Structural requirements for the binding of tRNA Lys3 to reverse transcriptase of the human immunodeficiency virus type 1.

Reverse transcription of the human immunodeficiency virus type 1 (HIV-1) RNA genome is primed by the cellular tRNA3Lys molecule. Packaging of this tRNA primer during virion assembly is thought to be mediated by specific interactions with the reverse transcriptase (RT) protein. Portions of the tRNA molecule that are required for interaction with the RT protein remain poorly defined. We have used an RNA gel mobility shift assay to measure the in vitro binding of purified RT to mutant forms of tRNA3Lys. The anticodon loop could be mutated without eliminating RT recognition. However, mutations in the TΨC stem were found to partially interfere with RT binding, and D arm mutants were completely inactive in RT binding. Interestingly, binding of the RT protein to tRNA3Lys facilitates the subsequent annealing of template strand to the 3′-terminus of the tRNA molecule. Consistent with this finding, we demonstrate that mutant HIV-1 virions lacking the RT protein do contain a viral RNA genome without an associated tRNA3Lys primer. We also found that a preformed primer tRNA-template complex is efficiently recognized by RT protein in vitro. Extension of the template molecule over the TΨC loop did result in complete inhibition of RT binding, suggesting the presence of additional recognition elements in the TΨC loop. These results, combined with a comparative sequence analysis of tRNA species present in HIV-1 virions and RNA motifs selected in vitro for high affinity RT binding, suggest that RT recognizes the central domain of the tRNA tertiary structure, which is formed by interaction of the D and TΨC loops.

Retroviruses contain large amounts of tRNA, which is a non-random subset of the cellular tRNA pool (reviewed in Ref. 1). One tRNA species can anneal to the viral RNA genome and acts as a primer for cDNA synthesis by the viral reverse transcriptase enzyme (RT), 1 and this priming species is generally dominant among the tRNAs included in the particles. Different viruses use a different tRNA primer; avian retroviruses (e.g. avian myeloblastosis virus) use tRNA Trp , most murine retroviruses and the human T-cell leukemia viruses (HTLV-I and HTLV-II) use tRNA Pro , and the human (HIV) and simian immunodeficiency viruses use tRNA 3 Lys (1)(2)(3)(4)(5). There is accumu-lating evidence that packaging of the correct tRNA primer is determined by the RT protein. First, the RT proteins of the avian myeloblastosis virus and HIV-1 retroviruses have been shown to bind to their respective tRNA primers in vitro (6,7). Second, the primer tRNA is apparently absent from virus particles that lack the RT protein (8 -10). Alternatively, it is conceivable that the primer is specifically co-packaged with the RNA genome through annealing of the 3Ј-terminal 18 nucleotides of the tRNA primer to a perfect complementary sequence on the viral genome, the so-called primer binding site (PBS). This complex may be further stabilized through additional base-pairing interactions between the two RNA molecules (11)(12)(13)(14)(15)53). For the Rous sarcoma virus and HIV-1, however, it was reported that the tRNA primer is included in particles that lack viral RNA sequences encoding the PBS (3,4,10,16,17). These combined results suggest that the affinity for RT determines, at least in part, which tRNA species will be packaged. A complex between the HIV-1 RT protein and the tRNA 3 Lys primer has been identified in vitro using a variety of experimental approaches (7, 18 -25). However, the question of binding specificity of HIV-1 RT toward its cognate primer remains unresolved. For instance, several studies reported binding of other tRNA species with equal affinity (13,20,23). Based on UV cross-linking experiments, Barat et al. (7) have reported that HIV-1 RT interacts with its cognate primer tRNA 3 Lys by virtue of specific contacts with the anticodon stem-loop (7). It remains to be established whether solely the anticodon domain of tRNA 3 Lys is in contact with the enzyme. For instance, nuclease footprinting analysis suggested mild protection of all three loops by RT (25).
It was demonstrated that an in vitro synthesized tRNA 3 Lys transcript can functionally substitute for its natural counterpart in RT binding studies and reverse transcription assays (13,18,24,25). It was also shown that synthetic tRNA 3 Lys adopts the correct L-shaped structure, suggesting that all base pairs and tertiary interactions (e.g. between D and T⌿C loops) are formed in the absence of base modifications (25). Apparently, synthetic tRNA transcripts contain all sequence and structure requirements for recognition by the RT enzyme, although modified nucleotides may be important for fine tuning tRNA identity (13,18,26). Therefore, to a first approximation, rules that apply to the selective recognition of tRNA 3 Lys by the HIV-1 RT protein may be obtained in in vitro experiments with synthetic tRNA species. This allows a mutational analysis of the sequence and structure requirements in tRNA 3 Lys for RT binding.
Here, we probed the binding site for RT on synthetic tRNA 3 Lys in bandshift binding experiments with mutated tRNA molecules and demonstrate that the anticodon loop is not important for RT binding. In contrast, we found that mutations in the D-stem loop abolished RT binding. Furthermore, we demonstrate that annealing of an antisense oligonucleotide mimick-ing the PBS sequence to tRNA 3 Lys was possible at 37°C with the RT-tRNA complex but not with free tRNA, suggesting that RT opens the acceptor stem to allow intermolecular base pairing. A preformed tRNA 3 Lys -PBS complex, in which both acceptor and T⌿C stems will be disrupted, was efficiently recognized by RT protein. Extension of this oligonucleotide by five nucleotides, thereby forming a duplex with the T⌿C loop nucleotides, was found to completely block RT binding. These data, combined with results of a recent SELEX experiment (27) and a comparative sequence analysis of the tRNA species present in HIV-1 virions (4), suggest that the tertiary tRNA structure and the sequence of the D arm of tRNA 3 Lys is critically important for recognition by the RT protein.

MATERIALS AND METHODS
Plasmids were constructed to facilitate transcription of the tRNA 3 Lys gene with T7 RNA polymerase. Clones for wild-type (wt) tRNA 3 Lys and several mutants were constructed from a series of three overlapping DNA oligonucleotides that contained the tRNA sequence flanked by an upstream T7 RNA polymerase promoter and a downstream BanI restriction site to allow run-off transcription. Oligonucleotide 1-wt contained the wild-type tRNA 3 Lys sequence 5Ј-CTCACTATAGGCCCGGAT AGCTCAGTCGGTAGAGCATCAGACTTTTAATCTGAGGGTCCAGGG TTCAAGTCCCTG-3Ј (overlap regions underlined). Similar oligonucleotides with specific mutations in different tRNA domains were synthesized (see Fig. 1). The central, sense oligomers 1 were individually annealed to 3Ј-antisense oligonucleotide 2, which encoded BanI and EcoRI restriction sites for transcription and cloning purposes, respectively: 5ЈATGGAATTCCCTGGCGCCCGAACAGGGACTTGAA-3Ј (sites in bold, overlap underlined). The DNA duplex synthesized in a PCR reaction with oligomers 1 and 2 was extended by the 5Ј-sense oligonucleotide 3, containing a T7 promoter and a BamHI site: 5ЈCATGGATCCTA-ATACGACTCACTATAGGC-3Ј (site in bold, overlap underlined). An initial PCR reaction was performed with 5 ng of central oligonucleotide 1 and a molar excess of 5Ј and 3Ј oligonucelotides 2 and 3 (100 ng each, 35 cycles of 1 min, 95°C; 1 min, 55°C; and 1 min, 72°C). A sample was subsequently used in a second PCR reaction with a 100 ng of primers 2 and 3 (100 ng each, PCR protocol as indicated above). The final PCR product was digested with BamHI and EcoRI and inserted into plasmid pUC9. All plasmids were checked directly by sequencing.
The BanI restriction site was used to allow run-off transcription of a 74-nucleotide-long tRNA 3 Lys transcript. We initially failed to obtain BanI cleavage, which was shown to result from methylation of an overlapping dcm recognition sequence (28). To circumvent this problem, we transformed all pUC-tRNA 3 Lys plasmids into the dcm Ϫ host GM48. Unlabeled T7 transcripts were synthesized according to standard methods (29,30). tRNA 3 Lys was internally labeled with [␣-32 P]UTP during in vitro synthesis in a 2-h reaction with 1 g of linearized DNA in 12 l of T7 buffer (20 mM Tris-HCl, pH 7.5, 2 mM spermidine, 10 mM dithiothreitol, 12 mM MgCl 2 ) containing 0.5 mM G/A/CTP, 0.16 mM UTP and 2 l of [␣-32 P]UTP (800 Ci/mmol), 50 units of T7 RNA polymerase and 10 units of RNase inhibitor. Upon DNase treatment and phenol extraction, the RNA was ethanol precipitated, dissolved in renaturation buffer (10 mM Tris-HCl, pH 7.5, 100 mM NaCl), and refolded by incubation at 85°C for 2 min, followed by slow cooling to room temperature. The recombinant HIV-1 RT enzyme was obtained from Dr. D. Stammers (Wellcome Research Labs, Beckenham, Kent). This purified protein is in the 66,000 homodimer form and supplied at a 0.13 g/l concentration (38,000 enzyme units/mg protein) in 0.8 M ammonium sulfate, 20 mM Tris, 100 mM NaCl, 0.1 mM EDTA, 0.5 mM dithiothreitol. Moloney murine leukemia virus RT was obtained from Life Technologies, Inc., and avian myeloblastosis virus RT was from Boehringer Mannheim.
The affinity of RT for tRNA was measured by gel-bandshift assay (31). In some experiments, tRNA was pre-incubated with oligonucleotides as indicated in the figure legends. Oligonucleotide PBS is 5Ј-TGGCGCCCGAACAGGGAC-3Ј, oligonucleotide 3 and PBS ext , which is identical to oligonucleotide 2, were described in the PCR protocol (see above). A standard RT binding reaction mixture (20 l) contained ϳ10 ng of uniformly labeled tRNA probe in buffer (20 mM Tris-HCl, pH 7.5, 50 mM NaCl, 0.1 mM EDTA, and 5% glycerol) with 0.26 g of RT protein.
The final concentration of HIV-1 RT was ϳ100 nM, that of the tRNA molecule was ϳ20 nM, and that of oligonucleotide was Ն500 nM. After incubation for 15 min at 20°C, the samples were separated on a 5% non-denaturing polyacrylamide gel in 0.25 ϫ TBE containing 5% glycerol. Electrophoresis at 30 V was for approximately 18 h at room temperature. Gels were dried and exposed to x-ray film at Ϫ80°C using intensifying screens. For quantitation, gels were exposed overnight in a Molecular Dynamics phosphorimager.
The full-length molecular HIV-1 clone pLAI was used to construct an RT-deficient viral genome (RT Ϫ ). Details of the DNA construction are presented elsewhere (32). All techniques (cell culture, DNA transfection, virus purification, isolation of HIV-1 genomic RNA, and subsequent tRNA or oligonucleotide primer extension assays) were previously described (33).

Binding of HIV-1 RT to tRNA 3
Lys Requires an Intact D Arm but No Specific Anticodon Loop Sequences-We have used gelbandshift assays to establish that in vitro made human tRNA 3 Lys (Fig. 1) can bind to the HIV-1 RT protein in the absence of either the primer binding site on viral RNA or additional protein factors. Complex formation was readily detected as a shift in the mobility of the riboprobe in the gel ( Fig.  2A, compare lanes 1 and 2). A variety of parameters affecting the binding of the RT protein to tRNA 3 Lys was studied, and binding conditions were optimized (results not shown). For instance, binding was found to be temperature independent (0 -37°C) and completed in a 10-min incubation. Complex formation was strongly inhibited by greater than 100 mM NaCl or 100 mM MgCl 2 . Binding was observed in the presence of vast excess of 5 S rRNA added as a nonspecific competitor (1 g or approximately 25 pmol). Furthermore, no RT-tRNA 3 Lys complexes were obtained with RT enzymes of the Moloney murine FIG. 1. Secondary structure models of natural and synthetic tRNA 3 Lys . The shaded regions were mutated in synthetic tRNA 3 Lys as indicated (å, deletion) and studied in this work. Base modifications within the D, anticodon, and T⌿C arms in natural tRNA 3 Lys are indicated according to standard nomenclature (36); A 1 , 1-methyladenosine; A9, N-((9-␤-D-ribofuranosyl-2-methylthiopurine-2-yl)-carbamoyl)threonine; C 5 , 5-methylcytidine; D, dihydrouridine; G 2 , N 2 -methylguanosine; G 7 , 7-methylguanosine; T 3 , 2Ј-O-methyl-5-methyluridine; U 9 , 5-methoxycarbonylmethyl-2-thiouridine; ⌿, pseudouridine. The 5Ј-and 3Ј-extended tRNA forms contained 24 and 5 additional nucleotides, respectively (5Ј, ggauccuaauacgacucacuauag; 3Ј, caggg). leukemia and avian myeloblastosis viruses (data not shown).
The specificity of the interaction between the HIV-1 RT protein and tRNA 3 Lys implies that the tRNA molecule contains features that distinguish it from other transcripts. In differentiating among cellular tRNAs, the unique nucleotide sequence of the anticodon loop may form such an identity element. Consistent with this idea, cross-linking experiments revealed contacts between RT and the anticodon loop of tRNA 3 Lys (7). To assess the sequence-specific contribution of this tRNA domain to RT recognition, we constructed a 6-base substitution mutant ( Fig. 1, AC mutant). Gel shift assays with wt and mutant tRNA 3 Lys are shown in Fig. 2A and quantitated in Fig. 2B. The AC mutant consistently showed normal levels of RT binding when compared to wt tRNA 3 Lys ( Fig. 2A, lanes 14 and 2, respectively). To further define the sequence and structure requirements for the binding of RT to tRNA 3 Lys , two additional mutants were made: one substituting 4 bases in the T⌿C stem ( Fig. 1, T⌿C mutant), and the other carrying a 6-nucleotide deletion in the D arm (D mutant). Mutations in the D arm affected RT binding most severely ( Fig. 2A, lane 4), and a partial reduction in RT binding efficiency was measured for the T⌿C mutant (lane 6).
We also analyzed the RT binding capacity of 3-point mutants that were fortuitously obtained in the context of the AC mutant (C3 3 U, G10 3 C, C13 3 U). Compared to the AC mutant, we consistently measured reduced RT recognition (ϳ20 -30%) for the two D arm mutants (G10 3 C and C13 3 U), whereas ϳ70% binding was measured for the acceptor stem mutant (C3 3 U). The results of these binding assays are summarized in Fig. 2B. The combined data suggest that the sequence and structure of the D arm in tRNA 3 Lys is critically important for RT binding.
To investigate whether RT could bind to extended forms of tRNA 3 Lys with additional nucleotides added to either its 5Ј-or 3Ј-end, we synthesized two different transcripts. First, we used an aberrant plasmid construct with two tandem T7 promoter elements. Transcription of this plasmid will produce a mixture of two RNAs, wt tRNA 3 Lys and a 5Ј-elongated form with 24 additional nucleotides ( Fig. 2A, lane 11). RT binding assays indicated that this 5Ј-extended tRNA 3 Lys did efficiently form a complex with the RT protein (lane 12). Second, we generated 3Ј-extended transcripts by using the downstream EcoRI restriction site for run-off transcription. This results in the synthesis of a 79-nucleotide-long transcript, which is 5 nucleotides longer than wt tRNA 3 Lys (Fig. 2A, lane 15). It should be noted that this 3Ј-elongated transcript is only 3 nucleotides longer than natural tRNA 3 Lys because our synthetic wt tRNA is lacking the 3Ј-terminal CA dinucleotide (Fig. 1). In contrast to the results obtained for 5Ј-extended tRNA, we consistently measured reduced RT binding for the 3Ј-extended molecule (lane 16). Furthermore, we measured no binding activity for a transcript with 171 additional 3Ј-nucleotides (data not shown).
We observed some surprising effects of the tRNA 3 Lys size variants on the electrophoretic mobility of the binary RT complexes. As expected, the free 5Ј-extended tRNA 3 Lys molecule migrated slower in the polyacrylamide gel compared to wt tRNA 3 Lys (Fig. 2A, lane 11). The complex of RT protein with this extended RNA mutant, however, ran ahead of the wt tRNA 3 Lys -RT complex (lane 12). This result was confirmed in binding experiments with a gel-eluted, purified form of the 5Ј-extended tRNA (data not shown). It seems plausible that it is primarily the conformation of the RNA-protein complex and not so much the molecular weight of its constituents that determines the migration in a native gel. Since it has been reported that the RT polypeptide is a flexible protein and that substantial conformational changes occur upon primer binding (34), our findings may suggest that this primer-induced conformational change in RT is affected by 5Ј-extension of the tRNA 3 Lys primer. No such migration effect was observed for the RT complexed with the 3Ј-extended form of tRNA 3 Lys (lane 16). Reverse Transcriptase Can Recognize a Preformed tRNA-PBS Complex-We next tested whether a preformed tRNA 3 Lys -PBS complex could still be specifically recognized by the RT protein. We therefore synthesized an oligonucleotide that mimics the PBS sequence of the HIV-1 template RNA. This PBS oligonucleotide can form a 16-base pair duplex with the 3Ј-end of synthetic tRNA 3 Lys (Fig. 1, nucleotide positions 59 -74), thereby disrupting both the acceptor and T⌿C stem regions. To follow both nucleic acid moieties, we performed binding assays with 32 P label on either the tRNA 3 Lys primer or the PBS template (Fig. 3, A and B, respectively). Formation of the tRNA-PBS complex was found to be dependent on a denaturation step (lanes 3), which is expected because the tRNA cloverleaf structure needs to be partially unfolded for the PBS oligonucleotide to gain access to its target sequence in the acceptor and T⌿C stems. Most importantly, we detected a "supershifted" RT-tRNA-PBS complex upon subsequent incubation with HIV-1 RT (lanes 4). This ternary complex can be visualized with either a labeled tRNA or PBS species. These results indicate that RT can specifically recognize the preformed tRNA-PBS complex. Alternatively, RT may have bound the tRNA and DNA probes as individual molecules, perhaps by using different protein domains. We can rule out this possibility because formation of the ternary complex is dependent on the presence of a pre-assembled tRNA-PBS complex and not seen with both RNA/DNA components present as individual molecules due to a 0°C pre-incubation (lanes 5). Furthermore, ternary complex formation was specific for the PBS oligonucleotide and not seen with unrelated probes that do not bind tRNA 3 Lys (e.g. oligonucleotide 3 in Fig. 4A, lane 6).
As described earlier, we observed some unexpected gel migration effects with RT-nucleic acid complexes. Although no linear relationship exists between the size of a nucleic acidprotein complex and its electrophoretic mobility on native gels, we do think that the shift seen upon inclusion of the relatively small PBS oligonucleotide in the RT-tRNA complex is rather dramatic. A plausible explanation is that the RT polypeptide adopts a different conformation in the binary versus the ternary complex. Consistent with this idea, it has been reported that RT binds a primer differently depending on the presence or absence of template (35).
We next tested whether the T⌿C loop needs to be accessible for interaction with the RT protein with an extended version of the PBS oligonucleotide; PBS ext , which forms a duplex with the 3Ј-terminal 21 nucleotides of synthetic tRNA 3 Lys , thereby blocking the complete T⌿C loop (Fig. 1, nucleotide positions 54 -74). In contrast to the results obtained with the PBS oligonucleotide, we found that RT could not recognize the preformed tRNA-PBS ext complex (Fig. 4A, lane 4; Fig. 4B, lane 5). These results suggest that the T⌿C loop nucleotides are important for recognition by the RT protein. It is important to note that the PBS ext probe does not exert a general toxic effect on RT activity because inhibition by PBS ext is restricted to the situation in which the oligonucleotide is annealed to tRNA 3 Lys in a 85°C pre-incubation (Fig. 4A, lane 4). The control 0°C pre-incubation showed efficient RT-tRNA 3 Lys complex formation in the presence of a free PBS ext probe (Fig. 4A, lane 3). Furthermore, unrelated oligomers were unable to interfere with the RT-tRNA interaction, even upon 85°C pre-incubation with tRNA (Fig. 4A, lane 6, and data not shown).
RT Protein Facilitates the tRNA-PBS Annealing-It has been proposed that binding of the RT protein to tRNA 3 Lys results in opening of the acceptor stem, thus facilitating the subsequent annealing to the PBS site (22). We tested this idea by incubation of either the preformed RT-tRNA 3 Lys complex or the free tRNA 3 Lys with the PBS-mimicking oligonucleotide at increasing temperatures (Fig. 5, panels A and B, respectively, with quantified data in panels C and D, respectively). The majority of the binary RT-tRNA 3 Lys complex was supershifted to the ternary RT-tRNA 3 Lys -PBS complex at a relatively low temperature of 30°C, and complete conversion was observed at 35°C (Fig. 5, A and C). We note that the ternary complex disappeared at higher incubation temperatures (50°C, material present in the slot), which is most likely due to the formation of insoluble protein aggregates. Most importantly, a 50°C incubation was required for significant PBS binding to the free tRNA molecule (Fig. 5, B and D). This functional test suggests that RT directly influences the PBS binding capacity of tRNA 3 Lys . One could argue that the in vitro binding assay does not accurately reflect the in vivo situation, where it is the Gag-Pol precursor protein instead of a processed RT form that is involved in packaging of cellular tRNA (10). Therefore, we analyzed the status of the tRNA-viral RNA complex in mutant virions lacking the RT protein. A large deletion was introduced into the RT gene of the molecular HIV-1 clone pLAI. As expected, this RT Ϫ mutant is replication incompetent, but normal levels of virions can be produced in transient transfection assays in HeLa cells (32). We isolated total RNA from purified virions and scored for the presence of tRNA and vRNA with primer extension assays (Fig. 6, lanes 1-3 and 4 -6, respectively). A normal level of viral RNA was detected in RT Ϫ compared with wild-type virus particles (lanes 6 and 4, respectively). In   FIG. 3. RT can recognize a preformed tRNA 3 Lys -PBS complex. A, gel shift assay with 32 P-labeled tRNA 3 Lys in the absence or presence of RT and the primer binding site oligonucleotide PBS (indicated on top of the panel). tRNA 3 Lys was pre-incubated with excess PBS oligonucleotide (3 g) either for 15 min at 0°C or for 2 min at 85°C, followed by slow cooling to 20°C. The subsequent incubation with RT was for 15 min at 20°C. The migration position of the individual RNA/DNA components and the binary and tertiary complexes are indicated. The slower migrating RNA species present in the tRNA sample (lane 1) is a tRNA conformer that disappeared under denaturing gel electrophoresis conditions (data not shown). B, gel shift assay with 32 P-labeled PBS oligonucleotide in the absence or presence of RT and the tRNA 3 Lys primer (indicated on top of the panel). tRNA 3 Lys was pre-incubated with PBS either for 15 min at 0°C or for 2 min at 85°C, followed by slow cooling to 20°C. The subsequent incubation with RT was for 15 min at 20°C. A longer exposure of lanes 3 and 4 is shown to allow identification of the 32 P PBS oligonucleotide in the ternary complex with RT and tRNA 3 Lys .

FIG. 4. Inhibition of the RT-tRNA 3
Lys binding by masking of the T⌿C loop. A, gel shift assay with 32 P-labeled tRNA 3 Lys in the absence or presence of RT, the PBS ext , or the control oligonucleotide 3 (indicated on top of the panel). tRNA 3 Lys was pre-incubated with 3 g of the oligonucleotides indicated either for 15 min at 0°C or for 2 min at 85°C, followed by slow cooling to 20°C. The subsequent incubation with RT was for 15 min at 20°C. The migration position of the individual RNA/DNA components and the binary complexes are indicated. The slower migrating RNA species present in the tRNA sample is a conformer that disappeared under denaturing gel electrophoresis conditions (data not shown). B, gel shift assay with the 32 P-labeled PBS ext oligonucleotide in the absence or presence of RT and the tRNA 3 Lys primer (indicated on top of the panel). tRNA 3 Lys was pre-incubated with PBS either for 15 min at 0°C or for 2 min at 85°C, followed by slow cooling to 20°C. The subsequent incubation with RT was for 15 min at 20°C. contrast, we measured a dramatic reduction in the amount of tRNA primer associated with the RNA genome of RT Ϫ compared to wild-type particles (lanes 3 and 1, respectively). A tRNA occupancy of only ϳ2% was estimated from overexposed gels (not shown). These combined results clearly demonstrate that the RT domain, which is not required for virion assembly and packaging of the viral RNA genome, is essential for the establishment of a functional tRNA primer-template complex in vivo.
A Comparative Analysis of RNA Molecules Specifically Recognized by RT Protein-It has been proposed that the RT protein is responsible for the in vivo selection of tRNAs from the pool of host cell tRNA species. This idea is supported by the apparent absence of primer tRNA in RT-defective HIV-1 particles (10). Wild-type HIV-1 virions contain 4 major-abundance tRNA species: tRNA 3 Lys primer, the tRNA 1 Lys and tRNA 2 Lys isoacceptors, and tRNA Ile (4). To identify features that the tRNAs may have in common, we present the sequence and cloverleaf secondary structure of the 4 tRNAs in Fig. 7A and a consensus tRNA sequence in Fig. 7B. Despite the fact that the D-stem loop is relatively well conserved among class III tRNAs (Fig. 7B), some sequence variation is allowed. Strikingly, the D-stem loop of the 4 tRNAs present in HIV-1 virions have similar loop sequences and share an identical stem region. In contrast, the anticodon stem-loop sequences are less conserved among the 4 tRNAs. This comparative analysis is consistent with our experimental data showing that the D-stem loop, but not the anticodon loop, is critical for RT binding. Another class of RT-binding RNA molecules was recently identified in a search for high affinity inhibitors of the HIV-1 RT protein. Tuerk et al. (27) used the SELEX procedure (systematic evolution of ligands by exponential enrichment) to isolate RT binding molecules from a randomized RNA population. This selection procedure did enrich for sequences with a pseudoknot structure, without an apparent homology to tRNA 3 Lys or other naturally occurring sequences available in the GenBank data base (27). Reexamination of this consensus motif, however, allowed us to recognize a striking similarity to the D arm of tRNA 3 Lys (Fig. 7C). First, both structures consist of a 4-base pair stem and a loop of 8 nucleotides. Second, multiple loop nucleotides in both structures are involved in tertiary base pairing interactions with other segments of the same molecule. The SELEX RNA motifs are characterized by a pseudoknot interaction between loop nucleotides and sequences 3Ј of the hairpin. Likewise, the D-loop nucleotides are known to interact with other nucleotides in the formation of the typical L-shaped tertiary tRNA structure (e.g. A14-U8 and G18-⌿55).
Although a significant variability in the primary sequence of the collection of selected pseudoknots was found, a well conserved stem and a strong bias for A nucleotides at multiple single-stranded positions was reported (27). Little sequence similarity is apparent for the stem regions of the SELEXpseudoknot and the D arm of tRNA 3 Lys (only 1 out of 4 base pairs is identical), but 4 A nucleotides are flanking both stems (circled in Fig. 7C). It should be noted that only two of these A nucleotides are absolutely conserved among different tRNA species (Fig. 7B), suggesting that this characteristic may be one of the features used by RT to discriminate between cellular tRNAs. The combined results of the SELEX approach and our mutational analysis suggest that RT may recognize certain features in the D-stem loop in the L-shaped tertiary structure of tRNA 3 Lys .

DISCUSSION
Whereas detailed biochemical studies of the HIV-1 RT-tRNA 3 Lys binary complex have been presented (7, 18 -25), a limited number of mutagenesis experiments have been performed. Two recent studies (38,39) analyzed the RT protein domain(s) involved in tRNA binding, and two studies reported binding experiments with mutated forms of synthetic tRNA 3 Lys (18,24). The binding experiments presented by Barat et al. (18) and the data presented in this manuscript indicate that the anticodon loop of tRNA 3 Lys is not directly involved in the interaction with RT. A similar conclusion was reached by Huang et al. (40) based on the efficient incorporation into virus particles of a mutant tRNA 3 Lys species with an altered anticodon sequence. Our in vitro binding data do suggest that the T⌿C and especially the D arm nucleotides play a critical role in selection of the tRNA 3 Lys primer for reverse transcription of the HIV-1 viral genome. We like to note that both stem loops are on the outside of a native tRNA molecule (41) and therefore readily accessible for interaction with RT amino acids. The finding that the D arm is important for RT binding is also supported both by a comparative sequence analysis of both the subset of tRNA species that are abundantly present in HIV-1 particles (4) (Fig.  7A) and the RNA molecules selected for RT binding in vitro (27) (Fig. 7C). The latter SELEX procedure did enrich for RNA pseudoknot motifs that resemble the structure and sequence of the D arm of tRNA 3 Lys . This RNA motif was also shown to selectively inhibit the HIV-1 RT function. These combined results strongly suggest an involvement of the D arm in specific RT binding. Furthermore, Tuerk et al. (27) used two randomized starting templates, either with or without the tRNA 3 Lys anticodon arm. No difference in the affinity of these two RNA populations for HIV-1 RT was found, and the subsequent selection process was indifferent to the presence of the anticodon sequences. This result confirms that the anticodon loop is not involved in the tRNA-RT interaction in a sequence-specific manner.
We showed efficient binding of RT protein to the binary tRNA 3 Lys -PBS complex, suggesting that RT binding requires neither the intact acceptor stem nor the intact T⌿C stem. It seems likely that the interaction of tRNA 3 Lys with the PBS sequence will still allow for the formation of both the anticodon and D-stem loop structures (42), but it is unknown whether the tertiary D-T⌿C loop interaction is maintained in the PBS-tRNA complex. The finding that extension of the PBS oligonucleotide over the T⌿C loop did completely block RT recognition may suggest that the T⌿C loop contains critical recognition nucleotides that interact with RT. Alternatively, the importance of the T⌿C loop may reflect recognition of the typical L-shaped structure of tRNA molecules, which is dependent on the D-T⌿C loop interaction (41,43). As pointed out in the results section (Fig. 7C), the RT SELEX experiment suggests that the tertiary RNA folding is critical for RT binding. Because all tRNAs are structurally quite similar, tRNA 3 Lys -specific nucleotides are also expected to contribute to the observed specificity of binding.
We cannot currently explain the differences between our results and the binding data of Weiss et al. (24), who reported efficient RT binding with the 3Ј-terminal 24 nucleotides of tRNA 3 Lys . It is possible that at least some of the experimental discrepancies can be attributed to the use of different RT and tRNA reagents. In general, binding studies have been performed with synthetic or natural tRNA 3 Lys molecules and with many different forms of mature RT protein (66,000 homodimer  1 and 4) or an RT deletion mutant (RT Ϫ ; lanes 3 and 6). A sample of mock-transfected cells was used as a negative control (lanes 2 and 5). Viral RNA genomes were extracted, and the associated tRNA primer was visualized in a tRNA-extension assay upon addition of RT enzyme and [ 32 P]dNTPs (lanes 1-3). The viral RNA was analyzed in a standard primer extension assay with a DNA oligonucleotide primer complementary to positions ϩ123/ϩ151 of HIV-1 RNA (lanes 4 -6). The position of the respective products is indicated; the tRNA-cDNA molecule is 257 nucleotides, and the cDNA product is 151 nucleotides in length. or 66,000/51,000 heterodimer, 66,000 or 51,000 monomer). In this respect, it should also be noted that, whereas all in vitro binding studies use these mature RT species, in vivo primer selection is believed to occur during the initial stages of virus assembly when only the precursor Gag-Pol fusion protein Pr160K is available. Experiments with RT-deleted HIV-1 virus particles clearly demonstrated the involvement of the RT domain in selective tRNA packaging (10) and annealing of the tRNA primer to the viral RNA genome (this study). We note that although it has been suggested that tRNA 3 Lys binding may involve both subunits of the RT dimer (44,45), it is currently unknown whether the Gag-Pol precursor exists as a dimer.
Upon packaging of the tRNA primer, the cloverleaf RNA structure should be partially melted to expose its 3Ј-end for binding to the complementary PBS sequence on the viral RNA genome. Using footprint analysis, Sarih-Cottin et al. (22) originally reported that HIV-1 RT binding resulted in unwinding of the acceptor stem. In contrast, Wöhrl et al. (25) saw little evidence for such an effect and suggested that excessive nuclease digestion could have hampered the former study. Our binding experiments indicate that RT protein stimulates the annealing of a PBS oligonucleotide to the binary RT-tRNA 3 Lys complex (Fig. 5). Consistent with this idea, we demonstrated that mutant RT Ϫ HIV-1 virions contain a viral RNA genome lacking the complementary tRNA 3 Lys primer. A critical role for the HIV-1 RT protein in packaging of the tRNA 3 Lys primer is now well established based on in vitro binding experiments (7, 18 -25) and the phenotype of RT Ϫ viruses (10,32). This may suggest that the complementary PBS sequence on the viral RNA is less important for encapsidation of the proper primer tRNA molecule. Interestingly, HIV-1 mutants with altered PBS identities were recently constructed and tested for replication competence (33,46). Such PBS mutants are forced to use primers other than tRNA 3 Lys and exhibit severe replication defects. Furthermore, reversion to the wildtype tRNA 3 Lys was observed in both studies upon prolonged culture (33,46). These results convincingly demonstrate that tRNA primer selection is determined primarily by the HIV-1 RT protein. This does not, however, rule out a role for other RNA/protein factors in packaging or annealing of the tRNA primer. First, distinct tRNA regions, especially the singlestranded loop regions, may initially anchor the primer on the template in the vicinity of the PBS by analogy to the "kissing" step in ColE1 plasmid replication (47), and several non-PBS base-pairing interactions between HIV RNA and tRNA 3 Lys have been proposed (11)(12)(13)(14)(15)53). Second, the nucleocapsid protein FIG. 7. Sequence similarities between tRNA 3 Lys and other HIV-1 virion tRNAs. A, the cloverleaf structure of the four tRNA species isolated from HIV-1 virus particles are shown (4). Sequence differences when compared to tRNA 3 Lys are shaded. tRNA 1 Lys and tRNA 2 Lys differ only by 1 base pair in the anticodon stem. tRNA Ile contains 1 additional nucleotide in the D loop, which is indicated by ϩD. B, consensus sequence of class III tRNA molecules. The sequence of absolutely conserved nucleotides is shown (37). Pyrimidine, purine, and random positions are indicated. C, comparison of the consensus pseudoknot structure identified in an in vitro SELEX experiment (27) and the D arm of tRNA 3 Lys . N is any nucleotide. For both structures, we circled the four A residues that flank the stem region. Tertiary interactions are indicated by thin lines. In the pseudoknot structure, four "loop" nucleotides base pair to sequences 3Ј to the hairpin. In the D loop, four nucleotides are involved in tertiary interactions, with the base pairing partner indicated in bold, e.g. U8.
NC has been suggested to bind and unwind tRNA (26,48), although this interaction lacks specificity for the tRNA 3 Lys molecule (49). Because both RT and NC domains are part of the Gag-Pol precursor polyprotein, these subunits may cooperate in the initial interactions with the tRNA primer.
Finally, our experimental data do suggest that tRNA packaging and tRNA-mediated initiation of reverse transcription are not necessarily performed by the same RT molecule since RT can efficiently recognize a pre-assembled tRNA-PBS complex. According to this scenario, a Gag-Pol precursor may be involved in tRNA packaging and annealing of the primer to the viral RNA template, whereas a second, mature RT protein may play an active role in initiation of reverse transcription on the pre-assembled tRNA-PBS complex. Consistent with this hypothesis, these two reactions are widely separated both in time and place. Whereas tRNA encapsidation occurs on the surface of virus-producing cells, reverse transcription is only initiated upon entry of the viral particle into cytoplasm of newly infected cells, although a low level of cDNA was found in HIV-1 virus particles using sensitive PCR techniques (50,51). In support of this "multi-RT" mechanism, retroviral particles have been reported to contain a molar excess of RT protein (1,52).