The Genetic Stability of a Conditional Live HIV-1 Variant Can Be Improved by Mutations in the Tet-On Regulatory System That Restrain Evolution*

Live attenuated human immunodeficiency virus type 1 (HIV-1) vaccines are considered unsafe because more quickly replicating pathogenic virus variants may evolve after vaccination. As an alternative vaccine approach, we have previously presented a doxycycline (dox)-dependent HIV-1 variant that was constructed by incorporating the tetracycline-inducible gene expression system (Tet-On system) into the viral genome. Replication of this HIV-rtTA variant is driven by the dox-inducible transcriptional activator rtTA and can be switched on and off at will. A large scale evolution study was performed to test the genetic stability of this conditional live vaccine candidate. In several long term cultures, we selected for HIV-rtTA variants that no longer required dox for replication. These evolved variants acquired a typical amino acid substitution either at position 19 or 37 in the rtTA protein. Both mutations caused rtTA activity and viral replication in the absence of dox. We designed a novel rtTA variant with a higher genetic barrier toward these undesired evolutionary routes. The corresponding HIV-rtTA variant did not lose dox control in long term cultures, demonstrating its improved genetic stability.

Live attenuated human immunodeficiency virus type 1 (HIV-1) vaccines are considered unsafe because more quickly replicating pathogenic virus variants may evolve after vaccination. As an alternative vaccine approach, we have previously presented a doxycycline (dox)-dependent HIV-1 variant that was constructed by incorporating the tetracycline-inducible gene expression system (Tet-On system) into the viral genome. Replication of this HIV-rtTA variant is driven by the dox-inducible transcriptional activator rtTA and can be switched on and off at will. A large scale evolution study was performed to test the genetic stability of this conditional live vaccine candidate. In several long term cultures, we selected for HIV-rtTA variants that no longer required dox for replication. These evolved variants acquired a typical amino acid substitution either at position 19 or 37 in the rtTA protein. Both mutations caused rtTA activity and viral replication in the absence of dox. We designed a novel rtTA variant with a higher genetic barrier toward these undesired evolutionary routes. The corresponding HIV-rtTA variant did not lose dox control in long term cultures, demonstrating its improved genetic stability.
AIDS vaccines based on a live attenuated virus have shown promise in the simian immunodeficiency virus-macaque model but are generally considered unsafe for use in humans (1)(2)(3)(4)(5). The major safety concern is that chronic low level replication of the attenuated virus may eventually lead to selection of fitter and more pathogenic virus variants (6 -9). Ideally, one would like to restrict replication of a vaccine virus to the time window that is needed to elicit a protective immune response. We and others previously presented a novel vaccine approach that uses a conditional live HIV-1 2 virus (10 -14). In this HIV-rtTA virus, the viral transcriptional activator Tat and its TAR-binding site (15,16) were inactivated by mutation and functionally replaced by the components of the Tet-On system. This system, in which gene expression is stringently controlled by the nontoxic effector doxycycline (dox), is widely applied to regulate gene expression in eukaryotes (17)(18)(19). The rtTA gene encoding the transcriptional activator was inserted in place of the nef gene, and the tet-operator (tetO) DNA-binding sites were placed in the viral LTR promoter. This HIV-rtTA virus does not replicate in the absence of dox. Binding of dox to rtTA triggers a conformational change that allows the protein to bind tetO DNA, resulting in transcriptional activation and subsequent viral replication. Upon vaccination with this virus, replication can be temporarily activated and controlled to the extent needed for induction of the immune system by transient dox administration. The initial HIV-rtTA virus has been greatly improved by viral evolution (20 -22), and we have shown efficient and dox-dependent replication of this virus not only in vitro in T cell lines but also ex vivo in human lymphoid tissue (23).
The potential use of this dox-dependent HIV-rtTA virus as a vaccine raises new safety questions concerning the genetic stability of the introduced Tet-On system. There are several hypothetical evolutionary routes toward a constitutively replicating virus. First, the virus may restore the function of the Tat-TAR system, despite the multiple inactivating mutations that were introduced in both elements to avoid simple reversion to the wild-type sequence. In this scenario, the dox-controlled rtTA-tetO system will become superfluous and may be inactivated over time by mutation or deletion. Second, the viral LTR promoter could become a constitutive transcription element, for instance by acquisition of a binding site for a constitutively expressed cellular transcription factor. Replication of such a virus is not dependent on a virally encoded transactivator, neither Tat nor rtTA. Third, the introduced rtTA-tetO axis may lose dox dependence, thereby creating an uncontrolled Tet-On system. This scenario is most likely to occur through acquired mutations in the rtTA protein that shift its conformation into the DNA binding mode, even in the absence of dox. To address these safety issues, we followed the evolution of HIV-rtTA in multiple, independent virus cultures. We observed the loss of dox control in several cultures, which in all cases resulted from a typical amino acid substitution either at position 19 or 37 in the rtTA protein. We developed a novel rtTA variant with alternative amino acids at these positions and demonstrated that the corresponding HIV-rtTA variant did not lose dox control in long term cultures. Therefore, we improved the genetic stability of the Tet-On system and the HIV-rtTA vaccine candidate by blocking two unwanted evolutionary routes.

EXPERIMENTAL PROCEDURES
Virus Cultures-The HIV-rtTA infectious molecular clone is derived from the HIV-1 LAI proviral plasmid (24) and was described previously (11,12). The HIV-rtTA used in this study is the KYK version, which contains the inactivating Y26A mutation in the Tat gene and five nucleotide substitutions in the TAR hairpin motif. This virus contains the rtTA2 S -S2 gene (19) in place of the nef gene and eight tetO sequences in the LTR promoter region. The HIV-rtTA-2⌬tetO clone is identical to HIV-rtTA but with the optimized 2⌬tetO promoter configuration (21,22). HIV-rtTA F86Y/A209T contains the LTR-2⌬tetO promoter and the recently described rtTA F86Y/A209T gene (20). SupT1 T cells were cultured at 37°C with 5% CO 2 in RPMI 1640 medium containing 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin. SupT1 cells (5 ϫ 10 6 ) were transfected with 5 g of the HIV-rtTA molecular clone by electroporation (250 V and 975 microfarad). Transfected cells were cultured with or without dox (Sigma D-9891). The CA-p24 level in the cell-free culture supernatant was determined by antigen capture enzyme-linked immunosorbent assay (ELISA) (25).
The 24-well evolution experiment was started with transfection of 40 g of the HIV-rtTA proviral plasmid into 2 ϫ 10 7 SupT1 cells. The cells were split into 24 independent cultures and maintained in the presence of 1 g/ml dox for up to 200 days. The virus-containing culture supernatant was passaged onto fresh SupT1 cells at the peak of infection, as determined by the massive appearance of syncytia. At regular intervals, supernatant samples were taken from the culture and tested in parallel infections with and without dox. The cell samples were stored at Ϫ80°C for subsequent analysis. Proviral DNA Analysis and Cloning of Evolved Sequences-Total cellular DNA from HIV-rtTA-infected cells was isolated as described previously (26). Proviral DNA sequences were PCR-amplified from total cellular DNA. The first exon of the tat gene was amplified with the primers KV1 (5Ј-CCATCGATACCGTCGACATAGCAGA-ATAGG-3Ј) and 3Ј-TAT (5Ј-CGGGAATTCTTACTGCTTTGATA-GAGAAAC-3Ј). The LTR-tetO region was amplified with the primers tTA-tetO1 (5Ј-CTCCCCGGGTAACTAAGTAAGGAT-3Ј) and C(N1) (5Ј-GGTCTGAGGGATCTCTAGTTACCAGAGTC-3Ј). The rtTA gene was amplified with the primers tTA1 (5Ј-ACAGCCATAGCAGTAGCT-GAG-3Ј) and tTA-rev2 (5Ј-GATCAAGGATATCTTGTCTTCGT-3Ј). All of the PCR fragments were sequenced with the BigDye Terminator cycle sequencing kit (Applied Biosystems). For the cloning of the G19Eor E37K-mutated rtTA sequences into the HIV-rtTA provirus, rtTA PCR fragments were digested with XcmI and SmaI and cloned into the corresponding sites of the shuttle vector pBlue3ЈLTRext-deltaU3-rtTA-2⌬tetO (20). The BamHI-BglI fragment of the shuttle vector was used to replace the corresponding sequence in HIV-rtTA-2⌬tetO.
Construction of Novel HIV-rtTA Variants and rtTA Expression Plasmids-HIV-rtTA variants with an alternative codon for G (GGU instead of GGA) at rtTA position 19 and with a wild-type (Glu) or alternative amino acid (Asp, Phe, Leu, Asn, Gln, Arg, and Ser) at position 37 were constructed by oligonucleotide-directed mutagenesis. The oligonucleotide G19 (5Ј-ATAACCATGTCTAGACTGGACAAGAGCAAAGTCAT-AAACTCTGCTCTGGAATTACTCAATGGTGTCGGTATCGAAGG-CCTGACGACAAGGAAACTCGCT-3Ј, mutated nucleotide underlined) was annealed to the oligonucleotide rev-37 (5Ј-AGCAGGGCCCGCTTG-TTCTTCACGTGCCAGTACAGGGTAGGCTGXXXAACTCCCAGC-TTTTGAGCGAGTTTCCTTGTCGTCAGGCCTTCGA-3Ј, with XXX corresponding to amino acid 37; this triplet is CTC for Glu, ATC for Asp, GAA for Phe, AAG for Leu, ATT for Asn, CTG for Gln, GCG for Arg, and AGA for Ser), both strands were completed with Klenow DNA polymerase in the presence of dNTPs, digested with XcmI and ApaI, and ligated into the similarly digested shuttle vector pBlue3ЈLTRext-deltaU3-rtTA F86Y/A209T -2⌬tetO (20). The BamHI-BglI fragment of the shuttle vector was used to replace the corresponding sequence in HIV-rtTA-2⌬tetO.
rtTA Activity Assay-HeLa X1/6 cells are derived from the HeLa cervix carcinoma cell line and harbor chromosomally integrated copies of the CMV-7tetO firefly luciferase reporter construct pUHC13-3 (27). The cells were cultured at 37°C and 5% CO 2 in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, minimal essential medium nonessential amino acids, 100 units/ml penicillin, and 100 g/ml streptomycin. HeLa X1/6 cells were grown in 2-cm 2 wells to 60% confluency and transfected by the calcium phosphate precipitation method (20). The DNA mixture consisted of 8 ng of pCMV-rtTA, 2.5 ng of pRL-CMV, and 990 ng of pBluescript as carrier DNA. The plasmid pRL-CMV (Promega) that constitutively expresses Renilla luciferase was used as an internal control to allow correction for differences in transfection efficiency. The cells were cultured after transfection for 48 h at different dox concentrations and then lysed in passive lysis buffer (Promega). Firefly and Renilla luciferase activities were determined with the dual luciferase reporter assay (Promega) using a LUMIstar luminometer (BMG LABTECH). The expression of firefly and Renilla luciferase was within the linear range, and no squelching effects were observed. The activity of the rtTA variants was calculated as the ratio of the firefly and Renilla luciferase activities and corrected for between session variation (52).

Selection of HIV-rtTA Variants with Reduced dox Dependence-We
have previously constructed a conditional live HIV-1 variant (11,12), in which the natural Tat-TAR elements that control viral gene expression and replication were inactivated by mutation and functionally replaced by the rtTA-tetO elements of the Tet-On system for inducible gene expression (Fig. 1A). This HIV-rtTA virus does not replicate constitutively, but exclusively in the presence of dox. We recently reported that long term replication of this virus resulted in rearrangement of the tetO elements and amino acid substitutions in the rtTA protein that significantly improve viral replication without losing dox control (20 -22). We anticipated that the HIV-rtTA virus could evolve in different directions (see the Introduction) and therefore focused this study on the appearance of virus variants that no longer depend on dox for replication. We started multiple long term virus cultures and followed the development of dox independence. The evolution approach and the flow chart of the subsequent analyses are illustrated in Fig. 1B. The HIV-rtTA virus was passaged extensively in the presence of dox in 24 independent cultures. At several time points, supernatant samples were taken from the culture and tested in parallel infections without dox to determine the dox dependence of the evolved virus. The results for all 24 cultures are summarized in Fig. 1C (black squares). We observed a significant reduction in the number of dox-dependent viruses within 50 days of culturing, and only three cultures remained dox-dependent after 125 days.
The replication curves of the original HIV-rtTA virus and two representative dox-independent virus cultures are shown in Fig. 2. Virus sample C5 did replicate without dox but can still be activated by dox to some extent, whereas virus sample C6 replicated with similar efficiency with and without dox. Total cellular DNA with integrated provirus was isolated from eight dox-independent HIV-rtTA cultures. We analyzed the sequence of both the "old" Tat-TAR motifs and the "new" rtTA-tetO motifs in the virus population. In all cultures, the Tat and TAR sequences contained the original mutations, indicating that the Tat-TAR axis of transcriptional activation had not been repaired. In contrast, we observed in all of the cultures the characteristic truncation of tetO elements that had previously been shown to improve dox-depend-  (20)). The HIV-rtTA G19F/E37L variant developed in this study does not escape from dox control. Plotted is the number of dox-dependent cultures as a function of the culture time. Each experiment was started with 24 independent cultures. D, amino acid substitutions observed in HIV-rtTA cultures that lost dox control. In all cases, the G19E substitution resulted from a GGA to GAA codon mutation and the E37K substitution from a GAG to AAG mutation. Replication of the original HIV-rtTA virus, the virus from culture C6, or the virus from culture C5 (both harvested after 50 days of culturing) was compared by infecting SupT1 T cells with equal amounts of virus (5 ng/ml CA-p24) in the absence or presence of dox (1 g/ml). Sequence analysis revealed that the C6 virus carried the G19E and E156K mutations in the rtTA gene, and the C5 virus carried the E37K mutation (Fig. 1D). ent HIV-rtTA replication (21,22). Moreover, viruses from all dox-independent cultures had acquired either a G19E or an E37K mutation in the rtTA gene (Fig. 1D). Two of the cultures (B6 and C6) contained additional amino acid substitutions. The repeated selection of G19E or E37K in multiple cultures, combined with their absence in the three cultures that remained dox-dependent (data not shown), strongly suggests their linkage to the acquired dox-independent phenotype.
Amino Acid Substitutions in rtTA Confer the Loss of dox Control-To demonstrate that these rtTA mutations are responsible for the observed viral replication without dox, we constructed HIV-rtTA molecular clones with the G19E or E37K mutation in the rtTA gene and assayed their replication at different dox concentrations (Fig. 3). The original HIV-rtTA virus did not replicate without dox and showed a graded increase in viral replication with increasing dox concentrations. HIV-rtTA G19E replicated efficiently both with and without dox. HIV-rtTA E37K also replicated without dox, but replication was more efficient with dox. These results demonstrate that the G19E or E37K mutation is sufficient to reduce the dox dependence of the HIV-rtTA virus.
The results described above were obtained with the original HIV-rtTA virus, which replicates relatively poorly. We also tested the genetic stability of two improved HIV-rtTA variants in a similar 24-well long term culture assay. HIV-rtTA-2⌬tetO is identical to HIV-rtTA, but with the improved LTR-2⌬tetO promoter (21,22), and HIV-rtTA F86Y/A209T contains in addition the improved rtTA F86Y/A209T gene (20). With both viruses we again observed the appearance of variants that replicated without dox, albeit at a significantly slower rate than the original HIV-rtTA (Fig. 1C). Whereas the original HIV-rtTA lost dox control in 50% of the cultures within 50 days, 50% of the HIV-rtTA-2⌬tetO cultures lost dox dependence in ϳ75 days, and more than 50% of the HIV-rtTA F86Y/A209T cultures were still fully doxdependent after 120 days, although the majority lost dox control upon prolonged culturing. Apparently, these new HIV-rtTA variants not only have an improved replication capacity but also have a lower tendency to lose dox control. Sequence analysis of two dox-independent HIV-rtTA F86Y/A209T cultures revealed the G19E mutation in both cases.
HIV-rtTA Variants with Alternative Codons at rtTA Positions 19 and 37-In the evolution experiments, we observed very specific amino acid substitutions that reduced dox dependence at only two rtTA positions (G19E and E37K). Introduction of alternative rtTA codons may make such specific amino acid substitutions more difficult or even prevent these unwanted evolutionary routes. For instance, the G19E mutation involves a GGA to GAA codon change, and the G-to-A transition is the most frequent error during HIV-1 reverse transcription (28 -30). Introduction of an alternative Gly codon (GGU or GGC) would require a much more difficult two-hit mutation, including one transversion, to create a Glu codon (GAA or GAG). We previously described that such a difference in genetic threshold can strongly influence the course of HIV-1 evolution (31,32). A similar strategy is not possible for E37K because all possible Glu codons (GAA and GAG) require only a single G-to-A mutation to turn into a Lys codon (AAA or AAG). As an alternative blocking strategy, we could replace the Glu 37 codon with a non-Glu codon that would be more difficult to transform into a Lys codon. However, such an amino acid substitution should ideally not affect the activity or dox dependence of the rtTA protein. We first examined natural variation at this position in the tet repressor (TetR). The rtTA protein is based on the Escherichia coli class B TetR (TetR B ), but there are six additional TetR classes (A, C, D, E, G, and H) (33). TetR from classes D, E, and H also have the Glu at position 37, but TetR from classes A, C, and G have a Gln instead. Evolution of a Gln codon (CAA or CAG) to a Lys codon (AAA or AAG) would require only a single C-to-A mutation, but this transversion is less frequently observed in HIV-1 evolution (28 -30). We therefore constructed an HIV-rtTA variant with a Gln codon (CAG) at position 37 (E37Q). In addition, we constructed variants with Asp (GAU; E37D), Asn (AAU; E37N), Ser (UCU; E37S), Arg (CGC; E37R), Phe (UUC; E37F), or Leu (CUU; E37L). The E37D substitution leaves the acidic nature of the residue intact. The E37N and E37S mutations, like the natural variant E37Q, result in polar, uncharged residues. The E37F and E37L mutations result in hydrophobic residues. The E37R substitution creates a basic residue that is similar to the E37K mutation selected through viral evolution. When allowed by the degeneracy of the genetic codon, we chose the codon that requires the most mutations to be converted to a Lys codon. For example, a CGC rather than an AGA codon was chosen for the E37R variant. Moreover, all new HIV-rtTA variants contain the alternative Gly codon (GGU) at position 19.
We tested replication of these novel HIV-rtTA variants in SupT1 cells with and without dox (Fig. 4). As expected, the Glu 37 virus with the silent codon change at position 19 replicated like the original HIV-rtTA in a dox-dependent manner. The E37L, E37N, E37F, E37Q, and E37R variants also showed efficient and dox-dependent replication. The E37D variant did not replicate with or without dox. Interestingly, the E37S variant replicated efficiently both with and without dox and thus has a phenotype similar to that of the E37K variant. This initial survey demonstrates that the HIV-rtTA phenotype is difficult to predict from the chemical nature of the residue at position 37, e.g. E37R is similar to E37K but does not reduce dox dependence. To construct a more stable dox-dependent virus, it seems necessary to know the impact of all possible amino acid substitutions at position 37.
A Complete Mutational Analysis of Residue 37 in rtTA-We constructed rtTA expression plasmids with all possible amino acids at position 37. The activity of these variants was assayed by transfection into HeLa X1/6 cells that contain stably integrated copies of the CMV-7tetO luciferase reporter construct (27). Transfected cells were cultured for 2 days in the presence of 0 -1000 ng/ml dox. We subsequently determined the intracellular luciferase level, which reflects rtTA activity. As shown in Fig. 5A, the activity of these 20 rtTA variants varies considerably. Most variants show no or very low activity in the absence of dox, and their activity increases with increasing dox levels.
Comparison of the rtTA activity data (Fig. 5A) with the replication capacity of the selected set of HIV-rtTA variants (Fig. 4) allows us to determine the level of rtTA activity that is required for viral replication. The 37F, 37L, 37N, 37Q, and 37R variants show no or very low activity in the absence of dox (less or equal to 0.06% of the wild-type rtTA activity at 1000 ng/ml dox), and viruses carrying these rtTA variants do not replicate without dox. The low activity (0.09%) of the 37D variant at 1000 ng/ml dox is not sufficient for viral replication either. The 37K and 37S variants show 0.19 and 1% activity without dox, respectively. This level of activity is apparently sufficient to drive a low level of virus replication. The threshold of rtTA activity that is sufficient for HIV-rtTA replication was therefore set at 0.1%. This would mean that not only HIV-rtTA E37K and HIV-rtTA E37S but also HIV-rtTA E37A should repli-cate in the absence of dox. The corresponding codons of these unwanted amino acids are colored red in the codon table (Fig. 5C), and evolution toward them should be prevented. All other variants, except for the inactive 37D mutant, show a phenotype similar to wild-type rtTA, i.e. activity below 0.1% without dox and much higher than 0.1% at 1000 ng/ml dox. HIV-rtTA viruses carrying these variants are expected to replicate in a dox-dependent manner. These amino acids are colored green in the codon table. The Asp and stop codons are marked in black, because the corresponding viruses will be replication-incompetent.
In the codon table, every change in row or column represents a single nucleotide substitution. This colored codon table (Fig. 5C) thus facilitates the identification of position 37 codons that preserve dox dependence (green) and that require multiple nucleotide mutations to convert into a codon that allows replication in the absence of dox (red). The Leu codons CUN meet these safety requirements.
A Complete Mutational Analysis of Residue 19 in rtTA-Like the E37K mutation, the G19E mutation causes viral replication in the absence of dox. To reveal whether other amino acid substitutions at this position would similarly result in a loss of dox dependence, we constructed rtTA expression plasmids with all possible amino acids at position 19. The activity of these rtTA variants was analyzed as described above for the position 37 variants. As shown in Fig. 5B, most variants show no or very low activity (less than 0.1%) without dox, and their activity increases with increasing dox levels. In contrast, the 19P variant is inactive, and the 19E variant shows 3% activity without dox. This relatively high basal activity of 19E is in agreement with the efficient replication of the corresponding HIV-rtTA virus without dox. There are multiple codons possible at position 19 that preserve dox dependence (colored green in Fig. 5D) and that require multiple nucleotide mutations to turn into a codon that allows replication in the absence of dox (colored red). For example, the Phe codon UUU meets these safety requirements very well, because it requires three transversions to convert into a Glu codon.
Novel rtTA Variant Prevents the Loss of dox Control-We constructed an rtTA variant that combines the Phe (UUU) at position 19 (G19F) and Leu (CUU) at position 37 (E37L). This rtTA variant shows very low basal activity (less than 0.1%), and its activity gradually increases with increasing dox levels (Fig. 6A). Although rtTA G19F/E37L is less active than wild-type rtTA at low dox concentrations, it is fully active at high dox levels. Accordingly, HIV-rtTA G19F/E37L does not replicate in the absence of dox but does replicate efficiently at 1000 ng/ml dox (Fig. 6C). We tested the genetic stability of this virus in 24 long term cultures with dox. The HIV-rtTA G19F/E37L variant never lost dox control up to 200 days of culturing (Fig. 1C). This result demonstrates the increased genetic stability, and possibly improved safety, of the novel HIV-rtTA variant.

DISCUSSION
The introduced Tet-On system has become an essential part of the replication machinery of the HIV-rtTA virus. This precludes the spontaneous loss of these foreign sequences, e.g. by simple deletion, as has been reported for gene inserts encoding thymidine kinase (34,35), interleukin 2 (36), interleukin 12 (37), interferon ␥ (38, 39), and green fluorescence protein (35). One other study reported the stable maintenance of an inserted IB-␣S32/36A gene, which presumably improved the virus-cell interaction during HIV-1 replication (40). Upon prolonged culturing of HIV-rtTA, the error-prone reverse transcription process may allow for the generation and outgrowth of faster replicating variants that harbor mutations in the components of the Tet-On system. We have earlier reported on the optimization of both the rtTA protein  (20). and the LTR-tetO promoter by viral evolution (20 -22). To test the safety of this dox-dependent HIV-rtTA vaccine candidate, we analyzed the genetic stability of the imposed dox control in long term evolution experiments. We observed a loss of dox control after several months of virus passage in a significant number of cultures. Molecular analysis of these viral isolates revealed that replication in the absence of dox is not caused by restoration of the viral Tat-TAR axis or by creation of a constitutive LTR promoter but rather by a G19E or E37K mutation in the rtTA protein. This suggests that amino acid substitutions at these two specific rtTA positions represent the only evolutionary routes A and B, rtTA activity was measured in HeLa X1/6 cells that contain stably integrated copies of the CMV-7tetO firefly luciferase reporter construct (27). The cells were transfected with the indicated rtTA expression plasmid (all of the rtTA variants contain the F86Y and A209T mutations that improve rtTA activity (20)) or pBluescript as a negative control (Ϫ), and a plasmid constitutively expressing Renilla luciferase to correct for differences in transfection efficiency. The cells were cultured in the presence of different dox concentrations (0 -1000 ng/ml). The ratio of the firefly and Renilla luciferase activities measured 2 days after transfection reflects rtTA activity. All of the values were related to the wild-type (37E in A and 19G in B) rtTA activity at 1000 ng/ml dox, which was arbitrarily set at 100%. toward the loss of dox control. This finding convinced us that it might be feasible to block such unwanted evolution. We developed the rtTA G19F/E37L variant that requires multiple mutations to allow viral replication in the absence of dox. Multiple nucleotide changes occur much less frequently than a single G-to-A transition that caused both the G19E and E37K substitutions (28 -30). These differences in mutational frequency strongly influence the course of HIV-1 evolution (31,32). Accordingly, the HIV-rtTA G19F/E37L variant was found to be much more stable, because it did not lose dox control in multiple long term cultures. The G19F and E37L mutations thus significantly improved the genetic stability of the Tet-On system, which is not only important for the development of a safe conditional live HIV-1 vaccine but also for other applications of the Tet-On system in which the rtTA protein may be subject to mutation through error-prone replication (e.g. applications in other viruses).
HIV-rtTA was cultured with dox in the long term evolution assay. We refreshed the culture medium every three to 4 days and added dox to 1000 ng/ml. Because the half-life of dox in cell culture is about 63 h (41), the dox level may have dropped to around 400 ng/ml during culturing. The original HIV-rtTA replicates efficiently at this dox level (Fig. 3). However, the repeated selection of the G19E and E37K variants argues for a replication advantage of these viruses over the original HIV-rtTA. Our results do indeed demonstrate that both of these variants are not only active in the absence of dox but also more active than wild-type rtTA at high dox levels (Fig. 5).
The G19E mutation in rtTA is a true reversion to the sequence of tTA, which is the transcriptional activator of the Tet-Off system. The activity of tTA is inhibited, instead of activated, by dox (42). rtTA is a variant of tTA with the reverse phenotype and differs from tTA by four amino acids (E19G, A56P, D148E, and H179R), of which the E19G and A56P combination is sufficient for the phenotype reversal (19). The 19E rtTA variant is still activated by dox, despite its relatively high basal activity. Therefore, the G19E mutation does not fully revert rtTA into a tTA phenotype. This is not surprising because rtTA variants that are inhibited by dox would not be selected in cultures with dox. Among all position 19 rtTA variants, only 19E allows viral replication without dox. The other variant with an acidic side chain (19D) also shows increased basal activity, but this level is not sufficient to support viral replication. Variants with an aromatic residue at position 19 (19F, 19W, and 19Y) demonstrate similar or lower activity than wild-type rtTA (19G), whereas most other variants are more active. These results indicate that the activity of rtTA is significantly affected by the chemical nature of residue 19. This amino acid is located in the DNA-binding domain of rtTA (Fig. 1A), and although it does not directly contact tetO DNA, it may adjust the orientation of the DNA-binding domain and thus indirectly affect the affinity of rtTA for tetO (43,44).
Residue 37 is also located in the DNA-binding domain of rtTA (Fig.  1A). It is the last amino acid in the turn of a classical helix-turn-helix motif (residues 27-44) (43,44). Glu 37 forms a hydrogen bond between its main chain amino group and a phosphate group of tetO DNA (45). Substitution of this amino acid may thus directly affect the DNA binding properties of rtTA. The helix following residue 37 is the recognition helix of the helix-turn-helix motif. Residues in this helix form numerous sequence specific and nonspecific interactions with tetO DNA (45). Mutations at position 37 may alter the orientation of the recognition helix and thus also indirectly affect DNA binding (46,47). Most rtTA variants with a polar side chain at position 37 (37C, 37K, 37Q, 37R, 37S, and 37T) are equally or more active than wild-type rtTA (37E). In contrast, hydrophobic side chains at this position often lead to variants that are poorly activated by dox (37F, 37I, 37L, 37M, 37P, and 37W). However, some variants with similar side chains display totally different phenotypes. For example, whereas the Asp residue is negatively charged like the wild-type Glu, the 37D rtTA variant is completely inactive. Detailed structural analyses of these rtTA variants would help to fully understand how specific amino acid substitutions at position 19 or 37 affect rtTA structure and function.
We improved the genetic stability of the conditional live HIV-rtTA virus by blocking the undesired evolutionary routes of the incorporated Tet-On system that were observed when the virus was cultured for up to 200 days at high dox levels. If HIV-rtTA would be used as a vaccine, virus replication can be limited to the extent required to mount a protective immune response by transient dox administration. Although FIGURE 6. Activity of the novel rtTA variant with G19F and E37L mutations. A, the activity of wild-type rtTA (rtTA F86Y/A209T ) and rtTA G19F/E37L (which also carries the F86Y and A209T mutations) was measured in HeLa X1/6 cells (see Fig. 5 for details). The cells were cultured in the presence of different dox concentrations (0 -1000 ng/ml). All of the values were related to the wild-type rtTA activity at 1000 ng/ml dox, which was arbitrarily set at 100%. The average values of two transfections are plotted with the error bars indicating the standard deviations. B and C, replication of HIV-rtTA F86Y/A209T and HIV-rtTA G19F/E37L . SupT1 cells were transfected with 5 g of the molecular clones and cultured with or without 1 g/ml dox. Virus replication was monitored by CA-p24 ELISA on culture supernatant samples. this period has to be determined empirically, studies with live attenuated simian immunodeficiency virus in macaques suggest that this period may be as short as 3 weeks (48). Subsequent dox-withdrawal will halt replication and thus prevent evolution. To exclude the possibility that the virus may find alternative evolutionary routes to escape from dox control when the dox level is reduced gradually, we will test this scenario in future experiments. The safety of HIV-rtTA can be further improved by attenuation through deletion of nonessential parts of the viral genome (mini-HIV approach) (49) or by incorporating a second regulatory system. We recently identified amino acid changes in the HIV-1 Envelope protein that make viral replication dependent on the T20 peptide (50). T20 (Fuzeon) is a new anti-HIV-1 drug that normally blocks viral entry into cells. T20 therapy not only resulted in Env mutations that mediate T20 resistance but also resulted in mutations that cause T20 dependence. Introduction of the latter mutations into HIV-rtTA generated a novel virus variant that replicates exclusively in the presence of both dox and T20 (51). Combining the novel rtTA G19F/E37L variant with this double regulatory circuit will prevent the emergence of constitutively replicating viruses and further improve the safety of HIV-rtTA as a conditional live HIV-1 vaccine.