Internal ribosome entry site-based vectors for combined gene therapy

Abstract

Gene therapy appears as a promising strategy to treat incurable diseases. In particular, combined gene therapy has shown improved therapeutic efficiency. Internal ribosome entry sites (IRESs), RNA elements naturally present in the 5’ untranslated regions of a few mRNAs, constitute a powerful tool to co-express several genes of interest. IRESs are translational enhancers allowing the translational machinery to start protein synthesis by internal initiation. This feature allowed the design of multi-cistronic vectors expressing several genes from a single mRNA. IRESs exhibit tissue specificity, and drive translation in stress conditions when the global cell translation is blocked, which renders them useful for gene transfer in hypoxic conditions occurring in ischemic diseases and cancer. IRES-based viral and non viral vectors have been used successfully in preclinical and clinical assays of combined gene therapy and resulted in therapeutic benefits for various pathologies including cancers, cardiovascular diseases and degenerative diseases.

Key Words: Vector; Gene transfer; Internal ribosome entry site; Gene therapy

Core tip: Combined gene therapy has emerged for a few years as a promising strategy to improve treatments of many diseases including cancer, cardiovascular diseases and degenerative diseases. In this context, internal ribosome entry site (IRES)-based vectors provide a powerful system to co-express several therapeutic genes from the same transcription unit. IRESs are translational enhancers, exhibiting tissue-specificity, and activated by stress. Different IRES-based vectors including plasmids, adeno-associated virus-derived and lentiviral vectors have been used successfully in many preclinical protocols of gene therapy. Moreover the few clinical assays launched with IRES-based multicistronic vectors resulted in therapeutic benefits.

INTRODUCTION

Combined gene therapy has appeared for a few years as an attractive approach to optimize the therapeutic benefits of gene transfer. In the field of cancer, the first examples of antitumoral cooperative effect have been provided by co-expression of the co-stimulation molecules CD70 and CD80, and of the two anti-angiogenic factors, angiostatin and endostatin, respectively[1-4]. Synergistical effects have also been obtained with co-expression of angiogenic growth factors generating therapeutic angiogenesis in ischemic diseases. This rational has been proven first with co-administration of vascular endothelial growth factor A (VEGFA) and angiopoietin as recombinant proteins as well as by co-administration of two plasmids coding these growth factors[5]. A few years later, combination of recombinant fibroblast growth factor 2 (FGF2) and PDGF-B also improved hindlimb ischemia in rats whereas a bicistronic vector expressing FGF2 and VEGFA efficiently induced vessel formation in a mouse angiogenesis assay[6,7]. These studies launched the concept of combined biotherapy. They also revealed that combined gene therapy is a promising therapeutic approach, allowing long term efficiency of treatments compared to recombinant proteins whose half life is often very short.

Internal ribosome entry sites (IRESs) are translational enhancers naturally present in a series of mRNAs, mediating internal initiation of translation when present between the genes of interest (Figure 1). IRESs thus allow the design of multicistronic expression cassettes resembling bacterial operons, able to drive translation of several genes coded by the same mRNA[8]. We have demonstrated that the use of IRES-based vectors co-expressing two genes of interest allows stable transgene expression with a constant ratio of the proteins of interest, in contrast to the use of two different plasmids expressing each transgene[9]. Actually, a bicistronic IRES-based vector co-expressing FGF2 and Cyr61 has revealed more efficient to generate therapeutic angiogenesis at low doses than the monocistronic vectors expressing large amounts of only one of these angiogenic factors[10]. It must be underlined that the IRES-based vector had no side effects on promotion of tumoral angiogenesis in contrast to the monocistronic ones, a very important feature for increased safety in clinical assays. These observations prompted us to deepen the features of IRESs applicable to vectorology and assess progress made in the field of gene transfer and combined gene therapy clinical assays using IRES-based vectors.

Figure 1 Cap-dependent and internal ribosome entry site-dependent initiation, two alternative mechanisms of translation. A: The so-called cap-dependent ribosome scanning mechanism predicts that ribosome 40S subunit binds to the mRNA 5’ end. Ribosome binding requires the initiation factor 4F (eIF-4F, composed of the three proteins eIF-4E, -4A and -4G). Then the mRNA is unwound under the control of the helicases eIF-4A and -4B, allowing the ribosome to scan the mRNA until recognition of an initiation codon (classically AUG)[11,12]; B: When an Internal ribosome entry site (IRES) is present in the mRNA 5’ untranslated region, IRES trans-acting factors (ITAFs) allow ribosome 40S internal recruitment, independently of the presence of cap and eIF-4F. The IRES-dependent mechanism occurs in the case of picornavirus uncapped mRNAs as well as for cellular capped mRNAs.

IRESS, TRANSLATIONAL ACTIVATORS FOR COMBINED TRANSGENE EXPRESSION

At a time when it was admitted that initiation of translation in eukaryotes required recognition of the capped mRNA 5’ end to recruit ribosomes, translation of the uncapped picornavirus mRNAs from an internal start codon remained a mystery. Indeed, the so-called ribosome scanning mechanism predicted that ribosomes bound to the mRNA 5’ end scanned the mRNA molecules until they recognized an AUG codon[11,12] (Figure 1). The event of internal ribosome binding was thought impossible. This puzzle raised by picornaviruses was solved by the discovery of RNA elements, called IRES, present in the 5’ untranslated regions of their mRNAs, which allow internal recruitment of ribosomes[13,14]. The dogma of the scanning mechanism was thus broken. In addition, it was quickly extended to cellular mRNAs as the first cellular IRES was discovered three years later in the BiP mRNA, coding for the immunoglobulin chaperone also known as GRP78[15]. This discovery was followed by the finding of several other IRESs in cellular mRNAs, in particular in the mRNAs of angiogenic growth factors such as FGF2, proto-oncogenes such as c-myc, pro and anti-apoptotic proteins such as X chromosome-linked inhibitor-of-apoptosis protein and apoptotic peptidase activating factor 1[16-20]. IRESs were also found in retroviruses, whose mRNAs are capped as cellular mRNAs, leading to the design of IRES-containing retroviral vectors[21,22].

The existence of IRESs in capped cellular mRNAs asked the question of their pathophysiological function[23]. Actually, several reports showed that IRESs from cellular mRNAs are regulated in various physiological processes including cell differentiation, spermatogenesis, neurone plasticity[24-27]. Several IRESs are also activated during cell cycle mitosis[28,29]. Recent reports have also shown that IRESs are aberrantly activated in tumor cells, and are thus involved in dysregulation of gene expression in cancer[30]. Furthermore, cellular IRES activity is stimulated in stress conditions such as apoptosis and hypoxia when cap-dependent translation is blocked[31-36].

IRES-dependent internal initiation of translation reminds the prokaryotic initiation mechanism which can translate polycistronic mRNAs[37,38]. This observation gave the idea that such operons could be created in eukaryotes using IRESs to design expression vectors[39]. A large majority of expression vectors allow co-expression of two genes under the control of two promoters. However such an approach has revealed that one of the genes may be silenced despite of the expression of the other one even though it expresses an antibiotic[40]. This can result from competition between the two promoters or counterselection of the gene of interest in case of toxicity or of cell growth inhibition. In such a context, IRESs have been used to generate transgene co-expression under the control of a single promoter (Figure 2).

Figure 2 Internal ribosome entry site-based multicistronic vector concept. The internal ribosome entry site (IRES)-based expression cassette contains several genes, separated by IRESs, under the control of the same promoter (Pr). This transcription unit gives rise to a single mRNA coding the different genes. Translation initiation occurs at the 5’ end by the cap-dependent mechanism, resulting in translation of the first open reading frame (ORF, Gene A). Internal initiations of translation occur at each IRES, resulting in translation of the other ORFs (Genes B and C). Thus the multicistronic mRNA generates several proteins from a single transcription unit, allowing more stable long term expression and stable transgene ratio[9,48]. For each ORF, initiation (AUG) and termination (STOP) codons are indicated.

The first retroviral tricistronic IRES-based vector appeared in 1992, providing an exciting potential for gene therapy[41]. This vector successfully co-expressed adenosine desaminase with neomycin (NEO) resistance and chloramphenicol acetyltransferase reporter genes, using the two picornavirus IRESs from poliovirus and encephalomyocarditis virus (EMCV), respectively. Two years later, a therapeutic tricistronic vector expressing the two interleukin-12 subunits with NEO validated the concept of IRES-based vectors to co-express two subunits of a protein with an adequate stoechiometry together with a resistance gene[42]. In the following years, bicistronic vectors were used successfully to select cell clones expressing a protein of interest with a resistance gene, preventing the problems generated by the use of two promoters[40,43].

TISSUE-SPECIFICITY OF CELLULAR IRESS

Most IRES-based vectors developed up to now use picornavirus IRESs, based on the strong efficiency of such IRESs in transient transfection, compared to cellular IRESs. It has been observed that cellular IRESs often exhibit a low efficiency in transiently transfected cells. Such a feature may result from the cell and tissue specificity of the cellular IRES activities. Actually, the FGF2 IRES activity varies with the cell type, the lowest being in fibroblasts, and the highest in neuroblastoma and osteosarcoma cells[44]. Similar variations have been observed for other cellular IRESs[45] (Creancier L and Prats AC, unpublished results). The strongest regulation of cellular IRESs has been shown in vivo, in transgenic mice expressing bicistronic dual luciferase constructs containing different IRESs. Clearly, the EMCV IRES was active in most tissues and organs, while the FGF2 IRES was very low in most organs except for testis and brain where its activity increased 200 to 400 times, at least 10 times higher than the EMCV IRES activity[44]. A similar behavior was observed with other cellular IRESs such as c-myc and VEGFA IRESs[31,45].

The tissue-specific features of cellular IRESs are useful to control transgene expression. Thus they can be considered as translational enhancers, if one makes a parallel with transcriptional enhancers upstream of promoters, governing the tissue-specificity of gene expression. The concept of translational tissue-specificity may be applied to gene transfer by coupling tissue-specific IRESs with tissue-specific promoters to create vectors with increased safety. This concept should also remember us that EMCV is not always the best IRES to be used. A recent study reports the failure of expression of the second cistron of a bicistronic adeno-associated virus (AAV) vector using the EMCV IRES, in murine cerebellar Purkinje neurons[46].

The advantage of using a cellular IRES has also been demonstrated for gene transfer into skeletal muscle. The FGF1 IRES is as efficient as the EMCV IRES in mouse muscle after plasmid DNA electrotransfer[47]. Moreover, when this IRES is used in a bicistronic AAV vector, its activity is significantly superior to that of the EMCV IRES in myoblasts and allows a transgene expression 10 times more efficient when this AAV is injected in mouse muscle[48]. Such a difference may be due to the presence of specific FGF1 IRES trans-acting factors (ITAFs) (Ainaoui et al, in revision). Alternatively, it can result from the lower ability of the EMCV IRES to maintain a stable long term compared to cellular IRESs, shown in a previous report[9].

On the basis of these different data, it can be recommended to choose the adequate IRES to be used according to the cell type or tissue to be targeted, rather than using systematically the EMCV IRES as presently proposed in all commercial IRES-based vectors.

IRES-MEDIATED GENE EXPRESSION IN STRESS CONDITIONS

In many diseases cells are subjected to different stresses such as hypoxia, apoptosis or ER stress. In stress conditions, translation initiation is inhibited by two ways: blockade the mammalian target of rapamycin pathway which affects ribosome recruitment on the cap, and phosphorylation of eIF2-α which prevents charged initiator Met-tRNA formation. Interestingly, IRES-dependent translation is not affected by these two ways of silencing[35,49,50].

As mentioned above, IRESs are naturally present in messenger RNAs coding for proteins involved in the stress response, especially apoptosis and hypoxia. In particular, an IRES is present in the mRNA of the hypoxia-induced factor 1α (HIF1α), the key of the cell response to hypoxia that induces transcription of all the genes containing a hypoxia responsive element (HRE) in their promoters[51]. This IRES allows HIF mRNA translation to be activated during hypoxia despite of the blockade of global translation[32,52]. Such activation occurs under the control of an ITAF, the pyrimidine tract binding protein, also known as a regulator for various IRESs[52,53].

An important consequence of hypoxia is the stimulation of angiogenesis in order to generate new vessels able to restore the cell supply with oxygen. This process occurs in cancers when cells in the tumor core are oxygen deprivated, as well as in ischemic diseases such as heart and lower limb ischemia when tissues are not any more irrigated due to artery occlusion. Strikingly, the major angiogenic factors VEGFA (vascular endothelial growth factor A), FGF1 and FGF2, possess IRESs in their mRNAs[20,47,54-56]. VEGFA expression, transcriptionally induced by HIF1α, is also translationally enhanced via the IRES in hypoxic tumors and in ischemic mouse legs[31,32,36]. In contrast to VEGFA, FGF2 is not induced transcriptionally by hypoxia but its synthesis is translationally induced by the IRES-dependent mechanism in ischemic tissues[31,33]. The same phenomenon has been observed for the major lymphangiogenic factor VEGFC, induced by hypoxia at the translational level via an IRES, but not at the transcriptional level, in tumors and lymph nodes[36,57]. FGF2 and VEGFC induction is exclusively translational and HIF-independent, revealing that IRESs provide an alternative HIF-independent way of response to hypoxia.

On a biotechnological point of view, the sensitivity of IRESs to hypoxia may be an advantage for several applications. Gene transfer vectors can benefit from this feature as the presence of IRESs allows increased transgene expression in ischemic conditions in vivo. Once again, one can see that data from basic research have to be taken into account in the design of optimized expression cassettes. The use of IRES-based vectors seems particularly adequate for gene therapy of ischemic diseases and cancer, as in both cases the transgenes have to be expressed in hypoxic conditions.

BIOMEDICAL APPLICATIONS OF IRESS

IRESs have found biomedical applications for several years. As mentioned above, the first biomedical use of IRESs in an expression vector has been co-expression of subunits of a therapeutic protein with a gene of resistance, as shown for interleukin 12 subunits with a gene of resistance[42]. However this application is limited to therapeutic genes composed of several subunits. In addition, the use of resistance genes is not recommended as it may prevent the use of the vector in a clinical assay.

Another application of IRESs raised during the last decade, resulting form the emerging concept of combined gene therapy. Several studies have validated this concept using a cocktail of two vectors to transfer two genes simultaneously. This has been particularly documented in the field of cardiovascular diseases and cancer, with therapeutic benefits obtained in different animal models using different combinations of angiogenic or anti-angiogenic factors[4,5,58-61] (Table 1). Interestingly the combination of VEGFA and PDGFB successfully induced therapeutic angiogenesis both in ischemic leg and in ischemic heart. In the field of rare diseases, two AAV vectors expressing microdystrophin and IGF1 resulted in increased muscle mass and strength, reduced myofiber degeneration and increased protection against contraction-induced injury in mdx mice[62]. These different studies were performed either with naked DNA or with recombinant adeno-associated virus vectors.

Table 1 Preclinical studies of combined gene therapy with co-administration of monocistronic vectors.

Pathology	Therapeutic genes	Animal model	Vector type	Ref.
Cancers
Leukemia, melanoma	Angiostatin + endostatin	Mouse	Retrovirus	Scappaticci et al[4], 2001
Ovarian cancer	Angiostatin + endostatin	Mouse	AAV	Ponnazhagan et al[3], 2004
Glioblastoma	VEGF-R1 + angio-endo (Statin AE)	Mouse	SB transposon	Ohlfest et al[60], 2005
Pancreatic cancer	TSP1 + endostatin	Mouse	AAV	Zhang et al[61], 2007
Cardiovascular diseases
Limb ischemia	VEGFA + angiopoietin-1	Rabbit	Plasmid	Chae et al[5], 2000
Limb ischemia	VEGFA + FGF2	Mouse	Plasmid	Lee et al[59], 2007
Limb ischemia	VEGFA + PDGFB	Rabbit	AAV	Kupatt et al[58], 2010
Heart ischemia	VEGFA + PDGFB	Pig	AAV	Kupatt et al[58], 2010
Rear diseases
DMD	Microdystrophin + IGF1	Mouse	AAV	Abmayr et al[62], 2005

DMD: Duchenne muscular dystrophy; VEGF: Vascular endothelial growth factor; FGF2: Fibroblast growth factor 2; AAV: Adeno-associated virus.

The use of two different vectors for multiple transgene expression exhibits disadvantages: on the one hand, the ratio of the therapeutic molecules cannot be controlled, leading in the loss of the cooperative effect: expression of one of the vectors often decrease or is silenced earlier than the other one[40]. On the other hand, the cost of two therapeutic vectors in a clinical perspective is higher than a single one. These disadvantages are still more important in case of a cocktail of three or more therapeutic genes.

The concept of IRES-based vectors for combined gene therapy has been validated for combined immunotherapy of cancer using a tricistronic retrovirus expressing the two co-stimulation molecules CD70 and CD80[2] (Table 2). In addition to the EMCV IRES, several cellular or retroviral IRESs were successful in this approach[63]. In vivo gene therapy has also been validated for the treatment of ischemic limb in a mouse model, following intramuscular injection and electrotransfer of a plasmid containing the FGF1 IRES for co-expression of FGF2 and Cyr61[10]. This study showed than the two angiogenic factors, although expressed at lower doses from the bicistronic vector than from the monocistronic ones, have a synergistical effect in stimulating therapeutic angiogenesis, rendering the bicistronic construct more efficient. More importantly, due to the lower doses of therapeutic molecules, the bicistronic vector induces no side effects on tumoral angiogenesis, in contrast to one of the monocistronic vectors expressing huge amounts of Cyr61. Thus combined gene therapy using IRES-based vectors is also a safer therapeutic approach.

Table 2 Preclinical studies of combined gene therapy using multicistronic vectors.

Pathology	Therapeutic genes	Animal model	IRES	Vector type	Ref.
Cancers
Fibrosarcoma	CD70 + CD80	Mouse	EMCV	Retrovirus	Couderc et al[2], 1998
Melanoma	Angio-endo fusion	Mouse	None (fusion)	Retrovirus	Scappaticci et al[79], 2001
Multiple myeloma	IL12 subunits + CD80	Mouse	EMCV + FMDV	Retrovirus	Wen et al[69], 2001; Li et al[67], 2003
Melanoma	CD70 + CD80	Mouse	EMCV, c-myc, FGF2, HTLV1	Retrovirus	Douin et al[63], 2004
Ovarian cancer	Angiostatin + endostatin	Mouse	EMCV	AAV	Isayeva et al[66], 2005
Head and neck cancer	Angio-endo fusion	Mouse	None (fusion)	Vaccinia virus	Tysome et al[80], 2011
Pancreas cancer	CXCL4L1 + fibstatin	Mouse	FGF1	AAV, Lentivector	Prats et al[68], 2013
Cardiovascular diseases
Limb ischemia	FGF2 + Cyr 61	Mouse	FGF1	Plasmid	Rayssac et al[10], 2009
Limb ischemia	VEGFA + BMP7	Rabbit	EMCV	AAV	Zhang et al[65], 2010
Limb ischemia	VEGFA + FGF4	Mouse	EMCV	AAV	Jazwa et al[64], 2013
Heart ischemia	Gata4 + Mef2C + Tbx5	Rat	None (2A element)	Lentivector	Mathison et al[75], 2014
Neurodegenerative diseases
Parkinson	TH + AADC + CH1	Rat	EMCV	Lentivector	Azzouz et al[73], 2002
					Stewart et al[74], 2011
AMD	Angiostatin + endostatin	Mouse	EMCV	Lentivector	Kachi et al[70], 2009

BMP7: Bone morphogenetic protein 7; Gata4: GATA binding protein 4; Mef2C: Myocyte-specific enhancer factor 2C; Tbx5: T-box transcription factor 5; TH: Tyrosine hydroxylase; AADC: aromatic L-amino acid decarboxylase; CH1: GTP cyclohydrolase-1; AMD: Age-related macular degeneration.

Additional studies have confirmed the successful use of IRES-based vectors for combined treatment of limb ischemia with VEGFA and FGF4 or bone morphogenetic protein7 (BMP7)[64,65]. Combined gene therapy of cancer was also reported using IRES-based vectors co-expressing IL-12 and CD80, as well as antiangiogenic factors angiostatin and endostatin, or CXCL4I and fibstatin[66-69] (Table 2). Combination of angiostatin and endostatin in an IRES-based vector was also successful to treat age-related macular degeneration in a mouse model[70]. In the field of degenerative diseases, mucopolysaccharidosis type IIIA has been addressed in presymptomatic MPSIIIA mice by intrastriatal administration of an AAV vector co-expressing N-sulfoglycosamine sulfohydrase (SGSH) with the sulfatase-modifying factor (SUMF1) (Winner et al, submitted). This study has resulted in a clinical assay[71] see below). Only one report has obtained better data with two separate AAV vectors to deliver FGF14 and a fluorescent protein into purkinje neurons, than with an IRES[46]. This study used the EMCV IRES previously reported to function in neurons[72]. However it must be underlined that the EMCV IRES is not very active in neurons in vivo, by comparison with the FGF2 IRES that is at least ten times more active[24,44]. In such a case, one can expect that the choice of the FGF2 IRES would provide better data.

Multigene transfer has also been validated for combina-tions of three genes. A tricistronic IRES-based lentivector expressing three catecholaminergic proteins, Prosavin, was administrated by bilateral striatal injection for treatment of Parkinson in rats, resulting in important therapeutic benefits[73,74] (Table 2). Moreover, a tricistronic 2A-based lentivector administrated in situ was also efficient in co-expressing Gata4, Mef2c and Tbx5 for postinfarct ventricular functional improvement in rats[75].

It is often mentioned that the IRES-driven translation of the downstream cistrons is lower than the cap-dependent first cistron translation. This issue can easily be addressed by intelligent vector design: First, one can take into account the tissue specificity of the IRES by choosing the most adequate IRES rather than using systematically the EMCV IRES. Most bi- and- tricistronic vectors use this IRES although it is far to be the best one in many tissues such as muscle or brain[24,48]. Second, the IRES efficiency can be improved. It must be noticed that the EMCV IRES activity is very sensitive to the position of the start codon of the gene of interest. This IRES, in contrast to the FGF1 IRES, exhibits no flexibility: the AUG must be positioned just downstream from the IRES. The insertion of a single restriction site between the IRES and the AUG codon is sufficient to inactivate the IRES[76]. The insertion of a spacer between the first gene and the IRES is also susceptible to enhance the IRES activity by preventing IRES structural alterations by RNA sequences located upstream[77]. In addition, mutations of the upstream AUG codons in the EMCV IRES improve its efficiency[78]. Finally, an important parameter is the IRES regulation by microenvironment. In particular, FGF or VEGF IRES activities are more sensitive to hypoxia than the EMCV IRES and may allow a more efficient transgene expression in ischemic diseases.

ALTERNATIVES TO IRESS FOR MULTICISTRONIC VECTORS

IRES-based vectors are not the only approach to co-express several gene products under the control of a single promoter. The first alternative is gene fusion. It has been successfully used to combine endostatin and angiostatin in a treatment of melanoma and of head and neck cancer[79,80]. A second alternative to IRESs is the use of alternative splicing-based vectors. Such an approach had been proposed many years ago using retroviral vectors, using the natural alternative splicing features of retrovirus genome[81,82]. This concept has been developed more recently in the purpose of co-expressing two immunoglobulin chains[83]. The interest of this system is the ability to adapt the ratio of the two transgenes by mutating the splicing sites. However one limit of this attractive system is that splicing site efficiency and consequently the ratio of the two proteins of interest, is influenced by the presence of exon splicing enhancers or silencers in the transgene sequences, preventing the design of vectors with a stable transgene ratio applicable to co-expression of any pair of therapeutic proteins.

A third exciting system of co-expression is provided by the 2A peptides. Such peptides, occurring in many viral genomes, are peptide sequences of about 19 amino-acid residues, which can produce a discontinuity in the translated polypeptide when encoded in a longer open reading frame (ORF)[84]. In contrast to what is currently admitted, 2A peptides do not catalyze a protein cleavage, but they catalyze termination of translation in the absence of a stop codon, followed by reinitiation. They are currently used as a tool to co-express two or more separate proteins from a single ORF[85]. 2A peptides thus constitute an alternative to IRESs, but do not work in all systems. By example, in the study in purkinje neurons mentioned above, a 2A peptide was used but did not function, resulting in detection of the longer ORF rather than the two expected proteins[46]. In another report comparing bicistronic constructs expressing Sox9 and EGFP separated by the EMCV IRES or by the FMDV 2A peptide, the authors detected 42% of Sox-EGFP fusion protein, reflecting an inefficient ribosome skipping mechanism[86]. Formation of such fusion proteins often occurs with proteins bearing N-terminal signal sequences[87]. In addition, no information is available about the 2A peptides tissue-specificity or behavior in response to stress, in contrast to IRESs.

CLINICAL APPLICATIONS OF IRES-BASED VECTORS TO GENE THERAPY

All the preclinical studies mentioned above show that IRES-based vectors represent an exciting tool to be used for combined gene therapy. Nowadays, very little clinical trials with such vectors have been reported. The first trial to be cited is the tricistronic IL12-expressing retrovirus, which gave significant decrease of tumor sizes on a few patients with melanoma or head and neck cancer[88,89].

A bicistronic IRES-based vector co-expressing FGF2 and VEGFA has been assessed in a clinical assay of gene therapy on patients with refractory coronary disease[90] (Table 3). The protocol corresponded to intramyocardial transfer of a plasmid expressing the bicistronic cassette. This study showed no improvement in myocardial perfusion, but treated patients exhibited improved exercice tolerance and clinical symptoms. Furthermore the bicistronic gene transfer was safe. This moderate benefit, although encouraging, may be due to the use of a plasmid, which does not provide long term expression in contrast to viral vectors, and also to the choice of the EMCV IRES which is not optimal to drive gene expression in hypoxic conditions[31,36].

Table 3 Clinical studies of combined gene therapy.

Pathology	Therapeutic genes	IRES	Vector type	Outcome	Ref.
Ischemic heart disease	VEGFA + FGF2	EMCV	Plasmid	Moderate benefits	Kukula et al[90], 2011
Parkinson	TH + AADC + CH1	EMCV	Lentivector	Benefits for 15/15 patients	Palfi et al[91], 2014
Mucopolysaccharidosis type IIIA	SGSH + SUMF1	EMCV	Lentivector	Benefits for 1/4 patients, stabilization for 3/4	Tardieu et al[71], 2014

TH: Tyrosine hydroxylase; AADC: Aromatic L-amino acid decarboxylase; CH1: GTP cyclohydrolase-1; SGSH: N-sulfoglycosamine sulfohydrolase; SUMF1: Sulfatase-modifying factor.

Very recently, two gene therapy clinical trials successfully used multi-cistronic IRES-based viral vectors. On the one hand, a gene therapyI/II phase clinical trial on patients with mucopolysaccharidosis type IIIA, a severe degenerative disease, has displayed neurocognitive benefits[71]. Four children received intracerebral injections of a bicistronic AAV vector expressing the SGSH and SUMF1 genes separated by the EMCV IRES. Neurocognitive evaluations suggest a cognitive benefit on the youngest patient, where as the other ones are stabilized. Importantly, the treatment was safe and well tolerated after 1 year in all the patients, validating the surgical approach for direct AAV delivery in the brain parenchyma. On the other hand, a phaseI/II assay was performed on 15 patients with Parkinson’s disease using Prosavin (see above), a tricistronic lentivector with EMCV IRESs administrated by intrastriatal delivery[91]. A significant improvement of motor scores was recorded in all patients at 6 mo. This is the first-in-man use of a lentiviral-based gene therapy vector for a neurodegenerative disease. These studies validate the clinical use of IRES-based viral vectors.

CONCLUSION

Many reports have shown that combined gene therapy is an attractive approach in animal models. This observation has justified extensive research on optimization of gene transfer vectors able to co-express several proteins. In this context, IRES-based vectors have now been validated in pre-clinical as well as in clinical studies by showing their safety and ability to generate therapeutic benefits.

In addition, the data available on IRES tissue-specificity and activation in response to stress provide promising perspectives of vector improvement, which may result in better efficiency of gene therapy.