Identifying and Characterizing a Functional HIV-1 Reverse Transcriptase-binding Site on Integrase*

Integrase (IN) from human immunodeficiency virus, type 1 (HIV-1) exerts pleiotropic effects in the viral replication cycle. Besides integration, IN mutations can impact nuclear import, viral maturation, and reverse transcription. IN and reverse transcriptase (RT) interact in vitro, and the IN C-terminal domain (CTD) is both necessary and sufficient for binding RT. We used nuclear magnetic resonance spectroscopy to identify a putative RT-binding surface on the IN CTD, and surface plasmon resonance to obtain kinetic parameters and the binding affinity for the IN-RT interaction. An IN K258A substitution that disrupts reverse transcription in infected cells is located at the putative RT-binding surface, and we found that this substitution substantially weakens IN CTD-RT interactions. We also identified two additional IN amino acid substitutions located at the putative RT-binding surface (W243E and V250E) that significantly impair viral replication in tissue culture. These results strengthen the notion that IN-RT interactions are biologically relevant during HIV-1 replication and also provide insights into this interaction at the molecular level.

A key step in the replication cycle of human immunodeficiency virus, type 1 (HIV-1) 2 involves the synthesis of doublestranded DNA copies of the diploid, single-stranded RNA genome by viral reverse transcriptase (RT) within the cytoplasm of infected cells. This newly synthesized viral DNA becomes an integral component of a large nucleoprotein complex termed the preintegration complex (PIC). PIC-associated proteins include HIV-1 matrix, Vpr, RT, and integrase (IN) (1)(2)(3)(4), as well as cellular components such as lens epitheliumderived growth factor p75 (5), barrier-to-autointegration factor (6), and the high mobility group I(Y) protein (7). The PIC traverses the nuclear envelope, and within the nucleus the viral DNA is subsequently inserted into the chromosomes of the infected host, an event specifically catalyzed by IN.
HIV-1 IN and RT (a heterodimeric enzyme of 66-and 51-kDa subunits) are initially synthesized as part of the Gag-Pol polyprotein, which is subsequently processed by HIV-1 protease during viral maturation to produce active IN and RT. IN (288 amino acids; 32 kDa) comprises three domains: an N-terminal zinc-binding domain (residues 1-50), a catalytic core domain (residues 51-212), and a C-terminal domain (CTD; residues 213-288) that binds DNA nonspecifically (reviewed in Ref. 8). IN has multiple effects throughout the viral life cycle, as perturbations within the IN coding sequence of the pol gene can impact not only integration but also PIC nuclear import, proper virion maturation, and reverse transcription (9 -20).
Purified HIV-1 IN and RT have been found to physically interact in co-immunoprecipitation, glutathione S-transferasebased pulldown, and surface plasmon resonance (SPR) experiments (15,18,21,22), and assays with IN deletion constructs showed the IN CTD is both necessary and sufficient for this association (18,21). The significance of this IN-RT interaction in the context of viral replication has not yet been firmly established. In vitro studies with purified IN and RT, however, have shown that HIV-1 IN can stimulate both the initiation and elongation modes of RT-catalyzed reverse transcription by enhancing RT processivity (23).
To aid our ongoing studies of functional interactions between IN and RT during HIV-1 replication, we have characterized in detail the RT-binding surface on the IN CTD using nuclear magnetic resonance (NMR) spectroscopy, identifying particular IN CTD amino acids that form contacts with RT through chemical shift perturbation experiments. In addition, we have assessed kinetic parameters and binding constants for both IN CTD-RT and IN-RT interactions using SPR. This experimental approach allows us to define and quantify the IN-RT interaction to a level that is not possible through coimmunoprecipitation or other fusion protein pulldown techniques.
With the intention of perturbing IN-RT interactions, we introduced an alanine substitution for lysine at position 258 (K258A) within the IN CTD and subsequently used SPR to characterize the effect of this substitution upon IN⅐RT complex formation. Lys 258 in IN lies at the putative RT-binding site identified here by NMR, and the IN K258A substitution has been previously shown to significantly impair reverse transcription in infected cells (20). We found that the K258A substitution substantially weakened IN CTD-RT interactions. Finally, we introduced various IN amino acid substitutions at the putative RT-binding surface into the NL4-3 viral clone, and replication of viruses containing these substitutions was significantly impaired in tissue culture assays. Collectively, the observations presented here suggest that IN-RT contacts are biologically relevant during HIV-1 replication and that an IN CTD⅐RT complex may provide a useful predictive model that can guide studies of IN-RT interactions in the viral life cycle. All IN constructs were expressed in CodonPlus Escherichia coli cells (Stratagene) and were purified using nickel-nitrilotriacetic acid immobilized metal affinity chromatography. RT was expressed in E. coli cells (M15::pDMI.1 strain) using a plasmid (p6HRT-PROT) with dual inducible expression cassettes for (6 ϫ His tag)-p66 and untagged HIV-1 protease. In this expression system, partial cleavage of the p66 product by protease generates the His-tagged p66/p51 RT heterodimer (26). RT was purified using Ni 2ϩ -nitrilotriacetic acid immobilized metal affinity chromatography followed by cation exchange chromatography, and SDS-PAGE analysis confirmed the presence of highly (Ͼ95%) pure RT heterodimer. Protein concentration was measured by Bradford assay (Bio-Rad) using bovine serum albumin (BSA) as a standard.

EXPERIMENTAL PROCEDURES
Virus Preparation and Replication Kinetics-Generation of the HIV-1 NL4-3 molecular clones (27) bearing the IN single amino acid substitutions W243E and V250E were prepared by first introducing the desired substitutions into pSK-SC001, a pBluescript-derived subcloning vector containing an engineered AgeI site and the 2.3-kilobase pair AgeI-EcoRI fragment of NL4-3 (nucleotide positions 3486 -5744), which encompasses the IN gene. Mutagenic primers and a PCR-based sitedirected mutagenesis protocol (25) were used to introduce W243E and V250E substitutions into the IN coding sequence in pSK-SC001. The modified AgeI-EcoRI fragment was then excised from the subcloning vector and ligated into NL4-3 that had been sequentially digested with AgeI and then EcoRI. Top10 cells (Invitrogen) were transformed with the ligation product, plated on LB agar containing 80 g/ml ampicillin, and incubated at 30°C. The presence of the desired mutation within the IN gene was confirmed by sequencing.
Virus stocks were prepared by transfecting 293T cells grown at 37°C in Dulbecco modified Eagle's medium containing L-glutamine, sodium pyruvate, and 4.5 g/liter glucose (Invitrogen) supplemented with 10% fetal bovine serum (HyClone), 100 IU/ml penicillin G, and 100 g/ml streptomycin. Transfections were performed in 75-cm 2 flasks with PolyFect Transfection Reagent (Qiagen) using 8 g of plasmid DNA bearing the wildtype or mutant HIV-1 clone. Medium containing virus was collected 36 -48 h later and gravity-filtered through 0.45-m cellulose acetate filters, and the levels of p24 antigen contained in the viral supernatant were determined by enzyme-linked immunosorbent assay (PerkinElmer).
One hundred nanograms of p24 equivalent of viral stocks were added to 2 ϫ 10 6 CEM cells grown at 37°C in 2 ml of RPMI 1640 (Invitrogen) having 100 IU/ml penicillin G, 100 g/ml streptomycin, and 10% fetal bovine serum. The medium was removed 12 h postinfection and replaced with 3 ml of fresh medium. Aliquots of the culture medium were then collected for p24 analysis every 2 days after the infection, with the cells being split 1:3 following these collections of viral supernatant. Mock infections were performed using wild-type virus that had been heat-inactivated by treatment at 95°C for 4 min.
NMR Experiments with 15 (29) were also employed in our NMR experiments. Prior to performing titrations with RT, a preliminary spectrum of free 15 N-labeled IN 220 -270 was collected for the purpose of establishing peak assignments, which were made by comparison of peak positions with the NMR data archived at the Protein Data Bank (Protein Data Bank code 1IHV). Structural representations for IN 220 -270 were produced using the program MacPyMOL (30). The cross-linking reaction was quenched by injecting 1 M ethanolamine (pH 8.5) across both flow cells. Concentrations of 500, 250, 125, 62.5, 31.3, 15.6, 7.8, 3.9, and 0 nM (serial 2-fold dilutions) of the various IN constructs were injected across both flow cells. All of these experiments at a given concentration were performed in triplicate, and double-referencing (i.e. both solvent subtraction and correction for nonspecific binding to the unmodified reference flow cell) was employed to produce signals resulting only from IN-RT interactions. All of the SPR experiments were performed at 25°C in a running buffer containing 20 mM Hepes (pH 7.5), 300 mM NaCl, 0.05% Tween 20, and 1 mM ␤-mercaptoethanol (SPR buffer) using a 50 l/min flow rate. Regeneration of the chip surface between individual experiments was accomplished using a 120 s pulse of 10 mM glycine (pH 2.5) at a flow rate of 10 l/min. In addition, the running buffers for experiments with IN and IN 220 -270 contained 1% and 0.011% glycerol, respectively, to match residual glycerol levels in the samples that were injected across the chip surface. The association and dissociation phases of all of these experiments were 240 and 300 s, respectively. Similar experiments using 0, 0.125, 0.25, 0.5, and 1 M BSA were performed in duplicate as negative controls. Assessment of the validity of 1:1 or two-state conformational exchange binding models, as well as the calculation of binding parameters, was performed using Biacore T100 Evaluation software (Biacore). Residual 2 values are a measure of the average squared difference between experimental data and fitted models and were calculated using the formula 2 ϭ ⌺(RU exp Ϫ RU fitted ) 2 /n, where the summation takes place over n time points in the sensorgrams, and RU exp and RU fitted are the experimental and fitted response unit values, respectively, at a given time point.
Analytical Ultracentrifugation-The oligomeric state of IN 220 -270 in both NMR and SPR buffers was assessed by sedimentation equilibrium analysis with a Beckman XL-A analytical ultracentrifuge using absorption optics at 280 nm. IN 220 -270 in NMR buffer was analyzed at a concentration of 150 M at 20°C in a 3-mm-path length double sector cell and at a concentration of 93 M at 25°C in a 12-mm path length six-sector cell. Absorbance values in this data set were corrected for differences in cell path length, and partial specific volumes for ensuing calculations were derived from the IN 220 -270 amino acid composition and corrected for temperature (v 25°C ϭ 0.734 ml/g and v 20°C ϭ 0.732 ml/g) (31,32). IN 220 -270 was analyzed in SPR buffer at concentrations of 150, 93, and 31 M at 25°C in a 12-mm-path length sixsector cell. Sedimentation equilibrium profiles for IN 220 -270 in both NMR and SPR buffers were measured at speeds of 20,000, 24,000, and 30,000 rpm. For the two buffer conditions, the absorbance profiles at each speed were initially fitted by a nonlinear least squares algorithm within Beckman Origin-based software (version 3.01) to an exponential curve corresponding to a single ideal species, yielding weight-averaged molecular weight values for the IN 220 -270 sample. The data set encompassing the entire collection of scans corresponding to different rotor speeds and protein loading concentrations were globally fitted to a monomer-dimer equilibrium model (using the "multifit" option) to obtain a dimer dissociation constant for each buffer condition.
Higher order association models ultimately resulted in poorer fits and thus were not analyzed further.  the peaks in the acquired spectrum became broadened to the noise level (data not shown), and so no further titrations were performed.

NMR Spectroscopy to Map the RT-binding
Amide backbone peaks associated with nine residues showed significant decreases in intensity during the titration, as compared with most other peaks in the spectra: Arg 231 , Leu 242 , Trp 243 , Gly 247 , Ala 248 , Val 250 , Ile 251 , Gln 252 , and Lys 258 (Fig. 1). With the exception of Arg 231 , these residues mapped to one side of the IN 220 -270 structure, thus identifying a putative RT-binding surface (Fig. 2). Arg 231 is located within a stretch of amino acids connecting strands ␤1 and ␤2 in the NMR structure (29) and is positioned close to the putative RT-binding surface. The signal decreases observed during the titration suggest that these nine residues experience intermediate exchange on the NMR timescale (i.e. chemical exchange occurs at a frequency that is on the order of the difference between free and bound amide group chemical shift) (33,34). Finally, the selective broadening of the peaks for the above nine amino acids indicates that binding of IN 220 -270 to RT is specific, as nonspecific binding would produce uniform changes across all of the peaks in the spectra. were performed. RT was immobilized via lysine residues to a CM5 chip using amine coupling chemistry. The crystal structure for RT (Protein Data Bank code 1HMV) shows that lysines lie throughout the RT surface; thus the cross-linking method is anticipated to attach RT to the chip surface in a random orientation. As such, any putative IN-binding site on RT is expected to be accessible for some percentage of the immobilized RT molecules. The p66/p51 heterodimer exhibits long term stability under the conditions employed for SPR experiments (35), and therefore the RT heterodimer species (rather than isolated p66 and p51 subunits) are anticipated to be immobilized on the biosensor chip surface.
Purified IN or IN 220 -270 was passed over the immobilized RT at concentrations of 0 -500 nM, and sensorgrams were recorded. The sensorgrams clearly show that both IN and IN 220 -270 bound immobilized RT under these experimental conditions (Fig. 3). Analogous experiments using BSA as a negative control indicated no binding to RT (Fig. 3), and BSA present in 10-fold molar excess of IN 220 -270 did not influence IN 220 -270 binding to RT (data not shown), indicating that IN⅐RT complex formation is not due to nonspecific protein-protein interactions.
A "two-state conformational exchange" binding model, where an initial bimolecular encounter to form a nascent species is followed by a conformational transition to a stable complex, has been recently described for the IN-RT interaction (22), and this model was fitted to the sensorgram data to extract binding parameters. The quality of the fit for the two-state conformational exchange model was assessed by a residual 2 value, a parameter that measures the difference between the fitted model and the experimental data. A simple bimolecular "1:1 binding" model was also fitted to the sensorgram data for comparison purposes.    Table 1). Both the kinetic parameters and the dissociation constant for the full-length IN-RT interaction were in good agreement with previously published measurements (22).
Although the dissociation constants for IN and IN 220 -270 binding to RT were similar, two of the kinetic parameters found for these two systems differed: both the initial on-rate and offrate values (k a1 and k d1 , respectively) for IN binding to RT were ϳ10-fold lower than those found for IN 220 -270 ( Fig. 4 and Table 1 (20). Furthermore, a K258A substitution has been previously shown to significantly impair reverse transcription in infected cells (20) IN 220 -270). The difference in binding affinities stems mostly from changes in k a1 (Fig. 4 and Table 1). As such, initial association of IN 220 -270 with RT may be mediated by electrostatic attraction, because removal of positive charge from the putative RT-binding surface adversely affected the on-rate for complex formation. The binding data show that the same IN K258A substitution that impairs reverse transcription (20) also results in significantly weaker IN-RT binding interactions, indicating that productive IN-RT interactions may be important for efficient reverse transcription during HIV-1 replication.

NL4-3 Viral Clones Bearing IN W243E and V250E Substitutions Are Replication-defective-Mutant viral clones encoding
IN amino acid substitutions W243E and V250E were constructed to test whether such substitutions at the putative RTbinding surface significantly impair viral replication in tissue culture. Viruses prepared by transient transfection in 293T cells were used to infect CEM cells, and the infectivity was monitored by measuring levels of p24 antigen in culture media every 2 days following infection. Both of the IN substitutions resulted in impaired HIV-1 replication (Fig. 5). The p24 levels for wildtype virus peaked after 6 days (ϳ600 ng/ml), whereas p24 levels for the mutant viruses and the heat-inactivated wild-type virus never exceeded 40 ng/ml during the course of the experiment. These observations show that nonconservative substitutions at the putative IN-RT interface are not well tolerated by replicating virus and suggest that such substitutions may inhibit viral replication by disrupting important IN-RT contacts.

DISCUSSION
Although the IN CTD has been long known to bind to DNA (36,37), additional studies indicate that this domain also promotes binding to various cellular proteins. Such proteins include the survival of motor neuron complex component Gemin2 (38), the nuclear import receptor importin 7 (39), the cytidine deaminase APOBEC3G (40), the human homolog of the mouse EED (embryonic ectoderm development) gene product (41), and the histone acetyl transferase p300 (42). The diverse array of IN CTD binding partners suggests that the IN CTD is a versatile protein component that can mediate a variety of physiological processes during HIV-1 replication.
Previous studies have also shown that the IN CTD is both necessary and sufficient for binding to RT (18,21), and both the NMR and SPR data presented in the current study further confirm these reports. In addition, Hehl et al. (21) identified regions within RT involved in forming IN contacts, demonstrating that either the p66 or the p51 subunit from HIV-1 RT is sufficient for binding HIV-1 IN. Additional experiments with RT deletion constructs narrowed the IN-interacting residues on RT to a bipartite region encompassing residues 1-242 in the fingers-palm domain and residues 387-422 in the connection domain. In vitro IN-RT interactions in murine leukemia virus have also been identified (43) but are less well characterized than in HIV-1.
The role of IN-RT interactions during viral replication is currently unknown. Nonetheless, diverse reports collectively suggest that IN and RT may be functionally intertwined, not only in HIV-1 but in other retroviruses. For example, both RT and RT-derived peptides can inhibit the 3Ј-end processing and strand transfer activity of IN in vitro (44 -46). In vitro assays also show that HIV-2 IN is inhibited by HIV-2 RT and that HIV-1 IN is inhibited by heterologous RTs from HIV-2 or murine leukemia virus (45). As such, the possibility that RT may serve as a viral cofactor that prevents suicidal autointegration events in retroviruses has been proposed (45,46). Furthermore, in avian sarcoma leukosis virus IN exhibits activity not only in its free form as a 32-kDa phosphoprotein, but also as part of a larger ␤ subunit that includes RT (47). The overt tethering of IN with RT in this case suggests that a tight IN-RT linkage may be an important feature for retroviral replication.
In addition, genetic studies have indicated that IN is important for reverse transcription during both HIV-1 and murine leukemia virus replication (15,48) and have identified a number of HIV-1 IN mutant viruses that are impaired in viral cDNA synthesis (9, 14 -20). Furthermore, continuous passage of a replication-impaired chimeric HIV-1 viral clone bearing the HIV-2 IN gene ultimately yielded compensating mutations within the IN and RT genes, resulting in increased synthesis of viral DNA and improved viral fitness (49). A recent study from our laboratory has shown that IN directly stimulates both the initiation and elongation modes of reverse transcription in vitro and that the processivity of RT in the presence of IN is enhanced. Such findings suggest that direct physical interactions between IN and RT may facilitate reverse transcription in vivo (23).
In all, the above studies suggest that IN and RT may exert mutual effects upon one another's activity during viral replication. Whether such effects may always proceed through direct IN-RT interactions during infection is currently unclear. For example, some IN mutations that impact reverse transcription in infected cells could conceivably act indirectly by perturbing other steps in the viral life cycle prior to reverse transcription (e.g. uncoating). IN mutations may also affect reverse transcription indirectly through altered interactions with cellular factors. One such cellular factor may be Gemin2, a host cell protein that binds IN and that also appears to play a role in reverse transcription (38).
Nonetheless, one straightforward explanation for the various findings described above is that IN and RT influence each other's activity through direct IN-RT contacts. The overall aim of this study is to investigate whether physical interactions between IN and RT are functionally relevant. A necessary step in achieving this aim is pinpointing the particular amino acids IN 220 -270 was modeled as a homodimer in the NMR structure (29). Complementation studies indicate that IN functions as a multimer (50 -52), and various models describing the IN oligomeric state have been proposed (53)(54)(55)(56)(57)(58) (other than the homodimer interface) were identified by the NMR experiments. Therefore, in this alternative binding model, a single amino acid (Arg 231 ) must mediate the entire interaction with the 117-kDa RT heterodimer. Although formally possible, such an event seems highly unlikely to provide sufficient binding energy to form a stable protein⅐protein complex. Altogether, the NMR data and structural analysis favor a sequence of events where (i) dimeric IN 220 -270 first dissociates into monomers, and then (ii) RT forms a complex with the monomeric form of IN 220 -270 at the homodimer interface, further shifting the IN 220 -270 monomer-dimer equilibrium toward the monomeric species. Whether a comparable mechanism describes the interaction between RT and full-length IN is unclear, because dimerization of the IN CTD in the context of full-length IN has not yet been established.
Our SPR data indicate that both IN and IN 220 -270 bind to RT with nanomolar affinity and are consistent with a two-state reaction process involving a conformational change step. Indeed, conformational flexibility is a prominent feature of RT recognition and binding of other substrates. Comparison of the x-ray crystal structures of unliganded RT (60) and complexes containing either nevirapine (61) or duplex DNA (62,63) show significant structural differences between free and substratebound enzyme. In particular, the thumb subdomain within p66 undergoes a substantial (Ͼ30°) rotation upon complex formation, and the p66 fingers and palm subdomains reorient as well. Shifts in position of the p51 palm subdomain are also noted between DNA-bound and unliganded RT. Molecular dynamics simulations of RT also indicate that the thumb and fingers subdomains in the p66 subunit, as well as the p51 thumb subdomain, show considerable mobility (64). Such mobility could conceivably confer on RT the conformational adaptability needed to recognize IN through a two-step, conformational exchange mechanism.
The sedimentation equilibrium data indicate that the experimental conditions employed in the SPR binding assays strongly favor the monomeric form of IN 220 -270, and previously published biophysical studies indicate that the SPR conditions should favor the monomeric form of full-length IN as well. IN has been previously shown to be monomeric at submicromolar concentrations in the presence of nondenaturing detergents at 25°C (65). Given the concentrations of IN (3-500 nM), temperature (25°C), and the levels of detergent (0.05% Tween 20) used in our SPR assays, we hypothesize that both IN 220 -270 and IN bind RT as a monomer in these SPR experiments.
IN can directly stimulate both RT-mediated synthesis of early viral cDNA products as well as RT processivity in vitro (23). The current study further suggests that direct binding between IN and RT may have functional consequences during viral replication. Previous infectivity studies indicated that HIV-1 harboring a K258A substitution within IN resulted in substantially reduced levels of late reverse transcription products (Ͻ10% that of wild-type virus) and no detectable virus replication in a CD4ϩ T-cell line (20). Our NMR data indicate that Lys 258 forms contacts with RT, and our SPR data indicate that introducing a K258A substitution into IN 220 -270 had a disruptive impact on the dissociation constant by primarily affecting the on-rate for IN⅐RT complex formation. Thus, the IN K258A substitution results in both impaired reverse transcription (20) and significantly weaker IN-RT binding interactions (this study). In addition, our NMR data indicate that IN Val 250 and Trp 243 are involved in RT contacts, and introduction of nonconservative substitutions at these IN positions dramatically reduces viral infectivity. Collectively, the data presented here and by Lu et al. (20)  HIV-1 RT and IN are responsible for catalyzing the essential steps of reverse transcription and integration, respectively, during the early stages of the retroviral life cycle. The physical interactions between RT and IN may serve to ensure that the PIC is assembled properly and can function efficiently. Characterization of the RT-IN interaction and determination of its biological significance may reveal new functional roles for IN, provide a mechanistic basis for the phenotypes observed with certain IN mutants, and identify new potential targets for anti-HIV therapy.