Mechanism of minus strand strong stop transfer in HIV-1 reverse transcription.

Retrovirus minus strand strong stop transfer (minus strand transfer) requires reverse transcriptase-associated RNase H, R sequence homology, and viral nucleocapsid protein. The minus strand transfer mechanism in human immunodeficiency virus-1 was examined in vitro with purified protein and substrates. Blocking donor RNA 5'-end cleavage inhibited transfers when template homology was 19 nucleotides (nt) or less. Cleavage of the donor 5'-end occurred prior to formation of transfer products. This suggests that when template homology is short, transfer occurs through a primer terminus switch-initiated mechanism, which requires cleavage of the donor 5' terminus. On templates with 26-nt and longer homology, transfer occurred before cleavage of the donor 5' terminus. Transfer was unaffected when donor 5'-end cleavages were blocked but was reduced when internal cleavages within the donor were restricted. Based on the overall data, we conclude that in human immunodeficiency virus-1, which contains a 97-nt R sequence, minus strand transfer occurs through an acceptor invasion-initiated mechanism. Transfer is initiated at internal regions of the homologous R sequence without requiring cleavage at the donor 5'-end. The acceptor invades at gaps created by reverse transcriptase-RNase H in the donor-cDNA hybrid. The fragmented donor is eventually strand-displaced by the acceptor, completing the transfer.

minus strand synthesis, RT simultaneously cleaves the copied RNA using its RNase H activity. This frees the ϪsssDNA, enabling its transfer to the homologous R sequence in the 3Ј-untranslated region and continuation of minus strand synthesis. This step is called minus strand strong stop transfer (minus strand transfer), and it is obligatory for reverse transcription and viral replication. Extensive studies have identified the RT-RNase H activity, R sequence homology, and viral nucleocapsid (NC) protein as important components for minus strand transfer (see Ref. 2
HIV-1 RT has been shown to partially cleave the template RNA during polymerization, leaving RNA fragments annealed to the cDNA (9,10). Since there are 50 -100 RTs and only two copies of the RNA genome in a viral particle (11), RTs not involved in polymerization are available for cleaving these fragments. The RNase H acts independently of polymerization in these reactions, serving to clear off the partially degraded RNA fragments left after the partial RNA removal associated with polymerization. Failure to detect significant amounts of ϪsssDNA in the infected cell cytoplasm indicates that minus strand transfer is an efficient event in vivo (12,13).
Minus strand transfer utilizes the homologous R sequences at both ends of the viral genome. In HIV-1, the R region is 97 nt long and contains all of the TAR hairpin and part of the poly(A) hairpin. Studies with mutations within the R segment indicate that transfers predominantly take place after the DNA synthesis reaches the 5Ј-end of the genome (14 -19). Deletion and mutational analysis of the R region shows that not all the R region is required for transfer in HIV-1. HIV-1 can still effectively complete replication with an R homology much shorter than 97 nt (20). In murine leukemia virus, an R region with as little as 12 or 14 nt does not affect transfer significantly (21,22). In addition to homology, RNA structure has also been implicated in the minus strand transfer mechanism (17,23,24).
The viral NC protein has been shown to stimulate minus strand transfer (see Ref. 1

and references therein and Ref. 2).
The RNA genome in retroviruses is coated with the NC protein, a proteolytic product of the gag polyprotein. It is estimated that one molecule of NC protein coats about 7 nt of genomic RNA in vivo (25,26). NC is a highly basic protein, containing one or two zinc fingers, which are essential for its function (see Refs. 27-29; see Ref. 30 and references therein). NC has been shown to function in several steps of viral replication, including genomic RNA maturation, reverse transcription, and integration. NC displays nucleic acid chaperone activity, catalyzing the melting, reannealing, and folding of nucleic acid into their most stable conformations (31)(32)(33)(34)(35)(36)(37)(38)(39)(40)(41)(42)(43)(44). All of these have important implications in strand transfer and reverse transcription. NC protein noticeably stimulates transfer in reconstituted in vitro systems. The strand annealing and chaperone activities of NC plus its effects on RT-RNase H have been implicated in facilitating the transfer reaction (34,35,39,(43)(44)(45)(46)(47)(48). Analyses in vitro have been used to show that NC inhibits the formation of dead end, self-primed products from the ϪsssDNA, allowing it to be used for transfers (38,49,50). In addition, NC has been shown to stimulate RT-RNase H activity, especially on a bluntended substrate (33,43). Although direct interaction between NC and RT has been reported (51,52), the nature of such interactions is unclear.
Two possible mechanisms can be envisioned for minus strand transfer (Fig. 1). In the first, transfer primarily involves switching of the 3Ј terminus of the fully extended strong stop DNA from the donor to the acceptor. Here, the degradation of the genomic RNA, specifically at the 5Ј-end, is required to release the primer terminus such that it is free to anneal to the acceptor. This is therefore referred to as the primer terminus switch-initiated pathway. The second, referred to as acceptor invasion-initiated pathway, proposes that RNase H cleavages within the internal regions of the donor RNA are critical in facilitating the transfer. The homologous acceptor RNA anneals to the cDNA at gaps created within the donor RNA template and initiates the transfer well behind the primer terminus. The acceptor template strand displaces the fragmented donor and captures the primer terminus, thereby completing the transfer.
Using a reconstituted HIV-1 R sequence-based transfer system, we set out to test the primary pathway for minus strand transfer in HIV-1. Templates with longer homology supported more transfers, and NC protein further enhanced the efficiency. Inhibiting internal RNase H cleavages within the donor reduced transfer with long homology. However, inhibiting cleavages at the donor 5Ј-end specifically inhibited transfers only for short (and not for long) homology template systems. In long homology templates, in the presence of NC, transfer products were formed before degradation of the donor 5Ј-end. The overall data support that minus strand transfer in retroviruses with long R sequences, such as HIV-1, occurs through an acceptor invasion-initiated mechanism that does not require terminal cleavages at the 5Ј-end of R. Whereas DNA synthesis is completed to the 5Ј-end of the R sequence, the transfer is initiated well behind the primer terminus and is promoted through internal RNase H cleavages with the donor template and the strand exchange and chaperone activities of NC. Plasmid Constructs-Genomic sequences from the HIV-1 NL4 -3 strain were amplified by PCR and cloned into LITMUS 28 to create the donor and the acceptors constructs for generation of RNA templates. The following PCR primers were used in the construction of the various constructs. To generate the pNL-71D donor construct, primers SpeI/ 1(ϩ)DpNL4 -3 (5Ј-GGACTAGTGGGTCTCTCTGGTTAGACCAGA) and XhoI/71(Ϫ)DpNL4 -3 (5Ј-CTATCTCGAGGGCTTAAGCAGTGGGTTC-CC) were used to amplify a 71-nt segment (ϩ1 to ϩ71) from the 5Ј-end of the genome. Restriction enzyme sites within the primers are underlined. PCR fragments were digested with SpeI and XhoI and cloned into LITMUS 28. Acceptor constructs were generated using the same approach. Primers SpeI/9056(ϩ)ApNL4 -3 (5Ј-CGACTAGTGCTGCTT-TTTGCCTGTACTGGG) and XhoI/9119(Ϫ)LApNL4-3 (5Ј-TATACTCG-AGGCCAGAGAGCTCCCAGGCTC) and primers SpeI/9056(ϩ)A-pNL4-3 and XhoI/9100(Ϫ)SApNL4 -3 (5Ј-GGCACTCGAGCAGATCTG-GTCTAACCAGAG) were used to amplify 64 nt (ϩ9056 to ϩ9119) and 45 nt (ϩ9056 to ϩ9100) segments from the 3Ј-end of the genome to construct the pNL-A45h and pNL-A26h acceptor constructs, respectively. All constructs were transformed into E. coli DH5␣ cells and sequenced.

Materials-Purified
Generation of RNA Templates-Donor RNA template D71 and acceptor RNA templates A26h and A45h were generated by run-off transcription in vitro from the corresponding linearized plasmid constructs using the Ambion T7-MEGAshortscript kit (Austin, TX) as per the manufacturer's protocol. The acceptor RNAs A19h and A12h were transcribed in vitro from a synthetic double-stranded DNA fragment containing the T7 promoter. The sequences of the templates were 5Ј-GGATCCTAATAC-GACTCACTATAGGGCTGCTTTTTGCCTGTACTGGGTCTCTCTGGT-TAGACC (for A19h) and 5Ј-GGATCCTAATACGACTCACTATAGGGC-TGCTTTTTGCCTGTACTGGGTCTCTCTGG (for A12h) and their complementary strands. The 5Ј-end 2Ј-O-methyl-modified 71-nt donor template, D71(OM4 -14), was generated by ligating a 31-nt 5Ј-oligomer containing the 2Ј-O-methyl modification to a 40-nt 3Ј-oligomer, using T4 DNA ligase as described previously (54). The 69-nt donor template with internal 2Ј-O-methyl modification, D69(OM17-40), was purchased from Dharmacon Research Inc. All RNAs used in the study were PAG-E-purified and resuspended in 10 mM Tris-HCl (pH 8.0), 1 mM EDTA buffer. RNAs were quantitated using the Ribogreen assay (Molecular Probes, Inc., Eugene, OR).
Synthesis Assay-Reactions were performed as described previously (56) with slight modifications. Final reactions contained 2.5 nM primertemplate, in 50 mM Tris-HCl (pH 8.0), 50 mM KCl, 1 mM dithiothreitol, 1 mM EDTA, and 32 nM HIV-1 RT. 3.75 nM (7.5 nM for time course assays and reactions with D69(OM17-40) donor) acceptor RNA was included in the transfer reactions. For reactions with NC protein, NC was added to the mix and incubated at 37°C for 5 min. RT was then allowed to prebind to substrate before initiating reactions with MgCl 2 and dNTPs at a final concentration of 6 mM and 50 M respectively. Reactions were incubated at 37°C and terminated at the appropriate time using 2ϫ termination dye (10 mM EDTA (pH 8.0), 90% formamide (v/v), and 0.1% each of xylene cyanol and bromphenol blue).
Synthesis Assay with the Internally Modified D69(OM17-40) Donor RNA-Reactions were performed as described above, with slight modifications. The ratio of primer/donor/acceptor was 1. RNase H Assay-Reactions were performed as previously described (43). Briefly, reactions contained 50 mM Tris-HCl (pH 8.0), 50 mM KCl, 1 mM dithiothreitol, 1 mM EDTA, 4 nM substrate, and 32 nM HIV-1 RT. The reaction mixture was incubated for 2 min at 37°C to allow prebinding of RT to substrate. Reactions were then started by the addition of the MgCl 2 at 6 mM final concentration and incubated at 37°C for the appropriate time.

RESULTS
Results from our laboratory and others have shown that during ϪsssDNA synthesis, cleavage of the genomic RNA at the 5Ј-end is relatively inefficient, leaving a 14 -15-nt fragment annealed to the 3Ј-end of the extended DNA (43, 57, 58). However, sequence analysis of the transfer product suggests that most RT copies all of the 5Ј R sequence before the transfer occurs (15)(16)(17)(18)(19). These observations led us to hypothesize two possible transfer processes shown in Fig. 1. The primer terminus switch-initiated pathway predicts that the transfer is initiated through annealing of the 3Ј terminal region (10 -20 nt) of the fully extended ϪsssDNA to the acceptor template. In this mechanism, degradation of the 5Ј-end genomic RNA fragments is important to allow the critical annealing step. The acceptor invasion-initiated pathway, on the other hand, is unaffected by the presence of the uncleaved 5Ј-end RNA fragments. Instead, transfers would depend on initiation of DNA-acceptor template interaction at a site internal to the primer terminus and a propagation of that interaction to the 3Ј terminus of the strong stop DNA. RNA in the path of the propagation, including the terminal segment, would be presumably displaced. An efficient invasion mechanism would therefore require an extended homology between the donor and acceptor templates.
Cleavage at the 5Ј-End of the HIV-1 Genome after ϪsssDNA Synthesis Is Poor and Is Facilitated by NC Protein-Characterization of RT-RNase H cleavage shows that on RNA fragments annealed to a DNA template, RT preferentially binds the RNA 5Ј-end, making a cut 18 -19 nt from the RNA 5Ј-end (59 -63). This initial fragment is further processed into 8 -10nt-long smaller fragments, which can dissociate from the template. We refer to the 18 -19-nt and 8 -10-nt cuts as the primary and the secondary cuts, respectively, based on their rate of formation (55). On blunt-ended hybrid substrates, where the RNA 5Ј-end is flush with the DNA 3Ј-end, secondary cleavages are less efficient and stimulated by NC protein (43). This was true for substrates comprising nonviral or HIV-1 R sequences (43).
In the current study, we analyzed the role of the RNase H cleavages during the course of minus strand DNA synthesis and transfer. A 71-nt RNA template (D71) corresponding to the 5Ј-end of the HIV-1 genome was used as the donor template, and minus strand DNA synthesis was initiated using a DNA primer (Fig. 2, A and B). Synthesis to the end of the donor RNA would create a blunt-ended hybrid substrate. Overall RNA degradation of the template (overall RNase H) and formation of the 5Ј-end 8 -10-nt cleavage product (secondary cut) were examined in the presence and absence of NC (Fig. 2, A and B). For comparison, RNase H cleavages were also examined on preformed blunt-ended and recessed substrates (Fig. 2, C and D). For the preformed substrates, D71 RNA was annealed to appropriate cDNAs to generate blunt-ended and recessed substrates. In the absence of NC, both the overall RNase H and secondary cut at the donor 5Ј-end were lower when examined during synthesis than with the preformed blunt-ended hybrid substrate (Fig. 2, compare A and C with B and D). This is not surprising, considering that in the synthesis reactions, the substrate for RNase H cleavage had to be created during the reaction. The presence of NC (at 200% coating; i.e. 1 NC/3.5 nt) during minus strand DNA synthesis increased the overall RNase H cleavage by 50% ( Fig. 2A) and the 8 -10-nt cut at the RNA 5Ј-end by 100% (Fig. 2B). With the preformed substrates in the absence of NC, the 8 -10-nt terminal cut was slower for the blunt-ended substrate as compared with the recessed substrate (Fig. 2D). This result was consistent with those observed previously with the shorter 28-and 41-nt viral and nonviral substrates (43). In comparison, NC stimulation of RNase H, in particular the 8 -10-nt cut, was more significant on preformed blunt-ended substrate as compared with those during synthesis (Fig. 2, compare B and D). At 200% NC coating, the 8 -10-nt cut efficiency on the blunt-ended substrate was enhanced by 4 -5-fold, making it comparable with that on the recessed substrate (Fig. 2D). NC did not alter the specificity of RNase H cuts on any of the substrates tested (data not shown). This suggests that NC not only stimulates RT-RNase H during synthesis but also specifically stimulates the 8 -10-nt cut at the blunt end of an RNA/DNA hybrid.
Effect of Homology on Transfer-To determine the most likely pathway for minus strand transfer in HIV-1, we designed a transfer system in vitro using the HIV-1 R sequences (Fig.  3A). The 71-nt RNA derived from the 5Ј-end of the HIV-1 genome was used as the donor template (D71), whereas the acceptor templates were derived from the 3Ј R region. All four acceptor templates (A12h, A19h, A26h, and A45h) shared the same 5Ј-end but differed at the 3Ј-end, such that they generated 12, 19, 26, and 45 nucleotides of homology, respectively, with the 5Ј-end of the donor template (Fig. 3A). This ensured that the transfer product remained the same size (90 nt) irrespective of the acceptor used.
Primer extension on the donor template resulted in a 71-nt full-length donor extension product (Fig. 3B, lane 1). In addition, an 85-nt product was also detected, resulting from selfpriming of the full-length extension product. A ladder of pause sites was observed from position 47 through 53. Analysis of the predicted secondary structure of this region suggested that the pausing ladder was due to displacement synthesis through an alternate hairpin structure formed from refolding of the template, as shown previously (50).
We next examined transfers, using the four different acceptor RNA templates (Fig. 3B). An expected 90-nt transfer product was detected in all of the reactions with acceptor template. Similar synthesis profiles were observed with the donor extension and transfer reactions, with respect to pausing sites, fulllength donor extension, and self-priming products. Quantitative values for the transfer efficiencies are shown in Fig. 3C. In the absence of NC, transfer products were almost undetectable, with the acceptor having the shortest homology, A12h ( RT copies to the 5Ј-end of 5Ј R (donor) before switching to 3Ј R (acceptor). The primer terminus switch-initiated pathway proposes that the donor RNA 5Ј-end needs to be degraded to release the primer terminus such that it is free to anneal to the acceptor and initiate transfer. The acceptor invasion-initiated pathway proposes that the homologous acceptor RNA anneals to the cDNA at gaps created within the donor RNA template and initiates the transfer well behind the primer terminus. The acceptor template strand displaces the fragmented donor and captures the primer terminus, thereby completing the transfer. Patterned lines represent donor RNA, gray lines represent acceptor RNA, and black lines represent DNA. The arrows in the DNA strands indicate the direction of synthesis. homology to 19 nt (Fig. 3B, lane 11) resulted in a 1.9% transfer efficiency (Fig. 3C), whereas both 26-nt (Fig. 3B, lane 16) and 45-nt (Fig. 3B, lane 21) homologies yielded transfer efficiencies of 3.6 and 4.2%, respectively (Fig. 3C). Data indicate that within a certain range, increasing the length of homology between donor and acceptor increased strand transfer efficiency.
NC protein has been shown to stimulate strand transfer and inhibit self-priming (3, 28, 33, 37-40, 47-50, 58, 64 -76). We therefore examined the effect of NC on transfers as template homology was increased. The amount of NC was titrated to achieve from 100 to 1200% coating of the substrates. The presence of NC did not alter the overall extension profile in the absence or presence of acceptor (Fig. 3B). 100% NC coating had no obvious effect on transfer efficiency and self-priming for the four substrates (data not shown). As NC concentration was increased to 200% coating, a noticeable increase in transfer product together with a slight decrease in self-priming product was generally observed (Fig. 3B, lanes 7, 12, 17, and 22). Subsequent increases in NC concentration further enhanced transfer efficiency (Fig. 3, B and C). With the 12-nt homology substrate, transfer efficiency increased 22-fold with 400 and 800% NC coating but decreased as NC was further increased to 1200% coating. For longer homologies of 19, 26, and 45 nt, transfer efficiencies increased up to 39, 65, and 46%, respectively, with 1200% NC coating (Fig. 3B (lanes 15, 20, and 25) and 3C). The slight decrease of transfer efficiency from 26-to 45-nt homology will be discussed later. This set of reactions was also performed using the 55-amino acid form of NC protein, and results with the same trend were obtained (data not shown). Overall, NC was found to increase transfer efficiency, while slightly decreasing self-priming, in a concentration-dependent manner.
Blocking Donor RNA 5Ј-End Cleavage Inhibits Transfer with Short but Not Long Homology-Although the enhancement in transfers observed with increased donor-acceptor homology would support the invasion-initiated pathway as the primary transfer mechanism, the approach does not exclude transfers initiated via the primer terminus switch. Therefore, to assess the contribution of the two pathways to the transfer mechanism, we attempted to specifically block transfers promoted via the primer terminus switch-initiated pathway. Since primer terminus transfer would require degrading the donor 5Ј terminus, inhibiting cleavages in this region should block transfer facilitated by this mode. Utilizing this idea, we created the 71-nt donor RNA template, D71(OM4 -14), containing 2Ј-Omethyl modifications from position 4 to 14 measured from the 5Ј-end of the RNA, while retaining the original sequence (Fig.  4). To test the susceptibility of this RNA to RNase H cleavage, the D71(OM4 -14) RNA was 5Ј-end-labeled and annealed to the 71-nt cDNA, to create the blunt-ended substrate and then subjected to the standard cleavage assay. As expected, the 2Ј-O-methyl modification inhibited RNase H cleavages from position 1 to 16 at the 5Ј-end of the donor RNA, in contrast to the unmodified substrate, which sustained normal cleavages (Fig. 4).
We then performed strand transfer assays using the D71(OM4 -14) RNA as the donor template in combination with the four different acceptors (Fig. 5). The 2Ј-O-methyl modification did not interfere with synthesis significantly in this substrate, and similar extension profiles were observed for both the modified and unmodified substrates (Data not shown). Interestingly, the 2Ј-O-methyl modification greatly reduced the self-priming product in both the donor extension and strand transfer reactions (data not shown). This is not surprising, since the uncleaved 16-nt RNA fragment would remain annealed to the 3Ј-end of cDNA and thereby inhibit it from folding back and self-priming, as shown previously by the Hughes group (50,58,77).
For the short 12-nt homology, inhibiting donor 5Ј-end cleavage completely blocked transfers even at high NC concentration (Fig. 5A, lanes 6 -10). With the 19-nt homology, transfers were detected only at high NC (800 and 1200% coating), but transfer efficiency was generally lower than that observed with the unmodified substrate (Fig. 5, A (lanes 14 and 15) and B,  3, 8, 13, 18, and 23), 800% (lanes 4, 9, 14, 19, and 24), and 1200% (lanes 5, 10, 15, 20, and 25) NC protein. L, the 10-bp DNA ladder. Transfer product (T), self-priming product (SF), full-length donor extension product (FL), and primer (P) are indicated. C, quantitation of transfer efficiencies from assays with the D71 donors. Quantitation was based on at least three independent experiments. Transfer efficiency was defined as 100 ϫ transfer product/(fulllength product ϩ self-priming product ϩ transfer product). compared with Fig. 3C). This suggests the primer terminus switch as the predominant pathway. For the 26-nt homology substrate (Fig. 5A, lanes 16 -20), transfer products were detected beginning at 200% NC coating, at about 50% of that observed with the unmodified substrate. At higher NC concentration, transfer efficiencies were similar to those observed with the unmodified templates (compare Fig. 5B with Fig. 3C). When the homology was further increased to 45 nt, the modified and unmodified substrates showed very similar transfer efficiencies (Fig. 5, A (lanes 21-25) and B; compare with Fig.  3C), suggesting that the primer terminus switch-initiated pathway contributed minimally to transfers in this system. Instead, transfers take place mainly via the acceptor invasioninitiated pathway.
Taken together, the results indicate that when the homology between the donor and acceptor is short (19 nt or less), transfers take place mainly through the primer terminus switchinitiated pathway. When the homology is longer, such as 45 nt, acceptor invasion-initiated pathway is the major pathway for minus transfer.
Blocking Donor Internal Cleavages Inhibits Transfer with Long but Not Short Homology Substrates-The above results suggest that in long homology substrates, cleavages well within the donor RNA create invasion sites for the acceptor RNA. Blocking internal cleavages in the donor RNA would then suppress transfers that are promoted through an invasion-initiated mechanism but not a primer terminus switch-initiated mechanism. To test this, we generated a 69-nt donor RNA from the 5Ј-end of the HIV-1 genome, D69(OM17-40), containing a 2Ј-O-methyl modification from position 17 to 40 from the 5Јend. RNase H assay with D69(OM17-40) showed that internal cleavages from position 17 to 43 were blocked (Fig. 6A). In addition, we also observed that the cleavage at the RNA 5Ј-end was slightly enhanced compared with that observed from the unmodified RNA. At 1 min, more 8 -14-nt 5Ј-end cleavage products were detected from the internal modified substrate than those from the unmodified substrate (compare Fig. 6A with Fig.  4) (data not shown).
When the DNA primer utilized above was used for synthesis, internal modified D69(OM17-40) RNA should result in fulllength extension product and transfer product with the same length as those from the D71 donor RNA. The internal modifications within the D69(OM17-40) donor template caused extensive pausing by RT during synthesis, resulting in poor accumulation of full-length extension product. Reaction conditions were therefore reoptimized for this substrate (see "Experimental Procedures"). Both unmodified and modified donor substrates were then tested using the 5Ј-end-labeled primer DNA. Fig. 6B shows an example of the synthesis reactions, in the presence and absence of acceptor and NC protein. The internal 2Ј-O-methyl modifications resulted in more pausing between positions 30 and 40 and a strong pause site around position 50 (compare Figs. 3B and 6B). With the unmodified donor, a ladder of pausing was observed from position 47 to 53 (Fig. 3B). Self-priming products were absent, presumably because uncleavable 2Ј-O-methyl internal fragments annealed to the full-length extension product, preventing it from folding back. As with the unmodified donor, NC also increased the transfer efficiency with the modified donor.
Comparing transfer efficiencies between the internally modified D69(OM17-40) donor system and the unmodified D71 donor, we observed an inhibition with the longer homology acceptors, A26h and A45h (Fig. 6C). The inhibition was more pronounced (40 -60%) with the 45-nt homology in the presence of NC. Interestingly, for transfers with the A12h and A19h acceptors, transfer efficiencies with the D69(OM17-40) donor were higher than those observed with the unmodified donor D71 (Fig. 6C). This is most likely, because 5Ј-end cleavages were slightly enhanced on the D69(OM17-40) donor (Fig. 6A). Such enhanced cleavages at the donor 5Ј terminus should improve primer annealing to the acceptor template. The finding that blocking donor internal cleavages inhibits transfer only in the long homology template systems supports our prediction that the long homology transfers take place mainly via an acceptor invasion-initiated pathway. Blocking donor internal cleavages may force the system to rely only on the primer terminus switch-initiated pathway. We therefore observe only a partial instead of a complete inhibition of transfers with the 26-and 45-nt homology systems. Since short homology transfers occur by primer terminus switch, blocking internal cleavages does not affect transfers in these systems.
Time Course Analysis of the 45-nt Homology Transfer-To examine the sequence of events during the transfer process, we followed the formation and accumulation of the various intermediates and end products during the course of the reaction. Time course assays were performed for the transfer and RNase H reactions using 5Ј-end-labeled primer (Fig. 7A) and donor (Fig. 7B), respectively. Reactions were performed at a primer/ donor/acceptor ratio of 1.5:1:3 under standard reaction conditions in the absence of NC or at 400% NC coating. For transfer reactions in the absence of NC, full-length donor extension product was detectable within 15 s, followed by self-priming product at 30 s (Fig. 7A). Transfer product was not detected until about 1 min (Fig. 7, A and C). In the presence of NC, full-length extension product again was detected within 15 s (Fig. 7A). Interestingly, the transfer product was now detected as early as 30 s (Fig. 7, A and D), whereas the self-priming product was not observed until 1 min (Fig. 7A). Formation of the self-priming product requires cleavage and degradation of the donor RNA at its 5Ј-end (38, 50, 77). The accumulation of transfer products before the formation of self-priming product, in the presence of NC, suggests that transfer occurred before the donor 5Ј-end fragment was cleaved. This argues that the end cleavage is not required to facilitate the transfer.
When RNase H cleavage was followed during transfer reac-tions, the donor template was almost completely degraded in reactions with NC (Fig. 7B). In the absence of NC, larger cleavage fragments accumulated at the earlier time points, which were degraded to smaller fragments in time. The absence of such products in reactions with NC suggests that NC enhances the overall RNase H efficiency. However, no substantive differences in the cleavage site distribution were observed. The 14 -15-nt product was detected within 15 s, whereas the 8 -10-nt secondary cut product was not detected until 30 s or 1 min. In all, during earlier time points, no significant changes were observed in the formation and accumulation of the 14 -15and 8 -10-nt fragments at the donor 5Ј-end, suggesting that NC does not influence donor 5Ј-end cleavages significantly during the transfer reaction.
To examine the effect of NC on transfers with the 45-nt homology system in more detail, formation of the transfer product and 8 -10-nt secondary cut products during the transfer reaction were plotted together as a function of time (Fig. 7,  C and D). At 1 min, in the absence of NC secondary cut efficiency was 1.5%, whereas transfer efficiency was only 0.4% (Fig. 7C). The overall time course showed a more rapid accumulation of 8 -10-nt cut products as compared with transfer products. In sharp contrast, in reactions with NC, a 4% transfer efficiency was detectable as early as 30 s (Fig. 7D). The overall time course of reaction with NC showed no significant change in the rate of formation of the 8 -10-nt cut product, whereas transfer product was formed at almost 20-fold higher efficiency. In the presence of NC, transfer products were formed even before the secondary cuts at the donor 5Ј-end were initiated, thereby suggesting that the secondary cut is not necessary for transfer. The overall data agree with our conclusion that for minus strand transfer with 45-nt homology, transfer occurs mainly via the acceptor invasion-initiated pathway. Most likely, the donor 5Ј-end fragment is strand-displaced by the acceptor template during the invasion process.
Time Course Analysis of the 19-nt Homology Transfer-From our results, we expected that transfers take place through the primer terminus switch-initiated pathway in templates with short homologies (12 and 19 nt). If so, cuts that remove the donor 5Ј-end fragments should be detectable prior to the formation of transfer products. To test this, we repeated the time course assays using the 12-nt (data not shown) and 19-nt homology (Fig. 8) acceptors in the presence and absence of NC. Transfer product formation and donor RNA degradation were followed by using 5Ј-end-labeled primer (Fig. 8A) and donor   FIG. 5. Synthesis with D71(OM4 -14). A, extension assays in the absence or presence of acceptor, with different levels of NC coating, 0, 200, 400, 800, and 1200%. Descriptions are the same as in Fig. 3B. Full-length (FL) and transfer (T) products are indicated. B, quantitation of transfer efficiencies from assays with the modified D71 (OM4 -14) donors. Quantitation was based on at least three independent experiments. Details are the same as in Fig. 3C.   FIG. 6. Synthesis and RNase H assay with D69(OM17-40). A, polymerase-independent RNase H cleavage with D69(OM17-40) RNA containing hybrid substrate. Details are the same as in Fig. 4. The RNA was annealed to its cDNA to create a blunt-ended hybrid. Assays were performed in the absence of NC and terminated after 1, 2, 4, 16, and 32 min of reaction. Lane C, a control reaction without NC; lanes L, a 10-bp DNA ladder. B, representative gels of extension assays with D69(OM17-40), showing donor extension (lanes 1 and 2) and transfer with A12h (lanes 3 and 4) in the absence (lanes 1 and 3) or with 400% coating (lanes 2 and 4)  RNA (data not shown), respectively. In the presence or absence of NC, full-length extension product was detected in 15 s, whereas the self-priming product was detected at 30 s. Transfer product formation occurred more slowly, detected only by 2-4 min in the absence of NC and by 1-2 min in its presence. Therefore, although NC stimulated transfer, self-priming product was always formed prior to formation of transfer product. This was the converse order to that found with the 45-nt homology system. Cleavage of the donor template during transfer was similar to that observed for the 45-nt homology template (data not shown), where NC increased the overall RNase H activity but did not alter the cleavage profile. Formation of the donor 5Ј-end cleavage products (namely the 14-nt fragment and the 8 -10 nt secondary cut products) was similar to that observed for the 45-nt homology template (see Fig. 7B).
Formation of the transfer product and the secondary cut products with time is plotted in Fig. 8, B and C. Unlike results with the A45h system, NC protein did not change the order of formation of the products in the 19-nt homology system, although it stimulated transfer rate. In the absence or presence of NC, full-length extension product appeared within 15 s, followed by the 8 -10-nt cut product by 30 s to 1 min. Transfer products were formed only by 1 min in the presence or 2 min in the absence of NC. At 2 min, transfer efficiency was 1.5 and 0.3% with and without NC, whereas secondary cut efficiency was about 5.5% in both cases (Fig. 8, B and C). During the whole time course, although NC stimulated transfers, the transfer efficiency was much lower than the 8 -10-nt secondary cut efficiency. The result that transfer product was formed only after the 8 -10-nt cut is consistent with the need to remove the terminal RNA segment to allow transfer by the primer terminus switch-initiated mechanism. DISCUSSION Understanding the mechanism of minus strand strong stop transfer, an obligatory step in reverse transcription has been the focus of numerous studies. The current study examines two possible pathways for the transfer mechanism. Previous obser- vations that the R 5Ј-end fragment is cleaved inefficiently (43,54,59,60) prompted us to consider the "acceptor invasioninitiated pathway," where transfer occurs independent of the terminal cleavage. In this pathway, transfer is initiated through annealing of the acceptor at internal sites in the cDNA. Strand exchange would proceed by a branch migration process, displacing other RNA fragments including the one at the terminus. The requirement for long homology for efficient transfers (20 -22) supports such an acceptor invasion-initiated mechanism. Additionally, the ability of NC to promote strand exchange of the primer from the cleaved donor template to the acceptor (34) could account for the stimulatory effects of NC on transfer by this mechanism. Alternately, NC can stimulate cleavage of the 14 -16-nt 5Ј-end fragment (43), thereby freeing the primer terminus. In such a scenario, transfer can occur through a "primer terminus switch-initiated" mechanism ( Fig.  1), where the free primer terminus anneals to the acceptor and facilitates transfer.
We have designed a minus strand transfer system in vitro using HIV-1 R sequences to examine features of primer terminus switch-initiated and acceptor invasion-initiated pathways and to determine which is the likely mechanism for minus strand transfer in HIV-1. The role of RNase H cleavages at the 5Ј-end versus internal regions of the donor template in the transfer mechanism were verified using 2Ј-O-methyl modifications. The modification specifically blocks RNase H cleavage at the modified site without affecting DNA polymerization. Blocking donor 5Ј-end cleavages almost completely inhibited transfers with the short homology ( Fig. 5; 12 and 19 nt), confirming that for short homology, degrading the donor 5Ј-end is critical for transfer. The absence of transfer products even at 400% NC coating further suggests that the NC-induced transfers in these templates do not occur by a simple, RNase H-independent, strand exchange mechanism. In the long homology template system, blocking 5Ј-end cleavage did not interfere with transfers (Fig. 5). Instead, transfers were reduced when internal cleavages within the donor template were blocked (Fig. 6). Taken together, data suggest that the primer terminus switchinitiated mechanism is not the primary pathway for transfers in the long homology templates.
In templates with long homology, inhibition of internal cuts did not completely block transfers. This was probably because alternate pathways such as primer terminus transfer are activated when the invasion pathway is blocked or because a portion of the transfers always occurred through the terminus switch pathway. Some transfers may also have resulted from NC-induced strand exchange with the partially cleaved donor. Interestingly, 5Ј-end cleavages were slightly enhanced in our system with the D69(OM17-40) donor (Fig. 6A), which would explain the slight increase in transfer efficiencies with the 12-nt and 19-nt homology templates with the D69(OM17-40) donor (Fig. 6C).
Relevant to the simultaneous employment of both pathways is our observation that the fate of the 5Ј-terminal RNA segment of the HIV-1 genome is different in a reaction in which ϪsssDNA is simply made versus a reaction in which it engages in transfer to an acceptor RNA. Our initial analysis showed that NC stimulated the 8 -10-nt cuts at the RNA 5Ј-end by 50 -100% in donor extension reactions without acceptor (Fig.  2B) and by 4 -5-fold on preformed hybrid substrates without synthesis (Fig. 2D) (43). In contrast, those same cuts at the donor 5Ј-end were not stimulated by NC during transfer reaction, especially at early time points. This was true for transfer systems with both short (19 nt; Fig. 8C) and long (45 nt; Fig.  7D) homologies. This difference is probably due to many of the end fragments being displaced by the acceptor during transfer and therefore no longer available for cleavage. However, in donor extension reactions and in RNase H assays in the absence of synthesis, the 14 -18-nt 5Ј-end fragment remains annealed to the cDNA, and NC stimulates its cleavage. This raises an interesting issue concerning the 19-nt homology transfers. If the 19-nt homology transfer occurs through the end transfer mechanism but NC does not stimulate donor 5Јend cleavages, then what is the mechanism of NC stimulation of transfer in this situation? Although cleavages out to 16 nucleotides were blocked in D71 (OM4 -14), very high levels of NC stimulated transfer (Fig. 5, lane 15), presumably by promoting strand exchange. This indicates that aspects of stimulation of strand exchange by NC are not related to terminal cleavages. We believe that the low efficiency of cleavages at the donor 5Ј-end is in fact a limiting factor for transfers in the 19-nt homology system, making the end transfer mechanism less efficient. The same is also true for the 12-nt system, with the reduced homology further exacerbating the situation. For short homologies, NC very likely has a general stimulatory effect on both efficient transfers from cleaved terminal regions and inefficient transfers from uncleaved terminal regions.
Analysis of transfers with different lengths of homology between donor and acceptor RNAs helped us to understand the structural features of the substrates that promote one mechanism of transfer versus the other. With the 45-nt homology transfer system, transfer products accumulated at a faster rate than the 5Ј-end cleavage product (Fig. 7), indicating that transfers do not depend on cleavage of the terminal fragment. Furthermore, in the presence of NC, transfer products were formed prior to formation of self-priming products (Fig. 7A). Formation of self-priming products is an indication of donor 5Ј-end cleavage (50,58). The Hughes group proposed that the uncleaved 14-nt 5Ј-end fragment serves to prevent the ϪsssDNA from folding back on itself and self-priming and that NC serves to keep the terminal 14-nt fragment annealed to the donor template (50,58). In contrast, with the shorter 19-nt homology system, the bulk of transfers occurred after substantial cleavage near the RNA 5Ј-end (Fig. 8), and self-priming products were formed before and during formation of transfer products (Fig. 8A). Peliska and Benkovic (57) found that the 14-nt fragment at the genome 5Ј-end needs to be degraded to facilitate transfer. Since their system was based on a 20-nt homology, the results agree with our "primer terminus switch-initiated" pathway. The overall data suggest that acceptor invasion rather than the primer terminus switch drives strand transfer in retroviruses with long terminal homologies such as HIV-1.
Recent studies in our laboratory have shown evidence for an acceptor invasion-initiated mechanism for internal transfer events that lead to recombination (56,78). Transfers occur through a two-step "dock and lock" mechanism, where the acceptor interacts (docks) with the donor-cDNA complex at regions where RNase H cleavage is enhanced (76,78). The acceptor-cDNA hybrid propagates through branch migration and ultimately catches up with the primer terminus, where transfer of the primer terminus (lock) completes the event. The acceptor invasion therefore appears to be the preferred pathway for transfers during both minus strand transfers and internal transfers leading to recombination, including repeat deletions. Minus strand transfer is a special case of the dock and lock mechanism in which the position of the lock step is determined by the termination of synthesis at the 5Ј-end of the viral genome.
The viral NC is a well studied nucleic acid chaperone protein shown to promote strand transfer (see Ref. 30 and references therein). As observed in this study, NC stimulates specific steps in minus strand transfer mechanism, two of which in-clude 1) increasing RNase H cleavages and 2) facilitating acceptor-cDNA interactions and strand exchange. NC enhanced overall RNase H activity during minus strand transfer, where full-length donor template and the large template fragments formed early in the reaction were rapidly degraded (Fig.  7B), creating opportunities for acceptor-cDNA interaction for long homology transfer. NC also stimulated transfers with short and long homology templates. However, the efficiency of the 8 -10-nt secondary cuts was not significantly affected by the homology length or presence of NC. This suggests that NC-stimulated transfers were not effected through the primer terminus transfer-initiated mechanism, involving cleavage of the donor 5Ј-end. The increased internal cleavages within the donor, together with the increased transfer efficiency and inefficient cleavage of the donor 5Ј-end suggest that NC stimulates transfers by facilitating acceptor-cDNA interactions followed by efficient strand exchange. The 5Ј-end fragment is thus displaced and remains uncleaved. NC has been shown to promote strand exchange (34) as well as accelerate the annealing between the ϪsssDNA and the HIV-1 3Ј R region by 3000-fold (35). Studies by Negroni and Buc (76) show that NC coating of the acceptor, but not donor, is important for recombination, suggesting that NC-induced changes in the acceptor conformation contribute to the enhanced transfers.
Minus strand transfer is facilitated through the homologous repeat (R) regions at both ends of the genome. Results of our study show that transfers are more efficient in systems with longer homology, although a low efficiency of transfers was still detectable with as little as 12-nt homology. This agrees with the observations of others that a minimum homology is required (20 -22). Moreover, the ability to reconstitute minus strand transfers by substituting the viral R sequence with nonviral sequences supports the argument that it is the homology rather than the specific sequence that is important for transfer (79). The reason why some retroviruses have evolved a shorter homologous region for ϪsssDNA transfer is not known, but the effects of differences in homology length may be compensated by a myriad of factors. Berkhout et al. (24) proposed that in addition to homology, structural features of the HIV-1 repeat regions facilitate minus strand transfer. They propose that "kissing interactions" between the anti-TAR and antipoly(A) hairpins in the ϪsssDNA and the TAR and poly(A) of the 3Ј-untranslated region (acceptor) initiate the first interactions between the cDNA and acceptor. In addition, studies have shown that template secondary structure, especially that of the acceptor RNA, is important (72,76,78). The acceptor needs to be accessible to invade and initiate transfer. In our study, although sharing longer homology with the donor template than A26h, the A45h acceptor forms a more stable secondary structure that apparently limits accessibility for invasion, resulting in lower transfers. Studies show that besides RT-RNase H activity, NC, and template homology, additional factors are also involved in facilitating minus strong stop transfer. Conformation of the dimeric RNA genome (23) as well as transient interactions between the tRNA 3 Lys and a conserved sequence in the U3 region (74) have been implicated in facilitating minus strand transfer in HIV-1. Topping et al. (80) propose that a base pairing-independent mechanism functions in the specific guidance of strong stop template switch to the U3/R junction in Moloney murine leukemia virus.
Based on the overall data, we conclude that transfers can be facilitated by two different mechanisms, depending on structure of the polymer substrates. When the homology is relatively short, transfers occur via the primer terminus switch-initiated pathway. This is likely to be the major pathway for minus strand transfer in retroviruses such as the mouse mammary tumor virus (15-nt R), type D viruses such as Mason-Pfizer monkey virus (25-nt R), and avian type C viruses (21-nt R), which have short repeat regions. In lentiviruses such as HIV (98 -174-nt R), spumaviruses (170 -191-nt R), murine leukemia virus (68-nt R), and the human T-cell leukemia virus/bovine leukemia virus group of viruses (ϳ230-nt R), which contain relatively long R sequences, transfers are likely to occur predominantly through an acceptor invasion-initiated mechanism. In the context of a long homology R, it is not clear where the invasion is initiated and whether the invasion initiates before or after the RT has copied to the end of the donor. Such details are currently under investigation. In addition, the templates used in this study are relatively short and simple compared with the 10-kb genome in HIV-1. We are currently in the process of testing the acceptor invasion-initiated pathway by utilizing a more complete system to mimic the minus strand transfer in HIV-1.