Desipramine restores the alterations in circadian entrainment induced by prenatal exposure to glucocorticoids

All experiments were conducted in agreement with the European legislation (2010/63/EU Directive) and Swedish national regulation following approval by the local Animal Ethics Committee (Stockholms Norra djurförsöksetiska nämnd).

DEX (Sigma-Aldrich, Stockholm, Sweden) was dissolved in sterile saline to a concentration of 5 µg/mL. Pregnant C57BI/6 dams (N = 15 females/group) (Charles River, SCANBUR Research, Sollentuna, Sweden) received a subcutaneous injection of 0.05 mg/kg/day DEX or equivalent volume of vehicle (10 mL/kg b.w. physiological saline), from gestational day 14 until delivery. The litters were culled to 4 pups per litter at post-natal day (PND) 3 and weaned at PND21. At the time of weaning, the animals were distributed in groups of 4–5 mice per cage so that each mouse originated in a different litter, and were tagged with subcutaneous radio frequency identification (RFID) transponders (Trovan™ Unique T-100A, Trovan, UK) under brief 4% isoflurane anesthesia. The RFID tags allowed unambiguous identification of individual animals throughout the experiment. Only the male pups were used for subsequent experiments, since female offspring did not show similar behavioral alterations (see also ref. ²⁷). The timeline of experiments is shown in Supplementary Fig. S1.

Mice aged 5 months (mo) were treated with DMI (Sigma-Aldrich, Stockholm, Sweden) dissolved in drinking water (10 mg/kg/day) for 28 days before recording spontaneous activity, and the treatment was continued throughout the recording period. We did not observe changes in fluid intake after adding DMI to drinking water (not shown).

All experiments involving circadian rhythms were performed in an isolated room (22 ± 1 °C; 50 ± 5% relative humidity), with 12:12 h LD cycle (light intensity 200 lx) and free access to food and water. The circadian zeitgeber (German: “time-giver”) time (ZT) 0 corresponds to the onset of the light phase. The interaction with human experimenters was limited to changing the cage and replenishing food and water, which occurred at random times throughout the experiment. The mice were kept in conditions similar to the environment in the holding facility (i.e., group-housed in the homecage as assigned at weaning), except the top of the cage was kept clear to ensure even illumination of the cage bottom.

We recorded the spontaneous activity of mice (N = 7–8/group) using the TraffiCage™ system (NewBehavior™, Zürich, Switzerland), as described elsewhere²⁷. The system consists of an array of antennas placed under the cage. The antennas read the RFID tags and provide the location of each animal with a time resolution of 20 ms. A “visit” was defined as the time interval during which an animal is detected constantly by the same antenna, and was used as activity count. Each recording session included two cages monitored simultaneously, one control and one test cage, placed randomly on either TraffiCage™ plates. After reaching stable circadian entrainment to the 12:12 h LD cycle, the light was turned off 6 h earlier (advance the onset of dark phase) to induce a phase advance in the circadian rhythms of activity. The activity was monitored for at least four LD cycles before, and at least five LD cycles after the phase shift.

The time series of activity counts were exported as ASCII files and analyzed using custom algorithm implementations in Matlab™ 9 (The MathWorks™, Natick, MD, USA). To identify the onset of active phase, the activity was binned in 5 min non-overlapping epochs, and smoothed by weighted average with a sliding Gauss window (2 h width). The epochs with activity above the individual’s detrended average were considered “active epochs”. The active phase was defined as a sequence of active epochs contiguous or separated by gaps no larger than 1 h. The onset of active phase was defined as the time corresponding to the beginning of the active phase (relative to subjective time, ZT; ZT0 = light on). We defined spontaneous activity as entrained to the LD cycle if the onset of the active phase occurred within 1 h from the onset of the dark phase. The individual time to re-entrain was estimated as the first day when spontaneous activity was entrained to the shifted LD cycle.

The periodicity of spontaneous activity was analyzed using the Lomb-Scargle periodogram³³ in non-overlapping contiguous 24 h epochs using a publicly available implementation of the algorithm in Matlab™ (https://se.mathworks.com/matlabcentral/fileexchange/20004-lomb–lomb-scargle–periodogram↗). The false alarm probability for periodogram peaks was estimated as described by Ruf³³ and later developed by Baluev³⁴. Intrinsic periodicity can be divided in circadian (i.e., regular fluctuations in spontaneous activity with a period of ~24 h) and ultradian (i.e., quasi-regular fluctuations in spontaneous activity with period shorter than 24 h, typically below 12 h). Both circadian and ultradian rhythms are modulated by the SCN^35–38, but circadian periodicity is typically entrained by environmental stimuli (e.g., light, feeding schedule), while ultradian periodicity is driven rather by instrinsic factors such as cycles in behavioral arousal and episodic glucocorticoid release³⁹.

We analyzed the variability of spontaneous activity over 96 h of continuous recording at baseline and after the phase advance by estimating the intra-daily variability (IV), and by means of detrended fluctuation analysis (DFA). IV was calculated as the ratio between the average squared differences between consecutive timebins (pooled activity over 5 min) and the global variance over the entire epoch^40,41. IV typically ranges between 0 and 2; low IV indicates smooth fluctuations in activity levels; high IV indicates fragmented activity patterns characterized by abrupt transitions between episodes of intense activity and inactive periods^40,41. For DFA, we ran linear regression analysis for residual variance in time series after detrending against the time scale used for detrending. The intervals for detrending ranged from 20 min to 21.3 h (i.e., in ultradian range) in equally spaced exponential increments. The correlation coefficient of a linear regression in double-logarithmic plot translates into a scaling exponent, and describes the long-term autocorrelation patterns embedded in the time series⁴². Complex time series with fractal-like patterned irregularity yield values between 0.5 and 1: positively correlated fluctuations yield a scaling exponent around 0.5; values around 1 suggest strong underlying regularity (such as diurnal rhythm); values approaching 1.5 indicate unbounded fluctuations resembling random walk (i.e., Brownian noise)^42,43. The scaling exponent in young, healthy rodents and humans is around 0.8^36,37.

We evaluated depression-like behavior in mice aged 6 or 12 mo (N = 8–12 per treatment) by means of forced-swim test (FST), as described elsewhere^27,44. Briefly, the animals were individually placed for 6 min in plastic cylinders (24 cm height, 20 cm diameter) filled with water (23.5 °C) to a depth of 16 cm, and videotaped. The footage was analyzed offline by one investigator who was blind to the treatment and exposure conditions. Immobility was defined as continuous passive floating lasting for at least 2 s. After testing, the mice were dried with paper towels and returned to the homecage to recover. The total immobility time over the last 5 min of recording was used for subsequent analyses.

The mice were killed by an overdose of anesthetic (sodium pentobarbital, 150 mg/kg) and perfused transcardially with ice-cold buffered saline followed by 4% paraformaldehyde. The brains were removed and postfixed with 4% paraformaldehyde overnight, then cryoprotected with 15% sucrose. The brains were cut along the midline and processed for immunohistochemistry as follows: the right hemisphere (injected with retroviral particles, see below) was embedded in TopVision™ Low Melting point Agarose (5% Thermo Scientific) and cut in 70 μm thick coronal sections with a vibratome (Leica VT1000s) for confocal microscopy imaging. The left hemisphere was quickly frozen on dry ice, and 20 μm thick sections were cut with a cryostat. For stereological counting of DCX + cells, we collected 2–3 slices every 200 μm throughout the hippocampal formation. Between the sections saved for immunohistochemistry we used the trim function of the cryostat to cut 140 μm thick sections for microdissection of hippocampal samples for RNA extraction.

The mice were killed by an overdose of anesthetic (sodium pentobarbital, 150 mg/kg) and perfused transcardially with ice-cold buffered saline followed by 4% paraformaldehyde. The brains were removed and postfixed with 4% paraformaldehyde overnight, then cryoprotected with 15% sucrose. For gene expression analysis, the SCN was obtained by dissecting a small piece of tissue (~0.5 × 0.5 × 0.5 mm³) from the anterior hypothalamus (located on the ventral extent of the brain, dorsal and posterior to the optic chiasm, in the anterior wall of the third ventricle; atlas coordinates bregma −0.84−46 mm, dorsal 0–0.5 mm; lateral 0.5 mm on either side of the midline). The accuracy of SCN sampling was verified by the relative mRNA expression levels for V1a and V1b AVP receptors (V1b expression hardly detectable; see ref. ⁴⁵). In our samples, the mRNA expression of V1b AVP receptors was hardly detectable, indicating negligible contamination with surrounding hypothalamic tissue. The hippocampus was sampled by microdissection from frozen sections (see Supplementary Materials and Methods). RNA extraction was performed using FFPE RNA Purification Kit (Norgen Biotek, Montreal, Canada) according to the manufacturer instruction. The concentration of RNA was measured by NanoDrop 1000 spectrophotometer (Thermo Scientific, Wilmington, DE, USA), and RNA degradation was assessed using an RNA 6000 Nano Kit (Agilent Technologies Inc., Santa Clara, CA, USA) running on an Agilent 2100 Bioanalyzer system (RIN range: 5.1–6.2). The RNA (1 µg template/sample) was reverse-transcribed into cDNA using 0.5 µg of oligo-dT primer according to the instruction of Maxima First Strand cDNA Synthesis kit (Thermo Fisher). Amplification reactions were performed on a QuantStudio 5 thermocycler (Applied Biosystems) using SYBRGreen PCR MasterMix (Thermo Fisher) under the following conditions: initial denaturation 50 °C for 2 min followed by 95 °C for 10 min; followed by 40 amplification cycles (95 °C for 15 s, then annealing temperature (60 °C) for 1 min). The specificity of the qRT-PCR reactions was evaluated by including a dissociation stage to the melting curve analysis. Primer sequences, annealing temperatures, and length of amplification product are available in Supplementary Table. The mRNA expression was normalized against the housekeeping gene glyceraldehyde 3-phosphate dehydrogenase (GAPDH), and the relative regulation was estimated using the 2^−ΔΔCt method.

Inactive GR is located in the cytosol in molecular complexes built around a chaperone protein (heat shock protein 90, Hsp90). Upon binding to GC, GR is released from Hsp90 and translocates to the nucleus, where it forms homodimers that bind to the DNA and regulate gene expression⁴⁶. GR complexes in the hippocampus were assessed using in situ proximity ligation assay (PLA)^47,48 The mice were injected with an overdose of sodium pentobarbital (150 mg/kg i.p.) and exsanguinated by transcardial perfusion with ice-cold PBS, followed by 4% paraformaldehyde. The brains were removed from the skull, postfixed in 4% paraformaldehyde at 4 °C overnight, then cryoprotected using 10% sucrose (0.02% sodium azide) for 3 days, followed by 30% sucrose (0.02% sodium azide) until they sank to the bottom. The brains were snap-frozen in isopentane (−55 °C for 30 s) before sectioning in a cryostat. Brain sections (35 µm thick) through the hippocampal formation were collected at bregma level −3.6 mm and stored in Hoffman solution (250 mL 0.4 M PBS, ethylene glycol 300 mL, 300 g sucrose, 10 g polyvinylpyrrolidone, 9 g NaCl, and high purity water to 1000 mL) at −20 °C until further use. Free-floating PLA was performed using the DUO92101 kit as instructed by the manufacturer. All procedures were performed at room temperature unless specified otherwise. The sections were washed 3 × 5 min in PBS, followed by 20 min of quenching in 10 mM glycine (Sigma-Aldrich), then washed 2 × 5 min in PBS. The sections were then permeabilized with 0.1% Triton X-100 for 30 min before a final washing step of 2 × 5 min in PBS and blocking step of 30 min in blocking buffer (SuperBlock, Thermo Scientific). The sections were incubated with primary antibodies diluted in blocking buffer overnight at 4 °C, with the following antibodies: mouse monoclonal anti-GR (1:100, SAB4800041), rabbit polyclonal anti-GR (1:50, sc-8992) and rabbit polyclonal anti-HSP90 (1:25, SAB4300541). Control experiments included only one primary antibody. The following day the sections were washed 2 × 5 min in blocking buffer and transferred to 37 °C humidified chamber with the probes (1:5, PLA Probes) for 2 h. Afterward the slices were washed 2 × 5 min under gentle agitation first in blocking buffer, then in PBS. Sections were incubated with the hybridization-ligation solution (1:5, Ligation Stock (5×) in dH₂O, 1:40, Ligase (1 U/uL)) for 1 h in a humidified chamber at 37 °C. The slices were washed twice in Buffer A (8.8 g NaCl, 1.2 g Tris base, 0.5 mL Tween 20 in 1000 mL water adjusted to pH 7.4 with HCl), before incubation with the rolling circle mixture (1:5, Amplification Stock (5×) in dH₂O, 1:80, Polymerase (10 U/ul)) for 100 min in humidified chamber at 37 °C. The slices were washed 2 × 10 min in Buffer B (5.84 g NaCl, 4.24 g Tris base, 26 g Tris-HCl in 1000 mL water adjusted to pH 7.5 using HCl) shielded from light at RT under gentle agitation, then mounted on microscope slides with mounting medium (Mounting medium with DAPI, Duolink, Sigma-Aldrich), coverslipped, sealed with nailpolish, and stored at −20 °C until imaging. The PLA signal was detected using a Leica TCS-SL confocal microscope (Leica) and quantified with Duolink ImageTool 1.0.1.2 (Sigma-Aldrich). A stack of 20 adjacent optical slices (1 µm thick) was acquired for blue (DAPI) and red (PLA) channels. The PLA signal was visible as intense spots (see Supplementary Fig. S2) located in the cytosol or inside the nucleus. The images used for analysis were generated from the red channel by maximum projection intensity of confocal stacks. We estimated the signal density within manually delineated regions of interest (ROI) inside the pyramidal cell layer of hippocampal CA as the proportion of area occupied by red signal after applying a threshold. A number of 3 images per area for each animal (N = 4 animals/group) was used for deriving the average signal density in the hippocampus and was used for subsequent analyses.

Animals were killed within an interval spanning 4.5 h around the expected diurnal trough in Bmal1 expression in the SCN (early subjective night)⁴⁹. The relative mRNA expression for core clock genes Bmal1 and Per1 was normalized to the average of all samples before estimating the fold regulation using the 2^−ΔΔCt method. We assessed the coupling between oscillations in core clock gene expression by means of state portraits⁵⁰, where the state of an individual animal (data point) is described either by the expression of Bmal1 and Per1 in the same brain region, or by the expression of Bmal1 in two brain regions (SCN and hippocampus). Synchronized oscillations are defined as oscillations which are phase-locked, i.e., they have the same period, and the time difference between occurrence of peaks is constant. Under steady photic entrainment, when the oscillations in clock gene expression are synchronized⁵¹, the state portraits approximate an arch of an ellipse, where the order of positions of datapoints follows the temporal sequence of sample collection. First we assessed the synchronization of clock gene expression within brain region by plotting the relative expression level of Per1 against Bmal1 mRNA in SCN and hippocampus, respectively. Second we assessed the synchronization between the oscillations in Bmal1 expression across regions by plotting the relative expression level of Bmal1 mRNA in hippocampus against the relative expression level in the SCN.

Tissue samples (~0.25 cm²) were harvested from the ear of adult (6 mo) control and DEX-exposed mice under terminal anesthesia. The tissue was rinsed in Hank’s Balanced Salt Solution (HBSS) (Life Technologies Europe BV, Stockholm, Sweden), then minced with sterile razor blade into Collagenase (Type XI-S) (Sigma-Aldrich, Sweden) (30 min at 37 °C). After digestion, 3 mL of DMEM Medium (Life Technologies) supplemented with 10% Fetal Bovine Serum and 1% Penicillin/Streptomycin (Life Technologies) was added to a 6 cm plate and the samples were incubated at 37 °C for at least 6 days. After passaging (0.05% Trypsin-EDTA; Invitrogen), the cells were plated in 35 mm dishes in MEF medium (DMEM Medium +10% FBS +1% pen/strep) at a density of at least 50 k/cm². After 24 h, the expression of clock genes was synchronized by exposing the fibroblasts to 1 µM DEX. The cells were collected between 36 and 63 h after synchronization. The relative expression of Bmal1 was assessed by qPCR with GAPDH as housekeeping gene. Circadian oscillations in clock gene expression were analyzed by means of cosinor rhythmometry^52,53.

The maturation of newborn granule neurons was investigated in cells expressing green fluorescent protein (GFP) delivered using a retroviral vector injected stereotactically in the dorsal hippocampus as described earlier⁴⁴. The injection site was located as follows, in relation to bregma: antero-posterior, −2.6 mm; medio-lateral, +1.75 mm; dorso-ventral, −2.0 mm (from dura)⁵⁴. The GFP-expressing cells were analyzed 4 weeks after the injection, which corresponds to the full morphological maturation of the retrovirus labeled granule cells (21–28 days post infection)⁵⁵. For the morphological analysis, all brain sections from the hemisphere injected with retroviral particles through the entire hippocampus (12–15 sections/animal) were collected in PBS for 10 min washing, and subsequently processed for free-floating immunohistochemistry as described above. Sections were then mounted on Microscope Slides cut edges frosted (VWR International) with Fluorescent Mounting Media (DAKO). We used rabbit anti-mouse GFP (1:1000, Thermo Fisher; USA) and Pierce^TM Donkey anti-Rabbit IgG (H + L) Cross Adsorbed Secondary Antibody 543 (1:200, Thermo Fisher; USA). An average optical thickness of 50 μm was imaged (from the slice thickness of 70 μm). Z-series at 1 μm intervals were acquired with a Plan-Apochromat 20 × 0.75 objective, digital zoom 1.5, on a Zeiss ZEN 2009 LSM 510 META confocal system. A total of 5–12 cells from each mouse were analyzed for each data point. All images were analyzed in FiJi⁵⁶ using a semiautomatic procedure. Images of dendritic arborization were deconvolved with Iterative deconvolve 3D plugin and manually traced with Simple Neurite Tracer plugin before running automated Sholl analysis. Granule cells displaying truncated dendrites, or grossly departing from the plane of the coronal sections, were excluded from the analysis. The number of neurons analyzed ranged between 4 and 12/animal (N = 7–12 animals/treatment). See also ref. ⁴⁴ for further methodological details.

Neurogenesis and the maturation of newborn neurons were analyzed as previously described⁴⁴ in mice aged 6 or 12 mo. Briefly, proliferating of neuronal progenitors were labeled by 5-ethynyl-2 deoxyuridine (EdU) uptake. EdU (50 mg/kg/day) was injected i.p. for 5 consecutive days before killing the animals at 6 months. The EdU uptake was visualized using the Click-iT™ EdU Alexa Flour™ 488 Imaging kit (Thermo Fisher) as instructed by the manufacturer. Postmitotic progenitors committed to the neuronal lineage were identified in 12 mo animals by immunohistochemical detection of DCX expression. The number of EdU or DCX-positive cells in the subgranular zone was assessed on 20 μm thick frozen coronal section collected every 200 μm from the left hemisphere (i.e., not injected with viral particles). The slices were mounted on SuperFrostPlus™ Microscope Slides (VWR International) and circled with Dako pen. After 10 min re-hydration with phosphate buffered saline (PBS), sections were treated with blocking mix (PBS, 5% Normal Donkey Serum, 0.3% Triton-X) for 2 h. Sections were then incubated overnight at 4 °C with primary antibody. After washing with PBS 3 × 5 min, sections were incubated with secondary antibody for 2 h at room temperature. After washing 2 × 5 min with PBS, sections were counterstained with 4′,6′-diamidino-2-phenylindole (DAPI, 1:1000, Sigma-Aldrich) for 5 min, washed with PBS for 3 × 5 min and mounted with coverslips. The primary antibody used: Guinea pig anti-mouse DCX antibody (1:1000, Millipore; CA). The secondary antibody used was Donkey anti-Guinea pig 488 (1:200, Alexa). The DCX-positive cells in the granule cell layer of the DG were counted throughout the entire hippocampal structure in a stereological design on equally spaced 20 μm sections (200 μm between consecutive slides). The total number of cells was estimated by multiplying the number of counted cells with the inverse sampling fraction.

All statistical analyses were performed in Statistica™ 13 (TIBCO Software Inc., Palo Alto, CA, USA). The onset of active phase and the variability in spontaneous activity were analyzed using a mixed model ANOVA with repeated measures, followed by Dunnett’s post hoc test against DEX as reference group. mRNA expression was analyzed first by one-sample t-test vs. “no-regulation” value (=1), then between-group comparisons were analyzed by two-sample t-test. The time to re-entrain, immobility time in FST, and PLA group differences were analyzed using simple ANOVA models followed by Dunnett’s post hoc test against DEX as reference group. The differences between Sholl curves were reported only when pointwise differences (t-test) were found significant for at least three consecutive points (see also ref. ⁴⁴). Differences are reported as significant if p < 0.05. All individual comparisons are one-sided. The data are shown as average ± SEM.

Next we asked whether the differential gene expression regulation was accompanied by changes in GR activation. We measured in situ PLA signal intensity for activated (GR-GR homodimers) and inactive (GR-Hsp90 heterodimers) fractions of GR in the Cornu Ammonis (CA) region of the hippocampus. The PLA signal for activated GR was localized within the nucleus (Supplementary Fig. S2A), while the inactive form was localized primarily in the cytosol (Supplementary Fig. S2B). The signal for activated GR was lower in the DEX-exposed mice than in controls, while the signal for inactive GR was not different between groups (Fig. 3c). After DMI treatment, the PLA signal for both activated and inactive GR was increased in DEX-exposed mice (Fig. 3c).

To investigate the response of peripheral oscillators to entrainment by GR-mediated signaling we analyzed clock gene expression in primary skin fibroblasts following synchronization by GR activation (see also ref. ²⁷). The amplitude of oscillations in core clock genes Bmal1 and Per1 was dampened in cells isolated from DEX mice. In contrast, skin fibroblasts isolated from DMI-treated DEX mice display increased amplitude of oscillations in clock gene expression as compared with cells isolated from untreated mice (Fig. 4c).

Supplementary Information