Substrate Specificity for 4-Thiouridine Modification in Escherichia coli *

The biosynthesis of 4-thiouridine (s 4 U) in Escherichia coli tRNA requires the action of both the thiamin path-way enzyme ThiI and the cysteine desulfurase IscS. IscS catalyzes sulfur transfer from L -cysteine to ThiI, which utilizes Mg-ATP to activate uridine 8 in tRNA and trans-fers sulfur to give s 4 U. In this work, we show through deletion analysis of unmodified E. coli tRNA Phe that the minimum substrate for s 4 U modification is a mini-helix comprising the stacked acceptor and T stems containing an internal bulged region. The size of the bulged loop must be at least 4 nucleotides and contain the target uridine as the first nucleotide. Replacement of the T loop sequence with a tetraloop in the deletion substrate increases activity and shows that the T (cid:1) C primary sequence is not a recognition element. An unmodified tRNA Phe transcript in which the 3 (cid:1) -terminal ACCA sequence is removed to give a blunt terminus has < 0.1% activity, although the addition of a single overhanging base essentially restores activity. In addition, reducing the distance of the 3 (cid:1) terminus relative to U8 by as little as 1 bp severely impairs activity. By dissecting a minimal RNA substrate in the T loop region, a two-piece system consisting of a substrate RNA and a “guide” RNA is efficiently modified. Our results indicate that outside of the modified U8, there is no primary sequence requirement for substrate recognition. However, the secondary and tertiary structure restrictions appear sufficient to explain why s 4 tRNA structure or trap a transient partially unfolded intermediate to catalyze the chemistry at O4. To gain insight on how ThiI recognizes its tRNA substrate, we report on the substrate specificity for s 4 U modification using variants of unmodified E.

in the tRNA of bacteria and archea. Its biological role appears to be the result of its photochemical reactivity. Irradiation of bacteria can cause a photocross-link in specific tRNAs between s 4 U at position 8 and cytidine at position 13 (8). Some of the cross-linked tRNAs are poor substrates for aminoacylation (9). The physiological result of near UV exposure in Escherichia coli (10) and Salmonella enterica Serovar typhimurium (11) is an s 4 U-dependent growth delay and induction of the stringent response to amino acid starvation. Thus, s 4 U in tRNA appears to be important in vivo as a physiological sensor for near UV radiation exposure. In vitro, s 4 U has found use as an effectively zero-length photocross-linking agent (12)(13)(14) as well as a chemoselective site for the attachment of biophysical probes (15,16).
Initial work on s 4 U biosynthesis in E. coli found a requirement for two enzymes ( Fig. 1) (17). The thiI gene was shown to be required for s 4 U synthesis in E. coli by Mueller et al. (18) and for thiazole biosynthesis in Salmonella typhimurium by Webb et al. (19). We later found that the cysteine desulfurase IscS was required for s 4 U synthesis both in vitro (20) and in vivo (21) and demonstrated in vitro s 4 U synthesis with an unmodified tRNA substrate. We have shown through gel mobility shift assays that ThiI and not IscS binds to tRNA and that ThiI binds ATP, most probably for activation of the uridine via adenylation (22). The mechanism involves mobilization of sulfur from cysteine by IscS in the form of an enzyme-bound persulfide (23,24), which is then transferred to a cysteine residue on ThiI before final insertion into U8 of tRNA (25)(26)(27). The nature of the final sulfur transfer step is unclear, but it has been recently shown that ThiI is oxidized to a disulfide under single turnover conditions (28). Thus, ThiI has the multiple roles of binding tRNA, activating O4 of uridine 8 using Mg-ATP while accepting a sulfur atom from IscS and transferring it to the 4-position to give the final product, s 4 U.
The targeted uridine 8 for s 4 U biosynthesis is essentially invariant in cytosolic tRNAs. Fig. 2 shows the position of U8 in different representations of the canonical L-shaped tRNA structure. In E. coli, all of the tRNAs contain s 4 U to some extent, although the levels of s 4 U in particular tRNAs are found to vary with growth rate (29). Because of the prevalence of the modification, it is reasonable to propose that the tertiary structure and not primary sequence is the major structural determinant for substrate activity. In the structure of most cytosolic tRNAs, U8 is involved in a reverse Hoogsteen base pair with A14, whereas the 2Ј-hydroxyl of U8 is hydrogen-bonded to A21 (30). The U8-A14 interaction is particularly important for the formation of the core of the tRNA structure by bringing the D loop in closer proximity for base pairing with residues in the T loop (see Fig. 2). An examination of a space-filling model of yeast tRNA Phe suggests that although O4 of U8 is not involved in direct tertiary interactions, access to O4 is restricted in the folded structure. Thus, ThiI to some extent must invade the tRNA structure or trap a transient partially unfolded intermediate to catalyze the chemistry at O4. To gain insight on how ThiI recognizes its tRNA substrate, we report on the substrate specificity for s 4 U modification using variants of unmodified E. coli tRNAs.

MATERIALS AND METHODS
General-ThiI and IscS were overproduced from expression plasmids in E. coli BL21(DE3) and purified as described previously (20). T7 RNA polymerase was purified from strain BL21(pAR219), which was a generous gift of F. Studier. Acryloyl aminophenylmercury(II) chloride (APM) for affinity electrophoresis was prepared as described previously (31) and used as a stock solution of 1 mg/ml in formamide. L-[ 35 S]Cysteine was obtained from PerkinElmer Life Sciences. DE81 DEAE filter discs were from Whatman. Oligodeoxynucleotides were from Integrated DNA Technologies and were purified by denaturing PAGE before use. Dideoxysequencing of cloned DNA templates was performed by the University of Wisconsin Biotechnology Center.
Preparation of RNA Substrates-RNA substrates were obtained by runoff transcription using T7 RNA polymerase as described previously (32). Templates for the preparation of short RNAs (Ͻ50 nt) were singlestranded DNA oligomers comprising the antisense of the desired RNA sequence and a 5Ј-17-nt region that is the antisense of the T7 promoter. For example, the template for TPHE39A is 5Ј-TGG TGC CCG GAC TTG  GTT TCC CAA GGA CAC TAT CCG GGC TAT AGT GAG TCG TAT  TA-3Ј where the underlined region is complementary to the T7 promoter. This single-stranded oligonucleotide is annealed to an 18-bp DNA oligonucleotide containing the sense strand of the T7 promoter and initial G, having the sequence 5Ј-TAA TAC GAC TCA CTA TAG-3Ј. For longer RNA substrates, double-stranded templates containing the RNA sequence preceded by the T7 promoter sequence were prepared by the method of Peterson et al. (33) using two overlapping oligodeoxynucleotides with a 12-bp region of complementarity. For tRNAs, the sense strand was 57 nt and contained the T7 promoter followed by the first 40 nt of the RNA, whereas the antisense strand was 48 nt and contained the last 48 nt of the tRNA, including the 12-nt overlap. The annealed oligonucleotides were extended with reverse transcriptase as described previously (33) and used as templates for the in vitro transcription. Double-stranded DNA templates were checked for accuracy by incubation with Taq polymerase for the addition of terminal nucleotide followed by cloning into vector pCR2.1 (Invitrogen) for sequencing. Transcription reactions contained 1 M template DNA in a mixture of 50 mM Tris, pH 7.5, 35 mM MgCl 2 , 5 mM each of NTPs, 1 mM spermidine, 0.01% (v/v) Triton X-100, 8 units of RNasin, and 4 g of T7 RNA polymerase in a total volume of 1 ml. Reactions were typically run overnight at 37°C. RNA was purified by denaturing PAGE (8 M urea, 1ϫ Tris borate EDTA), visualized by UV shadowing, and then excised and eluted in 0.5 M NaCl using the crush and soak method. RNAs prepared for 3Ј-terminal analysis were further purified on 1.5-mm-thick sequencing gels to achieve single nucleotide resolution. Transcription reactions typically resulted in 50 -70% RNA of the desired length (N), 20 -30% N ϩ 1, and lesser amounts of longer products. Gel-purified RNA was passed through a 0.2 M syringe filter, ethanol-precipitated, dissolved in Milli Q-purified water, and stored at 20°C. RNA concentrations were measured by UV absorbance at 260 nm. Yields for RNAs Ͼ50 nt in length were consistently Ն0.5 mg/ml transcription reaction. Yields for smaller RNAs varied from 0.2 to 0.4 mg/ml depending on sequence and length. For activity and binding studies, RNAs were folded by heating at 65°C in 10 mM Tris, pH 7.5, 0.5 mM Na 2 EDTA, and 5.5 mM MgCl 2 or 5.5 mM Mg(OAc) 2 for 5 min and slow cooling to room temperature over 30 min. s 4 U Assays-Assays for s 4 U synthesis were conducted using DEAE filter discs using a modification of a method described previously (20,34). This assay measures the incorporation of 35  For all of the substrates tested, rates of s 4 U formation at 10 and 20 M were the same within error. Reactions were allowed to proceed for 6 min at 37°C. Under these assay conditions, ThiI was rate-limiting and the reaction rate was linear with Ͻ10% formation of s 4 U. The reactions were initiated by adding ThiI immediately followed by IscS. The activity for each RNA substrate (20 M) was reported as the observed initial rate constant (k obs ) calculated as the rate of s 4 U formation divided by the ThiI concentration.
APM Affinity Gel Assays with Sulfide as Substrate-For reactions with sulfide as sulfur source, a gel shift assay using APM polyacrylamide gels was used based on the work of Igloi (31). This technique incorporates APM into the polyacrylamide gel. The mercury binds to thiouridine groups on the RNA with a K d ϭ 0.2 M (35) and acts to retard selectively the mobility of RNAs modified with s 4 U (31). Stock solutions of sodium sulfide were prepared shortly before performing the assays by dissolving solid Na 2 S nonahydrate in water and adjusting the pH to 8 with aqueous 6 N HCl. Reaction mixtures (50 l) were prepared as described above with the exception of sulfide substituted for IscS and [ 35 S]cysteine. Timed aliquots (25 l) of each reaction mixture were removed and quenched by the addition of 25 l of phenol/chloroform/ isoamyl alcohol (25:24:1). After vortexing and centrifugation for 3 min in a microcentrifuge, the top aqueous layer was passed through a 0.5-ml G-50 spin column. To the eluate was added 25 l of 8 M urea with 0.05% bromphenol blue and 0.05% xylene cyanol. An aliquot (25 l) of this mixture was loaded per gel lane. For tRNA Phe and minor variants, 8% polyacrylamide gels (0.75 mm) containing 100 M APM in Tris borate EDTA buffer were run on a MiniProtean II (Bio-Rad) apparatus at 170 V for 1 h at room temperature. The bands were visualized by ethidium bromide staining in Tris acetate EDTA buffer (10 min) and quantitated using a densitometer. Alternatively, 32 P-labeled RNA was utilized and quantitation achieved by PhosphorImager analysis of the dried gel using ImageQuant software from Amersham Biosciences.
Measurements of ThiI Binding to RNA Substrates-For binding studies, mixtures of 50 mM Tris, pH 7.5, 5 mM Mg(OAc) 2 , 50 mM KCl, 1 M RNA, and variable concentrations of ThiI were incubated for 10 min at room temperature. Aliquots of 20 l were then applied via an 8-channel pipette to a modified 96-well dot-blot filtration apparatus (Schleicher & Schuell). The apparatus contained both nitrocellulose (Protran BA85, Schleicher & Schuell) and DEAE filters (DE81, Whatman) essentially as described by Wong and Lohman (36). Wells were washed with 100 l of assay buffer both before and after application of binding mixtures. Filters were then blotted to remove excess liquid and wrapped in plastic wrap prior to phosphorimaging analysis. The fraction bound to protein was measured directly by comparing the amount of 32 P-labeled RNA bound to nitrocellulose relative to the amount bound to DE81 anionexchange filter paper. Measurements were performed in triplicate, and data were corrected for background-nonspecific RNA binding to the nitrocellulose filter. The corrected data were plotted as the fraction of total RNA bound to nitrocellulose versus ThiI concentration, and the resulting plot was fitted to a sigmoidal binding curve using Kaleidograph (Synergy).
Assay for Two-piece Trans-s 4  RNAs were first heated to 85°C for 2 min and slow-cooled for 1 h in the presence of Tris, MgCl 2 , and KCl before the remaining ingredients were added to initiate reaction. After 30 min at 37°C, an equal volume of phenol:chloroform:isoamyl alcohol (50:48:2) was added and the mixture was vortexed for 20 s. The layers were separated by centrifugation in a microcentrifuge for 4 min, and 25 l of the top aqueous layer was passed through a G-50 spin column and added to 25 ml of 8 M urea containing 0.1% bromphenol blue and 0.1% xylene cyanol FF. This mixture was mixed by vortexing and heated to 90°C for 3 min and loaded onto two duplicate 8% denaturing polyacrylamide gels. One of the gels contained 100 M APM, whereas the other had APM omitted. The gels were run for 45 min at 170 watts in 1ϫ Tris borate EDTA buffer. The gels were then soaked in a solution of ethidium bromide in 1ϫ TAE buffer for 10 min, and RNA was visualized on a UV transilluminator (Fotodyne).

RESULTS
Deletion Analysis of tRNA Phe Transcripts-We used a DEAE filter disc assay described previously (34) to obtain the specific activity for s 4 U modification of RNA variants. The assay was measured for the incorporation of 35 S from L-[ 35 S]cysteine into tRNA that was specifically bound on the discs, and the rates were assessed in the linear range of the progress curve at Ͻ10% of the product formation. Under these conditions, unmodified tRNA Phe gave an initial rate constant (k obs ) of 1.0 min Ϫ1 at 20 M (Table I). We found that 20 M RNA concentration is saturating for all of the RNA variants tested thus far that are substrates. We were not able to obtain accurate K m values because of the limitations of the assay and the apparent low values for K m (Ͻ0.1 M). Using a filter binding assay (36), we were able to measure a K d value for tRNA Phe of 1.9 M (Fig.  3), which is considerably higher than the K m estimate. Values for K d were measured in the absence of other substrates, and this may explain the discrepancy with the K m . In addition, K m may reflect rate constants that indicate a large commitment for catalysis.
As shown in Table I, all of the full-length E. coli tRNA transcripts we have studied are good substrates at 20 M. This is consistent with the study of Emilsson et al. (29) who found that all tRNAs contain s 4 U in vivo, although some are modified more efficiently than others at high growth rates. They re-ported that tRNA 2 Glu , now known as the active form of the single tRNA Glu in E. coli (37), was poorly modified at all of the growth rates studied. We measured the in vitro activity of  unmodified tRNA Glu as a substrate and found that it has a lower k obs by over 2-fold. Thus, although the magnitude of the reduced activity is small, it is consistent with their results. Modifications normally present in this tRNA in vivo may modify the activity further. The unmodified transcripts of tRNA Lys , tRNA Ser-2 , and tRNA Arg were also found to be good substrates, consistent with in vivo results (26).
Initial efforts to study the structure activity relationship of the tRNA were to introduce subtle disruptions of the tertiary fold in the vicinity of U8. We first mutated the target uridine 8 to make sure our in vitro system was specific for this base alone. An unmodified tRNA Phe transcript containing either a U8C or U8A mutation was inactive as a substrate. Binding studies of WT tRNA Phe and the U8C mutant surprisingly showed no difference in the value of K d within experimental error (Table I). This finding suggests that the U8C mutant can adopt a structure that is bound by ThiI, but because this structure lacks the target, uridine it is not modified.
Because U8 is involved in tertiary interactions with A14, we proceeded to make an analogous change in A14. An A14U mutation would be expected to disrupt the interaction and loosen the tertiary fold and reduce the effective concentration of properly folded tRNA. We found that this mutant had the same activity as the WT tRNA. Thus, relaxation of the tertiary fold has no effect on s 4 U modification, which may be expected if a nonnative structure is bound by the enzyme.
Defining the Minimal RNA Substrate-We subsequently introduced more drastic alterations of the tRNA structure. We sequentially replaced the D, anticodon (AC), and T stem loops with a two-nucleotide spacer and found that these RNAs were good substrates (data not shown). A tRNA mutant lacking both the D and AC stem loops was still processed at 50% of the rate of the full-length tRNA under saturating conditions. Replacement of the T⌿C loop of this RNA with a GAAA tetraloop resulted in a 39-nt RNA (Fig. 4, TPHE39A) that was an efficient substrate at saturation with activity nearly two-thirds that of the full-length tRNA Phe transcript (Table II). Binding measurements give a K d of 6 M, indicating that some binding interactions are lost in the ground state but have little effect during catalysis. This was confirmed by our estimate of the K m for TPHE39A to also be Ͻ0.1 M. A UUCG tetraloop gave similar results. Thus, the conserved T[GRAPHIC]⌿C sequence of the T loop is not a recognition element for s 4

U synthesis in vitro.
Other deletion constructs that retained the larger T loop of tRNA Phe had lower specific activity as substrates (data not shown). The tetraloop sequence may act to stabilize TPHE39A and prevent alternative folded structures. Mutation of the target uridine 8 in the minimal substrates abolished s 4 U activity (for example, see TPHE39C in Table II). This shows that the small substrates are still binding in broadly the same orientation in the active site as the full-length tRNA. Reduction in the length of either stem in TPHE39A led to a significant decrease in activity. Fig. 5 shows the relationship between internal loop size and activity with TPHE39A (7-nt loop) as reference substrate. The data show that a 6-nt loop with a base deleted from two different points in the loop (TPHE38B and TPHE38C) gives a slightly better substrate. The specific activity drops significantly with a 5-nt loop and diminishes rapidly as the loop is further contracted. The extreme case where a single uridine is bulged out of a long stem loop is inactive (not shown). Alternative structures predicted by MFOLD (38) were probed by sub-  stituting different bases in the 6-and 7-nt loops. Substitutions that decreased the possibility of base pairing within the internal loop gave modest increases in activity (data not shown).

Analysis of Internal Loop Containing the Target Uridine-
Because the structure of tRNA is complex in this region, these small substrates do not provide direct insight to the contouring of the active site but do show that multiple bases are required and that primary sequence is probably of little importance.
Effects of Changes at the 5Ј and 3Ј Termini-Because the modified U8 is relatively close to the acceptor terminus of the tRNA, this may be a site for recognition by ThiI. Extension of the small substrates in the 5Ј direction as shown in substrate TPHE45A (Fig. 4) leads to little reduction in activity (Table II). Extension of the full-length tRNA in the 3Ј direction is also well tolerated. 2 However, when the acceptor stem in either fulllength tRNA Phe or the small substrates are extended significantly as a duplex (TPHE-ZIP or TPHE47A), activity is drastically reduced.
The 3Ј Extension Is a Positive Determinant for s 4 U Activity-The only invariant primary structural element that remains in common between TPHE39A and natural tRNA Phe is the 3Јterminal NCCA. We first prepared a blunt-ended derivative of TPHE39A. This RNA was not a substrate for s 4 U synthesis at concentrations as high as 40 M. This indicated that in the context of a minimal substrate, the 3Ј overhang was absolutely required for activity. We then deleted the 3Ј-ACCA from the full-length tRNA Phe transcript (Fig. 6, TPHEBLUNT). The initially purified transcript contained mostly Nϩ1-length RNA and was a good substrate for the enzyme. However, when we purify the T7 transcript to single nucleotide homogeneity, we find that the blunt-ended tRNA is also no longer an observable substrate (Ͻ0.1%). Fig. 7 shows APM affinity gel analysis of s 4 U reactions with full-length tRNA Phe compared with the truncated substrate TPHEBLUNT. In both initial rate studies (Fig. 7, panel A) and during extended incubations (panel B), we see no s 4 U formation in the purified blunt tRNA.
We also prepared a series of 3Ј-truncated substrates lacking A, CA, and CCA from the 3Ј end of the tRNA and carefully purified the transcripts. As shown in Table III, Table III. For each structure, the remainder of the sequence is identical to tRNA Phe . concentration was 0.36 -0.5 min Ϫ1 . The primary sequence recognition of this single-stranded region is not stringent. A variant of tRNA Phe with a 3Ј-GUUG (Fig. 6, TPHE-GUUG) that replaces the ACCA is only reduced in activity by ϳ2-fold. Thus, ThiI appears to have a binding site for the 3Ј-overhanging discriminator base N73 but does not appear to interact with the conserved 3Ј-terminal CCA in a sequence-specific manner. All of the effects at the terminus are more pronounced in the small substrates, which may amplify defects in binding interactions with ThiI during catalysis.
Relative Orientation of 3Ј End and Modified U8 -Because the acceptor stem is helical and there is evidence of an interaction at both of its ends, the relative orientation between U8 and the 3Ј end of the tRNA may be an important recognition element for ThiI. Thus, we varied the distance between the two by adding or subtracting a single base pair in the acceptor stem. As shown in Table III, the addition of one base pair in the substrate TPHE ϩ1BP causes a relatively modest effect. This is not surprising because E. coli tRNA His , which contains an extra base pair in the acceptor stem (39) between C73 and G(Ϫ1) is reported to be modified with s 4 U in vivo (29). As mentioned above, extension of the acceptor stem by base pairing with the 3Ј-CCA (TPHE-ZIP) significantly reduces activity. Deletion of a single base pair in the acceptor stem causes a decrease in rate of nearly an order of magnitude. As in the case of the U8C mutant, K d is only slightly affected (not shown). In the context of the minimal substrate, deletion of a single base pair in the stem (TPHE37A in Fig. 4 and Table II) abolishes activity. Because it is unlikely that decreasing the acceptor stem by one bp affects the stability of the helix, this is evidence that orientation between U8 and the terminus is important for activity and may involve direct interaction between ThiI and position 73.
Sulfide Is Able to Replace IscS/Cysteine as a Substrate for s 4 U Synthesis-We have reported previously that sulfide is able to act as an alternative source of sulfur for s 4 U synthesis in vivo, but the level of s 4 U is only 1-2% of that typically found in E. coli when an iscS mutant is grown in the presence of up to 10 mM sulfide (40). In vitro, we find that sulfide is an excellent source of sulfur for s 4 U synthesis but only at millimolar concentrations. We used 32 P-labeled tRNA Phe as substrate and measured the fraction of modified tRNA using affinity electrophoresis (31) and PhosphorImager analysis. Reaction times were 2 min in the linear range of the progress curve. As shown in Fig. 8, the optimal sulfide concentration for s 4 U modification of tRNA Phe is ϳ50 mM. Higher sulfide concentrations are inhibitory. At 50 mM sulfide and 20 M tRNA, the initial rate constant for ThiI is ϳ1.4 min Ϫ1 , which is similar to the k obs of 1.0 for the reaction with IscS/cysteine. This result is in agreement with the lower levels of s 4 U produced in vivo at 10 mM concentration, although the free sulfide concentration inside the cells was not known.
A Two-piece Trans-system Is an Efficient Substrate-The simple structural requirements for s 4 U activity as demonstrated in the minimal substrate TPHE39A suggested that a two-piece RNA system may combine to act as a substrate. For practical reasons related to transcription yields, we used the extended substrate TPHE45A as a model substrate (Fig. 9, top) for the trans-system. Dissection of the phosphodiester backbone in the tetraloop of TPHE45A gives two RNAs that are expected to anneal to form the proper structure for recognition by ThiI. Indeed, when we prepared the "substrate" RNA 45cutA and the "guide" RNA 45cutB and tested them for s 4 U activity, we found that 45cutA was modified at nearly the same efficiency as the one-piece TPHE45A substrate. Fig. 9 (41) have recently separated these enzymes into two major groups. Group I enzymes do not require an intact tRNA for substrate recognition and instead recognize structural domains. This group includes the tRNA A37 N 6 -dimethylallyltransferase (MiaA) (42), G37 guanine trans-glycosylase in E. coli tRNA Tyr (43), the U54 methyltransferase RUMT (44), and 2-thiouridine modification of 5-methyl-U54 to give s 2 T54 (m 5 s 2 U54) in the archaeon Thermus Thermophilus (45). Each of these enzymes recognizes a small stem loop structure with a varying degree of primary sequence recognition. Group II modification enzymes require a properly folded tRNA for substrate recognition. These include the m 1 G37 methyltransferase, which modifies tRNA Leu in E. coli (46,47) and Salmonella (48), the m 2 G26 dimethyltransferase (49), and the ⌿35 synthase in Arabidopsis (50). Using tRNA Asp variants injected into Xenopus oocytes, Grosjean et al. (41) have found that ⌿13, ⌿40, and m 1 G37 are extremely sensitive to changes in tRNA tertiary structure, whereas m 2 G26 was less so and T54, ⌿55, m 1 A58, m 5 C49, and m 2 G6 were largely insensitive (40).
In this study, we show the specificity determinants for substrate activity in s 4 U modification. The minimal substrate requirements are a mini-helix containing an internal bulge of at least 5 nt with a uridine at the 5Ј end of the bulge. This mini-helix is similar to the mini-helices shown to be substrates for aminoacyl-tRNA synthetases (51), but in that system, there is no requirement for an unpaired U8 (52). ThiI could be characterized as a Group I modification enzyme, although the minimal substrate is more complex than that of stem loop-recognizing enzymes. Other tRNA-modifying enzymes have minimal recognition elements similar to that of ThiI. These include RNase P (53) and the archael G15 guanine trans-glycosylase that initiates archaeosine biosynthesis (54). Both recognize bulged mini-helix portions of the tRNA. For human RNase P, a mini-helix comprising the acceptor and T stems is inactive but becomes an efficient substrate with the insertion of a single nucleotide at position 8 (53). Recently, the crystal structure of the G15 guanine trans-  (55). This intriguing structure involves a rearrangement of the tRNA core to expose nucleotides U8-G22 while preserving the coaxially stacked minihelix comprised of the acceptor and T-stems. It is conceivable as the authors suggest that this alternate form of tRNA is also a preferred substrate for s 4 U modification (55). Binding an alter-nate tRNA conformation would solve the problem of inaccessibility to U8 in the L-shape tRNA. However, for s 4 U modification, the acceptor and T stem would be recognized rather than the acceptor-D stem loop region. Sequence homology searches have failed to show similarity between ThiI and the archael G37 guanine trans-glycosylase, although this does not rule out a similar three-dimensional fold. The 3Ј-NCCA overhang is also an important recognition element for s 4 U modification, although the sequence specificity is not stringent. The lack of activity of the blunt tRNA mutant is unlikely to be physiologically relevant because tRNA processing does not involve such a structure. However, the variant TPHE(-CCA) with a single base overhang is an intermediate in tRNA maturation and we show that it is nearly as active as the mature tRNA. Thus, all of the tRNA processing intermediates appear to be good substrates for s 4 U modification at a high concentration. Our results also suggest that the relative orientation of U8 with the tRNA terminus is important for substrate recognition. This discrimination may be a key component that ensures that only tRNA is modified to s 4 U in the cell.
In addition to RNA substrate specificity, we also investigated the specificity of the sulfur donor. We find that millimolar concentrations of inorganic sulfide can replace IscS/L-cysteine in the reaction. The optimal sulfide concentration was 50 mM, and higher concentrations were inhibitory under our assay conditions. This contrasts with our measurement of an apparent K m for cysteine in the coupled assay of 2.5 Ϯ 1.2 M. 2 Thus, sulfide binding by ThiI is weak (K m Ͼ20 mM) and physiologically irrelevant. However, the fact that sulfide is an efficient substrate at its optimal concentration provides some support for a mechanism of s 4 U synthesis in which nascent sulfide is produced by internal reduction of a ThiI persulfide (27). The alternative mechanism involves direct attack of a ThiI persulfide on the activated uridine before internal reduction. It is also possible that both mechanisms may be operative depending on the source of sulfur.
Finally, we have shown that by dissecting the substrate TPHE45A into two pieces, two RNAs can be annealed to form the correct structure for substrate recognition by ThiI. In this system, the modified RNA strand acts as the substrate, whereas the unmodified strand acts as a guide RNA to direct ThiI binding and catalysis via complementary regions. The efficient modification of the two-piece substrate suggests that the reaction may be engineered to direct modification at any uridine in a separate RNA molecule. We have recently found that the two-piece system can be extended in either direction with little loss in activity and have successfully introduced s 4 U at other specific sites in tRNA. 2 This finding suggests that ThiI may bind the tRNA in a clamp-like manner with both ends of the RNA exposed. Future work will concentrate on mapping the binding of tRNA onto ThiI and optimization of the transmodification system.