Avian Retrovirus U3 and U5 DNA Inverted Repeats ROLE OF NONSYMMETRICAL NUCLEOTIDES IN PROMOTING FULL-SITE INTEGRATION BY PURIFIED VIRION AND BACTERIAL RECOMBINANT INTEGRASES*

The U3 and U5 termini of linear retrovirus DNA con- tain imperfect inverted repeats that are necessary for the concerted insertion of the termini into the host chro- mosome by viral integrase. Avian myeloblastosis virus integrase can efficiently insert the termini of retrovirus- like DNA donor substrates (480 base pairs) by a concerted mechanism (full-site reaction) into circular tar- get DNA in vitro . The specific activities of virion-derived avian myeloblastosis virus integrase and bacterial re- combinant Rous sarcoma virus (Prague A strain) integrase ( (cid:59) 50 n M or less) appear similar upon catalyzing the full-site reaction with 3 (cid:42) -OH recessed wild type or mutant donor substrates. We examined the role of the three nonsymmetrical nucleotides located at the 5th, 8th, and 12th positions in the U3 and U5 15-base pair inverted repeats for their ability to modify the full-site and simultaneously, the half-site strand transfer reac- tions. Our data suggest that the nucleotide at the 5th position appears to be responsible for the 3–5-fold pref- erence for wild type U3 ends over wild type U5 ends by integrase for concerted integration. Additional muta- tions at the 5th or 6th position, or both, of U3 or U5 termini significantly increased ( (cid:59) 3 fold) the full-site reactions 5th nucleotide of LTR role in regulating integrase catalysis.

Upon retrovirus infection, the viral RNA genome is reverse transcribed into a linear blunt ended DNA genome. The retrovirus U3 and U5 DNA termini contain LTR 1 sequences with short imperfect inverted repeats located at the very end of the blunt ended LTRs (1). In vivo, the inverted repeats are necessary for virally encoded integrase to catalyze the removal of a dinucleotide from the 3Ј-OH termini and the subsequent fullsite integration reaction (2). The full-site reaction involves the concerted insertion of the two recessed LTR DNA termini into the host genome. This reaction also results in the formation of a small size host duplication at the site of insertion whose size is virus-specific (1).
In vitro, the mechanisms involved in the recognition of the blunt ended LTR termini by integrase for the 3Ј-OH processing reaction and for half-site strand transfer of the recessed LTR termini into target DNA have been investigated (2)(3)(4)(5)(6)(7)(8)(9)(10). The half-site reaction involves the insertion of only one LTR terminus into the DNA target. These in vitro analyses using purified integrase from several retrovirus species have established that the imperfect inverted repeat sequences located at the LTR termini are also necessary for catalysis (1,11,12). Besides the essential CA dinucleotide located 2 nucleotides downstream from the blunt ended viral termini, approximately 5-10 nucleotides internally also play varying roles in the 3Ј-OH processing and half-site strand transfer reactions (2,13).
The full-site integration reaction can be catalyzed efficiently using retrovirus-like donor substrates with integrase purified from virions (14,15). Study of the specific interactions of integrase with the U3 and U5 LTR termini for the full-site integration reaction in vitro would provide insights into the in vivo integration reaction. The bimolecular donor full-site reaction (see Fig. 1, bottom) catalyzed by AMV integrase produces the avian 6-bp host site duplication as demonstrated by DNA sequence analysis (14,15, this report). The unimolecular donor full-site reaction, where two ends of one molecule are used, is less than 5% as efficient as the bimolecular reaction (15)(16)(17)(18). The ability of bacterial recombinant RSV integrase (16) and AMV integrase (17,18) to produce the 6-bp host duplication with the unimolecular donor reaction is also high.
We have reported previously that the avian retrovirus U3 LTR terminus is preferred over the U5 LTR terminus for both the 3Ј-OH processing (19) and the half-site and full-site strand transfer reactions (14, 15, 18 -20). Three nonsymmetrical nucleotides are at the 5th, 8th, and 12th positions of the 15-bp inverted repeats located at the RSV LTR termini. It is unknown what role the three nonsymmetrical nucleotides have in modifying the recognition of integrase for either LTR terminus or their influence on the full-site integration reaction.
In this report we examined the role of the three nonsymmetrical nucleotides located in the avian U3 and U5 15-bp inverted repeat and several other LTR mutations in the formation of donor-target recombinants that are produced by the full-site integration reaction. The full-site reactions were catalyzed by either AMV integrase derived from virions or by recombinant RSV PrA integrase purified from bacteria. Of the seven LTR mutations examined, some had either a significant gain or a loss of function for the full-site reaction, whereas others had no affect. The interactions of integrase with wt and mutant LTR DNA termini for full-site strand transfer were evaluated by BglII digestion of labeled donor-target recombinants resolved by agarose gel electrophoresis and DNA sequence analysis of donor-target junctions. The specific activities of both virion and recombinant integrase appear equivalent when catalyzing the half-site and full-site strand transfer reactions with either wt or mutant LTR DNA sub-strates, or the 3Ј-OH processing reaction with similar blunt ended substrates.

EXPERIMENTAL PROCEDURES
DNA Donor Substrates-The wt and mutant LTR donor substrates were produced by cloning 60-bp double-stranded oligonucleotides containing terminal U3 and U5 LTR sequences into the NdeI site of pUC19 plasmids lacking a HindIII site (Fig. 1, top). The U3 and U5 ends were separated by a HindIII site. A genetic selection marker, the supF gene (420 bp), was cloned into the HindIII site of each plasmid. The clones were sequenced to verify the constructs. All plasmids were purified by velocity sedimentation before digestion with either NdeI or both NdeI and HinfI to isolate the appropriate LTR donor fragments. The donor fragments were isolated by either agarose or by polyacrylamide gel electrophoresis. All donor substrates contained the unique BglII site adjacent to the U3 LTR terminus (Fig. 1, bottom). With this same structural arrangement, all of the donor-target recombinants would give the same restriction pattern. The target DNA was supercoiled pGEM, which lacked a BglII site.
Labeling of Donor Substrates-The 5Ј-end labeling of NdeI-digested DNA fragments was performed by T4 polynucleotide kinase and [␥-32 P]ATP (15). We routinely labeled sets of wt and mutant donor substrates at the same time to produce labeled molecules that had very similar specific activities, usually at 2,000 cpm/ng of DNA. This labeling strategy allowed us to determine directly by PhosphorImager analysis the amount of input donor DNA which was used for strand transfer within each set of wt and mutant donor reactions.
The 3Ј-OH labeling of the NdeI sites on the 480-bp donors was accomplished by using Klenow polymerase with [␣-32 P]TTP and unlabeled deoxynucleotides (19). The fill-in reactions were always greater than 98% as verified by restriction enzyme cleavage of the labeled molecules and by subsequent examination of the labeled strand on DNA sequencing gels. After labeling, the donors were digested with HinfI producing two fragments of 248 bp (U5 end) and 198 bp (U3 end) in size. The HinfI sites were filled in with unlabeled deoxynucleotides.
Genetic Selection of Full-site Donor-Target Recombinants-Scale-up reactions (40ϫ) were performed for sequencing of donor-target recombinants. After isolation of the 3.3-kbp donor-target recombinants by BglII restriction enzyme digestion and agarose gel electrophoresis, the ligated DNAs were transformed into CA244 cells (15). Colonies were screened for plasmids that were analyzed by size, restriction enzyme digestion, and dideoxy DNA sequencing to determine the size of the host site duplications.
Assay Conditions for Strand Transfer and 3Ј-OH Processing-Standard reaction mixtures (20 l) for full-site strand transfer contained 20 mM HEPES buffer (pH 7.5), 1 mM dithiothreitol, 8% polyethylene glycol 6000, 15% dioxane, and 330 mM NaCl (14). Briefly, AMV or RSV integrase (50 nM) was first preincubated on ice for 10 min in the above assay mixture with 15 ng of donor DNA. The strand transfer reactions were initiated by the addition of target (100 ng) and immediate incubation at 37°C for 10 min. The molar ratio of dimeric integrase to donor ends was 12:1, respectively. The reactions were stopped, and the DNA products were separated on agarose gels (15). The amount of products produced was determined by a Molecular Dynamics PhosphorImager. The standard 3Ј-OH processing reaction conditions were 10 mM HEPES (pH 7.5), 2 mM dithiothreitol, 140 mM NaCl, and 20 mM MgCl 2 (19). AMV and RSV integrase was at either 10 or 20 nM. Integrase:blunt ended donor substrates ratios used for the 3Ј-OH processing reaction were similar to those described for strand transfer.
Integrase from AMV Virions and Recombinant RSV Integrase from Bacteria-AMV (21) and RSV integrase (data not shown) were purified to near homogeneity. RSV integrase was cloned from an infectious PrA viral DNA clone (22) and expressed in bacteria using a pET11 vector. The purification procedure and the physical characterization of RSV integrase will be published separately. The protein concentrations of each purified integrase preparation were determined by A 280 measurements.

RESULTS
Mutagenesis Strategy Used to Investigate the Full-site Integration Reaction in Vitro-In this report we examined the contribution of the three nonsymmetrical nucleotides that map to the 5th, 8th, and 12th positions of the RSV U3 and U5 LTR DNA termini for the full-site integration reaction (Fig. 1). The mutations were introduced onto the 3Ј-OH processing strand and are numbered with respect to the blunt ended terminus. The U3 nonsymmetrical nucleotides were modified singly to U5 sequences, and the U5 nonsymmetrical nucleotides were modified singly to U3 sequences; in each case, the other two nonsymmetrical nucleotides were unchanged. Several other gain or loss of function mutations in the LTR termini were also examined. In all cases, a wt LTR terminus was present on the opposite end of each LTR mutant donor substrate.
Measuring Catalytic Rates of Wild Type and Mutant LTR Donor Substrates-The full-site and half-site strand transfer reactions using the 480-bp wt U3/U5, wt U3/U3, and three mutant LTR donor substrates ( Fig. 1) with AMV integrase at 50 nM are shown in Fig. 2A. The wt U3/U5 and wt U3/U3 donors served as control substrates for measuring catalytic rates and for determining how the different wt LTR termini affect the full-site reaction. In a 10-min reaction at 37°C, the total incorporation of each input donor into the pGEM target was 10.8, 15, 3.7, 10.8, and 16.3% for wt U3/U5, wt U3/U3, U3P5-A/T, U3P8-A/G, and U5P5,6-TT/AA, respectively (Fig. 3, bottom line). As determined by PhosphorImager analysis, the above reactions were linear for at least 20 min, and in most cases, half or more of the DNA products were full-site recombinants. With the same labeled LTR donors, purified RSV integrase at 50 nM incorporated 9.4, 14.3, 1.8, 8, and 22% of the input substrates, respectively, into the pGEM target (Fig. 2B). With either integrase, fewer than 1% of the input donors were integrated into themselves as donor-donor recombinants. This is the first report of a recombinant integrase protein that has a specific activity similar to that of the virion-purified protein for catalyzing the full-site reaction using wt and mutant donor substrates. It should be noted that concentrations of AMV integrase greater than ϳ125 nM (14) or RSV integrase greater than ϳ60 nM (data not shown) in the standard reaction for full-site integration initiate inhibition of catalysis.
Modification of the 5th nucleotide (A to T) on the U3 LTR terminus (U3P5-A/T donor) decreased the ability of both AMV integrase and RSV integrase to catalyze the full-site and half-FIG. 2. Full-site and half-site integration reactions using wt and mutant U3 and U5 LTR donor substrates. Panel A, a standard strand transfer reaction with AMV integrase was performed, and the DNA products were resolved by 1% agarose gel electrophoresis. Standard amounts of each reaction followed by radioactive counting were analyzed, allowing for a direct comparison among the different input donor substrates having similar specific activities. The gel was dried and exposed to x-ray film and was also subjected to PhosphorImager analysis. The donor substrates are identified as five sets of two reactions each, one without (Ϫ) and one with (ϩ) integrase. The vertical lettering on the top of the gel above each set identifies the donor substrates. The various DNA recombinant products and input donors are marked on the left. The half-site reaction involved the insertion of one donor molecule into pGEM, whereas the full-site reaction involved the concerted insertion of two donor molecules into pGEM producing linear 3.8-kbp DNA (Fig. 1). The donor-donor recombinants are caused by integrase inserting donor DNA into other donor molecules. Panel B, the same DNA substrates and reaction conditions as described in panel A were used by RSV integrase to produce donor-target recombinants. Analysis of the reaction products was the same as in panel A. The BglII digestion of one donor molecule inserted into pGEM produces either "A" or "B" type structures depending on whether the U5 or U3 end was integrated, respectively. BglII digestion of the linear 3.8-kbp donor-target recombinant produces four linear restriction fragments with the two 3.3-kbp U3/U5 fragments comigrating. Using PhosphorImager data, the percentage of each digested product produced was determined relative to the total donor-target products produced by that donor (Fig. 4A). The BglII 3.7-kbp U5/U5 product (middle of first vertical column of numbers) obtained using wt U3/U5 donor was set to equal 1. The other numbers were derived by dividing the percentage of homologous U5/U5 full-site product of the wt donor into the percentage of each product obtained with a specific donor. For example, the numbers in the first column (wt U3/U5 donor) are very similar to the numbers in the fourth column (U3P8-A/G donor) because this mutation did not modify significantly the half-site and full-site reactions catalyzed by integrase. For visualization of the above comparison, compare lanes 2 and 8 of Fig. 4A. site reactions to one-third or less of that obtained with the wt U3/U5 donor (Fig. 2, A and B, lanes 2 and 6; Fig. 3). Modification of the nonsymmetrical 8th nucleotide of U3 from A to G (U3P8-A/G donor) did not modify the ability of either integrase to catalyze stand transfer compared with the wt U3/U5 donor (Fig. 2, A and B, lanes 2 and 8). A gain of function mutation in the U5 LTR, the U5P5,6-TT/AA donor relative to the wt U3/U5 donor, is shown in Fig. 2, A and B, lanes 2 and 10; and Fig. 3. The double mutation produces a U3 LTR sequence up to the 5th nucleotide with one additional nucleotide change upstream. Analysis of the data shows that the 5th and 6th nucleotides that are critical for half-site strand transfer also significantly influence the full-site reaction in vitro.
Quantitative Interactions between Wild Type and Mutant LTR Termini for Full-site Integration-We investigated if the individual mutations that were introduced into the recessed U3 and U5 LTR termini modified their interactions with other wt LTR termini or with themselves for full-site strand transfer and, simultaneously, for the half-site reaction. To investigate these interactions, the wt and mutant donor-target recombinants (Fig. 2, A and B) were subjected to BglII restriction analysis (AMV reactions, Fig. 4A; RSV reactions, Fig. 4B). It was established previously that BglII digestion of half-site and full-site donor-target recombinants gives specific cleavage patterns (14,15), as illustrated in Figs. 1 and 3. The quantity of each uncleaved and cleaved donor-target recombinant was determined by PhosphorImager analysis. For comparison among all donor-target recombinants, the full-site 3.7-kbp homologous U5/U5 recombinant obtained by BglII digestion of the wt U3/U5 donor (Fig. 4A, lane 2) for AMV integrase was arbitrarily set to 1 (Fig. 3), in relationship to the other digested donor-target recombinants obtained with the other donor substrates. The full-site U5/U5 recombinants were produced at the lowest level for any of the wt full-site recombinants. Similar quantitative data (data not shown) were obtained as shown in Fig. 3 with the RSV integrase reactions with the same donor substrates (Fig. 4B).
A control strand transfer reaction was analyzed first with the wt U3/U5 donor substrate (Fig. 4, A and B, lanes 1 and 2;  Fig. 3). With this donor, the recessed U3 LTR terminus was approximately 3-fold more effective than the recessed U5 LTR terminus for the half-site reaction (see column of wt U3/U5 donor in Fig. 3 and the column of BglII-digested products in Fig. 4, A or B, lane 2). The BglII-digested half-site donor-target recombinants were classified as "A" and "B" type structures that were produced by the use of either the U5 or U3 LTR terminus, respectively (Fig. 3). For the full-site reactions, the 2.9-kbp homologous U3/U3 recombinants were ϳ5-fold higher than the 3.7-kbp homologous U5/U5 recombinants whose quantity was set to 1. The full-site 3.3-kbp U3/U5 recombinants were ϳ3.6-fold higher than the U5/U5 recombinants. The data suggest that integrase recognizes and subsequently uses the U3 terminus at a significantly higher level than the U5 LTR terminus when the LTRs are presented at an equal molar ratio in the reaction mixture.
The BglII digestion of control wt U3/U3 donor reactions (Fig.  3, second column; Fig. 4, A and B, lane 4) demonstrated that the "A" and "B" half-site reactions were nearly equal, suggesting that sequences outside the 25-bp LTR region appeared not to influence strand transfer significantly. All full-site reactions with this donor would be homologous U3/U3 reactions only. The full-site 3.7-kbp and 2.9-kbp U3/U3 recombinants were nearly equal but were half the amount of the full-site 3.3-kbp U3/U3 recombinants. The individual wt U3/U3 donors can be inserted into pGEM in two orientations giving rise to the same size 3.3-kbp BglII restriction fragment. We cannot rule out the possibility of a minor negative effect exerted by the proximity of supF sequences located at the BglII side of the wt U3/U3 donor. The effect is seen when comparing the "B" with "A" half-site products or the full-site 2.9 kbp with 3.7 U3/U3 recombinants (Fig. 3, second column).
The two control reactions above demonstrate that BglII restriction analysis of the donor-target recombinants provides a quantitative and convenient method of simultaneously analyzing half-site and full-site integration reactions.
Effects of U3 Mutations on Full-site Strand Transfer-Because the U3 terminus is identical to the U5 terminus in sequence except for the nonsymmetrical nucleotides, the ability of integrase to recognize and to use recessed U3 over U5 LTR ends would be related to these nucleotides. We next examined how the individual nucleotide changes in the U3 LTR terminus modified the ability of integrase to catalyze both half-site reactions and full-site reactions with its own mutated U3 terminus and with wt U5 LTR ends. With the mutant U3P5-A/T donor and AMV integrase at 50 nM (Fig. 4A, lanes 5 and 6), the overall strand transfer products were decreased to one-third the level observed with the wt U3/U5 donor (Fig. 3). Increasing the concentration of AMV integrase to 88 or 120 nM did not FIG. 4. BglII digestion of wt and mutant LTR donor-target recombinants. The AMV and RSV reaction sets are shown in panels A and B, respectively. The DNA products produced by AMV and RSV integrase (Fig. 2) were subjected to phenol extraction and ethanol precipitation. Each reaction mixture was suspended in a constant volume of buffer. Approximately 20,000 cpm of each sample was (ϩ) or was not (Ϫ) subjected to BglII digestion. It should be noted that the 20,000 cpm of each sample included unused donor DNA as well as the donortarget recombinants, and therefore the observed donor-target recombinants on the gel reflect the catalytic rate of each reaction accurately. The samples were resolved on a 1.5% agarose gel. The donors are listed vertically above each set of two lanes. The half-site and full-site recombinants are indicated on the left, and the DNA products produced by BglII digestion are indicated on the right.
increase the efficiency of catalysis of donor U3P5-A/T nor change its BglII restriction pattern relative to the wt U3/U5 donor (data not shown).
The BglII digestion pattern observed with the mutant donor U3P5-A/T was changed significantly relative to the wt U3/U5 donor (Fig. 4A, lanes 2 and 6). The most striking observation is that the U3P5-A/T LTR end appears to have activity similar to the wt U5 LTR end located on the same donor molecule. Almost all of the catalytic reactions with the U3P5-A/T donor more closely parallel wt U5 LTR than wt U3 LTR activities (Fig. 3,  first and third columns). In addition, the half-site B U3P5-A/T LTR reaction of donor U3P5-A/T was inhibited 60% compared with the U3 LTR of wt donor U3/U5. The full-site 3.3-kbp U3/U5 and the 2.9-kbp homologous U3/U3 recombinant reactions of donor U3P5-A/T were decreased 50 and 85%, respectively, compared with the same wt U3/U5 donor reactions. Similar results were obtained with the BglII digestion patterns using donor U3P5-A/T with RSV integrase at 50 nM (Fig. 4B,  lane 6). With either AMV or RSV integrase (Fig. 4, A and B,  lanes 2 and 8), changing the 8th nucleotide of the U3 LTR from A to G (donor U3P8-A/G) (Fig. 3, fourth column) had no apparent affect on the strand transfer reactions compared with the wt U3/U5 donor. The 12th nonsymmetrical nucleotide in the U3 inverted repeat was not examined, although a single nucleotide deletion 2 bp downstream of the 8th nucleotide decreases half-site and full-site catalytic rates (see below). The data suggest that the nonsymmetrical nucleotide at the 5th position has a significant effect on the preferential recognition of U3 ends over U5 ends by integrase as well as on subsequent catalysis.
The wt U5 end of donor U3P5-A/T was able to relieve some of the inhibitory effects of the single nucleotide change in the U3P5-A/T terminus. This conclusion was reached if one compares the ratio of the full-site U3/U5 (3.3 kbp) to homologous U3/U3 (2.9 kbp) reactions observed (0.66) with the wt U3/U5 donor (Fig. 3, first column) to the ratio observed (2.2) with the full-site 3.3-kbp to 2.9-kbp reactions of donor U3P5-A/T (Fig. 3,  third column).
Effects of U5 Mutations on Full-site Strand Transfer-We next examined the effects of three U5 mutations individually directed against the nonsymmetrical nucleotides on the strand transfer reactions. We compared the wt U3/U5 donor with donors containing single mutations in U5 at the 5th, 8th, and 12th positions (U5P5-T/A, U5P8-G/A, U5P12-G/T, respectively) using both AMV integrase (Fig. 5A) and RSV integrase (Fig.  5B). To determine if sequences downstream of the 8th nonsymmetrical nucleotide of U3 were important, we also investigated whether a single nucleotide deletion (A at position 10) in the U3 LTR of donor U5P8-G/A had an effect on catalysis. For a 10-min reaction at 37°C, AMV and RSV integrase at 50 nM incorporated 10.6, 12.6, 6, and 12% and 10.6, 12.8, 8.4, and 14% of the above donors into pGEM, respectively. As shown previously, fewer than 1% of the donors were integrated into themselves, and greater than 50% of the donor-target recombinants were full-site products (Fig. 5 and data not shown).
The BglII digestion fragments observed in Fig. 5 with the U5 LTR mutations were subjected to PhosphorImager analysis as described in Fig. 3 (data not shown). The data demonstrated that the wt U3/U5 donor reactions with AMV integrase (Fig.  5A, lanes 1 and 2) and RSV integrase (Fig. 5B, lanes 1 and 2) were essentially identical to those observed in Fig. 4 and as calculated in Fig. 3. Several changes in the half-site and fullsite reactions were evident with the above set of U5 LTR mutations. As expected, the modification of the U5 LTR sequence to a U3 LTR sequence (T to A at the 5th position; donor U5P5-T/A) up-regulated the ability of both integrase proteins to use the mutated U5 terminus, which is now similar to the wt U3 terminus. For example, the half-site ("A") product and the full-site 3.7-kbp recombinants using the mutated U5 terminus of donor U5P5-T/A (Fig. 5, A or B, lane 4) were significantly higher (2-3-fold) than the same size products using the U5 end of the wt U3/U5 donor, shown in lane 2 of panels A and B. Similarly, the half-site wt U3 product ("B") of donor U5P5-T/A is nearly the same quantity as the half-site product ("A") with the mutated U5 terminus of the same donor (Fig. 5, A and B,  lane 4). In lane 4 of both integrase sets, the mutated U5 end interactions with the wt U3 end to produce the full-site 3.3-kbp recombinants were similar in quantity to the full-site 2.9-kbp homologous U3/U3 product obtained with the wt U3/U5 donor shown in lane 2. The changing of the 8th (donor U5P8-G/A) or 12th (donor U5P12-G/T) position of the U5 LTR to U3 sequences (Fig. 5, A and B, lanes 6 and 8, respectively) had little effect on the mutated U5 LTR strand transfer reactions compared with reactions observed with the U5 LTR of the wt U3/U5 donor (lane 2). The modification of the U3 LTR by a single nucleotide deletion at the 10th position of donor U5P8-G/A had a modest effect on the overall catalytic rates (compare lanes 5 and 6 of 5A and 5B with the wt U3/U5 donor reactions in lanes 1 and 2). The catalytic rate was decreased with donor U5P8-G/A, but the BglII restriction pattern was not modified were also used in this analysis of the mutant U5 LTR donor-target recombinants. The wt U3/U5 and different U5 LTR mutant donor substrates were labeled to similar specific activities. Standard integration reactions were performed, and the amount of each donor incorporated into pGEM was determined by 1% agarose gel electrophoresis. The remaining samples were suspended is a constant volume of buffer, and approximately 21,000 cpm of each set was (ϩ) or was not (Ϫ) subjected to BglII digestion. The donors are indicated vertically above each set of two lanes. The half-site and full-site recombinants are indicated on the left, and the DNA products produced by BglII digestion are indicated on the right.
significantly. The results suggest that the 5th nucleotide of the U5 LTR end also plays a significant role in regulating integrase recognition and catalysis.
Effects of Double Mutations at the 5th and 6th Positions of U5 and U3-As shown previously, modification of the 5th and 6th nucleotides of the U5 LTR from TT to AA (donor U5P5,6-TT/ AA) increased strand transfer for both the half-site and full-site reactions relative to that observed with the wt U3/U5 donor (Figs. 2 and 3). There was a 1.5-and 2.3-fold increase in the catalytic rates observed with AMV and RSV integrase, respectively, when comparing donor U5P5,6-TT/AA with the wt U3/U5 donor. The double mutation at the 5th and 6th nucleotide of U5 results in a U3 LTR sequence up to the 5th position with an additional T to A change at the 6th position. The U5P5,6-TT/AA mutation enhanced markedly (3-5-fold) both the mutated U5 LTR half-site ("A") and the full-site 3.7-kbp homologous U5/U5 reactions (Fig. 4A, lanes 9 and 10; Fig. 3, last column) compared with same size U5 LTR products obtained with the wt U3/U5 donor (Fig. 4A, lanes 1 and 2; Fig. 3, first column). In fact, the full-site 2.9-kbp homologous U3/U3 recombinant observed using donor U5P5,6-TT/AA was the poorest reaction, suggesting that integrase now has a much higher affinity for the modified U5 end than the wt U3 terminus in the same reaction mixture. The full-site 3.3-kbp U3/U5 product observed with donor U5P5,6-TT/AA was similar in quantity to the same size product obtained using the wt U3/U5 donor. Similar conclusions can be reached about the above strand transfer reactions using donor U5P5,6-TT/AA with RSV integrase (Fig. 4B, lanes 9 and 10). The changing of the two T pyrimidines to two A purines at the 5th and 6th positions of U5 enhanced markedly the ability of either AMV or RSV integrase to recognize and to use this modified LTR terminus compared with the wt U5 and U3 ends.
To establish further that the 5th and 6th positions of either U5 or U3 play critical roles in controlling full-site catalysis, we modified the sixth position of the U3 LTR from T to A (donor U3P6-T/A). The sequence of donor U3P6-T/A on the processed strand is 5Ј-GACAACA OH which is similar to the above donor U5P5,6-TT/AA (5Ј-GGCAACA OH ). AMV integrase and RSV integrase at 25 or 50 nM were able to use the donor U3P6-T/A 3-4-fold better for full-site and half-site strand transfer reactions than the wt U3/U5 donor (data not shown). For example, in a 10-min reaction with AMV integrase at 50 nM, 40% and 24% of the input donor were incorporated into full-site and half-site products, respectively. BglII digestion of the half-site reaction products demonstrated that the U3P6-T/A LTR end was used 5-fold better than the wt U5 end. The homologous mutated U3/U3 and homologous wt U5/U5 full-site reactions accounted for 68% and 2% of the full-site products, respectively, whereas the wt U5/mutated U3 end reaction was 30%. The data support the results that were obtained with the U5P5,6-TT/AA donor and that the affinity of integrase for wt U3 is most likely controlled by both the 5th and 6th positions for full-site catalysis.

Sequence Analysis of Wild Type and Mutant LTR Donor-Target
Junctions-AMV integrase produces the 6-bp avian host site duplication using a wt donor termed M-2 (14,15) which is very similar to the wt U3/U5 donor in this study. The ability of RSV PrA integrase to produce the avian 6-bp host site duplication with wt U3/U5 and mutant U5P5,6-TT/AA as donor substrates was investigated. Standard reactions were performed with RSV and AMV integrase, and the donor-target recombinants were subjected to BglII digestion. The linear 3.3-kbp restriction fragments were isolated from each donor set, ligated, and transformed into CA244 cells. Recombinants from each donor set were sequenced at the donor-target junctions (Table I). The results show that both integrase proteins were capable of producing the avian 6-bp host site duplication with some 5-bp and 7-bp duplications. All of the donor LTR ends that were inserted into the target ended with the conserved CA dinucleotide. Several small size deletions were also observed with both donor reactions. The results show that purified RSV and AMV integrase can produce faithfully the avian 6-bp host site duplication in vitro.
3Ј-OH processing of blunt ended wt and mutant LTR DNA termini by AMV and RSV Integrase-Different assay conditions are needed for AMV integrase to process effectively a dinucleotide from blunt ended wt LTR donors compared with using the recessed LTR donors for strand transfer (14,19). The aprotic solvents dimethyl sulfoxide or dioxane, or high NaCl concentrations (330 mM) severely inhibited the 3Ј-OH processing reaction. The NdeI sites of wt U3/U5 and mutant donors U3P5-T/A and U5P5,6-TT/AA were filled in with labeled TTP and unlabeled dATP, digested with HinfI, and the appropriate end-labeled fragments were isolated. Using assay conditions without aprotic solvents and 140 mM NaCl, AMV integrase at 10 or 20 nM preferentially trimmed the blunt ended wt U3 LTR terminus over the wt U5 LTR terminus (Fig. 6, left). AMV integrase also preferred the wt U5 terminus over the mutated U3 terminus of donor U3P5-T/A, whereas the wt U3 and the U5P5,6-TT/AA ends were hydrolyzed similarly (Fig. 6, middle and right, respectively). Similar data were obtained with RSV integrase at 10 nM. Higher concentrations of RSV integrase (data not shown) and AMV integrase (19) inhibit the 3Ј-OH processing reaction. Examination of the released nucleotides on FIG. 6. Site-specific cleavage of 3-end-labeled wt U3, wt U5, and U5P5,6-TT/AA ends by AMV integrase. The donors (wt U3/U5, U3P5-T/A, and U5P5,6-TT/AA) were filled in at their NdeI sites using labeled TTP. The blunt ended donor substrates were digested with HinfI, and the appropriate fragments were isolated on agarose gels. Standard 3Ј-OH processing reactions were performed as described under "Experimental Procedures." The concentration of integrase is shown for each set of reactions. Aliquots of each reaction were taken at the indicated times, and soluble trichloroacetic acid counts were determined. The percentages of donor substrate cleaved by integrase were determined (left). 23% polyacrylamide DNA sequencing gels showed that both AMV and RSV integrase produced the same dinucleotide (data not shown; 19). The data suggest that integrase does not show a marked preference for blunt ended LTR donor substrates for processing to the same degree as was shown with their recessed LTR donor substrates for strand transfer. DISCUSSION Single or double nucleotide changes close to the recessed termini of U3 and U5 LTR substrates modify the ability of AMV and RSV integrase to catalyze the full-site integration reaction. The 5th nucleotide was the only nonsymmetrical nucleotide within either the U3 or U5 inverted repeats which significantly affected integrase for the full-site reaction. The A nucleotide at the 5th position of U3 (T for U5 at this position) appears to be responsible for the large preference for wt U3 over wt U5 LTR ends by integrase for half-site and full-site strand transfer. The 8th and 12th nonsymmetrical nucleotides appear to have little effect on strand transfer. Taken together or independently, the U3 and U5 mutational analyses demonstrate that the 5th nucleotide of both LTR termini have a significant role in regulating integrase recognition and catalysis for full-site strand transfer. Significantly, it appears that the LTR mutations affect both the half-site and full-site reactions in nearly a parallel quantitative fashion.
Both recombinant RSV (Schmitt-Ruppin B strain) integrase (16) as well as AMV integrase (14,15) produce the avian 6-bp host site duplication at a high frequency. The specific activities of our RSV PrA integrase and AMV integrase preparations are similar for half-site and full-site strand transfer with wt and mutant donor substrates. However, compared with AMV integrase, RSV PrA integrase appears to prefer the donor U5P5,6-TT/AA, although both proteins prefer this substrate over the wt U3/U5 donor. The mutation in the U5 end of donor U5P5,6-TT/AA produced the sequence 5Ј-GGCAACA OH on the processed strand compared with the wt U5 sequence of 5Ј-GGCT-TCA OH . Cleavage by integrase does not occur at the internal CA dinucleotide with donor U5P5,6-TT/AA (data not shown), which is consistent with only the terminal CA dinucleotide being used for full-site strand transfer (Table I). Further studies are under way to determine if this mutation or any of the other LTR mutations studied in this report will allow us to determine the number of integrase subunits involved in fullsite reactions, as yet undefined.
It was reported previously that changing the 5th and 6th nucleotides on the U5 LTR processed strand from TT to AA completely inhibited 3Ј-OH processing and strand transfer reactions using double-stranded oligonucleotides substrates and purified bacterial expressed RSV integrase (Schmidt-Ruppin B strain) (23,24). Substitution of this U5 LTR mutation into an infectious RSV clone demonstrated that the virus was still infectious, showing that this U5 LTR mutation was not lethal. It was suggested that there were cooperative interactions between the wt U3 LTR terminus and the mutated U5 LTR terminus to permit virus replication. With either AMV or RSV integrase, the donor substrate containing the same TT to AA mutation at the U5 LTR terminus (donor U5P5,6-TT/AA) showed a significant increase in strand transfer activities relative to the wt U5 end. The mutated U5 LTR end of donor U5P5,6-TT/AA is fully capable of full-site strand transfer with the wt U3 end (Fig. 3, fifth column). Sequence analysis of the donor-target junctions in these recombinants verified that this mutation was still present in the donor after genetic selection in our CA244 cells. The blunt ended donor U5P5,6-TT/AA substrate was also trimmed correctly by integrase to produce a dinucleotide. Chemical degradation of the fill-in donor showed that it also contained the correct mutation. The reason for the difference between our in vitro data and those reported with oligonucleotide substrates containing the TT to AA mutation is unknown.
The 5Ј 2-bp overhang on the unprocessed donor strand appears to play a major stabilizing role for integrase-DNA complexes capable of half-site strand transfer reactions with oligonucleotide substrates and recombinant human immunodeficiency virus type 1 integrase (5). Although the 3Ј-OH trimming reactions using blunt ended donor substrates with AMV integrase are observed (Fig. 6), the resulting integrase-recessed DNA complexes are not used in an efficient matter for full-site reactions compared with reaction mixtures containing 3Ј-OH recessed donors (14). Similar data for the inefficient use of blunt ended donor substrates for the unimolecular full-site strand transfer reaction with AMV (6,15,18) or RSV integrase (16) have been reported. The wt and several mutant recessed donor substrates used in our study which contained the 5Ј 2-bp overhang are inserted efficiently into target DNA by a full-site reaction mechanism. These results suggest that the 5Ј 2-bp overhang on recessed LTR substrates may induce conformational changes in integrase which allow integrase to form stable protein-DNA complexes both in vitro and in vivo (5). Possibly, the molecular crowding and osmotic pressure effects of PEG and organic solvents (25)(26)(27) modify the ability of both AMV and RSV integrase to recognize and to use substrates with the 5Ј 2-bp overhang but not blunt ended substrates efficiently for the subsequent strand transfer activities. Whether the lack of efficient coupling of the 3Ј-OH processing reaction to the subsequent full-site integration reaction (but not half-site reactions) (14) is the result of the lack of physical contact between processed donors in the reaction mixture or other reasons is unknown. The potential role of the organic solvents in destabilizing the donor termini (14) thus allowing for more efficient strand transfer is also unknown. A recent report suggests that end fraying or DNA distortion of the blunt ended LTR termini by integrase is a required step in the 3Ј-OH processing reaction in vitro (28).
Mutations introduced into one LTR terminus influenced the ability of integrase to use the other wt LTR end for the 3Ј-OH processing reaction in vivo (29). It is difficult to compare this in vivo result with the interactions between wt and mutant recessed LTR ends in our study for full-site strand transfer. Our results with recessed LTR substrates suggest that a wt LTR substrate can increase the ability of a mutant LTR substrate to participate in full-site catalysis. Apparently, the wt U5 LTR end of donor U3P5-A/T diminishes the inhibitory effect of the mutated U3 LTR end on the same molecule to produce the 3.3-kbp U3/U5 recombinant (Fig. 3, third column). Comparison of this U3/U5 reaction with donor U3P5-A/T with the same reaction with the wt U3/U5 donor (Fig. 3, first column) is necessary to support this possibility. What is apparent with the wt U3/U5 donor reactions is that the homologous U3/U3 fullsite reaction is significantly better (ϳ5-fold) than the homologous U5/U5 reaction (set to 1), with the normal in vivo U3/U5 reaction at an intermediate level (ϳ3.6-fold). Whether there is a preferential or leading role of the U3 LTR end over the U5 LTR end for formation of stable preintegration complexes (30) in vivo is unknown. The above observations suggest that there may be significant interactions between the different LTR ends that are presumably coupled by integrase.