Replicative senescent human cells possess altered circadian clocks with a prolonged period and delayed peak-time

Replicative senescence of TIG-3 cells

To evaluate the changes in the circadian clock during senescence, we chose replicative senescence, in which cell cycle would be irreversibly arrested due to the limitation of cell division. TIG-3 cells, which are normal primary diploid human fibroblast cells derived from fetal male lung [31], were serially passaged until they ceased proliferation, that is, reached replicative senescence. When we obtained TIG-3 cells, they had already been passaged 20 times. Therefore, we performed serial passaging starting from Passage 22 (P22). In order to assess at which passage cells begin to cease proliferation, individual and cumulative population doubling level (iPDL and cPDL, respectively) were recorded (Fig.1A and B). Here, we defined a cease of cell proliferation (CCP) as the point at which iPDL value dips below 1.0. Proliferation of TIG-3 cells were fairly constant up to P32, before which iPDLs were above 2. However, iPDLs started to decrease from P33 and TIG-3 cells reached CCP at P37 (Fig.1B).

Senescent cells have unique features, such as enlarged, flattened cell shape, increased senescence-associated β-galactosidase (SA-β-Gal) activity [34,35], and increased expressions of cell cycle inhibitors (p16^INK4a, p19^ARF and p21^CIP1) [35–37]. We therefore investigated the aforementioned features to confirm that serially passaged TIG-3 cells indeed exhibit senescence features, and to use these features as the basis to group the cells into two categories: proliferative and senescent cells. Representative pictures of SA-β-Gal assay at P25 and P39 are shown in Fig.1C. Senescent cells, which were stained in blue, were observed at P39, in which 43.5% of cells were SA-β-Gal positive, but only 0.8% were positive at P25. Fig.1D represents the percentages of SA-β-Gal positive cells at several passages. SA-β-Gal positive cell number was dramatically increased from P36, at which iPDL was already below 2.0 (Fig.1B), suggesting that decrease in proliferation is due to replicative senescence. Moreover, we observed enlarged, flattened cells at P39 compared to the cells at P25 (Fig.1C), supporting that many cells have already turned senescent at P39. We next measured the expression levels of senescence marker genes, p21^CIP1 and p16^INK4a, which are known to initiate and maintain cellular senescence, respectively [38], at P27, P38 and P41. Expression levels of p21^CIP1 increased with passages (Fig.1E upper panel). Expression levels of p16^INK4a was high at P41, but not at P27 and P38 (Fig.1E bottom panel). These results indicated that P38 and P41 were at the senescence phase. All the results above indicate that the serially passaged TIG-3 cells indeed reached replicative senescence. Furthermore, we deduced that TIG-3 cells are in the proliferative phase until P32, followed by the transition phase from P33 to P35, and the senescence phase after P36. Hence, in the subsequent parts of this study, TIG-3 cells of P25-29 and of P36-41 were utilized as proliferative and senescent cells, respectively. We performed three independent proliferation assays and could reproduce similar results as shown in Fig.1.

Senescent TIG-3 cells show altered characteristics of circadian clock

To address the changes in the circadian clock during senescence, virus containing bmal1 promoter-driven luciferase gene [39] was infected to the proliferative and senescent TIG-3 cells obtained from three independent proliferation assays. The infected cells were synchronized with dexamethasone (Dex) and subjected to the real-time luciferase assay. Fig.2A shows the superimposition of relative oscillatory patterns of luciferase (See Fig. S1 for raw data of oscillation patterns). In order to exclude the possibility that the phase and/or luciferase intensity of the first peak are modified by acute effect by Dex treatment, we normalized oscillation patterns using the first trough value. The oscillation pattern in the senescent cells did not overlap with that in the proliferative cells and seemed to be delayed (Fig.2A). To investigate the trough-time difference that occurs between the proliferative and senescent cells, trough-times were extracted from each data. The 1^st trough-time of the proliferative and senescent cells were 23.7 ± 0.1 hr and 26.1 ± 0.2 hr, respectively (Fig.2B). The 1^st trough-time of the senescent cells was 2.4 ± 0.2 hr delayed with significant difference (Fig.2C). Differences of 2^nd trough-time between the proliferative and senescent cells were 4.0 ± 0.2 hr (Fig.2C), indicating that trough-time differences have increased. This result suggests that the period length of circadian clock in the senescent cells is longer compared to that in the proliferative cells. To evaluate the period length of the proliferative and senescent cells, we analyzed the duration between two subsequent trough-times. The 1^st period, which is the duration between 1^st and 2^nd trough-times, was 23.1 ± 0.1 hr in the proliferative cells. In contrast, the 1^st period in the senescent cells was significantly longer than that in the proliferative cells (24.7 ± 0.1 hr; p-value, 5.35 x 10^-14; Fig.2D). We also analyzed the peak-time differences and the period using 2^nd and 3^rd peak-times (Fig.S2). Consistent with trough-time differences, differences of 2^nd and 3^rd peak-time between the proliferative and senescent cells were 3.6 ± 0.2 and 4.0 ± 0.3 hr, respectively (Fig.S2). Again, the 2^nd period, which is the duration between 2^nd and 3^rd peak-times, in the senescent cells also demonstrated a longer period compared to the proliferative cells (23.6 ± 0.1 hr vs 24.2 ± 0.2 hr; p-value, 0.03; Fig.S2).

Since the amplitude of circadian clock has been reported to decline with aging at transcriptional [30,40], neural activity [41] and locomotor activity levels [24–26,28], we proceeded to analyze the range of oscillations in TIG-3 cells. To measure the range of oscillation, the detrended luciferase intensity between top to nadir was determined (Fig.S3). Range of the 1^st oscillation (A1) was set to 1 and the damping ratios of A2 and A3 were calculated. We found that, unlike previous reports at the tissue and animal levels, the damping ratios between the proliferative and senescent TIG-3 cells were comparable (Fig.2E).

To prove that period lengths and acrophase are different between the proliferative and senescent cells, we mathematically analyzed the data of Fig.2 using the Cosinor software, provided by Dr. R Refinetti [42]. Period lengths were 24.09 ± 0.20 hr and 25.47 ± 0.28 hr for the proliferative and senescent cells, respectively (p-value, 1.56 x 10^-4; Table 1). The acrophases were -238.71 ± 1.23 degree and -272.54 ± 1.69 degree for the proliferative and senescent cells, respectively (p-value, 5.79 x 10^-27; Table 1), indicating that the phase of senescent cells is significantly delayed compared to that of proliferative cells.

In summary, we revealed that the Dex-synchronized senescent TIG-3 cells display a longer period length and the delay of peak-time compared with the proliferative cells, suggesting that senescent cells possess the altered circadian clock.

Input pathways are comparable between the proliferative and senescent cells

Circadian clock consists of three components; an oscillator with a period of 24 hr, input pathways to the oscillator to allow entrainment of the clock, and output pathways to regulate circadian biochemical, physiological and behavioral rhythms [43,44]. In order to evaluate whether a longer circadian period and delayed peak-time in the senescent TIG-3 cells is the result of the altered oscillator or input pathways, we compared the circadian oscillation patterns of luciferase synchronized by Dex and Forskolin (Fsk). Forskolin is another well-known clock-entrainment factor [14,45,46], activates cAMP-CREB signaling pathway and transcribes per1 gene via the cAMP response elements (CRE) in per1 promoter [47]. In contrast, Dex binds to and activates glucocorticoid receptor and transcribes per1 and per2 genes via the glucocorticoid response element (GRE) in per1 and per2 promoters [48,49]. In the proliferative cells, real-time luciferase monitoring assays demonstrated that circadian oscillation patterns of luciferase activity synchronized by Dex or Fsk show almost similar phase with the same intensity (Fig.3A left panel and Fig.S4). However, consistent with the previous report [46], there was a significant peak-time difference triggered by these two treatments. The second peak of Fsk-treated proliferative cells appeared 2.07 hr earlier than that of Dex-treated proliferative cells (p-value, 0.02; Fig.3B). In the senescent cells, the peak-time of the Fsk-treated cells was also significantly earlier compared to the Dex-treated cells (the difference was 1.67 hr; p-value, 0.03; Fig.3B). Importantly, the peak-time difference between the proliferative and senescent cells were similar (Fig.3B). Although the first peak and the amplitude of Fsk-treated senescent cells was small and weak, respectively, the above result suggests that both input pathways triggered by Dex or Fsk might not be functionally impaired in senescent cells. Rather, the oscillator of circadian clock was altered with senescence.

Infection efficiency does not affect the oscillation patterns of luciferase

It is also interesting to note that the luciferase activity was higher in the senescent cells than in the proliferative cells (Fig.3A, S1 and S4), giving rise to the possibility that BMAL1 gene expression is upregulated in the senescent cells, because the luciferase gene used in this study was driven by the mouse bmal1 promoter [39]. To confirm this possibility, we performed qPCR to compare endogenous BMAL1 transcripts in the non-entrained proliferative and senescent TIG-3 cells. The transcript level of BMAL1 in the senescent cells was, however, similar or slightly downregulated compared to that in the proliferative cells (Fig.4A). Moreover, circadian expression profile of BMAL1 in the senescent cells after Dex-entrainment were also similar or lower than those in the proliferative cells (Fig.7D). These results indicate that the high luciferase activity in the senescent cells might not be due to the upregulation of luciferase transcription, suggesting that infection efficiency is high in the senescent TIG-3 cells. Therefore, to rule out the possibility that the infection efficiency affects luciferase oscillation patterns, cells were then infected with serially diluted viruses. As expected, luciferase intensities were decreased in a virus-dose dependent manner in both the proliferative and senescent cells (Fig.S5). However, the relative oscillation patterns completely overlapped within the same passages (Fig.4B), demonstrating that the infection efficiency has no effect on the oscillation patterns of luciferase.

Half-life of luciferase activity is comparable between the proliferative and senescent cells

Various cellular metabolisms/homeostasis are altered in senescent cells [50], for example, energy production in senescent fibroblast cells shifts to glycolysis with little to no change in oxidative phosphorylation [51]; while protein homeostasis through both ubiquitin proteasome pathway and autophagy is also impaired [2,52,53]. These findings raised the possibility that altered metabolisms/homeostasis in the senescent cells might have effects on luciferase activity, which is independent from the circadian clock. To address this possibility, the half-life of luciferase activity in the proliferative and senescent cells were monitored by performing a real-time luciferase monitoring assay (Fig.5A). Similar to the previous report [54], half-life of luciferase activity in the proliferative cells was 2.30 ± 0.10 hr, which was comparable to that in the senescent cells, 2.54 ± 0.08 hr (Fig.5B). This result showed that the senescent TIG-3 cells have no effect of altered metabolisms/homeostasis on the half-life of luciferase activity, demonstrating that circadian clock-independent, senescence-induced unknown metabolic alterations would not affect the circadian oscillation patterns monitored by luciferase intensity.

Taking all results in Figs.3-5, we concluded that a longer circadian period and delayed peak-time elicited by Dex are due to the altered oscillator of the circadian clock in senescent cells.

Alteration to features of circadian clock in the senescent cells is a general hallmark of senescence of the circadian clock

To elucidate whether the findings shown in Fig.2 are Dex-synchronization specific or can be generalized, we further analyzed the trough-time difference and the period of Fsk-synchronized TIG-3 cells. Superimposed relative oscillation patterns of luciferase demonstrated that the oscillation patterns in the senescent cells seemed to be delayed compared to the proliferative cells (Fig.6A, B and S1), similar to the results synchronized by Dex (Fig.2A and B). Statistical analyses revealed that the 1^st and 2^nd trough-times of senescent cells were significantly delayed (Fig.6C). In addition, the 1^st period length, the duration between 1^st and 2^nd trough-times, was longer in the senescent cells (23.7 ± 0.3 hr vs 25.8 ± 0.3 hr; p-value, 1.50 x 10^-4; Fig.6D). We also analyzed the peak-time differences and the period using 2^nd and 3^rd peak-times (Fig.S6). Consistent with trough-time differences, differences of 2^nd and 3^rd peak-time between the proliferative and senescent cells were 4.7 ± 0.3 and 7.1 ± 1.0 hr, respectively (Fig.S6B). Again, the 2^nd period, which is the duration between 2^nd and 3^rd peak-times, in the senescent cells also demonstrated a longer period compared to the proliferative cells (23.5 ± 0.4 hr vs 25.6 ± 0.8 hr; p-value, 0.01; Fig.S6C). Finally, cosiner analyses revealed that the period length in the senescent cells is longer than that in the proliferative cells (23.79 ± 0.05 hr vs 26.91 ± 0.27 hr; p-value, 1.52 x 10^-8; Table 2), and the acrophase in the senescent cells is delayed compared to that in the proliferative cells (-227.13 ± 2.37 degree and -257.50 ± 3.12 degree; p-value, 1.99 x 10^-6; Table 2). Consistent with the result of Dex-induced entrainment, we again found that the damping ratios between the proliferative and senescent TIG-3 cells were comparable (Fig.6E).

Taking all results together, we propose that the alteration to features of circadian clock in the senescent cells is not a Dex-synchronization specific phenomenon, but rather the general hallmark of senescence of the oscillator of circadian clock.

Acute response and circadian expression profiles of endogenous clock genes are attenuated in the senescent cells

Finally, we addressed expression profiles of endogenous circadian clock genes in TIG-3 cells. We first investigated the acute response of clock genes by Dex treatment. Although PER1 and PER2 are known to possess glucocorticoid response elements (GRE) in their promoters [48,49], their acute responses to Dex treatment in the senescent cells were different from those in the proliferative cells. Acute response of PER1 was delayed, the peak-time was 3 hr after Dex treatment in the senescent cells, although the peak-time was approximately 2.5 hr in the proliferative cells (Fig. 7A). In contrast, acute response of PER2 was severely weakened in the senescent cells (Fig.7B). Distinct responses of PER1 and PER2 were also observed by Fsk treatment (Fig.S7). Interestingly, acute responses of other clock genes, which have not been reported to possess GRE on their promoters, were grouped into two categories; one was BMAL1 and CRY2, which showed similar expression levels and responses between the proliferative and the senescent cells (Fig.S8), the other was CRY1 and REV-ERBA, which showed similar responses with different expression levels (Fig.S8). We further monitored circadian expression profiles of the endogenous clock genes in the proliferative and senescent cells. Expression levels of PER2, CRY1 and REV-ERBA mRNAs in the senescent cells decreased to less than half compared to the proliferative cells throughout the time course, and amplitudes of these genes were attenuated (Fig.7C and Fig.S9). On the other hand, expression levels and oscillation patterns of BMAL1, PER1 and CRY2 were similar or moderately weakened in the senescent cells (Fig.7D and Fig.S9). In summary, we revealed that in the senescent cells the acute response of PER1 and PER2 are altered, which may correspond to the following delayed-phase, and PER2, CRY1 and REV-ERBA demonstrate attenuated expressions and oscillation, which may give rise to the extension of circadian period in the senescent cells.

Cell culture

TIG-3 cells, which were kindly provided by Drs. T. Takumi and T. Akagi, were grown in Dulbecco's Modified Eagle Medium (4.5 g/L glucose) (Nacalai Tesque, Kyoto, Japan) supplemented with 10% fetal bovine serum (biowest, Nuaillé, France) and 10 U/ml penicillin/10 µg/ml streptomycin (Nacalai Tesque, Kyoto, Japan) at 37 °C and 5% CO₂.

Cell growth assay

TIG-3 cells were cultured in triplicates in 10 cm dishes at seeding density = 5x10⁵ cells/10 cm dish. Upon reaching semi-confluence; every 3 days till passage 34 and 4 days or more from passage 35, the cells were trypsinized, counted and seeded again at 5x10⁵ cells/10cm dish. This process was repeated until the cells stopped proliferation and three independent assays were performed. To calculate the Population Doubling Level (PDL), the following formula was used: PDL=3.32log(N_harvest/N_seeding), where N_seeding is the seeding cell number and N_harvest is the cell number at harvest when cells reach semi-confluence. Cumulative PDL (cPDL) of a particular passage number is the accumulation of all the previous PDL values, including that of the passage being measured. In this study cells are defined as senescent when the PDL value of the cells is below 1.

Senescence associated β-galactosidase (SA-β-Gal) activity

Cells were fixed with 2% formaldehyde and 0.2% glutaraldehyde in H₂O for 5 min and incubated at 37 °C without CO₂ for 16 hr in X-Gal staining solution (40 mM sodium citrate (pH 6.0), 5 mM potassium ferrocyanide, 5 mM potassium ferricyanide, 1 mg/ml 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal), 150 mM sodium chloride, and 2 mM magnesium chloride). 200 cells were counted under a bright-field microscope for each sample.

RNA extraction and qPCR

Total RNA from the TIG-3 cells were extracted using Sepasol RNA-I Super G (Nacalai Tesque, Kyoto, Japan). Total RNA was reverse-transcribed by using SuperScriptII Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA) with random primers in accordance with the manufacturer’s instructions. Quantitative PCR was performed in the presence of KAPA SYBR FAST Universal 2X qPCR Master Mix (Nippon Genetics, Tokyo, Japan) on the Light Cycler 480 (Roche, Basel, Switzerland) under the following conditions: 95 °C for 3 min, and 40 cycles at 95 °C for 10 sec, 60 °C for 20 sec and 72 °C for 1 sec. The sequences of the forward and reverse primers are as follows. p16^INK4a FW; ACC AGA GGC AGT AAC CAT GC, p16^INK4a RV; GGA CCT TCG GTG ACT GAT GA. p21^CIP FW; AGC ATG ACA GAT TTC TAC CAC TCC, p21^CIP RV; GCA GAA GAT GTA GAG CGG GC. PER1 FW; GAG AGC AGC AAG AGC ACA AA, PER1 RV; AGC TCT CGA AGT GCT GTC AT. PER2 FW;TCC AGT GGA CAT GAG ACC AA, PER2 RV; CGC TAC TGC AGC CAC TTG TA. CRY1 FW; CAG CAG TGG AAG TTG CTC TC, CRY1 RV; CTA GGA CGT TTC CCA CCA CT. CRY2 FW; GGG AGG AGA GAC AGA AGC TC, CRY2 RV; GAG GGA GTT GGC GTT CAT TC. REV-ERBA FW; GAC ATG ACG ACC CTG GAC TC, REV-ERBA RV; GCT GCC ATT GGA GTT GTC AC. BMAL1 FW; CAT GCA ACG CAA TGT CCA G, BMAL1 RV; GTG TAT GGA TTG GTG GCA CCT. The primer set of 18S rRNA was described preciously [35].

Lentivirus production and infection to TIG-3 cells

HEK293T cells were seeded at 4x10⁶ cells in 10 cm dish. After 24 hr of seeding, when the cells reached 80-90% confluency, transfection was performed as follows: two 1.5 mL tubes were taken and 200 μL of Opti-MEM (WAKO, Osaka, Japan) was put into each tube. Into one tube, 9 μg of psPAX2, 6 μg of pMD2.G and 30 μL of 1 mg/ml polyethylenimine (PEI) (Polysciences, Warrington, PA, USA) were added. Into the other tube, 14 μg of pLV6-Bmal1-luc and 28 μL of PEI were added. Then the components of each tube were mixed by repeated inversion and then incubated at room temperature for 15 minutes. The contents of each tube were then added to the HEK293T cell dish. After 24 hr, the medium was removed from the cells, washed with PBS once and 5 mL of fresh medium was added. 24 hr later, the viral supernatant was collected and stored in 4°C, and 5 mL of fresh medium was added to the dish again. After 24 hr the viral supernatant was collected and pooled with the viral supernatant of the previous day. The total volume of viral supernatant was centrifuged at 167.7 x g at room temperature for 5 min to get rid of any cellular debris. The supernatant was further filtered through a 0.2 μm filter to remove any remaining debris of HEK293T cells. The filtered viral suspension was then centrifuged at 8,000 x g at 4°C for 4 hr to concentrate the lentivirus particles. After 4 hr, the supernatant was removed, and the lentivirus pellet was resuspended in 2 mL of medium to obtain the viral suspension.

For infection of the target TIG-3 cells, which had been seeded on the previous day at 1.5x10⁵ cells/35 mm dish, the culture medium was replaced with the viral suspension supplemented with protamine sulfate. 24 hr later the cells were washed with PBS once and 2 mL of fresh medium was added. The cells were then cultured for an additional 48 hr.

Synchronization of circadian clock in TIG-3 cells for Real-time luciferase monitoring assay and qPCR

48 hr after medium change of the infected TIG-3 cells, the medium was replaced with fresh 2 mL medium containing 0.1 μM dexamethasone (Dex; Nacalai Tesque, Japan) or 20 μM forskolin (Fsk; Nacalai Tesque, Japan), and incubated for 1 hr at 37°C 5% CO₂ for the entrainment of the circadian clock. Afterwards, the medium was removed, cells washed with PBS twice and medium containing 100 μM D-luciferin (Nacalai Tesque, Kyoto, Japan) was added. The cells were placed in the KronosDio luminometer (ATTO, Tokyo, Japan) and bioluminescence of the cells were measured for 5 days continuously. After 5 days, raw data were collected and in order to observe the circadian gene expression pattern more precisely, detrended data were also retrieved from the KronosDio software.

For the evaluation of endogenous circadian expression profiles, TIG-3 cells at confluent conditions were treated with 0.1 μM dexamethasone for 1 hr. Then the medium was replaced with normal culture medium and cells were cultured until indicated time points. RNA extraction, cDNA synthesis and qPCR were described above.

Cosinor analysis

To perform Cosinor analyses, data of three-day detrended samples from real-time luciferase monitoring assay were used. Cosinor software was kindly provided by Dr. R Refinetti [42].

Half-life of luciferase activity

Bmal1-luc-infected TIG-3 cells were treated with 50 μg/ml cycloheximide, then cells were set to the real-time luciferase monitoring system. Half-life of luciferase activity was evaluated at the time when decay rate became 50%.

Statistics

Values are reported as mean ± SEM. Statistical differences were determined by a Student's two-tailed t test. Statistical significance is displayed as * (p < 0.05), ** (p < 0.01), or *** (p < 0.001).