Glucocorticoids (GCs) are steroid hormones produced by the adrenal cortex ^[1][2][1,2]. They have a broad range of effects including developmental, anti-inflammatory [3], metabolic ^[4][5][4,5] and many other functions [2]. GCs are hydrophobic molecules derived from cholesterol and hence they can easily pass cell membranes to enter in cells and exert their effects. GCs were first discovered in 1946 based on research on Addison’s disease, a rare disease, also known as primary adrenal insufficiency or hypoadrenalism [6]. The active form of GCs is known as cortisol in humans and corticosterone in rodents. Owing to their anti-inflammatory and immune-suppressive actions, and small molecular weight, GCs are among the most widely prescribed drugs worldwide and are used both in short-term and chronic settings, despite that their chronic use poses risks for side effects such as osteoporosis [7], or opportunistic infections due to the immune suppressive effects [8]. Furthermore, some patients do not respond to GC treatment and display so-called GC resistance [9], the occurrence of which varies between diseases and can also develop over the course of the treatment [9]. Over the years, much effort has been directed towards developing selective synthetic GCs with fewer side effects while keeping the therapeutic effects intact; however, success has been limited ^[10][11][10,11].

GCs bind the GC receptor (GR), also known as Nuclear Receptor 3 C1 (NR3C1). This is a soluble receptor belonging to the superfamily of nuclear receptors (NRs) ^[2][12][2,12]. It is estimated that 1000 to 2000 genes are subject to GR-mediated regulation, and some studies suggest that up to 20% of all genes are responsive to the GR. The GR regulates many pathways (e.g., gluconeogenesis ^[13][14][13,14], inflammatory response ^[3][15][3,15], fatty acid metabolism ^[16][17][16,17]) by regulating gene expression in various organs and tissues (e.g., liver ^[18][19][18,19], nervous system ^[20][21][20,21], adipocytes ^[17][22][17,22]). Over- or under stimulation of the GR results in severe phenotypes such as Cushing syndrome and Addison’s disease ^[23][24][23,24]. The NR family is composed of 49 members and includes other well-known receptors, such as the estrogen receptor (ER) and mineralocorticoid receptor (MR) as well as a number of orphan receptors (the ligands of which are unknown) [25]. Similar to other nuclear receptors, the protein domain configuration of the GR can be subdivided into the following four regions (Figure 1A), from N-terminus to C-terminus [25]: N-terminal domain (NTD), a DNA-binding domain (DBD), a short flexible hinge region and the ligand binding domain (LBD). The N-terminal domain is the least conserved of the four domains. It shows large differences between different nuclear receptors [25] and contains most polymorphisms in humans (mostly without deleterious effect) ^[26][27][26,27]. The NTD is also intrinsically disordered, making it difficult to study its structure, and no crystal structure of this region is available. However, there is some structural information known, mainly secondary structure information. The activator function domain within the NTD is an organized domain in vivo which may adopt variable conformation to induce specific responses by recruiting different cofactors based on its configuration [28]. Expression of the GR-coding gene (NR3C1) can start from several alternative transcription initiation sites, which give rise to receptor isoforms differing in the size of the NTD, each with a different transcriptional capability and tissue-specific expression ^[29][30][29,30]. Next to the multiple possible transcription initiation sites, the GR also has multiple possible alternative translation start sites in exon 2, leading to different translational isoforms. Finally, the GR also undergoes alternative splicing, leading to several alternative splice isoforms [31]. These isoforms are formed by splicing variations in the final introns/exons, involving the sequence coding for the LBD and sometimes the hinge region [31]. The main isoforms involve an alternative terminal splice acceptor site, resulting a different final exon used giving the GRα and GRβ isoforms ^[32][33][32,33]. GRα is the canonical active GR while GRβ lacks ligand binding and transactivation activities so that GRα/GRβ and GRβ/GRβ dimers cannot activate gene expression, but can bind DNA. As GRβ is always located in the nucleus, this results in an inhibition of GR activity ^[33][34][33,34]. Without ligand, the GR is kept in the cytoplasm in a multiprotein complex consisting of immunophilins and various chaperones in a configuration optimal for highly sensitive ligand binding [35]. When a ligand binds, the complex changes and recruits other factors ^[36][37][36,37], ultimately resulting in the import of GR into the nucleus through the nuclear pores, where it will exert its regulatory functions [38]. While there is one study that shows GR dimers in cytoplasm, this study was based on a strong overexpression system used in vitro [39]. Furthermore, David Bain’s group showed that GR does not form dimers spontaneously and that an additional factor, such as DNA, is needed [40].

It is generally accepted that GR exerts its function in the nucleus by forming a receptor homodimer and binding DNA at a specific sequence element. Initially it was thought that GR would form tetramers [41], but this was abandoned due to lack of direct evidence and the fact that support was found for the existence of dimers [42]. However, several recent studies, mainly by the group of Diego M. Presman, have once again brought the notion that GR might also exist as a tetramer, or as a dimer of dimers ^[43][44][43,44], to the forefront. The GR dimer model remains the most generally accepted model for GR function and GR tetramerization is an active avenue of research. Next to the dimeric and, and possible tetrameric functions of GR, the GR monomer is also thought to perform several functions. The exact functioning of GR monomers is not yet fully understood, but dimer-deficient mutants have been generated and used to infer monomeric activity. Likewise, some GR ligands that prevent dimer formation, such as compound A, can also be used. Decades ago, it was hypothesized that GR monomers would rather be transcription blockers, either by DNA binding or by inhibitory protein–protein interactions with other transcription factors, such as sequestration and tethering, while GR homodimers would be more responsible for gene upregulation via direct DNA binding. While direct monomeric DNA-binding activities have not been directly observed in a quantitative assay, there is other evidence in support (see further). Evidence for tethering and sequestration of other TFs by GR, which is believed to be a monomeric function, has been found. Since GR is supposed to bind pro-inflammatory transcription factors such as NF-kappaB as a monomer [45], and since GR likely induces expression of some genes involved in gluconeogenesis (Phosphoenolpyruvate carboxykinase 1 (PCK1) and Glucose 6-phosphate (G6PC), potentially leading to type 2 diabetes mellitus, an important side effect of GCs), as a dimer [14], this led to the paradigm that beneficial, anti-inflammatory, effects were monomer mediated and that undesirable side effects were dimer mediated. It was generally believed that skewing GR into a monomeric form or into a dimeric form could determine therapeutic outcome versus side effects profiles [46]. A balance in favor of monomers would be ideal for chronical treatment, with a minimum of side effects (from GR dimers) of long-term GR activation [47], and stronger dimerization would be better in acute settings (Figure 2) [48]. Indeed, several GR dependent gene products–thought to be dimer dependent–have been identified to have anti-inflammatory functions that can be responsible for the acute anti-inflammatory effects of GCs, as will be illustrated below.

Work has been carried out to search for ligands that push the GR towards monomer or dimer action as respectively Selective GR Activators and Modulators (SEGRAM), such as compound A [49], and Selective Dimerizing Glucocorticoid Receptor Agonists and Modulators (SEDIGRAM) [10]. While several effective synthetic GCs have been developed, so far this approach has not resulted in such “skewing” ligands at the bedside. Dissociating the side effects and anti-inflammatory effects of GCs solely on the basis of their monomeric or dimeric structure turned out to be unrealistic. Indeed, it has been shown that in an acute inflammatory setting, consensus GRE (dimer or higher order) gene expression is required for GC/GR effectiveness. Moreover, some side effects are not mediated via dimeric GR (see Table 1). A good understanding of the mechanisms and functions of receptor dimer formation is needed to help developing such ligands in a more educated way. TIn the resesearchers is paper, we will provide an overview of the current understanding of GR dimerization, with a focus on insights into dimerization mechanisms and the importance of intact GR dimers in pathophysiological conditions inferred from studies with mutant mice.