Cytokines play a critical role in regulating lymphoid cell responses to most antigenic stimuli, and it is therefore very likely that these molecules profoundly affect the pathogenesis of autoimmune diseases. Consequently, investigations of the role of cytokines as effectors or predisposing elements in these diseases have received prominent attention for many years. Early findings from such studies, however, were frequently contradictory and, more importantly, difficult to incorporate into a conceptual framework. A new impetus in this area was provided by the discovery of Mosmann and colleagues that T cells could be conveniently divided into two main subsets, ie Th1 and Th2, with distinct arrays of cytokine secretion patterns and functions [1,2]. A hypothesis was subsequently formulated according to which cell-mediated autoimmune diseases, such as insulin-dependent diabetes mellitus, are induced by Th1 cells and their cytokines, and humorally-mediated autoimmune diseases, such as systemic lupus erythematosus, are induced by Th2 cells and their cytokines. As reviewed below, findings on the primary role of IFN-γ in the pathogenesis of murine lupus and other observations clearly refute this hypothesis.

Many studies have now clearly established that, after antigen recognition, cytokines at the priming site, together with other factors such as the type of antigen-presenting cell, antigen dose, expressed costimulatory molecules, affinity, and duration of exposure, direct a polarized differentiation of a T helper cell clone into either the Th1 or the Th2 type [2,3,4,5] (Fig. 1). Th1 cells secrete IL-2, IFN-γ, and TNF-β, while Th2 cells secrete IL-4, IL-5.

IL-10, and IL-13. Th1 cells protect against intracellular pathogens, activate phagocytes, induce IgG_2a antibodies, and promote delayed-type hypersensitivity responses, whereas Th2 cells protect against extracellular pathogens, activate eosinophils, induce IgE-mediated allergic reactions, and promote other humoral responses in which IgG₁ predominates. Certain cell-surface molecules have been reported to be differentially expressed in these two subsets, including the chemokine receptors CCR5, CXCR3, and CCR1 on Th1 cells and CCR3 and CCR4 on Th2 cells [6]. Some of these markers are not specific, since they are not exclusive to one or the other subset. For example, CCR5 and CXCR3 are also expressed by Th2 cells, albeit at lower levels than by Th1 cells [7]. Moreover, a recently identified unique T-cell subtype does not fit either the Th1 or Th2 definition, and yet it expresses CXCR5 and preferentially localizes in B-cell follicles, where it provides help for antibody responses [8,9]. In addition to certain chemokine receptors, both IL-12Rβ2 [10,11,12] and IL-18R [13] are selectively expressed by Th1 cells.

Similar Th1- and Th2-like polarized cytokine secretion patterns have now been described for CD8⁺ and γ T cells, natural killer cells, dendritic cells, macrophages, mast cells, eosinophils, and even B cells [14]. With regard to the latter, it was recently found that naïve B cells that were stimulated in an antigen-dependent fashion with polarized, cytokine-secreting T effector cells could be induced to differentiate into two types of B effector cells, ie Be1, secreting IFN-γ (and other cytokines), and Be2, secreting IL-4 (and other cytokines) [15]. Moreover, IFN-γ secreted by Be1 cells and IL-4 secreted by Be2 cells, while presenting antigen to naïve T cells, regulated Th1 and Th2 development, respectively. In recognition of the fact that specific patterns of cytokine production are now applicable to many cell types, the main types of cytokine polarization are now defined as type 1 or type 2 rather than Th1 and Th2 [14]. Apparently, the immune system employs the same overall pattern of cytokine production regardless of the cell type involved in a particular response or setting. Therefore, the assumption that effects observed in a particular disease after treatment with cytokine agonists or antagonists are mediated by the corresponding subset of T cells is an oversimplification. In fact, a variety of other cellular components might be engaged, either directly or indirectly.

The molecular events associated with type 1 or type 2 cytokine polarization have not been fully elucidated. It is known that Th1 differentiation driven by IL-12 (a product of activated macrophages and dendritic cells) requires the IL-12-responsive signal transducer and activator of transcription (STAT)4, while Th2 differentiation requires the IL-4-responsive transcription factor STAT6 (Fig. 1) [16]. IL-18 acts synergistically with IL-12 in inducing IFN-γ production, but signalling by IL-18 is mediated by the IL-1 receptor-associated kinase pathway, not STAT4, leading to the nuclear translocation of the NF-κB complex selectively in Th1 cells [17]. Additional proto-oncogenes, kinases, and transcription factors have been implicated in Th1/Th2 differentiation, including the interferon regulatory factor 1 and the T-box expressed in T cell protein for Th1 cells, and the c-Maf proto-oncogene and the GATA3 zinc finger protein for Th2 cells [16]. Moreover, the JNK/MAP kinase pathway is induced in Th1 but not Th2 effector cells [18].

We and others have used a variety of techniques to measure cytokine levels in lymphoid organs and affected tissues of mice predisposed to lupus and have catalogued a variety of perturbations, with some cytokines upregulated and others downregulated compared with controls [19]. The most consistent result was high levels of IFN-γ, particularly at the late stages of the disease, with the most abundant levels being observed in the MRL-Fas ^lpr lupus strain [20,21,22,23,24,25]. These increases in IFN-γ were documented at both the RNA and protein levels, as well as by ELISPOT assays and by cloning kidney-infiltrating T cells. Increased expression of IL-12 in kidney-infiltrating mononuclear cells and tubular epithelial cells of MRL-Fas ^lpr mice have also been reported [26,27,28].

In addition to spontaneous mouse models of lupus, transgenic mice with a normal background overexpressing IFN-γ in the epidermis (under the involucrin promoter) developed inflammatory skin disease, as well as a T-cell-dependent lupus-like syndrome with antinuclear autoantibodies and kidney deposits of immune complexes [29,30].

Concentrations of IFN-γ and IL-12 are also increased in the serum of lupus patients, particularly those in the active stages of the disease [31,32]. Moreover, in such patients, increases in the Th1/Th2 ratio have been detected in the peripheral mononuclear cells, and a predominance of Th1 cells is particularly evident in the blood and kidneys of patients with diffuse proliferative nephritis [33]. Mutations in the IFN-γR1 and R2 have been identified in some lupus patients, and a combination of the IFN-γR1 Met14/Val14 with the IFN-γR2 Gln64/Gln64 genotypes has been suggested to be a risk factor for this disease [34]. Nonetheless, in vitro experiments have shown higher IFN-γ expression in peripheral blood mononuclear cells of lupus patients than in controls after mitogenic stimulation [35]. Of interest is the observation that severe lupus-like disease was induced in some patients treated with IFN-γ for unrelated autoimmune diseases or myeloproliferative disorders [36,37].

Definitive evidence that IFN-γ is required for pathogenesis in murine lupus was provided in a series of studies with MRL-Fas ^lpr or B × W mice in which the IFN-γ or IFN-γR gene had been deleted [41,42,43,44]. These studies uniformly reported significant reduction of histologic and serologic disease characteristics and extended survival. The authors of one of these studies [42] made two notable observations. First, they found that hypergammaglobulinemia was maintained in MRL-Fas ^lpr mice in which the IFN-γ gene had been deleted, with a switch from IgG_2a to IgG₁ predominance, but that the dramatic decrease in levels of the predominant IgG_2a anti-dsDNA autoantibodies was not associated with a compensatory increase in IgG₁ anti-dsDNA subclass autoantibodies. A second, remarkable, finding was that glomerulonephritis and early death were prevented even in mice heterozygous for this deletion (IFN-γ^+/-; that is, with about 50% reduction in IFN-γ concentrations), even though autoantibody levels and renal deposits of immune complexes were the same as those in the wild-type MRL-Fas ^lpr mice [42]. These important findings suggest that therapeutic interventions to reduce IFN-γ levels in lupus may selectively affect certain autoimmune responses without significantly compromising the capacity to respond to exogenous antigens. Further, even partial reduction in IFN-γ might curtail its deleterious effects locally in the afflicted organs. Similar uncoupling phenomena between autoantibody production and local inflammatory responses have been described in B × W mice in which the gene for the FcRγ-chain had been deleted [45]. Moreover, our genome-wide studies have shown that some loci associated with glomerulonephritis or mortality do not cosegregate with those associated with the production of anti-dsDNA autoantibody [46], indicating that other autoantibody specificities or additional locally acting mechanisms may be necessary for the development of glomerulonephritis. Overall, the findings suggest that IFN-γ may exert disease-promoting effects at both a systemic, humorally-mediated axis and a local, cell-mediated axis.

Encouraged by the beneficial effects of experiments blocking IFN-γ in murine lupus, we developed a new strategy to intercept the activity of this molecule by intramuscular injections of a nonviral vector containing cDNA encoding the IFN-γR/IgG₁Fc fusion protein [47]. We used this inhibitor instead of the truncated receptor alone, because fusion molecules secreted as disulfide-linked homodimers have been shown to have much longer half-lives than the truncated IFN-γR (40 vs 1–3 h, respectively) [48,49], and dimeric IFN-γR/Fc fusion protein shows higher ligand avidity than single-chain receptors [50]. This therapy significantly reduced serum levels of IFN-γ and all disease parameters in MRL-Fas ^lpr mice when it was initiated before appearance of the disease, particularly when expression of this biomolecule was enhanced by electroporation at the injection site. Remarkably, this treatment arrested and often ameliorated disease even when the treatment was initiated when the disease was at an advanced stage, an unprecedented result. It is impressive that inhibition of a single molecule has such a profound effect on this multifactorial disease.

Because of the highly pleiotropic properties of IFN-γ, it is very difficult to identify the exact mechanism(s) by which blockade of this molecule curtails the development of murine lupus. The most likely possibility is reduced expression of MHC class I and II molecules on both mononuclear and tubular epithelial cells, leading to inefficient self-peptide presentation and responses [42]. Additional mechanisms may include reduced expression of other inflammatory molecules, such as intercellular adhesion molecule 1 and monocyte chemoattractant protein-1 [47].

The delivery of IFN-γ inhibitory molecules by intramuscular injection of plasmid vectors is simple, and appears to be nontoxic and safe. This approach for gene therapy of lupus circumvents several problems encountered with viral vectors [51], and may have advantages over the use of sIFN-γR, such as provision of a depot of genetic material for long-term expression of a biomolecule as well as its expression in afflicted organs due to migration of the injected DNA to distant sites [52].

Although demonstrations of the primary role of IFN-γ in the pathogenesis of murine lupus challenge the Th1/Th2 paradigm in autoimmunity, this finding should not be taken as evidence that Th2 cytokines are without influence in this disease. On the contrary, several studies have shown that manipulation of IL-4, IL-6, or IL-10 can also affect the progression of murine lupus. For example, treatment of B × W or MRL-Fas ^lpr mice with IL-4R or anti-IL-4 antibody resulted in reduced mortality and disease [53,54]. Conversely, IL-4 was shown not to be a necessary component in the BXSB male lupus disease [55] and, as a further complication, transgenic high expression of IL-4 under the influence of an immunoglobulin promoter in (NZW × C57B/6.Yaa)F₁ mice was protective [56]. In addition, administration of IL-6 promoted, and anti-IL-6 inhibited, the B × W and MRL-Fas ^lpr disease [57,58]. Finally, anti-IL-10 treatment ameliorated disease in B × W mice, presumably by upregulating TNF-α [59].

The discovery of the Th1-vs-Th2 dichotomy provided the paradigm according to which cell-mediated autoimmune diseases engage Th1 cells and their respective cytokines, while humorally mediated autoimmune diseases engage Th2 cells and their respective cytokines. Evidence reviewed above, primarily on the role of IFN-γ in lupus pathogenesis, appears to contradict this paradigm. Challenges to this hypothesis have also been provided by evidence from other autoimmune diseases. For example, IFN-γ was also shown to be required in the classic autoantibody-mediated disease myasthenia gravis, whereas IL-4 was not [60,61]. Furthermore, both Th1 and Th2 reportedly affect the severity of disease in the cell-mediated autoimmune diabetes of nonobese diabetic mice. Thus, depending on the experimental conditions, IFN-γ was both detrimental and beneficial [62,63], pancreatic IL-4 expression was protective [64], islet-specific expression of IL-10 accelerated disease [65] while systemic treatment with IL-10 prevented it [66], and deletion of the IL-10 gene was without effect [67] in this model. It appears, therefore, that both Th1 and Th2 cytokines can modify a given autoimmune disease depending on various factors, such as stage of the disease and timing of treatment, local vs systemic expression, and genetic background. Overall, it seems that the effects of cytokines in autoimmune syndromes cannot be dogmatically predicted, and their effects can be much more complex than the simplistic Th1-vs-Th2 definition dictates.