The Escherichia coli tRNA-guanine transglycosylase can recognize and modify DNA.

tRNA-guanine transglycosylase (TGT) catalyzes the exchange of queuine (or a precursor) for guanine 34 in tRNA. The minimal RNA recognition motif for TGT has been found to involve a UGU sequence in the anticodon loop of the queuine-cognate tRNAs. Recent studies have shown that the enzyme is capable of recognizing the UGU sequence in alternative contexts (Kung, F. L., Nonekowski, S., and Garcia, G. A. (2000) RNA 6, 233-244) and have investigated the role of the first U of the UGU sequence in tRNA recognition by TGT (Nonekowski, S. T., and Garcia, G. A. (2001) RNA 7, 1432-1441). The TGT reaction involves the breakage and re-formation of a glycosidic bond. To rule out a potential chemical mechanism involving the 2'-hydroxyl at position 34, we synthesized and evaluated an RNA minihelix with 2'-deoxy-G at 34. The high level of activity exhibited by this analogue indicates that the 2'-hydroxyl of G(34) is not required for catalysis. Furthermore, we find that TGT can recognize analogues composed entirely of DNA, but only when 2'-deoxyuridines replace the thymidines in the DNA. The requirement for uridine bases for recognition is perhaps not surprising given the UGU recognition motif for TGT. However, it is not clear if the uracil requirement is due to specific recognition by TGT or due to the effect of uracils on the conformation of the oligonucleotide.

tRNAs contain a large number of modified nucleosides (1). One of the more elaborately modified nucleosides is queuine (7-(4,5-cis-dihydroxy-1-cyclopenten-3-yl-aminomethyl)-7-deazaguanine). tRNA-guanine transglycosylase (TGT) 1 catalyzes the exchange of queuine (or a precursor) for guanine 34 in the anticodon of certain tRNAs. The minimal RNA recognition motif for TGT has been found to involve a UGU sequence in the anticodon loop of the queuine-cognate tRNAs (tyrosine, aspartate, asparagine, and histidine) (2,3). The UGU sequence is undeniably the major determinant for tRNA recognition by TGT. However, provided that TGT can position the UGU sequence in the active site in the proper orientation, the UGU sequence need not reside in the anticodon loop to be recognized. Recent studies have shown that TGT can recognize the UGU sequence in at least 2 additional minihelical contexts, at the base of the T⌿C stem in yeast tRNA Phe and in the anticodon position in the absence of U33 (4,5). Thus, tRNA recognition by TGT is more flexible than previously believed. This observation prompted an examination of the ability of TGT to recognize RNAs containing modifications of the UGU sequence. The initial studies that identified the UGU sequence as the major identity element utilized only canonical base (C, G, A, U) replacements (2,3). Previous experiments have demonstrated that a DNA analogue of an RNA minihelix corresponding to the anticodon arm of Escherichia coli tRNA Tyr , ECYMH (dEC-YMH) was inactive. 2 However, there are two fundamental differences between RNA and DNA, the lack of the 2Ј-hydroxyls and the presence of thymidine rather than uracil in DNA. Therefore, dECYMH has a TGT sequence rather than a UGU sequence.
Given the importance of the UGU sequence in TGT recognition, it is possible that the inactivity of dECYMH was due to the presence of thymidine rather than the loss of the 2Ј-hydroxyls. To investigate the role of the 2Ј-hydroxyl in TGT recognition and catalysis, we have studied a deoxyguanosine 34 analogue of ECYMH ( Fig. 1, dG 34 ECYMH). This analogue is a substrate for TGT with less than a 10-fold reduction in activity. To further probe the ability of TGT to recognize RNA analogues lacking the 2Ј-hydroxyl, modified DNA analogues of the previously described minihelix ECYMH (2, 6) ( Fig. 1, dUdECYMH) and the alternative TGT minihelical substrates UGU ϩ1 (dUdUGU ϩ1 ) (4) and SCFMH(T⌿C) (dUdSCFMH(T⌿C)) (5) were synthesized and characterized. These analogues (containing deoxyuracil (dU) bases rather than thymidine bases) all serve as substrates for TGT, indicating that the tRNA-guanine transglycosylase from E. coli can recognize and modify DNA.
Synthesis of Minihelical Analogues-The RNA minihelix, ECYMH, was chemically synthesized by automated chemical synthesis performed on an Expedite nucleic acid synthesis system (model 8909, PerSeptive Biosystems) using the manufacturer's protocols and reagents. However, the RNA phosphoramidite monomers (G, C, A, and U) and CPG columns were from Glen Research. dG 34 ECYMH was the generous gift of Dr. Houng-Yau Mei (Bioorganic Chemistry, Pfizer Global Research and Development, Ann Arbor, MI). The dU-DNA analogues, dUdECYMH, dUdUGU ϩ1 , and dUdSCFMH(T⌿C) were from Invitrogen. The oligos were resuspended in 300 -800 l of HM 7.3 buffer (10 mM HEPES, pH 7.3, 1 mM MgCl 2 ). Concentrations of the analogues were determined spectrophotometrically using the extinction coefficients at 260 nm calculated from the base composition of each oligonucleotide.
Polyacrylamide Gel Electrophoresis-All native and denaturing PAGE were performed on a Phast System (Amersham Biosciences) as previously reported (2,7). Typical native band shift assays were performed as follows: TGT (3 M) was incubated with excess RNA (45-100 M) at 37°C for 30 min in a 10-l reaction mixture containing 10 mM HEPES, pH 7.3, 1 mM MgCl 2 , 1 mM dithiothreitol, and 1 mM sodium phosphate. The reaction mixtures were then analyzed by native PAGE using 8 -25% gradient polyacrylamide gels. Approximately 4 l were loaded onto each lane. The capability of the RNAs to form a stable complex with TGT was assayed via denaturing PAGE as follows: TGT (7 M) was incubated with excess RNA (45-100 M) in the presence of 400 M 9-methylguanine at 37°C for 30 min in a 10-l reaction mixture containing 10 mM HEPES, pH 7.3, 1 mM MgCl 2 , 1 mM dithiothreitol, and 1 mM sodium phosphate. SDS buffer (10 l of 60 mM Tris/HCl, pH 6.8, 2% SDS, 10% glycerol, and 0.01% bromphenol blue) was added to the reaction mixtures, and the incubation was continued for an additional hour at 25°C. Approximately 4 l were loaded onto each lane. In both native and denaturing PAGE, the gels were stained with Coomassie Blue to visualize the protein, although the gels were scanned in gray scale to generate the figures.
Kinetic Analyses-A guanine incorporation assay was used to obtain the steady-state kinetic parameters as previously reported (2,5,8). A slight modification was made to the precipitation step of the assays: the analogues were precipitated for 1 h at Ϫ20°C rather than at room temperature. This improved the assay by reducing the background radioactivity. The dU-DNA analogues, dUdUGU ϩ1 and dUdSCFMH-(T⌿C), were assayed at concentrations ranging from 0.2 to 20 M, whereas the concentration for dUdECYMH ranged from 0.2 to 40 M and the concentration for dG 34 ECYMH ranged from 0.1 to 60 M. The TGT concentration was 180 nM. Aliquots (70 l) were taken at 5, 10, 20, 40, and 80 min after the reaction had been initiated with enzyme and quenched by precipitation with 2 ml of ethanol and 10 l of 3 M sodium acetate, pH 5.3. After the addition of the ethanol, the test tubes containing the samples were covered with parafilm and placed in the freezer (Ϫ20°C) to chill for 1 h. The precipitated RNAs were collected on glass fiber filters (GF/C filter, Whatman), washed three times with EtOH, and dried. The amount of radioactivity on each filter was quantified by liquid scintillation. Disintegrations per minute (DPM) were converted into picomoles of radiolabeled guanine using the appropriate specific activities. The initial velocities (v i ) obtained from linear regression of guanine incorporation versus time were plotted versus substrate concentrations. V max and K m were obtained by nonlinear regression of these hyperbolic plots to the Michaelis-Menten equation. Values for k cat were obtained by dividing the V max values by the TGT concentration (180 nM) and the aliquot volume (70 l). Assays were conducted in triplicate, and the average of the data points (v i ) and the error bars generated from the standard deviation within each point were plotted. The initial velocity of ECYMH at 10 M was determined in all assays to normalize the specific activity of TGT from assay to assay.

RESULTS AND DISCUSSION
The observation that TGT recognition of tRNA is flexible enough to accommodate the UGU sequence in alternate contexts (4,5) prompted the investigation of the capability of TGT to recognize a modified RNA minihelix. A 2Ј-deoxyguanosine 34 analogue of ECYMH (dG 34 ECYMH) was synthesized and shown to be a substrate for TGT with an 8-fold decrease in k cat and a 3-fold decrease in K m with respect to ECYMH (Table I).
The substantial activity of dG 34 ECYMH rules out any requisite participation of the 2Ј-hydroxyl in the chemical mechanism of the TGT reaction. This conclusion is consistent with mutagenesis studies that implicate an aspartic acid residue as the enzymic nucleophile (9, 10). Thus, it seems that any potential H-bonding interaction between the 2Ј-hydroxyl of G 34 and TGT is not critical for binding or activity.
To determine the importance of the remaining 2Ј-hydroxyls, a DNA analogue of ECYMH (dUdECYMH) was synthesized and characterized. This analogue is comprised entirely of 2Јdeoxyribonucleotides. However, because of the importance of the UGU sequence in TGT recognition, it contains deoxyuracil bases in place of thymidine bases. This allows the effect of the ribose backbone to be examined exclusively. In addition to dUdECYMH, the dU-DNA analogues of the alternate substrates UGU ϩ1 (4) (dUdUGU ϩ1 ) and SCFMH(T⌿C) (5) (dUdSCFMH(T⌿C)) ( Fig. 1) were also synthesized and evaluated.
Native PAGE band shift experiments demonstrate that all of the dU-DNA analogues bind to TGT ( Fig. 2A). This indicates that none of the 2Ј-hydroxyls are critical for binding to TGT. The RNA analogue with the single deoxyribose substitution, dG 34 ECYMH, qualitatively exhibited the highest ratio of RNAbound TGT to free TGT ( Fig. 2A, lane 3). It is not clear why the removal of the 2Ј-hydroxyl of G 34 results in apparently tighter binding than the RNA analogue, ECYMH. Although it is unlikely that the overall conformation of dG 34 ECYMH differs significantly from ECYMH, it is possible that a local change in the sugar pucker of the G 34 could be responsible for the apparently tighter binding. 2Ј-deoxyribonucleotides typically favor a C2Ј-endo(N) sugar pucker, whereas ribonucleotides are frequently found in a C3Ј-endo(S) conformation (11). It seems likely that the predominantly RNA nature of dG 34 ECYMH yields a structure that is virtually identical to that of the native RNA substrate for TGT. The almost certain change in conformation at position 34 because of the deoxyguanosine could result in an orientation in the active site that is suboptimal for catalysis, but does not interfere with binding. dUdECYMH and the other dU-DNA analogues do show qualitatively less band shift than either ECYMH or dG 34 ECYMH (Fig. 2). Furthermore, the slightly higher K m values of the dU-DNA analogues are consistent with weaker binding. All of these results suggest that the loss of the 2Ј-hydroxyl at position 34 leads to a reduction in catalysis, possibly because of suboptimal orientation. The loss of the remaining 2Ј-hydroxyls appears to have a very small effect on binding, which may be because of a conformational effect.
All the analogues were able to form a complex that was stable to mild denaturing conditions (ϳ50 kDa, Fig. 2B). Previous reports strongly suggest that this complex is a covalent intermediate formed between RNA and aspartate 89 of TGT (9, 10). Consistent with its tighter binding, dG 34 ECYMH formed a significantly more intense covalent complex (Fig. 2B, lane 3). In fact, dG 34 ECYMH produces the largest amount of complex formation of any substrate analogue (RNA or DNA) tested to date, to our knowledge. Although it has not yet been firmly established that the stable complex represents a true mechanistic intermediate (e.g. chemical and kinetic competence of the covalent complex), there is a direct correlation between the formation of this complex and enzymatic activity (5). This correlation is consistent with our findings that all the dU-DNA analogues are substrates for TGT (Fig. 3). In general, the activity of the dU-DNA analogues mirrored their RNA analogues (Table I). For example, dUdECYMH, which was the most active DNA analogue, was derived from the "normal" RNA substrate ECYMH. The reduction in the k cat values for dUdECYMH (5-fold) and dG 34 ECYMH (7-fold) with respect to ECYMH suggests that the orientation of dG 34 is not optimal for catalysis. The decreased activities of the alternate RNA substrates UGU ϩ1 and SCFMH(T⌿C) demonstrate that catalysis for these analogues is also affected by suboptimal orientation of the UGU sequence as compared with the normal RNA substrate. The essentially identical k cat values of dUdUGU ϩ1 and dUdSCFMH(T⌿C), relative to their respective RNA analogues, indicate that the ribose backbone does not significantly aid in any rearrangement that might enhance catalysis. The most prominent result presented in Table I is that removal of all the 2Ј-hydroxyls has a relatively minor effect (5-to 14-fold reduction) on k cat . Despite the small increases in K m , the overall specificities (k cat /K m ) are quite comparable to the RNA analogues, especially for the alternate substrates UGU ϩ1 and SCFMH(T⌿C).
The activity of the dU-DNA analogues can be explained in one of two ways. The first explanation is that TGT is not sensitive to the differences between RNA and DNA. However, we know that this is not strictly accurate because the analogue containing thymidine (dECYMH) is inactive. Yet, replacement of the thymidine bases with deoxyuracil does restore activity with TGT; thus, TGT recognition is not strictly dependent upon the ribose backbone or any conformational effects of the 2Јhydroxyls. The capability of TGT to recognize the UGU sequence in alternate contexts (4,5) suggests that TGT recognition is fairly indiscriminate. However, recent results from our laboratory (4) show that TGT recognition is blocked when the UGU sequence is locked into certain conformations (e.g. the anticodon loop "U-turn").
An alternative explanation is that the dU-DNA analogues are able to emulate the RNA analogues despite differences in the preferred conformations of deoxyribonucleotides versus ribonucleotides. The most prevalent (and most recognizable) form of DNA is the B-form double helix in which the sugar conformation is 2Ј-endo (11). Because it lacks the 2Ј-hydroxyls, DNA can also exist in several different forms, including A-DNA and Z-DNA (11). Conversely, RNA exists predominantly in the A-form with a 3Ј-endo ribose conformation. Given the greater flexibility of DNA, it is possible that DNA analogues can adopt a conformation that is similar to the corresponding RNA in solution.
Paquette et al. (12) compared the conformation of a tRNA with its "tDNA" and "anti-tDNA" (the complement of tDNA) analogues. By examining the mung bean nuclease cleavage patterns of tRNA Met , tDNA Met , and anti-tDNA Met these authors demonstrated that the global structures of tRNA and tDNA were quite similar, although the core regions of the tDNAs are more exposed than the tRNA. Furthermore, tDNA Met and anti-tDNA Met are both cleaved by the restriction enzymes HhaI and CfoI, which verifies the presence of basepaired stems. Based on these studies, the authors conclude that nucleic acid conformation is largely determined by the interactions of the bases and those elements common to both DNA and RNA. One role of the 2Ј-hydroxyls (in tRNA at least) is to increase the stability of the molecule (12). Studies using transition metal complexes have confirmed that the global structure of tDNA Phe resembles that of tRNA Phe but that the base paired stem conformations differ somewhat (13). Those results are most consistent with a more B-like conformation for the tDNA double helical regions. Additionally, Holmes and Hecht have shown that Fe⅐bleomycin cleaves tRNA His and tDNA His at the same major site, U 35 or T 35 respectively (14). All of the above studies were conducted using canonical base replace-ments (A, C, G, T, and deoxyribose for the tDNA analogues) and demonstrate that tDNA and tRNA can have similar conformations in solution.
Additional studies have revealed that the presence of modified bases further increases the capability of DNA to mimic RNA (15)(16)(17)(18). For example, DNA analogues of the yeast tRNA Phe anticodon stem loop containing modifications such as deoxyuracil, 5-methylcytidine (m 5 C), and 1-methylguanine (m 1 G) are able to bind Mg ϩ2 . This binding is dependent upon the presence of the modified bases (15). The Mg ϩ2 is bound in the upper part of the DNA hairpin (16) in a position that is analogous to that seen in the x-ray crystal structure of yeast tRNA Phe (19). Furthermore, the modified DNA analogues were able to bind to poly(U)-programmed 30 S ribosomal subunits and competitively inhibit the binding of native tRNA Phe (17). They were also able to inhibit protein expression in a coupled transcription-translation system (18). The solution structure of the fully modified DNA analogue demonstrates that the helix is in the B-form. However, the conformation of the anticodon loop includes the formation of the U-turn and is strikingly similar to that of yeast tRNA Phe (18). This demonstrates that the structure of tRNA is not strictly dependent on the ribose backbone and underscores the importance of modified bases in tRNA structure and function.
In light of these experiments, it is perhaps less surprising that tDNA analogues of tRNA can be aminoacylated. E. coli tDNA Phe (with a 3Ј-terminal riboadenosine) and E. coli tDNA Lys were aminoacylated by their respective aminoacyl- tRNA synthetases (20). In a similar fashion, E. coli methionyl-tRNA synthetase will aminoacylate a tDNA fMet analogue (21). Giegé et al. (22) have utilized a mutant T7 RNA polymerase to selectively incorporate deoxyribose derivatives of each base (dA, dG, dC, or dU) in order to study the effect of replacing a subset of 2Ј-hydroxyls. Their results indicate that the yeast methionyl-tRNA synthetase will tolerate dA and dU substitutions, but large decreases in charging occur for dG or dC analogues. Similarly, yeast aspartyl-tRNA synthetase will efficiently charge dC and dA analogues but not dG or dU analogues (22).
The recognition of DNA analogues of RNA substrates also extends to some RNA editing and modifying enzymes. For example, a substrate containing all DNA residues (except for a single ribonucleotide at the cleavage site) can be cleaved by the hammerhead ribozyme, albeit with reduced efficiency (23)(24)(25). From a catalytic perspective, a predominately DNA analogue of the ribozyme domain is capable of cleaving an RNA substrate (26). Other examples of RNA catalysis with DNA substrates include Group I introns (27) and RNase P (28,29). There is also some evidence that other tRNA modification enzymes are capable of recognizing DNA analogues. A tDNA fMet analogue is threonylated to a small degree by a crude yeast extract and is able to inhibit the threonylation of tRNA fMet (21); however, these results need to be verified under more stringent conditions. A dU-containing DNA minihelix analogue of the T⌿C stem and loop of yeast tRNA Phe was reported to be a substrate for E. coli m 5 U 54 -tRNA methyltransferase (RUMT) (30). This analogue, presumably acting as a weak competitive substrate, was also able to inhibit the methylation of tRNA substrates and reduced the aminoacylation of yeast tRNA Phe . Although some activity was seen, this activity was not linear with respect to time. Furthermore, these authors later reported that RNA mutants with single (dU 54 or dU 55 ) or double (dU 54 dU 55 ) mutations were not substrates for E. coli RUMT (31). No explanation was given in this later report to account for the activity previously seen with the entirely deoxyribose analogue. Therefore, there is some ambiguity as to whether or not RUMT will recognize a DNA analogue. The tRNA editing enzyme that catalyzes the precise addition of the 3Ј-terminal CCA sequence to tRNAs (the CCA-adding enzyme, ATP(CTP):tRNA nucleotidyl-transferase) will recognize tDNA analogues provided they have a 3Ј-terminal ribonucleotide (32). Both full-length and minihelix DNA analogues of the T⌿C stem and loop of tDNA Val and tDNA Ala were substrates for the E. coli CCA-adding enzyme. Interestingly, the minihelix analogues were slightly better substrates than the full-length analogues for the CCAadding enzyme (32).
The activity of the deoxyuridine-containing DNA analogues with E. coli TGT clearly demonstrates that TGT recognition is not critically dependent upon the native ribose backbone. However, the 2Ј-hydroxyl probably does influence binding, either through conformational effects or direct interactions with TGT. These experiments also demonstrate that TGT is capable of recognizing DNA provided that a UGU sequence can be found. This suggests that under certain conditions (e.g. in the E. coli mutant that lacks the enzymes deoxyribouracil-triphosphatase (dUTPase) and uracil N-glycosylase (E. coli dut Ϫ ung Ϫ strain, Ref. 33) it is possible that there may be a physiological role for queuine modification of DNA.