Disrupting Circadian Homeostasis of Sympathetic Signaling Promotes Tumor Development in Mice

1Disruption of circadian rhythm increases spontaneous and carcinogen-induced mammary tumors in rodents,,,,,. Epidemiological studies have revealed that human night-shift workers show an increased risk of breast, colon, lung, endometrial and prostate cancer, hepatocellular carcinoma and non-Hodgkin's lymphoma,,,,,. Loss of circadian rhythm is also associated with accelerated tumor growth in both rodents and human cancer patients,,. These findings raise the question of how circadian dysfunction increases the risk of cancers.

2Circadian rhythms in mammals are generated by an endogenous clock composed of a central clock located in the hypothalamic suprachiasmatic nucleus (SCN) and subordinate clocks in all peripheral tissues. The SCN clock responds to external cues and drives peripheral clocks via circadian output pathways. Both the central and peripheral clocks are operated by feedback loops of circadian genes, including Bmal1, Clock, Period (Per1-3) and Cryptochrome (Cry1 and 2). Bmal1 and Clock encode bHLH-PAS transcription factors that heterodimerize and bind to E-boxes in gene promoters to activate Per and Cry transcription, whereas Per and Cry encode repressors of BMAL1/CLOCK. The alternating activation and suppression of the BMAL1-driven positive loop and the PER/CRY-controlled negative loop result in a circadian oscillation of the molecular clock,,.

3The molecular clock regulates clock-controlled genes (CCGs) to control tissue/organ function. Most CCGs follow tissue-specific expression patterns. Only a small group of CCGs, which include key cell cycle regulators and tumor suppressors, are expressed in all tissues studied. Such circadian control leads to the coupling of cell proliferation with key tissue functions in vivo,,,,,. Disruption of circadian rhythm in cell proliferation is frequently associated with tumor development and progression in mammals,,,,,,.

4Both positive and negative loops of the molecular clock are involved in cell cycle control. For example, BMAL1 suppresses proto-oncogene c-myc but stimulates the tumor suppressor Wee1,,, CRY2 indirectly regulates the intra S-check point,, and PER1 directly interacts with ATM in response to γ-radiation in vitro. In mice, mutation in Per2 leads to deregulation of DNA-damage response and increased neoplastic growth,,. In humans, deregulation or polymorphism of Per1, Per2, Cry2, Npas2 and Clock is associated with acute myelogenous leukemia, hepatocellular carcinoma, breast, lung, endometrial and pancreatic cancers, and non-Hodgkin's lymphoma,,,,,,,,,.

5However, peripheral clock-controlled gene expression is not sufficient to generate and maintain the circadian rhythm of cell proliferation because G1 cell cycle progression in normal somatic cells is strictly controlled by extracellular mitogenic signals. In vivo, these signals include hormones, growth factors, cytokines and neurotransmitters that follow cyclic changes over a 24 hour period since they are either directly released from circadian output pathways or encoded by CCGs in peripheral tissues,,. We hypothesize that cell proliferation in vivo is paced by both central and peripheral clocks. The central clock-controlled mitogenic signals simultaneously actives the cell cycle and peripheral clocks leading to a circadian coupling of cell cycle and tumor suppressor gene expression. Disruption of circadian rhythm promotes tumor development due to, at least in part, loss of the homeostasis of cell cycle control.

6To test our hypothesis, we studied the role of circadian homeostasis of the sympathetic nervous system (SNS) in tumor suppression. The central clock generates a robust circadian rhythm in SNS signaling via direct and indirect targeting of the presympathetic neurons located in the hypothalamic autonomic paraventricular nucleus. In vivo, the SNS controls all peripheral tissues by releasing the hormones epinephrine and norepinephrine that target adrenergic receptors (ADRs) on the cell membrane. Norepinephrine is directly released from postganglionic sympathetic neurons, whereas epinephrine is released from preganglionic sympathetic neuron-controlled chromaffin cells located in the adrenal medulla. We have previously reported that all circadian gene-mutant mice studied show hyperplastic growth of osteoblasts in bone, and that β-adrenergic receptor (ADRβ) activation leads to Per and Ap1 induction in primary osteoblasts. In addition, AP1 directly activates c-myc, whereas BMAL/CLOCK prevents myc over-expression. Since the induction of Per leads to peripheral clock activation and that the activation of c-myc initiates cell proliferation, SNS signaling could be a circadian time cue for both cell cycle and peripheral clocks. Therefore, SNS dysfunction may disrupt the circadian homeostasis of cell cycle control in peripheral tissues.

7Here we report that all circadian gene-mutant mouse models studied are cancer-prone and that hyperplastic growth of Per-mutant somatic cells is dependent upon extracellular mitogens. Rhythmic SNS signaling plays a key role in pacing cell proliferation rate in peripheral tissues by simultaneously activating Ap1-c-myc and ATM-p53 signaling. However, activation of Ap1 by SNS signaling is independent of the peripheral clock while activation of ATM is peripheral clock-dependent. Disruption of circadian rhythm desynchronizes the central clock-SNS-peripheral clock axis, suppresses peripheral clock function and abolishes peripheral clock-dependent ATM activation, leading to myc oncogenic activation and increased incidence of tumors in wild-type mice. Our studies identify a previously unknown molecular pathway that links disruption of circadian rhythm with oncogenesis and demonstrate that tumor suppression in vivo is a clock-controlled physiological process but not a non-clock function of a specific circadian gene.

8We found that when kept in 24 hour alternating light-dark conditions (24hr LD cycles), mice deficient in Bmal1 (Bmal1+/−), Cry1 and Cry2 (Cry1−/−;Cry2−/−), Per1 and Per2 (Per1−/−;Per2m/m) or Per2 alone (Per2−/−) were all cancer-prone. About 10–15% of Bmal1−/− mice also showed hyperplasia of salivary glands in spite of a significant reduction in the size of other major organ systems due to aggressive aging. Bmal1+/−, Cry- and Per-mutant mice all showed increased risk of ulcerative dermatitis and hyperplasia in the salivary gland, preputial gland, liver and uterus as well as spontaneous lymphoma, liver and ovarian tumor development, although spontaneous tumors in Cry-mutants were mostly identified after 50 weeks of age, later than that of Per1−/−;Per2m/m and Per2−/− mice (Fig. 1a, Table 1, Fig. S1a and data not shown).

9A sublethal dose of γ-Radiation induced premature aging on the external appearance of all circadian gene-mutant mouse models studied and further increased incidence of tumor and hyperplasia as well as ulcerative dermatitis in Bmal1+/−, Per- and Cry-mutant mice (Fig. 1b, Table 1 and Fig. S1b). About 8% of irradiated Bmal1−/− mice also developed lymphoma despite an average lifespan of 27 weeks due to a further accelerated rate of aging (Fig. 1c and Table 1). Irradiated Bmal1+/− mice showed a similar rate of tumor development as did irradiated Per2−/− mice (Fig. 1d–e and Table 1). Since all circadian gene-mutant mouse models show increased sensitivity to γ-radiation, we conclude that the molecular clock functions in tumor suppression in vivo.

10We then examined the role of the central clock in tumor suppression by studying the effects of jet-lag (an 8hr phase-advance followed by an 8hr phase-delay in the onset of the light period every 3 days) on tumor development in mice. We found that jet-lag further increased and hastened tumor development in wt, Cry- and Per-mutant mice and induced pancreatic, kidney and intestinal tumors in mutant mice. However, jet-lag did not significantly change the overall survival and tumor development of irradiated Bmal1−/− mice that were deficient in responding to circadian light cues in the central clock (Table 1, Fig. 2a, Fig. S1c and data not shown). Jet-lag also significantly changed tumor spectrum and induced osteosarcoma, liver and ovarian tumors, hyperplasia of the salivary gland, liver and uterus as well as severe seminal vesicle enlargement in wt mice (Table 1. Fig. 2b–g, Fig. S1d–l and data not shown). In addition, although when kept 24hr LD cycles, female circadian gene-mutant mice showed an earlier onset and a higher tumor incidence compared to male mutant mice, we did not found a significant gender difference in total tumor incidence and the time of tumor onset among irradiated and jet-lagged male and female mutant mice (data not shown).

11We found that chronic jet-lag disrupted the circadian rhythm in urine levels of creatinine and creatinine kinase and significantly increased the risk of renal failure in mice (Table 1 and data not shown). Mice that displayed fatal renal failure usually had only one kidney left at the time of pathological examination which completely lost filtration function due to severe hydronephrosis leading to pressure atrophy and loss of the renal cortex (Fig. 3a–e). Since renal failure has a strong correlation with sympathetic dysfunction, this finding indicates that jet-lag disrupts the homeostasis of multiple circadian output pathways including the SNS. To test this hypothesis, we first studied urine catecholamine levels in wt and Per1−/−;Per2m/m mice. We found that in 24hr LD cycles, both wt and Per1−/−;Per2m/m mice showed a two-peak oscillation in norepinephrine and a one-peak oscillation in epinephrine levels in urine (Fig. 3f–g). In constant darkness (24hr DD cycles), the urinary catecholamine levels still oscillated with a similar pattern as in 24hr LD cycles in wt mice, but were significantly elevated and arrhythmic in Per1−/−;Per2m/m mice (Fig. 3h–i). We conclude that one of the urine norepinephrine peaks is generated by central clock-controlled postganglionic sympathetic neuron activity, while the other is generated by preganglionic sympathetic neuron-controlled chromaffin cells that also rhythmically release epinephrine. Since the central clock in Per1−/−;Per2m/m mice is driven by environmental light in LD cycles but is not functional in DD cycles, these mice could maintain a circadian rhythm of urinary catecholamine levels in LD cycles but not in DD cycles.

12We then studied whether the SNS rhythmically activates gene expression in peripheral tissues. We found that among various SNS targets studied, Ucp1 mRNA, which is expressed in brown adipose tissue under the control of a CRE site in Ucp1 promoter, could be used as a molecular marker to analyze the activities of both central and peripheral clocks. In 24hr LD cycles, Ucp1 mRNA showed a two-peak expression in wt BAT, with the first peak appearing at ZT10 and the second at ZT22. In Per1−/−;Per2m/m BAT, only the ZT22 peak was observed (Fig. 4a–b). In 24hr DD cycles, Ucp1 mRNA still followed a two-peak expression in wt BAT, but was arrhythmic and high in Per1−/−;Per2m/m BAT (Fig. 4c–d). In wt mice, the activation of CREB peaked between ZT10 and ZT18 (Fig. S2a–b). Loss of function in Per1 and Per2 had no effect on CREB activation induced by ADRβ signaling (Fig. S2c), but abolished the peripheral clock activity as shown by lacking a rhythmic expression of Per2 and Bmal1 mRNAs in Per1−/−;Per2m/m BAT (Fig. 4e–g). Thus, we conclude that the ZT10 peak of Ucp1 mRNA is generated by a collaborative activity of the central and peripheral clocks, whereas, the ZT22 peak of Ucp1 mRNA is exclusively controlled by the central clock.

13We next examined the effect of jet-lag on Ucp1 mRNA expression in wt mice on the second day after one cycle of jet-lag. We found that although the ZT22 peak of Ucp1 mRNA was found in both control and jet-lagged mice, the ZT10 peak of Ucp1 mRNA was absent in jet-lagged mice, which correlated with a lack of rhythmic expression of Per2 and Bmal1 mRNAs in the same BAT (Fig. 4h–i, Fig. S2d–e). We then studied Ucp1 expression in wt mice treated with chronic jet-lag for 10 months and found that Ucp1 mRNA expression followed a similar pattern in these mice compared to wt mice treated with only one cycle of jet-lag (Fig. 4j–k). Together, our studies demonstrate that although the central clock in mice always couples to external light cues, acute jet-lag promptly disrupts the coordination of the central and peripheral clocks, whereas chronic jet-lag abolishes the peripheral clock function in the absence of circadian gene mutations.

14To investigate whether SNS signaling stimulates cell cycle progression, we treated wt and Per1−/−;Per2m/m calvarial osteoblasts (preosteoblasts) with isoproterenol (iso), a synthetic agonist for ADRβ2 and β3 expressed on the preosteoblast cell membrane,. We found that iso induced Ap1, c-myc and cyclin D1 mRNAs as well as Cyclin D1 protein in both wt and Per1−/−;Per2m/m preosteoblasts, but the level of c-myc and cyclin D1 induction was significantly elevated in Per-mutant cells (Fig. 5a–c). The induction of c-myc by iso was c-fos-dependent (Fig. 5d). Iso activated Per and c-fos in preosteoblasts via ADRβ3, or both ADRβ2 and ADRβ3 since ADRβ2−/− preosteoblasts showed an intact Per1 and c-fos induction by iso (Fig. 5e). We then studied whether iso stimulates osteoblast proliferation. We found that Per1−/−;Per2m/m osteoblasts showed a higher G2 cell population compared to wt controls when grown in media containing 10% serum (Fig. 5f), which indicates that these cells had an accelerated rate of proliferation. However, under serum-starved conditions, wt and Per1−/−;Per2m/m osteoblasts arrested equally well (Fig. 5g). Iso induced cell cycle progression in both wt and Per1−/−;Per2m/m preosteoblasts but iso-treated Per1−/−;Per2m/m osteoblasts needed less time to progress from G1 to G2 phase compared to wt controls (Fig. 5h). The accelerated proliferation of Per1−/−;Per2m/m osteoblasts correlated with the over-induction of c-myc and Cyclin D1 by iso in these cells (Fig. 5a–c). Thus, hyperplastic growth of Per1−/−;Per2m/m osteoblasts is dependent on extracellular mitogenic signals.

15Since p53 plays a key role in preventing Myc oncogenic activation, we studied whether iso also induces p53 expression in preosteoblasts. We found that p53 was induced by iso with kinetics similar to AP1 transcription factors in wt osteoblasts. However, iso-treated Per1−/−;Per2m/m osteoblasts showed AP1 overexpression in the absence of p53 induction (Fig. 6a). The expression of p53 is mainly controlled by its interaction with the E3 ubiquitin ligase MDM2 in vivo,. Thus, we examined MDM2 expression in iso-treated preosteoblasts. We found that MDM2 was suppressed by iso with the same kinetics as p53 induction in wt osteoblasts, but remained high in Per1−/−;Per2m/m osteoblasts (Fig. 6a). The MDM2-p53 interaction is best studied in γ-radiation response, which activates protein kinase ATM that phosphorylates p53 at Ser18 (S18) and MDM2 at Ser395 (S395) in mice. The phosphorylation at these two sites blocks MDM2-p53 interaction, leading to MDM2 autoubiquitination and p53 induction. Since we used an anti-MDM2 2A10 antibody raised against a MDM2 C-terminal region containing the S395 residue, the interaction of 2A10 with MDM2 could be blocked by MDM2 S395 phosphorylation. Thus, we studied MDM2 and p53 expression in iso-treated preosteoblasts using an anti-p53 S18 antibody and an anti-MDM2 AB4 antibody that interacts with MDM2 at the N-terminal region. We found that these two antibodies detected a similar rate of p53 induction and MDM2 degradation in wt osteoblasts as shown by the p53 PAb421 and MDM2 2A10 antibodies but failed to detect p53 S18 induction and MDM2 decrease in Per1−/−;Per2m/m osteoblasts (Fig. 6b). We then studied p53 induction in iso-treated wt and Atm−/− preosteoblasts. We found that in wt osteoblasts, ATM was activated by iso, as shown by ATM S1981 phosphorylation, in an iso dose-dependent manner and that the duration and the levels of ATM activation correlated with the level of p53 S18 phosphorylation and total p53 accumulation. In contrast, Atm−/− osteoblasts that lacked p53 S18 phosphorylation failed to show p53 induction by iso (Fig. 6c–d). Together, our findings indicate that the SNS activates p53 via ATM in a peripheral clock-dependent manner.

16We then studied p53 expression and S18 phosphorylation in vivo. We found that the expression of total p53 and p53 S18 phosphorylation followed a coupled circadian rhythm over a 24hr period and peaked around ZT10 in most wt thymuses studied. In Per1−/−;Per2m/m thymuses, p53 S18 phosphorylation was almost undetectable, and the total p53 level was significantly dampened over a 24hr period (Fig. 7a). In Atm−/− thymuses, the circadian rhythm of p53 expression and p53 S18 phosphorylation was completely abolished (Fig. 7b). In addition, in 24hr LD cycles, the circadian expression of p53 was coupled with c-Fos and c-Myc in all wt mouse tissues studied including thymus, liver, kidney and white adipose tissues (Fig. 7c and data not shown).

17We next studied p53 and c-Myc expression in the thymus of jet-lagged wt mice and found that one cycle of jet-lag was sufficient enough to uncouple p53 and Myc signaling, inhibit p53, and stimulate Myc expression (Fig.7d–f). To further study the role of p53 in clock-controlled tumor suppression, we crossed p53− /− mice with Per2−/− mice that have deficient peripheral clocks and overexpress c-myc. We found a Per2 and p53 dose-dependent reduction in survival time in mice due to increased tumor development. Complete loss of Per2 function reduced the average life span of p53− /− mice from 6 months to 4 months (Fig. 7g). We then treated p53− /− mice with jet-lag and found that jet-lagged p53− /− mice showed increased liver and salivary gland hyperplasia, kidney failure, and accelerated tumor development compared to untreated control p53−/ − mice. Lymphoma and osteosarcoma were identified among jet-lagged p53− /− mice, with 10% of tumors being osteosarcoma (Fig. 7h and data not shown).

18We have found that circadian gene-mutant mice share common aging and cancer-prone phenotypes although these symptoms may be more evident at different times during lifespan or under different physiological conditions in different mouse models. Disruption of circadian rhythm abolishes the peripheral clock which induces a tumor spectrum in wt mice similar to that found in circadian gene-mutant mice. These findings demonstrate that tumor suppression in vivo is a clock-controlled physiological function but not a non-clock function of a specific circadian gene.

19Epidemiological studies have shown that disruption of circadian rhythm increases the risk of breast, prostate, endometrial, ovarian, lung, colon and pancreatic cancers, as well as non-Hodgkin's lymphoma, osteosarcoma and hepatocellular carcinoma in humans,,,,,,,,,,,. The role of circadian dysfunction in mammary tumor development in rodents has also been well-documented,,,,. Our studies show that chronic jet-lag also increases the risk of ovarian, kidney, intestinal and pancreatic cancers, osteosarcoma, lymphoma and hepatocellular carcinoma in mice. Together, these studies suggest that the mechanism of clock-controlled tumor suppression is conserved during evolution.

20Irradiated and jet-lagged male wild-type mice show relatively lower tumor incidences and later onset of tumors than their female littermates. However, we did not find a significant gender-dependent difference in the time of tumor onset and total tumor incidence among irradiated and jet-lagged circadian gene-mutant mice. Although a larger sampling size may show a gender-dependent tumor development in irradiated and jet-lagged mutant mice, our studies indicate that gender-independent mechanisms are also involved in clock-controlled tumor suppression. Recently, nonsexual hormones leptin and growth hormones have been linked to the development of colon and breast cancers as well as non-Hodgkin's lymphoma in humans,,. A role for SNS dysfunction in the development and/or progression of various human cancers has also been reported, which include prostate, breast, pancreatic, colon and ovarian cancers, glioma, neuroblastoma, osteosarcoma, hepatocellular carcinoma, cholangiocarcinoma and non-Hodgkin's lymphoma,,,,,,,,,,. These findings in conjunction with our own studies indicate that circadian dysfunction of sympathetic activity could promote tumor development independently and/or synergistically with sex hormones.

21The direct and indirect neuronal control of SNS by the central clock and the multi-synaptic control of peripheral tissues by the SNS have been intensively studied,. Our studies demonstrate for the first time that in vivo, the rhythmic SNS signaling is a key central-peripheral timing mechanism that couples cell proliferation and tumor suppression with mammalian daily physiology. Such circadian control provides a gender-independent mechanism for tumor suppression. The identification of Ucp1 as an endogenous reporter for both central and peripheral clock activities in our studies also provides a powerful tool for analyzing and comprising the status of endogenous circadian homeostasis which could not be evaluated by behavioral studies due to the light masking effect on the central clock.

22Previous studies have shown that Per-mutant mice are cancer-prone, whereas Cry1−/−;Cry2−/− mice are deficient in cell proliferation in the first 72 hours of liver regeneration,,,. A similar deficiency in liver regeneration has been reported for mice lacking the nuclear receptor FXR which are also prone for spontaneous hepatocellular carcinoma,. These findings, along with our own studies, suggest that cell proliferation is differentially controlled under different physiological conditions in vivo. Using the central clock-SNS-peripheral clock axis as a model system, we propose that the central clock-controlled SNS signaling generates a coupled AP1, peripheral clock, and ATM activation. The activation of AP1 leads to myc-induced cell cycle progression, while the activation of the peripheral clock inhibits myc overexpression and is required for ATM activity. ATM then induces p53 to prevent Myc oncogenic signaling by blocking p53-MDM2 interaction. Disruption of circadian rhythm desynchronizes the central clock-SNS-peripheral clock axis which suppresses peripheral clock and peripheral clock-dependent ATM-p53 signaling but has no effect on c-myc activation. Together, these events lead to Myc oncogenic activation that promotes genomic instability and tumor development (Fig. 7i). Our model suggests that the circadian clock plays a dual role in cell cycle control and it suppresses tumor development by controlling the homeostasis but not the inhibition of cell proliferation.

23Our finding of the sympathetic control of ATM-p53 signaling indicates that the induction of p53 occurs as an integrated part of mammalian daily physiology in vivo. This finding has important therapeutic implications. Recent studies indicate that cancer chronotherapy improves therapeutic index. This may, at least in part, be explained by the fact that chronotherapy is usually applied during the sleeping phase when the endogenous p53 level is high, allowing host tissues to better tolerate genotoxic insults that induces p53 and p53-dependent apoptosis in somatic cells expressing low levels of p53. We have previously reported that wt thymocytes are less protected from radiation-induced apoptosis if mice are irradiated at ZT2 or ZT18 when the endogenous p53 level is low compared to thymocytes irradiated at ZT10 when the endogenous p53 level is high (Fig. 7a–c). Further study of clock-controlled ATM-p53 signaling is important for the development of novel strategies for cancer prevention and treatment.

24Animal experiments were approved by the Baylor College of Medicine Institutional Animal Care and Use Committee. Research was conducted in compliance with the Animal Welfare Act Regulations and other Federal statutes with regard to animals and experiments involving animals and adheres to the principles set forth in the Guide for Care and Use of Laboratory Animals, National Research Council, 1996. Mice were generated by heterozygous breeding of C57BL/6 inbred strains of Bmal1+/−, Cry1+/−;Cry2+/−, Per1+/−;Per2+/m, Per2+/−, Per2+/−;p53+/− and p53+/− mice, and 129/C57BL6 hybrid strain of Atm+/− mice, and maintained under standard pathogen-free conditions of 2–5 per cage at 21°C–23°C with a humidity of 50%–70%, an air-flow rate of 15 exchanges/hr and in 24hr L/D cycles (light on at ZT0 and off at ZT12) if no specific light/dark conditions were described in the text.

25Similar numbers of male and female mice of each genotype were used for all survival curve studies. Mice were monitored 2–3 times a week following the five Morton and Griffiths aspects of animal conditions as a guide. Terminally-ill mice showed multiple symptoms including lack of food-intake, difficulty in standing, walking and breathing, visible abnormal growth at localized body parts, dramatic loss of body weight, ulcerated skin infections that did not respond to antibiotic treatment, and bleeding from the mouth, vagina, and/or anus, etc. These mice were sacrificed before death and all their vital organ/tissues were inspected. The major organs and tumors were isolated and processed for histological analysis. More than 95% of terminally-ill mice identified were diagnosed with grave pathological conditions including metastasizing tumors that had caused extensive damage of affected tissues, severe hyperplasia of reproduction organs or renal failure that would have caused death. The estimated time of natural death for these mice was ≤1 or 2 days from the time of euthanasia.

26For collecting mouse tissues and urine samples in DD cycles, mice were maintained in standard LD cycles for two weeks and then placed in 24hr DD cycles for two days before being sacrificed for sample collection. For collecting mouse tissue samples from jet-lagged mice, mice were placed in 24hr LD cycles for two days after jet-lag prior to tissue isolation at indicated ZT times.

27Mice were shifted every 3 days from 6 weeks of age between a 24hr LD cycle of 6:00am to 6:00pm light/6:00pm to 6:00am dark and a 24hr LD cycle of 10:00am to 10:00pm dark/10:00pm to 10:00am light.

28Mice were irradiated at 6 weeks of age at ZT10 with a single dose of 4Gy (16.8 cGy/sec) sublethal γ-radiation in a cesium-137 Gammacell.

29Tissue and tumor samples were fixed in formalin. Paraffin sections were prepared and stained with haematoxylin and eosin following standard procedures. All tumor types were confirmed by histological diagnosis.

30Northern and Western blot analyses were performed following the standard procedures as described previously. RNA expression was detected by Northern blot analysis and quantified using a Molecular Dynamics Storm 860 PhosphorImager/Fluorima. Protein expression detected by Western blot analysis was quantified using the ImageJ program.

31Anti-MDM2 2A10 (Calbiochem), PAb421 (Oncogene), anti-c-Fos (Santa Cruz), anti-c-Jun (Santa Cruz), anti-CREB S133 (Santa Cruz), anti-c-Myc (Santa Cruz), anti-ATM-S1981 (Calbiochem), anti-p53-S18 (R&D Systems), anti-MDM2 AB4 (Calbiochem), anti-MDM2-H122 (Santa Cruz) and anti-β-Actin (Sigma).

32Calvarial osteoblasts (preosteoblasts) were isolated, cultured, and treated with iso (Sigma, final conc. 10µM or 100µM) as described previously.

33Cells were fixed in 70% ethanol, incubated with PBS containing 50µg/ml propidium iodide, 0.2% Tween 20, and 1 mg/ml RNase at 4°C overnight, and then analyzed by a Becton Dickinson FACScan flow cytometer using CellQuest software (Becton Dickinson) as described previously,.

34Urine norepinephrine and epinephrine levels were measured using the Catecholamine Research Biotrak Assay System (Amersham) by TLC following manufacturer's instructions.

35All survival curves were plotted using the Kaplan-Meier survival analysis software. Other statistical analysis was assessed by Student's two-tailed t-test. Values were considered statistically significant at p<0.05.