Viral Long Terminal Repeat Substrate Binding Characteristics of the Human Immunodeficiency Virus Type 1 Integrase*

A DNA binding assay was developed for the human immunodeficiency virus type 1 (HIV-1) integrase. The assay was capable of defining discrete complexes be- tween the enzyme and the viral long terminal repeat (LTR) substrate. DNA binding reflected the sequence re-quirements previously demonstrated for the enzyme’s 3’-end processing activity. Binding exhibited a nonlinear dependence on integrase concentration, suggesting that the enzyme functions as a multimer. The oligomeric state was investigated by UV-photo-cross-linking of in- tegrase-LTR oligonucleotide complexes using DNA substrates substituted with 5-bromo-2’-deoxycytidine within the integrase recognition sequence. In the ab- sence of divalent cation, integrase cross-linked to the LTR oligonucleotide as a single species whose mobility by SDS-polyacrylamide gel electrophoresis was consist- ent with the formation of tetramers. Using these techniques, analysis of the binding properties of integrase mutants demonstrated that the catalytic and sequence-specific DNA binding activities of the enzyme are dis- tinct, involving residues within the conserved “DD(S5)E’’ and zinc finger motifs, respectively. Integration of a the cell Integration a defined endonucleolytic

Integration of a copy of the viral genome into host cell DNA appears to be generally required for the replication of retroviruses (see Ref. 1 for review). Integration occurs in a defined series of endonucleolytic and DNA strand transfer reactions that are mediated by the virally encoded integrase protein (2)(3)(4)(5). The site-specific endonucleolytic activity of integrase removes the 3"terminal dinucleotide from the LTR' sequences at each end of the viral genome (5)(6)(7)(8). Subsequently, by a strandtransfer reaction, the enzyme joins the recessed 3"termini of the viral DNA to the 5' ends of target DNA which are generated by integrase-mediated nonspecific cleavage of the host cell genome (4, 6, 9, 10).
In several retroviral systems, including Moloney murine leukemia virus, avian sarcoma leukosis virus, Rous sarcoma virus (RSV), and the human immunodeficiency virus type 1 (HIV-1) (2, 4, 5, 7, 10-121, the development of in vitro enzyme assays, using oligonucleotide substrates and recombinantly derived integrase, has increased understanding of the aforementioned catalytic activities of integrase. The functions of specific and nonspecific endonucleolytic processing as well as DNA strandtransfer have been studied using appropriate oligonucleotide substrates and analyzing the reaction products by gel electro-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. $ To whom correspondence and reprint requests should be addressed. The abbreviations used are: LTR, long terminal repeat; RSV, Rous sarcoma virus; BddC), 5-bromo-2'-deoxycytidine; PAGE, polyacrylamide gel electrophoresis.
phoresis. The relative efficiency with which various oligonucleotide sequences serve as substrates in these reactions has been characterized in detail (11, 13-16).
Several investigators have also exploited these in vitro enzymatic reactions to study the effects of mutations in integrase in regions of the protein that are highly conserved between retroviruses (17)(18)(19)(20)(21)(22)(23)(24)(25)(26). Two motifs which are of particular interest include a putative amino-terminal zinc finger and the socalled DD(35)E motif in the enzyme's "central core." The zinc finger is distinguished by 2 His and 2 Cys amino acid residues at positions 12, 16, 40, and 43, respectively (21,27). The DD(35)E motif includes conserved Asp residues at positions 64 and 121 and a conserved Glu residue at 152 (17,21,26,28). Lesions within the zinc finger decrease the specific endonucleolytic and integration activities of the HIV-1 enzyme, but have limited effect on nonspecific DNA cleavage (17,18,261. In contrast, mutations in the DD(35)E motif eliminate all in vitro enzymatic activity (17,18,(24)(25)(26). Although these data suggest that the specific endonucleolytic and DNA strand-transfer activities of integrase involve a common active site, the trivial explanation that these mutations merely affect the protein's structural integrity cannot be discounted. In addition, previous studies did not address whether the substitutions affect substrate binding andor catalytic function. An additional factor complicating the interpretation of these studies derives from biophysical evidence that integrase is oligomeric (5,18,29). Depending on the source of enzyme and method of analysis, both dimeric and tetrameric forms of the enzyme have been observed (5,18,291. Although the precise relationship between oligomeric structure and function is unknown, biochemical studies using the RSV enzyme suggest that at least a dimer of integrase is required for both the processing and DNA strand transfer reactions (29).
Accordingly, a DNA binding assay for the HIV-1 integrase was developed to address questions regarding DNA substrate recognition and subunit composition. Unlike previously published binding assays (12, 211, this assay was able to discern a discrete, enzymatically relevant complex between integrase and the HIV-1 LTR substrate. The nature of the oligomeric state of integrase in these complexes in relationship to substrate binding and divalent cation composition was investigated by UV-photo-cross-linking. Finally, these techniques are used to analyze integrase mutants containing substitutions in either the zinc finger or DD(35)E domains, suggesting that the former participates in specific substrate recognition, while the DD(35)E motif is essential for enzymatic activity.

Cloning, Expression, and Purification of Wild-type and Mutant Inte-
gruse-Cloning of the wild-type HIV-1 integrase protein and expression by a T7 expression vector were described previously (11). Mutation of the integrase and characterization of the cleavage and strand-transfer properties of the mutant enzymes were also described (17). Both wild-
The integrity of the wild-type and mutant integrase proteins was demonstrated by silver stain analysis of each sample following separation by SDS-polyacrylamide gel electrophoresis (PAGE) (Fig. 1). All proteins were expressed at comparable levels and were purified to greater than 90% homogeneity. Western blot analysis using monospecific rabbit antisera (peptide 270-288, see below) was also performed to confirm the identity of the purified products and to verify that the relative amounts of mutant and wild-type enzymes were comparable prior to biochemical analysis.
DNA Binding Assay-A 20-base pair oligonucleotide (Midland Scientific, Midland, T X , Fig. 2), representing the terminus of the HIV-1 HxB2 U3 LTR, was labeled at the 5' end by [Y-"~PIATP b5000 Ci/mmol, Amersham Corp.) using T4 polynucleotide kinase (Pharmacia LKB Biotechnology Inc.) as described (17). Binding reactions between the oligonucleotide and the purified integrase were performed in the absence of either Mg2+ or Mn2+ in binding buffer (20 m Tris HCl, pH 7.9, 0.1 M NaCI, 0.05 mg/ml bovine serum albumin, 5.0 m 2-mercaptoethanol). Labeled oligonucleotides were used at a final concentration of 1.0 nM. Unlabeled competitor oligonucleotides were added to the reactions prior to integrase at the concentrations noted in the respective figure legends. Binding reactions were initiated by the addition of wild-type or mutant integrase, again as noted in the figure legends. Following 30 min at 4 "C, 5.0 p1 of loading buffer (20% glycerol, 0.1% w/v bromphenol blue) was carefully added, and each reaction mixture was loaded onto a nondenaturing 6.0% polyacrylamide gel (acry1amide:bisacrylamide. 37.5:l) in 0.25 x TBE (25 m Tris, pH 8.0,22 m borate, 0.5 m M EDTA). Gels were electrophoresed at 350 V in 0.25 x TBE at 4 "C for 2 h. Gels were dried, and the radiolabeled protein-nucleic acid complexes were visualized by autoradiography. Protein band quantification was performed using the AMBIS Radioanalytic Imaging System (AMBIS, San Diego, CA).
For experiments in which antiserum was used, 1 p1 of the appropriate antiserum was added subsequent to the 30-min binding reaction. The reactions were incubated for an additional 30 min on ice and then analyzed on nondenaturing gels as detailed above. Rabbit polyclonal antisera to HIV-1 integrase residues 1-16, 23-34, and 276288 were obtained from D. P. Grandgenett through the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH. These antibodies are both Western blot and immunoprecipitation-reactive. Rabbit antisera to HIV-1 integrase were also prepared against synthetic peptides encompassing residues 6 2 3 and 270-288. These antisera were shown to specifically recognize a 33-kDa protein in Western blot analysis of lysates from recombinant Escherichia coli expressing either wildtype or mutant integrase. No immunoreactive proteins were detected using control lysates or preimmune sera (not shown). The position of the photoreactive Br(dC) substitution is noted in the U3 substrate. The arrow indicates the cleavage site for the integrase's specific endonuclease activity. The mutant oligonucleotide contains nucleotide substitutions that adversely affect its ability to act as a substrate for integrase (see text).
UV-photo-cross-linking of Integrase-Substrate Complexes-Binding reactions were performed as noted above using a U3 LTR oligonucleotide substrate that contained a Br(dC) substitution within the integrase recognition sequence as shown in Fig. 2. Following binding at 4 "C for 30 min, the complexes were photo-cross-linked for 5 min at 300 nm using a UV transilluminator (Fotodyne W 300). The complexes were precipitated with trichloroacetate and analyzed on 10% Tricine gels using the buffers and protocol provided by the manufacturer (NOVEX, Encinitas, CA).  (29) reported t h a t the RSV integrase turns over catalytically in both the endonucleolytic and DNA strand-transfer reactions (29), suggesting that the enzyme-substrate complex may be metastable and not readily detected when integrase is enzymatically active.

H N -1 Integrase Forms Discrete Complexes with LTR
The catalytic activity of integrase displays an absolute requirement for divalent cations; Mg2+ or Mn2+ must be available for both the endonucleolytic and integrative functions of the enzyme (18,25). Bacterial type I1 restriction enzymes are also specific endonucleases that require divalent cations for activity. I n the absence of cation, although these enzymes are catalytically inert, they form stable, high affinity complexes with oligonucleotide substrates (30). By analogy to restriction enzymes, an attempt was made to produce stable complexes between the HIV-1 integrase and substrate oligonucleotides by eliminating divalent cations from the reaction.
The results of DNA binding assays performed in the presence or absence of divalent cations as analyzed by gel mobility shift are presented in Fig. 3. In the absence of divalent cation, the addition of increasing concentrations of integrase to the binding reaction resulted in the formation of a single complex with significantly reduced mobility (Fig. 3A, lanes 2-5). However, when either Mn2+ or M 2 + was included in the reaction, the amount of this discrete complex was substantially reduced, at some concentrations of integrase by greater than 100-fold ( Fig.  3B, also compare lanes 2-5 with lanes 6-9 and 10-13 in Fig.  3A). The addition of either Mn2+ or Mg2+ shifted this discrete complex to a large form(s) that did not resolve during electrophoresis.
The decrease in the discrete integrase-substrate complex in the presence of divalent cation did not result from a loss of B.
bindable substrate due to processing by the enzyme. Minimal catalysis occurs under the conditions used for the DNA binding reaction (i.e. a t 4 "C), and analysis of the substrate DNA isolated from either the discrete complex assembled in the absence of divalent cation or the undefined complexes formed in the presence of Mn2+ confirmed that the integrase-associated LTR sequences were neither processed nor integrated (data not shown). Moreover, integrase binds to LTR oligonucleotides with 3"recessed termini with an efficiency similar to unprocessed substrates (data not shown).
Although the preparation of recombinant integrase used in these studies was greater than 90% pure (Fig. l), it was possible that the observed retarded complex was the result of a n interaction between the LTR oligonucleotide and a low abundance but high affinity contaminant protein. To address this concern, antibodies elicited by peptides representing either the amino or carboxyl terminus of integrase were added to the binding reaction prior to analysis. As shown in Fig. 4, these antibodies both reduced the amount of the complex and further decreased its electrophoretic mobility. In contrast, preimmune sera had no effect (Fig. 4B, compare lanes 1 and 2 with lanes 3  and 4 ) . These data demonstrate that the complex detected in the gel mobility shift assay resulted specifically from the interaction of integrase with the LTR oligonucleotide and was not due to a contaminating protein.

Substrate Recognition Parallels Substrate Utilization-To
determine if the discrete integrase-substrate complex formed in the absence of M$+ or Mn2+ actually reflected relevant substrate recognition of viral DNA sequences by the enzyme, competition studies were performed using both an oligonucleotide representing wild-type U5 LTR sequence and a mutant U5 viral LTR oligonucleotide (Fig. 2) containing substitutions that significantly reduce its ability to be processed by integrase (11).
The ability of these oligonucleotides to compete with the radiolabeled U3 LTR substrate for integrase binding was assessed in the gel mobility shift assay. The results of these studies showed that the wild-type U5 LTR competed effectively with the U3 LTR for binding, while the mutant U5 sequence was significantly reduced (approximately 9-fold) in its ability to compete for binding to the wild type sequence (Fig. 4). It had been shown previously that the U5 LTFt is indistinguishable from the U3 LTR as a substrate for integrase-mediated cleavage, while the mutant LTR sequence is cleaved 20-fold less effciently (11). Therefore, in the gel mobility shift assay, the ability of the integrase to bind a specific substrate DNA sequence and form a discrete complex appears to correlate with the sequence-specific recognition required for enzymatic function.
Zntegrase Tetramers Bind to the LTR-Analysis of the relationship between integrase concentration and integrase-LTR substrate complex formation (Fig. 3B) suggested that the binding of the enzyme to the substrate is nonlinearly dependent on integrase concentration. Therefore, binding to the LTR may require a multimeric form of the enzyme. Both the endonucleolytic and DNA strand-transfer activities of the RSV integrase also display a distinct nonlinear dependence on enzyme concentration (29). Sedimentation studies have shown that the RSV enzyme exists in equilibrium among monomeric, dimeric, and tetrameric forms, with a K d for multimerization of about 1.0 (29). If the HJY-1 integrase is similar to the RSV protein in this regard, several multimeric forms of the HIV-1 enzyme should be present in the DNA binding reactions at the enzyme concentrations used. However, since only a single complex between the integrase and the LTR oligonucleotide was detected in the gel mobility shift assay (Figs. 3,4, and 5), only one form of the HJY-1 enzyme may be competent to bind the viral DNA.
To determine which multimeric form of the integrase is bound to the LTR substrate, photo-cross-linking studies were performed using a U3 LTR oligonucleotide that included a W photoactivatable group (Br(dC)) within the integrase recogni-Binding by H N -1 Integrase tion sequence (Fig. 2). The substitution affected neither the ability of the oligonucleotide to bind nor to be catalyzed by the integrase (data not shown). DNA binding reactions were performed using the radiolabeled photoreactive substrate, and the resulting protein-DNA complexes were UV-photo-cross-linked and analyzed by SDS-PAGE.
As was seen previously in the gel mobility shift experiments, in the absence of divalent cation, integrase bound and crosslinked to the Br-substituted substrate as a single complex, CI (Fig. 6 A , lanes 2 and 3). In contrast, in the presence of MnC12, three cross-linked complexes were detected, CI as well as CII and CIII (Fig. 6, lanes 5 and 6). The complexes, CI, CII, and CIII exhibited apparent molecular masses of 141, 70.8, and 35.5 kDa, respectively (Fig. 6B). In the gel system used for the analysis, monomeric integrase migrates as a 33-kDa species. Therefore, the apparent size of the three cross-linked complexes were compatible with monomeric (35.5 kDa), dimeric (70.8 kDa), and tetrameric (132 kDa) forms of the enzyme plus an additional contribution in mass by the linked oligonucleotide.
The LTR Binding and Catalytic Activities of Zntegrase Are Distinct-Having developed assays competent to assess both the DNA binding and multimerization properties of integrase, the functional lesions of integrase mutants previously shown to be deficient in catalytic activity can be analyzed. Mutations within the highly conserved regions of integrase, the putative zinc finger and the "DD(35)E" core, are of particular interest. Mutations in either of these conserved motifs adversely affect the specific endonuclease andor integrative properties of the enzyme (16, 17,[23][24][25]. However, while mutations in the DD(35)E motif affect both the "specific" and "nonspecific" catalytic activities of integrase, mutations in the zinc finger appear to influence only those activities of integrase which involve viral DNA sequences, i.e. cleavage of the LTR and integration.
We used the gel retardation and cross-linking assays to investigate the DNA binding phenotype of two integrase mutants, C43S and V151E,D152Q whose catalytic properties were described previously (11). The former mutant involves a consewed residue within the zinc finger, while the latter is a double substitution of highly conserved amino acids in the central core. As detailed elsewhere, these mutant enzymes were expressed and purified equivalently to the wild-type protein, suggesting that the mutations did not result in a gross structural alteration of the enzyme (11). All proteins were purified to greater than 90% homogeneity and were used at identical concentrations (Fig. 1).
Examination of the LTR DNA binding activity of the V151E,D152Q mutant by gel retardation demonstrated that the binding and catalytic functions of the integrase are separable. Although the mutant protein is enzymatically nonfunctional, it exhibited wild-type LTR substrate binding activity  60 m (lams 1-3 and 4-6, respectively). B shows the quantification and molecular weight determination from the results in A. (Fig. 7). Moreover, since the electrophoretic mobility of the mutant integrase complex was indistinguishable from the wild type integrase complex, the V151E,D152Q mutant and wildtype enzyme complexes assume the same multimeric state. As expected, UV-photo-cross-linking of the mutant protein to the Br(dC)-substituted LTR oligonucleotide showed that, like the wild-type enzyme, the mutant integrase forms a tetrameric complex with the oligonucleotide substrate (Fig. 8). Therefore, the inability of the mutant to express catalytic activity is not due to a defect in either appropriate substrate binding or multimer assembly.
The Zinc Finger of Zntegrase Mediates Specific Interaction with the Viral LTR Substrate-In contrast to the V151E,D152Q enzyme which is enzymatically inert, an integrase mutant containing a substitution (Cys + Ser) within the enzyme's zinc finger domain a t residue position 43, described by LaFemina et al. (17), exhibits normal levels of nonspecific endonucleolytic activity, but its ability to mediate specific removal of the terminal LTR dinucleotide is reduced by 90% (17). This mutant enzyme also exhibits significantly reduced DNA strand-transfer activity, a t least in part due to the inability to specifically process the integration substrate. As demonstrated both by the DNA binding assay shown in Fig. 7 and by the UV-photo-crosslinking experiment shown in Fig. 8, this mutant enzyme dis- played a greatly reduced capability to bind the substrate oligonucleotide. These data demonstrated a correlation between zinc finger-mediated sequence-specific recognition of the viral LTR substrate and specific endonucleolytic and DNA strandtransfer activities.

DISCUSSION
Two novel assay systems were developed to study the interaction between the HIV-1 integrase and its viral LTR substrate. The assays, performed in the absence of divalent cations using purified enzyme and oligonucleotides representing the U3 and U5 LTR ends, showed that integrase binds the LTR substrate forming a discrete complex in which the enzyme exists as a tetramer. Formation of the complex correlated with sequencespecific recognition of the substrate required for 3' end-processing activity.
This is the first report to demonstrate defined complexes between the HIV-1 integrase and viral LTR oligonucleotide substrates. Previous attempts to study LTR binding were typically performed under conditions favoring enzyme catalysis (12,21).
As a result, the stable formation of complexes may have been limited by normal enzymatic turnover. The assay described here minimized turnover by eliminating the divalent cations required for enzyme activity. The only other report of sequencespecific DNA binding by a retroviral integrase was published by Krogstad and Champoux (31) using the Moloney murine leukemia virus enzyme. These experiments were also performed under experimental conditions not favorable to enzymatic function.
In addition to limiting the enzymatic activity during binding, the concentration of labeled LTR substrate oligonucleotide used in the present studies was also lowered significantly (approximately 10-fold) relative to the concentrations reported for previously described assays (11,121. Kinetic analyses of DNAbinding suggest that the K, for the specific LTR substrate is 5.0 nM, while the K , for nonspecific DNA substrates is 6-to 10-fold higher.2 Therefore, as the concentration of the LTR substrate oligonucleotide approaches the K , for nonspecific binding, the difference between the two modes of enzyme-substrate interaction is diminished. Since most published assays rely on the LTR substrate for both the specific and nonspecific endonucleolytic events required for strand transfer, conditions have been optimized such that the concentration of the LTR oligonucleotide is not limiting as a nonspecific substrate for integration. In contrast, the lower concentration of the substrate used in the present assay should favor the specific endonucleolytic event. Lower LTR substrate concentrations do in fact promote the generation of the specific cleavage product while, at higher concentrations, a significant proportion of the cleavage products are nonspecific. 2 We have demonstrated that, under specific conditions, integrase forms a discrete tetrameric complex with the LTR substrate oligonucleotide. Previous biophysical studies have shown that retroviral integrases exist in solution as monomers and dimers, as well as tetramers (5, 18,291. The DNA binding and photo-cross-linking experiments reported here demonstrated that in the absence of divalent cation, a single oligomeric species of integrase interacts with LTR substrate oligonucleotides, suggesting that this species, the tetrameric form of the enzyme, is functional. However, this observation is limited by the substrate used and may apply only to the enzyme's specific endonucleolytic activity. It may not extend to DNA strand-transfer which requires integrase interaction with more than one DNA molecule. The assembly of the DNA strand-transfer complex would likely be mediated through protein-protein interaction between integrase oligomers bound to the respective donor and target DNA substrates. Since the largest multimer of the integrase identified through biophysical methods is the tetramer, it has been suggested that this form is also probably involved in the strand-transfer reaction (29). It should be noted, however, that both the sedimentation and gel filtration studies previously published were performed in high salt and in the absence of divalent cation (18,29). Since high salt inhibits enzymatic activity (18) and either Mn2+ or Mg2' are absolutely required for catalysis, it is possible that these higher order interactions can only be observed under appropriate buffer conditions. Alternately, higher order interactions may require assembly on the appropriate substrate(s). Our observation that the addition of Mn2+ or Mg2+ promotes the formation of large nucleoprotein complexes composed of monomeric, dimeric, and/or tetrameric forms of integrase bound to DNA suggests that the integration complex may require interactions beyond the tetramer. Whether these interactions require prior binding to the appropriate substrates is currently under investigation.
We used the DNA binding assay to show that a mutant form of the HIV-1 integrase, containing a substitution within the enzyme's DD(35)E motif, was indistinguishable from the wildtype protein in its ability to form a specific complex with the LTR substrate. Nonetheless, the mutant enzyme is deficient for all catalytic activity (17) as are other enzymes with amino acid substitutions in this motif (24). Hence, the DD(35)E motif, which is highly conserved among retroviral integrases, is essential for integrase catalytic function, but does not participate in substrate recognition.

Substrate Binding by H N -1 Integrase
Finally, the DNA binding assay was also used to demonstrate that a HIV-1 integrase mutant containing a substitution within the enzyme's zinc finger is deficient in its ability to form complex with the LTR substrate. This deficiency was consistent with the mutant's inability to catalyze the specific endonuclease reaction (17). However, the nonspecific endonucleolytic activity of this enzyme is not impaired (17), as is the nonspecific DNA binding of such zinc finger mutants (21). Therefore, the zinc finger of integrase is functionally analogous to the zinc finger of the TFIIIA transcription factor, in which zinc is essential for the specific interaction of the factor with the 5 S RNA gene but not for nonspecific interactions with DNA (32). Most sequence-specific DNA binding proteins in which zinc fingers have been identified function as multimers and/or have multiple zinc fingers (33). Studies involving deletion or mutation of individual zinc fingers in multiple-finger proteins have shown that more than one finger is needed to maintain the requisite number of correct base contacts (33). Therefore, the specificity and affinity of protein-DNA binding is probably attained through the cooperative effect of multiple zinc finger interactions. Given that the HIV-1 integrase monomer has only one zinc finger, multimerization of the enzyme may be essential for sequence-specific DNA binding.