We recently reported the development of a mini-HIV-1 variant for virotherapy of T-ALL 13. This minimized HIV-1 derivative carries five deletions (vif, vpR, vpU, nef and U3). The deleted genes/motifs contribute to virus replication in untransformed cells, but are dispensable for replication in leukemic T-cells. To obtain control over virus replication, we now combined the mini-HIV approach with the dox-dependent HIV-rtTA concept 1618. The nef gene in HIV-rtTA is replaced by the gene encoding the rtTA protein (Fig. 1). In the presence of dox, the transcriptional activator rtTA protein binds to tetO binding sites that were introduced in the U3 domain of the LTR promoter (Fig. 1, black box). Tat-mediated transcriptional activation is abrogated by an inactivating mutation in the tat gene (Tyr26Ala)1920 and multiple inactivating mutations in the TAR hairpin (indicated by crosses in Fig. 1) 16. HIV-rtTA also carries a deletion of a large upstream part of the U3 domain 21 and thus represents a Δ4 HIV-rtTA genome (Fig. 1, ΔTAR, tat, nef, U3). HIV-rtTA was further minimized by deletion of genes encoding the accessory proteins Vif and VpR. Additionally, the vpU gene was inactivated in rtTAΔ6^A (Fig. 1) by mutation of the startcodon (AUG to AUA) and in rtTAΔ6^B by gene deletion. Due to the vpU-cloning procedure, the wild type tat gene was restored. These minimized rtTAΔ6^A and rtTAΔ6^B variants express the basic set of HIV-1 proteins (gag, pol, env), the essential Rev and Tat proteins, but lack the accessory proteins Vif, VpR, VpU and Nef. The RNA genome of rtTAΔ6^B is 8,872 nt compared to 9,229 nt for full length HIV-1 LAI and 9,607 nt for the parental HIV-rtTA virus.

Viral gene expression and production of virus particles was tested by transfection of the mini-rtTA plasmids in C33A cells. These cells lack the CD4 receptor and are thus not susceptible for multiple rounds of HIV-1 replication. We measured no difference in virus production of the mini-rtTA viruses compared with the original HIV-rtTA construct (Fig. 2). All constructs are fully dependent on dox for gene expression. These results demonstrate that none of the deleted/mutated genes/motifs play an important role in viral gene expression (transcription, splicing, and translation) and the assembly of new virions. Virus production of all dox-dependent rtTA viruses is somewhat lower than that of the wild type LAI virus, consistent with our previous studies 1622.

The virus stocks produced in C33A cells were used to infect the HIV-susceptible leukemic T-cell line SupT1. Virus replication was followed by sampling of the culture supernatant and measurement of the CA-p24 concentration (Fig. 3, left panel). Surprisingly, replication of the minimized rtTAΔ6^A and rtTAΔ6^B variants is significantly faster than that of the parental HIV-rtTA virus and even faster than the wild type LAI virus. Similar results were obtained in multiple replication assays that were initiated either by virus infection or by transfection of the molecular clones (results not shown). Direct virus competition assays confirmed this ranking order, with rtTAΔ6^A being slightly more fit than rtTAΔ6^B, and both much more fit than HIV-rtTA (results not shown). This surprise finding will be dealt with in detail later in this paper. As expected, virus replication is fully dependent on dox addition.

The T-cell cultures were also analyzed for the cell killing capacity of these viruses. A time-limited FACS analysis was used to determine the relative number of live cells in the infected cultures and a mock-infected SupT1 culture as the control. The cell killing capacity was determined by dividing the number of cells in the infected culture by the number of cells in the control culture (Fig. 3, right panel). The LAI virus and the different HIV-rtTA variants are able to kill all SupT1 cells. The cell killing kinetics correlate nicely with the replication capacity of the respective viruses. There was no decrease in the number of live cells when the HIV-rtTA virus was tested without dox, confirming that the increase in cell death is the result of active virus replication.

We tried to set up experiments with patient derived primary leukemic T-cells but the high death rate of these cells in in vitro culture experiments (without any virus) prevented any significant conclusions to be reached about virus-induced cell killing (results not shown).

Dox-regulation should allow strict control over replication of the therapeutic viruses. To demonstrate the regulatory possibilities of this system, we followed several rtTAΔ6^A cultures with different dox regimens, ranging from no to continuous dox treatment. We also tested delayed dox addition and dox-withdrawal near the peak of infection. Virus infections were started with dox (Fig. 4, upper left panel) or without dox (Fig. 4, lower left panel) and virus production was followed by measurement of the CA-p24 concentration in the supernatant. After nine days, the cultures were split and either continued with the same treatment (Fig. 4, left panels) or switched from dox to no dox (Fig. 4, upper right panel) or vice versa (Fig. 4, lower right panel). The results show that virus replication is completely controllable by dox. In the cultures with dox a productive infection is started that can be turned off by withdrawal of dox. In the cultures that were started without dox, a single round of infection takes place that leads to the establishment of an integrated but silent provirus, which can subsequently be activated by the addition of dox.

The different HIV-rtTA viruses were further analyzed by testing their replication capacity on PBMC (Fig. 5, left panel). Killing of the CD4+ target cells was plotted as the CD4+/CD8+ ratio relative to that of the control culture without dox (Fig. 5, right panel). The wild type HIV-1 LAI isolate replicates efficiently, resulting in a high peak of CA-p24 production and complete removal of the CD4+ cells from the PBMC culture within 5 days. Due to the removal of target cells, the CA-p24 concentration reaches a maximum at 3 days post infection and subsequently levels off. The parental HIV-rtTA virus replicates slowly, but eventually reaches CA-p24 values similar to that of the wt virus. In this culture, a gradual reduction in CD4+ cells is scored, but HIV-rtTA replication is completely dependent on dox addition. No production of CA-p24 was measured in the PBMC cultures infected with the minimized rtTAΔ6^A and rtTAΔ6^B variants, and no significant reduction in the CD4+/CD8+ ratio was observed. Thus, these viruses are unable to cause a spreading infection in PBMC. Extending the time for replication by feeding these cultures with fresh PBMC did also not result in a spreading infection.

It cannot be excluded that the rtTAΔ6^A and rtTAΔ6^B viruses replicate at an extremely low level, and thus stay below the CA-p24 detection limit. To test for this, we used a very sensitive SupT1-based rescue assay to screen for viable virus in the PBMC cultures. PBMC were harvested at day 13, washed and subsequently co-cultured with SupT1 cells. Virus replication is readily observed in the control co-cultures derived from the LAI and HIV-rtTA infections. No virus could be detected in the cultures derived from the rtTAΔ6^A or rtTAΔ6^B infections, even with 1000-fold more input sample compared to the LAI or HIV-rtTA samples.

In a virotherapy setting, the blood of a patient will contain a mixture of leukemic and untransformed cells. The viral therapeutic agent should selectively replicate and kill the leukemic target cells without affecting the untransformed cells. To mimic this situation in our in vitro culture system, we started co-cultures of the SupT1 cell line and PBMC. These cells can easily be distinguished by FACS analysis using the CD4 and CD8 surface markers. SupT1 cells are double positive T-cells (CD4⁺CD8⁺), whereas PBMC contain a mixture of single positive CD4⁺CD8^- and CD4^-CD8⁺ cells (Fig. 6, left). A PBMC-SupT1 culture was split in five samples. These cultures were infected with an equal amount of HIV-rtTA, rtTAΔ6^A or rtTAΔ6^B virus. The two remaining cultures were used for a mock infection and a control rtTAΔ6^A infection without dox. The cell composition was followed over time by FACS analysis, showing the more rapid proliferation of leukemic SupT1 cells versus PBMC in the uninfected control (mock, upper panels). The same result was obtained for the rtTAΔ6^A control without dox (rtTAΔ6^A-dox, lower panels). In contrast, the SupT1 cells are selectively depleted in 8 days from the cultures containing rtTAΔ6^A or rtTAΔ6^B virus with dox. In agreement with the slower replication kinetics of HIV-rtTA in SupT1 cells (Fig. 3), SupT1-depletion is delayed for this virus. These results indicate that it is possible to selectively remove leukemic T-cells from a mixture with untransformed cells by the use of a dox-controlled mini-HIV-1 variant.

We constructed two dox-regulated viruses that specifically target leukemic T cells. A surprising finding was that these viruses, with many deletions (Δ6), replicated much better in SupT1 cells than the parental construct HIV-rtTA (Fig. 3). In the construction of rtTAΔ6^A and rtTAΔ6^B, the wild-type tat open reading frame is restored when compared to the rtTA virus that carries the Y26A inactivating Tat mutation. Although Tat-mediated transcriptional activation is not needed for replication of the dox-controlled virus, it is possible that Tat restoration enhances virus replication by other means, which may explain the enhanced replication of rtTAΔ6 variants.

To test this hypothesis, the Y26A mutation was reintroduced in the rtTAΔ6^A and rtTAΔ6^B background, yielding the rtTAΔ7^A and rtTAΔ7^B viruses, respectively. For comparison, the mutant tat gene (Y26A) in the HIV-rtTA virus was also replaced by the wild-type tat gene from the LAI isolate, yielding rtTAΔ3 (LAI), or the tat gene from the NL4-3 isolate, yielding rtTAΔ3 (NL4-3). This set of viruses was used to infect SupT1 cells to test their replication capacity (Fig. 8). Comparison of the replication capacity of rtTAΔ6^A versus rtTAΔ7^A, rtTAΔ6^B versus rtTAΔ7^B (Fig. 8, left) and rtTAΔ3 (LAI) versus HIV-rtTA (Fig. 8, right) demonstrate that the Y26A Tat mutation causes a small decrease in replication. Thus, a wild-type tat gene improves replication. The introduction of the NL4-3 tat gene in HIV-rtTA, however, improved replication much more than introduction of the tat gene of the LAI isolate (Fig. 8, left panel, compare rtTAΔ3 (NL4-3) with HIV-rtTA and rtTAΔ3 (LAI)). In fact, the rtTAΔ3 (NL4-3) variant replicated consistently better that the wild type LAI virus. Similar results were obtained in repeated infections and in replication studies that were initiated by DNA transfection (results not shown).

Full-length molecular HIV-1 clones are based on an improved variant of the dox-inducible HIV-1 variant described previously 38. We first deleted the accessory proteins vif and vpR in this HIV-rtTA virus. Plasmid pDR2483 39, which contains the 5' genome of the HIV-1 isolate NL4-3 with deletions in the genes encoding the vif and vpR proteins, was used as template in a PCR reaction with primers RJ001 (5' GGG CCT TAT CGA TTC CAT CTA 3') and 6 N (5'CTT CCT GCC ATA GGA GAT GCC TAA G 3'). The resulting PCR fragment was cut with ClaI and EcoR1 and ligated with a 9644 bp BclI-EcoR1 HIV-rtTA vector fragment and a 1816 bp BclI-ClaI fragment from pLAI-001 13 to generate the subclone rtTAΔvifΔvpR. We noticed a vpU startcodon inactivation (AUG to AUA) in one of the evolution cultures 13. Proviral DNA was PCR amplified from total cellular DNA of this culture with the primers Pol5'FM (5'TGG AAA GGA CCA GCA AAG CTC CTC TGG AAA GGT 3') and WS3 (5'TAG AAT TCA AAC TAG GGT ATT TGA CTA AT). The same PCR was performed on DNA from a vpU-deletion construct [pDR2484, 39]. The PCR fragments were cut with EcoRI and NdeI and ligated with a 2086 bp wild type (wt)t rtTA BamHI-NdeI fragment and the vector rtTAΔvifΔvpR cut with EcoRI and BamHI. The resulting molecular clones (Fig. 1) were named rtTAΔ6^A (vpU startcodon inactivation) and rtTAΔ6^B (vpU deletion).

As part of the vpU inactivation strategy, the Y26A inactivating mutation in the tat gene of HIV-rtTA is replaced by the wt tat gene of the NL4-3 isolate (first exon). The Y26A mutation was cloned back into the rtTAΔ6^A and rtTAΔ6^B molecular clones as follows. A PCR was done with HIV-rtTA as template and primers Pol5'FM and RJ036 (5'CTT TTG TCA TGA AAC AAA CTT GGC A 3'). The latter primer introduces a BspHI site that is also present in the wt NL4-3 sequence. The PCR product was digested with EcoRI and BspHI and used in a triple ligation with the 9028 bp EcoRI-BamHI vector and either the 2545 bp BspHI-BamHI fragment of rtTAΔ6^A or the 2428 bp BspHI-BamHI fragment of rtTAΔ6^B. For comparison, we also introduced the LAI and NL4-3 tat gene into the HIV-rtTA background. For NL4-3, this was done in a triple ligation with the rtTA vector cut with SphI and Asp718 I, the 4378 bp SalI-SpHI rtTA fragment and the 558 bp SalI-Asp718 I fragment of pDR2480 39. For LAI this was done by ligation of the 9646 bp NcoI-BamHI digested HIV-rtTA vector with the 2811 bp NcoI-BamHI fragment of LAI.

All constructs were verified by restriction enzyme digestion and BigDye terminator sequencing (Applied Biosystems, Foster City, CA) with appropriate primers on an automatic sequencer (Applied Biosystems DNA sequencer 377). Plasmid DNA isolation was done with the Qiagen Plasmid isolation kit according to the manufacturers' protocol (Qiagen, Chatsworth, CA).

Culture supernatant was heat inactivated at 56°C for 30 min in the presence of 0.05% Empigen-BB (Calbiochem, La Jolla, USA). CA-p24 concentration was determined by a twin-site ELISA with D7320 (Biochrom, Berlin, Germany) as the capture antibody and alkaline phosphatase-conjugated anti-p24 monoclonal antibody (EH12-AP) as the detection antibody. Detection was done with the lumiphos plus system (Lumigen, Michigan, USA) in a LUMIstar Galaxy (BMG labtechnologies, Offenburg, Germany) luminescence reader. Recombinant CA-p24 expressed in a baculovirus system was used as the reference standard.

C33A cervix carcinoma cells [ATCC HTB31, 40] were grown as a monolayer in Dulbecco's minimal essential medium supplemented with 10% (v/v) fetal calf serum (FCS), 100 U/mL penicillin, 100 μg/mL streptomycin, 20 mM glucose and minimal essential medium nonessential amino acids at 37°C and 5% CO₂. The cells were transfected by the calcium phosphate method as described previously 41.

The human T lymphocyte cell line SupT1 [ATCC CRL-1942, 42] was cultured in RPMI 1640 (Gibco BRL, Gaithersburg, MD) supplemented with 10% (v/v) FCS, 100 U/mL penicillin, and 100 μg/mL streptomycin at 37°C and in 5% CO₂. Transfections were carried out by electroporation as described 43 using a BioRad Gene Pulser II (BioRad, Hercules, CA). Infection with C33A produced virus stocks was performed with the indicated amount of virus.

PBMC were isolated from different healthy donors, each batch consists of a mixture of four different donors. PBMC were grown as the SupT1 cells, but with the addition of 100 U/mL human IL2 after an initial PHA (5 μg/mL) stimulation for 2 days. Infections were performed with C33A produced virus stocks with the indicated amount of virus.

Virus competition experiments were done as described previously 44. Competitions were initiated with C33A produced virus stocks. Each competition was done with virus corresponding to 6 ng virus with starting ratios of 5 to 1, 1 to 1 and 1 to 5. The competition was repeated with independent virus stocks.

Low-level replication in PBMC was analyzed with a virus rescue assay. At day 13 post infection of PBMC, the cells from 1 mL culture were collected (4 min 4000 RPM, eppendorf centrifuge), washed once with 1 mL PBS to remove any input virus and resuspended in medium. A dilution series (1, 10×, 100×, 1000×, 10.000×) was made and each sample mixed with one million SupT1 cells. The cultures were maintained for four weeks, regularly split and inspected for virus replication by CA-p24 Elisa on the culture supernatant and visual inspection for syncytia formation.

Flow cytometry was performed with RPE-conjugated mouse monoclonal anti-human CD4 (clone MT310, Dako, Glostrup, Denmark) and FITC-conjugated mouse monoclonal anti-human CD8 (clone DK25, Dako). Cells from a 1 mL culture sample were collected (4 min 4000 RPM, eppendorf centrifuge), incubated with a mixture of both monoclonal antibodies in fluorescence-activated cell sorting (FACS) buffer (PBS with 2% FCS) for 30 min at room temperature and washed with 800 μL FACS buffer. The cells were subsequently collected (4 min 4000 RPM, eppendorf centrifuge) and resuspended in 20 μL of 4% paraformaldehyde. After 5-minute incubation at room temperature, 750 μL FACS buffer was added and the suspension analyzed on a FACScalibur flow cytometer with CellQuest Pro software (BD biosciences, San Jose, CA). The machine was set for a 30-sec collection time. Cell populations were defined based on forward/sideward scattering and isotype controls were used to set markers. For mixed SupT1 plus PBMC cultures, the gates for PBMC (CD4⁺CD8^- and CD4^-CD8⁺) and SupT1 (CD4⁺, CD8⁺) were set using a separate control culture.

Freshly isolated PBMC were stimulated for 2 days with PHA (5 μg/mL), washed twice with medium and mixed with SupT1 cells. The cell mixture was analyzed by FACS staining for CD4 and CD8 as described above. The culture was divided into equal 10 mL samples, containing approximately 1 million PBMC and 2 million SupT1 cells, which were infected with different virus variants (input 40 ng CA-p24). Daily samples were taken for CA-p24 Elisa and anti-CD4/CD8 FACS analysis.