Immunomodulation of autoimmune arthritis by pro-inflammatory cytokines

Abstract

Pro-inflammatory cytokines promote autoimmune inflammation and tissue damage, while anti-inflammatory cytokines help resolve inflammation and facilitate tissue repair. Over the past few decades, this general feature of cytokine-mediated events has offered a broad framework to comprehend the pathogenesis of autoimmune and other immune-mediated diseases, and to successfully develop therapeutic approaches for diseases such as rheumatoid arthritis (RA). Anti-tumor necrosis factor-α (TNF-α) therapy is a testimony in support of this endeavor. However, many patients with RA fail to respond to this or other biologics, and some patients may suffer unexpected aggravation of arthritic inflammation or other autoimmune effects. These observations combined with rapid advancements in immunology in regard to newer cytokines and T cell subsets have enforced a re-evaluation of the perceived pathogenic attribute of the pro-inflammatory cytokines. Studies conducted by others and us in experimental models of arthritis involving direct administration of IFN-γ or TNF-α; in vivo neutralization of the cytokine; the use of animals deficient in the cytokine or its receptor; and the impact of the cytokine or anti-cytokine therapy on defined T cell subsets have revealed a paradoxical anti-inflammatory and immunoregulatory attributes of these two cytokines. Similar studies in other models of autoimmunity as well as limited studies in arthritis patients have also unveiled the disease-protective effects of these pro-inflammatory cytokines. A major mechanism in this regard is the altered balance between the pathogenic T helper 17 (Th17) and protective T regulatory (Treg) cells in favor of the latter. However, it is essential to consider that this aspect of the pro-inflammatory cytokines is context-dependent such that the dose and timing of intervention, the experimental model of the disease under study, and the differences in individual responsiveness can influence the final outcomes. Nevertheless, the realization that pro-inflammatory cytokines can also be immunoregulatory offers a new perspective in fully understanding the pathogenesis of autoimmune diseases and in designing better therapies for controlling them.

1. Introduction

Pro-inflammatory cytokines such as tumor necrosis factor-α (TNF-α), interferon-γ(IFN-γ), interleukin-1β (IL-1β), IL-6, and IL-17 play a vital role in the pathogenesis of rheumatoid arthritis (RA), which is characterized by chronic inflammation of the synovial tissue, joint dysfunction, and tissue damage in the joints [1–5]. Collectively, these cytokines facilitate the recruitment of leukocytes into the joints to maintain chronic inflammation; induce the proliferation of synovial fibroblasts that leads to pannus formation; and contribute to the processes of angiogenesis as well as cartilage and bone degradation in the course of arthritis [3, 6–9]. The roles of 3 of the pro-inflammatory cytokines, namely TNF-α, IFN-γ and IL-17, in autoimmune arthritis are discussed in detail below. Macrophages, monocytes, and CD4+ T helper 1 (Th1) cells produce TNF-α, a key driver of inflammation [9, 10].

Neutrophils, endothelial cells, and fibroblasts are among other cell types that can serve as a source of this cytokine. TNF-α acts on macrophages to enhance phagocytosis as well as the production of other pro-inflammatory cytokines and prostaglandin E2 (PGE₂) [9, 10]. It also serves as a chemoattractant for neutrophils, and induces chemokine expression on endothelial cell lining to facilitate trans-endothelial migration of neutrophils. TNF-α acts on fibroblast-like synoviocytes (FLS) to induce their proliferation and pannus formation, and upregulates collagenase and matrix metalloproteinases (MMPs), which participate in cartilage damage. TNF-α also activates osteoclasts, which promote bone demineralization [9, 10]. TNF-α also has systemic effects such as fever and cachexia. Regarding IFN-γ the natural killer (NK) cells, Th1 cells, CD8+ cytotoxic T cells, NK T (NKT) cells, and innate lymphoid cells 1 (ILC1) are the primary source of this cytokine [11–13]. Subset of dendritic cells (DCs) and B cells are among other cellular sources of IFN-γ. Like TNF-α, IFN-γ enhances chemokine expression for leukocyte recruitment by facilitating their transfer through the endothelial layer. IFN-γ also activates macrophages and FLS to increase antigen presentation; promotes T_h1 differentiation; and activates NK cells and inducible nitric oxide synthase (iNOS) [11, 12, 14]. Over the past decade or so, significant attention has been focused on IL-17, which has been shown to play a critical role in the pathogenic processes involved in arthritis in both RA patients [5, 15–19] and animal models of RA [20–24]. Th17 cells are one of the major sources of IL-17 in autoimmune arthritis [18]. The CD8+ T cells [24], γδ T cells [23–25], ILC3 [26], and other cell types [27] may also contribute IL-17 at the site of arthritic inflammation. IL-17 acts on FLS and other cells to increase the production/activity of other pro-inflammatory cytokines; of chemokines that attract T cells, macrophages, neutrophils and other cells into the joints; of new blood vessels (angiogenesis); and of osteoclast MMPs that contribute to joint damage [15, 17, 28–31].

While the pro-inflammatory cytokines can upregulate each other in the short term to promote acute inflammation, there are various negative feedback loops to dampen the inflammatory response with the progression of inflammation. For example, TNF-α can be translationally repressed by micro-RNAs. Both transgenic mice having the human TNF-α transgene with altered 3′ untranslated region (3′UTR) site [32] and mutant mice that lack 3′ adenylate-uridylate (AU)-rich element (ARE) in the TNF gene [33], a modification that prevents translational repression, develop spontaneous chronic polyarthritis. TNF receptor 1 (TNFR1) (also known as TNFR-I or p55) can also be cleaved and thus rendered soluble, which not only prevents further signaling but also permits soluble TNFR (sTNFR) to bind and sequester TNF-α. Interestingly, the prevention of TNFR1 shedding can lead to spontaneous development of arthritis [34]. For IL-17, typically IL-23 (produced by macrophage and dendritic cells) enhances IL-17 secretion, thus forming the IL-23/IL-17 axis [24]. Nevertheless, IL-17 can negatively regulate IL-23 production [35], and thus can be self-limiting. In addition, IFN-γ can inhibit IL-17 production [20, 36]. Furthermore, the production of TNF-α, IFN-γ and IL-17 can be modulated by the action of CD4+ CD25+ T regulatory (Treg) cells on the T helper cell subsets (Th1 and Th17) that produce these cytokines. (Additional details on T cell subsets are given below.) These self-regulatory mechanisms are supplemented by other mechanisms mediated via IFN-γ (Fig. 1, Table 1) and TNF-α (Fig. 2, Table 2) to control arthritic inflammation [37, 38], and these mechanisms are discussed below.

Figure 1. IFN-γ-mediated regulation of autoimmune arthritis.

IFN-γ has been shown to inhibit Th17 differentiation; induce IL-27 secretion (from DCs and macrophages) that can inhibit Th17 cells either directly or indirectly via induction of IL-10-secreting Tr1 cells; promote the differentiation of iTreg cells and enhance their function; prevent FLS from proliferating in response to TNF-α; and inhibit the secretion by FLS of various factors (e.g., collagenase, GM-CSF, MMPs, and PGE2) that promote inflammation as well as bone and cartilage damage.

Table 1.

IFN-γ-mediated modulation of autoimmunity

Models	Species	Mechanisms	References
CIA, AA	Rat, Mouse	IFN-γ suppresses IL-17 production.	20, 37, 64
CIA, EAE, EAM	Human, Mouse	IFN-γ-/IFN-γ-R deficient mice develop more severe CIA or EAE, in part because of impaired Treg activity. Similarly, aggravation of EAM involves expansion of activated T cells.	20, 69–72, 79, 85, 91
EAE, GVHD, Colitis	Mouse	IFN-γ can promote the conversion of naïve T cells into inducible Treg cells and/or enhance Treg activity.	85–88
RA-FLS, AIA, monocytes	Human, Mouse	IFN-γ has been shown to inhibit TNF-α dependent proliferation of synoviocytes, collagenase production, and GM-CSF secretion; neutrophil influx into the joints; PGE2 release and PGE2 receptor expression; and TLR- or IL-1β-induced MMP production.	92–96
AA, EAE, EAU	Rat, Mouse	IFN-γ enhances IL-27 secretion by macrophages and dendritic cells. IL-27 in turn inhibits osteopontin expression as well as the development of Th17 cells, but induces Tr1 cells	36, 80, 81
EAE	Mouse	IFN-γ induces ER stress response pathway in oligodendrocytes and prevents immune mediated damage during EAE. This protection is dependent on pancreatic endoplasmic reticulum kinase (PERK).	62
NOD (T1D)	Mouse	IFN-γ can induce apoptosis of diabetogenic T cells.	78
RA patients	Human	Direct administration of IFN-γ has shown some benefits in RA patients without significant side effects.	98–104
CIA, PGIA	Mouse	Direct injection of IFN-γ into the joints of mice after CIA induction exacerbates disease. In the PGIA model, IFN-γ is the central cytokine mediating pathology, while IL-17 can contribute to the disease process in the absence of IFN-γ. Furthermore, IL-27 induces Th1 response that drives the disease pathology in PGIA.	14, 73, 83

Figure 2. TNF-α mediated regulation of autoimmune arthritis.

TNF-α can induce the apoptosis of arthritogenic Th1/Th17 cells; inhibit the expression of shared p40 subunit of IL-12 and IL-23 in DCs and prevent the differentiation of Th1 and Th17 cells; and signal nTreg cells via TNFR2 and enhance their function. Furthermore, Adalimumab can bind to the membrane-bound TNF-α (mTNF-α) on monocytes of RA patients leading to upregulation of surface mTNF-α on the monocytes. This causes increased interaction of monocytes with Treg cells through TNFR2. This in turn leads to the expansion of Treg cells and an increase in their suppressive activity, along with a reduction in Th17 cells, which results in an altered Treg/Th17 cell ratio in favor of immune regulation.

Table 2.

TNF-α-mediated modulation of autoimmunity

Models	Species	Mechanisms	References
AA	Rat	Direct administration of TNF-α downmodulated AA by inhibiting IFN-γ production by arthritogenic T cells, but without elevating Treg cell number, soluble TNFR1, or indoleamine 2,3 dioxygenase (IDO) activity.	38
NOD (T1D)	Mouse	TNF-α can induce apoptosis of diabetogenic T cells.	78
CIA	Mouse	TNF-α can suppress Th1/Th17 response via inhibition of the p40 subunit of IL-12/IL-23; TNF-α blockade can prevent the migration of pathogenic T cells into the joints despite an increase in the Th1/Th17 cells in the periphery.	125
CIA, EAE	Mouse	TNFR1-deficient mice, TNFR1/2 double knock-out mice, or mice treated with TNFR1-selective blockade, all develop less severe disease, while TNFR2 knock- out mice develop more severe disease. TNFR1- selective blockade enhanced the activation and expansion of Treg. Furthermore, the nTreg cells express high levels of TNFR2, and their function is enhanced by TNFR2 signaling, which stabilizes Foxp3 expression.	129, 135, 136
Treg cells	Mouse, Human	Both positive and negative influences of TNF-α on the generation and/or function of Treg cells have been observed in different studies. In addition, some disparities on the effects of TNF-α on mouse versus human Treg cells have been reported.	112, 114–120
RA patient’s monocytes	Human	Anti-TNF-α antibody (adalimumab) can bind to the membrane-bound TNF-α (mTNF-α) on monocytes and cause upregulation of mTNF-α; and mTNF-α binds to TNFR2 on Treg cells to expand them as well as increase their suppressive activity coupled with reduction in Th17 cells.	113, 140
RA patients	Human	TNF-α can inhibit Treg activity in RA patients by activating protein phosphatase 1 (PP1) through NF-κB pathway, which in turn dephosphorylates Ser418 in the DNA-binding domain of Foxp3, leading to reduced Treg function.	119

To fully understand the roles of pro-inflammatory cytokines in autoimmunity, it is essential to consider the characteristics of, and the balance between, different T cell subsets, as well as the plasticity of T cell subsets [39–44]. Activation of naïve T cells under defined cytokine environment conditions facilitates the generation of distinct CD4+ T cell subsets that are characterized by the production of specific cytokines and expression of particular transcription factors. Among the T helper subsets, Th1 cells produce IFN-γ and TNF-α and express the transcription factor T-box transcription factor (T-box 21; also known as TBX21 or T-bet); Th2 cells produce IL-4 and express GATA binding protein 3 (GATA3); and Th17 cells produce IL-17 and express retinoic acid-related orphan receptor γt (RORγt; in humans, RORC) [39–43]. In contrast, Treg cells produce transforming growth factor-β (TGF-β) and IL-10 and express forkhead box P3 (FoxP3). The Th1 and Th2 cells can mutually cross-regulate each other, and Treg cells can suppress the activity of above-mentioned T helper subsets [39–44]. Interestingly, there is a reciprocal development of Th17 and Treg cells from naïve T cells, with TGF-β favoring Treg cell development and TGF-β and IL-6 facilitating Th17 cell development [41]. In RA, earlier studies showed that Th1 cells are enriched in the joints of these patients [45–47]. Subsequent studies revealed similar findings for Th17 cells in the joints of RA [48] and juvenile idiopathic arthritis (JIA) patients [49]. As previously observed for the Th1/Th2 imbalance, an imbalance in the Th17/Treg cell ratio has been suggested to be a critical factor in the pathogenesis of autoimmunity, as well as a target of new therapeutic approaches aimed at re-setting this balance [5, 39–43]. Furthermore, a subset of the Th17 cells infiltrating the joints of JIA patients were found to be of a dual Th1/Th17 phenotype that expressed both T-bet and RORC2 [50]. In addition, the conversion of Th17 cells into cells of a dual Th17/Th1 or Th1 phenotype was demonstrated in vitro in that study. The Th17 cells that express IFN-γ are associated with immune pathology in arthritis and multiple sclerosis (MS) [50, 51]. However, in mice, the expression of T-bet and IFN-γ has also been shown in natural Treg (nTreg) cells [52]. Therefore, further studies are needed to fully define the role of IFN-γ in situations involving plasticity of T cell subsets. In another study, retinoic acid and retinoic acid receptor-α have been shown to be critical for Th1 development and for repression of the genetic program for Th17 cell development [53].

Most of the discussion below is focused on autoimmune arthritis, particularly RA and its animal models. However, for completeness and relevant comparison, examples of a few other immune-mediated diseases are also discussed at appropriate places.

2. Regulatory roles of pro-inflammatory cytokines IFN-γ and TNF-α in adjuvant arthritis

Adjuvant arthritis (AA) is a well-characterized experimental model of human RA, and it can be induced in Lewis (LEW) rats by immunization with heat-killed M. tuberculosis H37Ra (Mtb) [54, 55]. The disease manifests as a polyarthritis, and it appears within about 8–10 days after Mtb injection. After reaching the peak phase, which lasts for about 4–5 days, there is a spontaneous regression of arthritis over the next 10–12 days. Arthritic rats raise T cell response against mycobacterial heat-shock protein 65 (Bhsp65) following Mtb injection [37, 55]. The epitope region 180–188 (B180), which is nested within the longer sequence 177–191 (B177), represents the arthritogenic determinant of Bhsp65 [37, 55]. Arthritic LEW rats also develop T cell response to self (rat) hsp65 (Rhsp65) [54, 55]. Most information on Rhsp65 relates to its immunoregulatory role in AA [54], although it has also been proposed that crossreactivity between self and foreign Hsp65 might be involved in disease induction [55]. However, the latter phenomenon has not yet been fully addressed and needs further work. We previously showed that unlike the LEW rats, the Wistar Kyoto (WKY) rats of the same major histocompatibility complex (MHC) haplotype are resistant to AA induction [37, 55].

Our previous studies revealed that the T cells against defined determinants within Bhsp65, namely the Bhsp65 C-terminal determinants (BCTD), as well as those within its self-homolog, namely the Rhsp65 C-terminal determinants (RCTD), are capable of downregulating AA [54, 55]. Examination of the cytokine secretion profiles showed that surprisingly, the disease-protective T cells against the C-terminal determinant(s) secreted predominantly Th1-type cytokines [37, 38, 56]. Furthermore, LEW rats (AA-susceptible) had increased IFN-γ and TNF-α response during regression from arthritis, while WKY rats (AA-resistant) had a similar type of response (Th1) but temporally it was detectable early after a potentially arthritogenic challenge (Mtb injection) [37, 38]. These results indicated that there was a positive correlation of enhanced Th1 response with recovery from AA in LEW rats as well as protection against AA in WKY rats. Our subsequent studies demonstrated that the treatment of rats with IFN-γ or TNF-α induced protection against AA [36–38, 57]. The results of these studies and the mechanisms by which the two key Th1-response related cytokines, IFN-γ(Fig. 1, Table 1) and TNF-α (Fig. 2, Table 2), regulate autoimmune inflammation are described below. Also discussed are studies by other investigators demonstrating the disease-protective effects of IFN-γ and/or TNF-α in AA, collagen-induced arthritis (CIA), and few other models of immune-mediated diseases.

3. IFN-γ-induced immune regulation

During AA, the T cells reactive against an arthritogenic determinant (B177) of Bhsp65 secrete moderate levels of IFN-γ, while expressing high levels of IL-17 [37]. Pre-immunization of LEW rats with an AA-modulatory peptide (R465) containing amino acid residues 465 to 479 of Rhsp65, as well as the adoptive transfer of R465-specific T cells given separately, not only alleviated arthritis but also decreased the level of IL-17 expression in B177-reactive T cells [37]. Surprisingly, R465 immunization induced predominantly Th1 cells as did priming with a mixture of BCTD/RCTD, the C-terminal determinants of Bhsp65/Rhsp65, respectively [37, 56]. Therefore, higher levels of IFN-γ secretion from the T cells reactive against the immunomodulatory BCTD/RCTD suppressed IL-17 response apparently by inhibiting the activation/development of arthritogenic Th17 cells. This reasoning is supported by the observation that the Th1, Th2, and Th17 cells can modulate each other’s development/activity in part via specific cytokines secreted by them [24, 58]. IFN-γ may play a role in the induction and maintenance of autoimmune inflammation under certain set of conditions [1, 59, 60], yet be protective against arthritis and other autoimmune diseases under another set of conditions [61–63]. In this context, we proposed a model in which a particular threshold of IFN-γ was required for the initiation of inflammation, whereas secretion of a critical, higher level of IFN-γ instead triggered regulatory mechanisms to suppress the ongoing disease [6, 37]. (The same model also was applicable to the dual role of TNF-α discussed below.) While simplistic, this model helps comprehend the dual action of IFN-γ, even though the body of knowledge regarding the cellular sources of these cytokines and their targets during arthritis, as well as the plasticity of T helper subtypes during autoimmune disease, has broadened over time.

Two of our subsequent studies have further validated the immunoregulatory role of IFN-γ in AA. In one study, we examined the impact of T cell tolerance induction by soluble Bhsp65 on the severity of AA [64]. The treatment of rats with Bhsp65 prior to arthritis induction led to significant reduction in the severity of subsequent arthritis. The T cells of Bhsp65-tolerized rats showed increased production of IFN-γ coupled with reduced IL-17 expression, as well as evidence of T cell anergy [64]. However, there was no measurable effect either on the production of anti-inflammatory/immunomodulatory IL-4 and IL-10, or on the frequency and suppressive activity of Treg cells. Moreover, there was no significant effect on the indoleamine 2,3 dioxygenase (IDO)-tryptophan pathway [64]. In another study discussed below in more detail, we showed that IFN-γ can induce IL-27, and that both these cytokines can downmodulate the course of AA [36].

We have described above our studies showing that IFN-γ a key Th1 cytokine, can downmodulate AA. Similar results have been reported by other investigators showing that treatment with IFN-γ reduced the severity of arthritis and facilitated recovery from this disease in the AA model [65]. Another study revealed that treatment with anti-IFN-γ antibodies to rats prior to active AA induction [66] or passive AA induction (via adoptive transfer of arthritogenic lymphoid cells) [67] enhanced disease severity. In a different study in AA, it was shown that the disease-regulating antigen-primed T cells produced higher concentration of IFN-γ than that of IL-10 [68]. The protective effect of IFN-γ was also evident from a series of studies in the mouse CIA model. Mice deficient in IFN-γ or IFN-γ receptor developed more severe arthritis than control mice [20, 69–72]. Furthermore, IFN-γ was shown to inhibit IL-17 production and thereby to regulate susceptibility to CIA. Collectively, these studies in the CIA model showed that IFN-γ inhibited IL-17 production, and that in the absence of IFN-γ IL-17 had stimulatory effects on granulopoiesis, neutrophil infiltration and bone destruction [20, 69–72]. These studies contrast with couple others showing a pathogenic effect of IFN-γ in arthritis. Injection of IFN-γ into the joints was shown to exacerbate the severity of arthritis in mice having CIA [14]. Similarly, IFN-γ was shown to be the dominant cytokine mediating immune pathology in proteoglycan-induced arthritis (PGIA) in mice [73]. However, in the absence of IFN-γ IL-17 contributed to the disease process. Above studies highlight the complexities of IFN-γ action under different conditions and in different experimental model systems [6, 74].

Other investigators have offered additional evidence for the immunoregulatory role of IFN-γ in experimental models of multiple sclerosis (experimental autoimmune encephalomyelitis (EAE)) [58, 61, 62, 75, 76], uveitis (experimental autoimmune uveitis (EAU)) [59, 63], insulin-dependent diabetes mellitus (IDDM) (Type 1 diabetes (T1D)) [77, 78], and myocarditis (experimental autoimmune myocarditis (EAM) [79]. Various mechanisms that have been proposed to explain the disease-protective attributes of IFN-γ in these studies include, but are not limited to, suppression of the initiation of T cell response, control of the expansion of activated T cells, induction of apoptosis in immune cells infiltrating the target organ, and inhibition of IL-17 production by the T cells against the disease-related antigens. Two additional mechanisms are elaborated below: the interplay between IFN-γ and IL-27 and the integrated stress response (ISR) in IFN-γ-induced protection against autoimmunity.

We [36] and others [80] have observed that IFN-γ can enhance the production of IL-27 in AA [36], EAE [80], and uveitis [81]. In AA, IFN-γ secretion by Th1 cells enhances IL-27 production by the spleen adherent cells (which are comprised of macrophages and dendritic cells), and both of these cytokines directly suppress the development of Th17 cells (Fig. 1). In addition, IFN-γ, which reduces IL-1β production by macrophages, has been shown to be increased in WKY rats compared to LEW rats in the incubation phase of AA after Mtb injection [82]. IL-1β is a vital contributor to arthritis development and progression, and therefore, a reduction in IL-1β levels by IFN-γ represents one of the mechanisms of IFN-γ-induced protection against arthritis. A reverse effect, namely the induction of IFN-γ by IL-27 leading to immune pathology, has been observed in mice with PGIA [83]. In EAE, it was shown that IFN-γ acted on dendritic cells to induce IL-27 production and inhibit osteopontin expression [80]. These cytokines in turn modulated the function of DCs such that IL-27 facilitated the induction of IL-10-producing T regulatory (Tr1) cells, whereas osteopontin inhibition reduced the IL-17-mediated T cell response. Acting in concert, these changes resulted in the attenuation of EAE progression [80]. In the EAU model in IFN-γ^−/− mice, adoptively transferred IFN-γ-producing NK cells have been shown to suppress uveitis by interacting with and inducing IL-27 secretion by CD11c⁺ DCs in the draining lymph nodes [81]. These DCs in turn promoted Tr1 cells to secrete IL-10 and inhibit Th17 response. However, for autoimmune arthritis, it remains to be examined whether NK cells contribute to the total pool of IFN-γ levels either in the lymph nodes or synovial tissue/fluid. On the other hand, the frequency of ILC1 has been shown to be increased in early RA patients that have biomarkers for RA but without any signs of the disease [84]. This finding suggests a possible role of ILC1 in RA, although more studies are needed to determine if they play any significant role in the control of inflammation via IFN-γ.

A study in EAE brought out a new aspect of IFN-γ-mediated protection against tissue damage [62]. It involved the induction of the endoplasmic reticular stress response pathway, which protected mature oligodendrocytes from immune-mediated damage in the course of EAE. This effect was observed when IFN-γ was administered to mice given an encephalitogenic challenge, but before the onset of EAE. Furthermore, the observed effect on the survival of mature oligodendrocytes was shown to be mediated via activation of the pancreatic endoplasmic reticulum kinase (PERK), which coordinates the integrated stress response (ISR) [62]. However, this effect (ISR) did not show any correlation with the effects of IFN-γ on the immune system, including cellular migration into, or cytokine production within, the CNS.

Foxp3-expressing Treg cells, which are known to mediate their suppressive function via cell-cell contact and secreted cytokines IL-10 and TGF-β, have also been shown to produce IFN-γ in an inflammatory milieu, including a Th1-type environment [85, 86]. Furthermore, IFN-γ thus produced contributes to the suppressive activity of these Treg cells. It was proposed that rapid and transient production of Treg cells permitted them to inhibit effector T cell proliferation and prevent further T cell activation in part by modulating the function of antigen presenting cells (APCs) [86]. Studies in EAE, graft-versus-host disease (GVHD), and few other experimental models have revealed that IFN-γ can induce the conversion of naïve CD4+CD25− T cells into Treg (CD4+CD25+Foxp+ T) cells and/or enhance the suppressive activity of these Treg cells (Fig. 1) [85–88]. Such converted Treg cells are referred to “induced Treg” (iTreg) cells to differentiate them from naturally occurring thymic Treg (nTreg) cells. The ability of IFN-γ to convert naïve T cells into iTreg cells was also evident in in vitro systems using murine or human T cells [85]. This explanation helped understand the observations of enhanced severity of EAE in IFN-γ-deficient mice [85] as well as IFN-γ-induced protection in GVHD [88]. IFN-γ has been shown to be necessary for long-term allograft survival [89]. In another study, alloantigen-specific Treg cells secreting IFN-γ were shown to induce protection against lethal GVHD [88]. Furthermore, neutralization of IFN-γ by specific antibody or the use of Treg cells from IFN-γ-deficient mice abrogated the beneficial effect of Treg cells against GVHD. In a study on human kidney transplant, serum IFN-γ and the frequency of IFN-γ-secreting Treg cells in biopsy specimen were reduced in patients with graft rejection compared to stable controls [90]. Using the colitis model, it was shown that IL-12 induced the conversion of Foxp3+ Treg cells that recognized microbiota antigen into IFN-γ-secreting cells, which were immunosuppressive and could afford protection against colitis [87]. In CIA, it was observed that the enhanced severity of arthritis in IFN-γ-receptor-deficient mice was associated with reduced suppressive efficacy of Treg cells following immunization with type II collagen [91]. Furthermore, this deficit could be counteracted by IFN-γ. Of note, unimmunized, naïve mice of that strain did not show any defect either in the number or suppressive action of Treg cells.

In addition to IFN-γ-mediated regulation of immune cells, IFN-γ can also directly affect the target organ. Arthritis is characterized by inflammation of the synovial tissue, formation of new blood vessels (angiogenesis), immune cell infiltration, and damage to bone and cartilage in the affected joint [3, 6–9]. These events are driven by a variety of mediators such as pro-inflammatory cytokines, chemokines, angiogenic factors (e.g., vascular endothelial growth factor; VEGF), and MMPs [3, 6–9]. The activity of some of these pathogenic mediators is counterbalanced by other defined biomolecules, for example, the tissue inhibitors of metalloproteinases (TIMPs) control the activity of MMPs. Taking together the results of several studies, it has been shown that IFN-γ inhibits: TNF-α-dependent synoviocyte proliferation, collagenase production, and granulocyte-macrophage colony-stimulating factor (GM-CSF) secretion (Fig. 1) [92]; neutrophil influx into the joints [93]; PGE2 release and PGE2 receptor (EP) receptor expression [94]; Toll-like receptor (TLR)-induced MMP expression [95]; and IL-1β-induced MMPs, thus altering the MMP/TIMP ratio [96]. Collectively, these effects of IFN-γ contribute to inhibition of the progression of arthritis. Further studies on IFN-γ-induced immune genes would lead to identification of additional novel targets of therapeutic utility [97].

Recombinant IFN-γ has been used as a therapeutic agent in RA patients during the 80’s and 90’s. Among the double blind controlled trials, while couple trials have shown statistically significant clinical efficacy of IFN-γ in the treatment of RA [98, 99], other trials showed a trend towards improvement but failed to show a statistically significant therapeutic result [100, 101]. A few earlier uncontrolled trials showed a beneficial effect of IFN-γ in a proportion of RA patients [102–104]. Overall, no major side effects of IFN-γ use were reported that might have precluded its testing in RA patients. Thus, despite extensive literature in the animal models of RA supporting the immunoregulatory role of IFN-γ mentioned above, there has been rather limited translation of that information into IFN-γ-based therapy of RA. It is hoped that new approaches based on increasing the half-life of IFN-γ and/or specific targeting of IFN-γ to FLS or other targets may be attempted in the future, and if found useful, further tested in RA patients.

4. TNF-α-induced immune regulation

TNF-α is a prototypic pro-inflammatory cytokine that drives inflammation in RA and many other diseases [9, 105]. TNF-α mediates its effects via TNF receptor 1 (TNFR1; also known as TNFRI or p55) and TNFR2 (also known as TNFRII or p75) [105, 106]. TNF-α is expressed as a type II transmembrane protein on macrophages, T and B cells, and NK cells, whereas TNFR1 and TNFR2 are expressed as type I transmembrane receptors but differ in their expression. TNFR1 is expressed on most normal and transformed cells, whereas TNFR2 is expressed on endothelial cells and immune cells. Another difference between the two receptors is that TNRF1, but not TNFR2, contains the death domain; accordingly, signaling via TNFR1 can lead to caspase-mediated cellular apoptosis. However, with some differences, both TNFRs can engage TNFR-associated factor 2 (TRAF2) and activate the p38 MAP kinase (p38MAPK), the activated protein-1 (AP-1), and the nuclear factor-kappa B (NF-kB) pathways, leading to inflammation and cell survival. Because of the differential expression of TNFRs, TNF-α signaling is primarily mediated via TNFR1 in majority of cells, but also via TNFR2 in immune cells. Furthermore, TNFR1 can be activated by both membrane-bound TNF-α (mTNF-α) and soluble TNF-α (sTNF-α), whereas TNFR2 can be optimally activated by mTNF-α.

TNF-α is known to mediate immune pathology in various autoimmune diseases, including RA, Crohn’s disease, and psoriasis [9, 105]. Accordingly, anti-TNF-α therapy is being used in these diseases [9]. However, subsets of patients with these disorders can suffer from aggravated disease when treated with anti-TNF-α drugs. Furthermore, such drugs can induce demyelination (as in multiple sclerosis; MS) or autoantibody production (as in lupus) [107, 108], and worsen the disease in subsets of patients having MS or psoriasis [107, 109]. On the contrary, treatment with TNF-α per se has been shown to be beneficial in animal models of autoimmunity, such as RA [38], lupus [110], and IDDM (T1D) [111]. In addition, there are conflicting reports on the influence of TNF-α on the generation and function of Treg cells [112, 113]; some reports emphasize a positive relationship between TNF-α and Treg cell expansion and/or function [112, 114–117], while others describe a negative effect of TNF-α on Treg cell number and activity [118–120]. There is additional disparity between murine and human Treg cells in regard to the effect of TNF-α on these cells [112, 113]. Additional details of this aspect of TNF-α are discussed below.

In our study in AA, we have observed an anti-inflammatory role of TNF-α [38, 57]. Lymph node cells (LNC) from LEW rats during recovery from AA as well as those of WKY rats early after Mtb immunization secreted TNF-α in response to Bhsp65 [38]. Therefore, we tested whether TNF-α by itself had any immunomodulatory activity. Surprisingly, exogenous administration of TNF-α downmodulated the severity of arthritis in LEW rats [38]. At the dosage tested, there was no evidence of elevated blood levels of Treg cells (which otherwise can suppress pathogenic effector T cells), soluble TNFR1 (indicating a shedding of the receptor, which could neutralize TNF-α), or anti-TNF-α antibodies (that might be generated as a result of an antibody response to the injected TNF-α and then neutralize it). Furthermore, TNF-α treatment neither increased the migration of T cells to the site of injection in the periphery (and thereby diverting the T cells away from the site of inflammation in the joints) nor did it elevate the level of indoleamine 2, 3-dioxygenase (IDO) expression (which can mediate tolerance induction or suppression of T cells) in splenic antigen-presenting cells. Instead, the main effect of systemic TNF-α injection was in the form of decreased IFN-γ production, specifically by the T cells against the arthritogenic determinant B177 of Bhsp65. Based on the work by other investigators [121, 122], we further suggested that the protective effect of TNF-α in arthritis might involve additional mechanisms such as TNF-α-mediated suppression of IL-12 secretion [121] and enhanced apoptosis of arthritogenic T cells [122] (Fig. 2). A similar effect of apoptosis of diabetogenic T cells by TNF-α and IFN-γ induced following adjuvant therapy has been reported in an experimental model of human IDDM [78].

An unexpected increase in Th1 and/or Th17 cells has been observed in RA patients following anti-TNF-α therapy [123, 124]; in CIA mice treated with anti-TNF-α as well as in TNFR1-deficient mice having CIA [125]; and in EAE mice treated with anti-TNF-α [126]. Studies in patients with RA and other rheumatic diseases have shown that in some patients, anti-TNF-α therapy can expand/activate Th1 and Th17 cells, leading to worsening of inflammation. In RA patients treated with infliximab (one of the anti-TNF-α therapies) [123], a positive correlation was observed between the expansion of Th1 and Th17 cells in the peripheral blood mononuclear cells (PBMCs) cultured with infliximab in vitro and a lack of clinical response to infliximab. Such expansion of Th1 and Th17 cells was not observed in PBMCs of RA patients who responded well to infliximab or in PBMCs of healthy controls. In another study [124], anti-TNF-α therapy led to an increase in the level of Th17 in PBMCs not only in RA patients, but also in those having ankylosing spondylitis (AS) and psoriatic arthritis (PsA). Apparently, such an unanticipated activation of Th1 and/or Th17 cells following anti-TNF-α therapy might contribute to the lack of responsiveness to therapy and/or worsening of the underlying inflammation and certain clinical symptoms. Indirectly, these observations also indicated that TNF-α has suppressive effects on Th1 and Th17 response. In fact, a study in the CIA model (discussed below) has demonstrated a novel aspect of TNF-α-induced control of Th1 and Th17 response via inhibition of p40 expression [125]. Furthermore, in our study in AA, we also showed that TNF-α-mediated suppression of AA involved inhibition of IFN-γ production by the T cells against the arthritogenic determinant B177 of Bhsp65 [38].

Examination of the effect of TNF-α blockade using TNFR-Fc or anti-TNF-α antibody in established CIA (which was induced in inbred, wild type mice) revealed that IFN-γ and IL-17 production by lymph node cells was increased [125]. Furthermore, a comparison of p55 (TNFR1)^−/− and p75 (TNFR2)^−/− mice revealed that Th1 and Th17 cells were increased after CIA immunization only in p55^−/− mice (but not in unimmunized mice), and that this effect was due in part to upregulation of IL-12 and IL-23, which share the p40 subunit and are involved in Th1 and Th17 development, respectively [125]. Interestingly, p40 blockade prevented the expansion of Th1 and Th17 cells in the periphery [125]. These results have unveiled one of the mechanisms of action of TNF-α in controlling Th1/Th17 response, namely the inhibition of p40 expression (Fig. 2). However, despite an increase in Th1 and Th17 cells following anti-TNF-α therapy, the arthritis severity was found to be reduced [125], and this was explained by the observation that TNF-α blockade prevented the pathogenic T cells from migrating into the joints. Based on these results, it can be speculated that in RA patients, an increase in the frequency of Th1 and Th17 cells in the periphery may not always lead to worsening of inflammation; instead, arthritis may be reduced in severity or remain unchanged.

A differential effect of the anti-TNF-α treatment in the periphery versus the target organ was also observed in EAE mice [126]. There was an increase in Th1 and Th17 in the periphery (spleen) but not in the target organ (central nervous system; CNS). Clinically, the onset of EAE was delayed and the incidence of disease was reduced, but without affecting the eventual severity of the disease that developed. In studies using a rat model of EAE, treatment with TNFR1-IgG (p55-IgG) was shown to prevent the development of the disease following an encephalitogenic challenge (active/passive EAE) [127, 128]. Another study in murine EAE model showed that mice deficient in p55 or both p55/p75 were resistant to EAE induction, whereas mice deficient in p75 developed aggravated EAE [129]. These results demonstrate the pathogenic effects of signaling via p55, but ameliorative effects of signaling via TNFR2. (This aspects of TNFR signaling is discussed below in more detail.) However, there is a disconnect between the effects of TNF-α blockade in mouse models of EAE compared with MS patients, and currently it is recommended that TNF blockers should be avoided in MS patients [130].

Interest in the role of TNFR2 in autoimmune pathogenesis and in therapeutic approaches based on either TNFR2 agonism or TNFR1 blockade specifically has been on the rise; the latter because there are many patients who do not benefit from the current anti-TNF-α therapies. There is evidence suggesting that TNFR2 may play an important role in the pathogenesis of autoimmunity. Genetic studies have linked TNFR2 polymorphisms to several autoimmune diseases, including RA [131–133] and ankylosing spondylitis [134]. Furthermore, TNFR1 blockade could suppress the development of autoimmunity by both selectively dampening the pathogenic T cell response and enhancing the Treg cell activity because of the residual signaling via TNFR2 [135]. Specifically, a selective blockade or genetic ablation of TNFRI has been shown to suppress CIA, and this effect was associated with the expansion and activation of Treg cells [135]. In comparison, the blockade of signaling via both TNFR1 and TNFR2 also resulted in the inhibition of arthritis development, but this effect was associated with relatively reduced level of Treg cells and unexpectedly enhanced level of T cell-derived cytokines compared to that after TNFR1 blockade. Apparently, a reduction in arthritis following the blockade of signaling via both TNFR1 and TNFR2 was owing to inhibition of the dominant pro-inflammatory effect of TNFR1. Taken together, these results show that TNFR1 is the major pro-inflammatory receptor for arthritis development in CIA. Furthermore, the blockade of TNFR1 signaling alone, without affecting TNFR2 signaling, can afford protection against arthritis [135].

TNFR2 expression can influence the activity of both Treg and CD4+ T effector (Teff) cells but in different ways [136, 137]. TNF-α has been shown to enhance Treg cell function in part through TNFR2 signaling [136, 138]. TNFR2 on Treg cells confers stability of FoxP3 expression in an inflammatory environment [115]. However, as discussed below, this enhancement may be restricted to natural Treg (nTreg) cells rather than inducible Treg (iTreg) cells [112]. In the case of Teff, TNFR2 helps them resist immunosuppression by Treg cells as well as undergo increased proliferation [137]. This effect perhaps might manifest under conditions when Treg cells also do not express high levels of TNFR2 and thereby are limited in their suppressive ability. A comparative study on the role of TNFR2 expression on the in vivo suppressive activity of murine nTreg and iTreg cells under inflammatory conditions revealed that TNFR2 expression was required for nTreg but not iTreg cells [112]. This was evident from the findings that TNFR2-deficient nTreg but not iTreg cells failed to suppress autoimmune inflammation in vivo. Furthermore, the requirement for TNRF2 expression in nTreg cells can be circumvented by preactivation with TGF-β [112]. These results expand the earlier findings that nTreg cells require activation at the sites of inflammation and depend on TNF-α, whereas iTreg cells are generated and activated in the lymph nodes and require TGF-β [139]. The role of TNF-α in selective activation of Treg cells and enhancement of their suppressive activity under inflammatory conditions is also supported by the results of another study using murine GVHD model of transplantation [138].

However, there also is evidence for the inhibitory effect of TNF-α on Treg cell activity in RA patients [119]. This effect is mediated via TNF-α-induced upregulation of protein phosphatase (PP1) through activation of the NF-kB pathway. PP1 in turn interacts with Foxp3 and dephosphorylates Ser418 in the DNA-binding domain of FoxP3 [119]. This dephosphorylation of Foxp3 led to reduced Treg cell activity. Interestingly, treatment with infliximab, an anti-TNF-α antibody, not only restored both Foxp3 phosphorylation and Treg cell activity, but also reduced PP1 expression as well as Th17 and Th1 cell numbers, thus altering the Treg/Th17 and Treg/Th1 cell ratio in favor of immune regulation. Contrary to the above study [119], it has been reported that TNF-α can enhance the expression of CD25 and FoxP3 in Treg cells co-cultured with IL-2 and TNF-α, and that Treg cells maintain their suppressive activity in the presence of TNF-α [117].

Additional insight into the impact of TNFR2 on Treg cell activity has been gained from a recent study on adalimumab, an humanized anti-TNF-α antibody [113]. Adalimumab can bind to the mTNF-α on monocytes of RA patients but not healthy controls [113]. This binding caused upregulation of mTNF-α on the monocytes and increased the interaction of monocytes with Treg cells through TNFR2 (Fig. 2). This interaction resulted in expansion of Treg cells and their suppressive activity, coupled with reduction in Th17 cells leading to an altered Treg/Th17 cell ratio. These outcomes involved IL-2/signal transducer and activator of transcription 5 (STAT5) signaling. The interplay between monocytes and Treg cells via increased mTNF-α offered a mechanistic explanation to the earlier observation that adalimumab treatment increased Treg cell numbers in RA patients [140]. Taken together, the above observations suggest that blocking TNFR1 without affecting TNFR2 might offer advantages over blocking of both receptors; the current anti-TNFα- therapies modulate signaling via both the receptors in immune cells.

5. Concluding remarks

The role of pro-inflammatory cytokines in promoting autoimmune inflammation is well established. What is surprising is their involvement in the attenuation of inflammation via active inhibition and/or immunoregulation. For example, using complementary approaches, several earlier studies in experimental model of RA have documented that IFN-γ has disease-protective activities. This action of IFN-γ was rather difficult to understand in the context of Th1/Th2 balance. Subsequently, information about newer cytokines (e.g., IL-17 and IL-27) and T cell subsets (e.g., Th17 and Treg cells) helped explain some of these observations. For example, IFN-γ can inhibit IL-17 response directly as well as via the induction of IL-27. Furthermore, IFN-γ can facilitate the generation of Treg (iTreg) cells. Enigmatically, sizable findings in experimental models of RA did not translate into an IFN-γ-based therapy in RA patients. Few studies showed a beneficial effect of IFN-γ in RA, but another study showed comparable effect of IFN-γ and placebo. Unlike for IFN-γ, there are relatively fewer studies in arthritis models that document the anti-inflammatory activity of injected TNF-α per se, but studies on TNFR signaling have yielded interesting information in this regard. Furthermore, TNF-α has been shown to enhance the activity of Treg cells in mice but most studies on human Treg cells have revealed the opposite. This issue needs further resolution. Nevertheless, signaling via TNFR2 has been shown to be anti-inflammatory. In addition, studies in RA patients given an anti-TNF-α therapy (Adalimumab) have unveiled additional mechanisms by which TNFR2-engagement via monocytes can contribute to the anti-inflammatory activity of this therapy. However, the disparate outcomes of anti-TNF-α therapy in different autoimmune diseases require further investigation and clarification. Taken together, now it is clear that IFN-γ and TNF-α have a dual role, pro- as well as anti-inflammatory. It is important to realize that the anti-inflammatory activity of these two cytokines is context-dependent at multiple levels, and therefore, it is not easy to generalize a particular set of conditions under which a predictable outcome can be expected for various immune-mediated diseases. It is hoped that additional studies to unravel the immunoregulatory attributes of IFN-γ and TNF-α would contribute significantly not only to advancing our understanding of the pathogenesis of RA and other autoimmune diseases, but also to devising better therapeutic approaches for these disorders.

Highlights.

• IFN-γ and TNF-α are prototypic pro-inflammatory cytokines involved in autoimmunity

• Paradoxically, both these cytokines may also exhibit immunoregulatory attributes

• Treatment with IFN-γ or TNF-α inhibits the progression of adjuvant arthritis in rats

• Both IFN-γ and TNF-α can induce Foxp3-expressing regulatory T cells in mice

• Signaling via TNFR2 has been shown to induce anti-inflammatory effects

Abbreviations

AA

Adjuvant-induced arthritis

BCTD

Bhsp65 C-terminal determinants

Bhsp65

Mycobacterial heat-shock protein 65

CIA

Collagen-induced arthritis

EAE

Experimental autoimmune encephalomyelitis

EAU

Experimental autoimmune uveitis

FLS

Fibroblast-like synoviocytes

Foxp3

forkhead box P3

GM-CSF

Granulocyte-macrophage colony-stimulating factor

GVHD

graft-versus-host disease

IFN-γ

interferon-γ

ILC

innate lymphoid cells

iTreg

induced Treg

JIA

Juvenile idiopathic arthritis

LNC

Lymph node cells

MMP

matrix metalloproteinases

Mtb

Mycobacterium tuberculosis H37Ra

mTNF

membrane-bound TNF

nTreg

natural Treg

PBMC

peripheral blood mononuclear cells

PGE2

prostaglanding E2

PGIA

Proteoglycan-induced arthritis

R465

Rhsp65 peptide 465 to 479

RA

Rheumatoid arthritis

RCTD

Rhsp65 C-terminal determinants

Rhsp65

Rat hsp65

RORγt

Retinoic acid-related orphan receptor γt

sTNF

soluble TNF

Teff

T effector

Th17

T helper 17

TNF-α

tumor necrosis factor-α

TNFR

TNF receptor

Treg

CD4+CD25+Foxp3+ T regulatory