Circadian regulation of hippocampal function is disrupted with corticosteroid treatment

Significance Circadian oscillations of glucocorticoids are a fundamental characteristic of adrenal hormone secretion in all mammals. These rhythms can be disrupted by chronic stress or synthetic glucocorticoid therapy, resulting in side effects including impaired memory, mood, and sleep. Here, we show that 5-d treatment with methylprednisolone induces prolonged activation of glucocorticoid receptors and consequent binding at glucocorticoid regulatory elements in the core clock gene Period 1, in the hippocampus. Furthermore, disruption of circadian hippocampal gene regulation was evident, along with loss of circadian variation in hippocampal long-term potentiation (LTP) and impaired hippocampal memory. Together, these data demonstrate dysregulated hippocampal function during long-acting glucocorticoid treatment and further support the importance of devising improved regimens of glucocorticoid therapy that recapitulate endogenous glucocorticoid rhythmicity.

Disrupted circadian activity is associated with many neuropsychiatric disorders. A major coordinator of circadian biological systems is adrenal glucocorticoid secretion which exhibits a pronounced preawakening peak that regulates metabolic, immune, and cardiovascular processes, as well as mood and cognitive function. Loss of this circadian rhythm during corticosteroid therapy is often associated with memory impairment. Surprisingly, the mechanisms that underlie this deficit are not understood. In this study, in rats, we report that circadian regulation of the hippocampal transcriptome integrates crucial functional networks that link corticosteroid-inducible gene regulation to synaptic plasticity processes via an intrahippocampal circadian transcriptional clock. Further, these circadian hippocampal functions were significantly impacted by corticosteroid treatment delivered in a 5-d oral dosing treatment protocol. Rhythmic expression of the hippocampal transcriptome, as well as the circadian regulation of synaptic plasticity, was misaligned with the natural light/dark circadian-entraining cues, resulting in memory impairment in hippocampal-dependent behavior. These findings provide mechanistic insights into how the transcriptional clock machinery within the hippocampus is influenced by corticosteroid exposure, leading to adverse effects on critical hippocampal functions, as well as identifying a molecular basis for memory deficits in patients treated with long-acting synthetic corticosteroids.

glucocorticoids | hippocampus | circadian rhythms | memory | methylprednisolone
Circadian rhythms regulate many physiological, biological, and behavioral processes in mammals (1)(2)(3)(4). In addition to the primary mammalian clock in the hypothalamic suprachiasmatic nucleus (SCN) (5), which is largely regulated by the light/dark period, clock gene transcriptional rhythms are observed in other regions of the brain, including the hippocampus (6)(7)(8), which remain synchronized to the SCN via a combination of neural and humoral signals (4,5,9) The endogenous corticosteroid hormone (predominantly cortisol in man, and corticosterone in rat) is under circadian control via SCN regulation of hypothalamic-pituitary-adrenal (HPA) axis function. Disruption to this, for example in Cushing's disease and during commonly prescribed steroid therapy for inflammatory diseases such as rheumatoid arthritis or asthma, results in elevated cortisol or exogenous steroid levels, respectively, throughout the day-and has been associated with depression, anxiety, and psychosis (10)(11)(12). Recent studies have demonstrated the importance of normal physiological oscillations of steroid availability for optimal cognitive and emotional responses (13), suggesting that the timing of corticosteroid secretion and therefore activity of glucocorticoid receptors (GRs) at specific times of day may be essential to neuropsychiatric health. How synthetic steroid use causes adverse neuropsychiatric and cognitive outcomes, however, remains unknown. In this study, we sought to mimic oral corticosteroid dosing in a rat model, in order to identify how the treatment may influence such adverse outcomes. Methylprednisolone (MPL), which is a long-acting (14), frequently prescribed synthetic corticosteroid (15) associated with circadian disturbances, psychiatric disorders, and memory impairment in humans, (16)(17)(18) was provided for 5 d ad libitum in drinking water at a concentration of 1 mg/mL; a treatment regime that will be referred to throughout the text as "MPL treatment. " We first sought to assess the effects of MPL treatment on the rats' daily activity profiles, behavior known to be under central clock control, according to zeitgeber time (ZT) (SI Appendix, Fig. S1A). Period 1 (Per1), a light-entrainable regulator of rhythmic function (19), constitutes a key component of the circadian clock machinery and can be targeted by the GR (Nr3c1) via a hypersensitive glucocorticoid response element (GRE) (20). Per1 is highly enriched in the SCN, where we found similarly robust circadian fluctuation in its OPEN ACCESS messenger RNA (mRNA) expression in both control and MPL-treated rats ( Fig. 1A and SI Appendix, Fig. S1 B and C). This finding indicates that SCN Per1 is resistant to transcriptional disruption by corticosteroids, presumably because the SCN is devoid of GR expression (21). Consistent with this, SCN-regulated physiological rhythms such as locomotor activity and core body temperature were unaffected by MPL treatment (Fig. 1 B-E), remaining primarily under the control of SCN transcriptional clock oscillations synchronized to the 12-h light/12-h dark cycle.
In contrast to the SCN, GRs are highly enriched in the hippocampus (21,22), where their activity displays strong circadian variation in activity (23), in line with the circadian rise and fall in adrenal corticosteroid secretion (Fig. 1F), and demonstrated by significantly increased nuclear pSer211 GR during the active (ZT 14) compared to the inactive phase (ZT 2) (Fig. 1G), along with significantly increased GR binding to the Per1 distal GRE (Fig. 1H), and subsequent Per1 transcription (Fig. 1I), during the active (ZT 14) but not inactive phase (ZT 2). RNA sequencing (RNAseq) data from rat hippocampi every 4 h (from ZT 2-22) identified 485 differentially expressed genes (DEGs) across time ( Fig. 2A and Dataset S1). Of these DEGs, 56 genes (10%) were found to be directly regulated by corticosteroids in an independent experiment where adrenalectomized (ADX) rats were infused with an acute infusion of 0.75 mg/mL corticosterone at ZT 2 (SI Appendix, Fig. S2 A-C and Dataset S2).
In nonstress conditions, endogenous corticosterone secretion reaches maximal levels near ZT 10, to modulate the awakening response just prior to the start of the active phase (24), when many other biological processes are initiated (25). Accordingly, it is quite well appreciated that during the early period of the active phase, learning and memory processes are more effective, with performance declining as the day progresses (26)(27)(28)(29)(30). Therefore, we used gene ontology (GO) analysis to identify the biological processes and molecular functions that were enriched in the DEG analysis at ZT 10, relative to the circadian nadir ZT 2 when endogenous corticosteroid secretion is lowest (Figs. 1F and 2 B-D). At ZT 10, DEGs included Per1, Per2, Per3, Bhlhe40, CamkII, Grin2a, and Grin2c, known mediators of circadian rhythms and entrainment, as well as modulators of hippocampal synaptic function (31). Per1, Grin2a, and Grin2b are examples of genes that were directly regulated by acute corticosteroid treatment in ADX rats (SI Appendix, Fig. S2 and Dataset S2), Indeed, GR regulatory sites were identified within or proximal to Per1, Per2, Bhlhe40, CamkII, Grin2a, and Grin2b (32). To prepare for the active period, many biological processes have a priming phase, suggesting that genes controlling synaptic function may be regulated by local clock mechanisms in line with the endogenous corticosteroid peak. We clustered the DEGs at ZT 10, by known association (Fig. 2C), discovering functional interactions between core clock gene products-Per1, Per3, Dbp, Bhlhe40, and Aanat-and the known synaptic function genes Grin2 and CaMKII, Example locomotor activity in a CTL (blue) and a MPL (green)-treated rat. (C) CTL and MPL-treated rats maintained rhythmic locomotor activity (F 1,12 = 1801, P < 0.0001: CTL; P < 0.0001, MPL; P < 0.0001). (D) Example core body temperature in a CTL (blue) and a MPL (green)-treated rat. Similar to activity data, (E) MPL treatment did not influence core body temperature rhythm (F 1,12 = 1,964, P < 0.0001: CTL; P < 0.0001, MPL; P < 0.0001). Two-way ANOVA with Sidak's post-hoc test. (F) MPL in drinking water was sufficient to suppress corticosterone release in rats. F 5,81 = 16.98, P < 0.0001, for time, and F 1,81 = 160.6, P < 0.0001, for treatment, by two-way ANOVA with Sidak's post-hoc test. (G) Western blot showing pSer211 GR and HDAC1 loading control in hippocampal nuclear extract samples prepared from MPL-treated and CTL rats. Associated graph displaying relative quantification of densitometry data. suggesting they are poised to mediate clock-coupled behavioral rhythms to influence learning and memory processes (33,34).
Grin2a and Grin2b encode for GluN2A and GluN2B, respectively. Endogenous corticosteroids have been reported to rapidly regulate GluN2B-NMDAR membrane trafficking through nongenomic mineralocorticoid (MR) signaling (35). However, corticosteroids bind to GRs in a circadian pattern, whereas MRs are near maximally occupied at both the circadian nadir and peak due to their higher corticosteroid affinity (36). Using a network connectivity map, we identified interactions between enriched genes of corticosteroid-mediated expression, clock gene expression, and synaptic function at ZT 10 ( Fig. 2D). Identification of genes under circadian and corticosteroid control from independent RNAseq analyses ( Fig. 2 and SI Appendix, Fig. S2: Per1, Klf9, Klf15, Sgk1, Zfp189, and Tsc22d3) supports corticosteroid-mediated regulation of clock gene expression (37,38). Importantly, peak corticosteroid secretion occurs in preparation for increased neurocognitive activity, which, in concert with peak Per1 clock gene expression in the hippocampus, facilitates action on synaptic plasticity target genes CamkII and Grin2 (Fig. 2D).
Indeed, corticosteroid action on memory performance is well established (26,39). However, seminal studies focusing on the role of corticosteroids have concentrated largely on the response to acute changes in corticosteroids in models of stress (39,40). Furthermore, although time of day is known to influence memory function (26,27), different models and tests have identified conflicting outcomes (28)(29)(30)(41)(42)(43). For example, mice and rats learn maze navigation more effectively during the active phase, yet tone-cued fear conditioning in mice is more readily formed during the inactive phase (41). Moreover, evidence that tasks with a high cognitive demand can themselves serve as zeitgebers suggests that the interaction with time of day is stimulus specific (44). Here, we have identified a molecular basis for the improved performance often observed at the start of the active phase. Given that the colocalization between CaMKII and GluN2B-containing NMDARs at the synapse is critical for long-term potentiation (LTP), a cellular correlate of long-term memory, we hypothesized that this process is under circadian control, and potentially driven by GR-mediated transcriptional rhythms. To test this hypothesis, ex vivo hippocampal slices were prepared from rats culled during either the early active or inactive period to allow time for gene transcription occurring at the respective peak or trough of endogenous corticosteroid secretion to translate to functional proteins at the synapse. Using a high-frequency stimulation protocol, the method used previously to show LTP is modulated by the addition of corticosterone (45), we found that LTP was robustly induced in hippocampal slices prepared from control rats culled during the early active period but not the early inactive period ( Given the GO pathway predictions for associations between glucocorticoids and the synaptic processes under circadian control here, we next sought to interrogate whether exogenous glucocorticoid treatment would have an impact on these functional outputs of the hippocampus. Indeed, corticosteroid-mediated memory improvement has previously been demonstrated in rats, where an acute corticosterone injection during the circadian nadir enhanced memory consolidation in a hippocampal-dependent task (46). Chronic corticosteroid exposure, on the contrary, has been reported to impair memory in both aversive and nonaversive tasks, and at different times of day (47).
We therefore next assessed the effect of 5-d 1 mg/mL MPL treatment (ad libitum in drinking water) on circadian transcriptome regulation and synaptic plasticity in the rat hippocampus. Considering that GR binding dynamics of MPL are distinct from that of endogenous corticosteroid (48), and that MPL treatment protocol suppresses adrenal corticosterone over the full circadian cycle (Fig. 1F) while inducing prolonged hippocampal GR activation throughout the full circadian cycle ( Fig. 1 G-I), we hypothesized that MPL treatment may also continuously influence glucocorticoid target genes throughout the day. To test this, we sequenced the RNA from rat hippocampi every 4 h (ZT 2-22), on the final day of MPL-treatment. RNAseq analysis identified 512 DEGs across time ( Fig. 3A and Dataset S3). Here, we discovered that many genes known to be important for memory processing in the hippocampus were dysregulated with MPL treatment, an important finding considering that circadian disturbances and memory impairments are common among individuals with major depression (49,50) or that are receiving steroid therapy (51).
Following this, we sought to assess whether the same functional pathways identified in control rat hippocampus were influenced by MPL treatment. GO analysis of DEGs from ZT 10 to ZT 2 comparison ( Fig. 3 B-E) predicted distinct target pathways compared to GO analysis of the equivalent comparison in control data (Fig. 3B). It was also notable that additional genes were differentially expressed at ZT 10 in the MPL-treated rats ( Fig. 3 C and D and Dataset S4). Interestingly, no circadian-entraining pathways were identified with GO analysis, suggesting a loss of hippocampal rhythm with MPL treatment. Synaptic plasticity pathways were identified in the MPL treatment ZT 10 vs. ZT 2 comparison, although identified DEGs were again distinct from those identified in the ZT 10 vs. ZT 2 comparison in control rats (Dataset S5). With differential pathway activity between control and MPL-treated groups, we quantified the expression of circadian rhythm and entrainment genes, as well as synaptic plasticity target genes at ZT 10 in control and MPL groups (Fig. 3E). Following the detection of a dissociation between clock regulating and synaptic plasticity genes at ZT 10, we sought to determine the peak amplitude of all DEGs across the day ( Fig. 3F and Dataset S6) (52). We found that most DEGs in control exhibited peak expression levels between ZT 5 and 7 (Fig. 3F), whereas following steroid treatment, most DEGs peaked between ZT 22 and 1 (Fig. 3F). After identifying a shift in peak gene expression, we sought to identify if this affected the expression patterns of target genes; mapping select expression profiles across time of known clock-regulating, corticosteroid target, and synaptic plasticity-associated genes (Fig. 3G). The striking pattern differences, and expression profiles of genes regulating clock and synaptic plasticity processes following MPL treatment, led us to investigate a potential functional output of this disruption.
In controls, LTP was robust during the early active period, as was the frequency of mPSCs (Fig. 2 E-I). Using the same high frequency stimulation (HFS) protocol, but this time in rats that received 5 d of MPL treatment, LTP could not be induced in either the early active or inactive period (Fig. 3 H-J). Remarkably, MPL treatment increased the frequency, but not amplitude, of mEPSCs recorded during the early inactive period, an inversion of the pattern observed in controls. This was not evident in the frequency of mIPSC events, instead eliminating circadian variation, suggesting that the MPL treatment disrupted the excitatory/ inhibitory (E-I) mPSC balance ( Fig. 3 K and L and SI Appendix, Fig. S3 D and E). Interestingly, alterations to the E-I balance in the forebrain have been implicated in several psychiatric disorders including Alzheimer's disease (53,54). Additionally, early-life stress accelerates developmental shifts in the E-I balance (55), emphasizing the role of corticosteroid signaling in potentially mediating the regulation of this phenomenon.
CaMKII is essential for modulating LTP and spatial memory formation (56), and following the striking disruption to the dynamic expression of both N-methyl-D-aspartate (NMDA) receptor and CaMKII subunits, we show that hippocampal LTP and mPSCs were disrupted following MPL treatment. Again, using the same HFS protocol that was sensitive to regulation by time of day in control rats ( Fig. 2 E-I), we found that MPL treatment abolished the formation of LTP during the early active period, when it could be readily elicited in controls (Fig. 4 A-C). Furthermore, in controls, LTP induction could be similarly blocked by application of the GluN2B-selective NMDAR antagonist, Ro-25-6981 ( Fig. 4 D-F). This suggests that ablation of the circadian regulation of NMDARs and CaMKII results in disruption of downstream plasticity processes preventing their association at a time critical to nonaversive learning. Consistent with our data demonstrating corticosteroid sensitivity, NMDARs are critical in mediating the dendritic atrophy and synaptic dysfunction caused by chronic stress, (57,58) with GluN2B appearing to play a pivotal role (59).
Following these robust effects of steroid treatment on the hippocampal transcriptome and synaptic physiology, we investigated whether MPL-induced disruption of NMDA receptor function had a consequent impact on hippocampal-dependent memory formation. After 5 d of MPL treatment, using a novel object location task, where training and testing were carried out during the early active period, short-term memory retrieval (1-h postacquisition phase) was unaffected (Fig. 4H). However, memory retrieval was impaired Data are mean ± SEM. *P < 0.05, by unpaired t test (Fig. 3 I and K), or Mann-Whitney U test Fig. 3J).

Discussion
Corticosteroids are widely used in clinical medicine to relieve signs and symptoms of many inflammatory and autoimmune disorders (62,63). However, in addition to this, their use is often reported with cognitive and psychiatric symptoms-inducing a range of psychiatric adverse effects including depression, mania, and cognitive and memory impairment. These relationships have been known for decades, yet the mechanisms underlying these disturbances and their clinical management have been poorly described.
In this reverse translational study, we demonstrate that 5 d of corticosteroid treatment disrupts the rhythmic activity of hippocampal function through GRs, induces NMDAR-dependent synaptic dysfunction, and subsequently impairs memory. This model is supported by several lines of evidence: 1) Corticosteroids do not influence the master clock (SCN), or light/dark entraining behaviors due to a lack of GR expression in the SCN; 2) the hippocampus is rich in GRs, which in the presence of corticosteroids, bind to GREs on the essential clock gene Per1 to mediate hippocampal activity in competition with central clock-mediated control; 3) synthetic corticosteroid treatment eliminates the circadian variation in circulating corticosteroids, and consequently the circadian variation in the activity of hippocampal GRs to influence clock gene expression and destabilize clock-entrained NMDAR/CamkII complexes; and 4) synthetic corticosteroid treatment blocks LTP, a cellular correlate of long-term memory, by disrupting NMDAR-dependent processes to impair long-term, but not short-term, hippocampal-dependent memory.
There are two regulators in normal chronophysiology. On the one hand, the SCN is the master regulator of circadian timing, entraining behaviors such as feeding, sleep, and memory processes to ~24 h (64). On the other hand, extra-SCN oscillators, such as those in the hippocampus and other forebrain regions, are entrained by SCN-dependent neural and humoral inputs (65). Pertinent to this study, one of the more powerful humoral signals is from adrenal corticosteroids (66), which exhibit a robust circadian secretory rhythm that peaks prior to wakening (2) in preparation for the neurocognitive activities that ensue. Synthetic long-acting corticosteroids, such as MPL, are therefore poised to mediate influence over these endogenous systems in the presence of the master clock.
Indeed, corticosteroids exert vast and variable effects on the body's physiology, both peripherally and centrally (2,67,68). Consistent with previous studies, we found that MPL treatment suppressed appetite, and subsequently decreased body weight (SI Appendix, Fig. S1 D-F). However, interestingly, we identified distinct and specific actions of MPL on the brain. For instance, the SCN remained protected from MPL-mediated disruption, as previously reported with dexamethasone (69). However, the GR-rich hippocampus was sensitive to corticosteroid interference (70). The identification of GREs upstream of the Per1 gene, and its corticosteroid-induced transcription, underpins a mechanism whereby GR-expressing extra-SCN oscillators can be disturbed by corticosteroids, and may account for the loss of circadian rhythms often seen in hypercortisolemic states of Cushing's disease and syndrome (71). Notably, our RNAseq dataset which encompassed whole hippocampus found that Per1 mRNA expression was up-regulated throughout the day with MPL treatment. Interestingly, Per1 (−/−) mice also exhibit spatial memory deficits. These targeted manipulations of Per1 are opposing (i.e., overexpression vs. knockout) yet led to a similar outcome, suggesting that any disruption to the rhythmic expression of Per1 may be detrimental to spatial memory performance.
It is well established that corticosteroids are secreted from the adrenal gland in a circadian pattern with underlying ultradian pulses, and common to several neuropsychiatric disorders are HPA axis hyperactivity or steroid therapy (72)(73)(74). We found that 5 d of MPL treatment induced a prolonged increase in activated nuclear GR levels throughout the full circadian cycle and disrupted the circadian expression of NMDAR and CAMKII subunits. Both NMDAR and CAMKII are known to be essential mediators of neural plasticity (75), a process that is causally linked with the successful encoding of mnemonic information (76). Further to this, CAMKII is essential for coupling time of day to behavioral rhythms (33) and improved memory performance (77).
Consistent with our findings, other synthetic corticosteroids, particularly dexamethasone, can impair memory (78). However, translational studies often assess memory performance during the early inactive period, when endogenous circulating corticosteroids are low, therefore not time relevant to biological rhythms (79) or peak cognitive performance of the participants. Similarly, many rodent studies assess both hippocampal synaptic plasticity processes and memory performance during the animals' inactive period, leading to the somewhat incomplete conclusion that effective memory consolidation requires a strong stress association. These tests use aversive or fear conditioning-based tasks such as inescapable foot shock, which induces a robust stress response in the animal, resulting in similar elevated glucocorticoid levels to those circulating at the onset of the active phase. Consistent with our interpretation that elevated glucocorticoids during the inactive phase can influence the molecular physiology of the brain, acute administration of corticosteroids has been demonstrated to enhance memory consolidation during nonaversive memory tasks performed during rats' inactive phase. (80,81). The specificity of corticosteroid, duration of action, as well as the timing and type of experimental testing, may contribute to differences in corticosteroid-mediated memory formation in the literature. Importantly in this study, we have identified a mechanism involving dysregulation of the hippocampal circadian clock, in which neurocognitive decline is mediated by a commonly prescribed synthetic corticosteroid. A limitation of this work is that a rescue strategy is yet to be established, and this will require future investigation. Moreover, how various corticosteroids (endogenous and synthetic) contribute to such differential outcomes on hippocampal function warrants further investigation. This study also used male rats, and therefore does not consider the action of corticosteroids on circadian processes and memory performance in females. Indeed, it is well established that in hippocampal CA1, LTP is dependent on locally synthesized estrogen (82), and that blocking estradiol synthesis disrupts LTP in females only (82,83). Therefore, understanding how endogenous and exogenous corticosteroids may interact with and disrupt these processes warrants investigation.
Nevertheless, our ex vivo physiology data suggest that time of day is a potent modulator of neuronal activity and plasticity processes in CA1 neurons. Crucially, we show that this circadian regulation is modified by long-acting corticosteroid treatment, disrupting the dynamic control of miniature postsynaptic currents (mPSCs) and preventing LTP induction. These data support the notion that circadian and ultradian corticosteroid fluctuations play an important role in maintaining the rhythmic activity of the hippocampus via dynamic transcription of target genes. This has major physiological consequences considering the importance of mPSCs in the maintenance of synaptic connections and dendritic spines (84), as long-term alterations to these dynamics can result in critical changes to hippocampal circuitry and function, as demonstrated by the prevention of hippocampal-dependent memory formation (Fig. 4).
In summary, a large proportion of patients prescribed with corticosteroids report cognitive decline and memory impairment (16). Our data reveal a GR-mediated pathway that underlies the circadian regulation of hippocampal-dependent memory formation which is vulnerable to long-acting corticosteroid treatment. Currently, there are no clinical guidelines for treating corticosteroid-induced adverse effects. It is, perhaps, quite surprising that the brain-specific effects of such treatment has had relatively little scientific investigation given the broad application, widespread clinical use, and adverse psychiatric outcomes (15,85,86). Advancing our knowledge of corticosteroid-mediated regulation of hippocampal function will take us a step closer to understanding the mechanisms underpinning several prevalent major mental illnesses and suggest tailoring treatment regimens for prevention and intervention of corticosteroid (endogenous and synthetic)-induced disorders.

Methods
Adult male Lister hooded rats (250 to 350 g, 9 to 11 wk) (Envigo, UK) were used in all experimental procedures. Male rats were used in this study to limit estrous cycle influence on corticosterone release (87). Animals were maintained in standard housing conditions under a 12:12 gradual light/dark cycle in soundattenuating rooms. A red light was used during the night phase to allow researchers to manipulate the animals. During the night phase, no external light could enter the room when opening the door. Food and water were available ad libitum. Throughout all experimental procedures, the same researcher took care of the animals to limit stress-induced effects and was blinded to treatment groups. All procedures were carried out in accordance with the UK Animals (Scientific Procedures) Act 1986 under PPL 30/3114 and PIL I04092F5F.

Adrenalectomy and Jugular Vein Cannulation.
For studies identifying glucocorticoid-specific control of hippocampal gene transcription (SI Appendix, Fig. S4), rats received balanced anesthesia using veterinary isoflurane (Merial Animal Health, Woking, UK) prior to bilateral adrenalectomy and cannulation of the right jugular vein for infusion of corticosterone. Rats recovered for 5 d postsurgery on 15 µg/mL corticosterone in 0.9% saline drinking solution to maintain isotonic levels. This solution was replaced 12 h prior to experiments with 0.9% saline to ensure washout of circulating corticosterone.

MPL Treatment.
For all studies assessing the role of MPL, rats (randomly assigned to treatment) were given MPL sodium succinate (Solu-Medrol; Pharmacia Ltd, Sandwich, UK) 1 mg/mL ad libitum in drinking water for 5 d (prolonged treatment protocol), prior to the start of all further procedures. Concentrations of MPL in drinking water (1 mg/mL) were optimized in previous experiments as the minimum dose required to reproducibly induce hippocampal GR activation and suppress endogenous corticosterone (20 mg/day). Body weight was monitored throughout all procedures. Ad libitum access to MPL in drinking water was designed to limit any external stimuli that may act as a zeitgeber (88) and also represents greater than 1% of the human population that are prescribed oral glucocorticoid treatment (15). At this dose, no difference in fluid consumption was observed across days, and circulating endogenous corticosteroids (measured at the end of 5-d treatment) were suppressed (SI Appendix, Fig. S3).
Radioimmunoassay. Using an automated gamma counter (PerkinElmer, US), a corticosterone radioimmunoassay measured endogenous levels in blood samples collected immediately following rapid decapitation. An 11-point standard curve of known corticosterone concentrations was prepared in B-buffer (25 mM tri-sodium citrate, 50 mM sodium dihydrogen orthophosphate, 1 mg/mL bovine serum albumin: pH3). Plasma obtained was diluted in triplicate at a ratio of either 1:10 or 1:50 in B-buffer. A specific corticosterone antibody (kindly provided by G. Makara, Institute of Experimental Medicine, Budapest, Hungary) was diluted at a ratio of 1:50 in B-buffer and 50 μL added to 100 μL standards, unknown samples, and Quality Control (QC20 and QC100) tubes. Tracer (Izotop, Institute of Isotopes, Hungary) was diluted in B-buffer to give total counts of 3,750 cpm in 50 μL and added to all tubes (50 μL/tube). Tubes were incubated overnight at 4 °C. Charcoal suspension (5 g charcoal added to 0.5 g dextran T70 dissolved in 1L B-buffer) was prepared and 500 μL was added to all tubes and briefly vortexed. Blocks were centrifuged at 4,000 rpm at 4 °C and the resulting supernatant was aspirated off. Unknown samples were determined from interpolation of the standard curve.
In Situ Hybridization Histochemistry. Whole brains were cryosectioned into coronal 12 μm sections, and thaw mounted on gelatin/chrome alum-coated slides. The location of the SCN was determined according to coordinates in the rat brain atlas (89). 35  CAG TGG TAC -3′) was used. The in situ hybridization protocol has been previously described in detail (90). Briefly, sections were fixed in 4 % (w/v) formaldehyde for 5 min and incubated in saline containing 0.25 % (v/v) acetic anhydride and 0.1 M triethanolamine for 10 min. Sections were then dehydrated in ethanol, delipidated in chloroform, and partially rehydrated. Hybridization with a total count of 1 × 10 6 cpm was performed overnight at 37 °C in 45 μL hybridization buffer under Nescofilm (Bando Kagaku, Osaka, Japan). After hybridization, sections were washed 4 times with SSC (150 mM NaCl and 15 mM sodium citrate) for 1 h at 65 °C and for an additional hour with two changes of SSC at room temperature. Hybridized sections were exposed for autoradiography (Hyperfilm, Amersham, Bucks, UK) for 1 wk. The amount of bound probe was analyzed in comparison to 14 C-labeled standards (Amersham, Bucks, UK) using image analysis software (NIH Image 1.6.2, W. Rasband, NIH, Bethesda, MD, USA). The obtained results were represented in arbitrary units setting the mean optical density (OD) obtained from sham-operated rats.
PCR primers (forward: 5′-CCAAGGCTGAGTGCATGTC -3′; reverse: 5′-GCGGC CAGCGCACTA -3′) were designed to amplify across a previously described GRE (20,91) in the rat Period 1 gene promoter. Samples were amplified with Sybr Green master mix (Applied Biosystems) in accordance with the manufacturer's instructions. GR binding for each sample was calculated relative to 1% input chromatin taken from each individual sample, using the %Input method, described in (ThermoFisher Scientific, http://bit.ly/ChIPAnalysisTFS). RNA and qPCR. Animals were anesthetized with isoflurane in the animal facility, and eight were killed every 4 h (six time points). The brain was quickly extracted, and the hippocampus was removed and rapidly frozen in liquid nitrogen. The time between decapitation and sample freezing was <1 min to limit RNA degradation. The collection of the hippocampal tissue per time point was <20 min (10 min either side of the hour mark).
Total RNA was extracted from frozen whole hippocampi using the miRNeasy total RNA extraction kit protocol (Qiagen, US) following the manufacturer's guide and included a DNase digestion step. Samples were stored at −80 °C. All samples were assessed for RNA quality and quantity using a Nanodrop (ThermoScientific, US). Samples sent for RNAseq were further assessed for RNA integrity on the 2200 TapeStation  Whole-Genome RNAseq and Analyses. RNAseq of hippocampal tissue was carried out using TruSeq Stranded Total RNA kit and protocols (Illumina, US). A 1 μg aliquot total RNA from each hippocampus was prepared for RNAseq with 48 samples in total. First, bioanalyzer traces were carried out on all samples for RNA quantity and quality check. RIN scores above 8.0 were deemed high enough quality to take through to RNAseq. Following library preparation, the samples were run on a HiSeq 2500 machine (Illumina, US). Eight samples were run in each lane, in a paired-end sequencing run, generating approximately 300 million reads across the eight samples.
High-throughput RNAseq raw data were uploaded to Galaxy (92), an openaccess portal for next-generation sequencing analysis. For the circadian sequencing experiments, an N of 4 was used in each group for a total of 48 whole hippocampi samples, and three lanes of data were collected for each sample.
For the acute corticosterone administration experiments, an N of 4 was used in each group for a total of 8 whole hippocampi samples. Sequencing files were aligned to the Rattus norvegicus (Rn6) genome-generating individual Binary Alignment Map (BAM) files. These BAM files from each of the three lanes were merged. The merged BAM files were analyzed for gene expression differences using Tophat2, Cufflinks, and CuffDiff analyses aligning to the Rn6 genome (93). The parameters included geometric library organization and pooled dispersion estimation, and the false discovery rate was set at 0.05. Minimum alignment count-10, multi-read correct, bias correction and Cufflinks effective length correction were also included. Differential expression was assessed across time using CuffDiff version 15 (2020-06-16), with all time points compared to ZT 2. Any gene that failed DE analysis because of low gene expression [based on minimum alignment count (94,95)] at any time point was removed from further analyses. Therefore, in the circadian RNAseq analysis-vehicle-treated rats, DE analysis was carried out on 13,939 genes. In MPL-treated rats, DE analysis was carried out on 16,269 genes. In the acute corticosterone administration to ADX rats, DE analysis was carried out on 16,266 genes. Differential gene expression was calculated from fragments per kilobase per million mapped reads values. Multiple hypothesis correction was carried out using the Benjamini-Hochberg test. Data deemed statistically significant, FDR < 0.05.
The Database for Annotation, Visualization and Integrated Discovery (96) was used to determine GO and pathways. GO categories tested included biological processes and pathways from the Kyoto Encyclopedia of Genes and Genomes (KEGG).

RNAseq Rhythmicity Analysis.
For CTL and MPL-treated groups, we focused on 24 h periodicity using harmonic regression on the log2 transformed signals, as previously described (52), from gene expression data sampled every 4 h. Each hour on the clock face refers to the total number of genes that peak at that time of day.
Brain Slice Preparation. The brain was rapidly removed at either ZT0030 (30 to 60 min following light change-inactive phase recording) or ZT1230 (30 to 60 min following light change-active phase recording) (SI Appendix, Fig. S5). These times were chosen based upon our RNAseq data. Peak Per1 mRNA expression in controls was between ZT 10 and ZT 14. Nadir Per1 mRNA expression in controls was between ZT 22 and ZT 2. A minimum of 2 h after peak and trough expression was chosen to allow for functional protein availability. Slices placed in ice-cold slicing solution contained (in mM): 52.5 NaCl, 2.5 KCl, 25 NaHCO 3 , 1.25 NaH 2 PO 4 , 5 MgCl 2 , 25 D-Glucose, 100 sucrose, 2 CaCl 2 , 0.1 kynurenic acid, bubbled with 95% O 2 /5% CO 2 . Parasagittal slices (400 μm) were cut using a vibrating blade microtome (Leica VT1000 S) while maintained in slicing solution. The hippocampus was then isolated and placed in a holding chamber containing artificial cerebrospinal fluid (aCSF) (containing (in mM) 124 NaCl, 3 KCl, 26 NaHCO 3 , 1.4 NaH 2 PO 4 , 1 MgSO4, 10 D-Glucose, 2 CaCl 2 ) where it was incubated at 32 °C for 30 min, followed by RT for at least 30 min (for whole-cell recordings), or kept at room temperature for at least 1 h (for field electrophysiology) before being transferred to the recording chamber. Extracellular Electrophysiology. Hippocampal slices were placed in a submersion style recording chamber maintained at 30 °C and continuously perfused with oxygenated aCSF at a flow rate of ~2 mL/min. Field excitatory postsynaptic potentials were evoked at 0.033 Hz by placing bipolar stimulating electrodes on the Schaffer collateral fibers with the recording electrode positioned in CA1 stratum radiatum. Recording electrodes were prepared by pulling borosilicate capillary tubes with a P-97 Flaming/Brown micropipette puller (Sutter Instrument Co) to a tip resistance of 2 to 6 MΩ and were then back-filled with aCSF. Signals were amplified using an AxoClamp 2B amplifier (Molecular Devices), digitized using a BNC-2110 (National Instruments) board, and 50/60 Hz noise eliminated by a Hum Bug (Quest Scientific). Data were acquired and analyzed using WinLTP software (97). Signals were averaged over a period of 2 min and a stable baseline recording of 30 min was acquired before LTP induction by delivery of 10 Hz stimulation for 90 s.
Whole-Cell Electrophysiology. Whole-cell recordings were taken from pyramidal neurons in the CA1 cell layer. Borosilicate pipettes (2 to 6 MΩ) were filled with an internal solution containing (in mM): 8 NaCl, 130 CsMeSO 4 , 10 HEPES, 0.5 EGTA, 4 MgATP, 0.3 NaGTP, and 5 QX-314. Recordings were accepted for analysis with an uncompensated series resistance of <2.5 times the pipette resistance. Recordings were not corrected for series resistance due to the small current amplitudes. During mEPSC and mIPSC recordings, 100 µM D-AP5 was added to the perfused aCSF to block NMDA receptor-mediated currents. mEPSCs were recorded at a membrane potential of −70 mV and for mIPSCs, membrane potential was held at 0 mV. Recordings were amplified using an AxoClamp 700B (Molecular Devices) for whole-cell voltage-clamp recordings. Data were acquired using WinLTP software at a sampling rate of 10 KHz, and filtered at 6 KHz, before being analyzed offline using ClampFit 9.2. mEPSCs and mIPSCs were identified when the rise time was faster than the decay time and had a peak amplitude >6 mV.
Object Location Memory (OLM) Task. For all memory testing, different rats were used at all time points and treatment. Rats were transferred to a soundattenuating behavior room in low light (40 to 50 lx on arena floor) at ZT11.
Rats were left to habituate to the room for at least an hour, prior to starting behavioral experiments. Similarly to our electrophysiology studies, this time was chosen based upon our RNAseq data. Peak Per1 mRNA expression in controls was between ZT 10 and ZT 14. A minimum of 2 h after peak expression was chosen to allow for functional protein availability. Rats were handled for 1 wk in the behavior room prior to experimental start, followed by habituation to the arena without stimuli for 10 min daily for 5 d. Training and testing occurred in an opentop arena (50 × 90 × 100 cm) made of wood. The walls inside the arena were black and floor covered in sawdust. An overhead camera recorded behavior for analysis. Exploration was scored when the rat head orientated toward the object and came within 1 cm of the object. The objects were constructed from Duplo blocks, which were too heavy for the animals to displace. During training, rats were exposed to two identical objects and allowed to explore for 4 min. These objects were placed in the far side of the arena, 10 cm away from the walls, to allow full access around the objects. During the retention test (1 h for short-term memory, 6 h for intermediate memory, or 24 h for long-term memory), rats were allowed to explore for 3 min. During testing, one of the objects was moved to a new location. The position of the object was counterbalanced between rats. Total exploration time was recorded, and preference for the novel item was expressed as a discrimination index. Activity and Temperature Recording. All rats were individually housed for telemetry recordings for technical reasons. Rats were implanted intraperitoneally with telemeters (PTD 4000 E-mitter, Starr Life Sciences Corp, US). Following recovery (>3 d), cages were placed upon receivers (ER-4000 receiver, Starr Life Sciences Corp, US). Locomotor activity and core body temperature data were collected every 10 min for five consecutive days. Data collected were analyzed with R CRAN package cosinor to measure bathyphase, acrophase, and circadian period under a 24-h period.
Statistical Analyses. Results are presented as mean ± SEM. All statistical analyses were performed using GraphPad Prism (v 9.1, GraphPad software Inc., US), with parametric and nonparametric tests used where appropriate. Details of specific tests are provided in the figure legends. Statistical significance was set at P < 0.05.