Cancer Research 67, 6247, July 1, 2007. doi: 10.1158/0008-5472.CAN-07-0342
© 2007 American Association for Cancer Research
Cell, Tumor, and Stem Cell Biology |
Quantitative Assessment of the p53-Mdm2 Feedback Loop Using Protein Lysate Microarrays
Sundhar Ramalingam1,
Peter Honkanen4,
Lynn Young3,
Tsutomu Shimura2,
John Austin4,
Patricia S. Steeg1,2 and
Satoshi Nishizuka1,5
1 Molecular Therapeutics Program, 2 Laboratory of Molecular Pharmacology, National Cancer Institute, and 3 Center for Information Technology, NIH, Bethesda, Maryland; 4 Aushon BioSystems, Burlington, Massachusetts; and 5 Laboratory of Proteomics and Analytical Technologies, Research Technology Program, Science Applications International Corporation-Frederick, Inc., National Cancer Institute at Frederick, Frederick, Maryland
Requests for reprints: Satoshi Nishizuka, Department of Surgery, Iwate Medical University School of Medicine, 19-1 Uchimaru, Morioka 020-8505, Japan. Phone: 81-19-651-5111, ext. 3627; Fax: 81-19-651-7166; E-mail: snishizu{at}iwate-med.ac.jp or nishizus{at}mail.nih.gov.
 |
Abstract
|
|---|
Mathematical simulations of the p53-Mdm2 feedback loop suggest that both proteins will exhibit impulsive expression characteristics in response to high cellular stress levels. However, little quantitative experimental evaluation has been done, particularly of the phosphorylated forms. To evaluate the mathematical models experimentally, we used lysate microarrays from an isogenic pair of 
-ray–irradiated cell lysates from HCT116 (p53+/+ and p53–/–). Both p53 and Mdm2 proteins showed expected pulses in the wild type, whereas no pulses were seen in the knockout. Based on experimental observations, we determined model parameters and generated an in silico "knockout," reflecting the experimental data, including phosphorylated proteins. [Cancer Res 2007;67(13):6247–52]
|
Introduction
|
|---|
Biological responses in cells are coordinated by signaling networks based on protein dynamics. A major goal in the elucidation of biological networks is to develop mathematical models of the dynamics of protein signaling pathways. However, mathematical modeling of such pathways is limited by the lack of an appropriate, high-throughput means to experimentally validate protein dynamics under different experimental conditions. For instance, the instantaneous state of a biological network will vary according to both the type of stimulus and the time of measurement. Therefore, acquiring quantitative information from different conditions or time points is extremely important for model development and validation. To achieve the monitoring of multidimensional (different conditions, times, and proteins) pathway dynamics, a study format allowing the simultaneous measurement of many proteins across many samples is ideal. "Reverse-phase" protein lysate microarrays (RPA) have been developed for quantitative proteomic monitoring in various biological contexts (1, 2). In principle, the RPA technique involves spotting whole-cell fraction lysates in a microarray format to detect particular proteins with specific primary antibodies, making it especially useful when testing many samples in a single experiment.
Using RPA technology, we captured the protein dynamics of the p53-Mdm2 feedback loop, one of the best-studied biological networks. An important tumor suppressor protein, p53, is mutated frequently in human cancers and is vital to cell mechanisms such as DNA repair, apoptosis, and cell cycle regulation. As a key transcriptional regulator, p53 binds to specific sequences in DNA and activates many downstream genes, including Mdm2. In turn, the Mdm2 protein inhibits the transcriptional activity of p53 and enhances its degradation. This negative feedback loop keeps the p53 protein level low yet stable under normal conditions and helps to switch off p53 at the end of a stress response. Accumulating knowledge of the p53-Mdm2 feedback mechanism has led investigators to explore mathematical methods to better understand this system (3–6). However, validating the plausibility of the mathematical models has not been a trivial task because of the many samples that are required from many time points and the difficulty in quantitative protein monitoring in a high-throughput fashion. In the present study, we produced a high-density RPA from a p53-knockout model cell line, HCT116, to validate mathematical models of the p53-Mdm2 feedback loop by quantitative protein monitoring.
|
Materials and Methods
|
|---|
Cell culture. An isogenic pair of HCT116 colon cancer cell lines was kindly provided by Dr. Bert Vogelstein (Johns Hopkins University, Baltimore, MD; ref. 7). Cells were grown at 37°C in the presence of 5% CO2 in McCoy's 5A medium containing 10% fetal bovine serum.
Collection of cell lysate. The isogenic pair of HCT116 with three biological replicates were grown and collected independently. Following exposure to 0.3, 3, and 30 Gy of -irradiation using the 137Cs Mark 1 irradiator (Shepard & Associates), protein lysates of p53+/+ and p53–/– cells at 10 time points over an 8-h period were collected with a modified cell lysis protocol (1).
Reverse-phase protein lysate microarrays. The Aushon 2470 microarrayer (Aushon BioSystems) was used to produce the RPAs for this study. The array surface was a nitrocellulose-coated glass slide for which manufacturing conditions were optimized for our RPA (Grace BioLabs). A set of nineteen 384-well microplates, each containing 10-time, 2-fold serial dilutions of each sample, was prepared. The robustness of the array design was maximized by including two technical replicates for each biological replicate as well as ensuring that no two biological replicates were spotted by the same pin.
Image processing. To acquire enough bit depth and pixels per feature, an optical flatbed scanner (Epson 4870, Epson America) with the resolution set to 16 bit, 2,400 dpi was used. To prevent the image intensity from being skewed, a Wedge Density Strip (Danes-Picta) was used for readout range calibration (8). The resulting images were processed by P-SCAN and Proteinscan software packages6 written in Matlab.
Immunocytochemistry. A half-million cells were plated onto a cover glass and placed in a 3.5-cm Petri dish to grow on the surface of both the cover glass and the Petri dish. After 16 to 24 h of incubation, the irradiated cells were processed for immunostaining with either mouse anti–p53 primary antibody (LabVision) or rabbit anti–p53-Ser15 antibody (Cell Signaling Technology), followed by fluorophore conjugated either with antimouse or with antirabbit immunoglobulin G secondary antibodies.
Mathematical models. The premise of our model comes from the systems developed by Lev Bar-Or et al. (5) and Ma et al. (6). MATLAB (MathWorks) language was used to program all simulations. The feedback loop circuits were divided into Oscillator and Stress modules. All equations and parameter settings are available in Supplementary data.
|
Results
|
|---|
Quantitative monitoring of p53 and Mdm2 proteins. Because the wild-type cells (p53+/+) have functional p53 activity, comparison with the knockout cells (p53–/–) provides useful p53-dependent information about cellular response to DNA damage. To examine the p53-mediated stress response, protein lysates from each cell line were collected at 10 time points during a period of 8 h following exposure to ionizing radiation, at doses of 0.3, 3, and 30 Gy, and subjected to RPA production (Fig. 1A
).
View larger version (30K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1. Protein expression dynamics measured by RPA. A, RPA slides stained with p53 and Mdm2 antibody. The slides depict 3,800 features showing 10-time, 2-fold serial dilutions and 2-time technical replicates of 190 different samples. The 190 samples consist of protein lysates taken from 10 control samples and 3 biological replicates of the 10 time points studied after the exposure of two cell lines to three radiation doses. The serial dilutions were used to produce dilution curves to facilitate quantitation. Protein expression levels of total p53 (B), total Mdm2 (C), p53-Ser15 (D), and Mdm2-Ser166 (E), generated by a DI25 algorithm, are plotted as a function of time. Each protein has a set of data from three doses and a pair of isogenic cell lines (p53+/+ and p53–/–). Points, average of three biological and two technical replicates. Horizontal axis, time (hours); vertical axis, protein expression levels in arbitrary units, calculated by dividing each data point by the value representing the maximum staining intensity found on the array. Bars, SE.
|
|
In p53+/+ cells at a dose of 30 Gy, both proteins showed discrete peaks at 
4 and 7 h (Fig. 1B and C). Oscillations of p53 and Mdm2 have been reported in computational simulations, with a delay caused by the time lag necessary for activation of Mdm2 by p53 (4–6); however, the discrete timing of our sample collections may have made this delay difficult to see. Evidence of a delay is better seen at the 2-h time point in p53+/+ cells exposed to 30 Gy, when p53 has begun to increase; however, Mdm2 does not increase until later. Mdm2 showed very little increase in p53–/– cells at all doses, suggesting involvement of p53 for Mdm2 induction after irradiation. Interestingly, p53–/– cells showed a gradual decrease in Mdm2 on exposure to radiation, presumably due to an increased rate of degradation. Thus, although p53–/– cells do not possess the p53 gene, p53-independent activity upstream of Mdm2 may still regulate Mdm2 stability (9, 10).
In addition to stabilizing p53 by preventing Mdm2 from binding, phosphorylation at Ser15 is known to aid p53 in transcriptionally activating other genes (Fig. 1D; ref. 11). Within the first hour following a high dose of radiation, the level of p53-Ser15 in p53+/+ cells increases more rapidly than total p53, implying that phosphorylation of the existing cellular pool of p53 is critical for its accumulation after stress. Total Mdm2 decreases (Fig. 1C) at all doses within the first hour after radiation, probably because of stress-induced degradation of Mdm2; however, Mdm2-Ser166 increases or stays constant, indicating that this form of the molecule might be more stable (Fig. 1E; ref. 12). Phosphorylation of Mdm2 at Ser166 by ionizing radiation has been shown by inducing protein kinase B/Akt activation, which presumably stabilizes Mdm2 and helps it enter the nucleus, where it inactivates p53 (12). Our observations of phosphorylated proteins with total p53 and Mdm2 levels indicate a possible correlation between phosphorylation at certain sites, conferring stability and activity of these species in this time period.
Mathematical models of the p53-Mdm2 feedback loop. The initial model generated by Lev Bar-Or et al. (5) showed damped oscillations of p53 and Mdm2 after stress through a mathematical representation of an unknown intermediary process leading to a delay in the p53-dependent induction of Mdm2. Recently, a model was developed by Ma et al. (6) depicting oscillations by showing the role of the DNA damage sensor, ataxia telangiectasia mutated, in both phosphorylating p53 and increasing the degradation of Mdm2 (13, 14). Based on these two studies, we generated an interaction map consisting of four proteins: p53, Mdm2, p53-Ser15, and Mdm2-Ser166 (Fig. 2A
). It should also be noted that other models have shown to generate similar oscillations (15). For instance, the model by Ciliberto et al. (4) showed oscillations of p53 as a result of both negative and positive feedback loops. Chikarmane et al. (3) proposed a model taking into account the role of additional proteins including Arf, Siah, and ß-catenin in p53 regulation.
View larger version (26K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2. Schematic diagram of the p53-Mdm2 feedback loop. Red and blue ovals, p53 and Mdm2 molecules, respectively. Gray ovals, those not incorporated in the mathematical models explicitly, but likely to play important roles in the feedback loop (18–20). A, under normal conditions, p53, Mdm2, p53-Ser15, and Mdm2-Ser166 are maintained at a steady-state level. In response to cell stress, oscillations in these protein levels are triggered as a result of activation signals and upstream kinase activity. HAUSP expression is not influenced by DNA damage (20). Solid arrows, signal transduction and conversion of the molecules. Mdm2-Ser166 inhibits p53 activity. Dashed arrows, dephosphorylation. B, in knockout cells, where there is no p53, continuous degradation of unphosphorylated Mdm2 may drive the equilibrium of Mdm2-Ser166 toward dephosphorylation.
|
|
Our experimental results guided the assignment of parameter values for the set of differential equations (Supplementary data), and notably of phosphorylated protein species. We fixed the time scale and the frequency of pulses of each protein to fit accordingly. Kinetic parameters were adjusted in the simulations of p53+/+ cells until we observed oscillations of p53 and Mdm2 at 4 and 7 h to coincide with our experimental results (Fig. 3A
). After making the appropriate parameter settings for p53 and Mdm2, peaks were seen in simulations of both p53 and p53-Ser15 at 4 and 7 h. As in our experiments, the level of p53-Ser15 increased more rapidly than that of total p53 within the first hour after stress was applied. Because the model incorporates the effect of upstream kinase activity on Mdm2, which induces rapid and selective degradation of the molecule, the simulations showed an initial dip in total Mdm2 level, consistent with simulations by others (4, 6) and with our RPA results. In contrast, we saw Mdm2-Ser166 stay at the initial level within the first 3 h before exhibiting oscillations, similar to our experimental observations, suggesting that phosphorylation at this site could prevent rapid degradation, thereby stabilizing the molecule (12).
View larger version (31K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3. Mathematical simulations of the p53-Mdm2 feedback loop in response to 30 Gy of -irradiation. Each column illustrates the protein expression levels of four protein species. A, RPA data of p53+/+ cells and the simulations of p53+/+ cells; B, RPA data of p53–/– cells and the simulations (in silico knockout) of p53–/– cells. Horizontal axes, time (hours); vertical axes, protein expression levels by linear scaled DI25 values (RPA results) and fold changes in concentration (simulations). B, insets, identical kinetics at the same scale as the wild type. Bars, SE.
|
|
Based on parameters that agreed with the RPA results in p53+/+ cells, we created an in silico "knockout" of the p53 gene by setting the p53 production rate of the mathematical models to zero (Fig. 3B). As in our experiments, total Mdm2 decreased after exposure to stress, whereas Mdm2-Ser166 increased briefly up to 1 h and decreased gradually afterwards. Interestingly, Mdm2 and Mdm2-Ser166 are similar in both p53–/– and p53+/+ cells within the first 3 h, as total Mdm2 drops and Mdm2-Ser166 is sustained. However, due to the lack of p53 production, total Mdm2 does not increase after 3 h in p53–/– cells. This finding suggests that, after 3 h, upstream kinase activity continues to degrade unphosphorylated Mdm2, causing the Mdm2-Ser166 equilibrium to shift toward dephosphorylation, which results in a concomitant decrease in Mdm2-Ser166 levels (Fig. 2B).
Protein expression synchronization in cell population. From our RPA results at lower radiation doses, we found that p53+/+ cells seemed to show no noticeable oscillation. However, previous studies have shown that individual cells respond with discrete pulses in a nonsynchronized manner, even at low radiation doses (16). To explain this discrepancy, we stained populations of cells for p53 and p53-Ser15 expression after radiation at doses of 0.3, 3, and 30.0 Gy. We found that more cells expressed higher levels of p53 in response to higher doses of radiation (Fig. 4
). At lower doses, fewer cells may need to respond to DNA damage, presumably because the dose does not reach a threshold level necessary for a response in a majority of cells (17). Therefore, pulses at lower doses may not be clearly visible at the full population level due to an averaging effect. We postulate that, at higher doses, most cells respond at closer-to-maximum capacity, exhibiting pulses that are closely synchronized and therefore detectable at the full population level. These observations suggest that protein expression kinetics at lower stress levels may not appropriately reflect the network at the single-cell level; however, at high stress levels, where most cells are showing response above the threshold level, population-driven quantitative data can still provide a reasonable approximation of the average stress-induced response of individual cells (17).
View larger version (53K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4. Expression of p53 and p53-Ser15 in response to different doses of ionizing radiation by immunochemistry in HCT116 p53+/+ cells 4 h after radiation. HCT116 p53+/+ cells carry wild-type p53 alleles and thus show visible basal levels of p53 staining in control (nonirradiated) cells. At 0.3 Gy, cells start responding by exhibiting p53, although most cells exhibit the same level of expression as the control. At 3.0 Gy, more cells show a distinct response than at 0.3 Gy, but there is considerable heterogeneity. At 30 Gy, most cells show a distinct response. High-level expression is seen at 30 Gy, including some very high responders that are presumably pulsating. Expression of p53-Ser15 is also distinct at the dose of 30 Gy, although a small fraction of cells shows a visible response at lower doses. 4',6-Diamidino-2-phenylindole (DAPI) staining highlights the corresponding nuclear DNA.
|
|
|
Discussion
|
|---|
Although successful mathematical modeling of signal transduction has been reported, details of the timing of events are largely missing because compensating quantitative experimental systems are lacking. The aims of our present study are (a) to provide highly quantitative measurements at the protein level as a function of time; (b) to simulate a gene knockout effect by mathematical modeling; and (c) to compare the in silico knockout model with an isogenic in vitro model. Although we are not explicitly incorporating all of the existing identified molecules in the pathway as part of the equations, using a combination of the RPA system and the knockout isogenic cell lines, we predicted the knockout effect in vitro with an in silico approach at the protein level, suggesting that the robust response of the feedback loop represented by p53 and Mdm2 is consistent, although it may be a consequence of complex processes.
Molecular network studies often focus on signaling events by using averaged or pooled samples from a heterogeneous population. This data can be easily misinterpreted because individual cells behave differently in response to external stimuli. For instance, recent extensive protein imaging analysis revealed considerable variability of the p53-Mdm2 feedback loop at the single-cell level when cells were exposed to low and medium doses of -irradiation (15). However, tracking multiple species of molecules with multiple parameters at the single-cell level is not always feasible. Although a high-stress dose may be necessary to synchronize responses of the cells, using RPAs can be an adequate complementary technique to imaging analysis. Accumulating such quantitative proteomic reference data represents a substantial contribution to understanding signaling networks at the systems level.
|
Acknowledgments
|
|---|
Grant support: National Cancer Institute, NIH, under contract no. N01-CO-12400.
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.
We thank Drs. Myong-Hee Sung and Toni Holway for critical discussions and reading of the manuscript, Dr. Bert Vogelstein for providing HCT116 isogenic pair of cell lines, and Maritta Grau for extensive manuscript editing.
|
Footnotes
|
|---|
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
Current address for S. Ramalingam: University of North Carolina-School of Medicine, Chapel Hill, NC 27599. Current address for T. Shimura: Biosignal Research Center, Kobe University, 1-1 Rokkodai-cho, Nada-ku, Kobe 657-8501, Japan. Current address for S. Nizhizula: Department of Surgery, Iwate Medical University School of Medicine, 19-1 Uchimaru, Morioka 020-8505, Japan.
Received 1/30/07.
Revised 3/27/07.
Accepted 4/25/07.
|
References
|
|---|
- Nishizuka S, Charboneau L, Young L, et al. Proteomic profiling of the NCI-60 cancer cell lines using new high-density reverse-phase lysate microarrays. Proc Natl Acad Sci U S A 2003;100:14229–34.[Abstract/Free Full Text]
- Paweletz CP, Charboneau L, Bichsel VE, et al. Reverse phase protein microarrays which capture disease progression show activation of pro-survival pathways at the cancer invasion front. Oncogene 2001;20:1981–9.[CrossRef][Medline]
- Chickarmane V, Nadim A, Ray A, Sauro HM. A p53 oscillator model of DNA break repair control. Quantitative Biology. 2005 Oct. Available from:http://arxiv.org/PS_cache/q-bio/pdf/0510/0510002v1.pdf.
- Ciliberto A, Novak B, Tyson JJ. Steady states and oscillations in the p53/Mdm2 network. Cell Cycle 2005;4:488–93.[Medline]
- Lev Bar-Or R, Maya R, Segel LA, Alon U, Levine AJ, Oren M. Generation of oscillations by the p53-2 feedback loop: a theoretical and experimental study. Proc Natl Acad Sci U S A 2000;97:11250–5.[Abstract/Free Full Text]
- Ma L, Wagner J, Rice JJ, Hu W, Levine AJ, Stolovitzky GA. A plausible model for the digital response of p53 to DNA damage. Proc Natl Acad Sci U S A 2005;102:14266–71.[Abstract/Free Full Text]
- Bunz F, Dutriaux A, Lengauer C, et al. Requirement for p53 and p21 to sustain G2 arrest after DNA damage. Science 1998;282:1497–501.[Abstract/Free Full Text]
- Nishizuka S, Washburn NR, Munson PJ. Evaluation method of ordinary flatbed scanners for quantitative density analysis. Biotechniques 2006;40:442, 444, 446 passim.
- Khosravi R, Maya R, Gottlieb T, Oren M, Shiloh Y, Shkedy D. Rapid ATM-dependent phosphorylation of MDM2 precedes p53 accumulation in response to DNA damage. Proc Natl Acad Sci U S A 1999;96:14973–7.[Abstract/Free Full Text]
- Maya R, Balass M, Kim ST, et al. ATM-dependent phosphorylation of Mdm2 on serine 395: role in p53 activation by DNA damage. Genes Dev 2001;15:1067–77.[Abstract/Free Full Text]
- Ashcroft M, Kubbutat MH, Vousden KH. Regulation of p53 function and stability by phosphorylation. Mol Cell Biol 1999;19:1751–8.[Abstract/Free Full Text]
- Feng J, Tamaskovic R, Yang Z, et al. Stabilization of Mdm2 via decreased ubiquitination is mediated by protein kinase B/Akt-dependent phosphorylation. J Biol Chem 2004;279:35510–7.[Abstract/Free Full Text]
- Pereg Y, Shkedy D, de Graaf P, et al. Phosphorylation of Hdmx mediates its Hdm2- and ATM-dependent degradation in response to DNA damage. Proc Natl Acad Sci U S A 2005;102:5056–61.[Abstract/Free Full Text]
- Stommel JM, Wahl GM. Accelerated MDM2 auto-degradation induced by DNA-damage kinases is required for p53 activation. EMBO J 2004;23:1547–56.[CrossRef][Medline]
- Geva-Zatorsky N, Rosenfeld N, Itzkovitz S, et al. Oscillations and variability in the p53 system. Mol Syst Biol 2006;2:2006.0033.
- Lahav G, Rosenfeld N, Sigal A, et al. Dynamics of the p53-2 feedback loop in individual cells. Nat Genet 2004;36:147–50.[CrossRef][Medline]
- Kringstein AM, Rossi FM, Hofmann A, Blau HM. Graded transcriptional response to different concentrations of a single transactivator. Proc Natl Acad Sci U S A 1998;95:13670–5.[Abstract/Free Full Text]
- Cummins JM, Rago C, Kohli M, Kinzler KW, Lengauer C, Vogelstein B. Tumour suppression: disruption of HAUSP gene stabilizes p53. Nature 2004;428:1 p following 486.[Medline]
- Li M, Chen D, Shiloh A, et al. Deubiquitination of p53 by HAUSP is an important pathway for p53 stabilization. Nature 2002;416:648–53.[CrossRef][Medline]
- Meulmeester E, Maurice MM, Boutell C, et al. Loss of HAUSP-mediated deubiquitination contributes to DNA damage-induced destabilization of Hdmx and Hdm2. Mol Cell 2005;18:565–76.[CrossRef][Medline]
This article has been cited by other articles:
|
|
|
|
 
R. G. Clyde, A. L. Craig, L. de Breed, J. L. Bown, L. Forrester, B. Vojtesek, G. Smith, T. Hupp, and J. Crawford
A novel ataxia-telangiectasia mutated autoregulatory feedback mechanism in murine embryonic stem cells
J R Soc Interface,
December 6, 2009;
6(41):
1167 - 1177.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
X.-P. Zhang, F. Liu, Z. Cheng, and W. Wang
Cell fate decision mediated by p53 pulses
PNAS,
July 28, 2009;
106(30):
12245 - 12250.
[Abstract]
[Full Text]
[PDF]
|
|
|
Top
Abstract
Introduction
