Currently there are several types of gene delivery vectors available to deliver one or two genes into target cells. An increasing demand for more complex multicistronic vectors has arisen in recent years for various applications both in basic research and clinical gene therapy. Herein we described a new method to coexpress multiple transgenes efficiently in HIV-1 based lentiviral vectors. We constructed bicistronic and tricistronic lentiviral vectors using combinations of a self-processing 2A cleavage factor and IRES and undertook systematic analysis of the expression of selected marker genes. In this report we describe bicistronic and tricistronic lentiviral vectors. These multigene vectors can successfully co-express 2 or 3 transgenes under the direction of a single promoter. All the vectors described in this study produced high titer vector stocks comparable to the monocistronic vectors. They were also able to transduce multiple target cells of human and murine origin efficiently. However, there were differences in the level of transgene expression among the vectors depending on the size, position and total number of transgenes placed within the expression cassette; and type of transgene involved. Bicistronic vectors based on the 2A cleavage factor were more efficient in the co-expression of two transgenes than IRES based vectors. Indeed, co-expression mediated by the 2A motif was superior to internal ribosome entry across a range of different vector MOIs, and it is of import that this differential was maintained at a limiting copy number. Thus, 2A represents an attractive alternative to currently used systems for the co-expression of two proteins in lentiviral vectors.
A major advantage of using the 2A cleavage factor in the construction of multicistronic vectors is its small size compared to internal promoters or IRES sequences. Given the packaging constraints on lentiviral vectors, minimizing the size of sequences required to enable co-expression is important in maximizing the capacity for therapeutic sequences. In addition, efficient co-expression of both genes is ensured as we have shown in the case of MGMT-2A-eGFP. The 2A sequence efficiently promoted the generation of predicted cleavage products from the artificial fusion protein in transduced cells. Previous studies with oncoretroviral [13, 23, 24] and AAV  vectors have shown the feasibility of using the 2A sequence for the expression of multiple transgenes. Incomplete cleavage of 2A mediated fusion products has previously been reported in AAV  and retroviral vectors [12, 25]. In our hands, the efficiency of cleavage was construct dependent, with the MGMT-2A-eGFP cassette leading to some (approximately 6–8%) uncleaved product, whilst those cassettes incorporating HOXB4 showed apparent 100% cleavage. Although the reason for incomplete cleavage remain obscure, it is not unreasonable to speculate that differences in fusion protein secondary structure might influence this.
In addition to efficient generation of cleavage products, it is important that these are transported to the appropriate compartment of the cell where their action is required. As shown by the nuclear localization of HOXB4 and MGMT in our study, the addition of 2A sequences did not adversely affect the trafficking of these two proteins. Recently Szymczak et al  reported the construction of a multicistronic retroviral vector using multiple 2A cleavage factors or similar sequences with efficient coexpression of complete T cell receptor complex proteins. They showed that a 2A like peptide linked retroviral vector could be used to express all of the four CD3 proteins (CD3ε,γ,δ,ζ), appropriately localized to the membrane and that this restored T cell development in CD3 deficient mice. However in another recent report, mistargetting of second gene products was observed dependent on the context in which they were expressed . It will be important; therefore, to empirically test any co-expression cassette to ensure that localization of transgene products is appropriate. Szymczak et al used four separate 2A sequences from different viruses, which share a conserved sequence. To avoid recombination they changed codon usage by introducing silent mutations within 2A sequences. A similar approach in lentiviral vectors might allow efficient delivery of multiple genes linked with multiple 2A cleavage factors without the need to use IRES sequences. However, whether or not recombination would be a problem if identical sequences were used, may be worth establishing.
One particular attraction of this 2A-based strategy is in applications in which it is desirable to coexpress two or more therapeutic genes in comparable amounts as in the case of two subunits of a functional protein (e.g. enzyme, cytokines). Previously described lentiviral vectors based on IRES or multiple internal promoters  have revealed inconsistent levels of expression of individual transgenes within the expression cassette. From our data summarized in Table 2, it is clear that the relative levels of MGMT and eGFP expression from the bicistronic 2A-based vector were higher than IRES based vector. In contrast, expression of eGFP from the IRES-containing vector was around one fifth that of MGMT. Although this is an improvement on other reports of IRES-containing lentiviral vectors , such a discrepancy in expression levels of the upstream and downstream genes would probably be detrimental to certain therapeutic applications. 2A based multigene vectors, thus offer the unique advantage of better coexpression of two or more desired transgenes. It is of particular interest that this comparison was made at limiting MOI using expression cassettes whose transcription was driven by a clinically relevant human cellular promoter. Hence we can conclude that a 2A mediated co-expression strategy is significantly improved over an approach using the EMCV IRES when lentiviral vectors are used to infect cells at a level which is appropriate to gene therapy applications, where a major concern may be minimizing the risk of insertional mutagenesis.
In addition to the potential for intracellular mislocalisation of protein, the addition of 2A peptide [ additional amino acids in this case) to the first gene product might also interfere with the function of a given protein, and again this will have to be determined empirically for each application. In our experience addition of the 2A peptide did not affect the function of MGMT protein as neither its DNA repair activity nor nuclear localization were altered. Moreover, recent studies indicated that HOXB4 expressed using the 2A strategy retains its ability to support hematopoietic reconstitution by murine hematopoietic stem cells [25, 27]. A further issue might be immunogenicity due to the attachment of the 2A peptide-adduct to a therapeutic protein. Although these problems are not encountered in two recent murine in vivo studies [24, 25], further studies in multiple species are needed to understand this issue. More recently Fang et al  successfully engineered a furin cleavage site next to the 2A sequence to eliminate any possible adverse effects that might be caused by having a 2A peptide residue on a therapeutic protein.