Inhibitors of Strand Transfer That Prevent Integration and Inhibit HIV-1 Replication in Cells

See allHide authors and affiliations

Science  28 Jan 2000:
Vol. 287, Issue 5453, pp. 646-650
DOI: 10.1126/science.287.5453.646


Integrase is essential for human immunodeficiency virus–type 1 (HIV-1) replication; however, potent inhibition of the isolated enzyme in biochemical assays has not readily translated into antiviral activity in a manner consistent with inhibition of integration. In this report, we describe diketo acid inhibitors of HIV-1 integrase that manifest antiviral activity as a consequence of their effect on integration. The antiviral activity of these compounds is due exclusively to inhibition of one of the two catalytic functions of integrase, strand transfer.

The development of chemotherapeutic agents for the treatment of HIV-1 infection has focused primarily on two viral enzymes: reverse transcriptase and protease. Although regimens including agents directed at each of these biochemical targets are effective in reducing viral load and morbidity and mortality, the long-lived nature of the infection and the genetic plasticity of the virus have made it apparent that new antiretroviral agents are required to deal with the appearance and spread of resistance (1). HIV-1 integrase catalyzes the insertion of the viral DNA into the genome of the host cell. Integration is essential for viral replication and is thus an attractive target for novel chemotherapy (2,3). Many inhibitors of HIV-1 integrase have been identified; however, their in vitro activity has not translated into antiviral activity in cells (4).

Integration is a multistep process that occurs in discrete biochemical stages: (i) assembly of a stable complex with specific DNA sequences at the end of the HIV-1 long terminal repeat (LTR) regions, (ii) endonucleolytic processing of the viral DNA to remove the terminal dinucleotide from each 3′ end, and (iii) strand transfer in which the viral DNA 3′ ends are covalently linked to the cellular (target) DNA (Fig. 1) (4). Each of the catalytic reactions (3′ processing and strand transfer) requires integrase to be appropriately assembled on a specific viral DNA (or donor) substrate (5). In general, compounds identified in assays with purified, recombinant integrase interfere with assembly in vitro (6, 7). Because assembly is a prerequisite for catalysis, such compounds may appear to inhibit 3′ processing and strand transfer, but they have no effect on either reaction when assayed subsequent to assembly on HIV-1–specific oligonucleotides (6). These compounds are also ineffective in assays wherein viral preintegration complexes isolated from HIV-1–infected cells are used (8).

Figure 1

Activity (IC50, in μM) of diketo acid inhibitors. PIC, preintegration complex.

To identify inhibitors of catalysis, we biased the strand transfer reaction by means of preassembling recombinant integrase on immobilized oligonucleotides as a surrogate for preintegration complexes (6) (Fig. 1). In a random screen of more than 250,000 samples, a variety of inhibitors was identified; however, the most potent and specific compounds each contained a distinct diketo acid moiety, and thus these inhibitors segregate into a single structural class (Fig. 1). The diketo acid functionality is an intrinsic feature of these inhibitors but is not sufficient for activity, as structural analogs exhibit a range of inhibitory potency. For most analogs, the activity observed in strand transfer assays with recombinant integrase correlated with their relative activity in assays using HIV-1 preintegration complexes (9). Analogs that were more potent in these biochemical assays also inhibited HIV-1 replication in cell culture.

L-731,988 and L-708,906 were two of the most active diketo acids in strand transfer assays with recombinant integrase. With 50% inhibitory concentrations (IC50's) of 80 and 150 nM, respectively, L-731,988 and L-708,906 are also the most potent inhibitors of preintegration complexes described to date. In a single-cycle assay for acute infection (10), L-731,988 and L-708,906 inhibited HIV-1 replication with IC50's of 1 to 2 μM; higher concentrations prevented the spread of HIV-1 in cell culture for several weeks (Fig. 2). L-731,988 and L-708,906 were comparably active against both macrophage- and T cell line–tropic strains of HIV-1, clinical isolates, and variants resistant to reverse transcriptase and protease inhibitors (11). Consistent with the effect of an early stage inhibitor, the compounds did not affect virus production from persistently infected cells (up to 50 μM) (11).

Figure 2

Effect of L-708,906 and L-731,988 on the replication of wild-type (WT) HIV-1 and diketo acid–resistant variants. Viral replication in H9 T-lymphoid cells was assayed by immunofluorescent staining for HIV-1–specific antigens (19). Cultures were incubated (A) in the absence of inhibitor or in the presence of (B) 20 μM L-708,906 or (C) 20 μM L-731,988. For each condition, the replication of wild-type HIV-1 IIIB and virus populations selected for resistance to L-708,906 and L-731,988 (20) was evaluated.

To validate integrase as the molecular target responsible for the antiviral effect, we selected HIV-1 variants resistant to L-708,906 and L-731,988. At concentrations of inhibitor sufficient to block replication of the wild-type virus (20 μM), the resistant variants replicated nearly as well as the wild-type (or resistant) virus in the absence of inhibitor (Fig. 2). Sequencing of the cDNA derived from four resistant populations consistently identified specific mutations within the integrase coding region (Table 1). Met154 → Ile154 (M154I) was detected in 18 clones from three populations selected independently for L-731,988 resistance. Although the L-708,906 population was more heterogeneous, M154I was identified in a subset of these variants. Ser153→ Tyr153 (S153Y) and Thr66→ Ile66 (T66I) were also routinely observed in the L-708,906–resistant population; T66I was identified in all but one sequence and was nearly always accompanied by either S153Y or M154I. Other changes in integrase were detected; however, only T66I, S153Y, and M154I were observed consistently, and every clone contained one or more of these mutations. The observation that Thr66, Ser153, and Met154 are conserved and proximal to the active site residues Asp64 and Glu152(12) suggests that these inhibitors bind at or near the enzyme active site. No changes were detected in either the U5 or U3 LTR.

Table 1

Mutations in HIV-1 integrase confer resistance to diketo acid inhibitors. ND, not determined; Und., undetermined.

View this table:

To demonstrate that the mutations observed in integrase were necessary and sufficient for resistance, we engineered each of the substitutions into an otherwise isogenic recombinant virus and into the recombinant integrase protein and evaluated susceptibility to L-708,906 and L-731,988. Each mutation conferred some degree of resistance to one or both inhibitors (Table 1). M154I engendered modest resistance (two- to fivefold) to both compounds, whereas S153Y affected the relative susceptibility to L-708,906 but was not sufficient to confer resistance to L-731,988. These results are consistent with the observation that only M154I was present in the L-731,988–resistant population. The data also suggest that the two inhibitors differ subtly in their interaction with the enzyme.

At best, individual mutations in integrase conferred a six- to sevenfold loss of susceptibility to either L-708,906 or L-731,988; however, increased resistance was achieved when the mutations were combined pairwise, as observed in the L-708,906–selected population. In a single-cycle HIV-1 infection assay (10), the combination of T66I and S153Y produced the largest effect with the IC50 for L-708,906 shifted by more than 20-fold, as compared with two- to threefold for either mutation alone (Table 1). The mutant viruses remained sensitive to inhibitors of protease and reverse transcriptase, e.g., L-697,661 (Table 1) (13). Although individual mutations did not appear to impair catalytic activity or infectivity, double mutants displayed decreased enzymatic activity and altered replication kinetics (14). The T66I/S153Y enzyme exhibited less than 5% of the activity of wild-type integrase; therefore, IC50's could not be determined.

To define the mechanism by which these compounds inhibit integrase and HIV-1 replication, we evaluated the activity of L-731,988 in relation to each of the discrete steps required for integration. In biochemical assays, the concentrations of L-731,988 required to inhibit 3′ processing were 70 times higher than the concentrations required to inhibit strand transfer (6 μΜ versus 80 nM, respectively) (Fig. 3). Disparate potency was also observed for disintegration (15), where the IC50 for L-731,988 was ∼20 μM with either the core domain (amino acids 50 through 212) or the full-length enzyme (16).

Figure 3

Inhibition of 3′ processing and strand transfer by L-731,988. (A) Integrase 3′ processing and (B) strand transfer activities are shown. HIV-1 HXB-2 integrase (1 μM) was incubated with 50 nM of a32P-labeled oligonucleotide representing the last 20 nucleotides of the HIV-1 U5 end. Reactions included varying levels of L- 731,988 (threefold dilutions from 100 μM, lane 12, to 5 × 10−4 μM, lane 1). Reactions were performed for 1 hour at 37°C and analyzed as described (18). The fraction of substrate converted to the specific 3′ processed product (18-nucleotide oligomer) or to strand transfer products (>20 nucleotides) was determined by PhosphorImager analysis. (C) Percent inhibition in relation to the DMSO control (lane 13); the IC50's determined for inhibition of 3′ processing (circles) and strand transfer (squares) (6 and 0.08 μM, respectively) are representative of multiple experimental determinations.

The concentration of L-731,988 required to inhibit 3′ processing (or disintegration) was greater than that required to inhibit acute infection (6 μM versus 1 μM). Therefore, to assess the effect on integrase activity in the context of HIV-1 replication, we analyzed preintegration complexes extracted at various times after infection. The kinetics of reverse transcription and 3′ processing of the U5 LTR (and U3 LTR) by integrase were identical in cells infected in the presence or absence of L-731,988. (Fig. 4A) (17). In contrast, preintegration complexes isolated from cells infected in the presence of L-731,988 were not competent to integrate when tested in vitro (Fig. 4B). The observation that L-731,988 and related inhibitors (e.g., L-708,906) affect integration activity without affecting synthesis or processing of the viral DNA is consistent with the selective profile that the compounds manifest toward strand transfer in biochemical assays. L-731,988 inhibits the final catalytic step in integration, and this specific effect on integrase is sufficient to account for the antiviral properties of the inhibitor.

Figure 4

Effects of L-731,988 on HIV-1 DNA in acutely infected cells. (A) Kinetics of reverse transcription and 3′ processing. Molt IIIB/Sup T1 cocultures were established, and cytoplasmic extracts were made at various times after infection (8, 21). Where appropriate, Sup T1 cells were preincubated with L-731,988 or L-708,906 for 1 hour before coculture. Analysis of the U5 3′ DNA end was done by indirect end-labeling of restriction digest fragments (21). Positions of the unprocessed (blunt) U5 ends, the 3′ processed U5 ends, and the plus strand strong-stop DNA (reverse transcription intermediate) are indicated. (B) In vitro activity of preintegration complexes from cells infected for 5 hours in the presence or absence of L-731,988. Preintegration complexes were isolated [as in (A)] and then incubated with (+) or without (−) ϕX174 target DNA; reactions were analyzed by Southern blotting (22). (C) HIV-1 DNA forms in infected cells in the presence or absence of L-731,988. After coculture for 24 hours [as in (A)], low-molecular-weight DNA was prepared. The DNA was digested with Sal I and Pst I and analyzed by Southern blotting with an HIV-1 LTR probe to detect all products or an internal probe. (21). Positions of linear and circular HIV-1 DNA forms are indicated. (D) Sup T1 cells were infected with wild-type and integrase-defective NL4-3 (18). DNAs were prepared and analyzed by Southern blotting [as in (C)] and by a quantitative TaqMan assay for HIV-1 two-LTR circular DNA. Diamonds, wild-type NL4-3 control; squares, wild-type NL4-3 with 10 μM L-731,988; and triangles, NL4-3 with the integrase D116N mutation (18).

To assess the outcome of this effect in infected cells, we investigated the nature of the HIV-1 DNA produced as a consequence of L-731,988 inhibition. Cells infected with wild-type HIV-1 contain both linear and circular viral DNA products (3). Circular forms of HIV-1 DNA are the products of cellular enzymes in the nucleus and are not competent for integration. With integrase-deficient viruses, circular DNA forms accumulate (3). In cells acutely infected with wild-type HIV-1, the presence of L-731,988 also led to an accumulation of circular products and a concomitant decrease in integration-competent linear viral DNA (Fig. 4C). The effect of L-731,988 on HIV-1 infection was indistinguishable from an integration-defective mutation [integrase Asp116 → Asn116 (D116N)] (18); in both infections, two LTR circles peaked within 24 hours at a level ∼10 times the level observed in the absence of inhibitor (Fig. 4D). Similar results were obtained for L-708,906 and other active diketo acids (17). By altering the kinetics of the strand transfer, the diketo acids bias the formation of replication-defective products and inhibit HIV-1 replication.

We have presented L-731,988 and the diketo acids as the archetype of a new class of integrase inhibitors and novel antiretroviral agents. The compounds are specific inhibitors of integration, which exert their antiviral effect on HIV-1 solely as a consequence of their ability to inhibit the strand transfer activity of integrase. The compounds exhibit a preference for the strand transfer reaction in vitro and inhibit integration without affecting synthesis or processing of the HIV-1 DNA in infected cells. Mutations proximal to the integrase active site residues engender resistance to the virus and the isolated enzyme, establishing the diketo acids as the first biologically validated inhibitors of integration. Although the same active site residues are required for 3′ processing and strand transfer (12), the ability of the diketo acids to discriminate between the two catalytic functions of integrase implies that a unique conformation of the enzyme may mediate strand transfer. The association of integrase with LTR sequences results in the formation of a stable, active strand transfer complex (5). We have noted that HIV-1 LTR sequences stimulate high-affinity binding of L-731,988 and related compounds to integrase, suggesting that integrase may adopt a distinct conformation subsequent to assembly (16). Together with the results presented, these observations provide biochemical and physical evidence dissociating the two catalytic functions of integrase and suggest that the diketo acids may be useful tools to probe the enzymatically active structure of integrase and to investigate the complexities of this reaction.

  • * To whom correspondence should be addressed. E-mail: daria−hazuda{at}


View Abstract

Navigate This Article