NeuroD Factors Discriminate Mineralocorticoid From Glucocorticoid Receptor DNA Binding in the Male Rat Brain

In the limbic brain, mineralocorticoid receptors (MRs) and glucocorticoid receptors (GRs) both function as receptors for the naturally occurring glucocorticoids (corticosterone/cortisol) but mediate distinct effects on cellular physiology via transcriptional mechanisms. The transcriptional basis for specificity of these MR- vs GR-mediated effects is unknown. To address this conundrum, we have identified the extent of MR/GR DNA-binding selectivity in the rat hippocampus using chromatin immunoprecipitation followed by sequencing. We found 918 and 1450 nonoverlapping binding sites for MR and GR, respectively. Furthermore, 475 loci were co-occupied by MR and GR. De novo motif analysis resulted in a similar binding motif for both receptors at 100% of the target loci, which matched the known glucocorticoid response element (GRE). In addition, the Atoh/NeuroD consensus sequence was found in co-occurrence with all MR-specific binding sites but was absent for GR-specific or MR-GR overlapping sites. Basic helix-loop-helix family members Neurod1, Neurod2, and Neurod6 showed hippocampal expression and were hypothesized to bind the Atoh motif. Neurod2 was detected at rat hippocampal MR binding sites but not at GR-exclusive sites. All three NeuroD transcription factors acted as DNA-binding-dependent coactivators for both MR and GR in reporter assays in heterologous HEK293 cells, likely via indirect interactions with the receptors. In conclusion, a NeuroD family member binding to an additional motif near the GRE seems to drive specificity for MR over GR binding at hippocampal binding sites.

T he endogenous glucocorticoid hormone of the rat, corticosterone, has a profound action on the brain. This action is mediated in a complementary manner by mineralocorticoid receptors (MRs) and glucocorticoid receptors (GRs), which are unevenly distributed over the brain, but coexpressed in abundance in the hippocampus (1). The high-affinity MRs are already substantially occupied with low corticosterone levels (2). In the initial response to stress, these MRs play a crucial role in retrieval of stressful information and the selection of an appropriate coping response (3)(4)(5). In contrast, the loweraffinity GRs become activated only at higher corticosterone levels, around the peak of the circadian rhythm and during a stress response. GR activation promotes memory storage of the stressful experience (6,7) and behavioral adaptation and recovery (1,8).
Much progress has been made in understanding the cellular mechanism of these coordinated MR-GR-mediated actions of corticosterone (9). Many of the effects depend on the transcriptional activity of the receptors. MR-mediated actions generally raise excitability in the hippocampus. In the most ventral part of the hippocampus, corticosterone prolongs excitability via GR, providing an extended period for encoding of new information. In the dorsal pyramidal cells, GR-mediated actions oppose those mediated by MR (10). That these MR-and GR-mediated effects of corticosterone are sometimes overlapping and, in other processes, are distinct is remarkable, given the large structural similarity between the two receptor types.
MR and GR are members of the nuclear receptor family, with a modular structure of an N-terminal domain, a DNAbinding domain (DBD) and C-terminal ligand-binding domain (LBD). Upon ligand binding, the receptors can dimerize and translocate to the nucleus, where they alter the transcription of their target genes. MR and GR can affect gene expression via tethering to other proteins such as AP-1 and NFkB (11), but in the hippocampus, at least under basal conditions, the main mechanism seems to be via direct DNA binding to the glucocorticoid response element (GRE), palindromic sequences that are variations of AGAACANN-NTGTTCT (12). Homodimers as well as heterodimers of the receptors may occur (13,14). The intrinsically unstructured N-terminal domain contains an Activator Function (AF)-1, and the LBD contains a ligand-dependent AF-2. Through these AF domains, the receptors can interact with coregulators, which can modulate the transcriptional effects by histone-modifying activity and recruitment/stabilization of the transcription factor complex (15). The fact that the two receptors are 94% identical in their DBD (16) suggests that other mechanisms must exist that confer transcriptional specificity underlying the differential effects of MR/GR. It has remained elusive to what extent genomic targets of MR and GR overlap and what determines the specificity of MR and GR DNA binding. We previously identified genomic loci for GR, using chromatin immunoprecipitation-sequencing (ChIP-seq) after a single injection of corticosterone (12). In this study, we aimed to characterize mechanisms that confer MR/ GR specificity by directly comparing their genomic binding sites in the same tissue. Our findings suggest that interactions between MR/GR and DNA-binding transcription factors from the NeuroD family are responsible for MR-selective signaling in the limbic brain and that NeuroD factors are able to potentiate transcriptional activity of both receptor types in vitro.

In vivo experiment
For the ChIP-seq experiment, adult male Sprague-Dawley rats (Harlan, The Netherlands) were housed on a 12-hour light/12-hour dark cycle (lights on at 7:30 AM) with food and water ad libitum. ChIP-seq with MR, GR, or control immunoglobulin G (IgG) antibody was performed on hippocampal tissue of three-day adrenalectomized animals 60 minutes after a single intraperitoneal injection of 300 or 3000 mg/kg corticosterone as a 2-hydroxypropyl-b-cyclodextrin complex, as described (12). ChIP-seq was done on pooled tissue from six animals per treatment, which was redivided, leading to four technical ChIP replicates for both MR and GR. All experiments were performed according to the European Commission Council Directive 2010/63/EU and the Dutch law on animal experiments and approved by the animal ethical committee from Leiden University.

ChIP-seq analysis and motif search
The MR binding data were generated and analyzed in parallel with the previously published data for GR (12). Illumina Genome Analyzer 35-bp single-end reads were uniquely mapped to the rattus norvegicus genome version 4. Peaks were called using model-based analysis of ChIP-Seq (MACS) (17) with the IgG antibody binding dataset as the background. Binding sites were considered overlapping if more than 4 bp were shared. Data were visualized by uploading wiggle files to the Integrative Genomics Viewer (18). Using the annotate peak function of HOMER, binding sites were associated to their nearest gene (19). The Database for Annotation, Visualization, and Integrated Discovery was used for gene ontology analysis (20). Binding sequences were analyzed for the presence of de novo motifs using multiple expectation maximization for motif elicitation (21). The motif size was set from 6 bp minimum to 20 bp maximum, searching also the reverse complement, with a maximum of 10 output motifs, using random, shuffled input sequences as the background model. Enriched motifs were compared against the JASPAR vertebrate database of known motifs using the TOMTOM motif comparison tool. Analysis of Motif Enrichment (AME) was used for enrichment analysis of known motifs in MRexclusive relative to GR-exclusive binding sequences, and Motif Alignment and Search Tool (MAST) was used for directed search of motifs of interest, under default settings (21).

ChIP-qPCR validation
For binding site validation, we performed ChIP-quantitative polymerase chain reaction (qPCR) on hippocampal tissue of intact rats killed at the time of their endogenous corticosterone peak. Antibodies used are listed in Supplemental Table 1. Protease inhibitors (Roche) were added to all buffers during tissue processing and the ChIP procedure. Hippocampal hemispheres were fixated with 1% formaldehyde for 12 to 14 minutes and were homogenized in Jiang buffer [0.32 M sucrose, 5 mM CaCl 2 , 3 mM Mg(Ac) 2 , 0.1 mM EDTA, 10 mM tris (hydroxymethyl)aminomethane-HCl, pH 8.0, 0.1% Nonidet P (NP)-40] using a glass douncer (Kimble-Chase). The following steps were performed in NP buffer [150 mM NaCl, 50 mM tris (hydroxymethyl)aminomethane-HCl, pH 7.5, 5 mM EDTA, 0.5% NP-40, 1% Triton X-100]. Chromatin was fragmented by sonication for 32 minutes with 30-second ON/30-second OFF cycles, using a Bioruptor Pico (Diagenode). Three processed hippocampal hemispheres were pooled and redivided to perform a ChIP for both MR and Neurod2. From each chromatin sample, an aliquot was taken as input material to be able to calculate the percentage of immunoprecipitated DNA.
Chromatin (500 mL) was incubated overnight with 6 mg antibody, after which 20 mL protein A Sepharose beads (GE Healthcare) was added for 2.5 hours. After several washing steps (Supplemental Methods), antibody-bound DNA was eluted from the beads using 10% Chelex 100 (Bio-Rad), further purified by phenolization and dissolved in 50 mL H 2 O. qPCR was performed on 4x diluted ChIP samples according to the protocol described below. Primers were designed to span the GRE of the discovered binding sites and are listed in Supplemental Table 2.

Reporter assays
For mechanistic insights into the effect of NeuroD factors on MR/GR promoter activity, we performed luciferase reporter assays. HEK293 cells (human embryonic kidney, female) were cultured in Dulbecco's Modified Eagle Medium with GlutaMax (Gibco) containing 100 U/mL penicillin, 100 mg/mL streptomycin (Gibco) and 10% fetal bovine serum (PAN-Biotech) at 37°C under 5% CO 2 . For the reporter assays, cells were seeded in a 24-well plate at a density of 80,000 cells/well and grown in medium supplemented with charcoal-stripped fetal bovine serum (Sigma) to exclude cortisol action from the serum. Cells were transfected on day 2 with luciferase construct (TAT1-Luc or TAT3-Luc: 25 ng/well; GRE-At, MRE-At, or GRE-MutAt: 30 ng/well), expression vector for one of the receptors (MMM, DMM, MMD, GGG, DGG, GGD: 10 ng/well), and pCMV-Myc-Neurod1/2/6 (0-1-3-10-50-100 ng/well) and completed with pcDNA3.1 to a total of 300 ng/well and 1.25 mL/well FuGENE (Promega) in unsupplemented Dulbecco's Modified Eagle Medium. Renilla luciferase was used to correct for transfection efficiency (1 ng/well, pRL-CMV; Promega). On day 3, cells were stimulated with corticosterone (10 -7 M or at indicated concentrations; Sigma) dissolved in ethanol, and diluted in medium with a final concentration of 0.1% ethanol. After 24 hours, the cells were washed with phosphatebuffered saline, and reporter protein was measured using the Dual-Luciferase Reporter Assay System according to the manufacturer's instruction (Promega). Briefly, 100 mL lysis buffer was added, and after 10 minutes, 10 mL lysate was transferred into a half-area 96-well plate. Luciferase levels were quantified with 25 mL luciferase assay substrate at 570 nm; subsequently, Renilla signal was measured at 470 nm after the addition of 25 mL Stop and Glo at a SpectraMax L microplate reader (Molecular Devices). All data are presented as mean 6 standard error of the mean. Reporter assays were done in triplicates and repeated at least once.

Real-time qPCR
To validate the NeuroD factor expression in the rat brain, we performed reverse transcription (RT)-qPCR measurements on Sprague Dawley tissue. Hippocampal hemispheres were homogenized in TriPure (Roche) by shaking the tissue with 1.0-mm-diameter glass beads for 20 seconds at 6.5 m/s in a FastPrep-24 5G instrument (MP Biomedicals). Total RNA was isolated with chloroform, precipitated with isopropanol, washed with 75% ethanol, and resuspended in nuclease-free H 2 O. The purity and concentration of the RNA samples were measured on a NanoDrop 1000 spectrophotometer (Thermo Scientific). cDNA was reverse transcribed from 1 mg RNA using random hexamers and M-MLV reverse transcription (Promega) and incubated for 10 minutes at 25°C, 50 minutes at 45°C, and 10 minutes at 70°C. RT-qPCR was performed in duplo on 103 diluted cDNA (5 ng/mL) with final primer concentrations of 0.5 mM using GoTaq qPCR master mix (Promega) in a CFX96 realtime PCR machine (Bio-Rad). The program consisted of 40 cycles of 10 seconds at 95°C and 30 seconds at 60°C, followed by a melting curve generation from 65°C to 95°C in steps of 0.5°C. Primer sequences are listed in Supplemental Table 2.

Allen Brain Atlas correlations
Lists of MR-exclusive, MR-GR overlapping, and GRexclusive genes corresponding to the intragenic and distal promoter (up to -5000 bp) ChIP-seq binding sites were evaluated for their coexpression with each studied NeuroD factor, using the mouse brain gene expression data from the Allen Brain Atlas (24). Pearson's correlation coefficient was used as a measure of similarity between the expression profile of the seed genes (Neurod1, Neurod2, and Neurod6) and every gene in the three aforementioned lists within an anatomical region of interest (25). Correlations were calculated in the hippocampus and its subregions cornu ammonis (CA)1 to CA3 and the dentate gyrus (DG) as well as the striatum. To assess the strength of the association between each gene list and a seed gene, we used a one-sided Wilcoxon rank-sum test.

Data deposition
ChIP-seq data have been submitted to the European Nucleotide Archive and will be available under accession number PRJEB18916.

MR-GR binding site overlap
ChIP-seq on hippocampus chromatin with MR and GR antibodies resulted in the generation of 1.3 to 1.9 3 10 7 reads per sample. After uniquely mapping 66.6% to 83.5% of these reads to the rat genome (rattus norvegicus genome version 4), MACS peak calling with a false discovery rate cutoff at 13.5% [conform (12); Supplemental Fig. 1(a)] resulted in 768 MR sites in the animals injected with 300 mg/kg (MR300) and 1465 MR sites and 2460 GR sites in the animals injected with 3000 mg/kg (MR3000 and GR3000).
We computed the overlap in binding-site genomic coordinates for MR and GR [ Fig. 1(a)]. Additional filtering of MR-and GR-exclusive sites demanded total absence of any peak (the MACS lists including those peaks with a false discovery rate above 13.5%) at the same locus in the GR and MR data, respectively. This resulted in 918 MR-exclusive sites (combined from the MR300 and MR3000 dataset), 475 MR-GR overlapping sites, and 1450 GR-exclusive sites (Supplemental Table 3). These correspond to 45.9% of the total MR sites and 58.9% of the total GR sites being nonoverlapping. ChIP-seq traces of an MR-exclusive, MR-GR overlapping, and GR-exclusive peak are shown in Supplemental Figure 2. The distribution of sites relative to nearest genes is similar for these subsets, with approximately 40% to 45% of the binding sites located within promoters and genes, mainly in introns [Supplemental Fig. 1(b)]. Limited overlap was found between the MR binding sites for the two different dosages, as only 30.6% of the MR300 sites were also found in the MR3000 dataset.

Validation of MR binding sites
The GR binding sites were thoroughly validated before (12). We performed ChIP-qPCR measurements for MR in the hippocampus of intact animals killed at the time of their endogenous corticosterone peak. MR binding was detected at all tested MR-exclusive sites, whereas no MR signal was found at any of the GR-exclusive sites [ Fig. 2(a)]. This demonstrates that the selectivity found in the pharmacological ChIP-seq experiment also occurs in a physiological context.

Processes associated with MR and GR target genes
The biological relevance of the hippocampal binding sites was examined by gene ontology enrichment analysis of target genes, under the assumption that expression of MR/GR bound genes will be regulated by the receptor. Intragenic and upstream (up to -5 kb) binding sites were annotated to generate lists of MR-exclusive, overlapping, and GR-exclusive target genes (Supplemental Table 4).
Functional annotation clustering using the Database for Annotation, Visualization, and Integrated Discovery showed enrichment of brain-related terms, such as "Regulation of cell projection assembly" (MR), "Synapse, Regulation of synaptic plasticity" (overlapping), and "Cell/neuron projection, Synaptic vesicle" (GR) (Supplemental Table 5). Interestingly, for those genes linked to specific MR binding, there was enrichment for "Sodium channel activity," "Calcium ion transport," and "Ion transport, voltage-gated channel activity." Another term specific for MR-exclusive target genes was "Cell adhesion." Furthermore, the annotated GR-exclusive target genes were associated with "Apoptosis and Response to oxidative stress."

An additional motif was found near MR-exclusive sites
To explore the biological mechanism underlying MR/ GR-selective binding, we performed de novo motif analysis on the binding site sequences. For the MR as well as the overlapping and GR datasets, all sites contained a GRE [ Fig. 1(b)]. This is in contrast to the aldosteroneinduced MR cistrome in a human renal cell line, where the majority of binding sites lack a GRE (26). The MRexclusive sites had a more degenerate GRE (lower probability of bases) than the GR-exclusive sites. All subsets also contained a motif that matched the ZNF263 binding site, which was present in 18% to 67% of the sequences. The MR-GR overlapping sites all contained a motif that resembles a GRE half site, suggestive of concomitant dimeric and monomeric (or multimeric) binding of the receptors.
Interestingly, we found a distinct motif near the MRexclusive sites that was not enriched near the GRexclusive or overlapping sites. This additional motif was present in 100% of the MR sites and matched to the Atoh1 binding sequence in the motif database. In a directed search, the Atoh1 motif was also enriched in MR over GR binding (AME; P = 1.11 3 10 -24 ), although in individual cases, we observed this site near GR-bound GREs (MAST; 1% of the GR-exclusive sites). The distance between the GRE and Atoh motif was normally distributed [ Fig. 1(c)] and independent of their respective orientation/strand (in or out of phase) or the binding site relative to genes (intergenic vs intragenic) [Supplemental Fig. 1(c)]. We supposed that another protein binding to this Atoh site can drive MR-specific binding.

NeuroD family members as candidate binders
According to the Allen Brain Atlas (24), Atoh1 is not expressed in the mouse hippocampus ( Fig. 3; Table 1) and is therefore not considered a candidate to bind the MRspecific motif found in the hippocampal ChIP-seq dataset. Atoh1 belongs to the basic helix-loop-helix (bHLH) family of transcription factors (27). Brain-specific family members Neurod1, Neurod2, and Neurod6 do show evident hippocampal expression (Fig. 3) and have been shown to bind the identified CAGATGG motif (28)(29)(30). We validated the very low expression levels (or absence) of Atoh1 and expression of the three NeuroD genes in the rat hippocampus by RT-qPCR (Table 1) and hypothesized (one of) these corresponding proteins could be responsible for the binding site selectivity for MR. By ChIP-qPCR, we demonstrated Neurod2 binding at the same sites at which we validated MR binding [ Fig. 2(b)]. It was however absent from GR-exclusive loci. This gives a proof of concept that Neurod2 might be binding to the Atoh site in vivo. Although Neurod2 was selected based on the availability of chromatin immunoprecipitation-grade antibodies, this result does not exclude involvement of Neurod1 or Neurod6 in MR-selective signaling.

In vivo coexpression of NeuroD factors with putative MR/GR target genes
To get an indication if the other two NeuroD factors could be (co)responsible for the MR-selective binding in vivo, we examined to what extent they are coexpressed with putative MR/GR target genes (as defined by intragenic or up to -5-kb binding of MR or GR). We assessed the spatial coexpression of the MR, overlapping, and GR target gene lists with each of the NeuroD family members based on their expression patterns across the brain using data from the Allen Brain Atlas (24). The MR targets had a stronger coexpression with Neurod6 than the overlapping or GR targets, while for Neurod2, there was no difference between the three lists, and the Neurod1 spatial correlation was highest for the GR targets (Supplemental Fig. 3). This could argue for Neurod6 as an in vivo determinant of MR-selective signaling. Nevertheless, all three NeuroD factors correlated strongly with the expression of MR-exclusive targets and were subsequently studied in vitro.

NeuroD family members potentiate MR/GR transactivation
The putative role of Neurod1, Neurod2, and Neurod6 in MR-specific signaling was further studied in reporter assays in HEK293 cells. All three proteins potentiated MR but unexpectedly also GR transactivation upon corticosterone treatment on a luciferase construct containing a GRE plus the additional Atoh motif in its promoter (GRE-At) by approximately fourfold and sevenfold to ninefold, respectively (Fig. 4). This effect was not observed at a control construct lacking the Atoh binding site (GRE-MutAt), and the NeuroDs could not enhance reporter expression without hormone stimulation. The NeuroD factors thus acted as MR/GR transcriptional coactivators via the identified Atoh motif. For Neurod6, a clear dose-response curve was observed for transfection with increasing doses of expression vector [Supplemental Fig. 4(a)]. We further tested a reporter driven by a more degenerate GRE, as found for the MRexclusive sites [ Fig. 1(b)], combined with the additional Atoh site (MRE-At). The receptors were less efficient in stimulating this luciferase promoter, and the NeuroD effect also did not differ for MR and GR on this reporter [Supplemental Fig. 4(b)].

NeuroD family members increase mainly the maximum transcriptional effect
As the mechanism of action of a receptor modulator can be deduced from both the change in maximum effect as well as the ligand concentration needed for 50% of this effect (EC50) (31), we generated corticosterone dose-response curves with and without cotransfecting Neurod6. The maximum MR/GR effect was increased by Neurod6 presence over the whole concentration range that activates the receptor (data not shown), as was seen before by increased luciferase expression at saturating corticosterone concentrations of 10 -7 M (Fig. 4). Besides, the EC50 was not changed for MR (2.24 6 0.06 3 10 -10 M vs 1.89 6 0.05 3 10 -10 M), while the GR showed a slightly decreased EC50 upon Neurod6 addition (1.03 6 0.05 3 10 -8 M vs 5.29 6 0.02 3 10 -9 M) [ Fig. 5(a)].

NeuroD family members interact with both N-and C-terminal domain-lacking receptors
To further investigate the mechanism of interaction between the MR/ GR and NeuroD factors, reporter assays were performed using truncated receptors [ Fig. 5(b)]. The transactivation by receptors lacking the LBD (MMD and GGD) could be potentiated by the different NeuroDs, although to a lesser extent than for the full-length receptors (MMM and GGG). The potentiation by NeuroDs was also seen without hormone treatment of these constitutively active receptors lacking the LBD. Besides, the NeuroDs could also increase transcriptional activity of the receptors that did not have an N-terminal domain (DMM and DGG). For MR, the NeuroD potentiation of the truncate was comparable to that for the full-length receptor, but for GR, the enhancement relative to nonstimulated cells was less than one-half that of the full-length receptor. Unexpectedly, the DMM and DGG were unresponsive to corticosterone treatment at this reporter, but we did confirm proper transactivation at TAT1-Luc and TAT3-Luc reporters (data not shown). This potentiation of both N-and C-terminal receptor truncations suggests that NeuroD factors have an indirect interaction with MR/GR.

Discussion
This study examined the overlap and specificity of MR vs GR regarding whole-genome hippocampal binding sites. We found both MR-specific, GR-specific, and joint sites, which all contained a GRE. Virtually all MR-specific sites had an Atoh consensus sequence within 400 bp of the GRE, whereas de novo motif analysis did not find this sequence near sites that showed GR occupancy (including overlapping sites). Neurod1, Neurod2, and Neurod6 are coexpressed with MR and/or GR in the principal The limited overlap found in MR and GR binding sites is in accordance with the distinct roles of the two receptors in the hippocampus (6,10,32). It should be noted however that the lower sequencing depth of our analysis might have precluded the detection of weaker binding sites. In addition, as we performed ChIP-seq on whole hippocampi, the small proportion of shared targets could also be a result of cell type-specific MR/GR loci as a consequence of the differential MR and GR expression patterns throughout the hippocampal area. Coexpression of MR and GR is observed in the majority of CA pyramidal and DG granular neurons, with the exception of CA3 pyramidal cells that have high MR but low GR levels (33). Besides, GR is also expressed in glial cells (34,35).
Limited overlap in the MR binding sites for the two different corticosterone doses (MR300 vs MR3000) could be explained partly by an insufficient depth of sequencing (limit of detection). In addition, it might reflect different concentrations of activated MR in the nucleus, in combination with differential affinity of binding sequences for the receptor, even if the majority of MR likely was occupied by the lower dose. A recent study suggests that high receptor occupancy does not necessarily translate into high DNA binding, and MR can show circadian variation in target site occupancy (36). Differences in sensitivity between MR-expressing cell types might also be of relevance. A last possibility may be opening up of chromatin domains via GR, making GREs available for MR binding. In the same line, heterodimerization of MR and GR could play a role (36).
The additional MR-selective motif could be bound by Neurod1, Neurod2, and Neurod6, as evidenced by response element-dependent transcriptional modulation. NeuroD proteins are members of the bHLH protein family and are known for their function in neuronal differentiation (28,37). Neurod1 knockout mice lack a dentate granule cell layer (38), and heterozygous Neurod2-deficient mice show impaired contextual and cued freezing in a fear-conditioning task (29). Our binding sites were detected in adult rat hippocampal tissue, suggesting that the NeuroD factors not only regulate neuronal differentiation during development, but also can be crucial in later processes such as cell survival or retaining differentiation status. As the hippocampal DG is the main site of adult neurogenesis (39), this might also provide a role for NeuroD factors in adulthood, although their expression is much wider than neurogenic zones. Furthermore, overexpression of Neurod2 in the ventral hippocampus has recently been shown to increase stress susceptibility in a chronic social defeat paradigm (40), posing a role for Neurod2 in depression.
Based on mouse brain expression data from the Allen Brain Atlas, we observed that Neurod6 expression is restricted to the CA subregions of the hippocampus, while the lower Neurod1 signal seems to be more pronounced in the DG (Fig. 3). Furthermore, Neurod2 expression is observed throughout the whole hippocampus and seems to be at levels similar to Neurod6, as we validated by RT-qPCR on rat hippocampal tissue. The three NeuroD proteins have a highly similar bHLH region (37), which makes it not surprising that all members can bind the additional Atoh motif derived from our ChIP-seq analysis and potentiate MR/GR transactivation in reporter assays. Based on our data, we cannot pinpoint which of the family members is/are responsible for the MR-specific binding, although Neurod2 was detected at rat hippocampal MR-exclusive sites [ Fig. 2(b)] and target gene correlations suggest that Neurod6 is also a likely candidate (Supplemental Fig. 3). We cannot exclude the possibility that another bHLH-containing protein binds to the Atoh motif and drives the exclusive MR action.
Neurod1-or Neurod2-deficient mice that also lack Neurod6 have more severe brain abnormalities than the single mutants, indicating cooperation and/or partial redundancy (41,42). A model in which Neurod1, Neurod2, and Neurod6 are each involved in MR-specific signaling within a certain subregion of the hippocampus might be considered. The in vivo-found, MR-exclusive motif does not discriminate in vitro in reporter assays. This discrepancy could be explained by the possibility that, in the luciferase assay, the receptors use different intermediate transcriptional proteins than in the hippocampus. The observed coactivation of both N-and C-terminally truncated receptors implies that NeuroD family members interact via the transcriptional complex of MR/GR rather than directly with the receptors. A side note is that we cannot rule out interactions via the DBD or hinge region of the receptors. Nevertheless, the suggested indirect interaction is also supported by the fact that the Atoh motif was found at a variable distance up to 400 bp from the GRE. It is likely that the HEK293 cells lack or do contain other variants of the proteins that are crucial to mediate the NeuroD effect on selective MR transcriptional activity. For example, the pool of coregulators present in a cell is highly tissue specific and can result in opposite effects on gene transcription (43). Also, bHLH protein heterodimerization partners might be responsible for an MR-specific effect (27). Besides, as the chromatin landscape is a crucial determinant of a transcription factor cistrome (44), the lack of chromatin context in the luciferase assay might make it difficult to mimic the exact conditions of in vivo binding and transcription. Interestingly, Neurod1 itself can also induce chromatin remodeling and increase neuronal gene accessibility (45).
In lung fibroblasts, the Atoh1 motif was detected, although nonsignificantly, near GR-bound sequences (46). Directed motif search by MAST showed the presence of a Neurod2 binding site in 1% of our GR-exclusive sites, but the Atoh motif was clearly enriched in MRexclusive over GR-exclusive sites using AME. It might be that a NeuroD factor through binding to the Atoh motif only excludes GR binding and subsequent transactivation when MR is present, which can be another reason that we do not find a difference in MR/GR Figure 5. NeuroD increases the maximum MR/GR effect via an indirect mechanism of action. (a) Dose-response curves for corticosterone stimulation of MR and GR in absence and presence of Neurod6 to determine the effect on EC50. The luciferase activity is presented as percentage of the maximum effect. Sigmoidal curves were fit by nonlinear regression using a variable slope model. (b) Effect of NeuroD factors on truncated receptors. HEK293 cells were transfected with full MR or GR (MMM, GGG) or variants lacking the N terminus (DMM, DGG) or C terminus (MMD, GGD), the GRE-At construct, and Neurod1, Neurod2, or Neurod6 (10 ng/well) and stimulated with corticosterone (10 -7 M). All nonstimulated cells were normalized to 1; for the constitutively active MMD and GGD, luciferase levels were normalized to nonstimulated control cells. a.u., arbitrary unit. potentiation in vitro when studying the receptors in isolation. In cotransfections of MR and GR combined with selective pharmacological activation, both receptors were also potentiated by Neurod6 (data not shown). Furthermore, the highly dynamic DNA-binding kinetics of nuclear receptors are not supportive of a competitionbased mechanism (47,48). A recent study also found motifs that were associated with absence of GR binding, and proteins recognizing these sequences could indeed decrease GR occupancy and transactivation (49).
In conclusion, we identified a motif that is associated with MR-selective signaling in the rat hippocampus. NeuroD factors could bind this motif and, via indirect interactions, were found to potentiate the MR/GR transcriptional activity in HEK293 cells. The data support a model in which NeuroD factors stabilize MR binding in vivo by interacting with cell-specific components of the MR-associated transcriptional complex. Further elucidation of distinct MR/GR downstream pathways will enable us to more specifically target aspects of glucocorticoid signaling for treatment of stress-related disorders.