A Lentivirus Packaging System Based on Alternative RNA Transport Mechanisms To Express Helper and Gene Transfer Vector RNAs and Its Use To Study the Requirement of Accessory Proteins for Particle Formation and Gene Delivery

Abstract

A lentivirus-based packaging system was designed to reduce the chance of recombination between helper and gene transfer vector sequences by using the constitutive transport element (CTE) derived from Mason-Pfizer monkey virus for expression of the viral proteins and the Rev-Rev response element (RRE) combination for expression of the gene transfer vector. Using this approach, we evaluated a series of human immunodeficiency virus type 1 packaging constructs that express one or more accessory proteins (Vif, Vpr, and Vpu), in addition to the Gag and Pol proteins, for particle formation and virus stock production for gene transfer. Constructs that also express Vpr or both Vpr and Vpu produced more particles, as measured by a p24 assay, than did plasmids that did not contain these sequences. Transactivation experiments showed that the packaging plasmids that encode Vpr or both Vpr and Vpu also expressed a functional single-exon Tat protein. For these constructs, high-titer virus stocks could be prepared in the absence of a cotransfected Tat-expressing plasmid. Amphotropic-envelope-pseudotyped virus stocks prepared with all of the packaging constructs, irrespective of whether any of the accessory proteins were coexpressed, were equally efficient in transducing growth-arrested HeLa cells. The combination/mixed packaging system was compared to systems that were based on either the CTE alone or Rev and RRE for expression of both the packaging plasmid as well as the gene transfer vector. The combination/mixed packaging system was comparable to the other systems for production of virus stocks, suggesting that this design may prove to be safer for the eventual deployment of lentivirus vectors for therapeutic purposes.

The use of the human immunodeficiency virus type 1 (HIV-1)-based packaging system for gene transfer to a variety of cell types, such as neuronal, muscle, liver, and retinal cells as well as hematopoietic stem cells, is growing in popularity (13, 23, 24, 26, 37, 38, 40). This is due to the ability of HIV-1 to efficiently enter nondividing or quiescent cells (16, 17). Attempts are being made in many laboratories to optimize the packaging and vector constructs to ensure safety without compromising vector titer, so that HIV-1 vectors can eventually be used in a clinical setting. To create a packaging system with HIV-1, one needs a helper construct(s) that produces all necessary HIV-1 proteins, as well as a gene transfer vector. Both helper and gene transfer constructs will require sequences that ensure nucleocytoplasmic transport of their RNAs (34). In the natural context, HIV-1 structural-protein expression is regulated by the Rev protein and its cognate RNA recognition sequence, the Rev-responsive element (RRE), which forms part of the structural-protein gene message. Rev and RRE ensure nucleocytoplasmic transport and expression of the HIV-1 partially spliced and unspliced mRNAs (9, 20). Without a transport mechanism, viral protein expression from partially spliced and unspliced messages is severely reduced. This requirement by HIV-1 can be overcome by using the RNA transport elements from other viruses. For example, the constitutive transport element (CTE) from Mason-Pfizer monkey virus (MPMV) or simian retrovirus type 1 can substitute for this function of Rev and RRE in proviral HIV-1 clones as well as in subgenomic constructs (3, 39). We felt that it would be advantageous to use alternative or mutually exclusive transport systems to express the helper and gene transfer vectors. We rationalized that this would reduce the chance of recombination between helper and gene transfer vectors. Toward that end, we created helper plasmids that contain the CTE and gene transfer vectors that contain the RRE. We compared this combination packaging system with one that uses only CTE or Rev and RRE for expression of both packaging plasmid and gene transfer vector RNAs. The results with amphotropic-envelope-pseudotyped HIV-1 vectors indicated that the combination system was comparable to the CTE- or the Rev-RRE-based systems for production of retroviral vector stocks. There was no evidence for formation of replication-competent retrovirus (RCR) with any of the packaging systems, and prolonged transgene expression was evident even after extended in vitro culture of transduced cells.

We have previously described packaging cell lines which were used to define the requirement of HIV-1 Tat, Rev, and Nef proteins for gene transfer by HIV-1 vectors (34). In this study, using the combination packaging system approach, we evaluated novel HIV-1 packaging constructs that express one or more accessory proteins (Vif, Vpr, and/or Vpu) in addition to the Gag and Pol proteins for particle formation and for production of virus stocks for gene transfer studies. The results of our experiments indicated that packaging constructs that express all three accessory proteins (Vif, Vpr, and Vpu) produced larger numbers of particles and provided higher transduction levels than the one with just the gag-pol sequence. Amphotropic-envelope-pseudotyped virus stocks produced with all packaging constructs were able to transduce growth-arrested cells, suggesting that Vif, Vpr, and Vpu are dispensable for transduction of some cell types. The studies also revealed that the truncated Tat protein expressed from the first coding exon in some of the packaging constructs was as efficient as full-length (86-amino-acid) Tat in transactivation and transduction assays.

HIV-1-packaging plasmids.

To introduce HIV-1 sequences into an expression vector, we created a plasmid vector, pCMC, containing the simian cytomegalovirus (sCMV) immediate-early promoter as well as the CTE and poly(A) signal of MPMV, with a multiple cloning site between the two. This plasmid expression vector was derived from another plasmid, pCMV-VSV-G.CTE (kindly provided by David Rekosh and Marie-Louise Hammarskjöld, University of Virginia, Charlottesville). The sCMV promoter-enhancer corresponds to bp 681 to 1349 of the IE94 gene (GenBank accession no. M16019), and the CTE-poly(A) sequence corresponds to bp 8007 to 8557 of MPMV (GenBank accession no. M12349).

As a first step in deriving the packaging constructs, the encapsidation sequence between bp 751 to 779 of pNL4-3 in pGEM-NL4-3 was deleted by PCR to obtain Δψ-NL4-3.(pGEM-NL4-3 was kindly provided by Antonito Panganiban, University of Wisconsin, Madison, and has been described previously [21].) Fragments from Δψ-NL4-3 of different lengths were introduced into the multiple cloning site of pCMC to obtain the packaging or helper plasmids (Fig. 1). All packaging constructs contain the gag-pol region of HIV-1, but they differ in the number of accessory proteins that they encode (Fig. 1A, Table 2). These vectors are referred to as either pgp or pgpxv, where the x stands for the number (1, 2, or 3) of “V” proteins (Vif, Vpr, or Vpu). The pgp plasmid encodes the HIV-1 gag-pol sequence and a partial sequence of vif. The gp1v plasmid contains the entire Vif coding region but a frame-shifted vpr sequence. The gp2v vector contains both vif and vpr and almost all of the first coding exon of tat. The pgp3v plasmid encodes the sequences for Vif, Vpr, and Vpu and the entire first coding exons of tat and rev.

FIG. 1.

Schematic representation of HIV-1-packaging constructs (A), the HIV-1 provirus (B), and the gene transfer vector used in this study (C). To create the packaging constructs pgp, pgp1v, pgp2v, and pgp3v, fragments from NL4-3 of different lengths (indicated by thick horizontal lines), extending from a restriction enzyme site upstream of gag to sites downstream of pol corresponding to nucleotides 5122 (NdeI), 5785 (SalI), 5999 (SacI) and 6399 (NdeI) of pNL4-3, were introduced into pCMC (described in the text) between the sCMV immediate-early promoter and the MPMV CTE and polyadenylation signal. All plasmids contain a deletion within the encapsidation sequence (Δψ) between bp 751 and 779 of pNL4-3. In the “irin” plasmids, rev and nef cDNAs were positioned downstream of the IRESes of Ha-MSV and EMCV, respectively. The Ha-MSV IRES sequence corresponds to bp 205 to 794 of pHa-MSV (described by Makris et al. [19]) but was derived from pHaMDR1/A (30). The EMCV IRES sequence corresponds to bp 361 to 860 of EMCV (GenBank accession no. X74312). The Nef coding sequence corresponds to bp 8787 to 9407 of pNL4-3. The rev cDNA coding sequence corresponds to bp 981 to 1331 of pCV1 (GenBank accession no. M11840). The Ha-MSV IRES-Rev-EMCV IRES-Nef cassette was first assembled in an intermediate cloning vector and subsequently introduced downstream of pgp, pgp1v, pgp2v, and pgp3v to obtain the vectors pgpirin, pgp1virin, pgp2virin, and pgp3virin, respectively. The accessory and regulatory proteins encoded by the packaging constructs are summarized in the adjoining table. The HIV-1 gene transfer vector pN-FS-sCMVluc (C) was derived from pTR167 (31). The HIV-1 coding sequences between the proximal NsiI site in gag and the distal NsiI site in env (shown) have been deleted in this vector. It contains the firefly luciferase gene driven from an internal sCMV immediate-early promoter. The reporter cassette was positioned between the BamHI and XhoI sites, thereby interrupting the Nef coding sequence. A frameshift mutation (FS) was introduced 27 bp into the gag coding sequence by inserting an A residue between codons 9 and 10. pN-FS-sCMVluc-cte contains the MPMV CTE (GenBank accession no. M12349) (bp 8007 to 8240) downstream of the luciferase coding sequence but in other respects is identical to pN-FS-sCMVluc. The 5′ splice donor site (5′ss) and 3′ splice acceptor site (3′ss) are shown. Details of plasmid construction will be provided on request.

TABLE 2.

Gene transfer into growth-arrested cells by virus stocks produced with the different packaging constructs

Packaging plasmid	Mean RLU ± SD for:
	Untreated HeLa targets	Aphidocolin-treated HeLa targetsa
Mock	166 ± 2	176 ± 6
pgpirin	15,530 ± 1,407	70,472 ± 6,499
pgp1virin	7,375 ± 692	28,022 ± 2,457
pgp2virin	61,945 ± 2,964	285,921 ± 11,151
pgp3virin	111,069 ± 1,364	549,497 ± 36,405
pSV-A-MLV-gagpol	31,979 ± 3,847	385 ± 77

^aAphidocolin was added to a concentration of 15 μg/ml at the time of seeding of the cells into six-well plates and was also included in the virus inoculum during virus adsorption. Aphidocolin-containing medium was replaced daily until the time of harvest. 

We also created versions of the pgp and pgpxv series of packaging constructs that contain, in addition, the rev and nef cDNA sequences. These were expressed via internal ribosome entry sites (IRESes) of Harvey murine sarcoma virus (Ha-MSV) and encephalomyocarditis virus (EMCV), respectively. These packaging plasmids are referred to as pgpirin and pgpxvirin, respectively, and are described in Fig. 1. Again the x in pgpxvirin stands for the number (1, 2, or 3) of “V” proteins of pNL4-3.

To detect viral-protein expression by the packaging plasmids, each packaging plasmid was separately transfected into CMT3-COS (a simian virus 40-transformed monkey cell line) (7) by the DEAE-dextran method (10). Cell lysates were prepared 72 h posttransfection and analyzed by an immunoblotting procedure (35) using either pooled HIV-1-positive human sera or a rabbit antiserum raised against Vif or Vpu. The results of the immunoblotting experiment using anti-HIV human serum indicated that all packaging constructs expressed the Pr55^gag precursor as well as the processed p24 (capsid [CA]) product (data not shown). The precursor proteins for all of the packaging constructs appeared to be processed similarly. pgp2v and pgp3v as well as their “irin” counterparts expressed larger amounts of Pr55^gag as well as the processed p24 CA protein than pgp and pgp1v (and their corresponding “irin” packaging plasmids). pgp3v produced more p24 than pgp2v. In contrast, with regard to the “irin” constructs, pgp2virin appeared to produce an equal amount of or more p24 than pgp3virin. These results were consistent with the results of the p24 assays of transfected-cell supernatants shown in Fig. 2 (see below).

FIG. 2.

(A and B) Particle production by (A) and efficiency of particle export into the medium of (B) 293 cells transfected with packaging constructs. A 3.75-μg quantity of each packaging plasmid was transfected into 293 cells together with 1 μg of pCMVβ-gal. (C and D) The effect of Vpr on particle production (C) and efficiency of particle export (D) by pgp in 293 cells. The indicated amounts of pVpr-wt or pVprAug(−) were cotransfected with a constant amount of pgp and pCMVβ-gal into 293 cells. The supernatants and cell lysates were harvested 72 h posttransfection and assayed for HIV-1 p24 antigen by using a commercial ELISA kit (Cellular Products, Buffalo, N.Y.) in accordance with the recommended protocol. The cell lysates were tested for β-galactosidase (β-gal) activity by using a luminescent β-galactosidase detection kit (Clontech, Palo Alto, Calif.) according to the recommended procedure. p24 levels in the medium were normalized to β-galactosidase activity in cell lysates and are shown in panels A and C. Efficiencies of particle export was estimated from the ratio of p24 in the medium to that seen in the cell lysates and are shown in panels B and D. The average amounts of p24 (in picograms per milliliter) found in the media of cells transfected with the different plasmids are shown above the respective bars. Error bars correspond to 1 standard deviation and were derived from duplicate experiments.

Vif and Vpu expression by the packaging plasmids in the transfected-cell lysates were also detected by the immunoblot procedure, using an antiserum obtained from the NIH AIDS Research and Reference Reagent Program. A 27-kDa protein corresponding to Vif was detected for pgp1v, pgp2v, and pgp3v and their “irin” counterparts, while none was detected, as anticipated, with pgp and pgpirin or in mock-transfected lysates (data not shown). pgp2v and pgp2virin plasmids seemed to express larger amounts of Vif than pgp3v or pgp3virin. Vpu was detected in lysates of cells transfected with the pgp3v packaging construct but not in the other lysates (data not shown). Since pgp3virin was derived from pgp3v, this construct may express a level of Vpu that is undetectable by the immunoblot procedure.

Due to a high background in the immunoblots with the Vpr antiserum, this protein could not be identified unambiguously in the lysates of cells transfected with any of the packaging constructs. Instead, to detect expression of Vpr, we used a functional approach, transfecting 293 cells with each of the plasmid expression constructs and analyzing the transfected cells for cell cycle arrest. The transfected cells were stained with propidium iodide and then analyzed by flow cytometry to determine the DNA content. The results showed, as expected, that pgp2v, pgp3v, and their “irin” counterparts arrested the cells in G₂/M phase of the cell cycle (data not shown). Cells transfected with pgp1v (but not pgp1virin, and both can express a frame-shifted Vpr) also exhibited a partial blockade in G₂/M.

Coexpression of Vpr results in increased particle production.

Previous studies had suggested that some of the accessory proteins might affect particle production by either enhancing transcription from the promoter (e.g., Vpr) (8) or increasing particle release (e.g., Vpu) (4, 15, 32). To test whether this occurred with our packaging constructs, pgp, pgp1v, pgp2v, and pgp3v were separately transfected into CMT3-COS cells or 293 cells (2). The supernatants and cell lysates were harvested 48 to 72 h posttransfection. The supernatants were cleared of cellular debris and assayed for HIV-1 p24 (CA) by using a commercial kit (Cellular Products, Buffalo, N.Y.). The cell lysates were assayed for p24 and for β-galactosidase activity (Clontech, Palo Alto, Calif.). Particle production was determined by normalizing p24 levels to β-galactosidase activity. The efficiency of particle release or export was determined by comparing the ratio of the particles in the supernatant to the particles associated with the cells.

Since similar results were obtained in CMT3-COS cells and 293 cells, only the results of experiments using 293 cells are shown in Fig. 2. The data indicate that cells transfected with pgp2v or pgp3v produced larger amounts of p24 in the medium than those transfected with pgp or pgp1v. This increase in p24 in the medium was most probably due to an increased efficiency of particle release, since the ratio of p24 in the medium to that associated with the cells, as a measure of particle export, was increased in cells transfected with constructs that express Vpr. Similar to what was observed with the pgp and pgpxv constructs, cells transfected with pgp2virin or pgp3virin released larger amounts of p24 into the medium than cells transfected with pgpirin or pgp1virin. Since pgp2v and pgp3v (and their “irin” counterparts) express full-length functional Vpr whereas pgp and pgp1v do not, these results suggest that Vpr may play a previously unrecognized role in particle production.

To directly test the effect of Vpr on particle production, we created two plasmids, one which expresses wild-type Vpr (pVpr-wt) and the other lacking the initiator codon [pVprAUG(−)]. Wild-type or mutant versions of vpr were amplified by PCR with pgp2virin as template and cloned into a expression plasmid that is similar to pCMC (see above) but contains the human CMV promoter-enhancer elements and a synthetic intron of pCI-neo (Promega, Madison, Wis.) in place of the sCMV immediate-early promoter. Cell cycle analysis following transfection of 293 cells with wild-type or mutant versions of Vpr-expressing plasmids confirmed that a functional Vpr was expressed by pVpr-wt but not by pVprAUG(−) (data not shown). To study the effect of Vpr on particle production, pgp (which expresses the Gag and Pol proteins but none of the accessory proteins) was cotransfected with either 1 or 5 μg of pVpr-wt or corresponding amounts of the mutant, pVprAUG(−), into 293 cells. The transfections also received pCMVβ-gal for normalization of the transfection efficiency. Transfected-cell lysates and cleared supernatants were harvested 48 h later and assayed for HIV p24. The cell lysates were also assayed for β-galactosidase activity by using a luminescent β-galactosidase detection kit. The results showed (after normalization to cell-associated p24 or β-galactosidase levels) that Vpr at the 1-μg level increased particle production and release by two- to threefold compared to particle production by pgp in the presence of the mutated Vpr (Fig. 2C and D). However, this effect was not as pronounced at the 5-μg level. These results support our contention that the increased particle production by pgp2v and pgp3v is most likely due to the expression of Vpr by the constructs.

Cells transfected with pgp3v produced the highest levels of particles in the medium of transfected cells (Fig. 2A), as measured by the p24 assay. pgp3v differs from pgp2v in that it codes for all three accessory proteins, Vif, Vpr, and Vpu, whereas pgp2v codes for Vif and Vpr but not Vpu. Several studies have shown that Vpu can accelerate particle release (32, 36). The increased particle production by pgp3v in comparison to pgp2v is most likely due to the expression of Vpu by pgp3v.

Vpu-expressing viruses are exported more efficiently in human cells than in monkey cells (32). We therefore tested the efficiency of particle release by pgp and pgpxv constructs in CMT3-COS (simian origin) and 293 (human origin) cells. Surprisingly, we saw similar increases in the efficiency of particle release by Vpu⁺ and Vpu⁻ constructs in the two cell types (data not shown). An observation similar to ours was recently made by Gasmi et al. (6), who also found that Vpu did not affect particle production in 293 cells. It may be that the effect of Vpu not only differs in simian and human cells but also varies among human cell lines.

In contrast to what we observed for pgp2v and pgp3v, pgp2virin and pgp3virin released similar levels of particles (as measured by the p24 assay) into the medium (Fig. 2A) of transfected cells. In addition, no significant difference in particle release (as measured by the ratio of particles in the medium to particles associated with cells) was noted for the two constructs (Fig. 2B). This may be due to the expression of lower levels of Vpu by pgp3virin than by pgp3v (see the results of immunoblot experiments above). pgp2virin and pgp3virin, nevertheless, showed higher levels of p24 than did pgpirin and pgp1virin. These results are consistent with the immunoblot data described earlier. We also noticed that pgp1v consistently gave two- to threefold-lower values than the plasmid that encoded no accessory or regulatory proteins in terms of both the absolute number of particles produced and the efficiency of particle release. These results were confirmed with independent clones of the packaging plasmids (data not shown). We are presently exploring this phenomenon in greater detail.

pgp2v and pgp3v express a single-exon Tat that is functional in transactivation assays.

The HIV-1 Tat protein in pNL4-3 is 86 amino acids long. The coding region of tat is separated into two exons. The first exon codes for 72 amino acids of the Tat protein. It was shown previously that almost all of the transactivation function of Tat is encoded in the first coding exon (5, 29). pgp3v and pgp3virin contain the entire first coding exon of tat and have the potential to express a functional Tat protein. pgp2v and pgp2virin encode the first 57 amino acids of Tat, but their sequences diverge downstream of amino acid 57. Since the Tat sequences encoded by pgp2v and pgp2virin retain the arginine-rich nucleic acid binding domain, pgp2v and pgp2virin should also have the potential to express a functional Tat protein. To determine if this is indeed the case, the different packaging constructs were transfected into CMT3-COS cells together with the reporter plasmid pLTR-luc-BGHpA, which expresses firefly luciferase under the control of the HIV-1 long terminal repeat (LTR). pCMVtat (which expresses full-length HIV-1 Tat) was included in parallel transfections with the packaging plasmids. All cells undergoing transfections, except the mock-transfected cells, also received pCMVβ-gal. Cell lysates were harvested and assayed for luciferase and β-galactosidase activities. The results of this experiment are shown in Fig. 3. Cells transfected with the luciferase plasmid alone gave a mean value ± standard deviation of the mean of 1,652 ± 865 relative light units (RLU), which corresponds to the basal promoter activity in CMT3-COS cells. On addition of a full-length-Tat-expressing plasmid (pCMVtat), the luciferase activity increased 44-fold to 73,159 ± 5,426 RLU. Transfection of cells with pgp, which does not express any accessory proteins, did not significantly augment basal levels of transcription from the HIV-1 LTR. Cotransfection of pgp and pLTR-luc-BGHpA with a Tat-expressing plasmid, pCMVtat, resulted in luciferase activity levels comparable to those for the control transfection with the reporter and pCMVtat alone. Cotransfection of pLTR-luc-BGHpA with pg1v (which contains an intact Vif coding sequence) gave about twofold-higher basal levels of expression from the HIV-1 LTR. Again, cotransfection of pCMVtat and pgp1v gave RLU values similar to those obtained with pLTR-luc-BGHpA and pCMVtat. With pgp2v and pgp3v, very high levels of transactivation of the HIV-1 LTR were noted even in the absence of cotransfected pCMVtat. Inclusion of pCMVtat with these constructs did not augment expression from the HIV-1 LTR, in contrast to what was noticed with the pgp and pgp1v constructs. These results demonstrate that pgp2v and pgp3v express a functional single-exon Tat protein and that saturating levels of Tat must be produced under the conditions used, since the transactivation of the LTR was not augmented by inclusion of a full-length Tat-expressing plasmid. Interestingly, Tat also increased transcription from the sCMV promoter, as determined by measuring the β-galactosidase activity in the cell lysates (Fig. 3B). This effect was not as dramatic as that seen with the HIV-1 LTR and resulted in an increase in β-galactosidase activity of about fivefold.

FIG. 3.

(A and B) Transactivation of the HIV-1 LTR (A) or the sCMV promoter (B) by the different packaging plasmids. Five micrograms of each of the indicated packaging plasmids was transfected separately into CMT3-COS cells together with 5 μg of pLTR-luc-BGHpA and 2 μg of pCMVβ-gal. pLTR-luc-BGHpA contains the firefly luciferase gene under the control of the HIV-1 LTR as well as the bovine growth hormone poly(A) signal downstream of the luciferase coding region. Parallel transfections also received 2 μg of a CMV-tat plasmid (pCMVtat) (34). Transfected-cell lysates were prepared 72 h posttransfection and assayed for luciferase and β-galactosidase activities. Cell lysates were prepared in 0.5 ml of a luciferase lysis buffer, and a 20-μl aliquot of a 1:100 dilution was used. Luciferase activity was assayed by using a kit and a luminometer (Analytical Luminescence Laboratory, Sparks, Md.) in accordance with the manufacturer’s protocol. β-Galactosidase activity was measured as described in the legend to Fig. 2. RLU are shown for transfections with (+) (cross-hatched boxes) and without (−) (boxes with wavy lines) cotransfected pCMVtat. The fold increase in the presence of Tat compared to the level in its absence is shown above the respective bar for each plasmid. Error bars correspond to 1 standard deviation and were derived from duplicate experiments.

The transactivation experiments were repeated in 293 cells and gave comparable results, indicating that pgp2v, pgp3v, and their “irin” counterparts expressed a functional Tat protein (data not shown). In contrast to the observed effect of Tat on the sCMV immediate-early promoter in CMT3-COS cells, no such effect was discernible in 293 cells. This may be due to the fact that the CMV immediate-early promoter is already functioning at a maximal rate in 293 cells (22, 28, 33).

Efficient transduction of target cells by virus stocks produced with single-exon-Tat-encoding packaging plasmids pgp2virin and pgp3virin.

In a previous study (34), we found that Tat was essential for production of high-titer virus stocks. This was attributed to either transactivation of the HIV-1 LTR leading to the production of increased amounts of packagable RNA from the gene transfer vector or the effect of Tat on reverse transcription during infection (11). The above-described transactivation experiments with the packaging plasmids (Fig. 3) indicated that two of the constructs (pgp2v and pgp3v) and their “irin” counterparts expressed a functional Tat protein. It was therefore likely that the production of high-titer virus stocks from pgp2virin and pgp3virin constructs would not require cotransfection with a Tat-encoding vector. To test this premise directly, virus stocks were prepared separately for each packaging plasmid by transfection of CMT3-COS cells alone or with pCMVtat. All cells undergoing transfection also received a gene transfer vector (pN-FS-sCMVluc) (Fig. 1) and an amphotropic murine leukemia virus envelope-expressing plasmid (pSV-A-MLV-Env). The virus stocks obtained from the transfected-cell supernatants were then tested on HeLa cells. The lysates of transduced cells were assayed for luciferase activity. The results are shown in Table 1. Virus stocks produced with pgpirin and pgp1virin (which do not encode a functional Tat protein) exhibited low transduction levels in the absence of a Tat-expressing plasmid. The titers increased by 88- and 11-fold, respectively, when pCMVtat was included during the production of virus stocks. In contrast, for virus stocks prepared with pgp2virin and pgp3virin, transduction efficiencies were similar with and without cotransfected pCMVtat, as expected from the transactivation experiments (Fig. 3). The highest levels of transduction were noted with virus stocks made by using the pgp3virin plasmid; they were about twofold higher than those obtained with the pgp2virin plasmid. The increased level of transduction evident with pgpirin and pgp1virin in the presence of Tat was most likely due to increased production of vector RNA for packaging and is in part also explained by the increase in p24 levels. The increase in particle production was most likely due to transactivation of the sCMV promoter, which drives gag-pol expression, by Tat (Fig. 3B). These experiments indicated that the 72-amino-acid Tat was sufficient for efficient transduction of target cells and that Tat was being produced by pgp2virin and pgp3virin in the transfected cells in adequate amounts for the production of virus stocks. Although there was roughly a 21-fold or greater increase in particle production by pgpirin in comparison to pgp2virin or pgp3virin, their gene transfer efficiencies differed by only 3- to 6-fold. From this it appears that the limiting factor may be the amount or accessibility of gene transfer vector RNA available for packaging by the viral Gag or Gag-Pol proteins. Similar observations have been made by other investigators (27).

TABLE 1.

Effect of Tat expression in the producer cell on gene transfer by virus stocks produced with the different packaging constructs

Packaging plasmida	pCMVtatb	p24 concn, pg/ml (mean ± SD)c	RLU
			In HeLa targetsd (mean ± SD)	Fold increase with Tate
Mock		0	367 ± 1
pgpirin	−	11 ± 11	392 ± 1
pgpirin	+	168 ± 16	34,666 ± 17	88
pgp1virin	−	155 ± 16	1,168 ± 191
pgp1virin	+	512 ± 76	12,497 ± 955	11
pgp2virin	−	3,037 ± 864	127,975 ± 4,873
pgp2virin	+	3,625 ± 249	108,593 ± 2,950	0.9
pgp3virin	−	3,139 ± 64	227,229 ± 1,013
pgp3virin	+	6,896 ± 1,440	223,132 ± 957	1

^aVirus stocks were prepared by cotransfecting 5 μg of packaging plasmid, 5 μg of an amphotropic-envelope-expressing plasmid (pSV-A-MLV-Env), and 10 μg of the gene transfer vector (pN-FS-sCMVluc [Fig. 1]) into CMT3-COS cells by the DEAE-dextran method (10). Parallel transfections received 2 μg of a CMV-tat plasmid, pCMVtat. Supernatants were harvested 72 h posttransfection and cleared by centrifugation at 2,500 rpm (1,430 × g) at 4°C for 15 min. The virus stocks were stored at −70°C in aliquots. 

^b+, pCMVtat added; −, pCMCtat not added. 

^cp24 levels were determined by using a commercial ELISA kit (Cellular Products, Buffalo, N.Y.). 

^dHeLa cells (2 × 10⁵) were seeded into each well of a six-well plate the day before infection. On the day of infection, the wells were rinsed twice with 2 ml of Tris-buffered saline, pH 7.4 (137 mM NaCl, 20 mM KCl, 25 mM Trizma base, 0.7 mM Na₂HPO₄, 0.9 mM CaCl₂, and 0.5 mM MgCl₂), and incubated overnight with 1 ml of virus-containing supernatant in the presence of 10 μg of DEAE-dextran or Polybrene/ml. The next morning, an additional 2 ml of complete medium (Iscove’s modification of Dulbecco’s modified Eagle’s medium with serum and antibiotics) was added, and the plates were incubated for 48 h prior to harvest. At harvest, the cells were washed twice with calcium- and magnesium-free phosphate-buffered saline and then lysed in 250 μl of a commercial lysis buffer (Pharmingen or Analytical Luminescence Laboratory). An aliquot was tested for luciferase activity (as described in the legend to Fig. 3), which is represented as RLU. 

^eRatio of RLU in transduced HeLa cells for virus stocks produced in the presence of Tat-expressing plasmid to that produced in its absence for each packaging plasmid. 

The pgpxvirin plasmids were designed to express rev and nef via IRESes. To determine if the “irin” plasmids were producing adequate amounts of Rev and Nef, virus stocks were produced by transfecting CMT3-COS cells with pgp3virin, as described above, but with out without added pCMVnef and pCMVrev. pCMVnef and pCMVrev, which express Nef and Rev, respectively, under the control of the sCMV immediate-early promoter, have been previously described (34). Virus stocks were tested on HeLa targets, and luciferase activity in transduced cells was determined as described earlier. Addition of pCMVnef and pCMVrev during virus production did not augment transduction of the luciferase marker gene, indicating that sufficient amounts of Rev and Nef were being produced by the pgp3virin construct (data not shown).

To estimate the titer of virus stocks produced by the pgp3virin construct, we used gene transfer vectors with other marker genes, such as hygromycin phosphotransferase or the green fluorescent protein gene. The titer obtained with the pgp3virin construct was around 1 × 10⁵ to 2 × 10⁵ transducing units/ml (data not shown).

Efficiency of gene transfer into nondividing target cells by virus stocks produced with the different packaging constructs.

To determine if coexpression of Vif, Vpr, and Vpu affected transduction of growth-arrested cells, virus stocks produced with each of the packaging plasmids were tested on HeLa cells that were either growing or growth arrested. HeLa cells were treated with the DNA polymerase alpha inhibitor aphidocolin (15 μg/ml) to bring about growth arrest in G₁/S phase (12). A preliminary study was done to ensure that the aphidocolin concentration used was adequate for blocking cell cycling. For metabolic labeling of S-phase active cells, the cells were incubated with the thymidine analog 5-bromo-2′-deoxyuridine (BrdU) and then stained with an anti-BrdU antibody and a fluorescein isothiocyanate-conjugated secondary antibody. Flow cytometry demonstrated that the aphidocolin concentration used was effective in blocking the cell cycle in G₁/S (data not shown). As an independent biological control, we used virus stocks prepared with Moloney murine leukemia virus (MoMLV). MoMLV has been previously shown to infect dividing cells but not growth-arrested cells (17). MoMLV virus stocks were prepared by transfection of CMT3-COS cells with pSV-A-MLV-gagpol (5 μg), pSV-A-MLV-env (5 μg), and pLgSVluc (10 μg). pLgSVluc is an MoMLV vector that expresses the luciferase gene under the control of an internal simian virus 40 early promoter. (Details of the construction of this vector will be provided on request.) The results of the gene transfer experiments are shown in Table 2 and indicate that virus stocks from all of the different HIV-1-packaging constructs were equally efficient in transducing aphidocolin-treated and untreated HeLa cells. In contrast, the MoMLV vector transduced only growing HeLa cells. Interestingly, in several independent experiments, aphidocolin-treated cells transduced with the HIV-1 vectors exhibited about two- to fivefold-higher luciferase values than untreated cells. We have no explanation for this observation.

The results of our studies are in agreement with the general conclusion of several studies which showed that virus stocks produced in the absence of one or more accessory proteins are capable of transducing certain proliferating and growth-arrested cell lines (such as HeLa, 293, or HOS cells) (14, 25, 40). In contrast, Kafri and coworkers (13) found that Vpr and Vif were required for efficient gene delivery into the liver but not for transduction of terminally differentiated neurons.

Comparison of packaging systems using different combinations of RNA transport mechanisms for expression of packaging and gene transfer vector RNAs.

To reduce the chance of RCR formation, it may be advantageous to use alternative RNA transport systems for expression of helper and gene transfer vector RNAs (Fig. 4). We therefore compared our combination system, in which the packaging plasmid uses the CTE while the gene transfer vector uses Rev and RRE, with a system that exclusively uses either CTE or Rev and RRE for expression of both helper and vector RNAs. To this end, we transfected 293 cells with the following combinations of packaging and gene transfer vector plasmids to produce virus stocks: (i) pgp3virin and pN-FS-sCMVluc (combination system); (ii) pgp3v, pN-FS-sCMVluc-cte, and pCMVnef (CTE system); and (iii) pCMVΔR9 and pN-FS-sCMVluc (RRE-Rev system). Since pgp3v does not express either Rev or Nef, pCMVnef was added during production of virus stock with this packaging plasmid, to allow comparison with pgp3virin. This ensured that the only difference between pgp3v and pgp3virin would be the absence of Rev during virus stock production with pgp3v. pCMVΔR9 expresses all HIV-1 proteins with the exception of Vpu and Env and is regulated by the Rev-RRE system (26). This plasmid was kindly provided by Didier Trono, University of Geneva Medical School, Geneva, Switzerland. All transfections included an amphotropic MoMLV envelope-expressing plasmid. Virus stocks obtained from these transfections were then tested on HeLa cells, and the transduction efficiency was determined by the luciferase assay. The luciferase activity was normalized to p24 levels for each virus stock. The results are shown in Table 3. Virus stocks produced with pgp3v gave 8 to 12 RLU per pg of p24. Virus stocks produced with pgp3virin, in contrast, gave 127 to 167 RLU per pg of p24, an increase of 14- to 16-fold. Virus stock production with both of these helper plasmids is based on the combination RNA transport system. The difference was the absence of Rev during virus production with pgp3v while Rev was expressed by pgp3virin during virus stock production with the latter helper plasmid. The gene transfer vector is a Rev-dependent vector since it has the RRE but no CTE. The results obtained with virus stocks produced with the two different helper plasmids are in accordance with our previous results showing that Rev was essential for obtaining high transduction levels for vectors that were based on RRE (34). To determine if we could restore the titer by rendering the gene transfer vector Rev independent, we used a gene transfer vector that contained the CTE, pN-FS-sCMVluc-cte (Fig. 1), instead of pN-FS-sCMVluc for deriving virus stocks with pgp3v. The normalized titer of virus stocks produced with the CTE-containing vector was 22- to 27-fold higher than that obtained with the Rev-dependent vector in the absence of Rev in the same system.

FIG. 4.

Strategies for construction of HIV-1-based packaging systems. (A) In the RRE-Rev-based packaging system, expression of RNA from both helper plasmid and gene transfer vector is regulated by Rev and RRE. (B) In the CTE-based system, helper and vector RNA expression is regulated by the MPMV CTE. (C) In one combination packaging system (1), the helper plasmid is regulated by the CTE while the gene transfer vector is controlled by RRE and Rev. In an alternative scenario (2), the helper plasmid is regulated by RRE and Rev while the gene transfer vector is controlled by the CTE. Possible sites of recombination between helper and vector constructs are indicated with bidirectional arrowheads or arrows. For clarity, the transgene expression cassette in the gene transfer vector and accessory and regulatory proteins in the helper constructs are not depicted.

TABLE 3.

Comparison of combination packaging system with CTE-based and RRE-Rev-based packaging systems for gene transfer efficiency

Packaging plasmid	Gene transfer vector	RNA transport signal in packaging plasmid/vector	Rev in system?	Luciferase activity in transduced HeLa cells, mean RLU ± SD (RLU/p24 ratio)
				Expt 1	Expt 2
Mock				220 ± 6	297 ± 32
pgp3va	pN-FS-sCMVluc	CTE/RRE	No	13,626 ± 606 (12)	17,697 ± 2,579 (8)
pgp3virin	pN-FS-sCMVluc	CTE/RRE	Yes	648,316 ± 45,932 (167)	635,787 ± 39,571 (127)
pgp3va	pN-FS-sCMVluc-CTE	CTE/CTE	No	103,232 ± 1,774 (326)	194,495 ± 49,736 (129)
pCMVΔR9	pN-FS-sCMVluc	RRE/RRE	Yes	59,584 ± 9,507 (113)	54,202 ± 2,456 (93)

^aOne microgram of pCMVnef (34) was added during transfection of 293 cells for virus stock production. 

Finally, we compared the combination packaging system and the CTE-based packaging system with a Rev-RRE-based system. The packaging plasmid used in the Rev-RRE system was the previously described pCMVΔR9 (26). Virus stocks produced with this packaging plasmid gave relative titers of 93 to 113 RLU per pg of p24. Since pCMVΔR9 has a mutation in Vpu, this may explain the titer difference between virus stocks produced with pgp3virin and those produced with pCMVΔR9. However, this premise needs to be tested directly.

Assay for RCR and persistence of transgene expression.

We checked the virus stock products of the packaging systems described in the previous section (Table 3) for RCR formation by using marker rescue assays and an enzyme-linked immunosorbent assay (ELISA) for the HIV-1 CA antigen. For these experiments, we used an extremely sensitive reporter gene (luciferase) in combination with 293 target cells, which allow high levels of expression from the internal sCMV immediate-early promoter in the HIV-1 vector. The transduction of 293 cells for marker rescue was done as follows. 293 cells (2.5 × 10⁶) in T25 flasks were transduced with 1-ml volumes of virus stocks produced by the three different types of packaging systems shown in Table 3. This corresponded to 10- and 3- to 6-fold-higher transducing units of virus stocks produced by the combination system than by the RRE-Rev-CTE-based systems, respectively. Following infection, for the marker rescue assay, the primary transductants were cultured in the presence of 4 μg of Polybrene/ml (to aid the spread of any replication-competent virus in the culture) for at least 2 days. The medium was then replaced with fresh medium lacking Polybrene, and culture supernatants were harvested the following day for use in infection of fresh 293 cells (secondary transductants). The secondary transductants were assayed for luciferase activity. The results of this experiment are shown in Table 4. While the primary transductants showed abundant levels of luciferase expression, no luciferase activity significantly above background was observed in the secondary transductants. This was true for virus stocks prepared by all three of the different packaging systems (the combination system, the CTE-based system, and the Rev- and RRE-based system). The primary transductants were maintained in culture by subculturing cells approximately twice a week at a density of 10⁶ per flask (which corresponded to a 1:5 to 1:10 split at each passage). No selection agent was used in the maintenance of these cells, since the vectors did not contain any antibiotic resistance genes. After 10 passages, the cells were harvested and assayed for luciferase activities, and the resultant values were compared to those of the cell lysate obtained from the first passage. The results are shown in Table 4. Luciferase activity could be readily detected in transduced cells at the 10th passage. To determine if the persistence of the transgene could be explained by the spread of the marker within the culture by a replication-competent virus, the supernatants from passages 1, 2, 5, and 8 were assayed for HIV-1 p24 by using a commercial ELISA kit. The results of the p24 assay indicated that the residual input p24 detected at the early passages from the initial inoculum quickly decreased to below detectable levels by passage 5 and remained so even at passage no. 8 (data not shown). This was true for virus stocks derived from the combination packaging system, which was tested at as large of an inoculum as possible (resulting in up to 10-fold-higher p24 levels than those of virus stocks produced with the other packaging systems). To further rule out the possibility that a virus not detectable by the p24 assay was spreading the transgene within the culture, the supernatant of the 10th passage was harvested and used for infection of naive 293 cells in the presence of 10 μg of DEAE-dextran/ml. Cell lysates of these secondary transductants were prepared 60 h postinfection and assayed for luciferase activity. As shown in Table 4, only background levels of luciferase activity could be detected in these secondary transductants. Taken together, these results suggest that no RCR was produced during culture of transduced 293 cells, that marker gene expression was maintained at high levels, and that this level of expression remained virtually unchanged in the transduced cells for over 10 passages. Finally, the results also suggested that there was no negative selection against the transduced cells in culture.

TABLE 4.

Duration of marker expression after gene transfer with virus stocks produced by using different packaging systems, and assay for RCR in virus stocks by marker rescue

RNA transport signal in packaging plasmid/ vector	Luciferase activity (mean RLU ± SD)
	Passage no. 1		Passage no. 10
	Primary transductants	Secondary transductants	Primary transductants	Secondary transductants
Mock	181 ± 1	180 ± 3	220 ± 70	119 ± 5
CTE/RRE	1,791,220 ± 161,831	226 ± 42	1,694,012 ± 25,362	112 ± 8
CTE/CTE	555,180 ± 84,476	174 ± 8	504,154 ± 30,390	124 ± 3
RRE/RRE	132,986 ± 17,088	168 ± 8	100,578 ± 6,976	116 ± 6

In this paper, we describe a strategy for designing lentivirus-packaging systems with a reduced chance of generating RCR during virus production for gene transfer experiments. The strategy is based on decreasing regions of homology between the HIV-1 packaging/helper construct and the gene transfer vector by using dissimilar RNA transport elements toward the 3′ ends of the two constructs. Figure 4 shows, schematically, lentivirus packaging systems that are based on Rev-RRE alone, CTE alone, or a combination thereof. In the Rev-RRE- and the CTE-based systems, the helper and gene transfer vectors exhibit considerable homology at both the 5′ and 3′ ends. In contrast, the combination system exhibits homology only at the 5′ end. There is, therefore, a higher likelihood of recombination between the packaging and gene transfer vectors in the case of packaging systems that use the same RNA transport elements at their 3′ ends than with the combination packaging system. For the combination RNA transport-based approach, one can use the CTE for expression of viral proteins and Rev-RRE for expression of the gene transfer vector, or vice versa. Using the CTE in the gene transfer vector can have a potentially negative consequence. A nonhomologous recombination between the gene transfer vector upstream of the CTE and viral protein coding sequence downstream of helper sequences, in conjunction with a homologous recombination at the 5′ end, can result in viral protein expression in transduced target cells by rendering HIV-1 protein expression Rev independent. Although we found no evidence of RCR formation, validation of the purported safety of the combination packaging system requires the development of more-sensitive in vitro and in vivo assay systems.

The packaging system we have used in the present studies is still not optimized to remove all regions of homology at the 3′ ends of the helper and gene transfer vectors. For instance, the vector and helper sequences overlap in the nef (approximate overlap, 500 bp) and rev (approximate overlap, 100 bp) regions in the case of the pgpirin and pgxvirin helper plasmids. However, a recombination at this site would essentially eliminate the RNA transport mechanisms (CTE or RRE) from the recombinant. Therefore, only proteins expressed from spliced messages would be expected to be produced in the recombinant. The system can be modified to further improve its safety. For instance, vesicular stomatitis virus G-pseudotyped vectors have been shown to transduce quite efficiently even in the absence of Nef expression in the producer cell (1, 18). Thus, one can reduce the chance of RCR formation by using vesicular stomatitis virus G protein to pseudotype the virus and eliminate Nef coding sequences from the packaging plasmid. It is not known whether the overlapping sequences in rev of the gene transfer vector can be removed without affecting transduction or expression from the internal promoter, since this would entail removal of the 3′ splice acceptor site of the second coding exon of rev and tat, which abuts the 5′ end of the reporter cassette.

In a previous study, we compared CTE-based and Rev-RRE-based packaging systems and found that they produced comparable titers. We cannot directly compare the titers of this study with those of the previous one, since the previous study used packaging cell lines whereas in this study we used a transient-transfection approach. In the previous study, we selected for cell lines that constitutively expressed high levels of p24, whereas with the transient-transfection method, one deals with a population of cells that may express variable amounts of p24. Thus, any differences that could have resulted from the use of different RNA transport mechanisms for expression may have been obscured during selection for efficient particle-producing cell lines in the earlier study.

Kim et al. (14) and Gasmi et al. (6) also compared packaging plasmids that were regulated by Rev-RRE or the CTE. In the study by Kim et al., the gene transfer vector was based on Rev and RRE and the vector also coded for Rev. In the study by Gasmi and coauthors, Rev was expressed by using a separate plasmid. Kim et al. found that the Rev- and RRE-based packaging system resulted in nearly a 100-fold-higher titer than the system that used the CTE, while Gasmi and coworkers found a difference of about 10-fold between the two packaging constructs. This is in contrast to our results showing that the highest titers were obtained with the combination system, which resulted in titers that surpassed those obtained with the Rev- and RRE-based system by 10- to 12-fold and those achieved with the purely CTE-based system by about 3- to 6-fold. The lack of agreement of results may be a reflection of the differences in the packaging and gene transfer vector plasmids used by the groups. We are presently conducting studies to reconcile these differences.

Several studies, including our own, have shown that Tat increases the gene transfer efficiency of vectors that contain the HIV-1 LTR as the promoter (14, 25, 34). This is possibly due to Tat’s increasing transcription from the viral LTR promoter to produce abundant vector RNA for packaging and also to its recently described effect on reverse transcription (11). In this study, we confirmed our previous observations, and in addition, the results indicated that Tat expressed from the first coding exon (72-amino-acid Tat) in a subset of the packaging plasmids (pgp2virin and pgp3virin) was sufficient for efficient gene transfer into dividing and growth-arrested target cells. This enables one to design safer packaging constructs, such as the ones described here, since it allows the deletion of the entire HIV-1 sequence downstream of the first coding exon of tat.

We compared particle formation by and gene transfer efficiencies of HIV-1-based packaging constructs that differ in the number of accessory proteins (Vif, Vpr, and Vpu) they encode. Although we found that coexpression of accessory proteins did not affect gene transfer into growth-arrested HeLa cells, it is clear from other studies, such as those by Kafri et al. (13), that accessory proteins may be essential for efficient gene delivery to some cells or tissues, such as the liver. The difference between our study and the previous studies is that we found that plasmid constructs that express Vpr and Vpu showed higher levels of particle production than those which did not express any of the accessory proteins, which was in turn reflected in the titers obtained with the virus stocks produced by using the various constructs. These results are more in line with the observations of Goh et al. (8), who showed that Vpr increases particle production by manipulating the cell cycle to up regulate gene expression from the virus LTR. Since the HIV-1 proteins in our packaging constructs were expressed under the control of the sCMV immediate-early promoter and particle formation was normalized to β-galactosidase expression from a plasmid containing the same promoter, it appears unlikely that the effect of Vpr is due to transcriptional transactivation of the sCMV promoter. Cotransfection experiments with the Vpr-expressing plasmid demonstrated that Vpr has a hitherto-unrecognized function in particle export. The series of packaging constructs we have created will thus be useful not only to address the role of viral accessory proteins in the biology of the virus but also in delineating the requirement for these proteins for gene transfer into a variety of target cells.