Novel Tat-Encoding Bicistronic Human Immunodeficiency Virus Type 1-Based Gene Transfer Vectors for High-Level Transgene Expression

Abstract

We describe bicistronic single-exon Tat (72-amino-acid Tat [Tat72])- and full-length Tat (Tat86)-encoding gene transfer vectors based on human immunodeficiency virus type 1 (HIV-1). We created versions of these vectors that were rendered Rev independent by using the constitutive transport element (CTE) from Mason-Pfizer monkey virus (MPMV). Tat72-encoding vectors performed better than Tat86-expressing vectors in gene transfer experiments. CTE-containing vectors, produced in a Rev-independent packaging system, had gene transfer efficiencies nearly equivalent to those produced using a combination RNA transport (CTE and Rev-Rev response element)-based packaging system. The Tat72-encoding vectors could be efficiently transduced into a variety of cell types, showed higher levels of transgene expression than vectors with the simian cytomegalovirus immediate-early or the simian virus 40 early promoter, and provide an alternative to HIV-1 vectors with internal promoters.

Lentiviruses can transduce terminally differentiated and growth-arrested cells (45, 46) and are therefore being developed for gene therapy purposes. Most human immunodeficiency virus type 1 (HIV-1)-based gene transfer vectors express marker genes under control of internal promoters. The HIV-1 long terminal repeat (LTR) in the presence of Tat is one of strongest promoters known. The Tat protein of HIV specifically binds to the transactivation response (TAR) RNA element at the 5′ end of nascent viral transcript to not only increase initiation of RNA synthesis from the viral LTR promoter (44) (a minor effect) but also enhance the efficiency of elongation or processivity of RNA polymerase II (a pronounced effect) (21, 25, 33–36). This allows accumulation of abundant full-length HIV messages in the infected cells. The full-length message and its spliced products are necessary for the synthesis of viral structural and regulatory proteins.

For situations that demand high levels of transgene expression, it would be advantageous to design HIV vectors that express transgenes under control of the viral LTR. Another potential problem of using heterologous internal promoters in retroviral vectors is the possibility of promoter interference between the internal promoter and the viral LTR, although the evidence for this thus far has been scanty for HIV vectors. Thus, using the viral LTR as a promoter may overcome any potential conflict between the LTR and internal promoters. An alternative approach to overcome promoter competition is by inactivation of the viral promoter by introducing promoter-debilitating mutations in the 3′ U3 region of the gene transfer vector. This results in the inactivation of viral LTR promoter during the process of infection. Several groups have developed such self-inactivating lentivirus vectors that give titers equivalent to vectors without these mutations (51, 74).

The HIV-1 Tat protein is encoded in two exons. The full-length protein is between 86 and 130 amino acids long, depending on the strain of HIV (33). Generally, gene transfer vectors derived from HIV-1 that express Tat use both coding exons coupled with a deletion within env (which is an intron for tat mRNA) (1, 42, 52, 56, 66). This results in the retention of a considerable portion of the HIV-1 sequence in the middle of the genome, a situation that increases the possibility of generating replication-competent HIV-1 by homologous recombination. It was previously demonstrated that the transactivation function of Tat resides almost entirely in the first coding exon (23, 53, 64), although sequences in the second coding exon can improve transactivation in certain cell types (68). There is also a report that indicates that Tat has another novel function, i.e., in the efficient reverse transcription of HIV-1 RNA as well. This function of Tat is also carried out by the protein encoded in the first-exon of tat (27, 30). It should thus be possible to construct a single-exon-Tat-producing HIV-1-based gene transfer vector.

Description of gene transfer vectors.

We created four Tat-expressing HIV-1 gene transfer vectors (Fig. 1) in which the Tat coding sequences were positioned downstream of an internal ribosome entry site (IRES) of encephalomyocarditis virus (EMCV) to allow expression of Tat when used in a bicistronic configuration. The vector pN-IT72 encodes a single-exon Tat (72-amino-acid Tat [Tat72]), while pN-IT86 encodes a full-length Tat (86-amino-acid Tat [Tat86]). We created versions of these plasmids containing the constitutive transport element (CTE) derived from Mason-Pfizer monkey virus (8) (MPMV) (pN-IT72CTE and pN-IT86CTE). In MPMV, the CTE performs a function that is analogous to that of Rev-Rev response element (RRE) of lentiviruses, which is the nucleocytoplasmic transport of unspliced or partially spliced viral mRNAs (20). The CTE can completely substitute for Rev-RRE function not only in subgenomic HIV-1 structural protein expression vectors but also in the context of the HIV-1 provirus itself. This is evidenced by the fact that a replication-competent HIV-1 that lacks functional Rev and RRE can be made by utilizing this RNA element in the proviral sequence (8, 73). Likewise, Rev-independent HIV-1-based gene transfer vectors that contain this RNA element have been described (63, 64).

FIG. 1.

Schematic representation of HIV-1 packaging constructs (A), provirus (B), and gene transfer vectors (C). The HIV-1 packaging constructs and gene transfer vectors were derivatives of the molecular clone pNL4-3 (GenBank accession no. M19921). For all packaging constructs, viral proteins were expressed under control of the sCMV immediate-early promoter-enhancer elements, and RNA transport was regulated using the MPMV CTE and poly(A) signal. The sCMV promoter-enhancer corresponds to bp 681 to 1349 of the IE94 gene (GenBank accession no. M16019), and the CTE and poly(A) region corresponds to bp 8007 to 8557 of MPMV (GenBank accession no. M12349). In pgpirin and pgp3virin, Rev and Nef were expressed using IRES derived from Harvey murine sarcoma virus (HaMSV) and EMCV, respectively. The HaMSV IRES-Rev and EMCV IRES-Nef were positioned upstream of MPMV CTE-poly(A). The viral accessory and regulatory proteins expressed by the constructs are tabulated at the right. Restriction enzyme sites and their nucleotide positions in pNL4-3 pertinent for the creation of the packaging plasmids are indicated above the provirus. The packaging constructs contain a deletion within the encapsidation signal (Δψ) and have been described previously (64). The gene transfer vectors were derived from pTR167 (61). They contain a deletion of HIV coding sequences between the proximal NsiI site in gag to the distal NsiI site in env. Restriction enzyme sites and their nucleotide positions in pNL4-3 that were used for the creation of the gene transfer vectors are indicated below the provirus. A frameshift mutation was introduced into the gag open reading frame between codons 9 and 10 by inserting an A residue (64). All transgene expression cassettes were positioned downstream of the 3′ splice acceptor site of Tat and Rev between the BamHI in the second coding exon of rev and XhoI site in nef. Tat72 or Tat86 sequences were amplified by PCR from pGEM-NL4-3 and pCMVtat, respectively, and cloned into a workshop vector to assemble the EMCV IRES-Tat cassettes, which were then introduced between the indicated restriction enzyme sites of the modified pTR167 to obtain pN-IT72 and pN-IT86 or between the BamHI and SalI sites of pTR167-CTE (64), to obtain pN-IT72CTE and pN-IT86CTE. CTE (GenBank accession no. M12349; bp 8007 to 8240)-containing Tat-expressing vectors do not have the frameshift mutation in gag. pN-sCMV contains the sCMV promoter (bp 681 to 1349 of the IE94 gene; GenBank accession no. M16019), whereas pN-SV contains the SV40 early promoter (bp 5175 to 5243 and 1 to 272; GenBank accession no. J02400). The promoters in both vectors are situated between the indicated restriction enzyme sites of the modified pTR167. Multiple cloning sites have been introduced downstream of these promoters to allow insertion of any foreign or marker gene of interest. Marker genes (luciferase or enhanced GFP) were introduced upstream of the IRES-Tat cassette between the unique BamHI and NotI restriction enzyme sites or downstream of the sCMV or SV40 promoter between unique BamHI and XhoI sites. Details of plasmid construction will be provided on request. 5′ss, 5′ splice acceptor site.

The Tat-encoding HIV-1 vectors were engineered to contain unique restriction enzyme sites to allow introduction of any gene of interest upstream of the IRES but situated downstream of the 3′ splice acceptor site of tat and rev to enable expression of the transgene off the spliced message. We also created HIV-1 vectors that express marker genes under the control of internal promoters (Fig. 1). In pN-sCMV, the marker gene is expressed under the control of the simian cytomegalovirus (sCMV) immediate-early promoter-enhancer elements; in pN-SV, the marker is expressed under control of the simian virus 40 (SV40) virus early promoter.

Transactivation of HIV-1 LTR by Tat-encoding vectors.

The Tat-encoding vectors were first tested for the production of functional Tat protein by cotransfecting 293 cells with each of the vectors together with an HIV LTR-luciferase reporter (pLTR-luc). If Tat is produced, it should lead to the transactivation of the HIV-1 LTR, resulting in increased luciferase activity. As controls, 293 cells were transfected with pLTR-Luc alone (for basal promoter activity) or pLTR-Luc and a plasmid (pCMVtat) (63) that encodes Tat86 under control of the sCMV immediate-early promoter as a positive control. Mock-transfected cells were used as negative controls. Parallel transfections with each gene transfer vector received pCMVtat. To normalize for transfection efficiency, all transfections also received pCMVβ-gal, which expresses β-galactosidase under control of the sCMV immediate-early promoter.

Cells lysates were prepared 48 h posttransfection and assayed for luciferase and β-galactosidase activities using commercially available reagents. Luciferase activity, normalized to β-galactosidase activity, for each vector is shown in Fig. 2A. The data indicate that all vectors expressed a functional Tat protein. While there was some variation in the transactivation levels produced by the different vectors, there was no clear indication for perturbation of transactivation function by the presence or absence of the CTE. Cotransfection with pCMVtat increased luciferase activity with all Tat-encoding vectors. This indicated that either the vectors were producing suboptimal levels of Tat or the reporter plasmid (pLTR-luc) was present in relative excess. A titration experiment using decreasing amounts of pLTR-luc with constant amount of Tat72-encoding vector (pN-IT72) confirmed that the reporter construct appeared to be in excess (data not shown).

FIG. 2.

Transactivation of HIV-1 LTR by tat expression gene transfer vectors. 293 cells were transfected with 1 μg of each vector together with either 2 μg of pLTR-luc and 1 μg of pCMVβ-gal (A) or 2 μg of each bicistronic vector with 1 μg of pCMVβ-gal (B) by the CaPO₄ transfection method (64). Parallel transfections also received 1 μg of pCMVtat (64). pLTR-luc contains the 3′ HIV-1 LTR of pNL4-3, inserted within the multiple cloning site of pBluescript II SK(+). The firefly luciferase gene is positioned downstream of the HIV-1 LTR. Cell lysates were prepared 48 h posttransfection and assayed for luciferase and β-galactosidase activities using commercially available kits according to the recommended protocols (Pharmingen, San Diego, Calif., for luciferase and Clontech, Palo Alto, Calif., for β-galactosidase). Error bars correspond to 1 standard deviation and were derived from duplicate experiments.

Additional support for the latter hypothesis was provided in experiments carried out using bicistronic vectors in which the marker gene was expressed within the gene transfer vector itself along with Tat72 or Tat86. In this situation, the Tat protein would upregulate expression from its own promoter, which would also lead to increased marker gene activity. To test this hypothesis, we introduced the luciferase gene upstream of the IRES in each of the Tat-encoding vectors (Fig. 1) to create pN-LIT72, pN-LIT86, pN-LIT72CTE, and pN-LIT86CTE. The bicistronic vectors were transfected into 293 cells together with pCMV-β-gal to normalize for transfection efficiency. Parallel transfections received pCMVtat. Cell lysates were harvested and assayed for luciferase and β-galactosidase activities as described earlier. The data (Fig. 2B) indicate that high levels of luciferase activity were observed with all bicistronic constructs but to various levels. Cotransfection with pCMVtat enhanced luciferase activity either marginally (pN-LIT86 and pN-LIT72CTE) or not at all (pN-LIT72 and pN-LIT86CTE). This suggested that some of the vectors were producing adequate amounts of Tat whereas others were producing less than saturating amounts of Tat. The CTE-containing vectors yielded lower luciferase activity than the vectors lacking CTE. This difference was statistically significant (P ≤ 0.05). A previous report showed that the two-exon Tat provides, depending on cell type, approximately two-fold-higher transcriptional activation in comparison to single-exon Tat (68). In contrast to this, other reports (23, 53, 64) have not found any such difference between the activities of Tat72 and Tat86. The results of our experiments are in agreement with those of the latter studies.

Gene transfer efficiency of Tat72- and Tat86-encoding vectors.

Next, we wished to determine the efficiency of gene transfer by each of the luciferase-encoding bicistronic tat expression vectors. To produce virus stocks, each vector was cotransfected into 293 cells with a helper plasmid, pgpirin or pgp (Fig. 1), that produced viral packaging proteins and a vesicular stomatitis virus G envelope glycoprotein (VSV-G)-expressing plasmid (pMD.G). The packaging plasmid, pgpirin, is a polycistronic construct that expresses HIV-1 Gag, Pol, Rev, and Nef. Rev and Nef were expressed via the Harvey murine sarcoma virus IRES and EMCV IRES, respectively. The other construct, pgp, expresses only HIV Gag and Pol. Both constructs express viral proteins using the sCMV immediate-early promoter, with RNA transport being regulated by the MPMV CTE and poly(A). These packaging plasmids have been described previously (64). Virus stocks, produced with each of the vectors, were then used for infection of 293 (adenovirus-transformed human embryonic kidney cell line) and D17 (canine osteosarcoma) cells. Cell lysates were prepared 48 h after transduction and assayed for luciferase activity (expressed as relative light units [RLU]). Since results were similar for both D17 and 293 targets, only the results of experiments with D17 cells are shown in Fig. 3.

FIG. 3.

(A) Gene transfer into D17 cells by luciferase-expressing HIV-1 vectors encoding Tat72 or Tat86 and either containing or lacking CTE. Virus stocks were produced separately for each indicated vector by transfecting 293 cells with 7.5 μg of each vector, 3.75 μg of indicated packaging plasmid, and 0.2 μg of pMD.G. Parallel transfections received 1 μg of pCMVtat. Supernatants were harvested 72 h posttransfection, cleared of debris by centrifugation at 2,500 rpm (1,430 × g) for 15 min at 4°C, and used for infection of D17 cells. For determining luciferase activity, transduced cell lysates were prepared 72 h postinfection, and an aliquot, after appropriate dilution, was assayed for luciferase activity using a commercial kit as described in the legend to Fig. 2. The luciferase activity, normalized to p24 levels, is shown. (B) Gene transfer into 293T by GFP-expressing vectors encoding Tat72 or Tat86 and either containing or lacking CTE. Virus stocks were produced separately for each indicated GFP-expressing vector by cotransfecting 293T cells with 3.75 μg of pgp3virin or pgp3 v, 0.5 μg of pMD.G, and 7.5 μg of each vector. Supernatants were harvested as for panel A and used for infection of 293T cells. The cells were harvested 72 h postinfection, fixed with 5% paraformaldehyde, and analyzed by flow cytometry, as described for Table 2, to obtain the number of GFC per milliliter. The GFP titers (as GFC per milliliter) were normalized to p24 levels (which ranged between 10 and 22 ng/ml for the different vectors) in the supernatants used for infection. The number above each bar is the average titer obtained for the corresponding vector in units of 100,000. Thus, the titer of pN-GIT72 was 2.9 × 10⁵ GFC/ml. The amount of p24 in the supernatant was quantitated using a commercial kit (Cellular Products or Zeptometrix Corp., Buffalo, N.Y.) (64). Error bars correspond to 1 standard deviation and were derived from duplicate experiments.

Virus stocks for pN-LIT72 and pN-LIT86 were produced using pgpirin as the packaging plasmid. Both vectors are expected to be Rev dependent since they contain RRE but no CTE. The results of gene transfer experiments with pN-LIT72 and pN-LIT86 in D17 cells indicate that the Tat72-encoding bicistronic vector was transduced at two- to threefold-higher efficiency than the Tat86-expressing vector. In contrast to pN-LIT72 and pN-LIT86, the CTE-containing vectors, pN-LIT72CTE and pN-LIT86CTE, are expected to be Rev independent since they contain the MPMV CTE (63, 64). We therefore wanted to test pN-LIT72CTE and pN-LIT86CTE in a packaging system devoid of Rev. To produce virus stocks, pN-LIT72CTE or pN-LIT86CTE was cotransfected with the helper plasmid pgp, which does not code for either Rev or Nef. To allow comparison with virus stocks produced with pgpirin, which expresses Nef and Rev, a CMV-Nef-encoding plasmid, pCMVnef (63), was included during virus stock production. This ensured that the only difference between packaging of pN-LIT72 or pN-LIGT86 by pgpirin and pN-LIT72CTE or pN-LIT86CTE was the absence of Rev in the latter case. Virus stocks produced with the Rev-independent vector (pN-LIT72CTE) in a packaging system lacking Rev had approximately 70% of the gene transfer efficiency of the vector stock produced with pN-LIT72 using pgpirin (Fig. 3). Again, pN-LIT72CTE gave two- to threefold-higher gene transfer efficiencies than pN-LIT86CTE.

Tat has been shown previously to be required for efficient reverse transcription in addition to its role in transactivation of the HIV-1 LTR (27). To confirm that levels of Tat produced by the vectors were adequate to allow maximal production of vector RNA for packaging, and also test if coexpression of full-length Tat affected gene transfer, virus stocks were produced for each vector with and without pCMVtat. Gene transfer efficiencies for all vectors were similar both in the presence and absence of a full-length-Tat-expressing plasmid (Fig. 3). This demonstrated that adequate amounts of Tat were being produced by the vectors and that the single-exon Tat was sufficient for gene delivery into D17 or 293 cells. These results were consistent with our previous results, which also showed that Tat72 was adequate for gene transfer into growing and growth-arrested target cells (64). In our previously reported experiments Tat72 was produced by the packaging plasmid, whereas in the experiments presented here, Tat is encoded by the HIV-1 gene transfer vector.

It was not clear from the preceding experiments if the lower levels of luciferase activity observed in transduced cells with the Tat86 vectors was due to lower levels of gene expression, to lower titers, or to a combination of the two. To address this question and to confirm and extend the results using an independent marker gene, we tested bicistronic tat expression vectors expressing the enhanced version of green fluorescent protein (GFP). These vectors were packaged using the packaging plasmid pgp3v or pgp3virin (Fig. 1). The packaging plasmid pgp3v encodes HIV-1 Gag, Pol, Vif, Vpr, Vpu, and Tat72, while pgp3virin encodes Rev and Nef in addition to those proteins present in pgp3v (Fig. 1). These plasmids yield higher titers than vector stocks produced with pgpirin or pgp (64). The pgp3v construct does not encode for Rev and can therefore be used for packaging gene transfer vectors that are Rev independent (i.e., those containing the CTE), while pgp3virin, which encodes both Rev and Nef, would be suitable for packaging Rev-dependent vectors (i.e., those containing RRE but no CTE). The four GFP-encoding vectors, pN-GIT72, pN-GIT86, pN-GIT72CTE, and pN-GIT86CTE, were used to produce separate virus stocks using either pgp3v or pgp3virin. The vectors were pseudotyped with VSV-G. A Nef-expressing plasmid was included for production of virus stocks with pgp3v to allow comparison with pgp3virin. Each of the vector stocks was used for infection of 293T cells. Transduced cells were harvested 3 days postinfection, fixed with 5% paraformaldehyde, and analyzed by flow cytometry. The titers, as deduced from the number of green fluorescent cells (GFC) per ml, normalized to p24 levels, are depicted in Fig. 3B. Consistent with the results for the luciferase-encoding vectors, Tat72-expressing vectors (pN-GIT72 and pN-GIT72CTE) gave approximately two- to threefold-higher titers than the Tat86 vectors (pN-GIT86 and pN-GIT86CTE). Rev-dependent Tat72 and Tat86 vectors (pN-GIT72 and pN-GIT86) packaged with pgp3virin had titers of 1.6 × 10⁴ and 0.7 × 10⁴ GFC/ng of p24, respectively. In contrast, the same vectors packaged in a system lacking Rev gave, as anticipated, 16.5- and 14.9-fold-lower titers, respectively. This result is in accordance with our previous observations (63, 64) showing the requirement for Rev for vectors that contain the RRE but no CTE. Also consistent with our previous observations (63, 64) was the finding that CTE-containing vectors (pN-GIT72CTE and pN-GIT86CTE) could be efficiently packaged and transduced into target cells by the pgp3v packaging plasmid. Thus, pN-GIT72CTE and pN-GIT86CTE, packaged in a system lacking Rev, gave titers of 1.5 × 10⁴ and 0.6 × 10⁴ GFC/ng of p24, respectively, which were essentially identical to titers of vectors lacking the CTE (pN-GIT72 and pN-GIT86) and produced in a Rev-containing packaging system. These same vectors (pN-GIT72CTE and pN-GIT86CTE) could also be efficiently packaged with pgp3virin and interestingly, provided 1.4- to 2-fold-higher titers than corresponding vectors without CTE packaged with the same packaging plasmid. These titers were also higher than the titers obtained with the same vectors packaged in a Rev-independent system using pgp3v. Statisfical analysis indicated that the differences in titers between the CTE-containing vectors and the non-CTE-containing vectors when packaged with pgp3virin were not significant (P > 0.05).

In a separate experiment, levels of GFP expressed by Tat72 and Tat86 vectors in transduced target cells were determined by measuring the geometric mean of fluorescent intensity (GMFI). D17 cells transduced with pN-GIT72 had a GMFI of 268 ± 40 (mean ± standard deviation), while pN-GIT86-transduced cells had a GMFI of 328 ± 24. Cells transduced with the CTE-containing vector, pN-GIT72CTE, had a GMFI of 221 ± 4.45, while those transduced with pN-GIT86CTE had a GMFI of 259 ± 13.2. Thus, Tat72-encoding vectors produced similar levels of GFP expression in transduced D17 cells as Tat86-expressing vectors.

Comparison of bicistronic Tat-encoding vector with vectors that express transgenes under control of internal promoters.

Next, we wished to compare the HIV-1 LTR as a promoter with the sCMV and SV40 promoters for transgene expression in a variety of cell lines. To do this we used two types of reporter genes. One series (pN-LIT72, pN-sCMVluc, and pN-SVluc) had the firefly luciferase gene, and a corresponding set (pN-GIT72, pN-sCMVGFP, and pN-SVGFP) contained the GFP gene under control of the HIV-1 LTR and the sCMV and SV40 promoters, respectively. Virus stocks were produced as described above, using pgp3virin to provide helper function. All vectors were pseudotyped with VSV-G. We first compared virus stocks produced using pN-LIT72, pN-sCMVluc (previously referred to as pN-FS-sCMVluc [64]), and pN-SVluc. Virus stocks of each of these vectors were used to transduce D17, 293, HeLa, and 3T3 cell lines. Transduced cell lysates were harvested 48 h postinfection and assayed for luciferase activity. The results of this experiment are shown in Table 1. D17 and 293 cells transduced with pN-LIT72 showed higher luciferase activity than cells transduced with either pN-sCMVluc or pN-SVluc. In HeLa targets, transduction with pN-sCMVluc resulted in higher luciferase activity than transduction with pN-SVluc or pN-LIT72. Low levels of luciferase activity were noted with all vectors in 3T3 cells.

TABLE 1.

Transduction of various cell lines by HIV-1 vectors encoding luciferase under the control of different promoters

Vectora (p24 level)	Mean ± SD of RLU (relative RLUc) in transduced cell line:
	D17	293	HeLa	3T3
Mock	164 ± 4	174 ± 0	188 ± 6	196 ± 21
pN-LIT72 (∼13 ng/ml)b	4,759,722 ± 590 (100)	8,127,200 ± 1,150,300 (100)	79,538 ± 10,864 (100)	5,272 ± 460 (100)
pN-sCMVluc (∼13 ng/ml)b	135,377 ± 12,294 (3)	2,584,900 ± 579,800 (32)	212,690 ± 31,587 (267)	6,445 ± 877 (122)
pN-SVluc (10 ng/ml)	41,466 ± 1,806 (1)	81,100 ± 1,500 (1)	19,304 ± 574 (24)	2,638 ± 270 (50)

^aAll vectors were pseudotyped with VSV-G. Transduced cells were harvested 48 h postinfection. Cell lysates were prepared in 250 μl of luciferase-lysis buffer, and a 50-μl aliquot was tested for luciferase activity represented as RLU. For some samples, where the RLU was beyond the linear range of the assay, the lysates were first diluted 100-fold prior to the assay (the final RLU was multiplied by 100 to allow comparison of the different vectors). Data were derived from duplicate experiments. 

^bFor pN-sCMVluc and pN-LIT72, only approximate p24 levels are given because the absorbance values were near the upper limit of the linear range of the assay. 

^cCalculated using the formula (RLU of each indicated vector/RLU of pN-LIT72) × 100. Values have been rounded off to the nearest whole number. 

The differences between the vectors could be either due to differences in titers or due to differences in the strength of each of the promoters in the various cell lines. To address this more directly, we used vectors with GFP as the reporter. GFP would allow us to detect and quantitate gene expression at a single-cell level. Again, virus stocks were prepared with each of these vectors and used for transduction of the various cell lines, some of which were also used as targets for luciferase-encoding vectors. Transduced cells were harvested 72 h postinfection by trypsinization, washed and fixed with 4% paraformaldehyde, and analyzed by flow cytometry. The results are shown in Table 2. In all cases, less than 10% of cells were transduced, judging from proportion of cells that were positive for GFP (data not shown), suggesting that most cells harbored only one virus genome. By enumerating the number of transduced cells per well for each virus stock, we were able to estimate the number of transducing units per milliliter for each vector.

TABLE 2.

Transduction of various cell lines by HIV-1 vectors encoding GFP under the control of different promoters

Vector (p24 level)	Mean ± SD of GFCa/ml or GFMIb/cell in transduced cell line:
	D17		293		HeLa		THP-1		3T3
	GFC/ml	GMFI/cell (relative FI)d	GFC/ml	GMFI/cell (relative FI)	GFC/ml	GMFI/cell (relative FI)	GFC/ml	GMFI/cell (relative FI)	GFC/ml	GMFI/cell (relative FI)
Mock	0	2 ± 0	0	1c	0	1 ± 1	0	2c	0	2 ± 0
pN-GIT72 (5.5 ng/ml)	(1.1 ± 0.2) × 105	1,032 ± 31 (100)	9.6 ± 0.4 × 104	834 ± 21 (100)	5.4 ± 0.1 × 105	607 ± 23 (100)	6.9 ± 2.4 × 103	71c (100)	1.4 ± 0.1 × 105	13 ± 0 (100)
pN-sCMVGFP (1.3 ng/ml)	(2.1 ± 0.2) × 104	160 ± 23 (15.5)	1.4 ± 0.4 × 105	30 ± 0 (3.6)	1.6 ± 0.1 × 105	171 ± 8 (28.2)	2.4 ± 0.1 × 103	14 ± 0 (19.7)	9.5 ± 0.8 × 104	27 ± 0 (207.7)
pN-SVGFP (4.1 ng/ml)	(1.5 ± 0.4) × 104	105 ± 8 (10.2)	7.7 ± 1.3 × 104	8 ± 0 (1.0)	1.8 ± 0.1 × 105	31 ± 0 (5.1)	1.6 ± 0.2 × 103	18 ± 1 (25.4)	1.1 ± 0.1 × 105	22 ± 0 (169.2)

^aIndicated cell lines were transduced with 0.5 ml of each virus stock as described previously (64). The cells in each well were trypsinized, quenched with serum-containing medium, and washed with phosphate-buffered saline (PBS). After treatment with 4% paraformaldehyde for 30 min at room temperature, the cells were washed with PBS and resuspended in 500 μl of PBS before analysis by flow cytometry. The GFC concentration was estimated by using the formula [(no. of GFP-positive cells/volume analyzed by flow cytometry) × total cell suspension volume] × dilution factor of virus stock. The titers were corrected for background (derived from mock-transduced cells). The titer is probably overestimated by a factor of 2 to 4 due to several cell divisions between the time of transduction and the time of harvest (approximately 60 h). Data were derived from duplicate experiments. 

^bDetermined from the GFP-positive fraction (gated based on the signal obtained from mock-transduced cells.) The GMFI of mock-transduced cells was derived from control cells gated only to eliminate dead cells and debris. Values have been rounded off to the nearest whole number. 

^cSingle determination. 

^dRelative fluorescence intensity (FI) was calculated using the formula (GMFI of each indicated vector/GMFI of pN-GIT72) × 100. 

Flow cytometry also allowed us to obtain the GMFI of transduced cells for each vector (Fig. 4). Representative flow diagrams obtained with three cell lines are shown in Fig. 4. In D17, 293, and THP-1 (human acute monocytic leukemia cell line) cells, pN-GIT72 achieved 4- to 104-fold-higher GFP expression than vectors with sCMV or SV40 promoters (as deduced from GMFI in transduced cells [Table 2]). In HeLa cells, although pN-sCMVGFP produced an overall lower GMFI than pN-GIT72, there was considerable overlap in expression levels of the two vectors (Fig. 4), with the peaks exhibiting similar GMFIs. In contrast, the LTR proved less efficient in 3T3 cells. All of the three promoters tested provided quite low GFP expression in 3T3 cells and in fact were difficult to visualize by fluorescence microscopy, but they could be distinguished by flow cytometry from untransduced cells, using a second detector for measuring emission between 564 and 604 nm. This allowed gating of untransduced cells exhibiting autofluorescence from cells that were transduced with the HIV vector and therefore expressing GFP (data not shown).

FIG. 4.

Flow cytometric analysis of 293, HeLa, and 3T3 cells transduced with pN-GIT72, pN-sCMVGFP, or pN-SVGFP. Virus stocks were prepared as described in the text and in the legend to Fig. 3. One half-milliliter of each virus stock was used for infection of indicated cell lines. The cells were harvested approximately 60 h postinfection, fixed, and analyzed by flow cytometry as described for Table 2. The horizontal axis shows GFP expression in logarithmic scale; the vertical axis shows cell number. The gate used for analysis (M1) is shown for each cell line. The GMFIs of the positive peaks are indicated.

The titers of the GFP-encoding vectors were lower in THP-1 cells than the other cell lines tested. This is probably due to the higher background or autofluorescence in these cells, which in turn interfered with the accurate enumeration of GFP-expressing cells by flow cytometry.

Thus, the results were comparable for both luciferase and GFP vectors in all cell lines tested except HeLa. In HeLa cells, the pN-sCMVluc vector showed higher luciferase activity than the others (Table 2), whereas pN-GIT72-transduced cells showed levels of GFP expression marginally higher than those in cells transduced by pN-sCMVGFP (Fig. 4; Table 2). We have no explanation for this discrepancy. Nevertheless, these results demonstrated that the Tat-encoding bicistronic vectors were in general more efficient than the vectors that expressed the transgene under the control of either the sCMV or SV40 internal promoter in many cell lines, but not in those cell lines where the HIV-1 LTR has been shown to be weak (e.g., mouse cells) (13, 43). Since most published work with HIV vectors has used the human CMV immediate-early promoter to drive transgene expression, we also tested a vector that expressed GFP under control of this promoter. The CMV promoter for this vector was derived from pcDNA3 (Invitrogen Corp., San Diego, Calif.). This vector achieved somewhat lower levels of gene expression (GMFI of 19.9 ± 0.04) than the sCMV promoter-containing vector in 293 T cells (GMFI of 29.1 ± 0.2).

Another puzzling observation was that both the sCMV and SV40 promoters, as part of the HIV-1 vector, seemed to work very poorly in 3T3 cells. The same promoters gave higher levels of gene expression in the context of murine leukemia virus (MLV)-based vectors in 3T3 cells (data not shown). This would imply that the HIV-1 vector backbone can negatively affect transgene expression under control of heterologous internal promoters in some cell types.

Test for pseudotransduction by VSV-G-pseudotyped vectors.

Several recent studies have shown the disconcerting possibility of pseudotransduction of marker genes by VSV-G-pseudotyped vectors (10, 22, 47). Pseudotransduction is a type of protein or DNA delivery in the absence of bona fide infection. The marker protein or plasmid DNA is incorporated into the virion or pseudovirion. Upon binding and fusion of the virus particles with target cell plasma membrane, the protein or DNA is delivered to the cytoplasm. If enough protein or DNA is delivered, one may wrongly conclude that the target cells were expressing the protein as a result of authentic retrovirus infection process. Pseudotransduction has been observed with highly concentrated preparations of VSV-G-pseudotyped vectors. To rule out pseudotransduction, we did the following experiment.

We prepared virus stocks of pN-GIT72 pseudotyped with either VSV-G or amphotropic MLV envelope glycoprotein. The amphotropic envelope-pseudotyped virus would serve as a negative control since pseudotransduction has been observed only with VSV-G-pseudotyped virus. The virus stocks were concentrated approximately 30-fold using Centricon Plus-20 centrifugal concentrators (Millipore Corp., Bedford, Mass.). To eliminate the possibility of plasmid DNA carryover from transfected cells, virus stocks were treated with RNase-free DNase (Promega, Madison, Wis.) and then used for infection of D17 target cells. Preliminary studies showed that DNase treatment did not adversely affect titer (data not shown). Infections were carried out with cells that were pretreated with 10 or 20 mM azidothymidine (AZT; Sigma, St. Louis, Mo.). AZT was present through out infection and up to the time of harvest. The treatment of target cells with AZT should allow the virus particles to bind and fuse with the plasma membrane but interfere with the reverse transcription of incoming virus RNA and thereby abort the infection process. Under these circumstances, pseudotransduction by protein or DNA delivery is still possible, but no true transduction can occur. Transduced cells were harvested 48 h after infection, fixed with 4% paraformaldehyde, and analyzed by flow cytometry. The results of this experiment are shown in Table 3. Amphotropic envelope-pseudotyped virus had a titer of 6.5 × 10⁶ transducing units/ml, while the VSV-G-pseudotyped virus had a titer of 1.2 × 10⁷ transducing units/ml on D17 cells. Treatment of cells with either 10 or 20 mM AZT decreased transduction of VSV-G-pseudotyped virus by 93 and 97% and that of amphotropic envelope pseudotyped virus by 96 and 98%, respectively, compared to gene transfer into untreated cells. This indicated that pseudotransduction was occurring at very low levels, if at all. The difference between our study and those described previously (10, 22, 47) may be related to the method of concentration of virus particles, the marker gene used within the gene transfer vector, the cell line used for production of virus stocks, or the amount of VSV-G used for production of virus stocks. It is also possible that concentration by ultracentrifugation may result in copurification of marker protein and/or plasmid DNA with the virus particles resulting in pseudotransduction of target cells, whereas ultrafiltration may not have these drawbacks.

TABLE 3.

Effect of AZT on gene transfer into D17 cells by amphotropic MLV envelope or VSV-G-pseudotyped HIV vectors

Envelope	GFC/ml			% Inhibition
	0 AZTa	10 μM AZT	20 μM AZT
VSV-G	1.2 × 107	8.7 × 105	3.8 × 105	93–97
Amphotropic	6.5 × 106	2.4 × 105	1.1 × 105	96–98

^aThe virus stocks were prepared by transfection of 293 cells with 20 μg of pN-GIT72, 10 μg of pgp3virin, and either 0.6 μg of pMD.G (expresses VSV-G) or 6 μg of pSV-A-MLV-env (expresses amphotropic envelope) as described in the legend to Fig. 3, but using a scaled-up transfection protocol using T75 flasks instead of T25 flasks, and concentrated approximately 30-fold using Centricon Plus-20 centrifugal concentrators according to the recommended protocol. D17 cells were pretreated with the indicated amounts of AZT for 1.5 h prior to infection with VSV-G or amphotropic MLV envelope-pseudotyped virus. Before infection, the virus samples were treated with RNase-free DNase in the presence of 10 mM MgCl₂ for 30 min at room temperature and then kept on ice until an aliquot was added to cells. AZT was maintained in the culture during virus adsorption and throughout the incubation of the cells until the time of harvest. The cells in each well were harvested, and the titer was estimated as described for Table 2. 

In this report, we describe novel Tat-encoding HIV-1-based gene transfer vectors that allow high levels of transgene expression using the HIV-1 LTR promoter. Vectors that expressed Tat72 were more efficient than those encoding Tat86. There are several possible explanations for why Tat86-encoding vectors give lower gene transfer efficiencies than Tat72-encoding vectors. One possible explanation is that inserting a cDNA encoding the complete Tat coding sequence downstream of the second exon of tat will result in duplication of sequences of a second coding exon on either side of the marker-IRES cassette (Fig. 1). Since retroviruses can delete sequences between repeats during reverse transcription (16, 37), this process could have resulted in the elimination of the marker-IRES sequence during the infection process, thus accounting for the lower gene transfer efficiency with the full-length Tat-expressing vectors. We are currently conducting experiments to confirm this hypothesis. Other possibilities include interference in the RNA secondary structures between the Tat86 sequences and IRES RNA or unknown function of the Tat86 sequences not present in Tat72.

Finally, the results demonstrated that the CTE-containing vector could transduce genes nearly as efficiently as the vector without the CTE. This Rev-independent system may prove useful for expressing high levels of transdominant Rev in HIV-1 susceptible targets. An additional benefit of the RNA produced from this vector is that it contains the TAR element and RRE, both of which can act as decoys in the event of infection of the intracellularly immune cells by wild-type virus (15). The CTE-containing vector would also be useful to define the requirement for envelope sequences (RRE) in HIV-1-based gene transfer vectors. The vectors described here still retain 1.7 kb of envelope sequence that contains the RRE and the tat/rev splice acceptor site. In the CTE-containing vectors, it should be possible to eliminate much of this envelope sequence and thereby render the vectors safer by removal of unnecessary sequences and also increase its capacity to accommodate larger transgenes.

Our results using CTE in the packaging system are at odds with those obtained by Kim et al. (42) and Gasmi et al. (24) but consistent with previously published results of our group. In the study of Kim et al., the gene transfer vector was based on Rev and RRE and the vector also coded for Rev, while the packaging plasmid contained the CTE. In the study by Gasmi and coauthors, again the packaging plasmid contained the CTE and the gene transfer vector contained the RRE, but Rev was expressed using a separate plasmid. Kim et al. found that the Rev- and RRE-based packaging system resulted in nearly 100-fold-higher titers than the system that used the CTE, while Gasmi and coworkers found a difference of about 10-fold between the two packaging constructs. One possible explanation that has been offered previously is that the placement or length of CTE used might have influenced titers of vectors containing this element (42). There are several significant differences between our study and the above-mentioned ones. We have used the CTE not only in the packaging plasmids but also in the gene transfer vector to create a truly Rev-independent system. Both of the above-mentioned studies required the coexpression of Rev to allow export of RRE-containing and thereby Rev-dependent vector RNAs. This is similar to the combination packaging system (using CTE for expression of packaging plasmid and RRE-Rev for expression of gene transfer vector) we have described in this study and in a previous one (64). Unlike the studies of Kim et al. (42) and Gasmi et al. (24), we have not compared a purely RRE-Rev-based packaging system directly with the CTE-based packaging system. However, in an earlier study (64), we found that the combination packaging system gave gene transfer efficiencies comparable to those of an RRE-Rev-based packaging system. We are now attempting to reconcile these differences among studies.

Tat-encoding vectors may have some unique advantages in overcoming two important hurdles to the implementation of gene therapy. The first is the gradual loss of transgene expression in the transduced target cells (26, 31, 39, 50, 55, 60). This effect has been ascribed to chromatin remodeling by methylation of DNA (26, 31, 32) or due to deacetylation of histones (11, 12). Tat has been shown to recruit histone acetyltransferases to the viral promoter (7, 40, 48, 72). This may keep the chromatin in an open configuration and allow prolonged transgene expression. Even if transgene expression diminishes over time, it may be possible to reverse this silencing by using inhibitors of histone deacetylases such as trichostatin A or sodium butyrate or its analogs such as phenylbutyrate (11, 40). It remains to be demonstrated that the Tat-encoding vectors can indeed allow long-term transgene expression.

The second hurdle for gene therapy is immune-mediated elimination of transduced cells expressing a protein recognized as foreign by the host. A biological property of Tat that may be useful in this context, is the ability of Tat to downregulate the expression of major histocompatibility complex (MHC) class I and class II genes (9, 28, 29, 38, 69), although there are other reports which seem to indicate that Tat possesses no such activity (49, 68). If Tat can indeed down modulate MHC expression, this may prevent immune-mediated elimination of transduced target cells by decreasing antigen presentation in the transduced cells. This property is, however, principally localized to the second coding exon of Tat (29). Therefore, one will have to use a full-length-Tat-encoding vector to prevent immune responses to transduced cells. Thus, Tat-encoding vectors may provide answers to the central problems currently besetting gene therapy.

On the other hand, it is also possible that Tat (particularly Tat72) may engender an immune response to the transduced cells leading to their elimination. Moreover, Tat, which is known to be secreted from infected cells and act in a paracrine manner on neighboring cells, may produce other untoward consequences (14, 53, 54, 57–59, 65, 67, 70). One concern is the observation that Tat may be involved in the induction of Kaposi's sarcoma (2–5, 17–19). Notwithstanding its effect on MHC expression, Tat may also produce other effects on the immune system. For example, Tat has been shown to induce interleukins 2 and 8 in peripheral blood lymphocytes and T-cell lines when used in conjunction with antibodies to CD3 and CD28 (53). These activities can have unforeseen consequences in vivo. Both of the above-mentioned activities of Tat (immune modulation and induction of Kaposi's sarcoma) most likely require that Tat be secreted from one cell and then act on neighboring cells. Although Tat can traverse through the plasma membrane via signals present in the first coding exon of Tat (41, 62), such activities as immune modulation and induction of Kaposi's sarcoma may require that Tat bind to the integrin receptors on the cell membrane via the RGD sequence present in the second coding exon (6, 53, 71). While the single-exon-Tat-encoding vector, which lacks the RGD sequence, may allow high levels of transgene expression without the possible untoward consequences envisaged with full-length Tat, it is not clear which of the Tat-encoding vectors will eventually prove useful for gene therapy. Clearly, the vectors need to be tested in suitable animal models to establish their safety and efficacy. Conversely, these vectors may prove useful for illuminating the role of Tat in the pathogenesis of AIDS in in vivo animal models.