Insulin-like growth factor-1 acts as a zeitgeber on hypothalamic circadian clock gene expression via glycogen synthase kinase-3β signaling

Brain and muscle ARNT-like protein-1 (BMAL-1) is an important component of the cellular circadian clock. Proteins such as epidermal (EGF) or nerve growth factor (NGF) affect the cellular clock via extracellular signal–regulated kinases-1/2 (ERK-1/2) in NIH3T3 or neuronal stem cells, but no such data are available for the insulin-like growth factor-1 (IGF-1). The hypothalamus expresses receptors for all three growth factors, acts as a central circadian pacemaker, and releases hormones in a circadian fashion. However, little is known about growth factor–induced modulation of clock gene activity in hypothalamic cells. Here, we investigated effects of IGF-1, EGF, or NGF on the Bmal-1 promoter in two hypothalamic cell lines. We found that only IGF-1 but not EGF or NGF enhanced activity of the Bmal-1 promoter. Inhibition of ERK-1/2 activity did not affect IGF-1–induced Bmal-1 promoter activation and all three growth factors similarly phosphorylated ERK-1/2, questioning a role for ERK-1/2 in controlling BMAL-1 promoter activity. Of note, only IGF-1 induced sustained phosphorylation of glycogen synthase kinase-3β (GSK-3β). Moreover, the GSK-3β inhibitor lithium or siRNA-mediated GSK-3β knockdown diminished the effects of IGF-1 on the Bmal-1 promoter. When IGF-1 was used in the context of temperature cycles entraining hypothalamic clock gene expression to a 24-h rhythm, it shifted the phase of Bmal-1 promoter activity, indicating that IGF-1 functions as a zeitgeber for cellular hypothalamic circadian clocks. Our results reveal that IGF-1 regulates clock gene expression and that GSK-3β but not ERK-1/2 is required for the IGF-1–mediated regulation of the Bmal-1 promoter in hypothalamic cells.

The cellular circadian clock is present in each one of our cells, with genes, proteins, and metabolites oscillating every 24 h. It acts as a biological timekeeper and manifold higher order physiological processes are subject to regulation by the circadian clock. Sleep/wake cycles, core body temperature, metabolism, and neuroendocrine secretion are controlled by diverse, cell-or organ-specific daily clocks (1)(2)(3)(4).
Much is known about the molecular components of the cellular clock in mammals, which features a network of feedback loops that span transcriptional, translational, and post-translational levels (5,6). Briefly, BMAL-1 3 and its partner CLOCK bind E-boxes within the promoter of many regulatory clock genes (Periods, Cryptochromes, Rev-Erbs, and RORs), thereby enhancing their expression. Several of the clock gene products exert a negative feedback on their own expression by inhibiting the transcriptional activation by BMAL-1 and CLOCK. The Bmal-1 promoter is controlled by two orphan nuclear receptor proteins: REV-ERB␣ and the retinoid orphan receptor A (RORA), with RORA activating and REV-ERB␣ repressing Bmal-1 expression. Because BMAL-1 is an activator of Rev-Erb␣, this represents an additional feedback pathway. Through a combination of repression and activation, BMAL-1 expression rises and falls as part of the circadian cycle (7).
As known for other transcription factors, post-translational modifications of clock components are critical for their function. Clock proteins are heavily phosphorylated, thus, second messenger-dependent kinases regulate the clock on the molecular level (8). cAMP-dependent kinase A-mediated activation of the cAMP-response element-binding protein (CREB) or ERK-1/2 has been shown to regulate PER expression (9 -12). GSK-3␤ has been described to modulate expression or cellular location of CLOCK, REV-ERB␣, or PER proteins (13)(14)(15)(16)(17)(18).
Although the key role of growth factors for cell proliferation and differentiation has been established, their role in the regulation of the cellular clock is not fully understood. Serum (a mixture of growth factors and cytokines) is well known to induce expression of clock genes (12,19,20). Induction of the c-FOS protein and activation of CREB, both acting on Per-1 or Ϫ2, has been proposed to be the molecular link between seruminduced signaling and clock gene expression (12). EGF or NGF enhanced expression of Per-2 in fibroblasts or neuronal stem cells via ERK-1/2 (9 -11, 21-25). Of note, their relative IGF- 1 has not yet been reported to regulate the cellular clock although it also activates ERK-1/2 in numerous cell lines (26). Likewise, despite the high impact of GSK-3␤ on clock gene expression described above, no effects on clock gene expression by growth factors via this kinase have been reported, although functional interactions between GSK-3␤ and growth factors are well established (26 -28).
Rhythms of circadian clock genes, as described above, occur ex vivo in cell culture, showing the cell-based nature of the circadian clock. Similar to organismal circadian clocks, cellular clocks in constant conditions show a free-running approximately 24-h rhythm. Cellular and organismal clocks can also be entrained to an external clue. In such a case, the clock adopts to the rhythm provided by this clue, which is then called zeitgeber (German for time emitter). Daily 24-h oscillations in core body temperature have been proposed to be such a zeitgeber for cellular and organismal clocks (29 -32). Clock gene expression regulating hormones are, thus, able to act as a zeitgeber for the cellular clock as shown for serum (12). EGF, did not entrain the clock in fibroblasts but in neuronal stem cells (9,21). NGF has been shown to entrain the clock in the suprachiasmatic nucleus of Syrian hamster (22,25).
Here we aimed at analyzing effects of IGF-1 on clock gene expression and compared its effects to other growth factors such as EGF and NGF. The hypothalamus is crucial for the generation of circadian rhythms, undergoes daily rhythmicity and expresses receptors for all three growth factors (22,(33)(34)(35)(36)(37). Thus, we used murine hypothalamic cells to investigate effects of growth factors on clock gene expression. We found that IGF-1 regulates Bmal-1 expression, in contrast to NGF and EGF, in these cells. It does so not via CREB, c-FOS, or ERK-1/2 but rather via GSK-3␤. When IGF-1 was used in the context of an entraining temperature cycle, it shifted the phase of Bmal-1 promoter activity, indicating that it functions as a zeitgeber for cellular circadian clocks via GSK-3␤.

IGF-1 but not EGF or NGF regulates clock gene expression in hypothalamic cells
Hypothalamic mHypoA-2/10 cells were used to investigate the effects of growth factors on Bmal-1 promoter activity (38 -40). These cells have recently been shown to express receptors for melanocortins and neuropeptide Y as well as the precursor of thyroliberin in a T3-dependent manner (39). Thus, mHypoA-2/10 cells resemble specialized neurons of the paraventricular nucleus. To measure promoter activity under constant conditions, we stably expressed a plasmid encoding the Bmal-1 promoter (containing binding sites for REV-ERB␣ and RORA) fused to a luciferase gene (Bmal-1-luc) in mHypoA-2/10 cells (mHypoA-2/10-Bmal-1-luc cells). Light emission in the presence of luciferin or luciferin ethyl ester was monitored in living cells (29). Although luminescence in mHypoA-2/10-Bmal-1luc cells was detectable in lysates, it was too low to be measured in living cells (Fig. S1). We thus used protein lysates for measurements of Bmal-1 promoter activity in these cells.
We first determined if hypothalamic cell cultures are capable of rhythmic clock gene expression. mHypoA-2/10-Bmal-1-luc cells were entrained by a protocol including serum starvation (1 h) or serum stimulation (40%, 30 min), followed by a 48-h incubation in 0.1% serum. Samples were harvested at 4-h intervals over these 2 days revealing two cycles of expression from the Bmal-1 promoter (Fig. 1A). These data suggest that the cellular clock in mHypoA-2/10-Bmal-1-luc cells was synchronized by a serum shock, showing that this cell line is a suitable model for the analysis of growth factor-induced effects on clock-regulated gene expression using the Bmal-1 promoter as a proxy. Because growth factor-induced signaling might depend on the cellular context, we also investigated the effects of growth factors on Bmal-1 promoter activity in hypothalamic cells producing the gonadotropin-releasing hormone, GT1-7 cells (41). Expression of the Bmal-1 reporter transiently in GT1-7 cells revealed that in these cells the Bmal-1 promoter was also synchronized by a serum shock (Fig. 1B). Thus, mHypoA-2/10-Bmal-1-luc and GT1-7 cells transiently expressing the Bmal-1-luc reporter are suitable for the analysis of growth factorinduced effects on the Bmal-1 promoter.
We next used the same protocol but replaced serum with IGF-1, EGF, or NGF. We first focused on acute Bmal-1 reporter activation and analyzed the effects of growth factors on both cell types 4 and 16 h post-stimulation. Strikingly, only IGF-1 activated the Bmal-1 promoter at both time points in either cell line (Fig. 1, C and D), highlighting a hitherto unappreciated functional interaction between IGF-1 and the Bmal-1 promoter. EGF failed to induce Bmal-1 reporter activation at any time point in both cell lines. NGF promoted significant Bmal-1 promoter activity only after 4 h in GT1-7 cells.
The IGF-1 concentration used (100 nM) activates both IGF and insulin receptors (42). A recent study revealed that insulin inactivates BMAL-1 by phosphorylation and subsequent nuclear extraction via insulin receptors in hepatic cells (43). Thus, to discriminate between insulin and IGF-1 receptors, we performed concentration-response curves with both hormones ( Fig. 1 E and F). EC 50 values of IGF-1 for activation of the BMAL-1 promoter were about 2 log orders lower compared with insulin in both cell lines (mHypoA-2/10-BMAL-luc: 1.2 Ϯ 0.4 nM for IGF-1 and 245 Ϯ 37 nM for insulin; GT1-7 cells: 1.5 Ϯ 0.5 nM for IGF-1 and 117 Ϯ 49 nM for insulin), clearly suggesting that IGF and not insulin receptors are involved.
Lack of Bmal-1 promoter activation by EGF is surprising in light of its effects on clock gene expression via the Per-2 promoter in other cell lines (9,21,22,25). Note that PER-2 hinders the Bmal-1 suppressor REV-ERB␣ from its inhibitory actions on this promoter. Thus activation of the Per-2 promoter should subsequently enhance Bmal-1 promoter activity. Similar, because BMAL-1 activates E-box elements of the Per-2 promoter, activation of the Bmal-1 promoter should subsequently enhance Per-2 promoter activity (Fig. 9C). Hence, we next tested the effect of all three growth factors on a Per-2 luciferase reporter in hypothalamic cells (Fig. 1, G and H). As expected from the functional interactions between BMAL-1 and the Per-2 promoter, IGF-1 induced significant Per-2 promoter activity 16 h but not at 4 h post-stimulation, indicating that it affects first the Bmal-1 and subsequently the Per-2 promoter.  In E and F, cells were stimulated with increasing concentrations of IGF-1 or insulin and analyzed after 4 h. Data from 5 independent experiments, each performed in quadruplicates, were compiled by setting values of untreated cells to 100% and are shown as the mean Ϯ S.D. Asterisks indicate significant differences relative to 100% (one-sample Student's t test) and hash marks between growth factors (one-way ANOVA).

IGF-1 is a zeitgeber
EGF and NGF failed again to activate this clock gene reporter at any time and in either cell line, suggesting that both growth factors do not affect clock gene expression in hypothalamic cells at all. These results could be because these cells do not express NGF or EGF receptors. Thus, we next analyzed growth factor-induced phosphorylation of ERK-1/2 and used a luciferase reporter gene that incorporated serum-response elements (Sre) and thus would respond to activated serum-response factor (SRF) and ternary complex factors (TCF). There were no significant differences between EGF-and IGF-1induced ERK-1/2 phosphorylation in both cell types (Fig. 2, A and B), suggesting that a lack of EGF receptor expression is not the cause for absent EGF-induced Bmal-1 or Per-2 promoter activation. NGF-induced phosphorylation of ERK-1/2 was significantly increased relative to unstimulated cells, but significantly weaker compared with IGF-1 in GT1-7 cells. All three growth factors similarly induced luciferase expression in cells containing the Sre reporter (Fig. 3, A and B, Fig. S2). These results indicate that appropriate receptors for all of these growth factors were expressed on these cells and were functional. To identify where in the cognate signal transduction pathways of IGF-1 acts differently from NGF and EGF, two additional reporter genes were used. A reporter containing the entire c-Fos promoter (harboring SRF/TCF-binding elements as well as binding sites for signal transducers and activators of transcription and for the activating protein-1) fused to luciferase was transfected into hypothalamic cells. As for the Sre reporter, treatments with serum or all three growth factors showed similarly increased gene expression (Fig. 3, C and D). When a reporter consisting of cAMP-response elements (Cre) (containing binding sites for CREB) fused to the luciferase gene was expressed, serum activated gene expression, but all three growth factors failed to induce luciferase activity in both cell types (Fig. 3, E and F). So far, our data suggest that IGF-1promoted circadian clock gene expression in hypothalamic cells is unique among the three tested growth factors and distinct from what is known in other cell models, because it does not correlate with the Cre promoter activation, c-FOS induction, or ERK-1/2 phosphorylation. To exclude a role for ERK-1/2, we used PD-184352, an inhibitor of the ERK-1/2 kinases mitogen-activated protein kinase-1/2. PD-184352 did not inhibit the effects of IGF-1 on the Bmal-1 promoter, indicating that ERK-1/2 signaling is indeed not involved in this process (Fig. 4, A and B). In contrast the inhibitor strongly decreased Sre reporter activation by all growth factors (Fig. 4, A and B). We next asked whether the distinct effects of IGF-1 on the Bmal-1 reporter are also found in cell models, where ERK-1/2-mediated activation of clock gene expression by EGF has previously One of 5 representative blots is shown; data were quantified by ImageJ, compiled by setting values of untreated cells to 100% and shown in the graphs below.
The area under the curve (restricted to 100%) is given in parentheses. Asterisks indicate significant differences between EGF or NGF and IGF-1 (two-sample Student's t test).  Hash marks indicate significant differences to DMSO (one-sample Student's t test) and asterisks between cells expressing different reporters (two-way ANOVA). In C and D, Bmal-1 reporter activation in NIH-3T3 cells was measured. In C, cells were serum-starved for 1 h, stimulated with serum (40%) or IGF-1, EGF, or NGF for 30 min and reporter activity was measured after 4 h. D, cells were serum-starved for 1 h in the presence of 10 M Pd-184352 or as a control with 0.1% DMSO followed by stimulation with IGF-1, EGF, or NGF (100 nM) for 4 h. Hash marks indicate significant differences to 100% (one-sample Student's t test).

IGF-1 is a zeitgeber
been shown. Thus, we tested the Bmal-1 reporter in NIH-3T3 fibroblasts for its sensitivity toward growth factors (9). Indeed, IGF-1 and EGF induced reporter activation in these cells (Fig.  4C) in a PD-184352-dependent manner (Fig. 4D). Hence, it appears that IGF-1-induced circadian gene expression in hypothalamic cells is unique among growth factors and cell types.

PI3K/GSK-3␤ signaling is required for IGF-1-induced Bmal-1 promoter activity in hypothalamic cells
GSK-3␤ acts on the molecular circadian clock and is an established component of growth factor signaling (13, 15, 26 -28, 44). Thus, we next investigated functional interactions between this kinase and growth factors. Phosphorylation of Ser-9 has been reported to inhibit GSK-3␤ activity. Analysis of hypothalamic cells revealed that NGF failed to phosphorylate GSK-3␤ in both cell lines (Fig. 5, A and B). EGF weakly phosphorylated this enzyme transiently. In contrast, IGF-1 induced high levels of GSK-3␤ phosphorylation persisting for at least 30 min after stimulation, with the area under the curve ϳ5 times larger in the case of IGF-1 compared with EGF stimulation in both cell types (Fig. 5, A and B). Thus, sustained phosphorylation of GSK-3␤ may represent a specific signaling pathway engaged by IGF-1 in hypothalamic cells consistent with its function on clock gene expression.

IGF-1 is a zeitgeber in hypothalamic cells
After having established the enhancing effects of IGF-1 on clock genes in hypothalamic cells, we next asked whether IGF-1 would modify the entrained phase of clock gene expression. Temperature is a near universal zeitgeber for circadian clocks,

IGF-1 is a zeitgeber
with 24-h temperature cycles within a physiological range synchronizing the expression of circadian clocks (30,31,46). Thus, mHypoA-2/10-Bmal-1-luc cells were either exposed to 24-h temperature cycles (12 h at 38.5 and 12 h at 36.0°C or 12 h at 36.0 and 12 h at 38.5°C) or, as a control, kept at constant temperature of 37.0°C. Luciferase activity was measured every 4 h for 48 h. As shown in Fig. 8, A and B, temperature cycles synchronize oscillations of Bmal-1 promoter activity with peaks in luciferase expression in the low temperature phase. Thus, we used temperature-entrained cultures to investigate if IGF-1 is capable of advancing or delaying the phase of Bmal-1 luciferase expression by applying IGF-1 to cells at ZT12 and ZT0. Note that for mice, ZT12 corresponds to high core body temperature which occurs in the dark period in a 12-h light/12-h dark cycle.
Addition of IGF-1 at the low to high temperature transition (ZT12) led to a 2.7 Ϯ 0.6-h phase advance (Fig. 8, A, C, and D).
Adding the growth factor at the high to low temperature transition (ZT0) resulted in a 5.5 Ϯ 0.5-h phase delay (Fig. 8, B-D).
Hence, IGF-1 acts as a zeitgeber for hypothalamic cells, as demonstrated by the change in the phase of temperature-entrained Bmal-1 promoter activity. One might assume that IGF-1 acts as zeitgeber in temperature-entrained cells by modifying distinct transcriptional regulators within the clock. As shown in Fig. 1, IGF-1 first activated the Bmal-1 (4 h) and subsequently the Per-2 (16 h) reporter, suggesting that Bmal-1 is the primary target of IGF-1 within the clock. To further test this hypothesis, we performed a checkerboard experiment whereby kinetics (2, 4, and 6 h) of IGF-1 or

IGF-1 is a zeitgeber
temperature-induced Bmal-1 and Per-2 promoter activation was compared with and without REV-ERB␣ (modulated by siRNA application). In the presence of REV-ERB␣, IGF-1 induced significant activity of the Bmal-1 promoter after 2 h and of the Per-2 promoter after 6 h, whereas temperature first activated the Per-2 and subsequently the Bmal-1 promoter (Fig.  9, A and B). When REV-ERB␣ expression was suppressed by siRNA, IGF-1 failed to activate Bmal-1 and the Per-2 promoter. Thus, the sequence of transcription factors affected by IGF-1 is REV-ERB␣, BMAL-1, and PER-2 (Fig. 9C). Temperature still activated the Per-2 promoter in the absence of REV-ERB␣, supporting the concept that temperature primarily affects the Per-2 promoter independently of the Bmal-1 suppressor (Fig. 9,  A and B). However, temperature-induced Bmal-1 promoter activation was significantly inhibited, indicating the following sequence of temperature-controlled transcription factor activation: PER-2, REV-ERB␣, and BMAL-1 (Fig. 9C).

Discussion
The IGF-1 system, consisting of the hormone itself, its receptors and binding proteins (IGFBP) is crucial for the growth, development, and plasticity of the central nervous system (37). IGF-1 is produced by many cell types in the central nervous system and IGF receptors are widely distributed (37). Here, we provide the first evidence that IGF-1 directly affects the expression of the circadian gene Bmal-1 and influences circadian clock gene expression in hypothalamic cells. Thus, regulation of hypothalamic circadian rhythms can be considered as a new biological role for IGF-1.
GSK-3␤ has previously been reported to affect circadian rhythms via multiple mechanisms (13)(14)(15)(16)(17)(18). This role for GSK-3␤ in clock gene expression has so far either been demonstrated by overexpression or pharmacologically (lithium). Thus, it remained elusive whether GSK-3␤ is also involved in the regulation of clock gene expression, when regulated by a physiological stimulus. Here, we show for the first time that hormone-controlled regulation of GSK-3␤ activity via PI3K affects circadian clock gene expression. First, this finding is supported by the strong correlation between the propensity of a given growth factor to phosphorylate GSK-3␤ and its ability to activate the Bmal-1 promoter. Second, down-regulation of GSK-3␤ via siRNA or inhibition by LiCl inhibited IGF-1induced Bmal-1 activation. Because both effects, GSK-3␤ phosphorylation and Bmal-1 reporter activation, were also blunted by a specific PI3K inhibitor, we suggest a new mechanistic model in which IGF-1 influences circadian rhythms via PI3Kpromoted inhibition of GSK-3␤ (Fig. 9C).
Interestingly, in both hypothalamic cell types analyzed this pathway is neither affected by EGF nor NGF, even though receptors for both growth factors are functionally expressed. EGF and NGF have been reported to regulate clock gene expression in various cell lines via ERK-1/2-mediated expression of PER-2 (9 -11, 21-24). Thus, either the Bmal-1 promoter is differentially regulated depending on the cell type or growth factors per se affect the Bmal-1 promoter differentially independent of the cellular context. We show that IGF-1 and EGF induce Bmal-1 reporter expression via ERK-1/2 in NIH-3T3 fibroblasts. Thus, ERK-1/2 appears to be key for growth factorinduced clock gene expression in fibroblasts. Of note, lithium affects clock gene expression in NIH3T3 cells and EGF or IGF-1 phosphorylate GSK-3␤ in these cells (47)(48)(49). However, for yet

IGF-1 is a zeitgeber
unknown reasons, growth factor-mediated GSK-3␤ activity is apparently less important for clock gene expression in NIH3T3 cells. On the contrary, our data indicate that in hypothalamic cells ERK-1/2 appears to be less important for growth factorinduced clock gene expression, whereas GSK-3␤ is pivotal. Because EGF and NGF only weakly phosphorylated GSK-3␤ in hypothalamic cell models, this situation allows hypothalamic cells to exclusively link IGF-1 to the Bmal-1 promoter. At this point we can only speculate why in both hypothalamic cells EGF and NGF only poorly phosphorylate GSK-3␤. Although EGF and NGF receptors mainly interact with the GRB2 adapter protein, IGF receptors prefer insulin receptor substrates. Thus, different functions and/or expression levels of both adapter proteins in distinct cells will affect growth factor signaling differentially. Furthermore, recent data propose a role for G␣ i proteins in EGF-but not IGF-1-induced PI3K activation, suggest-ing that distinct expression of G␣ i subtypes could also underlie different interactions between receptor tyrosine kinases and PI3K (50). Our work not only reveals that IGF-1 regulates Bmal-1 promoter activity via PI3K/GSK-3␤ but also suggests that IGF-1 changes the phase of temperature-entrained Bmal-1 promoter activity and thus acts as cellular circadian zeitgeber. When IGF-1 was applied prior to the cold-warm transition (ZT12), it induced an ϳ2 h advance in the Bmal-1 promoter activity phase. In clear contrast, it promoted a ϳ5 h delay when given before the warm-cold transition (ZT0). Thus, IGF-1 shows phase-specific modulation of clock gene expression similar to other zeitgebers. Recently serum IGF-1 levels in WT mice and mice deficient for the clock genes cryptochrome-1 and Ϫ2 were determined. A peak in IGF-1 serum levels was found at ZT0, where effects of IGF-1 on the Bmal-1 promoter in hypothalamic cells were most effective and advanced the phase of clock gene expression (51). Of note, circadian IGF-1 serum peaks were dramatically reduced in animals deficient for both cryptochrome genes. These results together with the data obtained here, suggest that the cellular clock regulates circadian IGF-1 levels and that the hormone feeds back on the clock by regulating its phase. Recent work by Shibata and colleagues (52) clearly shows that application of IGF-1 to liver modifies the expression of the clock gene Per-2. In vitro, the free running period is shortened. In vivo the phase of rhythmic gene expression is advanced following administration of the hormone at ZT5. Thus, serum IGF-1 levels could be a zeitgeber for both circadian entrainment of the liver based on feeding time and temperature-mediated entrainment in the hypothalamus. It will be interesting to discover what cell types integrate rhythmic IGF-1 expression for tuning circadian clock-regulated gene expression.
In addition to IGF-1, its receptors and binding proteins also show daily oscillations (53)(54)(55). Thus, the net effect of the IGF system on the cellular clock in vivo may depend on the circadian rhythms of IGF, IGF receptors, and IGFBP's in combination. Interestingly, EGF serum levels and EGF receptor expression and its signaling also show diurnal patterns with peaks in the resting phase, suggesting that feedback loops between the circadian clock and growth factors are not unique for IGF-1 (56,57), at least in the liver.
We provide data on the cellular level indicating that IGF-1 affects clock gene expression and thus potentially modulates clock-dependent processes in cells. Assuming that our data obtained on the cellular level can be extrapolated to the living human organism or to behavior, the impact of IGF-1 on the clock would be time-of-life specific. IGF-1 serum levels increase throughout childhood, reaching a plateau after puberty and subsequently declining in adults and the elderly (58 -66). Changes in the phase of circadian rhythm has been observed in adolescent and elderly individuals (25,26,(67)(68)(69)(70)(71). Re-synchronization of the "aged clock" may even slow down the aging process (72). Thus, the new link between IGF-1 and circadian rhythm, suggested by our data, provides an intriguing molecular mechanistic explanation for age-related changes in circadian rhythms and may offer a new strategy for therapeutic intervention. Asterisks indicate a significant difference relative to basal/control samples; hash marks between cells expressing the random or the selective siRNA (two-way ANOVA). C, a cartoon depicting a model illustrating distinct IGF-1-and temperature-induced signaling pathway leading to Bmal-1 or Per-2 promoter activation, respectively, is shown.

Cell culture and transfection
Hypothalamic or NIH-3T3 cells were routinely cultured in DMEM with 10% FBS, 2 mM L-glutamine and penicillin/streptomycin at 37°C and 5% CO 2 . mHypoA-2/10 cells (Clu-176) were obtained from Cedarlane, GT1-7 cells were from Dr. Weiner, University of California, San Francisco, CA, and NIH-3T3 cells were from Sigma (38,41). To obtain mHypoA-2/10 cells stably expressing the Bmal-1-Luc reporter (mHypoA-2/ 10-Bmal-1-luc cells), cells were co-transfected with the pABpuro-BluF9 plasmid and an empty pcDNA4 vector containing the resistance gene for zeocin. Transfected cells were then selected with 600 g/ml of zeocin for 4 weeks. For transient expression of reporter plasmids, as for Figs. 1-3 and 4B, ϳ20,000 GT1-7 cells per well were seeded in 12-well plates 24 h prior to transfection. 200 to 400-ng plasmids were transfected using Turbofect TM according to the manufacturer's protocol. After 24 h, cells were stimulated as described in the corresponding figure legends. To down-regulate expression of GSK-3␤ (sc-35525) or REV-ERB␣ (sc-61459), gene-specific siRNAs from Santa Cruz (Heidelberg, Germany) were used and tested against a control siRNA construct (sc-37003). To obtain sufficient transfection efficacy, siRNAs were used on mHypoA-2/10-Bmal-1-luc cells alone (Fig. 4) or together with the corresponding reporter plasmid into mHypoA-2/10 ( Fig. 6) or GT1-7 cells (Fig. 4D) in conjunction with electroporation using the neon transfection system from Invitrogen according to the manufacturer's protocol. Briefly, 500,000 cells together with 50 nM of the corresponding siRNA or with this siRNA and 2 g of the reporter plasmid were pulsed with 1450 V for 30 ms and then placed in 24-well plates. To culture cells under changing temperature conditions, two FRIOCELL-22 incubators were used. Cells were kept in DMEM/F-12 medium (50:50) containing 25 mM HEPES, 0.1% FCS, and penicillin/streptomycin. For experiments shown in Fig. 1, one incubator was set to a 12-h 38.5/36.0°C and the second to 12-h 36.0/38.5°C rhythm. A control incubator was kept at a constant 37°C (Hereaus, T6030). Before cells were cultured in the anti-phase temperature cycles or at constant temperature, they were first serumstarved for 1 h, stimulated or not with IGF-1 for 30 min at 37°C in DMEM/F-12 without FCS, and then washed and kept in DMEM/F-12 with 0.1% FCS for the duration of the experiment. For each time point, one 12-well plate (6 control and 6 IGF-1stimulated cells) was used for each time point in temperature cycles and one 12-well plate with 6 unstimulated control cell pools at constant temperature. Cells were removed at the indicated time points. For analysis, cells from temperature cycles (unstimulated or IGF-1 treated) were normalized to those from unstimulated cells at a constant temperature by setting luciferase measured to 100%. For the experiments shown in Fig. 6, A and B, cells were kept in DMEM/F-12 without FCS at 37°C for 18 h. One incubator was cooled down to 36°C, whereas another was heated to 38.5°C. For each time point, one 12-well plate was removed from each incubator.

Reporter assay
Cells were lysed in 100 -200 l of lysis buffer (25 mM Tris-HCl, pH 7.4, 4 mM EGTA, 8 mM MgCl 2 , 1 mM DTT, and 1% Triton X-100) and a volume of 90 -180 l was transferred to white-bottomed, 96-well plates. Luciferase activity was measured after automatically injecting the luciferase substrate (20 -40 l) using a FLUOstar Omega plate reader. Total light emission was detected every second for 10 s post-injection and average emission between 2 and 10 s was determined.

Western blotting
Cells were seeded on 6-well plates (ϳ100,000/well), cultured for 1 day, stimulated as indicated in the figure legends, and then lysed in Laemmli buffer. Lysates were subjected to SDS-PAGE (10%), proteins were transferred to nitrocellulose by Western blotting and probed for total histone-3 proteins as a loading control and for GSK-3␤ or ERK-1/2 phosphorylation. Immune reactivity was quantified by densitometry, with the ratios between p-GSK-3␤ or p-ERK-1/2 and histone signals were calculated using ImageJ. Ligand-induced protein phosphorylation was normalized to unstimulated cells.

Quantification and statistical analysis
Values represent the mean Ϯ S.D. of three to six independent experiments. Western blotting data are representative of a typical experiment repeated 5 times or more and also quantified by ImageJ. Statistical analysis was performed either by one-sample or two-sample Student's t test, or one-way or two-way ANOVA followed by Dunnett's post-test using the GraphPad Prism software. Circular wave fit was performed using the CircWave software version 1.4 (75). Data of fitted curves were then transferred to GraphPad Prism and further analyzed. One symbol indicates a p value of Ͻ0.05, two of Ͻ0.01, and three of Ͻ0.001.