Cardiac Atrial Circadian Rhythms in PERIOD2::LUCIFERASE and per1:luc Mice: Amplitude and Phase Responses to Glucocorticoid Signaling and Medium Treatment

In mammals, many processes in physiology show daily variation under normal conditions. When these daily patterns persist under constant conditions, they are described as circadian (∼ 24 hours) rhythms, and are driven by an endogenous clock. Circadian rhythms are governed by internal clocks that, in the presence of timing cues, are synchronized (entrained) to the external environment [1]. The central light-entrainable clock resides in the suprachiasmatic nucleus (SCN) of the hypothalamus, and the SCN transmits timing information in a hierarchical manner to clocks found in peripheral tissues [2]. Temporal orchestration of cellular processes is seen in many peripheral tissues, such as in the liver and cardiovascular tissue, and it is thought that these tissues use a similar molecular clock mechanism as described for the SCN.

For the heart, diel (observed under light:dark conditions) and circadian rhythms are apparent in physiology and behavior. Peaks in both arterial blood pressure and heart rate in mice are observed when they are active, around Zeitgeber time 15 (ZT; ZT 0 = lights on, ZT 12 = lights off) and at ZT 12, respectively [3]. While behavioral state is an important causal factor to diurnal variation in blood pressure and heart rate, there is a significant contribution of a functional circadian clock to this rhythm [4]. A temporal profile with an early morning peak in acute adverse cardiac events and sudden cardiac death is observed in man [5], [6] and cardiac vulnerability has been found to have a circadian component that is independent of behavior [7]. Interestingly, tau mutant hamsters, that exhibit significantly reduced intrinsic day-length, show significant cardiac hypotrophy, reduced blood pressure and impaired myocardial contractility [8]. Also, mice that have a functional knockout of the gene encoding vasoactive intestinal polypeptide, a major SCN neuropeptide, show dampened rhythms in heart rate under entrained conditions and loss of this rhythm under constant darkness conditions [9].

A defining characteristic of circadian clocks is the expression of canonical clock genes (e.g. clock, bmal1, per and cry genes) which, through regulatory feedback loops, orchestrate each other’s activity, resulting in rhythmically active transcriptional loops [10]. Cardiac tissue has been shown to exhibit rhythmic gene expression, including the expression of canonical clock genes, in vivo both in the presence of an entraining light-dark (LD) cycle and under constant conditions [11]–[14]. In vitro cultures of cardiac tissue also show robust rhythmic expression of per1 as reported by bioluminescence, indicating a functional circadian clock in the heart [15], [16]. Circadian variation in cardiac function persists in isolated heart tissue, highlighting the cell-autonomous activity of the cardiac clock. For example, rat hearts in vitro exhibit 24 hr variation in contractile responses, oxidative metabolism, oxidative stress tolerance and lipid peroxidation, and rhythmic electrochemical activity is maintained in isolated myocytes [17], [18]. Disruption of the circadian clock in the heart results in decreased diurnal variations in heart rate, sinus bradycardia, loss of diurnal variations in cardiac power and responsiveness of the heart to changes in workload, increased fatty acid oxidation, decreased cardiac efficiency, as well as other cellular functions [19].

Clocks in peripheral tissues are entrained in part, by the central light entrainable clock in the SCN, and there are factors in serum that can reset peripheral clocks such as in liver and fibroblasts cells [20]–[23]. Glucocortcoids (GCs) have been implicated as being important in synchronizing clocks in peripheral tissues in vivo and in vitro[24]. Indeed, there is evidence for direct binding of the glucocorticoid receptor (GR) to per1 and per2 genes, and binding of the GR to per2 is essential for amplitude regulation by the synthetic glucocorticoid dexamethasone (DEX) in in vitro cultures of mesenchymal stem cells, and maybe even for the maintenance of a peripheral rhythm per se[13], [25]. In response to acute stress, per1 mRNA, but not per2 mRNA is readily upregulated in the heart, and this is subject to the presence of the GR binding sequence in the promoter region of the per1 gene [13].

GCs are steroid hormones that are released by the adrenal gland. The release of GCs is under the control of the hypothalamic-pituitary-adrenal (HPA) axis and GCs show a daily rhythm in circulating plasma levels. Plasma levels of corticosterone, a major GC, peak at the onset of activity, both in nocturnal and diurnal subjects. As the main light-entrained circadian oscillator, the SCN drives rhythms in plasma GCs, and when the SCN is lesioned, the daily rhythm in circulating corticosterone is ablated [26]. The SCN acts on the adrenal gland, both through direct sympathetic connections to the adrenal cortex, and through modulating adrenocorticotropic hormone (ACTH) release via rhythmic release of corticotrophin-releasing hormone from the hypothalamic paraventricular nucleus (For review see [27]). The sympathetic SCN-adrenal pathway is hypothesized to control the adrenal sensitivity to circulating ACTH [28]. Besides SCN control of daily rhythms in GCs levels, there is also a role of clocks in the adrenal gland itself.

Most canonical clock genes show rhythmic expression in the adrenal gland and the adrenal clock regulates the sensitivity of the adrenal gland to ACTH, varying with time-of-day [29]. Moreover, the local adrenal clock has been reported to cause rhythmic expression of adrenal steroid production, and absence of the adrenal clock reduces daily amplitude of steroid production [30]. This reduced amplitude could be indicative of a role of GCs in synchronizing clocks in peripheral tissues [24], where desynchronized tissue cause an attenuation of overall amplitude of the clock and clock output.

Circadian rhythms are apparent in cardiac tissue at the molecular and functional level. It has been shown previously that DEX can affect clock gene expression in the heart [24], although this has been tested only at one particular time of day. Because of the reported susceptibility of clock genes to GCs, we set out to study the response of atrial tissue to dexamethasone (DEX) in a time informed manner. Using in vitro atrial tissue cultures of mice carrying bioluminescence reporters for per1 and PER2 activity, we monitored the circadian clock in this tissue, and it’s response to DEX treatment. In vitro culture systems allow for true examination of tissue-specific circadian responses independent of extrinsic systemic factors. In these studies we specifically focused on the cardiac atria. The cardiac atria are multifaceted structures that function in the circulatory system to promote continuous venous flow; they contain pacemaker elements required for timing of heart rate; and function as an endocrine organ, producing atrial natriuretic peptide in response to stretching of the atrial vessel walls associated with elevated blood pressure or volume [31], [32]. Because DEX reportedly does not shift the rhythm of per1 expression in the SCN (although only tested at one time of day), but has a known response in liver [24], we also cultured SCN and liver of the same individuals.

The peak phase of the circadian rhythm of in vitro bioluminescence shows variation as a result of the time of sacrifice and subsequent culturing. Figure 1 shows that in cultures of the SCN, liver and especially atrial tissues the phase of the first PER2^LUC driven peak in vitro is different between groups with different times of killing and culturing (One way ANOVA; SCN, F_3,26 = 18.442; liver, F_3,74 = 5.384; atria, F_3,101 = 499.783; P’s <0.002 for culture start time). In the case of atrial tissue, the peak phase of PER2^LUC bioluminescence rhythm is actually different between each of the four culture start times. In order to test for a possible ‘resetting’ of the clock, the time interval between start of culturing and the first peak was calculated, and shown to be different for different times of sacrifice for each of the tissues (Kruskal–Wallis one-way ANOVA; SCN, H₃ = 16.002; liver, H₃ = 56.333; atria, H₃ = 70.306; P’s <0.001 for cultures start time). These data indicate that an in vivo treatment followed by in vitro culturing at different times of day is not a suitable approach for testing the effects of DEX on atrial or other tissues in a time-informed manner, we therefore continued with an in vitro treatment approach, where time of treatment is determined by the phase of bioluminescence after culturing.

Using this in vitro approach we first cultured tissues and then subjected them to control or experimental conditions. Representative baseline subtracted traces for per1 and PER2 driven bioluminescence are shown in Figure 2A and 2B respectively. Traces are shown for SCN, liver and atrial cultures that received no treatment, or were treated with medium or DEX. As a first observation, no rhythms in per1^luc bioluminescence were detectable in the atria. In contrast, PER2^LUC bioluminescence was rhythmic in all tissues. This observation is quantified in Table 1, which indicates the significant lack of rhythmic cultures of per1^luc bioluminescence as compared to PER2^LUC bioluminescence (chi-square, χ₁ = 134.111, P<0.001). The fraction of cultures that showed rhythmicity was 97% for the SCN and 82% for liver cultures and was not different between per1 and PER2 driven bioluminescence (chi-square, χ₁ = 1.525 and χ₁ = 0.486 respectively, P’s >0.05).

As a first response to treatment, we quantified the change in the amplitude of the oscillation in bioluminescence in response to no treatment, and medium or DEX treatment. DEX treatment increased the amplitude of the per1^luc bioluminescence oscillation in cultures of liver, but not SCN (two way ANOVA, effect of tissue F_1,23 = 35.208; effect of treatment F_2,23 = 21.292; interaction effect F_2,23 = 16.071, P’s <0.001), an amplitude increase in atrium is apparent, but could not be quantified due to absence of a rhythm before treatment (Figure S1). Although already slightly increased in response to medium treatment, DEX treatment resulted in a 4-fold increase in amplitude, which was not dependent on time of treatment (Holm-Sidak; P>0.05; data not shown). The amplitude increase for atrial tissue in response to DEX was not quantifiable due to a lack of pre-treatment rhythmicity. Because of the lack of oscillations in atrial per1^luc bioluminescence, we continued to look at PER2^LUC bioluminescence. Figure 2C shows amplitude responses in PER2^LUC bioluminescence in SCN, liver and atrial cultures, where only in atrial tissue a significant increase in amplitude was observed (two way ANOVA, effect of tissue F_2,200 = 8.235, P<0.001; effect of treatment F_2,200 = 5.858, P<0.005; interaction effect F_4,200 = 11.725, P<0.001). The post hoc contrasts indicated that the only significant change in post-treatment amplitude is observed in the atrium after DEX treatment (Holm-Sidak; P<0.001 for DEX versus no treatment and medium treated). Interestingly, while there is an overall >3-fold increase in amplitude, the response of the atrium to DEX is time dependant, in contrast to the medium treatment (Figure 2D; PROC mixed: effect of treatment, F_1,51 = 47.78, P<0.0001; effect of time, F_21,51 = 2.04, P<0.05; interaction effect, F_19,51 = 1.79, P = 0.05), where DEX treatment shows a significant increase versus medium treatment when treated 2, 3, 5, 6, 9 & 10 hours after the last peak in bioluminescence. While there were already no significant increases in overall amplitude responses in liver and SCN, testing for differential treatment, time-of-day, or interaction effect for either SCN or liver tissue, revealed no significant effects (P’s >0.05), indicating that the overall absence of an amplitude response was not masking a time-of-day specific amplitude response in these tissue.

Not only amplitude, but also phase of the rhythms of tissues was affected by medium and DEX treatment in vitro. Figure 3A shows the phase shift response of individual atrial cultures when treated in vitro with medium or DEX at different time intervals after the peak in PER2^LUC bioluminescence. Within 0 to 9 hours following the peak phase of the rhythm, phase delays were observed in the atrium, whereas 9 to 15 hours after the peak phase, shifts in the rhythm were characterized by phase advances. Between 16 and 18 hours after the peak phase, the phase shifting responses of both medium and DEX consist of minor variations around 0 hr, which is typically referred to as the ‘dead zone’ of a phase response curve. These shifts are binned in 3 hours (Figure 3B) and the absolute magnitude of the phase shifts differs both between different times of day and between the treatment type (two-way ANOVA, effect of time F_7,96 = 6.198, P<0.001; effect of treatment F_1,96 = 5.327, P<0.05 for treatment; non-significant interaction, F_7,96 = 1.730, P>0.05). The phase responses for DEX were different from those of medium, and Figure 3C shows the phase responses that are attributable to DEX through subtraction of medium induced responses. In this format, a dead zone for DEX-specific responses extends from between 16 and 23 hours after the peak phase. Interestingly, there is a limited, but significant positive linear relationship between amplitude and phase response, where a larger amplitude increase after treatment predicts a larger absolute phase shift (Linear regression R² = 0.14; F_{1, 35} = 5.6409, P<0.05).

As a reference, we also looked at the phase shift responses of liver tissue in response to treatment with medium and DEX at different time points centered around 6, 12, 18 and 24 hours after the last peak in bioluminescence. The phase responses are shown in Figure 4A and a significant effect of time of treatment and an interaction between time*treatment was observed (two-way ANOVA: effect of time, F_3,40 = 7.419, P<0.001; effect of treatment, F_1,40 = 0.194, P>0.05; interaction effect, F_3,40 = 12.333, P<0.001). Around 6 hours after the last peak in PER2^LUC bioluminescence, the phase delay in response to medium treatment was significantly different from that of DEX, which caused a large phase advance (Holm-Sidak, P<0.001). When treated 24 hours after the last peak, DEX treatment now resulted in a large phase delay, which is significantly different from the very small phase advance in response to medium treatment (Holm-Sidak, P<0.005). The phase responses in the liver (Figure 4A) were different from those observed in the atrium, which is replotted on the same time-scale in Figure 4B. A significant difference in phase shifts was observed between treatment of atrium and liver with medium and DEX (PROC mixed: interaction effect of tissue, treatment and time F_6,144 = 4.26, P<0.001).The responses to medium treatment differ between atria and liver when treated between 9–15 hours only (LSmeans, t₁₄₄ = 3.34, P<0.005). Contrastingly, DEX treated liver and atria show tissue-specific responses when treated between 3–9 and 9–15 hours (LSmeans, t₁₄₄ = −5.49, P<0.001 and T₁₄₄ = 3.36, P<0.005 respectively, see Figure 4C).

Because the SCN has been reported to not shift in response to DEX treatment, we included SCN cultures as a control. We examined whether there was a phase shift in per1^luc and PER2^LUC bioluminescence in response to medium and DEX treatment. Both treatments resulted in minimal phase shifts (medium, 0.14±0.81 hr; DEX, −0.01±0.94 hr; mean ± SEM) and our data indicate that indeed there was no effect of DEX treatment, as compared to medium treatment (two-way ANOVA: effect of time, F_3,16 = 0.156, P>0.05; effect of treatment, F_1,16 = 1.417, P>0.05; Figure S2).

The large responses to medium treatment, especially in the cultures made from atrial tissue, were striking, and we wished to investigate whether these responses were related to the mechanical movement during the media change procedure and/or serum content or change in gas pressures in the air-tissue/medium interface. We cultured atrial tissue and removed them from the incubator at 3.75±0.28 hours (mean ± SEM; range of 1.0–5.0 hours) after the last peak, a time at which we expect to see a moderate phase delay in response to the media treatment. When we did not treat the atria but simply removed the tissue from the incubator, we observed no net phase shift, 0.16±0.52 hour (mean ± SEM), in PER2^LUC bioluminescence (Figure 5). Similarly, removing the tissue from the incubator but opening the lid of the sealed tissue culture dish to provide for air exchange, did not result in a phase shift comparable to that of medium treatment, −0.90±0.39 hours (mean ± SEM). When we treated the tissue with fresh medium containing 0%, 1% or 5% serum (5% being the standard concentration used in all other experiments) or its own medium, we observed phase delays, and with similar magnitudes, −4.13±0.04 hours (mean ± SEM). The phase delays seen in all four medium treatment groups was significantly different from the limited phase response of untreated (but moved) or lid opened atrial cultures (one-way ANOVA, F_5,102 = 10.836, P<0.001). Post hoc Holm-Sidak t-tests reveal that all four medium treatment groups showed significant phase delay responses to media treatment as compared to the removal from incubator only and lid opened groups (both no medium exchanged or lid opened only versus all media treatments, P≤0.001). Interestingly, no difference was observed in the magnitude of phase shifts between the four different medium groups (P>0.05), suggesting that the change in media type, namely fresh versus original media, or the presence/concentration/absence of serum in the media, is not causative for this phase shift response of the atria. Moreover, the absence of phase shifts in the lid opened group also suggests that the phase shift response is not due to partial pressure changes in O₂, CO₂ and H₂O in the air surrounding the surface of the tissue and culture medium.

Because of the differences observed in the atria in the pronounced occurrence of robust rhythms in PER2 expression but absence of per1 expression rhythms, we wished to examine this phenomenon in another in vitro peripheral tissue system. We used NIH3T3 fibroblast cultures to look at rhythmicity of per1 and per2 gene expression induced by serum treatment (Figure S3). Analysis of per1 revealed an acute effect of the serum treatment (one factor ANOVA for time = 0–48 h, F_13,42 = 2.6, P<0.01; time = 0–8 h, F_3,12 = 7.1, p<0.01) with mRNA levels being induced 1h following serum stimulation (post hoc Dunnett's t-test, p<0.05) and then decreasing back to non-fluctuating levels by 4 h (P<0.05). Analysis of per2 revealed an effect of the serum treatment upon mRNA levels (one factor ANOVA, F_13,68 = 8.4, P<0.001). The levels of per2 were induced transiently peaking 1 h following the start of the serum treatment (post-hoc Dunnett's t-test, p<0.05) and decreased to baseline levels by 8 h post serum (p<0.01), then rose to peak again 24 h later (12 h v. 24 h, p<0.01) and decrease to low levels again by 36 h (p<0.01) rising again at 48 h. These data reveal a significant acute induction of both per1 and per2 gene expression in response to serum treatment, but a significant rhythm was observed for per2 only.

We made one other observation on the peak phasing of the PER2^LUC bioluminescence rhythm in cultures made from atrial tissue. As shown in Figure 6, the mean phase of the PER2^LUC bioluminescence in the right atrium was more than 2 hours (and 7 minutes) later than the left atrium (t-test, t₃₅ = 4.1, P<0.001).

In our initial in vivo approach, we discovered that time of culturing influences post-culturing peak time of the PER2^LUC bioluminescence rhythm. This effect was strongest in atrial tissue, but was also evident in SCN and liver tissue. It is important to note that in the atrial tissue, phases of the PER2^LUC bioluminescence rhythm are shifted within 24-hour range when they were cultured at times with a 12 hour range. This indicates that the clock in the atrium is not ‘simply’ reset by culturing (which would result in a 12 hour range, instead of a 24-hour range of shifts), but that the phases found are the results of shifts in the atrial clock. That the clock is not reset is further corroborated by the fact that the interval between culturing and the first peak is dissimilar between the different times of culturing. Since PER2 represents a state variable of the clock, one might predict other clock components would also be affected similarly, if not immediately, certainly later [33].

Also in atrial cultures from per1:luciferase (per1:luc) rats, an effect of time of culturing has been observed when prepared at two different time points six hours apart [16]. The peak phase of cultured rat per1:luc SCN tissues also shows an effect of time of culturing, with a range of 6 hours [34], which is similar to the range we find in the SCN in the current study in mouse PER2^LUC bioluminescence. In the same study also rat pineal and pituitary (but not cornea) tissue peak phase in vitro exhibits modulation by time of culturing. The susceptibility of tissue to shifting effects due to culturing time appears to correlate with the amplitude of oscillation, which may be indicative of oscillator stability [34]. The large phase shifts that we see in the atrium may indicate that the clock in the atrium is less robust than those in the liver or SCN, but may also indicate a specific atrial sensitivity to mechanical stress (see later).

We continued our study with an in vitro approach, and cultured all tissues at the same time of the circadian day. An initial observation we made was that when we culture the mouse atrium, per1^luc bioluminescence showed no significant circadian rhythm. A 30 minute DEX treatment resulted in a large peak in bioluminescence, and in many cases one or more ‘cycles’ of per1^luc bioluminescence followed the DEX treatment of these atria. In contrast, in rat in vitro cultures of atrial tissue, per1 driven bioluminescence is reported to be rhythmic [16]. So and colleagues [25] also report DEX induced expression of per1 and per2 in mesenchymal stem cells in vitro, but the transcriptional levels of per1 in untreated cells is markedly less rhythmic than per2, if rhythmic at all, which is in line with our results. The induction of an oscillation in constitutively expressed per1 by DEX treatment, is in line with the notion that humeral signals phase-synchronize individual cells within a tissue [35], [36]. These data also suggest that some of the dampening of rhythms in per1^luc expression observed in peripheral tissue explants in culture may be explained by not only a dissociation of coordinated rhythmicity of individual cells but also by the loss of rhythmicity within the individual cells. Since this loss is most apparent in tissues derived from the per1:luc rat and less obvious in the PER2::LUC mouse [21], [37] would be consistent with our data that also show a robust rhythm in atrial PER2^LUC but absence in atrial per1^luc. Additionally, the per1 rhythm observed in vivo in the heart [11]–[14], but which is absent in our study in vitro, is likely to be driven by signals derived from outside of the tissue, i.e. it is a “systematically driven gene” [38]. A limitation to this interpretation is that whilst our study focused on the atria, the in vivo studies examined tissues harvested from the ventricle or whole heart.

Whilst the circadian clock in immortalized fibroblasts has been shown to be synchronized by serum treatment [20], [23], [39], in the current study in mouse NIH3T3 cells we observe a discrete rhythm in per2 gene expression but the absence of a significant rhythm in per1. This is consistent with other studies of synchronized NIH3T3 cells that show either an absence of significant per1 rhythmicity ([40]–[42]; Best, J.D., Duffield, G.E., Loros, J.J. and Dunlap, J.C., unpublished data) or a low amplitude/dampening rhythm [39], [43], [44], whilst simultaneously reporting discrete rhythms in other clock genes, e.g. per2, per3, cry1 and bmal1. Interestingly, this observation is consistent with the results of our PER2^LUC and per1^luc analysis in cultured atria, and provides yet more evidence for cell and tissue-dependent organization of the clock.

We further continued to explore the PER2 expression in atrial, liver and SCN tissue in a time informed manner. While the amplitude of PER2 driven bioluminescence is not significantly altered by DEX treatment in SCN and liver tissue, atrial tissue shows an on average more than 3-fold increase in bioluminescence. The observation that the SCN is not responsive is in line with the literature (see [24]). We are unaware of any reports that report the differential amplitude response of PER2 to DEX treatment in liver and atrium that we observe. On further analysis, we found that the amplitude response was strongly dependent on time of treatment. The amplitude response curve shows a maximal amplitude response when atrial tissue is treated with DEX around the time of intrinsic peaking of bioluminescence, and a low to absent amplitude response to DEX treatment is observed around 18 hours after the last peak.

Both the per1 and per2 genes contain glucocorticoid receptor binding sequences, and for per2, these binding sequences are essential for in vitro amplitude response to DEX in cultured primary mesenchymal stem cells [25]. Whether the amplitude increase in per1 in response to DEX is due to direct stimulation of the glucocorticoid receptor binding sequences in the per1 gene, or is the secondary results of per2 amplitude upregulation in our tissues is unclear. It is curious to see that in the liver per1, but not PER2 amplitude is upregulated in response to DEX treatment, indicating tissue specific regulation of per1 and PER2 genes to DEX, and per1 amplitude regulation independent of PER2. In in vitro cultures of primary mesenchymal stem cells, circadian rhythms in per2 show larger circadian amplitude in transcription, as well as overall increased levels of transcription in response to DEX treatment [25]. Interestingly, a restraint induced stress response, also only upregulates per1 mRNA in the heart and liver, and not per2[13].

Besides amplitude responses, we also looked at phase shift responses of PER2 reporting bioluminescence to medium and DEX treatment. We find that cultures made of atrial tissue show both a response to medium treatment and DEX treatment, but the treatment with DEX is always additive, and the magnitude of the phase response is similar to that reported for in vivo treatment of DEX in the liver [24]. To our knowledge, this is the first report of a time-dependent response to DEX in the atrium, or any other tissue than the liver.

As a positive control we also looked at the PRC to DEX treatment in the liver. We find that also the liver shows clear responses to medium and DEX treatment, but in the liver, the shapes of the respective phase response curves (PRCs) are much more different than they are in the atrium. Importantly, the direction of the phase responses (delays or advances) to DEX in the liver in vitro are as predicted from the in vivo DEX PRC [24], assuming that the peak phase of PER2 represents ∼CT18–19 [45]. In comparison to the atrial PRC to DEX treatment, the liver shows a significantly different type and size of response to DEX treatment. We also used the SCN as a negative control and as predicted this tissue exhibited limited responses to DEX treatment.

Interestingly, our data indicates the presence of a ‘dead zone’ in atrial phase responses to DEX, in which the stimulus does not result in a shift of the rhythm [24]; between 16–23 hours and 16–18 hours after the last peak in bioluminescence the observed DEX phase responses consist of minor variations around 0 hrs (with and without medium responses subtracted, respectively). This is in contrast to the liver, where the PRC for DEX treatment in vivo indicates the absence of such a large ‘dead zone’ [24]. However, our in vitro liver data does in fact show limited DEX induced phase shifts at 12.5–13.5 hours after the last peak in bioluminescence, suggesting the presence of an unresponsive component of the liver DEX PRC.

Medium treatment appears to have a smaller effect in the liver than in the atrium. In the atrium, where the medium treated PRC is relatively large and shows both advances and delays, the medium treated PRC may be have been due to the serum in the medium. However, we found no significant difference in response between 0, 1 and 5% serum in the fresh medium or treating with its own original medium. Also, the similar response to fresh medium and treatment with its own medium indicated that the response is not elicited by a factor that is present in the new medium. Only taking the culture out of the incubator does not result in a similar phase shift response, and opening the lid of the culture dish also does not result in a significant phase shift, suggesting that simply moving the culture, or gas exchange with the external environment is not the cause of the phase shift. Taken together, our results may suggest that some component of the medium treatment procedure, irrespective of the content, causes large shifts in the atrium. One explanation could be that the atrium has a high sensitivity for the mechanical stress it undergoes during the treatment. Such a hypothesis is intriguing, when taking into account the mechanical nature of cardiac tissue and the sensitivity of the atrial cardiomyocytes to stretch, and should be studied further. This hypothesis can also be adopted to explain the effects of timing of culturing (a mechanically stressful procedure), which also results in a larger range of phase shifts in the atrium, as compared to liver and SCN tissue. It may have some functional consequence in the intact animal with clock resetting of the heart, noting that day versus night differences in activity of the heart occur and are often associated with daily temporal exposure to exercise and stress. Furthermore, several components of the intracellular signaling cascade responsible for conveying the atrial myocardial stretch response for atrial natriuretic peptide secretion, are shared with those regulating the circadian clock, e.g. protein kinase C and mitogen-activated protein kinase (MAPK) [22], [32], [46]–[48].

Both the per1 and per2 genes contain glucocorticoid receptor binding sequences, making direct interaction between the glucocorticoid receptor and each gene promoter possible, and may be indicative that per1 and per2 are sites of action for DEX to alter the circadian clock [25]. Using liver specific knockout of the mouse circadian clock, Kornmann et al[38], [49] revealed that rhythmic expression of per2 in the liver, is under the control of both the local clock and systemic cues. This makes per2 an ideal candidate gene for an ‘entry point’ of systemic cues into the local liver clock. Our data indicate that indeed PER2 expression is altered in response to DEX in terms of phase in both the atrium and the liver, while per1 does not show rhythmic expression in the atrium in the absence of external signals. In contrast, per1 appears to be the more responsive of the two genes to DEX in both liver and atrium in terms of amplitude, while PER2 only shows an amplitude response in the atrium.

In the data presented here, we observe tissue dependent differences in per1 and PER2 reported bioluminescence: 1) per1 is not rhythmic in atria, while PER2 is rhythmic; 2) Amplitude responses in PER2 to DEX treatment are absent in liver and SCN, but show a clear time-dependent response in the atrium; 3) The PER2 phase response curve to DEX treatment in the atrium is significantly different from the liver. In addition there is a gene related observation, which is that we observe an amplitude response to DEX treatment in per1, but not PER2 in the liver, while in the atrium both per1 and PER2 respond to DEX treatment with an amplitude increase. These tissue dependent observations are in line with the finding of tissue-dependent effects of per2 and per3 knockout mice [50]. Conversely, it has been shown that blood-borne cues can elicit circadian rhythm in liver and kidneys of animals bearing SCN lesions, but not in heart, spleen, or skeletal muscle of the same animals [51]. These data highlight the diversity and complexity of the local tissue circadian clocks.

As a last observation, we noticed a significant phase difference between the PER2 reporting bioluminescence peaks, where the right atrium is significantly delayed as compared the left atrium. In an earlier study by Davidson et al [16] cultures of the left ventricle appeared to have a delayed phase compared to those of the right atrium in per1-driven bioluminescence in the rat. This finding may indicate that the functional differences in the left and right atria and ventricles is reflected in phase angle differences between the clocks in these cardiac compartments.

The physiological relevance of our findings are several: firstly, the GR specific effect on atrial amplitude of canonical clock components would ensure a robust pacemaker, achieved either through individual cell changes in amplitude, or through increased coordinated rhythms in expression in a population of cells, or a combination of the two (see above). In vivo, normal GR signaling on a daily basis may be an important mechanism in the cardiac system for maintaining normal cardiomyocyte 24 hr rhythmicity. Secondly, whilst the phase resetting effect of GR signalling in the circadian clock has been demonstrated [22], little is known of this mechanism of entrainment beyond the hepatic system. Differences in the phase response curves to DEX between atria (heart) and liver may explain the phase differences in rhythms of certain canonical clock genes observed in vivo between these two tissues [11]. For example, inspection of the phase estimate data in this study [11] reveal per1 and per2 to peak at earlier circadian times in the heart compared to the liver, with a phase difference of approximately 3 hours. Thirdly, the implications of our data are implicit if we consider that DEX and other glucocorticoid drugs are used in the broad treatment of inflammatory and autoimmune conditions [52], [53]: the effects of such treatment paradigms could alter internal circadian coordination between different organ systems and between the heart and the day-night/activity-rest cycle, and therefore might impact cardiovascular health. Fourthly, the apparent sensitivity to mechanical change highlights the possibility that enhanced cardiac activity in vivo, such as during exercise or stress, might in itself result in a resetting of the circadian clock of the atria.

In summary, we find that the clock in atrial tissue shows a strong phase shifting effect in response to time of culturing. Together with the phase shift observed after medium treatment, this could lead to the hypothesis of a time-dependent sensitivity of the atrial clock to mechanical treatment, which is highly relevant for a tissue that shows a daily variation in mechanical (stretch and contractile) activity. Atrial tissue also shows phase shifting responses to DEX, which is markedly different from that of the liver. Moreover, in atrial tissue per1^luc expression is not rhythmic, but can be induced by DEX, while PER2^LUC is rhythmic and shows a strong amplitude response at certain times within the circadian cycle. To our knowledge, this is the first report of a time dependent response to DEX in the atrium, or any tissue other than the liver. These data indicate that the clock in the atrium has a defined glucocorticoid sensitivity and circadian clock characteristics in vitro.

We thank Theresa Sikorski and staff at the Freimann Life Science Center, University of Notre Dame, for assistance with animal care, Deepa Mathew for assistance with data analysis, Jonathan Best with fibroblast analysis, and Cameron Pywell, Nathaniel Balmert and Nicolle Bonar for genotyping analysis. We thank Drs S. Yamazaki, M. Harrington, A. Davidson and E. Herzog for advice on culturing tissue explants.

The work was supported from grants to GED from the American Heart Association (Scientist Development Grant 10SDG4030011) and the National Institute of General Medical Sciences (R01-GM087508), and R01-GM083336. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.