Glutamate versus glutamine exchange swaps substrate selectivity in tRNA-guanine transglycosylase: Insight into the regulation of substrate selectivity by kinetic and crystallographic studies

Summary

Bacterial tRNA-guanine transglycosylase (Tgt) catalyses the exchange of guanine in the wobble position of particular tRNAs by the modified base preQ₁. In vitro, however, the enzyme is also able to insert the immediate biosynthetic precursor, preQ₀, into those tRNAs. This substrate promiscuity is based on a peptide switch in the active site, gated by the general acid/base Glu²³⁵. The switch alters the properties of the binding pocket to allow either the accommodation of guanine or preQ₁. The peptide conformer recognising guanine, however, is also able to bind preQ₀. To investigate selectivity regulation, kinetic data for Z. mobilis Tgt were recorded. They show that selectivity in favour of the actual substrate preQ₁ over preQ₀ is not achieved by a difference in affinity but via a higher turn-over rate. Moreover, a Tgt(Glu²³⁵Gln) variant was constructed. The mutation was intended to stabilise the peptide switch in the conformation favouring guanine and preQ₀ binding. Kinetic characterisation of the mutated enzyme revealed that the Glu²³⁵Gln exchange has, with respect to all substrate bases, no significant influence on k_cat. In contrast, K_M(preQ₁) is drastically increased while K_M(preQ₀) seems to be decreased. Hence, regarding k_cat/K_M as an indicator for catalytic efficiency, selectivity of Tgt in favour of preQ₁ is abolished or even inverted in favour of preQ₀ for Tgt(Glu²³⁵Gln). Crystal structures of the mutated enzyme confirm that the mutation strongly favours the binding pocket conformation required for the accommodation of guanine and preQ₀. The way this is achieved, however, significantly differs from what was predicted based on crystal structures of wild type Tgt.

Introduction

Transfer RNA-guanine transglycosylases (Tgt, E.C. 2.1.2.29) constitute a family of tRNA-modifying enzymes present in all three domains of life. They catalyse the base exchange of specific guanine residues in tRNAs by 7-substituted-7-deazaguanines in a domain specific manner.^1,2 Archaeal Tgt inserts preQ₀ (for chemical formula see Table 1) at position 15 of the majority of known archaeal tRNAs where it is further converted to archaeosine.³ Bacterial Tgt, in contrast, is involved in a pathway introducing the hypermodified base queuine (Q) into the anticodon position 34 (the “wobble position”) of just four different tRNAs⁴ (Figure 1). These are specific for His, Tyr, Asp and Asn, respectively, and share a U³³G³⁴U³⁵ sequence in common.^5,6 The enzyme performs the first tRNA dependent step in queuine biosynthesis replacing guanine³⁴ (G³⁴) by preQ₁⁴ (for chemical formula see Table 1). This premodified base is produced from GTP⁷ by means of the queC, queD, queE and queF gene products.^8,9 While the exact functions of QueC, QueD and QueE still have to be figured out, QueF has been identified as an NADPH-dependent oxidoreductase catalysing the reduction of preQ₀ to preQ₁.¹⁰ After incorporation into tRNA, preQ₁ is converted to the functional Q base in two subsequent steps performed by S-adenosylmethionine:tRNA ribosyltransferase-isomerase (QueA)^11–13 and an unknown coenzyme B₁₂-dependent enzyme¹⁴.

Table 1.

Structures of natural substrates of bacterial Tgt and the inhibitor 2-butyl-1H-imidazole-4,5-dicarboxylic acid hydrazide (BIH). In the line “conformation” the functional group of the Leu²³¹/Ala²³² peptide bond exposed to the binding pocket is given.

conformation	carbonyl Leu231H-bonded Glu235 (deprotonated)	amide Ala232H-bonded Glu235 (protonated)	amide Ala232 stacking Glu235 (protonated)
natural substrates	KM: 0.7 μM	KM: 1.2 μM	KM: 0.9  M
inhibitor	Kic: 62 μM

Figure 1. Queuine modification pathway.

AdoMet: S-adenosylmethionine; B₁₂: coenzyme B₁₂; oQ: epoxyqueuine.

The base exchange catalysed by bacterial Tgt follows a ping-pong reaction mechanism¹⁵ which is well documented by crystal structure analyses of Tgt from Zymomonas mobilis^16–18 (Figure 2). In a first step, tRNA binds to Tgt and the glycosidic bond of G³⁴ is cleaved via a nucleophilic attack of Asp²⁸⁰ producing a covalent Tgt·tRNA complex intermediate. Subsequently, preQ₁ replaces guanine within the binding pocket and is incorporated into tRNA in a reverse reaction step. During the base exchange step, replacement of guanine by preQ₁ requires an adjustment of the binding pocket geometry (Figures 2(b),(c)). Both guanine and preQ₁ are recognised by functional groups of Asp¹⁰², Asp¹⁵⁶, Gln²⁰³, and Gly²³⁰. However, to distinguish between the two bases, a peptide bond formed by Leu²³¹ and Ala²³² has to flip. G³⁴-tRNA binding results in a water mediated contact of the G³⁴-N7 to the main chain amide of Ala²³². Although the moderate resolution of the G³⁴-tRNA bound Tgt crystal structure does not allow to unambiguously extract this structural detail,¹⁷ the structure of Tgt bound to the guanine analogous inhibitor 2-butyl-1H-imidazole-4,5-dicarboxylic acid hydrazide (BIH) (Table 1), clearly suggests this binding mode.¹⁹ In order to permit efficient preQ₁ binding the composition of the binding pocket has to be altered. Proper recognition of the preQ₁ amino methyl group requires the CO acceptor functionality of the Leu²³¹/Ala²³² peptide bond to be exposed instead of the NH donor (Figure 2(c)). Both peptide bond geometries are stabilised from the protein interior using the side chain carboxyl of Glu²³⁵, a residue strictly conserved in all bacterial Tgts. Obviously depending on its protonation state, Glu²³⁵ either donates an H-bond to the backwards oriented carbonyl oxygen or accepts an H-bond from the amide of the flipped peptide bond (Figures 2(b),(c)). Consistently, the peptide switch is also observed in crystal structures of the ligand free enzyme in a pH dependent manner. Tgt crystals obtained at pH 5.5 show the amide of the Leu²³¹/Ala²³² peptide bond exposed to the binding pocket,¹⁸ while the opposite orientation is observed in crystals obtained at pH 8.5 (Figure 3(d)).

Figure 2.

Base exchange mechanism in bacterial Tgt.

Figure 3. Crystal structures of bacterial and archaeal Tgt.

(a) The superposition of bacterial Tgt in complex with the Tgt inhibitor BIH (C atoms in cyan) and guanine (C atoms in grey) shows in both cases an identical conformation for the Leu²³¹/Ala²³² peptide bond stabilised by an H-bond of the Leu²³¹ carbonyl to the protonated Glu²³⁵ side chain carboxyl. (b) Bacterial Tgt in complex with preQ₁. The Leu²³¹/Ala²³² peptide bond is switched compared to the guanine/BIH bound situation stabilised by an H-bond of the Ala²³² amide to the deprotonated Glu²³⁵ carboxyl. (c) Bacterial Tgt in complex with preQ₀. The Leu²³¹/Ala²³² peptide bond conformation corresponds to that observed for guanine/BIH bound Tgt, but stabilisation surprisingly occurs through a stacking interaction with the protonated Glu²³⁵ carboxyl concomitant with a Glu²³⁵/Gly²³⁶ peptide bond flip. (d) Superposition of ligand-free bacterial Tgt crystallised at pH 5.5 (C atoms in red) and pH 8.5 (C atoms in grey). Protonation of Glu²³⁵ at pH 5.5 leads to a binding pocket geometry observed upon binding of guanine/BIH. Deprotonation of Glu²³⁵ at pH 8.5 results in a geometry as induced upon preQ₁ binding. (e) Superposition of bacterial Tgt and Tgt(Glu²³⁵Gln) both in complex with preQ₀ reveals identical binding pocket geometries. The Glu²³⁵ side chain is stacking between the aromatic ring system of Tyr¹⁶¹ and the Leu²³¹/Ala²³² peptide bond (encircled). (f) Archaeal Tgt with bound preQ₀. The amide group of the Val¹⁹⁷/Val¹⁹⁸ peptide bond (corresponding to Leu²³¹/Ala²³² in bacterial Tgt) is unable to flip. It permanently exposes the Val¹⁹⁸ NH, since the opposing Val¹⁹⁷ CO is involved in an invariant backbone H-bond network.

In addition to preQ₁, bacterial Tgt in vitro also accepts the biosynthetic precursor of preQ₁, namely preQ₀, as a substrate.^18,20 This substrate promiscuity is a consequence of the above-mentioned peptide switch as binding of preQ₀ is facilitated by the Leu²³¹/Ala²³² peptide bond conformation required for the accommodation of guanine. Since in bacterial cells both preQ₀ and preQ₁ are present (even at unknown concentrations), bacterial Tgt must be able to discriminate between these substrates and preferentially incorporate preQ₁ into tRNA. For Escherichia coli Tgt there exists evidence that this is achieved through a lower affinity of the enzyme to preQ₀, since K_M determination revealed a six-fold higher value for preQ₀ (2.4 μM) compared to preQ₁ (0.4 μM)²⁰ (for Z. mobilis Tgt no kinetic data regarding preQ₀ and preQ₁ have been reported yet). Although the available complex structures of preQ₀ and preQ₁ bound to Z. mobilis Tgt do not provide an ostensible explanation for this selectivity, they reveal some significant differences between the binding modes of the two substrates.¹⁸

In the Tgt·preQ₁ complex structure the Glu²³⁵ carboxyl function is clearly present in a deprotonated state, promoting the formation of an H-bond (2.8 Å) to the main chain amide of Ala²³². This results in the exposure of the Leu²³¹ main chain carbonyl into the binding pocket enabling it to interact with the presumably protonated and thus positively charged amino methyl group of preQ₁ (Figure 3(b)). It should be noted that in this scenario the charges of the Glu²³⁵ carboxylate and the protonated preQ₁ amino methyl group compensate for each other.

In the Tgt·preQ₀ complex the Leu²³¹/Ala²³² peptide bond is flipped compared to the preQ₁ bound situation, i.e. the Ala²³² main chain amide is facing the binding pocket, where it H-bonds the ligand’s acceptor nitrile group. As mentioned above, this conformation of the Leu²³¹/Ala²³² peptide bond exactly corresponds to the one observed upon binding of the guanine analogue inhibitor BIH. However, the way this conformation is stabilised by the Glu²³⁵ side chain carboxyl, is significantly different. Indeed, the Glu²³⁵ carboxyl function does not H-bond in its protonated state the backwards oriented Leu²³¹ carbonyl. It rather stacks with its hydrophobic π face between the Leu²³¹/Ala²³² peptide bond and the aromatic ring system of the Tyr¹⁶¹ side chain. This stacking interaction is stabilised by H-bonds formed by the Glu²³⁵ carboxyl group to the main chain amide of Val²³³ (3.1 Å) and to the carbonyl oxygen of Gly²³⁰ (2.8 Å). The latter H-bond clearly indicates that also here the Glu²³⁵ carboxyl function is present in a protonated state. A shift of the Glu²³⁵ side chain by ca. 1.5 Å towards the binding pocket is prerequisite for the stacking interaction. This shift is rendered possible by a Glu²³⁵/Gly²³⁶ peptide bond flip which so far had not been observed in any other Tgt crystal structure¹⁸ (Figures 3(c),(e)).

The present work was aimed to confirm and further analyse the role of Glu²³⁵ in the Leu²³¹/Ala²³² peptide switch and in substrate promiscuity of bacterial Tgt. In order to mimic a permanently protonated state of the Glu²³⁵ carboxyl function we introduced a conservative Glu²³⁵Gln mutation into the Z. mobilis Tgt enzyme. This mutation was expected to firmly fix the peptide switch in its NH exposing form independent of the applied pH or any bound substrate. This article describes the consequences of this mutation on substrate specificity of Z. mobilis Tgt as studied by means of enzyme kinetics and crystal structure analyses.

Results

Crystal structure of wild type Z. mobilis Tgt in complex with guanine

The crystal structure of Z. mobilis Tgt in complex with the inhibitor 2-butyl-1H-imidazole-4,5-dicarboxylic acid hydrazide (BIH) had shown, for the first time, a conformation of the Leu²³¹/Ala²³² peptide bond, in which the main chain amide of Ala²³² was exposed to the binding pocket.¹⁹ This had been in contrast to all other crystal structures of Z. mobilis Tgt determined previously, where this peptide bond was oriented such that the main chain carbonyl of Leu²³¹ was facing the binding pocket. Since guanine just like BIH lacks the exocyclic aminomethyl group of preQ₁, it had immediately been postulated that binding of guanine to Tgt resulted in the same conformation of the Leu²³¹/Ala²³² peptide bond as binding of BIH. Due to the lack of a high resolution crystal structure of guanine-bound Tgt, this assumption had not yet been proven so far. Thus, in order to figure out the exact geometry of the Tgt binding pocket with guanine bound, we crystallised Z. mobilis Tgt in the presence of guanine hydrochloride and determined the complex structure at a resolution of 1.77 Å (Table 2). In the electron density map the bound ligand was clearly defined. As observed for the Tgt·BIH complex, the Leu²³¹/Ala²³² peptide bond was present in the NH exposing conformation stabilised by a strong H-bond (2.4 Å) which was formed between the backwards oriented Leu²³¹ carbonyl oxygen and the protonated carboxyl group of Glu²³⁵. Also in absolute accordance with the situation in the Tgt·BIH complex an interstitial water was bound H-bonding both the Ala²³² main chain amide and the guanine N7 (Figure 3(a)). Thus, the presented structure confirmed that binding of guanine by Tgt leads to the same Leu²³¹/Ala²³² peptide bond orientation and Glu²³⁵ conformation as binding of the BIH inhibitor.

Table 2.

Crystallographic data collection and refinement statistics

Crystal data	wild type·guanine (2PWU)	Glu235Gln pH 5.5 (2OKO)	Glu235Gln pH 8.5 (2Z1V)	Glu235Gln·guanine (2POT)
A. Data processing
Space group	C2	C2	C2	C2
a, b, c (Å)	90.2, 64.5, 70.8	90.4, 65.2, 70.5	90.8, 65.2, 70.5	89.1, 63.7, 71.1
β (deg.)	92.8	96.4	96.1	92.9
Resolution rangea (Å)	50-1.77 (1.80 -1.77)	50-1.50 (1.53-1.50)	20–1.55 (1.58–1.55)	50-1.80 (1.83-1.80)
Total no. of reflections	105,401	210,241	132,929	90,611
No. of unique reflections	36,424	59,104	56,093	33,979
Completenessa (%)	94.8 (83.9)	91.5 (85.1)	95.6 (78.5)	97.6 (83.4)
Redundancy	2.8	3.5	2.3	2.5
R(I)syma,b (%)	4.4 (38.9)	3.2 (15.9)	5.8 (23.9)	4.1 (46.7)
I/σ (I) a	22 (2.3)	33 (8.0)	16 (3.8)	21 (1.5)
B. Refinement
Rworkc/Rfreed (%)	20.0/27.1	12.5/17.4	14.3/20.3	21.9/28.6
No. of atoms/residues (molecules)
Protein	2,702/358	2,824/365	2,860/369	2,688/355
Water	101	346	305	86
Glycerol (cryo buffer)	18/3	96/16	24/4	42/7
Ligand	14/1	---	---	11/1
Mean B-factors (Å2)
Protein	32.8	21.8	21.6	39.0
Water	34.0	34.8	31.3	40.4
Glycerol (cryo buffer)	52.6	48.1	39.4	62.6
Ligand	60.4	---	---	53.5
rmsd angle (°)	2.0	2.2	2.2	2.0
rmsd bond (Å)	0.007	0.011	0.010	0.006
(An)isotropic refinement	isotropic	anisotropic	anisotropic	isotropic
Crystal data	Glu235Gln·BIH (2Z1W)	Glu235Gln·preQ0 (2PWV)	Glu235Gln·preQ1 (2Z1X)
A. Data processing
Space group	C2	C2	C2
a, b, c (Å)	90.9, 65.3, 70.6	90.5, 65.2, 70.4	89.0, 64.2, 70.6
β (deg.)	96.7	96.3	93.1
Resolution rangea (Å)	50-1.63 (1.66-1.63)	20–1.70 (1.73–1.70)	50-1.63 (1.66-1.63)
Total no. of reflections	170,346	124,483	103,473
No. of unique reflections	47,068	42,679	42,180
Completenessa (%)	92.5 (83.9)	98.3 (96.9)	89.8 (93.1)
Redundancy	3.6	2.8	2.3
R(I)syma,b (%)	3.2 (17.2)	5.9 (49.1)	7.8 (31.6)
I/σ(I) a	33 (8.3)	12 (2.2)	24 (2.0)
B. Refinement
Rworkc/Rfreed (%)	13.2/19.4	20.1/24.5	15.3/21.6
No. of atoms/residues (molecules)
Protein	2,801/362	2,799/361	2,790/364
Water	298	228	172
Glycerol (cryo buffer)	60/10	36/6	48/8
Ligand	15/1	13/1	13/1
Mean B-factors (Å2)
Protein	23.5	22.8	25.7
Water	34.0	28.6	32.9
Glycerol (cryo buffer)	54.8	48.5	53.0
Ligand	22.2	24.9	25.2
rmsd angle (°)	2.1	2.1	2.1
rmsd bond (Å)	0.010	0.009	0.008
(An)isotropic refinement	anisotropic	isotropic	anisotropic

^anumber in parentheses is for highest resolution shell

^bR(I)_sym = Σ |I −〈I〉|/ΣI with I representing the observed intensities and 〈I〉 representing the mean observed intensity

^cR_work = Σ_hkl |F_o − F_c|/Σ_hkl |F_o|.

^dR_free was calculated as for R_work but on 5 % of the data excluded from the refinement

Enzymatic characterisation of wild type Z. mobilis Tgt

Since no kinetic data were available for the Z. mobilis Tgt enzyme regarding preQ₀ and preQ₁ we determined and accordingly redetermined k_cat and K_M concerning guanine, preQ₀, preQ₁ and tRNA. Redetermination of the Michaelis/Menten parameters for guanine and tRNA was done by means of a well established assay monitoring the incorporation of radio-labelled guanine into tRNA.²¹ Since no radio-labelled preQ₀ and preQ₁ is commercially available, we used for these substrates a guanine washout assay which in essence was first described by Hoops et al. ²⁰ Here, the decrease of radioactivity in tRNA labelled in position 34 with [8-³H]-guanine due to the incorporation of the respective substrate is monitored. The kinetic parameters for all four substrates are summarised in Table 3. With respect to preQ₀ and preQ₁ the six-fold difference in K_M reported for the E. coli enzyme was not observed for Z. mobilis Tgt. Within the experimental errors the K_M-values of preQ₀ and preQ₁ were virtually identical and did not differ from that measured for guanine. Instead, the k_cat-values of preQ₀ and preQ₁ differed significantly by a factor of almost 10. While, compared to guanine, preQ₁ was incorporated into tRNA with a higher rate, the incorporation rate of preQ₀ was substantially lower than that of guanine. Obviously, in contrast to E. coli Tgt, in Z. mobilis Tgt the preferred insertion of preQ₁ into tRNA is not achieved by a higher affinity of preQ₁ to the enzyme compared to preQ₀, but by a clearly higher turnover number.

Table 3.

Kinetic parameters for wild type Tgt and Tgt(Glu²³⁵Gln)

wild type Tgt	tRNATyr*	[3H]-guanine	preQ1	preQ0
KM [μM]	0.9 ± 0.2	1.2 ± 0.2	0.7 ± 0.2	0.9 ± 0.2
kcat [s−1] **	5.4 · 10−2	5.6 · 10−2	10.2 · 10−2	1.2 · 10−2
kcat/KM [μM−1s−1]	6.0 · 10−2	4.6 · 10−2	14.6 · 10−2	1.3 · 10−2
Tgt(Glu235Gln)	tRNATyr *	[3H]-guanine	preQ1	preQ0
KM [μM]	1.0 ± 0.1	3.3 ± 0.3	32 ± 7	< 0.5
kcat [s−1] **	7.0 ·10−2	7.6 ·10−2	10.0 ·10−2	0.6 ·10−2
kcat/KM [μM−1s−1]	7.0 ·10−2	2.4 ·10−2	0.3 ·10−2	> 1.2 ·10−2

^*using [3H]-guanine as 2nd substrate

^**taking into account the presence of Tgt as a homodimer able to bind and convert only one substrate tRNA molecule at a time.³⁴

Construction and enzymatic characterisation of Z. mobilis Tgt(Glu²³⁵Gln)

To further investigate the influence of the Glu²³⁵ carboxyl on the Leu²³¹/Ala²³² peptide switch and thus on substrate selectivity we introduced a Glu²³⁵Gln mutation into Z. mobilis Tgt by site directed mutagenesis. The carboxamide of the mutated residue was intended to mimic a permanently protonated and electrically neutral Glu²³⁵ side chain carboxyl. Subsequently, we determined Michaelis/Menten parameters for the mutated enzyme concerning guanine, preQ₀, preQ₁ and tRNA. The results are summarised in Table 3. As expected, the mutation had no significant influence on the affinity of the tRNA substrate reflected by a virtually unchanged K_M compared to wild type Tgt. A slight increase of k_cat (7.0· 10⁻² s⁻¹ vs. 5.4 · 10⁻² s⁻¹ measured for the wild type enzyme) using guanine as second substrate was not considered significant with respect to the error range of the assay. Alike, the Glu²³⁵Gln mutation obviously had no significant influence on the turnover numbers for any of the three substrate bases. The k_cat values regarding guanine, preQ₀ and preQ₁ determined for the mutated enzyme agreed well with those determined for the wild-type enzyme. In contrast, however, the mutation caused a measurable change in K_M with respect to the substrate bases. While K_M(guanine) was only marginally changed, the most drastical impact of the mutation was observed for K_M(preQ₁). Although this base was still accepted as a substrate, K_M was increased by a factor of almost 50 suggesting a noticeable decrease in affinity. Unlike guanine and preQ₁, preQ₀ seemed to bind with an increased affinity to the mutated enzyme as K_M for this substrate was, compared to wild type Tgt, lowered from 0.9 μM to less than 0.5 μM. Due to the slow turnover rate achieved by Tgt, in particular with respect to this substrate, an exact determination of K_M was not possible. To assure an excess of substrate even at low concentrations the amount of enzyme used in the assay would have had to be kept too low to obtain a reasonable signal to noise ratio.

In summary, while the Glu²³⁵Gln mutation had no significant influence on the turnover number for any of the three substrate bases, K_M was drastically increased for preQ₁, marginally elevated for guanine and obviously lowered (even though to an unknown extent) for preQ₀. Thus, regarding k_cat/K_M as a measure of catalytic efficiency, the Glu²³⁵Gln mutation resulted in a clearing or even inversion of substrate preference from preQ₁ to preQ₀ (see Table 3). This agrees well with the hypothesis, that the Leu²³¹/Ala²³² peptide conformation necessary for guanine and preQ₀ accommodation is in wild type Tgt stabilised by the protonated state of Glu²³⁵, which was mimicked by the Glu²³⁵Gln mutation.

Crystal structure analyses of Tgt(Glu²³⁵Gln)

To allow the interpretation of the measured enzyme kinetic data at a structural level, we crystallised the mutated enzyme in its apo form as well as in complex with its substrates and the guanine analogous inhibitor BIH (Table 2).

Two structures of ligand-free Tgt(Glu²³⁵Gln) derived from crystals grown at pH 5.5 and pH 8.5, respectively, were determined at nominal resolutions of better than 1.6 Å. In the electron densities of both structures the residues of and near the active site were excellently defined. As expected, in contrast to wild type Tgt no pH-dependent conformational change was observed for the Leu²³¹/Ala²³² peptide bond. Independently whether crystallisation had been performed at pH 5.5 or 8.5, the carbonyl of this peptide bond was oriented backwards while the amide was exposed into the binding pocket. Surprisingly, however, the way this conformation was stabilised by the Gln²³⁵ carboxamide did not correspond to the way it is stabilised by the protonated Glu²³⁵ carboxyl in wild type apo Tgt crystallised at pH 5.5. Rather, the geometry of the binding pocket was virtually identical to the one observed in the structure of wild type Tgt with preQ₀ bound. No H-bond was formed between the Gln²³⁵ carboxamide and the Leu²³¹ carbonyl oxygen. Instead, the carboxamide stacked between the Leu²³¹/Ala²³² peptide bond and the side chain aromatic ring system of Tyr¹⁶¹. As a consequence, the Gln²³⁵ side chain was shifted towards the binding pocket concomitant with a Gln²³⁵/Gly²³⁶ peptide flip so far only observed in preQ₀ bound wild type Tgt (with Gln²³⁵ being replaced by Glu²³⁵). To investigate the influence of a bound substrate on the active-site geometry of Tgt(Glu²³⁵Gln), we determined the crystal structures of the mutated enzyme in complex with preQ₀, guanine, the BIH inhibitor, and preQ₁, respectively.

The Tgt(Glu²³⁵Gln) · preQ₀ complex structure was determined at a resolution of 1.70 Å. The electron density for the preQ₀ ligand was clearly defined. Superposition of the Tgt(Glu²³⁵Gln) · preQ₀ complex structure and the structures of the ligand-free mutated enzyme revealed that the binding pocket geometry was virtually identical in all three structures (Figure 4(a)) and exactly corresponded to the one observed in preQ₀ bound wild type Tgt (Figure 3(e)). The nitrile nitrogen of preQ₀ was present at a position occupied by a glycerol oxygen picked up from the cryo buffer in the structures of uncomplexed mutated Tgt (Figure 4(a)).

Figure 4. Crystal structures of Tgt(Glu²³⁵Gln).

(a) Superposition of Tgt(Glu²³⁵Gln) in its apo form (crystallised at pH 5.5; the structure obtained at pH 8.5 looks identical) and in complex with preQ₀ reveals equal binding pocket geometries. (b) 2 |F_o| − |F_c| electron density map of Tgt(Glu²³⁵Gln) in complex with guanine contoured at 1.5 σ (blue). The green density represents an |F_o| − |F_c| map with the ligand omitted from the calculation contoured at 3.0 σ. (c) 2 |F_o| − |F_c| electron density map of Tgt(Glu²³⁵Gln) in complex with preQ₁ contoured at 1.5σ (blue). The green density represents an |F_o| − |F_c| map with the ligand omitted from the calculation contoured at 3.0 σ. The electron density reveals a split conformation for the Asp¹⁰² side chain indicating an uncomplete occupancy of the binding pocket with the substrate base. The Leu²³¹/Ala²³² peptide bond is present in two different conformations. The occupancy of the ligand and of the predominant Leu²³¹/Ala²³² and Asp¹⁰² conformers refine to 0.7.

The crystal structures of Tgt(Glu²³⁵Gln) in complex with guanine and with the guanine analogous inhibitor BIH were determined at a resolution of 1.80 Å and 1.63 Å, respectively. In both structures the electron densities for the ligands as well as for the residues constituting the binding pockets were excellently defined (for the guanine complex see Figure 4(b)). It was clearly visible that also binding of guanine or BIH did, compared to the apo-form, not result in any conformational change of the Leu²³¹/Ala²³² peptide bond or of Gln²³⁵. As in guanine or BIH-bound wild-type Tgt, the exposed Ala²³² main chain amide formed an H bond to an interstitial water molecule, which was further H-bonded to N7 of guanine or the corresponding nitrogen of BIH, respectively. In contrast to wild type Tgt, however, this conformation was not stabilised by an H-bond formed between the Leu²³¹ carbonyl and the carboxamide of Gln²³⁵ but by a stacking interaction between the Leu²³¹/Ala²³² peptide bond and the Gln²³⁵ side chain.

In order to understand the way preQ₁ was bound by the mutated Tgt we determined the crystal structure of the Tgt(Glu²³⁵Gln) · preQ₁ complex at a resolution of 1.63 Å. Also in this structure the electron density attributable to the ligand was very well defined. A split conformation of the Asp¹⁰² side chain, however, indicated an incomplete occupancy of the binding pocket with preQ₁ (Figure 4(c)). In apo structures of Tgt the Asp¹⁰² side chain carboxylate is oriented towards the outside of the binding pocket H-bonding the side chain amide of Asn⁷⁰ as well as the main chain amide of Thr⁷¹. Once a substrate base has bound the Asp¹⁰² side chain rotates into the binding pocket towards the ligand in order to H-bond its exocyclic amine at position 3.²² In the Tgt(Glu²³⁵Gln) · preQ₁ complex structure ca. 30 % of Asp¹⁰² exhibited a side chain conformation pointing towards the outside of the binding pocket suggesting a portion of the binding pocket being unoccupied by the ligand. This was in line with the fact that K_M(preQ₁) was drastically increased for Tgt(Glu²³⁵Gln). As in all other structures of Tgt(Glu²³⁵Gln) the Gln²³⁵ side chain stacked on top of the Leu²³¹/Ala²³² peptide bond. Accordingly, the conformation of the Gln²³⁵/Gly²³⁶ peptide bond remained also here unchanged compared to all other Tgt(Glu²³⁵Gln) structures as well. Interestingly, however, the electron density attributable to the Leu²³¹/Ala²³² peptide bond could only been interpreted such that it was present in two different conformations. One conformer corresponded exactly to the one observed in the remaining Tgt(Glu²³⁵Gln) structures with the main chain amide of Ala²³² being exposed to the binding pocket. Indeed, this conformation was hardly compatible with a bound preQ₁ molecule due to unfavourable interactions between the Ala²³² main chain amide and the amino methyl group of preQ₁. The fact that this conformation could be refined to an occupancy of ca. 30 % agreed well with the assumption that a portion of the binding pocket was devoid of ligand. In the second conformer the Leu²³¹/Ala²³² peptide bond was flipped enabling the exposed Leu²³¹ carbonyl oxygen to H-bond the amino methyl group of preQ₁. This geometry combining the exposed Leu²³¹ carbonyl with the stacking conformation of Gln²³⁵ (or Glu²³⁵) had never been observed in any Tgt crystal structure before. As a consequence, exposure of the Leu²³¹ carbonyl into the binding pocket does not necessarily require the deprotonated carboxyl of Glu²³⁵ H-bonding the backwards oriented Ala²³² main chain amide. It is also compatible with a stacking interaction of the Gln²³⁵ side chain, although this mode of stabilisation is obviously energetically less favourable as reflected by the drastically increased K_M(preQ₁) measured for Tgt(Glu²³⁵Gln).

Discussion

In this study we have confirmed and further analysed the role of Glu²³⁵ as a general acid/base not directly involved in catalysis but facilitating the binding of various substrates with different H-donor and -acceptor properties. The high resolution crystal structure of Z. mobilis Tgt in complex with guanine presented in this work gives the final proof that binding of each of the three substrate bases (guanine, preQ₀ and preQ₁) triggers a different geometry of the Tgt binding pocket. Prerequisite of the substrate promiscuity observed in Tgt is the ability of the Leu²³¹/Ala²³² peptide bond to flip and thus to change its orientation depending on the bound substrate. Remarkably, the bound substrate not only influences the orientation of this peptide bond but, moreover, the way it is stabilised by the strictly conserved Glu²³⁵. Obviously, the highly flexible interplay of the Leu²³¹/Ala²³² peptide bond and the general acid/base Glu²³⁵ facilitates an optimal adaptation of the Tgt binding pocket to each of its substrate bases. This is reflected by the virtually identical K_M values of Z. mobilis Tgt for guanine, preQ₀ and preQ₁ as determined in this study. Moreover, our results show that the preferred incorporation of the “correct” substrate preQ₁ into tRNA over its biosynthetic precursor preQ₀ is achieved by a ca. 10 fold faster turnover rate. It should be noted, however, that to the best of our knowledge it has not been investigated so far, if QueF, the nitrile reductase reducing preQ₀ to preQ₁, is also acting on preQ₀ incorporated into tRNA. In this case preQ₀ could represent an in vivo substrate of Tgt as well. The observed differences in k_cat for the three substrate bases may hardly be explained by their binding modes observed in the corresponding crystal structures, but possibly may be deduced from their chemical properties. In each case nitrogen N9 of the respective base nucleophilically attacks C1 of the covalently bound tRNA ribose³⁴ during catalysis (Figure 2(c) and Table 1). In preQ₁ this nitrogen is the most potent nucleophile, as it is in no electronic conjugation with the exocylic amino methyl group. In guanine, N9 is less nucleophilic due to an electron withdrawing effect of N7 being in conjugation with N9. The preQ₀ substrate is the least reactive base in this series. Here, the exocyclic nitrile function is conjugated to the endocyclic nitrogen, resulting in an even more pronounced electron withdrawing effect. This is fully consistent with the observed decrease in turnover in the order preQ₁ > guanine > preQ₀. The virtually identical K_M-values for the three substrate bases indicate that the conformation of the Leu²³¹/Ala²³² peptide bond as well as the way it is stabilised by the Glu²³⁵ side chain carboxyl may have little influence on the energetic state of the substrate binding pocket.

With intent to create a bacterial Tgt with altered substrate specificity Glu²³⁵ was mutated to Gln in an attempt to mimic a permanently protonated Glu²³⁵ side chain carboxyl. Since the Leu²³¹/Ala²³² peptide bond conformation required for the accommodation of preQ₁ involves stabilisation by a deprotonated Glu²³⁵ the mutation was expected to strongly disfavour or even disable preQ₁ binding while leaving unaffected or even improving the binding of guanine and preQ₀. Kinetic characterisation of Tgt(Glu²³⁵Gln) showed that the change of K_M(guanine) caused by the mutation was at the border of significance, while K_M(preQ₀) seemed to be (to an unknown extent) decreased compared to wild type Tgt. Although K_M(preQ₁) of the mutated enzyme was increased by a factor of almost 50, the remaining affinity to this substrate was still unexpectedly high. The crystal structures of Tgt(Glu²³⁵Gln) in its apo-form as well as in complex with various ligands provided an explanation to the observed kinetic data. The same geometry of the binding pocket was observed, no matter if the mutated enzyme was crystallised in its apo-form (independent of the applied pH) or in complex with guanine, BIH, or preQ₀. In each case, the Ala²³² main chain amide was exposed to the binding pocket, stabilised by a stacking interaction with the Gln²³⁵ side chain carboxamide. This conformation was analogous to that observed in wild ₂₃₁ type Tgt bound to preQ₀. Seemingly, a surprisingly strong H-bond formed between the Leucarbonyl and the protonated Glu²³⁵ side chain carboxyl (2.4 Å) plays an important role in stabilising the binding pocket geometry observed in wild type Tgt crystallised at pH 5.5 or in complex with guanine/BIH. Due to its less electronegative nitrogen and the presence of two hydrogens over which the positive partial charge is distributed the carboxamide of Gln²³⁵ is obviously not able to form as strong an H-bond to the Leu²³¹ carbonyl rendering the stacking interaction energetically more favourable. Hence, the binding pocket of Tgt(Glu²³⁵Gln) seems predestined for the accommodation of preQ₀, since independent of any pH no structural rearrangement is required for its binding. A priori the binding pocket is present in a geometry identical to the one induced upon preQ₀ binding in wild-type Tgt. This finding is supported by the (even though to an unknown extent) reduced K_M(preQ₀) determined for Tgt(Glu²³⁵Gln). As a surprise, the crystal structure of Tgt(Glu²³⁵Gln) in complex with preQ₁ revealed a binding pocket geometry featuring an exposed Leu²³¹ carbonyl stabilised by a Gln²³⁵ stacking interaction. This unexpected geometry showed that the stacking interaction still allows the Leu²³¹/Ala²³² peptide flip. As a result, the substrate specificity achieved by the Glu²³⁵Gln mutation was not as strict as that observed for archaeal Tgt which is absolutely unable to bind preQ₁. It permanently exposes the amide group of the Val¹⁹⁷/Val¹⁹⁸ peptide bond (corresponding to Leu²³¹/Ala²³² in bacterial Tgt) allowing both the binding of guanine and of preQ₀ but not of preQ₁. The peptide flip necessary to accommodate this base is rendered impossible, since the opposing carbonyl of this peptide bond is involved in an invariant backbone H-bond network which does not imply the carboxyl group of an acidic amino acid^1,23 (Figure 3(f)).

In summary, the mutated Tgt(Glu²³⁵Gln) investigated in this study behaved in several details different from what was expected, although a more conservative exchange than glutamic acid/glutamine can hardly be performed. The present work clearly shows the indispensability of crystal structure analyses for the correct interpretation of kinetic data obtained from mutated proteins.

Furthermore, this study provides an example that mutation of a residue which does not even directly interact with any ligand enables the inversion of substrate selectivity of an enzyme. This may generally underline the importance to consider residues in the “second sphere” of binding pockets during attempts of creating enzymes with new binding properties.

The successful modulation of selectivity is an important step towards the understanding of selectivity and specificity determining features in Tgts. In E. coli Tgt replacement of Asp¹⁵⁶ (Z. mobilis numbering) by Asn altered the specificity towards xanthine even though at the expense of a reduced catalytic activity.²⁴ In this context it might be interesting to study mutational exchanges in the close neighbourhood of the attacking nucleophile Asp²⁸⁰. In bacterial Tgt this residue is tightly kept in position by H-bonds formed with Gly²⁶¹ and Tyr^258.17 In archaeal Tgt the tyrosine is replaced by a strictly conserved histidine.¹ Likely, this exchange will have pronounced influence on the catalytic properties.

These considerations provide a perspective towards an ambitious goal: the modification of substrate specificity towards bases other than preQ₀ or preQ₁ and their efficient incorporation into the wobble position of tRNAs. This would allow to study the translational process in more detail via well designed modulation of the accuracy of the wobble base pairing.

Materials and Methods

Cloning and Tgt preparation

The QuikChange^™ site-directed mutagenesis kit (Stratagene) was used to introduce a Glu²³⁵Gln mutation into the wild-type tgt expression plasmid pET9d-ZM4²⁵ following the vendor’s protocol. The primers used for mutagenesis were E235Q-s (5′-GGG GGA TTG GCT GTG GGT CAA GGA CAG GAT GAA ATG-3′) and E235Q-a (5′-CAT TTC ATC CTG TCC TTG ACC CAC AGC CAA TCC CCC-3′) (mutation underlined). Sequencing of the entire tgt gene (MWG Biotech, Ebersberg) confirmed the presence of the desired mutation as well as the absence of any further unwanted mutation. Subsequently, the mutated plasmid pET9d-ZM4-E235Q was transformed into E. coli BL21(DE3) pLysS cells. These cells were used for the preparation of the Tgt enzyme following the method described by Romier et al. .²⁶

Preparation of tRNA^Tyr

Preparation of E. coli tRNA^Tyr (ECY2) via in vitro transcription was done following the method described by Curnow et al. .²⁷

Kinetic parameters

Kinetic parameters for Tgt and Tgt(Glu²³⁵Gln) have been determined for the different substrates essentially as described in Meyer et al. .²¹

Michaelis-Menten parameters for tRNA and guanine were determined monitoring incorporation of radioactively labelled guanine into tRNA^Tyr in position 34. Kinetic data for both substrates were determined separately in triplicate and average values were calculated. Kinetic parameters for guanine were measured using 150 nM Tgt, 15 μM unlabelled tRNA^Tyr, and variable concentrations of [8-³H]-guanine (0.9 Ci/mmol; Hartmann Analytic) in the range of 0.5 – 20 μM in buffer solution (200 mM HEPES pH 7.3, 20 mM MgCl₂) and 2.95 μM (≡ 5 % of critical micellar concentration) Tween 20 (Roth). Kinetic parameters for tRNA were measured using 150 nM Tgt, 20 μM [8-³H]-guanine, and variable concentrations of tRNA^Tyr (0.25 – 15 μM). Initial velocities of the base exchange reaction in counts per minute were converted to [μM/min] using a calibration constant derived from liquid scintillation counting of [8-³H]-guanine solutions with variable concentrations. Kinetic parameters were determined via double-reciprocal linearization using the method of Eadie-Hofstee and linear regression using GraFit.²⁸

Michaelis-Menten parameters for preQ₀ and preQ₁ were calculated via monitoring the loss of [8-³H]-guanine from tRNA^Tyr radio-labelled in position 34. To produce radioactively labelled tRNA 50 μM unmodified tRNA^Tyr was incubated with 0.5 μM Tgt and 10 μM [8-³H]-guanine (11.8 Ci/mmol; Hartmann Analytic) in Tgt assay buffer for 2 h. Tgt and free guanine/[8-³H]-guanine were extracted from the reaction mixture by the addition of equal volumes of Roti-Phenol (Roth) and chloroform : isoamylalcohol (24:1). The aqueous supernatant was once more treated with an equal volume of chloroform : isoamylalcohol (24:1). From the aqueous supernatant containing the radioactively labelled tRNA samples were retained to determine the specific activity of the radio-labelled tRNA. An additional purification step was performed via gel filtration using NAP-columns (GE Healthcare, Life Science) and Tgt assay buffer. From the eluted solution again samples were retained to derive the final concentration of the radio-labelled tRNA from its specific activity. For each preparation it amounted to ca. 20 μM. Kinetic parameters for preQ₀ and preQ₁ were measured using 150 nM wild type Tgt or Tgt(Glu²³⁵Gln) and 8 μM radio-labelled tRNA^Tyr. The concentration of preQ₀ was varied in a range of 0.5 – 15 μM. For preQ₁ variable concentrations in a range of 0.5 – 15 μM for wild type Tgt and 2 – 80 μM for Tgt(Glu²³⁵Gln) were applied. Initial velocities in counts per minute were calculated from the decreasing tritium labelling level of tRNA due to the incorporation of the respective substrate bases. Initial velocities in counts per minute were converted to [μM/min] using a calibration constant derived from liquid scintillation counting of guanine/[8-³H]-guanine solutions with variable concentrations. Kinetic parameters were determined via double-reciprocal linearization using the method of Eadie-Hofstee and linear regression using GraFit.²⁸

Crystal structure analyses

Tgt(Glu²³⁵Gln) crystals suitable for ligand soaking were produced in a two step procedure. Droplets were prepared by mixing 2 μL of concentrated protein solution (14 mg/mL) with 2 μL reservoir solution [100 mM MES, pH 5.5, 1 mM DTT, 13 % (w/v) PEG 8,000, 10 % (v/v) DMSO]. Micro-crystals were grown at 291 K using the hanging-drop vapour diffusion method in the presence of 1.0 mL of reservoir solution of the respective seeding buffer. Micro crystals grew within two weeks.

Subsequently, macro-seeding was performed under similar conditions to obtain crystals of the uncomplexed protein. Again droplets were prepared by mixing 2 μL of concentrated protein solution with 2 μL macro-seeding reservoir solution [100 mM MES, pH 5.5, 1 mM DTT, 8 % (w/v) PEG 8,000, 10 % (v/v) DMSO]. One micro-crystal was transferred into this solution. Single crystals with a size of approximately 0.7 × 0.7 × 0.2 mm³ grew within two to four weeks.

To obtain crystals of apo-Tgt(Glu²³⁵Gln) at pH 8.5, buffers were used identical to those described above but with 100 mM TrisHCl pH 8.5 instead of 100 mM MES pH 5.5 as buffer system.

To allow cocrystallization of the mutated Tgt in complex with preQ₀ the compound was dissolved in DMSO and added to the macro-seeding droplet to a final concentration of 2 mM. Crystals of mutated Tgt complexed to preQ₁, BIH and guanine as well as crystals of wild type Tgt bound to guanine were produced by a soaking procedure. The compounds were dissolved in DMSO and added to 2 μL of a crystallization droplet to a final concentration of 10 to 20 mM. Finally, a single Tgt crystal was transferred into the droplet and soaked for 30 min.

For data collection, crystals were cryoprotected using glycerol. The glycerol and PEG 8,000 concentrations of the macro-seeding buffers were increased stepwise by transferring the crystals to six different 2 μL cryo-droplets each with 5 to 30 min incubation times [glycerol concentrations (v/v): 5 % → 10 % → 15 % → 20 % → 25 % → 30 %; and PEG 8,000 concentrations (w/v): 5.0 % → 6.3 % → 7.5 % → 8.0 % → 8.8 % → 9.8 %, respectively]. These droplets also contained the ligands at equivalent concentrations compared to soaking and co-crystallization conditions. The cryo-soaked crystals were flash-frozen in liquid N₂. Data sets were collected at cryo conditions (100 K) with CuK_α radiation (λ = 1.5418 Å) using a Rigaku RU-300 rotating-anode generator at 50 kV and 90 mA equipped with Xenocs focussing optics and an R-AXIS IV detector. All crystals tested exhibited monoclinic symmetry in space group C2 containing one monomer per asymmetric unit with Matthews coefficients of 2.3 – 2.4. Data processing and scaling was performed using the HKL2000 package.²⁹ For all refined structures unit cell dimensions for the crystals, data collection and processing statistics are given in Table 2.

For crystals grown at pH 5.5 coordinates of the apo Tgt crystal structure grown at pH 5.5 (PDB-code: 1P0D) were directly applied for initial rigid-body refinement of the protein molecule followed by repeated cycles of conjugate gradient energy minimisation, simulated annealing and B-factor refinement using the CNS program package.³⁰ For the crystal grown at pH 8.5 the coordinates of the apo Tgt crystal structure grown at pH 8.5 (PDB-code: 1PUD) were applied. Refinement at the later stages for all other structures was performed with SHELXL.³¹ Here, up to 50 cycles of conjugate gradient minimisation were performed with default restraints on bonding geometry and B-values: 5 % of all data were used for R_free calculation. Amino acid side chains were fit to σA-weighted 2 |F_o|−|F_c| and |F_o|−|F_c| electron density maps using Coot.³² Water, glycerol molecules, and ligands were located in the difference electron density and added to the model for further refinement cycles. Anisotropic conjugate gradient refinement resulted in a significant improvement of the models built for apo Tgt(Glu²³⁵Gln) (pH 5.5 and 8.5), Tgt(Glu²³⁵Gln)·BIH, and Tgt(Glu²³⁵Gln)·preQ_1. During the last refinement cycles, riding H-atoms were introduced for the protein residues (not for the ligand) without using additional parameters. All final models were validated using PROCHECK.³³ Data refinement statistics are given in Table 2.

Protein Data Bank Accession Codes

The following Protein Data Bank (PDB) accession codes were allocated to the crystal structures determined during this work:

Tgt·guanine: 2PWU; Tgt(Glu²³⁵Gln) pH 5.5: 2OKO; Tgt(Glu²³⁵Gln) pH 8.5 2Z1V; Tgt(Glu²³⁵Gln) · guanine 2POT; Tgt(Glu²³⁵Gln) · BIH 2Z1W; Tgt(Glu²³⁵Gln) · preQ₀ 2PWV; Tgt(Glu²³⁵Gln) · preQ₁: 2Z1X (also given in Table 2).

Alignment and Figures

Alignment of structures with similar or identical sequences was performed with the alignment function implemented in Pymol (http://www.pymol.org).

Figures were prepared using Isis Draw (MDL, San Leandro, USA) and Pymol.