Timing of expression of the core clock gene Bmal1 influences its effects on aging and survival

See allHide authors and affiliations

Science Translational Medicine  03 Feb 2016:
Vol. 8, Issue 324, pp. 324ra16
DOI: 10.1126/scitranslmed.aad3305

For clock ticking, timing matters

Ironically, antiaging product advertisements often promise to “slow down the clock.” But abolishing the circadian clock—for example, by knocking out Bmal1, a core clock gene—accelerates aging and shortens the life span in mice. As a result, Bmal1 knockout mice often serve as a model system in studies of the role of circadian rhythms in the aging process. Now Yang et al. show that the developmental timing of Bmal1 expression influences the circadian clock’s effects on aging and survival.

To assess the role of circadian rhythms in the aging process, the authors made conditional Bmal1 knockout mice that are missing the BMAL1 protein only during adult life. Unlike knockout mice that perpetually lack Bmal1 expression, the new conditional Bmal1 knockout mice displayed loss of circadian rhythm in wheel-running activity, heart rate, and blood pressure, but exhibited normal life spans, fertility, body weight, blood glucose levels, and age-dependent arthropathy; in fact, atherosclerosis and hair growth actually improved, despite obliteration of clock function. Another surprising observation was little changes in overall gene expression in the livers of adult-life Bmal1 knockout mice, even though there’s a quelling of expression of oscillating genes. Both prenatal and postnatal knockout mice displayed similar ocular abnormalities and brain astrogliosis.

Taken together, these findings reveal that many phenotypes thought to be caused by circadian rhythm disruption in conventional Bmal1 knockout mice apparently manifest as a result of clock-independent BMAL1 functions. Thus, the systemic role of the molecular clock in the biology of aging requires reinvestigation in order to increase the likelihood of translation for preclinical studies of the aging process.


The absence of Bmal1, a core clock gene, results in a loss of circadian rhythms, an acceleration of aging, and a shortened life span in mice. To address the importance of circadian rhythms in the aging process, we generated conditional Bmal1 knockout mice that lacked the BMAL1 protein during adult life and found that wild-type circadian variations in wheel-running activity, heart rate, and blood pressure were abolished. Ocular abnormalities and brain astrogliosis were conserved irrespective of the timing of Bmal1 deletion. However, life span, fertility, body weight, blood glucose levels, and age-dependent arthropathy, which are altered in standard Bmal1 knockout mice, remained unaltered, whereas atherosclerosis and hair growth improved, in the conditional adult-life Bmal1 knockout mice, despite abolition of clock function. Hepatic RNA-Seq revealed that expression of oscillatory genes was dampened in the adult-life Bmal1 knockout mice, whereas overall gene expression was largely unchanged. Thus, many phenotypes in conventional Bmal1 knockout mice, hitherto attributed to disruption of circadian rhythms, reflect the loss of properties of BMAL1 that are independent of its role in the clock. These findings prompt reevaluation of the systemic consequences of disruption of the molecular clock.


Circadian rhythms are biological processes that display endogenous and entrainable oscillations of about 24 hours. They are driven by a group of clock genes that are widely conserved across plants, animals, and bacteria (1). In mammals, the core clock genes, including Bmal1, Clock, Cry, and Per, are rhythmically expressed in both the suprachiasmatic nucleus (SCN)—the master clock that resides in the hypothalamus—and almost all peripheral tissues, in which they control the expression of numerous target genes in a circadian manner, influencing many physiological and biochemical processes (2). Chronic disruption of circadian rhythms has been associated with cardiometabolic morbidity in humans (3). In mice, deletion of Bmal1 prenatally disrupts clock-dependent oscillatory gene expression and behavioral rhythmicity coincident with reduced body weight, impaired hair growth, abnormal bone calcification, eye pathologies, neurodegeneration, and a shortened life span (47). However, although Bmal1 is the sole nonredundant gene in the core molecular clock, the degree to which phenotypes expressed in conventional knockout (cKO) mice reflect disruption of clock function is unknown.

Here, we describe the characterization of inducible Bmal1 KO (iKO) mice that express the gene during embryogenesis but not after birth. Despite ablation of clock function in both iKO and cKO mice, we observed striking physiological differences between the two model systems, prompting reevaluation of the systemic consequences of disruption of the molecular clock.


Loss of circadian rhythms

Bmal1 deletion in various tissues of the iKO mice (Bmal1f/f-EsrCre), including brain, heart, lung, liver, kidney, spleen, skeletal muscle, and epididymal fat, was confirmed by quantitative reverse transcription polymerase chain reaction (qRT-PCR). The iKO mice showed an 80% (liver) to 99% (brain and muscle) reduction of Bmal1 mRNA levels at Zeitgeber time 0 (ZT0) when Bmal1 expression is high (fig. S1A). Disruption of circadian behavior in iKOs was confirmed using running wheels. Before tamoxifen treatment, the Bmal1f/f-EsrCre mice showed normal rhythmic locomotor activities under both 12-hour/12-hour light/dark (LD) and constant darkness (DD), which is indistinguishable from their Cre Bmal1f/f control littermates (Fig. 1A). When mice were administered tamoxifen, Bmal1f/f mice (Ctrl) maintained their circadian behavior under DD, whereas all Bmal1f/f-EsrCre mice lost rhythmicity immediately after the treatment (Fig. 1A), which is similar to cKO mice (fig. S2A). Loss of circadian behavior in iKOs was still evident 15 months after Bmal1 deletion (Fig. 1C), suggesting the permanent disruption of circadian rhythms. The reduction in overall locomotor activity in cKOs (fig. S2B) (8) was not recapitulated in iKOs (Fig. 1, B and D), indicating that adult-life deletion of Bmal1 does not predispose mice to the usual age-related decline of wheel-running activity (9).

Fig. 1. Loss of circadian rhythms in iKO mice.

(A) Representative double-plotted actograms of wheel-running activity of 3-month-old Bmal1f/f and Bmal1f/f-EsrCre mice (red dots, tamoxifen treatment). Similar results were obtained in n = 8 to 9 mice per group. (B) Counts of wheel revolutions per hour from control mice (Ctrls) and iKO mice under conditions of DD (n = 8 to 9, Student’s t test; ns, no significant difference). (C) Representative double-plotted actograms of wheel-running activity from 18-month-old Ctrl and iKO mice under DD. Similar results were obtained in n = 6 to 7 mice per group. (D) Counts of wheel revolutions from 18-month-old Ctrls and iKOs under DD (n = 6 to 7; Student’s t test). (E) Representative radiotelemetry results of locomotor activity, systolic BP (SBP), and HR in Bmal1f/f and Bmal1f/f-EsrCre mice (red inverted triangles, tamoxifen treatment). Similar results were obtained in n = 3 mice per group. (F) Hepatic mRNA levels of canonical clock genes and clock-controlled gene Dbp were determined by qRT-PCR [n = 4 per genotype per time point; x axis, circadian time (CT); y axis, relative mRNA levels; *P < 0.05; **P < 0.01; ***P < 0.001, two-way analysis of variance (ANOVA)].

Consistent with disruption of core clock function, diurnal variation in heart rate (HR) and blood pressure (BP) was lost in iKOs (Fig. 1E), and circadian expression of hepatic clock genes was dampened (Fig. 1F). The variation of clock gene expression between ZT0 and ZT12 was also dampened in other tissues in iKOs (fig. S1B). Thus, behavioral, physiological, and molecular evidence for molecular clock disruption was present in the iKOs, consistent with what has previously been reported in cKOs (8, 10).

Conserved life span, weight, fertility, and blood glucose

Despite permanent loss of circadian rhythmicity in iKOs, the mice had a normal average life span of more than 2 years (Fig. 2A and fig. S3A). By contrast, the average life span of cKOs was just 9 months (fig. S3B) (6). Except for ocular abnormalities, the iKO mice generally exhibit no gross morphological defects, and body weight was conserved in both genders (Fig. 2B and fig. S2C). Similarly, the weight of organs examined in the iKOs did not differ from controls, except for the liver at 2 months after Bmal1 deletion (Fig. 2C). Although iKO mice are less fertile than normal mice (tamoxifen untreated), the fertility percentage was comparable to their tamoxifen-treated littermate controls (36% versus 30% in male, and 22% versus 27% in female; Fig. 2D), suggesting that the defect in fertility resulted from the tamoxifen treatment, not the consequent gene deletion or disruption of circadian rhythms. In contrast, the cKOs were completely sterile (fig. S2D). Glucose tolerance tests (GTTs) and insulin tolerance tests (ITTs) did not differ between Ctrls and iKOs (Fig. 2E).

Fig. 2. General status of iKO mice.

(A) Life span of Ctrls and iKOs with medians of 745 and 765 days, respectively (P = 0.9852, log-rank test). TAM, tamoxifen. (B) Body weight of Ctrls and iKOs for both genders (Student’s t test). (C) Ratio of organ to body weight in 5- and 11-month-old Ctrls and iKOs (**P < 0.01, multiple t tests with Holm-Sidak correction). (D) Fertility analysis in both male and female Ctrl, iKO, and untreated Cre+ mice (**P < 0.01; ***P < 0.001, χ2 test). (E) Blood glucose concentrations of Ctrls and iKOs that underwent GTT and ITT (n = 6 to 7; two-way ANOVA; no significant difference between Ctrls and iKOs at any time point).

Hair growth and arthropathy

Loss, graying, and growth inactivity of hair (telogen) are hallmarks of aging (11, 12). Indeed, conventional Bmal1 and Clock mutant mice demonstrate an increase in telogen follicles compared to controls (6, 13). Here, in an assay to assess hair follicle cycling in which hair is shaved and new hair regrowth is assessed (Fig. 3A), cKOs were shown not to enter the hair growth phase (anagen) as frequently as did wild-type mice (Fig. 3C), as previously reported. Unexpectedly, however, hair follicles in iKO mice entered anagen more frequently than did controls (Fig. 3B) even in aged (6 months and 1.5 years) mice (Fig. 3D). In young cKO mice, anagen follicles were absent 7 weeks after shaving (Fig. 3B), indicating that the follicles stayed in telogen. Anagen was also rare in wild-type and Ctrl mice; however, two-thirds of the iKO mice entered anagen at 7 weeks after shaving (Fig. 3B). Similarly, 12 weeks after shaving, all wild-type, Ctrl, and iKO mice had entered anagen, but more than half of the cKOs still had not done so (Fig. 3C). Expression of genes associated with anagen (Ccnd1 and Mki67) was up-regulated in iKOs compared with controls (fig. S4).

Fig. 3. Hair cycling and arthropathy.

(A) Six-week-old Bmal1f/f and Bmal1f/f-EsrCre mice were treated with tamoxifen. Meanwhile, dorsal hair was shaved. Pictures were taken 7 weeks after hair shaving. Mice with obvious hair regrowth are highlighted with orange boxes. (B and C) Contingency bar graphs show frequency distribution of hair regrowth in both cKO and iKO strains. The hair regrowth was observed 7 weeks (B) or 12 weeks (C) after hair shaving. WT, wild type. (D) Hair regrowth assay in 6-month-old and 1.5-year-old mice. The hair regrowth was observed 7 weeks after hair shaving. (E) Representative photographs of Alizarin Red–stained ribcages and hindlimbs from both cKO and iKO strains. Blue circles indicate calcification.

Accelerated age-related arthropathy has been reported in cKOs and was studied here using Alizarin Red–stained ribcages and hindlimbs, as previously described (4, 5). As expected, we observed abnormal calcification in the costosternal junctions and calcaneal tendons of all cKOs (n = 5) at 24 weeks of age (Fig. 3E). By contrast, abnormal calcification was not evident in any 30-week-old iKOs (n = 5, 24 weeks after Bmal1 deletion). Calcification of the calcaneal tendon was observed in some 48-week-old iKO mice (3 of 8, 24 weeks after Bmal1 deletion) but was indistinguishable from that in littermate controls (4 of 10).

Differential impact on atherogenesis

The molecular clock has been implicated in various aspects of cardiovascular function, including BP homeostasis and thrombogenesis (14). Atherogenesis is accelerated in ClockΔ19/Δ19 mutant mice (15). Here, we compared conventional and adult-life Bmal1 deletion in mice on a hyperlipidemic low-density lipoprotein receptor–deficient (LDLR−/−) background that were fed a high-fat diet (HFD). Similar to ClockΔ19/Δ19 mutant mice, en face aortic lesion analysis revealed marked acceleration of atherogenesis in both male and female cKOs (Fig. 4A), whereas, unlike ClockΔ19/Δ19 mutant mice, cKOs show plasma total cholesterol and triglyceride levels comparable to those of wild-type mice (fig. S5A). By contrast, atherogenesis relative to Ctrl mice was retarded when Bmal1 was deleted in adulthood (Fig. 4B). Adult-life deletion of Bmal1 did not affect BP or HR, at least at ZT8 to ZT9 (fig. S5B), whereas the increase in body weight in HFD-fed mice was reduced in male iKOs. Female iKOs showed a trend toward a reduction in body weight (fig. S5C). Consistently, aortae from iKOs showed reduced expression of proinflammatory genes, including CD68 (cluster of differentiation 68), MCP-1 (monocyte chemotactic protein 1), and iNOS (inducible nitric oxide synthase), compared to controls (fig. S5D). Among these genes, MCP-1 is a direct target gene of Bmal1, as revealed by ChIP-Seq (chromatin immunoprecipitation sequencing) in mouse liver (16).

Fig. 4. HFD-induced atherosclerosis.

(A and B) Both cKO (A) and iKO (B) mice and their littermate controls were in an LDLR−/− background and fed an HFD (42% kcal from fat) from 12 weeks of age. Sixteen weeks after HFD treatment, mice were euthanized and whole aortas were dissected and stained with oil red O. Plaque areas (shown in red) were quantified with Image-Pro. The percentage of plaque area relative to the entire area was calculated (*P < 0.05; **P < 0.01; ***P < 0.001, Student’s t test).

Ocular abnormalities and astrogliosis

Despite the differential effects on hair regrowth, arthropathy, and atherogenesis, ocular abnormalities similar to those observed with prenatal Bmal1 depletion were evident in the iKOs (Fig. 5, A to D). By 8 months of age (5 months after Bmal1 deletion), about 80% of iKOs developed cloudy changes in one or both eyes, whereas none of their littermate controls displayed any visible abnormalities. Histological examination revealed various pathological changes, including corneal neovascularization, keratinization, and progressive inflammation in the iKOs.

Fig. 5. Ocular abnormalities and astrogliosis in iKO mice.

(A to D) Representative gross images (left) and hematoxylin and eosin (H&E)–stained sections of eyes (middle and right) from Ctrl (A) and iKOs (B to D). (A) Unremarkable globe from a Ctrl mouse. AC, anterior chamber; Co, cornea; Ir, iris; Le, lens; Re, retina. (B) Pathologic changes in a male mouse eye. Grossly, there is a leukoplakic plaque on the cornea (left, arrow). Histologically, the cornea appears thickened with keratinization of the epithelium (right, arrow) and chronic inflammation and neovascularization of the stroma (right, arrowhead). (C) In the contralateral eye of the same mouse, the corneal surface appears irregular with a flattened chamber. Histologically, the cornea appears thickened. The retina is adherent to the lens. The corneal epithelium is attenuated (right, arrow) with chronic inflammation and neovascularization in the stroma. A subcapsular anterior cataract is present (right, arrowhead). (D) A female mouse eye shows an irregular corneal surface with leukoplakia (left, arrow) and corneal neovascularization (right, arrow). (E) Immunofluorescence staining of 4′,6-diamidino-2-phenylindole (DAPI) (all nuclei), NeuN (neuronal nuclei), BMAL1, and GFAP (activated astrocytes) in the motor cortex from cKO, iKO, and nKO mouse strains. Representative images show loss of Bmal1 immunoreactivity in all KOs, which is accompanied by severe astrogliosis (GFAP panels). Similar results were obtained in n = 5 mice per group. Scale bar, 150 μm.

We have reported that cKOs and neural-specific Bmal1 KO mice (Bmal1f/f-NestinCre; nKO) exhibit severe astrogliosis (7). Here, we examined the brains of 11- to 12-month-old iKO mice and found them to be nearly indistinguishable from those of cKO mice (aged 6 months) or nKO mice (aged 11 to 12 months), with undetectable BMAL1 immunoreactivity in neurons and severe astrogliosis, which was reflected by GFAP (glial fibrillary acidic protein) staining throughout the cerebral cortex (Fig. 5E). Transcriptional analysis revealed similar changes in Gfap mRNA expression (fig. S6). However, such neurodegenerative changes did not affect the life spans of iKOs (Fig. 2A) or nKOs (fig. S3C). It is not surprising that this pathological brain phenotype did not lead to premature death, because similar levels of inflammation and synapse loss are observed in some mouse models of amyloid β–mediated cerebral amyloidosis, and these mice have normal life spans (17).

Loss of the circadian transcriptome

To characterize gene expression profiles, we performed RNA-Seq using samples of liver, where most circadian genes are expressed (18). As expected, all clock genes oscillated in control mice, but were expressed at relatively constant levels in iKOs (fig. S7). This expression pattern was highly consistent with the qRT-PCR results (Fig. 1F). A circadian analysis was performed using JTK (Jonckheere-Terpstra-Kendall)_CYCLE algorithm (19) to explore 24-hour oscillations in transcript abundance. We used a 5% false discovery rate (FDR) threshold to detect circadian genes in control mice and found that 14.5% of genes (5457 among 37681; data file S1) were rhythmically expressed (Fig. 6A). By contrast, even those genes that top the list in control mice displayed no circadian expression pattern in iKOs (table S1). Only one gene, Erh, showed circadian variability at the mRNA level in iKOs (Fig. 6B), illustrating how few genes are rhythmically expressed independent of the classic circadian clock.

Fig. 6. Hepatic transcriptome.

RNA samples from iKOs and their littermates under DD condition were used for RNA-Seq. (A) Of the 37,681 transcripts, 5457 exhibited circadian variability in Ctrl mice. By contrast, only one gene in iKOs exhibited a circadian pattern. Heat map rendering of the temporal gene expression pattern of these 5457 circadian genes is shown with the average gene expression level measured for each time point. (B) Erh is the only hepatic gene that shows a circadian expression pattern in iKOs, and the q value (JTK_CYCLE analysis) and fold change (ratio of peak/trough) are shown [x axis, circadian time; y axis, fragments per kilobase of transcript per million mapped reads (FPKM)]. (C) Venn diagram (sizes not to scale) depicting the number of differentially expressed genes in cKO and iKO strains. (D) Mouse Genome Informatics Mammalian Phenotype Level 3 enrichment analysis for cKO differentially expressed genes. The top 17 (adjusted P < 0.05) phenotypes related to these genes are given. Overlap, number of appearing gene/number of background genes. (E) Mup3, Serpina3k, and Apoa1 are the only differentially expressed genes in the iKO strain.

Phenotype enrichment analysis of differentially expressed genes

To understand the potential reasons why there is a striking phenotypic difference between iKOs and cKOs despite a shared ablation of circadian rhythms, we compared hepatic gene expression profiles between wild-type versus cKOs and Ctrl versus iKOs, irrespective of time (data file S2). An FDR threshold of 5% was used for detection of differentially expressed genes for both KO strains, although the specific value of this cutoff did not alter the fact that the iKOs had many fewer differentially expressed genes than did the cKOs (table S2); from these analyses, we found 462 differentially expressed genes in cKOs and 3 in iKOs (Fig. 6C). Enrichment analysis of these 462 genes revealed multiple potential phenotypes that are consistent with what might be expected, including abnormal cardiovascular function, induced morbidity and mortality, abnormal skin morphology, abnormal metabolism, abnormal survival, abnormal muscle physiology, and abnormal immune function (Fig. 6D and data file S3). However, only three genes in the iKOs (without overlap with the cKOs) were identified as differentially expressed (Fig. 6, C and E). Enrichment analysis showed that no category reached the cutoff (P < 0.05), perhaps because of the small number of genes involved (data file S3).


Among canonical clock genes, Bmal1 is the only one whose deletion results in complete loss of circadian rhythms (8); therefore, Bmal1 KO mice have become a valuable resource for study of the effects of circadian disruption. Other than loss of circadian rhythms, however, these mice also exhibit a greatly reduced life span and develop various features consistent with premature aging (47). Because of the absence of Bmal1 expression during development in cKO mice, it is not clear whether these phenotypes relate to its absence during embryogenesis or postnatal development or, indeed, whether these phenotypes reflect disruption of the molecular clock or loss of other properties of BMAL1. Here, we compared iKOs with cKOs and found that both approaches to gene deletion similarly disrupt behavioral and transcriptional indicators of clock-dependent circadian function, and both types of KO mice exhibit markers consistent with accelerated aging [such as ocular abnormalities and brain astrogliosis (table S3)]. In contrast, other biomarkers of aging and premature death, so evident in cKOs, including a markedly shortened life span, infertility, reduced body weight, organ shrinkage (6), impaired glucose homeostasis (20, 21), and abnormal bone calcification (4), are not replicated in the iKOs. Further, two aging-related phenotypes, hair growth inactivity and a predisposition to atherosclerosis, were reversed in iKOs. To characterize Bmal1 KO mice at the level of gene expression as well as to determine potential mechanisms underlying the general difference between cKO and iKO strains, we performed RNA-Seq using liver samples. Although the various phenotypes may not be explained simply by changes of the transcriptome in any given tissue, liver affords an appropriate initial opportunity to address this question, because it expresses most circadian genes (18). All time-effect genes were dampened in iKO mice without obvious changes in time-independent genes. These striking differential effects between the two strains of mice, despite a shared ablative effect on clock-dependent circadian biology, question the attribution of many phenotypes in the widely used cKOs to disruption of the molecular clock.

Circadian recruitment of clock proteins including BMAL1 to DNA binding sites is believed to contribute to the oscillation of target genes (16). However, Bmal1 is expressed in mice throughout embryogenesis (22) but does not oscillate, either in the SCN (23, 24) or in peripheral tissues (25), suggesting that clock regulatory loops may not be well established until late in development. Hence, Bmal1 expression might exert important nonclock functions during early development, and its deletion during this period might account for premature aging and various related phenotypes observed in the adult mice. Alternatively, the physiologic effects of Bmal1 may vary across tissues and defects in one tissue may compensate for those in another when analyzing multi-tissue mutants. This possibility is not addressed in the present study. However, although shortened life span has yet to be reported in any prenatal tissue-specific Bmal1 KO mice, some do exhibit phenotypes that are not evident in adult-life global KOs. For example, impaired glucose tolerance was reported in prenatal hepatocyte–specific (19, 25) and pancreatic β cell–specific (26) Bmal1 KO mice. This is consistent with cKOs, but is not seen in adult-life global KOs.

To our surprise, our cKO and iKO mice showed opposite phenotypes in hair growth and atherogenesis. We found increased growing follicles in all young, middle-aged, and old iKO mice as well as consistent expression of hair growth–promoting genes Ccnd1 and Mik67 in the skin. ChIP-Seq has revealed that the Ccnd1 gene contains binding sites for all core clock proteins including BMAL1 (16), which implies that Ccnd1 is directly clock-controlled. However, its expression in cKOs was not changed. It is possible that this regulation is masked and hair growth is retarded by the overall premature aging phenotype. Pan et al. found impaired cholesterol metabolism and enhanced atherosclerosis in HFD-fed Clock mutant mice generated in an LDLR−/− background, implicating the circadian clock in atherogenesis (15). Consistent with their finding, we observed increased atherogenesis in Bmal1 cKOs. However, unlike the case for Clock mutant mice, Bmal1 deletion does not alter cholesterol metabolism, which suggests that disruption of the two clock-directed genes influences atherogenesis by distinct the mechanisms. By contrast, adult-life Bmal1 depletion restrains atherogenesis coincident with dampened expression of proinflammatory genes in the vasculature and a reduction in the weight gain induced by the HFD. The relative contribution of local and systemic effects to this phenotype remains to be determined. Ocular abnormalities and brain astrogliosis were preserved in iKO mice, consistent with Bmal1 exerting a direct role in the eye and neural system irrespective of developmental issues. However, it remains unknown if these phenotypes are caused by loss of circadian rhythms or loss of nonclock functions of Bmal1, because deletion of core clock genes not only disrupts rhythmicity of genes relevant to expression of these phenotypes but also may have striking consequences on their overall expression levels (26, 27).

The apparent restraining influence of Bmal1 expression during embryogenesis on aging-related phenotypes and, indeed, life span in the adult mice is reminiscent of the Barker hypothesis, which postulated that factors impinging on fetal development would influence the expression of aging and life span in adult humans (28). Insight into the impact of environmental factors and maternal behavior, such as smoking and alcohol consumption, on postnatal health has suggested an epigenetic basis for this hypothesis (29). Here, in the case of Bmal1, we raise the additional possibility of a direct genetic effect on such phenomena. A provocative consideration is that the anticipatory role of the clock might condition the impact of environmental insults, attenuating them at certain times of day.

In summary, despite a shared impact on circadian function, many phenotypes observed in conventional KOs are not observed when Bmal1 is deleted in adult mice. This prompts reevaluation of the systemic role of the molecular clock and its importance in the biology of aging.


Study design

The primary objective of this study was to investigate the role of circadian rhythms in aging in adult mice. We used Bmal1-inducible KO mice to bypass possible issues about developmental defects in conventional KOs. For all experiments, the number of independent replicates is outlined in each figure or legend unless individually plotted. Sample sizes were based on the desired magnitude of the difference to be detected, and the variance of the estimates was based on previous data where available and setting α = 0.05 and 1 − β = 0.9. Mice in all experiments were randomized into various groups. The studies were not performed under double-blind conditions. The experimenters were not blinded to the identities of individual groups. Animals were not withdrawn from the studies according to predetermined criteria in the Institutional Animal Care and Use Committee (IACUC) protocols, nor were they excluded from any of the analyses.


Both cKO and Bmal1f/f mice (30) were obtained from C. Bradfield (University of Wisconsin, Madison, WI). The floxed mice were crossed with a tamoxifen-inducible universal Cre (EsrCre) mice or NestinCre mice (The Jackson Laboratory) to yield Bmal1f/f-EsrCre mice and Bmal1f/f-NestinCre (nKO) mice, respectively. To generate iKO mice, 3-month-old (unless specified) Bmal1f/f-EsrCre mice were treated with tamoxifen as previously described (31). Cre Bmal1f/f littermates treated with tamoxifen served as controls (Ctrl). For atherogenesis, all mice were crossed into an LDLR−/− background. All mice were housed under 12-hour/12-hour LD conditions with free access to food and water unless specified. All procedures were approved by the University of Pennsylvania IACUC.

Wheel-running activity

Young (3 months, both cKO and iKO strains) and aged (18 months, iKO strain only) mice were individually housed in cages equipped with running wheels (Coulbourn Instruments). Cages were kept within a ventilated and light-tight isolation chamber with a computer-controlled lighting system. Locomotor activity was continuously recorded and analyzed using ClockLab software (Actimetrics). For the young group, mice from both cKO (n = 4 per genotype) and iKO (n = 8 to 9 per genotype) strains were housed in 12-hour/12-hour LD and then released into DD. After 1 week, Bmal1f/f and Bmal1f/f-EsrCre mice were treated with 5 mg of tamoxifen by gavage for five consecutive days followed by recording for 10 days.

Radiotelemetry reading

The implantation of radiotelemetry (model no. TA11PA-C20, Data Sciences International) was performed as previously described (31). After 1 week of recovery, BP, HR, and locomotor activity were monitored continuously in 3-month-old Bmal1f/f (n = 3) and Bmal1f/f-EsrCre (n = 3) mice using the Dataquest LabPro Acquisition System. Mice were housed under LD and then released into DD. After 1 week, mice were treated with 5 mg of tamoxifen by gavage for five consecutive days followed by recording for 15 days.

Fertility analysis

Both genders were used for fertility analysis. Eight-week-old Bmal1f/f and Bmal1f/f-EsrCre mice were treated with tamoxifen. At 10 weeks old, each iKO, control, or untreated Cre+ littermate was mated with a wild-type C57Bl/6 mouse. All breeding cages were monitored for birth daily for 9 months.

Glucose tolerance test

Twelve-month-old mice (10 months after tamoxifen treatment) were fasted for 16 hours before the GTT. Blood was obtained from a tail cut and was assessed for fasting glucose levels using a OneTouch Ultra 2 (LifeScan, Johnson & Johnson) glucometer. The mice then received a glucose solution (0.2 g/ml; 2 g/kg body weight) delivered by oral gavage. At 15, 30, 60, 90, and 120 min after the administration, dried blood was quickly removed from the tail wound and fresh blood was collected again to measure the glucose concentration.

Insulin tolerance test

Twenty-month-old mice (18 months after tamoxifen treatment) were fasted for 4 hours before ITT. After baseline blood glucose measurement, insulin (1 U/kg) (Eli Lilly) was injected intraperitoneally and blood glucose was measured again at time points of 30, 60, 90, and 120 min after injection.

Hair regrowth assay

To characterize hair growth, the backs of mice were shaved and animals were monitored for hair regrowth for 3 months. For RNA analysis, dorsal skins from 8-week-old mice (1 week after tamoxifen treatment for iKO strain) were collected for RNA extraction and quantitative RT-PCR analysis for Ccnd1 and Mki67.

Bone staining

Skeletons were stained using Alizarin Red as previously described (5) with modifications. Briefly, skinned and eviscerated animal carcasses were submerged in 1% KOH for 4 hours followed by incubation in 2% KOH overnight. Carcasses were then stained in 2% KOH containing 0.004% Alizarin Red (Sigma) for 2 days. After clearing in 2:1:2 glycerin/benzyl alcohol/70% ETOH, stained skeletons were photographed.


All mice were on an LDLR−/− background and fed an HFD (42% kcal from fat, Harlan Laboratories) for 16 weeks from 12 weeks of age. Mouse aortic trees were prepared and stained, and atherosclerotic lesions were quantified as described (32). Briefly, the entire aorta from the aortic root to the iliac bifurcation was collected and fixed in formalin-free fixative Prefer (Anatech Ltd.). Adventitial fat was removed. The aorta was opened longitudinally, stained with Sudan IV (Sigma), and pinned down on black wax to expose the intima. The extent of atherosclerosis was determined by the en face lesion area percentage to the entire intimal area.

Lipid analysis in plasma

Blood was collected from the vena cava of CO2-euthanized mice. Plasma total cholesterol and triglycerides were measured by commercial kits from Wako Chemicals.

Tail-cuff BP and HR measurement

SBP and HR were measured in conscious mice at ZT8–9 using a tail cuff system (Visitech Systems Inc.) as previously described (33).

Ocular pathology

Animals were euthanized, and both eyes were harvested and fixed in 10% paraformaldehyde solution. After gross examination, globes were routinely processed and embedded in paraffin. Sections were obtained, stained with H&E, and reviewed by an ocular pathologist.

Immunostaining of brain sections

Cryosections from the prefrontal cortex to caudal hippocampus were prepared as previously described (7). Sections were washed in tris-buffered saline (TBS), blocked for 30 min in TBS containing 3% donkey serum and 0.25% Triton X-100, and then incubated overnight in TBS plus 0.25% Triton X-100 with 1% donkey serum containing rabbit GFAP (Dako), BMAL1 (Novus), or mouse NeuN primary antibodies at 4°C. Sections were washed, then incubated in TBS containing Alexa Fluor 488–conjugated anti-rabbit or Alexa Fluor 568–conjugated anti-mouse secondary antibodies for 1 hour at room temperature, and then washed again in TBS. Sections were mounted and imaged using an epifluorescent microscope.

Quantitative RT-PCR

Total RNA from various tissues was isolated using the Qiagen RNeasy Kit. Reverse transcription was performed using an RNA-cDNA kit (Applied Biosystems). Real-time PCR was performed using ABI TaqMan primers and reagents on an ABI Prism 7500 thermocycler according to the manufacturer’s instructions. All mRNA measurements were normalized to β-actin mRNA levels.

Sample preparation for RNA-Seq

Twenty-four male iKO mice aged 4 to 6 months (2 weeks after tamoxifen treatment) and their littermates were maintained under 12-hour/12-hour LD conditions and then released into DD condition. Starting at 36 hours after release, four mice per genotype were euthanized in darkness every 4 hours for 20 hours (six time points). Liver samples were quickly excised and snap-frozen in liquid nitrogen. Total RNA was isolated using Qiagen RNeasy Kit, and the RNA integrity was checked using an Agilent Technologies 2100 Bioanalyzer. RNA samples were subjected to two rounds of hybridization to oligo(dT) beads. The resulting mRNA was then used as template for double-stranded complementary DNA (cDNA) synthesis. After purification, cDNA samples were fragmented with nebulization technique to yield fragment sizes of 100 to 300 base pairs and subjected to end-repair, adenylation, ligation of Illumina sequencing adapters, and PCR amplification. The PCR product was purified and used directly for cluster generation and sequencing (HiSeq 2000).

When comparing iKOs with cKOs irrespective of time points, we randomly picked one mouse per genotype per time point from iKOs and littermates. To match these samples, six male cKO mice and six littermate wild-type controls (one mouse per genotype per time point) were euthanized for liver collection. Samples for RNA-Seq were prepared as described above.

RNA-Seq analysis

RNA-Seq data were aligned to the mouse genome build mm9 by STAR version 2.4.2a (34). Data were normalized using a resampling strategy as described previously (35); the code is available at The normalized counts thus produced were then used for the subsequent statistical analyses. JTK_CYCLE was used to explore 24-hour oscillations in transcript abundance (19). Differential expression analysis was performed by standard permutation methods to control the FDR (36). The list of resulting differentially expressed genes was used for enrichment analysis using Enrichr ( All data have been submitted to the Gene Expression Omnibus (GEO) database (accession nos. GSE70497, GSE70499, and GSE70500).

Statistical analysis

All statistical tests were two-sided. A log-rank test was used for survival curve analysis, and χ2 test was used for contingency data. For other comparisons except RNA-Seq results, Student’s t test or one-way ANOVA was used when a single variable was compared between two or more genotypes, and two-way ANOVA with Bonferroni’s posttest was used when multiple variables were compared between genotypes. The cutoff for significance was P < 0.05 (*). #P < 0.1, **P < 0.01, ***P < 0.001. In all figures with error bars, the graphs depict means ± SEM.


Fig. S1. Validation of Bmal1 deletion and dampening effect of other clock genes.

Fig. S2. Wheel-running activity, body weight, and fertility in cKO mice.

Fig. S3. Life span of iKO, cKO, and nKO mice.

Fig. S4. mRNA levels of Ccnd1 and Mki67 in skin.

Fig. S5. Other parameters in HFD-fed mice.

Fig. S6. iKO mice display similar Gfap induction in the brain as cKO.

Fig. S7. RNA-Seq results revealed dampening effect in core clock genes in iKO mice.

Table S1. Top 20 oscillating hepatic genes (JTK_CYCLE q value) in Ctrl mice show no circadian pattern in iKO mice.

Table S2. The ratio of differentially expressed gene numbers in cKO strain to the numbers in iKO strain.

Table S3. Summary of phenotypes of cKO and iKO mice.

Data file S1. Circadian transcriptome.

Data file S2. Differentially expressed genes irrespective of time points.

Data file S3. Phenotype enrichment analysis of differentially expressed genes.


  1. Acknowledgments: We thank C. A. Bradfield (University of Wisconsin, Madison, WI) for conventional Bmal1 KO mice and Bmal1f/f mice. We thank E. J. Kim, N. Lahens, K. Hayer, F. Colden, A. Pizarro, and A. Srinivasan (University of Pennsylvania, Philadelphia, PA) for their support in RNA-Seq analysis. We thank Y. Zheng (University of Pennsylvania, Philadelphia, PA) for discussions about hair regrowth. We thank W. Yan (University of Pennsylvania, Philadelphia, PA) for technical assistance. We thank A. Sehgal for reading and commenting on this manuscript. Funding: This work was supported by a grant from the NIH (HL097800) and the University of Pennsylvania Genomics Frontiers Institute’s Translational and Personalized Genomics Centers Initiative. G.A.F. is the McNeil Professor of Translational Medicine and Therapeutics. Author contributions: G.Y. and G.A.F. conceived the project. G.Y. and L.C. performed and analyzed most of the experiments in this study. G.P. prepared cKO samples for RNA-Seq. E.S.M. and V.L. conducted staining of brain and eye, respectively. W.-L.S. performed atherosclerosis experiment for cKO mice. S.C.M. collected iKO organs for weight measurement. T.G. designed and conducted the RNA-Seq experiments, and G.R.G. did the data analysis. G.C. discussed the data of hair growth assay. G.R.G., E.S.M., V.L., and G.C. provided critical intellectual input in manuscript preparation of their pertinent parts. G.Y., L.C., and G.A.F. wrote the paper. Competing interests: The authors declare that they have no competing interests. Data and materials availability: RNA-Seq data can be retrieved from the GEO database under accession nos. GSE70497, GSE70499, and GSE70500.

Stay Connected to Science Translational Medicine

Navigate This Article