Induction of circadian clock function in B16 melanoma cells
First we characterized circadian rhythm generation in B16 melanoma cells. Reporter constructs consisting of the luciferase gene under the control of the Bmal1 or the Per2 promoter (Bmal1-Luc, Per2-Luc) were stably transfected in B16 cells. The bioluminescence of these cells was arrhythmic (Fig. 1a, from −44 to 0 h). However, upon addition of DEX, an agonist of the glucocorticoid receptor (which is expressed in B16 cells [28]) known to induce circadian rhythms in cultured cells, we observed rhythmic Bmal1-Luc and Per2-Luc promoter activity with the expected opposite phases (Fig. 1a) [34], which, however dampened earlier than in various non-cancer cell lines [35, 36].
To address whether the loss of rhythmicity after a few cycles resulted from circadian dysfunction in individual cells or from a desynchronization among rhythmic cells, we analyzed Per2-Luc bioluminescence at the single-cell level (Fig. 1b, Additional file 3, Additional file 4A–E). Although rhythms in the circadian range (first to second peak: 24.2 ± 1.4 h, second to third peak: 25.3 ± 2.0 h, third to fourth peak: 24.3 ± 1.5 h; Additional file 4E, F) were induced by DEX in ~90% of the cells, they quickly dampened within two to three cycles: 50% and 36% of the cells showed a second or third circadian peak, respectively, but circadian rhythms in bioluminescence were greatly suppressed in all cells after 132 h (Fig. 1b, Additional file 4A–D, G). Similarly the amplitude of single-cell rhythms dropped by about 80% within three cycles and the phase distribution became dispersed after 48 h (Additional file 4H, I). Repeated treatments showed that clock gene suppression was not due to cell death (Additional file 4J). Overall, these data indicated a suppression of clock gene expression in B16 cells rather than a desynchronization between single cells.
Additional file 3: Induction of rhythmic clock gene expression in B16 cells. Per2-Luc single-cell bioluminescence of cultured B16 cells for 142 h after dexamethasone treatment. (AVI 1831 kb)
Rhythmic clock gene expression was also observed when employing alternative methods for clock gene activation [30, 37]: FSK, an activator of adenylyl cyclase and of the cAMP/PKA pathway (Fig. 1c), heat shock (Fig. 1d) and serum shock (Fig. 1e) all induced transcript oscillations in B16 cells. In addition to Bmal1 and Per2, the mRNAs of clock genes Per1, Cry1 and Nr1d1 also fluctuated rhythmically in B16 cells after DEX treatment but not in untreated cells (Fig. 1f–j).
Altogether, these results indicated that the B16 cells harbor an unstable but inducible circadian oscillator. We decided to take advantage of this property to address the role of the tumor cell-intrinsic clock in cell proliferation and tumor growth.
The cell cycle is under circadian control after dexamethasone treatment in vitro
Because the molecular clockwork was shown to regulate the expression of genes encoding cell cycle regulators [2], we set out to study the expression of such genes in B16 cells. B16 cells were treated with DEX for 2 h, harvested over 24 h, and tested for mRNA expression of WEE 1 homolog 1 (Wee1), Cyclin-dependent kinase 1 (Cdk1), Cdk2, Cyclin E, Myelocytomatosis oncogene c (c-Myc) and Cyclin-dependent kinase inhibitor 1A (p21). All six genes showed predominant 24-h rhythm components in transcript abundance as a consequence of DEX treatment (Fig. 1k–p).
Because these factors are involved in cell cycle checkpoints, we assessed the effect of DEX treatment on the distribution of B16 cells among cell cycle stages. Cells were collected at different time points over 24 h, stained for BrdU incorporation and with 7AAD, and analyzed by flow cytometry (Fig. 2a, b). The proportion of cells in G0/1, G2/M and S phases was found to be rhythmic in cells treated with DEX, while control cells showed no circadian variation (Fig. 2c–e). Interestingly, fewer cells entered the S phase 24, 36 and 42 h after DEX treatment (Fig. 2e), indicating less DNA replication, while more cells were found in G0/G1 phases (Fig. 2c).
Induction of the circadian clock slows down B16 cell proliferation in vitro
Given that activation of the circadian clock in B16 cells triggered rhythms of cell cycle genes and phases, we next tested whether activation of the clock in B16 cells influences their proliferation. Thirty-six hours after DEX treatment, we counted significantly fewer live cells than without treatment (Fig. 2f), while the amount of dead cells in the medium was unchanged after 12–48 h (Additional file 4N). The effect was even more pronounced after 2 days, with ~50% fewer live cells but similar numbers of dead cells after DEX treatment (Fig. 2g, Additional file 4O). Of note, the numbers of live and dead cells were similar before and immediately after the end of the 2 h DEX treatment, suggesting that the treatment did not acutely induce cellular toxicity (Additional file 4K).
Because rhythms quickly dampened after treatment (Fig. 1) we tested whether repeated DEX treatment could further inhibit cell proliferation. While a single DEX treatment significantly reduced cell numbers after 50 h, a second treatment further reduced the amount of cells after 96 h (Fig. 2h). Population doubling time (PDT) was increased to 22.7 h in single DEX-treated cells and further to 23.0 h in double DEX-treated cells compared to 16.5 h PDT in untreated control cells. Annexin V staining indicated that the differences in cell counts were not due to differences in levels of apoptosis (Additional file 4P). Total cell numbers were even more reduced after three DEX treatments administered every 48 h (Additional file 4L, M).
Similar experiments were also conducted with other stimuli activating the B16 clock (Fig. 1). Both a single FSK treatment (Fig. 2f, g) and repeated FSK treatments (Additional file 4M) were as effective as DEX in slowing down B16 cell proliferation. Also, FSK treatment did not affect cell death rates (Additional file 4N, O), nor did it acutely impact on cell numbers right after the 2 h treatment (Additional file 4K). Moreover, exposing the cells to a 30 min heat shock was sufficient to reduce proliferation without affecting apoptosis (Fig. 2i, Additional file 4Q). PDT was increased to 20.7 h after heat shock compared to 17.2 h in untreated cells. These data suggested that the reduction in cell proliferation might be due to action on intrinsic clock function, as activating clock gene expression was the common denominator of all three treatments.
Dexamethasone activates the circadian clock in B16 tumors
B16 melanoma cells, which represent a well-established and widely used mouse model for human melanoma [38], form lung and s.c. tumors when injected in the tail vein and subcutaneously, respectively (Additional file 5A, B). The analysis of circadian clock gene expression in B16 lung tumors unveiled suppressed or arrhythmic Per1, Per2 and Bmal1 expression, while there was robust circadian oscillation in the neighboring lung tissue (Fig. 3a). Similarly, Per1, Per2, Bmal1 and Nr1d1 expression was arrhythmic in s.c. B16 tumors (Fig. 3b).
DEX, an agonist of the glucocorticoid receptor, is well known to reset cellular clocks by inducing Per gene expression [39]. Upon repeated addition of DEX, we observed rhythmic Bmal1-Luciferase reporter expression in slice cultures of explanted B16 lung tumors (Fig. 3c). Moreover, monitoring of single cells within the same lung tumor slice indicated an absence of Bmal1 rhythms before treatment, but circadian oscillation of Bmal1 after DEX treatment (Additional file 6, Fig. 3d). Thus, DEX induces de novo rhythmic Bmal1 gene expression in single B16 cells rather than synchronizing cellular oscillators. This is consistent with the single-cell data obtained on B16 cells in vitro (Fig. 1, Additional files 3 and 4).
Additional file 6: Bioluminescence monitoring in the absence and after induction of rhythmic clock gene expression in cells of a B16 lung tumor. Bmal1-Luc single-cell bioluminescence in a B16 lung tumor slice for 44 h in the absence of DEX treatment and for 68 h after DEX treatment. (AVI 2935 kb)
To further confirm an activation of the tumor clock by DEX in vivo, s.c. tumors were injected intra-tumorally with DEX every 48 h to ensure maintenance of tumor clock activation. Rhythmic clock gene expression in tumors was found after DEX but not PBS treatment (Fig. 3e–i). Per1, Per2, Cry1 and Nr1d1 expression showed a significant effect of time and treatment. Rhythmicity of Bmal1 mRNA expression did not reach significance (Fig. 3i). However, the ANOVA analysis showed an effect of time (see Additional file 1), and immunohistochemistry on s.c. tumor slices revealed a significant rhythm of BMAL1 protein levels in DEX-treated tumors, but not in PBS-treated tumors (Fig. 3j, Additional file 5C, D).
The efficiency and specificity of the intra-tumoral injections were evaluated by injecting methylene blue in tumors. The injected fluid spanned the whole tumor tissue 6 h after intra-tumoral injection, but was absent in surrounding tissues (Additional file 5E). To further assess the specificity of the response, clock gene expression in the liver - a non-cancerous, peripheral tissue - was compared between PBS- and DEX-injected mice: we found no significant differences between the treatment groups (Additional file 5F).
Altogether, these results indicated that the B16 tumors harbor a suppressed but inducible circadian oscillator. Moreover, these experiments showed that repeated DEX injection consistently induced rhythmic clock gene expression in B16 tumors rather than re-synchronized them. The ability to compare the B16 tumors with or without circadian clock function gave us the opportunity to test the role of the tumor-intrinsic clock in regulating cell cycle and tumor growth.
The cell cycle is under circadian control after dexamethasone treatment of B16 tumors
We measured the expression of cell cycle regulators in s.c. B16 tumors harvested over 24 h. Similar to the data obtained in vitro (Fig. 1k–p), rhythmic protein expression of CYCLIN E, p21 and c-MYC was found in DEX-treated tumors, but not in PBS-treated tumors (Fig. 4a–c, Additional file 7A–C). On the other hand, WEE1, CDK1, CDK2 and p57, a direct glucocorticoid receptor target [40], did not show a daily variation in the DEX-treated tumors (Fig. 4d–g).
Homogenized tumor cells collected at different time points were stained for BrdU incorporation and with 7AAD. The proportion of cells in G0/1 and S phases was found to be rhythmic in DEX-treated control tumors, whereas PBS-treated tumors showed no circadian variation (Fig. 4h–j, Additional file 7D). Interestingly, fewer rhythmic cells entered the S phase in DEX-treated tumors, indicating less DNA replication, whereas more cells were found in G0/G1 phases. Importantly, DEX treatment reduced rather than increased the number of cells in G0 arrest as indicated by Ki67/7AAD staining in vitro (see control cells in Additional file 8). Consistent with results obtained from BrdU stainings, cell numbers in G1 phase were increased, while cells distributed in S/G2/M phases were slightly reduced (compare Figs. 2 and 4 and Additional file 8). Thus, reduced cell numbers after DEX treatment were not caused by cell cycle arrest. Moreover, even though entrance to G2/M phase was not rhythmically controlled in the tumor, mitotic index assessed by pHH3 staining indicated a rhythmic percentage of cells undergoing mitosis (Fig. 4k, Additional file 7E). Interestingly, cells from DEX-treated tumors underwent apoptosis in a circadian manner (Fig. 4l, Additional file 7F).
Dexamethasone treatment slows down B16 tumor growth
Given that activation of the circadian clock in B16 tumors triggered rhythms of cell cycle genes and phases, we then evaluated whether the activation of clock function in B16 tumors was paralleled by a reduction in tumor growth. To this end, s.c. B16 tumor growth was compared between DEX- and PBS-treated tumors. DEX treatment significantly slowed down tumor growth (by ~60% after 8 days) compared to PBS treatment (Fig. 4m). Importantly though, Annexin V staining indicated that the differences in tumor growth were not due to differences in the levels of apoptosis, because a similar proportion of apoptotic cells were found in DEX- and PBS-treated tumors (when averaged over the 24-h day, PBS: 14.6 ± 2.3% versus DEX: 11.5 ± 4.6, extra sum-of squares F-test: p = 0.25; Fig. 4l).
To rule out that reduced tumor growth after DEX treatment may be caused by DEX-induced immune cell infiltration in the tumor, we repeated this experiment in immune-deficient NSG mice, which lack T cells, B cells and natural killer cells. Similarly to results in C57BL/6J hosts, tumor growth was strongly reduced by DEX in NSG mice (Fig. 4n), and no differences were found in the levels of infiltration of remaining immune cells between DEX- and PBS-treated tumors in NSG mice (Additional file 9A–C). Again, the proportion of cells undergoing apoptosis was not significantly different between DEX- and PBS-treated tumors collected at Circadian Time (CT)14 in NSG mice (Additional file 9D). Moreover, the proportions of cells undergoing cell division (G1/S/G2/M) and of those in G0 arrest were unchanged by DEX treatment of tumors in NSG mice (Additional file 9E). Together with results obtained in vitro that excluded cell cycle arrest (Additional file 8A) and in C57BL/6J mice that showed more cells in G0/G1 phase at CT14 upon DEX treatment, these results indicated that cells remain more in G1 phase and transit less to the S phase.
Knockdown of Bmal1 abolishes the effects of dexamethasone on clock genes and cell cycle genes
To test whether the activation of the clock in B16 tumors is the causal link between DEX treatment and the inhibition of tumor growth, experiments were repeated in B16 tumors with a disrupted circadian clock. Using a lentiviral vector, we stably introduced into B16 cells an shRNA against Bmal1, a necessary component of the circadian clock [4]. Bmal1 shRNA-transfected cells expressed ~80% less Bmal1 RNA and ~65% less BMAL1 protein than Scrambled shRNA control cells (Fig. 5a, b). B16 cells stably expressing shRNA against Bmal1 or Scrambled shRNA were injected subcutaneously into mice to form either tumors lacking a functional clock or control tumors. Similarly to non-transfected tumors, clock gene expression showed significant circadian rhythms in DEX-treated but not PBS-treated Scrambled shRNA tumors (compare Fig. 3 and Fig. 5c–g). In Bmal1 shRNA tumors, Bmal1 expression was reduced by ~70% in both DEX- and PBS-treated tumors (Fig. 5c). Accordingly, the expression of BMAL1 target genes Per1, Per2, Cry1 and Nr1d1 was suppressed and arrhythmic, indicating effective disruption of the clock machinery in the tumors (Fig. 5d–g). Thus, Bmal1 knockdown completely prevented the induction of circadian rhythms by DEX in the Bmal1 shRNA tumors.
Rhythmic gene expression of Cyclin E, p21 and c-Myc was only found in DEX-treated Scrambled shRNA tumors; Cdk1, Cdk2 and Wee1 did not show a daily variation, in line with cell cycle protein data (compare Fig. 5h–m and Fig. 4a–f). Importantly, abrogation of DEX-induced activation of the circadian clock in Bmal1 shRNA tumors was reflected by the arrhythmic expression of cell cycle genes in these tumors. Consistently, Bmal1 knockdown in vitro also prevented the DEX-induced rhythms in cell cycle phases (Additional file 8A–E).
Knockdown of Bmal1 prevents the inhibitory effect of dexamethasone on tumor growth
Next we sought to determine whether clock function in the B16 tumors is needed for DEX-induced reduction in tumor growth. As expected, B16 tumors expressing the Scrambled shRNA grew more slowly during DEX treatment: after 7 days, DEX-treated tumors were ~60% smaller than PBS-treated tumors (Fig. 5n) and this is consistent with the growth of untransfected tumors (Fig. 4m, n). In contrast, tumor volumes were indistinguishable between DEX- and PBS-treated mice harboring Bmal1 shRNA tumors (Fig. 5n). Thus, DEX had no effect on tumor growth after Bmal1 knockdown. Volume doubling time analysis of Scrambled and Bmal1 shRNA tumors confirmed these results (Scrambled shRNA: PBS: 53.3 h, DEX: 84.0 h; Bmal1 shRNA: PBS: 50.1 h, DEX: 50.0 h). Similar results were obtained in vitro using Bmal1 and Scrambled shRNA-transfected cells (data not shown).
Consistent with the data obtained using NSG mice (Fig. 4n), DEX did not affect the levels of immune cell infiltration in B16 tumors inoculated in the syngeneic C57BL/6J mice or in Bmal1 knockdown tumors, indicated by similar values for B cells, dendritic cells, CD4+ and CD8+ T cells, macrophages, neutrophils, and monocytes (Additional file 8F–L). These data confirmed that the slower tumor growth was not caused by effects of DEX on immune cell infiltration. Overall, these findings are in strong support of the notion that activating circadian clock function within B16 tumors slows down their growth.
Enhancement of circadian rhythms slows down HCT-116 tumor growth
To test the link between the circadian clock and tumor growth beyond our model using mouse melanoma, we treated human HCT-116 colon carcinoma cells with DEX and measured clock gene expression as well as proliferation and apoptosis up to 48 h after the treatment. HCT-116 cells exhibited rhythmic clock gene expression upon DEX treatment (Fig. 6a, b), which could be due to either synchronization of individual cells’ clocks or de novo activation of clocks in the cells. Upon this treatment HCT-116 cell proliferation was strongly decreased while apoptosis levels were unaffected (Fig. 6c, d).
Similar to the B16 s.c. tumors in C57BL/6 J and NSG mice, growth of tumors formed in NSG mice after HCT-116 cell inoculation was strongly slowed down after DEX treatment (Fig. 6e). These data underscore the possibility that controlling cell division rate by circadian clock enhancement and thus controlling tumor growth is not restricted to mouse melanoma cells, and may be generally applicable to human cancer cells and to other cancer types.