Dr Gimble serves as a consultant or collaborator with Artecel, Inc., Vesta Therapeutics, Vet-Stem, Inc., and Zen-BioA, Inc. All other authors state that they have no conflicts of interest.
The genes encoding the core circadian transcription factors display an oscillating expression profile in murine calvarial bone. More than 26% of the calvarial bone transcriptome exhibits a circadian rhythm, comparable with that observed in brown and white adipose tissues and liver. Thus, circadian mechanisms may directly modulate oxidative phosphorylation and multiple metabolic pathways in bone homeostasis.
Introduction: Although circadian rhythms have been associated historically with central regulatory mechanisms, there is emerging evidence that the circadian transcriptional apparatus exists in peripheral tissues. The aim of this study was to determine the presence and extent of circadian oscillation in the transcriptome of murine calvarial bone.
Materials and Methods: Cohorts of 8-week-old male AKR/J mice were maintained in a controlled 12-h light:12-h dark cycle on an ad libitum diet for 2 weeks. Groups of three mice were killed every 4 h over a 48-h period. The level of gene expression at successive times-points was determined by quantitative RT-PCR and Affymetrix microarray. Data were analyzed using multiple statistical time series algorithms, including Cosinor, Fisher g-test, and the permutation time test.
Results: Both the positive (Bmal1, Npas2) and negative (Cry1, Cry2, Per1, Per2, Per3) elements of the circadian transcriptional apparatus and their immediate downstream targets and mediators (Dbp, Rev-erbα, Rev-erbβ) exhibited oscillatory expression profiles. Consistent with findings in other tissues, the positive and negative elements were in antiphase relative to each other. More than 26% of the genes present on the microarray displayed an oscillatory profile in calvarial bone, comparable with the levels observed in brown and white adipose tissues and liver; however, only a subset of 174 oscillating genes were shared among all four tissues.
Conclusions: Our findings show that the components of the circadian transcriptional apparatus are represented in calvarial bone and display coordinated oscillatory behavior. However, these are not the only genes to display an oscillatory expression profile, which is seen in multiple pathways involving oxidative phosphorylation and lipid, protein, and carbohydrate metabolism.
Circadian rhythms are evident throughout the phylogenetic tree, from algae to mammals. These oscillations in gene expression synchronize the organism's physiology, allowing it to anticipate, adapt, and respond to daily changes in its external environment.(1)
The coordinated action of basic helix-loop-helix/Per-Arnt-Simpleminded (bHLH-PAS) domain proteins, encoded by Clock (or its paralog Npas2) and Bmal1, transcriptionally regulate the circadian patterns of gene expression.(2,3) CLOCK heterodimerizes with BMAL1 to drive the rhythmic expression of PAS domain proteins encoded by Period (Per) 1–3 and flavoproteins encoded by Cryptochrome (Cry) 1 and 2.(4,5) Translated PER and CRY proteins form heterodimers in the cytoplasm and translocate to the nucleus to complete the transcriptional/translational feedback loop regulating the activity of CLOCK:BMAL1.(6–8) Consequently, these two distinct sets of genes oscillate in antiphase of one another. CLOCK: BMAL1 dimers also drive the expression of circadian effector genes, such as the genes encoding transcription factors DBP and REV-ERBα, believed to govern circadian oscillations observed in liver, adipose tissue, and other metabolically active sites.(9–11)
Whereas the classical circadian studies identified the suprachiasmatic nucleus (SCN) in the brain as the core circadian oscillator, investigators have long appreciated the existence of circadian mechanisms in peripheral tissues.(12,13) This notion has been substantiated in recent studies of mPer2 promoter:Luciferase reporter mice.(14) In a manner parallel to the SCN, liver and muscle explants from these transgenic mice display a persistent oscillatory Luciferase profile for >20 days ex vivo.(14) In vivo, oscillators in peripheral tissues continue to operate even in animals where the SCN has been surgically ablated, showing that independent circadian oscillators exist within peripheral tissues. Recent studies from our laboratory and others have shown a pervasive activity of circadian mechanisms in adipose depots, liver, and other peripheral tissues.(15–20) Our transcriptomic analyses have shown that at least 15% or more of adipose tissue genes display a circadian expression profile.(15,16) In light of the extensive studies supporting the role of circadian mechanisms in bone metabolism,(21–40) we applied a combined quantitative RT-PCR (qRT-PCR) and Affymetrix microarray approach to examine the hypothesis that gene expression of multiple metabolic pathways in murine calvarial bone undergoes circadian oscillation. Because the mRNA levels of Bmal1, Clock, Cry, Dbp, Npas2, Per, Rev-erbα, and Rev-erb β displayed a robust and characteristic circadian rhythm in adipose tissues and liver, we focused our initial attention to their expression in calvarial bone.(15,16)
MATERIALS AND METHODS
In vivo circadian studies
All protocols were reviewed and approved by the Pennington Biomedical Research Center Institutional Animal Care and Use Committee. Studies used 8- to 10-week-old male AKR/J mice obtained from the Jackson Laboratories (Bar Harbor, ME, USA). This murine strain has been extensively characterized with respect to its metabolic phenotype(41) and had been used in our previous analyses of adipose and liver circadian rhythms.(15,16) For these reasons, we continued our examination of circadian biology in calvarial bone in this murine strain. The animals were acclimated to a regular chow diet (Purina 5015) ad libitum, under a strict 12-h light:12-h dark cycle for 2 weeks. After the acclimation period, animals were killed in groups of three (December 2003) or five (September 2004) every 4 h over a 48-h period. All animals were killed by CO2 asphyxiation and cervical dislocation. From the individual animals, we harvested calvarial bone and serum, inguinal white adipose tissue (iWAT), epididymal WAT (eWAT), brown adipose tissue (BAT), and liver. The calvarial bone was dissected using scissors and forceps; grossly visible skeletal muscle and connective tissue was removed before preserving the bone tissue by submersion in liquid nitrogen.
Total RNA was purified from tissues collected using TriReagent (Molecular Research Center) according to the manufacturer's specifications. Approximately 2 μg of total RNA was reverse transcribed using Moloney murine leukemia virus reverse transcriptase (MMLV-RT; Promega), with Oligo dT at 42°C for 1 h in a 20-μl reaction. Primers for genes of interest were identified using Primer Express software (Applied Biosystems). A complete list of primers used in these studies is listed in Table 1. qRT-PCR was performed in triplicate on diluted cDNA pools prepared with equal quantities of reverse transcribed RNA harvested from all animals at each individual time-point with SYBR Green PCR Master Mix (Applied Biosystems) using the 7900 Real Time PCR system (Applied Biosystems) under universal cycling conditions (95°C for 10 minutes; 40 cycles of 95°C for 15 s; 60°C for 1 minute). All results were normalized relative to a cyclophilin B expression control.
Table Table 1.. Oligonucleotide Primers
Periodicity of the circadian data obtained by qRT-PCR was tested with Time Series Analysis-Single Cosinor v. 6.0 software (Expert Soft Technologie) as described in.(15,16) Each data set was fitted to a general cosine equation model:
where A is the amplitude, T is the period (24 h), and M is the MESOR (midline estimating statistic of rhythm).(42,43) Model (ANOVA) was set as valid at the 0.950 probability level. The goodness of fit for each data set was tested with Kolomogorov and Smirnov (K-S), k2, average, and Q (Ljung-Box Q-statistic lack-of-fit hypothesis) tests; each individual data set reported has met acceptance criteria for each of these tests.
RNA integrity was assessed by electrophoresis on the Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA). Pools of total RNA were prepared, combining equal amounts of RNA isolated from the calvarial bone of each of the three individual mice for each time-point in the circadian study. Double-stranded cDNA was synthesized from ∼9 μg of pooled total RNA using a Superscript cDNA Synthesis Kit (Invitrogen, Carlsbad, CA, USA) in combination with a T7-(dT)24 primer. Biotinylated cRNA was transcribed in vitro using the GeneChip IVT Labeling Kit (Affymetrix, Santa Clara, CA, USA) and purified using the GeneChip Sample Cleanup Module. Ten micrograms of purified cRNA was fragmented by incubation in fragmentation buffer (200 mM Tris-acetate, pH 8.1, 500 mM potassium acetate, 150 mM magnesium acetate) at 94°C for 35 minutes and chilled on ice. Fragmented biotin-labeled cRNA (6.5 μg) was hybridized to the Mouse Genome 430A 2.0 Array (Affymetrix), interrogating over 14,000 substantiated mouse genes. Arrays were incubated for 16 h at 45°C with constant rotation (60 rpm), washed, and stained for 10 minutes at 25°C with 10 μg/ml streptavidin-R phycoerythrin (Vector Laboratories, Burlingame, CA, USA) followed by 3 μg/ml biotinylated goat anti-streptavidin antibody (Vector Laboratories) for 10 minutes at 25°C. Arrays were stained once again with streptavidin-R phycoerythrin for 10 minutes at 25°C. After washing and staining, the arrays were scanned using a GeneChip Scanner 3000. Pixel intensities were measured, expression signals were analyzed, and features were extracted using the commercial software package GeneChip Operating Software v.1.2 (Affymetrix). Data mining and statistical analyses were performed with Data Mining Tool v.3.0 (Affymetrix) algorithms. Arrays were globally scaled to a target intensity value of 2500 to compare individual experiments. The absolute call (present, marginal, absent) of each gene expression in each sample, and the direction of change, and fold change of gene expressions between samples were identified using the above-mentioned software.
Spectral analysis of microarray data
Series of microarray expression values for gene x with N samples of the form x0,x1,x2, … xN−1 were converted from time-domain to a frequency domain using the discrete Fourier transform (DFT) algorithm(15,16):
Time series with a significant sinusoidal component with frequency ω∈ [0, π] showed a peak (periodogram) at that frequency with a high probability, unlike the purely random series whose periodogram approaches a flat line.(44) The significance of the observed periodicity was estimated by Fisher g-statistics, as recently recommended.(45) To account for multiple testing problems we used the false discovery rate (FDR) method as a multiple comparison procedure.(46) This method is less conservative compared with the classic Bonferroni correction, which makes it more applicable for testing large numbers of relatively short time series produced by microarray experiments. This method is adaptive to the actual data(45) and has been shown to control the FDR.(46)
Circadian gene identification and annotation
Circadian-expressed genes detected by Affymetrix microarray analysis were identified and annotated by matching the probe-set number with the gene information in the DAVID database.
qRT-PCR analysis of core circadian transcription factors in calvarial bone
The initial studies examined the expression profile of core circadian transcription factors and their immediate targets by qRT-PCR in total RNA samples from murine calvarial bone, harvested at 4-h intervals over a 48-h period (Fig. 1). Times are displayed relative to the start of the “lights-on” period, given as Zeitgeber time 0 (ZT 0), and the start of the “dark” period (ZT 12). The gene expression of positive regulatory factors, Bmal1 and Npas2, displayed a zenith at ZT 0 and a nadir at ZT 12 (Fig. 1A). As seen in other metabolic tissues, the Npas2 homolog, Clock, did not display a consistent oscillatory profile (Fig. 1A). In contrast, the negative regulatory factors, Per 1–3 and Cry 2, displayed phase shifts relative to Bmal1 and Npas2, with zeniths between ZT 8–12 and a nadir at ZT 0 (Fig. 1A). Likewise, the zenith of Cry 1 was at ZT 20 and the nadir at ZT 4 (Fig. 1A). The immediate downstream targets and mediators of the BMAL1/ NPAS2 transcriptional complex include Rev-erb〈, Rev-erb, and Albumin D-Binding Protein (DBP) (Fig. 1B). Each of these genes exhibited an oscillatory circadian profile with zeniths at ZT 8–12 and nadirs at ZT 20–24. The periodic oscillatory pattern of expression for each gene was confirmed by fitting the data to cosine curves as mathematical models using Cosinor software analyses. With the exception of Clock (p = 0.5528) and Cry 2 (p = 0.0687), all mRNA expression patterns fit the cosine curves with p < 0.05.
Microarray analysis in calvarial bone
The pooled total calvarial bone RNA from each time-point, described in Fig. 1, was examined by Affymetrix microarray. The datasets were analyzed as a time series using multiple algorithms (Fisher g-test and permutation test). Fisher g-test(47) relies on signal-to-noise ratio to identify periodicity. The highest peak of a periodogram (which could be circadian, i.e., corresponding to one complete period in 24 h) is related to the sum of all other peaks in the periodogram. This method is one of the most widely used for the analysis of expression profiles.(45) The alternative permutated time (PT test) has been developed specifically for the analysis of very short time series with low sampling rate.(16) Unlike Fisher g-test, it can identify periodicity even in the presence of other significant (noncircadian) oscillations. The noise level in the PT test is estimated through multiple random permutations of points in a time series. Although it takes much longer time to complete compared with Fisher g-test, the PT test is more sensitive. Based on the Fisher g-test, >2000 or 10%+ of the microarray-represented transcriptome exhibited evidence of circadian oscillation. The PT test identified close to 6000 or >26% of the microarray-represented transcriptome as oscillating (the complete list of these genes is available under the “Research” link on the Stem Cell Biology website at http://labs.pbrc.edu/stemcell). The heat map analysis in Fig. 2 displays all genes clustered based on the time of their peak expression at ZT0 (I), ZT4 (II), ZT8 (III), and ZT16 (IV). Genes associated with multiple biochemical pathways displayed circadian oscillations; those with ≥10 oligonucleotides identified are summarized in Table 2. The list identified by the PT test included multiple genes directly related to bone metabolism. Among these were genes potentially related to bone formation, bone remodeling, and hematopoietic function (cataloged in Table 3).
Table Table 2.. Most Common Biochemical Pathways Identified as Oscillatory by Permutation Testing
Table Table 3.. Partial List of Genes Identified by PT Test With Circadian Oscillation and Potential Association With Bone Metabolic Functions
It should be noted that only 814 of the oligonucleotides detected on the microarray were classified based on KEGG analysis, whereas the remaining >5000 were unclassified. Consistent with the qRT-PCR results, all 11 of the genes associated with the core circadian transcriptional apparatus represented on the microarray were detected by the permutation test. Genes related to oxidative phosphorylation were most abundant, and the majority of pathways identified were related to amino acid, glucose, or lipid metabolism. In addition, the permutation test detected 63 oligonucleotides cataloged as heat shock proteins, chaperones, and/or immunophilins. All are associated with protein folding, protein processing, and stress responses.
Circadian microarray relationships among metabolically active peripheral tissues
In previous publications, we described the circadian expression profile of BAT, iWAT, and liver isolated from this same cohort of mice.(15,16) The oscillating components from these permutation test datasets were compared with that of the calvarial bone (Fig. 3). The calvarial oscillatory gene profile overlapped 17.8%, 16.9%, and 22.4% with those of BAT, iWAT, and liver, respectively (Fig. 3). All four tissues shared a core group of 174 microarray targets with a circadian expression profile (Fig. 3). The KEGG biochemical pathways associated this core group of shared genes are summarized in Table 4.
Table Table 4.. Classified Expressed Genes Conserved Between Calvarial Bone, Brown Adipose Tissue, Inguinal White Adipose Tissue, and Liver (of 174 Total Classified and Unclassified Genes)
This study shows by both qRT-PCR and microarray analysis that the core circadian transcriptional apparatus in murine calvarial bone exhibits a robust oscillatory expression profile. These findings are consistent with previous reports detecting the circadian transcriptional apparatus in multiple peripheral tissues, including cortical bone.(14–20,40) Fu et al.(40) provided the first evidence for the involvement of circadian transcriptional regulators in bone homeostasis when they found that Per and Cry deficient mice had an elevated bone mass and an altered response to intracranial leptin administration. Even before the advent of molecular biological tools, there have been clues linking circadian rhythms to bone metabolism. Clinical studies have detected 24-h oscillations in human serum levels of osteocalcin and alkaline phosphatase, both markers of osteoblast activity, and C-telopeptide, released during collagen turnover in resorbing bone.(21–28) In a goat mandibular model, osteoblast proliferation markers displayed a circadian expression profile; this became more pronounced by a growth promoting procedure known as distraction osteogenic.(29,30) In a rat mandibular model, retractive forces shortening bone had stronger inhibitory effects on chondrocyte differentiation and proliferation during the daylight hours rather than at night.(31) Additional publications have indicated that the diurnal variation in bone resorption may be related to the release of PTH.(32–39)
Genes encoding enzymes from multiple metabolic pathways display an oscillatory expression profile in calvarial bone; these account for 26.2% of the genes represented on the microarray (5956 of 22,690). This includes a number of genes with functions closely related to bone formation and remodeling and their related signal transduction pathways, including the bone morphogenetic proteins and their receptors and transcription factors such as Runx2/Cbfa1 and NF-κB (Table 3). Future studies will need to focus on the circadian expression profile of these osteogenic genes and their encoded proteins. Based on the permutation tests, a similar percentage of genes display an oscillatory expression profile within other metabolically active tissues such as BAT (16.2%), iWAT (15.1%), and liver (20.8%).(15,16) Nevertheless, a cross-comparison with calvarial bone identifies only a small percentage (2.9–5%) of genes in common among all four tissues. This shared subset of circadian expressed genes includes few, if any, of the transcripts that directly relate to osteogenesis as summarized in Table 3. As would be predicted, the KEGG list does include seven of the circadian transcription factors (Bmal1, Cry 1, Dbp, Npas2, Per1, Per2, Per3) and Rev-erbα and Rev-erbβ (Table 4). An additional gene identified on the list, RORγ, is a potential circadian transcriptional target because the circadian transcriptional factor BMAL1and multiple RAR orphan receptors, RORγ, RORα1, and RORα, reciprocally regulate each other's promoter function.(48,49)
Mesenchymal stem cells (MSCs) contribute to the formation of calvarial bone and the shared gene list contains genes related to MSC differentiation. Among these are the immediate early genes, JunB and DAZ associated protein 2 (also known as proline rich transcript of the brain), which are induced by serum shock in MC-3T3 E1 osteoblasts.(50) Additional genes relate to the adipogenic differentiation potential of bone marrow–derived MSCs. Glucocorticoid-induced leucine zipper (GILZ) has been reported to bind to C/EBPα, thereby inhibiting MSC adipogenesis.(51) Whereas C/EBPα is not included in the gene list, the related C/EBP transcription factor, C/EBPγ, is present. Likewise, eight members of the heat shock proteins/chaperones/immunophilin family are represented. Induction of the heat shock proteins/chaperone family is associated with adipogenic differentiation by stem cells from adipose tissue, which display many similarities to bone marrow–derived MSCs.(52) Similarly, the rate limiting gluconeogenic enzyme, phosphoenol pyruvate decarboxylase (PEPCK), is induced during adipogenesis. The cytokine preB colony enhancing factor 1, also known as visfatin, is expressed on adipogenic differentiation of adipose-derived stem cells (ASCs) and is elevated in the serum of obese individuals.(53) It should be noted that there are few, if any, adipocytes within the marrow cavity of calvarial bone in 8- to 10-week-old mice. Nevertheless, because adipogenic progenitor cells can be found in adipose tissue, bone, and liver, the presence of adipogenic-associated genes in the shared oscillatory profile is not a surprise.
These microarray data extend our appreciation of the circadian nature of calvarial bone metabolism. To support this point, in Fig. 4, we show a subset of interrelated oscillating genes involved in the glucocorticoid receptor signaling pathway. It is well established that pathological or pharmacological elevation of serum glucocorticoid levels can lead to osteonecrosis and osteoporosis. Classical studies in the circadian literature have documented the oscillatory profile of physiological levels of serum glucocorticoids. In our previous study on the cohort of mice used in this analysis, we found that the serum corticosterone zenith was detected between ZT 8 and ZT 12 and its nadir was at ZT 20.(15,16) In parallel with this, the calvarial bone mRNA levels of the glucocorticoid receptor and its associated chaperones, heat shock proteins 70 and 90, all displayed an oscillatory expression profile by microarray. These findings have potential practical implications relating to bone homeostasis and patient care. The time of day when patients receive pharmacological doses of corticosterone may accentuate or ameliorate the net glucocorticoid effect on bone homeostasis because of innate circadian rhythms in gene expression. We conclude that (1) calvarial bone possesses an active circadian apparatus, potentially capable of independent function, similar to those proposed to exist in other peripheral tissues, and (2) circadian rhythms may be involved in the regulation of multiple metabolic pathways in calvarial bone, aside from the core circadian transcriptional apparatus.
The authors thank Paula Polk (Louisiana State University Health Science Center-Shreveport, Genomic Core Laboratory) for assistance with the microarrays, Dr Randall Mynatt and the members of the Stem Cell Biology Laboratory (Pennington Biomedical Research Center) for discussions and comments on the final manuscript, Laura Dallam for administrative assistance, and funding support from the Pennington Biomedical Research Foundation and the National Institutes of Health (NIDCR R21 DE016371 nd NIDDK CNRU DK072476).