Abstract
The circadian clock exists in virtually every cell and regulates key biological processes in cells and tissues. Even in cancer cells, DNA synthesis, cell division and tumor growth are gated by the circadian clock. This study examined the gene expression profiles of transplanted mouse breast tumor cells under normal light-dark (LD) as well as constant dark (DD) conditions. It was found that the overall percentage of rhythmic transcripts in breast tumor tissue was lower than that in normal tissue. Few transcripts had unaltered rhythmic expression patterns under both LD and DD conditions. Most rhythmic transcripts in DD displayed 12h or shorter periods. These results suggest that in addition to the circadian clock control of gene transcription, altering light, feeding, physical activity and other factors characteristically affect the expression of many genes.
Circadian organization is highly conserved across species and is essential to an organism's well-being (1-3). Circadian rhythms are generated both directly and indirectly by a set of core circadian clock genes. The mammalian circadian clockwork mechanism includes both negative and positive transcription-translation feedback loops which generate approximately 24 h physiological cycles. Two transcription factors, CLK (or NPAS2), and BMAL1, form complexes that activate the transcription of Per (Per1, 2) and Cry (Cry1, 2) genes. The resultant PER and CRY proteins form dimers which, in turn, suppress the transcription of their own genes. As a result, the expression of the Per and Cry genes oscillates with a near 24 h periodicity (1, 3). The second circadian clock loop involves the rhythmic expression of the Rorα and Rev-erbα genes, whose expressions are also activated by CLK/BMAL1 and suppressed by PER/CRY. RORα and REV-ERBα control the transcription of the Bmal1 gene (4). In this way, the rhythmic expression of the Bmal1 gene stabilizes the circadian oscillator. Light is a potent and the most well studied environmental time cue that sets the circadian clock. Daily light-dark cycles entrain the mammalian clock by influencing directly the master clock located in the suprachiasmatic nuclei (SCN), which entrains the clocks within peripheral tissues. Circadian rhythmicity is a self-sustained process which remains organized under constant conditions, such as constant darkness (DD), in both the organism and its cells (1, 3).
Virtually all fundamental biological processes are circadian regulated. In proliferating tissues, DNA synthesis, cell division, DNA damage repair and apoptosis are gated by the circadian clock (5-9). In humans, the activity of thymidylate synthase (TS), one of the most frequently targeted enzymes in proliferation pathways that control cancers, peaks in the early afternoon in oral mucosa and gut epithelium (8, 10, 11). The expression of cell cycle genes and growth factor genes as well as tumor growth rates are also under the control of the circadian clock (12). In a mouse breast cancer model, the protein content of VEGF, tumor blood flow rate, TS enzyme activity, tumor cell mitotic index and tumor growth rate display robust and reproducible rhythmic daily variations (6, 7, 13).
Microarray data have indicated that more than 10% of the genes in peripheral tissues are expressed rhythmically each day (14-16). The majority of these rhythmically expressed genes are tissue specific, indicating that the circadian clock is tightly coupled to key tissue specific functions. Since it has been shown that tumor growth is gated by the circadian clock (17), this study examined the gene expression profiles of mouse breast cancer cells growing within the C3H mouse under normal 12 h light and 12 h dark (LD) as well as under constant darkness (DD) conditions. It was found that the expression of approximately 4% and 6% of total RNA transcripts within this breast cancer type varies in LD and DD conditions, respectively. However, most of them did not demonstrate a 24 h or a 12 h period. Of these transcripts with a 24 h or a 12 h period, very few demonstrated a constant expression pattern under both LD and DD conditions, indicating that in addition to the circadian clock control of gene transcription, altering light/dark also characteristically affects the expression of many breast cancer genes.
Materials and Methods
Animals and tumors. C3HeB/FeJ (C3H) mice were purchased from Jackson Laboratory (Bar Harbor, ME, USA) and kept under a 12 h light and 12 h dark (LD) schedule. When the mice were 13-15 weeks of age, 2×104 MTCL (murine mammary tumor cell line) cells were injected subcutaneously in the backs of the mice, as described previously (17, 18). When the tumor size reached 600-700 mm3 (11 to 12 days after tumor inoculation), the mice were divided randomly into two groups, one of which was kept under LD conditions and the other under DD conditions. After 48 h, tumors from both groups were collected at six time points (zeitgeber time (zt) and circadian time (ct) 2, 6, 10, 14, 18, and 22) and flash frozen using liquid nitrogen. All experiments were performed in compliance with the NIH and VA guidelines for care and use of laboratory animals.
Reverse transcription and real-time PCR. Total RNA was isolated from tumors using Trizol. Two μg of RNA was reverse-transcribed by M-MLV reverse transcriptase using random primers (Invitrogen, Carlsbad, CA, USA). Real time quantitative PCR analyses were performed on an iCycler iQ PCR system (BioRad, CA, USA). PCR reactions were performed in triplicate. The amount of the target gene expression was computed with respect to the endogenous reference (Gapdh gene). The primers for each gene were reported previously (19).
Microarray experiment. RNA extraction, reverse transcription, and hybridization to the Affymetrix 430.20 mouse whole genome array were performed by Paradigm Array Labs (Morriville, NC, USA). The Affymetrix 430 2.0 array has probes for 45,101 mouse transcripts. Equal amounts of RNA from three tumors collected at the same time-point were pooled.
Data analysis. The microarray data were analyzed by Genespring GX software (Agilent Technologies, Santa Clara, CA, USA). Relative gene expression was determined by comparing intensity levels among samples collected at 6 time points (LD or DD). With respect to zt/ct 2, transcripts with fold-change >1.3, log10 intensity >–1.2 and p<0.001 within 6 time points were considered to be differentially expressed transcripts. Gene ontology analysis was performed to identify the biological processes associated with each differentially expressed transcript.
Study design and definition of rhythmic expression patterns. Figure 1 shows the study design used. The gene expressions were compared among tumor RNAs from six time points in LD and DD. The expression level of each transcript at zt/ct 2 was set as the reference point. Differentially expressed transcripts were indentified by comparing their RNA levels of zt/ct 6, 10 14, 18 and 22 with that of zt/ct 2. If the expression of a particular transcript differed only at one time point, this transcript was not considered to be differentially expressed within each 24 h period because it was not possible to determine what caused this singular alteration of expression. If the expression of a transcript significantly differed at two or more time points compared with its expression level at zt/ct 2, this transcript was considered to be differentially expressed within a 24 h period. Among these differentially expressed transcripts, if the expression levels displayed one peak and one trough during each 24 h, they were considered to have a 24 h circadian rhythm. If the expression levels had two peaks or two troughs within each 24 h, and the two peaks or the two troughs were 12 h apart, these transcripts were defined to have a 12 h rhythm.
Results
Rhythmically expressed transcripts in mouse breast tumors under LD and DD conditions. The size and calculated growth rate of MTCL mouse breast tumor demonstrates a robust daily rhythm in vivo under 12 h light 12 h dark conditions (17). However, the tumor growth rate displays a two-peak instead of typical one-peak circadian pattern. One of the peak growth rates occurs in the middle of the day and the other in the middle of the night (17).
The pattern of a two-peak daily tumor size and growth rate prompted an examination of the in vivo circadian transcription profile of this MTCL mouse breast tumor in this study. An Affymetrix microarray was used to compare the relative RNA amount of 45,101 transcripts (22,000 genes) of MTCL tumors collected at six time points during a 24 h period. Approximately 1.8% of the transcripts (819 and 824 of 45,101 transcripts in total under LD and DD conditions, respectively) may be rhythmically expressed because their expressions vary at multiple time points (Table I, see Materials and Methods for the definition of rhythmic expression). Among these rhythmically expressed transcripts, some of them may have periods shorter than 12 h. However, sampling every 4 h does not allow a determination of whether these transcripts have a rhythm with a period of 4 h, 6 h or 8 h. Therefore, this study focused on the transcripts with either a 12 h or a 24 h rhythm (414 and 212 transcripts in LD and DD respectively). Among these, 38 transcripts had either a 12 h or a 24 h rhythm under both LD and DD conditions (Figure 2).
Expression patterns of transcripts only rhythmic in LD. A total of 376 transcripts displayed a 12 h or a 24 h expression rhythm only under the LD condition. Most of them (341 of 376) had a 24 h rhythm, while less than 10% (35 of 376) of the transcripts had a 12 h rhythm under the LD condition (Table II). Of those with a 24 h rhythm, more than half demonstrated a `classic' circadian pattern, i.e. their highest expression and lowest expression points were 12 h apart. The rest had a peak-trough either 8 h or 16 h apart. Most transcripts with 12 h expression patterns under LD conditions either had peaks at zt 2 and zt 14 or troughs at zt 2 and zt14 (Table II).
Further analysis reveals that the majority of these rhythmic transcripts in LD can be categorized into seven groups. The first two groups represented most of the classic 24 h circadian transcripts. They had peaks at zt2 and troughs at zt14 (Figure 3a) or peaks at zt14 and troughs at zt2 (Figure 3b). Unlike the expression patterns in the first two groups, the transcripts in the third and fourth group had the peak-trough in anti-phase, 8 h apart, either having the peak at zt2 and trough at zt10, or the peak at zt10 and trough at zt2 (Figure 3c, d). The expression patterns of the fifth and sixth group were also anti-phase. Their expression levels were highest at zt2 and lowest at zt18, or highest at zt18 and lowest at zt2 (Figure 3e, f). The transcripts of the last group displayed 12 h rhythms under LD conditions. These transcripts had two daily peaks at zt2 and zt14, and two troughs at zt6 and zt18 (Figure 3g).
Transcript expression patterns of transcripts only rhythmic in DD. The gene expressions of DD samples displayed unique circadian patterns. Unlike the rhythmically expressed transcripts in LD, most of the transcripts that were rhythmic only in DD had 12 h rhythms (113 out of 174 transcripts, 65%) while 35% of them (61 out of 174) had 24 h rhythms (Table II). The expression patterns of these DD samples can be characterized into six main cluster patterns (Figure 4). The first two groups had typical 24 h circadian rhythms with their peak-trough 12 h apart. They either had peaks at ct2 and troughs at ct14 (Figure 4a), or peaks at ct14 and troughs at ct2 (Figure 4b). The next two groups had two high expressions at ct2 and ct14. They also had two low expressions at ct6 and ct18 (Figure 4c) or at ct10 and ct22 (Figure 4d). These two groups represented most of the transcripts with 12 h rhythms in DD. The transcripts in the fifth group had peaks at ct10 and ct22, and the transcripts in the sixth group had peaks at ct6 and ct18. These two groups of transcripts both had troughs at ct2 and ct14 (Figure 4e, f).
Expression patterns of transcripts that have rhythms under both LD and DD. Among the 38 transcripts that displayed 12 h or 24 h rhythms under both LD and DD conditions, 16 of them (42%) had 24 h rhythms and 12 of them (31.6%) had 12 h rhythms. The expression patterns of those transcripts that maintained their 24 h rhythms in LD and DD can be categorized into 3 groups (Figure 5, Panel A). The first group of transcripts had peaks at the day/night transition time (zt/ct 10, 14) and troughs at night/day transition time (zt/ct 2, 22). The second group had the anti-phase expression pattern of the first group. The transcripts in the third group displayed anti-phase expression patterns in LD and DD. The transcripts that had 12h rhythms in both LD and DD all had low expression levels in the middle of the day and in the middle of the night, and had high expression levels at day/night and night/day transition times (Figure 5 Panel B). Ten transcripts (26.4 %) had different time structures in LD and DD. Nine of them demonstrated 24 h rhythms in LD, with expression peaks at the day/night transition time, troughs at the night/day transition time, and 12 h rhythms in DD. Only one transcript demonstrated a 12 h rhythm in LD but a 24 h rhythm in DD (Figure 5 Panel C).
Pattern analysis revealed another feature of these rhythmic transcripts in LD and DD. The overall amplitude of rhythmic transcripts under the DD condition was higher than that under the LD condition. The amplitudes of these 38 transcripts were compared with daily rhythms in both LD and DD. These transcripts demonstrated more robust rhythms under constant darkness (Figure 6).
Functional categories of rhythmically expressed transcripts. The transcripts with 12 h and 24 h daily rhythms were grouped according to their major associated functions (Figure 7, and data not shown). Most of the transcripts that cycle daily in LD, DD, or both are involved in regulating cell cycle, cellular metabolism, cytoskeleton and immune function (data not shown). There were more rhythmically expressed cell cycle regulators in LD conditions, while more cytoskeleton and structure related genes had circadian rhythms in DD. Compared with those cell cycle regulators that were rhythmic only in LD or DD, a larger proportion of rhythmic genes that had rhythms in both LD and DD were immune related. Genes involved in angiogenesis are directly related to tumor cell proliferation. There were three and five angiogenesis-related genes that had daily rhythms in LD and DD respectively. Of them, two were rhythmic in both LD and DD.
Expression of core clock genes in tumors under LD and DD conditions. Because the circadian clock genes cycle with lower amplitudes in this tumor (19), the microarray analysis did not identify the rhythmic expression of core circadian genes in either LD or DD tumor samples. Real-time PCR, a more sensitive method, was used to compare the expression patterns of core clock genes and a clock output gene, Dbp, in LD and DD tumor samples (Figure 8). Per1, Per2, Cry1, Bmal1, Rev-erbα and Dbp have a damped but statistically significant 24 h rhythm in LD tumor samples. The expression patterns with respect to peak timing of these clock genes were very similar to those reported in normal tissues (20). In DD tumor samples, Bmal1 expression lost its circadian rhythm. Dbp, a clock controlled gene, tended to have a 12 h circadian rhythm under DD condition compared to a 24 h rhythm in LD. The circadian rhythms of Per1 and Rev-erbα in DD were slightly damped compared with those in LD samples. Per2 expressions had identical patterns in LD and DD tumor samples. Increased amplitudes of clock gene expressions were not seen in DD.
Discussion
In this study, daily rhythmic gene expression patterns of transplanted mouse breast tumors were compared under LD and DD conditions. This is the first demonstration of the circadian gene expression profile in mouse breast cancer under both LD and DD conditions. It has been previously demonstrated that despite the fact that cancer cells proliferate rapidly, tumor growth in vivo is still under the control of the circadian clock (7, 13, 17). However, the rhythmic expressions of clock genes were not detectable by the microarray analysis due to their low amplitudes. The percentage of transcripts with daily rhythms in the tumor was somewhat lower than that reported in normal peripheral tissues (14). This may reflect the fact that the circadian clock is not well organized in tumors compared to normal tissues. Interestingly, the number of differentially expressed tumor transcripts was higher in DD than in LD. However, more than half of these LD transcripts and more than two third of these DD transcripts were only differently expressed at one time point during the day. Their expressions did not differ among the other five time points within a 24 h period. It is unclear what causes these singular altered expressions. Therefore, these transcripts were not considered to have circadian expression patterns. The numbers of transcripts that were differently expressed at multiple time points in these 6 time points under LD and DD conditions were similar. Some of them may have had ultradian expression patterns with shorter periods that could not be defined accurately because the sampling frequency was not high enough to determine the existence of the ultradian rhythm. Therefore, this study only focused on the transcripts with either a 12 h or a 24 h rhythm. It is provocative that only a few of these oscillated under both LD and DD conditions. The differential daily gene expression pattern in LD and DD is not unique to this tumor; similar results have been reported in chick pineal gland and Drosophila head (21-23). The expression of those genes that had 24 h circadian rhythms only under LD may have been driven solely by the light (24-27). On the other hand, the expression of those genes that had 24 h rhythms only in the DD condition may have been controlled by the endogenous circadian clock and masked by the daily light cycle. It is also possible that the expression of some genes was induced by darkness, which also acts as a circadian cue (28).
One major difference between the rhythmic expression pattern of LD and DD tumor samples in this study was that the amplitudes of many rhythmic genes were increased under the DD condition. It is possible that the expression of such genes was controlled mainly by the clock, and light may have suppressed or stimulated their expressions. Therefore, their expressions will have had more robust and uniform rhythms without the interference of light.
Most cyclic genes under LD conditions have a single peak within each 24 h period, but a significant proportion of cycling gene transcripts (9% in this study) have two daily peaks. In fact, the 12 h rhythm is fairly common in normal tissues under normal light and dark conditions. Gene expressions in normal chick pineal gland and mouse distal colon under LD conditions exhibit both 24 h and 12 h rhythms (29). Under constant conditions (DD), most rhythmically expressed genes have two peaks within every 24 h (65%). Similar expression patterns have been demonstrated in normal tissues under DD conditions (30). One possibility is that the expression of these clock-controlled genes may respond to other stimuli such as light, feeding and physical activity. The effects of these stimuli and endogenous clock may then generate gene expression patterns with 12 h, 8 h or 6 h periods. Since light is such a potent environmental cue, under LD conditions, it can mask or attenuate the effects of other stimuli. Therefore, when light is removed, the expression of these genes is controlled by the combination of multiple factors including the circadian clock, feeding, physical activity etc. which may lead to ultradian expression patterns under DD conditions.
It is known that cell division can desynchronize the circadian clock and tissues containing rapidly dividing cells have less robust rhythms of clock gene expression as a population (20, 31). Even though the expressions of core circadian clock genes in tumors have damped 24 h rhythms, tumor growth has a robust 12 h rhythm. Several genes related to cell proliferation and apoptosis display 12 h rhythmic expression patterns under LD conditions, including Ephb3, Ctnnd2, Rbm3, Plrg1, Kif20a, Ccng2 and Casp6. EPHB3 is one of the Eph receptor tyrosine kinases that regulate tumor growth and angiogenesis in vivo (32). CTNND2, or delta-catenin, is required for vascular assembly and tumor growth (33). The RNA-binding protein RBM3 is required for cell proliferation and cell death inhibition under serum starvation (34, 35). PLRG1 is a component of splicesome and is essential for cell proliferation (36). KIF20A belongs to the kinesin family of motor protein. Down-regulation of Kif20a has been shown to inhibit the growth of pancreatic cancer cells (37). Ctnnd2 gene encodes cyclin G2 which is a negative regulator of cell growth (38, 39). Caspase 6 is one of major caspases functioning in apoptosis (40, 41). It is unclear whether there is a causal relationship between the expression pattern of these genes and the tumor growth pattern. At the molecular level, it has been shown that other cell cycle indicators such as the tumor mitotic index, thymidylate synthase activity (S phase-specific), and cyclin E protein also clearly have 2 daily peaks (7, 13). Since this study identified more rhythmic genes with 2 or 3 peaks per 24 h under constant dark conditions, it would be important to determine the tumor growth rate and rhythm under the DD condition in vivo. A careful review of other circadian gene expression patterns in both normal and malignant tissues confirms the primacy of 12 h periodicity. It is possible that the unimodal circadian clock gene rhythm can generate and coordinate multimodal gene expressions. Finally, this study revealed that the variation of gene expression occurs most prominently at the light-dark and dark-light transition zone.
In summary, this study indicates that in tumors, as in normal tissues, the expression of a subset of genes is controlled by the daily light-dark cycle, by the endogenous circadian clock, or both. This finding should provide information for the further study of the role of circadian clock in tumor biology.
Footnotes
-
↵* Both authors contributed equally to this work.
- Received September 9, 2010.
- Revision received September 27, 2010.
- Accepted September 27, 2010.
- Copyright© 2010 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved