In this study we performed liver gene expression profiling of WT and GRdim mice after prednisolone administration. Our aim was to chart the biological processes in the liver that are affected by GCs and to study their dependence on DNA-binding and dimerization of the GR. CoPub keyword enrichment analysis with prednisolone-regulated genes in WT mice showed enrichment of keywords associated with glucose, lipid and amino acid metabolism, the cell cycle and apoptosis (Table 1 and Figure 3). This is in agreement with cellular processes known to be affected by GCs [73, 74]. Interestingly, the forkhead transcription factors Foxo1 and Foxo3a are regulators of these cellular processes and induced by prednisolone in WT mice. Together with the observation that a subset of the prednisolone-induced genes are Foxo1 and Foxo3a target genes and are similarly regulated in Foxo1 and Foxo3a expression profiling experiments [54, 59, 60], this suggests that the GR synergizes with these transcription factors in mouse liver to control lipid and glucose metabolism and the cell cycle. This concept is further supported by the recent finding that the MurF1 (Muscle RING finger 1) promoter contains adjacent binding sites for the GR and Foxo transcription factors and is synergistically activated when both are co-expressed . However, it is likely that a subset of the prednisolone-induced genes is under direct control of Foxo1 and Foxo3a without the need for synergy with the GR. Other studies, addressing small gene sets or individual genes, had also identified a role for Foxo1 and Foxo3a in GR-induced gene expression [58, 75–77].
The list of differentially regulated genes by prednisolone in WT mice overlaps with that of an earlier study by Phuc Le et al, which combined chromatin immunoprecipitation (ChIP) data with gene expression data to identify direct GR target genes in mouse liver . Most of the genes found in our study displayed the same up and downregulation of gene expression, such as Tat, Foxo1 and Fkbp5 that are upregulated and Adipor2, CCAAT/enhancer binding protein alpha (C/EBPα) and Tkt that are downregulated. From a candidate gene set of 302 genes, Phuc Le et al identified metabolism, cell proliferation and programmed cell death as important processes in their GO-term analysis, which is in agreement with our findings.
Interestingly, in contrast to our study, Phuc Le et al did not observe upregulation of bona fide GR target genes, such as Pck and Igfpb in fed CD1 mice and explain this by lack of response due to a potential inhibition by insulin signaling in the fed state . The fact that we observed upregulation of these genes indicates that in our experimental setup, which differs in several aspects from the setup used by Phuc Le et al, such as mouse strain used for the study, GC dosage and the fed state of the mice, the inhibitory effects of fed-induced insulin signaling do not play a role.
The overall induction of gene expression was strongly reduced in prednisolone-treated GRdim mice compared to WT mice (Figure 4). Nevertheless, in many cases residual gene induction by prednisolone was still observed. This indicates that GR dimerization is indeed an important mechanism for activation of these genes, although some regulation can take place even in the absence of GR homodimers. In fact, GR monomers are in principle capable of binding to GC response elements (GREs) and evoking a basal induction of gene expression. However, due to the lack of cooperative binding they are less potent than GR homodimers. This is in line with the regulation reported for Pnmt (Phenylethanolamine-N-methyltransferase) and Amy2 in which binding of GR monomers to GREs or a GRE half-site (i.e. only one-half of the classical palindrome) was sufficient to confer induction by GCs [80, 81]. In case of the Pnmt gene, multiple GRE half-site have been identified in the promoter region allowing receptor clustering and thereby stable binding of GR monomers independent of the DNA-binding domain (DBD) interface . This also explains why expression of the Pnmt gene is not compromised in GRdim mice despite the lack of GR dimer formation . Another explanation for the residual gene induction by prednisolone in GRdim mice is that the mutant GR still forms homodimers but that these are far less stable than in WT mice.
A recent study showed that the specific sequence of a GRE differently affects the conformation of the GR and thereby its activity towards specific target genes . Mutation analysis of overexpressed GR domains that are involved in transcriptional activation, namely the dimerization region in the DBD (Dim) as well as AF1 and AF2 (activation function 1 and 2), showed that the dependency on each of them was specific for the sequence of the GR binding site and that genes differed in their dependence on dimerization . The fold induction of Tat and Fkbp5 by dexamethasone in the Dim mutant was around 30% for Tat and 50% for Fkbp5 of that for WT GR, which is in gross agreement with our own observations in vivo (Figure 4).
The effect of the GRdim mutation was found in all major pathways that were induced by prednisolone, i.e. for genes in inflammatory pathways, gluconeogenesis and cell cycle. Moreover, the attenuating effect of the GRdim mutation was found for genes that were upregulated as well as for genes that were downregulated by prednisolone. This suggests that genes, which fail to be repressed by prednisolone in GRdim mice, are either regulated through GR binding to negative GRE elements (nGREs) or indirectly regulated via other GR target genes. Interestingly, analysis of the magnitude of gene regulation by prednisolone in WT mice compared to GRdim mice showed that the cell cycle-related genes are more dependent on the dim interface than genes related to gluconeogenesis (Table 2 and 3). This can also be appreciated in Figure 4; several genes that are on the x-axis (i.e. show no differential gene expression upon prednisolone treatment in GRdim mice), are cell cycle-related.
We identified several genes that showed strong upregulation by prednisolone in GRdim mice compared to WT mice, such as Amy2, Car8 and Zdhhc23, and several genes that showed equal induction of gene expression by prednisolone in WT and GRdim mice, amongst them are Dst and Ppargc1a (Figure 4). These genes are potential candidates for having GRE half-sites in their promotor regions, which could explain why these genes show equal or even higher induction of gene expression by prednisolone compared to WT mice. As mentioned earlier, the presence of GR half-sites in the promotor region for Amy2 was indeed experimentally confirmed . For the other genes however, we did not find literature evidence for the presence of GR half-sites in their promotor regions. Therefore, follow-up studies on these genes to determine the functional GRE sites in their regulatory region would be of interest to study the significance of dimer-interface independent GRE binding. Genes that showed equal or stronger upregulation by prednisolone in GRdim mice compared to WT mice can also be secondary response genes; under transcriptional control of transcription factors other than the GR. Keyword enrichment analysis performed with this set of genes did not identify enrichment for a particular cellular process.
We identified several genes that are absolutely dependent on GR dimerization for the induction of gene expression by prednisolone, such as Cdkn1a, Gadd45b, Dusp1, Plk3 and Foxo3a (Figure 4). Interestingly, all of these have a strong relationship with p53: some are under direct transcriptional control of p53 (Cdkn1a , Gadd45b  and Dusp1 ) or physically interact with p53 (Foxo3a [85, 86] and Plk3 ). The cellular responses mediated by Foxo3a and p53 are highly similar, share some of their target genes (e.g. Cdkn1a and Gadd45b) and use similar mechanisms to regulate post-translational modification [51, 88]. Therefore, Foxo3a and p53 can be regarded as partners that positively as well as negatively regulate each other, depending on the context . This observation might indicate that GC-induced activation of Foxo3a and/or p53 is hampered in GRdim mice. In line with this hypothesis Foxo3a was recently shown to be required for GC-induced apoptosis in lymphocytes . Interestingly, this process is defective in GRdim mice highlighting a possible link between the GR, Foxo3a and induction of lymphocyte apoptosis .
Fam107a showed the largest induction of gene expression by prednisolone in WT mice (Figure 4 and 5). Analysis of protein-protein interactions revealed that Fam107a interacts with Tada2a [90, 91], a protein that together with binding partner Tada3a (Transcriptional adaptor 2 and 3 alpha) are core proteins of the histone acetyltransferase (HAT) complex . HAT complexes are involved in chromatin structure modification for initiation of gene transcription, but can also acetylate non-histone proteins to modify their activity and stability [92, 93]. Interestingly, Fam107b a paralog of Fam107a with 84% protein similarity (not regulated in mouse liver by prednisolone) was shown to interact with Tada3a .
The observation that Fam107a inhibits cell proliferation and induces apoptosis when overexpressed [65, 67], suggests that Fam107a, like Foxo3a, Dusp1, Gadd45b, Cdkn1a and Plk3 play a role in regulating the cell cycle. Furthermore, the association of Fam107a with a core protein of the HAT complex might indicate that Fam107a may serve as a cofactor in the transcription machinery complex.
The activity and function of Foxo3a and p53 are strongly modulated by acetylation [51, 88, 95]. Hence, Fam107a is an interesting candidate gene for follow up experiments to study whether it modulates GR-induced gene expression and/or acetylation of GR-associated transcription factors such as Foxo3a, p53, C/EBPα and C/EBPβ (CCAAT/enhancer binding protein, beta).
For a more comprehensive view on GC regulated process in the liver, experiments with multiple time points, different doses of GCs and using one or more inflammatory stimuli, could be considered.