Multimerization determinants reside in both the catalytic core and C terminus of avian sarcoma virus integrase.

We have shown previously that the active form of avian sarcoma virus integrase (ASV IN) is a multimer. In this report we investigate IN multimerization properties by a variety of methods that include size exclusion chromatography, chemical cross-linking, and protein overlay assays. We show that removal of the nonconserved C-terminal region of IN results in a reduced capacity for multimerization, whereas deletion of the first 38 amino acids has little effect on the oligomeric state. Binding of a full-length IN fusion protein to various IN fragments indicates that sequences in both the catalytic core (residues 50-207) and a C-terminal region (residues 201-240) contribute to IN self-association. We also observe that the isolated C-terminal fragment (residues 201-286) is capable of self-association. Finally, a single amino acid substitution in the core domain (S85G) produces a severe defect in multimerization. We conclude from these analyses that both the catalytic core and a region in the nonconserved C terminus are involved in ASV integrase multimerization. These results enhance our understanding of intergrase self-association determinants and define a major role of the C-terminal region of ASV integrase in this process.

The integration of viral DNA into the genome of the host cell is a unique and vital step in the normal life cycle of retroviruses (for recent reviews, see Refs. 1 and 2). The retroviral integrase (IN) 1 is both necessary and sufficient for the integration of a linear DNA with viral ends into a target DNA in vitro (3,4). Our knowledge of retroviral integrase structure and function continues to be refined by identification of domains of the enzyme that contribute to the various functions required during integration (Fig. 1). These include DNA binding, catalytic activities, and the multimerization required for the coordinated joining of processed viral ends to the site of integration in host DNA. Several investigations have mapped a nonspecific DNA binding domain to C-terminal portions of both ASV and HIV-1 IN (5)(6)(7)(8)(9). Catalytic functions have been localized to the central core domain, which is resistant to limited proteolysis (10), and contains a conserved triad of acidic residues [D,D (35)E] that is presumed to bind the divalent cations that are required for catalytic activity (5,11). The isolated catalytic domain is competent to perform a concerted cleavage-ligation activity (12)(13)(14) and contributes to the recognition of conserved CA residues present at the 3Ј ends of retroviral long terminal repeats and other transposable elements (15). However, it cannot perform the viral DNA end processing and joining reactions required for insertion of viral DNA into target DNA sequences. The crystal structures of the catalytic cores of both HIV-1 IN and ASV IN have been solved recently (16,17).
Many enzymes that catalyze DNA recombination require the formation of multimeric protein-DNA complexes. Detailed structural information is available for some of these (18 -22). However, important questions concerning the stoichiometry of IN protein and DNA substrates in the nucleoprotein complex that is competent for integration remain unanswered. In addition, the role of protein multimerization in substrate binding and catalysis is yet to be clearly defined.
Both biochemical and genetic studies indicate that multimerization is a functionally important property of retroviral integrases. ASV IN, purified from viral particles, and bacterially expressed HIV-1 IN were observed to migrate in glycerol gradients at a position consistent with a dimer molecular weight (23,24). Gel filtration studies also suggest that purified HIV-1 IN forms dimers (25). We have used sedimentation and kinetic studies to demonstrate that purified ASV IN, which exhibits a reversible mass action between monomer 7 dimer 7 tetramer, must multimerize to perform its catalytic function (26). Similar sedimentation analyses have been performed with HIV-1 IN (27). The coordinated action of Moloney murine leukemia virus IN on both ends of viral DNA was demonstrated by mutagenesis studies and analysis of intermediates produced in vivo (28). Our laboratory has recently obtained similar results with ASV IN in vitro, using DNA substrates that link two viral DNA end sequences (29). Multimerization has also been inferred from the enzymatic complementation of two defective HIV-1 IN mutants when incubated together (30,31). Finally, a yeast transcriptional reporter system was used to analyze HIV-1 IN homomeric interactions and to determine a minimal domain required for self-association (32).
The work presented here includes physical analyses of the multimeric state of purified ASV IN and various IN fragments using size exclusion chromatography (SEC) and chemical crosslinking techniques. To further delineate the regions of this enzyme that contribute to self-association, we have also employed a modification of a protein overlay blot technique which tests directly for binding between two potential associating proteins. In addition to demonstrating a role for the conserved catalytic core domain in ASV IN dimerization, our results un-* The work was supported in part by National Institutes of Health Grants CA-47486 and CA-06927, a grant for infectious disease research from Bristol-Myers Squibb Foundation, and also by an appropriation from the Commonwealth of Pennsylvania. 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  cover an important determinant for multimerization located in the less conserved C-terminal domain.

Plasmids and Cloning
The cloning of nonfused ASV IN and IN fragments has been described (15). To produce plasmid DNA clones encoding IN fragments with the indicated boundaries fused to glutathione S-transferase (GST), polymerase chain reaction was performed with Vent polymerase (Boehringer) for 25 cycles from the template pRC23p32 (33), using primers with flanking BamHI and EcoRI sites suitable for insertion in the expression vector, pGEX2TK (Pharmacia Biotech Inc.). This expression vector encodes a kinase labeling site at the junction between GST and the cloned insert of the fusion protein. Polymerase chain reaction fragments were purified prior to digestion and ligation; cloning and screening procedures followed standard practice (34).
The IN(201-286)st clone, which expressed the IN fragment 201-286, included a streptavidin epitope tag at the C terminus. It was constructed in three steps as follows. 1) A DNA duplex fragment encoding a kinase/strep-tag sequence (35) of 15 amino acids (Arg-Arg-Ala-Ser-Val-Ser-Ala-Trp-Arg-His-Pro-Gln-Phe-Gly-Gly) with compatible ends was inserted into BamHI/HindIII-digested pSE380 (36). 2) A polymerase chain reaction fragment encoding IN amino acids 201-286 (using primers with flanking NcoI and BamHI sites) was inserted into the pSE380 strep-tag derivative.
3) The IN-strep-tag encoding DNA fragment was purified and ligated to NcoI/HindIII-digested expression vector pET-20b (Novagen). The polypeptide expressed from this construct includes a leader sequence (pelB) from the pET-20b vector, which is intended for periplasmic targeting of overexpressed proteins. However, the most effective purification of IN(201-286)st was from whole cell lysates. The apparent molecular weight from SDS gels indicate that this leader is cleaved by a signal peptidase. The IN sequences of all clones were confirmed by sequencing, and fusion proteins of expected molecular weight were synthesized in all cases.

Protein Expression and Purification
GST Fusion Protein Expression, Purification, and Labeling-An overnight culture of Escherichia coli MC1061 bearing GST-IN-encoding plasmids was diluted 1:20 in Luria Broth medium containing ampicillin and grown at 30°C to an A 600 between 0.8 and 1.0 before expression was induced by addition of isopropyl-1-thio-␤-D-galactopyranoside to a final concentration of 0.2 mM. Growth was continued for 4 h at 30-35°C prior to harvesting cells by centrifugation for 25 min at 5000 ϫ g at 4°C. Cell pellets were frozen and stored at Ϫ70°C. Lysis was accomplished by two passes through a French press at 20,000 p.s.i. in TNE buffer (20 mM Tris-Cl, pH 8, 1 M NaCl, 1 mM EDTA). After lysis, Nonidet P-40 was added to a final concentration of 1% (v/v), and the preparation was incubated with inversion at 4°C for 10 min, prior to clearing by centrifugation at 10,000 ϫ g for 15 min at 4°C. The supernatant fraction was collected and then diluted with an equal volume of 20 mM Tris-Cl, pH 8, 1 mM EDTA, to bring the final NaCl concentration to 0.5 M. A 50% (v/v) slurry of glutathione-agarose (Sigma) was added for affinity purification by a batch procedure. Fusion proteins were incubated with the resin for Ͼ1 h at 5°C prior to washing 5 times with NETN buffer (0.5 M NaCl, 1 mM EDTA, 20 mM Tris-Cl, pH 8, 0.5% Nonidet P-40). GST-fusion proteins could be stored at 4°C immobilized on the glutathione resin (for several weeks) or, alternatively, eluted from the resin with 20 mM glutathione (reduced form, Sigma) in elution buffer (50 mM HEPES, pH 8, 0.4 M NaCl, 0.1 M LiCl, 1 mM EDTA, 0.5% (v/v) Brij-35). Eluted fusion protein preparations were then dialyzed overnight at 5°C against 50 mM HEPES, pH 8, 0.5 M NaCl, 0.1 mM EDTA, 1% (v/v) thiodiglycol, 40% (v/v) glycerol for storage at Ϫ20°C or Ϫ70°C. Throughout the manuscript, GST fusion proteins are denoted with the prefix of a lowercase g, i.e. gIN .
For labeling purposes, glutathione beads with bound fusion protein were washed with kinase buffer (20 mM Tris-Cl, pH 7.5, 0.1 M NaCl, 12 mM MgCl 2 , 10 mM dithiothreitol). Then, 10 to 20 units of protein kinase (catalytic subunit from bovine heart, Sigma) and 330 Ci of [␥-32 P]ATP were added, and the mixture was incubated with agitation for 30 -60 min at 4°C. After quenching, the glutathione resin was washed 4 times with NETN buffer prior to elution of the labeled fusion protein as described above.
Strep-tag Protein Expression and Purification-Induction of expression and cell lysis were performed essentially as described for GSTfusion proteins above. Streptavidin-agarose (Pierce) was added to cleared lysates for affinity purification by a batch procedure. Fusion proteins were exposed to the resin for longer than 1 h at 5°C. The resin was then washed 5 times with 20 mM HEPES, pH 8, 0.5 M NaCl, 1% (v/v) glycerol, 1 mM EDTA. Strep-tag-fusion proteins were eluted from the resin with 1 mM D-biotin in elution buffer. Eluted fusion protein was then dialyzed overnight at 5°C against 50 mM HEPES, pH 8, 0 Expression and Purification of Full-length Nonfused IN and IN Fragments-Full-length ASV IN and various IN fragments were expressed in Escherichia coli MC1061 and purified as described previously (15,26) using an immunoaffinity column as the final purification step.

Chemical Cross-linking
Cross-linking of IN and IN fragments was carried out using a variety of chemical reagents (Pierce) with buffer and protein concentrations noted in the figures. Reactions using the reagents DSP and BS 3 were performed in 100 mM HEPES buffer, pH 8.0. Reactions with the reagent EDC were performed in 100 mM MES, pH 5.6, and 5 mM N-hydroxysuccinimidyl ester. Unless otherwise noted, all reactions contained 600 mM NaCl and were quenched with a molar excess of either glycine or lysine prior to the addition of an equal volume of SDS sample buffer containing 280 mM 2-mercaptoethanol. The 2-mercaptoethanol was omitted if the cross-linker contained a disulfide bond for cleavage, i.e. DSP. Samples were subsequently heated at 95°C for 10 min. Covalently linked multimers were detected by separation in 12% or 15% SDS-polyacrylamide gels and silver staining.

Size Exclusion Chromatography of Full-length and Truncated IN Proteins
Size exclusion chromatography was performed using a Superdex 75HR 10/30 column (Pharmacia) on a Rainin HPLC system with native IN or nonfused IN fragments. A flow rate of 0.5 ml/min with a mobile phase of 50 mM HEPES, pH 7.5, 0.5 M NaCl, 1% (v/v) glycerol was used for all experiments. Typically, 25-50 l of purified IN polypeptides at a concentration of 30 M were injected. Samples were subjected to centrifugation at 10,000 ϫ g prior to injection on the column. Absorbance of the column eluate was monitored at both 280 and 220 nm. Samples from peak fractions were monitored by SDS-PAGE for the presence of the expected protein species. The column was calibrated using seven different globular proteins as molecular weight standards, and the apparent molecular weight of each sample peak was determined using linear regression of the log of known molecular weight versus the elution behavior (K av or elution time).
The SEC analyses presented do not attempt to account for the dissociation kinetics of IN multimers. The observation of discrete monomer and dimer peaks in some chromatograms (for example, Fig. 4, traces 4 and 5) indicates that the dissociation rate of dimer to monomer must be slow relative to the column run time. The intent of the experiments presented here was to compare the behavior of different polypeptides and not to establish absolute quantitative constants for IN self-association. Sedimentation equilibrium analysis is more suited to the quantitative determination of association constants, and the results of such studies (in progress) will be reported separately.

Protein Overlay Binding Assay
To test the ability of a labeled protein probe to bind to a defined set of target IN fragments, standard SDS-PAGE was performed on the target polypeptides prior to electrophoretic transfer to either a polyvinylidene difluoride or nitrocellulose membrane using a 3-buffer semidry technique (according to Millipore recommended procedures). Subsequent steps of blocking, renaturing, and probing were carried out at 4°C. For buffers requiring dithiothreitol, this reagent was added just prior to use. The transfer membrane was blocked in binding buffer (25 mM HEPES-KOH, pH 7.7, 25 mM NaCl, 5 mM MgCl 2 , 1 mM dithiothreitol) containing 5% (w/v) powdered milk and 0.05% (v/v) Nonidet P-40 (blocking solution) for at least 1 h (typically overnight). Target polypeptides were denatured on the membrane by two successive incubations in binding buffer containing 6 M guanidine HCl for 10 min. Slow renaturation was accomplished by 6 successive dilutions with an equal volume of binding buffer, each with a 10 -15-min incubation. After two washes with binding buffer, the membrane was treated once more with block-ing solution for 1 h and for 30 min in blocking solution with only 1% powdered milk. A molar excess of labeled probe protein was diluted into probe buffer (20 mM HEPES, pH 7.7, 75 mM KCl, 0.1 mM EDTA, 2.5 mM MgCl 2 , 1% powdered milk, 0.05% Nonidet P-40, 1 mM dithiothreitol). This probe mixture also included at least a 2-fold molar excess of unlabeled GST protein to block nonspecific binding to the GST-fusion targets. The probe was incubated with the renatured blot for greater than 6 h, followed by three successive 10-min washes in probe buffer alone. The membrane was then dried and the radioactivity quantitated on a Fuji BAS1000 phosphoimaging system. The amount of radioactive probe bound to target bands of interest was normalized to the quantity of the target polypeptide loaded on the gel, as assessed by densitometric quantitation of an identically loaded Coomassie-stained gel.
The experiment in Fig. 6 uses GST-fusion proteins for both probe and targets. GST has been shown to be a dimer in the active state (37). Although we included excess unlabeled, nonfused GST as competitor, this assay may detect some GST-GST interactions between probe and targets. The level of probe binding to GST alone is indicated and should be considered the background for each probe. Note that the binding of the gIN(1-286) probe to the full-length target is greater than 20-fold above binding to GST alone, demonstrating an adequate signal-to-noise ratio under our assay conditions. Experiments were repeated with several of the constructs using nonfused protein targets (data not shown), and similar results were obtained.

Integrase Multimers Revealed by Chemical Cross-linking-
Chemical cross-linking has been employed successfully to examine the protein-protein associations of many multisubunit enzymes. We have used a variety of chemical cross-linking reagents to identify the multimeric forms of ASV IN that exist in solution. As shown in Fig. 2A, titration of the cross-linker DSP revealed dose-dependent formation of covalently linked forms of IN (lanes 2-6). As the concentration of cross-linker was increased, the amount of dimers, tetramers, and higher order multimers increased while the amount of free monomer decreased. The dominant multimer observed was a dimer. Optimal concentrations of DSP were between 1 and 2 mM; higher concentrations of the cross-linker produced aggregates which failed to enter the gel (Fig. 2A, lane 7). The DSP cross-linker contains a disulfide bond which can be reduced, thereby breaking the covalent link between subunits. As expected, treatment with ␤-mercaptoethanol led to the disruption of cross-linked multimers ( Fig. 2A, compare lanes 6 and 8). Fig. 2B shows results of treatment with increasing concentrations of the "zero-length" cross-linker EDC. Due to the short length of the covalent bond formed (the length of a typical amide bond), detection of cross-linked multimers using this reagent (lanes 2-5) provided strong evidence that the subunit association is structurally significant (38,39).
Other reagents with different chemistries (e.g. glutaraldehyde and dimethyl 3,3Ј-dithiobispropionimidate) were also examined for their ability to covalently cross-link IN multimers. Multimeric complexes similar in composition and amount to those observed with DSP and EDC were detected with these reagents (data not shown), suggesting that a variety of reactive residues (basic, acidic, and others) must be present at or near the interface between monomer subunits.
Experiments in which the IN concentration was decreased while the cross-linker concentration was held constant showed that dimers are present at IN concentrations as low as 70 nM (Fig. 2C). These data also reveal a concentration dependence for tetramer formation (Fig. 2C, compare lanes 2 and 3). These results are consistent with previous estimates of a K d (monomer-dimer) in the 1-5 M range (26).
Cross-linking Analysis of Truncated Derivatives of IN-We have previously described a series of N-and C-terminal deletions in ASV IN (15). Two of these, IN(1-207) and IN(52-207), were shown to have lost normal processing and joining activities, but retained endonuclease and cleavage-ligation activity with unimolecular substrates that represent an integration intermediate (15). Another truncated protein, IN(39 -286), which lacks amino acids 5-38 from the N-terminal region, retains both processing and joining activities. 2 The cross-linker BS 3 was used to determine the ability of these three truncated forms of IN to form covalently linked dimers in solution (Fig. 3). The results showed significantly less cross-linked dimer with IN(1-207) (Fig. 3A, lanes 7-9), compared to full-length IN (Fig. 3A, lanes 3-5). At a concentration of 2 mM BS 3 in which all full-length IN was covalently linked in multimeric forms, most of IN(1-207) remained monomeric. We conclude that IN(1-207) is deficient in multimerization. Analysis of IN(52-207), which also lacks amino acids from the N terminus, showed a similar deficiency (Fig. 3B, compare lane 2 with lane 8). In contrast, deletion of residues 5-38 from the N-terminal segment alone did not cause reduction in multimerization (Fig. 3B, compare lane 2 and lane 4).
No heterodimerization was observed in mixtures of fulllength and IN(1-207) polypeptides in this assay (data not shown). It is possible that under these experimental conditions, exchange of monomer subunits proceeds too slowly for heterodimers to form. However, this could also reflect the fact that the affinity of an IN(1-207) subunit for the full-length IN is significantly lower than two full-length IN monomers for each other.
Multimerization Detected by Size Exclusion Chromatography-The oligomeric composition of IN and IN truncated proteins in solution was also investigated by size exclusion chromatography. In these experiments, the properties of full-length and truncated proteins were compared at similar molar concentrations. Under the conditions described for Fig. 4 Incubations were for 15 min at 22°C. Labeling is as described in Fig. 2. In the course of site-directed mutagenesis studies, our laboratory has prepared a number of altered ASV IN proteins that contain single amino acid substitutions in residues that are highly conserved in retroviruses and certain other transposable elements (11). A number of these proteins (D64E, T66A, F126A, D121E, L163A, H9N, K206A, R227A) were examined by SEC (data not shown). Only one of these, S85G, exhibited a significant difference when compared with wild type protein.
The results of these SEC analyses are in general agreement with those of the chemical cross-linking studies. Both sets of data indicate that the catalytic core domain of ASV IN can dimerize, but with reduced efficiency compared to the fulllength protein. Addition of the C-terminal region appears to restore full multimerization capability. We conclude that selfassociation determinants are located in both the core and Cterminal regions of ASV IN. The inability of the S85G mutant to dimerize suggests that substitutions in this residue alter the catalytic core structure, or the way in which the core interacts with the C-terminal domain. Analysis of the crystal structure of ASV IN(52-207) reveals that the side chain of this residue participates in a network of hydrogen bonds in a tight turn between two ␤-strands (17).
Localization of Self-associating IN Domains-In order to map the self-association domains of ASV IN and to gain a better understanding of their relationship to one another, we used deletion mutagenesis coupled with a protein overlay technique. This technique (40, 41) employs a labeled protein to probe a Western blot of target proteins that are first denatured and then renatured. We modified this procedure to investigate the self-association potential of individual regions of ASV IN. For the first experiment, a labeled GST-IN fusion protein, gIN(1-286), was the probe for a series of renatured targets which included full- length IN, IN(1-207), and other non-integrase protein controls. The data showed efficient binding of the probe to full-length IN (Fig. 5B, lane 3) and no detectable binding to the bovine serum albumin and molecular mass standards (Fig. 5B, lanes 1 and 2). Probe binding to full-length IN could be competed by incubation of the blot with unlabeled, full-length IN fusion, but not with the GST portion of the fusion protein alone (data not shown). Therefore, we conclude that the binding reaction is specific for IN-derived polypeptides.
The C-terminal truncation protein showed reduced binding (Fig. 5B, lane 4), even though equal molar amounts of this polypeptide were used in the assay (Fig. 5A, compare lanes 3  and 4). The relative amount of probe bound was quantitated by radioanalytic imaging and normalized to the amount of protein present on the filter. These calculations indicated that the C-terminal truncation protein bound to the probe with 5-to 10-fold lower efficiency than the full-length IN. Since these results were consistent with those from our physical assays of IN multimerization, we used this method to screen a series of nested N-terminal truncated IN proteins. For ease and uniformity of purification, these truncated proteins were constructed as GST-fusion proteins (see Fig. 6 and "Materials and Methods"). The fusion proteins were expressed in E. coli, affinity-purified, and used as targets in the protein overlay assay. The relative binding capacity of these proteins was tested with two probes, full-length gIN(1-286) and gIN(201-236), and quantitated as described above. Fig. 6 (Fig. 6, last column). The C-terminal truncation IN(1-236) bound this probe as well as or better than the full-length IN, but no binding above background was observed to the IN(1-207) target.
It is possible that this method could detect both quaternary interactions between IN monomers and tertiary interactions that reflect the folding of domains within an IN monomer polypeptide. However, results with the gIN(201-236) probe, which showed equivalent binding to all targets that contained residues 201-240, make it unlikely that this reaction is simply mimicking tertiary interactions. They suggest, instead, that this C-terminal peptide is capable of specific association with the homologous region in another target polypeptide.
An IN(201-286) Fragment Can Self-associate-As a final test of the self-associating capability of the C-terminal domain, we expressed and partially purified IN(201-286)st as a streptagged protein (35) and analyzed this polypeptide by SEC and chemical cross-linking (Fig. 7). Under the SEC conditions used, a monomer of this fragment is expected to elute at 24.5 to 25 min. The results showed that this C-terminal fragment eluted at 20.5 min, consistent with a formation of a multimer (trimer or tetramer). Material in the void volume contained other protein and nucleic acid contaminants, whereas the multimer peak contained the bulk of the IN(201-286)st fragment (confirmed by SDS-PAGE analysis of fractions, data not shown). This C-terminal fragment was also tested in chemical crosslinking experiments, performed as in Fig. 3B. These results showed that the C-terminal polypeptide is able to form dimers, trimers, and tetramers in solution (Fig. 7 inset at top). The self-association properties of this fragment are not as limited as those observed with other fragments analyzed. It could be that this isolated C-terminal fragment is less sterically restrained, allowing it to form multimers (e.g. trimers) not found with other larger fragments. We conclude that the C-terminal fragment, IN(201-286), can self-associate as an independently expressed polypeptide. distinct advantages and limitations. Chemical cross-linking is the least stringent of the tests for multimerization because protein concentration can be controlled to favor association. SEC is the most stringent assay presented here because dissociative forces due to dilution are prominent during migration of multimeric proteins through a column, and the detection of multimers depends on the rate of dissociation relative to column run times (42). Accordingly, chemical cross-linking of the ASV IN catalytic core [IN(52-207)] clearly reveals dimer formation (Fig. 3), whereas in SEC, this fragment runs predominantly as a monomer.
Results obtained with these nonequilibrium methods are consistent with independent estimates of dissociation constants determined from sedimentation equilibrium experiments. Whereas full-length ASV IN has a K d (monomer-dimer) of 1-5 M (26), ASV IN(52-207) has a K d (monomer-dimer) in excess of 500 M. 3 The HIV-1 catalytic core (residues 50 -212) has been reported to have a stronger association than that of the analogous ASV fragment, and dimers have been observed with chemical cross-linking, SEC, and sedimentation analysis of the HIV-1 IN core fragment (27,30,43).
Both the cross-linking and SEC methods record the behavior of the majority of molecules in the protein preparations, whereas multimerization inferred from enzymatic complementation (30,31) or a transcriptional reporter system (32) could reflect the activity of a small fraction of protein present. In addition, the latter assays include DNA substrates which could facilitate the formation of multimers of IN. We have performed cross-linking experiments in the presence of various DNA substrates and have failed to detect enhancement of ASV IN mul-timerization (data not shown). However, this could reflect the adverse affects of the high salt conditions required to keep the protein soluble. We note also that enzymatic complementation cannot identify regions of IN that contribute to self-association if they do not include the minimal region necessary for catalytic activity. The analyses reported here do not require overlap with catalytic regions and represent the first evidence that important determinants of self-association reside in the Cterminal region of a retroviral integrase.
Multimerization Detected by Protein Overlay-The protein overlay technique provides an intermediate level of stringency for detection of protein-protein associations. It differs from the first two solution methods in several ways: protein probes are tested for binding to immobilized renatured targets, binding conditions are easily manipulated, and the specificity and sensitivity are high. This method offers an opportunity to investigate binding to partially purified target proteins and permits rapid screening of many potential partners for protein-protein associations. However, unlike the first two methods, it does not allow direct determination of the multimeric state and stoichiometry of the interacting partners. From these overlay experiments, we identified both the catalytic core and C-terminal domains as contributing to multimerization. We also observed that the C-terminal domain can specifically associate with itself. This latter property was confirmed by SEC and chemical cross-linking experiments with an isolated C-terminal fragment. Considered together, results from our analyses indicate that the two self-association domains of ASV IN can act independently, predominantly through interactions between homologous domains in each monomer. However, our data do not rule out cooperativity of the two self-association domains in the native full-length protein, nor can we exclude the possibility of other interactions not identified by these methods that could also contribute to the stability of a multimer.
Since multimerization is required for IN function, inhibitors of self-association may be of potential use in antiviral therapy. The protein overlay method may be particularly suited for the identification of peptide inhibitors that interfere with this property and presumably viral integration. There is precedent for such an inhibitor strategy with the retroviral protease (44,45).
Relationship between IN Structure, Multimerization, and Function-Several aspects of the domain structure of IN proteins revealed in these studies are consistent with recent information obtained from x-ray crystallographic analysis of the HIV-1 IN(50 -212) (16) and ASV IN(52-207) (17) catalytic cores. In both analyses, an extensive interface between two monomers in the unit cell was observed, with a large solventexcluded surface area. The presence of an extensive dimer interface in both structures is consistent with the contention that the catalytic core of one monomer interacts with the core of another. Despite the similar size of the two core domains, the ASV IN structure lacks the sixth C-terminal ␣ helix observed in HIV-1. In the HIV-1 dimer, this helix from one monomer extends out from the structure and interacts across the dimer interface with the analogous helix from the second monomer. It is possible that the added stability conferred by this interaction accounts for the tighter association of the HIV-1 core dimer relative to the ASV core noted above. The topology of both dimers suggests that the C-terminal domains of the full-length proteins would be in close proximity and available for interaction with each other across a multimeric interface.
It is still unclear whether retroviral integrase functions as a dimer or a tetramer. Formation of a tetramer might require interactions across two separate protomer interfaces, one for dimerization and a second for the association of two dimers into 3 S. Eaton, M. Andrake, and T. Laue, unpublished observations.  (Figs. 2 and 6). Whether this is due to the absence of the C-terminal domain remains to be investigated.
The topology of their folding places the retroviral integrases in a family of nucleases that includes RNase H, the RuvC resolvase, and the MuA transposase (46). Despite sharing a similar fold, and probably similar reaction chemistries, these diverse nucleases differ in substrate specificity, multimeric structure, and the requirement for coordination of cleavages performed. For example, RNase H is known to act as a monomer, whether as an independent domain or in the context of HIV-1 reverse transcriptase (47), and, correspondingly, its function does not require coordination of multiple cleavages. In contrast, RuvC is known to act as a dimer and performs two DNA cleavages to resolve a Holliday junction, but is not involved in joining of DNA strands (48). It is apparent from comparison of their crystal structures that the RuvC dimeric interface (49) differs from that of the IN core structures. A primary challenge for the future will be to determine how aspects of protein sequence and structure give rise to the specific quaternary interactions that allow each of these proteins to perform their specialized functions. Further study of ASV IN self-association should help to identify the relevant distinguishing features of integrase structure.