Optimization of Feline Immunodeficiency Virus Vectors for RNA Interference

Abstract

RNA interference (RNAi) occurs naturally in plant and animal cells as a means for modulating gene expression. This process has been experimentally manipulated to achieve targeted gene silencing in cells, tissues, and animals, using a variety of vector systems. Here, we tested the hypothesis that vectors based on feline immunodeficiency virus (FIV) could be used for coexpression of reporter constructs and RNAi expression cassettes. We found, unexpectedly, in our initial constructs that placement of RNAi expression cassettes downstream from a polymerase II (pol II)-expressed reporter gene inhibited reporter expression but not vector titer. Through a series of intermediate vector constructs, we found that placement of the RNAi expression cassette relative to the Rev response element and the pol II expression cassette was critical for efficient RNAi and reporter gene expression. These results suggested that steric factors, including RNA structure and recruitment of competing transcriptional machinery, may affect gene expression from FIV vectors. In a second series of studies, we show that target sequence silencing can be achieved in cells transduced by FIV vectors coexpressing reporter genes and 3′ untranslated region resident microRNAs. The optimized FIV-based RNAi expression vectors will find broad use given the extensive tropism of pseudotyped FIV vectors for many cell types in vitro and in vivo.

RNA interference (RNAi) refers to posttranscriptional control of gene expression by inhibitory RNAs. RNAi can be induced in cells by naturally occurring endogenous microRNAs (miRNAs) or through artificial expression of inhibitory RNAs (e.g., small inhibitory RNAs [siRNAs], short hairpin RNAs [shRNAs], and synthetic microRNAs) designed to specifically reduce expression of a gene of interest (4, 11, 29, 54). Over the past several years, inhibitory RNAs have been used in basic research to elucidate gene function or as tools in gene therapy to develop preclinical antivirals (5, 24, 31, 40, 50) and treatments for cancer (36, 44, 46) and dominant genetic diseases, such as Huntington's disease and spinocerebellar ataxia type 1 (13, 32, 33, 35, 52). Numerous methods have been developed to introduce inhibitory RNAs into cells, including lipid-based transfection of in vitro-synthesized siRNAs and gene transfer using viral vectors containing promoters that constitutively transcribe shRNAs or microRNAs (10, 13, 32, 33, 49, 52, 53). In principle, inhibitory RNAs delivered by viral vectors are longer lasting than siRNAs, since the latter are synthesized in vitro and their abundance within the cell is thus limited by the finite amounts of double-stranded RNA administered in a single-bolus injection or via osmotic infusion pumps (10, 49). In contrast, inhibitory RNAs delivered by viral vectors are transcribed within the cell from engineered promoters and should constantly be produced as long as the vector DNA is present and the promoter is active. This feature is advantageous in instances where long-term expression of inhibitory RNAs may be warranted, such as in generating stable cell lines that are hypomorphic for a gene of interest and in treating a dominant genetic disorder, such as Huntington's disease, where chronic gene suppression may be necessary. In addition, viral vector delivery of RNAi helps avoid the activation of interferon immune responses that may occur upon exogenous siRNA application via binding of Toll-like receptors (19).

Lentiviral vectors are well suited for gene transfer because they provide long-term, heritable gene expression by integrating into host cell chromosomes (28). In addition, they can be modified to transduce numerous cell types, including postmitotic neurons and difficult-to-transduce primary neural progenitor cells (NPCs) (9, 15, 18, 23, 28, 37, 38, 42). The most well developed lentiviral vector systems are based on human immunodeficiency virus type 1 (HIV-1), and several publications have described delivery of inhibitory RNAs using HIV-1-based lentiviral vectors in vitro and in vivo (20, 22, 26, 32-34, 45, 48). Vectors derived from nonhuman pathogens, for example, feline immunodeficiency virus (FIV), pose less risk of recombining into replication-competent viruses in humans (17, 30, 43) and may have less associated stigma with the lay public than HIV-based vectors. Despite these advantages, compared to HIV-based vectors, less is known about the biology and construction of FIV-based vectors.

Here, we describe the development of an FIV-based lentiviral vector system to deliver inhibitory RNAs. These vectors will be useful in future gene therapy experiments for cancer and dominant disorders or examination of gene function in primary NPCs and other cell lines. Significant optimization was required to generate FIV vectors that retained high-level reporter gene and inhibitory RNA expression. Our results show that the position of the shRNA expression cassette within the FIV vector has a significant impact on reporter gene and shRNA expression. Finally, we further refined the vector system to codeliver primary microRNAs and reporter genes from a single transcript. Together, these studies help to better define FIV vector development in general and, more specifically, for use as an RNAi delivery vehicle.

MATERIALS AND METHODS

Plasmid construction.

Numerous DNA fragments used in this study were generated by PCR. Unless otherwise noted, all products were amplified using Pfu Turbo polymerase (Stratagene), cloned into pCR Blunt II TOPO vectors (Invitrogen), and sequence verified. The shLacZ shRNA sequences as well as the PCR-based method for producing U6 shRNA expression cassettes were previously described (12). The shLuc sequence is cognate to bases 375 to 393 of Renilla luciferase (5′-GGCCTTTCACTACTCCTAC-3′) and was kindly provided by Promega Technical Services. The 472-bp fragment of firefly luciferase used in the 3′ Lucfrag vector was cloned by PCR using Taq polymerase (Biolase) with BamHI-tagged primers sh395 (5′-CGGAATTCGGATCCTTGCCGAAAATAGGGT cgc-3′) and sh396 (5′-CGGAATTCGGATCCTGAACACCAACCACCGCATC-3′). The product was then cloned into pCR2.1 TOPO (Invitrogen). To generate 3′ U6, 3′ U6shLacZ, 3′ U6shLuc, and 3′ Lucfrag vectors, respective inserts were excised from TOPO vectors using BamHI and cloned into the FIV proviral plasmid pVET_LCGFP at the BglII site located 3′ to the green fluorescent protein (GFP) coding sequence (pVET_LCGFP is identical to pVET_LCβ [17] except that the LacZ gene is replaced by GFP). To make the 5′ U6, 5′ U6shLacZ, and 5′U6shLuc vectors, respective inserts were EcoRI digested from pCR Blunt II TOPO vectors and ligated into the MfeI site located in the FIV gag gene of pVET_LCGFP. To create the 5′ Rev response element (RRE) FIV proviral plasmids, the FIV RRE and 3′ splice acceptor (SA) along with some flanking sequence was amplified from pVETLcβ by PCR using a forward primer (5′-GACGGTGTCGAAATACAACCACAAATGG-3′) tagged with a Tth111I site and a reverse primer (5′-GCGGCCGCGAAGGCTTTCTTCTTTC-3′) tagged with a NotI site. The resulting product was cloned into pCR Blunt II TOPO, sequenced, and cloned into the Tth111I and NotI sites located in pVETLcβ to generate the FIV proviral plasmid pVET_LCntβ0.5′RRE.SA. From this, ntLacZ was removed using NheI-BamHI digestion and replaced with GFP excised from pVET_LCGFP, using NheI-BglII to create pVET_LCGFP.5′RRE SA. This series of cloning steps destroyed the BglII site located 3′ to GFP (described above) but added a unique EcoRI site 16 bp downstream. To create the FIV proviral plasmid with a 5′ RRE but without the SA, the RRE was amplified by PCR with an MfeI-tagged forward primer (5′-CAATTGAGGAGAAATGGTAGGCAATG-3′) and a NotI-tagged reverse primer (5′-GCGGCCGCTTTGATTCGAAATGGATTCATATGACA-3′). The product was then ligated into MfeI-NotI-digested pVET_LCGFP.5′RRE.SA to generate pVET_LCGFP.min5′RRE. EcoRI-digested U6, U6shLacZ, or U6shLuc fragments were then cloned into the 5′ RRE vector backbones at the MfeI or EcoRI site located upstream or downstream, respectively, of the CMV.GFP cassette. To produce the neomycin resistance (Neo^r) vector pVET_LCneo^r.5′RRE.SA, the Neo^r gene was PCR amplified from a pcDNA3.1 vector using a forward Neo^r primer containing AgeI-SpeI restriction sites and a consensus Kozak sequence (sh356, 5′-ACCGGTACTAGTCCACCATGGGATCGGCCATGGAAC-3′) and an EcoRI-SalI-tagged reverse primer (sh357, 5′-GAATTCGTCGACTCAGAAGAACTCGTCAAGAAGGC-3′). The resulting PCR product was ligated into AgeI-EcoRI-digested pVET_LCGFP.5′RRE.SA. MicroRNAs with flanking chromosomal sequences were amplified by PCR from C57BL/6 mouse genomic DNA and inserted into pVET_LCneo^r.5′RRE.SA at the EcoRI site located 3′ to the Neo^r gene. The microRNA primer sequences were as follows: mouse mir34a, 5′-AGGAGTGTGTCATACCTCGG-3′ (forward) and 5′-TTAGCCAGAAGTGCTCACAC-3′ (reverse); mouse mir21, 5′-TGACTGCAAACCATGATGC-3′ (forward) and 5′-CATTAAGCCCCAGCAAACC-3′) (reverse). The mir34a target sequence was generated by PCR using a forward primer (5′-ATCTAGAACAACCAGCTAAGACACTGCCAGAGCTCGCTTCGAGCAGACATGATAAGATAC-3′) containing an XbaI site, a mir34a target site (underlined), and 25 nucleotides that anneal to base pairs 4211 to 4235 of Psicheck2 (Promega), together with a reverse primer (5′-AAAAAACCTCCCACACCTCCC-3′) located 3′ to the MfeI site located in Psicheck2. The subsequent PCR product was cloned into XbaI-MfeI-digested Psicheck2, creating the plasmid Psicheck2-mir34target, which contains a mir34a target site in the 3′ untranslated region (UTR) of the firefly luciferase gene.

Vector production and titration.

All vectors were pseudotyped with vesicular stomatitis virus G protein envelopes using previously described transient transfection methods (17). Following a 250-fold concentration, vectors were titered using visual methods and/or quantitative PCR (QPCR), using limiting dilutions on HT1080 cells. GFP transducing units were determined at 5 days posttransduction by fluorescent microscopy using a DM RBE microscope (Leica), and images were captured with a SPOT RT camera (Diagnostic Instruments). Integrated genome copies were determined by Taqman assay and a Prism 7000 or Prism 7900 sequence detection system (Applied Biosytems). Briefly, 100- and 1,000-fold dilutions of concentrated vector preparations were applied to HT1080 cells in duplicate and genomic DNA was extracted at 3 days posttransduction (Wizard SV genomic DNA purification system; Promega). Quantitative PCR was performed in triplicate using Taqman universal PCR master mix (Applied Biosystems) and a primer/probe set that produces a 77-bp amplicon which begins 75 bp downstream of the 5′ long terminal repeat (LTR) (forward primer, 5′-AGCAGAACTCCTGCTGACCTAAA-3′; reverse primer, 5′-TCGAGTCTGCTTCACTAGAGATACTC-3′; probe, 5′-ACTGTTAGCAGCGTCTGCTACTGCTTCCCT-3′). Standard curves representing FIV genome copy numbers ranging from 10⁵ to 10⁸ copies are run in triplicate with each titration plate and used to determine the copy number of individual viral preps. The reported genomic titers are the averages of the duplicate runs from the 100- and 1,000-fold-diluted transductions.

β-Galactosidase detection using enzymatic assay and QPCR.

HT1080 cells were transduced with FIV vectors at a multiplicity of infection (MOI) of 10 in Dulbecco modified Eagle’s medium (DMEM) containing 2% fetal bovine serum (FBS) and 1% penicillin-streptomycin (P-S). Four days later, transduced cells were seeded onto poly-l-ornithine (Sigma)-coated 24-well plates at a density of 150,000 cells per well in DMEM containing 10% FBS and 1% P-S. The next day, cells were transfected with 200 ng of a CMV.LacZ expression plasmid and Lipofectamine 2000 (Invitrogen) following the manufacturer's instructions. RNA or protein was harvested 20 h later using TRIzol (Invitrogen) or Galacto-Light buffer supplied in the Galacto-Light assay kit (Applied Biosystems), respectively. To perform QPCR, 1 μg total RNA was DNase treated (DNA-free, Ambion) for 1 h at 37°C. Following DNase inactivation, 500 ng was used to synthesize cDNA, per the manufacturer's protocols but with a 1-h extension time, using Taqman reverse transcription reagents (Applied Biosystems). The other 500 ng was used in control reactions lacking reverse transcriptase. The 50-μl reverse transcription reaction mixtures were then diluted 1:4, and 9 μl was added to 20-μl QPCR reaction mixtures (Taqman PCR master mix; Applied Biosystems) using a custom primer/probe set for LacZ (Assay-By-Design; Applied Biosystems) or control human GAPDH (glyceraldehyde-3-phosphate dehydrogenase) (Hs99999905_m1; Applied Biosystems). QPCR was performed in triplicate runs per experiment using a Prism 7900 real-time PCR machine (Applied Biosystems), and LacZ gene expression was determined by the relative standard curve method. QPCR data are presented as means ± standard errors of the means (SEM) from three experiments. To determine β-galactosidase (β-gal) enzyme activity, Galacto-Light assays were performed following the manufacturer's instructions, using 1:20 total cell extract per sample. Individual assays were performed in triplicate using a MonoLight 3010 (Pharmingen) luminometer. Relative light units (RLUs) were normalized to total protein determined by the Lowry assay (Bio-Rad DC protein assay and Molecular Devices ThermoMax microplate reader), and data are presented as means ± SEM from two independent experiments. t test statistical analyses were performed using Excel 2003 software (Microsoft).

Western blot.

HT1080 cells were infected with FIV and transfected with a LacZ expression plasmid as described above. Protein extract (25 μg) was run on a 10% acrylamide gel, transferred to nitrocellulose, and probed with rabbit primary antibodies to β-galactosidase (1:5,000; Biodesign) and β-catenin (1:9,000; Abcam), followed by horseradish peroxidase-coupled goat anti-rabbit secondary antibodies (1:10,000; Jackson Immunochemicals). Blots were developed using ECL Plus reagents (Amersham Pharmacia). The blot is representative of two experiments.

Small-transcript Northern blotting. (i) shLacZ vectors.

HT1080 cells were infected as described above. Small RNA was extracted using a mirVana miRNA isolation kit (Ambion), and 900 ng was loaded on a 15% acrylamide-bisacrylamide (19:1) gel containing 8 M urea (48%, wt/vol) and 1× Tris-borate-EDTA. The decade marker (Ambion) was radiolabeled using the manufacturer's protocols and diluted 1:50, and 2.5 μl was loaded on the gel as a size reference. DNA oligonucleotides (1 and 0.1 pmol) containing shLacZ guide strand sequences were loaded as positive controls. Following a 30-min prerun, electrophoresis proceeded at 20 mA until the bromophenol blue loading dye reached the gel halfway point. RNA was then electrotransferred (Bio-Rad Transblot SD) to Hybond N⁺ nylon membranes for 45 min at 200 mA in 0.5× Tris-borate-EDTA, and then membranes were UV cross-linked (Stratalinker; Stratagene). Following overnight prehybridization, the blot was hybridized overnight in ULTRAhyb-Oligo hybridization buffer (Ambion) at 36°C with a ³²P-end-labeled (Ready-To-Go T4 polynucleotide kinase; Amersham) oligonucleotide probe that detects the active guide strand shLacZ RNA. The blot was then washed in 2× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate) and exposed to film (Kodak BioMax MS).

(ii) Mir34a vectors.

HT1080 cells were infected as described above, and RNA was TRIzol extracted 4 days later. Northern blots were run as described above except that 15 μg total RNA was loaded per sample. Radiolabeled probe was complementary to the mature mir34a sequence.

Luciferase assays. (i) U6.shLuc.

HT1080 cells were infected with the FIV.shLuc vectors and split 4 days later as described for shLacZ vectors. FIV-transduced cells were then transfected using Lipofectamine 2000 and 10 ng of Psicheck2 vector, which separately expresses Renilla and firefly luciferases. The next day, cell extracts were harvested in 100 μl passive lysis buffer (Promega). Firefly and Renilla luciferase activities were measured using the Dual-Luciferase reporter assay system (Promega) and a MonoLight 3010 (Pharmingen) luminometer following the manufacturers' instructions. Transfections and measurement of luciferase activity were performed in triplicate for each experiment. Normalized RLUs were calculated as the quotient of Renilla/firefly luciferase activities. Data are presented as means from two separate experiments.

(ii) Mir34a.

HEK293 cells were infected at ∼10 MOI with FIV.neo^r or FIV.neo^r.mir34a, and neo-resistant cells were selected using 1.2 mg/ml Geneticin (Invitrogen) in DMEM containing 10% FBS and 1% P-S. Following selection, 500,000 cells were seeded onto 24-well plates and then transfected in triplicate with 25 ng Psicheck2 or Psicheck2-mir34atarget. Luciferase activity was measured the following day as described above except that mean normalized RLUs were calculated as the quotient of firefly/Renilla luciferase activities.

Fluorescence-activated cell sorter (FACS) quantification of GFP expression in FIV-transduced cells.

HT1080 cells were infected with FIV.5′RRE.GFP or FIV.5′RRE.GFPmir21 at 0.1 MOI. Five days later, cells were trypsinized, pelleted by centrifugation at 1,000 rpm for 5 min, and resuspended in phosphate-buffered saline at a concentration of 10⁶ cells/ml. Cells were then filtered (Falcon 2350 mesh filters), stained with propidium iodide (1 μg/ml final concentration), and run on a Becton Dickinson FACScan instrument in the University of Iowa Flow Cytometry Facility. Gating and compensation were set using GFP-positive and GFP-negative cells with and without propidium iodide. The percentage of live GFP-positive cells and mean fluorescence intensity were determined by comparing infected cells to uninfected, GFP-negative samples. FIV.GFPmir21 data are reported as means from two separate FIV infections.

RESULTS

Position of U6 promoter within FIV vector genome impacts reporter gene expression.

To generate FIV vectors expressing both an shRNA and a reporter gene from the same vector, we modified the FIV shuttle plasmid pVET_LCGFP. This plasmid contains deletions in the FIV pol, vif, orf2, rev, and env genes but maintains the FIV bipartite packaging signal (Ψ), a portion of gag, the RRE, and the 3′ LTR (Fig. 1A). Deleted viral genes were replaced with a cytomegalovirus (CMV) promoter driving expression of GFP. Initially, constructs were made to test the effectiveness of an shRNA expression cassette placed between the CMV.GFP sequences and the 3′ RRE (Fig. 1B). Compared to FIV.GFP virus alone, this vector (3′ U6shLacZ) had a statistically significant 132-fold decrease in GFP expression (P < 0.002, t test), although viral genomes/ml packaged were not significantly affected (2.6-fold decrease) (Table 1). Reduction of GFP expression was not exclusive to U6, since placement of a human H1 promoter in the same location had a similar impact on transduction titer (data not shown). Thus, reduced transducing unit titer did not result from self-silencing of the FIV by the expressed hairpin during viral production.

FIG. 1.

FIV proviral constructs. (A) Wild-type FIV genome. FIV genes required for FIV virus production are indicated. The bipartite rev gene (not pictured) is located in env. The RRE (gray box) is naturally located at the 3′ end of env. A polyadenylation signal (AATAAA), required to terminate the FIV RNA genome, is located in the R region of the 3′ LTR. Ψ indicates the FIV packaging signal. (B) Complete or near-complete deletions of FIV gag, pol, vif, orf2, and env genes are incorporated in FIV vectors. All vectors contained a GFP gene transcribed by the CMV promoter. In plasmids used to generate FIV vectors, the U3 region of the 5′ LTR is replaced by a CMV promoter (not pictured) and is reconstituted in proviral DNA during reverse transcription. SD, splice donor. SA, 3′ splice acceptor moved to 5′ end in some 5′ RRE constructs. The MfeI cloning site, located in gag, is indicated. (C) Photomicrographs demonstrating that different FIV vectors expressed various GFP levels depending on the presence and location of the U6 promoter, RRE, and shLacZ hairpin with respect to the CMV.GFP expression cassette. HT1080 fibrosarcoma cells were transduced with equivalent numbers of vector genomes, and images were captured 5 days later. Bar, 50 μm.

TABLE 1.

Effect of cloning site and DNA composition on vector titer

Vector name	TU titer (no. of vector preps)a	Genome titer (no. of vector preps)b
FIV.GFP	(5.0 ± 1.0) × 107 (6)	(2.0 ± 0.5) × 109 (7)
3′ U6	(1.6 ± 1.5) × 106 (2)	(2.9 ± 1.1) × 109 (3)
5′ U6	(8.0 ± 4.0) × 107 (3)	(4.8 ± 2.7) × 109 (3)
3′ Lucfrag	(9.5 ± 3.5) × 107 (2)	(7.2 ± 2.4) × 108 (2)
3′ U6shLacZ	(3.8 ± 1.6) × 105 (4)c	(7.7 ± 6.8) × 108 (2)
5′ U6shLacZ	(3.1 ± 0.7) × 107 (13)d	(9.8 ± 1.6) × 109 (5)
5′ RRE.GFP	(3.5 ± 0.9) × 107 (4)	(4.4 ± 1.1) × 109 (4)
5′ RRE/3′ U6shLacZ	(6.4 ± 3.2) × 106 (5)e	(2.2 ± 1.0) × 109 (5)
5′ U6shLacZ/5′RRE	(2.4 ± 1.0) × 107 (5)e	(2.4 ± 1.2) × 109 (5)
5′ U6shLacZ/min5′RRE	(5.1 ± 0.9) × 107 (2)	(6.0 ± 3.3) × 108 (2)

^aMean transducing unit (TU) titer ± SEM based on GFP fluorescence in HT1080 cells at 5 days posttransduction (expressed as transducing units/ml of FIV vector stock).

^bNumber of integrated vector genomes ± SEM present in HT1080 genomic DNA at 3 days posttransduction (determined by quantitative PCR and expressed as vector genomes/ml of FIV vector stock).

^cTwo shLacZ hairpins and two human huntingtin-specific shRNAs.

^dSix shLacZ hairpins, five huntingtin-specific shRNAs, and two luciferase-specific shRNAs.

^eThree shLacZ shRNAs and two luciferase-specific shRNAs.

Next, we tested the hypothesis that the shLacZ hairpin could impose strong RNA secondary structure, which could reduce GFP expression by impairing RNA polymerase read-through to the 3′ LTR poly(A) signal. This hypothesis was based on our previous observations that some shRNAs form secondary structures strong enough to prevent polymerase processivity during cycle sequencing reactions (12). To address this, we cloned the U6 promoter alone into the same site (3′ U6), generated virus, and determined titer. Compared to FIV.GFP alone, the U6 promoter without an shRNA significantly reduced GFP transducing unit titer (P < 0.002, t test), though not as dramatically as that observed for 3′ U6shLacZ (32-fold compared to 132-fold reduction). These data support that transducing unit titers determined from reporter gene activity are reduced by the presence of a polymerase III (pol III) promoter or an shRNA expression cassette cloned 3′ of an upstream GFP expression cassette. Interestingly, the effect was on upstream reporter expression only; neither the U6shLacZ cassette nor the U6 promoter alone impacted virus production since integrated vector genome titers determined by QPCR were unaffected relative to those of FIV.GFP (Table 1).

We next tested whether close physical association of promoter elements located downstream of CMV.GFP interfered with GFP mRNA expression from integrated proviral DNA. The U6 promoter contains at least three cis-acting DNA sequences known to bind transcription factors (6, 55). There is also promoter activity associated with the FIV LTRs (41, 47). In addition, although the RRE is best defined as an RNA export element, it may also contribute promoter activity for an antisense transcript produced in FIV proviruses (3). Thus, we hypothesized that the concentration of these elements between the GFP coding region and the polyadenylation signal in the FIV 3′ LTR of integrated genomes could reduce GFP expression by recruiting transcription factors that sterically interfere with RNA polymerase II transcription of GFP mRNA. We undertook two strategies to test this. The first was to replace the U6 promoter in the 3′ U6 vector with a similarly sized fragment (472 bp) of nonpromoter stuffer DNA derived from the firefly luciferase gene (Lucfrag). In the second strategy, we moved 3′ elements (U6, RRE, or both) to the 5′ end of the vector genome, upstream of the CMV.GFP cassette. We generated two different 5′ U6shLacZ/5′RRE constructs which differed by the presence or absence of an upstream splice acceptor and the space between the U6.shLacZ cassette and the RRE (5′ U6shLacZ/5′RRE and 5′ U6shLacZ/min5′RRE) (Fig. 1). As shown in Fig. 1 and Table 1, replacement of the U6 promoter with the luciferase fragment restored GFP titers to normal values. Furthermore, GFP titers were restored to normal when the U6 promoter or U6shLacZ was moved upstream of GFP alone or in combination with the RRE (Fig. 1; Table 1). Interestingly, moving the RRE to the 5′ end of CMV.GFP had an intermediate restorative effect on GFP titers when the U6 cassette remained 3′ of GFP (17-fold improvement; P < 0.06, t test). However, the 5′ RRE/3′ U6shLacZ titers were still significantly reduced compared to those of FIV.GFP (8-fold difference; P < 0.002, t test) or FIV.5′RRE.GFP (5.4-fold difference; P < 0.009, t test). Again, virus production was unaffected, since there was no statistically significant difference between any of the virus constructs and FIV.GFP as determined by QPCR for integrated vector genomes (Table 1). Taken together, these data suggest that the combination of the U6 promoter, an shRNA, and the RRE located 3′ to the GFP reporter was responsible for the 132-fold decrease in GFP expression. Importantly, cloning U6 alone or U6shLacZ into the MfeI site located at the 5′ end near the packaging signal had no impact on either GFP expression or vector production (Fig. 1; Table 1).

FIV vectors support shRNA expression and gene silencing.

We next tested the effects of U6 promoter position within the FIV vector on shRNA expression and gene silencing. HT1080 cells were transduced with vesicular stomatitis virus G protein pseudotyped FIV vectors. At 5 days posttransduction, cells were split and transfected with CMV.β-gal plasmid. β-Galactosidase expression was measured by an enzyme assay, Western blotting, and quantitative PCR, and siRNA expression was measured by small-transcript Northern blotting. As expected, viruses lacking shLacZ had no impact on β-gal expression and expressed no mature siRNA (Fig. 2A to D). Interestingly, viruses with U6shLacZ cloned at the 3′ position supported statistically significant silencing of β-galactosidase although viral GFP expression was reduced. Specifically, compared to 3′ U6, the 3′ U6shLacZ virus silenced β-gal protein and mRNA by 78% (P < 0.009, t test) and 64% (P < 0.004, t test), respectively (Table 1; Fig. 2A to C). Abundant shLacZ guide strand siRNA expression was confirmed by Northern blotting (Fig. 2D). Importantly, the 5′ U6shLacZ vector similarly silenced β-gal protein (73% reduction in β-gal activity compared to that of 3′ U6; P < 0.02, t test) (Fig. 2) and mRNA (60% reduction; P < 0.02, t test) (Fig. 2A to C) and supported mature shLacZ siRNA production (Fig. 2D) while allowing for normal vector production and GFP expression (Fig. 1C and Table 1).

FIG. 2.

Gene silencing and shRNA expression are affected by the presence and position of the U6 promoter, RRE, and shLacZ hairpin. HT1080 cells were infected with equivalent doses of indicated FIV vectors expressing no shRNA, shLacZ, or shLuc and then transfected with plasmids expressing LacZ or luciferase genes. LacZ gene silencing was assessed by (A) a β-gal enzymatic assay, (B) Western blotting for β-gal protein, and (C) QPCR to detect LacZ mRNA. In panel B, β-catenin was used as a loading control, and protein sizes are indicated. (D) Small-transcript Northern blot showing shLacZ guide strand expression. Ten- to 30-nucleotide RNA size standards are indicated. + Cont, positive control DNA oligonucleotide at 0.1 (left) and 1 pmol (right). (E) Renilla luciferase gene silencing was determined by a luciferase enzymatic assay in cells transduced with FIV.GFP, 5′ RRE/3′ U6shLuc, or 3′ U6shLuc FIV vector.

In our initial 5′ RRE clone, the FIV 3′ splice acceptor was moved to the 5′ end along with the RRE (5′ U6shLacZ/5′RRE) (Fig. 1B). The cloning strategy brought the 3′ splice acceptor in close proximity to the 5′ splice donor located upstream of gag. If spliced, removal of a portion of the packaging signal would reduce vector production. We thus generated a second 5′ RRE construct in which the 3′ splice acceptor was deleted. In turn, this reduced the space between U6shLacZ and the 5′ RRE from 75 nucleotides to 6 nucleotides (5′ U6shLacZ/min5′RRE) (Fig. 1B). Interestingly, moving the 3′ splice acceptor to the 5′ end had no effect on vector production, as determined by vector genome QPCR titers (Table 1). However, gene silencing and shRNA expressions were reduced when the U6 promoter was brought to within 6 bp of the RRE at the 5′ end of the vector genome. Specifically, 5′ U6shLacZ/5′RRE vectors caused 59% and 53% reductions in β-gal protein and mRNA, respectively (Fig. 2A to C). In contrast, 5′ U6shLacZ/min5′RRE viruses caused only 24% and 34% reductions in protein and mRNA, respectively, which were significant differences between the two groups (P < 0.04 and P < 0.05, respectively; t test) (Fig. 2A to C). Not surprisingly, Northern blotting confirmed that 5′ U6shLacZ/min5′RRE viruses produced lower levels of mature shLacZ guide strand RNA than 5′ U6shLacZ/5′RRE vectors (Fig. 2D). Together, these data suggest that close proximity of U6 and RRE may impose steric hindrance of the U6 promoter after integration (Fig. 2D).

Interestingly, cells transduced by the 5′ RRE/3′ U6shLacZ vectors displayed very little LacZ gene silencing (22% reductions in protein and RNA compared to those of 3′ U6 control) (Fig. 2A to C). Small-transcript Northern blots showed that this was due to reduced mature shRNA production (Fig. 2D). In addition to effects on shRNA expression, GFP titers from the 5′ RRE/3′ U6shLacZ vector were lowered, though not as markedly as when the RRE was present at the 3′ end (Table 1; Fig. 1C). This surprising result was confirmed using a second set of viruses containing a hairpin directed to Renilla luciferase (shLuc) (Fig. 2E). Similar to the results obtained with the 5′ RRE/3′ U6shLacZ virus, 5′ RRE/3′ U6shLuc reduced Renilla luciferase gene expression by only 25% compared to a control vector expressing no shRNA. In contrast, 5′ U6shLuc, which had a reciprocal formulation (i.e., 5′ shRNA and 3′ RRE), caused significantly greater reduction in Renilla luciferase activity (62%; P < 0.02, t test) (Fig. 2E).

FIV vectors support efficient reporter and miRNA codelivery.

Another method for inducing RNA interference in cells is by expressing naturally occurring or synthetic primary or pre-miRNAs. Recent studies suggest that miRNAs modulate gene expression during the development of neural and other tissues (7, 14, 16, 21, 25, 27, 39, 51). Still, very little is known about their biological roles. One method to examine miRNA function is to ectopically express or overexpress them in a model for development and assess phenotypic outcomes, such as in undifferentiated NPCs. Our laboratory has previously shown that FIV efficiently transduces primary NPCs without significantly affecting cell fate following differentiation (15).

An FIV vector for coexpressing a reporter gene or selectable marker along with a portion of a primary miRNA transcript was generated based on prior feasibility studies published by Cai and colleagues (4). In their study, transfection of plasmid DNA encoding a single, heterologous luciferase/mir30 transcription unit produced luciferase and mature microRNA expression in cultured cells (4). In our vector, mouse mir21, including several hundred base pairs of flanking genomic DNA, was cloned as an artificial 3′ UTR of GFP (Fig. 3A) and the RRE was cloned 5′ to the CMV.GFP cassette. The transducing unit titer of the GFPmir21 virus (FIV.5′RRE.GFPmir21) was reduced fivefold relative to FIV.5′RRE.GFP (P < 0.01, t test) but retained normal genome titers (Table 2). These data were confirmed using FACS analysis of HT1080 cells transduced with each virus. As shown in Fig. 3B, there were 2.5 times fewer GFP-positive cells in samples transduced by the FIV virus expressing GFP.mir21 than GFP alone. In addition, cells expressing GFPmir21 were half as intense as those transduced with FIV.5′RRE.GFP alone (Fig. 3B and C). Expression of mature mir21 was confirmed by small-transcript Northern blotting (data not shown). We next generated FIV vectors expressing the neomycin resistance gene and a different miRNA, mir34a. Similar to our results using GFP viruses, FIV vectors expressing the Neo^r gene alone or together with an miRNA had normal integrated genome titers (Table 2). In addition, the Neo^r gene contributed resistance to Geneticin in HEK293 cells and NPCs, as indicated by the observation that cells transduced with FIVs expressing the Neo^r gene or neo^r.mir34a were viable after drug selection while untransduced cells or cells transduced by FIV.GFP were killed. Small-transcript Northern blotting confirmed that mir34a was enriched in unselected FIV.neo^r.mir34a-infected cells compared to that in cells transduced by FIVs containing the Neo^r gene or GFP alone (Fig. 3D).

FIG. 3.

FIV vectors support efficient reporter gene and miRNA expression from the same transcript. (A) FIV proviral constructs expressing a heterologous mRNA consisting of GFP or neomycin resistance reporter genes and primary microRNAs. In these constructs, the RRE was moved to the 5′ end of the FIV genome. (B) Representative FACS plots and (C) tabulation of FACS data from HT1080 cells transduced by indicated vectors (0.1 MOI). (D) Small-transcript Northern blot for mir34a. Cells were mock infected or transduced with FIV.neo^r.mir34, FIV.GFP-, or FIV.neo^r (MOI, 10). Equal loading was demonstrated by ethidium bromide staining of tRNA. (E) Psicheck2-mir34target plasmid contained a perfect mir34 binding site in the 3′ UTR of firefly luciferase (FF Luc). TKp, herpes simplex virus thymidine kinase promoter. A transfection control expression cassette consisting of the simian virus 40 promoter (SVp), an artificial intron, and the Renilla luciferase gene (RLuc) was present on the same plasmid. Black boxes indicate poly(A) signals. (F) Mir34-directed silencing. HEK293 cells were infected with FIV.neo^r or FIV.neo^r.mir34 vector, and stable integrants (Neo^r cells and Neo^r.mir34 cells) were selected with Geneticin. Cells were then transfected with Psicheck2 (vector) or Psicheck2-mir34target, and gene silencing was determined using a luciferase enzymatic assay.

TABLE 2.

Titers of FIV vectors coexpressing reporter genes and microRNAs

Vector name	TU titer (no. of vectors)a	Genome titer (no. of vectors)b
FIV 5′RRE.GFP	(3.5 ± 0.9) × 107 (4)	(4.1 ± 1.1) × 109 (4)
FIV.5′RRE.GFPmir21	(7.0 ± 4.0) × 106 (3)	(5.0 ± 2.0) × 109 (2)
FIV.Neo	ND (ND)	3.3 × 109 (1)
FIV.Neo.mir34a	ND (ND)	2.2 × 109 (1)
FIV.Neo.mir23b	ND (ND)	1.8 × 109 (1)
FIV.Neo.mir182	ND (ND)	6.5 × 109 (1)
FIV.Neo.mirs average	ND (ND)	(3.5 ± 1.5) × 109c

^aMean transducing unit (TU) titer ± SEM based on GFP fluorescence in HT1080 cells at 5 days posttransduction (expressed as transducing units/ml of FIV vector stock). ND, not determined.

^bNumber of integrated vector genomes ± SEM (where indicated) present in HT1080 genomic DNA at 3 days posttransduction (determined by quantitative PCR and expressed as vector genomes/ml of FIV vector stock).

^cAverage genome titer of FIV.Neo.mir34a, FIV.Neo.mir23b, and FIV.Neo.mir182.

Next, we tested whether mir34a expressed from the FIV vectors was biologically functional. For this, reporter plasmids encoding normal firefly luciferase (Psicheck2), or with a mir34a binding site placed in the 3′ UTR (Psicheck2-mir34target), were transfected into cells stably expressing FIV.neo^r.mir34a or FIV.neo^r. (Fig. 3E), Luciferase activity from the mir34 target plasmid was reduced 47% in cells stably overexpressing FIV.neo^r.mir34a compared to that in control cells expressing only the Neo^r gene (P < 0.0008, t test) (Fig. 3F). No gene silencing was observed when Psicheck2 was transfected into either cell line. These data demonstrated that the neo^r.mir34a transcript is appropriately processed to a functional mir34a after FIV transduction.

DISCUSSION

Here, we present the first extensive characterization of FIV-based lentiviral systems for inhibitory RNA delivery. We have identified an optimal vector construct for shRNA delivery which maintains normal titers, supports high level reporter gene and shRNA expression, and elicits significant gene silencing. We further refined this vector system to codeliver primary microRNAs and reporter genes from a single transcript.

The RNA genome of a lentiviral vector encoding an shRNA expression cassette could be subject to self-induced silencing during viral production, thereby reducing titer. Though several publications have described HIV-based lentiviral vectors expressing inhibitory RNAs, only two noted reduced titers as an outcome. One study reported a modest ∼2- to 3-fold drop in GFP transducing units when pol III-driven shRNAs were included in the vectors (22), and a second suggested that an ∼5- to 6-fold reduction in p24 titer resulted from silencing of the viral genomic RNA during production (1). In contrast, our FIV genome titer data support that viral production was unaffected by the U6.shRNA cassette, regardless of cloning position (Table 1).

Importantly, we found that the location of the U6 promoter, shRNA, and RRE sequences, with respect to the CMV.GFP reporter, significantly impacted GFP expression, shRNA production, and subsequent gene silencing. This has not previously been reported for HIV-based shRNA vector studies in which the RRE was always located at the 5′ end of the viral genome and shRNA expression cassettes were cloned in various locations upstream, downstream, or in place of a reporter gene.

In our initial vectors, cloning of U6.shLacZ between CMV.GFP and the 3′-located FIV RRE (3′ U6shLacZ) resulted in a 132-fold drop in GFP expression compared to that of the parental vector (FIV.GFP) but shRNA expression and gene silencing were normal. Deleting the shLacZ hairpin from 3′ U6shLacZ (3′ U6) resulted in partial restoration of GFP expression. In contrast, a similarly sized, nonpromoter sequence (Lucfrag) had no impact on GFP levels. These results demonstrated that the U6 promoter by itself was sufficient to reduce transducing unit titer when cloned 3′ to the CMV.GFP cassette, but maximal reduction required that an shRNA be present in this location as well. Our data also suggest that the RRE may contribute to GFP inhibition in the 3′ U6shLacZ vector, since moving the RRE upstream of the CMV.GFP cassette (5′ RRE/3′ U6shLacZ) partially relieved the GFP expression inhibition observed in the 3′ U6shLacZ vectors (17-fold improvement). It is possible that the U6 promoter, shRNA, and RRE inhibit GFP through DNA- and/or RNA-based mechanisms. In the integrated provirus, the pol III-responsive U6 promoter and the RRE will recruit transcription factors between the CMV.GFP cassette and the poly(A) signal located in the 3′ LTR which could sterically interfere with pol II processivity. Separately, the highly structured FIV RRE and the shLacZ hairpin sequences present on the GFP transcript could contribute a potentially inhibitory secondary structure, which may also reduce pol II processivity. Finally, the shLacZ target site present on the GFP transcript might be subject to shLacZ gene silencing.

Interestingly, there was an inverse correlation between GFP and shRNA expression in the 5′ RRE/3′ U6shLacZ vector compared to that in 3′ U6shLacZ. In particular, moving the RRE to the 5′ end of the FIV genome caused a significant improvement in GFP expression that was coincident with markedly reduced shRNA expression and gene silencing capability (Table 1; Fig. 2). This was confirmed using a second set of viruses expressing shLuc. These data suggest that there may be steric competition for proximal DNA sequences and that having the RRE at the 3′ end favors transcription of U6 over the CMV promoter. Movement of the RRE to the 5′ end may somehow relieve the U6 competitive advantage, which results in increased GFP levels and decreased shRNA expression. When both the RRE and U6shLacZ cassettes were located 5′ to the CMV.GFP reporter (5′ U6shLacZ/5 ′RRE), GFP expression, shRNA production, and gene silencing were normal. In contrast, decreasing the space between the RRE and U6shLacZ cassettes at the 5′ end of the viral genome (5′ U6shLacZ/min5′ RRE) reduced shRNA expression and gene silencing but had no effect on GFP titer. These data further support that the locations of the RRE, U6 promoter, and shRNA significantly impact transducing unit titer and shRNA expression/gene silencing.

The 5′ U6shRNA construct is our most optimal pol III-based shRNA vector because it maintains normal titers, produces high-level reporter gene and shRNA expression, and supports gene silencing. In addition, the 5′ U6shRNA vectors are easy to produce since the digested overhang from the MfeI cloning site used to create these vectors is compatible with EcoRI-digested U6.shRNAs prepared using our PCR method (12). The inclusion of cis-acting insulating elements, such as scaffold/matrix attachment regions and posttranscriptional regulatory elements (e.g., woodchuck hepatitis virus posttranscriptional regulatory element), to these second-generation FIV vectors may further improve shRNA expression and gene silencing.

The U6 promoter (and other related pol III promoters) is constitutively and ubiquitously active. These features may be disadvantages when restriction of RNAi to a particular cell or tissue type is desired. Our laboratory has previously demonstrated that shRNAs can be expressed from the pol II-based, constitutively active CMV promoter, but the distance between the transcription start site, shRNA, and polyadenylation signal was critical to achieve gene silencing (53). These data suggested that tissue-specific pol II promoters could theoretically be engineered to express shRNAs. However, unlike the CMV promoter, critical cis-acting elements and transcription start sites have not been mapped for most tissue-specific promoters, which could limit their usefulness for RNAi delivery. Furthermore, the required close proximity of a pol II polyadenylation signal and shRNA cannot be easily achieved in FIV vectors, since the FIV poly(A) is necessarily located in the R region of the 3′ LTR and placement of it at an internal location negatively affects vector production. Indeed, we found that expressing a reporter gene and a pol II-driven shRNA from two separate promoters in the same lentiviral vector could significantly reduce expression of the upstream gene (our unpublished results). In previous studies using HIV-based lentiviral vectors, two different pol II genes were expressed from the same lentivirus by incorporating internal ribosomal entry site sequences, protease cleavage sites within fusion proteins, or bidirectional expression cassettes, with some success (2, 8, 56). The first two approaches could not be used for shRNA delivery, and the last approach has not yet been developed for FIV vectors. To overcome this obstacle and retain the ability to express inhibitory RNAs from a pol II promoter, we cloned miRNA sequences in the 3′ UTR of our reporter genes. This strategy was based on a paradigm first demonstrated by Cai et al., who showed that luciferase expression and biologically active microRNAs were produced in cells transfected with plasmids expressing heterologous luciferase microRNA transcripts (4). The microRNAs were considered to be 3′ UTRs because they were cloned at the 3′ end of the reporter gene and were not translated. Consistent with the previous report by Cai et al. (4), our FIV lentiviral vectors expressed detectable, though modestly reduced, reporter genes and, importantly, supported significant gene silencing in transduced cells. These data demonstrate that FIV vectors can deliver active RNAi species from a pol II transcript. These vectors can be further modified to express any primary or artificial shRNA or microRNA of interest by using any known tissue-specific promoter. In addition, controlled RNAi could be expressed by incorporating well-defined inducible pol II promoters into this system.

In summary, this report extensively characterizes FIV vectors for RNAi delivery. Our studies identified an optimal vector for shRNA delivery and established proof of principle that FIV vectors could produce active, pol II-driven microRNAs and reporter genes. In addition, our studies revealed important data about the molecular biology and construction of FIV vectors and suggest that multiple factors may influence their ability to efficiently deliver high-level gene expression. This work is an important first step in further developing FIV vectors for RNAi delivery.