LLC cells readily produced solid tumors when injected subcutaneously into mPer1::luc mice, as described previously.18 Over 95% of the mice had palpable tumors at the time of imaging (21 of 22 mice). Nevertheless, not all of these mice showed detectable luminescence from the area of the tumor mass. Some light was likely blocked by the dark fur, filtered by skin pigments, or scattered by the skin to an extent that limited easy detection. Even in the anesthetized mice the shaved tumor areas did not always reveal bright luminescence from the tumor stroma, suggesting that the skin is also an impediment to imaging. Although imaging conditions were not ideal, significant gene expression in the tumor stroma could be monitored reproducibly under three different conditions—euthanized, anesthetized, or shaved and anesthetized mice (Fig. 1). One advantage of stromal imaging is that it may be effective with a wide range of syngeneic cancer lines without needing to transfect the cancer cells with a fluorescent or bioluminescent reporter gene. In addition, the same host mice could be used to compare the effects of different cancer cells on the stromal response to these cells.
Without direct imaging of the tumor, it seemed possible that distension of the skin by the tumor may have made the bioluminescence of the tissues more visible than in the corresponding region of the opposite leg. To verify that the signal did indeed originate in the tumor stroma and not in the surrounding tissue, which produces a faint background signal, the tumor was exposed and imaged in euthanized mice. This procedure revealed significant bioluminescence in all tumors examined in this manner. In every case the tumor stroma was brighter than the corresponding area of the opposite leg, after removing skin from this area as well. The center of the tumor mass was often darker than the surrounding cells, because LLC cells cannot express luciferase. Considering that there was little necrosis in the tumor core, the best interpretation seems to be that the lack of signal near the center of the tumor was due to an absence of stromal cells. Finally, tumors removed from the mice showed a bioluminescence throughout much of the stroma, indicating that the light was not generated only in the outer edges of the tumor or in a few distinct structures such as blood vessels.
Source of the stromal bioluminescence
It was in some ways surprising that the exposed tumors did not show visibly luminescent blood vessels. It is possible that microvessels were present but these were below the resolution limit of the bioluminescence imaging system. Circadian rhythms in rPer1 expression have been recorded in explant cultures made from several veins and arteries along with the atrium and ventricle,27 in which case ex vivo bioluminescence recordings were made from tissues harvested from transgenic rats expressing the firefly luciferase gene regulated by the rPer1 promoter. The optimal scheduling of antiangiogenic treatments is an important factor in the control of LLC tumors in mice to avoid drug resistance.15 The specific role of core clock genes such as Per1 in the cardiovascular system is not known, and higher resolution imaging of tumors might reveal circadian rhythms in these structures.
Tumor histology was used to help characterize the distribution of mPer1::luc expression in the stroma. Cells scattered throughout the population of LLC cells likely were immune cells that had migrated into the tumor. Some neutrophils were also present near the perimeter. Macrophages have been identified in LLC tumors and these appear to alter tumor growth through neovascularization.28 Tumor-activated macrophages (TAMs) and tumor-infiltrating lymphocytes (TILs) are known to facilitate growth of other tumors as well.23 These cells may not be adequate for generating the entire stromal signal because of their scarcity in the tumors examined here (Fig. 4f). Nearly all of the cells were closely packed LLC cells, as described previously by others.19
LLC cells and a diverse set of stromal cells showed mPER1 and mPER2 immunoreactivity (Fig. 4). Therefore, the mPer1 expression assayed in tumor stroma by bioluminescence imaging represents gene activity in several different cell types. Additional imaging and cell phenotyping would be needed to determine the relative contribution of each of these host-derived components. The capsule surrounding the tumor was also positive for mPER1 and mPER2 suggesting that it is another source of bioluminescence.
Although the immunohistochemistry suggests that two of the approximately seven core clock genes29 are present in the stroma and LLC cells, molecular approaches targeting clock genes and clock-driven genes are needed. We are confident that the two commercial antibodies used in this study provided good localization of mPER proteins based on the observed mPER1-like immunostaining of neurons in the SCN of the hypothalamus, a positive control (Fig. 4),30, 31 and the many published reports using the affinity-purified anti-mPER2 antibody that was applied here.32–37
The appearance of the capsule tissue was suggestive of muscle-like cells. Many tumors contain cells with myofilaments that are evident at the ultrastructural level.38 Circadian clock genes are expressed in muscle cells,39, 40 and it appears that the abnormal muscle-like cells of the tumor capsule also express these genes. The immunolabeling results indicating clock gene expression in LLC cells and stromal cells suggest that the tumor may contain one or more circadian oscillators, although mPER1 and mPER2 expression alone is not sufficient evidence of a biological clock. Nevertheless, circadian rhythms have been described in other tumors,3 and these could be generated by clock cells within the tumor rather than driven by a circadian oscillator located elsewhere in the body. We were not able to image the mice repeatedly to identify circadian rhythms in bioluminescence in the stroma because of the limitations of our imaging methods and the rapid expansion of the tumors during the last two weeks of growth. Alternatively, methods in which bioluminescence imaging is performed continuously with freely-moving mice (without anesthesia) might identify rhythms in the stroma. Similarly, molecular approaches in which stromal tissue is harvested from many mice at time points across the circadian cycle could be effective as well.
Tumor cell communication with other body regions
An additional question addressed by this study is whether tumor growth alters mPer1::luc expression in areas well outside the tumor region such as the paws, tail, ears, and snout. All of these areas showed spontaneous mPer1::luc expression in all normal and LLC-injected mice. If we had observed induction or suppression of peripheral gene expression by the tumor, this would suggest that growth of the LLC tumor can have a systemic impact on mPer1 and therefore an influence on the circadian oscillators in peripheral tissues and, by extrapolation, the clock cells of the major circadian pacemaker in the SCN of the hypothalamus. One reasonable set of candidate mediators of this effect are the pro-inflammatory cytokines that are known to be elevated in the bloodstream as tumors grow.41 Cytokines can shift the phase of circadian rhythms,42 suggesting that they could act in signaling between clock cells within the tumor, by inducing clock genes, or they could act between the tumor and oscillators in the body.
We did not, however, find any evidence of communication between the tumor and other host cells located farther from the tumor site than the stroma. The mPer1::luc expression in the rest of the animal was not affected in a way that could be detected by bioluminescence. Similarly, the behavioral assay used here—the circadian rhythm in locomotor activity—did not show an altered free-running period. We did observe a decline in the average free-running period of both the tumor and control groups over time. A similar decrease in period during constant darkness has been described in other circadian studies using the same mouse species.43 The wheel-running rhythm primarily reflects the circadian rhythms of the SCN and is sensitive to subtle changes in the intensity of light, food availability, reproductive cycles, aging, and disruptions of metabolic state. For example, Vipr2−/− mice, which lack a key neuropeptide receptor in the SCN, show concomitant disruptions of both circadian wheel-running rhythms and circadian rhythms in SCN gene expression.44
Several studies have shown that the circadian system alters tumor growth,9, 45–51 and disrupted circadian rhythms in DNA synthesis have been described in mice carrying LLC tumors.52 The SCN interacts with other circadian oscillators of the body, thereby altering their timing, and it is conceivable that tumors might also communicate with the circadian timing system as circadian oscillators. Our results suggest, however, that any interactions between the circadian system and these tumors are mostly unidirectional if at all. Therefore, effects of the circadian system on tumor growth might be studied using the LLC tumor model without confounding effects of the tumor acting on the body's circadian clocks.
It seems most likely that the observed elevated expression of the mPer1::luc transgene is associated with the accelerated growth of the stroma relative to surrounding tissues. We observed that rapid tumor expansion occurs mostly during the last two weeks of the three weeks of tumor growth. Cell growth is associated with cell signaling pathways that can also induce mPer1 expression, e.g., ones acting through cAMP response element-binding protein (CREB).53–58 The timing ability of circadian oscillators is disrupted when Per1 expression is elevated experimentally, in which case the circadian period lengthens and the animal is less able to entrain to light cycles.59 If the elevated mPer1 expression we detected in the stroma was due to the effects of the LLC cells, then tumor growth might have disrupted the circadian clock gene functions in these cells.
Alternatively, the stromal signal may be generated in a circadian rhythm with a peak occurring during the night like many peripheral circadian oscillators (those located outside the SCN). Our preliminary experiment using a single mouse yielded a signal more than four standard deviations above the average stromal signal recorded during the day and agrees with this hypothesis. Any future stromal imaging experiments planned with these mice should take into consideration the extra requirements for imaging at night, including maintaining the mice in darkness up to the time of imaging to avoid phase shifts, and whether this effort is offset by the possibly higher signal available at this phase. What is most important for cancer studies examining the intensity of the stromal bioluminescence is to perform imaging as close to the same time of day as possible, thereby minimizing effects from circadian or daily oscillations. The phases used here for most imaging, during the second half of the daytime, may be optimal for identifying both increases and decreases in the day-to-day signal, perhaps in response to an anti-tumor treatment.
It will remain unresolved whether mPer1 expression is circadian in the stroma until a more extensive study can be performed examining mice, ideally during at least six phases of their circadian activity rhythms displayed while in constant darkness. As an alternative possibility, it should also be considered that increased mPER1 protein levels from persistent induction might disrupt the circadian oscillation. Evidence of this effect has been identified in transgenic rats.59 A key event in the circadian cycle is suppression of the period genes by their own protein products, followed by a decline in PER protein levels through degradation. If signals from the tumor act as strong inducers of mPer1, then the gene might remain active independent of circadian modulation. If mPER1 protein remains elevated in the tumor, then it should be asked whether it can serve a non-circadian role in this particular location. Interestingly, evidence indicates that mPER1 protein remains cytoplasmic in a subset of retinal cells in mice,60 and without entering the nucleus it is difficult to argue that the protein serves in the circadian timing mechanism. Although we observed nuclear mPER1-like immunostaining in the LLC cells of tumors, nuclear staining in the many stromal cells was not confirmed. Further studies are needed to determine specifically whether mPer1 plays a circadian or non-circadian role in the stroma and tumor growth.
Overall, the importance of circadian effects on cancer development or growth is not well understood. Growth of tumors from LLC cell injections in mice is inhibited when they are made to over express mPER2.4 The circadian system regulates inflammatory cytokines, such as interferon-gamma,61 and may also control natural killer cell responses.62 It is interesting to consider how the pattern of mPer1 gene expression described here could be part of an ongoing defense mechanism between the host and cancer cells, particularly if this process cycles throughout the day. Circadian clock gene expression might play an important role in these tumor-infiltrating immune cells. Additional exploration of circadian timing within the stroma is needed.