem I with tRNA[14]. U33was modeled in a U-turn configuration to reflect the solvent protection and chemical shift properties of the U33NH proton.

Fig. 3. (A) NH region of the1

H 1D difference spectra of ASLGly

in free (left) and SD-bound (right) states. The15

N carrier positions of 145 and 161 ppm are centered on the guanine and uridine regions, respectively, for low power CW15_{N decoupling.} The top spectra were acquired using broadband (GARP)15

N decoupling centered at 153 ppm. In the complex, the U33resonance intensiﬁes and collapses to a single peak with a frequency that corresponds to a U in a U-turn motif[20,21]. (B)15

N–1 H HMQC spectrum of the NH region of SD in complex (black) with two mole equivalents of the U33A mutant of ASLGly. Peaks highlighted in red correspond to resonances of the Mg2+

-bound SD molecule. The G26, G29, and G30resonances (blue boxes) appear to be in intermediate exchange and are not present in the spectrum. With the addition of one equivalent of native ASLGly

, G26, G29, and G30appear and peaks from the free SD disappear.

(5)

In both the solution model and crystal structure, U33is not

posi-tioned for intermolecular pairing. However, increased protection of the U33NH proton from solvent exchange and reduced dynamics of

the U33ribose in the complex, as evidenced by SHAPE results[13],

suggest U33adopts a more intra-helical orientation in solution than

the looped-out conformation observed in the co-crystal structure. An intra-helical orientation of A33may impair binding due to steric

clash or loss of key hydrogen bonds unique to U33 [22], both of

which would be disruptive to the Speciﬁer–anticodon helix. A looped-out conformation of residue 33 is expected to accommo-date the U33A substitution with little effect on binding. However,

the current data do not permit deﬁnitive determination of the spe-ciﬁc cause for weaker binding of A33–ASLGly.

In agreement with biochemical studies, the solution model and crystal structure of the complex indicate tRNA is selected through complementary pairing of Speciﬁer and anticodon bases[4,5]and the rejection of non-cognate tRNA appears dependent on intermo-lecular helix stability. In the ribosome, invariant nucleotides G530,

A1492, and A1493 of 16S rRNA form a hydrogen bond network

involving 20_{-OH groups that interrogate the minor groove of the}

co-don–anticodon helix[23]. This network is critical to selection of cognate tRNAs and rejection of non-cognate tRNAs[23]. Although presumably less critical than in translation, a similar ‘‘proof-read-ing’’ mechanism to ensure cognate tRNA selection is not apparent for the Specifier–anticodon interaction. Instead, cognate tRNA selection may rely on the geometry of the Specifier–anticodon he-lix to promote favorable interactions, such as ordering of the SD loop E motif upon ASL binding and optimized stacking of the flank-ing purines, to slow tRNA dissociation and permit dockflank-ing of the tRNA elbow and apical loop of stem I[13,14].

4. Conclusion

The solution and crystal forms of the SD–ASLGly_{complex share}

many structural features including a three-base pair intermolecu-lar helix ordering of the SD loop E motif. In solution, the U-turn motif may facilitate the 30_{base stack of the anticodon loop}

nucle-otides G34–A37and stabilize of the SD–ASLGlyminihelix. The highly

conserved purine nucleotide at the 30_{end of the Speciﬁer sequence}

provides a surface for stacking of the intermolecular minihelix at the base of the Speciﬁer domain and may further stabilize the com-plex. Although the three-base pair interaction mirrors the mRNA– tRNA interaction in the ribosome, the rejection of non-cognate tRNA by the T box riboswitch utilizes a different mechanism. Acknowledgments

We thank Malgorzata Michnicka for preparation of the T7 RNA polymerase and synthesis of the labeled 50_-nucleotide

triphos-phates. The 800 MHz NMR spectrometer was purchased with funds from the W. M. Keck Foundation and the John S. Dunn Foundation.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.febslet.2013. 09.003.

References

[1] Henkin, T.M. (2000) Transcription termination control in bacteria. Curr. Opin. Microbiol. 3 (2), 149–153.

[2] Henkin, T.M. (2008) Riboswitch RNAs: using RNA to sense cellular metabolism. Genes Dev. 22 (24), 3383–3390.

[3] Grundy, F.J. and Henkin, T.M. (1994) Conservation of a transcription antitermination mechanism in aminoacyl-tRNA synthetase and amino acid biosynthesis genes in gram-positive bacteria. J. Mol. Biol. 235 (2), 798–804. [4] Grundy, F.J. and Henkin, T.M. (1993) TRNA as a positive regulator of

transcription antitermination in B. subtilis. Cell 74 (3), 475–482.

[5] Grundy, F.J., Hodil, S.E., Rollins, S.M. and Henkin, T.M. (1997) Speciﬁcity of tRNA–mRNA interactions in Bacillus subtilis tyrS antitermination. J. Bacteriol. 179 (8), 2587–2594.

[6] Grundy, F.J., Moir, T.R., Haldeman, M.T. and Henkin, T.M. (2002) Sequence requirements for terminators and antiterminators in the T box transcription antitermination system: disparity between conservation and functional requirements. Nucleic Acids Res. 30 (7), 1646–1655.

[7] Rollins, S.M., Grundy, F.J. and Henkin, T.M. (1997) Analysis of cis-acting sequence and structural elements required for antitermination of the Bacillus subtilis tyrS gene. Mol. Microbiol. 25 (2), 411–421.

[8] Wang, J., Henkin, T.M. and Nikonowicz, E.P. (2010) NMR structure and dynamics of the Speciﬁer loop domain from the Bacillus subtilis tyrS T box leader RNA. Nucleic Acids Res. 38 (10), 3388–3398.

[9] Wang, J. and Nikonowicz, E.P. (2011) Solution structure of the K-turn and Speciﬁer loop domains from the Bacillus subtilis tyrS T-box leader RNA. J. Mol. Biol. 408 (1), 99–117.

[10] Vitreschak, A.G., Mironov, A.A., Lyubetsky, V.A. and Gelfand, M.S. (2008) Comparative genomic analysis of T-box regulatory systems in bacteria. RNA 14 (4), 717–735.

[11] Yousef, M.R., Grundy, F.J. and Henkin, T.M. (2005) Structural transitions induced by the interaction between tRNAGly

and the Bacillus subtilis glyQS T box leader RNA. J. Mol. Biol. 349 (2), 273–287.

[12] Nelson, A.R., Henkin, T.M. and Agris, P.F. (2006) TRNA regulation of gene expression: interactions of an mRNA 50_{-UTR with a regulatory tRNA. RNA 12}

(7), 1254–1261.

[13] Grigg, J.C., Chen, Y., Grundy, F.J., Henkin, T.M., Pollack, L. and Ke, A. (2013) T box RNA decodes both the information content and geometry of tRNA to affect gene expression. Proc. Natl. Acad. Sci. USA 110 (18), 7240–7245.

[14] Zhang, J. and Ferre-D’Amare, A.R. (2013) Co-crystal structure of a T-box riboswitch stem I domain in complex with its cognate tRNA. Nature 500 (7462), 363–366.

[15] Nikonowicz, E.P., Sirr, A., Legault, P., Jucker, F.M., Baer, L.M. and Pardi, A. (1992) Preparation of13

C and15

N labelled RNAs for heteronuclear multi-dimensional NMR studies. Nucleic Acids Res. 20, 4507–4513.

[16] Salim, N.N. and Feig, A.L. (2009) Isothermal titration calorimetry of RNA. Methods 47 (3), 198–205.

[17] Chang, A.T. and Nikonowicz, E.P. (2012) Solution nuclear magnetic resonance analyses of the anticodon arms of proteinogenic and nonproteinogenic tRNAGly_{. Biochemistry 51 (17), 3662–3674}_.

[18] Denmon, A.P., Wang, J. and Nikonowicz, E.P. (2011) Conformation effects of base modiﬁcation on the anticodon stem-loop of Bacillus subtilis tRNATyr_{. J.}

Mol. Biol. 412 (2), 285–303.

[19] Correll, C.C., Freeborn, B., Moore, P.B. and Steitz, T.A. (1997) Metals, motifs, and recognition in the crystal structure of a 5S rRNA domain. Cell 91 (5), 705–712. [20] Cabello-Villegas, J., Tworowska, I. and Nikonowicz, E.P. (2004) Metal ion stabilization of the U-turn of the A37 N6-dimethylallyl-modiﬁed anticodon

stem-loop of Escherichia coli tRNAPhe

. Biochemistry 43 (1), 55–66.

[21] Sundaram, M., Durant, P.C. and Davis, D.R. (2000) Hypermodiﬁed nucleosides in the anticodon of tRNALys

stabilize a canonical U-turn structure. Biochemistry 39 (41), 12575–12584.

[22] Quigley, G.J. and Rich, A. (1976) Structural domains of transfer RNA molecules. Science 194 (4267), 796–806.

[23] Ogle, J.M., Murphy, F.V., Tarry, M.J. and Ramakrishnan, V. (2002) Selection of tRNA by the ribosome requires a transition from an open to a closed form. Cell 111 (5), 721–732.

[24] Murphy, F.V.T., Ramakrishnan, V., Malkiewicz, A. and Agris, P.F. (2004) The role of modiﬁcations in codon discrimination by tRNALys,UUU_{. Nat. Struct. Mol.}