Mutating a Region of HIV-1 Reverse Transcriptase Implicated in tRNALys-3 Binding and the Consequences for (−)-Strand DNA Synthesis*

Recently, tRNALys-3 was cross-linked via its anticodon loop to human immunodeficiency virus type 1 (HIV-1) reverse transcriptase (RT) between residues 230 and 357 (Mishima, Y., and Steitz, J. A. (1995) EMBO J. 14, 2679–2687). Scanning the surface of this region identified three basic amino acids Lys249, Arg307, and Lys311 flanking a small crevice on the p66 thumb subdomain outside the primer-template binding cleft. To assess an interaction of this region with the tRNA anticodon loop, these p66 residues were altered to Glu or Gln. p66 subunits containing K249Q, K311Q, K311E, and a dual R307E/K311E mutation formed a stable dimer with wild type p51. All mutants showed reduced affinity for tRNALys-3 and supported significantly less (−)-strand DNA synthesis from this primer than the parental heterodimer. In contrast, these variants efficiently synthesized HIV-1 (−)-strand strong-stop DNA from oligonucleotide primers and had minimal effect on RNase H activity, retaining endonucleolytic and directed cleavage of an RNA/DNA hybrid. Structural features of binary RT·tRNALys-3 complexes were examined by in situ footprinting, via susceptibility to 1,10-phenanthroline-copper-mediated cleavage. Unlike wild type RT, mutants p66K311Q/p51 and p66K311E/p51 failed to protect the tRNA anticodon domain from chemical cleavage, indicating a significant structural alteration in the binary RT·tRNA complex. These results suggest a crevice in the p66 thumb subdomain of HIV-1 RT supports an interaction with the tRNALys-3 anticodon loop critical for efficient (−)-strand DNA synthesis.

Initiation of reverse transcription is a complex step in the retroviral life cycle. With human immunodeficiency virus type 1 (HIV-1), 1 several specific events, occurring prior to virus entry, precede processive synthesis of (Ϫ)-strand DNA. The first of these is preferential packaging of tRNA Lys isoacceptors and a dimer of the RNA genome into the assembling virion by gag-pol and gag precursors (p160 gag-pol and p55 gag , respectively) (2)(3)(4). tRNA Lys-3 is selectively placed on the viral RNA, serving as primer for (Ϫ)-strand DNA synthesis, by hybridization of 18 nt at its 3Ј terminus to the primer binding site (PBS), a complementary sequence immediately adjacent to the unique 5Ј region (U5) (5)(6)(7)(8)(9). Beyond this interaction, PBS-bound tRNA  has been shown in vitro to interact with additional sequences of the viral RNA genome, most notably an -AAAAsequence in the U5-inverted repeat (IR) loop that contacts the U-rich anticodon loop (10,11). These events are most likely established in an immature virus particle, prior to proteasemediated maturation of p55 gag and p160 gag-pol . However, mature HIV-1 proteins are required for initiation of (Ϫ)-strand DNA synthesis from tRNA Lys-3 (12,13). In vitro, initiation of (Ϫ)-strand synthesis involves addition of up to 6 deoxynucleotides to the tRNA Lys-3 3Ј terminus (14 -16), after which such products are elongated to full-length (Ϫ)-strand strong-stop DNA containing the U5 and repeat (R) sequences of the genome. Recently, it was shown that although the initiation event, i.e. addition of two deoxynucleotides to tRNA Lys-3 , can be accomplished in the mature virus (13,17), the level of (Ϫ)strand strong-stop DNA in HIV-1 was ϳ100-fold less than the amount of tRNA Lys-3 alone or tRNA Lys-3 extended by 2 nt, suggesting further elongation of (Ϫ)-strand DNA in the virion is a rare event (12,13).
HIV-1 proviral DNA synthesis is accomplished by reverse transcriptase (RT), a multifunctional, heterodimeric enzyme (p66/p51) containing a catalytically competent 66-kDa subunit and an inactive 51-kDa subunit (Fig. 1A). Although all retroviral RTs perform comparable functions, many complex steps, such as initiation of (Ϫ)-strand synthesis, require virus-specific substrates and catalysts (17,18). For example, we recently observed that of several retroviral enzymes, only HIV-1 and avian myeloblastosis virus RT efficiently initiated (Ϫ)-strand DNA synthesis from unmodified or natural tRNA Lys-3 primers annealed to the HIV-1 genome (18). This finding was unexpected, since the RTs of HIV-2, simian immunodeficiency virus, feline immunodeficiency virus, and equine infectious anemia virus used in our study also exploit tRNA Lys-3 as primer for initiation of reverse transcription (19). However, disrupting the U5-IR loop-tRNA Lys-3 anticodon loop complex in the HIV-1 template-primer substrate restored the capacity for (Ϫ)-strand DNA synthesis to the heterologous enzymes (18). Interaction of this A-rich sequence 5Ј to the PBS with the tRNA primer appears unique to HIV-1, since no barrier to tRNA-primed (Ϫ)-strand DNA synthesis by heterologous RTs from PBS-containing RNA fragments of the equine infectious anemia virus or feline immunodeficiency virus genome was observed (18).
We and others (14 -16, 18) have demonstrated that wild type HIV-1 RT is required to disrupt the U5-IR stem and U5-IR loop-anticodon loop complex by extending tRNA Lys-3 by 1-6 deoxynucleotides, after which processive synthesis from the extended tRNA yields full-length (Ϫ)-strand strong-stop DNA. The first HIV-1 RT mutant displaying a defect in this initiation program contained a 13-amino acid deletion at the C terminus of p51 while retaining a full-length p66 subunit. This mutant, p66/p51⌬13 RT, exhibits reduced affinity for the free replication primer in a binary complex and an inability to catalyze (Ϫ)-strand DNA synthesis from PBS-bound tRNA while accomplishing the same event from a PBS-bound RNA primer (20,21). In addition, the 13-residue p51 truncation indirectly influences the architecture of the p66 RNase H domain, evidenced by a dysfunctional RNase H activity, which surprisingly could be restored by addition of HIV-1 nucleocapsid protein (21,22). For the present study, our goal was to identify amino acids in p66 HIV-1 RT that might be specifically implicated in the interaction with tRNA Lys-3 and subsequent initiation of (Ϫ)strand DNA synthesis without compromising any other enzymatic function.
Although HIV-1 RT has been shown to interact with tRNA Lys-3 at all three loops (23)(24)(25)(26), its affinity for most tRNA isoacceptors is similar (27,28), suggesting structure is the primary determinant of this bimolecular interaction rather than sequence. Recently, a 127-amino acid cyanogen bromide (CNBr) cleavage fragment of p66 HIV-1 RT (residues 230 -357 and depicted in white in Fig. 1A) was cross-linked to synthetic tRNA Lys-3 thiolated at U 36 of the anticodon loop (1). By analyzing this peptide fragment in the context of the three-dimensional structure of HIV-1 RT (29,30), we tentatively identified three basic residues (Lys 249 , Arg 307 , and Lys 311 ) conserved in several HIV-1 isolates (31) and adjacent to a small hydrophobic crevice in the thumb subdomain of p66 (Fig. 1A) as targets for mutagenesis. In contrast to wild type RT, these mutants showed greatly reduced affinity for tRNA Lys-3 and failed to protect regions in the anticodon stem-loop from chemical cleavage, the latter of which was evaluated via a novel in situ footprinting approach. Although mutant enzymes retained the capacity to initiate (Ϫ)-strand DNA synthesis from PBS-bound tRNA Lys-3 , subsequent elongation events were severely reduced. General defects to polymerase function were ruled out by the observation that the same mutants retained their ability to (i) synthesize (Ϫ)-strand DNA from PBS-bound RNA and DNA primers and (ii) catalyze synthesis-dependent and -independent modes of RNase H activity. Collectively, our findings suggest that mutating Lys 249 , Arg 307 , or Lys 311 of the thumb subdomain of the catalytically competent p66 subunit induces structural alterations in p66/p51 HIV-1 RT which disrupt interactions with its cognate tRNA replication primer.
Primers and RNA Templates-Four primers were utilized for this study and placement of the radiolabel differed for the various experiments. 18-nt DNA and RNA primers were 5Ј-end-labeled with [␥-32 P]ATP and T4 polynucleotide kinase according to standard procedures. Synthetic tRNA Lys-3 for electrophoretic mobility shift analyses was internally labeled with ␣-[ 32 P]UTP during its in vitro synthesis with T7 RNA polymerase (Ambion). Fully modified, natural tRNA Lys-3 was labeled by extension with a single deoxynucleotide (␣-[ 32 P]dCTP) utilizing the Klenow fragment of E. coli DNA polymerase on a 60-nt DNA template homologous to the region surrounding the PBS in HIV-1 genomic RNA (21). 3Ј-[ 32 P]dCMP-labeled tRNA Lys-3 was purified as described (21). In situ footprinting employed natural and synthetic tRNA Lys-3 3Ј-end-labeled with [ 32 P]pCp by RNA ligase. Labeled nucleotides were obtained from NEN Life Science Products and molecular biology reagents from Boehringer Mannheim. Finally, a 237-nt HIV-1 PBS-containing RNA template was prepared by in vitro transcription of HgaI-linearized pHIV-1 PBS vector as described (21). This template lacks the dimer initiation sequence located between the PBS and gag initiation codon.
Gel Mobility Shift Analyses and K d Determination-Wild type and mutant enzymes were incubated for 10 min at room temperature with 3Ј-end-labeled natural, 32 P internally labeled synthetic tRNA Lys-3 or an end-labeled tRNA/PBS RNA duplex, in binding buffer (25 mM KCl, 50 mM Tris, pH 7.4, 10% glycerol, and 1 mM dithiothreitol), then fractionated through a 5% non-denaturing polyacrylamide gel. Following electrophoresis, gels were dried and visualized by autoradiography. The stability of RT bound to free tRNA Lys-3 or tRNA Lys-3 -template complexes was evaluated using a 4:1 ratio of RT (16 nM) to tRNA (4 nM), i.e. an RT concentration below the K d for wild type RT binding to free tRNA Lys-3 . Higher ratios of RT:tRNA resulted in a complete shift of free tRNA Lys-3 by wild type RT, whereas lower ratios prevented an analysis of tRNA Lys-3 binding by mutant RTs (data not shown).
In order to determine dissociation constants, the concentration of synthetic tRNA Lys-3 in reactions remained constant at 0.1 nM, whereas the RT concentration was increased from 0.1 nM to 1.3 M. The binary RT⅐tRNA complex was separated from free tRNA on a 5% non-denaturing polyacrylamide gel, and the amount of shifted complex calculated from phosphorimaging analyses of dried gels and plotted against RT concentration. A hyperbolic curve was fitted to the plots, and dissociation constants were estimated from the RT concentration resulting in 50% shifting of synthetic tRNA Lys-3 . Dissociation constants were determined as the average of three independent experiments.
1,10-Phenanthroline-Copper Footprinting of RT⅐tRNA Lys-3 Complexes in Situ-Higher concentrations of RT (200 nM) and 32 P 3Ј-end-labeled tRNA Lys-3 (33 nM) were necessary to increase the amount of the binary RT⅐tRNA complex resolved by non-denaturing polyacrylamide gel electrophoresis as described above. Wet gels were exposed to Kodak Biomax MS film for 20 min to visualize the shifted complexes and free tRNA. Gel slices containing these bands were excised, immersed in 100 l of 50 mM Tris, pH 7.4, and then probed with chemical reagents according to Kuwabara and Sigman (34).
Modified tRNA was eluted from gel slices overnight at 37°C after adding 270 l of gel elution buffer (0.5 M ammonium acetate, 1 mM EDTA, 0.1% SDS). Cleaved tRNA was precipitated with ethanol, resuspended in loading buffer (90% formamide, 1 mM EDTA, 0.05% bromphenol blue, and 0.05% xylene cyanol), and then fractionated on a 12% denaturing polyacrylamide gel. Gels were dried and visualized by autoradiography. Ladders of tRNA Lys-3 cleavage products generated by NaOH treatment were used as size markers. In order to control for nuclease contamination, RT⅐tRNA complexes and free tRNA were eluted from unprobed gel slices and fractionated in parallel with the experimental samples.
RNA-dependent DNA Polymerase Activity-RNA-dependent DNA polymerase activity of wild type and mutant enzymes was measured using an HIV-1 PBS-containing RNA template to which a 32 P 5Ј-endlabeled 18-nt DNA or RNA primer or 77-nt chimera of natural tRNA Lys-3 containing radiolabeled dCTP at its 3Ј terminus (21) was hybridized. Primers were annealed to the 237-nt RNA template prior to addition of HIV-1 RT (57 nM), the appropriate Mg 2ϩ -containing buffer, and dNTPs (18,21). Reaction mixtures were incubated at 37°C for 1, 10, and 80 min and then quenched with loading buffer. (Ϫ)-Strand DNA products were resolved on a 6 or 12% denaturing polyacrylamide gel and visualized by autoradiography.
RNase H Activity-RNase H activity was determined as described previously (20). Briefly, HIV-1 RT (57 nM) was incubated with a 36-nt DNA primer (40 nM) annealed to a 32 P 5Ј-end-labeled 90-nt RNA template (40 nM) prior to the addition of Mg 2ϩ . Aliquots were removed at 10 s and 1, 10, and 30 min and supplemented with gel loading buffer. Hydrolysis products were fractionated on a 10% denaturing polyacrylamide gel which was subsequently dried and subjected to autoradiography.

Potential tRNA Lys-3 -binding Sites in HIV-1 RT-Mishima
and Steitz (1) have successfully cross-linked synthetic tRNA Lys-3 , containing thiouridine at position 36 of the anticodon loop, to p66/p51 HIV-1 RT. CNBr cleavage of this complex indicated covalent linkage to a 127-amino acid fragment encompassing residues 230 -357 of the p66 subunit (represented in white in Fig. 1), corresponding to the entire thumb and a portion of the connection subdomain. A prominent feature of anticodon loop of mammalian tRNA Lys-3 are two heavily modified nucleotides at positions 34 (5Ј-methoxycarbonylmethyl-2Јthiouridine, mcm 5 s 2 U 34 ) and 37 (2Ј-methylthio-N 6 -threonylcarbamoyl adenosine, ms 2 t 6 A 37 ) (35). By using a space-filling model of HIV-1 RT (29), we tentatively identified 5 of 13 positively charged amino acids in this cross-linked fragment (Lys 249 , Lys 259 , Lys 287 , Arg 307 , and Lys 311 ) which might lie on the surface of the p66 thumb subdomain. Of these Lys 249 , Arg 307 , and Lys 311 appear clustered around a small crevice in the thumb opposite to the primer-template binding cleft ( Fig.  1). Two proposed flaps, designated flap a and flap b in Fig. 1, have hydrophilic residues on the surface (marked by dotted line), whereas the crevice floor contains mainly hydrophobic residues.
A sequence alignment of HIV-1 isolates from four different clades (or subtypes), as well as isolates of HIV-2, simian immunodeficiency virus, feline immunodeficiency virus, and equine infectious anemia virus was undertaken (data not shown). Residues Lys 249 , Arg 307 , and Lys 311 are conserved in 50 HIV-1 strains, comprising every HIV-1 subtype (a mutation frequency of Ͻ3.6 ϫ 10 Ϫ5 substitutions/nt for these three residues compared with 2.5 ϫ 10 Ϫ2 substitutions/nt for entire pol gene) (31). These three positively charged residues are not conserved in the coding regions of other lentiviral RTs. However, hydrophilicity in the flaps and hydrophobicity in the crevice appear to be maintained by semi-conservative substitutions in HIV-1, HIV-2, and simian immunodeficiency virus. Interestingly, HIV-2 and simian immunodeficiency virus RT had reduced affinity for tRNA Lys-3 compared with HIV-1 RT and failed to disrupt the tRNA anticodon loop-HIV-1 RNA U5⅐IR loop complex for productive (Ϫ)-strand DNA synthesis (18). From this analysis, we speculated that Lys 249 , Arg 307 , and Lys 311 in HIV-1 RT might be responsible for accommodating the electronegative anticodon loop of tRNA Lys-3 and important for initiation and/or elongation of (Ϫ)-strand DNA synthesis.
Lys 249 , Arg 307 , and Lys 311 were substituted with Glu or Gln in a p66 RT expression vector by site-directed mutagenesis. The hydrophilic Glu and Gln residues were designed to remove the positive charge while maintaining the flap crevice structure determined for wild type enzyme (Fig. 1). Mutated p66 subunits were reconstituted with a polyhistidine-extended, wild type p51 subunit and purified by metal chelate chromatography, which also monitors dimerization efficiency (20,36). Mutants p66 K249Q /p51, p66 K311Q /p51, p66 K311E /p51, and p66 R307E/K311E /p51 RT were stable heterodimers and reconstituted efficiently with a 1:1 subunit stoichiometry. In contrast, p66 K249E , p66 R307Q , and p66 R307E RT failed to efficiently reconstitute with p51, suggesting the alterations may have affected protein stability and dimerization.
Affinity of HIV-1 RT Mutants for tRNA Lys-3 -To determine the effects of the p66 point mutations on tRNA binding, we performed gel mobility shift analyses with the wild type enzyme and the above-mentioned variants. For comparison, the selectively deleted mutant p66/p51⌬13 was included. The results of our analysis are presented in Fig. 2. A complex characteristic of wild type HIV-1 RT bound to natural tRNA Lys-3 (23, 37) is shown in Fig. 2A. At the RT:tRNA ratio which was necessary to visualize complexes containing mutant RTs, two retarded species (labeled a and b in Fig. 2A) were identified for the wild type enzyme, the slower-migrating of which (labeled a in Fig. 2A) most likely represents a dimer of the binary complex. 2 Reduced stability of the binary complex was clearly evident with all mutants. With p66 R307E/K311E /p51 RT and, to a lesser extent, p66 K311Q /p51 and p66 K311E /p51 RTs, a third species (labeled c in Fig. 2A) migrating faster than the tRNA⅐RT complex was observed. The migration position of this complex correlates with a binary complex of p51 RT and tRNA Lys  (1) as participating in tRNA binding, i.e. K 249 , R 307 , and K 311 , appear surface-exposed and clustered around a small crevice on the thumb subdomain opposite to the primer-template binding cleft. Lower, expanded illustration of flaps (a and b) surrounding this crevice which may be important in accommodating the tRNA replication primer. these mutations might induce dissociation. Although retaining a stable heterodimeric organization, data of later sections indicate p66 R307E/K311E /p51 RT has substantially reduced DNA polymerase and RNase H activity, consistent with a more pronounced structural perturbation.
In order to extend the gel mobility shift data, synthetic tRNA Lys-3 was used to determine dissociation constants (K d ) for each enzyme-tRNA Lys-3 complex, the results of which are presented in Fig. 2B. An increase in ionic strength (25-100 mM KCl) had little effect on the amount of shifted tRNA-wild type RT complex. In addition, a K d of 59 Ϯ 11 nM, determined for wild type RT and synthetic tRNA Lys-3 at 25 mM KCl, was in agreement with the value of 40 nM obtained with a buffer of slightly higher ionic strength (50 mM KCl) (28). This value increased 3-6-fold with mutant RTs, representing decreased affinity for free tRNA Lys-3 . As predicted p66 R307E/K311E /p51 and p66/p51⌬13 RTs had substantially reduced affinity for tRNA Lys-3 which again may be related to more severe architectural alterations induced by these mutations. Use of synthetic tRNA Lys-3 in place of natural tRNA Lys-3 is justified by a parallel experiment indicating proportional amounts of binary complex with wild type and mutant RTs (data not shown). Moreover, it has been demonstrated by Thrall et al. (28) that the affinity of HIV-1 RT for natural or synthetic tRNA Lys-3 is comparable when determined in solutions of slightly higher ionic strength.
In Situ Footprinting Analysis of RT⅐tRNA Lys-3 Complexes via 1,10-Phenanthroline Copper Cleavage-To determine how mutations in the p66 thumb subdomain might affect an interaction with the replication primer, free tRNA Lys-3 and binary RT⅐tRNA Lys-3 complexes were excised from a non-denaturing polyacrylamide gel and probed in situ with 1,10-phenanthroline copper (OP-Cu) according to the procedure of Kuwabara and Sigman (34). Cleavage products were eluted from the gel slice and fractionated by high resolution denaturing polyacrylamide gel electrophoresis, the results of which are given in Fig.  3. OP-Cu preferentially intercalates and cleaves singlestranded RNA. Both the A conformation of double-stranded RNA and bulky ribonucleoside modifications (such as those in the anticodon domain of natural tRNA Lys-3 ) can prevent intercalation of OP-Cu (38). Thus, enhanced OP-Cu cleavage in both the anticodon and T⌿C loops was observed in some singlestranded regions of synthetic tRNA Lys-3 (Fig. 3A) although less pronounced with the natural tRNA Lys-3 counterpart (Fig. 3B).
In situ footprinting of tRNA Lys-3 complexed to wild type RT revealed weak protection around the anticodon loop. This protection was consistently found in repeated OP-Cu footprinting experiments and in previous UV cross-linking (23) and enzymatic footprinting studies (25). Despite reduced affinity of the p66 K249Q /p51 RT for free tRNA, the pattern of tRNA protection with this mutant is unaltered from that observed with wild type RT (Fig. 2B). In contrast, RT mutants p66 K311Q /p51 and p66 K311E /p51, which display reduced affinity for tRNA Lys-3 , provide a strikingly different OP-Cu hydrolysis profile. Rather than protect the anticodon loop these p66 alterations give rise to augmented OP-Cu cleavage throughout this domain (enclosed by the upper bars in Fig. 3, A and B). Since OP-Cu does not intercalate into double-stranded RNA due to a more compact minor groove, cleavage in the anticodon stem may reflect (i) increased binding of OP-Cu at the loop-stem interface, permitting short distance diffusion of the hydroxyl radical and cleavage of the bases closest to the loop (38) or (ii) disruption of base pairing within the anticodon stem accompanying the interaction of mutant RT with tRNA Lys-3 . The latter is more probable considering OP-Cu cleavage in the 3Ј anticodon stem of natural tRNA Lys-3 does not occur immediately adjacent to the anticodon loop. Mutants p66 K311Q /p51 RT and p66 K311E / p51 RT also give rise to enhanced cleavage within the TC loop but only with synthetic tRNA Lys-3 . Cleavage of the TC loop was limited to the last two bases (G 59 and U 60 ) of free tRNA Lys-3 or tRNA Lys-3 complexed to wild type RT suggesting either a D-loop-T⌿C loop interaction (Fig. 3C) or protection by wild type RT, respectively. Increased cleavage in the TC loop of synthetic tRNA Lys-3 complexed with p66 K311Q /p51 RT or p66 K311E /p51 RT may be the result of differential binding by these mutants at the anticodon loop. Augmented OP-Cu cleavage in the anticodon stem and loop of both tRNA species complexed with p66 K311Q /p51 RT and p66 K311E /p51 RT suggests the mutations influence a specific interaction of RT with this region. Due to weak binding, insufficient amounts of p66 R307E/K311E /p51 RT-tRNA Lys-3 and p66/p51⌬13 RT⅐tRNA Lys-3 complexes could be recovered for in situ footprinting. It is important to note that extensive manipulations of free tRNA Lys-3 in solution and in gel make it impossible to completely prevent some weak hydrolysis of the tRNA. This weak hydrolysis was only found in single-stranded RNA regions suggesting a low level RNase A contamination.
(Ϫ)-Strand DNA Synthesis Supported by tRNA Lys-3 -Viral RNA Duplexes-Previous studies indicated that mutant or non-HIV-1 RTs with reduced affinity for tRNA Lys-3 do not support efficient synthesis of (Ϫ)-strand strong-stop DNA from tRNA Lys-3 annealed to an HIV-1 RNA template (18,21). Since p66 K249Q /p51, p66 K311Q /p51, p66 K311E /p51, and p66 R307E/K311E /  (37). The tRNA binding mutant p66/p51⌬13 has been described previously (20) and is included as a control. B, tRNA dissociation constants for wild type and mutant enzymes. In view of the amounts of tRNA required, these experiments were conducted with the in vitro synthesized version of tRNA Lys-3 . p51 RT showed reduced affinity for tRNA Lys-3 , we compared their ability to synthesize (Ϫ)-strand DNA from oligonucleotide primers and natural tRNA Lys-3 annealed to an HIV-1 PBScontaining RNA genome. The substrates employed in this assay and the predicted (Ϫ)-strand strong-stop products are schematically represented in Fig. 4.
As described previously, wild type HIV-1 RT supported efficient synthesis of tRNA Lys-3 -primed (Ϫ)-strand strong-stop DNA, whereas mutant p66/p51⌬13 was considerably less active ( Fig. 5A and Ref. 18). The absence of full-length (Ϫ)-strand strong-stop DNA with p66/p51⌬13 RT in our previous report (21) may be accounted for by the shorter incubation time in these experiments, whereas in the present experiment prolonged incubation and a slightly higher RT::template-primer ratio allows limited extension of prematurely terminated products. Under the same conditions, p66 K249Q /p51, p66 K311Q /p51, p66 K311E /p51, and p66 R307E/K311E /p51 RT support substantially less (Ϫ)-strand strong-stop DNA synthesis from a PBS-bound tRNA Lys-3 . In support of the data of Fig. 5A, a similar effect could be demonstrated with in vitro synthesized tRNA (data not shown). In order to determine whether this decrease in (Ϫ)-strand DNA synthesis was related to reduced affinity of mutant enzymes for the tRNA Lys-3 ⅐HIV-1 PBS RNA complex, we performed a gel mobility shift analysis with a pre-annealed tRNA Lys-3 -viral RNA duplex in the absence of DNA synthesis. In contrast to the differential binding observed with mutant RTs and free tRNA Lys-3 , the data of Fig. 5B show similar amounts of shifted RT⅐tRNA Lys-3 -viral RNA complex, with the exception of p66 R307E/K311E /p51 RT. Since accommodating the PBS-containing RNA duplex in the nucleic acid binding cleft and extended interactions with the viral RNA template outside this cleft (involving the p66 fingers subdomain and RNase H domain) will enhance the strength of this interaction, the data of Figs. 2 and 5B are not necessarily contradictory. Moreover, the strength of this "pre-initiation" complex is evidenced by our observation that bound enzyme could not be displaced with an excess of tRNA Lys-3 or viral RNA template (data not shown).
On closer examination of Fig. 5A, DNA synthesis products migrating in the immediate vicinity of the tRNA primer indicated that, despite a reduction in full-length (Ϫ)-strand strongstop DNA, mutant enzymes retained their capacity to initiate DNA synthesis. High resolution analysis of each (Ϫ)-strand were 3Ј-end-labeled with [ 32 P]pCp. Binary complexes were separated from free tRNA by non-denaturing polyacrylamide gel electrophoresis. Gel slices containing the RT⅐tRNA complexes (Fig. 2) were excised, immersed in 50 mM Tris, and then treated with OP-Cu (see "Experimental Procedures"). The OP-Cu cleavage products were then eluted from the gel slice, ethanol-precipitated, separated on a 12% denaturing polyacrylamide gel, and subjected to autoradiography. In order to validate its integrity natural or synthetic tRNA was loaded onto the denaturing polyacrylamide gel directly (lane 1) or after electrophoresis through a non-denaturing polyacrylamide gel and elution from the gel slice (lane 2). Lane 3, free tRNA probed with OP-Cu in situ. WT, p66/p51 HIV-1 RT; K249Q, p66 K249Q /p51 RT; K311Q, p66 K311Q /p51 HIV-1 RT; K311E, p66 K311E /p51 HIV-1 RT. Labels on either side of A and B indicate domains of tRNA Lys-3 . Solid lines in the middle of each panel represent regions of tRNA displaying altered OP-Cu cleavage when complexed with RT mutants. C, provides a schematic representation of the tertiary structure of tRNA Lys-3 modeled after the crystal structure of tRNA Met (50). An additional diagram presents its L-shaped structure and sequence. DNA synthesis reaction (Fig. 5C) indicated accumulation of prematurely terminated cDNAs corresponding to addition of 3-5 deoxynucleotides to the tRNA primer. The majority of these nascent transcripts was chased into full-length (Ϫ)strand strong-stop DNA over 80 min by wild type p66/p51 RT (Fig. 5, A and C). However, for mutant enzymes, initiation products predominate over full-length (Ϫ)-strand DNA (Fig.  5C). Percentages of initiation and elongation products to total (Ϫ)-strand DNA products were calculated from phosphorimaging analysis and are presented in Table I. With the exception of the p66 R307E/K311E /p51 RT, similar total amounts of (Ϫ)-strand DNA products were catalyzed by each enzyme at each time point. Approximately 80% of (Ϫ)-strand DNA products catalyzed by p66 K249Q /p51 RT represented initiation products, whereas only 22% of tRNA Lys-3 utilized by wild type RT to initiate (Ϫ)-strand DNA synthesis was found as T ϩ 3 Ϫ T ϩ 5 initiation products after 80 min. A similar inability to extend initiation products was observed with p66 K311Q /p51, p66 K311E / p51, and p66/p51⌬13 RT. As might be expected, p66 R307E/K311E / p51 RT initiated substantially less (Ϫ)-strand DNA synthesis from tRNA Lys-3 , none of which could be extended to full-length (Ϫ)-strand strong-stop DNA.
Since very few (Ϫ)-strand DNA intermediates are detected, the results depicted in Fig. 5 suggest that altering residues FIG. 4. Schematic representation of the substrates and products of (؊)strand strong-stop synthesis. A 237nt, R/U5/PBS-containing HIV-1 RNA template, derived from the 5Ј end of the HIV-1 genome, is indicated. Template RNA was annealed via the PBS to one of three primers, namely 5Ј-end-labeled 18-nt RNA or DNA oligonucleotides or a 3Ј-end-labeled fully modified natural tRNA Lys-3 , giving rise to the RT-catalyzed (Ϫ)-strand strong-stop DNA synthesis products indicated. P]-tRNA Lys-3 -template (4 nM) and cold excess template (10 nM) was incubated for 10 min at room temperature and then fractionated on a 5% non-denaturing polyacrylamide gel. Migration positions of the tRNA-viral RNA duplex and nucleoprotein complex have been indicated. C, fractionation of the reaction products of A on a 12% denaturing polyacrylamide gel, illustrating the T ϩ 1 Ϫ T ϩ 5 initiation products. Lane notations are as in A. Products of both A and C were quantified by phosphorimaging analysis and are reported in Table I. Lys 249 , Arg 307 , and Lys 311 of p66 HIV-1 RT compromises the transition from initiation to elongation during (Ϫ)-strand DNA synthesis. However, mutations at these sites also lead to reduced and altered affinity for the tRNA primer (Figs. 2 and 3). Based on cross-linking studies of RT to the anticodon loop of tRNA Lys-3 (1) and OP-Cu in situ footprinting of RT⅐tRNA Lys-3 complexes presented here (Fig. 3), an altered interaction of HIV-1 RT with the anticodon loop of tRNA following disruption of the viral RNA⅐tRNA loop-loop complex may impair the transition from initiation to elongation during (Ϫ)-strand DNA synthesis.
(Ϫ)-Strand Strong-stop Synthesis from PBS-bound RNA and DNA Primers-Although the RT mutant p66/p51⌬13 is incapable of supporting tRNA Lys-3 -primed (Ϫ)-strand strong-stop DNA synthesis, it efficiently copies the same viral RNA template when this is substituted by RNA or DNA primers (18). Since important tRNA-viral RNA loop-loop interactions outside the PBS (10,11,15) cannot be established under the latter conditions, substitution of oligonucleotide primers removes the intermolecular structural barrier to initiation. We therefore elected to evaluate our mutants in the context of RNA-and DNA-primed (Ϫ)-strand strong stop DNA synthesis, the results of which are presented in Fig. 6.
Under these conditions, all enzymes supported (Ϫ)-strand strong-stop DNA synthesis from an 18-nt RNA primer annealed to the PBS (Fig. 6A). Surprisingly, this included p66 R307E/K311E /p51 RT, which was shown to be severely impaired in tRNA Lys-3 -dependent events (Figs. 3 and 5). Decreased synthesis of (Ϫ)-strand strong-stop DNA by this mutant reflects reduced initiation from the tRNA primer and not a general defect to processive DNA synthesis, since enhanced accumulation of prematurely terminated cDNAs was not observed. In addition, heterodimer instability observed in the tRNA gel shift analyses with the p66 R307E/K311E /p51 RT is not evident when examining RNA-dependent DNA polymerase activity. This mutant catalyzed nearly the same amount of (Ϫ) strand DNA from a DNA primer as the other mutant or wild type RT. These studies suggest that binding to or initiation from tRNA Lys-3 and not DNA polymerase activity may be associated with the instability of the p66 R307E/K311E /p51 dimer.
The data of Fig. 6A thus suggests that the double mutation R 307 E/K 311 E in p66 induces defects manifesting themselves almost exclusively in tRNA-mediated events, indicating the conformation of the reconstituted heterodimer is not globally compromised. In support of the notion that the amino acids Lys 249 , Arg 307 , and Lys 311 in p66 RT influence specifically tRNA-mediated events, we observed little difference in the RNA-dependent DNA synthesis profiles (comparing both fulllength and prematurely terminated DNAs) of wild type RT and mutants p66 K249Q /p51, p66 K311Q /p51, and p66 K311E /p51. A similar pattern emerged when (Ϫ)-strand strong-stop DNA synthesis was catalyzed by these mutant and wild type RTs from a DNA primer hybridized to the PBS (Fig. 6B). For reasons presently unclear, considerably more prematurely terminated products were evident in DNA-primed events, an observation we have also made in related studies (18,21,39,40). Collectively, the data of Fig. 6 indicate that alterations to the nucleic acid binding cleft formed by p66 and p51 do not account for the inability of mutant heterodimers to use tRNA as primer.
Evaluation of RNase H Activity-The multifunctional nature a Total amount of (Ϫ)-strand DNA products primed from tRNA Lys-3 (Fig. 5) was measured by phosphorimaging analysis. These values represented a relative measure of (Ϫ)-strand DNA product. b WT, wild type. of HIV-1 RT allows mutations influencing DNA polymerase function to be evaluated in the context of RNase H activity. The importance of monitoring both activities was recently evidenced by an RT mutant containing a point mutation at position 232 of the DNA polymerase catalytic center (Y232A) which directed RNase H cleavage almost exclusively to template nucleotide Ϫ8 rather than Ϫ17 (36). In light of this, the RNase H activities of all mutants were evaluated on a defined heteropolymeric RNA-DNA hybrid that monitors the hydrolysis products of endonuclease (DNA synthesis-dependent) and directional processing (DNA synthesis-independent) events. Substrate for this analysis is outlined in Fig. 7A and the cleavage products derived from each enzyme in Fig. 7B.
In our assay, the primary hydrolysis product of a reaction containing wild type p66/p51 RT is 71 nt, defining endonucleolytic cleavage at template nucleotide Ϫ17 (according to Metzger et al. (41), template nucleotide Ϫ1 is defined as that hybridized to the 3Ј-terminal base of the primer). With prolonged incubation, this is replaced by a 62-nt fragment, reflecting directional processing as far as template nucleotide Ϫ8. This hydrolysis pattern is qualitatively preserved in RT mutants p66 K249Q /p51, p66 K311Q /p51, and p66 K311E /p51, with only minor differences in the rates of hydrolysis. Relative amounts of the 71-and 62-nt hydrolysis products with the p66 R307E/K311E /p51 RT at 30 min (lane 4, Fig. 7B) was similar to that observed with the wild type RT at an earlier time (10 s; lane 2, Fig. 7B) indicating a decrease in rate but not a change in specificity of RNase H hydrolysis. In contrast, mutant p66/p51⌬13 completely lacks directional processing activity. Retention of RNase H function but reduction in the rate of hydrolysis supports the contention that the R 307 E/K 311 E mutation influences primarily positioning of the tRNA replication primer for extension of its 3Ј terminus at the DNA polymerase catalytic center.

DISCUSSION
Initiation of HIV-1 reverse transcription is a multi-step process, involving intricate interactions between the retroviral polymerase and its cognate tRNA primer, intertwined with the HIV-1 RNA genome at sites including and surrounding the PBS. The anticodon loop of tRNA Lys-3 appears essential to these interactions and concomitant initiation of (Ϫ)-strand DNA synthesis from the PBS (10,11,15,16,18,21,23). Independent experiments have demonstrated that this domain can be chemically cross-linked to RT in a binary nucleoprotein complex (1,23). Moreover, the anticodon loop has the capacity for intermolecular base pairing with the A-rich U5-IR loop of HIV-1 RNA to provide an additional level of control/selectivity (10,11,15,16,18,21). Although this A-rich sequence is not FIG. 7. RNase H activity of tRNA Lys-3 binding mutants. A, schematic representation of the 36-nt DNA primer-32 P 90-nt RNA template and cleavage products resulting from endonuclease and directional processing RNase H activities. RT (shaded ellipsoid) occupies an area approximately 7 nt upstream of the primer terminus (ϩ7) on the template strand to 24 nt behind the primer terminus (Ϫ24) in the primer-template duplex (49). The C-terminal RNase H domain cleaves endonucleolytically at Ϫ17 on the template to generate a 71-nt hydrolysis product. Subsequent directional processing as far as template nucleotide Ϫ8 reduces the 32 P-labeled RNA template to 62 nt. B, RNase H hydrolysis profiles for wild type and mutants enzymes. The nature of the mutation in the p66 subunit is indicated above each panel. phylogenetically conserved in non-primate lentiviruses or other retroviruses (18,42), it does appear that this element is important in HIV-1 reverse transcription and replication (43). For example, a mutant virus with a deletion of these four adenosines supported substantially less (Ϫ)-strand DNA synthesis (43). Long term cultures of this mutant HIV-1 resulted in the replacement of two or three adenosines and the restoration of near wild type levels of (Ϫ)-strand DNA synthesis and virus replication (43). The significance of the tRNA anticodon loop during (Ϫ)-strand DNA synthesis is quite evident in studies showing that the use of a tRNA His isoacceptor as a replication primer on a mutated HIV-1 RNA genome (i.e. a PBS complementary to the 3Ј end of tRNA His ) was greatly facilitated by substitutions of U5-IR loop bases for those complementary to the tRNA His anticodon loop (44).
In this study, we have combined cross-linking (1) and crystallographic data of the p66/p51 heterodimer (29,30,45) to identify p66 residues potentially involved in tRNA Lys-3 binding events, beyond the duplex formed with the PBS, which promote efficient initiation of (Ϫ)-strand synthesis. Our studies have focused on three residues conserved in a large number of HIV-1 isolates, Lys 249 , Arg 307 , and Lys 311 , which appear to line a crevice of the p66 thumb subdomain (residues 244 -322). We show here that substitutions at these positions affect the following: (a) the affinity of RT for tRNA; (b) the manner in which the tRNA primer is accommodated in a binary complex with RT; and (c) the transition from initiation to elongation of tRNAprimed (Ϫ)-strand synthesis. The same mutations have minimal impact on either oligonucleotide-primed (Ϫ)-strand synthesis or RNase H activity, suggesting their effects are localized to the manner in which tRNA Lys-3 is utilized by HIV-1 RT.
From cross-linking and biochemical studies on HIV-1 RT-tRNA Lys-3 interactions (23-26, 35), we had predicted that these mutations in the p66 component of p66/p51 RT might alter the conformation of binary RT⅐tRNA complexes. Indeed, enhanced reactivity over the entire anticodon domain was highlighted with two mutants via in situ footprinting with the chemical nuclease OP-Cu. Based on the reactivity of OP-Cu toward single-stranded RNA, including bulges and loops (46), the data of Fig. 3 suggest that much of the anticodon domain has been unwound following substitution of p66 residue Lys 311 with either Glu or Gln. This effect is clearly evident with synthetic tRNA Lys-3 and to a lesser extent with the fully modified natural species (which may be attributable to the presence of hypermodified bases in the latter) (38). In addition to altered protection of the anticodon domain, the T⌿C loop of synthetic tRNA is rendered hypersensitive when bound to p66 K311Q /p51 or p66 K311E /p51. The tRNA structure depicted in Fig. 3C indicates intramolecular pairing between bases of the T⌿C and D-loops, which also appears to be disrupted following substitution of Lys 311 . Similarities in the OP-Cu hydrolysis profiles of mutant p66 K249Q /p51 and wild type RT complexed with tRNA Lys-3 suggests this residue may not play the same role in tRNA binding, although a contribution is suggested by its 6-fold reduction in affinity for free tRNA (Fig. 2B). The inability of mutant RTs to extend PBS-bound tRNA Lys-3 into full-length (Ϫ)-strand strong-stop DNA (Fig. 5A) appears unrelated to the initiation event, since T ϩ 3 Ϫ T ϩ 5 products accumulate in all cases (Fig. 5C). Moreover, the observation that (i) the affinity of mutant enzymes for a tRNA Lys-3 -viral RNA duplex (Fig. 5B) and (ii) DNA-and RNA-directed (Ϫ)-strand synthesis (Fig. 6B) are largely unaltered, an unfavorable interaction of the tRNA anticodon domain with our RT mutants during or immediately following initiation might account for the reduced efficiency with which this primer is utilized.
A consequence of pre-annealing tRNA Lys-3 to the viral genome will be the establishment of intermolecular loop-loop interactions (10,11,15,16,18,21) prior to RT binding. Under such conditions, viral template sequences ahead of the DNA polymerase catalytic center, together with the 18-base pair PBS/tRNA duplex, will become major determinants of the strength of the preinitiation nucleoprotein complex (evidenced by the gel mobility shift data of Fig. 5B). Thus, reduced affinity of RT for free tRNA appears unlikely to account for the inability of our mutants to overcome the tRNA-directed transition from initiation to elongation. An alternative explanation for loss of tRNA-primed synthesis would be that mutant enzymes, having initiated (Ϫ)-strand synthesis, cannot disrupt the U5-IR loop-tRNA anticodon loop interaction, with the consequence of a stalled RT. Such a general defect also seems unlikely, considering the same mutants efficiently extend a PBS-bound RNA primer through several highly structured regions of the viral genome (Ref. 47 and Fig. 6A). In light of this, an explanation for our findings might lie in the position the polymerase domain occupies on the template as the tRNA primer is extended at the DNA polymerase catalytic center.
Independent lines of evidence indicate that the p66 N-terminal fingers subdomain (residues 1-84 and 120 -150) encompasses the single-stranded template overhang approximately 5-7 bases ahead of the polymerase catalytic center (48,49). In the context of the tRNA-viral RNA structure proposed by Isel and co-workers (10,11,15), extending tRNA Lys-3 by three deoxynucleotides would translocate the fingers subdomain such that it is positioned directly over the intermolecular U5-IR loop-tRNA anticodon loop duplex. Should the fingers participate in destabilizing ("unwinding") inter-and intramolecular template duplexes for presentation at the polymerase catalytic center, adding 3-5 dNTPs to tRNA Lys-3 3Ј terminus would be accompanied by disruption of this intermolecular duplex. Once the tRNA anticodon loop is released from viral RNA within the p66 fingers, it may be involved in an unfavorable interaction with an alternate subdomain of the mutated RT (as depicted in Fig. 3), thereby inducing stalling. Such an inhibition would suggest the need for a specific interaction between wild type RT and the tRNA Lys-3 anticodon loop during initiation. Alternatively, once released from the A-rich loop of HIV-1 RNA, these mutants fail to fulfill a second interaction with the anticodon domain necessary for allosteric activation of the replication complex. While speculative, preliminary data suggest that tRNA-primed synthesis by our mutants is inhibited on a viral template lacking the U5-IR loop -AAAA-sequence, which prevents intermolecular loop-loop interactions from being established and again leaves the tRNA anticodon domain available to freely interact with RT. 3 Moreover, an additional site for interaction of the tRNA anticodon loop is suggested by recent studies on the strand transfer complex (50). These hypotheses could be evaluated experimentally by determining the location of the tRNA anticodon loop in "pseudo-initiation complexes" where a chimera of tRNA containing the first 3-5 deoxynucleotides of (Ϫ)-strand DNA is hybridized to the viral template (18,21). Experiments of this nature are presently underway.