Sex-Dependent Changes of miRNA levels in the hippocampus of adrenalectomized rats following acute corticosterone administration

Abstract

There is a complex interaction between sex and stress hormones in the hippocampus, a key brain area for cognitive/emotional processes. Interestingly, compelling evidence has pointed out a sex-biased effect of the glucocorticoids, a primary stress hormone, on miRNAs -key posttranscriptional regulators- in non-neuronal tissues. Here, we explored sex-biased effects of the endogenous glucocorticoid corticosterone on the miRNA expression profile in the rat hippocampus. Adult adrenalectomized female and male rats received a single corticosterone (10 mg/kg) or vehicle injection and after 6 h hippocampi were collected for microarray-based miRNAs analyses. Corticosterone triggers a sex-biased differential expression of hippocampal miRNAs derived from genes located in chromosomes 6 and X, despite of the sex. Putative promoter analysis unveiled that most of corticosterone-responsive miRNAs genes contained putative motifs for either direct or indirect glucocorticoid actions in both sexes. Validated target analysis of corticosterone-responsive miRNAs showed a more complex miRNA-mRNA interaction network in males, compared to females. Functional prediction analysis revealed that several hippocampal-relevant pathways were affected in both sexes, such as neurogenesis and neurotrophin signaling. Altogether, our results indicate a sex-biased effect of corticosterone in hippocampal miRNA expression, which may engage in sculpting the sex differences observed at more complex levels of hippocampal functioning.

INTRODUCTION

The hippocampus is a key brain structure for learning and memory processes and its physiology displays important sex differences across numerous mammalian species [1]. In rodents, different approaches aimed at evaluating differences in synaptic efficacy indicated that males exhibit a larger early and late- long term potentiation (LTP) than females in the dentate gyrus and the CA3 and CA1 regions of the hippocampus [2, 3]. These dissimilarities may be linked to sex differences observed in episodic and spatial memory [4]. For instance, men show better spatial memory, whereas women show better object memory (recently revised in [5]). These distinctions may arise from sex-linked neuronal complexity, connectivity with other brain areas and the neurogenesis rate of the hippocampus [4]. Interestingly, a sex-bias in hippocampal physiology can also be visualized following stress exposure [4], suggesting that the actions of stress hormones -such as glucocorticoids (GCs)- may be involved in establishing these differences.

GCs -through the binding to glucocorticoid (GR, low affinity) and mineralocorticoid (MR, high affinity) receptors- modulate the functioning of the hippocampus and other brain regions [6–8]. Balanced activity of the MR and GR is crucial for homeostasis in the hippocampus, in which low levels of GCs increase neuronal excitability by activating the MR [9]. In contrast, GRs in the hippocampus modulate L-type calcium currents, facilitating the slow after-hyperpolarization of neurons (Myers, McKlveen et al. 2014). Moreover, GCs may modulate several processes in the rodent hippocampus, including the performance on spatial memory tasks [10], both LTP and long-term depression (LTD) [11], the proliferation rate of neural progenitors, and the survival of mature granular cells [12]. Furthermore, GCs may trigger changes in the expression of neurotrophic factors and dendritic spine remodeling in the hippocampus [13]. Thus, GCs have an ample impact on hippocampal neuroplasticity through changes in gene expression, involving either direct or indirect actions of the GR and MR [14, 15].

Interestingly, an interaction between sex and GCs has been described in non-neuronal tissues [16–18]. For instance, an extensive down-regulation of mature microRNAs (miRNA) is produced in lymphocytes in response to GC treatment [19]. miRNAs are essential regulators of gene expression at the post-transcriptional level. This regulation is mediated by their interaction within the 3’UTR region of mRNAs by interfering with either translation and/or accelerating targeted mRNAs decay [20, 21]. More recent reports have indicated that miRNAs participate as regulators of gene transcription [22]. Based on sequence complementarity, each miRNA has the probability of regulating the translation of hundreds of different mRNAs [23]. More interestingly, different brain regions of adult rodents display an uneven distribution pattern of miRNAs, suggesting a differential role in brain functioning [24]. Remarkably, the neonatal brain displays an extensive sex bias in miRNA expression, involving regulation by both sexual hormones and sex chromosomes [25]. Additionally, sex-specific microRNAome responses are observed in the hippocampus after exposure to genotoxic mediators, suggesting a sex-biased regulation of miRNA expression [26]. Importantly, the miRNA expression is altered in a sex-dependent manner by GCs in non-neural models [27]; however, the effect of GCs on the regulation of miRNA levels in the brain, particularly in the hippocampus, is still unknown.

In this study, we hypothesized that acute GC administration to adrenalectomized rats triggers a sex-biased miRNA expression pattern in the hippocampus. Then, we searched for expressed miRNAs in the female and male hippocampus and determined if these targets were enriched for a particular functional process.

RESULTS AND DISCUSSION

To define the effect of GCs on the levels of miRNAs expressed in the hippocampus, male and female rats underwent ADX, and after seven days, we administered a single injection of vehicle or corticosterone (10 mg/kg). Hippocampal RNAs were obtained 6 h post hormone injection and we used microarray analysis to determine whether corticosterone triggers a delayed expression of miRNAs. The effectiveness of ADX and variation in corticosterone levels after its administration is shown in Fig. S1.

Sex-bias in miRNA responsiveness to corticosterone in rat hippocampus

We used the ADX-corticosterone/ADX-saline ratio of the normalized chip intensities to identify the differentially expressed miRNAs by using a microarray containing both 680 mature and 486 pre-miRNA rat probes. We identified 17 miRNAs that were differentially expressed in female rat hippocampus (adjusted p-value < 0.05 and fold change ≥ 1.5; Table S1). The volcano plot shows that five miRNAs decreased (blue dots) and 12 miRNAs increased (red dots) their levels after 6 h of corticosterone administration in females (Figure 1A). On the other hand, the same analysis conducted in males identified 27 miRNAs that were differentially expressed in the hippocampus, of which 16 decreased, and 11 increased their expression after 6 h of corticosterone administration (Figure 1A; Table S1). After comparing the differentially expressed miRNAs in both sexes 6 h after corticosterone treatment, we identified only 3 common miRNAs; miR-376c and miR-450 were up-regulated in both male and female hippocampus, however, miR-758 was up-regulated in female, but down-regulated in male rats (Figure 1B). These results indicate that 6 h after its administration, corticosterone triggers a sex-biased miRNA expression profile in the hippocampus. Compelling evidences have unveiled that male and female adaptations to chronic stress are at least in part facilitated by interactions among sex steroids and the GR [28]. Here, we report for the first time that miRNAs change their expression in a sex-biased manner in the hippocampus following systemic corticosterone administration, suggesting that these cellular regulators may participate in establishing sex differences that involve the hippocampus.

Figure 1. Differentially expressed miRNAs in female and male rats after corticosterone administration and chromosome location of miRNA genes.

(A) Volcano plot of differentially expressed miRNAs in female and male rats. The horizontal axis represents a log2 fold change (ADX-Corticosterone/ADX-Vehicle) and the vertical axis represents the negative log10 of the p-value. miRNAs that decreased (blue) or increased (red) their expression are shown. Gray indicates those miRNAs whose expression did not change significantly, or when their fold change was outside the cut-off value. Fold change ≤ −1.5 or ≥ 1.5 and p-value < 0.05. (B) Venn diagram of differentially expressed miRNAs in female (purple circle) and male (teal circle) rats. (C) Distribution of miRNA genes sensitive to corticosterone in autosomic chromosomes and the X chromosome (D) Chromosomal location of miRNA-coding genes (Chi-square, df 43,42, 14; P < ,0001).

Chromosomal locations of miRNA genes responsive to corticosterone

miRNA genes are distributed throughout the mammalian genome within a gene or in intergenic regions [29]. We found that the distribution of corticosterone-responsive miRNAs genes in autosomic and X chromosomes (chr) was similar in males and females (Figure 1C; χ²: 43,42, df: 14, P=0.9198). However, the expression of miRNAs after corticosterone administration was associated with different chromosomes. In females, we detected that miRNAs sensitive to corticosterone were hosted in 8 different autosomal chromosomes (1, 2, 3, 4, 6, 9, 10, 13; see Figure 1C and Table S1) and the X chromosome. Additionally, and only in females, we observed that chromosomes 3, 4, and 9 showed some miRNA genes responsive to corticosterone. For instance, we observed miR-296-3p (intergenic region of chromosome 3) levels were reduced, while miR-671 [located in the Chpf2 ( chondroitin polymerizing factor 2) exon in chromosome 4] and miR-26b-3p [located in the Ctdsp2 (CTD Small Phosphatase 2) intron in chr 9] levels were increased (Table S1 and Figure 1C).

In the male hippocampus, in contrast, we found that miRNAs sensitive to corticosterone were in 11 different autosomal chromosomes (1, 2, 5, 6, 8, 10, 11, 13, 17, 18, and 19) and the X chromosome. Moreover, some miRNAs in chromosomes 5, 8, 11, 17, 18, and 19 were sensitive to the hormone only in males (Table S1 and Figure 1C). Among these, we identified miR-455-3p [located in the Col27a1 (a type XXVII collagen) intron in chr 5), miR-3588 (chr11, intergenic), miR-143-5p (chr18, intergenic), miR-328a-3p [Elmo3 (engulfment and cell motility protein 3) intron in chr19], and miR-3596a (chr8, intergenic), whose levels were reduced by corticosterone administration. In contrast, miR-466c-5p (chr17, intergenic), miR-133a-5p (Mib1 (E3 ubiquitin protein ligase 1) intron in chr18] and miR664-1-5p (chr18, intergenic) increased their levels following corticosterone administration (Table S1 and Figure 1C).

On the other hand, some miRNA genes located in chromosomes 1, 2, 6 and X were sensitive to corticosterone both in the male and female hippocampus. We identified an increased expression of miR-204-3p in females, while in males the transcript from the opposite strand (i.e., miR-204-5p) was also increased, both derived from the same locus in chromosome 1. As has been described, both miRNA strands can be functional in some tissues, whereas in other tissues there is a strong selection for one of the strands [30]. Besides, the miRNA strand selection seems to be a tightly controlled process that may be influenced by cell phenotype, developmental status among other factors [31]. In non-neuronal cells, for example, the transcript from both strands of miR-204-5p participates in the regulation of cell function [32, 33]. In females, we also detected increased miR-186 levels, hosted within the Zranb2 gene that encodes the splicing factor ZRANB2 in chromosome 2. In males, we detected the expression of miRNAs whose genes were in chromosome 2: miR-582–5p, hosted in the intron of LOC108350211 and miR-2985, hosted within the Gria2 gene, which encodes one subunit of the glutamate AMPA receptor.

Additionally, we found that 6 miRNAs hosted in chromosome 6 increased their levels after corticosterone administration in the female hippocampus, with five of them located in close proximity (<10 kb): miR-323-3p, miR-758-5p, miR-1193-3p, miR-376c-3p and miR-539-3p (Table S1 and Fig. 2A). Interestingly, in males we also observed that corticosterone administration increased miR-376c-3p and decreased miR-758-5p levels. Remarkably, the aforementioned miRNAs that were in close proximity belong to the miR-379-410 cluster located in the imprinted Dlk1-Dio3 gene cluster locus in chromosome 6 [34] (Fig. 2B and Table S1), which is orthologous to the mouse distal chromosome 12 and to the human chromosome 14q32 [35]; evidencing an evolutionary conservation across different species. The Mirg is the host gene for miR-379-410 members (Fig. 2B), and is expressed in several brain areas of adult mice, including the cortex and hippocampus (Allen Brain Atlas; http://mouse.brain-map.org/experiment/show?id=70946557). Additionally, in primary hippocampal cultures, the expression of Mirg is limited to MAP2-positive neurons, and absent from non-neuronal cells, such as glial cells [36]. These evidences suggest that the changes observed in the miR-379-410 cluster members could be related to neuronal, rather than glial function.

Figure 2. Genomic location of miRNA-coding genes.

(A) The location of miRNAs in chromosomes are shown as circular ideograms. (B) miRNA gene clusters located in chromosomes 6 and X are sensitive to corticosterone. Arrows alongside miRNAs denote their differential expression (either up- or down-regulation) upon corticosterone administration in female (purple) and male (teal) hippocampus.

Additionally, in males, we detected another two miRNAs: miR-409a-3p and miR-412-3p, which increased and decreased their levels with hormone treatment, respectively. These miRNAs belong to the miR-379-410 cluster, whose genes are in the coding region of Mirg. Moreover, we detected other corticosterone-responsive miRNAs that also belong to the imprinted Dlk1-Dio3 gene locus in males, but outside of the miR-379-410 cluster: miR-673-5p and miR-3544, whose genes are located between the Gtl2 and anti-Rtl1 coding regions, and miR-1188-3p and miR-370-3p, which are located within the Rian coding region.

The expression of miRNAs belonging to the miR-379-410 cluster plays an important role in neocortex development [37] and synaptic plasticity [35]. Interestingly, the deletion of the cluster enhances anxiety-related behavior [38], but increases the sociability in mice [36]. The regulation of both gene and miRNA expression in the Dlk1-Dio3 region is poorly understood. Some studies have shown that the expression of the miR-379-410 cluster is restricted to developmental and early postnatal stages [34]. Additionally, in adult male rats, acute immobilization stress [39], or sub-chronic and chronic restraint stress, trigger variations in the levels of some miRNAs from the miR-379-410 cluster in the hippocampus [40] These evidences reinforce that stress mediators may regulate the expression within the cluster. Additionally, Fiore et al. demonstrated that the expression of several miRNAs belonging to the miR-379-410 cluster are up-regulated by neuronal activity in a Mef2-dependent manner [41].

In the female hippocampus, four differentially expressed miRNAs after corticosterone administration are hosted in the X chromosome (Fig. 2; Table S1). miR-764 levels, which are reduced upon hormone treatment, derive from the 5′ untranslated region of the pre-mRNA 5HT_2C receptor (intron II), along with other miRNAs [42]. The 5HT2c receptor is expressed in the brain, predominantly in neurons [43] and other non-neuronal cells [42]. Evidences have indicated that the miRNAs linked to intron II of the pre-mRNA 5HT_2C receptor are cell-type specific, suggesting a differential processing of the pre-miRNA and/or cell-type-specific stability [42]. On the other hand, miR-743b-3p and miR-871-3p are derived from genes located in intergenic regions of the X chromosome and their expression is reduced by corticosterone exposure. Notably, miR-743b-3p and miR-871-3p are part of the FX-MIR cluster/miR-743-465 cluster located in the Fragile-X region of the X chromosome. miRNAs encoded by this cluster may target the immediately adjacent gene, Fmr1 [44] -an important regulator of neuronal function- and participate in the regulation of rat cells reprogramming to a pluripotent state [45]. Finally, the miR-450a-5p -which also increased its levels with corticosterone- belongs to the miR-322-503 miRNA cluster within the X chromosome; a cluster described in humans, mice, and bovines and composed of six miRNAs: miR-322 (miR-424 in humans), miR-503, miR-542, miR-450a-1, miR-450a-2 and miR-450b [46]. This cluster promotes differentiation and induces G1 arrest by targeting an overlapping set of cell cycle regulators, and its expression is sensitive to estrogen in MCF-7 breast epithelial cells [47], suggesting sexual hormone-sensitive regulation.

In the male hippocampus, 22% of the differentially expressed miRNAs by acute corticosterone treatment are hosted in the X chromosome. The miR-223-3p, which was up-regulated, mediates the differentiation of neural stem/progenitor cells [48]. miR-532-3p -which belongs to miRNA-532-502 cluster- was down-regulated by corticosterone administration and, interestingly, this miRNA changed its levels in sperm derived from rats exposed to stress [49]. Similarly to females, miR-450a-5p, which belongs to the miR-322-503 miRNA cluster, is up-regulated by corticosterone in males. miR-351 levels are reduced by corticosterone; its gene is only found in rodents and is derived from the primary transcript of the miR-322-503 cluster [50, 51]; miR-351-3p has varied roles in development, cancer and inflammation [51]. We also found that miR-881-5p and miR-3580-3p increased and decreased their levels after corticosterone administration, respectively, and they have been identified as part of the miR-743-465 cluster [45].

Our data indicate that the most abundant miRNAs responsive to corticosterone both in females and males are hosted in the same clusters: the miR-379–410 cluster (Dlk1-Dio3 region) in chromosome 6 and the miR-322-503 and Fx-mir (FX-MIR) (miR-743-465) clusters in the X chromosome. Furthermore, we detected a glucocorticoid-induced reduction in the levels of miRNAs hosted in the miRNA-532-502 cluster only in males. miRNA expression associated with clusters of both autosomic and sex chromosomes represented almost 48% of the total differentially expressed miRNAs. Intriguingly, only 3 miRNAs derived from these clusters were shared between males and females (miR-758-5p, miR-376c-3p, and miR-450a-5p), suggesting a sex-biased effect on their expression, processing, or stability. In females, the expression of miRNAs belonging to the same cluster displayed the same direction change; i.e., members of miR-379-410 were up-regulated. Additionally, members of both the miR-743-465 and miR-322-503 clusters were down-regulated and up-regulated respectively (Figure 2). In males, the expression of miRNAs within a cluster (miR-379-410, miR-322-503 and miR-743-465) showed dissimilar regulations. Considering the variability in the identity of cluster miRNAs expressed in female and male hippocampus, it is plausible that the expression of miRNAs within a cluster displays a sex × context interaction. In a similar vein, we detected that miR-758-5p -part of the miR-379-410 cluster- raised its expression with hormone treatment in females, but decreased in males, suggesting a peculiar expression mechanism associated with the animal’s sex. Interestingly, a sex-dependent change in miRNAs expression has been reported in BNST (bed nucleus of stria terminalis) following social isolation stress through adolescence [52]. These authors reported that many stress-responsive miRNA genes were located in chromosome 6 and, particularly, some of them were found in the Dlk1-Dio3 region, but the identity of those miRNAs differs from our data [52]. These findings suggest that early-life stress changes the miRNA expression profile due to a variation of local regulators that control miRNA expression within a cluster. In line with this idea, another report has shown that the expression of two other miRNAs (miR-134 and miR-485) of the miR-379-410 cluster increase their levels postnatally, which led the authors to propose that this cluster may not be expressed as a polycistronic transcript [34].

Putative transcriptional regulation of miRNA genes sensitive to corticosterone

The expression of microRNAs may change dependently (intragenic miRNAs), or independently (intergenic miRNAs) of their host genes. Intergenic miRNAs have their own promoters, are expressed autonomously and can be controlled by different transcription factors [53]. Furthermore, clustered miRNA genes are frequently co-expressed in order to control biological processes [54]. The diversity of the corticosterone-responsive miRNAs detected both in male and female hippocampus that we found highlighted the relevance of evaluating putative regulatory elements in promoter regions of miRNA genes, which could help explain a co-regulatory mechanism. We used in silico approaches to identify cis-elements (motifs) in the putative promoters of the differentially expressed miRNAs in female and male datasets, separately, by using MEME software [55]. We chose the 1 kb sequence upstream of the transcription start site of each miRNA precursor as putative regulatory regions, based on other studies [56]. Using the FIMO algorithm [57], we first searched for glucocorticoid response elements (GREs) and mineralocorticoid response elements (MREs), which can be recognized by activated GR and MR, respectively. In addition, it is well-known that GR uses DNA-binding-independent mechanisms to accomplish context-specific transcriptional outcomes. For instance, tethering of GR to the transcription factor NF-κB through protein-protein interactions results in the transrepression of targeted genes [58]. Interestingly, the activity of NF-κB is related to neurogenesis and apoptosis processes in the hippocampus [59]. Therefore, we also looked for NF-κB binding motifs. The motif scanning with FIMO revealed that, in females, 8 corticosterone-responsive miRNA genes retrieved a known motif (Figure 3): 1 for GR binding (miR-497), 3 for MR binding (miR-1193, 671 and 871), and 4 for NF-κB binding (miR-26b, 296, 323 and 758). On the other hand, in the male dataset, we identified known motifs in 13 genes of differentially expressed miRNAs (Figure 3): 3 for GR (miR-328a, 351 and 881), 1 for MR (miR-324), and 9 for NF-κB (miR-351, 3544, 409a, 412, 466c, 582, 664, 673 and 758).

Figure 3. Transcription factor binding motifs in putative promoter regions of differentially expressed miRNAs.

Differentially expressed miRNAs in female (top, purple) and male (bottom, teal) rats whose putative promoter region contained binding motifs for the glucocorticoid receptor (GR), mineralocorticoid receptor (MR), or NF-κB. The position frequency matrix (PWM) logo used for each transcription factor is indicated on top of the table. Motif occurrences with a p-value < 0.0001 are shown.

As an extension of these findings, we conducted an untargeted search for overrepresented sequence motifs of variable-length using the motif discovery algorithm MEME, in order to identify motifs recognized by known transcription factors whose activity could be regulated in some way by corticosterone administration. Another approach was to find new motifs represented in several putative promoters of miRNA genes, which may correspond to a novel transcription factor.

Motif discovery revealed at least five recurring transcription factor-binding motifs (top 5 of MEME algorithm analysis), and many of them were heterogeneously distributed in tandem along the promoter sequences of differentially expressed miRNAs in the female hippocampus (Figure 4A). The motifs 1 and 3 were present in 5 and 4 miRNA gene putative promoter sites, respectively, and showed a consensus motif with a low percentage of CG and a similar number of bases (40 and 39, respectively) (Figures 4A and 5A). After comparing motifs 1 and 3 with binding profiles of known transcription factors, we were not able to identify any known regulatory elements. In contrast, motif 2 had a consensus sequence of 21 bases, was the most prevalent among the miRNA gene putative promoters (9 sites) and may bind the Hepatocyte nuclear factor 4 alpha (HNF4a), Sp1, and Sp4 transcription factors. Some evidences have shown that dexamethasone -a specific GR agonist- increases Hnf4α transcript levels in the liver, making feasible a relationship between GR and HNF4a [60]. Sp1 and Sp4 proteins have approximately 50% identity in primary structure and bind the same DNA motifs with similar affinity [61]. Sp1 is essential for embryogenesis, and its disruption has been linked to neurodegenerative diseases [62]. In adult mice, Sp4 expression is restricted to pyramidal neurons and dentate gyrus in the hippocampus [63]. Motif 4 was represented in 5 putative promoters, with a consensus sequence of 40 bases with low CG content, and it may bind the transcriptional enhancer factors TEF-1 (TEAD1) and −2 (TEAD2). A report has indicated that the GR may interact with a member of the TEAD transcription factors, probably regulating gene expression [64]. Motif 5, present in 6 sites, is 26 bp-long, with a medium CG content, and it may bind the Mothers Against Decantaplegic homologue-3 (SMAD3). The GR can interact with SMAD3 and SMAD4 in vitro and in vivo through its C-terminal portion [65]. Thus, it seems that motifs 2, 4, and 5 found in putative promoters of differentially expressed miRNA in females by corticosterone administration may be recognized by transcription factors that, in some way, modulate their activity through the participation of GR.

Figure 4. Discovered motifs in hypothetical promoter regions of differentially expressed miRNAs.

(A) Top five discovered motifs in promoter regions of female miRNAs. (B) Top five discovered motifs in promoter regions of male miRNAs. Sites indicates the number of matches contributing to the construction of the corresponding motif. Only motifs statistically significant are shown (E-value < 0.05). N, X (any base); V (not T); H (not G); D (not C); B (not A); M (amino matches A and C); R (purine); W (weak matches between A and T); S (strong matches between C and G); Y (pyrimidine); K (keto matches G and T).

Figure 5. Motif distribution in promoter regions of differentially expressed miRNAs.

Novel motif discovery was performed with the MEME algorithm, and specific motif scanning (for GR, MR and NF-kB transcription factors) was performed using the FIMO algorithm, both available at the MEME-Suite website (Bailey et al., 2009). P-values indicated in the figure correspond to the combination of the individual matching p-values of each motif in the respective sequence (individual p-value <0.05). Only statistically significant interactions with transcription factors are shown. Motifs in promoter regions of differentially expressed miRNAs in (A) female and (B) male rat hippocampus. Color code: motif 1 (red), motif 2 (blue), motif 3 (green), motif 4 (purple), motif 5 (yellow), containing potentially binding sites for the indicated transcription factors.

On the other hand, a similar analysis with the male dataset evidenced a more homogeneous motif distribution (Figure 4B), with a high recurrence of motif 1 in almost all the putative promoters analyzed. This motif was found in 21 sites and had a width of 15 bases, with a high content of CG. TomTom analysis revealed that motif 1 may be recognized by the MYC Associated Zinc Finger Protein (MAZ) transcription factor, which is abundant in the hippocampus [66]. On the other hand, motifs 2, 3, and 4 had a low percentage of CG content, with sites ranging from 4–5 and 40 bases of width; nonetheless, we were not able to recognize any known regulatory proteins for these motifs. Conversely, motif 5 -with intermedium CG content and 38 bp of width- was present in 4 putative promoters and can be bound by SMAD2 and SMAD3 transcription factors (Figure 5B). The transforming growth factor-β (TGF-β) superfamily of growth factors lead the phosphorylation of selected SMAD proteins that, in turn, translocate into the nucleus to regulate the expression of different target genes involved in cell proliferation and neuronal survival [67]. We have previously described that ADX in males promotes an increase in TGF-β1 expression in the hippocampus [68]. Interestingly, a high concentration of GCs inhibits the proliferation and neuronal differentiation in immortalized human hippocampal progenitor cell lines, an effect explained by a reduction in TGF-β-SMAD2/3-signaling [12]. Thus, acute corticosterone administration may trigger a crosstalk with TGF-β signaling, inducing changes in miRNAs expression.

The location of each putative response element for GR, MR, and NF-κB, as well as the discovered motifs, either in female or male corticosterone-responsive miRNA putative promoters, are shown in Figure 5. A more detailed analysis of our findings in females revealed that the miR-186 putative promoter contains motif 2, with a predicted binding capacity for HNF4a, SP1, and SP4; however, no response elements for corticoids were found. Nonetheless, as we described above, there is a direct relationship between the GR and HNF4a, which may explain, in part, the enhanced expression of miR-186 after corticosterone administration [60]. miR-153 -located in an intron of the Ptprn2 gene- raised its levels after hormone treatment and contained motif 5, bestowing the possibility to interact with SMAD3. Interestingly, the levels of miR-488 -hosted in an intron of the Astn1 gene- exhibited a rise and its putative promoter harbored all discovered motifs, i.e., motifs 1 and 3, along with motifs for HNF4a, SP1, SP4 and TEAD1/2 and SMAD3. Moreover, miR-26b (Ctdsp2 intron in chr9) presented motifs with binding capacity for HNF4a, SP1, SP4, and NF-κB; and its levels increased with corticosterone. Following acute restraint stress, miR-26a/b increases its levels in the frontal cortex, but not the hippocampus, of old male mice [69]. These results are in accordance with the positive effect of corticosterone on miR-26b levels. On the other hand, considering the miRNAs for which we detected either known or novel motifs, only 3 reduced their levels upon corticosterone treatment. For instance, miR-497 (exon RGD1308134 in chr 10,) showed 3 motifs, which may be recognized by TEAD1/2, HNF4a, and GR. miR-871 (miRNA 743–465 cluster) also showed binding capacity for the MR, HNF4a, SP1, and SP4. On the other hand, the putative promoter of the miR-296 gene -located in the intergenic region of chromosome 3- has two binding sites for NF-κB and one for TEAD-1/2. Considering that miR-296 levels were reduced in response to corticosterone, this effect may be produced by a tethering mechanism involving the GR.

Finally, in the female hippocampus, miR-323, miR-758, miR-1193 and miR-539 -all of which increased their levels after corticosterone treatment- are members of the miR-379-410 cluster located in chromosome 6 and seem to be controlled by response elements located upstream of miR-323, represented by two binding sites for NF-κB, TEAD1/2 and SMAD3 and HNF4a. Interestingly, transcription of the miR-379-410 cluster is induced by increasing neuronal activity in primary neuronal cultures and depends on the transcription factor MEF2. Fiore’s group identified a MEF2-binding element about 20 kb upstream of the miR-379-410 cluster and they suggested that this cluster is transcribed as a polycistronic precursor [41]. Nonetheless, it is well known that members of this cluster are differentially expressed in the adult brain [34], suggesting the presence of other regulatory elements.

Regarding our analyses in males, we detected that several corticosterone-responsive miRNAs whose expression change profiles showed the same direction, also shared certain motifs. For instance, most of the miRNAs whose putative promoters contained a MAZ-binding motif tended towards a down-regulation (Figure 5), suggesting MAZ as a master negative regulator in males. Interestingly, this down-regulation trend of miRNAs containing MAZ-binding motifs is also observed when one or more of the other regulatory motifs (SMAD, GR, MR, NF-κB) is present (Figure 5). However, some miRNA genes containing a MAZ motif were up-regulated following corticosterone treatment (e.g., miR-664-2, 466c and 204). More interestingly, the positive effect of corticosterone was also observed in the presence of other regulatory motifs (e.g., miR-664-2 and miR-466c also contained NF-κB-binding motifs, apart from MAZ). In males, some miRNAs belonging to the imprinted Dlk1-Dio3 gene locus; namely, miR-673-5p and miR-3544 (located between the Gtl2 and anti-Rtl1 coding regions), showed NF-κB and MAZ-binding sites and their levels were reduced by corticosterone. Furthermore, miR-1188-3p and miR-3703p -which are located within the Rian coding region of the Dlk1-Dio3 gene locus- displayed a negative regulation by corticosterone and showed MAZ-binding sites. We have found that several miRNAs that are very close to each other within the miR-379-410 cluster share common motifs, a fact in agreement with the overlapping of miRNA putative promoters. For instance, miR-758-5p and miR-376c-3p seem to be regulated by NF-κB and MAZ motifs, but in opposite directions. In addition, miR-409a-3p and miR-412-3p -which are in close proximity to the miR-379-410 cluster- have NF-κB response elements, followed by MAZ and SMAD3–4, but their levels changed in opposing directions. Interestingly, the hepatic induction of the mammalian miR-379-410 genomic cluster has been described as a key component of glucocorticoid/GR-driven metabolic dysfunction [70]. Additionally, in hepatocytes, the glucocorticoid-induced variation in miRNA expression of the miR-379/410 cluster seems to require GR dimerization [70]. GR ChIP-Seq assays have identified GR-binding sites around the miR-379 gene, and DNase-sensitivity experiments have proposed that the binding of GR occurs through the cooperative action of transcriptional activators, such as p300, CEBPB and the HNF4a [70]. Interestingly, we identified a common motif for HNF4a in miRNAs from the miR-379-410 genomic cluster. Finally, miRNA genes hosted in cluster within the X chromosome also displayed some recurrent motifs. For instance, miRNA-532 -which reduced its levels with the hormone- derived from the miR-532–502 cluster in the X chromosome and exhibited a MAZ binding site. On the other hand, miR-351-1 and miR-450a -which are closely located within the miR-322-503 cluster- exhibited MAZ, GR and NF-κB binding motifs; although their levels changed in opposite directions with hormone treatment. Altogether, these observations suggest that factors other than the presence or combination of regulatory motifs -such as genomic/epigenomic contexts and posttranscriptional processes- may impact miRNA levels in the hippocampus following corticosterone administration.

Remarkably, we did not detect a conserved or novel motif within the 1 kb upstream sequence of the start of the miRNA precursors. In females, only five (miR-204, miR-376c, miR-764, miR-743b and miR-450) of the 17 corticosterone-responsive miRNAs did not retrieve a novel or known motif. In the male data set, we could not identify response elements in only five (miR-133a, miR-455, miR-532, miR-223, and miR-3580) of the 27 corticosterone-responsive miRNAs. Therefore, it is plausible that promoters in these miRNA genes may be located upstream of the analyzed sequence.

According to the previous analyses, we want to point out that the GR and NF-κB share a large proportion of genomic regulatory elements and co-regulate many genes in a mutual antagonistic or synergistic manner; evidences that may explain either an up or down-regulation by corticosterone, by acting through direct and indirect mechanisms [14]. Additionally, MAZ is highly expressed in the mouse brain, especially in the hippocampus, reinforcing the idea of other authors that this factor could correspond to a novel element in the crosstalk with the MR/GR in a neuronal context [71]. The sex-biased effect of corticosterone could be explained, at least in part, considering that promoters are themselves targets of sex-biased regulation, as suggested by others [72]. Alternatively, the sex-bias in mature miRNA expression levels may occur without variation in the expression pattern of the precursor forms (pri- and pre-miRNAs), supporting the existence of regulatory mechanisms acting on mature miRNA stability [73]. Additionally, active degradation of mature miRNAs has been identified as another relevant mechanism for miRNA homeostasis [73].

Target prediction to identify cellular processes regulated by differentially expressed miRNAs

Experimentally validated mRNA targets obtained from TarBase and miRTarBase databases (Chou et al. 2018; Karagkouni et al. 2018) were used to identify the regulatory network of differentially expressed miRNAs in the male and female rat hippocampus. Considering that information is not available for all the miRNAs annotated, for the successive analyses we used a set of miRNA-mRNA interaction pairs of 12 miRNAs for the ADX-female rats and 16 miRNAs for the ADX-male rats. We found 588 mRNAs regulated by 12 differentially expressed miRNAs in ADX-female rats and 2,508 mRNAs regulated by 16 miRNAs in the case of ADX-male rats (Figure 6). Interestingly, an overlap of ~17% (437) of the total number of mRNAs, was found between male and female datasets. Given the larger set of regulated mRNAs in the male dataset, a more complex miRNA-mRNA interaction network was observed in males, compared to females (Figure 6). In addition, several highly regulated mRNA nodes, with a potentially pivotal role were recognized in males. We highlighted the components of Wnt pathway including frizzleds receptors (FZD3, FZD5), the co-receptor (LRP6) and β-catenin (CTNNB1). Interestingly, this pathway regulates adult hippocampal neurogenesis [74]. Also, we identified components of BDNF-TrkB signaling (NTRK2, MAPK1, PIK3CA, PLCG1),which modulates the hippocampal synaptic plasticity in the hippocampus [75, 76]. In females, we highlighted MAPKs (MAP3K1, MAP3K2), the receptor of Wnt (FZD3) and the kinase BRaf (BRAF) which is required for neuronal differentiation in the postnatal hippocampus [77].

Figure 6. miRNA-mRNA Interaction Networks.

Interaction networks of differentially expressed miRNAs in both (A) ADX-female and (B) ADX-male rats after corticosterone administration. The blue nodes represent the negatively regulated miRNAs, while the red nodes represent the positively regulated miRNAs. The nodes in gray represent the experimentally validated targeted mRNAs extracted from TarBase v8.0 and miRTarBase v7.0 [86, 87]. The solid gray line indicates the regulatory associations between miRNAs and mRNAs.

Interestingly, in females, we observed that miR-186-5p, miR-376c-3p, miR-450a-5p, and miR-743b-3p are part of an interactive network; but only miR-186-5p showed a known regulatory domain (HNF4a). Moreover, from the interaction information obtained from males, several miRNAs with a high number of regulatory interactions were identified, such as miR-204-5p, miR-324-3p, miR-532-3p, and miR-450a-5p, which had a common MAZ regulatory sequence. miR-223-3p and miR-133a-5p also showed a broad regulatory network, but these miRNAs did not display a common regulation as the previously mentioned miRNAs. Interestingly, miR-223-3p, which is up-regulated with corticosterone treatment, showed a higher density of targets, but did not share a common motif in its putative promoter with the other miRNAs. A recent study has described that miR-223-3p is up-regulated in response to inflammation (model of autoimmune encephalomyelitis) to trigger neuroprotection, by inducing a gene expression profile that protects neurons from glutamate [78].

Further analysis performed with the Enrichr platform on the same experimentally validated mRNAs (Chen et al. 2013; Kuleshov et al. 2016) revealed several enriched pathways (p-value < 0.05) in both males and females groups, related to cytoskeleton dynamics, cellular growth, survival, proliferation, and differentiation (Figure 7), with some common terms in males and females, such as the FoxO, MAPK, and mTOR signaling pathways, proteoglycans in cancer, and ubiquitin-mediated proteolysis. Furthermore, we detected axonal guidance and PI3K-Akt pathways in females; and apoptosis and neurotrophin signaling in males. Several investigations have reported that depletion of endogenous GCs induces cell death in the hippocampus [79–82]. Andrés et al. [82] and Greiner et al. [81] demonstrated that ADX surgery changes the expression of apoptotic-related genes, such as Bcl-2, and that corticosterone treatment prevents these effects. In contrast, in females, we detected an enrichment in the PI3K-Akt signaling pathway. In vitro assays with a hippocampal cell line have shown that cortisol stimulates the proliferation and decreases neurogenesis markers, such as MAP2 and DCX. These effects were suppressed by treatment with spironolactone, an MR antagonist [12]. This is interesting because miR26b, up-regulated by corticosterone treatment, represses the translation of its host gene Ctdsp2, which encodes for a CTD small phosphatase that regulates the expression of genes involved in neurogenesis [83].

Figure 7. Putative functional role of differentially expressed miRNAs.

(A) Comparison by divergent bar plot of functional enrichment analysis of target mRNAs regulated by differentially expressed miRNAs in female (purple) and male (teal) rats. The top 15 KEGG terms are indicated, based on the combined Fisher exact test score (horizontal axis). (B) Interaction networks of differentially expressed miRNAs in female and male rats after 6 h of corticosterone administration. Colored chords represent the regulatory interaction between each miRNA and respective target mRNAs. Experimentally validated target mRNAs were extracted from TarBase v8.0 and miRTarBase v7.0 databases (Chou et al. 2018; Karagkouni et al. 2018).

Taken altogether, the evaluation of the sex differences in miRNA expression triggered by treatment of ADX rats with corticosterone may provide significant clues to unveil the sex-biased effect of this important stress hormone in hippocampal physiology and pathology. Because miRNA plays such an important role in the regulation of gene expression, sex-specific regulation of miRNA in response to corticosterone level variations may subsequently lead to sex-specific control of gene expression and the subsequent effect on cellular functions in the hippocampus.

MATERIALS AND METHODS

Animals

Adult male and female Sprague-Dawley rats maintained at the Faculty of Chemical and Pharmaceutical Sciences, Universidad de Chile, were used in this study. Efforts were made to reduce both the number of animals used and their suffering. The rats were handled in compliance with the National Institutes of Health Guide for Care and Use of Laboratory Animals (NIH Publication, 8th Edition, 2011), under an experimental protocol approved by the Ethical Committee of the Faculty of Chemical and Pharmaceutical Sciences, Universidad de Chile, and the Science and Technology National Commission (CONICYT). Two-month-old Sprague Dawley male and female rats were given free access to water and pelleted food and were maintained at 22 ºC with 55% of humidity and photoperiod cycle of 12 h (lights on from 0700 to 1900h).

Adrenalectomy

Male (n=6) and female (n=6) rats underwent bilateral adrenalectomy (ADX) surgery under isofluorane (1.5% v/v air) anesthesia, as we previously described [68]. The animals were kept with a 0.9% NaCl drinking solution to compensate for the sodium deficit induced by the ADX. Seven days after surgery, half of the animals were injected with saline and the other half, with corticosterone previously dissolved in ethanol and then diluted in saline (10 mg/kg, i.p.). Considering that GCs promote genomic actions, we chose to euthanize the animals 6 h post injection as a period in which it is highly probable to observe the effects of the hormone. At the time of sacrifice, trunk blood was collected for the determination of serum corticosterone levels. Hippocampi were rapidly dissected and flash-frozen in liquid nitrogen and kept at −80 °C until processing for RNA isolation.

Serum Corticosterone Levels

Hormone level determination was carried out using Corticosterone ELISA Kit (Enzo, New York, USA; Cat. ADI-900-097), according to the kit’s instructions.

Isolation and preparation of RNA for Microarray Analysis

Purified RNAs from one hippocampus were isolated as we have previously described [40]. In brief, frozen hippocampus tissue was disrupted using an UltraTurrax homogenizer in QIAzol Lysis Reagent (Qiagen, California, USA). Enriched RNA fractions of less than 200 nucleotides were then obtained by using the RNeasy Mini Kit (Qiagen, California, USA), followed by the RNeasy MinElute Cleanup Kit (Qiagen, California, USA). RNA concentration and purity were determined at OD260/280 and samples with an absorbance ratio between 1.8 and 2.0 were chosen. RNA integrity was evaluated by nondenaturing agarose gel electrophoresis.

After labeling the RNAs with Affymetrix FlashTag™Biotin HSR kit, samples were hybridized on Affymetrix GeneChip™ Version 3 miRNA arrays (680 mature and 486 pre-miRNA rat probe sets), according to the Affymetrix hybridization protocols. Array slides were then stained with streptavidin/phycoerythrin and after washing, they were scanned using the Affymetrix GCS 3000 7G and GeneChip® Operating Software (AGCC; Version 3.2). The microarray data will be available in the Gene Expression Omnibus repository at the National Center for Biotechnology Information.

Bioinformatic microarray analysis

Microarray data analysis was performed using two bioinformatic tools: (i) Oligo v1.48.0 and Limma v3.40.0 R packages [84, 85], and (ii) Transcriptome Analysis Console (TAC, Affymetrix). For Oligo/Limma, background correction and normalization with the Robust Multichip Average method (RMA) were performed between all samples. Additionally. for TAC, the Detection Above Background (DABG) algorithm was used. Log2 values of normalized probe intensities were then computed. In order to assess the quality of the microarray data and to identify abnormal arrays, several graphic representations (box plots, density plots, MA plots, and pseudo images) were generated for both raw and normalized data. The identification of differentially expressed miRNAs in each experimental group was achieved using an empirical Bayesian approach. Probes with adjusted p-value < 0.05 and with a |fold change| ≥ 1.5 were considered as differentially expressed between experimental groups and controls.

Identification of mRNA-miRNA interactions based on validated data

To evaluate the interaction network of differentially expressed miRNAs, we developed a customized pipeline in Bash and R languages to extract and analyze experimentally validated miRNA-mRNA interactions from TarBase v8.0 and miRTarBase v7.0 databases [86, 87]. In order to obtain a list of genes with reported expression in rat hippocampus, a filter step was performed on the interaction information, using the Expression Atlas from the European Bioinformatics Institute (Petryszak et al. 2016). The filtered target mRNAs were further analyzed with the functional enrichment platform Enrichr [88, 89] to assess their putative influence in several biological processes and pathways relevant to hippocampal function. Only the terms statistically significant according to Fisher exact test (p-value < 0.05) were considered for further analysis. Circular visualization was produced using R package circlize [90].

Search of common elements in the promoter of co-expressed miRNA genes

We looked for possible control elements within the upstream region of genes corresponding to differentially expressed miRNAs. The sequence of the putative promoter considered an extension of 1 kb upstream of the transcription start site, and was analyzed using the MEME software suite (Bailey et al. 2009), which includes both the MEME algorithm for novel motif discovery and the FIMO algorithm for known motif scanning, in particular for MR and GR binding. Additionally, and considering the crosstalk between GR with other transcription factors, either in a mutual antagonistic or synergistic manner, we also included NF-κB in the binding profile analysis. Furthermore, in order to determine putative protein partners for each novel motif, the output of the MEME algorithm was subsequently compared against the JASPAR database with the TomTom tool [91].

Data Availability Disclosure Statement

The authors declare that all supporting data and method descriptions are available within the article, or from the corresponding author, upon reasonable request.