Starter

Raoul Dufy received corticosteroid therapy in 1950 as one of the first patients to be treated with cortisone to relieve symptoms of rheumatoid arthritis (RA).1 The disease strongly affected his artistic expression, and due to its rapid progression, he could no longer paint. Within a few days after starting a novel therapy with cortisone, first synthesised in 1948, he regained full mobility, experienced less pain and had enough energy to squeeze his paint tubes unassisted. Impressed by this miraculous recovery and his ability to paint again, Raoul Dufy created ‘La Cortisone’. For more than 75 years after synthetically produced GCs had been applied for the first time and published in this journal,2 cortisone-like substances have been used in many chronic inflammatory diseases. Despite the severe metabolic side effects, they are still frequently used today for RA treatment and are included in treatment guidelines. Surprisingly, our understanding of the molecular actions of glucocorticoids (GCs), including cortisone and its receptors, is incomplete and continues to grow even after 75 years (figure 1).

Figure 1

Selected key discoveries and technologies of GR function for 75 years. Since the first successful usage of cortisone (during this time known as compound E) in patients with rheumatoid arthritis, the mode of action of GCs was studied intensively. Since then, many scientific discoveries and novel technologies enhanced our understanding of how GCs act via their nuclear receptors. Strikingly key features, such as a quaternary structure of GR and the regulation of immunometabolism of GR were just identified recently. ChIP Seq, chromatin immunoprecipitation DNA sequencing; DBD, DNA binding domain; GC, glucocorticoids; GRE, glucocorticoid response elements; GR, glucocorticoid receptor. Created with BioRender.com.

Introduction

Since RA is an inflammatory disease, it has long been assumed that immune cells and their products are the main target cells for RA therapy. There are anti-inflammatory and immune suppressive therapies targeting tumour necrosis factor (TNF)-α (eg, adalimumab or etanercept), anti-interleukin drugs (eg, anakinra as interleukin (IL)-1 receptor antagonist or tocilizumab targeting the IL-6 receptor), or abatacept that prevents T-cell stimulation.3 However, therapy with these biological disease-modifying anti-rheumatic drugs is not sufficient for 30–40% of patients with RA.3–6 Therefore, to this day, GCs are the mainstay of RA therapy. However, long-term application of GCs causes severe side effects on metabolism and bone, leading to diabetes, osteoporosis, muscle wasting, dyslipidaemia and cardiovascular disease.7–14

The models describing how GCs orchestrate the beneficial anti-inflammatory effects to fight against inflammation have changed dramatically in the last 75 years (figure 1). With this review, we will unravel the evolution of our understanding of GC in inflammatory disease and continue to focus on their actions in inflammatory arthritis.

Insights into the molecular symphony: advancements in glucocorticoid receptor biology

In general, active corticosteroids act via their receptors: the glucocorticoid receptor (GR) and the mineralocorticoid receptor (MR). Previously, steroids were considered until the 1960s to regulate the energy status in cells. Pioneering work by Elwood Jenkins, involving labelling estradiol with tritium, demonstrated the action of steroids in the nucleus.15 During the 1970s, nuclear receptors were postulated for GCs, and eventually, their binding sites on DNA were characterised in the 1980s, after which their receptors were cloned.16 The identification of the GR and the MR, which has an even higher GC affinity, not only enhanced the understanding of the molecular biological action of GCs but also our knowledge of gene transcription control in general. This discovery positioned these members of the nuclear receptor family as hormone-inducible transcription factors and thus excellent tools for inducing transcription in experimental systems in a defined manner.

Whereas MR serves as a receptor for GC only in a few tissues, including the kidney and certain regions of the brain, the GR, encoded by the NR3C1 gene, is expressed in almost all tissues and organs. Inactive cortisone can be activated by the enzyme 11β-HSD1 (11 beta-hydroxysteroid dehydrogenase 1), which transforms it into cortisol, the active form that binds with high affinity to the GR. 11β-HSD1 is upregulated and highly active at inflammatory sites, amplifying the beneficial effects of endogenous and many commonly used therapeutic GCs.17 By the end of the 1980s, the mechanism by which GCs activate the GR had emerged as follows: The lipophilic steroids enter the cell through a still unknown mechanism. The GR resides in the cytoplasm within a multiprotein complex containing heat shock proteins (HSP90, 70, 23) and immunophilins. On ligand binding, a conformational change and remodelling of the complex expose the GR nuclear localisation sequences, allowing rapid import into the nucleus. Inside the nucleus, the GR binds to DNA at palindromic consensus sequences and triggers interactions with co-regulators. Chromatin remodelling and the subsequent recruitment of the basal transcription machinery enable messenger RNA synthesis and, therefore, gene expression. Immediately after cloning in 198516 it became clear that the different functions of GR are represented in distinct receptor domains:18 the N-terminal transactivation domain with a transactivation function domain (AF1), followed by the DNA binding domain (DBD) connected with a hinge region to the C-terminal ligand binding domain (LBD) with an AF2 domain (figure 2A). The AF1 domain interacts in a ligand-independent manner with basal transcription coactivators.19 In contrast, the AF2 domain interacts in a ligand-dependent manner and recruits members of the p160 family that remodel and open the chromatin with their histone acetyltransferase action and thus recruit further remodelling complexes.20 These complexes consist of coactivators such as steroid receptor coactivator 1 also named nuclear receptor coactivator 1 (NCOA1/SRC-1) or GR-interacting protein (NCOA2/SRC-2/Glucocorticoid receptor-interacting protein 1 (GRIP1)) and NOCA3/SRC-3 that bind to GR in the LBD with an LXXLL binding motif (figure 2B, left).21–24 These are the docking platforms to further large chromatin remodelling complexes and proteins, such as switch/sucrose-non-fermented complex (SWI/SNF), CREB binding protein (CBP) and p300 that help remodel the chromatin and are a prerequisite to change the GR transcriptional state (figure 2B, left).25–27 In addition to recruiting coactivators, the GR also recruits nuclear receptor corepressors (NCOR1 and NCOR2/SMRT). These corepressors interact with their consensus sequence LXX I/H I XXX I/L (L: leucine, X: any amino acid, I: isoleucine and H: histidine) with the AF-2 domain of the nuclear receptors (figure 2B, right).28–31 NCOR1 and NCOR2 repress genes by chromatin condensation mediated by recruited histone deacetylase 3.32

Figure 2

Primary structure and model of the glucocorticoid receptor (GR) and its interaction with coregulators. (A) The human GR alpha protein consists of 777 amino acids (numbers 1–777) and is organised in the N-terminal transactivation domain (NTD) that includes the activation domain 1 (AF1), the DNA-binding domain (DBD) and the ligand binding domain (LBD) that includes the activation domain 2 (AF2). (B) After binding to the DNA (open chromatin) to so-called GR binding sites (GBS), the GR recruits coregulators and interacts with other transcription factors. The coactivator interaction NCOA1 (or SRC-1), NCOA2 (or SRC-2/GRIP1), or NCOA3 (or SRC-3) takes place within the LXXLL motif of the LBD. Hence, coactivators like p300, CBP, SWI/SNF complexes and others are recruited. The nuclear receptor corepressor family (NCOR1 and 2) interact with the GR through the consensus sequence LXX I/H I XXX I / L and have histone deacetylase (HDAC) activity. (C) Classical concept of GR dimer/monomer action. GR monomer interacting with pro-inflammatory transcription factors (AP-1, NF-κB) resulting in repression of respective target genes. The GR dimer is able to bind to either the GREs and induce gene expression or to negative GREs (nGREs) resulting in gene repression. (D) Further binding motifs revealed by studies with chromatin immunoprecipitation sequencing including GR monomer binding GRE half sites, GR binding next to NF-κB or AP-1 sites and lineage-induced transcription factors. (E) Concept of potential tetrameric complexes binding of GR to DNA thus bringing two GBS into close proximity. AP-1, activator protein 1; BS, binding site; CBP, CREB binding protein; GBS, glucocorticoid receptor binding site; GC, glucocorticoids; GRE, glucocorticoid response element; GRIP1, glucocorticoid receptor interacting protein 1; NCOR1, nuclear receptor coactivator 1; NCOR2, nuclear receptor coactivator 2; NCOR3, nuclear receptor coactivator 3; NF-κB, nuclear factor kappa-Β; SWI/SNF, switch/sucrose-non-fermented; TF, transcription factor. Created with BioRender.com.

Until the end of the 1990s, many prominent coregulators of the GR had been identified. Although the structure for individual domains of the GR, such as the DBD in the year 199133 and the LBD in the year 2002,34 had been solved decades ago, the quaternary structure was just recently resolved.35

A new twist in the models of gene regulation by the GR emerged at the beginning of the 1990s with insights into negative gene regulation.36–38 The GR was thought to act in two different modes:

1. As a monomer, the GR tethers with pro-inflammatory transcription factors like activator protein 1 (AP-1) or nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) at their respective DNA binding sites and thus repress target genes (transrepression) encoding matrix metalloproteinases and cytokines and this eventually would explain the repression of inflammation and maybe even processes involved in carcinogenesis (figure 2C, left).39

2. As a dimer, the GR directly binds to DNA at palindromic sequences (glucocorticoid response element (GRE)40) and was proposed to induce transcription of genes (transactivation), often involved in glucose metabolism and thus likely being responsible for some side effects as diabetes (figure 2C, middle).41 In addition, transrepression of gene transcription was observed for GR dimers when binding to so-called negative GREs (figure 2C, right). This is discussed in more detail elsewhere.7 42

This simplified view of GR action—transrepression for anti-inflammatory effects and transactivation of target genes for inducing side effects—prompted academic and industrial research to develop selective GR agonists that are agonistic for transrepression but antagonistic for transactivation.41 Non-steroidal ligands (such as compound A, LDG552, ZK216348) and many selective GR agonists (SEGRAs, eg, RU24858) were developed and tested for their desired anti-inflammatory effects and reduced side effects.41 43 44 Unfortunately, only a few selective ligands made it into clinical trials.41 However, it became evident that transrepression is not sufficient for reducing most inflammatory responses in preclinical animal models.44 This was demonstrated with engineered mice carrying a point mutation on the GR dimerisation interface of the DNA ligand domain45 46 that impaired dimerisation, although not completely abrogating it.46–48 These GR mutant mice, known as GRdim mice, were still able to reduce the activity of AP-149 and NF-κB.50 Indeed, phorbol ester-induced ear inflammation was efficiently reduced by GC exposure50 in these GRdim mice. However, GRdim mice failed in most inflammatory preclinical models to reduce inflammation. GRdim mice were completely refractory to GC-mediated suppression of inflammation in models of RA, allergic airway inflammation, contact allergy, during lipopolysaccharide (LPS) or cecal ligation and puncture induced sepsis, TNF-α induced shock and acute lung inflammation.51–61 Furthermore, evidence emerged that GC-induced genes with anti-inflammatory effects are essentially contributing to anti-inflammatory actions by GCs, such as Dusp1 (dual specificity phosphatase 1),62 Anxa1,63 Gilz (glucocorticoid-induced leucine zipper)64 and A20.65 This challenged the simplified dichotomous model regarding the GR monomer mediating anti-inflammatory effects and GR dimer function mediating side effects.

In the mid-2000s, advanced techniques like chromatin immunoprecipitation (ChIP) followed by array hybridisation and sequencing technologies (such as DNase I hypersensitive site sequencing (DNase seq), Formaldeyde-Assisted Isolation of Regulatory Elements sequencing (FAIRE seq) and Assay for Transposase-Accessible Chromatin using sequencing (ATAC seq)) began to reveal how the GR interacts with the chromatin landscape during exposure to GCs. Intriguingly, the classical tethering mechanism towards NF-κB, AP-1 and other transcription factors was not so common, as anticipated in the 1990s but rather a direct binding of GR next to AP-1 and NF-κB sites66 67 (figure 2D, middle) and direct binding of even monomeric GR on GRE half sites were reported48 68 (figure 2D, left). In these assays, the GR was found often in close proximity with pioneering, lineage-specific transcription factors, which are necessary to first open the chromatin (figure 2D, right). Furthermore, GR facilitates the binding and interaction of further transcription factors.48 69 70 This might explain the cell type specificity of GR binding to certain DNA sequences.69 These DNA binding sites are known as hot spots of enhancer.69 Interestingly, after cessation of GC exposure, these open chromatin sites tend to persist for some time.71 These ‘memory effects’ are generally reversible but can vary in a gene-specific manner.72 The potential application of this knowledge for RA therapy has not yet been explored and remains an area of uncharted research. Finally, next to dimeric sites, also tetrameric formation of GR binding to two GBS in a DNA looping association had been proposed, but their biological relevance remains to be proven (figure 2E).73 74

The era of single cell (sc)75 and single molecule analysis76 77 adds a completely novel dimension to our understanding of the complexity of GR target gene regulation. Previously most knowledge was derived from bulk analysis with a mixture of different cell types or cell activities which obscured differences in type, status and polarisation. With the scRNA sequencing and single cell Assay for Transposase-Accessible Chromatin using sequencing (scATACseq), it was revealed that the response to the synthetic GC dexamethasone (Dex) and hence chromatin states vary already in a clonal cell line.78 Sc approaches revealed novel cell types in the course of arthritis.79 How GCs affect gene regulation in this distinct cell population has not yet been solved and will be discussed in the following.

Understanding of cell type-specific actions of GCs

GCs have diverse effects on cells, acting in a cell-specific manner. For example, GCs can trigger apoptosis, autophagy or senescence in certain cell types, such as in lymphocytes80 and osteocytes.81 82 Conversely, GCs even can accelerate proliferation, for example, in the process of stress erythropoiesis.83 The effect of GCs on the transcription of nine different primary human cell types (including innate and adaptive immune cells as well as stromal cells) from healthy donors was recently investigated in a comparative manner at the level of signalling pathways via RNA sequencing.84 This research, conducted using RNA sequencing, revealed that the impact of GCs varies significantly depending on the cell type. Methylprednisolone, a type of GC, influenced the expression of about 9000 distinct genes, or approximately 17% of the human transcriptome, depending on the cell type. Over the last 25 years, the functional relevance of target cells was elegantly demonstrated with conditional GR mutant mice using the Cre/LoxP (flanking the Nr3c1 at exon 385 or exon 286) system to delete the GR in various cells and tissues at different time points in life. This approach has provided deep insights into the cell type-specific effects of GCs, particularly in conditions like rheumatic and inflammatory arthritis. We will now explore some of the well-studied effects of GCs on major cell types and the insights gained from Cre/LoxP mutant mice (figure 3). We focus on the most well-studied cell types in RA towards their response to GCs and are aware that we disregard other important cell types that shall be further exploited in the future.

Figure 3

The synovial joint in RA and effects of GCs in different cell types. (A) In RA, the synovial environment forms a hyperplastic pannus tissue composed of resident and infiltrating cells. FLS and macrophages are central players and communicate with other immune cells, such as B cells, T cells, dendritic cells via cytokines and chemokines. Also, cells from bone and cartilage are affected and activated via soluble factors, resulting in increased bone erosion by enhanced activity of osteoclasts and chondrocytes expressing, for example, IL-6 and MMPs. Details are explained in the text. (B) GCs like cortisol rewire these cellular interactions. GC treatment leads to less activity and invasiveness of FLS and reduced release of cytokines and chemokines and therefore, cause less pro-inflammatory activation of other cell types. Furthermore, GCs induce anti-inflammatory phenotypes, for example, ‘M2’-like in macrophages and less invasiveness of T cells. How GCs act on B cells is poorly understood, but reduced IgG production of B cells and enhanced apoptosis are reported. Nonetheless, the reduced activity and invasiveness of FLS and macrophage may reduce the hyperplasticity of the synovial lining layer. The effects of GCs on bone and cartilage in RA are complex. There are GC-induced changes in bone formation by less differentiation and induced apoptosis of osteoblasts and osteocytes. GCs as well as the induced RANKL expression from osteoblasts and osteocytes leads to a high activity of bone degrading osteoclasts. Details are explained in the text. AnxA1, annexin A1; Axl, AXL receptor tyrosine kinase; FLS, fibroblast-like synoviocytes; GCs, glucocorticoids; GM-CSF, granulocyte-macrophage colony-stimulating factor; IFN-γ, interferon γ; IL-1, interleukin-1; IL-6, interleukin-6; IL-17, interleukin-17; MerTK, tyrosin-protein kinase MER; MMP, matrix metalloproteinases; OPG, osteoprotegerin; RA, rheumatoid arthritis; RANKL, receptor activator of nuclear factor kappa-Β ligand; TNF, tumour necrosis factor; Treg, regulatory T cells. Created with BioRender.com.

GCs affect crosstalk of immune and stromal cells

RA is orchestrated by numerous cells of haematopoietic and stromal origin that communicate with each other and contribute all to the disease. GCs affect this crosstalk by its cell type specific actions (figure 3). In particular, this affects the polarisation states of immune cells and their respective release of inflammatory mediators as well as their cellular functions such as inducing their repair phenotypes.

In T cells GC-induced target genes, such as annexin A1 (Anxa1)87 88 and GC-induced TNFR-related protein (Gitr)89, were suggested as important modulators of T cell function and polarisation. It was shown that GCs inhibit the T helper cell response and their cytokine production.90 Notably, GCs showed an effect on murine cytotoxic T cells via reduced cytokine levels91 and elevated apoptosis.92 Regulatory T cells (Tregs) whose loss is associated with auto-immune diseases were described to be amplified by GC in concert with IL-293 and the key regulator FoxP3 was supposed to be enhanced. In contrast, Tregs were also observed to be decreased by GCs in lung inflammation94 and in an antigen-induced arthritis (AIA) model.52 Specific ablation of the GR in Tregs in mice leads to a failure of GC-mediated resolution of inflammation in models of multiple sclerosis and airway inflammation.95 Complete deletion of GR in T cells abrogated the GC response in a model of arthritis, and here the decrease of the high pro-inflammatory IL-17 cells rather than the regulation of Tregs appeared to be crucial, since IL-17 knockout (KO) mice were refractory to GC exposure.52 However, to our knowledge, studies of mice with specific ablation of GR in Tregs only in arthritis models are lacking to finally clarify this aspect.

The impact of corticosteroids on B cells and antibody production in inflammatory diseases are poorly understood. Short-term treatment with GCs or methotrexate does not influence B cell subpopulations, particularly pre-switch memory B cells, in patients in the initial stages of RA.96 Furthermore, prednisolone medication does not affect IgG levels in human individuals with RA.97 In contrast, Dex reduced IgG synthesis in the collagen-induced arthritis (CIA) model98 but not in mice with AIA or glucose 6-isomerase-induced arthritis.52 Studies involving a mouse model with a specific deletion of GR in B cells revealed no significant differences in clinical outcomes compared with controls, suggesting that B cells are not the main target for GC-induced immune suppression in the AIA model.52 This supports the notion that the effects of GCs in these models are sufficient to reduce inflammation without directly targeting B cells.

Dendritic cells (DCs) play as professional antigen-presenting cells a central role in autoimmune diseases.99 Several types exist, including conventional DCs, plasmacytoid DCs or monocyte-derived DCs. DCs activate T cells to drive the adaptive immune response. GCs can lead—in conjunction with IL-4 signalling—to the development of so-called tolerogenic DCs, DCs that dampen T cell expansion and express an anti-inflammatory gene signature like anti-inflammatory macrophages.100 To our knowledge, however, a functional impact of DCs as GC target to dampen the disease course in murine models with the help of conditional knockout mice could not be demonstrated yet. This may be due to limitations of precise targeting of the distinct DC populations in the current models used so far.

Macrophages are innate mediators of the immune system, are highly active during RA, and contribute to inflammation of the synovium, as well as cartilage and bone degradation.101 Tissue-resident macrophages form an internal immunological barrier in the synovial lining, whereas infiltrating mononuclear cells eventually differentiate into macrophages seem to drive the disease.102 Cytokines like TNF-α and IL-6 are produced by macrophages, and their elevated levels contribute to the continued inflammation and communication with other cell types.103 Endogenous GC production in macrophages itself might affect RA disease. A mouse model with a specific 11β-HSD1 knockout in macrophages and no local GC activations exhibited high levels of bone erosion and more synovial hyperplasia.104 Intriguingly 11β-HSD1 deletion affected systemic activation of therapeutic GCs, partially via its action in myeloid cells.17 Exogenous GC treatment increases non-classical, non-activated anti-inflammatory macrophages in terms of the expression of anti-inflammatory marker gene expression such as Cd163, Cd36, Anxa1, Axl receptor tyrosine kinase (Axl) and MER proto-oncogene, tyrosine kinase (Mertk) in a mouse model of serum transfer-induced arthrits (STIA).105 However, this effect was rather dependent on the GR function in stromal cells in this model than on the GR function in macrophages itself. GC treatment leads to decreased macrophage cytokine production, which may result in less activation of fibroblast-like synoviocytes (FLS), preventing the construction of a hyperplastic pannus tissue and as well as inhibition of osteoclastogenesis, reducing the damage to bone and cartilage. GC-stimulated macrophages exhibit an increased efferocytosis and phagocytosis capacity.106 107 Thus, GC effects on macrophages may be direct or depend on crosstalk with stromal cells, particularly FLS during RA.

FLS actively participate in the beginning of the development of RA. They contribute to joint inflammation by producing pro-inflammatory cytokines, chemokines and matrix-degrading enzymes, such as matrix metalloproteinases (MMPs) 108 and hence promote inflammation and bone degradation.109 Because of their similar activities to human FLS in culture, it has been demonstrated that murine FLS isolated from arthritic mice are convenient candidates for in vitro research.110 In RA, FLS adopt an invasive hyperplastic phenotype accompanied by the high release of pro-inflammatory cytokines and matrix metalloproteases and promote further infiltration by immune cells.111 The crucial role of FLS as direct responders to GC therapy in vivo was demonstrated by bone marrow chimeric GR mutant mice, with impaired GR function in either irradiation-sensitive (immune cells) or irradiation-resistant compartment (stromal fraction).105 In addition, the pivotal role of GR in FLS for the general anti-inflammatory effect of GCs in an arthritis model (STIA) was shown by conditional mice with GR deletion in stromal cells.105 Mice without GR function in stromal cells were completely refractory to GC treatment and the induction of anti-inflammatory macrophages was defined by their marker gene expression (Cd163, Cd36, Anxa1, Axl and Mertk).105 Interestingly, Croft et al described two different FLS subsets in the STIA model:112 the inflammatory FAPα+THY1+ fibroblasts and the tissue-destructive FAPα+THY1− fibroblasts. These subsets are spatially separated in the synovial lining (FAPα+THY1−) and sublining (FAPα+THY1+) layer, and the subsets display a distinct gene expression signature, while both are harmful to their environment. The inflammatory FAPα+THY1+ fibroblasts show a high expression of chemokines and cytokines and a reduced effect on bone and cartilage damage. The tissue-destructive FAPα+THY1− fibroblasts are characterised by a higher expression of RANKL (receptor activator of nuclear factor kappa-Β ligand) are hardly involved in inflammatory processes. How GC influence the distribution of these subsets is not yet known. Recently, during the resolution of inflammation in humans, a novel resolving inflammatory subtype was discovered that occurs after anti-cytokine therapy.79 It is tempting to speculate that such anti-inflammatory subtypes also emerge during steroid therapy. The identification of GC-sensitive and critical FLS subsets in the future might be a key feature for developing more targeted and effective treatments in the future.

GCs affect cartilage and bone homeostasis and the cellular interactions

The relationship between GC, bone biology and RA is particularly complex to investigate due to the detrimental effects both GC and inflammation have on bone health. Direct effects of GCs on bone depend on an initial enhanced resorption, in part, due to GC-induced changes in osteoclast-inducing factors and their life span, as well as inhibition of bone formation.113 In addition, chondrocytes, osteoblasts/osteocytes and osteoclasts participate in joint inflammation and subsequent erosion of the joints (figure 3). In AIA as well as STIA chondrocyte-specific GR knockout mice showed an elevated joint inflammation concerning joint swelling and histological scores.114 However, cartilage degradation was attenuated in GR-specific chondrocyte-knockout mice, accompanied by decreased hypoxia factor 2 alpha expression and associated senescence and catabolic signalling.115 Hence, endogenous GR signalling decreases the pro-inflammatory effector function of chondrocytes, such as reducing cytokine and inflammatory mediator production, but unfortunately trigger cartilage degradation. Exogenous GCs can have beneficial and detrimental effects on chondrocytes by limiting cytokine and MMP expression in chondrocytes,116 while impairing proliferation and differentiation, as well as matrix generation, inducing autophagy and apoptosis.117 118 This highlights the complexity of GC function in chondrocytes and that timing in their administration matters. Local inflammation results in high numbers of osteoclasts at the cartilage-pannus side. These processes are caused by the secretion of various cytokines (eg, TNF-α, IL-1β and IL-17) promoting monocyte migration and differentiation into the joint.119 GCs while reducing the pro-inflammatory milieu might therefore also interfere with excessive osteoclast generation, as well they are reducing osteoclast progenitors.120 But GCs are acting themselves on prolongation of osteoclast lifetime120–122 and might indirectly enhance osteoclastogenesis by inducing osteoclast stimulating RANKL and CSF-1 (colony-stimulating factor 1) and decreasing the expression of the osteoclast inhibitor osteoprotegerin (OPG)123 in osteoblasts and osteocytes. In postmenopausal women, GC exposure is associated with osteoporosis.124 Here GCs play a dual role that to our knowledge has not been addressed sufficiently in arthritis mechanistically so far and needs further investigation.

The accumulation of immune cells in inflamed joints leads to a high level of IL-1β and TNF-α, which results inhibition of osteoblast bone formation leading to arthritic bone loss.125 While acute anti-inflammatory GC effects might be beneficial for bone integrity, long-term therapy shows harmful effects on these cell types: GCs lead to a decrease in the osteogenic cell fate of stromal progenitor cells, suppressed proliferation of osteoprogenitors, inhibited differentiation of osteoblast precursors into mature osteoblasts126 and inhibition of bone formation, which is sufficient for bone loss. Osteocytes undergo autophagy and apoptosis in response to GC exposure.82 127 Interestingly, harmful effects of GC on bone are mediated independently of a functional GR dimerisation.126 The role of endogenous GC signalling in osteoblasts towards the inflammatory response of arthritis remains controversial. In the Col 2.3-11β-HSD2-tg transgenic mouse line, 11β-HSD2 activity is exclusively expressed in osteoblasts and osteocytes, leading to inactivation of GCs in these cells. These mice are protected from GC-induced bone loss,128 but strikingly showed a slightly decreased inflammatory score in STIA,129 in contrast, this could not be reproduced in an AIA model. Col2.3-11β-HSD2-tg mice did not show differences in clinical parameters compared with the control group (arthritic wildtype (WT)).130 This demonstrates the complexity of GC signalling in osteoblasts and their contribution to attenuating the inflammatory response in arthritis is still completely obscure.

Taken together, how GCs affect the cellular crosstalk during arthritis despite some seminal studies with cell type-specific manipulation of GC signalling requires further investigations. The advent of scRNA sequencing, proteomics and metabolomics will help to clarify how these anti-inflammatory hormones affect the interaction of different cells in arthritis leading to the resolution of inflammation and GC impact on musculoskeletal tissues.

The impact of the GR on metabolism

Recent progress in inflammatory research placed the regulation of cellular energy metabolism of immune cells (immunometabolism) at the centre stage from the transition of pro-inflammatory states to the inflammation-resolving and repair-inducing anti-inflammatory stage. In general, pro-inflammatory immune cells undergo aerobic glycolysis as a catabolic metabolic step that leads to stalling of the TCA (tricarboxylic acid) cycle and a decrease in oxidative phosphorylation. Certain metabolites in the TCA cycle accumulate due to stalling of the TCA cycle, such as succinate.131 This supports the stabilisation of hypoxia-induced factor 1 alpha (Hif1α) that is also accelerated in response to pro-inflammatory signals to mediate a glycolytic response.132 Itaconate, another increased metabolite, counteracts succinate function and is therefore one of the entry points for the anti-inflammatory action (figure 4).133 Under pro-inflammatory conditions, mitochondria also change their fusion and fission behaviour. Anti-inflammatory stimuli (like GCs) intriguingly restore mitochondrial function and are important in triggering macrophage anti-inflammatory repair phenotypes,134 which occurs in an AMP-activated protein kinase (AMPK) dependent manner.135 GCs, despite being known to regulate energy metabolism for decades, have just emerged as a pivotal player in cellular energy metabolism in immune cells.136 137 The anti-inflammatory action of GC depends on the activity of AMPK in macrophages to resolve models of lung inflammation and to induce muscle repair.138 Indeed, LPS-induced disruption of mitochondrial networks is restored by GCs in macrophages in a GR-dependent manner.139 More importantly, GCs reduce lactate production induced by inflammatory stimuli, first observed 1992140 but then systemically investigated in 2022 and the following years.139 141 142 Metabolic flux analysis revealed that LPS-induced stalling of the TCA cycle is relieved by GC treatment in part by antagonising Hif1α.139 141 142 This occurs in part due to the non-genomic actions of GR, which interacts directly with the components of pyruvate dehydrogenase at low abundance of steroids.142 143 On exposure to GCs, this complex disrupts and leads to an acceleration of metabolite flux in TCA including enhanced itaconate production (figure 4).142 Most importantly, manipulating itaconate production occurs in patients with steroid-treated human RA, and inhibition of itaconate production in aconitate decarboxylase 1 knockout mice renders mice completely resistant to GC effects in an STIA model.142 Thus, reprogramming of catabolic metabolism by GCs is an essential feature in resolving inflammation and is of great pharmacological interest in the treatment of inflammatory diseases.

Figure 4

GCs regulate the immunometabolism. GCs reduce inflammation by altering macrophage mitochondrial metabolism in particular by inhibiting glycolysis and Hif1, boosting itaconate production while dampening the inflammatory response. This is mediated via partially non-genomic effects of the GR interacting with PDH. GC, glucocorticoids; GR, glucocorticoid receptor; Hif1, hypoxia-induced factor 1 alpha; PDH, pyruvate dehydrogenase; TCA, tricarboxylic acid. Created with BioRender.com.

Translational aspects and outlook

Corticosteroids are despite numerous side effects in active use to treat autoimmune diseases for 75 years. During this period, our understanding of corticosteroid action changed dramatically and was a surrogate of changing paradigms in biomedical research in general. GCs were first reported to be an important regulator of energy metabolism in the late 1950s to 1960s,144 145 when the discoveries of metabolite to generate ATP in cells were established and this was considered to be the major function. With the advent of principles of gene regulation at the end of the 1960s, followed by the concepts of Jacob Monod in bacteria146 and the transfer to mammalian systems by the example of cortisone action on liver enzymes147 the field shifted to transcriptional control by GCs until today. During this time, the discovery of DNA binding sites for GR, cloning of GR, interacting proteins of chromatin and epigenetic regulators was discovered as described above. Just lately, genetic tools mainly in mice allowed dissecting cell type-specific responses towards GCs in various disease models including those of arthritis. It was recognised that some cell types may be more relevant for the resolution of inflammation than others, also that crosstalk between cells and organs is decisive in the resolution of inflammation by GCs, for example, in arthritis the murine FLS-macrophage crosstalk.105 Sc genomics has just begun over the past years and has revealed a plethora of cell types and states, where the respective GC response in different cellular populations is being elucidated and will provide even more insight into the cell type-specific complex interplay. Finally, just in the last 2 years, it was (re)discovered that GCs regulate energy metabolisms of immune cells for therapeutic activities, so that the field might move again to concepts with improved methods that were vaguely phrased already in the 1960s of the last century.

All these exciting discoveries, which are in part still ongoing, provide intriguing concepts for optimised steroid therapy (figure 5). Whereas in the past selective GR agonists were only a little successful, partial agonists that allow tissue-specific activity recently came into focus. Some compounds seem to act antagonistically in the liver but agonistically in immune cells, and therefore are promising agents (figure 5).148 Cell type specificity can be further addressed by linking steroids to immune cell-specific surface molecules149 or encapsulating them in nanoparticles (figure 5).150 Furthermore, anti-inflammatory metabolites and other mediators that are modulated by GCs could be exploited to dampen inflammation, where the nuclear function of GR might be spared and rather cytoplasmic/mitochondrial functions shall be addressed (figure 5). There will be no simple solution, but rather the correct mix of advanced steroids together with other compounds in a personalised manner that will help to improve health in autoimmune disorders with lesser side effects, until real cures for these diseases will be found. This would allow drawing a novel even more promising flowering of ‘La Cortisone’ if today Dufy would live 75 years after the discovery of the therapeutic effects of GCs.

Figure 5

Novel ste