INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

A characteristic feature of retroviruses is the high degree of genetic variability, which is the result of the unique mechanism by which they replicate (43). The high rate of mutation of the viral genome has important consequences for the development of antiviral drugs. One of the sources of genetic variability in human immunodeficiency virus (HIV) is the reverse transcription process. Reverse transcriptase introduces errors that include frameshift mutations, base-pair substitutions, or deletions (for a recent review, see reference 30). After reverse transcription, the viral DNA is integrated into the human genome by the integrase (IN) protein. This integration process consists of two enzymatic reactions. The first is endonucleolytic cleavage of 2 nucleotides from the 3' ends of the viral DNA, which occurs in the cytoplasm of the infected cell. The second is integration of the viral DNA into a chromosome. The hydroxyl groups of the processed 3' ends of the viral DNA are coupled to the phosphate groups in the target DNA. This joining step obviously takes place in the nucleus of the host cell. Subsequent trimming of the 5' ends of the viral DNA and repair of the single-stranded gaps flanking the viral DNA complete proviral integration into the human genome and are presumably done by cellular enzymes (for reviews on retroviral integration, see references 21, 24, 48, and 52).

IN catalyzes cleavage and integration. These reactions can be studied in vitro with recombinant IN purified from Escherichia coli and oligonucleotides that mimic the viral DNA ends (6, 27, 41, 50). In the cleavage reaction, IN makes a specific phosphodiester bond of the viral DNA accessible for nucleophilic attack. In the presence of Mg²⁺, IN uses as a nucleophile primarily water, which is presumably also the nucleophile in vivo. However, when Mn²⁺ is present in the cleavage reaction, several nucleophiles, such as glycerol or the hydroxyl groups of the viral DNA ends, can attack the phosphodiester bond (20, 51). The latter case results in a cyclic dinucleotide as a by-product of the cleavage reaction (20). Similarly, when glycerol is used to attack the phosphodiester bond, the dinucleotide becomes attached to glycerol (51). HIV IN (20, 51) and equine infectious anemia virus IN (16) predominantly use water as a nucleophile in the Mn²⁺-dependent cleavage reaction, whereas IN from feline immunodeficiency virus (49), Moloney murine leukemia virus (10), and Rous sarcoma virus (34) prefer the hydroxyl groups of the viral DNA ends as a nucleophile. HIV IN can alter its preference when certain residues around the active site are substituted (17, 45).

In vitro, IN can perform the apparent reversal of integration, disintegration, in which the viral DNA is released from the target DNA (9). The substrate resembles the primary product of the concerted integration of both viral DNA ends into the target DNA. The viral DNA can be released in two different ways. In the intermolecular disintegration reaction, the phosphodiester bond between the viral DNA and the target DNA is attacked by the hydroxyl group of the target DNA of the opposite half molecule. In the intramolecular disintegration reaction, the attacking hydroxyl group is from the same half molecule as the phosphodiester bond, which has to be cleaved (see Fig. 3A for a schematic drawing of the reaction products) (31).

Whereas full-length IN is required for cleavage and integration (40, 47), the catalytic core alone (spanning amino acids 50 to 194) can perform the disintegration reaction (7, 47). The catalytic domain contains the conserved triad DD(35)E. These acidic residues coordinate a divalent metal ion. Substitution of one of the conserved residues of the active site abolishes both cleavage and integration (7, 11, 17, 26, 28, 45, 47), indicating that one active site is involved in the catalysis of both reactions. Structural studies of the catalytic core of HIV type 1 (HIV-1) IN (12) and of avian sarcoma virus IN (5) have shown that their structures are similar to those of other enzymes that catalyze phosphoryl transfer reactions, such as RNase H, RuvC, and MuA transposase (for recent reviews, see references 22, 33, 36, and 54).

The C terminus of IN, spanning amino acids 220 to 270, has nonspecific DNA binding activity (18, 35, 47, 53). The solution structure revealed that it has a Src homology 3-like fold (13, 29). Recently, the structure of the N terminus (amino acids 1 to 55) was solved by nuclear magnetic resonance (NMR) spectroscopy (8, 14); it is a three-helix bundle stabilized by a zinc binding unit. Both the N and the C termini are required for cleavage and integration.

Although the structures of the individual domains of IN are known, not much is known about specific amino acids that are involved in different functions of IN or about residues that can be mutated without a loss of activity. To gain insight into the functional organization of IN at the amino acid level, we analyzed over 100 mutants of HIV type 2 (HIV-2) IN. These mutants were obtained by a deliberately mutagenic PCR as well as by site-directed mutagenesis of the IN gene. Mutant proteins were tested for (i) the cleavage reaction and choice of nucleophile, (ii) integration activity, (iii) disintegration activity, and (iv) DNA binding activity of proteins which were catalytically inactive. We found substitutions that affected the choice of nucleophile in the cleavage reaction. These alterations in the IN protein clustered around the active site. We found mutants that showed a change in integration activity, while cleavage activity was not affected. In addition, we identified specific substitutions close to the dimer interface that increased intermolecular disintegration activity, while other catalytic activities remained at wild-type levels. This study shows that for a protein of this size (32 kDa), it is possible to scan amino acid functions by replacing the majority of the amino acids.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Mutagenesis. Deliberate random PCR mutagenesis was carried out with Taq polymerase (Gibco BRL) in the presence of MnCl₂ and a reduced dGTP concentration. First, a PCR of seven cycles was carried out (1 min at 94°C, 1 min at 55°C, and 2 min at 72°C with a reaction mixture containing 1 µM each primer, 0.4 mM each deoxynucleoside triphosphate [dNTP], 0.035 U of Taq polymerase/µl, and reaction buffer [Gibco BRL]). After the seven cycles, 10 µl of this PCR mixture was added to a 90-µl reaction mixture containing 1 µM each primer, 0.4 µM each dATP, dTTP, and dCTP, 0.08 µM dGTP, 0.5 mM MnCl₂, 3.5 U of Taq polymerase, and reaction buffer (Gibco BRL). PCR was continued for an additional 23 cycles. The following primers were used: a forward primer in the T7 promoter (5'-CGAAATTAATACGACTCACTATAGG-3') and a reverse primer in the HIV-2 gene introducing a stop codon at amino acid 279 (5'-CTAGGATCCTATCAACTATCCATCTCTTTGTCTTCC-3'). PCR products were cloned into the pET-15b vector (Novagen, Madison, Wis.). Mutant proteins (amino acids 1 to 279) were expressed as His-tagged fusion proteins in E. coli BL21(DE3) to allow one-step protein purification to near homogeneity. All substitutions within the HIV-2 IN gene were identified by sequencing with an ABI sequencer.

Protein expression and purification. Proteins were expressed in E. coli BL21(DE3) and purified by metal chelate chromatography at 4°C. Cells were lysed and sonicated in buffer A (10 mM sodium phosphate [pH 7.2], 0.1 mM EDTA, 3 mM -mercaptoethanol). After centrifugation (45 min at 15,000 × g in an SS34 rotor [Beckman]), the pellet was subjected to Dounce homogenization in buffer B (buffer A with 1 M NaCl). Tween 20 (0.1%) and 5 mM imidazole (pH 8.0) were added prior to 15 min of rotation at 4°C. The supernatant was cleared by a 30-min spin at 15,000 × g (SS34 rotor [Beckman]) and bound to Ni²⁺-nitrilotriacetic acid beads (Qiagen). After batchwise binding of the protein to the column material for 2 h, the column material was washed in buffer B with 0.1% Tween 20 and 20 mM imidazole (pH 8.0) and then in the same buffer without Tween 20 but with 2 mM imidazole (pH 8.0). The protein was eluted in buffer B with 200 mM imidazole (pH 8.0). Top fractions were dialyzed against buffer C (750 mM NaCl, 20 mM Tris [pH 7.6], 0.1 mM EDTA, 1 mM dithiothreitol, 40% glycerol) and stored at 80°C.

Activity assays. Cleavage and integration assays were done with oligonucleotides that mimic the ends of the HIV-2 U5 long terminal repeat as described previously (45). The disintegration substrate consisted of oligonucleotides representing an integration intermediate as described previously (substrate IV5 in reference 44). Reaction mixtures for the cleavage reaction contained 0.02 µM oligonucleotide substrate, 20 mM morpholinepropanesulfonic acid (MOPS; pH 7.2), 75 mM NaCl, 3 mM MnCl₂, 10 mM dithiothreitol 10% (vol/vol) glycerol, and approximately 100 ng of HIV-2 IN. For the integration and disintegration reactions, the concentration of MnCl₂ was reduced to 1 mM and 1 µg of bovine serum albumin was added instead of 10% glycerol. Cleavage and integration reactions were done at 37°C, and disintegration reactions were done at 30°C. Reactions were stopped after 1 h by the addition of 10 µl of formamide loading dye. After being heated to 80°C for 3 min, 5 µl of the samples was loaded onto a polyacrylamide-8 M urea-1× TBE (89 mM Tris, 89 mM boric acid, 2 mM EDTA) gel and electrophoresed. Reaction products from the integration reaction were separated on a 12% polyacrylamide gel, those from the disintegration reaction were separated on a 20% polyacrylamide gel, and those from the cleavage reaction were separated on a 24% polyacrylamide gel.

Alignment. An alignment was made among 53 different types of IN (from SWISSPROT) (3). We used the homology-derived structure prediction program (39) to determine the variability for amino acids with equivalents in HIV-2 IN (32), and we used the PHD server (37). IN sources for the alignment (according to EMBL/SWISSPROT) are listed in decreasing order of identity to HIV-2 IN (Hv2rodIV), and the identity to HIV-2 IN is indicated as a percentage after the IN source: polhv2ro (100%), polhv2st (94%), polhv2sb (94%), polhv2ca (94%), polhv2nz (93%), polhv2g1 (92%), polhv2be (92%), polhv2d1 (92%), polsivsp (87%), polsivs4 (86%), polsivm1 (85%), polhv2d2 (84%), polsivmk (80%), polsivag (64%), polsivgb (63%), polsiva1 (62%), polsivat (61%), polsivai (61%), polhv1jr (59%), polhv1a2 (59%), polhv1b5 (59%), polhv1br (59%), polhv1oy (59%), polhv1nd (59%), polhv1pv (59%), polhv1el (59%), polhv1rh (59%), polhv1b1 (59%), polhv1n5 (59%), polhv1ma (59%), polsivcz (58%), polhv1z2 (58%), polhv1y2 (58%), polhv1mn (58%), polhv1h2 (58%), polhv1u4 (57%), polhv1z6 (54%), polfivt2 (38%), polfivsd (38%), polfivpe (38%), polbiv27 (37%), polbiv06 (37%), poleiavy (36%), poleiavc (36%), poleiav9 (36%), polvilv1 (35%), polvilv2 (35%), polvilvk (35%), polvilv (35%), polomvvs (34%), polmpmv (30%), polcaevc (30%), and polsrv1 (30%).

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Methodology. Although the structures of the three domains of IN are known, little is known about the individual roles of specific amino acids in the functional interplay among these domains. In order to analyze the IN protein at the amino acid level, the HIV-2 gene was extensively mutated by deliberately mutagenic PCR and by site-directed mutagenesis. To obtain random mutations, the HIV-2 gene was amplified with Taq polymerase (Gibco BRL). The concentration of dGTP was reduced relative to the concentrations of the other dNTPs, and MnCl₂ was added to increase the error rate for Taq polymerase. The conditions were such that of the tested PCR-generated clones, the vast majority contained more than one base-pair substitution resulting in an amino acid change. Although only the dGTP concentration was reduced, we did not find a significant bias in the types of substitutions (see below).

Variability of IN. In total, 103 different residues were substituted, and of those, 77 amino acids could be replaced without affecting integration activity. This result indicates that the IN gene can be mutated extensively without a loss of in vitro activity of IN. To examine whether we could correlate tolerance of mutations found in our mutant series with the degree of conservation of amino acids of natural isolates of immunodeficiency viruses, we plotted the degree of conservation for every amino acid of IN and selected positions at which we found substitutions that did or did not alter integration activity. We aligned 53 IN protein sequences from different viruses or different natural isolates. The sequence variability at a specific position within the protein was determined by use of the program of Sander and Schneider (38). This program weights differences, in the sense that changes within a class of related residues (e.g., acidic residues) result in a lower variability score than changes between classes of nonrelated residues. The variability was plotted against the amino acid residues of HIV-2 IN (Fig. 1). Conserved residues, such as His-12 or Asp-64, scored low on the variability plot. As shown in Fig. 1, substitutions that resulted in an inactive protein (6 of 10) were in the low-variability portion (<10 on the variability plot); in other words, mutations that inactivated the protein were often in conserved residues. Most of the substitutions that had no effect on IN activity in vitro coincided with residues showing a variability of more than 10% in vivo (66 of 77). A few changes in conserved amino acids that had no effect on IN activity in vitro, e.g., A8V or P261V, were found. A change in these residues in vivo may not result in a fully replication-competent virus, as is suggested by the low variability of these residues. Presumably, they have another important role in vivo which was not detected in our biochemical assays. They could be involved in positioning of IN within the preintegration complex or in virion maturation (19).

View larger version (60K): [in this window] [in a new window]   FIG. 1.   Histogram representation of relative entropies of variability for all amino acids of HIV-2 IN. Percent variability (y axis) is plotted against residue number (nr) of HIV-2 IN (x axis). Relative entropies were derived from homology-derived structure prediction by use of the PHD server (37). A small relative entropy indicates a high degree of sequence homology. Asterisks and bullets indicate residue which can be substituted in HIV-2 IN without a loss of in vitro IN activity, as determined in this study (including single and multiple mutants) and as reported in the literature (11, 15, 17, 19, 28, 35, 45), respectively. Daggers indicate residues for which substitution results in a protein with <10% wild-type activity. Letters above the arrows indicate highly conserved residues.

Substitutions that influence the choice of nucleophile. Previously, it was shown that water, glycerol, or the hydroxyl ends of DNA can serve as a nucleophile in the cleavage reaction. The flexible loop between the second and third active-site residues [D(35)E], as well as residues around Asp-116, has been shown to be involved in the presentation of the nucleophile. Substitutions of several of these residues, such as Asn-117, Ser-147, or Gln-148, increase the formation of circular dinucleotides (46). As shown in Tables 1 and 2, substitution of Gln-146 by Arg, both as a single amino acid substitution and in the presence of other substitutions (H181Y or R187K), also resulted in an increase in circular dinucleotide products. A similar effect is observed when Ala replaces Gln-146 (46). Previously, only one mutant (N120L) that reduces the level of circular dinucleotide products (46), was identified. In this study, we found another mutant (F121I) that also produces fewer circular dinucleotide products (Table 2).

View larger version (56K): [in this window] [in a new window]   FIG. 2.   Ribbon diagram of the active site of HIV-1 IN (4). Residues indicated in dark grey are the active-site residues Asp-64, Asp-116, and Glu-152. Substitutions of residues indicated in light grey resulted in an increase in the preference of HIV-2 IN for the 3' hydroxyl group of the viral DNA as a nucleophile in the cleavage reaction. Substitutions of residues indicated in black had the opposite result. This image was generated with the program MOLSCRIPT (25).

Substitutions that affect integration and disintegration activities. Two mutants with reduced integration activity but with wild-type cleavage activity, S93P and F121I, were found. Replacing Trp-131, Leu-177, and Phe-185 slightly increased integration activity, whereas 3' processing was not affected.

View larger version (39K): [in this window] [in a new window]   FIG. 3.   Integration and disintegration activities of several mutants. (A) Intermolecular and intramolecular disintegration reactions. Intermolecular disintegration is the result of a phosphoryl transfer reaction from one half molecule (thick lines) to the other half molecule, resulting in a 30-nucleotide product. Intramolecular disintegration is the result of a phosphoryl transfer reaction within one half molecule, resulting in a hairpin of 25 nucleotides. *, radioactive label. Disintegration substrate was incubated without protein (), with wild-type protein (WT), or with mutant proteins as indicated above the lanes. (B) Integration activities of some of the mutants shown in panel A.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

The HIV-2 IN gene was randomly mutated, resulting in a collection of mutated genes in which 103 of 293 residues were substituted. Mutants were tested for in vitro activity. We found that IN can be extensively mutated without a loss of its in vitro activity. More than 67% of the mutants could still catalyze the integration reaction at wild-type levels.

We compared the tolerance of mutations of IN with the degree of conservation of amino acids in 53 natural isolates of immunodeficiency viruses. As expected, most of the mutants (86%) that had wild-type integration activity contained amino acid substitutions in residues that had variability of more than 10%. A few catalytically active mutants that contained substitutions of residues that had no variability at all in vivo were found. For instance, conserved Ala-8 could be substituted by a valine, with no affect on in vitro activity. Ala-8 is located in the hydrophobic core of the three-helix bundle at the N terminus. The conserved A8V substitution presumably did not cause a major change in the folding of IN. Another conserved substitution with no effect on in vitro activity was P261Q. Pro-261 is located just outside a  sheet before a helical turn in the C terminus. It has been suggested that Pro-261 and Lys-258 are both involved in DNA binding by the C-terminal domain (29). Although DNA binding by the C-terminal domains including substitutions at those residues was not tested, mutated full-length IN retained its catalytic activity. It is possible that IN containing one of these mutations was rapidly degraded in vivo or that IN has an additional role in virus maturation.

Nevertheless, certain alterations of conserved residues, e.g., Trp-61, His-78, and Gly-149, do abolish in vitro activity. Substitution of one of these residues is detrimental to the in vitro activity of IN. The close proximity of Trp-61 and Gly-149 to the active site may make these residues essential for normal activity.

Mutant L104P also had severely reduced catalytic activity. The introduction of a proline in an  helix often results in distortion of that helix. Since Leu-104 is located in the middle of a helix, a Pro substitution of Leu-104 probably results in a different folding of that helix, thereby affecting proper IN function.

The choice of a nucleophile is changed by several substitutions, e.g., F121I or Q146R. The formation of cyclic dinucleotide products requires nonpairing of the two base pairs at the outer ends of the viral DNA prior to nucleophilic attack of the internal phosphate group by the 3' hydroxyl group of the viral DNA. The increased level of circular dinucleotide products could be the result of disruption of base pairing at the viral DNA ends by the mutant proteins. Alternatively, the mutants could induce a conformational change which facilitates the entry of only specific nucleophiles into the cleft of the active site. Interestingly, thus far only neutral or basic amino acids have been found to alter the preference for the hydroxyl group of the viral DNA as a nucleophile. For reasons that we do not understand, substitutions with acidic residues do not change the use of the viral DNA ends as a nucleophile.

As shown in Table 2, certain amino acid substitutions resulted in a decrease in integration activity, while cleavage activity was not affected. Mutant S93P had reduced integration activity. Since Pro substitutes for Ser at the beginning of a helix, the conformation of IN is probably altered. Therefore, we cannot conclude that Ser-93 is directly involved in the integration process. Mutant F121I also had reduced integration activity and had wild-type cleavage activity. In addition, it had a decreased level of circular dinucleotide products. This same phenomenon was observed for mutant N117I (45). The formation of circular dinucleotide products and the integration reaction itself require nucleophilic attack of a phosphodiester bond by the 3' hydroxyl group of the viral DNA. Since both reactions were impaired by the N117I and F121I substitutions, these residues could be involved in positioning of the attacking hydroxyl group.

In this study, certain amino acid substitutions that affected intra- and intermolecular disintegration differently were identified. The two half molecules of the disintegration substrate anneal by 5 bp. In order to catalyze a phosphoryl transfer reaction from one half molecule to the opposite half molecule (intermolecular disintegration; Fig. 3A), IN should keep the two half molecules together during the reaction, by protein-protein interactions. Therefore, it is conceivable that the increased intermolecular disintegration activity was the result of stronger oligomerization. Mutants which showed increased intermolecular disintegration activity were indeed clustered around the dimer interface, e.g., I110R, T111I, and C182S.

Although the structures of the individual domains of IN have been elucidated, not much is known about the interaction among the domains and between IN and its substrates. In this study, we analyzed HIV-2 IN biochemically, demonstrating residues that are involved in presenting the nucleophile in the cleavage reaction and residues that are located in the dimer interface and affect intermolecular disintegration. The increase in the speed of protein purification by single-step affinity protocols and automated sequencing has made possible what was inconceivable a few years ago: a totally undirectional mutational scan of a gene, followed by functional analyses of the enzymatic properties of the purified proteins.