The Transcription Factor Runx2 Is under Circadian Control in the Suprachiasmatic Nucleus and Functions in the Control of Rhythmic Behavior

Runx2, a member of the family of runt-related transcription factors, is rhythmically expressed in bone and may be involved in circadian rhythms in bone homeostasis and osteogenesis. Runx2 is also expressed in the brain, but its function is unknown. We tested the hypothesis that in the brain, Runx2 may interact with clock-controlled genes to regulate circadian rhythms in behavior. First, we demonstrated diurnal and circadian rhythms in the expression of Runx2 in the mouse brain. Expression of Runx2 mRNA and protein mirrored that of the core clock genes, Period1 and Period2, in the suprachiasmatic nucleus (SCN), the paraventricular nucleus and the olfactory bulb. The rhythm of Runx2 expression was eliminated in the SCN of Bmal1−/− mice. Moreover, by crossbreeding mPer2Luc mice with Runx2+/− mice and recording bioluminescence rhythms, a significant lengthening of the period of rhythms was detected in cultured SCN of Runx2−/− animals compared to either Runx2+/− or Runx2+/+ mice. Behavioral analyses of Runx2 mutant mice revealed that Runx2+/− animals displayed a significantly lengthened free-running period of running wheel activity compared to Runx2+/+ littermates. Taken together, these findings provide evidence for clock gene-mediated rhythmic expression of Runx2, and its functional role in regulating circadian period at the level of the SCN and behavior.


Introduction
The mammalian timekeeping system that drives rhythmic processes occurring on a 24 hour cycle is organized by a hierarchy of oscillators whose phase is coordinated by the suprachiasmatic nucleus of the hypothalamus (SCN) [1,2]. In individual SCN cells, as in peripheral cellular oscillators, circadian rhythms are driven by networks of transcriptional-translational auto-regulatory loops (TTL) having a cycle lasting approximately 24 hours. The basic helix-loop-helix transcription factor Bmal1 (brain and muscle ARNT-like 1) and Clock (circadian locomotor output cycle kaput) heterodimerize and drive transcription of core clock genes Period (Per1, Per2 and Per3) and Cryptochrome (Cry1 and Cry2) genes by binding E-box elements within gene sequences [3]. Per and Cry proteins accumulate in the cytoplasm where they form heteromultimeric complexes with other proteins and translocate into the nucleus to inhibit the activity of Clock/Bmal1. As the levels of Per and Cry expression and protein concentration decrease, the inhibitory complexes are degraded thus restarting the cycle. In addition to core clock factors, genes involved in basic cellular events are also tightly regulated by the TTL (for examples, see [4][5][6][7]). By regulating the rhythmic expression of these clock-controlled genes (CCGs), the circadian system coordinates rhythmic physiological processes at both cellular and tissue levels.
Runt-related transcription factor 2 (Runx2) is one of three members (Runx1-3) of the mammalian family of transcription factors harboring a runt domain, a highly conserved DNA-binding sequence first described in Drosophila melanogaster [8][9][10]. Recent studies point to a putative interaction between Runx2 and the circadian clock network in osteogenesis and bone homeostasis [11,12]: a) clock gene mutant mice show deregulated bone formation parameters largely attributed to increased numbers of osteoblasts, a cell type whose development from its precursors is highly dependent on Runx2-mediated events [13,14]; b) Runx2 expression and several of its downstream gene targets are rhythmically expressed in bone [15]; and c) oscillations in molecular clock genes are detectible in bone and in cultured osteoblasts [14,16]. Since Runx2 expression is also detectable in non-osseous tissues including pituitary, gonads and the brain [17][18][19], these findings suggest that Runx2 may be a CCG in multiple tissues.
In this study, we first examined temporal expression of Runx2 mRNA and protein in the SCN and other brain regions known to be rhythmic, and compared its circadian profile to that of the core clock genes Per1, Per2 and Bmal1. Second, we determined whether Runx2 was under control of the circadian clock by examining its expression in the SCN of Bmal1-null mice. Third, we compared molecular rhythms between SCNs cultured from Runx2 +/+ , Runx2 +/2 and Runx2 2/2 mice in real-time by crossbreeding Runx2 +/mice with mice harboring the circadian reporter gene construct, mPeriod2::Luciferase (mPer2 Luc ). Finally, we examined activity rhythms in Runx2 +/2 mice to determine whether Runx2 has a role in regulating behavioral clock outputs. Taken together, the findings from these experiments provide insight to the relationship between Runx2 and the circadian system including evidence for clock gene-mediated rhythmic expression of Runx2, and the functional role for Runx2 in regulating circadian period at the level of the SCN and behavior.

Animals
All animal care and experimental procedures were approved by the Canadian Council of Animal Care and the Animal Use Subcommittee of the University of Western Ontario.
All mice strains, including mutants, were generated on a C57BL/6 background. Adult, male C57BL/6 mice (8-10 weeks old; Charles River Laboratories, Montreal, Canada) were used for gene and protein expression studies. Male Bmal1 knockout mice (Bmal1 2/2 ) mice were generated by breeding Bmal1 +/2 mice (founder mice generously provided by Dr. C. Bradfield, University of Wisconsin Medical School, Madison, WI) [20]. Offspring genotyping was performed similar to the methodology previously described [20]. However, a non-multiplex PCR approach was used in place by of multiplex PCR since the mutant allele is expressed at lower levels than the wild type allele, which can lead to competitive primer binding during multiplex PCR resulting in instances of false negative genotyping results. Primers used for genotyping PCR: OL2646, 59-CCACCAAGCCCAG-CAACTCA-39; OL2647, 59-ATTCGG-CCCCCTATCTTC-TGC-39; and OL278, 59-TCGCCTTCTATCGCCTTCTT-GACG-39. The wild-type allele is amplified by primers OL2647 and OL2646 (WT primer set), which produce a 400 bp PCR product. The gene knock-in product (600 bp) is amplified by the neomycin-specific primer set OL278 and OL2646 (M primer set) since the region targeted by OL2647 is deleted from the mutant Bmal1 gene. In brief, PCR reactions for each animal were performed using genomic DNA isolated from tail biopsies with 26 PCR buffer (10 ml; Promega) supplemented with 2.5 mM MgCl 2 and either WT or M primer sets (1 mM of each primer). PCR reaction conditions consisted of 40 cycles of 95uC for 15 s; 60uC for 15 s; 72uC for 1 min. PCR products were separated by electrophoresis on a 1% agarose gel in 16 TAE buffer [40 mM Tris; 19 mM Acetic Acid (EMD Biosciences, San Diego, CA)]. Two PCR reactions were performed for each animal, one for each primer set, to determine which alleles were present.
Mice homozygous for the mPeriod2::Luciferase fusion protein gene (mPer2 Luc ; Jackson Laboratory, Bar Harbor, ME) were crossbred with Runx2 +/2 mice to generate mPer2 Luc /Runx2 +/2 mice for in vitro bioluminescence recording experiments. Offspring genotyping was conducted for mPer2 Luc as previously described [22]. In brief, multiplex PCR reactions were performed using 250 ng of genomic DNA isolated from tail biopsies with 26 PCR buffer (10 ml; Promega) supplemented with 2.5 mM MgCl 2 and 2.5 mM of each primer. Primers used for genotyping PCR: P1, 59-CTGTGTTTACTGCGAGAGT-39; P2, 59-GGGTCC-ATGT-GATTAGAAAC-39; P3, 59-TAAAACCGGGAGGTAGAT-GAGA-39. The wild-type allele is amplified by primers P1 and P2, which produce a 230 bp PCR product. The luciferase knockin allele PCR product (680 bp) is amplified by primers P1 and P3. PCR reaction conditions consisted of 35 cycles of 95uC for 45 s; 55uC for 45 s; 72uC for 1 min. Genotyping for Runx2 was also performed, as described above. Crossbreeding Runx2 +/2 /mPer2-Luc+/+ produced litters with predicted Mendelian ratios for the three possible genotypes: Runx2 +/+ , Runx2 +/2 and Runx2 2/2 . Brain and liver were cultured in random order without prior knowledge of Runx2 genotype; overall, viability of tissues in culture was similar regardless of genotype and did not correlate with the order in which tissues were collected from the uterine sac.
Transgenic breeding colonies were group housed and kept under a 12 h: 12 h light: dark (LD) cycle until experimental manipulations were conducted. During experimentation, animals were individually housed at constant room temperature and humidity in cages equipped with running wheels (MiniMitter, Bend, OR) and food and water available ad libitum. Running wheel activity was monitored and recorded continuously in 5 min intervals using VitalView software (MiniMitter). Actograms, periodograms and activity levels were analyzed using CLOCKLAB software (Actimetrics, Evanston, IL).
For studies of diurnal and circadian rhythms in gene and protein expression, animals were housed in LD until running wheel activity was observed to be stably entrained (10-14 days) and subsequently either maintained in LD (7-10 days) or housed under constant dim red illumination (lux,5; DD) until stable freerunning behavior was established (10-14 days). For animals housed in LD, by convention, lights ON is designated as zeitgeber time 0 (ZT0) and lights OFF as ZT12. For animals housed in DD, the daily onset of wheel running was used to estimate circadian time 12 (CT12).

Tissue collection
Animals housed in running wheel cages were sacrificed at six defined times of day under either LD following stable entrainment or DD after stable free-running behaviour was established: ZT and CT 1, 5, 9, 13, 17 and 21, respectively (n = 12 per time of day). For Bmal1 2/2 mice, animals were transferred into DD at ZT12 and sacrificed at 17 h and 29 h later, reflecting ZT5 and ZT17 in the following circadian cycle. True circadian time cannot be established in these animals because behavior reflects their inherent molecular arrhythmicity in the absence of a daily light cue. Since photic stimulation of the SCN may itself stimulate expression of Runx2, animals were sampled within the first day without photic stimulation in order to sample tissues prior to the breakdown of entrained rhythms.
For tissue collection for gene or protein expression analyses, animals were overdosed with sodium pentobarbital; brains were rapidly removed and deeply frozen under sterile RNase-free conditions. Microdissections of distinct brain regions were performed by isolating tissue from frozen coronal sections using a sterilized blunt-end needle. Brain regions included olfactory bulb, suprachiasmatic nucleus (SCN), paraventricular nucleus of the hypothalamus (PVN) and hippocampus (HP). Tissue punches were stored at 280uC until further processed.
For immunohistochemistry analyses, animals were deeply anaesthetized with sodium pentobarbital and perfused transcardially with 50 mL of 4% paraformaldehyde in 0.1 M phosphate buffer. Brains were removed, post-fixed in 4% paraformaldehyde for 4 h at 4uC and then stored in 20% sucrose in 0.1 M sodium phosphate buffer (PB) with 0.01% sodium azide. Brains were cut using a freezing microtome into 4 parallel series of 25 mm thick coronal sections and stored at 220uC in cryoprotectant [23].
Total mRNA isolation and real-time polymerase chain reactions Each reaction was performed using the following PCR conditions: hold at 50uC for 2 min and 95uC for 10 min, 40 cycles of 95uC for 15 sec and 58uC for 1 min. The cycle threshold (C t ) values were determined from an optimal threshold line generated from standard curve dilutions for each gene analyzed. In brief, a series of 5-log dilutions of cDNA reverse transcribed from RNA isolated from the brain region of interest was used to perform qPCR reactions for each probe to be used for analyses. Standard curves were deemed optimal when the line of best fit through the C t values of the dilutions had a R 2 $0.99 and an efficiency value between 0.8-1. The C t of each experiment sample was determined from the intersection between the baseline subtracted amplification curve of the sample and the threshold line established by the optimal standard curve. The gene of interest levels of expression within a sample was normalized to their respective levels of the expression of the housekeeping gene, glyceraldehyde-3-phosphate dehydrogenase (GAPDH), within each sample; this value corresponds to the DC t value. In the current study, the DDC t value represents the difference between gene expression in an individual sample at a given time of day and the overall average level of gene expression measured at all-time points sampled rather than a comparison to a control group as typically performed (DC t(average) 2DC t(sample) = DDC t ). Relative mean fold changes for each sample were calculated relative to the average DDC t across samples at all times of day sampled with the value 1 representing a baseline level of expression. One-way Anovas and two-tailed t-tests were conducted to identify significant daily rhythms of each gene within each brain region, with significance levels of p,0.05.
For analyses of SCN tissues from Bmal1 mutant mice, data were normalized to GAPDH levels and are represented as a relative mean fold change from the average gene expression detected at both time points with the value 1 representing a baseline level of expression. One-way Anovas and two-tailed t-tests were conducted to identify significant daily rhythms. Values were considered significantly different with p,0.05.

Total protein isolation and western blotting
Protein samples were generated by pooling tissue isolated from three mice in the same treatment group to generate a total of n = 4/time point (CT9 and CT21). Methodology applied for protein isolation, preparation of protein lysates, PAGE, electrophoretic blotting and immunodetection of proteins of interest has been previously described [24]. In brief, protein samples of equal volume and concentration were loaded on a 10% polyacrylamide gel and separated under reducing conditions [24]. All samples were run in duplicate and balanced by time point and genotype across individual gels. Precision Plus Protein All Blue Standards (Bio-Rad Laboratories Ltd.) were used as molecular weight markers. Following protein transfer onto Millipore Immobilon-FL polyvinylidene difluoride membranes (Millipore, Billerica, MA), membranes were individually incubated for 16 h on a shaker at 4uC with an antibody raised against the loading control protein, GAPDH (mouse anti-GAPDH; 1:20,000; MAB374, Chemicon International, Temecula, CA), and one of the following rabbit anti-mRunx2 (mouse) antibody M-70 (1:200; Santa Cruz Biotechnology Inc., Santa Cruz, CA); and rabbit anti-Bmal1 (1:200; PA1-523, Affinity BioReagents, Golden, CO. The following secondary antibodies were used to visualize immunolabeled blots: IRDye 800CW-conjugated polyclonal goat anti-mouse (1:20,000; LI-COR Biosciences, Lincoln, NE); Alexa 680-conjugated polyclonal goat anti-rabbit (1:5000; Invitrogen); and IRDye 800CW-conjugated polyclonal donkey anti-mouse (1:20,000; LI-COR Biosciences). Images were captured using an Odyssey 2.1 scanner (LI-COR Biosciences). Quantification of protein concentration was performed by determining ratios of protein of interest/ GAPDH intensity levels for each lane and averages of the duplicates were calculated for each protein sample. The overall fold change in protein across the day was determined by normalizing the ratios for each sample to the overall mean at CT9 of the control (WT) groups. Post hoc Tukey analyses were performed where appropriate to determine time of day effects and values were considered significantly different with p,0.05.
All primary antibodies resulted in single bands of immunoreactivity at their respective expected molecular weights. Labeling of western blots of protein isolated from calvarial bone and SCN tissue with the Runx2 antibody resulted in labeling of a single band at the predicted molecular weight for Runx2 (,60 kDa; Figure 1B) further confirming the specificity of the antibody recognition over Runx1 (,50 kDa) and Runx3 (,45 kDa).The specificity of the Runx2 antibody used in the current study has been previously documented [25][26][27].

Immunohistochemistry
For immunohistochemical analyses, a series of sections throughout the forebrain, including sections containing the SCN and PVN, from perfused tissue was processed for Runx2. Immunostaining was performed at room temperature, with gentle agitation of the free-floating sections, and with extensive rinses with 0.

Histological analysis
To determine the global distribution of Runx2, a qualitative analysis was performed of Runx2 immunoperoxidase labeled sections of wild type mice (n = 4 at CT5 and CT9 each). Optical density analyses of Runx2 immunoperoxidase labeled sections were conducted using images of the SCN and PVN that were captured with a Leica DFC420 camera connected to a Leica DM5000B microscope (Leica Microsystems, Wetzlar, Germany). All images (2-3 images per animal/area of interest) were captured using identical microscope and camera settings. Density analyses of immunolabeled cells were conducted using ImageJ software (NIH, Bethesda, MD) with fixed threshold settings and were expressed as the area above threshold. Mean areas above threshold (6 SEM) were calculated for each area in individual animals and used to calculate the mean over all animals (6 SEM) for each brain region at a given time of day (CT1, 5, 9, 13, 17, and 21). The data were expressed as the mean fold change relative to the overall mean across times of day analyzed (6 SEM; 6 time points). Anova analyses were performed to determine the overall effects of time of day followed by Tukey post hoc tests for pair-wise comparisons between each time of day when appropriate. Values were considered significantly different when p,0.05.

Confocal Microscopy
Cortical tissues were imaged and captured from coronal slice of adult mouse brain (25 mm) using a Zeiss LSM-510 ConfoCor laser scanning confocal microscope system (Carl Zeiss MicroImaging GmbH, Göttingen, Germany). Alexa 488 fluorescence (green) was imaged with a 505-530 nm band pass filter and Argon laser and Alexa 555 fluorescence (red) with a 560-615 nm band pass filter and a HeNe1 laser. Z-stack images of 1.0 mm thick optical sections through sampled sections were captured with x63 oil immersion objective. Images were used for qualitative analyses only. Colocalization of NeuN and Runx2 immunostaining in cells was determined based on overlapping fluorescence detected in overlays of images captured from the same Z-stack images with each fluorescence filter. Images were viewed using Zeiss LSM Image Browser (Carl Zeiss MicroImaging GmbH) and were not altered in any way except for adjustment of brightness.
Bioluminescence data analyses were performed using the LumiCycle Analysis Program (Actimetrics, Wilmette, IL). Cycle 0 data (slice preparation start through midnight of the first day) were excluded from the period analyses and only recordings containing at least three mPer2 Luc peaks were used for analyses. Significant rhythms were determined by x 2 periodogram analyses conducted following subtraction of a baseline determined by fitting a low-order polynomial to each data set. Statistical analyses of period length and amplitude of bioluminescence were performed using the baseline-subtracted rhythms. Differences between genotypes were determined using one-way Anovas and Tukey post hoc analyses were conducted where appropriate. Values were considered significantly different with p,0.05. Genotype of pups was determined using gDNA extracts from tail biopsies performed during brain dissection (see Animals section for genotyping details).
Behavioral analysis of Runx2 +/2 mice Behavioral entrainment of Runx2 +/2 mice to a LD schedule and ability to establish stable free-running activity rhythms in constant darkness (DD) were assessed and compared to age and sexmatched Runx2 +/+ littermates. Behavioral analyses of Runx2 2/2 mice could not be evaluated due to the fact that these animals lack proper ossification and do not survive past post-natal day 1 [21]. For phase shifting studies, animals were subjected to an Aschoff type II phase shifting paradigm (Mrosovsky, 1996). In brief, animals were initially housed under a strict 12 h:12 h LD schedule for 7 days, subsequently released into constant dim red illumination at ZT12 (lux,5), and left undisturbed (control) or given a 15 min light pulse (same intensity as for light period during 12 h:12 h LD entrainment) at CT6, CT15, or CT21. Following each manipulation, the animals remained in DD for 7 days. All animals were exposed to each treatment and were allowed to stably re-entrain for at least 7 days prior to the each photic pulse.
Phase shift responses to light were calculated from running wheel activity records. CLOCKLAB software (Actimetrics) was used to detect the daily onset of wheel running and to fit a regression line to the 7 daily onsets prior to the photic manipulation. Phase shift magnitude (hours 6 SEM) was calculated by subtracting the observed activity onset on day 6 (phase advance) or day 3 (phase delay) of DD from the predicted onset prior to light treatment. Positive values represent phase advances, negative values represent phase delays. To control for individual variability in free-running period, the magnitude of the phase shift produced by the pulse treatments was subtracted by the inherent shift observed in each animal following the no light control condition. Mean normalized phase shift values for each treatment and genotype group were determined and comparisons between genotypes for a given photic manipulation were conducted using one-way Anova and paired t-test analyses.

Expression of Runx2 in the mouse brain
Given the limited description of the expression of Runx2 in the adult mammalian brain [19], we first established whether detectable levels of Runx2 are expressed in the adult C57BL/6 mouse SCN. Qualitative RT-PCR assays validated expression of Runx2 in this brain region ( Figure 1A). Consistent with previous observations that expression of Runx1 is ubiquitously expressed in the mammalian central nervous system in contrast to Runx3, another member of this transcription factor family, RT-PCR assays detected expression of Runx1 in SCN samples; however, Runx3 was undetectable. In addition, Osteocalcin, a known target of Runx2 transcription factor activity [30], was detected in the SCN (not shown). However, a second well established target of Runx2 in osteoblasts, Osteopontin [31], was not found to be expressed in the SCN. These findings suggest that the Runx2 gene product might have distinct functionality in the murine SCN compared to bone.
Next, we determined whether RUNX2 protein was detectable in the SCN of C57BL/6 mice. Western blot analyses of protein extracts from the SCN of adult mice resulted in a band of immunoreactivity at the predicted molecular weight of RUNX2 (,60 kDa) ( Figure 1B) with a corresponding band also present in tissue isolated from mouse calvaria on embryonic day 15 mice [32][33][34]. In order to determine if RUNX2 is expressed by neurons, coronal sections of fixed mouse brain tissue were immunolabeled for RUNX2 protein and the neuronal nuclei marker, NeuN [35]. Confocal analysis revealed that RUNX2-positive staining was exclusively localized to NeuN-positive cells in all brain areas ( Figure 1C). As others have observed Runx2 promoter activity in several brain structures including the cortex and hippocampus [19], we also determined whether RUNX2 protein was detectable in these and other brain areas. A qualitative analysis of RUNX2 distribution in the brain revealed that RUNX2 is not homogenously expressed. Specifically, dense immunolabeling was observed in all areas of the cortex (Figure 2A), except piriform cortex, where labeling was light to moderate. Dense labeling was also observed throughout the thalamus ( Figure 2D) and in numerous areas within the hypothalamus, including the paraventricular nucleus of the hypothalamus (PVN), SCN, ventromedial hypothalamus ( Figure 2F), preoptic area, anterior and later hypothalamus, but was absent in the arcuate nucleus and light in dorsomedial nucleus of the hypothalamus. Closer analysis of immunolabeling in the SCN and PVN revealed homogenous distribution of RUNX2 in both of these areas important in regulation of circadian functions (Figure 3). Moderate labeling was evident in the amygdala ( Figure 2E), but was light or absent in the bed nucleus of the stria terminalis and lateral septum. RUNX2 immunolabeling was absent or light throughout the dorsal and ventral striatum ( Figure 2B), including caudate putamen, nucleus accumbens, and globus pallidus. RUNX2 was present in the hippocampus, with dense labeling in CA3, and light scattered labeling in CA1-2 and dentate gyrus ( Figure 2C). Finally, immunolabeling for RUNX2 was dense in most areas of the midbrain and brainstem, including ventral tegmental area and substantia nigra, dorsal raphe and red nucleus. Runx2 gene expression is rhythmic in the mouse brain To examine the circadian regulation of Runx2, gene expression was examined in the SCN and several other brain regions previously shown to be rhythmic under both LD and DD conditions [36]. In the SCN, under LD conditions, there was a significant main effect of time of day upon Runx2 expression (F (5, 20) = 5.652; p,0.01) (Figure 4). Post hoc analyses revealed significantly greater expression of Runx2 during the early to midday than during the night (ZT1 vs. ZT17 and 21, p = 0.03 and 0.009; ZT5 vs. ZT13, 17 and 21, p = 0.025, 0.009, 0.0017; Figure 4A). The rhythmic expression of Per2 (F (5, 21) = 6.439; p,0.005) was in phase with the Runx2 rhythm, with peak levels observed during midday (ZT5 vs. ZT13 and 21, p = 0.004 and 0.0014; Figure 4A). Rhythmic expression of Bmal1 (F (5, 21) = 14.171; p = 0.015) was in antiphase compared to the expression of both Per2 and Runx2, with greater levels of Bmal1 expression occurring during the night than day (ZT5 vs. ZT17-21, p = 0.002; ZT5 vs. ZT13, p = 0.05; Figure 4A). Under DD conditions, Runx2 expression in the SCN was also rhythmic (F (5, 21) = 6.117; p = ,0.005) with peak levels in the early to midday (CT1-5) and a nadir in the mid to late night (CT1 vs. CT17 and 21, p = 0.004 and 0.0016; CT5 vs. CT17 and 21, p = 0.004 and 0.003; Figure 4B). Likewise, the Per2 expression rhythm (F (5, 20) = 16.509; p = ,0.001) was similar in phase to that observed for Runx2 (peak levels observed during early to midday; CT1 vs. CT13, 17 and 21, p = 0.0007, 0.0011 and 0.0033; CT5 vs. CT13, 17 and 21, p = 0.0003, 0.0001 and 0.0003) and antiphase to the rhythm of Bmal1 (peak in late night; CT1 vs. CT13, 17 and 21, p = 0.000004, 0.0008 and 0.009; CT5 vs. CT13, 17 and 21, p = 0.00005, 0.0005 and 0.0123; CT9 vs. CT13 and 17, p = 0.0002 and 0.007; Figure 4B). In addition, another member of the period gene family, Per1, was also expressed in phase with Runx2, with peak levels observed during early to midday (CT1 vs.  Figure 5A and B). Per2 expression mirrored that of Runx2 with greatest levels detected during mid to late day and nadirs observed during the night under both conditions ( Figure 5A and B). The peaks in Runx2 and Per2 expression in the PVN were phase delayed by 4 h compared to the SCN ( Figure 4A and B, 5A and B). In contrast to the SCN, Bmal1 expression was not rhythmic in the PVN under LD conditions (F (5, 20) = 1.856, p = 0.162) with a trend for rhythmicity in DD (F (5, 19) = 3.041, p = 0.058; Figure 5A and B). As in the SCN, Bmal1 appeared to be in antiphase to Per2 and Runx2 in the PVN under DD conditions ( Figure 5B).
The phase relationship between Runx2 and Per2 expression observed in the OB under LD conditions was similar to that seen in the SCN and PVN but was not as tightly linked when examined under DD. Specifically, the rhythmic expression of Per2 rhythm in the OB under LD conditions (F (5, 22) = 8.455, p = ,0.001) was similar to that of Runx2 ( Figure 5C). However, the Per2 expression peak was phase delayed by 4 h compared to the SCN while the Runx2 peak did not shift ( Figure 5C). In DD, Runx2 expression was shifted by 4 h compared to detected levels in the SCN and elevated levels were extended over the mid day to the early night in the OB in DD ( Figure 5D). In contrast, the peak in Per2 rhythm (F (5, 21) = 7.406, p = ,0.001) was delayed by 8 h under DD  Figure 5E and F). As observed in the PVN, the peak in Per2 expression was shifted by 4 h in the HP compared to the SCN. This peak was 4 h earlier than the peak in Runx2 expression, demonstrating a difference in phase between Per2 and Runx2 in the HP under LD conditions ( Figure 5E). Bmal1 expression appeared to be in antiphase to Per2 under LD conditions but displayed only a trend toward rhythmicity in this brain area (F (5, 20) = 2.675, p = 0.06; Figure 5E). Per2 expression was also rhythmic in the HP under DD conditions (F (5, 20) = 8.206, p = ,0.001) and displayed a similar phase in expression to Runx2, with shifts in peak levels of expression by 4-8 h compared to the SCN. Expression levels of Runx2 remained elevated for a longer interval than Per2 levels ( Figure 5F). This was opposite of what was observed under LD conditions wherein Per2 had prolonged elevated levels of detectible expression compared to a single clear peak for Runx2 ( Figure 5E).

Runx2 protein expression is rhythmic in the mouse SCN
Next, immunohistochemistry was used to determine whether RUNX2 protein, like its mRNA, was rhythmic in the SCN. In the SCN, there was a significant effect of time of day on levels of RUNX2 immunoreactivity (F (5, 43) = 17.563, p = 0.042; Figure 6), under DD conditions. Optical density levels were significantly greater at CT9 than at any time assessed during subjective night (CT9 vs. CT13, 17 and 21, p = 0.009, 0.032 and 0.003). The time of greatest optical density of RUNX2 immunoreactivity corresponded to ,4 h following peak Runx2 gene expression levels in the SCN of animals housed in either LD or DD conditions (Figure 4). Overall protein concentration of RUNX2 in the SCN at the peak and nadir times of immunoreactivity density were also found to be significantly different as determined by western blot analyses (CT9 vs. CT21, p = 0.023).
Rhythms in the optical density of RUNX2 immunostaining were also observed in the PVN (Figure 6). In the PVN (F (5, 31) = 11.019, p = ,0.01), peaks in optical density were seen in the early and late night compared to all times during the subjective day (CT13 vs. CT1, 5, 9, and 1, p = 0.002, 0.0006, 0.0056 and 0.0069; Figure 6). As in the SCN, the initial peak in optical density of RUNX2 immunostaining in the PVN occurred ,4 h following the observed peak in Runx2 gene expression ( Figure 5).

Rhythmic expression of Runx2 is absent in the SCN of Bmal1 knockout mice
The similarity in phase observed between Runx2 and Per1/2 gene expression rhythms suggested that Runx2 may be a clockcontrolled gene in neurons. Therefore, to determine whether rhythmic Runx2 expression was dependent on a functional molecular clock system, gene expression within the SCN of mice lacking one or both functional Bmal1 alleles were assayed at ZT/ CT5 and ZT/CT17. Specifically, animals were transferred into DD at ZT12 and sacrificed at 17 h and 29 h later, reflecting ZT/ CT5 and ZT/CT17 in the following circadian cycle. True circadian time cannot be established in these animals because their behavior becomes arrhythmic in the absence of a daily light cue. Since photic stimulation of the SCN may itself stimulate expression of Runx2, animals were sampled within the first day of DD without photic stimulation in effort to sample tissues prior to the breakdown of entrained rhythms. The times selected represent times of peak and nadir levels of expression of Runx2 and clock genes as observed in WT animals ( Figure 4).
In homozygote Bmal1 2/2 mice, the time of day difference in Runx2 gene expression was completely absent (CT5 vs. CT17, p = 0.147; Figure 7). As expected, time of day differences in gene expression of Per1 and Per2 were also absent in the SCN of Bmal1 2/2 mice (p = 0.388 and 0.867, respectively; Figure 6). In contrast, time of day differences in fold changes in gene expression of Per1, Per2 and Bmal1 observed in heterozygote Bmal1 +/2 mice mirrored those observed in WT animals [20]. Similar to previous reports, expression of both Period genes, Per1 and Per2, were significantly greater at ZT5 than ZT17 (p = 0.012 and 0.00001, respectively) and the expression of Runx2 was observed to be greater at ZT5 compared to ZT17 (p = 0.0007) in the SCN of Bmal1 +/2 mice (Figure 7) [20]. Together, these findings suggest that the circadian rhythms in Runx2 gene expression within the mouse SCN are dependent on expression of a functional Bmal1 gene.
Circadian rhythms of mPer2 Luc -mediated bioluminescence in isolated SCN from E19/20 Runx2 +/+ , Runx2 +/2 and Runx2 2/2 mice In order to determine whether a complete absence of Runx2 alters SCN circadian clock gene rhythms, Runx2 +/2 mice homozygous for the mPer2 luc allele were generated in order to measure molecular rhythms using real-time monitoring of in vitro bioluminescence from the SCN. Tissues were isolated on embryonic day 19-20 (E19/20) thus circumventing the neonatal lethality of the homozygote Runx2 2/2 genotype [21].
Overall, there was a significant difference in the mean period length of mPer2 Luc rhythms in SCN cultures from Runx2 +/+ , Runx2 +/2 and Runx2 2/2 mice (F (2, 28) = 3.537, p = 0.043; Figure 8). Post hoc analyses confirmed that the mean period of the mPer2 Luc rhythms in the SCN of Runx2 2/2 E19/20 mice (n = 7, 25.360.26 h) was significantly longer than those observed in the SCN of either Runx2 +/2 (n = 13, 24.2060.26 h, p = 0.039) or Runx2 +/+ (n = 11, 23.9260.34 h, p = 0.047) littermates ( Figure 8A and B). No significant difference in the mean period of the mPer2 Luc rhythms was seen between Runx2 +/2 and Runx2 +/+ littermates (p = 0.328). We did not observe a difference in the mean amplitude of bioluminescence levels between genotypes (p.0.1). There were no significant differences in any of the amplitude parameters, but this may be partially due to variability between cultured samples. Moreover, in liver tissue which does not express Runx2, there was no difference in the mean period of bioluminescence between genotypes (data not shown; p = 0.473).
Runx2 +/2 mice display lengthened period of free-running activity To determine whether Runx2 deficiency leads to in vivo consequences in circadian behavior, the wheel-running rhythms of Runx2 +/2 mice were compared to those of Runx2 +/+ mice. As noted above, heterozygote animals were used since Runx2-null mice are postnatal lethal [21].
Exposure to environmental cues, such as light, can induce phase-dependent shifts in activity rhythms. To investigate whether Runx2 plays a role in photic resetting of the clock, Runx2 +/+ and Runx2 +/2 mice were subjected to phase shifting light pulses during either subjective day, or early or late subjective night. As expected, running wheel activity was significantly shifted following exposure to a light pulse during late (CT21 vs. no pulse; p = 0.000003) or early subjective night (CT 15 vs. no pulse; p = 0.0005) but not to light exposure during subjective day (CT6 vs. no pulse; p = 0.704; Table 1). Genotype comparisons revealed that direction and amplitude of phase shifts in response to light pulses at these different circadian times did not differ between Runx2 +/2 and  Runx2 +/+ littermates (CT6 pulse, p = 0.566; CT15 pulse, p = 0.975; CT21 pulse, p = 0.159; Table 1).

Discussion
We found that Runx2 mRNA and protein expression is rhythmic and under the control of circadian clock network in the SCN, as well as in several other brain regions previously shown to exhibit rhythmic clock gene expression [36,37]. In addition, Runx2 appears to play a functional role in the regulation of cellular and behavioral rhythms, with lengthened free-running period being observed in Runx2 mutant SCN tissues and animals.
In the SCN, the phase of peak Runx2 expression was similar to that of Per1 and Per2 suggesting the possibility that Runx2, like Per1 and Per2, may be a target for Bmal1/Clock transactivation. Indeed, in the absence of a functional Bmal1 gene product, a rhythm in Runx2 mRNA in the SCN of mice housed under DD conditions was undetectable, again mirroring the effects on Per1 and Per2. This effect of the Bmal1 knockout on Runx2 gene expression suggests that Runx2 is a clock-controlled gene dependent on Bmal1 activity. Consistent with this, a putative Bmal1response element (E-box) has been described in the region upstream of the Runx2 promoter, although the functionality of this site remains to be validated [38]. A comprehensive study of the promoter regions of Runx2 would provide greater insight into the possible mechanisms of Bmal1-mediated Runx2 expression [39].
Alternatively, rhythmic expression of the Runx2 gene may not be directly mediated by Bmal1 but rather under the transcription control of one or several other CCGs. One such CCG, known to be rhythmically expressed in the SCN, is transforming growth factor ß (TGF-ß) [40][41][42]. In osseous tissues, Runx2 expression and activity is repressed by the signalling cascade activated by TGF-ß [41,42]. Since TGF-b expression and signalling activity are elevated during the subjective night in the SCN, a time when Runx2 expression is at its lowest, TGF-b may be having a similar repressive effect on the Runx2 transcription in the SCN [40]. Given the evidence for the non-photic phase shifting properties of TGF-ß signalling in peripheral tissues [43], it is interesting to speculate its capacity to do so in the SCN is attributable to its effects on Runx2.
The results of the in vitro and in vivo experiments reported here suggest that Runx2 participates in regulating circadian period length through its influence on molecular and/or signalling pathways occurring at the level of the SCN. Within the SCN and other tissues, period length is largely due to the timing of intracellular molecular rhythms [44], and there are several possible mechanisms by which Runx2 may act to lengthen the period of cellular rhythms. First, the timing of the intracellular rhythms may be modified by actions of Runx2 on transcription of clock genes or their regulators leading to a lengthened period of the individual cellular oscillators comprising the SCN. The runtdomain of the Runx2 protein binds DNA in a site-specific manner by recognizing sequences identified as of 59-puACCpuCA-39 [45]. A cursory analysis identified a number of putative Runx2 binding sites within the sequences of clock genes including Bmal1 and Per2 (NCBI reference sequence: AC_000029.1, NC_000067). Further studies are needed to identify functional Runx2 binding sites in clock gene sequences and to determine the influence of Runx2 on the expression of these genes. Runx2 also contains a number of protein-protein interaction domains that act to recruit proteins to downstream gene targets, including transcription activators (i.e. p300, CCAAT/enhancer binding proteins) and repressors (i.e. HDACs) [46][47][48]. The diversity of the multi-protein complexes formed through Runx2-mediated scaffolding is further increased by a number of potential post-transcriptional modifications to Runx2 [49,50]. Thus, in addition to acting directly through its DNA-binding sequences, Runx2 may act indirectly to regulate transcription of clock genes and/or CCGs by facilitating recruitment of other factors that, in turn, serve to facilitate or repress target genes.
The period lengthening effect of Runx2 +/2 observed in vivo was also observed in the period of the bioluminescence rhythms detected from cultured SCN tissues collected from Runx2 2/2 mice. Unlike activity rhythms measured in vivo, no significant change in the period length was detected in Runx2 +/2 SCN cultures. However, when the mean period length value of in vitro bioluminescence rhythms measure from the SCN tissues of Runx2 +/2 mice is compared to the mean period of free-running rhythms of Runx2 +/2 mice, both show a similar magnitude of period lengthening (16.8 min in vitro vs.14.4 min in vivo). The lack of statistical significance between the periods of molecular rhythms measured from Runx2 +/+ and Runx2 +/2 SCN tissues therefore likely lies in the variability of the measured bioluminescence rhythms in vitro. This variability could be due to a number of factors including a relatively smaller sample number for each group under in vitro conditions versus the in vivo sample sizes or the discrepancies in period lengths measured from tissue cultures and behavioral parameters [51][52][53][54]. This latter phenomenon was observed in the animals harboring the mutations in the circadian Figure 8. Mean period of mPer2 Luc -mediated bioluminescence rhythms is lengthened in the SCN of Runx2 deficient mice. A) Mean period (6 SEM) of mPer2 Luc rhythms in SCN cultures of Runx2 +/+ (black), Runx2 +/2 (grey) and Runx2 2/2 (white) determined from realtime monitoring of bioluminescence from isolated E19/20 mice. Inlayed number represents n value of experimental group. B) Representative trace of mPer2 Luc -mediated bioluminescence observed over several cycles from Runx2 +/+ (black), Runx2 +/2 (grey) and Runx2 2/2 (dashed) SCN cultures. *: significant difference in levels of gene expression between times of day; p,0.05. doi:10.1371/journal.pone.0054317.g008 clock factors, Clock and CK1e [54,55]. Previous studies correlating period in behavioral rhythms in vivo to molecular rhythms in vitro have examined animals harboring mutations with dramatic differences in period length [56][57][58] and under conditions where both gene alleles are compromised. Another factor may have been the differences in age of the mice in the in vivo (adult) and in vitro experiments (embryonic). Thus it is possible that the effects we observed in haplodeficient animals are only manifested in adulthood perhaps due to compensatory mechanism(s) during development that either wane or are insufficient later in life.
Beyond the SCN, detectible expression levels of Runx2 were rhythmic in the PVN, OB and hippocampus. The phase of the Runx2 rhythm was shifted by ,4 h in each of these brain regions under DD conditions compared to its phase in the SCN. This phase lag may be of physiological importance since the synchronization of oscillators to one another may be critical in coordinating neural activity and responses to external cues (i.e. light, food). Such a phase lag (3-9 h) relative to their maximal expression in the SCN is typical of genes expressed in peripheral tissues whose circadian oscillations are driven and/or synchronized by a signal led by the circadian pacemaker [22,59]. Support for SCN-influenced expression of Runx2 in these peripheral tissues can be derived from the observation that shifts in the phase of Runx2 oscillations were similar to the lag observed for the circadian clock gene, Per2. Whether the oscillations of Runx2 in these brain regions are driven and/or synchronized by the SCN remain to be elucidated. It may be that the rhythmic expression of Runx2 in the PVN, a target of SCN efferents, is driven by the SCN and would become arrhythmic in its absence [36,[60][61][62][63]. In contrast, the OB has been convincingly shown to contain independent, selfsustained oscillators that do not rely on the SCN [64,65]. Since rhythms in the OB are synchronized rather than driven by the SCN, oscillations in Runx2 in the OB would likely persist in SCNlesioned animals [65]. Rhythmic expression of clock genes, Figure 9. Free-running period of running wheel activity is lengthened in Runx2 +/2 mice. A) Mean (6 SEM) free-running period of Runx2 +/+ (grey) and Runx2 +/2 (white) mice. Inlayed number represents n value of experimental group. B) Mean 6 SEM wheel revolutions of male Runx2 +/+ (grey) and male Runx2 +/2 (white) entrained to a 12:12 h LD schedule. Inlayed number represents n value of experimental group. C) Representative running wheel activity records (upper panels) for Runx2 +/+ (left panels) and Runx2 +/2 (right panels) mice. Activity records are in single-plotted format with each horizontal line represents a 24 h period. In 12 h:12 h LD conditions, time of day is represented by white and black horizontal bars indicating periods of day and night, respectively, with dark onset marking ZT12. Black arrow with white asterisk: transition into 24 h DD. *: significant difference between genotypes; p,0.05. doi:10.1371/journal.pone.0054317.g009 Table 1. Runx2 +/2 mice display responses to phase shifting light pulses indistinguishable from Runx2 +/+ mice.

Genotype
Light Pulse at CT6 Light Pulse at CT15 Light Pulse at CT21 including Bmal1 and Per2, has previously been described within the HP; however, the dependency of these rhythms on the SCN has not been thoroughly evaluated [37,66]. Finally, Runx2 protein was found to expressed in numerous other areas within the brain, as are clock genes suggesting that Runx2 may influence circadian modulation of a variety of physiological and behavioral functions [24,37,66]. Runx2 deficiency is known to have clinical consequences in skeletal development and bone homeostasis. Cleidocranial dysplasia (Online Mendelian Inheritance in Man # 119600) is a rare genetic disorder resulting from an autosomal dominant mutation(s) or complete loss of a functional Runx2 gene copy [67]. In general, patients with this disorder have abnormal osteogenic development and present with a number of skeletal and dental deformities which vary in severity depending on the causal gene mutation [68]. To date, no circadian or sleep disorders have been linked with CCD. Given our findings on the effects of Runx2 on period length, it would be worthwhile to assess the status of circadian biomarkers and sleep patterns in individuals with CCD to determine if the effects of Runx2 haplodeficiency we observed in mice translate to humans.
In summary, Runx2 appears to be a circadian clock-controlled gene that is expressed rhythmically in the SCN and other regions of the murine brain. This expression occurs in a circadian fashion with greatest levels of expression occurring during the subjective day and this rhythm is dependent on the expression of Bmal1 suggesting that Runx2 is a circadian controlled gene. Decreased Runx2 gene expression leads to significant lengthening of the period of behavioral rhythms in vivo and lengthened mPer2 Luc rhythms in vitro. Future studies will be needed to identify the precise mechanisms by which Runx2 is regulated by circadian clock genes within the SCN as well as in bone and other peripheral tissues, and its potential impact on other clock-controlled genes.