HIV-1 preintegration complexes preferentially integrate into longer target DNA molecules in solution as detected by a sensitive, polymerase chain reaction-based integration assay.

After entering a cell and undergoing reverse transcription, the retroviral genome is contained in a preintegration complex (PIC) that mediates its integration into host cell DNA. PICs have been shown to prefer torsionally strained DNA, but the effect of target DNA length has not been previously examined. In this report, concatemerization of a repeating 105-base pair unit was used to vary target DNA length independently from basic DNA sequence, while maintaining both PICs and target DNAs in solution. Integration junctions were quantified by real-time fluorescence-monitored polymerase chain reaction amplification using primers in the viral long terminal repeat and the target DNA. Unreacted target DNA severely inhibited the post-reaction polymerase chain reaction detection step, requiring its removal using lambda exonuclease digestion. Integration into a 32-unit concatemer of target DNA was markedly more efficient than integration into a monomeric unit, indicating that longer target DNA was preferred. This substrate was used to construct a simple, robust, and adaptable assay that can serve as a method for studying the host cell factors that enhance PIC integration, and as a drug discovery platform for integration inhibitors active against PICs.

After entering a cell and undergoing reverse transcription, the retroviral genome is contained in a preintegration complex (PIC) that mediates its integration into host cell DNA. PICs have been shown to prefer torsionally strained DNA, but the effect of target DNA length has not been previously examined. In this report, concatemerization of a repeating 105-base pair unit was used to vary target DNA length independently from basic DNA sequence, while maintaining both PICs and target DNAs in solution. Integration junctions were quantified by real-time fluorescence-monitored polymerase chain reaction amplification using primers in the viral long terminal repeat and the target DNA. Unreacted target DNA severely inhibited the post-reaction polymerase chain reaction detection step, requiring its removal using exonuclease digestion. Integration into a 32-unit concatemer of target DNA was markedly more efficient than integration into a monomeric unit, indicating that longer target DNA was preferred. This substrate was used to construct a simple, robust, and adaptable assay that can serve as a method for studying the host cell factors that enhance PIC integration, and as a drug discovery platform for integration inhibitors active against PICs.
The integration of HIV into the host cell genome requires the integrase enzyme (1,2) and mutations which destroy integrase activity block viral replication (3). These observations established integration as an important target for the development of new antiretroviral drugs (4). Recently, Hazuda et al. (5) found a new class of integrase inhibitors by screening a library of 250,000 compounds with a strand-transfer assay. Several di-keto compounds were found which inhibited HIV infection in CD4ϩ T cells in vitro, igniting the hope that integration inhibitors could become a new treatment modality (5).
In the strand-transfer assay used by Hazuda et al. (5), an artificial, preassembled complex was formed between recombinant integrase protein and oligonucleotides designed to model the ends of the viral cDNA. Then, candidate inhibitors were added, followed by the target DNA (6). Although more reliable than the original strand transfer assay (7), this assay still scores as positive occasional compounds which fail to inhibit integration in vitro by authentic preintegration complexes (PICs) 1 isolated from the cytoplasm of infected cells (5,8) or virus-mediated integration in cultured cells (5). This suggests important roles for the other components present in PICs, which include viral proteins (reverse transcriptase, MA, NC, and Vpr) (9, 10) together with cellular factors that are essential for the efficient integration of retroviral cDNA (11)(12)(13)(14)(15). These host cell factors may provide additional drug targets or otherwise influence the ability of a putative inhibitor to successfully inhibit integration. However, relatively little is known about PIC-mediated integration as an enzymatic process, and it has been difficult to construct a high-throughput assay for integration by PICs. Several groups have studied the influence of the target DNA upon PIC integration. PICs prefer to integrate into regions of distorted DNA (16) such as nucleosomes (17,18) and tend to avoid integration sites upstream of a pyrimidine nucleotide (19). Not addressed in these studies, however, is the issue of target DNA size, which has inherent effects on substrate mobility.
The present report describes an entirely fluid phase assay in which integration junctions are directly detected using PCR amplification. The only molecules that score as positive in this assay result from an actual joining of viral cDNA and target DNA. A uniquely designed DNA substrate is used to maintain amplicon size within the optimum limits for real-time fluorescence-monitored PCR detection. By comparing a single target DNA sequence of 105 bp with longer concatemers of the same sequence, the preference of PICs for longer target DNA was clearly demonstrated. A model is presented which suggests that long DNA is preferred because its decreased rate of diffusion allows more time for the rate-limiting "target commitment" stage of PIC integration. Incorporating this information into the assay yielded a sensitive, robust, and adaptable platform for the discovery of drugs that inhibit integration by PICs. This assay was used to detect host cell factors that reconstituted the integration competence of salt-stripped PICs.

EXPERIMENTAL PROCEDURES
Integration Substrates-Plasmid DNA containing head-to-tail concatemers encoding for drag-line silk protein was kindly provided by Dr. David Kaplan, Biotechnology Center, Tufts University (20,21). Plasmid DNA was isolated from transformed Escherichia coli DH5a and purified using a Maxi Prep Kit (Qiagen Inc., Valencia, CA). The insert in the plasmid contained monomeric units of 105 bp repeated 32 times. Insert DNA was prepared by overnight digestion with BamHI (which creates the protruding 5Ј ends favored by exonuclease, described below), purified by electrophoresis in agarose, and isolated using a QIAquick Gel Extraction Kit (Qiagen). To prepare single monomeric units of the 105-bp sequence, insert DNA was digested with StyI (which cleaves between each unit) and similarly gel purified.
Preparation of PIC-containing Cytoplasmic Extracts-PICs were prepared from the detergent lysates of cells acutely infected by HIV-1 LAI (LAV-1 strain) (22,23). Briefly, CEM cells were infected by HIV-1 LAI under BSL-3 conditions and cultured for 7-10 days until cytopathic changes (ballooning and apoptosis) were evident in the majority of cells by phase microscopic examination (24). Then, the cells were counted and co-cultured with a 10-fold excess of SupT1 cells (American Type Culture Collection, Manassas, VA) in RPMI 1640 with 10% fetal bovine serum at 5% CO 2 in a flask placed on end so that the infected CEM cells and the uninfected SupT1 cells came into contact. After 5-7 h, when the cultures consisted of about 80% syncytia just large enough to be visible to the unaided eye, the cells were pelleted by centrifugation at 160 ϫ g for 5 min. The cells were washed once by gently resuspending them in 10 mM Tris-HCl (pH 7.4), 150 mM KCl, 5 mM MgCl 2 , and 20 g/ml aprotinin (Buffer A of Ellison et al. (23)) and pelleting them again by centrifugation. Then, based on the original cell counts of the culture, the pellet was resuspended in Buffer A plus 0.025% digitonin (Calbiochem-Novabiochem Corp., San Diego, CA) at a concentration of 2 ϫ 10 7 cells/ml for 10 min at room temperature. The nuclei in the cell lysate were then pelleted at 1,000 ϫ g at 4°C for 3 min in an Avanti 30 centrifuge with a F2402H rotor (Beckman Coulter Inc., Fullerton, CA). The post-nuclear supernatant was moved to a new microtube and the remaining cellular debris was removed by centrifugation at 8,000 ϫ g at 4°C for 10 min. Forty percent sucrose was added to the supernatant to a concentration of 8% sucrose final, and the PIC-containing cytoplasmic extract was aliquotted, snap frozen in liquid nitrogen, and stored at Ϫ80°C. PIC activity was stable at this temperature for at least six months. Typically, these extracts contained ϳ0.8 mg/ml total protein measured using the BCA reagent (Pierce Chemical Corp., Rockford, IL), and 0.5-2.0 ϫ 10 8 /ml copies of HIV-1 cDNA as quantified by real-time fluorescence-monitored PCR using the LTR-specific primers and Taq-Man probe described by Rossio et al. (25).
In Vitro Integration Reactions-Integration reactions (15 l total volume) were prepared on ice in 200-l thin-walled PCR tubes (either in 8-tube strips or in 96-well plates) by adding 5 l of a master mixture consisting of 1 l of target DNA (3.3-100 ng/l, with 11-33 ng/l S32 concatemer being optimal for general use), 1.5 l of 10 ϫ integration buffer (200 mM HEPES, pH 7.4, 50 mM MgCl 2 , 10 mM dithiothreitol), and 2.5 l of 30% PEG-8000 (Sigma). Then, 10 l of PIC-containing cytoplasmic extract was added and the tube contents were mixed by pipetting up and down. The tubes were kept on ice and placed in a thermocycler (GeneAmp 9600, Applied Biosystems, Foster City, CA) and incubated at 4°C for 10 min, 37°C for 45 min, and then at 60°C for 5 min.
Post-reaction Processing-Unreacted target DNA was removed by adding 3-5 units of exonuclease in 15 l of a 2 ϫ concentration of its supplied buffer (New England Biolabs, Beverly, MA). The tubes (now containing 30 l) were returned to the thermocycler and incubated at 37°C for 45 min (which allows the exonuclease to degrade the target DNA) followed by 75°C for 10 min (to heat-inactivate the exonuclease). The processing was completed by adding 20 l of proteinase K (1 mg/ml in water) and incubating in the thermocycler at 60°C for 30 min, followed by heat inactivation at 95°C for 15 min. The processed integration reactions (now 50 l in volume) could be stored at 4°C for at least 1 week prior to TaqMan analysis. In some experiments, processed and unprocessed reactions were analyzed by polyacrylamide gel electrophoresis, staining with SYBR Gold (Molecular Probes, Eugene, OR), and visualization with UV light.
Real-time Fluorescence-monitored PCR Detection of Integration Junctions-In some experiments, the exonuclease and proteinase K-processed integration reaction mixtures were diluted 10-fold in water and then analyzed by real-time PCR. For integration reactions containing colored integration inhibitors, the inhibitor was removed by concentration/filtration of the processed integration reaction mixtures using Microcon PCR 96-well plates according to the manufacturer's protocol (Millipore, Bedford, MA). TaqMan reactions were set up in optical grade 96-well thermocycler plates by adding 20 l of a master mixture containing 12.5 l of TaqMan Universal PCR Master Mix 2X (Applied Biosystems), 2.5 l each of 9 M LTRTaq5 and SILKREV1a amplification primers (900 nM final concentration), and 2.5 l of 2 M LTRTaq5P probe (200 nM final concentration). Five l of the processed integration reaction was added to each well, and the tube contents were mixed by pipetting up and down. Real-time fluorescence-monitored PCR reactions (TaqMan) were performed on an Applied Biosystems Model 7700 Sequence Detection System. The temperature profile for the reaction was: 50°C for 2 min, 95°C for 10 min, and then 95°C for 15 s and 60°C for 1 min for 45 cycles. Using the manufacturer's software, the cycle number at which fluorescence exceeded background (C t ) was determined for each well. For each real-time PCR analysis, a standard curve was generated using dilutions in water of a cloned integration junction (below) calculated to provide 3, 10, 100, and 1000 integration junctions per well. The reactions were set up as 2-or 3-fold replicates and typically differed by less than 0.4 C t .
Data Analysis of the Integration Assay-To deduce the number of integration junctions in each sample, its C t value was compared with the cloned integration junction standard curve following linear regression analysis. Because each TaqMan reaction used only 5 l of the 50 l of processed integration reaction, the number of integration junctions was then multiplied by 10 to arrive at the number of integration events produced by the original 10 l of PIC-containing cytoplasmic extract. If the processed reaction was diluted prior to TaqMan analysis, then an additional correction was made for the dilution factor. No correction was made, however, for the presumably equal number of integrations into the other, unmeasured strand of the target DNA. Statistical calculations were performed using the InStat software program (Graph-Pad Software Inc., San Diego, CA).
Hemi-nested PCR for Cloning and Sequencing of Representative Integration Junctions-Hemi-nested PCR was used to isolate PCR products in order to confirm that the method identifies integration junctions. For the first round of PCR, integration reactions were diluted 1:1,000 in water and 5 l was amplified in a 25-l reaction using HIVout and SILKREV1a (above) and HotStarTaq (Qiagen) according to the manufacturer's instructions. The reactions were heated to 95°C for 15 min; then 95°C for 15 s, 55°C for 15 s, and 72°C for 30 s for 25 cycles; and finally 72°C for 10 min. The first round reaction products were then diluted 1:100 with water and re-amplified in an identical manner using HIVin2 and SILKREV1a. The final products were electrophoresed in agarose, isolated using a using a QIAquick Gel Extraction Kit (Qiagen), and cloned into the pCR4-TOPO vector using a TOPO TA Cloning Kit (Invitrogen Corp., Carlsbad, CA). Dye-terminator sequencing of eight clones was performed by the Molecular Biology Core of the UCSD Center for AIDS Research using the kit-supplied primers. One of the sequenced clones was used to create an integration junction standard curve in the TaqMan reaction above.
Preparation of Salt-stripped PICs-Salt-stripped PICs were prepared by a modification of the methods of Lee and Craigie (26) and Chen and Engelman (15). 500 l of PIC-containing cytoplasmic extract was diluted in an equal volume of Buffer A (above) containing 0.025% digitonin but formulated without KCl to reduce the salt concentration to 75 mM. The diluted PICs were incubated at 4°C for 30 min and then pelleted at 8,000 ϫ g for 20 min in a refrigerated microcentrifuge. The barely visible pellet was then resuspended in 170 l of Buffer A with 1.2 M KCl and incubated for 30 min on ice. Then, the hypertonic PIC solution was loaded onto a Sepharose CL-4B column (2.2 ml bed volume) that had been pre-equilibrated in the same buffer (but not pretreated with bovine serum albumin). The column was placed in a tube and centrifuged at 800 ϫ g for 3 min at 4°C. About 1,500 l was retrieved from each column and then concentrated to about 40 l by centrifugation at 800 ϫ g in a Microcon-100 ultrafiltration unit (Millipore), taking care not to allow the sample to concentrate to dryness. The amount of HIV LTR cDNA was then quantified by TaqMan (above) so that samples could be adjusted to equal numbers of PICs before placing them in the integration assay.
Reconstitution of Salt-stripped PICs-Cytoplasmic extracts were prepared from uninfected SupT1 cells using exactly the same protocol used to prepare PIC-containing extracts. 5 l of this extract was added to 2 l of salt-stripped PICs and incubated on ice for 15 min. Then, 5 l of Buffer A without salt was added to bring the KCl concentration down to about 262 mM, and the mixture was incubated on ice for another 15 min to allow the cytoplasmic proteins to assemble onto the salt-stripped PICs. Then, 10 l of this mixture was analyzed in the integration assay above.

Design of a Concatemeric Target DNA and Primers for the Detection of Integration Junctions
To investigate the effect of target DNA length on the efficiency of integration, we chose a concatemeric DNA substrate containing 32 head-to-tail monomeric units encoding dragline silk, a natural protein composed of repeating units (27). To amplify integration junctions, PCR primers were designed to hybridize to the 3Ј LTR of HIV-1 (forward primer) and the silk target DNA (reverse primer). The concatemeric arrangement of the target DNA places a binding site for the reverse primer within a short distance (Յ183 bp) from the LTR forward primer and effectively limits amplicon size to a length that is optimal for TaqMan detection. Although both the 5Ј and 3Ј LTRs integrate into the target DNA during concerted integration, only the integration of the 3Ј LTR was analyzed in these studies. Also, although HIV can integrate into the target DNA in either orientation, the assay only detects integration junctions in one of the two target DNA strands (shown as the top strand in Fig.  1).

Integration Junctions Are Correctly Amplified by the PCR Assay System
As a first test of the PCR system depicted in Fig. 1, amplicons containing putative integration junctions were isolated by hemi-nested PCR and cloned into a plasmid. Sequencing analysis revealed that the integrations had occurred into various sites of the target DNA, with a possible "hot spot" immediately 5Ј to the reverse primer (data not shown). In several cases, the cloned amplicons contained an additional monomeric unit of target DNA on their 3Ј end resulting from the hybridization of the reverse primer to a binding site in the next unit adjacent to the integration site.

Real-time PCR Detection of Integration Events
To validate the real-time PCR detection step, control reactions were set up lacking either PICs or target DNA. As expected for a 45-cycle real-time PCR, these negative control integration reactions yielded C t values equal to 45, indicating that no integration junctions were detected. To quantify positive reactions, dilutions of a cloned integration junction were prepared to create the standard curves used to relate the C t values obtained from an unknown sample to the number of integration events that occurred. When reaction mixtures were diluted prior to detection, the final number of integration events reported was calculated by multiplying by the dilution factor. No correction was performed to account for the presumably equal number of integrations that occurred into the opposite, unmeasured strand of the target DNA. Fig. 2 presents data representative of five experiments. When PICs and target DNA were both present in an integration reaction and the exonuclease processing step was omitted, a signal corresponding to 500 -1,000 integration events was detected. If the reaction mixtures were diluted prior to PCR, however, the number of integration events detected (corrected for the dilution factor) increased, indicating that the reaction mixtures contained an inhibitor for the PCR detection step. This disturbing nonlinearity likely explains why there has been no previous report of a quantitative PIC assay based upon the PCR quantification of integration junctions. After an exhaustive analysis, it was determined that the principal cause of this inhibition of detection is the carryover into the PCR detection step of unreacted target DNA, which contains a very large number of binding sites for the reverse primer. For example, the addition of target DNA, but not irrelevant plasmid DNA, to a cloned integration junction dramatically inhibited real-time PCR detection. As a first approach to solve this problem, the concentration of the reverse primer was increased 10-fold (to 9 M final concentration) which completely reversed the inhibitory effects of target DNA on the detection of 10 and 100 copies of a cloned integration junction. However, for the integration junctions generated by PICs in the reaction mixtures, PCR amplification critically depends upon the reverse However, in part because the great excess of binding sites for the reverse primer interferes with the subsequent PCR reaction, unreacted target DNA is removed using exonuclease which digests DNA in the 5Ј 3 3Ј direction. Proteins bound to the LTR protect the integration junction from exonuclease digestion (see text for discussion). Next, proteinase K is used to remove the exonuclease and other proteins, following which the proteinase K itself is inactivated by heat treatment. Bottom panel depicts the PCR detection step. The first cycle is critical as it depends solely on the reverse primer to create the strand complementary to the integration junction. Once the double-stranded amplicon has been formed (containing a binding site for the forward primer at the position shown by the open box in the LTR), succeeding cycles of PCR can utilize both primers and exponential amplification results. During the real-time PCR detection process, a TaqMan probe binds to an additional site in the LTR (shown as the hatched box). The 5Ј exonuclease activity of Taq polymerase then cleaves the fluorophore (F) in the probe away from the quencher (Q), generating the signal measured by fluorescence detection.

Target DNA Preferences of HIV-1 Preintegration Complexes
primer creating the complementary strand to which the forward primer binds (whereas this strand pre-exists when a cloned integration junction is used). In this case, a simple increase in the reverse primer concentration was not sufficient to eliminate the nonlinearity of the assay (data not shown).
A close examination of the amplification plots revealed an additional problem caused by the carry-over of target DNA into the TaqMan reaction. In these cases, the slope of the amplification plot was markedly less than that of the cloned integration junction that was used as a standard. In effect, the target DNA prevented the amount of product from doubling with each cycle of PCR. The atypical shape of the amplification plots from the samples containing target DNA reduced the sensitivity of the TaqMan analysis by raising the apparent C t , and also precluded the possibility of converting the assay to an end point detection format.
Fortunately, two strategies were found which succeeded in removing the nonlinearity of the assay by reducing or eliminating the carryover of target DNA into the PCR detection step. The first strategy is simply to dilute the integration reaction mixtures prior to PCR detection. However, the drawback of diluting the reaction mixtures is that this also dilutes the integration junctions that the real-time PCR aims to detect. For example, diluting the reaction mixtures beyond 100-fold prior to TaqMan detection generally resulted in C t values outside of the dynamic range of the assay. A second strategy is to selectively remove the target DNA in order to prevent it from inhibiting the PCR detection step. This approach was not associated with the loss of sensitivity that occurred when the dilution method is employed. Exonuclease, which degrades double-stranded DNA containing phosphorylated, protruding 5Ј ends, was employed for target DNA removal prior to the PCR detection step. An examination of the exonuclease-treated reactions by polyacrylamide gel electrophoresis and SYBR Gold nucleic acid staining showed that all visible DNA in the reactions had been removed (data not shown). However, integration junctions evidently survive this exonuclease step because PIC proteins bound to the DNA block the 5Ј 3 3Ј digestive action of this exonuclease as it approaches the integration site (12). These PIC proteins may be part of the complex (the "intasome") that has been detected on the ends of integrated Moloney murine leukemia virus LTR DNA using a sensitive DNA footprinting method (28). Supporting the concept that exonuclease effects the selective removal of target DNA, pretreatment of an integration reaction with exonuclease prior to detection consistently resulted in a 4 -10-fold increase in detection sensitivity (Fig. 2). Also, the amplification plots of the exonuclease-treated samples showed a normal slope in parallel to the cloned integration junction standards, which could be important for converting the assay into a simplified end point detection format.

A Longer, Concatemerized Target DNA Permits the Detection of a Higher Number of Integration Events
To study the effects of target DNA length on the efficiency of PIC integration in vitro, target DNAs consisting of either 1 (S1) or 32 (S32) monomeric units of the same silk coding sequence were used in the integration assay system (Fig. 3). Two major conclusions were drawn from these experiments: 1) for a given number of monomeric units of target DNA present in the integration reaction, the concatemerization of those units into a longer molecule resulted in a 2-to Ͼ10-fold increase in the integration efficiency. 2) Overall, integration efficiency rises with an increase in the amount of target DNA molecules (in terms of monomeric units) (Fig. 3). This indicates that the amount of target DNA in the integration reaction must be kept high in order to maximize the number of integration events, even though the carryover of target DNA inhibits the subsequent real-time PCR detection step. As a counterbalancing factor, the linearity of the assay for PICs is best at Յ9.27 ϫ 10 10 monomers/15 l reaction (see below and data not shown), a concentration where only the 32-mer can be used.

Applications of the PIC Integration Assay
Quantifying Integration-competent PICs-To validate that the integration assay system using long target DNA concatemers can actually measure integration-competent PICs, PIC extracts were serially diluted and then assayed. Based on the data obtained in the experiments described above, the number of integration events was quantified in undiluted integration reactions using 32-mer concatemeric target DNA and exonuclease treatment in order to maximize detection efficiency. As shown (Fig. 4), the number of integration events decreased proportionally to PIC dilution. For the PIC extract tested, dilutions beyond 9-fold produced C t values that were outside of the linear range of the assay. In other experiments, the target DNA concentration was reduced resulting in less inhibition of the real-time PCR detection step. In these cases, integration events were detectable in PIC extracts diluted 27-fold, even though the absolute number of integration events at each dilution was lower. These results indicate that the PIC assay system using long concatemeric target DNA can reliably quantify the number of integration competent PICs in a sample. The system is so sensitive that a single 50-ml culture of infected SupT1 cells generates enough PICs for over 10,000 assays, which makes it possible to use authentic HIV-1 PICs in a drug discovery program. Additionally, the assay can be used to guide the purification of PICs by ultracentrifugation in a sucrose gradient (data not shown).

Detection and Characterization of Integrase-directed PIC Inhibitors (IPIs)-A number of compounds have been shown to inhibit HIV-1 integration in vitro.
To determine whether PICs could be preincubated with candidate inhibitors, the stability of PIC extracts at 37°C was first examined. As shown Fig. 5 (inset), PIC-containing cytoplasmic extracts were stable at 37°C for about 20 min, but became unable to integrate into target DNA by 1 h. Next, two known inhibitors of HIV-1 integrase, purpurin and quinalizarin, were studied using the PIC integration assay system. As expected, both integrase inhibitors were clearly identified as integration inhibitors using PICs (Fig. 5), indicating that an assay based on long, concatemeric target DNA can be used to identify IPIs. (11) introduced the technique of removing necessary factors from PICs under conditions of high salt. These "salt-stripped" PICs can then be recombined with extracts from uninfected host cells in order to identify cellular factors that may be important for retroviral integration. Consequently, HIV-1 PICs were saltstripped and found to have no detectable integration competence using the PIC integration assay system with long, concatemeric target DNA. However, when cytoplasmic extracts of uninfected SupT1 cells were added to these salt-stripped PICs, their integration activity was restored (Fig. 6). This indicates that this PIC integration assay system can be applied to the study of the host cell factors that contribute to PIC activity. In addition, this configuration of the assay can be used to identify a putative new category of antiretroviral compounds which would inhibit the functioning of PIC-activating host cell factors. Such nonintegrase-directed PIC inhibitors (NIPIs) might work either by preventing the assembly of an active PIC or by disrupting the activity of an already fully assembled PIC. No previous high-throughput assay has been reported to be applicable for the detection of NIPIs. In particular, the strandtransfer assay that is used in current drug discovery efforts can only be used to identify IPIs but not NIPIs. DISCUSSION PCR amplification is a highly sensitive and time-efficient approach to detect the covalent joining of two DNA substrates. The presence of an amplification product in a PCR reaction with primers specific to viral and target DNA molecules participating in an integration reaction would indicate that an integration event had occurred. Recently developed, highly sensitive, real-time fluorescence-monitored PCR technology provides a way to obtain accurate quantification of a PCR product within a 2-3-h period. This approach has maximum amplification efficiency and sensitivity for small amplicons. However, host genomic DNA, the natural target of retroviral integration, has substantial length and structures that are characteristic for longer DNA molecules. To mimic this long target DNA, plasmid DNA was previously employed as a target DNA in conjunction with PCR-based detection using LTR-specific and plasmid DNA-specific oligonucleotides (29). However, the FIG. 4. Enumeration of integration-competent PICs. PICs were serially diluted in the extraction solution and placed into integration reactions using 11 ng/reaction of target DNA. The reaction mixtures were treated with exonuclease and analyzed without dilution in the real-time fluorescence-monitored PCR assay. The data shown are the means of paired duplicate samples. A clear dose-response relationship is apparent between the number of PICs added and the number of integration events detected. The linear trend of this dose-response data was highly significant as determined by one-way ANOVA (r 2 ϭ 0.9545, p Ͻ 0.0001). As an additional statistical measure, a single sample was analyzed in 8 separate reactions and found to have a mean of 479 integration events with a co-efficient of variation (c.v.) of 31.45%.

FIG. 5. Detection and quantification of inhibitors of PIC integration.
In this experiment, PICs were preincubated at 37°C prior to the addition of target DNA. Thereafter, the integration reaction mixtures were treated with exonuclease and diluted 10-fold prior to real-time fluorescence-monitored PCR detection. As shown in the inset, PIC-containing cytoplasmic extracts are unstable when incubated at 37°C for more than 20 min. With this stability limitation in mind, PICs were preincubated with various concentrations of two known inhibitors of HIV-1 integrase, purpurin (f) or quinalizarin (OE), at room temperature for 10 min. Then, the target DNA was added, the reaction mixtures were treated with exonuclease, and the number of integration events was determined in undiluted samples following the removal of the interfering pigmented integration inhibitors by ultrafiltration. A dosedependent inhibition of PIC integration by the integrase inhibitors was seen.
length of the PCR product amplified as a result of integration varies significantly in this format, since the retroviral genome integrates nearly randomly into the target DNA. Thus, an integration assay utilizing plasmid DNA as a target DNA could not be adopted for the real-time PCR detection of integration events.
Alternatively, a short DNA molecule could be utilized as a target DNA in an integration assay. In this case, one can place an upper bound on the length of the PCR amplicon containing the integration junction. On the other hand, a short DNA molecule has little resemblance to host genomic DNA and is a very poor substrate for PIC integration (Fig. 3). This contrasts with the strand-transfer assay for integrase, where short target DNAs are typically used for the integration reaction (6,7).
Utilization of long, concatemeric DNA as an integration target overcomes these limitations. By designing a forward oligonucleotide primer near the very end of the 3Ј LTR and a reverse primer for a site in each 105-bp monomeric unit, amplicon size was maintained within the limits optimal for efficient real-time PCR detection (Fig. 1). Several thousand integration events were routinely detected in 10 l of PIC extract using real-time fluorescence-monitored PCR detection based on measurements for 3Ј LTR integration only.
Concatemeric target DNA provided an opportunity to conduct a systematic study of the relationship between target DNA length and the efficiency of HIV-1 cDNA integration. This analysis was performed under conditions where the target DNA sequence remained constant but only the degree of multimerization varied. The results of these experiments were normalized in terms of an absolute number of monomeric units present in an integration reaction. From this analysis, concatemerization of the target DNA resulted in a 2-to Ͼ10-fold increase in the number of integration events detected, which also depended on the concentration of target DNA (number of monomers) present in the integration reaction (Fig. 3). In this context, a recent study on the diffusion coefficients of doublestranded DNA molecules showed that they were highly dependent on DNA length and the media in which the DNA was tested. In the cytoplasm of HeLa cells, a large DNA fragment (Ͼ2,000 bp) diffused 20 times slower than a smaller, 100-bp DNA fragment (30). This raises the possibility that PICs prefer to integrate into a longer 32-mer target DNA (3,360 bp) because a slowly moving target DNA favors the formation of a PIC-target DNA complex. Maintaining target DNA in solution may be vital for optimum sensitivity, however. In a previous highthroughput PIC assay, target DNA was immobilized to the bottom of 96-well plates to capture integrated PICs which were subsequently detected by real-time fluorescence-monitored PCR with primers specific solely for the viral LTR, not the integration junctions (31). In this assay, which we estimate is 20 times less sensitive than the assay in the present report, PICs must diffuse to the bottom of the wells in order to interact with the immobilized target DNA, making diffusion a ratelimiting factor for integration.
Although it was originally reported that excess DNA does not affect PCR amplification reactions including real-time PCR (32), recent studies have shown that large amounts of doublestranded DNAs can completely suppress the PCR reaction (33,34). In the present study, the excess target DNA that was required for the integration reaction was carried over to the PCR detection step, where it became the major inhibitor of PCR amplification. This complication is the likely reason why there has been no previous report of a quantitative PIC integration assay using PCR to detect integration junctions. Two independent ways were devised to overcome the inhibition of PCR detection caused by target DNA carryover: 1) dilution of the integration reaction mixture prior to detection; and 2) selective degradation of target DNA by exonuclease. Overall, the combination of exonuclease treatment and 10-fold dilution resulted in the highest number of integration events detected in a given sample, after adjustment was made for the dilution.
This study also demonstrated that the ability of PICs to integrate decays within 1 h at 37°C under the conditions studied (Fig. 5, inset). This may be due to autointegration by PICs resulting from intramolecular or intermolecular interactions with HIV-1 cDNA (35). In this context, the preference for longer target DNA may reflect an improvement in target commitment caused by decreased diffusion of the target DNA away from the PIC. Through target DNA competition studies, Miller et al. (36) established that cytoplasmic extracts contain factors which interfere with the ability of PICs to quickly integrate into a target DNA molecule with which they may come into contact. The decreased diffusion of a long target DNA molecule may act to prolong the period of contact with a PIC, thereby facilitating the proposed target-induced conformational changes in the arrangement of integrase molecules on the LTR ends (37), before the decay of PIC activity becomes a limiting factor.
The PIC integration assay was successfully used to quantify the number of integration-competent PICs in a sample over a dynamic range of 1-1.5 orders of magnitude, depending on amount of target DNA used. Integration events from 10-l samples containing as few as 5 ϫ 10 4 copies of cDNA could be measured with a high degree of reproducibility.
The reliable measurement of integration events is particularly important for the development of integration inhibitors and for studies on the mechanism of retroviral integration. The activity of two previously described integrase inhibitors, purpurin and quinalizarin, was demonstrated in this assay, with IC 50 values that were similar to previously published values (8). In addition, the experiments showing the reconstitution of salt-stripped PICs by cytoplasmic extracts from uninfected host cells (Fig. 6) demonstrate the possibility of using this assay to detect NIPIs, a putative new category of antiviral drugs.
In conclusion, we have developed a quantitative assay for the direct detection of authentic integration junctions. The advantage of utilizing a long, concatemeric target DNA was shown. In its final form, the assay consists of only three steps performed in a single tube and yields quantitative data within 4 -5 h. The Host cell factors were removed from PICs by salt-stripping as described under "Experimental Procedures" and analyzed in the integration assay. The dotted line shows the lower limit of sensitivity in the real-time PCR detection system. By themselves, the salt-stripped PICs were not competent to integrate. However, with the addition of a cell extract from the SupT1 CD4 ϩ T cell line (15), integration competence was reconstituted. assay was successfully applied to measure the number of integration-competent PICs, to guide the purification of PICs, to detect host cell factors which restore the integration competence of salt-stripped PICs, and to characterize integration inhibitors. Converting from real-time PCR detection to an end point detection system will enable this assay to be used in a high-throughput screen for novel retroviral integration inhibitors of either the IPI or NIPI categories.