Gene Therapy (2001) 8, 864–873  2001 Nature Publishing Group All rights reserved 0969-7128/01 $15.00 www.nature.com/gt RESEARCH ARTICLE Recombinant AAV vectors containing the foot and mouth disease 2A sequence confer efficient bicistronic gene expression in cultured cells and rat substantia nigra neurons

S Furler, J-C Paterna, M Weibel and H Bu¨eler Institut fu¨r Molekularbiologie, Universita¨tZu¨rich, Winterthurerstrasse 190, CH-8057 Zu¨rich, Switzerland

Recombinant adeno-associated (rAAVs) are promis- generate the mature FMDV proteins. We have generated ing vectors for since they efficiently and stably several different rAAV genomes in which two coding regions transduce a variety of tissues of immunocompetent animals. are fused in-frame via the FMDV 2A sequence. We show The major disadvantage of rAAVs is their limited capacity to that FMDV 2A efficiently mediates the generation of the package foreign DNA (р5 kb). Often, co-expression of two expected cleavage products from the artificial fusion proteins or more genes from a single is desirable to achi- in cells. Furthermore, we find that both EGFP and ␣- eve maximal therapeutic efficacy or to track transduced cells synuclein are expressed at substantially higher levels from in vivo by suitable reporter genes. The internal ribosome 2A vectors than from the corresponding IRES-based vec- entry site (IRES) sequence of encephalomyocarditis virus tors, while SOD-1 is expressed at comparable or slightly has been widely used to construct bicistronic viral vectors. higher levels. Finally, we demonstrate for the first time, that However, the IRES is rather long and IRES-mediated trans- the 2A sequence results in effective bicistronic gene lation can be relatively inefficient when compared with cap- expression in vivo after injection of 2A-dependent rAAVs into dependent translation. As an alternative to the IRES for in the rat substantia nigra. We conclude that 2A-containing vivo gene expression, we studied the 16 amino-acid long 2A rAAVs may represent an attractive alternative to IRES- peptide of foot and mouth disease virus (FMDV). The 2A dependent vectors for ex vivo and in vivo gene expression peptide mediates the primary cis-‘cleavage’ of the FMDV and gene therapy. Gene Therapy (2001) 8, 864–873. polyprotein in a cascade of processing events that ultimately

Keywords: adeno-associated virus; FMDV 2A; IRES; bicistronic vector; brain

Introduction adenovirus/AAV,21 hybrid HSV/AAV,22 and trans- complementing AAV23 vectors have been generated as a Recombinant adeno-associated viruses (rAAVs) can potential solution to this problem. Nevertheless, the low infect both dividing and quiescent cells, are well tolerated insert capacity of rAAV remains a problem, in particular 1–5 and show low immunogenicity. Numerous studies when two or more genes need to be expressed from a have demonstrated the ability of these vectors to mediate single rAAV. efficient and long-lasting gene expression in different The internal ribosome entry site (IRES) of encepahlo- 6–14 tissues of immunocompetent animals. Recently, pro- myocarditis virus (EMCV) has been widely used to link gress has been made in identifying positions within the two genes transcribed from a single promoter within AAV proteins that tolerate small insertions of tar- recombinant viral vectors.24–26 However, the efficiency of geting ligands without significantly affecting resultant IRES-dependent translation may vary in different cells 15–17 viral titers. Moreover, rAAV genomes can be pack- and tissues27–29 and IRES-dependent second gene aged into derived from other AAV serotypes, expression can be significantly lower than cap-dependent rendering them more efficient in infecting certain cell first gene expression in bicistronic vectors.30,31 Further- 18,19 types. It is expected that these developments and more, in consideration of the low insert capacity of rAAV ongoing efforts to obtain detailed structural information vectors, an element promoting bicistronic expression but on the AAV capsid will expand the number of tissues being shorter than the IRES may be preferred. and cells that can be targeted by AAV in the future. Picornavirus genomes contain a single open reading Unfortunately, rAAVs can only encapsidate transgene frame encoding a polyprotein of about 225 kDa. This 20 cassettes smaller than 5 kb in length. Hybrid polyprotein undergoes co-translational cleavages in cis to produce the primary cleavage products, and then is further processed to generate the mature viral proteins.32 Correspondence: H Bu¨eler The first cleavage is mediated by the 2A sequence which Received 8 December 2000; accepted 27 March 2001 cleaves at its own C-terminus (cardioviruses and 2A-dependent adeno-associated viral vectors S Furler et al 865 aphthoviruses) or N-terminus (enteroviruses and rhinoviruses). The 2A region in foot and mouth disease virus (FMDV) spans only 16 amino acids.33,34 When engi- neered into heterologous fusion proteins, the FMDV 2A sequence has been shown to mediate cleavage both in vitro using eukaryotic translation systems,33–35 as well as in plants36,37 and cultured mammalian cells.38,39 Recently, the FMDV 2A sequence has been incorporated into retro- viral vectors, alone or combined with different IRES sequences to construct bicistronic, tricistronic and tetraci- stronic retroviral vectors. These vectors were shown to co-express resistance and reporter genes in producer lines and infected cultured fibroblasts.40,41 However, to date no results on the efficiency of 2A-mediated gene expression in animals and the potential of this system for in vivo gene therapy have been reported. To further investigate the usefulness of the 2A sequence for the coordinate expression of two genes from a bicistronic viral (gene therapy) vector, we generated recombinant adeno-associated viral genomes encoding ␣- synuclein (syn) and EGFP or Cu/Zn superoxide dismu- tase (SOD-1) and EGFP linked via the FMDV 2A sequence. We assessed 2A-mediated cleavage by analyzing the various gene products by Western blot and confocal microscopy in transiently transfected cells. Moreover, we quantitatively compared 2A- and IRES- mediated second gene expression for three different proteins. Finally, we generated rAAVs encoding syn-2A- Figure 1 The vectors used in this study. (a) CMV-SOD1–2A-EGFP, (b) EGFP and EGFP-2A-syn under the control of the neuron- CMV-syn-2A-EGFP, (c) CMV-syn-2A*-EGFP (NPGP mutated to specific human PDGF-␤ promoter, and compared them AAAA), (d) CMV-EGFP-2A-SOD1, (e) CMV-EGFP-2A-syn, (f) CMV- with a rAAV encoding ␣-synuclein and EGFP separated syn-IRES-EGFP, (g) CMV-EGFP-IRES-SOD1, (h) CMV-EGFP-IRES- by the IRES element of EMCV with respect to their syn, (i) PDGF-EGFP-2A-syn, (j) PDGF-syn-2A-EGFP, (k) PDGF-syn- abilities to confer gene expression in rat substantia nigra IRES-EGFP. neurons in vivo. or downstream of EGFP. The plasmids CMV-syn-2A- EGFP and CMV-EGFP-2A-syn resulted in the generation Results of both the unprocessed precursor proteins, as well as the expected cleavage products syn-2A (Figure 3a, lane 1), Immunoblot analysis of FMDV 2A-mediated protein Pro-EGFP (Figure 3B, lane 1), EGFP-2A (Figure 3b, lane expression in transfected and rAAV-infected cells 3) and Pro-syn (Figure 3a, lane 3). 2A-mediated cleavage The CMV promoter-controlled plasmids encoding ␣- was estimated to be 50–60% efficient. Extracts of 293T synuclein and EGFP or SOD1 and EGFP linked in-frame cells transfected with plasmid CMV-syn-2A-EGFP also via the FMDV 2A protease sequence (Figure 1a, b and d, contained an unexpected protein with an estimated size e) were transiently transfected into 293T cells, and of 35 kDa that reacted with the monoclonal antibody expression of EGFP and ␣-synuclein or SOD1 was against ␣-synuclein (Figure 3a, lanes 1 and 2). In order assessed by Western blot analysis. As shown in Figure 2, to find out whether this protein was generated by cleav- plasmid CMV-SOD1–2A-EGFP resulted in the generation age of the 2A ‘protease’ at a second cryptic site within of both expected cleavage products, SOD1–2A (Figure 2a, the N-terminal portion of EGFP, we mutated the NPGP lane 1) and Pro-EGFP (Figure 2b, lane 1) from the SOD1– motif of 2A to AAAA (four alanines) by PCR 2A-EGFP precursor protein. Likewise, a lysate of 293T mutagenesis. The NPGP motif has previously been cells transfected with plasmid CMV-EGFP-2A-SOD1 con- shown to be essential for 2A-mediated cleavage.42 tained the two cleavage products EGFP-2A (Figure 2b, Mutation of NPGP to AAAA in plasmid CMV-syn-2A*- lane 2) and Pro-SOD1 (Figure 2a, lane 2). Since Pro-SOD1 EGFP (Figure 1c) completely prevented the generation of (P-SOD1) co-migrates with endogenous SOD-1, we attri- both syn-2A (Figure 3a, compare lanes 1 and 2) and Pro- bute the increased signal to the presence of Pro-SOD1 EGFP (Figure 3b, compare lanes 1 and 2). In contrast, the (Figure 2a, lane 2). In addition to the proteins generated 35 kDa ␣-synuclein immunoreactive protein was unaffec- by 2A-mediated cleavage, both plasmids also resulted in ted by the mutagenesis of the conserved NPGP motif the synthesis of the full-length, unprocessed precursor (compare lanes 1 and 2 in Figure 3a), suggesting that this proteins SOD1–2A-EGFP and EGFP-2A-SOD1 (Figure 2a, protein was not generated by 2A-mediated cleavage, and b). Based on the relative intensities of the immunoreac- may be a degradation product of the unprocessed syn- tive bands on the Western blots, we estimated that about 2A-EGFP precursor (see also Discussion). Finally, to dem- two-thirds of the precursor proteins were cleaved. onstrate processing of a 2A-containing fusion protein To demonstrate the utility of the FMDV 2A protease after rAAV infection, cells were transduced with a rAAV within the context of another fusion protein, we inserted vector carrying the SOD1–2A-EGFP cassette, and 2A- the human ␣-synuclein coding sequence either upstream dependent cleavage products were analyzed by Western

Gene Therapy 2A-dependent adeno-associated viral vectors S Furler et al 866

Figure 3 Synthesis and processing of syn-2A-EGFP and EGFP-2A-syn fusion proteins in transiently transfected 293T cells. 20 ␮g total protein per lane was separated by SDS-PAGE, transferred to a PVDF membrane and analyzed by immunoblot using a monoclonal antibody against ␣- synuclein (a) or a polyclonal antibody against EGFP (b). Cells were trans- fected with plasmid CMV-syn-2A-EGFP (lanes 1), CMV-syn-2A*-EGFP (NPGP mutated to AAAA; lanes 2), CMV-EGFP-2A-syn (lanes 3), posi- tive control plasmid CMV-syn (lanes 5) or CMV-EGFP (lanes 6), or remained untransfected (lanes 4).

that was present on the transfected plasmids (Figure 4b, Figure 2 Synthesis and processing of SOD1–2A-EGFP and EGFP-2A- SOD1 fusion proteins in transfected and infected 293T cells. 40 ␮g total c). In addition, we performed FACS analyses for EGFP protein per lane was separated by SDS-PAGE, transferred to a PVDF expression to confirm comparable transfection membrane and analyzed by immunoblot using polyclonal antibodies efficiencies (Figure 4a, c). In all cases, between 85% and against SOD1 (a and c) or EGFP (b). Cells were transfected with plasmid 90% of the cells were transfected. Bands specific for EGFP CMV-SOD1–2A-EGFP (lanes 1), CMV-EGFP-2A-SOD1 (lanes 2), posi- (Figure 4a), SOD-1 (Figure 4b) and ␣-synuclein (Figure tive control plasmid CMV-SOD1 (lanes 4) or CMV-EGFP (lanes 5), or 4c), as well as the bands corresponding to the internal remained untransfected (lanes 3). In Figure 2c, lane 1, cells were infected with a rAAV (MOI of 3000 particles per cell) carrying the same CMV- control proteins were quantitated by densitometry (for SOD1–2A-EGFP cassette that was transfected in Figure 2a, lane 1. details, see Materials and methods). Densitometric quan- titation revealed that EGFP was expressed at nine-fold higher levels, ␣-synuclein at 11.9-fold higher levels and blot with an antibody to SOD1 (Figure 2c). The same pat- SOD-1 at 3.2-fold higher levels than from the correspond- tern of proteins was observed in transfected (Figure 2a) ing IRES vectors. Similar results were found in repeat and rAAV-infected (Figure 2c) cells. Taken together, experiments. These results indicate that 2A-mediated these results demonstrate that the FMDV 2A sequence cleavage can confer at least comparable and, in some efficiently promotes cis-cleavage of different heterologous cases, significantly higher levels of second gene fusion proteins encoded within the context of rAAV expression than IRES-dependent translation. vector genomes, resulting in the generation of the corresponding cleavage products in transfected and The C-terminal 2A tail does not inhibit SOD-1 enzymatic rAAV-infected cells. activity In our experiments, 2A-mediated cleavage of a fusion Quantitative comparison of 2A- and IRES-mediated protein results in the addition of 23 (2A) amino acids to second gene expression the C-terminus of the first gene product, and the incor- To compare the amounts of second gene product poration of a single proline at the N-terminus of the expressed from either a 2A-dependent or IRES- second protein. To study whether these modifications dependent vector, we transiently transfected 293T cells may inhibit protein activity, we measured the SOD-1 with plasmids expressing EGFP, SOD-1 or ␣-synuclein activities in lysates of 293T cells transfected with downstream of either the 2A or IRES sequence (all other psubCMV-SOD1–2A-EGFP or psubCMV-EGFP-2A-SOD1 elements and sequences were identical; vectors com- using a colorimetric kinetic assay. As controls, we used pared: Figure 1b versus 1f, Figure 1d versus 1g; Figure 1e psubCMV-SOD1-transfected and psubCMV-EGFP-IRES- versus 1h). Proteins were analyzed in whole cell lysates SOD1-transfected cells that express full-length wild-type by Western blot (Figure 4a–c). As an internal standard, SOD-1. In addition, psubCMV-EGFP-transfected cells we simultaneously detected either human SOD-1 that is were used to determine endogenous SOD-1 activity. In constitutively expressed in 293T cells (Figure 4a) or EGFP all cases, transfection efficiencies were comparable (85–

Gene Therapy 2A-dependent adeno-associated viral vectors S Furler et al 867 sample (lysate) and subtraction of endogenous SOD-1 activity revealed that neither the C-terminal 2A tail (SOD1–2A) nor the N-terminal proline (Pro-SOD1) inhibited SOD-1 activity (Table 1). Similarly, our results described above show that both EGFP-2A and Pro-EGFP are green fluorescent. Since the function of ␣-synuclein is unknown, we were not able to carry out experiments to address the influence of the 2A tail on this protein.

Confirmation of co-expression of ␣-synuclein and EGFP in transfected cells by confocal microscopy To demonstrate that ␣-synuclein and EGFP fused in- frame via the 2A sequence were co-expressed in the same cells, we identified EGFP-positive cells by direct confocal fluorescence microscopy and ␣-synuclein-expressing cells by immunocytochemistry with a monoclonal antibody against ␣-synuclein. Most, if not all cells transfected with the plasmid CMV-syn-2A-EGFP expressed both ␣- synuclein and EGFP (Figure 5a–c). However, EGFP was expressed uniformly in the cytoplasm whereas ␣- synuclein appeared to be enriched around membranes and at the periphery of the cells (Figure 5b, c). The expression patterns and levels of the two proteins were comparable to those observed in cells transfected with the corresponding positive control plasmids CMV-EGFP (Figure 5d) and CMV-syn (Figure 5e), respectively.

Bicistronic rAAV vectors based on the FMDV 2A sequence confer efficient gene expression in the rat brain To assess whether 2A-based bicistronic expression vec- tors can mediate efficient gene expression in vivo (ie in the rat brain) and to compare them side by side with an Figure 4 Quantitative comparison of 2A- and IRES-mediated second gene IRES-dependent vector, we generated rAAVs expressing expression. 2.5 × 106 293T cells in 60-mm plates were transiently trans- ␣-synuclein and EGFP linked via the 2A or IRES ␮ fected with 8 g of the various plasmids by the calcium-phosphate copre- sequence under the control of the human PDGF-␤ chain cipitation method. Comparable transfection efficiencies were confirmed by determining the percentage of EGFP-positive cells by flow cytometry. In promoter (Figure 1i–k). This promoter was chosen all cases, between 85 and 90% of the cells expressed EGFP. 20 ␮g total because it has been successfully used to target the protein per lane was separated by SDS-PAGE, transferred to a PVDF expression of proteins to neurons in transgenic models membrane and analyzed by immunoblot. (a) Blot probed with a rabbit of neurodegenerative disease43,44 Comparable infectious polyclonal antibody against EGFP and (simultaneously) a sheep poly- titers (105 infectious units) of HPLC-purified rAAV par- clonal antibody against human SOD1 (internal standard). Cells were ticles were injected unilaterally into the substantia nigra transfected with plasmid CMV-syn-2A-EGFP (lane 1), CMV-syn-IRES- EGFP (lane 2), positive control plasmid CMV-EGFP (lane 4), or remained of adult rats (three rats per vector type). Injection untransfected (lane 3). (b) Blot probed with a sheep polyclonal antibody of either rAAV-EGFP-2A-syn or rAAV-syn-2A-EGFP against human SOD1 and a rabbit polyclonal antibody against EGFP resulted in readily detectable and comparable expression (internal control). Cells were transfected with plasmid CMV-EGFP-2A- of both ␣-synuclein and EGFP in all animals 4 weeks after SOD1 (lane 1), CMV-EGFP-IRES-SOD1 (lane 2), positive control plas- virus injection (Figure 6a–d). In contrast, while mid CMV-SOD1 (lane 4), or remained untransfected (lane3). (c) Blot expression of ␣-synuclein by rAAV-syn-IRES-EGFP was probed with a mouse monoclonal antibody against human ␣-synuclein and a rabbit polyclonal antibody against EGFP (internal control). Cells similar to the 2A vectors, EGFP levels were much lower were transfected with plasmid CMV-EGFP-2A-syn (lane 1), CMV- than those achieved with the 2A vectors (Figure 6e, f). EGFP-IRES-syn (lane 2), positive control plasmid CMV-syn (lane 4), or There was a weak signal for ␣-synuclein in the uninjected remained untransfected (lane 3). hemisphere representing endogenous rat protein due to cross-reactivity of the polyclonal antibody to rat ␣- synuclein (Figure 6g). The in vivo expression data are in 90%). Table 1 shows that transfection increased SOD-1 good agreement with the immunoblot results that also activity 2.0–3.6-fold relative to the endogenous activity, showed higher protein levels when EGFP was encoded depending on the construct. No difference in SOD-1 downstream of the 2A sequence as compared with being activity was seen between psubCMV-SOD1-transfected positioned downstream of the IRES element (Figure 4a). and psubCMV-SOD1–2A-EGFP-transfected cells, indicat- To confirm co-expression of ␣-synuclein and EGFP in ing that the C-terminal 2A tail of SOD1–2A does not neurons transduced with the 2A-dependent rAAVs, we inhibit the enzymatic activity of SOD-1. SOD-1 activities analyzed higher magnifications of the brain sections by were also comparable in psubCMV-EGFP-2A-SOD1- confocal microscopy. The two proteins were co-expressed transfected and psubCMV-EGFP-IRES-SOD1-transfected in the majority of infected neurons (Figure 7a–f). How- cells. Correction for the protein concentration of each ever, occasional single-positive cells were also observed,

Gene Therapy 2A-dependent adeno-associated viral vectors S Furler et al 868 Table 1 SOD-1 enzymatic activity

Transfected plasmid SOD-525 units/ml Protein concentration SOD-525 units/ml/mg protein lysate of lysate (mg/ml) Including endogenous After subtraction of SOD-1 activity endogenous SOD-1 activity

SOD1 (30 ␮g) + EGFP (10 ␮g) 4.72 ± 0.72 9.6 ± 0.7 0.49 0.32 SOD1-2A-EGFP 5.13 ± 0.71 8.4 ± 0.3 0.61 0.44 EGFP-2A-SOD1 3.86 ± 0.02 9.2 ± 1.6 0.42 0.25 EGFP-IRES-SOD1 3.70 ± 0.81 10.8 ± 1.3 0.34 0.17 EGFP 1.90 ± 0.15 10.9 ± 1.2 0.17 0

5 × 106 293T cells were transfected in 10-cm plates by the calcium-phosphate precipitation method with 30 ␮g of the indicated plasmids. EGFP served as an internal control for transfection efficiency. In all cases, transfection efficiencies were comparable (85–90%). Thirty hours after transfection, the cells were collected, washed with PBS and lysed in 1% Triton X-100, 10 mm Tris-HCl pH 8.0, 0.14 m NaCl containing complete protease inhibitor cocktail (Roche Molecular Biochemicals). Protein concentrations were determined with the BCA protein assay kit (Pierce). Lysates were cleared by centrifugation and extracted with 1.6 volumes of ethanol/chloroform (6.25/3.75). The SOD-1 activity of each sample was determined in triplicate in the aqueous phase using the superoxide dismutase assay kit (Calbiochem) according to the manufacturer’s instructions. Values represent mean ± s.d. of triplicate measurements.

a b c de

Figure 5 Co-expression of ␣-synuclein and EGFP in transfected 293T cells. Cells were fixed with 4% paraformaldehyde. EGFP was detected by direct confocal fluorescence microscopy with a FITC filter. ␣-Synuclein was detected by incubation with an ␣-synuclein-specific mouse monoclonal antibody and a Cy3-conjugated secondary antibody against mouse immunoglobulins, and visualized by confocal scanning with a Cy3-selective filter. Cells were transfected with plasmid CMV-syn-2A-EGFP (a–c) or positive control plasmid CMV-EGFP (d) or CMV-syn (e). (a and d) EGFP expression; (b and e) ␣-synuclein expression; (c) overlay of a and b. Untransfected cells were negative for both EGFP and ␣-synuclein expression (not shown).

perhaps owing to the apparently distinct subcellular occurs in cis and requires the 16 amino-acid long 2A localization of ␣-synuclein and EGFP. sequence that mediates cleavage at its own C- terminus.32,34 The mechanism by which the 2A sequence Discussion promotes this cleavage is not fully understood. It has been proposed that 2A represents either (a) a recognition Many acquired diseases are now being considered for site for a fairly ubiquitous protease associated with euka- treatment by gene therapy, including AIDS, cancer and ryotic ribosomes; (b) an unusually short novel element of neurodegenerative diseases. Often, the underlying defect protein cleavage; or (c) a sequence that results in a failure and pathogenic mechanisms are unknown or poorly to synthesize a peptide bond, yet allows the continuation defined in these disorders. Consequently, treatments may of translation.33,46 Regardless of the actual mechanism, attempt to interfere with various (postulated) pathways the utility of the 2A sequence for the coordinate simultaneously (eg oxidative stress and apoptosis in expression of two heterologous proteins has been demon- neurodegeneration) or inhibit virus propagation at differ- strated previously in plants36,37 and cultured mammalian ent levels (eg inhibition of replication and selective killing cells.38–40 of virus-infected cells). Thus, it may be desirable to co- In this work, we have further explored the potential of express two or more genes simultaneously in the same the FMDV 2A sequence for the co-expression of two cells. Furthermore, co-expression of a reporter gene with genes in cultured cells, and quantitatively compared the a therapeutic gene is frequently desired to select geneti- levels of second gene expression mediated by 2A- and cally modified cells ex vivo or identify transduced cells in IRES-dependent vectors for three different proteins. Fur- vivo. Previously, different genes have been delivered thermore, we generated 2A-dependent recombinant either by separate viral vectors14,45 or bicistronic vectors adeno-associated viruses and studied, for the first time, incorporating the IRES.24–26 While the first approach is the utility of the 2A system for direct in vivo gene laborious and time-consuming, the second strategy can expression by intracerebral injection of rats with rAAV be limited by the reduced expression of the gene inserted vectors carrying ␣-synuclein and EGFP coding sequences downstream of the IRES.28,31 In addition, the IRES ocup- fused in-frame via 2A. ies a relatively large portion of the cloning capacity of The 2A sequence efficiently promoted the generation some vectors (eg rAAV). of the predicted cleavage products from four different In FMDV, the primary cleavage of the polyprotein fusion proteins in transfected cells. We also show that

Gene Therapy 2A-dependent adeno-associated viral vectors S Furler et al 869

a b a bc

c d d e f

Figure 7 Co-expression of ␣-synuclein and EGFP in substantia nigra e f mediated by 2A-dependent rAAV vectors. Brain sections of rats injected with rAAV-PDGF-EGFP-2A-syn (a–c) or rAAV-PDGF-syn-2A-EGFP (d–f) were analyzed by confocal microscopy at ×200 magnification. EGFP expression was visualized directly. ␣-Synuclein was detected with a rabbit polyclonal antibody against human ␣-synuclein and a donkey anti-rabbit IgG conjugated to Cy3. (a and d) EGFP expression; (b and e) ␣-synuclein expression; (c and f) overlays.

g derived from the syn-2A cleavage product by post- translational modification, since syn-2A disappeared upon mutation. Therefore, we believe that the 35 kDa protein most likely represents a degradation product of the syn-2A-EGFP precursor. We found that genes can be efficiently expressed from 2A vectors irrespective of whether they are fused upstream or downstream to the 2A sequence (half to two- thirds of the precursors cleaved, Figures 2 and 3). Some Figure 6 Comparison of ␣-synuclein and EGFP expression in neurons of the rat substantia nigra after injection with 2A- or IRES containing rAAV (but not all) previous studies with in vitro translation sys- 32 37 vectors. Rats were stereotactically injected above the left substantia nigra tems and transgenic plants reported nearly complete with 105 infectious units of rAAV-PDGF-EGFP-2A-syn (a, b), rAAV- cleavage of the precursor protein by 2A. However, simi- PDGF-syn-2A-EGFP (c, d), or rAAV-PDGF-syn-IRES-EGFP (e, f). lar to our findings, within the context of a poliovirus vec- ␣ Expression of EGFP (a, c, e) and -synuclein (b, d, f, g) was analyzed tor, 2A-mediated cleavage was (also) not complete.47 It is by fluorescence microscopy in coronal sections of the substantia nigra 4 conceivable that cleavage efficiency is influenced to some weeks after rAAV injection, either directly (EGFP), or after incubation with a mouse monoclonal antibody against ␣-synuclein and a Cy3- degree by the surrounding sequences and structural pro- 48 conjugated secondary antibody against mouse immunoglobulins. Note tein domains, the amount of precursor expressed, and that panel e has been overexposed for easier recognition of EGFP the cell type. A model for the nonstoichiometric, cotrans- expression. This can be seen from the increased diffuse green background lational protein scission by 2A has recently been pro- compared with panels a and c. (g) Section of the right uninjected hemi- 46 ␣ posed that may explain some of these observations. sphere, showing low level expression of endogenous rat -synuclein (same Importantly, despite incomplete processing, the exposure conditions as in b, d and f). Magnification ×32. expression levels of both EGFP and ␣-synuclein were consistently and substantially higher in transfected cells 2A-mediated cleavage functions in cells infected with a when the proteins were encoded downstream of 2A rAAV carrying the SOD1–2A-EGFP expression cassette. instead of being translated from an IRES element. SOD- In agreement with a previous report,42 2A-mediated pro- 1 expression was comparable or slightly higher after the cessing was completely dependent on the conserved 2A element. Moreover, in the brain EGFP fluorescence NPGP motif, since mutation of NPGP to AAAA fully pre- after injection of the syn-2A-EGFP rAAV was signifi- vented the generation of the cleavage products. In con- cantly stronger than after injection of a comparable infec- trast, the presence of an additional 35 kDa protein, that tious dose of the syn-IRES-EGFP rAAV. We conclude that was detected with the anti-␣-synuclein antibody in cells 2A-dependent rAAVs confer at least comparable and, in transfected with plasmid syn-2A-EGFP, was not affected some cases, substantially higher expression levels of the by the NPGP to AAAA mutation. This indicates that the downstream-positioned gene when compared with an 35 kDa protein was not generated by 2A-dependent otherwise identical IRES vector. Our experiments do not cleavage at a ‘cryptic’ position within EGFP. This con- address the mechanism resulting in increased protein clusion is consistent with the absence of any EGFP amino expression. While we believe that this is due to increased acid sequences similar to those at the C-terminus of 2A translational efficiency of the 2A system,46 we can not (as revealed by close inspection of the EGFP primary rule out a differential effect of the 2A and IRES elements sequence). In addition, the 35 kDa protein can not be on transcription and mRNA stability.

Gene Therapy 2A-dependent adeno-associated viral vectors S Furler et al 870 Most, if not all neurons infected with the 2A- after gene delivery by a different route or by more dependent rAAVs co-expressed EGFP and ␣-synuclein at immunogenic viral vectors, remains to be addressed. high levels. Although we have not formally shown 2A- Infection of cells by picornaviruses results in a rapid mediated cleavage in the brain (this is technically not shut-off of host protein synthesis due to virus-induced feasible), the distinct subcellular localizations of EGFP cleavage of specific eukaryotic translation initiation fac- (homogenous in cytoplasm) and ␣-synuclein (enriched tors (eIF-4␥).32 In enteroviruses and rhinoviruses this around membranes) in transduced neurons strongly sug- cleavage is dependent on the large 2A proteases.32 In gests that cleavage indeed occurs (if only the fusion pro- addition, coxsackie B virus protease 2A has been shown tein were expressed, we would expect colocalization of to result in the cleavage of dystrophin.53 The situation in the two signals for EGFP and ␣-synuclein). Moreover, we aphtoviruses (ie FMDV) is different; eIF-4␥ cleavage is observed the same subcellular pattern of EGFP and ␣- mediated by the virally encoded L proteinase and not by synuclein expression by confocal microscopy in trans- the 2A sequence.54–57 In the case of FMDV, the 2A fected cells, where we convincingly show cleavage by 2A sequence is much shorter (only 16 amino acids) and it on immunoblots. lacks any homology to other known viral and cellular Our findings that 2A-mediated expression of genes is proteases, indicating that FMDV 2A is in fact not a pro- at least as efficient as from the IRES, combined with the tease, but most likely mediates protein scission by a small size of the 2A sequence (72 bp as opposed to 500 unique mechanism.32 This is supported by experiments bp of the IRES), makes the 2A element particularly showing that the Gly-Pro amide bond (within NPGP) is attractive in gene therapy vectors with limited cloning actually not synthesized, and that the observed ‘cleavage’ capacity like rAAV or retroviral vectors.40 The small size may be the result of hydrolysis mediated by a ribosom- of the 2A element may be useful for the coordinate ally bound 2A polypeptidyl-tRNA.46 Finally, FMDV 2A- expression of several enzymes to reconstitute a complete mediated cleavage of the viral polyprotein and of arti- biosynthetic pathway, which may result in increased ficial target proteins is absolutely dependent on the con- therapeutic benefit in some instances.14,45 In tumor gene served NPGP motif.42 This motif is different from the therapy, combinations of immunostimulatory, anti- known recognition motifs of the other picornaviral 2A angiogenic and suicide genes may be expressed simul- proteases.53,57 Within current databases, the NPGP motif taneously from a single vector to maximize antitumor is only found in proteins from cardioviruses, aphto- effects.49–51 Moreover, effective vaccination against two or viruses (FMDV) and group C porcine retroviruses. Taken more tumor or viral antigens expressed at high levels together, we consider it extremely unlikely that the from one vector may be feasible. Other applications may FMDV 2A peptide will result in the ‘cleavage’ of include the coordinate synthesis of IL-12 subunits that endogenous host proteins. must be expressed at similar levels in order to achieve In summary, our results extend previous findings of maximal cytokine activity.38 Finally, the 2A system 2A-mediated gene expression, specifically show that 2A- should be useful to co-express therapeutic genes with a dependent vectors confer at least comparable protein lev- readily detectable reporter gene or cell-surface marker, in els when directly compared with IRES-containing vectors order to identify and track transduced cells in vivo, and and, most importantly, provide the first demonstration of to determine accurately the titer of transducing units in vivo gene transfer and expression with a 2A-dependent of different types and batches of rAAVs by flow gene therapy vector. We conclude that 2A-containing cytometry.52 bicistronic rAAVs may be an attractive alternative to A potential drawback of the 2A system may be that IRES-dependent vectors for ex vivo and in vivo gene the 23 amino acids being added to the C-terminus of the expression, in particular when the cloning capacity is lim- protein encoded upstream of 2A might influence the ited and high levels of the second gene product are function or activity of some proteins. However, this has required. Future studies addressing the function and not been observed for proteins and enzymes tested to safety of 2A-dependent vectors in additional tissues and date, including SOD-1 and EGFP (this study), chloram- their suitability to co-express proteins targeted to differ- phenicol acetyltransferase,37,39 IL-12,38 puromycin N- ent cellular compartments will be important to further acetyl transferase, ␤-glucuronidase and alkaline phospha- advance this system. tase.40,41 Furthermore, our finding that genes can be expressed effectively when fused upstream or down- stream of 2A suggests that modification-sensitive pro- Materials and methods teins could be encoded from the second cistron after 2A. This would result in the addition of only one proline to Plasmid constructions the N-terminus, which is not expected to interfere with Plasmid pTG39533 was used to amplify by PCR a protein function, in agreement with our findings and sequence encoding the 24 amino-acid FMDV peptide previous reports.37,38,40,41 APVKQTLNFDLLKLAGDVESNPGP (termed 2A). The The immunogenicity of the 2A peptide remains to be PCR product was subcloned between the NheI and MluI determined. A priori, we do not expect the 2A peptide to sites of psubCMV-WPRE and psubPDGF-WPRE52 to gen- be more immunogenic than other foreign epitopes erate the vectors psubCMV-2A-WPRE and psubPDGF- expressed as part of a transgene product. Our results 2A-WPRE. Subsequently, the coding sequences of human show that the 2A sequence does not compromise rAAV- ␣-synuclein (the cDNA was a kind gift of M Goedert, mediated transgene expression in vivo for at least 1 Medical Research Council Laboratory of Molecular month, since EGFP and ␣-synuclein levels in the brain Biology, Cambridge, UK), human SOD-1 (ATCC No. were comparable to those achieved with rAAVs lacking 61647) and EGFP (from pEGFP-N1, Clontech, Palo Alto, 2A.52 However, whether or not the 2A peptide may CA, USA) were amplified by PCR and cloned in-frame increase immunity against other transgene products, or upstream (NheI) or downstream (MluI) of 2A. In each

Gene Therapy 2A-dependent adeno-associated viral vectors S Furler et al 871 case, the upstream open reading frame (ORF) lacked the Switzerland). Prestained SDS-PAGE low range standard termination codon, while the downstream ORF was full- (BioRad, 161–0305) was used as a protein size marker. length (Figure 1a–e and i–j). psubCMV-syn-2A*-EGFP (Figure 1c) was generated by PCR-mediated mutagenesis Densitometry of NPGP to AAAA. The vector psubCMV-syn-IRES- For quantitative comparison of second gene product EGFP (Figure 1f) was constructed by insertion of the expression by 2A-mediated cleavage and IRES- Ј EGFP fragment at the 3 end of the EMCV IRES contained dependent translation, the specific bands on the auto- 58 within pBLSK-IRES to generate plasmid pBLSK-IRES- radiographs were scanned in the Personal Densitometer EGFP. The IRES-EGFP fragment was recovered from (Molecular Dynamics, Sunnyvale, CA, USA). We scanned pBLSK-IRES-EGFP and subcloned into psubCMV-WPRE short exposures of the blots to ensure that all signals were to generate psubCMV-IRES-EGFP-WPRE. Finally, the within the linear range of measurable optical densities ␣ PCR-amplified -synuclein ORF was inserted into of the instrument. Endogenous gene expression levels of psubCMV-IRES-EGFP-WPRE to generate psubCMV-syn- untransfected cells were subtracted where necessary IRES-EGFP (Figure 1f). To generate plasmids expressing (SOD-1 quantitation), and the signals were adjusted rela- ␣ SOD-1 and -synuclein (syn) from the IRES, the two open tive to those of internal standards. Internal standards reading frames were amplified by PCR, generating frag- were EGFP (Figure 4b, c) and endogenous SOD-1 (Figure ments flanked by NcoI sites. The NcoI-digested PCR pro- 4a). For 2A-dependent constructs, the sum of the two ducts were inserted into the NcoI site following the IRES EGFP-immunoreactive bands was taken for standard cal- 58 element in pBLSK-IRES to generate pBLSK-IRES-SOD1 culation. In addition, transfection efficiencies were con- and pBLSK-IRES-syn. Subsequently, the SmaI/HpaI trolled by flow cytometry for EGFP-expressing cells and EGFP fragment from pEGFP-N1 (Clontech) was inserted found to be between 85 and 90% for all samples. into the Klenow polymerase-filled XbaI site of pBLSK- IRES-SOD1 and pBLSK-IRES-syn to generate plasmids SOD-1 activity assay pBLSK-EGFP-IRES-SOD1 and pBLSK-EGFP-IRES-syn. SOD-1 activity was determined with the superoxide dis- Finally, the EGFP-IRES-SOD1 and EGFP-IRES-syn frag- mutase assay kit from Juro AG. Briefly, 5 × 106 293T cells ments were recovered from the Bluescript vectors and were transfected in 10-cm plates by the calcium-phos- inserted into the PmlI site of psubCMV-WPRE52 to gener- phate precipitation method with 30 ␮g of psubCMV- ate the plasmids psubCMV-EGFP-IRES-SOD1 (Figure 1g) SOD1–2A-EGFP, psubCMV-EGFP-2A-SOD1, psubCMV- and psubCMV-EGFP-IRES-syn (Figure 1h). psubPDGF- EGFP-IRES-SOD1, psubCMV-SOD1 plus psubCMV- syn-IRES-EGFP (Figure 1k) was obtained by isolating the EGFP or psubCMV-EGFP. Since EGFP was expressed syn-IRES-EGFP-WPRE fragment from psubCMV-syn- from all constructs, we used it as an internal control for IRES-EGFP, and subcloning it into psubPDGF.52 All con- transfection efficiency. In all cases, transfection structs used in expression analyses contained the wood- efficiencies were comparable (85–90%). Thirty hours after chuck hepatitis post- transcriptional regulatory element59 transfection, the cells were collected, washed with PBS after the second coding sequence and before the SV40 and lysed in 1% Triton X-100, 10 mm Tris-HCl pH 8.0, polyadenylation signal. All plasmids were sequenced to 0.14 m NaCl containing complete protease inhibitor cock- confirm sequence identity. tail (Roche Molecular Biochemicals). Lysates were cleared by centrifugation and extracted with 1.6 volumes of Transfections ethanol/chloroform (6.25/3.75). The SOD-1 activity of For Western blot experiments, 2.5 × 106 293T cells in 60- each sample was determined in triplicate in the aqueous mm plates were transiently transfected with 8 ␮g of the phase according to the manufacturer’s instructions, various plasmids by the calcium-phosphate coprecipi- adjusted for the protein concentration of each sample and tation method. Comparable transfection efficiencies were corrected for endogenous SOD-1 activity of psubCMV- confirmed by determining the percentage of EGFP- EGFP-transfected cells. positive cells by flow cytometry. In all cases, between 85 and 90% of the cells expressed EGFP. For confocal ␣ microscopy, triplicate wells of 293T cells grown on glass Analysis of EGFP and -synuclein expression in cover slips in 12-well plates were transfected with 2 ␮g transfected 293T cells by confocal microscopy plasmid DNA per well. 293T cells were fixed for 15 min with 4% paraformal- dehyde in PBS. EGFP was detected by direct fluorescence ␣ Western blot analysis microscopy in the FITC channel. -Synuclein was Twenty or 40 ␮g of total cellular proteins per lane were detected by immunocytochemistry with 1:200 diluted ␣ separated on 12.5% SDS polyacrylamide gels, and trans- mouse monoclonal antibody against -synuclein ferred to PVDF membranes (BioRad, Hercules, CA, USA) (Zymed) and secondary Cy3-conjugated donkey anti- by semi-dry electroblotting. Primary antibodies were mouse IgG (Jackson ImmunoResearch Laboratories, West ␣ 1:1000 diluted mouse monoclonal antibody against ␣- Grove, PA, USA) at a dilution of 1:200. -Synuclein synuclein (Zymed, South San Francisco, CA, USA), 1:300 expressing neurons were visualized using a Cy3-selective diluted rabbit polyclonal peptide antibody against EGFP filter set. (Clontech) and 1:1000 diluted sheep polyclonal anti-SOD- 1 antibody (Juro AG, Lucerne, Switzerland). Secondary Preparation of rAAV and infection of cells reagents were 1:4000 diluted anti-mouse immunoglob- Recombinant AAVs were generated as described9 and ulin-HRP (Roche Molecular Biochemicals, Rotkreuz, purified by iodixanol gradient centrifugation and heparin Switzerland) and 1:3000 diluted protein G-HRP (BioRad). affinity chromatography.60 Infectious titers of the purified The blots were developed using the ECL enhanced rAAVs were determined by the replication center assay.61 chemiluminescence system (Amersham, Du¨ bendorf, Titers were 5.2 × 107 IU/ml for rAAV-PDGF-syn-2A-

Gene Therapy 2A-dependent adeno-associated viral vectors S Furler et al 872 EGFP, 4.8 × 107 IU/ml for rAAV-PDGF-EGFP-2A-syn, in experimental stroke and epilepsy. Science 2000; 287: 1453– and 4.6 × 107 IU/ml for rAAV-PDGF-syn-IRES-EGFP. 1460. The titer of the rAAV-CMV-SOD1–2A-EGFP vector was 8 Fisher KJ et al. Recombinant adeno-associated virus for muscle determined by slot blot hybridization61 and found to be directed gene therapy. Nat Med 1997; 3: 306–312. 1011 particles/ml. Cells were infected with 3000 particles 9 Glatzel M et al. Adenoviral and adeno-associated viral transfer of genes to the peripheral nervous system. Proc Natl Acad Sci of the vector per cell (for the experiment shown in USA 2000; 97: 442–447. Figure 2c). 10 Herzog RW et al. Long-term correction of canine hemophilia B by gene transfer of blood coagulation factor IX mediated by Stereotactic injections and analysis of gene expression adeno-associated viral vector. Nat Med 1999; 5: 56–63. in brain 11 Lewin AS et al. Ribozyme rescue of photoreceptor cells in a Nine-week-old male Wistar rats were stereotactically transgenic rat model of autosomal dominant retinitis pig- injected with 2.2 ␮l rAAV (105 IU) above the left substan- mentosa (published erratum appears in Nat Med 1998; 4: 1081). tia nigra (AP −5.0 mm; ML +3.0 mm; DV −7.2 mm relative Nat Med 1998; 4: 967–971. to bregma) at a speed of 0.5 ␮l/min using a sp200i syr- 12 Peel AL et al. Efficient transduction of green fluorescent protein in spinal cord neurons using adeno-associated virus vectors con- inge pump (World Precision Instruments, Boston, MA, taining cell type-specific promoters. Gene Therapy 1997; 4: 16–24. USA) and a 10 ␮l Hamilton syringe fitted with a 26-gauge ␮ 13 Snyder RO et al. Correction of hemophilia B in canine and steel cannula. Four weeks later, 40 m coronal cryosec- murine models using recombinant adeno-associated viral vec- tions of the substantia nigra were cut on a microtome tors. Nat Med 1999; 5: 64–70. (Microm, Walldorf, Germany). EGFP expression was vis- 14 Szczypka MS et al. Viral gene delivery selectively restores feed- ualized by direct fluorescence microscopy. ␣-Synuclein ing and prevents lethality of dopamine-deficient mice. Neuron was detected by immunohistochemistry with a rabbit 1999; 22: 167–178. polyclonal antibody against human ␣-synuclein (1:500 15 Girod A et al. Genetic capsid modifications allow efficient re- dilution) (Chemicon, Temecula, CA, USA), followed by targeting of adeno-associated virus type 2. Nat Med 1999; 5: 1:800 diluted donkey anti-rabbit IgG conjugated to Cy3 1052–1056. (Jackson ImmunoResearch Laboratories). 16 Rabinowitz JE, Xiao W, Samulski RJ. Insertional mutagenesis of AAV2 capsid and the production of recombinant virus. 1999; 265: 274–285. Imaging 17 Wu P et al. Mutational analysis of the adeno-associated virus Images were captured by a Zeiss microscope (Carl Zeiss, type 2 (AAV2) capsid gene and construction of AAV2 vectors Feldbach, Switzerland) connected to a Hamamatsu CCD with altered tropism. J Virol 2000; 74: 8635–8647. camera for low magnification brain sections, or by con- 18 Davidson BL et al. Recombinant adeno-associated virus type 2, focal microscopy (Leica, Glattbrugg, Switzerland) for 4, and 5 vectors: transduction of variant cell types and regions high magnification brain sections and 293T cells, using in the mammalian central nervous system. Proc Natl Acad Sci FITC and Cy3 filter sets, respectively. Images were trans- USA 2000; 97: 3428–3432. ferred to a Macintosh computer, and processed using the 19 Zabner J et al. Adeno-associated virus type 5 (AAV5) but not AAV2 binds to the apical surfaces of airway epithelia and facili- Adobe Photoshop 5.0 software. tates gene transfer. J Virol 2000; 74: 3852–3858. 20 Dong JY, Fan PD, Frizzell RA. Quantitative analysis of the pack- Acknowledgements aging capacity of recombinant adeno-associated virus. Hum Gene Ther 1996; 7: 2101–2112. We thank Martin Ryan for plasmid pTG395, Thomas 21 Recchia A et al. Site-specific integration mediated by a hybrid Ba¨chi and Zhizhong Dong for help with confocal adenovirus/adeno-associated virus vector. Proc Natl Acad Sci microscopy and Fritz Ochsenbein for preparing figures. USA 1999; 96: 2615–2620. We are also grateful to Lee Martin for helpful discussion 22 Costantini LC et al. Gene transfer to the nigrostriatal system by hybrid herpes simplex virus/adeno-associated virus amplicon and critical reading of the manuscript. This work was vectors. Hum Gene Ther 1999; 10: 2481–2494. supported by grants from the Swiss National Science 23 Nakai H, Storm TA, Kay MA. Increasing the size of rAAV- Foundation to HB. mediated expression cassettes in vivo by intermolecular joining of two complementary vectors. Nat Biotechnol 2000; 18: 527–532. 24 Klein RL, Lewis MH, Muzyczka N, Meyer EM. Prevention of 6- References hydroxydopamine-induced rotational behavior by BDNF 1 Kaplitt MG et al. Long-term gene expression and phenotypic somatic gene transfer. Brain Res 1999; 847: 314–320. correction using adeno-associated virus vectors in the mam- 25 Levenson VV, Transue ED, Roninson IB. Internal ribosomal malian brain. Nat Genet 1994; 8: 148–154. entry site-containing retroviral vectors with green fluorescent 2 McCown TJ et al. Differential and persistent expression patterns protein and drug resistance markers. Hum Gene Ther 1998; 9: of CNS gene transfer by an adeno-associated virus (AAV) vec- 1233–1236. tor. Brain Res 1996; 713: 99–107. 26 Martinez-Salas E. Internal ribosome entry site biology and its 3 Xiao X, Li J, Samulski RJ. Efficient long-term gene transfer into use in expression vectors. Curr Opin Biotechnol 1999; 10: 458–464. muscle tissue of immunocompetent mice by adeno-associated 27 Borman AM, Le Mercier P, Girard M, Kean KM. Comparison virus vector. J Virol 1996; 70: 8098–8108. of picornaviral IRES-driven internal initiation of translation in 4 Jooss K, Yang Y, Fisher KJ, Wilson JM. Transduction of dendritic cultured cells of different origins. Nucleic Acids Res 1997; 25: cells by DNA viral vectors directs the immune response to trans- 925–932. gene products in muscle fibers. J Virol 1998; 72: 4212–4223. 28 Roberts LO, Seamons RA, Belsham GJ. Recognition of picorna- 5 Snyder RO. Adeno-associated virus-mediated gene delivery. J virus internal ribosome entry sites within cells; influence of Gene Med 1999; 1: 166–175. cellular and viral proteins. RNA 1998; 4: 520–529. 6 Bennett J et al. Stable transgene expression in rod photoreceptors 29 Harries M et al. Comparison of bicistronic retroviral vectors con- after recombinant adeno-associated virus-mediated gene trans- taining internal ribosome entry sites (IRES) using expression of fer to monkey retina. Proc Natl Acad Sci USA 1999; 96: 9920–9925. human interleukin-12 (IL-12) as a readout. J Gene Med 2000; 2: 7 During MJ et al. An oral against NMDAR1 with efficacy 243–249.

Gene Therapy 2A-dependent adeno-associated viral vectors S Furler et al 873 30 Mizuguchi H et al. IRES-dependent second gene expression is 48 Ypma-Wong MF, Filman DJ, Hogle JM, Semler BL. Structural significantly lower than cap-dependent first gene expression in domains of the poliovirus polyprotein are major determinants a bicistronic vector. Mol Ther 2000; 1: 376–382. for proteolytic cleavage at Gln-Gly pairs. J Biol Chem 1988; 263: 31 Zhou Y, Aran J, Gottesman MM, Pastan I. Co-expression of 17846–17856. human adenosine deaminase and multidrug resistance using a 49 Chen SH et al. Combination suicide and cytokine gene therapy bicistronic retroviral vector. Hum Gene Ther 1998; 9: 287–293. for hepatic metastases of colon carcinoma: sustained antitumor 32 Palmenberg AC. Proteolytic processing of picornaviral polypro- immunity prolongs animal survival. Cancer Res 1996; 56: 3758– tein. Annu Rev Microbiol 1990; 44: 603–623. 3762. 33 Donnelly ML et al. The cleavage activities of aphthovirus and 50 Moriuchi S et al. Enhanced tumor cell killing in the presence cardiovirus 2A proteins. J Gen Virol 1997; 78: 13–21. of ganciclovir by herpes simplex virus type 1 vector-directed 34 Ryan MD, King AM, Thomas GP. Cleavage of foot-and-mouth coexpression of human tumor necrosis factor-alpha and herpes disease virus polyprotein is mediated by residues located within simplex virus thymidine kinase. Cancer Res 1998; 58: 5731–5737. a 19 amino acid sequence. J Gen Virol 1991; 72: 2727–2732. 51 Pizzato M et al. Production and characterization of a bicistronic 35 Ryan MD, Drew J. Foot-and-mouth disease virus 2A oligopep- Moloney-based retroviral vector expressing human interleukin tide mediated cleavage of an artificial polyprotein. EMBO J 1994; 2 and herpes simplex virus thymidine kinase for gene therapy 13: 928–933. of cancer. Gene Therapy 1998; 5: 1003–1007. 36 Gopinath K et al. Engineering cowpea mosaic virus RNA-2 into 52 Paterna JC et al. Influence of promoter and WHV post- a vector to express heterologous proteins in plants. Virology transcriptional regulatory element on AAV-mediated transgene 2000; 267: 159–173. expression in the rat brain. Gene Therapy 2000; 7: 1304–1311. 37 Halpin C et al. Self-processing 2A-polyproteins – a system for 53 Badorff C et al. Enteroviral protease 2A cleaves dystrophin: evi- co-ordinate expression of multiple proteins in transgenic plants. dence of cytoskeletal disruption in an acquired cardiomyopathy. Plant J 1999; 17: 453–459. Nat Med 1999; 5: 320–326. et al 38 Chaplin PJ . Production of interleukin-12 as a self-processing 54 Kirchweger R et al. Foot-and-mouth disease virus leader pro- 2A polypeptide. J Interfer Cytok Res 1999; 19: 235–241. teinase: purification of the Lb form and determination of its 39 Percy N, Barclay WS, Garcia-Sastre A, Palese P. Expression of a cleavage site on eIF-4 gamma. J Virol 1994; 68: 5677–5684. foreign protein by influenza A virus. J Virol 1994; 68: 4486–4492. 55 Lamphear BJ, Kirchweger R, Skern T, Rhoads RE. Mapping of 40 de Felipe P et al. Use of the 2A sequence from foot-and-mouth functional domains in eukaryotic protein synthesis initiation disease virus in the generation of retroviral vectors for gene factor 4G (eIF4G) with picornaviral proteases. Implications for therapy. Gene Therapy 1999; 6: 198–208. cap-dependent and cap-independent translational initiation. J 41 De Felipe P, Izquierdo M. Tricistronic and tetracistronic retrovi- Biol Chem 1995; 270: 21975–21983. ral vectors for gene transfer. Hum Gene Ther 2000; 11: 1921–1931. et al 42 Palmenberg AC et al. Proteolytic processing of the cardioviral 56 Piccone ME . Identification of the active-site residues of the P2 region: primary 2A/2B cleavage in clone-derived precursors. L proteinase of foot-and-mouth disease virus. J Virol 1995; 69: Virology 1992; 190: 754–762. 4950–4956. 43 Games D et al. Alzheimer-type neuropathology in transgenic 57 Lamphear BJ et al. Mapping the cleavage site in protein syn- mice overexpressing V717F beta-amyloid precursor protein. thesis initiation factor eIF-4 gamma of the 2A proteases from Nature 1995; 373: 523–527. human Coxsackievirus and rhinovirus. J Biol Chem 1993; 268: 44 Masliah E et al. Dopaminergic loss and inclusion body formation 19200–19203. in alpha-synuclein mice: implications for neurodegenerative dis- 58 Hemmi S et al. The presence of human coxsackievirus and aden- orders. Science 2000; 287: 1265–1269. ovirus receptor is associated with efficient adenovirus-mediated 45 Fan DS et al. Behavioral recovery in 6-hydroxydopamine- transgene expression in human melanoma cell cultures. Hum lesioned rats by cotransduction of striatum with tyrosine Gene Ther 1998; 9: 2363–2373. hydroxylase and aromatic L-amino acid decarboxylase genes 59 Donello JE, Loeb JE, Hope TJ. Woodchuck hepatitis virus con- using two separate adeno-associated virus vectors. Hum Gene tains a tripartite posttranscriptional regulatory element. J Virol Ther 1998; 9: 2527–2535. 1998; 72: 5085–5092. 46 Ryan MD et al. A model for nonstoichiometric, cotranslational 60 Zolotukhin S et al. Recombinant adeno-associated virus purifi- protein scission in eukaryotic ribosomes. Bioorg Chem 1999; 27: cation using novel methods improves infectious titer and yield. 55–79. Gene Therapy 1999; 6: 973–985. 47 Mattion NM, Harnish EC, Crowley JC, Reilly PA. Foot-and- 61 Snyder RO, Xiao X, Samulski RJ. Production of recombinant mouth disease virus 2A protease mediates cleavage in attenu- adeno-associated viral vectors. In: Dracopoli NC et al (eds). Cur- ated Sabin 3 poliovirus vectors engineered for delivery of rent Protocols in Human Genetics. John Wiley: New York, 1997, foreign antigens. J Virol 1996; 70: 8124–8127. pp 12.1.1.–12.1.20.

Gene Therapy