Gene Therapy (1999) 6, 198–208  1999 Stockton Press All rights reserved 0969-7128/99 $12.00 http://www.stockton-press.co.uk/gt Use of the 2A sequence from foot-and-mouth disease in the generation of retroviral vectors for gene therapy

P de Felipe1, V Martı´n1, ML Corte´s1, M Ryan2 and M Izquierdo1 1Departamento de Bioquı´mica y Biologı´a Molecular, Centro de Biologı´a Molecular Severo Ochoa, Universidad Auto´noma de Madrid, Facultad de Ciencias, Cantoblanco 28049, Madrid, Spain; and 2University of St Andrews, Department of Biochemistry, St Andrews, Fife, UK

We describe the construction of retroviral plasmid vectors lation gives rise to a bicistronic mRNA and two inde- in which two genes are linked by a minimum of 96 nucleo- pendent protein products. The system offers advantages tides encoding the 2A sequence from the foot- to other alternative ways to create polycistronic mRNAs and-mouth disease virus (FMDV). Transcription and trans- and can be used in gene therapy delivery vectors.

Keywords: bicistronic retroviral vectors; FMDV 2A; pac; HSV1tk; gene therapy

Introduction cap-binding complex (eIF-4F) used in the translation of the vast majority of ‘capped’ cellular mRNAs. This p220 Several systems have been described to co-express two breakdown results in the ‘shut-off’ of host cell protein or more genes; splicing, internal promoters, internal ribo- synthesis. Initiation of translation of the viral RNA on some entry sites (IRES), fusion proteins, etc. All these sys- the other hand is directed by the IRES (Figure 1a) and, tems are being used by different wild-type and therefore, is not inhibited by the p220 cleavage. have been introduced by genetic engineering in retro- In contrast, in the cardio- and aphthoviruses the cleav- 1 virus and other contexts for various applications. Never- age site between the protein and replicative pro- theless, there remain several other viral mechanisms, not tein regions occurs at the 2A/2B junction. The car- yet exploited by genetic engineering, that could be used diovirus 2A region is about 140 amino acids long whilst in the construction of viral vectors. In this article, we aphthovirus 2A is only 18 amino acids long. The N- describe a strategy based on the incorporation of the 2A terminal end of cardiovirus 2A is dispensable and its sequence from FMDV. function is unknown.5–7 The C-terminal region of car- have a single-stranded RNA diovirus 2A is, however, highly similar to the (shorter) 7.5–8.5 kb long. The viral RNA has an oligopeptide (Vpg aphthovirus 2A. In this region, the last three residues of Ј or 3B) covalently attached to the 5 terminus. The genome 2A (Asn-Pro-Gly) and the N-terminal residue of 2B (Pro) Ј is organised into three regions: (1) a long 5 non-coding are completely conserved as part of the consensus octa- region; (2) a single, long open reading frame encoding a mer Asp-Val/Ile-Glu-X-Asn-Pro-Gly and Pro.3 This Ј polyprotein of some 225 kDa; and (3) a short 3 non- sequence is extremely rare within current databases, the coding region with a poly(A) tail. Picornavirus polypro- only examples are from cardiovirus, aphthovirus and teins undergo co-translational cleavages to produce the group C porcine rotaviruses and is always associated 2–4 primary processing products P1, P2, P3. Subsequent with a ‘cleavage’ activity.8 Mutations in this sequence for proteolytic processing of these precursors gives rise to EMCV (encephalomyocarditis virus) impair cleavage at the ‘mature’ proteins (Figure 1a). The position of picorna- the glycyl-prolyl site.5,8 There are several theories to virus proteins within the polyprotein is indicated by the explain aphto and cardiovirus 2A-mediated cleavage.3,4 domain (1, 2 or 3) and by the position within the individ- The first proposes that 2A could be an entirely novel ual domain (A, B, C or D). oligopeptide autoproteinase. The second that 2A could In the case of two major genera of the picornaviruses, be the target for an unknown cellular proteinase (whose the enterovirus and rhinoviruses, 2A is a proteinase that activity should be tightly coupled to translation); and cleaves at its own N-terminus. In this way, the P1 (capsid thirdly that aphtho- and cardiovirus 2A could function proteins) and P2 (replicative proteins) regions of the by impairing the normal peptide bond formation, polyprotein are separated by processing at the 1D/2A although allowing translation to proceed.4 Many obser- pro site. The 2A proteinase (2A ) is also able to cleave the vations support the latter hypothesis – in particular an p220 cellular component (also called eIF-4G) from the over-accumulation of the translation product N-terminal of 2A with respect to the C-terminal of 2A (not due to differential protein degradation rates; Donnelly MLL, Correspondence: M Izquierdo Gani D and Ryan MD, personal comunication). This Received 5 June 1998; accepted 16 September 1998 hypothesis proposes ribosomes may either continue Retroviral vectors using the 2A sequence from FMDV P de Felipe et al 199

Figure 1 Genomic structure of (a) aphthovirus FMDV; arrows represent primary polyprotein cleavage sites and virus proteins responsible. The 5Ј non- coding region is capped by the oligopeptide Vpg (3B). A poly(C) tract and an IRES (responsible for the translation initiation of the single open reading frame) are located within the 5Ј non-coding region. There is also a 3Ј untranslatable region that ends with a poly(A) tail. (b) pMD2, pTG394 and details of 2A variant sequences, shaded box represents FMDV derived sequences: 2A or ⌬1D2A. Underlined restriction enzymes are unique sites in the plasmid. (c) 2A sequence included in pMD2 and (d) ⌬1D2A sequence cloned in pTG394. In (c) and (d): rectangular boxes represent the flanking genes cat and gus; rounded rectangular boxes the FMDV-derived regions 2A and ⌬1D. The first codon downstream of the 2A region (-CCC-, Pro) is the first codon of the 2B sequence. Underlined sequences are restriction sites. Corresponding amino acids are shown over the nucleotide sequence. Arrows represent 2A–2B cleavage site. In (c), the CAG codon upstream to the 2A region is the last codon derived from the 1D sequence (aa, amino acids; nt, nucleotides). (*) There are two variants of this plasmid: one with the 2A sequence carrying a HindIII site, and the other with the ⌬1D2A sequence that has an AgeI site at the end of ⌬1D sequence but lacks the HindIII site. translation, or dissociate at the 2A region such that lower cleavage activity was also used in the construction of levels of downstream translation products are synthe- recombinant influenza virus (17 amino acid insertion)12 sized. Aphthoviruses seem to contain the minimum and foamy virus (14 amino acid insertion).13 FMDV 2A length of 2A region necessary for this ‘cleavage’. In cleavage activity was detected when these viruses were FMDV, the protease that induces cleavage of p220 is the introduced in Madin–Darby bovine kidney cells leader protein (Lpro; Figure 1a) and not 2A, as is the case (MDBK)12 and baby hamster kidney cells (BHK-21),13 for enterovirus and rhinovirus.9,10 The small 2A region respectively. has been used successfully in heterologous contexts,11 In this article, we describe the construction of a new obtaining Ͼ80% cleavage activity in rabbit reticulocyte kind of bicistronic retroviral vector using the 2A region lysate when this region was cloned between the reporter of FMDV. It represents an alternative to systems cur- genes cat (cholramphenicol acetyltransferase) and gus (␤- rently used for co-expression of two proteins in retroviral glucuronidase). 2A-mediated cleavage was demonstrated vectors. In the present procedure, two genes are cloned in human HTK-143 cells.11 Using similar constructions to flank the 2A sequence forming a single open reading (Figure 1b), other authors6 obtained 85% cleavage frame. The cleavage of the polyprotein product takes between cat and gus after introducing FMDV 2A (19 place at the C-terminal end of the 2A region, leaving this amino acids, Figure 1c). The percentage could be peptide fused to the upstream protein and releasing the increased to 99% when the fragment FMDV 2A also downstream protein intact (with the addition of a N- included the 39 amino acids at the C-terminal end of 1D terminal Pro). (Figure 1d). These cleavages were not detected in equiv- We also describe a new chimeric gene in which the alent constructions expressed in E. coli, showing ‘cleav- expression of a dominant positive puromycin selectable age’ specificity for eukaryotic systems alone.6 FMDV 2A marker gene pac (puromycin N-acetyl transferase) is Retroviral vectors using the 2A sequence from FMDV P de Felipe et al 200 fused in-frame with the dominant negative ganciclovir for LP-⌬1D2A-XIT with a little lower value). Supernatant (GCV) selection gene HSV1tk (herpes simplex virus type derived from the fusion LPT also showed a high titer. 1 thymidine kinase). Expression of the genes cloned downstream from 2A or ⌬1D2A Results HSV1tk expression was estimated by GCV sensitivity in the different cell lines. In transfected ⌿CRE producer

Construction of 2A-containing retroviral vectors cells, the concentration at which 50% of the cells die (IC50) Plasmids pMD2 (containing the FMDV 2A region was, in all cases, lower than 1 ␮g/ml (Figure 3a). Bicis- between the cat and gus genes) and pTG394 (containing tronic and putative tricistronic vectors with 2A or ⌬1D2A a ⌬1D2A insert) were used as starting points for sub- showed good sensitivities to GCV, in a range similar to sequent molecular biological manipulations (Figure 1b– that observed for the LPIT and LTSP control vectors; bici- d).6 The cat gene was substituted by the 3Ј end of pac stronic vectors showing slightly better results than puta- amplified by PCR (and removing the stop codon) to tive tricistronics. ⌿CRE cells transfected with the fusion obtain pMD2pac and pTG394pac (Figure 2a). From these LPT also showed a good sensitivity. The GCV sensitivity plasmids and the vector pLPIT (Figure 2b) universal of infected ⌿CRIP producer cells was also determined putative tricistronic constructions pLP-2A-XIT and pLP- (Figure 3b). The results were similar to those obtained ⌬1D2A-XIT (where the X position can be filled by a third previously, however, an improvement from transfected gene to become tricistronic) were obtained (Figure 2c). ⌿CRE cells can be observed. All bicistronic and putative Putative tricistronic vectors have a stop codon before the tricistronic vectors with 2A or ⌬1D2A showed good GCV SalI site (in-frame with 2A sequence) – a small peptide sensitivities – similar to the positive controls. Again, bici- being produced (see amino acids sequence in Materials stronic constructions were slightly better (IC50 from LP- and methods). The removal of IRES from these two con- 2A-T is 5 × 10−2 ␮g/ml and from LP-⌬1D2A-T, 7 × 10−2) × −2 ␮ structions gives rise to the bicistronic vectors pLP-2A-T than putative tricistronic (IC50 :13 10 g/ml for both and pLP-⌬1D2A-T (Figure 2d). In addition, we have con- LP-2A-XIT and LP-⌬1D2A-XIT) and the fusion LPT gave × −1 ␮ structed a true tricistronic vector with the insertion of high drug sensitivity (IC50:2 10 g/ml). enhanced green fluorescent protein (EGFP) upstream of In order to obtain a better approximation of 2A and the IRES-HSV1tk cassette eliminating the stop codon, ⌬1D2A cleavage efficiency, we analyzed the protein pro- pLP-⌬1D2A-GIT (Figure 2c). ducts derived from these vectors using antibodies against The retroviral plasmid pLPT was constructed with a HSV1TK by Western blotting (Figure 4a). In LPIT and direct fusion between pac and HSV1tk genes (Figure 2e). LTSP we detected a single HSV1TK band migrating at The vector pLPX, used as control together with pLPIT the same position and with similar intensity (lanes 2 and and pLTSP, includes only the gene pac (Figure 2f). pLPIT 3). In the LPIT vector (lane 2) and the putative tricistronic has an IRES-HSV1tk cassette downstream pac gene (lanes 4 and 5), HSV1TK translated from the IRES (Figure 2b). pLTSP was constructed introducing HSV1tk (originally from EMCV) whilst LTSP vector (lane 3) uses gene upstream of the pac gene. Transcription of pac gene the 5Ј long terminal repeat (LTR) for initiation of tran- is directed from SV40 promoter (Figure 2g). scription. The same HSV1TK protein band, migrating at the same position is observed in the case of the bicistronic Expression of the gene upstream of 2A or ⌬1D2A, in vectors (lanes 6 and 7). We concluded therefore, that the retroviral producer cells 2A and ⌬1D2A sequences efficiently ‘cleave’ the artificial Murine ecotropic retroviral packaging cells ⌿CRE were polyprotein. The high molecular weight HSV1TK- transfected and their supernatants were used to infect 2A/⌬1D2A-PAC fusion protein could not be detected amphotropic retroviral packaging cells ⌿CRIP.14 Trans- from these constructs. There is a non-specific band that fection efficiencies were similar in all cases (Table 1). can be seen also in the negative control with LPX vector Transfected and infected retroviral producer cells were infected cells (lane 1) in a position close to, but migrating selected by growing them in 2 ␮g/ml of puromycin for slightly slower than, the PAC-HSV1TK fusion protein about 3 weeks. All cells grew well despite the C-terminal (lane 8). The weaker HSV1TK band in lane 8 could be extension of 2A or ⌬1D2A to the PAC protein. The same explained by a slight instability of fusion protein and, effect was observed when estimating the titer of super- perhaps, by the method of sample preparation before natants harvested from producer cells by infection of Western blotting. NIH/3T3 and cultured in 2 ␮g/ml puromycin. Titers In the tricistronic vector in wich EGFP is placed after from ⌿CRE transfected cells were clearly lower (in the ⌬1D2A, we were able to detect green fluorescence in a range of 0.5–1.3 × 104 colony forming units/ml), than tit- significant number of cells, when a lawn of ⌿CRIP- ers of ⌿CRIP infected cells (ranging from 1.2 to 3.4 × 106 infected cells (selected for 1 month in 2 ␮g/ml c.f.u./ml) (Table 1). ⌿CRE titers were similar in all cases puromycin) was subjected to flow cytometry (Figure 4b). except those derived from bicistronic vectors which This shows that both PAC and EGFP proteins are always produced lower values than the others. In general expressed from this construct. we can confirm that transfected cells have lower titers than infected cells.14 We have also observed that ⌿CRE Behavior of retroviral producer cells after long-term cells have more difficulty in reaching confluence than culture in selective medium ⌿CRIP cells. ⌿CRIP titers were well within the range of The present study was completed with long-term culture normal values using retroviral vectors. LPIT and LTSP of the producer cells in order to determine whether the showed lower titers (about half) than the monocistronic new retroviral constructs retained the original titer and control LPX. Bicistronic and putative tricistronic vectors drug sensitivity. ⌿CRIP/LP-2A-T, ⌿CRIP/LPIT and with 2A or ⌬1D2A showed similar values to LPX (except ⌿CRIP/LTSP producer cells were cultured under con- Retroviral vectors using the 2A sequence from FMDV P de Felipe et al 201

Figure 2 Structure of the different plasmid vectors constructed: (a) pMD2pac and pTG394pac; (b) pLPIT (plasmid, LTR, pac, IRES, HSV1tk); (c) pLP- 2A-XIT (plasmid, LTR, pac, 2A, undetermined gen, IRES, HSV1tk) and pLP-⌬1D2A-XIT, (plasmid, LTR, pac, ⌬1D2A, undetermined gen, IRES, HSV1tk); (d) pLP-2A-T (plasmid, LTR, pac, 2A, HSV1tk) and pLP-⌬1D2A-T (plasmid, LTR, pac, ⌬1D2A, HSV1tk) and (e) pLPT (plasmid, LTR, pac, HSV1tk). Retroviral vectors using the 2A sequence from FMDV P de Felipe et al 202 Table 1 Transfection efficiencies of ⌿CRE cells and titers of producer cells ⌿CRE and ⌿CRIP

Vectors Transfection efficiencies Titers of producer cells

⌿CRE cells Transfected Infected (colonies/␮g DNA) ⌿CRE cells ⌿CRIP cells (c.f.u./ml) (c.f.u./ml)

LP-2A-T 2.6 × 104 0.5 × 104 2.7 × 106 LP-⌬1D2A-T 3.4 × 104 0.8 × 104 2.6 × 106 LP-2A-XIT 2 × 104 1.2 × 104 3.4 × 106 LP-⌬1D2A-XIT 3 × 104 1.3 × 104 1.7 × 106 LTSP 2.1 × 104 1.2 × 104 1.2 × 106 LPIT — — 1.2 × 106 LPT 2.2 × 104 1.3 × 104 2.8 × 106 LPX 4.1 × 104 1.1 × 104 2.4 × 106

Figure 3 (a) Sensitivity to GCV (expressed as percentage cell survival) for ⌿CRE retroviral producer cell populations. ⌿CRE cells were transfected with different retroviral vectors and selected 1 month in 2 ␮g/ml puromycin. (b) Sensitivity to GCV (expressed as percent cell survival), for ⌿CRIP retroviral producer cell populations. ⌿CRIP cells were infected with supernatants from transfected ⌿CRE producer cell populations and selected 1 month in 2 ␮g/ml puromycin. As the experiments are performed, it is not very unusual to obtain survival values higher in the low-concentration drug wells than in the samples without drug of a particular curve (see Materials and methods). Retroviral vectors using the 2A sequence from FMDV P de Felipe et al 203 tinuous selection with 2 or 10 ␮g/ml of puromycin. After 2 months selection in 2 ␮g/ml puromycin (from initial transfection–infection), cultures were divided in two parts. One part was maintained at the same selection con- centration while the other half was increased to 10 ␮g/ml puromycin. After 1 more month, titration of producer cells was repeated (except for ⌿CRIP/LTSP which, due to slow growth, was left for 2 additional weeks before supernatants were harvested) and GCV sensitivity determined. The data presented in Table 2 shows that long-term culture (about 3 months) in 2 ␮g/ml puromycin can be correlated with a general maintenance of all titers meas- ured, which remained close to 106 c.f.u./ml. Increasing puromycin selection (10 ␮g/ml) in the third month led to titers similar to those obtained previously from bicis- tronic vectors (with 2A or IRES), whilst LTSP retrovirus producer cells reduced in titer about one log unit (105 c.f.u./ml) from the initial value. Three months culture in 2 ␮g/ml puromycin resulted, however, in a small decrease of GCV sensitivity (LP-2A- × −2 × −2 ␮ T: the IC50 increases from 5 10 to 10 10 g/ml; LPIT: from 13 × 10−2 to 26 × 10−2 ␮g/ml), except for the LTSP vector where the effect was more marked (with × −2 × −2 ␮ increments of the IC50 from 6 10 to 230 10 g/ml; Figure 5). When selection was increased to 10 ␮g/ml of puromycin in the last month, GCV sensitivity was main-

tained at the previously observed values (IC50 LP-2A-T: 5 × 10−2 ␮g/ml; LPIT: 6 × 10−2 ␮g/ml). In the case of the LTSP producer cells, however, the improvement did not × −2 ␮ reach the initial IC50 value (40 10 g/ml). Figure 5 shows that whilst 2A or IRES bicistronic retrovirus pro- ducer cells behaved similarly in long-term culture (they lost GCV sensitivity with 2 ␮g/ml puromycin and reco- vered it with 10 ␮g/ml), the LTSP retrovirus producer cells (with an internal promoter directing pac gene expression) increased their IC50 by more than a half logar- ithmic unit despite increase of puromycin selection. These results suggest that vectors having IRES or 2A maintain higher titer and better sensitivity to the drug, than vectors with internal promoters.

Vector expression in human glioma cells Hs 683 In vectors that may be used for gene therapy of cancer cells, it is interesting to know whether tumoral cells can be efficiently infected as NIH/3T3 derived cells, and whether in vivo situations (infection but no selection) is relatively similar to cell culture conditions in which selec- tion generally precedes gene expression estimates. ⌿CRIP supernatants (LP-2A-T, LP-⌬1D2A-T, LTSP and Figure 4 (a) Western blot using anti HSV1TK antibodies in cell extracts ␮ of different retroviral producer cells (all lanes are from ⌿CRIP producer LPIT harvested after 1 month in 2 g/ml puromycin cells except lane 8, which is from ⌿CRE producer cells). Protein gel load- selection) were used to infect the Hs 683 human glioma ings are 70 ␮g for all lanes except for lane 4 (35 ␮g). HSV1TK is detected cell line. in all lanes except in the negative control ⌿CRIP-LPX lane 1, and in GCV sensitivity, using retrovirus-infected Hs 683 cells, lane 8 where an extract from cells carrying the fusion construct is loaded. was determined 3 days after infection in the absence of In lane 8, it is possible to see PAC-HSV1TK fusion product close to an unspecific band, also present in the negative control (lane 1). (b) Flow puromycin selection (Figure 6a). A lawn of Hs 683 cells cytometry analysis of ⌿CRIP cells infected with pLPX (upper panel) and (a mixture of infected and uninfected cells) was cultured ⌿CRIP cells infected with pLP-⌬1D2A-GIT (lower panel), both selected for 72 h with GCV. In spite of this, the observed IC50 was in 2 ␮g/ml puromycin. Vertical lines show the setting of the electronic lower than 1 ␮g/ml. Note that cells had been cultured in gate to distinguish between EGFP-negative and postive cells. Percentages the presence of GCV 1 day more than in other assays of total cells and fluorescence mean values with standard deviations are since in only 48 h, cell mortality was not so marked. The indicated for each gate. results do not exclude a component related to infection efficiency. In parallel experiments, Hs 683-infected cells selected with 2 ␮g/ml puromycin for 10 days showed significant Retroviral vectors using the 2A sequence from FMDV P de Felipe et al 204 Table 2 Long-term titers of infected retrovirus producer ⌿CRIP cells

Vectors Culture in 2 ␮g/ml puromycin Culture with puromycin: 2 months in 2 ␮g/ml and 1 more ␮ After 1 month After 3 months month in 10 g/ml

LP-2A-T 2.7 × 106 0.9 × 106 3 × 106 LPIT 1.2 × 106 1 × 106 1.4 × 106 LTSP 1.2 × 106 0.6 × 106 0.1 × 106

Figure 5 Sensitivity to GCV, as a percentage of cell survival, for ⌿CRIP retroviral producer cell populations infected with LP-2A-T, LTSP and LPIT vectors after 3 months of culture in different selective conditions (see also Table 2). 2P, 3 months with 2 ␮g/ml puromycin selection; 10P, 2 months with 2 ␮g/ml puromycin selection and 1 month with 10 ␮g/ml puromycin selection. Assay conditions are the same as in Figure 3.

GCV sensitivity after 48 h incubation (Figure 6b). A lower (hygromycin B phosphotransferase) 1 kb, this fusion pro- IC50 value has been obtained without puromycin (Figure tein is smaller than those previously described: HSV1tk- 6a) than with the selection (Figure 6b) provided the drug neo,15,16 hph-HSV1tk.17 The risk with these types of con- (GCV) is kept for 24 h longer in the culture. Nevertheless, structs is a lack of efficiency in one or both functions due ⌿ ⌿ their IC50 is not as low as that of CRE or CRIP pro- to structural changes, or to different cellular location ducer cells (Figure 3a and b). We believe the variations from which one or both gene products normally operate. observed reflect the different origins of the cell lines. In our fusion protein, the pac domain is functional and we only see a small decrease in GCV sensitivity (Table Discussion 1, Figure 3). In our particular case, the cellular location of the expressed protein is not a problem since both pro- Retroviral vectors to be used in gene therapy must ensure teins are normally located in the cytoplasm. The an efficient expression of therapeutic genes, as well as expression of membranous or extracellular proteins, selection markers. When different internal promoters are however, might prove more difficult with such fusion placed in these vectors, interference among them may proteins. cause inactivation of some of the exogenous genes. The The new alternative system we describe here is based presence of an IRES, instead of a second promoter, is a on the introduction of the 2A small region from FMDV well known and widely used alternative for the co- in phase between pac and HSV1tk genes or between pac expression of two genes. We have constructed these two and egfp in the context of a widely used defective retrovi- types of retroviral vectors using the selectable marker pac ral vector derived from Moloney murine leukemia virus and the potential killer suicide HSV1tk gene (Figure 2b (Figure 2). Previous viral constructions carrying 2A and g). Both constructions produced viruses to high titers sequence showed the utility of this small region to cleav- and GCV sensitivities (Table 1, Figure 3). age a polyprotein.12,13 In each case, the recombinant viral A third possibility would be to make a hybrid gene had the marker cat gene linked to another gene with no separation between the two coding regions, that via 2A. In the case of the influenza virus construct, the will give a single protein with a combined activity. Here CAT-2A cassette was inserted N-terminal of the neurami- we have fused pac and HSV1tk to obtain a positive and nidase to form a single CAT2ANA open reading frame negative selection protein. This fusion contained the com- (the 2A fragment used comprises 17 amino acids).12 In plete sequence of both genes (Figure 2e). Because pac the other study, 2A-CAT was inserted C-terminal of Bel- gene is smaller (0.6 kb) than other selection genes: neo 1 in foamy virus again forming a single open reading (neomycin phosphotransferase) 0.8 kb and hph frame. In this virus, a small region of non-essential genes Retroviral vectors using the 2A sequence from FMDV P de Felipe et al 205

Figure 6 Sensitivity to GCV, as a percentage of cell survival, for infected human glioma Hs 683 cells in the absence or presence of selection. Hs 683 cells were infected with four retroviral supernatants, adjusted at the same viral concentration, from cells: ⌿CRIP/LP-2A-T, LP-⌬1D2A-T, LTSP and LPIT for 48 h. (a) Infected Hs 683 cells, cultured 3 days without puromycin selection, were incubated with GCV for 72 h. (b) Infected Hs 683 cells, cultured with 2 ␮g/ml puromycin for 10 days, and then incubated with GCV for 48 h. was deleted to accommodate the insert. The FMDV 2A the activity of the attached gene product PAC (Tables 1 sequence used here was smaller (only 14 amino acids) and 2), or to other gene products such as CAT or GUS than that used in the construction of the bicistronic (data not shown). It should be noted that the second gene influenza virus RNA segment. will posses an N-terminal proline residue. In our con- The 96 nucleotide insert encoding the 2A region (54 structions a sequence coding for several restriction nucleotides plus the codon for the N-terminal proline of enzyme target sites was present at the 5Ј end of this 2B; Figure 1c) or the 201 nucleotide insert carrying the second (downstream of 2A) gene. Nevertheless, activities ⌬1D2A (Figure 1d) is shorter than the SV40 early pro- of HSV1tk (Figures 3 and 5) or EGFP (Figure 4b) were moter (343 nucleotides) used in a variety of retroviral not affected by these additional sequences. One of the vectors.18 Even the IRES, widely used in bicistronic vec- advantages of this bicistronic system is to obtain two dis- tors has a minimal sequence of 422 (poliovirus) or 426 crete protein products (similar to the IRES system). This (EMCV) nucleotides.2 The inserts carrying the IRES may be of paramount importance if the proteins have dif- sequences usually have an addition of 634 nucleotides ferent cellular localizations, a disadvantage of the fusion (pSBC-1)19 or 513 nucleotides (pSXLC-TK used in this protein strategy. We have show that our 2A-containing study).20 In vectors such as retrovirus where the coding vectors direct the synthesis of independent products capacity to express therapeutic genes is not very large (7– (Figure 4a). 8 kb), a small sequence such as 2A could be preferable It is possible to generate a small amount of fused pro- to other alternatives. tein that may or may not conserve the activity of one or Some caution should be employed when using the 2A both proteins. In vitro studies have estimated a cleavage system. The two cloned genes must be in a single open of Ͼ85% for 2A and 99% for ⌬1D2A.6 We have compared reading frame (eliminating the stop codon from the end the presence and activity of pac and HSV1tk when placed of the first gene). Additionally, following cleavage the 2A before and after both the short (2A) or long (⌬1D2A) 2A peptide remains attached to the first gene product as a sequences and found no significant differences. The C-terminal extension. This has not been a limitation for results are also comparable to the activities obtained for Retroviral vectors using the 2A sequence from FMDV P de Felipe et al 206 these gene products (pac and HSV1tk) using the SV40 These plasmids were digested with SacII and BglII to internal promoter or the IRES from the EMCV (Table 1 obtain 3Ј pac-2A and 3Ј pac-⌬1D2A fragments that were and Figure 3). inserted in vector pLPIT (Figure 2b) digested in the same The design of effective retroviral vectors and their way. Resulting vectors are putative tricistronic construc- possible use in vivo requires provirus stability. There are tions pLP-2A-XIT and pLP-⌬1D2A-XIT (Figure 2c). These several factors affecting long-term functional and struc- putative tricistronic constructions contain two genes and tural stability such as the nature of the infected cells, the a small polylinker free (with unique BglII and SalI restric- integration site of the provirus, the reporter gene and the tion sites) for cloning the second gene downstream 2A or presence or absence of selection pressure. Activity of ⌬1D2A. Before the SalI site, there is a stop codon in the HSV1tk was long-term maintained for at least 3 months polylinker (in frame with 2A sequence) delimiting a short (in the presence of puromycin selection) but we have peptide (predicted sequence: Pro-Phe-Phe-Phe-Thr-Ser- observed that the presence of an internal promoter before Arg-Ser-Glu-Leu-Leu). the selection gene (probably valid also for the reporter IRES within these constructs were removed (digesting gene) is likely to producer a higher instability than the with BglII and NcoI and blunting the ends with Klenow presence of 2A or IRES (Table 2, Figure 5). fragment) during the production of the bicistronic vectors The 2A system does not work in E. coli,6 but it works pLP-2A-T and pLP-⌬1D2A-T (Figure 2d). The NcoI site is well in eukaryotic cell culture from organisms as diverse reformed after ligation and can be used to remove the as insects (Spodoptera frugiperda)21 or mammalian cell lines HSV1tk gene. (HTK-143,11 MDBK,12 BHK-2113). In this work we have Tricistronic vectors were constructed as follows. EGFP demonstrated the correct functioning of the 2A system in gene was excised from pEGFP-N1 (Clontech, Palo Alto, other mammalian cells (⌿CRE, ⌿CRIP, NIH/3T3), CA, USA) with BamHI and HpaI and inserted into pBa- including human glioma cell line Hs 683 (Figure 6). bepuro digested with BamHI and SnaBI. This plasmid The small and efficient 2A sequence can be used in was then digested with BamHI and SalI, and the resulting combination with an IRES for the construction of tricis- EGFP fragment ligated into plasmid pLT-⌬1D2A-XIT tronic vectors eliminating the need for internal promoters (Figure 2c). Finally, this construct was cleaved with NcoI that may interfere with the potent LTR initial promoter and the resulting fragment containing the EGFP-IRES (eg LP-⌬1D2A-GIT; Figure 2c). Two different IRESs (one sequence was inserted into plasmid pLP-⌬1D2A-T from ECMV and the other from poliovirus) can also be (Figure 2d) giving rise to the construct pLP-⌬1D2A-GIT used to reduce the chance of putative recombination that (Figure 2c). could get rid of the internal gene; nevertheless it has been The vector pLPT (Figure 2e) contains a direct fusion shown that two IRESs close together may cause a between pac and HSV1tk genes in phase. Plasmid pLP- decrease in the expression of the coupled genes.22 ⌬1D2A-XIT was digested with SpeI and XbaI and the The 2A system appears to be a useful complement or restriction fragment containing the pac sequences pur- alternative to an IRES in the construction of bicistronic ified. This fragment was ligated into plasmid pLP-2A-T or higher order cistronic vectors. The vectors we describe restricted with SpeI and dephosphorylated with alkaline here could be useful as an alternative in gene therapy phosphatase, producing plasmid pLPT. The amino acid constructs, especially tricistronic systems for the sequence between pac and HSV1tk is, therefore, Ala-Ser- expression of heterodimeric proteins, such as certain Ser-Arg-Ser-Met (Ala being the last pac-derived amino cytokines.23 Improvements in the design of retroviral acid and Met the first HSV1tk-derived amino acid). vectors similar to the ones we propose here, offer the Vector LPX (Figure 2f) was obtained deleting the SV40 possibility to improve the efficiency of gene therapies. promoter (SalI/HindIII) and the polylinker (BamHI/EcoRI) from pBabepuro, and introducing in the ClaI site, the polylinker from plasmid pCla 12,24 also Materials and methods digested with ClaI. The retroviral plasmid pLPIT (Figure 2b) was derived Construction of recombinant plasmids from pBabepuro deleting the SV40 promoter The backbone of all retroviral plasmids was taken from (SalI/HindIII) and inserting IRES-HSV1tk cassette down- pBabepuro.18 HSV1tk and IRES sequences were obtained stream of the pac gene via the ClaI site using the adaptor from pSXLC-TK20 and FMDV sequences from plasmids plasmid pCla 12. pMD2 and pTG394.6 Finally, vector pLTSP (Figure 2g) was constructed Plasmid pMD2 contains the FMDV 2A sequence introducing the HSV1tk gene (excised via BamHI and inserted between the cat and gus genes, and plasmid XhoI from pTK) into the pBabepuro polylinker. Plasmid pTG394 contains ⌬1D2A insert in the same position pTK was obtained deleting IRES from pSXLC-TK with (Figure 1b–d). Regions 2A and ⌬1D2A are delimited by enzymes XbaI and NcoI. In pTK and pLTSP, the HSV1tk XbaI and ApaI restriction sites. Downstream from the gene has an initial ATG codon in a context (AGCATGG) ApaI site, and before the gus initiating ATG codon, there similar to the consensus Kozak sequence. are two unique sites (SpeI and BglII) that can be used for engineering constructs. The 3Ј region from pac was obtained by PCR and Cell culture cloned starting at the SacII restriction site (BamHI site was Murine fibroblasts NIH/3T3, ecotropic packaging cell introduced upstream for convenience). The stop codon of line ⌿CRE and amphotropic packaging cell line ⌿CRIP the pac gene was removed and an XbaI site introduced. were grown in Dulbecco’s modified Eagle’s medium The BamHI/XbaI pac fragment replaced the cat gene (DMEM) supplemented with 10% heat inactivated new- within the previous constructions to obtain pMD2pac and born calf serum. Human glioma cell line Hs 683 was pTG394pac (Figure 2a). grown in DMEM supplemented with 10% heat inacti- Retroviral vectors using the 2A sequence from FMDV P de Felipe et al 207 vated fetal calf serum. All cells were cultured at 37°C, 7% Flow cytometry ⌿ CO2 and 97% relative humidity. CRIP cells were infected with viruses released from ⌿CRE transfected cells with tricistronic EGFP containing ⌿ Transfection, infection and titration vector. These CRIP-infected cells were selected with 2 ␮g/ml of puromycin for 1 month. Cells were trypsinized, Producer cells were prepared by the transfection of eco- resuspended in DMEM supplemented with 1% heat inac- tropic packaging ⌿CRE cells, and subsequently using the tivated newborn calf serum and analyzed by cytometry supernatant to infect amphotropic packaging ⌿CRIP cells with EPICS Profile II Analyzer (Coulter Electronics, according to a previously described method.14 Trans- Hialeah, FL, USA). fected ⌿CRE cells and infected ⌿CRIP cells were selected in 2 ␮g/ml puromycin and colonies obtained were pooled. Acknowledgements To determine efficiency of transfection 48 h after trans- fection, cells were split 1:10 and cultured in 2 ␮g/ml pur- We gratefully acknowledge Jean-Michel Heard for the ⌿ ⌿ omycin. Resistant colonies were counted and results packaging cell lines CRIP and CRE, Hartmut Land for expressed as number of colonies per microgram DNA the plasmid pBabepuro, Ira Pastan for plasmid pSXLC- (the number of cells when selection was initiated was TK, Stephen H Hughes for plasmid pCla 12 and Juan A 106).14 de Carlos for plasmid pEGFP-N1. Anti-HSV1tk anti- Titers were estimated by infecting NIH/3T3 cells as bodies were a gift from Steven M Albelda and Amin described previously.25 Each supernatant was tested Kunjlata. The work was supported by the Plan Nacional once. de Salud (PNICyD, grant number 96–0037) from the Spanish Government, and by the local government: Comunidad de Madrid (grant number 8.1/0006/1997). Retroviral infection of Hs 683 cells The Centro de Biologı´a Molecular is the recipient of an × 3 Hs 683 human glioma cells (5 10 ) were infected in a institutional grant from the Fundacio´n Ramo´n Areces. PF × 6 60-mm plate with 1.2 10 retrovirus in 1 ml of culture is supported by a FPI grant from the Comunidad Auto´- ␮ medium with 8 g/ml of polybrene for 5 h. After that, noma de Madrid (Consejerı´a de Educacio´n y Cultura) another 0.5 ml of medium were added and cells were and VM has an FPI grant from the Ministerio de Educa- cultured for 48 h, before being split in two plates, one cio´n y Ciencia. with 2 ␮g/ml of puromycin selection and the other with- out selection. Resulting populations were used for sub- sequent experiments, after 10 days in the first case and 3 References days in the second. 1 Brown AMC, Dougherty JP. Retroviral vectors. In: Glover DM, Hames BD (eds). DNA Cloning, vol 4. Oxford University Press: GCV sensitivity Oxford, 1995, pp 113–142. Cells were seeded (5 × 103 per well) in a 96-well plate. 2 Wimmer E et al. Genetics of poliovirus. Annu Rev Genet 1993; Following 24 h incubation, increasing GCV concen- 27: 353–436. ␮ 3 Palmenberg AC. Proteolytic processing of picornaviral polypro- trations were added in a final volume of 200 l of normal tein. Ann Rev Microbiol 1990; 44: 603–623. medium. Two days later, the percentage of surviving 4 Ryan MD, Flint M. Virus-encoded proteinases of the picorna- cells was determined to evaluate GCV sensitivity. Cells virus super-group. J Gen Virol 1997; 78: 699–723. were fixed with 12% glutaraldehyde, stained with 0.05% 5 Palmenberg AC et al. Proteolytic processing of the cardioviral methylene blue and eluted in 0.33 N HCl. Absorbance at P2 region: primary 2A/2B cleavage in clone-derived precursors. 630 nm was determined in a MR 5000 microplate reader Virology 1992; 190: 754–762. (Dynatech, West Sussex, UK). Assays were performed by 6 Donnelly MLL et al. The cleavage activities of aphthovirus and triplicate and the estimated mean value with standard cardiovirus 2A proteins. J Gen Virol 1997; 78: 13–21. deviations presented in survival curves. Cells grown 7 Michiels T et al. Protein 2A is not required for Theiler’s virus replication. J Virol 1997; 71: 9549–9556. without GCV were considered as 100% cell survival. In 8 Hahn H, Palmenberg AC. Mutational analysis of the encephalo- these conditions, it is not unusual to obtain survival myocarditis virus primary cleavage. J Virol 1996; 70: 6870–6875. values higher in the low-concentration drug wells than 9 Devaney MA et al. Leader protein of foot-and-mouth disease in the samples without the drug of a particular curve. virus is required for cleavage of the p220 component of the cap- binding protein complex. J Virol 1988; 62: 4407–4409. 10 Medina M et al. The two species of the foot-and-mouth disease Western blotting virus leader protein, expressed individually, exhibit the same Retrovirus producer cells were sonicated and total pro- activities. Virology 1993; 194: 355–359. tein concentration was determined by BCA Protein Assay 11 Ryan MD, Drew J. Foot-and-mouth disease virus 2A oligopep- (Pierce, Rockford, USA). Proteins were analyzed by 12% tide mediated cleavage of an artificial polyprotein. EMBO J 1994; SDS-PAGE and electroblotted on to a nylon membrane. 13: 928–933. The membrane was blocked and incubated with a 1:2000 12 Percy N et al. Expression of a foreign protein by influenza A dilution of anti-rHSV1TK polyclonal antiserum. After virus. J Virol 1994; 68: 4486–4492. extensive washing, the membrane was incubated with a 13 Schmidt M, Rethwilm A. Replicating foamy virus-based vectors 1:10 000 dilution of HRP labelled anti-rabbit antibody. directing high level expression of foreing genes. J Virol 1995; 210: 167–178. The blot was washed again and the distribution of HRP- 14 Miller D, Rosman GJ. Improved retroviral vectors for gene labelled anti-rabbit antibody determined using the transfer and expression. BioTechniques 1989; 7: 980–988. Enhanced ChemiLuminiscence system (Amersham, Little 15 Schwartz F et al. A dominant positive and negative selectable Chalfont, UK). The signal were imaged by 10-min gene for use in mammalian cells. Proc Natl Acad Sci USA 1991; exposure. 88: 10416–10420. Retroviral vectors using the 2A sequence from FMDV P de Felipe et al 208 16 Veelken H et al. Systematic evaluation of chimeric marker genes proteins in insect cells using baculovirus expression vectors. J on dicistronic transcription units for regulated expression of Gen Virol 1990; 71: 1703–1711. transgenes in vitro and in vivo. Hum Gen Ther 1996; 7: 1827–1836. 22 Morgan RA et al. Retroviral vector containing putative internal 17 Lupton SD et al. Dominant positive and negative selection using ribosome entry sites: development of a polycistronic gene trans- a hygromycin phosphotransferase-thymidine kinase fusion fer system and applications to human gene therapy. Nucleic Acid gene. Mol Cell Biol 1991; 11: 3374–3378. Res 1992; 20: 1293–1299. 18 Morgenstern J, Land H. Advanced mammalian gene transfer: 23 Tahara H et al. Effective eradication of established murine high titre retroviral vectors with multiple drug selection mark- tumors with IL-12 gene therapy using a polycistronic retroviral ers and a complementary helper-free packaging cell line. Nucleic vector. J Immunol 1995; 154: 6466–6474. Acid Res 1990; 18: 3587–3596. 24 Hughes SH et al. Adaptor plasmid simplify the insertion of 19 Dirks W et al. Dicistronic transcription units for gene expression foreign DNA into helper-independent retroviral vectors. J Virol in mammalian cells. Gene 1993; 128: 247–249. 1987; 61: 3004–3012. 20 Sugimoto Y et al. Efficient expression of drug-selectable genes 25 Izquierdo M et al. Gene therapy in brain tumours: implication in retroviral vectors under control of an internal ribosome entry of the size of glioblastoma on its curability. Acta Neurochir 1997; site. Bio/Technology 1994; 12: 694–698. 68 (suppl): 111–117. 21 Roosien J et al. Synthesis of foot-and-mouth disease virus capsid