Central core domain of retroviral integrase is responsible for target site selection.

Integration of retroviral DNA can occur into many sites on target DNA with a wide variation in preference. One factor known to affect target site selection is integrase, the viral protein required for the integration reaction. In this study, assays that measure the distribution and frequency of retroviral DNA integration showed that purified integrases of human immunodeficiency virus type 1 (HIV-1) and feline immunodeficiency virus (FIV) had different patterns of target site usage. The integrase domain involved in target site selection was mapped by analyzing the integration pattern of chimeric proteins formed between HIV-1 and FIV integrases and of deletion variants of the two wild-type integrases. The results indicate that the domain responsible for target site selection resides in the central core region of integrase.

Integration of retroviral DNA can occur into many sites on target DNA with a wide variation in preference. One factor known to affect target site selection is integrase, the viral protein required for the integration reaction. In this study, assays that measure the distribution and frequency of retroviral DNA integration showed that purified integrases of human immunodeficiency virus type 1 (HIV-1) and feline immunodeficiency virus (FIV) had different patterns of target site usage.

The integrase domain involved in target site selection was mapped by analyzing the integration pattern of chimeric proteins formed between HIV-1 and FIV integrases and of deletion variants of the two wild-type integrases. The results indicate that the domain responsible for target site selection resides in the central core region of integrase.
Integration of a double-stranded DNA copy of the viral RNA genome into a chromosome of the host cell can be divided into three steps: 3Ј-end processing, 3Ј-end joining, and 5Ј-end joining. Purified integrase alone can catalyze the first two steps of the reaction (1-7). 3Ј-End processing involves the removal of two nucleotides from the 3Ј-end of each strand of linear viral DNA so that the viral 3Ј-ends terminate with the conserved CA dinucleotide. 3Ј-End joining or strand transfer is a concerted cleavage-ligation reaction during which integrase makes a staggered cut in the target DNA and ligates the recessed 3Јends of the viral DNA to the 5Ј-ends of the target DNA at the cleavage site. Integrase also catalyzes a reversal of the 3Ј-end joining reaction, termed disintegration, in which a substrate mimicking one end of viral DNA joined to target DNA is resolved to its viral and target DNA parts (8).
Mutational analyses of several retroviral integrases have identified at least three distinct functional domains (9 -11): (i) an HHCC domain in the N terminus, named for its zinc fingerlike motif HX 3-7 HX 23-32 CX 2 C, (ii) a central domain located within a protease-resistant core containing the DD35E motif (DX 39 -58 DX 35 E) that is highly conserved in retroviruses and retrovirus-like elements (12), and (iii) a C terminus domain that represents the least conserved region among retroviral integrases. The HHCC domain is critical for forming a stable complex with viral DNA, and the domain may be involved in multimerization of integrase (13,14). The central core domain is important for catalysis (9,(15)(16)(17) and dimerization (18,19). The C-terminal domain is capable of binding DNA nonspecifically and is believed to contain the site for binding target DNA (20 -25).
One characteristic feature of retroviral DNA integration is that many sites on target DNA can be used for integration. However, the frequency and distribution of integration are not random, and there are hot spots and cold spots of integration on target DNA (Refs. 26 and 27 and references therein). The mechanism for target site selection is not well understood. The transcriptional status of DNA, methylation, association of DNA with histones or other DNA-binding proteins, and DNA bending are known to affect target site specificity (28 -32). In vitro studies showed that integrases from different retroviruses, such as human immunodeficiency virus type 1 (HIV-1), 1 murine leukemia virus, feline immunodeficiency virus (FIV), and visna virus, produce different integration patterns (as defined by frequency and distribution), indicating that integrase is a major factor in determining target site preference (30,33). 2 However, the integrase domain involved in target site selection has not been defined. Similarities in the integration patterns among wild-type and various deletion derivatives of HIV-1 integrase suggest that the domain involved in target site selection resides in the central core region of integrase (34). Analysis of the patterns of nonspecific alcoholysis obtained from chimeras between HIV-1 and visna virus integrases also implicates the core region in selecting target sites (33). It is not clear, however, whether the determinant for nonspecific alcoholysis is equivalent to that for target site selection.
In the present study, we prepared chimeric proteins between HIV-1 and FIV integrase to identify domains responsible for target site selection. FIV is a lentivirus with a complex genomic organization resembling that of primate lentiviruses (35). The amino acid sequence of FIV integrase is 37% identical to that of HIV integrase, and mutational analysis of FIV integrase showed that its enzymatic activities and domain organization are similar to that of HIV-1 integrase (Fig. 1). 2 Since the integrases of HIV-1 and FIV produce different integration patterns and thus may possess different target site determinants, 2 analysis of integration patterns of chimeras between HIV-1 and FIV integrases represents a useful strategy of mapping integrase domains involved in target site selection. Comparison of the patterns of 3Ј-end joining of the various chimeras indicates that the central core domains of both HIV-1 and FIV integrases contain the determinants for target site selection.
The C terminus, although it binds DNA nonspecifically, is not involved in target site selection.

EXPERIMENTAL PROCEDURES
Enzymes and Reagents-Wild-type and deletion mutants of HIV-1 and FIV integrases were purified as described previously (34). 2 A rabbit antipeptide antiserum to FIV integrase was obtained from Dr. John H. Elder at the Scripps Research Institute. T4 polynucleotide kinase, T4 DNA ligase, and restriction endonucleases were purchased from New England Biolabs; AmpliTaq DNA polymerase was from Perkin-Elmer; modified T7 DNA polymerase (Sequenase version 2.0) and exonucleasefree Klenow fragment of Escherichia coli DNA polymerase I were from U.S. Biochemical Corp. Deoxyribonucleotides were purchased from Pharmacia Biotech Inc. [␥-32 P]ATP and [␣-32 P]TTP were obtained from Amersham at a specific activity of 6,000 Ci/mmol. Oligonucleotides were purchased from Operon Technologies, Inc., and were purified by electrophoresis through a 15% denaturing polyacrylamide gel before use.
Cloning and Expression of HIV/FIV Chimeric Integrases-HIV/FIV chimeric integrases were prepared by constructing plasmids encoding the chimeric integrase genes. The chimeras H/H/F and F/F/H, which contained a swap at the C terminus, were prepared by exchanging the AvaII-HindIII DNA fragment between pT7-7/H-IN, which encodes HIV-1 integrase, and pT7-7/F-IN, which encodes FIV integrase (34). 2 The chimeras containing a swap at the N terminus, H/F/F and F/H/H, were prepared by performing two sequential polymerase chain reactions (PCRs). For H/F/F, in the first PCR, pT7-7/H-IN was used as the template. The 5Ј-primer was H5ЈN (5Ј-GAAGGAGATATACATAT-GTTTTTAGATGGA-3Ј) that contains an NdeI site (underlined) and anneals to the N terminus of HIV-1 integrase. The 3Ј-primer was H48F (5Ј-TTCAATTGTCCTCCCACTTGTTCCCCTTTTAGCTGACATT-3Ј), and it contains DNA sequences derived from HIV-1 (roman type) and FIV (italic type) integrases. The PCR product was purified by a QIAquick spin column (Qiagen) and was used as the 5Ј-primer in a second PCR, which included pT7-7/F-IN as the template and F3ЈB (5Ј-GGTC-ACGGATCCTTACTCATCCCCTTCAGG-3Ј) as the 3Ј-primer. F3ЈB contains a BamHI site (underlined) and a stop codon (boldface type), and it anneals to the C terminus of FIV integrase. A similar strategy was used to prepare the chimera F/H/H using F5ЈN (5Ј-CCAGTGCAT-ATGTCCTCTTGGGTTGACAGA-3Ј) and F51H (5Ј-TCTACTTGTCCAT-GCATGGCTTCTCCTATGATTCTGCATA-3Ј) as the 5Ј-and 3Ј-primers, respectively, in the first PCR. The 3Ј-primer for the second PCR was H3ЈH (5Ј-GCTAGAAAGCTTTTAATCCTCATCCTGTCTACT-3Ј), containing a HindIII site (underlined) and a stop codon (boldface type). The chimeras containing a swap of the central core domain, H/F/H and F/H/F, were also prepared using the method of two sequential PCRs. The first PCRs were identical to those described previously for H/F/F and F/H/H. For H/F/H, the template and the 3Ј-primer for the second PCR were the F/F/H DNA and H3ЈH, respectively. For F/H/F, the template and the 3Ј-primer for the second PCR were the H/H/F DNA and F3ЈB, respectively. The PCR products were digested with NdeI and BamHI or NdeI and HindIII, purified by agarose gel electrophoresis and ligated to pT7-7(His) plasmid DNA, previously cut with NdeI and BamHI or NdeI and HindIII. The plasmid pT7-7(His) is derived from pT7-7, a T7 RNA polymerase promoter system (36), and it contains an ATG initiation codon and seven consecutive histidine codons upstream of the unique NdeI site of pT7-7 (34). The resulting plasmids were cleaved with NdeI, and a double-stranded oligonucleotide (5Ј-TATCG-TTCCGCGTGGATC-3Ј and 5Ј-TAGATCCACGCGGAACGA-3Ј) encoding a peptide with a thrombin cleavage site was inserted. These manipulations resulted in the attachment of 15 amino acid residues (MHHHHHHHIVPRGSM) at the N terminus of the integrase coding sequence.
The sequences of all of the PCR-amplified DNA fragments were verified by restriction analysis and the dideoxynucleotide chain termination method. Sequencing reactions were carried out with a modified T7 DNA polymerase according to the manufacturer's specification.
The DNA constructs were transformed into E. coli BL21 (DE3). The cells were grown at 35°C in 4 liters of LB medium containing 50 g of ampicillin/ml. At an optical density of 0.8, isopropyl-1-thio-␤-D-galactopyranoside was added to 0.3 mM for expression induction, and the culture was grown for an additional 5 h. After harvesting, the cell pellet was frozen at Ϫ80°C.
Purification of Chimeric Integrase-The frozen bacterial pellet was resuspended in 100 ml of a buffer containing 20 mM HEPES, pH 7.5, 0.2 mM EDTA, 1 M NaCl, 10% glycerol, 5 mM 2-mercaptoethanol, 1 mM phenylmethylsulfonyl fluoride, and 0.5% Nonidet P-40. The cell suspen-sion was sonicated and centrifuged at 100,000 ϫ g for 1 h at 4°C. The supernatant, after dialysis against buffer A (20 mM HEPES, pH 7.5, 1 M NaCl, 10% glycerol, 5 mM 2-mercaptoethanol, 0.1% Nonidet P-40), was incubated on ice for 2 h with 2.5 ml of Ni 2ϩ -nitrilotriacetic acidagarose resin (Qiagen). The resin was washed five times with 10 ml of buffer A plus 50 mM imidazole and then packed in a column. The protein was eluted in a total volume of 20 or 30 ml with a linear gradient from buffer A plus 50 mM imidazole to buffer A plus 500 mM imidazole. The fractions containing the protein were pooled and dialyzed against buffer C (20 mM HEPES, pH 7.5, 0.1 mM EDTA, 0.5 M NaCl, 20% glycerol, 1 mM dithiothreitol (DTT), and 10 mM CHAPS.
To remove the His tag, the isolated protein was incubated in buffer C with bovine thrombin (Sigma; 25 National Institutes of Health units/mg of integrase) at 30°C for 4 h. The sample (1 mg) was then diluted with 10 volumes of buffer D (20 mM HEPES, pH 7.0, 0.1 mM EDTA, 10% glycerol, 10 mM DTT, and 10 mM CHAPS) before loading onto a column containing 2 ml of SP-Sepharose (Pharmacia). The column was washed with 5 ml of buffer D plus 60 mM NaCl, and the protein was eluted in a total volume of 40 ml with a linear gradient from buffer D plus 60 mM NaCl to buffer D plus 800 mM NaCl. The fractions containing the desired protein were pooled, concentrated on a Centricon-10 column (Amicon), and dialyzed against storage buffer (20 mM HEPES, pH 7.5, 0.1 mM EDTA, 0.3 M NaCl, 20% glycerol, 10 mM DTT, and 10 mM CHAPS). Protein concentrations were determined by the Bradford method (Bio-Rad) using bovine serum albumin as a standard.
The substrates used to assay the 3Ј-end processing activity of wildtype and chimeric integrases were double-stranded oligonucleotides containing sequences derived from the U5 end of the HIV-1 (H-U5V1-2/H-U5V2) or FIV (F-U5V1-2/F-U5V2) long terminal repeat terminus. The substrate was prepared by annealing the H-and F-U5V1-2 strands with their complementary oligonucleotides H-and F-U5V2, respectively. The substrate was singly labeled at the 3Ј-end of the H-and F-U5V1-2 strands using exonuclease-free Klenow fragment of E. coli DNA polymerase I, dGTP and [␣-32 P]TTP (39).
The 3Ј-end joining activity was measured using a modified assay that uses separate oligonucleotides as the donor and target substrates (38,40). The donor substrate, prepared by annealing H-U5V1-2 strand with H-U5V2 strand or annealing F-U5V1-2 with F-U5V2, resembles the viral U5 end after 3Ј-end processing. The target DNA was prepared by annealing T5 and T6 and was singly labeled at the 3Ј-end of the T5 strand using exonuclease-free Klenow fragment of E. coli DNA polymerase I and [␣-32 P]TTP.
The substrate for nonspecific alcoholysis, prepared by annealing T5 and T6 strands, was identical to the labeled target DNA described earlier in the 3Ј-end joining assay.
In all assays, unless indicated otherwise, 0.1 pmol of the labeled DNA was incubated with integrase for 60 min at 37°C in a reaction buffer containing a final concentration of 20 mM HEPES, pH 7.5, 10 mM MnCl 2 , 30 mM NaCl, 10 mM DTT, 0.01 mM EDTA, 20% glycerol, 1 mM CHAPS, and 0.05% Nonidet P-40. The reaction volume was typically 20 l. The reaction was stopped by adding 2.0 l of 0.2 M EDTA, pH 8.0. The reaction products were mixed with an equal volume of loading buffer (98% deionized formamide, 10 mM EDTA, pH 8.0, 0.05% bromphenol blue, 0.05% xylene cyanol) and heated at 90°C for 3 min before analysis by electrophoresis on 15% polyacrylamide gels with 7 M urea in Tris-taurine-EDTA buffer. Alternatively, the reaction products were precipitated and washed with ethanol and resuspended in formamide loading buffer before analysis by electrophoresis on 15% denaturing polyacrylamide gels in Tris borate-EDTA buffer. Quantitation of the products was carried out with a Molecular Dynamics PhosphorImager.
PCR-based Assay for Distribution and Frequency of Integration Sites-The distribution and frequency of integration at individual DNA sites were measured by a PCR-based assay performed as described previously (29,30,34). The donor DNA substrate was the preprocessed U5 DNA substrate (F-U5V1-2 annealed with F-U5V2, or H-U5V1L-2 annealed with H-U5V2L). The target DNA was the plasmid pBluescript KSIIϩ (Stratagene). One microgram of plasmid DNA was incubated with the protein on ice for 5 min in the standard reaction buffer plus 100 g/ml bovine serum albumin. The integration reaction was started by adding 15 nM of preprocessed U5 DNA and incubating the sample at 37°C. After 60 min, the reaction was stopped by the addition of 15 mM EDTA. The sample was extracted with phenol-chloroform, ethanolprecipitated in the presence of 10 g of tRNA, and washed with 70% ethanol. The pellet was resuspended in 50 l of 10 mM Tris-HCl and 1 mM EDTA, pH 7.5. A 5-l aliquot of the reaction mixture was amplified for 25 or 30 cycles of PCR: 1 min at 94°C, 1 min at 55°C, and 2 min at 72°C. For analysis of the integration events occurring in the plus strand of the plasmid DNA, the PCR primers used were 0.2 M unlabeled H-U5V1-2, 0.05 M 5Ј-end-labeled H-U5V1-2, and 0.25 M BSϩ (5Ј-CATTAATGCAGCTGGCACGA-3Ј; nucleotides 988 -969), which is complementary to the plus strand of the plasmid DNA. For analysis of the integration events occurring in the minus strand, the BSϩ primer was replaced by the primer BSϪ (5Ј-TAATACGACTCACTATAGGG-3Ј, nucleotides 625-644), which is complementary to the minus strand of the plasmid DNA. The PCR reaction was performed in a buffer containing a final concentration of 10 mM Tris-HCl, pH 8.3, 50 mM KCl, 0.001% (w/v) gelatin, 1.5 mM MgCl 2 , 200 M dNTPs, and 1 unit of Taq polymerase (Perkin-Elmer Corp.) in a final volume of 20 l. The labeled PCR products were analyzed on a denaturing 5% polyacrylamide gel and visualized by autoradiography.

RESULTS
Experimental Plan-Since the patterns of preferred target sites (determined by frequency and distribution of integration events) are different between HIV-1 and FIV integrases (see Figs. 4 and 5), 2 one approach to identify the integrase domain responsible for target site selection is through analysis of HIV/ FIV chimeric proteins. If the selection of target sites by the wild-type integrase is mediated by a discrete domain, then exchanging that domain between a pair of chimeric proteins should produce a corresponding exchange of integration patterns. We chose to form chimeras between HIV-1 and FIV integrases because the two proteins are related by sequence and size, and they share similar reaction conditions for optimal activities in vitro (35,37,41). 2 Based on the amino acid alignment (Fig. 1) and the domain organization of HIV-1 and FIV integrases (9 -11), 2 the protein was divided into three domains: N terminus, central core, and C terminus. Swapping these domains between the two integrases produced a total of six chimeric proteins ( Fig. 2A). Each chimeric protein is identified by a three-letter code, wherein the first, second, and third letters represent the N-terminal, core, and C-terminal domains, respectively. The source of the domain is indicated by letters F and H, representing FIV and HIV-1 integrases, respectively.
Purification and Activities of Chimeric Proteins-The wildtype HIV-1 and FIV integrases and the chimeras were expressed with a short extension at the N terminus that includes seven histidine residues (His tag) and a thrombin cleavage site. Purification by nickel-chelating affinity chromatography and the subsequent removal of the His tag by cleavage with thrombin produce proteins containing at their N termini three extra residues: glycine, serine, and methionine. Removal of the His tag is critical because, as shown previously, the tag alters the pattern of target site usage. 2 The apparent molecular masses of wild-type and chimeric integrases determined by SDS-polyacrylamide gel electrophoresis were consistent with those predicted by the amino acid sequences (Fig. 2B). Wild-type FIV integrase and several chimeric proteins that contained the core region of FIV integrase, such as H/F/H, F/F/H, and H/F/H, showed prominent products of about 68 kDa in size (Fig. 2B) that reacted positively with an antipeptide antiserum to FIV integrase by Western blot analysis (data not shown). Based on their sizes and reactivity to anti-integrase antibody, the slowly migrating products were presumably dimers of the respective proteins.
The activities of all six chimeric proteins were analyzed in vitro using oligonucleotide-based assays (Fig. 3), and the results are summarized in Table I. All chimeric proteins could mediate disintegration, indicating that their catalytic sites were functional. The ability of the proteins to carry out 3Ј-end processing and 3Ј-end joining was less than that of the wildtype HIV or FIV integrase. No processing or joining activity was detected with H/H/F or F/H/F using the oligonucleotidebased assay. A weak 3Ј-end joining activity for both chimeric proteins could be detected using the more sensitive PCR-based assay when a preprocessed viral DNA end was the donor DNA (see Fig. 5).
Pattern of Target Site Usage Was Independent of the Sequence of Donor DNA-Previous studies have shown that selection of target sites by HIV-1 integrase is dependent on the target DNA and independent of the donor DNA (30,38,40). Whether the same is true with FIV integrase has not been investigated. The earlier finding that HIV-1 and FIV integrases have different preferred target sites 2 also needed to be confirmed. The target site usage of wild-type HIV-1 and FIV integrases was determined using an oligonucleotide-based and a PCR-based 3Ј-end joining assay (Fig. 4).

FIG. 4. Differences in distribution of integration sites between HIV-1 and FIV integrases.
A, oligonucleotidebased assay. Fifteen nanomolar preprocessed HIV-1 U5 (lanes 1, 3, and 5) or FIV U5 (lanes 2, 4, and 6) was preincubated in the absence of integrase (lanes 1 and 2) or in the presence of 200 nM of HIV-1 integrase (lanes 3 and 4) or FIV integrase (lanes 5 and 6) for 10 min at room temperature in the standard reaction buffer. This was followed by the addition of 5 nM 3Ј-end labeled target DNA (T5/T6) and incubation at 37°C for 60 min. The symbols are the same as in Fig. 3B. B, PCRbased assay. Preprocessed HIV-1 U5 (lanes 2, 4, and 6) or FIV U5 (lanes 3, 5, and 7) was preincubated with HIV-1 integrase (lanes 4 and 5) or FIV integrase (lanes 6 and 7) as described above. The reaction was started by adding 50 g/ml of pBluescript SK IIϩ plasmid DNA and incubating at 37°C for 60 min. The reaction products were amplified by PCR as described under "Experimental Procedures." Oligonucleotide BSϩ was used as the 3Ј-primer, and H-U5V1-2 (lanes 2, 4, and 6) or F-U5V1-2 (lanes 3, 5, and 7) was used as the 5Ј-primer. Lanes 1 and 2 contain DNA size markers with lengths in nucleotides indicated on the right. H, HIV-1; F, FIV; IN, integrase.

TABLE I Catalytic activities of chimeric HIV-1/FIV integrases and deletion
derivatives of FIV integrase Activities were determined using oligonucleotide-based assays as described under "Experimental Procedures," and the results are expressed as the percentage of the substrate converted to the product under standard reaction conditions. ϩ, 20% of conversion; ϩ/Ϫ, less than 2% of conversion; Ϫ, no conversion. The 3Ј-end joining activity of two chimeric proteins, H/H/F and F/H/F, and two delection mutants, F-IN53-281 and F-IN53-235, although undetectable with the oligonucleotide-based assay, was detected by the PCR-based assay (see Fig. 5). The concentrations of DNA substrates and wild-type and chimeric integrases were as described in the legend to Fig. 3 Ϫ Ϫ ϩ cleotides were used as donor and target molecules. The target oligonucleotide contains arbitrary DNA sequences and was singly labeled at the 3Ј-end of the T5 strand. The donor oligonucleotide, which contains sequences derived from the U5 end of HIV-1 or FIV long terminal repeat, was not radiolabeled. Therefore, the slowly migrating products were generated from 3Ј-end joining of the donor DNA to the labeled target DNA (Fig.  4A). With HIV-1 integrase (Fig. 4A, lanes 3 and 4; see also Fig.  7), the most preferred position for integration was nucleotide 13 (counting from the 5Ј-end), and other preferred sites were 14, 16, 17, 18, and 19. With FIV integrase, the most preferred position was nucleotide 17, and other preferred sites were 13, 16, 18, and 19 (Fig. 4A, lanes 5 and 6). The integration pattern was unchanged as the concentration of both integrases varied from 25 to 500 nM (data not shown). The result from the oligonucleotide-based assay confirmed the previous observation that HIV-1 and FIV integrases have different target site preferences. 2 However, the selection of target sites by HIV-1 or FIV integrase was identical regardless of whether the donor DNA was derived from HIV-1 U5 (Fig. 4A,  lanes 3 and 5) or FIV U5 (Fig. 4A, lanes 4 and 6). Therefore, using an identical target DNA, the integration pattern depended on the source of integrase and was independent of the donor DNA.
In the PCR-based 3Ј-end joining assay, the donor DNA was identical to the one used in the oligonucleotide assay, and the target DNA was a Bluescript plasmid. After the reaction, the integration site in the recombinant product was amplified by PCR using two primers. One primer anneals to the plasmid, and the other primer, which was labeled at the 5Ј-end, anneals to the donor DNA. The PCR products were analyzed on a denaturing polyacrylamide gel (30,34). Each band on the gel corresponds to an integration event at a given phosphodiester bond (Fig. 4B). The frequency of integration at a particular site is proportional to the intensity of the band, and its position can be determined with use of DNA size markers or a sequencing ladder.
Similar observations to those of the oligonucleotide-based assay were made with the PCR-based assay; HIV-1 and FIV integrases had different integration patterns, and the integration pattern was independent of the donor DNA (Fig. 4B). The usage of target sites of some chimeric proteins was also analyzed using the PCR-based assay and again was found to be independent of the source of donor DNA (data not shown). In all subsequent experiments, the integration pattern was determined using HIV-1 U5 as the donor DNA.
Central Core Domain of Integrase Controls Selection of Target Sites-The integration patterns of chimeric integrases, determined using the oligonucleotide-based (Fig. 3B) and the PCR-based 3Ј-end joining assays (Fig. 5), were compared with those of wild-type HIV-1 and FIV integrases to deduce the domain responsible for target site selection. The results, obtained either by oligonucleotide- (Fig. 3) or PCR-based assay (Fig. 5), clearly showed that swapping the N-or C-terminal domain had no effect on target site selection. The integration patterns of chimeric proteins containing a swap at the N ter-  (Fig. 3, lanes 4 and 8; Fig. 5, lanes 5 and 11), were similar to those of the wild-type HIV-1 (Fig. 3,  lane 2; Fig. 5, lane 3) and FIV integrase (Fig. 3, lane 6; Fig. 5, lane 9), respectively. Likewise, the integration patterns of chimeric proteins containing a swap at the C terminus, H/H/F and F/F/H (Fig. 3, lanes 3 and 7; Fig. 5, lanes 4 and 10), were similar to those of the wild-type HIV-1 and FIV integrases, respectively. Swapping of the central core domain, however, produced a reciprocal exchange of integration patterns between the two resultant chimeric proteins, F/H/F (Fig. 5, lane 6) and H/F/H (Fig. 3, lane 9; Fig. 5, lane 12). The result indicates that the domain responsible for target selection is not in the C or N terminus and instead resides in the core region of integrase.

minus, F/H/H and H/F/F
The role of the core domain in target site selection was further supported by the results obtained from various deletion mutants. Like some of the chimeric integrases, the 3Ј-end joining activity of the deletion mutants could only be detected by the PCR-based assay (Fig. 5). HIV-1 integrase containing a deletion in the C terminus (H-IN1-234) had an integration pattern (Fig. 5, lane 8) similar to that of the full-length HIV-1 integrase, whereas FIV integrase containing a deletion in the N (F-IN1-235) or C terminus (F-IN53-281) had an integration pattern (Fig. 5, lanes 13 and 14) similar to that of the fulllength FIV integrase. Most strikingly, similar integration pat- Patterns of Nonspecific Alcoholysis of Wild-type and Chimeric Integrases-In addition to having sequence-and sitespecific 3Ј-end processing activity, the integrases of HIV-1, Rous sarcoma virus, and visna virus exhibit a nonspecific endonuclease activity (42). The nonspecific alcoholysis of different integrases shows different preferences for target sites. The nonspecific cleavage pattern has been used previously to map the domain responsible for target site selection (33). Consistent with previous observations (42), the preferred sites for nonspecific alcoholysis of FIV integrase were different from those of HIV-1 integrase (Fig. 6, lanes 3 and 7; see also Fig. 7). We therefore examined the patterns of nonspecific alcoholysis of chimeric integrases of HIV-1 and FIV as another means of mapping the domain for target site selection (Fig. 6). For proteins with a core domain derived from FIV integrase (Fig. 6, lanes 8 -10), the nonspecific alcoholysis patterns were similar to that of the wild-type FIV integrase (Fig. 6, lane 7). Of the three chimeric proteins that contained the core domain of HIV-1 integrase, only F/H/H (Fig. 6, lane 5) showed a nonspecific alcoholysis pattern that was identical to the wild-type HIV-1 integrase (Fig. 6, lane 3). The nonspecific alcoholysis patterns of H/H/F and F/H/F (Fig. 6, lanes 4 and 6) were identical to each other but were different from that of wild-type integrases of HIV-1 or FIV. The result suggests that domains other than the core can influence the selection of nonspecific alcoholysis sites. Furthermore, the core domain of HIV-1 integrase or FIV integrase had a very weak to undetectable level of nonspecific cleavage activity (Fig. 6, lanes 11 and 12), suggesting that nonspecific alcoholysis requires domains in addition to the core.

DISCUSSION
In vivo and in vitro studies showed that integration of retroviral DNA occurs into many sites on target DNA (Ref. 26 and references therein). The process, however, is not entirely random; integration into some sites occurs at a frequency several hundred times random (27). One factor known to affect target site selection is integrase, a viral enzyme that catalyzes the 3Ј-end processing and 3Ј-end joining steps of the integration reaction. Even when tested under identical reaction conditions and using an identical DNA substrate as the integration target, Only the sequence of the T5 strand is shown, and the 32 P label is located 5Ј of the outlined thymidine. The sites of 3Ј-end joining and nonspecific alcoholysis were deduced from the sizes of the reaction products shown in Fig. 3B and Fig. 6, respectively. Results obtained from reactions carried out with wild-type HIV-1 integrase are depicted above the DNA sequence and results obtained with wild-type FIV integrase are shown below the DNA sequence. Filled arrowheads denote the sites of 3Ј-end joining, and open arrowheads denote the sites of nonspecific alcoholysis. The relative intensity of the reaction product was determined by Phos-phorImager analysis and was approximated using arrowheads of different sizes. integrases from different retroviruses have different target site preferences (30,33). 2 To identify the domain responsible for target site selection, the integrases of HIV-1 and FIV were divided into three domains, N terminus, central core, and C terminus, and a total of six chimeras were prepared by exchanging each of the three domains between the two proteins Analysis of the integration patterns of the various chimeras between HIV-1 and FIV integrases mapped the domain responsible for selection of target sites to the central core region of integrase. Using both oligonucleotide-and PCR-based assays for determining the distribution and frequency of integration events, we found that a reciprocal exchange of the core domain resulted in a corresponding exchange in the integration pattern of the resultant chimeras. In contrast, an exchange of the N or the C terminus did not alter the integration patterns of the chimeras from their respective wild-type integrase. The chimera result is further corroborated by analysis of deletion mutants of HIV-1 and FIV integrases. HIV-1 integrase containing a deletion at the C terminus or FIV integrase containing a deletion at the N or C terminus showed that the deletion did not appreciably change the integration pattern. Moreover, similar integration patterns to those of the wild-type integrases could be obtained using only the core domain of HIV-1 or FIV integrase. Taken together, the results indicate that the central core domain of integrase is responsible for target site selection.
Close examination of the integration patterns obtained by the PCR-based assay revealed that subtle differences existed between the patterns of the chimeras and those of the wild-type integrases. The differences were reproducible and manifested mainly as a change in integration frequencies in a small fraction of integration sites. The difference was more apparent in the patterns obtained from the core swap chimeras and the deletion mutants. The subtle difference in integration patterns may be caused by slight changes in the core structure induced by the presence of exogenous terminal domains or the absence of terminal domains. Alternatively, the subtle difference implies that the terminal domains may contain secondary determinants for specifying target site usage.
All of the HIV/FIV chimeric proteins tested had poorer activities than the wild-type HIV-1 and FIV integrases. Overall, chimeric integrases containing a core domain derived from FIV integrase had higher activities than those containing an HIV-1 integrase core domain. Since the core domains of HIV-1 and FIV integrases alone had similar activities (data not shown), we speculate that the core domain of FIV integrase can be better complemented by the exogenous N-and C-terminal domains than that of HIV-1 integrase. The best example is the chimeric protein F/F/H, which possessed activities ranging from 50 to 80% of the wild-type integrases.
The finding that the central core domain controls the target site selection is consistent with the previous data showing that various N and C terminus deletion variants of HIV-1 integrase retain a similar integration pattern to that of the full-length integrase (34). Mutational analysis of human immunodeficiency virus type 2 integrase also showed that target site preference is altered by single amino acid substitutions of the asparagine at position 120 within the central core domain (43). We are not aware of any mutations in the N or C terminus of integrase that can lead to a change in target site preference.
Two regions of HIV-1 integrase have been shown to bind DNA: The core domain, which requires the presence of a divalent metal ion for DNA binding (23), and the C terminus, which does not require a divalent metal ion (22)(23)(24)(25). Besides HIV-1 integrase, the C terminus of several integrases, including FIV, human immunodeficiency virus type 2, and avian sarcomaleukosis virus, is capable of binding DNA nonspecifically (20,22,24). 2 Because it binds DNA and is required for efficient 3Ј-end joining, the C terminus is generally regarded as the target DNA-binding domain. The present finding that exchanging or deleting the C terminus of integrase did not alter the integration pattern indicates that the C terminus, although capable of binding target DNA, is not involved in target site selection. Whether the metal ion-dependent, DNA-binding region in the core domain (23) is equivalent to the region for target site selection awaits further investigation.
In addition to the site-specific 3Ј-end processing activity, integrase possesses a nonspecific nuclease activity termed nonspecific alcoholysis (42,44). As in the joining reaction, the cleavage site on the DNA substrate is not entirely random, and different integrases exhibit different preferred cleavage sites (33). By analyzing the pattern of nonspecific alcoholysis of chimeric proteins between HIV-1 and visna virus integrases and assuming nonspecific alcoholysis reflects target site selection during integration, the domain responsible for determining target site usage was mapped to the central core region (33). However, the integration pattern of the HIV-1/visna chimeric integrase was not examined. In the present study, the patterns of nonspecific alcoholysis of HIV-1/FIV chimeric integrases were determined and correlated with the patterns of 3Ј-end joining. For the most part, nonspecific alcoholysis and 3Ј-end joining activities parallel each other, but several discrepancies exist. First, the patterns of nonspecific alcoholysis of H/H/F and F/H/F were different from those of wild-type HIV-1 integrase and F/H/H, although their core domains were all derived from HIV-1 integrase. Second, with both HIV-1 and FIV integrases, the preferred DNA sites for nonspecific alcoholysis were different from those for 3Ј-end joining (Fig. 7). Third, the core domain alone of both HIV-1 and FIV integrases had no significant nonspecific alcoholysis activity. These results lead us to conclude that the determinants for the two activities are not identical and that the pattern of nonspecific alcoholysis is not a reliable marker for identifying the minimal domain responsible for target site selection.
Since integration of retroviral DNA is essential for the subsequent expression of viral genes and production of progenies (45)(46)(47), selection of target sites may have a significant effect on the fate of the infecting retrovirus. Analysis of the preferred target sites may provide information on the interaction between integrase and the sequence and structure of target DNA. Future studies will be focused on identifying the minimal peptides or amino acids involved in target site selection and examining target site usage of infectious viruses bearing integrase with altered target specificity.