Interest in applying gene therapy to the treatment of rheumatic diseases began in the early 1990s with attempts to deliver cDNAs to the synovial linings of joints 2627. The basic premise is quite simple (Figure 2). The sustained intra-articular expression of a cDNA encoding a secreted anti-arthritic product will treat the joint locally without the need for readministration and avoid the peaks and troughs of traditional routes of drug delivery. No competing technology is able to do this. If gene transfer is sufficiently efficient, cDNAs encoding nonsecreted products are also possible. Treating individual diseased joints rather than the entire patient reduces costs and lowers opportunities for adverse systemic side effects. A number of different types of transgene have been suggested for this purpose, including those encoding cytokine antagonists, immunomodulators, antiangiogenic factors, apoptotic agents, antioxidants, inhibitors of mitosis, as well as molecules that modulate cell signalling and the activities of transcription factors (reviewed in 28).

By the time the first issue of Arthritis Research appeared, a phase I clinical trial was under way (Figure 1 and Table 1). This used a retrovirus (MFG-IRAP) to deliver the human interleukin-1 receptor antagonist (IL-1Ra) cDNA by an ex vivo protocol to the metacarpophalangeal (MCP) joints of patients with advanced RA 29. Among the strict safety requirements of this study was the need to recruit subjects who needed MCP joint replacement surgery, so that the genetically modified cells could be surgically removed 1 week after injection.

This study confirmed that genes could be safely transferred to human rheumatoid joints and expressed within them, at least for 1 week 30. Although several subjects reported symptomatic improvement, the study was not designed to measure efficacy. A small similar German study, involving just two subjects, is thus of interest because it included preliminary outcome measures based on pain and swelling, using a joint that did not receive the IL-1Ra cDNA as an intrapatient control. Both subjects responded to gene transfer, one of them dramatically so, and the clinical improvement lasted for the entire 4 weeks of the study, despite one subject experiencing flares in nontreated joints 31.

The occurrence of leukemia in humans as a result of insertional mutagenesis using retrovirus vectors, coupled to the high cost of ex vivo gene therapy using passaged autologous cells, has curtailed future trials of this kind. Instead, investigators are concentrating on in vivo gene delivery to joints. Based upon promising preclinical data in rabbits 32, Roessler and colleagues treated one subject with plasmid DNA encoding herpes simplex virusthymidine kinase and followed this with administration of ganciclovir to effect a genetic synovectomy. Although there were no adverse events associated with this procedure, the trial overlapped with the death of Jesse Gelsinger in 1999, which hindered recruitment, and the study was terminated. Since then, the emphasis for in vivo delivery to joints has shifted to AAV for the reasons described in the previous section.

There have been two clinical trials using AAV, both sponsored by Targeted Genetics Corporation (Seattle, WA, USA). The vector (tgAAC94) comprises AAV2 with a single-stranded DNA genome encoding etanercept. Expression is driven by a human cytomegalovirus immediate early promoter. The trials have been discussed in detail 23. In the first phase I study, 14 subjects with RA and 1 with ankylosing spondylitis were administered vector 33. Fourteen knee joints and one ankle were injected with 10¹⁰ or 10¹¹ virus particles per milliliter; knee joints received 5 mL and the ankle received 2 mL. A subsequent phase I/II study enrolled 127 subjects with a dose escalation of 10¹¹, 10¹², or 10¹³ virions per milliliter to be injected into symptomatic knee, ankle, wrist, MCP, or elbow joints. The protocol allowed subjects to receive a second injection of tgAAC94.

The phase I/II trial attracted considerable attention last year when a subject died soon after receiving a second injection of vector into her knee joint 23. The case aroused controversy because, in addition to receiving cDNA encoding etanercept, the subject was on adalimumab, having previously taken etanercept until this was discontinued because of a flare. The subject died from histoplasmosis, a known risk factor with anti-tumor necrosis factors (anti-TNFs), in conjunction with a massive retroperitoneal hematoma. After a lengthy investigation by the FDA and the Recombinant DNA Advisory Committee of the NIH, the trial was allowed to proceed in a slightly modified fashion. Preliminary efficacy data suggest that some subjects had symptomatic improvement in response to the gene treatment 34.

A number of groups are now interested in using AAV to deliver genes to joints. Research is focusing on the choice of serotype and the host immune response to the vectors. Serotypes 1, 2, 5, and 8 have attracted the most scrutiny. According to Apparailly and colleagues 35, AAV5 is superior to AAV1 or 2 in the knee joints of mice. This was confirmed in rats, and AAV2 and 5 were shown to have equal efficiency in transducing cultures of human synovial fibroblasts 36. Another study indicates the following order of preference: AAV2 > 1 > 5 > 8 37. However, when human synovial fluids were screened for pre-existing immunity to AAV, neutralizing antibodies to serotypes 1 and 2 were more common than antibodies to serotype 5, suggesting to Boissier and colleagues 37 that AAV5 may be more useful in humans, despite lower transduction efficiency. Humoral reactions to AAV2 were noted in the trial of tgAAC94, mentioned above, but possible cell-mediated immunity was not measured. Studies of AAV2-mediated gene delivery to the knee joints of rabbits confirm a neutralizing immune reaction that prevents redosing 18. AAV has been used to express soluble TNF receptors 38, beta interferon (IFN-β) 39, angiostatin 40, dominant negative Iκκβ (inhibitor of kappa B kinase β) 41, and IL-1Ra 18 in the joints of experimental animals, with an associated antiarthritic effect.

In most species, conventional AAV vectors containing a single-stranded DNA genome have only a modest ability to transduce articular tissues. Transduction efficiency can be enhanced by irradiation, a process that provokes second-strand synthesis 42. The need for the latter can be obviated with the use of scAAV, and recent findings confirm the superiority of these vectors in the rabbit knee joint 18. According to data from the same study, only 10% to 20% of AAV genomes that enter synovial fibroblasts appear in the nucleus. This identifies a second constraint to transduction efficiency that helps to account for the relatively high number of AAV virions (10⁴ to 10⁵ particles per cell) needed for useful levels of transgene expression. Proteosome inhibitors improve the nuclear uptake of AAV genomes in human synovial cells, leading to greatly enhanced transgene expression 43. In agreement with this, mutations to the AAV coat protein that prevent ubiquitination also increase transduction efficiency 44. According to Traister and colleagues 45, transgene expression from AAV vectors is increased in human synovial fibroblasts in the presence of inflammatory cytokines. A similar effect was reported some years ago by Pan and colleagues 4647 in rat knee joints but this has been difficult to reproduce 18.

For intra-articular gene therapy to be a clinical success, there is a need for extended periods of transgene expression. This has proved difficult in animal models. Recent work by Gouze and colleagues 48 identified immune reactions to non-homologous proteins as the major barrier to prolonged transgene expression. Using the rat knee joint as a model system, they showed that cDNA encoding a rat protein delivered by an immunologically silent vector can be expressed in a stable prolonged fashion. Of interest, long-term transgene expression does not require an integrating vector and is independent of the promoter. Instead, it relies on the presence of long-lived nonmitotic cells within certain dense collagenous tissues in and around the joint (Figure 3).

An ability to achieve long-term transgene expression opens the way for regulated expression. Two approaches have been investigated. One makes use of endogenous cues to ensure that the level of expression tracks disease activity within the joint. These strategies use inducible promoters based upon upstream regulatory sequences that control the expression of acute-phase proteins and inflammatory cytokines, such as IL-1 and IL-6 384950. A related method uses a sequence containing multiple nuclear factor-kappa-B (NF-κB)-binding sites 38. A second approach to regulated transgene expression uses exogenous molecules, such as doxycyline, to manipulate the level of production 515253. The latter approach provides greater insurance against inappropriate transgene expression as might occur during an infection. Though not strictly gene therapy, a related clinical trial injects decoy oligonucleotides that inhibit the activity of the transcription factor NF-κB into rheumatoid joints 54. So far, there have been no adverse events and some evidence of a clinical response in certain subjects (Tetsuya Tomita, personal communication).

In a polyarticular condition such as RA, an intra-articular gene therapy might require the injection of large numbers of joints. Moreover, a local gene therapy might not address systemic extra-articular manifestations of the disease. Thus, there is interest in a more general approach to therapy in which a transgene is introduced into a site where a secreted gene product will have access to the systemic circulation (Figure 2). Proof of principle has been established using intramuscular, intravenous, intraperitoneal, and subcutaneous routes of delivery by in vivo and ex vivo methods (reviewed in 28). Although this approach has obvious attractions, it provides only an incremental advance over what is already achieved by traditional methods of protein delivery and is accompanied by increased risk of adverse events. For these reasons, it has not achieved widespread popularity. One interesting possible exception, however, is the parenteral administration of naked DNA.

There are several reports in the refereed literature ascribing potent antiarthritic properties to plasmid DNA when delivered by intramuscular, intraperitoneal, intravenous, and intranasal routes 55565758596061. Because levels of transgene expression are low when DNA is administered in these ways, an alternative explanation for their efficacy in animal models of RA is needed. One possibility is the uptake of DNA by antigen-presenting cells (APCs) which then travel to sites of antigen presentation where sufficient transgene is expressed to modulate immune reactivity locally. This is an example of facilitated local therapy, described in the next section.

DNA can also be used to vaccinate. There are several examples using animal models of RA in which DNA vaccines that express arthritogenic antigens, such as heat-shock proteins 62, or mediators of arthritis, such as TNF 63, are protective. It is also possible to induce tolerance by DNA immunization in the absence of adjuvant 64. Though effective in animal models, such strategies may be risky in humans.

The ability to target multiple diseased joints selectively by a single parenteral injection is known as facilitated local therapy (Figure 2). This was first noted as a contralateral therapeutic effect in the knee joints of rabbits with bilateral antigen-induced arthritis 65. It occurs with both in vivo and ex vivo 66 gene delivery and is thought to reflect immune modulation via APCs that are exposed to appropriate transgene products as they present arthritogenic antigens to T lymphocytes (Figure 4).

Studies using the murine delayed-type hypersensitivity reaction as a model 67 showed that genetically modified dendritic cells and macrophages could migrate to sites of inflammation and inhibit immune-driven pathology in an antigen-specific manner. In subsequent studies, dendritic cells expressing IL-4 were shown to migrate to the paws of MHC-matched mice with collagen-induced arthritis and quell disease activity, even in established disease 68. The anti-arthritic effect was stronger than when the same adenovirus vector was used to deliver IL-4 systemically. A variety of additional transgenes, including IL-10, indoleamine 2,3-dioxy-genase, and IκB (inhibitor of kappa-B), are effective in this manner.

A related strategy produces selective ablation of autoreactive T lymphocytes by modifying APCs to express inducers of apoptosis on their cell surfaces. When the APC expresses an arthritogenic antigen to reactive T lymphocytes, the latter undergo apoptosis. Although this has been shown in murine models using Fas ligand 6970 as the transgene, TRAIL (tumor necrosis factor-related apoptosis-inducing ligand) is a better candidate because its receptor has more limited distribution, thus reducing opportunities for unwanted side effects. Impressive proof of principle has been demonstrated in murine collagen-induced arthritis using dendritic cells that express TRAIL 71. The therapeutic effect was improved by pulsing the dendritic cells with type II collagen before injection. The equivalent manipulation in RA will be complicated because the inciting antigens are not known, although a regulatory bystander effect could be achieved. The response to TRAIL gene transfer is enhanced when RNA interference is used to knock down expression of its decoy receptor, DcR2 72.

T lymphocytes also home to sites of inflammation and immune reactivity. Like APCs, they may be genetically modified and used to target multiple sites of disease by parenteral administration, although lymphocytes are more difficult to transduce than APCs. However, proof of principle has been shown in several animal experiments using IL-4, IL-10, IL-12 p40, and anti-TNF single-chain antibody as transgenes 7374. In most animal models, the arthritogenic antigen is known and T lymphocytes with the appropriate T-cell receptor can be used to maximize the effect. In RA, however, the inciting antigen is not known and enrichment is difficult. As one response to this limitation, Annenkov and Chernajovsky 75 engineered a T-cell receptor whose extracellular domain contains a type II collagen-binding motif.

Isis Pharmaceuticals (Carlsbad, CA, USA) sponsored phase I, IIa, and II clinical trials in which anti-sense RNA directed against TNF was injected intravenously and subcutaneously into subjects with RA. Anti-sense RNA was shown to traffic to the synovium of diseased joints, suggesting facilitated local delivery. The phase II study involved 157 subjects with RA who received 200 mg of anti-sense RNA twice a week, once a week, or once a fortnight. Subjects in the two highest dosing groups showed improvement in ACR20 (American College of Rheumatology 20% improvement criteria) scores. These studies were reported on the company's website 76 but were never published in the peer-reviewed literature. The company is no longer pursuing this project.

The recent emergence of RNA interference provides the ability to knock down cytokine synthesis in a highly specific fashion. Khoury and colleagues 77 have delivered short interfering RNA molecules targeted to IL-1, IL-6, and IL-18 in murine collagen-induced arthritis 77. Knockdown of each cytokine was effective in reducing the incidence and severity of disease, but a dramatic therapeutic effect was observed when all three were inhibited together.

When urate crystals are injected into subcutaneous air pouches on mice, they induce inflammatory responses of the type seen in human gout. Ex vivo delivery of prostaglandin D synthetase has a strong anti-inflammatory effect, suggesting that this could serve as the basis of a gene treatment for crystal induced arthropathy 86. A small clinical study indicates that recombinant IL-1Ra (Kineret^®) has a beneficial effect in human gout 87. This suggests an additional clinical target for the genetic therapies, discussed above, that presently use IL-1Ra cDNA to treat RA and OA.

Like diarthrodial joints, salivary glands are discrete isolated structures that lend themselves to local gene transfer. Vectors can be introduced through a cannulated duct and reach the luminal surfaces of the epithelial cells. Because the salivary gland is well encapsulated, there is little risk of vector escaping to nontarget organs. Many of the principles described above in the context of RA also apply to Sjögren syndrome 88.

Although adenovirus vectors transfer genes to the salivary glands very efficiently, AAV is proving to be the vector of choice because it is safe and noninflammatory. Serotypes 2 and 5 show promise, and efficacy has been demonstrated in animal models using IL-10 89 and vasoactive intestinal peptide 90 as the transgene products. Because the salivary gland has an exocrine function, it can also be used as a site of gene transfer for systemic delivery purposes 91. Gene therapy for Sjögren syndrome is reviewed in references 88.

Several of the strategies we have already discussed in the context of RA have also been applied to lupus. Unlike RA therapy, lupus therapy has not benefitted in a dramatic fashion from the introduction of biologics. Moreover, because lupus is accompanied by a large increase in mortality, the risk-to-benefit ratio is more favorable toward gene therapy. Lupus is thought to involve excess production of type 2 cytokines, so a number of investigators have introduced type 1 cytokines, such as IL-2 and IL-12. Encouraging results followed the intramuscular injection of plasmids encoding these cytokines in murine models 9293. Injection of plasmids encoding an IFN-γ receptor: Fc construct has also shown promise 94. In some experiments, the efficiency of transfection has been increased by electroporation, leading to efficacy with a cDNA encoding a dominant negative mutation of monocyte chemoattractant protein-1 95. The DNA vaccination approach has also worked using a cDNA encoding a consensus peptide from anti-DNA immunoglobulins.

Other investigators have used adenovirus to deliver the immunoinhibitory receptor PD-L1, TACI (transmembrane activator and CAML [calcium modulator and cyclophilin ligand] interactor, an inhibitor of B-lymphocyte stimulator [BLyS]), CTLA4, and a soluble form of the TGF receptor II 96979899. CTLA4 and CD40Ig have also been delivered in animal models using AAV8 100101. Data from animal models thus provide numerous examples of successful gene therapy in murine models of lupus. The challenge is to translate these into clinically useful human protocols. Gene therapy for lupus is reviewed in reference 101.

DNA vaccination has been used to generate antibodies to TNF with associated improvement in an animal model of antiphospholipid syndrome 102.