Daily rhythms are retained both in spontaneously developed sarcomas and in xenografts grown in immunocompromised SCID mice

Abstract

The circadian clock generates and regulates many daily physiological, metabolic and behavioral rhythms as well as acute responses to various types of stresses including those induced by anticancer treatment. It has been proposed that modulatory function of the clock may be used for improving the therapeutic efficacy of established anti-cancer treatments. In order to rationally exploit this mechanism, more information is needed to fully characterize the functional status of the molecular clock in tumors of different cellular origin; however, the data describing tumor clocks are still inconsistent. Here we tested the status of clock in two models of tumors derived from connective tissue: sarcomas spontaneously developed in p53-deficient mice and human fibrosarcoma cells grown as xenografts in immunocompromised severe combined immunodeficient (SCID) mice. We show that both types of tumors retain a functional clock, which is synchronized in phase with normal tissues. We also show that spontaneously developed tumors are not only oscillating in the context of an organism where they receive hormonal and metabolic signals but continue oscillating ex vivo in tissue explants demonstrating that tumors have functional clocks capable of timing all their functions. We also provide evidence that similar to liver, tumors can be synchronized by food availability independent of the central pacemaker in the suprachiasmatic nuclei (SCN). These data provide the basis for the design of anticancer therapies that take into account the circadian metabolic and physiological patterns of both the tumor and normal tissues.

INTRODUCTION

The circadian clock generates and regulates many daily physiological and behavioral rhythms with a period of 24 h at the cell, tissue and organism levels. Processes such as sleep, wakefulness, immune activity, hormone levels, metabolism, tumor growth, apoptosis and cellular proliferation display circadian variations (Takahashi et al., 2008). In mammals, the central pacemaker is located at the suprachiasmatic nuclei (SCN), which is synchronized to the 24 h environmental light–dark cycle mainly via the light input received by photoreceptors of the retina that is directly transmitted via the retinohy-pothalamic tract (RHT) to the SCN. Moreover, clocks are known to be present in virtually all individual cells in the organism (Balsalobre et al., 1998). Under normal conditions multiple peripheral clocks are synchronized by the SCN via neural and hormonal cues and rhythms such as feeding-fasting behavior and rest activity cycles are in synchrony (Dibner et al., 2010). However, peripheral clocks can be uncoupled from the central pacemaker in response to other synchronizing signals (Mohawk et al., 2012). Several studies have shown that food availability is an SCN-independent synchronizing cue for the liver and, as a consequence, restricting food administration to the rest period of the circadian cycle alters the phase relationship between liver and SCN (Damiola, 2000; Hara et al., 2001; Stokkan et al., 2001).

At the molecular level, the circadian clock consists of tightly controlled transcriptional activation/repression of positive and negative components of the feedback loops. The positive component comprises the heterodimer formed by CLOCK (or its closest paralog NPAS2) and BMAL1 proteins and this heterodimer activates the transcription of genes involved in the negative component of the loop, Period (Per1, Per2) and Cryptochrome (Cry1, Cry2). After transcription of the genes, the PER and CRY proteins dimerize and are translocated back to the nucleus where they repress the activity of the positive components, creating a negative feedback loop. In addition, there is another loop in which CLOCK/BMAL1 activates the transcription of REV-ERBα and RORα. REV-ERBα and RORα proteins can bind to the promoter of Bmal1 and respectively activate or repress its transcription. One full cycle of oscillation lasts for about 24 h which is due to the delays between transcription, translation and subsequent transcriptional repression (Partch et al., 2013). The CLOCK/BMAL1 complex regulates the expression of multiple clock-controlled genes either directly or indirectly through other transcription factors.

Among the clock-controlled genes are the ones regulating cell cycle, apoptosis, DNA damage and genotoxic stress response including the one induced by anticancer treatment (Antoch & Kondratov, 2013). Previous work performed in animal models demonstrated that side effects of anticancer treatment can be minimized by manipulating the schedule of drug administration and adjusting it to times of the day associated with lower toxicity, a concept implicated in clinical practice as chronotherapy (Levi & Schibler, 2007; Lévi et al., 2010). Initial observations of successful use of chronotherapeutic approach were further extended by demonstrating that drug-induced toxicity in tumor-free mice correlate with the functional status of the major circadian regulator, CLOCK/BMAL1 transcriptional complex (Gorbacheva et al., 2005). Furthermore, drug-induced toxicity in vivo can be reduced by treatment with pharmacological activators of CLOCK/BMAL1 functional activity as it has been shown for organic selenium compounds (Hu et al., 2011). Together, these data provide a proof-of-principle for potential use of clock-targeting pharmaceuticals for reducing damaging effects to normal tissues. However, in order to fully explore therapeutic potential of the modulatory function of the circadian clock in cancer therapy, it is important to characterize its functional status in tumors and to understand whether sensitivity of tumors can be modulated through the circadian mechanism similar to normal tissues. Although this information is critical for future translational applications directed for the improvement of therapeutic indexes of existing regimens, the data on tumor clocks remain very limited and inconsistent. Here, we tested the functional status of clock in two models of tumors derived from connective tissue: sarcomas spontaneously developed in p53-deficient mice and human fibrosarcoma cells grown as xenografts in immunocom-promised severe combined immunodeficient (SCID) mice. We show that both types of tumors retain a functional clock, which under normal conditions (12:12 light: dark cycle and food available ad libitum) is synchronized in phase with other peripheral tissues. Moreover, similar to liver, circadian gene expression in xenografts can be uncoupled from the SCN-derived synchronization signal under restricted feeding conditions. Together, these data suggest that development of clock-targeting therapeutic strategies should take into account circadian mechanism in tumors and that tumor clocks are amenable for manipulation. This knowledge in critical for refining future approaches for improving the therapeutic index of cancer treatment.

MATERIALS AND METHODS

Animals

p^53−/− mice on a C57B1/6J background were purchased from Jackson Laboratories (Bar Harbor, ME). PER2::LUC mice were obtained from Dr Takahashi. To generate Per2-Luc reporter line on p⁵³⁻deficient background, PER2::LUC female mice were crossed to p^53−/− male mice to obtain heterozygous F1 mice, which were then inter-crossed to obtain homozygous F2 mice. Animals were genotyped for the presence of PER2::LUC fusion and p⁵³-deficiency according to standard protocols (http://jaxmice.jax.org/protocolsdb). After weaning, PER2::LUC/p^53−/− mice were group-housed at 12 h light:12 h dark cycle (LD12:12) with food and water ad libitum in a sound attenuated room, at a temperature of 23 ± 1 °C. Animals were weekly observed until tumor appearance was visually detected or dramatic loss of weight, indicative of tumor appearance, happened. SCID mice were obtained from RPCI colony (C.B-Igh-1bIcrTac-Prkdcscid/Ros, Buffalo, NY). All procedures involving animals were approved by the Institutional Animal Care and Use Committee of Roswell Park Cancer Institute and are in accordance with the international ethics standards (Portaluppi et al., 2010).

Wheel-running behavior recording

PER2::LUC/p^53−/− were individually housed in cages equipped with a running wheel. Activity data were collected as previously described (Antoch et al., 1997; Vitaterna et al., 1994). A group of mice were synchronized to 12 h light:12 h dark cycle (LD12:12). Another group was synchronized to LD12:12 for 7 days and then were released into constant darkness (DD). Actograms were plotted using ClockLab software (Actimetrics Inc, Evanston, IL).

Cell culture and real-time luciferase recording

HT1080 cells were purchased from ATCC and were maintained in DMEM supplemented with 10% fetal bovine serum. Lentiviruses expressing Bmal1-Luc or Per2-Luc (generously provided by Dr S. Kay) were generated using Invitrogen ViraPower™ HiPerform™ Promoterless Gateway^® Expression System (Grand Island, NY) according to manufacturer’s protocol as described by Liu et al. (2008). HT1080 cells were plated on 24-well plates at 50% confluency. Viral stocks were added at 1:10 dilution for 3 days after which media was replaced with fresh media containing 3 μg/ml Blasticidin. After 2 weeks of selection, individual clones were picked up and tested for luciferase expression. Clones with the highest luciferase signal per μg of protein were used for further analysis.

For in vitro luminescence imaging cells were plated on 35 mm dish at 4×10⁵ cells/plate in DMEM and allowed adhering overnight. Next day cells were treated with 100 nM dexamethasone for 2 h, then media was replaced with recording media (DMEM without phenol red supplemented with 0.35 g/L Sodium Bicarbonate, 5% FBS, 10 mM HEPES, pH 7.2, 3.5 g/L D-Glucose, 100 U/mL penicillin-100 μg/mL streptomycin, and 0.1 mM Beetle Luciferin (Promega, Madison, WI) as described by Yamazaki & Takahashi (2005). Bioluminescence was recorded continuously; data were analyzed by Actimetrics software (Actimetrics, Wilmette, IL) and presented as photon counts per second. Baseline correction was calculated using a 24-hour average.

For the real-time monitoring of circadian profile of Per2-driven luciferase expression in explant cultures of normal tissues and spontaneously developed sarcomas, PER2::Luc mice were sacrificed one hour before lights-off and tissues were prepared for several days of bioluminescence real-time measurement as previously described by Yoo et al. (2004) and Yamazaki & Takahashi (2005). Bioluminescence was measured for several cycles using a LumiCycle (Actimetrics, Wilmette, IL). LumiCycle software (Actimetrics Inc., Wilmette, IL) was used to analyze the data. The circadian period was calculated performing a periodogram analysis to three consecutive cycles. The peak phase was determined from the first peak of Per2-luc expression.

Tumor cells inoculation and in vivo bioluminescence imaging

Bmal1-Luc and Per2-Luc expressing HT1080 cells were inoculated subcutaneously into left and right flanks, respectively, of SCID) male mice at 5×10⁶ cells/site. After the tumors reached 5 mm bioluminescence, imaging was performed using IVIS 50 imaging system Caliper LifeSciences (Xenogen Corp., Alameda, CA) at two time points, ZT2 and ZT14. Mice were injected i.p. with firefly D-Luciferin (150 mg/kg, Caliper Life Sciences, Alameda, CA), anesthetized with isoflurane and imaged 15 min later using 10 s integration time and medium binning. Data were quantified as the sum of photon flux within the region of interest using Living Image software (Xenogen Corp., Alameda, CA). The same procedure was used to measure daily profiles of Per2-driven luciferase in normal tissues and tumors of PER2::LUC-p^53−/− mice with the exception that each individual mouse was profiled every 4 h.

Isolation of total RNA and real-time PCR

Total RNA from liver and inoculated tumors was isolated using Trizol^® Reagent (Life Technologies, Grand Island, NY) and purified using Qiagen RNeasy kit (Valencia, CA) following manufacturer’s protocols. Quantity and quality of the RNA was assessed with the Nanodrop^® ND-1000 UV-Vis Spectrophotometer (Nanodrop Technologies, Wilmington, DE). RT-qPCR analysis was performed with total RNA using the ABI Prism 7700 and TaqMan EZ RT-PCR core reagent kit (Applied Biosystems, Grand Island, NY) with pre-made mouse-specific TaqMan gene Expression Assays for Bmal1 (Mm00500226_m1), Per2 (Mm00478113_m1) and 18S rRNA (Mm03928990_g1), which was used as the reference gene. Relative gene expression was calculated using ΔΔCt method as previously described (Panda et al., 2002).

Tumor histology

Tissues were fixed in 10% neutral formalin for 24 h, and then transferred to 70% ethanol. Samples were embedded in paraffin, sectioned and stained with hematoxylin and eosin. Histopathological examination was performed using Zeiss AxioImager A1 with Axiocam MRc digital camera (Thornwood, NY). In determining the diagnosis the guidelines of Bethesda classification was used (Morse et al., 2002). Pictures were taken using Zeiss AxioImager A1 with Axiocam MRc digital camera.

Statistical analyses

Significance of the rhythmicity of the in vivo mPer2-driven luciferase expression profiles was determined with the CircWave v1.4 software using a linear harmonic regression fit (http://webpage2.woelmuis.nl/downloads.htm; R.A. Hut, University of Groningen, NL; Comas et al., 2008; Lincoln et al., 2006). For other statistical analysis Statistix 7.0 (Analytical Software, Tallahassee, FL) was used. Independent t tests were used for comparison of two groups of data. Values of p<50.05 were considered significant.

RESULTS

Spontaneously developed sarcomas in p^53−/− mice show rhythmic circadian profile of Per2 gene expression both in vivo and ex vivo

To evaluate the functionality of the molecular clock in tumors we used Per2-Luciferase reporter mice (PER2::LUC) (Yoo et al., 2004) crossed to tumor-prone p^53−/− mice, which develop sarcomas and lymphomas early in life (Donehower et al., 1992). The presence of circadian-driven luciferase in this model allows for simultaneous accessing of daily variations in clock function both in normal tissues and in spontaneously developed tumors. Importantly, the locomotor activity assay shows that tumor-bearing mice retain their ability to synchronize to the light–dark cycle and remain rhythmic under constant darkness conditions indicating that presence of tumors does not affect the function of the central clock in the SCN (Supplementary Figure 1A and B). To access the functional status of the clock in normal tissues and tumors we monitored daily temporal profiles of Per2-driven luciferase in livers and salivary glands (normal tissues that are amenable for in vivo imaging) and spontaneously developed visible solid tumors using the Xenogen IVIS Imaging System (Perkin Elmer, Walham, MA). Imaging of each individual animal was performed at six different time points every four hours throughout a day. As shown in Figure 1(A) and (B), young tumor-free PER2::LUC-p^53−/− mice display significant mPer2-luciferase circadian profile with a peak of expression in the middle of the night.

FIGURE 1.

Rhythmic Per2-driven luciferase circadian profiles measured in vivo. (A) Representative example of series of images of Per2-driven luciferase of a non-tumor bearing mouse imaged throughout a day. (B) Quantification of the mPer2-luciferase reporter circadian activity profile measured in vivo in non-tumor bearing mice. Statistical significance of the circadian oscillations obtained was determined with the CircWave v1.4 software. Salivary glands and liver display significant Per2-luciferase circadian profile with a peak of expression in the middle of the night (salivary glands p<0.001; liver p<0.001). For both tissues, two sines (fundamental wave and the first harmonic) contributed significantly to the explained variance. Values represent mean (n = 8) ± SEM. (C) Representative example of series of images of Per2-driven luciferase of a sarcoma-bearing mouse imaged throughout a day. (D) Circadian profiles measured in tumor-bearing mice in normal liver, normal salivary glands and tumor. The three tissues analyzed display a significant Per2-luciferase circadian profile with a peak of expression at midnight (salivary glands p<0.001; liver p<0.001; tumor p<0.05). In the circadian curves obtained from tumor-bearing animals, only the fundamental wave significantly contributed to the explained variance. Values represent mean (n = 7) ± SEM.

This profile was retained in normal tissues of tumor-bearing mice. Importantly, Per2-luciferase expression was rhythmic in solid tumor with the phase of expression similar to normal tissues (Figure 1C and D) but with reduced amplitude (2 to 4-fold). Histological analysis performed after completion of the experiment confirmed that all solid tumors that were profiled represent spontaneously developed sarcomas (Supplementary Figure 2). Thus, spontaneously developed sarcomas retain rhythmic expression of circadian genes, which is in phase with the oscillations found in normal tissues.

It has been previously demonstrated that individual peripheral tissues contain circadian oscillators capable of persistent rhythmicity for up to 20 days independent of SCN input (Yoo et al., 2004). In order to test whether this characteristic is retained in tumors, we monitored Per2-luciferase expression ex vivo in explant cultures of normal tissues (liver, spleen and lung) and tumors using real-time luciferase recording. After entraining animals to a LD12:12, peripheral tissues (liver, lung, spleen) and tumors were dissected one hour before lights-off and placed in cultures containing media supplemented with luciferin, and luminescence was measured for at least four cycles (Figure 2). Similar to what has been previously reported (Yoo et al., 2004), explanted liver, lung and spleen display prominent circadian oscillation in Per2 gene expression with a peak in the dark phase of the cycle. Interestingly, both the phase and the amplitude of oscillation in normal tissues were identical between tumor-free and tumor-bearing animals. Consistent with our in vivo data, explanted sarcomas show rhythmic expression of Per2 gene with a peak phase similar to liver, lung and spleen but with reduced amplitude. Thus, spontaneously developed sarcomas are not only oscillating in the context of an organism but also in isolation suggesting that these tumors retain a functional clock.

FIGURE 2.

Circadian profiles of Per2-driven luciferase from explanted tissues monitored ex-vivo. (A) Representative examples of rhythmic mPer2-driven luciferase records measured over several cycles from explant cultures of normal tissues (liver, lung, spleen) in non-tumor bearing and tumor-bearing mice as well as in sarcoma tumors. (B) Quantification of period of the Per2-driven luciferase records in the different tissues. In non-tumor bearing mice, the period of the oscillation of tissues measured over the first three cycles is close to 24 h. Tumor-bearing mice display a period in normal tissues not significantly different from those obtained in non-tumor bearing animals. (C) Quantification of peak phase of mPer2-driven luciferase records in different tissues. Explanted pieces of tissue of liver, spleen and lung from non-tumor and tumor-bearing mice peak in the dark phase. Sarcoma peak phase occurs, similar to the peripheral tissues measured in the dark phase. Values represent mean (n = 11) ± SEM.

Human fibrosarcoma HT1080 cells retain functional clock, which could be synchronized both in vitro and in vivo

One of the concerns in the interpretation of the spontaneous tumors data is that tissue explants represent a mixture of cells and the oscillation may originate from normal stromal cells rather than from tumor cells. Also, the conclusions obtained in the first set of experiments are solely based on the PER2-LUC reporter construct and the Per2 gene expression is known to be influenced by temperature cycles of the mouse itself (Buhr et al., 2010). In addition, the procedure of preparation of the explants itself (e.g. slicing tissue, placing explants into cold Hank’s balance salt solution or culture media) can reset the clock and phase relationship between different tissues cannot be accurately measured. To address these issues, we used another model – human fibrosarcoma HT1080 cells stably expressing luciferase under control of either Bmal1 or Per2 promoters. Real-time bioluminescence recording showed that both Bmal1-Luc and Per2-Luc exhibit cell autonomous oscillations after synchronization by dexamethasone (Figure 3). As expected, the oscillations are 180-degree out of phase. Since HT1080 cells can be grown as xenografts in immunocompromised mice, they represent an ideal model for testing the potential ability of their clock to oscillate in the context of an organism and to evaluate the phase of expression of Bmal1 and Per2 promoters in implanted tumors with respect to normal tissues. To explore this, each individual SCID mouse was subcutaneously injected with Bmal1-Luc expressing HT1080 cells in the left flank and Per2-Luc expressing HT1080 cells in the right flank. After tumor sizes reached 0.5–1 cm, in vivo imaging was performed at ZT2 and ZT14. These time points were chosen as known peak times for the expression of Bmal1 and Per2, respectively, in peripheral tissues (Preitner et al., 2002). Figure 4 shows representative images of two individual animals performed at ZT2 and ZT14. As it can be observed, both genes display prominent oscillation in their expression with the corresponding peaks and troughs that are characteristic of normal tissues (i.e. liver, Figure 1).

FIGURE 3.

HT1080 cells are synchronized in circadian phase by dexamethasone treatment. Real-time luciferase recording of HT1080 cells stably expressing Bmal1-Luc (A) or Per2-Luc (B) promoters. Cells were plated on 35 mm dishes at ~50% confluency and allowed to adhere overnight. After a 2 h treatment with 100 nM dexamethasone media was replaced with luciferase recording media and placed in Lumicycle. HT1080 cells demonstrate at least two cycles of circadian oscillation of Bmal1- and Per2-driven luciferase, which are 180-degree out of phase.

FIGURE 4.

HT1080 xenografts are synchronized in phase with normal peripheral tissues. Representative in vivo images of mice performed at ZT2 and ZT14. Each mouse bears HT1080-Bmal1-Luc xenograft on the right flank and HT1080-Per2-Luc – on the left flank. Both Bmal1 and Per2 display prominent oscillation in their expression with the corresponding peaks and troughs that are characteristic of normal tissues, with Bmal1 peaking at ZT2 and Per2 peaking at ZT14 within an individual animal.

HT1080 transplanted tumors can be synchronized by food availability independent from the SCN

It has been long shown that the glucocorticoid hormone analog dexamethasone induces circadian gene expression in fibroblasts in vitro and phase shifts in peripheral tissues, including liver, in vivo but not in the SCN (Balsalobre et al., 2000b). These results suggested that glucocorticoid hormones, similar to restricted feeding, could uncouple peripheral oscillators from the master clock in the SCN. Since HT1080 reporter cells can be synchronized by dexamethasone in culture (Figure 3), we hypothesize that when transplanted to SCID mice, HT1080 cells may respond to synchronization by food availability similar to the liver. To test this prediction, Bmal1-Luc and Per2-Luc expressing HT1080 cells were inoculated following the same protocol as the previous experiment. In the first (control) group, food was available ad libitum whereas the second group was maintained on restricted feeding schedule (food available at lights-on time only) starting from 2 weeks prior to tumor inoculation. Differences in feeding schedule had no effect on the rate of tumor growth and both groups of animals were imaged at the same time at ZT2 and ZT14. Consistent with our previous experiment (Figure 4), mice on ad libitum food schedule showed prominent oscillation of both circadian promoters that are out of phase with each other with Per2-driven signal increasing and Bmal1-driven luciferase signal decreasing in the same animal at ZT14 compared to ZT2 (Figure 5A). In contrast, animals that were maintained on restricted feeding schedule demonstrated a shift in the phase of both Bmal1 and Per2 genes expression (Figure 5B). More detailed quantitative analysis performed using RT-PCR on tumor tissue extracts collected at ZT2 and ZT14 at the completion of the experiments confirmed the imaging data. Thus, it showed that normal condition (food available ad libitum) tumor xenografts display significant oscillation in Per2 and Bmal1 genes expression, which is reduced in the amplitude compared to normal tissues (liver) but retain the same phase of oscillation (Figure 5C). Restricted feeding reverses the phases of Bmal1 and Per2 genes expression both in liver and in tumor xenografts, such that Per2-driven luciferase expression is decreased whereas Bmal1-driven signal is increased at ZT14 compared to ZT2. These data clearly demonstrate the ability of tumor clock to be manipulated by external signal (feeding schedule).

FIGURE 5.

HT1080 xenografts are synchronized in phase by food availability independent from the SCN. Representative in vivo images of mice fed either ad libitum (A) or during lights-on time only (restricted feeding schedule, (B) performed at ZT2 and ZT14. Each mouse bears HT1080-Bmal1-Luc xenograft on the right flank and HT1080-Per2-Luc – on the left flank. Bmal1 and Per2 expression in HT1080 cells are synchronized with normal tissues in mice fed ad libitum. Regular pattern of oscillations is disturbed in mice on restricted feeding schedule. (C) Quantitative analysis of Bmal1 and Per2 gene expression in excised tumors (top panel) and liver (low panel) collected at ZT2 and ZT14. AL, ad lib; RF, restricted feeding. Luciferase signal in tumor lysates is normalized by protein concentration. Values represent mean (n = 6) ± SEM. In ad libitum fed mice Bmal1 expression is higher and Per2 expression is lower at ZT2 compared to ZT14 (*p = 0.06 and **p<0.0001 respectively, Student’s t-test). Both Bmal1 and Per2 expression are reversed in phase in xenografts of mice on restricted feeding schedule with Bmal1 expression getting lower and Per2 expression getting higher at ZT2 compared to ZT14 (#, ##p=0.001, Student’s t-test). These changes match phase shifts induced by restricted feeding in the liver (bottom panel). *p = 0.02; **p = 0.002; #p = 0.0004; ##p = 0.004; Student’s t-test.

DISCUSSION

It is well established that successful and less debilitating anti-cancer treatment relies on a subtle balance between the drug-induced damage to tumors and drug-induced toxicity to normal tissues. Therefore, finding novel approaches for selective modulation of stress response between normal and malignant cells could significantly improve therapeutic indexes of existing strategies. In this respect, the components of the molecular clock machinery represent attractive therapeutic targets. Previous work showed that clocks are present in every mammalian tissue where they orchestrate rhythmic expression of hundreds of genes involved in multiple cellular processes including those that are related to various aspects of cancer treatment (Antoch & Kondratov, 2013). It is believed that these oscillations underlie circadian variations in drug metabolism, therapeutic efficacy and tolerability (Lévi et al., 2010; Ortiz-Tudela et al., 2013).

The major question that yet remains unanswered is whether tumors retained functional clocks and, if so, whether they are functioning in synchrony with normal tissues and whether they are amenable for manipulations. Recent studies addressing the expression levels of clock proteins in different types of tumors have suggested that the circadian clock of tumor cells may be disrupted (Yu & Weaver, 2011). However, the data remain inconsistent, and as a result, inconclusive. Whereas some studies show a down-regulation of clock gene expression compared to normal surrounding tissue (Cao et al., 2009; Gery et al., 2006; Hsu et al., 2012; Krugluger et al., 2007; Lin et al., 2008; Mostafaie et al., 2009; Pogue-Geile et al., 2006; Relles et al., 2013; Winter et al., 2007; Xia et al., 2010; Yeh et al., 2005), others report up-regulation of clock gene expression in certain tumors (Geusz et al., 2010; Sato et al., 2009; Yu et al., 2013). Tokunaga and colleagues showed that in ovarian cancer some clock genes are down-regulated (Per1, Per2, Cry2, Clock and CKIɛ) whereas others are up-regulated (Cry1, Per3 and Bmal1) compared to normal ovaries (Tokunaga et al., 2008). Although these studies give some insight into the relationship of the circadian clock of different tumor types compared to normal tissue, they remain correlative as it is still not clear whether they represent the cause or the consequence of malignant tumor development. In addition, the functionality of tumor clock (i.e. the ability to sustain rhythmicity in gene expression and the ability to respond to changing environment) has not been tested yet.

Since tumors often loose pathways that are either detrimental or neutral for their survival and growth, it has been generally accepted that the functionality of tumor clocks is likely to be disrupted and this may contribute to deregulation of clock-controlled aspects of cell cycle, apoptosis, migration, differentiation in favor of tumor progression (Yu & Weaver, 2011). However, our data show that two types of tumors of connective tissue origin (spontaneously developed sarcomas in p^53−/−mutant mice and HT1080 human fibrosarcoma cells grown in SCID mice) display significant rhythmic profiles of Per2- and Bmal1-driven luciferase expression with the peaks and troughs coinciding with normal tissues. Although reduced in amplitude, Per2-driven luciferase recorded from sarcoma explants ex vivo displays oscillations that are as robust as those found in explants of normal tissues and the peak phase occurs, similar to liver, lung and spleen, in the dark phase suggesting that like in other peripheral tissues (Yoo et al., 2004), these clocks are self-sustained.

Another important finding that comes out of our study is that molecular clocks in tumors can be manipulated in the context of an organism. This is an important result that opens new avenues for translational applications. Hence, it suggests that therapeutic efficacy of anti-cancer treatment may be improved not only by clock-dependent protection of normal tissues but also through clock-dependent regulation of tumor-specific functions (efficiency of drug delivery, sensitivity, etc.). Interesting initial steps in this direction has been reported recently in application to photodynamic therapy (PDT). Thus, it has been shown that intratumoral accumulation of photosensitizers conjugated with anti-VEGF antibody in transplanted Lewis lung carcinoma and sarcoma-180 followed circadian pattern of VEGF expression (Gamaleia et al., 2012). As a consequence, PDT efficacy was higher when delivered at time of daily peak in VEGF expression. Our data also further supports the idea of potential value of clock-targeting pharmaceuticals in combinational anti-cancer therapies by applying it to clock components of tumors.

One of the confines of our study is that it is limited to tumors of connective tissue origin, which initially possess robust molecular clock both in tissues in vivo and in isolation. Thus, normal fibroblasts can be synchronized by variety of treatments, including glucocorticoids (Balsalobre et al., 2000a), glucose (Hirota et al., 2002), temperature (Sladek & Sumova, 2013), insulin (Tahara et al., 2011) and others. At this point it is not clear whether the idea of functional clocks in tumors applies to other tumor types. Recent studies of mammary epithelial cells have demonstrated robust oscillation in normal cells and the loss of rhythmic gene expression in cancer cell lines (Rossetti et al., 2012). It is important to note that these studies are limited to cell culture and the ability of breast tumor cells to be entrained in the context of an organism has not been tested.

Another important question that still remains unanswered is whether the robustness of tumor clock decreases with tumor progression. The inverse regulation (effect of cancer progression on an organism’s circadian function) is well-documented in cancer patients showing altered circadian variation of corticosterone, cortisol, melatonin, prolactin, temperature or rest-activity rhythms characterized by decrease in amplitude, changes in the period and phase of activity, etc. (Chevalier et al., 2003; Grutsch et al., 2011; Innominato et al., 2009a, 2009b; Lévi et al., 2007; Mormont & Lévi, 1997; Mormont et al., 2000; Sephton & Spiegel, 2003). In our study, development of spontaneous tumors in mice did not affect wheel-running activity rhythm under LD12:12 or DD conditions (Supplementary Figure 1), which is in line with previously reported data (Geusz et al., 2010). It also had no effect on robustness of tissue-specific clocks, which showed very similar pattern of oscillation in tumor-free and tumor-bearing animals; however, it is still not clear whether the lack of the effect is due to presumably early stages of tumorigenesis, at which mice were tested or it represents the general phenomenon.

Taken together, our results open up a new spectrum of possibilities to develop new therapies where clock-targeting pharmaceutical compounds are directed at manipulation of the tumor clock instead of normal tissues. Further research is needed to investigate whether tumors sensitivity to drugs can be manipulated by circadian clock. For now, our work establishes a new vision of the role of circadian clock in tumors.