Mechanisms regulating the p53 core module (Fig. 5). Under normal unstressed conditions the negative regulation of MDM2 keeps p53 order 76932-56-4 activity at low levels; but under various stress conditions, upstream mediators such as ATM and Chk2 kinases are activated 22948146 and induce post-translational modification on p53 and MDM2 [50]. These modifications lead to stabilization of p53 and an increase in p53 activity. Experimental studies in populations of cultured cells showed that p53 and MDM2 undergo damped oscillatory behavior followingModeling of Memory ReactionsFigure 2. Stochastic simulations of single-gene expression using the same rate constants. 
(A) Gene On/Off states; (B) mRNA numbers; (C) protein numbers. Two simulations when the lengths of memory windows are constants (length of transcription window l1 10 min and length of gene inactivity window l2 50 min). (D) Gene On/Off states; (E) mRNA numbers; (F) protein numbers. Two simulations when the lengths of memory windows follow the exponential distributions with mean li . (G) Gene On/Off states; (H) mRNA numbers; (I) protein numbers. Two simulations when the lengths of memory windows follow the Gaussian distributions N(li ,s2 ) with s 0:2: doi:10.1371/journal.pone.0052029.gDNA damage caused by gamma irradiation [51]. However, the protein dynamics observed in single cells was similar to digital clock behavior [9,52]. Although mathematical models have been designed to simulate the network dynamics either at population level [50,51,53,54] or at single-cell level [50,52,55], it is still a challenge to BIBS39 site realize experimental observations in single cells and population of cells simultaneously [56]. To tackle this challenge, a stochastic model with memory reactions (see Supporting Information S1) was designed to describe the dynamics of the p53 core circuit using rate constants estimated from experimental data that were given in STable 2. The transcription process of MDM2 follows the same assumptions in Fig. 1. We used two memory reactions to represent the gene activation and inactivation windows. Following experimental observations, it was assumed that the expression of gene MDM2 is activated continuously over a period of ,1 h and then an inactivated window of ,5.5 h follows [9]. Using the activity of ATM kinase as the upstream signal [50], Fig. 6 gives simulated protein numbers of p53 and MDM2 that were activated by the upstream signal with different pulse numbers. Simulations precisely realized experimentally measured p53 and MDM2 molecular numbers [57]. The sustained upstream signal maintained continuous oscillations of p53 activity that led to the corresponding expression cycles of gene MDM2. Simulations suggested that the feedback regulations between p53 and MDM2 are not sufficient to continue the expression oscillations. The p53 activities gradually return to the basal levels after one expressioncycle if the upstream signal ceases. When the p53 activity is below a threshold value, the TF activity is not adequate to stimulate another expression cycle of gene MDM2. Although the decrease of MDM2 activity contributes to the accumulation of p53 proteins, this negative regulation is not critical for the increase of the p53 transcriptional activity. We have demonstrated that the proposed gene activation window play a key role in inducing gene expression bursts with fairly constant width and height at the single cell level. The next question is whether the proposed stochastic model can realize the damped o.Mechanisms regulating the p53 core module (Fig. 5). Under normal unstressed conditions the negative regulation of MDM2 keeps p53 activity at low levels; but under various stress conditions, upstream mediators such as ATM and Chk2 kinases are activated 22948146 and induce post-translational modification on p53 and MDM2 [50]. These modifications lead to stabilization of p53 and an increase in p53 activity. Experimental studies in populations of cultured cells showed that p53 and MDM2 undergo damped oscillatory behavior followingModeling of Memory ReactionsFigure 2. Stochastic simulations of single-gene expression using the same rate constants. (A) Gene On/Off states; (B) mRNA numbers; (C) protein numbers. Two simulations when the lengths of memory windows are constants (length of 
transcription window l1 10 min and length of gene inactivity window l2 50 min). (D) Gene On/Off states; (E) mRNA numbers; (F) protein numbers. Two simulations when the lengths of memory windows follow the exponential distributions with mean li . (G) Gene On/Off states; (H) mRNA numbers; (I) protein numbers. Two simulations when the lengths of memory windows follow the Gaussian distributions N(li ,s2 ) with s 0:2: doi:10.1371/journal.pone.0052029.gDNA damage caused by gamma irradiation [51]. However, the protein dynamics observed in single cells was similar to digital clock behavior [9,52]. Although mathematical models have been designed to simulate the network dynamics either at population level [50,51,53,54] or at single-cell level [50,52,55], it is still a challenge to realize experimental observations in single cells and population of cells simultaneously [56]. To tackle this challenge, a stochastic model with memory reactions (see Supporting Information S1) was designed to describe the dynamics of the p53 core circuit using rate constants estimated from experimental data that were given in STable 2. The transcription process of MDM2 follows the same assumptions in Fig. 1. We used two memory reactions to represent the gene activation and inactivation windows. Following experimental observations, it was assumed that the expression of gene MDM2 is activated continuously over a period of ,1 h and then an inactivated window of ,5.5 h follows [9]. Using the activity of ATM kinase as the upstream signal [50], Fig. 6 gives simulated protein numbers of p53 and MDM2 that were activated by the upstream signal with different pulse numbers. Simulations precisely realized experimentally measured p53 and MDM2 molecular numbers [57]. The sustained upstream signal maintained continuous oscillations of p53 activity that led to the corresponding expression cycles of gene MDM2. Simulations suggested that the feedback regulations between p53 and MDM2 are not sufficient to continue the expression oscillations. The p53 activities gradually return to the basal levels after one expressioncycle if the upstream signal ceases. When the p53 activity is below a threshold value, the TF activity is not adequate to stimulate another expression cycle of gene MDM2. Although the decrease of MDM2 activity contributes to the accumulation of p53 proteins, this negative regulation is not critical for the increase of the p53 transcriptional activity. We have demonstrated that the proposed gene activation window play a key role in inducing gene expression bursts with fairly constant width and height at the single cell level. The next question is whether the proposed stochastic model can realize the damped o.
