INTRODUCTION

A variety of modified nucleosides has been found in tRNA (1, 2), but their functions and, in particular, their biosynthetic pathways are still largely unknown (3). Many modified nucleosides are highly conserved with respect to their sequence locations in tRNA (4), and some are characteristic of the evolutionary origin (2, 5), namely, archaea, bacteria, or eukarya (6). Perhaps the most phylogenetically specific nucleoside in tRNA is archaeosine, which occurs only in archaeal tRNA at position 15, a site that is not modified in tRNAs from the other two primary domains (7). Archaeosine was first discovered by Kilpatrick and Walker (8) during sequencing of tRNA from Thermoplasma acidophilum, and it was subsequently shown to be present in many archaeal species (9); in the most extensively studied archaeal tRNA, from Haloferax volcanii, archaeosine occurs in tRNAs specifying more than 15 amino acids (10). Subsequently, the structure of archaeosine was determined to be the non-purine, non-pyrimidine nucleoside 7-formamidino-7-deazaguanosine (Fig. 1A) (11).

The only other known examples of tRNA nucleosides with 7-deazaguanosine structures are the members of the Q¹ nucleoside (12) (Fig. 1E) family (13), which includes precursors in its biosynthesis, such as 7-cyano-7-deazaguanine (preQ₀; Fig. 1D) (14), 7-aminomethyl-7-deazaguanine (preQ₁; Fig. 1C) (15), and oQ (16) from bacterial tRNAs, and mannosyl and galactosyl derivatives of Q (17, 18) from mammalian tRNAs. In contrast to archaeosine, members of the Q nucleoside family are located at the first position of the anticodon (position 34) in bacterial and eukaryotic tRNAs that are specific for only four amino acids (Tyr, His, Asp, and Asn) (19). The key enzyme in the biosynthesis of the Q nucleoside in tRNA is tRNA-guanine transglycosylase (TGT; EC 2.4.2.29), which catalyzes a base-exchange reaction by cleavage of the N-C glycosidic bond at position 34 (20). In bacteria, TGT catalyzes the exchange of guanine at position 34 in tRNA with either guanine base, preQ₁ base, or preQ₀ base (20, 21). preQ₁ base is presumed to be synthesized de novo from GTP (1) and was identified as the physiological substrate of Escherichia coli TGT (21). After incorporation of preQ₁ into tRNA, it is further modified to oQ by transfer of the ribosyl moiety from S-adenosylmethionine (22), then finally to yield Q in the polynucleotide chain (23). In contrast, in eukarya, TGT can incorporate fully modified Q base into the first position of the anticodon by a base-replacement reaction (24, 25). Animals cannot synthesize Q-related compounds de novo and must obtain Q base as a nutrient from their diet or gut flora (26, 27).

Here we report the isolation of a new type of TGT from H. volcanii; it catalyzes the incorporation of preQ₀ base into position 15 of tRNA, replacing guanine originally located at that site. Further, we have demonstrated that free preQ₀ base is present in H. volcanii cells, implying that TGT utilizes preQ₀ as a substrate leading to the biosynthesis of archaeosine in archaeal tRNAs.

EXPERIMENTAL PROCEDURES

H. volcanii (ATCC 29605) was grown aerobically at 37 °C on a 500-liter scale in Gupta's medium (10), until the absorbance at 600 nm reached 0.8-1.0. About 1.0 kg of cells was collected.

Exchange between guanine and various 7-deazaguanine analogues, catalyzed by TGT, was assayed as described previously (20) except that the final ionic condition of the reaction mixture was 1.5 M KCl and 1.5 M NaCl. The 7-deazaguanines were synthesized as described previously: preQ₀ (28), preQ₁ (29), and archaeosine base (30).

Frozen H. volcanii cells (100 g) were suspended in 200 ml of buffer A (50 mM Hepes (pH 7.5), 10% glycerol, 1.0 mM dithiothreitol, and 0.5 mM phenylmethylsulfonyl fluoride) plus DNase I (2.5 µg/ml), and were broken by sonication. The S-100 fraction was obtained by centrifugation at 105,000 × g for 1 h, dialyzed against buffer A, and then adsorbed onto a DEAE-Sepharose FF column (2.5 × 20 cm) (Pharmacia Biotech Inc.), which was eluted by a linear gradient of NaCl from 0.02 to 0.5 M in buffer A. The eluate containing the active fraction was brought to 40% ammonium sulfate and then applied to a Butyl-Sepharose FF column (2.5 × 20 cm) (Pharmacia), which was eluted with a linear gradient of ammonium sulfate from 40 to 0% in buffer A. The active fraction was next applied to a Butyl-Sepharose 4B column (1.5 × 15 cm) (Pharmacia) and eluted as described above for the Butyl-Sepharose FF column. The active fraction was then applied to a Superdex 200 column (1.6 cm × 60 cm) (Pharmacia), and then eluted with buffer A containing 300 mM NaCl. Finally, the TGT fraction was applied to a Mono Q column (0.50 × 5 cm) (Pharmacia) and eluted with a linear gradient of NaCl from 300 mM to 1 M. This TGT fraction was stable for at least 1 month when stored at 4 °C. The activity of the enzyme was monitored by incorporation of [8-¹⁴C]guanine into unfractionated E. coli tRNA (20). Amino acid sequences of peptide fragments generated by digestion with lysylpeptidase were determined as described previously (31).

Two synthetic DNA oligomers, namely Lys-FOR (5- TAATACGACTCACTATAGGGCCGGTAGCTCAGTTAGGCAGAGCGTCTGACTCTT-3) and Lys-REV (5-TGGTGGGCCGGACGCGATTTGAACACGCGACCGTCTGATTAAGAGTCAGACGCTCTGCCTA-3) were annealed via the complementary region, and both of the 3 ends were extended by Tth DNA polymerase (Toyobo). After extension, two synthetic DNA primers, namely T7 (5-TAATACGACTCACTATA-3) and Halo-Lys3 (5-CCTGGTGGGCCGGACGCGATTT-3) were added, and a polymerase chain reaction was performed to yield the gene for H. volcanii tRNA^Lys(CUU) containing the promoter sequence for T7 RNA polymerase (Takara). We cloned the product of a polymerase chain reaction in pUC19; digestion of plasmid DNA with MvaI generated a CCA end for the tRNA gene, which was transcribed in vitro using T7 RNA polymerase (32).

For preparation of the T7 transcript into which preQ₀ base was incorporated, 200 µl of a reaction mixture containing 300 pmol of T7 transcript, 20 µl of TGT (15 units) (20), and 5 nmol of preQ₀ base (under ionic conditions of the guanine exchange reaction; see above) were incubated at 37 °C for 1.5 h. The sequence of the RNA was determined as described elsewhere (33, 34).

A reaction mixture containing T7 transcript and TGT in the presence of preQ₀ base or an aliquot of acid-soluble extract of H. volcanii was incubated at 37 °C for 1.5 h. After digestion of the T7 transcript by RNase T2, the preQ₀ nucleotide was analyzed by post-labeling using T4 polynucleotide kinase and [-³²P]ATP (21, 35). The enzymes used (RNase T2, T4 polynucleotide kinase, and yeast hexokinase) were inactivated by phenol extraction instead of boiling. After incubation with nuclease P1, the digestion product was applied to a cellulose thin layer plate (20 × 20 cm) and was subjected to two-dimensional chromatography (15).

H. volcanii cells were suspended in a solution of 0.2 M formic acid and shaken for 2 h at 4 °C. After centrifugation, the supernatant was filtered through a Millipore filter. After neutralization with NaOH, soluble substances were extracted with tetrahydrofuran. The organic phase was evaporated, and the material was used for the identification of preQ₀ base.

RESULTS

E. coli TGT can be assayed by its ability to incorporate [8-¹⁴C]guanine into Q-unmodified tRNAs (typically unfractionated yeast tRNA, which constitutively lacks Q, is used) by replacing the guanine base located at the first position of the anticodon (20). By analogy with E. coli TGT, we searched for such an enzymatic activity in a crude extract of H. volcanii using E. coli tRNA as a substrate (see below). H. volcanii TGT was purified to near homogeneity following successive column chromatographies. Table I shows the recovery and the purification factor at each step, and Fig. 2 shows the pattern of SDS-polyacrylamide gel electrophoresis at each step. The molecular mass of the enzyme was deduced to be 78 kDa from a profile of the gel (Fig. 2, lane 6). Like E. coli and eukaryotic TGT, H. volcanii TGT does not require ATP for the base replacement reaction. High salt concentration (approximately 2.4 M) is necessary for maximum activity. Magnesium ion is required for activity with the T7 transcript as a substrate, but activity with unfractionated E. coli tRNA does not require magnesium ion. These results suggest that magnesium ion may be responsible for conformational rigidity of the tRNA, but not for the enzymatic activity itself. Optimum activity occurs near pH 7.5. Purified TGT was digested with lysylpeptidase, and the sequences of several resultant peptide fragments were determined (see below).

Table I. Purification of tRNA-guanine transglycosylase from H. volcanii Fraction Protein Specific activitya Purification factor Recovery mg ×102 units/mg % S-100 5500 2.4 1 100 DEAE-Sepharose FF 1100 13.8 6 118 Butyl-Sepharose FF 350 36.9 16 102 Butyl-Sepharose 4B 50 34.8 15 14 Superdex 200 7 321.3 139 18 Mono Q 0.4 975.0 422 3 a 1 unit = 20 pmol/h.

To examine the specificity for tRNA substrate, we constructed a plasmid clone containing the sequence of H. volcanii tRNA^Lys(CUU) and that of T7 promoter upstream of the gene. Its T7 transcript (Fig. 3A) was found to be a good substrate for the enzyme (Fig. 3B). The labeled T7 transcript was isolated, and the site at which [8-¹⁴C]guanine had been incorporated was determined by RNA sequencing to be position 15,² the exclusive location of archaeosine nucleotide in archaeal tRNA. This result suggested that the enzymatic activity is involved in the biosynthesis of archaeosine nucleotide in tRNA. Unfractionated tRNA from E. coli was also found to be a good TGT substrate, whereas unfractionated H. volcanii, yeast, and bovine tRNAs were not (Fig. 3B), although we did not quantitatively measure the efficiency of unfractionated E. coli tRNA and of the T7^Lys transcript as substrates. These results further suggest that position 15 of H. volcanii tRNAs is fully modified to archaeosine nucleotide.

The ability of various bases to serve as substrates for incorporation into tRNA by H. volcanii TGT was examined using the procedure of Okada et al. (21). First, the T7 transcript was labeled with [8-¹⁴C]guanine by incubation with TGT. To a reaction mixture that contained this 8-¹⁴C-labeled tRNA and the TGT enzyme, we added various 7-deazaguanine bases and monitored the decrease in acid-insoluble radioactivity of the tRNA due to release of [8-¹⁴C]guanine by replacement with the added base (Fig. 4). Unexpectedly, neither archaeosine base itself (Fig. 1B), nor preQ₁ base (Fig. 1C), which is the physiological substrate for E. coli TGT (21), were incorporated into the tRNA transcript. Among 7-deazaguanine derivatives, only preQ₀ base (Fig. 1D) was efficiently incorporated. We attribute the small amount of apparent archaeosine base incorporation into tRNA to be due to preQ₀ base, and not archaeosine base, since approximately 20% of archaeosine base is chemically converted to preQ₀ base after incubation of the reaction mixture under the conditions used. Furthermore, the nucleotide at position 15 of the tRNA product after incubation with archaeosine base was found to be preQ₀ nucleotide by RNA sequencing² (see "Discussion").

To investigate whether preQ₀ base is directly incorporated into tRNA, as well as whether incorporation occurs at position 15 in the D-loop, the sequence of the D-loop region in the T7 transcript after incubation with preQ₀ base was determined by the post-labeling method (33, 34). The RNA was subjected to partial digestion with alkali and the 5 ends of resultant RNA fragments were labeled by using polynucleotide kinase and [-³²P]ATP, followed by separation by electrophoresis in a polyacrylamide gel (Fig. 5A). RNA was extracted from each band in the gel and digested with nuclease P1. The resultant ³²P-labeled nucleotide 5-monophosphate was analyzed by thin-layer chromatography. Fig. 5B shows clearly that preQ₀ base was incorporated at position 15 of the tRNA, and also shows that more than 90% of the nucleotide at position 15 is a preQ₀ nucleotide, indicating that the base-replacement reaction by H. volcanii TGT was efficient under the present conditions.

If preQ₀ base is the physiological substrate for H. volcanii TGT, free preQ₀ base could be present in H. volcanii cells. To test this hypothesis, we prepared an acid-soluble extract of H. volcanii and incubated an aliquot of the extract with the T7 transcript of H. volcanii tRNA^Lys(CUU) and H. volcanii TGT under the same conditions described in Fig. 4. After the reaction, we analyzed modified nucleotides in the treated tRNA using the post-labeling method (21, 35). As shown in Fig. 6, preQ₀ 5-monophosphate was detected in the tRNA transcript following incubation in the presence of the acid-soluble extract (Fig. 6B), but it was not detected following incubation with the enzyme alone (Fig. 6C). Further, similar acid treatment of isolated H. volcanii tRNA did not release preQ₀ by the criterion of failure of the T7 transcript to incorporate preQ₀ when incubated with the extract and TGT.² Although archaeosine base is unstable under conditions of high temperature and high salt (see above), archaeosine appears stable when present as a nucleotide in intact tRNA (10). These results suggest that free preQ₀ base is present in H. volcanii cells and that it may serve as the physiological substrate for H. volcanii TGT (see "Discussion"; Fig. 7A).

The normal growth medium for H. volcanii (10) contains Tryptone, which, as a whole meat extract, is a source of Q nucleoside and, therefore, a potential source of preQ₀. To rule out the possibility that H. volcanii may not synthesize archaeosine de novo, tRNA was isolated from cells grown in a chemically defined (Q-free) medium (36) and analyzed for archaeosine; archaeosine content in tRNA from cells grown in the normal growth medium and in chemically defined growth medium was identical.²

Recently, the complete genome sequence of the methanogenic archaeon, Methanococcus jannaschii, has been reported (37). Among 1738 protein-coding genes predicted is a putative M. jannaschii TGT gene (MJ#0436) that exhibits 30% identity to E. coli TGT (38). We determined the amino acid sequences of three peptide fragments, generated from purified H. volcanii TGT by digestion with lysylpeptidase, and compared them with the sequence of the putative M. jannaschii TGT. As shown in Fig. 8, fragments 1 and 2 from H. volcanii TGT appear to be closely related to the M. jannaschii sequence, with identities of 53.5 and 38.5%, respectively, although the C-terminal portion of fragment 3 diverges from that in M. jannaschii. These results suggest that the H. volcanii tRNA-guanine transglycosylase characterized here is the counterpart of the putative TGT whose sequence is present in M. jannaschii (37).

DISCUSSION

It is well established that TGT is involved in biosynthesis of Q nucleotide in E. coli (Fig. 1E) by exchange of guanine at position 34 by preQ₁ base in tRNAs specific for Tyr, Asp, Asn, and His ((20, 21); see Introduction). The resultant preQ₁ nucleotide in tRNA is then modified to the epoxide oQ by the S-adenosylmethionine-requiring enzyme QueA (22), and finally, oQ is converted to Q by an unknown vitamin B₁₂-dependent enzyme (23). These processes are schematically represented in Fig. 7B. In the present study, we provide evidence that, in contrast with the primary substrate of bacterial TGT (preQ₁), preQ₀ base is the normal substrate for H. volcanii TGT. Presumably, the incorporated preQ₀ base then is further converted to archaeosine by (net) addition of ammonia, at the polynucleotide level (Fig. 7A). Therefore, both E. coli and H. volcanii TGTs catalyze a very similar reaction, namely, the exchange of guanine base in a polynucleotide chain with a free 7-deazaguanine derivative; however, their actual substrates (in terms of base, tRNAs, and the site of replacement in tRNA) are different.

Archaeosine is present at position 15 (D-loop) in most archaeal tRNAs (7), whereas Q and its derivatives are present at position 34 (first position of the anticodon) of four specific tRNAs in bacteria and eukarya (19) (see Introduction). Accordingly, these conserved differences in structure and sequence location suggest differences in function. Q has been proposed to be involved in codon recognition (39) and has been shown to prevent stop codon readthrough in tobacco mosaic virus RNA in a codon context-dependent manner (40). A correlation between the presence of Q-undermodified tRNAs and frameshifts of some retroviruses including human immunodeficiency virus was proposed (41). Other functional implications of Q, such as in virulence of Shigella (42), signal transduction (43), ubiquitin-dependent proteolytic pathway (44), and tumor differentiation (45-48), have also been suggested.

The functional role of archaeosine has not been established, but has been proposed to involve enhanced stabilization of tRNA tertiary structure as a consequence of the unique charged imidino side chain (11). Earlier work has demonstrated that hydrogen bonding interactions between G-15 in the D-loop and C-48 in the T-loop, stabilized by stacking with purine-59, constitute a generally conserved mechanism for stabilization of the universal folded L-shape of tRNA (49, 50). These structural features (G-15, C-48, purine-59) are basically met by nearly all reported archaeosine-containing tRNA sequences (7), to which would be added the strong potential for electrostatic interactions between phosphate and the "arginine fork" imidino side chain of archaeosine.

Interestingly, precursor bases used as substrates for both bacterial and archaeal TGTs participate in analogous biosynthetic pathways. Free preQ₁ base has been isolated from E. coli (21), and here we provide evidence for the presence of free preQ₀ base in H. volcanii. We believe that this free preQ₀ base is likely to be the precursor exchanged into tRNA in the normal biosynthetic pathway leading to archaeosine, although, at present, we cannot strictly exclude the possibility that free preQ₀ base detected is instead derived from archaeosine in tRNA. In E. coli, preQ₁ base is synthesized from GTP (13), possibly via preQ₀ (51), although there is presently no direct evidence for any precursor-product relationship between these two 7-deazaguanine bases. Presumably a similar pathway is present in H. volcanii for biosynthesis of preQ₀ base from GTP. The key substrates following base replacement at the tRNA level, then, are preQ₁ nucleotide (leading to queuosine in E. coli) and preQ₀ nucleotide (leading to archaeosine in H. volcanii). It is noted that preQ₀ nucleoside is present in tRNA of certain mutants of E. coli (51), the meaning of which has not yet been rationalized (14, 21). The occurrence of these 7-deazaguanine precursor bases in both primary phylogenetic domains, archaea and bacteria, prompts us to speculate a more general role for them in cellular functions. In this respect, more detailed characterization of free preQ₀ base (and possibly free preQ₁ base) in H. volcanii cells is required.

tRNA structural requirements for enzyme recognition remain to be identified. Preliminary experiments² showed that an 18 nucleotide minihelix containing the D-loop and D-stem of H. volcanii tRNA^Lys(CUU) does not serve as a substrate for H. volcanii TGT, implying the existence of higher order recognition elements for the archaeal TGT. By contrast, bacterial TGT recognizes the anticodon loop sequence U³³-G³⁴-U³⁵, which is the minimum requirement for recognition by the enzyme, and minihelices containing this triplet sequence are good substrates for the enzyme (52, 53).

By x-ray crystallography, the tRNA-guanine transglycosylase from Zymomonas mobilis has been determined to be an irregular (/)₈ barrel with a tightly attached C-terminal zinc-containing subdomain (54). Further, the structure of Z. mobilis TGT in complex with preQ₁ suggests a binding mode for tRNA where the phosphate backbone interacts with the zinc subdomain and the U³³-G³⁴-U³⁵ sequence is recognized by the barrel. The zinc binding motif (CXCX₂CX₂₅H) is highly conserved in prokaryotic TGTs known so far (52), and the homologous region in M. jannaschii is (CXCX₂CX₂₂H). These results demonstrate a structural and functional conservation of the archaeal and bacterial/eukaryotic TGT binding mode with tRNA, despite archaeal modification of the D-loop and bacterial/eukaryotic modification of the anticodon loop. The utilization of 7-deazaguanine derivatives for tRNA processing by interrelated TGT enzymes suggests an evolutionarily fundamental role for 7-deazaguanine.

In contrast to bacterial TGT (52, 55), productive recognition of tRNA by eukaryotic TGT requires not only the U³³-G³⁴-U³⁵ sequence of the anticodon loop but also a correctly folded tRNA architecture (56). In addition, eukaryotic TGT is believed to be a heterodimer, although this is not conclusive at present (44, 57). More detailed examination of the substrate recognition properties of TGTs from archaea, bacteria, and eukaryotes will elucidate the domain structures of these proteins for the tRNA binding site, as well as further define their evolutionary relationship.