High mobility group box protein 1 in complex with lipopolysaccharide or IL-1 promotes an increased inflammatory phenotype in synovial fibroblasts

The highly conserved protein high mobility group box protein 1 (HMGB1) exerts vital functions in the nucleus of all eukaryotic cells. When tissue injury is inflicted and inflammation is induced, HMGB1 can be released extracellularly and can then convey inflammatory functions. Extracellular HMGB1 may induce cytokine production, up-regulation of adhesion molecules on endothelial cells and activation of dendritic cells and T cells [1‐11]. The reported presence of extracellular HMGB1 in multiple inflammatory conditions and the beneficial effects of HMGB1 blockade in preclinical models of inflammatory diseases have thus led to the acknowledgement of HMGB1 as an inflammatory mediator with pathogenic functions in several inflammatory diseases (reviewed in [12]).

HMGB1 interacts with the receptor for advanced glycated end products (RAGE), Toll-like receptor (TLR) 2 and with the TLR4 signalling complex. All three receptors are known to be involved in inflammatory processes and to possess the ability to activate NFκB translocation. RAGE-HMGB1 interaction has mainly been studied regarding induction of cell migration while HMGB1 interaction with TLR2 and TLR4 mediates immune activation. We recently reported that HMGB1-induced cytokine production in macrophages is mediated via TLR4 and requires a reduced cysteine with a thiol group in amino acid position 106, supplementing the findings of Kazama et al. that HMGB1 released from apoptotic cells contains an oxidized cysteine in position 106 that induces tolerance rather than immune activation [13, 14].

A second mechanism for the proinflammatory function of HMGB1 is due to the ability of HMGB1 to form complexes with inflammation-inducing agents such as LPS, IL-1β, CpG-DNA (short single-stranded synthetic DNA molecules that contain a cytosine followed by a guanine) and the TLR2-ligand Pam₃CSK₄. Such complexes have been demonstrated to strongly enhance cytokine production in cell cultures. Additionally, in an experimental model of systemic lupus erythematosus HMGB1 was detected in circulating nucleosome complexes and the necessity of HMGB1 for these complexes to be immunogenic and to induce production of anti-DNA antibodies were demonstrated [15‐20]. The molecular mechanism underlying the inflammatory activity of HMGB1 complexes and their ability to induce an enhanced response as compared to the partner molecule alone has not previously been addressed. Interestingly, it appears to be independent of the HMGB1 redox status as HMGB1, unable to induce cytokine production per se, still has the ability to induce such enhancement.

We and others have demonstrated an extracellular expression of HMGB1 in synovial tissue biopsies from rheumatoid arthritis (RA) patients and in joints from mice and rats with adjuvant-induced arthritis or collagen type II-induced arthritis [21‐24]. Additionally, extranuclear HMGB1 localisation has been described in synovial tissue from osteoarthritis (OA) patients and in bovine osteoarthritic cartilage specimens [25, 26]. Evidence for an active role of HMGB1 in arthritis pathogenesis is provided by studies demonstrating that a single injection of recombinant HMGB1 into knee joints of mice induces chronic synovitis [27] and, conversely, neutralisation of HMGB1 by treatment with antibodies or with a specific HMGB1 peptide antagonist significantly suppresses arthritis development in several studies [24, 28‐31].

Synovial fibroblasts (SFs) have been demonstrated to play a central role in arthritis pathogenesis, promoting both inflammation and bone and cartilage destruction [32, 33]. SFs display an activated phenotype with up-regulated expression of multiple TLRs and interleukin 1 receptor type I (IL-1RI) [34‐37].

We investigated whether the arthritogenic properties of HMGB1 could involve stimulation of SFs by HMGB1 complexes. We chose to study complexes formed by HMGB1 and endogenous mediators already described to be present in arthritic joints, that is, IL-1α and IL-1β, and with LPS which may also appear in arthritic joints [23, 38‐42]. We could demonstrate that SFs obtained from RA or OA patients responded to HMGB1 in complex with IL-1α, IL-1β or LPS, respectively, with enhanced production of tumor necrosis factor (TNF), IL-1, IL-6, IL-8 and MMP-3 and that the enhancement was mediated by interaction with IL-1RI or with TLR4, respectively. Knowing that uncomplexed HMGB1, depending on its redox status may or may not stimulate cytokine production, we initially tested the suitability of various HMGB1 batches for the present studies. We observed that every tested HMGB1 preparation, regardless of its inherent function to stimulate cytokine production, was capable to act in synergy in complexes with either LPS or IL-1α or β. In order to facilitate the read-out of the HMGB1-complex experiments we thus chose to base our studies on HMGB1 batches that did not induce cytokine formation per se. These experiments have enabled us to propose a mechanism by which HMGB1 contributes to both inflammatory and destructive processes activated during arthritis.

Herein we reveal a mechanism by which HMGB1 may contribute to both inflammatory and destructive processes present during arthritis. Synovial fibroblasts stimulated with HMGB1 in complex with IL-1α, IL-1β or LPS increased their cytokine production. Additionally, HMGB1-IL-1β complexes also increased MMP-3 production. Previous studies have demonstrated that HMGB1 is released from activated immune cells and from stressed synoviocytes in arthritic joints and that blockade of extracellular HMGB1 suppresses disease progression in experimental models. Here we demonstrate that HMGB1 potentiates the effects of two endogenous molecules reported to be present in arthritic joints, namely IL-1α/IL-1β and the microbial mediator LPS. The enhancing effects are caused by complex formation between HMGB1 and the partner molecules. Such immunostimulatory features of HMGB1 in complex with IL-1β and LPS have previously been reported by us and others, while the synergistic effects of HMGB1 and IL-1α are described for the first time in this study.

So, in addition to the direct cytokine-inducing effects of HMGB1 previously reported, our results suggest that the arthritogenic features of HMGB1 can also be mediated by the enhanced activity of molecules in complex with HMGB1. HMGB1 that is actively secreted by activated macrophages or passively released from necrotic cells signals via TLR4 since the TLR4-binding epitope of the HMGB1 molecule expresses its cysteine in position 106 (C106) in reduced form, which is a prerequisite for activation of this signal pathway [47, 13]. The C106 may then later be oxidized in the inflammatory milieu and will lose its capacity to signal via the TLR4 complex. However, this pacified version of HMGB1 may still act as a proinflammatory molecule if the environment contains danger molecules like IL-1α, IL-1β or LPS. HMGB1 will then act as an extracellular sensor and form complexes with these molecules that will enhance subsequent cytokine production in fibroblasts and other cells.

The ligands for complex formation with HMGB1 in this study, IL-1α, IL-1β and LPS, were chosen for three reasons; I) we and others have previously demonstrated a cytokine-enhancing effect of such complexes in macrophages [16, 17, 48]; II) HMGB1 [23], LPS [42], IL-1α and IL-1β [38‐40] have all been detected in RA and OA synovial samples; and III) fibroblasts are pivotal cells in arthritic inflammation that express the suggested receptors for HMGB1 [23, 35, 49‐51] in addition to the LPS receptor TLR4 and IL-1RI [34‐37].

Complexes of HMGB1 with IL-1α, IL-1β or LPS each strongly enhanced the production of TNF, IL-6 and IL-8, while the production of both IL-10 and IL-1β was not affected. It is of interest to note that fibroblasts retrieved from both OA and RA patients shared a similar ability to respond to HMGB1-complex stimulation. Previous studies have reported a difference in extracellular HMGB1 levels in RA and OA synovial fluid with HMGB1 levels being significantly higher (54.1 ± SD 73.0 ng/ml) in RA synovial fluid than in OA synovial fluid (12.0 ± SD 17.7 ng/ml [23]. Similarly, the IL-1β levels recorded in synovial fluid levels from RA patients are roughly 10 times higher than those recorded in OA patients [39]. One can thus assume that HMGB1-IL-1β complexes are more likely formed in vivo during RA than during OA. This could affect the activation status of synovial fibroblasts contributing to a more inflammatory and destructive disease course in RA than in OA.

The amounts of HMGB1 and IL-1β used in our study correspond to levels recorded in RA synovial fluid; ranging from 10 to 300 ng/ml and 5 to 193 pg/ml, respectively [23, 40, 52]. A study by Garcia-Arnandis et al. [25] demonstrated that simultaneous addition of HMGB1 and IL-1β induced enhanced IL-6 and IL-8 production in OA synovial fibroblast cultures, which we did not observe in our study. However, the IL-1β concentration used in their experiments was 20-fold higher than in our experimental setup, and also higher than levels recorded in RA synovial fluid. It is plausible that the high IL-1β levels used could lead to HMGB1-IL-1β complex formation during the cell stimulation and thus their findings are in agreement with our results.

LPS, as well as other constituents of various pathogens, have been reported to be present in arthritic joints [41, 42]. This has led to the hypothesis that infections can be both a cause of arthritis onset and also of disease exacerbation. However, no infectious agent in particular has been pinpointed to be associated with chronic arthritis. The data presented in this paper together with earlier studies on the interaction of HMGB1 with different TLR-ligands suggest that HMGB1 might be a unifying factor for the contribution of various infections to arthritis pathogenesis.

Our data clearly demonstrate the striking ability of HMGB1 complexes to enhance both cytokine production and MMP-3 production by SFs when compared to equivalent doses of the ligand molecules alone. We had originally hypothesized that the enhancing effects would be mediated by simultaneous engagement of an HMGB1 receptor (RAGE or TLR4) and the partner ligand receptor. By blocking IL-1RI and TLR4 with the respective receptor antagonists (IL-1 receptor antagonist or detoxified LPS) we could demonstrate that the stimulatory activities of the HMGB1-IL-1α and IL-1β complexes were mediated via the IL-1RI and that the stimulatory activity of the HMGB1-LPS complex was mediated via TLR4. Interestingly, blockade of TLR4 did not suppress the stimulation induced by HMGB1-IL-1β complexes, thus ruling out that the synergistic effects were mediated by a simultaneous interaction of TLR4 and IL-1RI. This conclusion is also supported by the fact that the HMGB1 used in our studies did not alone possess an endogenous cytokine-inducing capacity, this otherwise being mediated through TLR4 interaction [13, 46]. Attempts to block RAGE, the most studied receptor for HMGB1, using a receptor antagonist failed as we could not define a functional antagonist. Results from studies when soluble RAGE (sRAGE) was added to the cell culture (data not included) demonstrated that sRAGE could suppress the activity of the HMGB1 complexes. However, this only confirms that HMGB1 can bind to RAGE; the suppressive effects were most likely caused by steric hindrance rather than by an inactivation of RAGE signalling. Data from our laboratory (H. Hreggvidsdottir et al., submitted manuscript) indicate that RAGE is not involved in HMGB1 complex signaling as macrophages from RAGE-deficient mice respond equally well to HMGB1 complex stimulation as from wild type mice. However, a remaining possibility for the mechanism of HMGB1 complex-induced enhancement could be the involvement of an as yet undefined HMGB1 receptor in a receptor-pair with the partner ligand receptor. A second possibility could be a multiaggregation of ligand receptors caused by the HMGB1-ligand complex leading to enhanced activity. Both scenarios deserve further investigations.

Preformed complexes of HMGB1 with IL-1α, IL-1β or LPS have the ability to strongly enhance production of both proinflammatory mediators and of tissue destructive enzyme by synovial fibroblasts derived from RA and OA patients. HMGB1 thus acts as an endogenous amplifier endowed with a capacity to magnify responses to trace amounts of endogenous and exogenous danger signals. This effect is mediated via the reciprocal ligand receptors, IL-1RI and TLR4, for the ligands complexed to HMGB1 investigated in this study. HMGB1 without direct cytokine-inducing effects on its own might be present in arthritic joints as HMGB1 is released by apoptotic cells. Furthermore, exposure of cytokine-inducing, reduced HMGB1 to an oxidative burst during inflammation may downregulate its direct proinflammatory features by changing its redox status. We demonstrate that non-cytokine-inducing HMGB1 can form strongly inflammation-enhancing complexes with inflammatory mediators present in arthritic joints. These HMGB1 complexes act on both synovial fibroblasts and on monocytes and enhance their activation status. Thus in addition to the direct cytokine-inducing effect of HMGB1 previously described, we, herein, demonstrate a second mechanism by which HMGB1 may contribute to the arthritogenic process.

Through this study we have increased knowledge of the proinflammatory functions of HMGB1 in arthritis in both RA and OA settings. We have demonstrated enhancing effects of HMGB1 on both inflammatory and destructive disease mechanisms and further consolidated HMGB1 as a putative target for successful therapy.