Interaction of HIV reverse transcriptase with structures mimicking recombination intermediates.

Interactions between human immunodeficiency virus (HIV) reverse transcriptase (RT) and structures mimicking intermediates proposed to occur during recombination (strand transfer) were investigated. One mechanism proposed for strand transfer is strand exchange in which a homologous RNA (acceptor) "invades" a donor RNA.DNA duplex (replication intermediate) on which DNA synthesis is occurring. The acceptor displaces the donor of the duplex and binds to the DNA. During exchange a transient trimeric structure forms. A model structure was designed with a replication intermediate to which an acceptor RNA was bound. The acceptor was bound to the 5'-end of the DNA over a 54-base region, whereas the donor associated with the DNA 3'-end over a 28-base region. The dimeric constituents of the trimer (acceptor RNA.DNA and donor RNA.DNA) were also constructed. The acceptor RNA.DNA formed a branched structure in this case. Results showed that RT could cleave the RNA portion of all the structures examined. Association with junction substrates was less stable as determined by off-rates. On the trimer, RT cleaved both RNAs but showed a clear preference for cleaving the donor RNA region. This preference was accentuated by HIV nucleocapsid protein (NC). Results suggest that during recombination RT generally associates with the donor-RNA portion of the trimer and the acceptor RNA is protected but not immune from cleavage. The partial protection likely allows the acceptor RNA to more easily complete strand exchange and shield this RNA to provide a means to salvage replication if the DNA were to dissociate from the cleaved donor RNA.

Strand transfer is an essential step in retroviral replication and has been shown to occur from terminal regions and internal regions of RNA templates. Minus strand DNA strand transfer is proposed to potentially occur by two different mechanisms (1). In one, DNA being synthesized on an RNA template (referred to as the donor in this case) dissociates from the RNA and binds to a homologous region on a second RNA (referred to as the acceptor) where synthesis continues. A second mechanism proposes that the acceptor RNA actively participates in the displacement of the DNA strand by associating with the DNA and dislodging the donor. Because retroviruses are diploid, the donor and acceptor RNAs would represent the two genome copies in the capsid. Results from previous experiments suggest that viral nucleocapsid protein (NC) 1 stimulates the binding of a complementary RNA acceptor template to the DNA displacing the donor RNA in the process (2). Consistent with the second mechanism above, this implies that strand transfer could proceed through a trimeric complex consisting of the nascent DNA bound to both the RNA template it was being made on (donor) and a second homologous RNA (acceptor). Such a structure has been shown to occur during DNA transfer in vitro (2,3). Models developed in vivo (4) and in vitro (5,6) also indicate the likely existence of a trimeric intermediate as one of the mechanisms used for strand transfer during recombination. Little is known about what activities RT manifests on such a structure. The binding and orientation of RT to such a structure would become important, because both the RNA templates are accessible to the enzyme. The fate of the acceptor template is important, because degradation of the acceptor prior to transfer would eliminate a potential pathway to continue DNA synthesis if the nascent DNA were to dissociate from the donor template.
In this report, experiments were designed to characterize the activity of RT on strand transfer intermediates with the goal of determining if RT has a preference for associating with the donor or acceptor RNA on a trimeric strand transfer intermediate and if the RNase H activity of RT cleaves these RNAs. Three separate templates were designed to mimic the entire strand transfer intermediate (trimeric structure) or various parts of this structure. One template consisted of the donor RNA bound to the DNA (donor RNA⅐DNA), which mimics the "replication intermediate." The second template consisted of the acceptor RNA bound to the DNA such that the bases at the 3Ј-end of the DNA are not hybridized to the acceptor (acceptor RNA⅐DNA). This mimics the binding of the acceptor RNA to the DNA of the replication intermediate. The third structure, the strand transfer intermediate, contains both the donor and acceptor RNAs bound to DNA. Previous results have indicated that RT is anchored to nucleic acid substrates by a 3Ј recessed DNA terminus or a 5Ј recessed RNA terminus (7)(8)(9)(10)(11)(12)(13). With this in mind, RT would be expected to bind preferentially to the donor RNA of the trimer, because the 3Ј-end of the DNA strand is recessed on the donor RNA. In contrast, there is no 3Ј recessed DNA end on the acceptor RNA and the 5Ј-end of the acceptor is not associated with the DNA suggesting that RT would not efficiently cleave the acceptor RNA. Results showed that RT cleaved both RNAs of the trimer but with a clear preference for cleaving the donor RNA region. In the presence of HIV nucleocapsid protein (NC), this preference was even stronger. The findings imply that during recombination RT generally associates with the donor-RNA portion of the trimeric recombination intermediate and the acceptor RNA is protected but not immune from cleavage. The partial protection likely allows the acceptor RNA to more easily complete strand exchange and shield this RNA to provide a means to salvage replication if the DNA were to dissociate from the cleaved donor RNA.

EXPERIMENTAL PROCEDURES
Materials-Wild-type recombinant HIV-RT was graciously provided to us by Genetics Institute (Cambridge, MA). Aliquots of the enzyme were stored at Ϫ70°C, and a fresh aliquot was used for each experiment. T4 polynucleotide kinase was from United States Biochemical. HIV-1 nucleocapsid protein (NC) was obtained from Enzyco (Denver, CO). SP6, T3 RNA polymerases, calf intestinal alkaline phosphatase, and rNTPs were obtained from Roche Molecular Biochemicals. PvuII was obtained from American Allied Biochemical. HaeIII, placental RNase inhibitor, and Escherichia coli RNase H were obtained from Promega. Deoxyoligonucleotides were synthesized by Genosys Inc. (The Woodlands, TX). Proteinase K was obtained from Sigma. All other chemicals were from Fisher Scientific or Sigma. Radiolabeled compounds were from PerkinElmer Life Sciences.
Construction of Substrates-Run-off transcription was performed according to the instructions of the enzyme manufacturer. The donor RNA (5Ј-GGGAACAAAAGCUAAUUCGCCCUAUAGUGAGUCGUAUUAC-AAUCACUGG) was made from HaeIII cleavage of pBSM13(⌬) (14), and T3 RNA polymerase was used to prepare run-off transcripts 50 nucleotides in length. For the acceptor RNA (5Ј-GAAUACUCAAGCUUAU-GCAUGCGGCCGCAUCUAGAGGGCCCGGAUCAAUUCGAGCUCG-GUACCCGGGGAUCCUCUAGAGUCGACCUGCAGGCAUGCAAG), the vector plasmid pGEM-11ZFϩ (Promega) was cleaved with BamHI, and plasmid pBSM13ϩ was cleaved with EcoRI and HindIII. All termini were filled in with Klenow polymerase. The EcoRI/HindIII insert from pBSM13ϩ was isolated and ligated into the plasmid vector pGEM-11ZFϩ and transformed into E. coli XL1 blue cells. Blue white screening was used to isolate white colonies that contained the pGEM-11ZFϩ plasmid with the insert. The plasmid with the insert was isolated and used to obtain the acceptor RNA. The plasmid was cleaved with PvuII, and SP6 RNA polymerase was used to prepare run-off transcripts 262 nucleotides in length. Reactions for both the donor and acceptor RNA were extracted with phenol:chloroform:isoamyl alcohol (25:24:1, v/v) and precipitated with ethanol. The acceptor RNA was further cleaved to obtain an ϳ99-base pair template as follows: To the acceptor RNA, a DNA oligonucleotide 20 bases long was hybridized by mixing the RNA and the specific primer deoxyoligonucleotide (5Ј-acctcgagggatcagcttgc) at a 1:5 (RNA:DNA) ratio in 50 mM Tris-HCl (pH 8.0), 1 mM dithiothreitol, and 80 mM KCl. The DNA bound to bases 94 -114 from the 5Ј-end of the RNA. The mixture was heated to 65°C for 5 min and then slowly cooled to room temperature. The RNA was then treated with E. coli RNase H (4 units) at 37°C for 30 min in 50 mM Tris-HCl (pH 8.0), 1 mM DTT, 10 mM MgCl 2 , and 80 mM KCl. The final length of the RNA is approximate, because cleavage can theoretically occur anywhere within the 20-base hybrid region. In practice most RNAs differed by only a few bases in length. The precipitates from the donor RNA and the acceptor RNA after its treatment with E. coli RNase H were resuspended in 1ϫ formamide dye (45% formamide, 5 mM EDTA (pH 8), 0.05% xylene cyanol, and 0.05% bromphenol blue). The two RNA were gel-purified on denaturing polyacrylamide gels. The electrophoresed RNA was located by ultraviolet shadowing, excised, and eluted overnight by the crush and soak method (15) in a buffer containing 50 mM Tris-HCl (pH 8), 80 mM KCl, and 1 mM DTT. The eluate was separated from the polyacrylamide by centrifugation and subsequent filtration through a 0.45-m disposable syringe filter. The RNA was recovered by precipitation in ethanol with 300 mM sodium acetate.
The donor and acceptor RNAs were then dephosphorylated using calf-intestinal alkaline phosphatase. The alkaline phosphatase was heat-inactivated at 65°C for 10 min. Reactions were extracted with phenol:chloroform:isoamyl alcohol (25:24:1, v/v) and precipitated with ethanol. The RNA precipitates were labeled at the 5Ј-end with 32 P using T4 polynucleotide kinase according to the manufacturer's instructions. A DNA oligonucleotide 93 bases in length (5Ј-GAGGGATCAGCTTGC-ATGCCTGCAGGTCGACTCTAGAGGATCCCCGGGTACCGAGCTCG-AATTGGCCAGTGAATTGTAATACGACTCACTATA) was also radiola-beled at the 5Ј-end. Two other DNAs were also used in the construction of dimeric substrates, a second 93-base DNA (5Ј-GAGGGATCAGCTT-GCATGCCTGCAGGTCGACTCTAGAGGATCCCCGGGTACCGAGCT-CGAATTGATCCGGGCCCTGTAATACGACTCACTATA) was used to construct the acceptor RNA⅐DNA (ϩ10 hybrid) substrate (see "Results"), and a 64-nucleotide DNA (5Ј-GAGGGATCAGCTTGCATGC-CTGCAGGTCGACTCTAGAGGATCCCCGGGTACCGAGCTCGAAT-TG) was used to construct the acceptor RNA⅐DNA (recessed 3Ј-DNA) substrate.
The RNA⅐DNA hybrid structures were made by mixing together the radiolabeled RNAs and the radiolabeled 93-mer DNA at a ratio of 1:2. The mixture was heated at 70°C for 5 min and slowly cooled down to room temperature. After hybridization, 6ϫ native gel loading buffer (40% (w/v) sucrose, 0.25% (w/v) xylene cyanol and bromphenol blue) was added, and the mixture was electrophoresed on a non-denaturing 6% polyacrylamide gel (29:1 acrylamide:bisacrylamide) as described previously (15). The hybrid complex was located by autoradiography, excised, and eluted in a buffer containing 50 mM Tris-HCl (pH 8) and 1 mM DTT. The eluate was centrifuged and filtered as described above and used directly for experiments.
The trimeric complex consisted of the donor and acceptor RNA bound to the 93-nucleotide DNA. The three radiolabeled components in the trimer had equal specific activities. The trimer was constructed by first making a hybrid of the donor RNA and the DNA at a ratio of 1:0.5. The hybrid was made as described above. To this hybrid, acceptor RNA was added in a solution containing 50 mM Tris-HCl (pH 8.0), 20 mM KCl, 1 mM DTT, 100 M ZnCl 2 , and 6 mM MgCl 2 . The reaction was incubated at 37°C for 15 min in the presence of 5 M NC. Proteinase K was then added to the mixture and placed at room temperature for an additional 20 min. The reaction was stopped by the addition of 6ϫ native dye, and the sample was then run on a 6% native non-denaturing polyacrylamide gel. Other methods, including mixing all three components together at once or making the acceptor RNA⅐DNA first, were also tried in both the presence and absence of NC, but the above method gave the highest yields and most reproducible results. The trimeric complex was located by autoradiography, excised, and eluted at 4°C in a buffer containing 50 mM Tris-HCl (pH 8) and 1 mM DTT. The eluate was centrifuged and filtered as described above and used directly for experiments.
Quantification of Nucleic Acids-Donor and acceptor RNA templates were quantified spectrophotometrically by measuring the absorbance. The molecular weight of the RNA was used to determine the molar concentrations of the RNA templates. The amount of radiolabeled heteroduplex and trimeric substrates recovered from native gels was determined by specific activity. Quantification of RNase H cleavage products was accomplished by scanning the dried polyacrylamide gels with a phosphorimaging device (Bio-Rad, GS 525).
RNase H Cleavage Assays-In the standard RNase H cleavage assay, 5 nM of either the heteroduplex or the trimeric substrate was incubated in a volume of 10.5 l containing the following components at their final concentrations: 50 mM Tris-HCl (pH 8), 80 mM KCl, 7.1 mM MgCl 2 , 0.2 unit/l RNasin, 1 mM DTT, 0.1 mM EDTA (pH 8), and 6 mM AMP at 37°C. Reactions were initiated with the addition of 2 units (0.4 pmol) or in some 0.2 unit (0.04 pmol) of HIV-RT in a 2-l solution containing 50 mM Tris-HCl (pH 8.0), 1 mM DTT, and 80 mM KCl. In RNase H assays performed in the presence of NC, the NC was allowed to preincubate with the substrate for 2 min before the initiation of the reaction. In these reactions, the amount of NC used is indicated in the figures. All reactions were terminated at the indicated times with 12.5 l of 2ϫ gel loading buffer (90% formamide, 10 mM EDTA (pH 8), 0.1% xylene cyanol, and 0.1% bromphenol blue). Samples were electrophoresed on 8% polyacrylamide (19:1 (w/w) acrylamide:bisacrylamide)-7 M urea gels as described below. The gels were dried and used for cleavage product quantification by phosphorimaging using a Bio-Rad GS-525 phosphorimaging device.
Cleavage "Trap" Assays-In assays designed to allow only a single binding event between the enzyme and template ("trap assay"), 2 units (or as indicated) of enzyme was allowed to preincubate with 5 nM template for 3 min in 10.5 l of buffer containing 50 mM Tris-HCl (pH 8), 80 mM KCl, 0.2 unit/l RNasin, 1 mM DTT, 0.1 mM EDTA (pH 8), and 5 mM AMP at 37°C. The reactions were initiated with 2 l of a supplement in the above buffer containing 37.5 mM MgCl 2 , and 2.5 g of poly(rA)-oligo(dT 20 ) primer-template was used as the trap. The reactions were incubated at 37°C for the indicated times and were terminated by the addition of 12.5 l of 2ϫ gel loading buffer and electrophoresed as described above.
Off-rate (k off ) Determination-Experiments designed to determine the dissociation rate constant were performed by pre-binding the enzyme (2 units) with the substrate (5 nM) in a volume of 8.5 l of buffer containing 50 mM Tris-HCl (pH 8.0), 1 mM DTT, 10 mM KCl, 0.2 unit/l RNasin, 0.1 mM EDTA (pH 8), and 5 mM AMP at 37°C. Two microliters of a solution containing 2.5 g of poly(rA)-oligo(dT 20 ) trap in the above buffer was added to the reaction and allowed to incubate further. At various times after trap addition, 2 l of a solution containing 37.5 mM MgCl 2 in the above buffer was added to initiate the cleavage reaction, and samples were incubated for 5 min. The reactions were terminated by the addition of 12.5 l of 2ϫ gel loading buffer, and electrophoresis was performed as described above. For the time zero point, trap and MgCl 2 were added at the same time. Determinations of k off values were performed by fitting the data for the relative amount of cleaved RNA versus time to an equation for exponential decay (f (x) ϭ ae Ϫbx , where a ϭ 1, f (x) ϭ y (relative cleavage), x ϭ time, and b ϭ k off value). In these experiments a slight modification was made. The value at 10 s was set equal to the starting relative value of 1 rather than the time 0 value. This allowed a significantly better fit of the data to the equation. This is probably due to the addition of MgCl 2 along with the trap for time 0. Divalent cation greatly stabilizes binding of RT to the substrate (16). Because HIV-RT likely exists in a closed (tight binding) and open (weak binding) conformation (17)(18)(19), those bound in open conformation may rapidly switch to closed conformation upon Mg 2ϩ binding while they would dissociate in the presence of trap and absence of Mg 2ϩ . This would result in a disproportionately high cleavage amount at the time 0 point as was observed.

Construction of Substrates-
The substrates used for RNase H cleavage assays were described under "Experimental Procedures" and are depicted schematically in each autoradiogram. Five different structures were used. The main structure from which the others were derived is referred to as the "trimer" (Fig. 1). It consisted of three nucleic acid strands, the donor and acceptor RNAs, and a DNA strand. The donor RNA and acceptor RNA were transcribed in vitro and radiolabeled at the 5Ј-end with 32 P. The DNA oligonucleotide was also 5Ј-endlabeled. The donor RNA was 50 nucleotides in length, acceptor RNA approximately (see "Experimental Procedures") 99 nucleotides in length, and the DNA 93 nucleotides in length. The acceptor RNA⅐DNA structure was designed such that a hybrid region of 54 bases formed between the 3Ј-end of the RNA and an internal portion of the DNA that started 30 bases from the 3Ј-end of the DNA. The donor RNA was designed such that a hybrid region of 28 bases formed between the 3Ј-ends of the DNA and donor. This design essentially left one unhybridized base (29th base from the 3Ј-DNA-end) between the donor and acceptor RNAs. There were also 10 bases at the 5Ј-end of the DNA that were not hybridized to RNA. Note that the 28-base hybrid formed between the donor RNA and the DNA is longer than the presumed 18 bases between the polymerase and RNase H active sites of HIV-RT. However, it is still within the realm of what is likely to exist in vivo. Because RNase H activity is generally only about one-tenth the polymerase activity (20), hybrid regions between the nascent DNA and genomic RNA are likely to range from as long as about 30 bases down to sizes too small to remain stably associated (21,22). In addition to the timer, dimeric constituents consisting of the acceptor (acceptor RNA⅐DNA) or donor (donor RNA⅐DNA) bound to the DNA were also examined. Finally, two additional structures based on the acceptor RNA⅐DNA were also used. In one the hybrid region between the acceptor and DNA was increased to 64 bases by moving the junction 10 bases toward the 3Ј-end of the DNA (termed acceptor RNA⅐DNA (ϩ10 hybrid)). In the second the DNA was shortened by 29 bases at the 3Ј-end to produce a 64-nucleotide DNA. The resulting structure (termed acceptor RNA⅐DNA (recessed 3Ј-DNA)) eliminated the branch point of the original acceptor RNA⅐DNA structure and produced a recessed 3Ј-DNA terminus.
To ensure that the appropriate structures were used, it was important to determine that the DNA remained associated with the RNA for both the duplexes and the trimeric structures. For all preparations of substrates obtained, this was verified by examination of the recovered material by native gel electrophoresis. As can be seen from Fig. 2, the various complexes and single-stranded nucleic acids migrated to unique positions on the native gel. This allowed recovery of pure material and confirmation of the integrity of the isolated complexes. It should be noted that the acceptor RNA⅐DNA and trimer structures represent "model substrates" of intermediates occurring during strand transfer. An authentic trimeric transfer intermediate would contain an acceptor RNA that was complementary to the 3Ј-end of the DNA. Various conditions were used in an attempt to isolate "authentic" structures by first forming donor RNA⅐DNA, then adding acceptor RNA that was complementary to the complete DNA of the donor RNA⅐DNA. In all cases an acceptor RNA⅐DNA structure devoid of donor RNA was resolved on the native gels. Regardless, the model substrates that were produced should yield valuable information concerning the recognition of strand transfer intermediates by RT and the mechanism of RNase H cleavage on the substrates.
Cleavage of RNA in Heteroduplex Structures-RNase H cleavage assays done on heteroduplex structures provide information about the binding site of RT. When the donor RNA was bound to the DNA in the heteroduplex (donor RNA⅐DNA), results were as expected given the distance between the polymerase and RNase H active sites on RT and the 3Ј-5Ј directional cleavage activity of RT (data not shown) (7,8,21,22). The enzyme initially cleaved the RNA ϳ18 bases behind the 3Ј-end of the DNA, and the larger initial cleavage products were subsequently chased into smaller products over time by the 3Ј-5Ј nuclease activity of the RT. This results in a shortening of the hybrid region between the DNA and RNA and eventually can cause dissociation of the nucleic acids.
In the heteroduplex structure mimicking the binding of the acceptor RNA to the DNA, cleavage of the acceptor RNA was observed (Fig. 3, acceptor RNA⅐DNA). This was in contrast to what had been reported previously. Work by others using RNA bound to single-stranded circular DNA suggested that RT could cleave structures with the 5Ј-end of the RNA not bound to the DNA. However, this was providing that the unbound 5Ј-RNA end was less than 12 nucleotides in length (11). Based on these results alone it would be predicted that the acceptor RNA⅐DNA duplex would be resistant to cleavage, because the acceptor RNA has an unbound 5Ј-end that is 45 nucleotides away from the hybrid region (see "Discussion" also). Despite this results indicated that the RNA was cleaved. The product sizes at the early time points were diverse indicating that RT was initially bound to the structure such that the RNase H active site contacted the RNA at several positions. The shortest (and most prominent) RNA observed corresponded to cleavage 15 bases behind the junction of the RNA⅐DNA hybrid (60-nucleotide product). Products longer than 60 nucleotides were processed over time to produce more of the 60-base product, whereas small amounts of 56-nucleotide and shorter products were observed later in the reaction. Assays in which E. coli RNase H was added after RT digestion indicated that the 60-nucleotide RNA was still bound to the DNA (data not shown). Overall results suggested that, after somewhat random (although with a preference for sites near junction) initial cleavages, subsequent cleavages proceed toward the RNA⅐DNA junction. Processing at the junction to produce a small unstable hybrid region appeared to be slow.
To compare the cleavage of the acceptor RNA⅐DNA substrate to cleavage of a more standard substrate with a 3Ј recessed DNA termini, a 64-nucleotide DNA that bound the RNA and terminated at the end of the hybrid region was constructed (acceptor RNA⅐DNA (recessed 3Ј-DNA)). The DNA was identical to the first 64 bases at the 5Ј-end of the 93 base DNA used in the acceptor RNA⅐DNA substrate but was missing the 29 bases at the 3Ј-end. This configuration should direct RT cleavage ϳ18 bases behind the DNA 3Ј terminus (Ϫ18 position). Results of a cleavage assay are shown in Fig. 3 (acceptor RNA⅐DNA (recessed 3Ј-DNA)). Early time points revealed initial cleavages 19 (64-nucleotide product) and 16 (61-nucleotide product) bases behind the 3Ј-end. There were also some nonspecific cleavages that produced longer products but the Ϫ19 and Ϫ16 cleavages were clearly the most prominent. At later points cleavages at Ϫ11 (56 bases) were observed as the levels of the larger products decreased. The results indicate that cleavage on the standard substrate with a recessed 3Ј terminus leads to the expected results with a large proportion of initial cleavages occurring from RT molecules bound at the 3Ј terminus. Subsequent 3Ј-5Ј processing results in many of these products being processed to smaller products that likely destabilize the RNA⅐DNA hybrid. In contrast, on the acceptor RNA⅐DNA the RNA⅐DNA junction only loosely directed cleavage and processing was partially inhibited (see above).
To determine how the position of the junction affected cleavage a substrate was constructed in which the position of the junction was advanced by 10 nucleotides toward the 3Ј-end of the DNA. This produced an acceptor RNA⅐DNA substrate with a 64-base as opposed to a 54-base hybrid region. Results with this substrate are shown in Fig. 3 (acceptor RNA⅐DNA (ϩ10 hybrid)). At early time points, as with the original acceptor RNA⅐DNA, an array of products were seen with the distribution shifted toward products produced by cleavages nearer the junction. Prominent products 56 (Ϫ21 relative to the junction), 54 (Ϫ19), 50 (Ϫ15), 44 (Ϫ9), and 43 (Ϫ8) nucleotides were observed. Most products were processed to the 44-and 43-base products at latter time points. This was in contrast to the original acceptor RNA⅐DNA where processing to smaller products was slowed. Results indicate the junction structure does not directly inhibit processing to shorter products.
In the above experiments cleavages were carried out under conditions that allowed multiple binding and rebinding events between RT and the substrates. This makes it difficult to determine if the cleavages occurring even at early time points were initial or secondary cleavages. To more clearly determine the initial cleavage positions trap assays were performed with the substrates (Fig. 4). In these assays RT is preincubated with the substrate then reactions are initiated by addition of MgCl 2 and poly(rA)-oligo(dT) trap. The trap sequesters enzymes that dissociate from the substrate thus limiting cleavage to a single binding event (21). Because RT can catalyze both primary (initial) and secondary cuts in one binding event, the technique does not show only initial cleavage positions, however, secondary cuts occur slowly, so at early time points primary cleavages represent the majority of the observed products (7,10,12,13). A trap control assay in which RT was preincubated with trap was performed for each substrate (lanes E). A low level of cleavage was observed in these lanes in comparison to the starting material (lane F, shown for acceptor RNA⅐DNA only) indicating that the trap was not 100% effective. However, the level of trap read-through was very low and did not interfere with interpretations of the results. For the substrate with the recessed 3Ј termini (acceptor RNA⅐DNA (recessed 3Ј-DNA)) the Ϫ19 and Ϫ16 (64 and 61 nucleotides, respectively) products identified above were clearly the dominant cuts at the 2-and 5-s time points, in agreement with initial cleavage being directed by the position of the 3Ј recessed end. The preference was not as striking on the acceptor RNA⅐DNA substrates, however, cleavage at the Ϫ15 (60 nucleotide product) position relative to the junction was most prominent on the original acceptor RNA⅐DNA. On the acceptor RNA⅐DNA (ϩ10 hybrid), cleavages at Ϫ21 (56 bases) and Ϫ15 (50 bases) were most prominent. This suggests that RT has some preference to bind and cleave near the junction.
The stability with which RT bound to the junction versus substrate with recessed 3Ј terminus was also investigated. To eliminate possible sequence-specific effects, the acceptor RNA⅐DNA and the acceptor RNA⅐DNA (recessed 3Ј-DNA) rather than the donor RNA⅐DNA were used. A modified version of the trap assay was used to measure the rate of dissociation (k off ) of RT from the substrate. For these experiments the KCl concentration was reduced to 10 mM to decrease the rate of dissociation. Using 80 mM KCl RT dissociates rapidly in the absence of MgCl 2 (16). Divalent cation must be omitted in the preincubation phase of the assay, because it would stimulate cleavage of the substrate. For this assay, RT was preincubated with substrate then a supplement containing poly(rA)-oligo(dT) trap was added. Incubation was continued for a fixed time then a supplement containing MgCl 2 was added to initiate cleavage and incubations were continued for 2 min. During the incubation with trap RT molecules that dissociate from the substrate are sequestered. The greater the time between trap and divalent cation addition the more dissociation occurs. This results in a progressive decrease in cleavage that can be used to determine the half-life of the complex (see "Experimental Procedures"). Fig. 5 (A and B) shows the results of a typical k off experiment. Clearly dissociation is considerably slower for acceptor RNA⅐DNA (recessed 3Ј-DNA). This indicates that the presence of a recessed 3Ј-DNA terminus increases the stability of RT binding relative to the junction substrate. Values for k off were 0.015 Ϯ 0.003 and 0.008 Ϯ 0.001 (average of three experiments Ϯ S.D.) for acceptor RNA⅐DNA and acceptor RNA⅐DNA (recessed 3Ј-DNA), respectively, indicating that RT dissociated about twice as quickly from the substrate with no 3Ј recessed terminus.
Cleavage of RNA in the Trimer Structure-RNase H cleavage assays performed on the trimer structure indicated that both the donor RNA and the acceptor RNA were cleaved (Fig. 6). The design of the trimeric structure allowed for the identification of cleavage product derived from both donor and acceptor RNAs in the same reactions. Cleavage products from the acceptor were longer than the donor RNA allowing resolution in indi- The numbers above the arrows refer to the number of nucleotides behind either the junction (acceptor RNA⅐DNA and acceptor RNA⅐DNA (ϩ10 hybrid)) or 3Ј recessed terminus (acceptor RNA⅐DNA (recessed 3Ј-DNA)) that the cleavage occurred. Refer to Fig. 1 for a more detailed drawing of the substrates. Lanes A, partial digestion of template RNA with RNase T1; lanes B, partial base hydrolysis of the template RNA for 15 s; lanes C, partial base hydrolysis of the template RNA for 1 min. vidual reactions. This allowed all three nucleic acids to be radiolabeled in the trimer. In addition, the specific activity of each labeled component was approximately the same allowing a direct comparison of the amount of each RNA that was cleaved. Trimer preparations typically showed a low level of RNA breakdown products that migrated below the donor RNA in the substrate (time point 0 in Fig. 6). These products did not interfere with interpretation of the results. The observed cleav- age products in Fig. 6 indicated that RT bound each RNA in the same manner as it bound the RNAs in the duplex structures (see above). Note that the amount of enzyme used in the assay shown in Fig. 6 was low in comparison to cleavages with the substrates above (0.1 unit per reaction as opposed to 2 units in the assays for the other substrates). This accounts for the slower rate of cleavage observed and allows a more careful determination of the rate of cleavage of the individual RNAs. The results showed that the 50-base donor RNA was cleaved faster than the 99-base acceptor (see Fig. 10 below). This suggests that RT associates more frequently with the region of the trimer that has a recessed 3Ј terminus. This is despite the fact that this region contains only about one-half as much RNA⅐DNA hybrid in comparison to the region bound to the acceptor.
A second experiment was performed to measure the off-rate of RT from the two RNAs in the trimer. The experiment was performed as described above with 10 mM KCl in the reactions. Consistent with the more rapid dissociation of RT from the branched versus 3Ј recessed substrates above, RT dissociated more rapidly from the acceptor region of the trimer (Fig. 7, A and B, k off equals 0.005 and 0.014 for donor and acceptor regions, respectively). Taken together, results with the trimer indicate that RT associates with both regions of the trimer but binding is more stable to the donor region.
The observed cleavages of both RNAs in the trimer could occur in many ways. A single enzyme molecule could cleave both RNAs in succession without dissociating from the trimer. There could be a preference as to which RNA is bound first, or it may be random. Alternatively, RT may bind without preference to either RNA on the trimer and then dissociate after cleaving that RNA. If both RNAs were bound with nearly equal preference, then equal levels of donor and acceptor RNA cleavage would be observed throughout the reaction. Finally, two enzyme molecules could bind each trimer molecule in a cooperative fashion, one on the donor and a second on the acceptor. The latter mechanism is unlikely, because results have shown that RT does not bind cooperatively to substrates. 2 However, at high enzyme concentrations it is possible that more than one RT molecule could be bound to a single trimer simply due to a substantially greater number of enzyme versus substrate molecules. This could lead to cleavage of both RNAs even if RT 2 A. Raja and J. J. DeStefano, unpublished data.  7. A and B, off-rate (k off ) determination for the trimer. Isolated trimer, 5Ј-end-labeled on all three nucleic acid strands to approximately the same specific activity, was incubated with 2 units of HIV-RT as described under "Experimental Procedures" for the off-rate (k off ) determinations. Panel A shows an autoradiogram from a typical experiment. After RT pre-binding, poly(rA)-oligo(dT) trap was added, and the incubations continued for the time indicated. Digestion was then initiated by the addition of MgCl 2 , and reactions were continued for 5 min. Lane A, control incubated without enzyme; lane B, trap control in which RT was pre-mixed with the trap before addition to the substrate and incubated for 100 s; lane C, complete digest of the material with 5 units of RT for 20 min in the absence of trap. Other markings are as indicated in Fig. 6. Panel B shows a plot of relative cleavage versus time derived from the experiment in A. Values are relative to the 10-s time point in the assay, which was set equal to 1 (see "Experimental Procedures"). The k off values from this experiment were 0.014 and 0.005 s Ϫ1 for the acceptor and donor RNAs, respectively. The experiment was repeated with and yielded similar results.
preferred one RNA to the other. To eliminate this possibility reactions were performed with a limiting amount of enzyme under conditions allowing only a single binding event (trap reaction) between the enzyme and the substrate as described above. In trap assays it was observed that both the donor and acceptor RNAs of the trimer were cleaved even at low enzyme concentrations when only a fraction of the total substrate was cleaved (Fig. 8). The fact that only a small fraction of the substrate was cleaved at low enzyme amounts implies that the reactions were not overloaded with enzyme, such that more than one molecule was bound to the substrate. This conclusion assumes that RT is not binding cooperatively. Therefore, these results suggest that either one enzyme molecule cleaves both RNAs in succession or RT binds the RNAs randomly and dissociates after cleavage. In other words, at any given time during the reaction, RT may be bound to the donor RNA of one trimer molecule and acceptor RNA on another. After cleaving the RNA that is bound, RT dissociates without cleaving the second RNA.
In an attempt to differentiate between these possibilities, a time course trap reaction was performed (Fig. 9, A and B). In this experiment reactions were terminated at time points ranging from 2 to 160 s. Both RNAs were cleaved even at the earliest time point with more cleavage of donor versus acceptor. Both RNAs showed an increase in cleavage between 2 and 10 s and no further significant increase (Fig. 9B). This suggests that cleavage is due to individual enzyme molecules bound to either RNA and not to a single molecule cleaving the RNAs in succession (see "Discussion"). If this mechanism is correct, given that the level of donor cleavage was greater than for acceptor, then the implication is that RT associates better with the donor portion of the trimer as was noted above.
Effect of NC on Cleavage of RNA in the Trimer Complex-It has been well documented that NC affects the stability and formation of hybrid structures and can stimulate strand transfer by promoting the formation of a trimeric intermediate (see Introduction). Therefore, it is possible that NC would affect the binding and orientation of RT to the trimer complex. To determine if NC influenced RT binding to the trimer, RNase H assays were performed in the presence of NC. Two different amounts of NC were used corresponding to one molecule of NC for every seven nucleotides of nucleic acid (a final concentration of about 0.17 M) and about one NC per two base (about 0.6 M). A ratio of 1:2 NC:nucleotide was previously determined to be optimal for strand transfer reactions in vitro, although the actual in vivo ratio is ϳ1 NC per 7 nucleotides (23). In the presence of NC the site of cleavage did not change, but there was a decrease in the amount of cleavage for the acceptor RNA (Fig. 10, A and B). Cleavage inhibition increased as the concentration of NC increased but was evident at both concentrations tested. The level of inhibition was small for cleavage of the donor RNA and was only observed at the higher NC concentration (Fig. 10B). The results are consistent with NC decreasing the binding of RT by competing with the enzyme for binding sites on the substrate (2). Notable is the large difference between inhibition on the donor and acceptor. Even without NC the donor was cleaved more rapidly than the acceptor (see Fig. 6 also), and this difference was greatly accentuated in the presence of NC. DISCUSSION In these experiments the binding of RT to a trimeric substrate and its constituent dimer components was investigated. The trimer was designed to mimic a strand transfer intermediate in which the acceptor RNA is "invading" the donor RNA⅐DNA complex (see Introduction). Overall the experiments showed that the two RNAs involved in a recombination event that proceeds through the formation of a trimer are recognized differently by RT. RT binds the acceptor RNA, especially in the presence of NC, relatively poorly. Binding and cleavage of the donor RNA is clearly favored. The partial protection of the acceptor as well as cleavage of the donor may serve to promote more effective strand transfer as well as provide a salvage pathway for replication should the DNA and donor RNA dissociate due to RT RNase H cleavage.
HIV-RT bound both the donor and acceptor RNAs of the trimer while showing a preference for associating with the donor region (Figs. 6 -10). Cleavage of the acceptor was not anticipated because other experiments indicated that the 5Јend of an RNA molecule must be bound or in close proximity to the DNA to observe RT-directed RNA cleavage (11). Results showed that cleavage can be independent of the presence of a DNA 3Ј-end and is directed by the 5Ј-RNA end providing the unbound 5Ј-RNA end was short (less than 12 bases). In our experiments cleavage was clearly not directed by the 5Ј-RNA terminus, because it was too far away (45 bases from the RNA⅐DNA hybrid). In fact initial cleavages were only "loosely" directed by the position of the junction suggesting that RT binds the junction substrate with some tendency to bind nearer the junction (Figs. 3 and 4). The apparent contradiction between these and the previous results may be due to the difference in the substrates used. Palaniappan et al. (11) used a substrate in which RNA primers were annealed to large circular DNA templates. The presence of DNA termini on the substrates used in our experiments may have promoted the association of RT with the acceptor. A second possibility is that the difference in the size of the DNA templates used in the experiment affected binding. The circular DNAs employed by Palaniappan et al. were several thousand nucleotides long. This could have resulted in binding competition for RT, because the RNA⅐DNA hybrid regions were relatively small. It has been shown that RT has some affinity for single-stranded nucleic acid (24,25). In the absence of a bound 5Ј-RNA end the singlestranded region of the substrate could have out-competed the hybrid region for RT binding. This argument is strengthened by the fact the binding to the junctioned acceptor RNA⅐DNA was less stable than binding to a comparable substrate with a recessed 3Ј-end (Fig. 5). Finally, it is possible that the particular substrate used in our experiments allowed cleavage due to an undefined sequence or structural property of the substrate. This seems unlikely because another substrate with a similar configuration but unrelated sequence was also cleaved by RT (data not shown).
As was noted in the results, cleavage of both the donor and acceptor RNA in a trimeric complex could occur in several ways. The results presented here support a mechanism where an RT molecule binds a given substrate on either the donor or acceptor with a preference for binding donor. After cleavage of the bound RNA, RT most likely dissociates without cleaving the second RNA. Several lines of evidence support this mechanism: 1) Results indicate that RT shows no cooperativity in binding substrate. 2 This indicates that two RTs would not associate with a single substrate under conditions in which the concentration of enzyme is limiting. Under such conditions cleavage of both RNAs was still observed in a trap assay (Fig.  8). Therefore, cleavage must have resulted from a single RT molecule cleaving both RNAs in succession or to the mechanism noted above. 2) In trap assays that limited binding to a single event, cleavage of both RNAs was observed at early time points, and the extent of cleavage did not increase after a short time (Fig. 9). This makes it unlikely that cleavage occurred by one RT molecule cleaving both RNAs in succession. If this were FIG. 10. A and B, autoradiogram and graph of a time course cleavage experiment with the trimer in the presence of absence of NC protein. A, isolated trimer, 5Ј-end-labeled on all three nucleic acid strands to approximately the same specific activity, was incubated with 0.2 unit of HIV-RT in the presence or absence of NC protein at 0.17 (1 NC per 7 nucleotides total substrate) or 0.6 M (1 NC per 2 nucleotides total substrate) as described under "Experimental Procedures." Digestion was continued for the indicated time, and samples were electrophoresed on a denaturing polyacrylamide gel to generate the autoradiogram. Other markings are as indicated in Fig. 6. B, graph of relative substrate remaining versus time derived from the experiment shown in panel A. This experiment was repeated with similar results. the case an increase in the level of cleavage over time should have been observed for one of the RNAs. There was an initial increase between 2 and 10 s in the trap reactions but no further increase. Others have shown that it takes as long as 10 s for single RT molecules to catalyze an initial cleavage event on an RNA⅐DNA hybrid, although most cleavages occur within a few seconds. Following initial cleavages RT remains associated with the RNA⅐DNA hybrid for up to several minutes catalyzing further cleavages of the initial product in a 3Ј-5Ј direction (21,22). Given that no significant increase in total cleavage products was observed beyond 10 s in trap assays, it is unlikely that RT molecules cleaved one RNA then migrated to and cleaved the second RNA. If this occurred it would have to have taken place within the first 10 s of the trap reactions. Because the previous results indicate that RT remains associated with the RNA potentially for several minutes, cleavage of two RNAs in succession within 10 s seems highly unlikely.
Because these data support non-successive cleavage of the two RNAs this implies that RT must have some preference for binding to the donor over the acceptor RNA on the trimer. A greater proportion of donor versus acceptor RNA was cleaved in the trap assay indicating that RT associated preferentially with the donor (Fig. 9). In addition, cleavage in assays with low amounts of enzyme showed the donor RNA of the trimer being degraded more rapidly than the acceptor (Fig. 10). The k off values from experiments with the trimer (Fig. 7) as well as dimeric substrates (Fig. 5) support this notion, because RT bound more stably to substrates with 3Ј recessed termini, consistent with the role of these termini in stabilizing binding (26). However, the tendency to cleave acceptor or donor in the trimeric substrate likely depends on the overall affinity (K d ) of RT for that particular region rather than the off-rate. Given that the rate of cleavage by RT is several times greater than the off-rate (20), RT will presumably cleave either RNA nearly 100% of the time that the enzyme binds to it. Therefore, RT tends to associate more frequently with donor RNA rather than the acceptor in the trimer. Whether this is due to binding stability or to the overall configuration of the timer is not clear.
Cleavage in the presence of NC showed the same trend as experiments without NC. Cleavage of both RNAs occurred with faster cleavage of the donor RNA. The level of cleavage was slightly reduced on the donor RNA by NC, whereas cleavage on the acceptor was significantly inhibited (Fig. 10). This decrease likely resulted from NC competing with RT for binding to the substrate as has been suggested by others (2,27). The greater inhibition of acceptor cleavage may result from RT binding this region of the substrate more weakly as indicated above. The weaker binding may make it more difficult to out-compete NC for binding sites.
The nature of the trimeric substrate used in this work did not allow a direct comparison to what might occur on an authentic strand transfer intermediate. As was noted under "Results," the acceptor RNA 5Ј-end would be complimentary to the 3Ј-end of the DNA in the authentic intermediate. The effect of the complementarity would possibly be to further promote dissoci-ation of the donor RNA from the intermediate. Other experiments using a replication intermediate similar to the donor RNA⅐DNA duplex showed that an acceptor RNA could displace the donor in the presence of NC (see the introduction and Ref. 2). This occurred in the absence of RT, indicating that no RNase H activity was required for the strand exchange. However, at the same time, the current work clearly shows that the acceptor would not be immune to degradation during this process. Extensive degradation before completion of strand exchange could result in dissociation of the acceptor from the trimer complex. But the acceptor could still potentially bind the nascent DNA strand if the donor dissociated, because the complementary regions near the 3Ј-ends of the acceptor RNA and DNA would be preserved, but shortened. In this regard NC may play two important roles in the exchange process, promotion of the association of the acceptor and DNA as well as limiting RNase H cleavage (see Fig. 10), which would favor exchange over acceptor dissociation.