We used a computational approach to scan the entire HIV-1 genome for hybridization potential with tRNA₃^Lys 18,19. Two regions in HIV-1 stood out in this analysis, having substantial hybridization stability with tRNA₃^Lys. These are the PBS and a region in U3/R within the long terminal repeat (LTR) that encompasses the originally observed nine-nucleotide sequence, but that extends the length of the tRNA (Fig. 1a). The only interruption of the complementarity is a 29-nucleotide segment inserted in approximately the position expected for a tRNA intron (Fig. 1b). To test the statistical significance of this discovery, we generated ten thousand random sequences with nucleotide content identical to tRNA₃^Lys by shuffling the tRNA₃^Lys sequence. The PBS and U3/R sequences hybridize to tRNA₃^Lys with stabilities that are unlikely to occur by random chance. Figure 2 shows a histogram plot for the lowest folding free energy of hybridization of the shuffled sequences to the entire HIV genome and indicates the stabilities of the tRNA₃^Lys hybridization to the PBS (−80.5 kcal mol⁻¹) and to the U3/R region (−74.4 kcal mol⁻¹). It is also worth noting that the third most stable site to which tRNA₃^Lys can hybridize has a predicted stability of −62.7 kcal mol⁻¹, which is close to that expected by random chance and suggests that there are only two true hybridization sites. The full length near-complementarity between tRNA₃^Lys and the U3/R region of HIV-1 suggests that an entire tRNA gene was incorporated during evolution of the HIV-1 genome.

Since the first discovery of HIV, numerous relatives of this virus have been identified. HIV diverged into several lineages upon multiple cross-species infections involving SIV (simian immunodeficiency virus) from the nonhuman primates20,21. The HIV-2 and HIV-1 groups M, N, and O likely represent zoonotic transmission of different SIV groups22. FIV (feline immunodeficiency virus), infecting cats, is thought to be the virus from which SIV had originated23, and therefore an ancestor of HIV.

The high genomic diversity of HIV results from the error-prone nature of reverse transcriptase and from recombination events1. The LTR sequences do not encode proteins, which would tend to restrict evolution, thus evolution of this region could be fast. In order to determine whether the tRNA gene-like sequence is preserved within the HIV groups, we aligned their LTR sequences. We also analyzed the homology of long terminal repeats in SIV and FIV, to establish whether the tRNA gene-like sequence is also present in these two related immunodeficiency viruses.

The LTR alignments revealed that the tRNA gene-like sequence is well preserved among different subtypes of the HIV-1 group and also in SIV infecting chimpanzees (Fig. 3a), which is considered the most closely related immunodeficiency virus from nonhuman primates20. The LTR regions of the HIV-2 group are substantially longer than in HIV-1, but evidence for a tRNA gene-like sequence in U3/R region was apparent. Similarly, a tRNA gene-like sequence is present in the SIV that infects the monkey sooty mangabey (SIVsmm), a virus from which HIV-2 is thought to originate21 (Fig. 3b). However, LTR sequences of FIV are much shorter than those of SIV and HIV, and show no sequence evidence of the tRNA-like sequence. It is possible that the tRNA gene-like sequence was never present in the LTR of FIV, or, because FIV is thought to have infected cats for millions of years, it could have been eliminated during evolution.

Since the SIV viruses encompass appreciable diversity, we also performed the analysis of SIV genomes representing different species, including the recently discovered ancient lemur lentivirus pSIVgml identified at the base of the phylogenetic relationship tree of the primate lentiviruses24. For comparison we also analyzed sequences of HIV-1 and HIV-2. The results from scanning of their genomes against tRNA₃^Lys (P-values of homology) are summarized in Table 1, and indicate that significant homology to tRNA₃^Lys can be found in seven of twenty analyzed SIV species including SIVcpz tan, SIVmnd-2, SIVdrl and SIVlhoest with the lowest P-values of homology. These SIV species do not group together in the SIV family tree. Overall, acquisition of the tRNA gene-like sequence is evident in viruses from the SIV family, but it is not uniformly preserved, suggesting that biological roles of the sequence vary in different viruses. Moreover, while the tRNA-like sequence is evident in both HIV-1 and HIV-2 from alignment, its preservation is not uniform.

Our prediction of hybridization showed three major segments of interaction with the tRNA gene-like sequence, any or all of which could form biologically relevant interactions with the tRNA₃^Lys primer. Two segments are located in the U3 region; one is complementary to the 3′ region of tRNA₃^Lys, and is referred to here as segment 1 (S-1), and the other is the nine-nucleotide segment (9nt) previously analyzed8,9 (Fig. 4a). The third segment (S-3) is within the transacting responsive (TAR) hairpin of the R element, present at both ends of the HIV RNA genome, and already a fundamental part of the invasion mechanism of minus strand transfer25. Our initial experiments evaluated whether segment 1 can promote minus strand transfer. To analyze the role of tRNA/U3 interactions in driving minus strand transfer we previously developed a reconstituted system in vitro9. It consists of a primer (DNA oligonucleotide or human tRNA₃^Lys) and two RNA templates with sequences of the 5′ end (donor) and 3′ end (acceptor) of the HIV-1 RNA genome. The donor RNA template spans 199 nucleotides from the genome 5′ end including the R element, U5 region and PBS. The newly synthesized cDNA extended from the primer by purified viral RT can transfer to the acceptor RNA template that also has an R sequence, but with a 5′ extension with the sequence of the natural U3 region. Because the donor extension product is shorter than the product of transfer and subsequent extension on the acceptor, these two products are readily distinguishable by gel mobility.

Our previous analysis showed that an acceptor A97(−54) representing sequence between 9022-9172 of HIV-1, which includes the R element and a fragment of U3 including the nine-nucleotide motif (9033-9041), exhibited highly efficient minus strand transfer, when tRNA₃^Lys was used as a primer9. Transfer efficiency reached 80%. This high efficiency was proposed to occur through cooperation of an invasion-driven mechanism, occurring within the R elements, and a proximity effect achieved by interactions of tRNA₃^Lys with the nine complementary nucleotides in the U3 region. To determine whether segment 1 in U3 can also influence transfer, we extended the acceptor at its 5′ end to 76 nucleotides beyond the R element to include this sequence. In addition, we performed the transfer reactions for this study using donor-acceptor template combinations sufficiently short at the 3′ ends that they could not utilize the invasion mechanism. The designed acceptor, A19(−76), contained only the part of the R element not involved in the invasion-driven mechanism, but with 19nt of homology to the 5′ end of the donor template (Fig. 4b). Direct terminus transfer was the only allowed mode1. Enhancement of transfer could then only occur by a proximity effect mechanism, presumably mediated by sequences complementary to the tRNA₃^Lys in the U3 region. The advantage of designing the system in this manner is that the basal level of transfer was low. This increases the sensitivity with which enhancement by proximity effects could be measured.

In initial experiments we used the acceptor A19(−54) as a control. This acceptor encompasses the nine-nucleotide transfer-enhancing motif in U3 and allows for terminal transfer with efficiency of about 20%, similar to that previously reported9 (Figs. 4b and 5). Our mutation or deletion of the complementary nine nucleotides in this acceptor reduced the transfer efficiency to below 10% (Fig. 5). A similar low level of transfer was also observed for acceptor A19(−20)9, which does not have any complementarity to tRNA₃^Lys (data not shown). However, the presence of additional sequences 5′ of the nine-nucleotide motif in acceptor A19(−76) caused a significant increase in terminal transfer efficiency, namely up to 70%. Such high efficiency of transfer for systems in vitro was previously reported only in reactions with substrates capable of utilizing the invasion mechanism9. Mutation or deletion of the nine-nucleotide segment in acceptor A19(−76) caused a drop of transfer efficiency, but only to about 45% (A19(−76)m1; Fig. 5) and 56% (A19(−76)△; Fig. 5), respectively, indicating that segment 1 alone can significantly contribute to efficient minus strand DNA transfer.

In order to confirm that segment 1 indeed stimulates the minus strand DNA transfer reaction, we designed an acceptor with a mutation that eliminated potential tRNA interaction with segment 1, but allowed interaction with the nine-nucleotide segment (A19(−76)m3; Fig. 6). We also designed the acceptor with mutations in both segments that did not allow for any interactions with tRNA₃^Lys (A19(−76)m2). The results are displayed in Figure 6, and confirm that segment 1 contributes significantly to minus strand transfer efficiency. Mutations only in this segment reduced the stimulation to about 25%, a level similar to that measured with the shorter acceptor, A19(−54), having only the nine-nucleotide motif. Few transfer products were observed when the acceptor was long but had mutations in both segments. The transfer efficiency in that case was about 10%, comparable to that of acceptor A19(−54) with a mutated or deleted nine-nucleotide sequence.

As an additional control we performed reactions with the same acceptors, but primed with a DNA oligonucleotide complementary to primer binding site. This system does not allow interaction of the U3 sequences in the acceptor RNA with the donor-primer. For all acceptors the transfer products were barely detected, instead intense bands reported fold-back products. These products were present at a level similar to those in the control reaction performed without acceptor RNA, which, of course, precludes transfer (Fig. 6). The fold-back products represent the donor extension products folded back at their 3′ ends, allowable because their sequence is the complement of the TAR hairpin, and then further extended by copying the donor cDNA template to its 5′end. Fold-back products, prevalent when transfer is inefficient, are a dead-end alternative to transfer, that the natural transfer system has evolved to suppress1,26.

Taken together, these results demonstrate that segment 1 can promote efficient minus strand transfer. Interestingly, segment 1 corresponds to the sequence in tRNA₃^Lys known to be involved in interaction with the PBS region. This suggests that the 3′ end of tRNA₃^Lys used as a primer to start reverse transcription is already released from interactions with the PBS by the time it interacts with segment 1 to facilitate transfer.

We purchased purified human tRNA₃^Lys from Bio S&T (Lachine, QC) and DNA oligonucleotides from Integrated DNA Technologies, Inc. (Coralville, IA). We obtained pNL4-3 molecular clone (from Dr. Malcolm Martin) through the AIDS Research and Reference Reagent Program (Division of AIDS, NIAID, National Institutes of Health). Dr. Robert J. Gorelick generously provided HIV-1 NC (71 amino acids). We purified HIV-1 reverse transcriptase (p66/p51 heterodimer) (RT) as described previously55. We purchased the 32P labeled substrates from Perkin-Elmer Life Sciences.

We performed bimolecular RNA secondary structure prediction using the Bimol program from RNAstructure19. This program predicts base pairs between interacting RNA sequences without allowing unimolecular base pairs and predicts the lowest free energy hybridization site and suboptimal sites of hybridization18. We calculated P-values to determine significance of homology using a previously described method56.

We took the consensus sequences for each HIV and SIV group from the HIV Database (http://www.hiv.lanl.gov/). In order to perform alignments, we used a multiple sequence alignment program T-coffee57 available at the EMBL-EBI website.

In order to prepare the donor and acceptor RNA templates, we first amplified the genomic sequences of HIV-1 (pNL4-3) by PCR using Vent DNA polymerase (BioLabs). We generated the DNA templates for transcription in vitro with the following sets of oligonucleotides: oligomers 1 and 2 (see below) for the donor template corresponding to the first 199nt at 5′end of the HIV-1 RNA genome; oligomers 3 and 4 for acceptor template A19(−54) corresponding to region 9022-9094 in the HIV-1 RNA genome; oligomers 5 and 4 for A19(−54)m with mutations in the 9nt motif; oligomers 6 and 4 for A19(−54)△ with deletions of the 9nt motif; oligomers 7 and 4 for A19(−76) corresponding to region 9000-9094 in the HIV-1 RNA genome. We used the set of two overlapping oligonucleotides to generate DNA templates for the following acceptors: oligomers 8 and 4 for A19(−20) corresponding to region 9055-9097 in the HIV-1 RNA genome; oligomers 9 and 10 for A19(−76)m1 with mutation in the 9nt segment; oligomers 11 and 10 for A19(−76)△ with deletion of the 9nt segment; oligomers 12 and 13 for A19(−76)m2 with mutation of segment 1; oligomers 14 and 13 for A19(−76)m3 with mutations in segment 1 and the 9nt motif. The forward oligonucleotides (1, 3, 5, 6, 7, 8, 9, 11, 12, 14) for all DNA templates contain the sequence of T7 RNA promoter. Using the Ambion T7-MEGAshortscript kit (Applied Biosystems) we transcribed in vitro the RNA templates, then purified by polyacrylamide/urea gel electrophoresis and resuspended in 10 mM Tris-HCl (pH 8.0), 1 mM EDTA buffer. RNAs were quantitated by UV absorption using a GeneQuant II from Amersham Biosciences.

1. 5′-GGA TCC TAA TAC GAC TCA CTA TAG GGT CTC TCT GGT TAG ACC AGA TC-3′; 2. 5′-GTC CCT GTT CGG GCG CCA CTG CTA GAG A-3′; 3. 5′-GGA TCC TAA TAC GAC TCA CTA TAG GGA GTG GCG AGC CCT CAG ATG-3′; 4. 5′-GGT CTA ACC AGA GAG ACC CAG TAC AGG CAA AAA GCA G-3′; 5. 5′-TAA TAC GAC TCA CTA TAG GGA GTG GCG AGT ATA TAT ATG CTG CAT ATA AGC AGC TGC TTT TTG CC-3′; 6. 5′-GGA TCC TAA TAC GAC TCA CTA TAG GGA GTG GCG AGG CTG CAT ATA AGC AGC TGC TTT TTG-3′; 7. 5′-GGA TCC TAA TAC GAC TCA CTA TAG GGC GTG GCC TGG GCG GGA CTG GGG AG-3′; 8. 5′-GGA TCC TAA TAC GAC TCA CTA TAG GGC TGC TTT TTG CCT GTA CTG-3; 9. 5′-GGA TCC TAA TAC GAC TCA CTA TAG GGC GTG GCC TGG GCG GGA CTG GGG AGT GGC GAG CTC CTC GAT GCT GCA TAT AAG CAG CTG CTT TTT GC-3′; 10. 5′-GGT CTA ACC AGA GAG ACC CAG TAC AGG CAA AAA GCA GCT GCT TAT ATG CAG C-3′; 11. 5′-GGA TCC TAA TAC GAC TCA CTA TAG GGC GTG GCC TGG GCG GGA CTG GGG AGT GGC GAG GCT GCA TAT AAG CAG CTG CTT TTT GC-3′; 12. 5′-GGA TCC TAA TAC GAC TCA CTA TAG GGC GTG GCC AAC CAG GCC AAG GGG AGT GGC GAG CCC TCA GAT GCT GCA TAT AAG CAG CTG CTT TTT-3′; 13. 5′-GGT CTA ACC AGA GAG ACC CAG TAC AGG CAA AAA GCA GCT GCT TAT ATG CAG CAT C-3′; 14. 5′-GGA TCC TAA TAC GAC TCA CTA TAG GGC GTG GCC AAC CAG GCC AAG GGG AGT GGC GAG CTC CTC GAT GCT GCA TAT AAG CAG CTG CTT TTT-3′ ; 15. 5′-TGG CGC CCG AAC AGG GAC TTG AAC CCT GGA CCC-3′; 16. 5′-GTC TGA TGC TCT ACC GAC TGA GCT ATC CGG G-3′

We prepared 5′-radiolabeled tRNA₃^Lys as described previously with slight modifications9. We heat annealed a 1.5-fold excess of two DNA oligonucleotides 15 and 16 to tRNA₃^Lys at 65°C for 5 min in buffer 50 mM Tris-HCl (pH 8.0), 75 mM KCl, and gradually cooled to 37°C. The purpose of this annealing process was to prevent the tRNA from folding into a configuration that would inhibit labeling. Next, we treated the reaction mixture with shrimp alkaline phosphatase (SAP, Fermentas) at 37°C for 60 min., and then incubated at 65°C for 25 min to inactivate the enzyme. Following slow cooling to 37°C, we treated the reaction mixture with [γ-32P]ATP (6000 (222TBq) Ci mmol⁻¹). We separated the radiolabeled tRNA₃^Lys and DNA oligomers by electrophoresis in a 6% polyacrylamide/urea gel, and then eluted the tRNA₃^Lys overnight from the gel, precipitated with ethanol, and resuspended in 10 mM Tris-HCl (pH 8.0), 1 mM EDTA buffer. We labeled the DNA primer with sequence complementary to the PBS region (5′-GTC CCT GTT CGG GCG CCA-3′) at the 5′-end using [γ-32P]ATP (6000 (222 TBq) Ci mmol⁻¹) and T4 polynucleotide kinase (Roche) and separated it from unincorporated radionucleotides using a Micro Bio-Spin column (Bio-Rad).

We heat-annealed the tRNA₃^Lys primer or DNA primer to donor RNA by incubation at 95°C for 5 min and slow cooling to 37°C. Then, we added acceptor templates and incubated with 200% NC (100% NC is 7 nt per NC molecule) for 3 min. Next we added RT to the substrate and after 4 min of incubation at 37°C, we initiated reactions with MgCl₂ and dNTPs. We mixed primer, donor, and acceptor at a ratio of 2:1:2. Final reaction mixtures contained 50 mM Tris-HCl (pH 8.0), 50 mM KCl, 1 mM DTT, 1 mM EDTA, 32 nM HIV-1 RT, 6 mM MgCl₂, and 50 μM dNTPs. We incubated reactions at 37°C, and terminated them at appropriate times with 1 volume of termination dye (10 mM EDTA, pH 8.0, 90% formamide (v/v), and 0.1% (w/v) each xylene cyanole and bromophenol blue). We then resolved products by polyacrylamide-urea gels and analyzed using a PhosphorImager (GE Healthcare) and ImageQuant (version 2.1). We estimated sizes of DNA products by using a 5′-radiolabeled 10 bp DNA ladder (Invitrogen).