Orexin signaling regulates both the hippocampal clock and the circadian oscillation of Alzheimer’s disease-risk genes

All mice used in this paper were housed at 22 ± 2 °C, with 60 ± 5% humidity, and maintained with a LD 12:12 photoperiod (12 h light, 12 h dark, lights on at 07:00). Mice were fed a normal diet and provided water ad libitum. Clock^delta19/+ mice49 and homozygous mPer2::luciferase knock-in mice (mPer2^luc/luc)50 were purchased from the Jackson Laboratory. Clock^delta19/+ mice were crossed to mPer2^luc/luc reporter mice. From heterozygous offspring, we created double homozygous Clock^{delta19/delta19}; mPer2^luc/luc mice. In this study, APP/PS1dE9 transgenic mice were used to evaluate the mechanism through which the circadian clock contributes to AD51. These mice express a chimeric mouse/human APP (Mo/HuAPP695swe) and a mutant human PSEN1 (PS1-dE9). APP/PS1dE9 mice were also crossed with mPer2^luc/luc reporter mice to create APP/PS1dE9; mPer2^luc/luc mice. mPer2^luc/luc, Clock^{delta19/delta19}; mPer2^luc/luc, and APP/PS1dE9; mPer2^luc/luc mice were generated for hippocampal dissection and real-time recoding of the hippocampal oscillation. All experiments for this study were carried out with 2–4 month old male mice, except as otherwise noted. Wild-type (WT) mice were maintained in a LD 12:12 photoperiod condition with free access to food and water for 2 weeks before being kept in complete darkness (DD) for an additional 48 h. WT Mice (n = 3) were sacrificed every 4 h throughout the course of one circadian cycle (both under LD and DD condition). The hippocampus were dissected quickly from brains. Hypothalamus samples were collected every 6 h for one circadian cycle from young (age 4 months, n = 3–5) or aged (aged 12–15 months, n = 3–5) WT and APP/PS1dE9 transgenic mice brains. Animal experiments were performed in accordance with the NIBS institutional regulations, after approval by the Institutional Animal Care and Use Committee (IACUC).

mPer2^luc/luc, Clock^{delta19/delta19}; mPer2^luc/luc, and APP/PS1dE9; mPer2^luc/luc mice were anesthetized with 2,2,2-Tribromoethanol (Sigma) and sacrificed at ZT12-15 to reveal the bioluminescence rhythm in the hippocampus; these protocols were performed as previously described49,50. The brain was rapidly removed from the mouse and placed in ice-cold Hanks’ balanced salt solution (HBSS) (Thermo Fisher, pH = 7.2–7.4). The brain was then cut into slices of 220 μm thickness with a vibrating-blade microtome (VT1000S, Leica Microsystems). The slices were maintained in ice-cold HBSS during this procedure until the point when explants were placed into the experimental medium for luciferase recording. The hippocampus was carefully and quickly isolated from the brain slices using scalpels and was then explanted onto a culture membrane (Milli-CM 0.4 μm, EMD Millipore) on top of the liquid surface of a 35 mm Petri dish (Corning) and sealed with a greased 40 mm coverslip. Samples were then cultured with 1.3 mL of HEPES-buffered explant medium supplemented with 1 μM luciferin (Promega) and B-27 supplements (Thermo Fisher). The explants were incubated at 36 °C, and bioluminescence was monitored for one minute in each 10-minute interval using a dish-type luminometer (Actimetrics). The assessment of circadian periods and phases were performed as described in previous reports49,50,51.

HEK293 cells were grown in regular DMEM supplemented with 10% FBS (Hyclone, GE Healthcare Life Sciences) and antibiotics at 37 °C, 5% CO₂. For transfection, rapidly growing cells were trypsinized and re-suspended in DMEM containing 10% FBS lacking antibiotics at a 0.1 × 10⁶ cells/ml concentration. We next added 50 μl of transfection reagent mixture (0.5 μl/well Lipofectamine 2000 in Opti-MEM; Thermo Fisher) to wells containing pre-spotted plasmids. We incubated the wells at room temperature for 20 min and subsequently added 100 μl of cells (0.1 × 10⁵ cells/well). Approximately 6 h after transfection, we replaced this medium with 150 μl of pre-warmed fresh DMEM containing 10% FBS and antibiotics and allowed the cells to grow for an additional 24–30 h. 36 h post-transfection, we replaced this medium with 150 μl of HEPES-buffered explant medium supplemented with luciferin (1 μM) and B-27 supplements; the plates were sealed with an optically clear film. We next loaded these plates into a 36 °C incubator and recorded bioluminescence expression with an Infinite^® 200 PRO series microplate reader (Tecan, Thermo Fisher).

The hippocampal slices were treated with final concentrations of 10, 50, and 100 nM orexin A (Abcam) dissolved in DMSO. Forskolin (Sigma) was dissolved in DMSO and the hippocampal slices were treated with a final concentration of 10 μM. Orexin B (Genscript) was dissolved in DMSO; the hippocampal slices were treated with a final concentration of 500 nM. EMPA (Sigma), a high-affinity, reversible, and selective Hcrtr2 antagonist, was dissolved in DMSO52; the hippocampal slices were treated with a final concentration of 10 μM. All compounds, drugs and peptides were titrated in the explant medium to the final concentration and then the prepared medium was added to the 35-mm Petri dish with the slices on the insert. To express E4BP4, the coding sequence of the E4BP4 gene (NM_001289999.1↗) was amplified from cDNA and subcloned into the pcDNA3.1 plasmid (Thermo Fisher). The human 1.0-kb BACE1-promoter (NC_000011.10↗) and the 1.4-kb BACE2-promoter (NC_000021.9↗) were amplified from DNA extracted from HEK293 cells; these amplification products were cloned as pGL3-basic plasmid reporter constructs (Promega) and named, respectively, P(BACE1)-luc and P(BACE2)-luc. The primers used for the PCR amplification of target sequences are detailed in Supplementary Table 4.

Total RNA was extracted from the hippocampus and hypothalamus using Trizol reagent (Thermo Fisher). A 500 ng aliquot of total RNA was reverse transcribed into cDNA using PrimeScript™ RT Master Mix (Takara) and then analyzed with SYBR GREEN qPCR mix (Kapa Biosystems) using a CFX96 instrument (Bio-Rad). The relative quantification of expression levels was performed using a previously-described ^ΔΔCT calculation method53. Beta-Actin was used as a reference gene. The specific primer pairs used for the analysis of the core clock genes and the AD-risk genes were designed using Primer3 (Supplementary Tables 1–3).

OriginPro 2016 software (OriginLab) was used for statistical analyses. Period change data were analyzed with one-way/two-way ANOVA followed by Tukey’s HSD test. To validate whether or not the AD-risk genes displayed circadian oscillations under the LD and DD conditions in the hippocampus, we measured 24 h oscillations in transcript abundance using the JTK_CYCLE algorithm; we set a 5% false discovery rate for detection1,54. We have here reported the results as means with the standard error of the mean (mean ± s.e.m.), and have used P < 0.05 as the criterion for evaluating the inferential statistical significance of differences.

Mice were anesthetized with 2,2,2-Tribromoethanol and fixed by perfusion of 4% paraformaldehyde. Whole brains were dissected and post-fixated with 4% paraformaldehyde for an additional 2 h before dehydration in a 30% sucrose solution overnight at room temperature. Thoroughly dehydrated whole brains were frozen in powdered dry ice. Brain tissues were sliced using a cryostat instrument (CM3050S, Leica) to a thickness of 45 μm. The sections were stained with anti-orexin A (1:125, Santa Cruz) overnight at 4 °C, then incubated with Alexa Fluor 594 labeled anti-goat IgG (1:500, Thermo Fisher) and DAPI (Sigma, 1 mg/mL, 1:1000) for 1 h at room temperature in darkness. Images were obtained via confocal microscopy (Zeiss confocal LSM800). To visualize fibrillary amyloid plaques, sections were stained with Thioflavin T55,56. Images were obtained with a virtual scanning system for microscopy slides (Olympus VS120).

We first evaluated whether there was a self-sustained circadian oscillator in the hippocampus. The hippocampus of mPer2^luc/luc mice, which we were able to isolate completely from other brain regions (Fig. 1A), clearly maintains oscillations in ex vivo culture (Fig. 1B–D). We found that the oscillations could be synchronized by changing the growth medium and by treatment with forskolin, which activates adenylyl cyclase and then triggers cyclic AMP signaling (Fig. 1C,D). After one week in ex vivo culture, the hippocampal oscillation is dampened and becomes de-synchronized (Fig. 1C,D). This oscillation could be synchronized by treatment with 10 μm forskolin. We also found that Clock malfunction results in defects in maintaining the hippocampal oscillator. Clock^{delta19/delta19}; mPer2^luc/luc mice carry a mutation in the Clock gene that results in a dominant-negative protein that cannot activate transcription57. We dissected the hippocampus and recorded the oscillation of the Clock^{delta19/delta19}; mPer2^luc/luc mice. We found that deficiency in Clock led to an attenuation of circadian rhythms in the hippocampus (Fig. 1B).

To confirm our finding that a self-sustained circadian oscillator exists in the hippocampus, qPCR analysis was conducted to measure the expression levels of core clock genes in the hippocampus in the DD condition. We used the JTK_CYCLE algorithm to characterize cycling variables, including period, phase, and amplitude, for the core clock genes. The expression of 6 out of 7 core clock genes tested here (with the exception being Cry1) exhibited circadian rhythms (Fig. 2B and Table 1), indicating that an intact oscillator exists in the hippocampus. The JTK_CYCLE algorithm results revealed an exact 24 h period for the expression of core clock genes in the hippocampus (Table 1). The mRNA acrophase of Clock and Bmal1, as predicted by the JTK_CYCLE algorithm, occurred at CT0 and CT6, respectively; the JTK_CYCLE algorithm-predicted mRNA acrophase of Per1, Per2, Dbp, and Cry2 occurred, respectively, at CT10, CT14, CT0, and CT10 (Table 1).

Intact circadian oscillators can be regulated by input signals such as hormones, metabolites, and neuropeptides7. We evaluated whether oscillation in the hippocampus can be regulated by such signals. Orexins are essential regulators of sleep/wake cycles and are typically thought to be involved in the circadian clock, hippocampus-dependent social memory, and Aβ pathology. We therefore hypothesized that orexin signaling might be able to regulate the hippocampal oscillator. As expected, we found that treatment with orexin A or B (OR-A and OR-B, for short) did indeed regulate the hippocampal clock, leading to a shortened period of hippocampal oscillation in ex vivo culture (Fig. 3A). The period of explants was 23.68 ± 0.12 h with 100 nM OR-A treatment and was 24.4 ± 0.09 h with control DMSO treatment (Fig. 3B, Supplemental Fig. 1A–C). Treatment with 500 nM OR-B also shortened the period of hippocampal oscillation ex vivo (Fig. 3C). The period of explants was 23.57 ± 0.08 h with 500 nM orexin B treatment and was 24.54 ± 0.03 h with control DMSO treatment. Treatment with 10 μM forskolin did not alter the period of the hippocampal oscillator (Fig. 3C,D).

qPCR analyses of orexin receptor genes revealed that Hcrtr2 levels are higher than Hcrtr1 levels in the hippocampus, suggesting that Hcrtr2 is likely the primary receptor for orexins in this brain region (Fig. 3G). To block the binding of orexins to the receptors, we performed co-treatment experiments with hippocampus slices that tested combinations of EMPA and OR-A or OR-B. In the presence of EMPA, both OR-A and OR-B completely failed to shorten the period of the hippocampal clock (Fig. 3E,F,H; Supplemental Fig. 1D,E). All of these results demonstrate that both the OR-A peptide and the OR-B peptide can function, in redundant roles, in shortening the hippocampal clock.

The key AD-risk gene Psen1 has previously been shown to have rhythmic expression in the liver, which is thought to possess a functional circadian oscillator1. Since AD occurs in the hippocampus and the cerebral cortex of the brain, not in peripheral organs, we sought to measure whether, or how, the clock regulates the expression of AD-risk genes in the hippocampus. We therefore monitored the expression of a group of 77 candidate genes (AD RT2 Profiler PCR Array, QIAGEN)58,59 that are known to be involved in most of the cellular AD-related signaling pathways (Supplementary Table 1) to determine whether these genes merit classification as CCGs. The JTK_CYCLE algorithm was used to identify and characterize the cycling variables of the qPCR dataset for these 77 candidate genes. We found that nearly half of the AD-risk genes exhibited rhythmic expression patterns. We found that 39 and 34 of the candidate genes had rhythmic expression patterns in the LD condition and the DD condition, respectively (Fig. 4A). To narrow down the size of our target gene list, we concatenated the genes that were classified by JTK_CYCLE to be CCGs under both LD and DD conditions. We found that at least 12 out of 77 AD-risk genes are under circadian control in the hippocampus (Fig. 4A; Table 2). Gnb5, Sncβ, Casp3, ApoE, Cdc2, Bace1, Gng1, Psen2, Gsk3α, Apba1, Bace2, and Prkcδ are the genes most likely to be under clock regulation, as the expression of all of these genes was found to be rhythmic under both the LD and DD conditions in the hippocampus, and the statistical confidence levels for all of their predicted JTK_CYCLE algorithm cycling variables were high (Fig. 4B; Table 2).

The expression of both Bace1 and Bace2 was classified as rhythmic in the hippocampus based on our qPCR analysis. We found that the promoter regions of BACE1 and BACE2 contain putative cis-elements that are potential binding sites for core clock genes. There is a putative D-box in the promoter region of BACE1 (Fig. 5A). There are four potential E-boxes in the promoter region of BACE2 (Fig. 5A). We sought to confirm, in vitro, whether the BACE1 and BACE2 promoters could be regulated by core clock transcription activators such as CLOCK:BMAL1 and ROREs or be regulated by a repressor like E4BP4. Our co-transfection experiments showed that P(BACE2)-Luc expression was activated by the CLOCK:BMAL1, and we found that the activation of BACE2 depended strictly on the amount of CLOCK:BMAL1 present (Fig. 5B, right panel). P(BACE1)-Luc expression was inhibited by the E4BP4 repressor (Fig. 5B, left panel). Further qPCR analysis found that Bace2 decreased the expression at ZT5 and ZT17 in Clock mutant mice, as did the canonical E-box gene (Fig. 5C). All these data confirm that Bace1 and Bace2 are potential CCGs and could be directly regulated by the hippocampal clock.

The deposition of the Aβ peptide is a major pathological aspect of AD. We confirmed this phenotype in our AD-model mice (APP/PS1dE9). Thioflavin T (TFT) is a benzothiazole dye that exhibits enhanced fluorescence upon binding to amyloid fibrils60. Our TFT staining results indicated that there were massive amyloid fibrils in the cortex and hippocampus regions of APP/PS1dE9 mice compared with WT controls (Fig. 6A). We next used qPCR analysis to examine the expression patterns of core clock genes and AD-risk genes, and evaluated how AD progression affected the expression profiles of these genes. As predicted, aged AD mice experience changes in the expression of risk genes, including Bace2 and ApoE. Interestingly, a subset of the tested core clock genes of aged APP/PS1dE9 mice, including Per1, Per2, Cry1, and Dbp, had altered expression patterns compared with WT controls. Dbp expression was elevated at ZT5 and ZT17 in the hippocampus of aged APP/PS1dE9 mice, but was decreased at ZT23 compared with WT control; however, the expression of Per1, Per2, and Cry1 was decreased at ZT11 but elevated at ZT23 (Fig. 6D). Furthermore, the expression of Bace2 and ApoE, both of which have been verified by us and by other researchers to be E-box genes that contain non-canonical E-box motifs in their promoter regions61, was also altered in the hippocampus of aged APP/PS1dE9 mice at ZT5, ZT11, and ZT17 (Fig. 6C). Non-E-box-containing core clock genes, including Bmal1 and Clock, showed only minor, if any, changes in their mRNA levels (Supplemental Fig. 3A,C). There were no differences between aged APP/PS1dE9 mice compared with WT controls in the expression of Nr1d1 and Nr1d2 (Supplemental Fig. 3B), two additional E-box genes, possibly because these genes, which are representative of a secondary loop in the circadian network62, are not susceptible to the AD pathology.

Finally, we wondered whether the expression of orexins in the brain is governed by the clock. Intriguingly, the expression of the orexin precursor gene was found to be rhythmic in the hypothalamus (Fig. 6B and Supplemental Fig. 2A). Further immunofluorescence experiments confirmed that orexin A exhibits the same diurnal expression pattern at the protein level: In the lateral hypothalamus area, the immunostaining signal for the orexin A peptide is higher at ZT17 (Supplemental Fig. 2B), a finding consistent with our qPCR analysis of the orexin precursor gene (Supplemental Fig. 2A). Interestingly, the expression pattern of the orexin precursor gene is altered in the hypothalamus in APP/PS1dE9 mice. The transcription of this gene is higher at ZT5 in APP/PS1dE9 mice than in WT control mice (Fig. 6B). At other time points, the expression of this gene did not differ significantly between the two mouse genotypes (Fig. 6B).