Real-time Evaluation of p53 Oscillatory Behavior In vivo Using Bioluminescent Imaging

p53 is a prototypical tumor suppressor gene, mutated in nearly 50% of human cancers (1, 2), with a key role in cellular pathways, including regulation of the cell cycle and/or apoptosis following DNA damage (1–4). A central aspect of p53 function is its ability to act as a sequence-specific transcription factor modulating the expression of dozens of target genes (3). A prototypic p53 consensus binding sequence has been identified, and p53-responsive genes contain sequences that often vary from this prototype by as little as a few nucleotides (5). The regulation of p53 is itself complex being expressed at low levels at baseline with a large number of known phosphorylation, acetylation, and ubiqitination sites, which alter p53 stability and function (3).

The best studied activator of p53 is DNA damage secondary to UV radiation, ionizing radiation, or chemical toxins. A series of “sensor” and “effector” proteins are likely involved in the detection of such DNA damage events and the induction of p53 function. There are also several additional proteins that regulate p53 stability and function, keeping this p53-dependent DNA damage pathway tightly in check. The most notable of these p53 regulators is the murine double minute 2 (MDM2) protein, which acts as a p53-specific E3 ubiquitin ligase (6, 7) leading to proteasome-mediated p53 degradation. The interaction between p53 and MDM2 is rendered even more complex by the fact that, in addition to facilitating p53 degradation, MDM2 is also a prototypical p53-dependent transcript (8). The internal P2 promoter has little activity at baseline but is potently up-regulated by p53 on its activation (8). Through a combination of transcriptional and post-translational events, a feedback loop is established where p53 activation leads to an increase in MDM2 expression, which in turn destabilizes p53 leading to its degradation and the release of the p53-mediated MDM2 transcriptional up-regulation (see Fig. 1A).

The transcriptional dependence of p53 target genes can be extraordinarily complex with several different theories postulated on how these p53-dependent genes are differentially modulated (3). Although experiments in cultured cells or purified systems are starting to identify key differences in the regulatory features of p53-dependent transcripts, the process is likely even more complex in whole organisms (5, 9–12). The ability of p53 to modulate the expression of target genes following DNA damage has generally been studied using invasive biochemical techniques (10, 13, 14). In addition, transgenic mice expressing a p53-dependent β-lactamase gene either under the MDM2 P2 promoter (15) or a synthetic p53-dependent promoter (16) showed distinct tissue and embryologic differences in p53 gene expression. However, these studies were limited by their dependence on fixed and stained tissues to ascertain differences in p53 function at a single point following the induction of DNA damage. As such, these model systems were not able to dynamically assess the temporal range of p53 activation in vivo in a living organism. Noninvasive bioluminescent techniques have been used to follow p53 function in stable cell lines (17) or following adenoviral infection in tumor xenografts (18). Given the importance of p53 in both the maintenance of DNA stability within the organism and in its response to DNA-damaging events, we developed a transgenic mouse model wherein the firefly luciferase gene was placed under the transcriptional control of the MDM2 P2 promoter to noninvasively monitor p53 function in whole animals (see Fig. 1B).

Transgenic animals. All animals were cared for using protocols approved by the University Committee for the Use and Care of Animals. The shuttle plasmid containing the MDM2 P2 promoter driving β-lactamase expression was obtained from Moshe Oren (Weizmann Institute of Science, Rehovot, Israel; ref. 15), and the ApaI/NsiI fragment containing the promoter sequence (see Fig. 1B) was cloned in front of the cDNA for firefly luciferase in the pGL3-Basic plasmid (Promega, Madison, WI). This expression cassette was then used to generate transgenic animals in the Transgenic Animal Core at the University of Michigan Medical Center (Ann Arbor, MI) on a BALB/C background. Transgenic luciferase expression was characterized using standard PCR techniques with primers specific for the luciferase transgene (5′ luciferase, GTGCCAACCCTATTCTCCT and 3′ luciferase, TGTTCGTCTTCGTCCCAGTAAG). A total of seven distinct founder transgenic lines were obtained, and a single founder was selected, which showed transgene expression in all tissues studied and was then backcrossed to achieve a homozygous population. All experiments used animals from the F3 or later generation and were between ages 6 to 12 weeks at the time of experiments. p53 heterozygous mice were provided by Bart Williams, PhD. (Van Andel Research Institute, Grand Rapids, MI; ref. 19). This is a strain of mice that contains a p53 allele that deletes 40% of the p53 coding region and completely abrogates p53 protein expression from this allele. p53 heterozygous mice were crossed with the MDM2-luciferase homozygous mice to generate mice heterozygous for both MDM2-luciferase transgene and for p53. These were in turn backcrossed to generate the representative p53 genotypes, homozygous wild-type (WT; +/+), heterozygous (+/−), and homozygous null (−/−).

Animal treatments. For total body irradiation (TBI), animals were immobilized in a clear plastic restraint with irradiations carried out using 300 KeV photons in the Experimental Irradiation Core of the University of Michigan Cancer Center using a Pantak Therapax DXT 300 Model X-ray unit (Pantak, East Haven, CT) at a dose rate of ∼3 Gy/min. Dosimetry was done using an ionization chamber connected to an electrometer system, which is directly traceable to a National Institute of Standards and Technology calibration.

Bioluminescent imaging. For bioluminescence imaging, mice were anesthetized with a 2% isofluorane/air mixture and given a single i.p. dose of 150 mg/kg D-luciferin in normal saline. Bioluminescent imaging (BLI) was initiated 10 minutes after luciferin administration using a Xenogen Ivis Imaging Series 100 system (Xenogen, Alameda, CA). During image acquisition, isofluorane anesthesia was maintained by using a nose cone delivery system, and animal body temperature was regulated with a temperature-controlled stage. A gray-scale body image was collected (field of view, B; exposure, 0.2 seconds; binning, small (high resolution); and f/stop, 16) followed by acquisition and overlay of a pseudocolor image representing the spatial distribution of the detected photons emitted from the animal (field of view, B; exposure, 300 seconds; binning, large (high sensitivity); and f/stop, 1) with all luminescent data recorded as photons per minute. Signal intensity was quantified as the sum of all detected photon counts within a region of interest (ROI) prescribed over the entire animal or over individual uniform-sized ROIs that were manually placed during postdata acquisition image analysis with minimum and maximum thresholds for the pseudocolor scale set at 2,000 photons/min and 10,000 photons/min for all experiments.

For analysis of organ-specific differences in bioluminescence, animals were radiated and then sacrificed at the indicated time after radiation by cervical dislocation. Ten minutes before euthanasia, the animals were injected with D-luciferin; after sacrifice, animals were quickly dissected, and individual organs were assessed by BLI. A piece of small intestine (∼2 cm) just distal to the pylorus was then resected with half placed in RNeasy (Qiagen, Valencia, CA), and the other half was snap frozen on dry ice for later analysis.

Western blotting. For protein analysis, a 50 mg piece of intestine from each condition was lysed in 500 μL CellLytic-MT Mammalian Tissue Lysis/Extraction Reagent (Sigma-Aldrich, St. Louis, MO). Each sample was then homogenized on ice using a handheld tissue tearer and centrifuged at 14,000 rpm for 10 minutes to remove the remaining cellular debris. Protein samples were quantified and normalized using the Bio-Rad DC Protein Assay (Bio-Rad Laboratories, Hercules, CA). The samples were fractionated by 10% SDS-PAGE and electroblotted to polyvinylidene difluoride membranes. After blocking, the membranes were initially probed for p53 (Ab6; CalBiochem, San Diego, CA) or Ser¹⁸-phosphorylated p53 (16G8; Cell Signaling Technology, Inc., Danvers, MA). Blots were then probed with secondary antibodies conjugated with horseradish peroxidase (Jackson ImmunoResearch, West Grove, PA) and visualized using enhanced chemiluminescence (Pierce Biotechnology, Inc., Rockford, IL). They were then stripped and reprobed for actin (A2066, Sigma-Aldrich).

Quantitative PCR. RNA was isolated from the tissue samples using the RNeasy kit, and subsequent cDNA was made using iScript cDNA synthesis kit (Bio-Rad Laboratories). Quantitative PCR was done using the Quantitech SYBR Green PCR kit (Qiagen) according to the manual. Thirty-five cycles were measured on the Opticon Real-time PCR Detection System (MJ Research/Bio-Rad). The following primers used were β-actin 5′ AGAGGGAAATCGTGCGTGAC and β-actin 3′ CAATAGTGATGACCTGGCCGT; MDM2 5′ GCGCTTCCTAGGTCACTTTG and MDM2 3′ CCAGCAACTGTTCAGACGAA; and p21 5′ CCTGACAGATTTCTATCACTCCA and p21 3′ GCAGGCAGCGTATATCAGGAG.

Statistics. All statistical analysis was done using MEDCALC V8.0 (MedCalc Software, Mariakerke, Belgium). Differences between groups with continuous variables were done using Student's t test (two-tailed, unequal variance) with a P < 0.05 considered statistically significant.

In vivo bioluminescence is increased in a dose- and time-dependent manner following DNA damage. Transgenic animals expressing the luciferase cDNA under the control of the MDM2 P2 promoter (see Fig. 1B) were generated (see Materials and Methods). The MDM2-luciferase mice were evaluated by BLI (Fig. 2A), and before treatment, there was low baseline total body bioluminescence (3.7e⁵ ± 0.2e⁵ photons/min) consistent with low basal expression of the reporter gene (20, 21). As early as 30 minutes after TBI or 20 hours after treatment with doxorubicin (data not shown), there was an increase in total body bioluminescence reporter signal (12.2e⁵ ± 0.6e⁵ photons/min following 10 Gy TBI or 6.4e⁵ ± 0.7e⁵ photons/min following 14 mg/kg i.p. doxorubicin). Given the near instantaneous nature of DNA damage following ionizing radiation compared with the complicated pharmacokinetics after systemically administered chemotherapy, we chose to use ionizing radiation for subsequent experiments.

The increase in bioluminescence after TBI was observed across all regions of the animal but was greatest for ROI placed over the abdomen, oral cavity, and paws. This regional increase in BLI is consistent with the documented increase in the expression of both p53 and its downstream targets after ionizing radiation in the intestinal and oral mucosa (10–12). Peak BLI was dose dependent for doses from 2 to 10 Gy (Fig. 2A-C) and was consistently observed 30 minutes following radiation with the signal rapidly decaying toward baseline after this peak (Fig. 2B). The BLI from a ROI centered over the abdomen was induced 12.8 ± 1.3–fold (average ± SE) for 10 Gy TBI down to 1.8 ± 0.5–fold at 2 Gy, which were all measurably above that for unirradiated mice (P < 0.05 for comparison of each peak to sham-irradiated animals). This dose-dependent increase in BLI exhibited a clear exponential function (Fig. 2C) for both total body BLI and the abdomen-specific ROIs for the doses studied from 0 to 10 Gy with a steeper exponential dose response for the abdominal induction.

Bioluminescent induction is dependent on p53 genotype. Having ascertained the dose-dependent nature of the BLI induction following ionizing radiation, we next verified the p53 dependence by crossing the MDM2-luciferase mice with p53 heterozygous animals (19). The three different p53 genotypes, homozygous WT (+/+), heterozygous (+/−), and homozygous null (−/−), showed baseline BLIs before any DNA-damaging event that were slightly different [total body BLI, 3.2e⁵ ± 0.3e⁵ photons/min for p53 (−/−), 3.4e⁵ ± 0.4e⁵ photons/min for p53 (+/−), and 3.8e⁵ ± 0.3e⁵ for p53 (+/+)], but this did not achieve statistical significance (P > 0.05 for heterozygous or homozygous null when compared with p53 WT; see Fig. 3B and C). Following 10 Gy TBI p53 homozygous, WT mice exhibited an increase in BLI (10.8 ± 1.4–fold for an abdominal ROI), whereas p53 homozygous null mice had no appreciable increase (1.2 ± 0.8–fold). Confirming the results observed in β-lactamase-dependent p53 reporter mice (15), there was haploinsufficiency for p53 with only a 2.6 ± 1.2–fold increase in the abdominal BLI following 10 Gy TBI in p53 heterozygous mice. Thus, the loss of one allele of p53 resulted in a 76% decrease in the p53-dependent BLI response to 10 Gy TBI.

Correlation of bioluminescent induction with p53 expression and function. Given the dose, time, and p53 dependence of the MDM-luciferase mouse model, we correlated BLI induction with both p53 expression and function. Thirty-minutes after irradiation, the BLI induction was greatest in the intestine (11.3 ± 1.2–fold) and spleen (5.1 ± 1.1–fold) with no significant increase observed from heart (0.85 ± 0.05–fold) or kidney (1.3 ± 0.3–fold), whereas the liver showed only a modest increase in p53-dependent BLI at this time point (2.2 ± 0.8–fold; Fig. 4A). A similar trend was observed in all organs at 4 hours (Fig. 4A); however, there was a decrease in all BLI signals at this time relative to the 30-minute time point.

To evaluate the correlation of BLI induction with expression and function of p53, animals were either sham irradiated or treated with 10 Gy TBI. A clear temporal pattern of p53-dependent BLI induction was again seen following TBI with the greatest induction over the abdominal region, whereas there was no significant increase in BLI for sham-irradiated mice (Fig. 4B). Western blots for both p53 and the Ser¹⁸-phosphorylated form of p53 in the small intestine showed an increase in both total p53 and activated p53, respectively, 30 minutes after TBI (Fig. 4C). Following the initial peak, both total and activated p53 declined at the later time points such that they were below the baseline expression 4 hours after radiation. There were no appreciable changes in total p53 or Ser¹⁸-phosphorylated p53 expression in sham-irradiated animals (data not shown). Finally, when the expression levels of two different p53-dependent transcripts were monitored by quantitative PCR (Fig. 4D), there was a similar temporal pattern of induction of the cDNA for MDM2 with a peak at 30 minutes (7.1 ± 1.1–fold), which approached baseline 4 hours after radiation (1.7 ± 0.5–fold). The pattern of induction of the endogenous MDM2 gene was analogous to the induction seen for our BLI reporter, which were both transcribed from the MDM2 P2 promoter. In contrast, the transcript for p21 showed a somewhat lesser induction (3.1 ± 0.8–fold) at 30 minutes, which similarly approached that of preradiation levels by 4 hours (0.8 ± 0.2). Thus, the expression of both p53 and activated p53 in addition to its transcriptional function all seem to correlate well with the BLI induction documented following TBI.

Observation of the oscillatory behavior of p53 in vivo following radiation. Given the feedback loop between p53 and MDM2 (see Fig. 1A), it was previously observed in vitro that following radiation p53 and MDM2 expression increased and decreased in a series of interdigitated oscillations (22, 23). This process has never been evaluated quantitatively and temporally in live animals; therefore, the MDM2-luciferase mice were irradiated with 5 Gy TBI or sham treatment and followed with close temporal BLI. A distinct oscillatory pattern was observed in radiated mice that was not present in unirradiated controls (Fig. 5A and B). The greatest increase in BLI occurred over the abdominal cavity 30 minutes following ionizing radiation (9.3-fold induction) with the first nadir at 4 hours (2.5-fold induction) followed by a second peak at 6 hours (3.7-fold induction; Fig. 5A and B). Second (2.2-fold) and third (2.6-fold) nadirs were also observed 8 and 12 hours following radiation with a third peak (4.4-fold) between these at 11 hours. In the control animal, there was no significant induction of p53-dependent bioluminescence with values ranging between 0.8- to 2.1-fold of baseline that were all consistently below that observed in the irradiated animal (Fig. 5B).

When quantified for the abdominal ROI in a series of animals, this oscillatory behavior was remarkably consistent (Fig. 5C) with the first two peaks observed at 30 minutes and 6 hours in all instances, whereas the third peak diverged in time (range, 9-13 hours; average 11.2 ± 0.7 hours). The nadirs followed a similar temporal fashion with the first nadir at 2.5 ± 0.5 hours (range, 1-3 hours) and the second nadir at 7.2 ± 0.2 hours (range, 7-8 hours) from the time of irradiation. Overall, peaks occurred with a wavelength of 5.3 ± 0.7 hours. The amplitude of the peaks followed a pattern of dampened oscillations with an induction of 6.8 ± 1.4–fold, 3.5 ± 0.2–fold, and 2.7 ± 0.3–fold at the first, second, and third peaks, respectively. The subsequent nadir values also approached closer to the baseline at each successive value with the first and second nadirs of 2.4 ± 0.5–fold and 1.6 ± 0.2–fold, respectively, which were on average 40.0 ± 5.8% as large as their preceding peaks. No similar oscillatory behavior was observed in control animals with the largest peak observed 1 hour after radiation of 1.6 ± 0.3–fold and the lowest overall value of 0.9 ± 0.2–fold of baseline (see Fig. 5C). There was no apparent difference in this oscillatory pattern for animals radiated early in the day [between 6:00-10:00 a.m. eastern standard time (EST)] or those radiated later (2:00-5:00 p.m. EST; data not shown), which is support for the absence of a circadian behavior of p53 expression as documented previously in murine tissues (24).

Alteration of the p53-dependent bioluminescent response with radiation modulators. One of the most commonly used radiation sensitizers is 5-fluorouracil (FU; ref. 25) where it has been shown that, by decreasing DNA synthesis, it inhibits repair and potentiates the DNA double-strand breaks induced by ionizing radiation (26). Pretreatment of mice with FU 1 hour before radiation (at a dose that did not induce a BLI response alone) resulted in no significant change in the first peak following 5 Gy TBI (4.7 ± 1.0–fold for X-ray therapy (XRT) versus 4.2 ± 0.7–fold for FU/XRT; P > 0.05; Fig. 6A). However, in animals pretreated with FU and radiation following the initial peak, all subsequent values were elevated compared with the XRT alone such that, after the second peak, they were statistically significantly elevated (P < 0.04 comparing XRT to FU/XRT from 7 to 13 hours after radiation) by 1.8 ± 0.1–fold (range, 1.3-2.1). Therefore, although pretreatment with FU did not seem to alter the initial peak induction of p53 following ionizing radiation, it did significantly enhance the overall transcriptional function over time consistent with the known radiation enhancing role of FU.

Pretreatment of MDM2-luciferase mice with amifostine, a thioyl-containing prodrug, which has been approved for protection of salivary glands during radiotherapy due to its ability to act as a free radical scavenger (27), 30 minutes before radiation attenuated the BLI induction following 5 Gy TBI with no significant change in the distribution of this induction (Fig. 6B). This amifostine-dependent suppression of the p53 BLI response following TBI manifested itself as a persistent decrease in the BLI response throughout the period evaluated (P < 0.001; Fig. 6B and C). In addition, confirming previous reports, there seemed to be a somewhat greater protection of the abdominal region (55% decrease in BLI) compared with the total animal (25% decrease in BLI; ref. 28).

p53 is a key player in the maintenance of cellular homeostasis and the response to a diverse array of physiologic insults (3); however, despite its discovery >20 years ago, the exact scope and mechanism of its action are still being actively investigated. Part of the difficulty in understanding p53 function is that it is different both within different tissues in the body and stages of development (11, 12, 15, 16). Previous studies on β-lactamase-dependent p53 transgenic mice showed tissue-specific roles for p53 during embryonic development (15, 16); however, the dependence on fixed and stained tissues greatly limited the ability of these models to dynamically follow p53 function. In addition, the presence of endogenous β-lactamase activity in adult murine tissues (such as intestine and salivary gland) rendered this technique insensitive in adult animals. The MDM2-luciferase mouse model presented here was similarly able to document tissue dependent differences in p53-dependent bioluminescent activity, which were consistent with observed tissue dependence of p53 activation following ionizing radiation (4, 11, 29). However, given the magnitude of the induction within the intestinal tract and the large mass of intestine within the abdominal cavity, we chose to focus on the BLI-measured role of p53 in this organ.

The role of p53-mediated apoptosis of intestinal crypt cells in acute radiation toxicity has been shown (10, 11), where it was observed that, within 4 hours after radiation, intestinal crypt cells undergo cell cycle arrest and inhibition of bromodeoxyuridine incorporation and begin showing morphologic features consistent with apoptosis, phenotypes that were blocked in p53-null mice (10). Mice exposed to TBI of doses <12 Gy typically die of bone marrow depletion secondary to p53-dependent apoptosis, whereas those treated to higher doses of radiation succumb to gastrointestinal toxicity (30). Attenuation of p53 function with small-molecule inhibitors or in knockout mice greatly ameliorated this bone marrow–dependent death; in contrast, in the intestine, a lack of p53 interfered with cell cycle arrest and repair processes, and as a result, after high-dose TBI, there was near complete loss of the intestinal epithelium in p53-null mice, which subsequently underwent a more rapid death compared with their p53 WT littermates (30). In contrast to p53-null mice, pharmacologic inhibition of p53 using daily administration of a small-molecule inhibitor of p53 did not cause increased gastrointestinal toxicity, potentially suggesting a clinical role for p53 inhibitors (30). The mechanism by which p53 and p53 inhibitors differentiate between these cellular pathways is still under investigation (4, 11, 12).

One potential mechanism to explain these tissue and temporal differences in p53 function has come recently to light where it was determined that p53 activation and inactivation involves a complex interplay between the p53 protein and at least one of its target genes, MDM2 (6, 22, 23). Because MDM2 is up-regulated by p53 and in turn leads to degradation of p53, a negative feedback loop is present between these two proteins (Fig. 1A). Several mathematical models have postulated that activation of p53 will result in an oscillatory pattern between p53 and MDM2 (22, 31–34), and this phenomenon was shown in cells in culture following ionizing radiation both by Western blotting of p53 and MDM2 (22, 23) and using fluorescent fusion proteins (23). Interestingly, Western blots for p53 done on a population of cells revealed dampened oscillations analogous to what we observed in the MDM2-luciferase mice (22, 23). In contrast, in single-cell experiments, although the population as a whole exhibited dampened oscillations on a cell-by-cell basis, p53 was up-regulated in a pulsatile manner with each pulse of similar amplitude and wavelength independent of radiation dose. With increasing dose of radiation, the percentage of cells showing p53-dependent pulses increased as did the number of these pulses per cell such that the summation across the population resulted in the observed dampened oscillations. The mechanisms underlying this pulsatile behavior and its physiologic significance have yet to be determined, but it has been suggested that, by activating p53 in discrete quanta for a limited period of time, it might allow DNA repair enzymes the capacity to respond to DNA damage before the cells are inextricably sent down an apoptotic pathway (22, 23, 31, 34). This is consistent with the data presented here that pretreatment with FU did not alter the initial BLI response to ionizing radiation but caused a persistent activation corresponding to an interference with the physiologic DNA repair (26). In addition, the MDM2-luciferase mice were also able to show quantifiable differences in p53-dependent BLI following treatment with amifostine, a known free-radical scavenger, which was shown previously to decrease the incidence of DNA damage following ionizing radiation (28). Therefore, the MDM-luciferase mice might be useable to study the kinetics of the p53 response following the systemic administration of agents that modulate the response to ionizing radiation, which may be even more important when investigating the role of p53, MDM2, or proteasome inhibitors, which are all currently of significant clinical interest (30, 35, 36).

At present, the ability to monitor distinct areas within an animal using two-dimensional BLI limits the ability to accurately depict the site or tissue of origin. For instance, at later time points after radiation, we observed an increase in BLI from the thorax of irradiated animals, and it is possible that this represents a different temporal pattern of p53 response in these tissues, or, alternatively, this may be a function of late cytokine release that has been postulated following radiation therapy (37). However, newer techniques based on differential attenuation of the produced photons from the animal (38) and the development of tomographic reconstruction of these optical images (39) might increase the use of this or other transciptionally based transgenic mouse systems to elucidate physiologic processes in real time. The current system is also constrained by the half-life of luciferase (which has been reported as between 30 minutes to 2 hours in vivo; refs. 40–43) such that more rapid sampling of data was not possible. However, the generation of luciferase molecules with markedly attenuated half-life might make rapid sampling feasible to more carefully evaluate cellular physiology (44). In addition, a recent report showed that the administration of a slow infusion of luciferin allowed near continuous sampling of BLI data, although at the expense of a significant reduction in overall sensitivity (45). Thus, the MDM2-luciferase model points to the use of transcriptionally based mouse reporters to reveal underlying cellular physiology in a fashion that could not have been explored using traditional biochemical techniques and which may be further facilitated by advancing molecular imaging technology.

Grant support: NIH grants P01CA85878, P50CA01014, and R24CA83099, Varian Medical Systems/Radiological Society of North America (RSNA) Holman Pathway Research Resident Seed grant, and Philips Medical Systems/RSNA Holman Pathway Research Resident Seed grant (D.A. Hamstra).

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.