Circadian Oscillations Persist in Cervical and Esophageal Cancer Cells Displaying Decreased Expression of Tumor-Suppressing Circadian Clock Genes

This article is featured in Highlights of This Issue, p. 1255

The circadian clock aligns metabolic and physiologic processes, at cellular and systemic levels, to the external light and dark 24-hour cycle. At the molecular level, the circadian clock comprises a core set of rhythmically expressed genes and gene products, BMAL1, CLOCK, CRYPTOCHROME (CRY), PERIOD (PER), REV-ERB, and ROR, which drive rhythmic expression of clock-controlled genes to generate overt circadian rhythms (1). CLOCK and BMAL1 proteins heterodimerize and activate the expression of CRY and PER genes, which once translated repress CLOCK/BMAL1 activity to inhibit their own transcription (1). BMAL1 levels are further controlled by negative and positive regulatory transcription factors, REV-ERBα/β and RORα (1). All nucleated cells harbor a circadian clock mechanism and express circadian clock genes (2). The suprachiasmatic nucleus (SCN) is the master clock in the brain that receives light signals via the retino-hypothalamic tract and generates neural and hormonal signals accordingly (2). This is turn synchronizes peripheral clocks, which are those molecular clock networks expressed in all non-SCN cell types. Up to 40% of the genome is regulated in a circadian fashion, highlighting the crucial role of the circadian clock in cell biology (3). Its dysregulation is implicated in several disease states, including cancer.

Circadian disruption has been correlated with increased cancer risk and progression. This has been evident in shift workers and the International Agency for Research on Cancer (IARC) classified shiftwork that results in circadian disruption as a probable carcinogen (group 2A carcinogen; ref. 4). With recent epidemiologic data suggesting that more than 80% of the population may be living a lifestyle of chronic circadian disruption, this has widespread implications (5). In mouse studies, disruption of the light cycle accelerates tumorigenesis (6) and meal timing inhibits cancer growth by approximately 40%, as compared with mice fed ad libitum (7). Studies have shown that cancer cells in general commonly display a disrupted circadian rhythm or exhibit deregulated circadian clock gene expression (8), where for example loss of circadian clock gene expression has been correlated with poor prognosis in various cancer types (9–11). It has recently been shown that tumor hypoxia aggravates the extent of circadian dysregulation (12). Furthermore, a link between metabolic perturbations of cancer cells and circadian clock disruption has been described previously (13). These studies highlight how the circadian clock is entwined in important cellular processes implicated in cancer development.

While disruptions in the circadian clock have been described in certain cancers, there are no documented reports to date on the circadian machinery of cervical cancer or esophageal cancer cells, and with these cancer types being prevalent in the developing world they require further investigation. Cervical cancer was classified as the fourth most common cancer among women worldwide in 2018, and the second most common cancer among women in countries with a low human development index (HDI; GLOBOCAN 2018, IARC; ref. 14). Esophageal cancer was classified as the seventh most common cancer among men worldwide in 2018, and the sixth most common cancer among men and seventh among women in countries with a low HDI (14). Knowledge of circadian clock functioning in these cancer types could provide important information to benefit diagnosis and treatment. With many drug targets and drug-metabolizing enzymes expression in a circadian fashion, knowledge of the circadian rhythm exhibited by cancer cells can inform when to optimally administer chemotherapy, enhancing its efficacy and minimizing toxicity.

In this study, we show that expression of multiple circadian clock genes is downregulated in cervical and esophageal cancer cells. These cells, however, maintain functional circadian oscillations. Furthermore, we show that interfering with circadian clock expression/activity in these cancer types has an inhibitory effect on cancer biology. It has been reported that circadian clock genes are potential “druggable” targets, due to their dysregulation in cancer cells (5). Our study supports further investigation into the targeting of circadian clock machinery as an anticancer strategy.

Human primary (ARPE19) and immortalized (hTERT-RPE-1) retinal epithelial cell lines were cultured in DMEM:F12 (Gibco) supplemented with 10% FBS (Gibco), penicillin, and streptomycin (and 0.01 mg/mL Hygromycin B for hTERT-RPE-1 cells). Human fibroblasts (WI38 and SVWI38), cervical cancer cell lines (HeLa, CaSki and ME180), esophageal cancer cell lines (WHCO1, WHCO5, and KYSE30), and osteosarcoma cell line (U2OS), were maintained in DMEM (Gibco) supplemented with 10% FBS and penicillin and streptomycin. All cells except U2OS cells, which were kindly provided by Prof. J. Hapgood, University of Cape Town (Cape Town, South Africa), were obtained from the ATCC. Cells were maintained in a humidified incubator at 37°C and in 5% carbon dioxide. Cancer cell lines were authenticated by DNA profiling using the Cell ID System (Promega) and all experiments were conducted on cells within 10 passages after thawing. Mycoplasma testing was performed routinely, using Hoechst fluorescent stain, and all cells were confirmed to be Mycoplasma free.

Stable cell lines were generated using the Tol2 transposon system as described by Yagita and colleagues, 2010 (15). The PER2 promoter fused to destabilized luciferase (PER2-dLUC) was kindly provided by Prof. K. Yagita (Kyoto Prefectural University of Medicine, Kyoto, Japan). The BMAL1-dLUC construct was generated by excising the BMAL1 promoter-luciferase construct from the pGL4.27 plasmid, kindly provided by Dr. J. Hogenesch (University of Pennsylvania, Philadelphia, PA; ref. 16), and inserting it into the Tol2 plasmid. Cells were transfected using Genecellin (Celtic Diagnostics) and stable PER2-dLUC- or BMAL1-dLUC–expressing cells selected for in media containing between 100 and 500 μg/mL Zeocin (Invitrogen).

Cells were treated with RORα agonist, SR1078 (Merck), REV-ERBα agonist, SR9011 (Sigma), and the methylation inhibitor, 5-aza-2-deoxycytidine (Sigma), dissolved in DMSO. Drug stocks were prepared at concentrations of 5 mmol/L, 4.175 mmol/L, and 10 mmol/L, respectively, and stored at −20°C.

Freshly frozen patient tissue samples were obtained from patients at Groote Schuur and Tygerberg hospitals. This study was authorized by the University of Cape Town Research Ethics Committee (REC REF: 153/2004 and 040/2005) and written informed consents were obtained from all patients.

Overexpression plasmids were obtained as follows: myc-tagged mPer2 (pCS2+mt-6myc-Per2) was kindly provided by David Virshup (Duke-MED Medical School, Singapore; ref. 17), flag-tagged mClock (pCMV10/3Xflag-Clock) was purchased from Addgene and a kind gift from Joseph Takahashi (Addgene plasmid #47334; ref. 18), flag-, myc-, and his-tagged hCRY1 in pcDNA4 (pfmh-CRY1) was purchased from Addgene and a kind gift from Aziz Sancar (Addgene plasmid #25843), and Venus-tagged mBmal1 (pcDNA3.1-Venus-Bmal1) was kindly provided by Kyungjin Kim (University of Pennsylvania School of Medicine, Philadelphia, PA; ref. 19). Empty vector controls were generated by excising the circadian clock genes from their respective plasmids, and religation using the Quick Blunting and Quick Ligation kits (New England Biolabs).

Total RNA was extracted from patient tissue and cultured cell lines (2 hours post-dexamethasone synchronization) using Qiazol Reagent (Qiagen) according to the manufacturer's instructions. RNA was reverse transcribed using ImPromII Reverse Transcriptase (Promega). Quantitative PCR reactions were performed using the Kapa Fast SYBR Green qPCR reagent (Kapa Biosystems) and primers shown in Table 1. GAPDH, β-GLUCURONIDASE, and CYCLOPHILIN D were applied as the endogenous controls for normalization, and the 2-^ΔΔC_t was used to calculate the relative mRNA expression.

Protein was extracted from cells 2 hours after synchronization with dexamethasone, using RIPA lysis buffer and the Bicinchoninic acid (BCA) Assay Kit (Pierce) used to determine protein concentration. Protein was separated by SDS-PAGE and transferred onto nitrocellulose membranes (Millipore). Membranes were incubated with primary anti-CLOCK (1:2000; Pierce #PA1-520), anti-BMAL1 (1:2000; Abcam #ab93806), anti-CRY1 (1:1000; Santa Cruz Biotechnology #sc-33177), anti-PER2 (1:1000; Abcam #ab179813), anti-RORα (1:750; Abcam #ab70061), anti-REV-ERBα (1:500, Millipore #AB10130), or anti-MAPK14 (p38) (1:5000, Sigma #M0800) antibody overnight at 4°C, followed by species-matched secondary antibodies, and bands detected using Lumiglo reagent (Thermo Fisher Scientific). For PARP-1 cleavage analysis, protein was harvested from cells and cell floaters and Western blot analysis performed using anti-PARP-1 (1:500; Santa Cruz Biotechnology #sc-7150) antibody.

Bisulfite treatment of genomic DNA was carried out using the EpiTect Fast LyseAll Bisulfite Kit (Qiagen), according to the manufacturer's instructions. Bisulfite-modified DNA was amplified by PCR using primer sets designed to detect either methylated or unmethylated DNA, designed using Methprimer (Table 1). PCR was carried out using 200 nmol/L each primer, 200 μmol/L dNTPs, 1.5 mmol/L MgCl₂ 1.25 U GoTaq G2 Flexi DNA polymerase (Promega), and 100 ng bisulfite-converted DNA. Cycling conditions were: 95°C for 2 minutes, followed by 40 cycles of 95°C for 30 seconds, 52°C or 55°C (for unmethylated or methylated primer reactions, respectively), and 72°C for 30 seconds, followed by 72°C for 5 minutes. PCR products were visualized by 1.5% agarose gel electrophoresis. CpG methylase (SssI)-treated sodium bisulfite–modified genomic DNA was used as a positive control for methylation-specific primers. To determine the effect of 5-aza-2-deoxycytidine on promoter methylation, cells were treated with 5-aza-2-deoxycytidine every day for 3 days, and DNA harvested and bisulfite-converted. DNA amplification was performed by real-time PCR, using the Kapa Fast SYBR Green qPCR reagent (Kapa Biosystems) and 250 nmol/L each methylation-specific primer. PCR amplification was stopped before fluorescence generated from untreated cancer cell lines had reached saturation (the “plateau effect”); that is at 35 cycles for CLOCK and 28 cycles for RORα and CRY1.

Five-thousand cells per well were plated in 96-well plates and the following day either transfected with overexpression plasmids, using Genecellin (Celtic Diagnostics), or treated with RORα/REV-ERBα agonists. Proliferation was assessed 3 to 4 days after treatment or transfection, respectively, by addition of MTT in PBS at 5 mg/mL. After incubation at 37°C for 4 hours, MTT solvent was added and absorbance measured at 595 nm.

Cells were plated in 24-well plates and treated with RORα/REV-ERBα agonists for 72 hours. Cell floaters and adherent cells were collected by trypsinization, stained using 0.4 % Trypan blue stain, and dead (stained) and live (unstained) cells counted using a hemocytometer.

Cells were plated in 96-well plates and treated with RORα/REV-ERBα agonists for 72 hours. The Caspase-Glo 3/7 assay (Promega) was performed, according to the manufacturer's instructions. Luminescence was measured using the Veritas microplate luminometer (Promega) and normalized to OD595 readings of MTT experiments performed in parallel.

For imaging, 5000 PER2-dLUC- or BMAL1-dLUC–expressing cells were seeded in white 96-well plates with clear bottom. The next day, cells were synchronized with dexamethasone (Sigma) for 2 hours, following which media were replaced with Leibovitz L-15 (Thermo Fisher Scientific) containing 10 % FCS and 150 μg/mL d-Luciferin (Promega). Luciferase activity was measured using the IVIS Lumina II (PerkinElmer). Images were collected at intervals of 60 minutes, with 10-minute exposure duration, for up to 4 days. Images were acquired using Living Image 3.2 Acquisition software (PerkinElmer), and analyzed using Time Series Analysis 6.3 software (Expert Soft Tech). Phase, period, and amplitude were determined using detrended and smoothed bioluminescence data.

Statistical analyses were performed using Microsoft Excel, using Student t tests to assess differences in sample means. For analysis of continuous luminometer readings, Time Series Analysis 6.3 software (Expert Soft Tech) was used. Period was calculated using the Jenkin and Watts autoperiodogram test and cosinor curves generated using the population-mean cosinor method. Cosinor curves shown depict the mean of biological replicates.

A microarray study we previously conducted, comparing the gene expression profiles of tumor tissue and normal epithelial tissue from patients with cervical cancer and healthy controls, identified the circadian clock gene, PERIOD 2 (PER2), as one of the most significantly differentially expressed genes in cervical cancer and normal cervical tissue (20, 21). Its expression was among the ten most significantly downregulated genes in cancer compared with normal (*, P < 0.0005; Supplementary Fig. S1A). Downregulation of PER2 suggests that the circadian clock is altered in cervical cancer. Datamining of the Oncomine gene expression database, a platform aimed at facilitating discovery from genome-wide expression analyses (22), confirmed downregulation of PER2 in cervical cancer tissue, and revealed significant downregulation of other core circadian clock genes, CLOCK, BMAL1, PER1, CRY1, REV-ERBα, and RORα in high-grade squamous intraepithelial lesion (HSIL) and/or cervical cancer tissue specimens compared with normal cervical epithelium, using data derived from two independent studies (Fig. 1A; refs. 23, 24). Downregulation of circadian clock genes in HSIL tissues suggests that deregulation of circadian clock gene expression occurs as an early rather than late event in disease progression (Fig. 1A, ii). Similarly, circadian clock gene expression analysed using other valuable resources of gene expression data, The Cancer Genome Atlas (TCGA), and The Genotype Tissue Expression (GTEx) project, after removal of study-specific biases (25), revealed significant downregulation of several circadian clock genes in cervical tumor tissue compared with normal (Supplementary Fig. S2A). To show that there was not global downregulation of all genes using data derived from TCGA and GTEx studies, expression of KI67 and CDKN2A (p16) was evaluated, and both were found to be at significantly increased levels in cervical cancer tissue compared with normal, as expected (Supplementary Fig. S2B).

Because of the observed downregulation of multiple members of the circadian clock gene family in cervical cancer tissue, circadian clock gene expression in esophageal cancer patient tissue was next investigated by Oncomine datamining, to determine whether this effect was specific to cervical cancer or a more general phenotype of cancer cells. Similar to the results obtained for cervical cancer, multiple circadian clock genes were found to be downregulated in esophageal squamous cell carcinoma (OSCC) tumor tissue compared with matched normal epithelium in two independent datasets (26, 27), including CLOCK, PER1, PER2, PER3, CRY1, CRY2, REV-ERBα, and RORα genes (Fig. 1B). Circadian clock gene expression was similarly found to be downregulated in esophageal tumor tissue compared with normal esophageal epithelium, using data obtained from TCGA and GTEx project resources (Supplementary Fig. S2C). Again, expression of KI67 was found to be at increased levels in the tumor tissue compared with normal, as well as expression of a suggested biomarker for esophageal cancer, Karyopherin alpha 2 (KPNα2; ref. 28; Supplementary Fig. S2D).

Having access to esophageal tumor tissue and matched normal epithelial tissue from patients with esophageal squamous cell carcinoma, we next performed real-time RT-PCR analysis using RNA isolated from frozen tissue specimens, to independently validate Oncomine and TCGA data findings in a South African subset of patients. Real-time RT-PCR analysis revealed significant downregulation of CLOCK, PER1, PER2, PER3, CRY1, CRY2, REV-ERBα, and RORα in esophageal cancer tissue compared with normal (Fig. 1C). Interestingly, BMAL1 expression was unchanged in both Oncomine and real-time RT-PCR analyses. To address possible correlations between gene expression changes, pair-wise correlation analyses were performed using the RT-PCR data (for genes which showed significantly altered expression), where Pearson correlation coefficients were determined comparing the fold changes in cancer versus normal specimens. Strong positive correlations were identified between several circadian clock genes, where downregulation of one gene significantly correlated with downregulation of another (Supplementary Table S1).

These results show that collective downregulation of multiple circadian clock genes associates with cervical and esophageal tumor development.

Circadian clock gene expression levels were next measured in cells grown in culture. The WI38 and SVWI38 cell lines are a useful pair in which to compare gene expression levels, as the SVWI38 cell line is an SV40 large T antigen (T-Ag) transformed derivative of WI38 normal human fibroblast cells, hence any differences in gene expression observed can be attributed to the process of cellular transformation rather than heterogeneity between cell lines. As cells in long-term culture have been removed from the entraining influence of the SCN, the master clock in the brain, their circadian clock oscillators become out of phase with each other and require synchronization to become rhythmic. Thus, to more accurately compare circadian clock gene expression patterns between cell lines, cells were synchronized, by treatment with the glucocorticoid analogue dexamethasone, for 2 hours, which is known to be sufficient to synchronize the circadian clock of cultured cells, and circadian clock gene expression monitored immediately following synchronization. Interestingly, real-time RT-PCR analyses revealed significant downregulation of all circadian clock genes in the transformed compared with normal WI38 cells (Fig. 2A), apart from REV-ERBβ and RORβ, which displayed unchanged expression (Supplementary Fig. S3A). RORγ was difficult to detect due to its low levels of expression in tissues outside of lymphoid and skeletal muscle origin (Human Protein Atlas). Synchronization itself did not significantly alter circadian clock gene expression patterns from that observed in unsynchronized cells, besides PER1 expression, which was upregulated after synchronization (data not shown). This is consistent with the reported effect of dexamethasone in inducing PER1 expression (29). PER1 expression was thus not included in subsequent analyses. The downregulation of circadian clock genes in transformed cells suggests that transformation induces a global downregulation of the circadian clock machinery.

Circadian clock gene expression levels were next measured in a panel of normal and cancer cell lines by real-time RT-PCR, immediately after synchronization of cells for 2 hours with dexamethasone. ARPE-19 normal primary epithelial cells were used as the representative normal cell line, with immortalized hTERT-RPE-1 cells representative of an immortalized cell line, and HeLa, CaSki, and ME180, alongside WHCO1, WHCO5, and KYSE30, representative of cervical cancer cell lines and esophageal cancer cell lines, respectively. CLOCK, BMAL1, CRY1, and RORα showed significantly decreased expression in nearly all the cancer cell lines compared with normal (Fig. 2B). Expression of PER2 was decreased in two of the three cervical cancer cell lines but increased in esophageal cancer cell lines. Interestingly, the negative regulator REV-ERBα was expressed at increased levels in the cancer cell lines compared with normal, contrary to the positive regulator RORα (Fig. 2B). CRY2 and PER3 expression was largely unchanged in the cancer cell lines compared with normal, consistent with the cervical cancer patient data, as well as expression of REV-ERBβ and RORβ (with RORγ again difficult to detect; Supplementary Fig. S3B). Immortalized hTERT-RPE-1 cells similarly showed reduced expression of CLOCK, BMAL1, CRY1, and RORα, with REV-ERBα at increased levels. Overall, these results reveal deregulation of several circadian clock genes in cancer cell lines compared with normal and suggest that the process of cell immortalization itself might be sufficient to deregulate circadian clock gene mRNA expression.

At the protein level, protein was harvested from cells immediately following dexamethasone synchronization, and CLOCK, CRY1, REV-ERBα, and RORα were found at decreased levels in transformed SVI38 cells compared with normal, with CRY1 barely detected in the transformed cells (Fig. 2C). PER2, on the other hand, was present at increased levels in the transformed cells, and a dominant upper band was present, likely indicative of posttranslationally modified PER2. There is extensive posttranslational regulation of circadian clock genes that occurs in the circadian clock pathway, including ubiquitination, phosphorylation, glycosylation, acetylation, and sumoylation (1), and SV40 T-Ag–induced transformation appears to alter posttranslational control of PER2. In the cancer cell lines, CLOCK, CRY1, REV-ERBα, and RORα were at decreased levels in most cancer cell lines compared with normal, while BMAL1 levels were largely unchanged and PER2 found at elevated levels in most cancer cell lines compared with normal (Fig. 2D). These results largely support the mRNA findings, with certain exceptions, and highlight the fact that mRNA and protein levels do not always correlate, as demonstrated in previous studies (30, 31).

Together, these results reveal a significant deregulation of circadian clock gene family members, particularly CLOCK, CRY1, and RORα in transformed and cancer cell lines, consistent with findings using patient tumor tissue. These differences in expression between cell lines were evident not only immediately following synchronization but also at 12 and 24 hours postsynchronization (Supplementary Fig. S4). Their altered expression in cancer cells suggests these genes might be important players in mediating cancer biology.

Previous reports describe hypermethylation of circadian clock gene promoters in breast cancer and ovarian cancer cells (32, 33). To investigate whether methylation might be contributing to circadian clock gene downregulation in cervical and esophageal cancer cells, cells were treated with the methylation inhibitor, 5-aza-2-deoxycytidine (5-aza-2-dc), and circadian clock gene expression determined. Cells were treated daily for 3 days, whereafter gene expression was monitored by real-time RT-PCR. Results showed that CRY1 and RORα were consistently upregulated after treatment with 5-aza-2-dc, in all cancer cell lines, and CLOCK was upregulated in two of the three cancer cell lines (Fig. 3A–C). PER2 expression was increased in HeLa cells only (Fig. 3D), while BMAL1 and REV-ERBα expression was unchanged in response to 5-aza-2-dc treatment (data not shown). Treatment of normal ARPE-19 cells did not result in upregulation of circadian clock genes, but rather decreased expression of circadian clock genes, suggesting that the effect of methylation in downregulating circadian clock gene expression is cancer specific (Fig. 3A–D). At the protein level, increased expression of CLOCK, RORα, and PER2 (modified and unmodified forms) was observed in representative HeLa cells after methylation inhibition with 5-aza-2-dc (Fig. 3E). CRY1 protein levels were marginally increased, but difficult to examine due to their low levels of expression in cancer cells (Fig. 3E). In contrast, ARPE19 noncancer cells did not display increased levels of CLOCK, RORα, or CRY1 protein after 5-aza-2-dc treatment; in fact, levels were decreased after treatment, in line with the mRNA findings. PER2 protein, on the other hand, was present at increased levels after 5-aza-2-dc treatment of ARPE19 cells, as were levels of its modified form, suggesting that PER2 is regulated by methylation in both normal and cancer cells. The unchanged PER2 mRNA levels in ARPE19 cells upon 5-aza-2-dc treatment (Fig. 3D) suggest that this effect on PER2 protein levels is indirect, for example, via its stabilization in response to the inhibition of methylation.

Because cancer cells displayed increased expression of CLOCK, RORα, CRY1, and PER2 in response to methylation inhibition with 5-aza-2-dc, the methylation status of these genes was next investigated. CpG island prediction software (Methprimer) identified CpG islands in the 5′ DNA regulatory sequences of CLOCK, RORα, CRY1, and PER2 genes (predominantly in the region -1000 to +1000, with the transcription start site +1; Supplementary Fig. S5). Methylation-specific PCR was performed, where primers were designed to bind CpG-rich sites. Two sets of primers were designed for each gene to discriminate between methylated and unmethylated DNA. Cell line genomic DNA was bisulfite converted to allow for conversion of all unmethylated cytosines to uracil, and amplification carried out using both primer sets. CpG methylase (SssI)-treated bisulfite-converted HeLa and ME180 genomic DNA served as positive controls for the methylation-specific primers and PCR analysis revealed that methylation-specific primers positively amplified CpG methylase–treated DNA, as expected (Fig. 3F, first two panels).

For analysis of the methylation status of each cell line, methylation-specific PCR revealed that CLOCK, RORα, and CRY1 genes were all methylated to varying extents in their 5′ upstream regulatory regions in the CpG-rich regions examined (Fig. 3F). PER2 was not methylated in any of the cell lines tested (Fig. 3F). While the CRY1 gene promoter was fully methylated in all five cancer cell lines (no unmethylated DNA detected), CLOCK and RORα genes were found to be partially methylated, where DNA was amplified using both unmethylated and methylated primer sets. This could be due to the methylation of only one allele per gene.

To further investigate whether methylation was responsible for decreased circadian clock gene expression in the cancer cells, it was determined whether 5-aza-2-dc treatment could reduce methylation of CLOCK, RORα, and CRY1 gene promoters, thereby contributing to their reactivated expression in response to 5-aza-2-dc treatment. Cells were treated with 5-aza-2-dc for three days and DNA bisulfite–converted before real-time methylation-specific PCR was performed. Positive amplification of CLOCK, RORα, and CRY1 was clearly detected in untreated cancer cells using methylation-specific primers, which decreased after treatment with varying concentrations of 5-aza-2-dc. Methylation of CLOCK, RORα, and CRY1 gene promoters was also investigated in ARPE19 cells, but it was found that ARPE19 cells displayed substantially reduced methylation of CLOCK, RORα, and CRY1 gene promoters compared with cancer cells. There were, however, low levels of promoter methylation detected in these cells, which became apparent as the number of amplification cycles increased (Fig. 3G). However, unlike in the cancer cells, methylation of CLOCK, RORα, and CRY1 gene promoters was not reduced upon treatment of ARPE19 cells with 5-aza-2-dc.

Together, these results demonstrate that cancer cells lines exhibit low levels of CLOCK, CRY1, and RORα gene expression due, at least in part, to hypermethylation of CpG islands in their promoter regions. While the PER2 promoter was not found to be methylated, the increased expression of PER2 in response to methylation inhibition could be an indirect effect, and possibly due to the regulation of PER2 expression/stability by CLOCK, CRY1, and RORα gene family members.

As cancer patient tissue and cell lines displayed reduced circadian clock gene expression, it was next investigated whether there was a link between downregulated circadian clock gene expression and cancer biology. Circadian clock genes were overexpressed using circadian clock gene expression plasmids, and cell proliferation monitored using the MTT assay. HeLa, CaSki, ME180, and WHCO5 cancer cell lines displayed significantly reduced proliferation after CLOCK and PER2 overexpression, compared with cells transfected with corresponding empty vector controls (Fig. 4A and B). ARPE19 cell proliferation, on the other hand, was unaffected by CLOCK or PER2 overexpression. BMAL1 and CRY1 overexpression did not significantly affect cell proliferation in any of the cell lines (Fig. 4C and D). Western blot analysis confirmed successful overexpression of all circadian clock genes (Fig. 4E), and the more slowly migrating modified form of PER2 was again observed in the PER2-overexpressing cell lysates.

To explore the effect of enhancing REV-ERBα and RORα activity on cell proliferation, cells were treated with REV-ERBα and RORα agonists, SR9011 (34) and SR1078 (35), respectively. These agonists have been reported to specifically enhance REV-ERBα and RORα receptor function. Cancer cells treated with agonists for 72 hours showed significantly reduced proliferation, in a dose-dependent manner (with the effects of the RORα agonist SR1078 being more potent than the REV-ERBα agonist SR9011 at equivalent concentrations; Fig. 4F and G). Normal epithelial ARPE19 cells were significantly less affected by treatment, with SR9011 having no inhibitory effect on cell proliferation, and SR1078 only able to reduce cell proliferation at the highest concentration tested (lower concentrations of SR1078 in fact enhanced ARPE19 cell proliferation; Fig. 4F and G). Proliferation assays performed on WI38 and SVWI38 cells treated with agonists too revealed that SR9011 was more potent at killing transformed cells compared with normal counterparts, although SR1078 was effective at killing both cell lines (Supplementary Fig. S6). Trypan blue assays verified results in representative cancer cell lines, revealing a significantly reduced number of cancer cells upon SR9011 and SR1078 treatment, and a significantly increased number of dead cells, signifying that REV-ERBα and RORα activation results in the induction of cancer cell death (Fig. 4H and I).

To corroborate the induction of cell death upon REV-ERBβ and RORα agonist treatment, caspase 3/7 activity was measured in cancer cells after treatment with SR9011 and SR1078. Caspase-glo 3/7 activity was significantly induced in HeLa, ME180, and WHCO5 cells upon treatment with agonists, confirming the induction of apoptosis in response to REV-ERBβ and RORα activation (Fig. 4J). Increased cleavage of PARP-1 protein was also observed in cancer cells upon treatment, again confirming apoptosis as the mechanism of cell death (Fig. 4K).

Together, these results show that the overexpression/activation of specific circadian clock gene family members, PER2, CLOCK, RORα, and REV-ERBα, acts to suppress the proliferation and viability of cancer cells, resulting in cancer cell death via apoptosis, while having a lesser effect on noncancer cells.

The functional impact of circadian clock gene downregulation in cancer cells was further investigated by measuring the circadian rhythms of cervical cancer and esophageal cancer cells. Luminescence was measured in cells harboring a destabilized luciferase (dLUC) gene under the control of the PER2 or BMAL1 promoter. These stable PER2-dLUC or BMAL1-dLUC cells were generated using the Tol2 transposon system (15). Real-time monitoring of luciferase activity revealed fluctuations in PER2-dLUC activity in unsynchronized HeLa and WHCO5 cancer cells (Supplementary Fig. S7), which was significantly enhanced upon synchronization with dexamethasone, where robust rhythmic expression of PER2-dLUC was observed (Fig. 5A). Data were detrended and the best corresponding cosinor curve fit, whereafter the period could be calculated using the Jenkins and Watts autoperiodogram test. A period of 26.0 ± 0.83 hours and 25.9 ± 0.84 was determined for HeLa and WHCO5 cell lines, respectively, close to the intrinsic 24-hour timing of normal cells. BMAL1-dLUC activity showed an inverse pattern of oscillation to PER2-dLUC activity and was antiphase, as is expected of positive and negative circadian clock regulators (Fig. 5B). Plots of peak activity are shown in Fig. 5C.

Real-time monitoring of luciferase activity in additional cancer cell lines revealed distinct circadian oscillation of PER2-driven bioluminescence in all the cervical cancer (Fig. 5D, i) and esophageal cancer (Fig. 5D, ii) cell lines tested. Cosinor curves fitted to each dataset are shown, where the population-mean cosinor method was used (using TSA-Cosinor software). While amplitude varied between cell lines, this could be due to variability in total luminescence emitted from cells (due to different transfection efficiencies of cell lines and varying cell confluency), rather than heterogeneity between cell lines. Circadian oscillations are better compared by measuring the circadian phase and period. Interestingly, all cancer cells displayed a period of between 24 and 27 hours, and similar acrophase (peak in PER2-dLUC activity), except for the metastatic cervical cancer cell line, CaSki, which displayed a significantly delayed phase and increased period, compared with the nonmetastatic cervical cancer, HeLa and ME180, and esophageal cancer, WHCO5 and KYSE30 cells (Fig. 5E). This is consistent with previous reports describing a delayed circadian phase and increased circadian period in metastatic cancer cell lines (36). As a control, the circadian rhythm of human osteosarcoma cells, U2OS, was analyzed, as this is a widely used in vitro model to study properties of the mammalian circadian clock (37). Importantly, similar circadian rhythms were observed in the cervical cancer and esophageal cancer cells as the U2OS cells (Fig. 5E). Together, results reveal that despite dysregulated circadian clock gene expression, cervical cancer, and esophageal cancer cell lines maintain functional circadian oscillations.

This study identifies suppressed expression of circadian clock gene family members in cervical and esophageal tumor tissue and cell lines, compared with normal. Furthermore, it reveals that despite dysregulated gene expression, cervical cancer and esophageal cancer cells display overt circadian rhythms, based on their oscillating profiles of PER2 and BMAL1 promoter activity after synchronization. It also shows that perturbation of the circadian clock pathway, via overexpression of CLOCK and PER2 genes, or activation of RORα and REV-ERBα using small molecules, inhibits cancer cell biology (Fig. 6).

Circadian clock gene expression has been reported to be downregulated in breast cancer (9), head and neck squamous cell carcinoma (10), gastric cancer (11), and colorectal cancer (38), among others. There are multiple mechanisms by which downregulation may occur, for example, via the action of oncogenes or cancer/testis antigens. Relogio and colleagues (2014) showed how RAS perturbation can disrupt the circadian clock and revealed differential expression of core clock genes in normal and RAS-transformed cells (36), while Michael and colleagues (2015) showed how the cancer/testis antigen PASD1 represses BMAL1:CLOCK activity in cancer cells (39). Hypermethylation of the promoter regions of circadian clock genes has also been described, acting to downregulate circadian clock gene expression in breast cancer tissue (32) and ovarian cancer cells (33). Our study shows decreased expression of core circadian clock genes in cervical and esophageal cancer cells, and reveals that methylation plays a direct role in mediating downregulated CLOCK, CRY1, and RORα gene expression, and indirectly affects PER2 expression levels. The role of other factors such as oncogenes and cancer/testis antigens in further controlling circadian clock gene expression remains to be determined.

It was noted in our study that circadian clock gene mRNA and protein profiles did not always correlate. Shu and colleagues (2004) describe how circadian clock gene oscillations are controlled by four-step expression, where transcription, translation, degradation of mRNA, and degradation of protein are all vital in controlling gene expression (40). It is possible that differing protein/mRNA half-life in different cell lines may lead to the lack of correlation between mRNA and protein. Furthermore, a modified, more slowly migrating form of PER2 protein was observed in transformed SVWI38 cells (Fig. 2C), cells treated with 5-aza-2-dc (Fig. 3E), and PER2-overexpressing cells (Fig. 4E). It is reported that the size of PER2 fluctuates after synchronization of WI38 cells (it increases in size up to approximately 200 kDa: the size of modified PER2 in our study) and this is due to phosphorylation of PER2 by CKI (Casein Kinase I) proteins (41). It is also reported that CKI proteins can phosphorylate overexpressed PER1 and PER2 in vitro and induce an increase in the apparent molecular size of PER proteins (42). Future work is required to investigate why increased PER2 phosphorylation might be occurring under the various cellular conditions observed.

Chang and Lai (2019) recently showed that circadian clock genes that confer tumor-suppressing effects in one cancer type can play an opposing role and exhibit tumor-promoting effects in another cancer type (12). We show that in cervical and esophageal tumor cells PER2, CLOCK, RORα, and REV-ERBα display tumor-inhibiting properties, as their overexpression/activation (in cells displaying low levels of expression) results in reduced cell proliferation. Furthermore, we show that while cancer cells undergo cell death upon activation of RORα and REV-ERBα, normal cells respond differently and are less sensitive to SR9011, where concentrations that cause cancer cell death do not inhibit ARPE19 or WI38 cell proliferation. In addition, APRE19 proliferation is only inhibited at SR1078 concentrations of >30 μmol/L, although WI38 cells are sensitive to lower SR1078 concentrations. These results suggest that PER2, CLOCK, RORα, and REV-ERBα circadian clock genes might have potential as druggable targets, whereby their activation could have therapeutic potential. REV-ERBα agonists have been suggested to be potential treatment options for different types of cancer (43), where REV-ERBα activation appears to be nondamaging to healthy tissue, and can act to rid the tumor of not only actively proliferating cells, but also oncogene-induced senescent cells (which mediate chemotherapy resistance and relapse; ref. 43). It has been proposed that REV-ERBα activation suppresses CYCLIN A expression in breast cancer cells, contributing to cell-cycle arrest (44). The activation of RORα has been shown to induce p53 expression and activate cell death in HepG2 cells (35). Our results support the use of REV-ERBα and RORα agonists, SR9011 and SR1078, against cervical and esophageal cancers. Further work investigating the use of these agonists in treating these types of cancer is required. Particularly, further studies are needed to address whether the overexpression/activation of circadian clock components leads to the observed antitumoral effects via interfering with the circadian rhythm of cancer cells, or via alternate pathways.

Our study also examined the circadian oscillatory profiles of cervical and esophageal cancer cells. Literature suggests that the circadian rhythm is often suppressed in cancer cells. Breast cancer cells have been shown to display arrhythmic patterns of circadian clock gene expression, and while serum shock can induce oscillation of some circadian clock genes, the amplitude is greatly reduced compared with normal breast epithelial cells (45, 46). More recently, however, Lellupitiyage and colleagues (2019) showed that low malignancy MCF7 cells do display circadian oscillations of PER2 and BMAL1, while high-grade MDA-MB-231 cells do not (47). In tumor tissue in vivo absent or very weak rhythmic profiles of circadian clock genes are observed (7, 48). Interestingly, Relogio and colleagues (2014) describe how colon cancer cell lines show a rich variety of circadian phenotypes, where some show strong and others weak to no-oscillation phenotypes (classifying a strong oscillator as a cell line with a clear circadian period and amplitude variation of at least 20%; ref. 36). These authors identified a list of genes able to discriminate between weak and strong oscillator cell lines, and the roles of these genes in both circadian clock and oncogenic pathways suggests these pathways are strongly connected. Our results show that the cervical and esophageal cancer cells examined display strong oscillating phenotypes, despite deregulated circadian clock gene expression. There was little difference in the period and acrophase observed in the different cell lines, besides CaSki metastatic cancer cells having a significantly increased period and delayed phase. Together, these findings reinforce that cancer cells can maintain rhythmic circadian profiles and suggests that these rhythmic profiles might be contributing to cancer biology.

It has been shown that the circadian rhythm of clock genes becomes markedly impaired in senescent or “aging” cells, and this can be reversed by telomerase reconstitution (49). As cancer cells display increased telomerase activity, this could, in part, explain the overt circadian rhythms observed in the cancer cell lines used in our study. It has also been shown that telomerase mRNA expression and activity exhibits endogenous circadian rhythmicity and is under the control of the CLOCK-BMAL1 heterodimer (50), revealing a feedback mechanism between telomerase and the circadian clock.

In conclusion, the disruption of circadian clock components in cancer, or “circadian reprogramming” can be a critical player in tumorigenesis, while maintenance of circadian rhythms might also act to uphold cancer-related processes. Chang and Lai (2019) demonstrate that circadian reprogramming of tumor genomes plays an important role in influencing disease progression and patient outcomes (12). Ongoing efforts at investigating the circadian clock in cancer development are needed.

No potential conflicts of interest were disclosed.

P.J. van der Watt: Conceptualization, supervision, funding acquisition, investigation, methodology, writing-original draft, project administration. L.C. Roden: Methodology, writing-review and editing. K.T. Davis: Investigation. M.I. Parker: Resources. V.D. Leaner: Resources, writing-review and editing.

This work is supported through funding from the National Research Foundation (NRF) of South Africa (Research Career Advancement (RCA) award and grant number 120428).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.