Highly Efficient Multicistronic Lentiviral Vectors with Peptide 2A Sequences

Abstract

Gene discovery and gene therapy call for advanced technologies to reliably assess gene expression; efficient coupling of gene expression to the expression of reporter genes is critical. Various noninvasive molecular imaging modalities have emerged to track biological processes in animal models. Here, we evaluate various strategies to link transgene expression with that of an (imaging) reporter gene. Using lentiviral vectors containing internal ribosomal entry sites (IRES), 2A-like peptides, or a bidirectional promoter, we compared their ability to ensure efficient coexpression of multiple reporter genes. Although the encephalomyocarditis virus (EMCV) IRES yielded functional bicistronic vectors, the expression level of the reporter downstream of IRES was consistently lower than that of the upstream transgene. Interestingly, peptide 2A constructs performed best in vitro and in vivo, providing effective noninvasive follow-up of transgene expression and having reporter gene expression levels in line with that of the single reporter constructs. The intrinsic “cleavage” property of the peptide 2A sequences allows each protein to be produced at proportional levels, opening ample possibilities for functional genomics and future gene therapeutic applications. Last, using various peptide 2A sequences, we engineered the triple reporter LV-3R (i.e., eGFP, fLuc, HSV1-sr39tk), enabling efficient multimodality readouts in vivo.

Introduction

H IV-derived lentiviral (LV) vectors provide one of the most efficient methods for stable gene delivery in vivo. LV vectors are frequently used to transfer genes into target cells to study their biological function or to permanently mark cell populations in vivo, opening possibilities for gene therapeutic applications (Biffi and Naldini, 2005; Verma and Weitzman, 2005). Because LV vectors efficiently transduce nondividing postmitotic and quiescent cells, they are ideal vehicles for gene transfer into the CNS, targeting mainly differentiated neurons but also astrocytes and endogenous neural stem and progenitor cells (Consiglio et al., 2004; Geraerts et al., 2006; Reumers et al., 2008). LV vectors have been used to achieve sustained and stable expression or knockdown of genes in the brain (Van den Haute et al., 2003; Hioki et al., 2007) and have been successfully applied to create locoregional transgenic small animal models to study neurodegeneration (Kirik and Bjorklund, 2003; Lauwers et al., 2003, 2007; Jakobsson and Lundberg, 2006).

Functional genomics of novel genes in animal models, either in the brain when studying neurodegenerative disorders or in other organs, often necessitates coexpression of multiple genes encoding different subunits of a protein complex or multiple independent genes involved in a pathophysiological pathway. Gene therapy as well may require coexpression of different genes involved in a metabolic pathway. A viral vector-mediated enzyme replacement strategy in Parkinson's disease, for example, requires coexpression of tyrosine hydroxylase, aromatic l-amino acid decarboxylase, and GTP cyclohydrolase-1 to restore striatal dopamine synthesis (Carlsson et al., 2007; Eberling et al., 2008). Moreover, therapeutic genes may be linked to a marker or resistance gene, and even a suicide gene. Consequently, stable and efficient coexpression of multiple transgenes by a single viral vector is necessary.

Gene expression in general is often difficult to detect in vivo, especially when novel genes are studied. Therefore coexpression of a reporter gene is often required to allow indirect detection. Ideally, expression of the reporter gene follows that of the gene of interest as closely as possible. Although many reporter genes exist, there is none that meets the needs of multimodality imaging and unequivocal detection of gene expression at the cellular level. Various molecular imaging modalities for noninvasive monitoring of gene expression have emerged (micro-positron emission tomography [PET], micro-magnetic resonance imaging [MRI], and bioluminescence imaging [BLI]; for a review see Massoud and Gambhir, 2007). Molecular imaging is a key technology in translational research that can be applied (MRI and PET) to human patients. In animal models, BLI technology allows noninvasive follow-up of gene expression, which obviates the need to kill laboratory animals and reduces interindividual variability. To follow up transgene expression one ideally combines multiple imaging reporter genes in a single vector construct, allowing multimodality imaging of transduced cells.

Although there have been significant advances to generate versatile multicistronic vectors expressing multiple proteins from a single transcript, several hurdles still need to be overcome (Szymczak and Vignali, 2005; Martinez-Salas et al., 2008). One of the main issues is the limited packaging capacity of viral vectors, thereby dictating limitations to the size and number of distinct proteins to be expressed. Ways have been devised to increase the capacity of viral vectors to express multiple genes from a single transcript, each having advantages and disadvantages. Here, we focus on the generation of efficient multicistronic vectors. We evaluated the most commonly used strategies to coexpress multiple transgenes from a single lentiviral vector construct. Such a multicistronic vector should result in concerted expression of the transgene and all reporters included, ideally expressing each protein at levels comparable to that obtained with a monocistronic construct. Current strategies to create multicistronic constructs include fusion proteins, internal ribosome entry sites (IRESs), viral “self-cleaving” 2A-like peptides, and bidirectional promoters (Szymczak and Vignali, 2005; Martinez-Salas et al., 2008). Although researchers have successfully expressed proteins as fusions with several reporters, fusion may affect protein folding and/or trafficking and consequently its function or activity.

The most commonly used technique to express multiple genes is by an internal ribosomal entry site, or IRES. IRESs are structural elements found predominantly at the untranslated regions (5′ UTR) of RNA viruses, mediating translation of a downstream open reading frame (ORF) via a Cap-independent mechanism (Martinez-Salas et al., 2008). The IRES of the encephalomyocarditis virus (EMCV IRES, 561 bp), for instance, has been broadly used for dual expression of multiple genes driven from a single promoter (Martinez-Salas, 1999). Nevertheless, Cap-independent translation directed from the IRES element is reported to be considerably less efficient than 5′ Cap-directed translation (Mizuguchi et al., 2000) and expression levels can vary depending on the specific cell type and tissue context (Hellen and Sarnow, 2001), the nature of the IRES, and the arrangement of cistrons (Hennecke et al., 2001).

To overcome nonstoichiometric coexpression, a 10-fold repeat of a nine-nucleotide fragment (10CI, 202 bp) found in the 5′ UTR of the Gtx homeodomain mRNA that can function independently as an IRES, was demonstrated to significantly enhance the expression level of the second cistron as compared with the EMCV IRES (Chappell et al., 2000; Wang et al., 2005).

In addition, viruses provide an alternative solution to IRES elements. Positive-strand RNA viruses encode all their proteins in a single polyprotein and produce mature proteins by means of specific proteases and self-cleaving 2A-like peptide sequences that are encoded within the multicistronic RNA (Ryan and Flint, 1997). Self-cleaving 2A-like peptides are known to yield equimolar expression of separate proteins on translation from one multicistronic mRNA (for a review, see Szymczak and Vignali, 2005). Porcine teschovirus, a member of the Picornaviridae, and Thosea asigna insect virus, a betatetravirus of the Tetraviridae family, contain ∼18- to 22-amino acid-long 2A-like peptides sharing little sequence homology, which renders them attractive candidates to mediate multiple gene expression in viral vectors. The 2A peptide contains a canonical motif at the C-terminal [consensus D-V/I-E-X-N-P-G^(2A)-↓P ^(2B)], which results in cleavage between the 2A glycine and the 2B proline. The cleavage occurs through a ribosomal “skipping” activity during translation rather than through active proteolysis, resulting in release of the upstream protein while translation of the downstream messenger RNA is continued (de Felipe and Ryan, 2004). For all peptide 2A constructs, the protein that is positioned first will be tagged with a small 2A tag at its C-terminal end, whereas the second protein remains unaffected. As the “cleavage” occurs at the end of the 2A peptide, all but one amino acid remains attached to the N-terminal end of the second protein. Nevertheless, the presence of the 2A tag sequence attached to the C-terminal part of the first protein can be advantageous because it allows distinction between endogenous and recombinant protein on Western blot (Szymczak et al., 2004). Moreover, specific 2A tag antibodies have been generated (Ryan and Drew, 1994). At the start of this study the suitability of 2A-like peptides for multiple gene expression was still controversial as conflicting data have been reported on gene expression levels and cleavage efficiency (Donnelly et al., 1997; Furler et al., 2001; Amendola et al., 2005; Fang et al., 2005; Chinnasamy et al., 2006). Ultimately, expression of two mRNAs driven by a bidirectional synthetic promoter in a single lentiviral vector (Amendola et al., 2005) also results in efficient dual gene transfer in vitro. However, the in vivo gene transfer potential may be restricted to a subset of cells or tissues because of promoter or transcriptional interference (Amendola et al., 2005; Curtin et al., 2008).

Initially, we measured two reporter genes, enhanced green fluorescent protein (eGFP) and firefly luciferase (fLuc). We constructed LV vectors using EMCV IRES, 10CI, a bidirectional promoter, and various peptide 2A-like sequences for bicistronic gene expression of eGFP and fLuc. We studied the performance of these elements with respect to achieving the most efficient and combined expression and activity of both reporter genes in cultured cells and in the mouse brain. The 2A vectors outperformed the other bicistronic vectors and mediated efficient and comparable expression of both imaging reporters irrespective of whether they were cloned upstream or downstream to the 2A sequences. Next, 2A peptide sequences were used to link the ORFs of three reporter genes, eGFP, fLuc, and herpes simplex virus thymidine kinase (HSV1-sr39tk), to allow noninvasive imaging of reporter expression combined with ganciclovir-induced cell death. In conclusion, by using 2A peptides we generated a high-performance triple reporter (LV-3R) for polycistronic gene transfer with multimodality readout.

Materials and Methods

Cell culture

293T (human embryonic kidney) and mouse GL261 glioma cells were grown as monolayers and propagated in Dulbecco's modified Eagle's medium (DMEM) with Glutamax (GIBCO BRL; supplied by Invitrogen, Merelbeke, Belgium), supplemented with 10% heat-inactivated fetal calf serum (FCS; Sigma-Aldrich, Bornem, Belgium) and gentamicin (50 μg/ml; GIBCO BRL), and were cultured at 37°C in a humidified atmosphere containing 5% CO₂. Trypsin was purchased from Sigma-Aldrich.

Cloning of the various transfer plasmids

Construction of the pCHMWS-eGFP-IRES-fLuc, pCHMWS-eGFP, and pCHMWS-fLuc transfer plasmids has been described previously (Deroose et al., 2006). The remaining transfer plasmids were constructed using these transfer plasmids as templates. The bidirectional transfer plasmid was a kind gift from L. Naldini (Vita Salute San Raffaele University, Milan, Italy) (Amendola et al., 2005).

The eGFP-fLuc fusion was constructed by digestion of the fLuc cDNA from the pCHMWS-fLuc transfer plasmid, using BamHI and XmaI, and subsequently cloned in pEGFP-C1 (Clontech, Erembodegem, Belgium) digested with BglII and XmaI, resulting in the cDNA for an eGFP-fLuc fusion protein. The cDNA encoding the fusion protein was amplified by PCR, using eGFP_s1 and fLuc_as primers (for primer information see Supplementary Table S1 at http://www.liebertonline.com/hum). The resulting fragment was digested with BamHI and XmaI and cloned into BamHI- and XmaI-digested pCHMWS-eGFP plasmid, resulting in pCHMWS-eGFP-fLuc.

A 10-fold repeat of the nine-nucleotide cellular IRES, 10CI for short, in the pRST plasmid was kindly provided by S. Gambhir (Stanford University, Stanford, CA) (Wang et al., 2005). The 10CI fragment was amplified by PCR, using 10CI_s and 10CI_as primers, and cut with XhoI and EcoRV. The resulting fragment was subsequently inserted into pCHMWS-eGFP-IRES-fLuc digested with XhoI and MscI, replacing the EMCV IRES with 10CI, generating pCHMWS-eGFP-10CI-fLuc.

Szymczak and colleagues previously reported that the cleavage efficiency of 2A-linked constructs can be influenced by the N-terminal protein and they introduced a Gly-Ser-Gly linker preceding the 2A peptide to improve the cleavage (Szymczak et al., 2004). In this work we designed 2A adapters to contain a unique restriction site at either end of the peptide sequence (XhoI and BamHI, respectively), allowing subsequent cloning of any transgene of interest, but also enlarging the peptide with four amino acids (Leu-Glu for XhoI and Gly-Ser for BamHI).

To generate the peptide 2A constructs, two primers were designed encoding the sense and antisense sequences of the T2A (Thosea asigna virus: T2A_s and T2A_as) or F2A peptide 2A (foot-and-mouth disease virus [FMDV]: F2A_s and F2A_as), respectively, 25 and 27 amino acids long. Both primers were annealed for the respective 2A constructs, resulting in adapters with overhangs that are compatible with XbaI and BamHI restriction sites. Furthermore, we introduced additional unique restriction sites at the 5′ end (XbaI-BsiWI-XhoI). The annealed adapters were inserted upstream of fLuc cDNA in pCHMWS-fLuc, using XbaI and BamHI. In a second step, eGFP was amplified by PCR using eGFP_s2 and eGFP_as2 and cloned into XbaI–XhoI, resulting in the pCHMWS-eGFP-T2A-fLuc transfer plasmid. All cloning steps were verified by DNA sequencing.

Additional peptide 2A sequences were designed on the basis of published sequences (Szymczak et al., 2004). For cloning pCHMWS-eGFP-T2A-fLuc was digested with XhoI and BamHI, replacing the T2A sequence with that of E2A (equine rhinitis A virus, 27 amino acids), P2A (porcine teschovirus-1, 26 amino acids), or Ta2A, respectively. Ta2A stands for an alternative T2A, having an altered DNA sequence without affecting the amino acid composition of the peptide. Positive colonies were isolated and verified by DNA sequencing. All peptide sequences, as well as primer information, are presented as supplementary material (see http://www.liebertonline.com/hum).

To generate pCHMWS-fLuc-T2A-eGFP, XbaI-fLuc-XhoI PCR product was generated with primers fLuc_s2 and fLuc_as2. The digested fragment was ligated into the backbone XbaI-XhoI pCHMWS-eGFP-T2A-fLuc, resulting in the intermediate pCHMWS-fLuc-T2A-fLuc. A similar cloning procedure was used to clone eGFP downstream of the T2A peptides. Briefly, PCR product BamHI-eGFP-XmaI was amplified with eGFP_s3 and eGFP_as3 primers. Digested PCR product was inserted into BamHI–XmaI pCHMWS-fLuc-T2A-fLuc, resulting in pCHMWS-fLuc-T2A-eGFP. To make pCHMWS-eGFP-P2A-fLuc-T2A-HSV1-sr39tk (LV-3R), the eGFP cDNA in the inverted construct pCHMWS-fLuc-T2A-eGFP was exchanged with the HSV1-sr39tk reporter gene, a kind gift from S. Gambhir (Stanford University) (Gambhir et al., 2000), using BamHI and MluI restriction sites. The pCHMWS-fLuc-T2A-HSV1-sr39tk construct was then cut with BclI and MluI, and the digested BclI-fLuc-T2A-HSV1-sr39tk-MluI segment was inserted into BclI-MluI pCHMWS-eGFP-P2A-fLuc.

Lentiviral vector production and transduction

Lentiviral vector production was performed as described earlier (Geraerts et al., 2005) with minor modifications. Filtered vector particles were concentrated with Vivaspin 15 columns (Vivascience, Hannover, Germany), aliquoted, and stored at −80°C. For transduction experiments cells were seeded in a 96-well plate at 20,000 cells per well. The next day, the cells were transduced with vector preparations serially diluted in DMEM supplemented with 10% FCS. After 6 hr of incubation, the vectors were washed from the cells and medium was replaced. Cells were passaged (1:10) at least four times to exclude pseudo-transduction. Stably transduced cells were seeded in quadruplet in 96-well plates at 1.5 × 10⁴ cells per cup in 200 μl of DMEM with 10% FCS and used in the various experimental settings.

Flow cytometric analysis

Transduced cells were trypsinized and fixed in a final concentration of 2% paraformaldehyde. Overall eGFP expression (mean fluorescence intensity [MFI] multiplied by the percentage of eGFP-positive cells) was measured with a FACSCalibur flow cytometer (BD Biosciences, Erembodegem, Flanders) and analyzed with the CellQuest software package provided with the instrument.

In vitro luciferase enzyme activity assay

For measuring the luciferase activity in the cells, transfected or transduced cells were washed with phosphate-buffered saline (PBS) and subsequently lysed with 40 μl of lysis buffer containing 50 mM Tris (pH 7.5), 200 mM NaCl, 0.2% Nonidet P-40 (NP-40), 1 mM phenylmethylsulfonyl fluoride (PMSF), and 10% glycerol. Five microliters of this cell lysate was assayed for luciferase activity according to the manufacturer's protocol (ONE-Glo luciferase assay system; Promega, Madison, WI) and the light produced was measured at an integration time of 5 sec with a GloMax luminometer (Promega Benelux, Leiden, The Netherlands). Data were normalized to the total protein concentration, which was determined by the bicinchoninic acid (BCA) assay (Pierce Biotechnology, Rockford, IL).

Western blot analysis

Extracts of stably transduced cells were prepared in 1% sodium dodecyl sulfate (SDS) supplemented with Complete protease inhibitors (Roche Diagnostics, Mannheim, Germany). Equal amounts of total protein, as assayed by BCA assay (Pierce Biotechnology), were separated in a 12.5% SDS–polyacrylamide gel and subsequently transferred to a polyvinylidene difluoride (PVDF) membrane (Bio-Rad, Watford, UK). After blocking in PBS with 0.1% Triton X-100 and 5% milk powder, the PVDF membrane was incubated with a primary rabbit polyclonal eGFP antibody (1:10000) (Baekelandt et al., 2003) or goat polyclonal fLuc antibody (1:1000; Promega). After several washing steps, the membrane was incubated with horseradish peroxidase (HRP)-conjugated goat anti-rabbit or rabbit anti-goat secondary antibodies, respectively (1:10000; Dako, Gostrup, Denmark). The immunoreactive proteins were visualized by chemiluminescence (ECL Plus kit; Amersham Biosciences, Uppsala, Sweden). To confirm equal loading of the protein samples, blots were reprobed with a mouse monoclonal primary antibody to α-tubulin (1:10000, T5168; Sigma-Aldrich, St. Louis, MO) followed by incubation with an HRP-conjugated goat anti-mouse secondary antibody (1:10,000; Dako). For detection and densitometry, an LAS-3000mini image reader (Fujifilm, Düsseldorf, Germany) and AIDA image analyzer version 4.19 software (Raytest Isotopenmessgeräte, Straubenhardt, Germany) were used.

Confocal microscopy

Stably transduced SHSY5Y neuroblastoma cells were propagated in DMEM supplemented with 15% FCS and 1% nonessential amino acids. For confocal microscopy, cells were seeded in an 8-chamber slide (Nunc, Roskilde, Denmark) at 45 × 10³ cells per well. Live cells were imaged 48 hr later by laser scanning confocal microscopy (ConfoCor2; Carl Zeiss, Oberkochen, Germany).

Stereotactic surgery

Adult female NMRI mice were housed under 14-hr light/10-hr dark cycle with free access to food and water. All animal experiments were approved by the bioethics committee of Katholieke Universiteit Leuven (Leuven, Flanders, Belgium). For surgery, 8- to 9-week-old mice were anesthetized and stereotactically injected with 3 μl of lentiviral vector or mouse GL261 cells, using striatal coordinates as described previously (Deroose et al., 2006). The animals were kept in individually ventilated cages until killed on day 14.

Suicide assay and tumor implantation

293T cells were stably transduced with various dilutions of LV-3R. Four days afterward, transduced cells were seeded in sextuplets and in half of the replicates ganciclovir (GCV, 5 μg/ml) (Cymevene; Roche) was added to cell culture medium. The effect of ganciclovir on cells was estimated by fluorescence microscopy and by flow cytometry 4 days after addition of the drug. To study LV-3R-mediated HSV1-sr39tk reporter expression in vivo 1 × 10⁵ or 5 × 10⁵ GL261 cells transduced with LV-3R were implanted in the brain or in the hind leg, respectively, of severe combined immune deficiency (SCID) mice (n = 3). BLI was performed on days 4, 8, and 14 after implantation. Ten days after implantation, the mice were treated intraperitoneally with ganciclovir (100 mg/kg), and the tumor was analyzed by BLI.

Radiolabeled FHBG uptake experiment

293T cells stably transduced with LV-3R or LV-T2A as a negative control were plated in sextuplicates in 24-well plates (1.5 × 10⁵ cells per well). The medium was discarded and replaced with 250 μl of fresh medium containing 5 μCi of ¹⁸F-radiolabeled 9-[4-fluoro-3-(hydroxymethyl)butyl]guanine (¹⁸F-FHBG) per well. The cells were then incubated at 37°C in a 5% CO₂ atmosphere for 60, 120, and 180 min. After incubation and removal of the medium, the cells were washed two times with 300 μl of ice-cold PBS and lysed with 250 μl of lysis buffer for 10 min. The cell lysate fractions as well as the wash fractions (medium and PBS) were collected separately from each well and the amount of cellular radioactivity was measured with an automated γ counter.

In vivo bioluminescence imaging

The mice were anesthetized and imaged in an IVIS 100 system (Xenogen, Alameda, CA) as described previously (Deroose et al., 2006). The mice were intravenously injected with d-luciferin (126 mg/kg; Xenogen) dissolved in PBS (15 mg/ml). Subsequently, they were placed in the prone position in the IVIS and frames were acquired with a field of view of 10 cm. Consecutive scans with acquisition times ranging from 1 to 60 sec were performed until the maximal signal was reached. For the quantification of bicistronic reporter gene activity in vivo, the data are reported as the photon flux (p/sec) from a 1-cm² circular region of interest (ROI) on the head. For quantification of the BLI signal from gliomas in the brain and in the hind leg, the data are reported as the photon flux from 0.72-, 1.5-, and 3-cm² circular regions of interest (ROIs), respectively.

Histology and stereological counting

To assess lentiviral transduction, the mice were killed and the brains processed for immunohistochemistry for eGFP and stereological counting (Cavalieri method) as described previously (Deroose et al., 2006).

Western blot analysis of brain extracts

Three adult female NMRI mice were injected stereotactically via the right striatum with LV-T2A (3 μl, 3210 pg of p24) as described previously. Two weeks postinjection, the mice were killed. Coronal brain sections (1.0 mm thick) were subjected to ex vivo brain bioluminescence imaging as described previously (Deroose et al., 2006). Brain slices with maximal photon flux were selected and the striatal regions were dissected. The contralateral striatum was used as a control. Extracts were made by homogenization in Tris–sucrose buffer (10 mM Tris, 1 mM EDTA, 0.25 mM sucrose, and protease inhibitors). The Western blotting procedure was performed as described previously.

Results

Construction of various bicistronic lentiviral transfer plasmids

Initially, various transfer plasmids were constructed to generate lentiviral vectors with two reporter genes, enhanced green fluorescent protein and firefly luciferase (eGFP and fLuc), both driven from the human cytomegalovirus (CMV) immediate-early promoter (see Materials and Methods). HIV-derived lentiviral vectors expressing monocistronic eGFP or fLuc, LV-eGFP and LV-fLuc, were produced as references (Fig. 1). In addition, the open reading frames (ORFs) of fLuc and eGFP were fused, generating an eGFP-fLuc fusion construct (LV-Fusion).

FIG. 1.

Schematic representation of LV constructs. Reporter genes are driven by the human cytomegalovirus immediate-early promoter (hCMVie) in most of the constructs and are followed by the woodchuck hepatitis posttranscriptional regulatory element (WPRE). The promoter is preceded by the central polypurine tract (cPPT) and the LV cassette is flanked by the 5′ long terminal repeat (LTR) and 3′ self-inactivating (SIN) LTR. (a and b) Two monocistronic constructs, expressing eGFP and fLuc, respectively, were used as controls (LV-eGFP and LV-fLuc). (c) In the LV-Fusion plasmid the eGFP and fLuc ORFs are expressed as a fusion protein. (d) In the LV-IRES plasmid both reporters are linked by an EMCV IRES sequence; (e) in LV-10CI they are linked by a 10-fold repeat of a nine-nucleotide cellular mRNA sequence with IRES activity (10CI). For LV-T2A (f) and LV-F2A (g), expression of eGFP and fLuc is coordinated by two different self-cleaving 2A sequences, T2A and F2A. (h) In the LV-BiDir construct, coordinated expression of eGFP and fLuc is driven by a bidirectional promoter, consisting of the human phosphoglycerate kinase (PGK) promoter driving eGFP expression, joined to the minimal CMV promoter (mCMV) driving fLuc expression (Amendola et al., 2005).

The various bicistronic constructs always encode eGFP in the first position and fLuc in the second position. Both ORFs are joined by the internal ribosomal entry site of the encephalomyocarditis virus (EMCV IRES) or by a 10-fold repeat of a nine-nucleotide cellular IRES (Wang et al., 2005) and are referred to as LV-IRES and LV-10CI, respectively. Furthermore, we constructed lentiviral vector constructs in which the ORFs of both reporter proteins were separated by viral peptide 2A sequences. We used the sequence of foot-and-mouth disease virus (F2A) and the sequence of Thosea asigna insect virus (T2A), resulting in LV-F2A and LV-T2A, respectively. Finally, we also included a bidirectional promoter that allows expression of two individual mRNAs at either side of the promoter (Amendola et al., 2005). The human phosphoglycerate kinase (PGK) promoter drives eGFP expression and is joined to the minimal CMV promoter driving fLuc expression in the antisense direction. This system is referred to as LV-BiDir. The exact nucleotide sequences linking the eGFP and fLuc ORFs are listed in Supplementary Table S1.

Evaluation of the various bicistronic lentiviral vector constructs in cultured cells

First, we verified reporter protein expression after transient transfection of the various transfer plasmids in 293T cells. Western blot analysis confirmed the correct size of all proteins expressed, and eGFP and fLuc reporter activity was confirmed by flow cytometry and luciferase activity assay, respectively (data not shown). Subsequently, lentiviral vectors were produced and concentrated. 293T cells were transduced with the various lentiviral vectors normalized for p24 antigen (1.2 pg of p24 per cell). The transduced cells were subcultured at least four times (1:10). Western blot revealed protein bands for most of the transduced cells (Fig. 2a). After blotting the upper half of the membrane was incubated with fLuc antibody, whereas the lower half was incubated with eGFP antibody. Transduction with the monocistronic constructs LV-eGFP and LV-fLuc resulted in a single prominent band recognized by the eGFP or fLuc antibody, respectively. Transduction with LV-Fusion resulted in expression of an eGFP-fLuc fusion protein, corresponding to the expected molecular weight (MW) and recognized by both the fLuc antibody (Fig. 2a, top) and the eGFP antibody (data not shown). However, the expression level of the fusion protein was lower than for the individual reporter proteins. LV-IRES mediated high expression of both eGFP and fLuc. Furthermore, we also observed good eGFP and fLuc expression in extracts of cells transduced with LV-F2A and LV-T2A. Whereas fLuc protein remained unaffected, eGFP carried a 2A tag of 27 or 25 amino acids, respectively, explaining the shift of 2–3 kDa in the migration pattern compared with eGFP (Fig. 2a, middle). In addition, the LV-F2A- and LV-T2A-transduced cells contained a band of higher molecular weight, recognized by both the fLuc and eGFP antibodies, corresponding to the “uncleaved” eGFP-F2A-fLuc and eGFP-T2A-fLuc proteins. The high molecular weight band is more prominent for LV-F2A and detectable only after long exposure for LV-T2A. The cleavage efficiency [ratio fLuc/(fLuc + eGFP-2A-fLuc)] was estimated by densitometry, which demonstrated that only 45% of the F2A construct was properly processed as compared with 91% of the T2A construct. Comparable data were obtained with an eGFP antibody [ratio eGFP/(eGFP + eGFP-2A-fLuc)] (data not shown). In the cells transduced with LV-10CI and LV-BiDir we observed eGFP expression by Western analysis, although the signal was less pronounced as compared with the other constructs and fLuc expression was near detection limits. Only with LV-BiDir could we detect fLuc protein, albeit at longer exposure times. The latter plausibly relates to the weaker promoter activity in the bidirectional construct as compared with the CMV promoter that is present in the other constructs.

FIG. 2.

Western blot analysis of LV-transduced cells. (a) 293T cells transduced with various LV vectors (1.2 pg of p24 per cell; vectors are indicated at the top) were lysed and 15 μg of protein was subjected to Western blot analysis. After blotting, the membrane was cut and the upper half was incubated with fLuc antibody (top), whereas the lower half was incubated with eGFP antibody (middle). The eGFP antibody also recognized the fusion proteins eGFP-fLuc and eGFP-2A-fLuc (data not shown). Equal loading was confirmed by α-tubulin staining (bottom). (b) Western blot analysis of 293T cell lysates transduced with various peptide 2A constructs (indicated at the top). Cells were transduced with the respective LV vector (indicated at the top) at 0.7 pg of p24 per cell and lysed, and subsequently 15 μg of protein was subjected to Western blot analysis. The same procedure was performed as in (a). Estimated protein sizes for the respective proteins are as follows: eGFP, ∼26 kDa; eGFP-2A, ∼29 kDa; fLuc, ∼61 kDa; eGFP-fLuc, ∼87 kDa; eGFP-2A-fLuc, ∼90 kDa; α-tubulin, ∼58.5 kDa.

To explore the subcellular localization of the eGFP reporter, stably transduced SHSY5Y cells were subjected to confocal microscopy. eGFP displayed a uniform distribution throughout the cells for most of the vector constructs, localizing equally well to the nucleus and to the cytoplasm. However, its localization was restricted primarily to the cytoplasm for LV-Fusion and LV-F2A constructs (Fig. 3b–d). Similar data were obtained in 293T and Neuro2A cells (data not shown). The nuclear exclusion observed for both LV-Fusion and LV-F2A can be explained by the high molecular weight fusion protein, which cannot pass the nuclear membrane by passive diffusion (87 and 90 kDa for eGFP-fLuc and eGFP-F2A-fLuc, respectively), as demonstrated in Western blot analysis (see Fig. 2).

FIG. 3.

Analysis of subcellular localization of eGFP fluorescence in transduced SHSY5Y cells. Shown are confocal microscopy images of eGFP fluorescence in stably transduced SHSY5Y cells. Cells were transduced with (a) LV-eGFP, (b) LV-Fusion, (c) LV-IRES, (d) LV-F2A, (e) LV-fLuc-T2A-eGFP, (f) LV-P2A, (g) LV-E2A, (h) LV-Ta2A, (i) LV-T2A, (j) LV-10CI, and (k) LV-BiDir. Cells were transduced with 28,000 pg of p24 and subcultured at least four times (1:10). Cells were plated into 8-well chamber slides at 20,000 cells per well and visualized by confocal live imaging. eGFP fluorescence is evidenced in the cytoplasm and the nucleus of cells transduced with LV-eGFP, LV-IRES, LV-10CI, and LV-T2A. Nuclear exclusion of eGFP fluorescence occurs in cells transduced with LV-Fusion and LV-F2A. All other peptide 2A-like LV vectors (LV-P2A, LV-E2A, and LV-Ta2A) displayed homogeneous eGFP fluorescence throughout the cell, as did the reverse construct LV-fLuc-T2A-eGFP. Similar data were obtained with 293T and Neuro2A cells (data not shown). Color images available online at www.liebertonline.com/hum.

Evaluation of reporter gene activity

To investigate the efficiency of expression of both reporters by the various bicistronic LV vectors, serial dilutions of either monocistronic (eGFP, fLuc, and eGFP-fLuc fusion) or bicistronic LV vectors were used to transduce 293T cells, after p24 normalization. Serial dilution of the vectors allowed us to compare cells transduced at low and high multiplicities of infection (ranging from 5–20 to 80–90% transduced cells). At a low multiplicity of infection the majority of cells contained only one integrated proviral copy. Transduced cells were passaged at least three times (1:10) to minimize pseudo-transduction. eGFP and fLuc activities were measured by fluorescence-activated cell sorting (FACS) and luciferase activity assay, respectively. Because the eGFP ORF is positioned first in all constructs, one would expect only minor effects on eGFP activity. The percentage of eGFP-positive cells was comparable for all vector constructs (data not shown), whereas the fluorescence intensity differed largely. As demonstrated in Fig. 4a, the overall eGFP fluorescence (percent eGFP-positive cells × mean fluorescence intensity [MFI]) decreased proportionally with the vector dose and was comparable to LV-eGFP both for LV-IRES and LV-T2A (1.2- and 1.1-fold difference, respectively), thereby outperforming LV-Fusion, LV-F2A, LV-10CI, and LV-BiDir. The 35-fold difference in overall eGFP fluorescence for LV-Fusion can be explained by the fact that GFP fluorescence may be affected in a fusion protein. As demonstrated earlier, LV-F2A-transduced cells also produce a significant amount of fusion protein (cf. Figs. 2a and 3d) accounting for the observed decline in eGFP fluorescence. Surprisingly, the eGFP expression levels were significantly lower for LV-10CI (10-fold compared with LV-eGFP) and LV-BiDir (50-fold compared with LV-eGFP). The latter can be attributed to its different promoter.

FIG. 4.

Cellular reporter gene activities in transduced 293T cells. 293T cells were transduced with four 3-fold dilutions of each LV vector (4500 pg of p24 for the first dilution) normalized to p24. (a) Plot of overall eGFP fluorescence (equivalent to transduction efficiency [TE] × mean fluorescence intensity [MFI]), representing the mean of triplicate samples. (b) Plot of fLuc activity, representing the mean of triplicate samples and normalized for protein content (RLU/μg protein). (c) Comparative analysis of eGFP expression mediated by the various LV vectors. Percentage of transduction and MFI of eGFP-positive cells are indicated. Error bars represent the SE between triplicate samples.

In parallel, fLuc activity was evaluated for all these cell lines (Fig. 4b). As expected, more substantial differences were detected. LV-T2A demonstrated the best luciferase activity, and was comparable to LV-fLuc control (1.7-fold better than LV-fLuc), followed by LV-IRES and LV-F2A (1.7- and 1.9-fold lower, respectively), at high and at lower vector dose. This result is in accordance with previously published work describing considerable attenuation of the gene downstream of an IRES element (Mizuguchi et al., 2000; Deroose et al., 2006). The luciferase reporter activity from LV-Fusion and LV-10CI was 18- and 175-fold lower, respectively, as compared with the LV-fLuc control.

To monitor the expression of a therapeutic gene using a bicistronic vector, it is essential to demonstrate coordinated, high-level expression of both transgenes in the majority of target cells. In our study the second reporter (fLuc) reports on the activity of the therapeutic gene represented here by eGFP. An ideal construct would promote efficient coexpression of eGFP and fLuc, ultimately reaching the level of the individual monocistronic constructs. Both LV-IRES and LV-T2A demonstrated fLuc and eGFP activities that were comparable to those of individual LV-eGFP and LV-fLuc constructs (Fig. 4a and b). Although eGFP fluorescence was comparable for both vector constructs, luciferase activity was at least twice as high for LV-T2A.

In addition to 293T cells, we evaluated reporter gene activities for the most promising bicistronic vectors (LV-IRES and LV-T2A) in SHSY5Y and Neuro2A cells and compared them with control vectors (LV-eGFP, LV-fLuc, and LV-Fusion). Although the overall expression levels were significantly lower in these cell lines compared with 293T, similar results were obtained (data not shown). Again, LV-IRES was outperformed by LV-T2A, having fLuc activity comparable to that of the LV-fLuc control vector.

Construction and evaluation of additional 2A-like peptide sequences

Taken together, our data demonstrate that both EMCV IRES and 2A-like peptides are suitable cis-acting elements to orchestrate concerted reporter protein expression to sufficient and detectable levels in an LV vector context. However, for several gene therapeutic applications the introduction of multiple cDNAs of interest is necessary. Moreover, multimodality imaging uses different imaging techniques, and hence multiple reporter genes may be enclosed in a single vector construct.

Because of the data obtained with LV-T2A, we engaged in testing other self-cleaving 2A-like peptides. Three additional 2A-like sequences were cloned to link the eGFP and fLuc ORFs in the transfer plasmid, resulting in lentiviral constructs encoding eGFP-E2A-fLuc (LV-E2A), eGFP-P2A-fLuc (LV-P2A), and eGFP-Ta2A-fLuc (LV-Ta2A) (see Supplementary Table S1). All 2A-like peptide sequences originate from picornaviruses: E2A from equine rhinitis A virus, P2A from porcine teschovirus-1 (Lorens et al., 2004; Szymczak et al., 2004). Ta2A is an alternative T2A sequence (originally from Thosea asigna virus), the coding sequence of which was modified at the DNA level without altering the amino acid composition of the peptide itself (see supplementary information at http://www.liebertonline.com/hum). It has been reported earlier that the peptide sequence following and/or preceding the 2A-like peptides may affect the peptide 2A cleavage efficiency (Szymczak et al., 2004). Therefore, we also constructed an LV transfer plasmid with the fLuc and the eGFP ORF in reversed order, resulting in LV-fLuc-T2A-eGFP.

Again the various transfer plasmids were evaluated by transient transfection in 293T cells and reporter gene activity was confirmed by flow cytometry and luciferase assay (data not shown). Subsequently, LV vectors were produced, normalized for p24 antigen (0.7 pg of p24 per cell), and stably transduced in 293T cells. The self-cleaving capacity of all 2A constructs was evaluated by Western blot analysis demonstrating coexpression of eGFP and fLuc proteins, although some minor uncleaved 2A fusion protein could be detected at longer exposure times for LV-E2A, LV-P2A, LV-Ta2A, and LV-fLuc-T2A-eGFP with either eGFP or fLuc antibodies (Fig. 2b). In particular, LV-P2A repeatedly performed well with only 3.5% of the total protein present as eGFP-P2A-fLuc fusion. Furthermore, reversing the fLuc-eGFP order in the lentiviral construct resulted in more efficient skipping of the T2A peptide, with only minute amounts of the fusion protein being detected on Western blot. Consequently, cells transduced with LV-E2A, LV-P2A, LV-Ta2A, and LV-fLuc-T2A-eGFP also displayed uniform eGFP fluorescence throughout the cells (Fig. 3, bottom).

In addition, eGFP and fLuc activities were evaluated for all three 2A-like peptide constructs and compared with LV-T2A showing comparable reporter activities (Fig. 5a and b), demonstrating the usefulness of all these peptides to orchestrate the expression of multiple transgenes from a single LV construct.

FIG. 5.

Reporter gene activities with additional 2A peptide sequences. 293T cells were transduced with four 3-fold dilutions of each LV construct (28,000 pg of p24 for the first dilution), normalized to p24. (a) Plot of overall eGFP fluorescence, representing the mean of triplicate samples. (b) Plot of fLuc activity, representing the mean of triplicate samples and normalized for protein content (RLU/μg protein).

Bicistronic reporter gene activity in vivo

Noninvasive imaging of reporters to monitor therapeutic gene expression will be an enormous asset in translating preclinical gene transfer into gene therapy in human patients. We and others demonstrated previously that injection of LV vector into adult mouse brain results in stable transduction of neurons and astrocytes for up to 1 year (Baekelandt et al., 2002; Scherr et al., 2002). Furthermore, we proved the feasibility of bioluminescence imaging (BLI) to monitor LV vector-mediated gene transfer in the murine brain (Deroose et al., 2006). Here we applied BLI to report on eGFP expression in the mouse brain and compared the various bicistronic constructs.

Consequently, the individual bicistronic LV vectors were evaluated after locoregional injection in the mouse striatum. Three NMRI mice were stereotactically injected via the right striatum with 2500 pg of p24 of LV-eGFP, LV-fLuc, LV-Fusion, LV-IRES, or LV-T2A. Because the LV-F2A construct displayed only low cleavage efficiency in cell culture experiments, it was not included for in vivo evaluation. Likewise, we excluded LV-10CI and LV-BiDir because of their low fLuc activity. Two weeks postinjection the mice were imaged noninvasively for in vivo fLuc activity with the IVIS system. As presented in Fig. 6a, the brightest bioluminescent signal was detected projecting from the right striatum of mice injected with LV-fLuc or LV-T2A. The average photon flux for LV-T2A-injected animals was comparable to that from LV-fLuc-treated mice, whereas animals injected with LV-IRES or LV-Fusion displayed a 10- to 20-fold lower photon flux, respectively (Fig. 6b).

FIG. 6.

In vivo evaluation of reporter gene expression after transduction of mouse striatum. (a) Top: Representative graphic images of in vivo fLuc activity via noninvasive bioluminescence imaging of a representative animal from each group (n = 3–6) are shown. Bottom: The corresponding light microscopy image of eGFP IHC is presented at low magnification (×1.6). (b) Bar diagram (log scale) of in vivo fLuc activity (photon flux). The fLuc activity for LV-eGFP equals background. (c) Bar diagram of the eGFP-transduced volume in the mouse striatum (Cavalieri). Error bars represent the SD of the data of three animals, except for LV-eGFP (n = 4), LV-Fusion (n = 6), and LV-T2A (n = 5). Color images available online at www.liebertonline.com/hum.

After BLI quantification, the mice were killed and the brains were processed for immunohistochemistry (IHC). As anticipated, eGFP expression was observed exclusively in the right striatum and absent at the contralateral side, except for LV-fLuc, where eGFP was absent (Fig. 6c). Subsequently, the eGFP-transduced volume was determined by stereological counting (Cavalieri method), yielding an eGFP-transduced volume for LV-IRES and LV-T2A that was comparable to that of LV-eGFP.

To confirm that transduction of the mouse brain with LV-T2A resulted in the expression of two separate reporter proteins, we reinjected LV-T2A (3 μl, 3200 pg of p24) into the right striatum of NMRI mice. Two weeks postinjection, fLuc activity was confirmed by BLI. Immediately after imaging, the brain was removed, sliced, and imaged again under a charge-coupled device (CCD) camera. Subsequently, both the right and left striatum were dissected from the slice projecting the highest photon flux, with the left striatum serving as a negative control. Protein extracts were prepared and analyzed by Western blot (Fig. 7). A prominent band corresponding to eGFP-2A (∼29 kDa) and fLuc was recognized with eGFP and fLuc antibodies, respectively, whereas no higher molecular weight band corresponding to unprocessed eGFP-T2A-fLuc fusion protein was detected. In addition, we also evaluated the reversed LV-fLuc-T2A-eGFP construct in the mouse brain and compared its performance with that of the LV-T2A vector. Both the in vivo BLI signal and eGFP-transduced volume were comparable for the two constructs (data not shown). In conclusion, LV-T2A transduction of the mouse brain results in high transgene expression and efficient cleavage of peptide 2A-like sequences, resulting in separated eGFP and fLuc reporter proteins.

FIG. 7.

Functionality of peptide 2A sequence in LV-T2A injected into the mouse striatum. (a) Western blot of protein extracts originating from transduced (LV-T2A, 3200 pg of p24) and nontransduced contralateral striatum. Membranes were probed with eGFP antibody (left) and fLuc antibody (right). (b) Loading was controlled by reprobing the membranes with a β-tubulin antibody.

Construction of a triple reporter vector with 2A-like peptides

After demonstrating that several 2A-like peptides allow efficient expression of individual reporter genes, we set out to develop a lentiviral vector encoding three separate reporter genes. We constructed a viral vector using the best 2A-like peptides from our study to express eGFP, fLuc, and a mutated form of the herpes simplex virus thymidine kinase (HSV1-sr39TK) (Gambhir et al., 2000) in a concerted fashion. The resulting triple reporter construct (LV-3R) allows, in addition to fluorescence and bioluminescence imaging, also suicide therapy and/or PET imaging.

The construct was first validated in cultured 293T cells, showing proper expression of eGFP and fLuc as demonstrated by fluorescence microscopy and luciferase activity, respectively (Fig. 8a and b). Similar results were obtained in GL261 glioma cells (data not shown). To demonstrate the enzymatic activity of HSV1-sr39TK, 293T cells stably transduced with various dilutions of LV-3R were tested for sensitivity to ganciclovir (Yaghoubi et al., 2005). Four days posttreatment ganciclovir (GCV)-induced cytotoxicity was quantified by FACS analysis. A substantial loss of eGFP (and HSV1-sr39TK)-positive cells was evidenced (Fig. 8c). Because HSV1-sr39tk is also used in PET imaging, we performed a cell uptake assay with ¹⁸F-FHBG. We incubated LV-3R cells and LV-T2A control cells with ¹⁸F-FHBG probe and measured the kinetics of radioactivity uptake in the cell fractions as a percentage of the total dose. The ¹⁸F-FHBG-specific signal accumulated over time in LV-3R-transduced 293T cells as compared with LV-T2A control cells, indicating exclusive uptake by HSV1-sr39tk-positive cells. Surprisingly, no accumulation of radioactive signal was observed in the labeled GL261 glioma cells (data not shown).

FIG. 8.

LV-3R, a multimodality reporter based on 2A peptides. 293T cells were transduced with various dilutions of LV-3R (eGFP-T2A-fLuc-P2A-HSV1-sr39tk) and reporter activities were analyzed by (a) fluorescence microscopy or (b) luciferase assay. Stably transduced LV-3R cells were treated with or without ganciclovir (5 μg/ml) for 4 days and analyzed by fluorescence microscopy (b) or by flow cytometry (c) to assess HSV1-sr39tk activity and drug-induced cell toxicity. (d) Noninvasive detection of HSV1-sr39TK expression in cells. Radiolabeled ¹⁸F-FHBG uptake was monitored for 60, 120, and 180 min in 293T cells stably transduced with LV-3R or LV-T2A. Radioactivity retention in the cell fraction is shown as a percentage of the total dose. Values represent means + SEM, calculated by nonparametric t test (n = 6). Color images available online at www.liebertonline.com/hum.

Evaluation of the triple reporter lentiviral vector in vivo

For in vivo evaluation of the triple reporter construct, we opted for the established subcutaneous and intracranial GL261 experimental tumor models (Zagzag et al., 2003; Newcomb et al., 2004). GL261 glioma cells labeled with LV-3R were stereotactically implanted in the brain (striatum) of NMRI mice, as well as in the hind leg of SCID mice. Tumor growth was monitored by BLI on days 4, 8, and 14 after implantation. GCV treatment was started 10 days after implantation. The BLI signal emerging from tumors in the brain and the hind leg increased over time, reflecting tumor growth (Fig. 9a and d). A short GCV treatment (4 days) reduced this increase in BLI signal both in the brain and the hind leg (Fig. 9b and e). The intracranial tumors were stained for eGFP and tumor volumes were estimated by Cavalieri counting (Fig. 9c). Striatal and hind leg tumors were modestly to severely affected by the GCV treatment (Fig. 9f), indicating that cells expressing the HSV1-sr39tk reporter efficiently phosphorylate ganciclovir into its toxic metabolite, resulting in efficient cell killing.

FIG. 9.

Ganciclovir treatment reduces growth of LV-3R-transduced tumor cells. (a and d) Representative images of the BLI signal originating from murine GL261 cells stably transduced with LV-3R and implanted intracranially (a) or injected into the hind leg (d). Mice were either treated with GCV (n = 3) or not treated (n = 3) and treatment was started 10 days after implantation. BLI was performed on day 4 (data not shown) and on days 8 and 14 postimplantation. (b and e) To monitor relative tumor growth, the fold increase in BLI signal relative to the previous measurement was calculated. In GCV-treated animals the relative increase in the BLI signal was lower than in nontreated animals both in the brain (b) and in the hind leg (e). (c and f) After killing the animals, intracranial tumors of treated and nontreated animals were stained for eGFP (c, left) and tumor volume (mm³) was estimated by Cavalieri (c, right). Tumors in the hind limb were dissected (f, left) from GCV-treated and untreated mice and weighed (f, right). Color images available online at www.liebertonline.com/hum.

Discussion

The introduction of multiple genes into a single cell is required for a wide range of applications, for example, gene correction, selection, marking or (conditional) ablation of genetically altered cells, the reconstruction of multisubunit receptors, and reporting on transgene expression. Likewise, functional genomics requires efficient coupling to a variety of reporters to gain insights concerning the function of the gene of interest. The emerging technology to detect and quantify transgene expression in animal models is noninvasive, small-animal imaging using BLI, micro-PET, or micro-MRI reporter genes, which requires multicistronic constructs.

A straightforward and efficient way to introduce transgenes and reporter genes into cells of interest is to use stably integrating lentiviral vectors. Several approaches to proportionally coexpress multiple genes by a single viral vector have previously been pursued. Here, we describe an in-depth evaluation of various gene-coupling strategies, comparing their usefulness for multicistronic vector design in the lentiviral vector context. Therefore various lentiviral vectors were constructed harboring the ORFs for eGFP and fLuc driven by the human CMV ubiquitous promoter or by a chimeric bidirectional promoter (hPGK-minCMV). The separately expressed reporters and the fusion of both were used as control and compared with an EMCV IRES (LV-IRES), a 10-fold repeat of a nine-nucleotide cellular IRES sequence with IRES activity (LV-10CI) (Chappell et al., 2000), and two 2A-like peptide sequences, derived from FMDV (LV-F2A) and from Thosea asigna insect virus (LV-T2A).

Initial characterization of bicistronic LV vectors in cells showed persistent, high expression of both eGFP and fLuc reporters for LV-IRES and LV-T2A. However, fLuc activity, originating from the second cistron, was considerably lower for LV-IRES (approximately 59%) compared with LV-fLuc (Fig. 4b). It is known that IRES-dependent expression is significantly lower than Cap-dependent expression (Mizuguchi et al., 2000), explaining the apparent discrepancy. In addition, IRES-dependent translation is highly variable among different cell lines. Still, the EMCV IRES provides the second best choice for dual gene expression from a single transcript.

The use of T2A sequences, however, resulted in high eGFP fluorescence and high fLuc enzymatic activity, both comparable to the expression levels of the respective monocistronic eGFP and fLuc vectors. When these two reporters were linked by an F2A peptide or expressed as a fusion, the unprocessed eGFP-fLuc protein was excluded from the nucleus. In addition, we observed low eGFP fluorescence and luciferase activity for LV-Fusion, LV-10CI, and LV-BiDir. Although Amendola and colleagues validated lentiviral vectors expressing two transgenes in concert from a chimeric bidirectional promoter in a variety of cells (Amendola et al., 2005), in our hands LV-BiDir exhibited only low levels of eGFP fluorescence and fLuc activity in 293T cells or neuronal cells such as SHS5Y5 or Neuro2a (data not shown), but still above background and in a concerted fashion. Low expression levels are likely due to the nature of the promoters used in BiDir. The hCMV promoter used in all other constructs outperformed the hPGK promoter of the bidirectional system. Likewise, fLuc activity driven from the minimal CMV promoter was weak compared with the other vectors. Nevertheless, this construct is suited for applications in which low expression levels are desirable, as suggested in the original publication (Amendola et al., 2005). To enhance reporter expression downstream of the IRES element, we attempted to use LV-10CI, reported to possess IRES activity and to lead to 12-fold greater reporter activity when compared with the EMCV IRES (Wang et al., 2005). In the present work, LV-10CI mediated low eGFP and fLuc expression levels in various cell types, barely above the background signal. One possible explanation is that the 10CI sequence affects overall transcription because of competition between Cap-dependent and IRES-mediated translation for a translation initiation factor, as has been suggested earlier (Svitkin et al., 2005). As has been shown for the EMCV IRES, the gene context may also influence IRES activity (Hennecke et al., 2001).

Taken together, on transduction of 293T, SHSY5Y, and Neuro2A cells (Fig. 4 and data not shown) LV vectors carrying an EMCV IRES or T2A yielded good expression levels of eGFP and fLuc on serial dilution of the vector, with LV-T2A outperforming LV-IRES by more efficient translation of the second cistron.

The excellent performance of the T2A peptide sequence, the controversy about peptide 2A sequences (Amendola et al., 2005), and the need to use heterologous sequences in multicistronic lentiviral vector design to prevent reverse transcriptase-mediated recombination during vector production prompted us to test additional 2A-like peptide sequences. Next to the T2A and F2A sequences, two alternative sequences were evaluated, E2A and P2A, as well as an alternative T2A sequence (Ta2A) that encodes the same peptide using alternative codons. Because E2A, P2A, and Ta2A sequences performed equally well, four different 2A-like peptide sequences can now be used for multicistronic lentiviral vector design. In addition, we switched the position of the eGFP and fLuc ORF. LV fLuc-T2A-eGFP performed as well as the original LV-T2A construct, demonstrating that the position of the respective reporter gene in our setting does not affect its performance and/or the peptide 2A-mediated cleavage. In answer to the existing controversy in the literature about the usefulness of 2A-like peptides for multicistronic vector construction, we demonstrate that 2A-like peptides are the best means for multicistronic vector construction, warning that a given sequence ought to be checked for “cleavage” efficiency in a certain gene context. Even limiting amounts of uncleaved product may influence the subcellular localization of the protein and consequently may complicate interpretation of the results (as indicated in Fig. 3). In addition, for all peptide 2A constructs the protein in the first position is tagged with a 2A tag at its C-terminal end (see supplementary data at http://www.liebertonline.com/hum), which may affect activity, stability, and immunogenicity of the corresponding protein products. Most importantly, these results demonstrate the ability to use the expression of a reporter gene as a proportional measure to score the expression of another transgene of interest, which can be a disease-related or therapeutic gene. Next, we extended our comparative study to the mouse brain, for a side-by-side comparison of the LV vector that performed best in the cellular experiments. The promising results for both LV-IRES and LV-T2A were corroborated in vivo by high eGFP and fLuc expression levels. Although IRES-mediated fLuc expression is also less efficient in vivo than with T2A, the expression levels are high enough to allow in vivo BLI as shown in this study and in previous work (Deroose et al., 2006; Reumers et al., 2008). After transduction of the mouse brain with LV-Fusion an in vivo BLI signal was detected, but eGFP expression in the brain was low. When comparing in vitro and in vivo data, we should take into account that by estimating in vivo eGFP expression via the eGFP-transduced volume (Cavalieri method), no complete integration of eGFP fluorescence per cell and number of eGFP-positive cells is made, contrary to the in vivo fLuc activity measurements.

Ultimately, we constructed a reporter construct harboring three different reporter genes: eGFP, fLuc, and HSV1-sr39TK (LV-3R). The eGFP protein allows sensitive immunohistochemical detection and eGFP fluorescence; fLuc allows noninvasive bioluminescence imaging; and HSV1-sr39TK performs a dual role, allowing noninvasive imaging with specific PET tracers, and on the other hand allowing specific ablation of transduced cells on ganciclovir treatment, which adds safety to the current gene therapeutic vectors. Linking all three reporter genes with distinct peptide 2A sequences resulted in efficient ex vivo labeling of 293T and G261 glioma cells that in turn could be monitored in vivo by BLI imaging. Depending on the study, reporter genes can easily be exchanged to accommodate those reporter genes that are best suited.

Taken together, lentiviral vector-mediated bicistronic eGFP and fLuc expression in the murine brain can be achieved with an EMCV IRES sequence or a 2A-like linker peptide. Nevertheless, T2A is our method of choice because of its higher reporter gene expression levels in vitro and even more prominently in vivo, and because of its small sequence size. In conclusion, these data suggest that multicistronic LV vectors using peptide 2A-like sequences can be applied not only for in vitro studies to orchestrate the expression of two or more genes, but also to allow follow-up of gene delivery in vivo to address specific biological questions. In addition, these techniques will allow construction of safer vectors for gene therapeutic applications.

Footnotes

Acknowledgments

The authors thank Frea Coun, Martine Michiels, Sylvie Deswaef, and Marly Balcer for excellent technical assistance. A.I. is a postdoctoral fellow supported by the Molecular Small Animal Imaging Centre at the KU Leuven (MoSAIC); G.V.V. is funded by the European FP6 project StrokeMAP.VR; and C.V.d.P., J.T., and S.V. are funded by a grant from the Institute for the Promotion of Innovation through Science and Technology in Flanders (IWT-Vlaanderen). C.M.D. is a postdoctoral fellow of the Clinical Research Fund of the University Hospitals Leuven. R.G. is a postdoctoral fellow of the Flemish Fund for Scientific Research (FWO Vlaanderen). Research was funded by MoSAIC, SBO grants (IWT-30238 and 60838) of the IWT, an IDO grant (IDO/02/012) from the KU Leuven, FWO grant G.0406.06, IAP grant NIMI, and the EC grants DIMI (LSHB-CT-2005-512146) and Strokemap (LSHC-CT-2006-037186).

Author Disclosure Statement

No competing financial interests exist.

References

Amendola

, Venneri

M.A.

, Biffi

, Vigna

, Naldini

2005. Coordinate dual-gene transgenesis by lentiviral vectors carrying synthetic bidirectional promoters. Nat. Biotechnol., 23:108–116.

Baekelandt

, Claeys

, Eggermont

, Lauwers

, De Strooper

, Nuttin

, Debyser

2002. Characterization of lentiviral vector-mediated gene transfer in adult mouse brain. Hum. Gene Ther., 13:841–853.

Baekelandt

, Eggermont

, Michiels

, Nuttin

, Debyser

2003. Optimized lentiviral vector production and purification procedure prevents immune response after transduction of mouse brain. Gene Ther., 10:1933–1940.

Biffi

, Naldini

2005. Gene therapy of storage disorders by retroviral and lentiviral vectors. Hum. Gene Ther., 16:1133–1142.

Carlsson

, Bjorklund

, Kirik

2007. Restoration of the striatal dopamine synthesis for Parkinson's disease: Viral vector-mediated enzyme replacement strategy. Curr. Gene Ther., 7:109–120.

Chappell

S.A.

, Edelman

G.M.

, Mauro

V.P.

2000. A 9-nt segment of a cellular mRNA can function as an internal ribosome entry site (IRES) and when present in linked multiple copies greatly enhances IRES activity. Proc. Natl. Acad. Sci. U.S.A., 97:1536–1541.

Chinnasamy

, Milsom

, Shaffer

, Neuenfeldt

, Shaaban

, Margison

, Fairbairn

, Chinnasamy

2006. Multicistronic lentiviral vectors containing the FMDV 2A cleavage factor demonstrate robust expression of encoded genes at limiting MOI. Virol. J., 3:14.

Consiglio

, Gritti

, Dolcetta

, Follenzi

, Bordignon

, Gage

F.H.

, Vescovi

A.L.

, Naldini

2004. Robust in vivo gene transfer into adult mammalian neural stem cells by lentiviral vectors. Proc. Natl. Acad. Sci. U.S.A., 101:14835–14840.

Curtin

J.A.

, Dane

A.P.

, Swanson

, Alexander

I.E.

, Ginn

S.L.

2008. Bidirectional promoter interference between two widely used internal heterologous promoters in a late-generation lentiviral construct. Gene Ther., 15:384–390.

10.

de Felipe

, Ryan

M.D.

2004. Targeting of proteins derived from self-processing polyproteins containing multiple signal sequences. Traffic, 5:616–626.

11.

Deroose

C.M.

, Reumers

, Gijsbers

, Bormans

, Debyser

, Mortelmans

, Baekelandt

2006. Noninvasive monitoring of long-term lentiviral vector-mediated gene expression in rodent brain with bioluminescence imaging. Mol. Ther., 14:423–431.

12.

Donnelly

M.L.

, Gani

, Flint

, Monaghan

, Ryan

M.D.

1997. The cleavage activities of aphthovirus and cardiovirus 2A proteins. J. Gen. Virol., 78:13–21.

13.

Eberling

J.L.

, Jagust

W.J.

, Christine

C.W.

, Starr

, Larson

, Bankiewicz

K.S.

, Aminoff

M.J.

2008. Results from a phase I safety trial of hAADC gene therapy for Parkinson disease. Neurology, 70:1980–1983.

14.

Fang

, Qian

J.J.

, Yi

, Harding

T.C.

, Tu

G.H.

, Vanroey

, Jooss

2005. Stable antibody expression at therapeutic levels using the 2A peptide. Nat. Biotechnol., 23:584–590.

15.

Furler

, Paterna

J.C.

, Weibel

, Bueler

2001. Recombinant AAV vectors containing the foot and mouth disease virus 2A sequence confer efficient bicistronic gene expression in cultured cells and rat substantia nigra neurons. Gene Ther., 8:864–873.

16.

Gambhir

S.S.

, Bauer

, Black

M.E.

, Liang

, Kokoris

M.S.

, Barrio

J.R.

, Iyer

, Namavari

, Phelps

M.E.

, Herschman

H.R.

2000. A mutant herpes simplex virus type 1 thymidine kinase reporter gene shows improved sensitivity for imaging reporter gene expression with positron emission tomography. Proc. Natl. Acad. Sci. U.S.A., 97:2785–2790.

17.

Geraerts

, Michiels

, Baekelandt

, Debyser

, Gijsbers

2005. Upscaling of lentiviral vector production by tangential flow filtration. J. Gene Med., 7:1299–1310.

18.

Geraerts

, Eggermont

, Hernandez-Acosta

, Garcia-Verdugo

J.M.

, Baekelandt

, Debyser

2006. Lentiviral vectors mediate efficient and stable gene transfer in adult neural stem cells in vivo. Hum. Gene Ther., 17:635–650.

19.

Hellen

C.U.

, Sarnow

2001. Internal ribosome entry sites in eukaryotic mRNA molecules. Genes Dev., 15:1593–1612.

20.

Hennecke

, Kwissa

, Metzger

, Oumard

, Kroger

, Schirmbeck

, Reimann

, Hauser

2001. Composition and arrangement of genes define the strength of IRES-driven translation in bicistronic mRNAs. Nucleic Acids Res., 29:3327–3334.

21.

Hioki

, Kameda

, Nakamura

, Okunomiya

, Ohira

, Nakamura

, Kuroda

, Furuta

, Kaneko

2007. Efficient gene transduction of neurons by lentivirus with enhanced neuron-specific promoters. Gene Ther., 14:872–882.

22.

Jakobsson

, Lundberg

2006. Lentiviral vectors for use in the central nervous system. Mol. Ther., 13:484–493.

23.

Kirik

, Bjorklund

2003. Modeling CNS neurodegeneration by overexpression of disease-causing proteins using viral vectors. Trends Neurosci., 26:386–392.

24.

Lauwers

, Debyser

, Van Dorpe

, De Strooper

, Nuttin

, Baekelandt

2003. Neuropathology and neurodegeneration in rodent brain induced by lentiviral vector-mediated overexpression of α-synuclein. Brain Pathol., 13:364–372.

25.

Lauwers

, Beque

, Van Laere

, Nuyts

, Bormans

, Mortelmans

, Casteels

, Vercammen

, Bockstael

, Nuttin

, Debyser

, Baekelandt

2007. Non-invasive imaging of neuropathology in a rat model of α-synuclein overexpression. Neurobiol. Aging, 28:248–257.

26.

Lorens

J.B.

, Pearsall

D.M.

, Swift

S.E.

, Peelle

, Armstrong

, Demo

S.D.

, Ferrick

D.A.

, Hitoshi

, Payan

D.G.

, Anderson

2004. Stable, stoichiometric delivery of diverse protein functions. J. Biochem. Biophys. Methods, 58:101–110.

27.

Martinez-Salas

1999. Internal ribosome entry site biology and its use in expression vectors. Curr. Opin. Biotechnol., 10:458–464.

28.

Martinez-Salas

, Pacheco

, Serrano

, Fernandez

2008. New insights into internal ribosome entry site elements relevant for viral gene expression. J. Gen. Virol., 89:611–626.

29.

Massoud

T.F.

, Gambhir

S.S.

2007. Integrating noninvasive molecular imaging into molecular medicine: An evolving paradigm. Trends Mol. Med., 13:183–191.

30.

Mizuguchi

, Xu

, Ishii-Watabe

, Uchida

, Hayakawa

2000. IRES-dependent second gene expression is significantly lower than cap-dependent first gene expression in a bicistronic vector. Mol. Ther., 1:376–382.

31.

Newcomb

E.W.

, Tamasdan

, Entzminger

, Arena

, Schnee

, Kim

, Crisan

, Lukyanov

, Miller

D.C.

, Zagzag

2004. Flavopiridol inhibits the growth of GL261 gliomas in vivo: Implications for malignant glioma therapy. Cell Cycle, 3:230–234.

32.

Reumers

, Deroose

C.M.

, Krylyshkina

, Nuyts

, Geraerts

, Mortelmans

, Gijsbers

, Van den Haute

, Debyser

, Baekelandt

2008. Non-invasive and quantitative monitoring of adult neuronal stem cell migration in mouse brain using bioluminescence imaging. Stem Cells, 26:2382–2390.

33.

Ryan

M.D.

, Drew

1994. Foot-and-mouth disease virus 2A oligopeptide mediated cleavage of an artificial polyprotein. EMBO J., 13:928–933.

34.

Ryan

M.D.

, Flint

1997. Virus-encoded proteinases of the picornavirus super-group. J. Gen. Virol., 78:699–723.

35.

Scherr

, Battmer

, Eder

, Schule

, Hohenberg

, Ganser

, Grez

, Blomer

2002. Efficient gene transfer into the CNS by lentiviral vectors purified by anion exchange chromatography. Gene Ther., 9:1708–1714.

36.

Svitkin

Y.V.

, Herdy

, Costa-Mattioli

, Gingras

A.-C.

, Raught

, Sonenberg

2005. Eukaryotic translation initiation factor 4E availability controls the switch between cap-dependent and internal ribosomal entry site-mediated translation. Mol. Cell. Biol., 25:10556–10565.

37.

Szymczak

A.L.

, Vignali

D.A.

2005. Development of 2A peptide-based strategies in the design of multicistronic vectors. Expert Opin. Biol. Ther., 5:627–638.

38.

Szymczak

A.L.

, Workman

C.J.

, Wang

, Vignali

K.M.

, Dilioglou

, Vanin

E.F.

, Vignali

D.A.

2004. Correction of multi-gene deficiency in vivo using a single “self-cleaving” 2A peptide-based retroviral vector. Nat. Biotechnol., 22:589–594.

39.

Van den Haute

, Eggermont

, Nuttin

, Debyser

, Baekelandt

2003. Lentiviral vector-mediated delivery of short hairpin RNA results in persistent knockdown of gene expression in mouse brain. Hum. Gene Ther., 14:1799–1807.

40.

Verma

I.M.

, Weitzman

M.D.

2005. Gene therapy: Twenty-first century medicine. Annu. Rev. Biochem., 74:711–738.

41.

Wang

, Iyer

, Annala

A.J.

, Chappell

, Mauro

, Gambhir

S.S.

2005. Noninvasive monitoring of target gene expression by imaging reporter gene expression in living animals using improved bicistronic vectors. J. Nucl. Med., 46:667–674.

42.

Yaghoubi

S.S.

, Barrio

J.R.

, Namavari

, Satyamurthy

, Phelps

M.E.

, Herschman

H.R.

, Gambhir

S.S.

2005. Imaging progress of herpes simplex virus type 1 thymidine kinase suicide gene therapy in living subjects with positron emission tomography. Cancer Gene Ther., 12:329–339.

43.

Zagzag

, Miller

D.C.

, Chiriboga

, Yee

, Newcomb

E.W.

2003. Green fluorescent protein immunohistochemistry as a novel experimental tool for the detection of glioma cell invasion in vivo. Brain Pathol., 13:34–37.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.21 MB