The Ex vivo approach

This consists of engineering genetically modified cells in vitro that are subsequently injected directly into the affected joint. This procedure allows (1) controlling the quality of the injected material, especially when using a vector that might cause insertional mutagenesis, (2) sorting the transduced cells to be injected back and (3) knowing precisely the level of transgene expression, to adapt the number of genetically modified cells to be injected to the clinically relevant level of transgene desired. Such an approach has been achieved using genetically modified fibroblasts as well as immune cells, such as fibroblast-like synoviocytes, myoblasts, mesenchymal stromal cells, T cells and dendritic cells .

The In vivo strategy

This strategy comprises direct intra-articular administration of the vector and presents potential advantages over systemic routes, such as targeting of the diseased tissue and sustained suppression of inflammation in the joint. Because of the easy access to the joint space, local gene delivery has been achieved with a large panel of viral and non-viral vectors, mainly to target the synovium . The obvious advantage is to obtain in a single injection the production of the therapeutic agent directly by resident cells, with minimal manipulation of the targeted cells and spreading of the vector. It is a very simple medical procedure that does not involve complex surgical procedures or anaesthesia and allows the treatment of as many joints as necessary. Independently of the vector and transgene used, its efficiency has been proved by many groups, in several different animal models of RA. Moreover, few groups reported the observation of a contra-lateral effect, that is, an amelioration of the clinical features in the non-injected contra-lateral joint . Although the cellular and molecular mechanisms underlying such an effect have not been completely unravelled, it might represent a method to treat several joints by a single injection. For initial clinical trials, local therapy will be the reasonable approach to be followed, regarding the higher risk of systemic therapy. However, preclinical monitoring of the vector spreading and the immune-response induction against the vector following intra-articular administration showed that the technique is not devoid of some lacunae. These are major concerns that will need a close follow-up to avoid any setbacks for the future application of gene therapy strategies. The ultimate goal of gene therapy will be the injection of a vector that has a specific target cell.

Non-viral gene transfer methods

These methods used nucleic acids associated either with chemical (liposomes and polymers) or mechanical (gene gun) means to increase transduction efficiency. One alternative to enhance the entrance of nucleic acids into the targeted tissue has been the use of in vivo electrotransfer (ET) . It is based on the same principle as for in vitro electroporation, that is, applying an electric field on each side of the target organ following nucleic acid injection (plasmid DNA, siRNA and oligodeoxynucleotide (ODN)), and optimising parameters to each organ’s characteristics. This technique has been validated for the muscle as systemic approach and for the joint as local approach , overexpressing antagonists of pro-inflammatory agents or anti-inflammatory molecules. Since 2005, three publications have investigated the use of ET for in vivo transfection of siRNA against pro-inflammatory mediators (interleukin (IL)-1, IL-6, tumour necrosis factor (TNF) and receptor activator for nuclear factor kappa B ligand (RANKL)) into joint tissue in arthritic rodents to achieve local RNA interference (RNAi) . In a comparative study, our group showed that direct gene transfer into arthritic joints using ET is less efficient in treating collagen-induced arthritis (CIA) in mice than intra-muscular ET . In general, the major drawback of the ET approach is that the transgene expression is mostly silenced after 1 week and has to be performed under anaesthesia. Moreover, such strategy is restricted to experimental research and is not yet transposable to the clinic.

Numerous different nanocarriers have been applied as delivery vehicles for in vivo RNAi. Among them, intra-peritoneal administration of chitosan/anti-TNF siRNA complexes showed an efficient knockdown of TNF in peritoneal macrophages, reducing both local and systemic inflammation in mouse CIA . More recently, enforced expression of miR-15a within arthritic joints using local atelocollagen-mediated transfer of miR-15a showed efficient induction of synovium apoptosis in the autoantibody-mediated arthritis mouse model . The effect was mediated by the down-regulation of the anti-apoptotic Bcl-2 protein and the increased expression of caspase 3 locally.

Liposome formulations have been used to protect nucleic acids from enzymatic degradation, renal or macrophage-mediated clearance, and thus to enhance gene transfer efficacy. Our group has developed a synthetic vector based on the cationic liposome RPR209120/DOPE to formulate therapeutic nucleic acids (plasmid DNA or siRNA). We have demonstrated the potential of such anti-cytokine lipoplexes in the mouse CIA model following systemic injection. Targeting TNF, IL-1, IL-6 or IL-18 using siRNA-containing lipoplexes was very efficient in decreasing the severity of established arthritis, either individually or in combination . There are data suggesting that liposomes can be endocytosed by systemic monocytes entering the joints and by joint macrophages . Although liposome-based formulations could be envisaged for systemic non-viral anti-cytokine strategy in RA, pharmaco-toxicological studies need to be performed before planning clinical trials. This strategy can also present a valuable tool for in vivo screening of new potential therapeutic targets. In conclusion, non-viral vectors allow local delivery, but their efficiency is reduced compared with viral vectors due to drug bioavailability issues, and would thus require repetitive administrations. Most of the formulations tested are, moreover, themselves inflammatory.

Viral vectors

These are very efficient for synovial gene delivery. Most of the retroviral (RV) vectors used in clinical gene therapy trials were derived from the very well studied and characterised Moloney murine leukaemia oncoretrovirus (MMLV) . They are small RNA viruses that replicate through a DNA intermediate and require cell division for infection and integration . Thus, they have only been used in ex vivo gene transfer approaches in which isolated cells can be propagated in culture, genetically modified after RV infection, and then implanted back into a recipient patient. Although the synovium of inflamed joints produced higher RV-mediated transduction efficiency than normal joints, human rheumatoid synovium does not contain enough proliferating cells to support efficient retroviral transduction. Direct injection of an RV vector encoding β-galactosidase into the engrafted human synovium of severe combined immune deficiency (SCID) mice resulted in less than 1% of transduced cells. Ex vivo infection could however reach 35% of synoviocytes in the presence of TNFα or 50% when concentrating RV supernatants . A few groups have also explored the possibility to use lentiviral (LV) vectors from several species and showed that they transduce efficiently the non-dividing rheumatoid synovium, resulting in some clinical improvement in experimental arthritis models. So far, only one group has explored the feasibility of using intra-articular injection of LV-expressing short hairpin RNAs (shRNAs) for local long-term suppression of pathogenic genes. As proof of concept, this group targeted the TNF superfamily member B cell-activating factor (BAFF). Intra-articular injection of LV vectors targets dendritic cells in the joint tissue and BAFF gene silencing inhibits pro-inflammatory cytokine expression, suppresses generation of plasma cells and Th17 cells and ameliorates overall joint histopathology in mouse CIA .However, the same potential problem of insertional mutagenesis exists as for the entire family of RV; their low titre of production is still an important technical limitation, and they pose ethically dilemmas for in vivo use.

Local in vivo gene transfer has been achieved with recombinant vectors, such as adenoviral vector (AdV), adeno-associated virus (AAV) and herpes simplex virus (HSV), using various animal models of RA. Since RA is a chronic inflammatory disorder, immunogenic vectors such as AdV vectors have been eliminated from the panel. Although the serotype 5 transduces the rheumatoid synovium very efficiently, allowing the rapid expression of therapeutic genes to high levels, AdVs are highly immunogenic and thus only useful for transient transgene expression. To circumvent this main limitation, a few groups explored two tracks: the use of non-human AdV, such as canine AdV vectors and the use of AdV from the 51 different human serotypes for which there is very few or no occurrence of a pre-existing humoural response in the human population . Alternatively, targeted gene transfer to human synovial tissue using explants and in vivo transfection in the mouse CIA Model were efficiently increased with fibre-modified AdV . Despite these interesting strategies, AdVs are definitively not the gold standard approach for gene transfer in RA; but they still provide very interesting experimental tools for proof-of-concept studies.

Although interesting data have been generated using HSV for in vivo transfer of anti-arthritic genes to joints , they remain poorly studied in experimental models for RA. Indeed, these vectors showed high infectivity of joint tissues, can be produced at high titres and have a large packaging capacity allowing the inclusion of multiple anti-arthritic genes. A few groups have also attempted to develop hybrid vectors that would represent the synthesis of positive features for each vector, but the achievement of high-titre viral batches remains the major obstacle. Finally, the capacity of ‘gut-less’ vectors has been evaluated in vitro on human rheumatoid arthritis-fibroblast like synoviocytes (RA-FLS), but no concrete results were obtained during in vivo validation, although they are less immunogenic.

One of the most promising viral vectors for human gene therapy that emerged over the past decade is AAV. The biology of these parvoviruses, non-pathogenic and non-toxic in humans, has been extensively investigated, and methods of production, purification and titration for their clinical use highly improved their intrinsic limitations . Recombinant (r)AAV vectors have been shown to direct efficient, prolonged and safe transgene expression in several tissues, with distinct tropism for each serotype . Cross-packaging of serotypes allowed the transduction of a large panel of tissues and cell types . Moreover, physicochemical stability of rAAV facilitates storage and the clinical administration in RA. Currently, rAAV2 is used in a number of gene therapy clinical trials in haemophilia B, leucodystrophy, cystis fibrosis, low-density lipoprotein (LDL) deficiency and in RA . At least 46 clinical trials have been conducted or are in progress with rAAV vectors, all showing a good safety profile. Interest was aroused by the capacity of the wild-type vector to infect both dividing and non-dividing cells, as well as to stably integrate into a site-specific locus (q13.3) on chromosome 19 (AAVS1), conferring theoretical long-term and safe transgene expression. This property is lost when using rAAV and most of the transgene expression comes from episomal forms . In vitro , most of the cell types found in the RA joints can be transduced by rAAV2, including RA-FLS, chondrocytes and macrophages . A few studies used systemic injection of the rAAV vector, either intra-muscularly or peri-articularly , and showed high and therapeutically efficient transgene expression, detectable in sera for at least 4 months. The feasibility of direct intra-articular gene transfer to rat and mouse arthritic joints has been well demonstrated . The pattern of expression for AAV2 following local injection has been variable according to studies, from synovial lining cells to synoviocytes and chondrocytes and muscle and synoviocytes . More importantly, in rat and mouse models of arthritis, local transgene expression was shown to last for at least 7 months after joint delivery of rAAV5, mostly in synoviocytes . The serotype 5 capsid mediated rapid, high and stable gene transfer of a reporter gene into CIA mouse knee joints for at least 4 months ; Tak et al. showed that transgene expression was already detectable 7 days after injection and lasted for at least 4 weeks in the adjuvant-induced arthritis (AIA) rat model and Ghivizzani et al. reported that the rheumatic joints of horses were also efficiently transduced using rAAV5 . Using two different TNF-blocking agents (the tumour necrosis factor receptor (TNFR1)-mIgG1 fusion protein or a dimeric sTNFR2) as a proof of concept, two studies have demonstrated the feasibility of rAAV5-mediated gene therapy in mouse CIA and rat AIA models of arthritis. When the anti-TNF molecule expression is under a strong constitutive promoter (cytomegalovirus (CMV)), the antagonist molecule was rapidly (within 2 weeks), highly and stably expressed for 9 weeks when delivered intra-articularly by a CMV-driven rAAV5 vector. This was associated with a decrease in arthritis incidence and severity in both animal models. More importantly, the transgene was expressed under a nuclear factor-kappaB (NF-κB)-responsive promoter inducible by inflammation; the clinical effect was associated with a transient expression of the anti-TNF molecule, only detectable during disease flares. These results suggest that the local rAAV5-mediated gene delivery of a disease-inducible therapeutic agent may be instrumental in achieving successful treatment of RA by gene therapy . More recently, the feasibility of using rAAV5 for intra-articular expression of siRNA sequences targeting TNF to interfere with an ongoing arthritis has been demonstrated in the mouse CIA model . The rational design of AAV capsid mutants, and strategies such as the use of self-complementary vector genomes , might increase the potential of using AAV vectors for RA gene therapy even more.

Non-viral gene transfer methods

These methods used nucleic acids associated either with chemical (liposomes and polymers) or mechanical (gene gun) means to increase transduction efficiency. One alternative to enhance the entrance of nucleic acids into the targeted tissue has been the use of in vivo electrotransfer (ET) . It is based on the same principle as for in vitro electroporation, that is, applying an electric field on each side of the target organ following nucleic acid injection (plasmid DNA, siRNA and oligodeoxynucleotide (ODN)), and optimising parameters to each organ’s characteristics. This technique has been validated for the muscle as systemic approach and for the joint as local approach , overexpressing antagonists of pro-inflammatory agents or anti-inflammatory molecules. Since 2005, three publications have investigated the use of ET for in vivo transfection of siRNA against pro-inflammatory mediators (interleukin (IL)-1, IL-6, tumour necrosis factor (TNF) and receptor activator for nuclear factor kappa B ligand (RANKL)) into joint tissue in arthritic rodents to achieve local RNA interference (RNAi) . In a comparative study, our group showed that direct gene transfer into arthritic joints using ET is less efficient in treating collagen-induced arthritis (CIA) in mice than intra-muscular ET . In general, the major drawback of the ET approach is that the transgene expression is mostly silenced after 1 week and has to be performed under anaesthesia. Moreover, such strategy is restricted to experimental research and is not yet transposable to the clinic.

Numerous different nanocarriers have been applied as delivery vehicles for in vivo RNAi. Among them, intra-peritoneal administration of chitosan/anti-TNF siRNA complexes showed an efficient knockdown of TNF in peritoneal macrophages, reducing both local and systemic inflammation in mouse CIA . More recently, enforced expression of miR-15a within arthritic joints using local atelocollagen-mediated transfer of miR-15a showed efficient induction of synovium apoptosis in the autoantibody-mediated arthritis mouse model . The effect was mediated by the down-regulation of the anti-apoptotic Bcl-2 protein and the increased expression of caspase 3 locally.

Liposome formulations have been used to protect nucleic acids from enzymatic degradation, renal or macrophage-mediated clearance, and thus to enhance gene transfer efficacy. Our group has developed a synthetic vector based on the cationic liposome RPR209120/DOPE to formulate therapeutic nucleic acids (plasmid DNA or siRNA). We have demonstrated the potential of such anti-cytokine lipoplexes in the mouse CIA model following systemic injection. Targeting TNF, IL-1, IL-6 or IL-18 using siRNA-containing lipoplexes was very efficient in decreasing the severity of established arthritis, either individually or in combination . There are data suggesting that liposomes can be endocytosed by systemic monocytes entering the joints and by joint macrophages . Although liposome-based formulations could be envisaged for systemic non-viral anti-cytokine strategy in RA, pharmaco-toxicological studies need to be performed before planning clinical trials. This strategy can also present a valuable tool for in vivo screening of new potential therapeutic targets. In conclusion, non-viral vectors allow local delivery, but their efficiency is reduced compared with viral vectors due to drug bioavailability issues, and would thus require repetitive administrations. Most of the formulations tested are, moreover, themselves inflammatory.

Viral vectors

These are very efficient for synovial gene delivery. Most of the retroviral (RV) vectors used in clinical gene therapy trials were derived from the very well studied and characterised Moloney murine leukaemia oncoretrovirus (MMLV) . They are small RNA viruses that replicate through a DNA intermediate and require cell division for infection and integration . Thus, they have only been used in ex vivo gene transfer approaches in which isolated cells can be propagated in culture, genetically modified after RV infection, and then implanted back into a recipient patient. Although the synovium of inflamed joints produced higher RV-mediated transduction efficiency than normal joints, human rheumatoid synovium does not contain enough proliferating cells to support efficient retroviral transduction. Direct injection of an RV vector encoding β-galactosidase into the engrafted human synovium of severe combined immune deficiency (SCID) mice resulted in less than 1% of transduced cells. Ex vivo infection could however reach 35% of synoviocytes in the presence of TNFα or 50% when concentrating RV supernatants . A few groups have also explored the possibility to use lentiviral (LV) vectors from several species and showed that they transduce efficiently the non-dividing rheumatoid synovium, resulting in some clinical improvement in experimental arthritis models. So far, only one group has explored the feasibility of using intra-articular injection of LV-expressing short hairpin RNAs (shRNAs) for local long-term suppression of pathogenic genes. As proof of concept, this group targeted the TNF superfamily member B cell-activating factor (BAFF). Intra-articular injection of LV vectors targets dendritic cells in the joint tissue and BAFF gene silencing inhibits pro-inflammatory cytokine expression, suppresses generation of plasma cells and Th17 cells and ameliorates overall joint histopathology in mouse CIA .However, the same potential problem of insertional mutagenesis exists as for the entire family of RV; their low titre of production is still an important technical limitation, and they pose ethically dilemmas for in vivo use.

Local in vivo gene transfer has been achieved with recombinant vectors, such as adenoviral vector (AdV), adeno-associated virus (AAV) and herpes simplex virus (HSV), using various animal models of RA. Since RA is a chronic inflammatory disorder, immunogenic vectors such as AdV vectors have been eliminated from the panel. Although the serotype 5 transduces the rheumatoid synovium very efficiently, allowing the rapid expression of therapeutic genes to high levels, AdVs are highly immunogenic and thus only useful for transient transgene expression. To circumvent this main limitation, a few groups explored two tracks: the use of non-human AdV, such as canine AdV vectors and the use of AdV from the 51 different human serotypes for which there is very few or no occurrence of a pre-existing humoural response in the human population . Alternatively, targeted gene transfer to human synovial tissue using explants and in vivo transfection in the mouse CIA Model were efficiently increased with fibre-modified AdV . Despite these interesting strategies, AdVs are definitively not the gold standard approach for gene transfer in RA; but they still provide very interesting experimental tools for proof-of-concept studies.

Although interesting data have been generated using HSV for in vivo transfer of anti-arthritic genes to joints , they remain poorly studied in experimental models for RA. Indeed, these vectors showed high infectivity of joint tissues, can be produced at high titres and have a large packaging capacity allowing the inclusion of multiple anti-arthritic genes. A few groups have also attempted to develop hybrid vectors that would represent the synthesis of positive features for each vector, but the achievement of high-titre viral batches remains the major obstacle. Finally, the capacity of ‘gut-less’ vectors has been evaluated in vitro on human rheumatoid arthritis-fibroblast like synoviocytes (RA-FLS), but no concrete results were obtained during in vivo validation, although they are less immunogenic.

One of the most promising viral vectors for human gene therapy that emerged over the past decade is AAV. The biology of these parvoviruses, non-pathogenic and non-toxic in humans, has been extensively investigated, and methods of production, purification and titration for their clinical use highly improved their intrinsic limitations . Recombinant (r)AAV vectors have been shown to direct efficient, prolonged and safe transgene expression in several tissues, with distinct tropism for each serotype . Cross-packaging of serotypes allowed the transduction of a large panel of tissues and cell types . Moreover, physicochemical stability of rAAV facilitates storage and the clinical administration in RA. Currently, rAAV2 is used in a number of gene therapy clinical trials in haemophilia B, leucodystrophy, cystis fibrosis, low-density lipoprotein (LDL) deficiency and in RA . At least 46 clinical trials have been conducted or are in progress with rAAV vectors, all showing a good safety profile. Interest was aroused by the capacity of the wild-type vector to infect both dividing and non-dividing cells, as well as to stably integrate into a site-specific locus (q13.3) on chromosome 19 (AAVS1), conferring theoretical long-term and safe transgene expression. This property is lost when using rAAV and most of the transgene expression comes from episomal forms . In vitro , most of the cell types found in the RA joints can be transduced by rAAV2, including RA-FLS, chondrocytes and macrophages . A few studies used systemic injection of the rAAV vector, either intra-muscularly or peri-articularly , and showed high and therapeutically efficient transgene expression, detectable in sera for at least 4 months. The feasibility of direct intra-articular gene transfer to rat and mouse arthritic joints has been well demonstrated . The pattern of expression for AAV2 following local injection has been variable according to studies, from synovial lining cells to synoviocytes and chondrocytes and muscle and synoviocytes . More importantly, in rat and mouse models of arthritis, local transgene expression was shown to last for at least 7 months after joint delivery of rAAV5, mostly in synoviocytes . The serotype 5 capsid mediated rapid, high and stable gene transfer of a reporter gene into CIA mouse knee joints for at least 4 months ; Tak et al. showed that transgene expression was already detectable 7 days after injection and lasted for at least 4 weeks in the adjuvant-induced arthritis (AIA) rat model and Ghivizzani et al. reported that the rheumatic joints of horses were also efficiently transduced using rAAV5 . Using two different TNF-blocking agents (the tumour necrosis factor receptor (TNFR1)-mIgG1 fusion protein or a dimeric sTNFR2) as a proof of concept, two studies have demonstrated the feasibility of rAAV5-mediated gene therapy in mouse CIA and rat AIA models of arthritis. When the anti-TNF molecule expression is under a strong constitutive promoter (cytomegalovirus (CMV)), the antagonist molecule was rapidly (within 2 weeks), highly and stably expressed for 9 weeks when delivered intra-articularly by a CMV-driven rAAV5 vector. This was associated with a decrease in arthritis incidence and severity in both animal models. More importantly, the transgene was expressed under a nuclear factor-kappaB (NF-κB)-responsive promoter inducible by inflammation; the clinical effect was associated with a transient expression of the anti-TNF molecule, only detectable during disease flares. These results suggest that the local rAAV5-mediated gene delivery of a disease-inducible therapeutic agent may be instrumental in achieving successful treatment of RA by gene therapy . More recently, the feasibility of using rAAV5 for intra-articular expression of siRNA sequences targeting TNF to interfere with an ongoing arthritis has been demonstrated in the mouse CIA model . The rational design of AAV capsid mutants, and strategies such as the use of self-complementary vector genomes , might increase the potential of using AAV vectors for RA gene therapy even more.