Peptides Derived from the Reverse Transcriptase of Human Immunodeficiency Virus Type 1 as Novel Inhibitors of the Viral Integrase* □ S

Recent studies have shown that the integrase (IN) of HIV-1 is inhibited in vitro by HIV-1 reverse transcriptase (RT). We further investigated the specific protein sequences of RT that were involved in this inhibition by screening a complete library of RT-derived peptides for their inhibition of IN activities. Two 20-residue peptides, peptide 4286, derived from the RT DNA polymerase domain, and the one designated 4321, from the RT ribonuclease H domain, inhibit the enzymatic activities of IN in vitro . The former peptide inhibits all three IN-associated activities (3 (cid:2) -end processing, strand transfer, and disintegration), whereas the latter one inhibits primarily the first two functions. We showed the importance of the sequences and peptide length for the effective inhibition of IN activities. Binding assays of the peptides to IN (with no DNA substrate present) indicated that the two inhibitory peptides (as well as several non-inhibitory peptides) interact directly with IN. Moreover, the isolated catalytic core domain of IN also interacted directly with the two inhibitory peptides. Nevertheless, only peptide 4286 can inhibit the disintegration activity associated with the IN core domain, because this activity is the only one exhib-ited by this domain. This result was expected from the lack of inhibition of disintegration of full-length

Two viral encoded enzymes play central roles in the early stages of the replication of retroviruses and retrotransposons. The first one, reverse transcriptase (RT), 1 converts the single-stranded viral RNA into double-stranded DNA in a relatively complex process, reverse transcription. This step is catalyzed by the two catalytic activities of RT, the DNA polymerase (capable of copying both RNA and DNA into DNA) and the RNase H activity, which concomitantly hydrolyzes the RNA strand in the DNA-RNA heteroduplex formed (1). Subsequently, the RT-produced double-stranded DNA is transported into the nucleus, as part of the nucleoprotein complex (designated the preintegration complex), where it integrates into the genomic target DNA by the second viral enzyme, the IN. IN identifies the ends of the linear viral DNA, trims them (by removing two or three extra nucleotides located 3Ј to the highly conserved CA 3Ј termini), and then accompanies the DNA into the nucleus to catalyze integration into the target cellular DNA (1)(2)(3). There are several examples for potential linkages between RT and IN. First, the DNA product of RT is the substrate for IN, the next enzyme in the line of the viral replication cycle. Second, both proteins are proteolytic products of the same polyprotein precursor encoded by a single retroviral gene, the pol (1). In some cases, as in avian sarcoma leukosis virus, the IN sequence appears in two forms, one as part of the large ␤-subunit of the RT and the other as a free IN protein designated pp32 (1,4). Moreover, PICs, which are capable of performing in vitro integration, contain the viral DNA, IN, RT, and other proteins (5)(6)(7)(8). Third, the INs and RTs of HIV-1 and MLV were shown to exhibit physical interactions (9 -11). These direct contacts between the IN and RT of HIV-1 were recently confirmed by us by using surface plasmon resonance technology. 2 Finally, we and others (12,13) have shown recently that RT can inhibit in vitro the enzymatic activities of IN, suggesting functional roles for these interactions.
After the completion of reverse transcription in the cytoplasm of the infected cell, there is a significant delay in the process of integration depending on the rate by which the PICs are transported into the nucleus (1). Because all possible catalytic components for integration are likely to be present in the PICs, the viral DNA can serve as donor DNA as well as the target DNA for integration. Such a potential auto-integration process is suicidal for the virus, as it destructs the viral genome. For that reason, it is imperative to understand the mechanisms that regulate the integration of the retroviral genome and learn how to control it. Several cellular proteins are known to be involved in the integration process in HIV-infected cells (14 -17). Most interestingly, one of these factors serves also as a barrier to auto-integration in MLV-infected cells (18).
Because RT can also inhibit IN activities (12,13), it is possible that an alternative pathway of inhibiting auto-integration without the involvement of cellular proteins exists as well. This course of action involves the functional interactions between the authentic viral proteins, RTs and INs, prior to integration into the cellular DNA.
Here we have extended our previous in vitro study on the inhibition of IN by RT (12) by studying specific RT-derived sequences that are involved in binding and inhibiting IN. This was done through a systematic in vitro screening of a library of HIV-1 RT-derived peptides for their capacity to inhibit HIV-1 IN activities and to interact directly with IN. We have identified two 20-residue-long peptides that inhibit IN activities in vitro. The first peptide is derived from the DNA polymerase active site, and the second peptide is derived from the RNase H domain of RT. In addition, we have shown by a novel dot-blot analysis that the full-length RT, as well as several RT-derived peptides, bind directly to HIV-1 IN. The localization of the two inhibitory peptides in the three-dimensional structure of the RT and the putative docking (of the DNA polymerase-derived peptide) into IN-CCD suggest that the inhibition of IN activities results from a steric hindrance. These findings can lead to insights into the development of novel peptide-based, specific, and highly potent IN inhibitors. Such inhibitors are expected to interfere with HIV infectivity and, thus, serve as novel prodrugs for the treatment of AIDS.

EXPERIMENTAL PROCEDURES
The sequences of all studied peptides are given in the Supplemental Material.

HIV-1 RT-derived Peptides
Two sets of synthetic peptides libraries were generous gifts from the NIH-AIDS Research and Reference Reagent Program. The first set was the HIV-1 HXB2R Pol peptides complete set (catalog number 4358). These peptides are each 20 residues in length with 10 amino acid overlaps between the sequential peptides. All peptides derived from the DNA polymerase domain were prepared to a final concentration of 900 M, by dissolving in either DDW or in a minimal amount of 100% Me 2 SO and then diluted with DDW to 6.5% Me 2 SO, all according to the recommended instructions (www.aidsreagent.org). All peptides derived from the RNase H domain were dissolved in a minimal amount of 100% Me 2 SO and then diluted with DDW to 6.5% Me 2 SO (yielding final concentrations of 400 M). The second peptide set was the complete set of the HIV-1 clade B consensus pol peptides (catalog number 6208). These peptides are each 15 residues in length with 11 amino acid overlaps between the sequential peptides. They were also dissolved as above to a final concentration of 400 M each in 6.5% Me 2 SO. All peptides were designated according to the numbering given by the NIH-AIDS Research and Reference Reagent Program (www.aidsreagent.org).
Peptides 4286Ј-1 (15-mer) and 4321Ј-1 (14-mer) were custom-synthesized. The first one was diluted to a final concentration of 3.5 mM in DDW and the second one as described above to a final concentration of 5.7 mM in 6.5% Me 2 SO.

Recombinant Proteins
Bacterial Expression and Purification of HIV-1 IN and HIV-1 RT Proteins-All HIV-1 IN and RT versions used were highly purified as judged from their pattern after analysis by SDS-PAGE (data not shown).
HIV-1 IN-HIV-1 IN from the BH-10 strain of HIV-1, carrying amino-terminal His 6 tag, was expressed and purified as described in detail (20).
HIV-1 IN-Catalytic Core Domain (the Double Mutant W131E, F185K)-This large IN fragment (designated IN-CCD) was expressed as an amino-terminal His 6 -tagged protein and purified as described in detail (21).
HIV-1 RT-The expression and purification of heterodimeric (p66/ p51) HIV-1 RT with a His 6 tag, attached to the carboxyl-terminal end of the p66, were described in detail previously (22). The expression and purification of a non-tagged heterodimeric HIV-1 RT were described previously (23,24).

Assays for Enzymatic Activities of HIV-1 IN Oligonucleotides Used
The following gel-purified oligonucleotides were used in the enzymatic assays of HIV-1 IN: A (21-mer), 5Ј-GTGTGGAAAATCTCTAG-CAGT-3Ј; B (21-mer), 5Ј-ACTGCTAGAGATTTTCCACAC-3Ј; C (19mer), 5Ј-GTGTGGAAAATCTCTAGCA-3Ј; D (38-mer), 5-TGCTAGTTC-TAGCAGGCCCTTGGGCCGGCGCTTGCGCC-3. Oligonucleotides A-C correspond to the U5 end of the HIV-1 long terminal repeat (25). Boldface letters indicate the highly conserved CA/TG dinucleotide pair. Oligonucleotide C is identical to A, after the removal of the GT dinucleotides from its 3Ј-end and thus after annealing to oligonucleotide B, creating a dinucleotide overhang at the 5Ј-end of oligonucleotide B. Oligonucleotide D, termed "dumbbell," folds to form a structure mimicking the integration intermediate. This substrate, used for assaying the disintegration activity, has 5 bp of a viral sequence and 10 bp of a non-viral sequence, as described earlier (26). In order to test the 3Ј-end processing and the resulting strand transfer activity of IN, the 5Ј-endlabeled oligonucleotide A, annealed to its complementary strand, oligonucleotide B (both 21 nucleotides long), was used. We have used the duplex of oligonucleotides C and B and for assaying the 3Ј-end processing activity (12).
5Ј-End Labeling and Substrate Preparations-Fifty pmol of oligonucleotides A, C, or D were 5Ј-end-labeled using 1 unit of T4 polynucleotide kinase and 50 Ci of [␥-32 P]ATP, in a final volume of 50 l of the appropriate buffer (supplied by the manufacturer) for 30 min at 37°C. The samples were then heat-inactivated. The 5Ј-end-labeled oligonucleotides designated A or C were annealed each to an equimolar amount of oligonucleotide B in 55 mM Tris-HCl (pH 7.5) and 0.27 M NaCl. For the disintegration assay, we have used 5Ј-end-labeled oligonucleotide D, which forms a dumbbell structure after self-annealing (12).
Assays of the 3Ј-End Processing, Strand Transfer (or DNA Joining), and Disintegration Activities-In the strand transfer assays described, the labeled 5Ј-end substrate employed served as both the target and donor DNA leading to an increase in the molecular size of the substrate, whereas in the 3Ј-end processing and disintegration assays, we have followed the unique cleavage of the labeled substrates (12,26). All reactions were performed in 10-l reaction mixtures with 0.33 pmol of the labeled DNA substrate and the reaction buffer, containing 1 mM HEPES (pH 7.5), 54 mM NaCl, 2.5 mM MnCl 2 , 50 M EDTA, 2 mM dithiothreitol, 0.1 mg/ml bovine serum albumin ,0.1 mM spermidine, 25 mM MOPS (pH 7.2), 5% glycerol, and 1% Me 2 SO (12). We have assayed 250 ng of HIV-1 IN (which equals 4 pmol, assuming IN dimers of the 32-kDa subunits) or 1 g of HIV-1 IN-CCD (which equals 28 pmol, assuming dimers of the 18-kDa subunits). This difference in the amounts of proteins assayed results from the lower activity of IN-CCD relative to the full-length IN. The two versions of HIV-1 IN were preincubated on ice for 5 min in the presence or the absence of HIV-1 RT or the HIV-1 RT-derived peptides. Reactions were initiated after adding the labeled DNA substrate in the reaction buffer, incubated for 30 min at 37°C, and then stopped by adding 10 l of formamide loading buffer (90% formamide, 10 mM EDTA, 1 mg/ml bromphenol blue, 1 mg/ml xylene cyanole). The samples were heat-denatured, cooled on ice, and loaded onto 6 M urea, 14% polyacrylamide denaturing gels, followed by electrophoresis (urea-PAGE). The gels were dried and subjected to autoradiography at Ϫ80°C or at room temperature to obtain close to linear exposures.

Quantitative Analyses of the Inhibition of HIV-1 IN by HIV-1 RT and by RT-derived Peptides
The films were scanned, and the levels of IN activity were calculated using the densitometric software TINA (version 2.07d; Raytest Isotopenmessgeraete, GmbH). The 3Ј-end processing activity was determined as a percentage of the total 5Ј-end-labeled DNA oligonucleotides converted to 19-mer DNA. Strand transfer (or DNA joining) activity was calculated as a percentage of the total amount of labeled DNA found in DNA bands of 22 nucleotides or more in length. Disintegration activity was calculated as a percentage of the total 5Ј-end-labeled 38-mer DNA oligonucleotide converted to 14-mer DNA. All residual activities were relative to the control activity of the IN (with no RT or RT-derived peptides present) and were expressed as percentage of the initial IN activities.

Dot-blot Binding Assay
HIV-1 RT-derived Peptides Bound to HIV-1 IN-Nitrocellulose filters were soaked in DDW followed by transfer buffer, containing 25 mM Tris, 192 mM glycine, and 20% (v/v) methanol, for 5 min and then transferred to a dot-blot apparatus (Bio-Rad). Aliquots of RT-derived peptides or His 6 -tagged HIV-1 RT (10 g each) were bound to the filter under vacuum for 15 min. To ascertain the binding of the protein and peptides to the nitrocellulose filter, the blots were stained with Ponceau following by blocking with 5% MTBST (5% milk powder in 30 mM Tris-HCl (pH 8.5), 125 mM NaCl, and 0.1% (v/v) Tween 20) for ϳ16 h at 4°C. The experimental blot was then incubated with HIV-1 IN (17 g/ml in 5% MTBST) for 1 h at room temperature, and the control blot was not incubated with IN. The filters were then washed with a solution similar to MTBST but with no milk powder (TBST). In one set of experiments (set A), the filters were incubated with mouse-anti HIV-1 IN antibodies (diluted 1:1000 in 5% MTBST) for 90 min at room temperature, following by washing with TBST. After that the incubation was performed with horseradish peroxidase-conjugated goat anti-mouse IgG antibodies (from Santa Cruz Biotechnology, diluted 1:500 in 5% MTBST) for 1 h at room temperature and then washed again, followed by an enhanced chemiluminescence reaction by mixing the two stock solutions at 1:1 ratio before use. These solutions are as follows: 25 mM luminol, 400 M paracoumaric acid in 100 mM Tris-HCl (pH 8.5), and 5.4 mM H 2 O 2 in 100 mM Tris-HCl (pH 8.5). In a second set of experiments (set B), both blots were incubated with anti-His 6 -horseradish peroxidase antibody (from Sigma, diluted 1:1000 in 5% MTBST) for 90 min followed by an enhanced chemiluminescence reaction.
HIV-1 RT-derived Peptides Bound to HIV-1 IN-CCD-This experiment was identical to the one described above for set B. The HIV-1 RT used here had no tag. The experimental blot was incubated with HIV-1 IN-CCD (20 g/ml in 5% MTBST) and the other one without IN-CCD.

RESULTS
We have shown previously (12) that all three enzymatic activities performed by HIV INs, namely the 3Ј-end processing, the strand transfer (termed together integration), and the reverse process, the disintegration, are inhibited in vitro by RTs. Subsequently, the goal of this study was to identify the sequences in HIV-1 RT that inhibit and interact with the fulllength HIV-1 IN, as well as with the IN catalytic core domain. To this end, we have used a large set of synthetic peptides that cover the whole pol gene of HIV-1 (the HXB2R isolate). The peptides were each 20 residues long with 10 amino acids overlaps between the sequential peptides. These peptides were tested for their ability to bind IN (12). Consequently, we first tested the sequences located within this domain of HIV-1 RT. A total of 42 peptides, designated from 4269 up to 4310 (starting from the amino-terminal peptide, see Supplemental Material), were screened for their ability to inhibit HIV-1 IN. These peptides were first tested in groups, each containing a mixture of three contiguous and partially overlapping peptides (every one at a final concentration of 90 M). All mixtures were preincubated on ice for 5 min with 0.4 M purified HIV-1 IN. The enzymatic assays were then initiated by adding the DNA substrate, used for testing the combined 3Ј-end processing and DNA joining activities of IN (oligonucleotides A and B, see under "Experimental Procedures"), followed by incubation at 37°C for 30 min. The results obtained show that only the mixture containing peptides 4284, 4285, and 4286 inhibited the integration activity of IN (data not shown). Consequently, each of these peptides was then tested separately (at a final concentration of 270 M) for its ability to inhibit the integration activity. The results indicated that the inhibition of IN was caused primarily by a single peptide, 4286. This peptide inhibits both IN-mediated strand transfer and 3Ј-end processing activities (Fig. 1). It is apparent that both activities were substantially inhibited by the peptide at a final concentration of 27 M. As a result, the extent of inhibition was also quantified to determine the peptide concentrations inhibiting 50% of the initial IN activity (IC 50 value, Table I). The value, calculated for the inhibition of the strand transfer, was about 4.5 M, and the one for the 3Ј-end processing was ϳ4.8 M (Table I).
Because every two adjacent peptides share 10-residue overlaps, we have also calculated the IC 50 values for peptides 4285 and 4287 (that share sequences with the inhibitory peptide 4286). The analyses of the reaction products, which led to the calculations of the IC 50 values (summarized in Table I), indicate that both 4285 and 4287 were considerably less effective than peptide 4286. Furthermore, we have also checked whether the two peptides 4285 and 4287, when tested together, demonstrate a synergistic effect by increasing their IN inhibiting capacity to a level as efficient as peptide 4286. The results showed that this is not the case, and their combined effect is still far below that of peptide 4286 (data not shown). This indicated that the inhibitory sequence of 4286 can be active only as one continuously linked sequence, which is likely to be structurally different from that of the other two peptides.
Peptide  necessarily exclude the possibility that the p15 by itself can also exhibit inhibitory effects. To address this issue, we have tested all 15 RNase H domain-derived 20-mer peptides (peptide numbers from 4311 up to 4325). In each test, the peptides were analyzed in groups of two adjacent and partially overlapping ones (each at a final concentration of 60 M). Only the mixture containing peptides 4321 and 4322 inhibited IN activities. Consequently, we have tested each peptide individually, and we found that only peptide 4321 was responsible for inhibiting both the 3Ј-end processing and strand transfer activities of HIV-1 IN (Fig. 2). The quantitative analyses indicated that the apparent IC 50 values calculated for both activities were quite similar (Table I). These values are close to those calculated for peptide 4286. It should be noted that neither peptide 4320 nor peptide 4322 (that share sequences with peptide 4321) inhibited the enzymatic activity of HIV-1 IN in vitro (Table I).
The  (Table I) Peptide 4286 Strongly Inhibits the Disintegration Activity of HIV-1 IN-All studied retroviral INs possess in vitro, in addition to the 3Ј-end processing and the forward strand transfer activities, the reverse activity of disintegration. The authentic disintegration activity was already studied extensively as a major IN activity (26 -28). The disintegration activity involved a specific endonucleolytic excision of the "donor viral" DNA (with the conserved 3Ј-end "CA" dinucleotide) from its target DNA. Given that HIV-1 RT also inhibited this activity of HIV-1 IN (12), it was of interest to extend the study by testing the inhibitory effects of peptides 4286 and 4321 on the disinte-gration activity. We have used the 32 P 5Ј-end-labeled 38-nt "dumbbell" substrate (oligonucleotide D, see "Experimental Procedures") and followed the generation of the specific 14-nucleotide DNA product (26). It is apparent that the two RT-derived peptides were capable of inhibiting the disintegration activity of HIV-1 IN (Fig. 3). Peptide 4286 effectively inhibited the disintegration activity with full inhibition ob-

Inhibition of HIV-1 Integrase by HIV-1 RT-derived Peptides
tained at a concentration of 27 M (Fig. 3A). However, peptide 4321 was less effective, because full inhibition was not accomplished even at a high peptide concentration of 120 M (Fig.  3B). Accordingly, the quantitative analysis showed that the apparent IC 50 values for 4286 and 4321 were ϳ9.4 and 100 M, respectively (Table I).

HIV-1 RT and RT-derived Peptides Interact Directly with HIV-1 IN-
In order to evaluate whether the inhibitory effect of RT-derived peptides 4286 and 4321 involved physical interactions with IN, we have tested all 20-residue peptides that span the whole 560-residue-long RT molecule for their direct IN binding capacity. Because no macromolecules other than IN and the peptides (or RT) were present in these reactions, every binding result was likely to reflect direct physical interactions between these components. To this goal, we have employed a dot-blot binding assay that involved the pre-binding of HIV-1 RT, or the RT-derived peptides, to a nitrocellulose filter, followed by a reaction with HIV-1 IN. Specific anti-IN antibodies were used to detect the bound IN (see "Experimental Procedures"). The observed interaction between the whole HIV-1 RT with HIV-1 IN (Fig. 4A) supported previous findings (9 -11).
Out of the 57 peptides tested, only 13 have shown a significant binding. The bindings by the positive peptides and two nonbinding peptides (peptides 4269 and 4300), along with their test for nonspecific bindings, are shown in Fig. 4A. As three of these peptides (4321, 4316, and 4308) showed some nonspecific binding to the secondary antibodies used in this particular experiment, we have confirmed the specificity of interaction of the inhibitory peptide, 4321 (along with a positive control of peptide 4286) with IN, by using a different set of antibodies (Fig. 4B). Among the peptides shown to bind IN (Fig. 4A and Table II), only three peptides (4286, 4321, and to a lesser extent  (Tables I and II). The fact that there are additional RT-derived peptides that bind IN but do not inhibit any of the IN activities indicated that not all peptide-IN interactions necessarily led to inhibition of catalysis. Because we have assumed (see below) that the inhibition was mediated primarily by steric hindrance of the IN catalytic domain, it is likely that these non-inhibitory peptides bound IN sites that were not directly involved in catalysis.

4287) also inhibited IN enzymatic activities
Both HIV-1 RT and Peptide 4286 Inhibit the Disintegration Activity of HIV-1 IN-CCD-In addition to the inhibition of the 3Ј-end processing and forward strand transfer activities, peptide 4286 and to a lesser extent peptide 4321 were also shown to inhibit the disintegration activity of HIV-1 IN (Fig. 3). As the isolated core domain of HIV-1 IN exhibited, in vitro, a disintegration activity (2,3,27,29,30), it was possible that the two peptides interacted with IN-CCD and interfered with its activity. This recombinant segment of HIV-1 IN was 163 residues in length and was derived from residues 50 -212 of the 280-residue full-length HIV-1 IN. The disintegration activity of IN-CCD was known to be substantially lower than that of the full-length enzyme (27,30).
Because it was never tested whether the complete HIV-1 RT molecule inhibited HIV-1 IN CCD, we have first tested whether RT inhibits the disintegration activity of this IN domain. It is apparent that such an inhibition does exist (Fig. 5A) (Fig. 5B). On the other hand, peptide 4321 had no effect on this activity, even at a concentration as high as 270 M (Fig. 5B).  (42)(43)(44)(45), from random sequences identified in synthetic libraries (45)(46)(47), from combinatorial libraries of chemically synthesized peptides (45,48), or from cellular IN-interacting proteins (45,49,50). In contrast to many anti-RT drugs used in the treatment of AIDS patients, there is only one group of nonpeptide anti-IN drugs that has entered clinical trials (39,51), and none is used routinely so far in patients.

HIV-1 RT and RT-derived Peptides 4286 and 4321 Physically
It was recently reported that the HIV-1 IN activities are inhibited by HIV-1 RT (12,13). Moreover, we have proposed that this phenomenon has a critical biological significance, as it is possible that the self-destructive process of auto-integration of the viral DNA into itself is blocked in the cytoplasm of infected cells by the observed effect of RT on IN. In the present study, we have defined the precise sequences within HIV-1 RT that inhibit and interact with HIV-1 IN. We show that two peptides inhibit in vitro the enzymatic activities of IN and, in addition to few other peptides, can bind IN directly.
Two inhibitory 20-residue peptides, 4286 and 4321, derived from the RT DNA polymerase and RNase H domains, respectively, were identified in the synthetic peptides library that spans the whole pol protein of HIV-1 (Figs. 1-3 and Table I). The extent of inhibition of the strand transfer and 3Ј-end processing activities by both peptides are similar (Table I). On the other hand, the disintegration activity of IN is by far more sensitive to inhibition by peptide 4286 than by peptide 4321. This suggests that the mechanisms of IN inhibition by the two peptides might be different. For that reason, we have checked whether the inhibitory effects of these two peptides are coupled, namely whether their combined effect is stronger than the effect of each separately. The results of this experiment show no such an apparent augmentation (data not shown), suggesting that inhibitions by both peptides might be still interrelated.
We have previously predicted that the inhibition of IN by RT results from direct interactions between the proteins (12). To confirm this, we have conducted the dot-blot experiment, described in Fig. 4. Similar to the whole RT protein, the two peptides 4286 and 4321 interact also with HIV-1 IN. The fact that the in vitro binding assay was a direct one indicates that the observed interactions are not mediated by nucleic acids or other cofactors, further supporting our hypothesis that the inhibition of IN by RT-derived peptides is caused by physical interactions. However, the same assay also shows that noninhibitory peptides can also bind IN. This may indicate that the regions containing the sequences of these specific peptides (in   TABLE II Summary of the RT-derived peptides that bind HIV-1 IN and the inhibition data The binding was performed by a dot-blot assay (described as set A in "Experimental Procedures"). The inhibition of the strand transfer activity of HIV-1 IN was determined according to "Experimental Procedures." ND, could not be determined due to non-specific deformations of the electrophoresis patterns of the reaction products.
HIV-1 RT-derived peptides (20- ). An alternative explanation can result from the potential differences between the folding of free peptides and the structures of the same sequences in the context of the RT protein. It is likely that only a fraction of the free peptide molecules can mimic the folding within the whole protein, and only this portion is effective in inhibition. Moreover, the proper conformation that leads to inhibition by peptides 4286 and 4321 depends on the length of the peptides. It is apparent that the full 20-residue sequences of both inhibitory peptides are crucial for the inhibition of HIV-1 IN, as shorter peptides, derived from sequences of the same peptides (or peptides with partial sequence overlaps with these peptides), exhibit a substantial reduction in the inhibitory effects (Table I).
Peptide 4286 is derived from a sequence within the DNA polymerase active site of HIV-1 RT, which is located in the "palm" subdomain (as parts of ␣-helix E and ␤-sheet 9 (52)). This subdomain contains the highly conserved YXDD motif, typical of polymerases (53). This sequence exists almost entirely in peptide 4286 (except for the last D residue). This may explain our previous findings that HIV-1 and HIV-2 INs are inhibited by both HIV-1 and HIV-2 RTs and by the RT from another retroviral family, MLV RT (12).
Peptide 4321 (residues 516 -535 in HIV-1 RT) is derived from the RNase H domain (across ␣-helix DЈ and ␤-sheet 5Ј (52)). The fact that this peptide interacts with IN seems not to be in line with a very recent study (54), which suggests that the free RNase H domain of HIV-1 RT (residues 422-560, p15) is not capable of binding IN. This may result from differences in the methodologies used in the two studies. We have tested the binding of merely peptides, employing the dot-blot analysis, a test confirmed by IN inhibition. In contrast, Hehl et al. (54) employed for the p15 protein fragment a pull-down experiment, which might be less sensitive than our dot-blot assay. It is also important to note that, apart from peptide 4321, we have found that five additional RNase H-derived peptides bind directly HIV-1 IN (Fig. 4A and Table II).
An additional potential aspect, relating to the inhibition of IN by the RNase H-derived peptide, can stem from the three- dimensional similarities between IN-CCD and the RNase H of HIV-1 RT and Escherichia coli (55,56). This can explain why RNase H inhibitors were found to inhibit IN activities as well (57). It is then possible that the RNase H-derived peptide 4321 interacts with IN due to similarities in structure, thereby interfering with the dimerization of IN. A similar finding (that peptides derived from IN-CCD interfere with IN dimerization and, consequently, inhibit IN activity) was already reported (42,43).
Another issue of interest is to identify the locations of the sequences of peptides 4286 and 4321 in the surroundings of the globular structure of the heterodimeric p66/p51 RT protein. The DNA polymerase domain of the p66 subunit of RT resembles a right-hand structure, with fingers, thumb, and palm subdomain, connected to the RNase H domain by the connection (52,58). Fig. 7 shows that both peptide segments (peptide 4286 in the p66 and p51 subunits and peptide 4321 in the RNase H domain of p66) are largely facing outwards, as they are on the surface of the protein, thus allowing potential interactions with other proteins. Yet the poly-merase-derived peptide also penetrates from the "backhand" into the base of the palm subdomain (in the "forehand"), where the polymerase active site is located (Fig. 7B). This model supports the validity of the findings presented in this work about the inhibitory peptides, as only segments facing outwards of the RT can potentially interact with the IN molecule.
The DNA polymerase-derived peptide 4286 effectively interacts with IN-CCD and inhibits its disintegration activity, which is the only enzymatic activity expressed in vitro by this protein ( (55,59), crystal form I (21). Crystal structures of the active core domain (crystal forms II and III (21,60)) are very similar, with the exception of region 141-148. This region comprises a flexible loop, which is close to the active site. The conformation of this loop is strongly affected by the crystal packing and, presumably, by the binding of the substrate or inhibitor. By taking into account the above considerations, we have used the coordinates of HIV-1 IN-CCD from crystal form II (1.95 Å resolution) with the 141-148 region omitted (21). In the docking model (Fig. 8), we assume that the folding of peptide 4286, by which it interacts with IN, adopts a conformation similar to that one in the structure of HIV-1 RT The right-hand three-dimensional structure of the DNA polymerase domain is shown from two angles. Peptide 4286 is located at the base of the palm close to the DNA polymerase active site and penetrates outwards into the backhand surface. A, front view of the back of the right-hand conformation of the RT, with the different subdomains indicated. B, after rotating the molecule, an "inside" view of the palm subdomain of the right-hand p66 RT model.  (21). The peptide conformation was as in Protein Data Bank entry 1REV (61). In gray is the space-filling diagram of IN-CCD. The three highly conserved acidic-residues catalytic triad, DX(39 -58)DX 35 E, are indicated in red-orange. HIV-1 RT-derived peptide 4286 is represented by the pink stick and ball diagram and is colored by atom types as follows: carbon, magenta; nitrogen, blue; oxygen, red; and sulfur, orange. segment are residues Tyr-143 and Gln-148, known to be involved in viral DNA binding (62,63). Tyr-143 was also shown to play a secondary role in catalysis (64). Hence, the binding of peptide 4286 to this region may cause a steric interference with the DNA binding and consequently inhibit IN activities. The proposed docking model (Fig. 8) involves several non-covalent interactions between the peptide and IN-CCD; residue Gln-174 of the peptide (according to the full RT numbering) makes a hydrogen bond with Ser-57 of the IN-CCD. In addition, residues Phe-171, Leu-168, and Ile-178 in the peptide interact by van der Waals interactions with amino acids Val-79, His-114, and Gly-149 in the IN-CCD, respectively. The proposed docking model between the HIV-1 IN-CCD and peptide 4286 suggested here is supported by a similar model that was already found in the crystal structure of avian sarcoma leukosis virus IN-CCD complexed with an HIV-1 IN inhibitor (65). Given that peptide 4321 was not effective in inhibiting the disintegration activity of IN-CCD, a similar docking of this peptide into the catalytic site of the IN is likely not to be valid and hence was not performed.
The study presented here still leaves several unanswered questions. First, little is known about the precise mechanism of inhibition, despite our prediction that it involves steric hindrance. Second, the model proposed for the interaction of peptide 4286 with IN-CCD does not exclude the possibility that RT residues, other than those shown to inhibit IN, are critical for the inhibition. Therefore, it will be important to study the involvements of the RT-derived peptides that bind but do not inhibit IN, as part of the whole RT molecule. Future sitedirected mutagenesis of RT should indicate which residues are critical for binding and inhibition and whether the proposed docking model is correct. Likewise, mutagenesis of IN might also help in pinpointing the residues important for molecular interactions with RT. Accordingly, we have recently identified IN-derived peptides that interact with RT. Last of all, our final goal is to see whether the studied peptides and related ones can inhibit integration in HIV-infected cells and, hence, inhibit viral infectivity (and, as such, serves as a novel group of potent and highly specific anti-AIDS drugs). To this goal, the uptake of the peptides by human cells and their effects on viral infectivity will be studied.