We are searching data for your request:
Upon completion, a link will appear to access the found materials.
When scientists mutate bacteria or even embryos of lambs and other animals, why doesn't the p53 reverse whatever mutation the scientists cause? I know that the p53 stops the DNA from mutating and since the restriction enzymes essentially mutate the DNA, why doesn't the p53 stop it? Does the p53 gene have to be temporarily turned off?
I am no expert in this area, and this answer is only based on a reading of the Wikipedia article on p53, which you should perhaps have read carefully. I welcome edits or correction by persons more knowledgeable than me on this topic.
First, although the Wikipedia article neglects to mention it, you can forget about bacteria as the protein is only found in eukaryotes.
Second, one needs to consider what exactly p53 does in relation to protection of DNA from mutation, how it is activated to do this, and hence whether this would occur in the circumstances of the introduction of mutated DNA into a eukaryotic genome.
From the article it appears that “It (p53) can activate DNA repair proteins when DNA has sustained damage”, and that “p53 becomes activated in response to… stressors, including… DNA damage (induced by either UV, IR, or chemical agents such as hydrogen peroxide)… ”
The manipulation of animal DNA will be performed outside of the target organism, and the modified DNA re-introduced into the genome by one of several mechanisms that are unlikely to evoke the stress signals that are evoked by agents that damage DNA 'naturally'. Hence there is no reason to expect the activation and action of p53 in response to this type of experiment.
The P53 pathway: what questions remain to be explored?
The p53 pathway is composed of hundreds of genes and their products that respond to a wide variety of stress signals. These responses to stress include apoptosis, cellular senescence or cell cycle arrest. In addition the p53-regulated genes produce proteins that communicate these stress signals to adjacent cells, prevent and repair damaged DNA and create feedback loops that enhance or attenuate p53 activity and communicate with other signal transduction pathways. Many questions remain to be explored in our understanding of how this network of genes plays a role in protection from cancers, therapy and integrating the homeostatic mechanisms of stress management and fidelity in a cell and organism. The goal of this chapter is to elucidate some of those questions and suggest new directions for this area of research.
BIO 327 HW 7 CH 18
Chromosomes can not be pulled apart until every chromosome has been positioned correctly on the mitotic spindle. If the kinetochores are not attached, they send a signal that say "STOP", which inhibits APC/C.
Bak and Bax are responsible for initiating apoptosis by releasing cytochrome c into the cytosol
But if researchers inject the cytochrome c into the cytosol already, Bak and Bax will not need to initiate cell death because the cytochrome c will already initiate that and are thus bypassed
Condensins function as rings that hold loops of DNA together to compact chromosome structure during mitosis.
Cyclin levels rise slowly because they are transcribed and translated and it takes time for the proteins to accumulate prior to the onset of their cell-cycle phase.
Rb blocks the passage from G1 into S phase by inactivating transcription factors required for the expression of genes that allow entry into S phase.
p53 can arrest the cell cycle and allow time for the cell too repair its DNA damage.
The mitotic spindle is mostly disassembled during telophase, before cytokinesis begins
The mitotic spindle is composed of two spindle poles with microtubules radiating from them.
Chromosomes that are attached to opposite poles are under tension because both microtubules are pulling and pushing on the chromosomes
The levels of M cyclin gradually rise due to the time required for expression of the gene and synthesis of the proteins.
Once M cyclin is expressed, it can bind to Cdk, but this is not sufficient to switch it on
As cells enter M phase, they weaken their grip on the surface on which they are growing and round up.
Apoptotic stimuli activate the apoptotic signaling cascade, which is a series of proteolytic cleavages.
During embryonic development, asymmetric division sometimes produces daughter cells of different sizes
The centromere is the chromosomal region upon which kinetochores assemble.
The cleavage furrow must form directly between the two spindle poles to ensure that the divided chromosomes are separated into two equivalent daughter cells.
The M-Cdk complex is held in an inactive state by an inhibitory phosphate while it accumulates throughout G2 phase
For example, the discovery of cyclins and Cdks was enabled by studying frog eggs that divided synchronously after fertilization extracts from the embryos were thus representative of the cell-cycle stage at which the extract was made. Researchers have devised means to synchronize asynchronous populations of cultured cells.
As cells enter M phase, they become rounded and decrease adherence to other cells and the substratum
DNA double-strand breaks: signaling, repair and the cancer connection
To ensure the high-fidelity transmission of genetic information, cells have evolved mechanisms to monitor genome integrity. Cells respond to DNA damage by activating a complex DNA-damage-response pathway that includes cell-cycle arrest, the transcriptional and post-transcriptional activation of a subset of genes including those associated with DNA repair, and, under some circumstances, the triggering of programmed cell death. An inability to respond properly to, or to repair, DNA damage leads to genetic instability, which in turn may enhance the rate of cancer development. Indeed, it is becoming increasingly clear that deficiencies in DNA-damage signaling and repair pathways are fundamental to the etiology of most, if not all, human cancers. Here we describe recent progress in our understanding of how cells detect and signal the presence and repair of one particularly important form of DNA damage induced by ionizing radiation—the DNA double-strand break (DSB). Moreover, we discuss how tumor suppressor proteins such as p53, ATM, Brca1 and Brca2 have been linked to such pathways, and how accumulating evidence is connecting deficiencies in cellular responses to DNA DSBs with tumorigenesis.
Previously, we found that symmetries in p53 REs encode several possible binding modes ( 5 ). Nevertheless, the ways that p53 recognizes and accommodates DNA instructions and what happens in the cases of p53 mutations remain unknown. Here we move a step further to study the molecular mechanisms of various binding modes, focusing on native p53REs without insertion and considering DNA structures and p53 mutation. We were encouraged by the converged computational results, which are consistent with experimental findings and provide new insight into p53–DNA interactions.
H14 mode explains many experimental observations
The picture emerging from this work corroborates and provides a molecular mechanism of the highly preferred H14 mode found in DNA analysis for p53REs without spacer ( 5 ). (I) In the H14 mode, p53 may use only one type of protein–protein interface existing in the p53 trimer–DNA complex. (II) The core domain tetramerization in the H14 mode induces DNA bending. Subsequently, the H14 mode fits the DNA supercoil structure with sequence-specific recognition. (III) With adjustment of angles between two p53 dimers, the same p53 tetramer can recognize the p53RE sequence and the Holliday junction geometry. (IV) The highly stable salt-bridges in the H14 mode verify the experimental finding that the E180–R181 interaction is crucial for p53's function. (V) The specific interactions of p53 with only two quarter-sites (one and four) in the H14 mode are consistent, in that at least three p53 mutant molecules are required to inactivate a tetramer ( 36 ). Our computational mutagenesis study confirmed that the p53 tetramer can bind puma BS2 with two R273H mutant monomers and two wild-type p53 monomers, while three R273H mutant monomers disable p53 tetramer–DNA recognition.
In the H14 model, the p53 DNA-binding site surface is partially exposed in the DNA–p53 tetramer complex. Such a feature allows other proteins to regulate p53–DNA interaction. The binding site overlapping with the core domain DNA-binding site on the p53 protein was observed to be promiscuous ( 38 ) since it interacted with different proteins which regulate its function, for example, the 53BP1 protein ( 39 ). In another hand, the quarter-sites two and four are still possible to binding other parts (like C-terminal domain) of p53. For example, variation in the placement of the C-terminal domain can modulate the affinity to DNA by adopting two distinct binding modes: a high-affinity mode at the nanomolar range, and a low-affinity mode at the micromolar range ( 40 ).
Q1234 mode may be stabilized with spacer between half-sites
The specific binding modes and the details of the sequence-specific interactions may depend on the spacers between two half-sites ( 9 ), on gene-specific p53 functions ( 41 ) and may be affected by p53 mutations ( 42 ). The half-site palindrome dominates the p53REs with insertions of 3 bp spacers for the human genome and 8 bp spacers for the mouse genome ( 5 ). The corresponding binding mode Q1234 may be expected to bind p53REs with insertions of 3 bp spacers for the human genome and 8 bp spacers for the mouse genome. In fact, two available structures of p53 dimer–DNA half-site complexes may already corroborate with the role of insertion to stabilize the half-site palindrome mode. While one p53 dimer–DNA structure is helped by chemically cross-linking of p53 to DNA ( 10 ), another is stabilized by the p53 dimer–dimer interaction, which is made possible by insertions between two isolated DNA half-sites ( 9 ).
For the five native p53 REs without spacer studied here, we found that the Q1234 mode is a linear DNA-binding mode. The Q1234 binding probabilities are lower and are structurally less stable than other modes. The Q1234 mode may have better probabilities for the p53REs with highly symmetric half-sites. The p21-5′ site is one of the native p53RE with the highest half-site palindrome. Consistently, the Q1234 mode in the p21-5′ complex has the best energetic ( Table 1 ). The half-sites in the two dimer DNA complexes( 9 , 10 ) are 100% symmetric, while half-site in the p53-trimer DNA complex is less symmetric. Nevertheless, insertion of spacer increases the probability of the Q1234 mode greatly. In the future, it needs to systematically examine the effects of insertion of 3 bp and 8 bp spacer on the binding mode probability, which are beyond the scope of the current work.
Variation of binding modes depends on biological functions: T13 and T24 modes
P53 regulates hundreds genes with verities of biological functions. We have shown that the degree of the H14 coupling distinguishes transcription mechanisms of p53, with the H14 couplings being much stronger for positive regulation than for negatively regulated p53REs ( 5 ). Accordingly, p53REs in different functional group might have different binding modes. Inverse repeat p53REs may favor the H14 mode and direct repeat p53REs may have high possibilities of T13 and T24 modes. As may be seen in Table 1 , both DNA sequence analysis and molecular simulations indicated that T13 is favorable for pDINP1. For p53Aip1, DNA sequence analysis indicated that the T24 is the most favorable binding mode, while molecular simulations shows the T24 is only slightly less favorable than the H14 mode.
Similarly to p53REs, T-box genes also have palindromic response elements with two 10 bp half-sites. In the T-box proteins, the binding modes of transcriptional factors may vary with the arrangement of the DNA-binding site and the insertion of spacers ( 43 ). Xenopus T-box protein Xbra, VegT and Eomesodermin bind two half-site T-box response elements with different symmetrical arrangements: The Xbra binds DNA sites with inverted palindrome without insertion Vegt selects mirror palindrome sites with insertion of 4 bp spacer and Emoes binds translational repeats with spacers of 3-, 4-, 5- and 12 bp ( 43 ). The Xbra binds DNA as a dimer to two half-sites ( 44 ) in a way similar to the H14 mode of p53–DNA interaction (Supplementary Figure 8).
Balance of p53-binding modes and genome stability
p53–DNA recognition necessitates flexibility and specificity ( 45 ). The possibility of multiple binding modes is consistent with biological experiments suggesting that p53–DNA binding can be modulated by factors that can alter the p53 conformational equilibrium ( 46 ). c-Abl can selectively stabilize the p53 tetramer conformation to enhance p53 binding and transcription for p21 but not Bax ( 47 ). mAb 421 differentially affects the binding of p53 to two p53REs, suggesting that optimal binding of p53 may require different conformations of the p53 tetramer ( 41 ).
The existing and balance of several p53-binding modes may be needed to maintain genome stability. The p53 mutation will shift the balance maintained by native p53. Even though both experiments and our simulations confirm that at least three R273H mutant molecules are needed to completely disable the p53 tetramer, the existence of merely one R273H mutant monomer greatly shifts the possibility of a binding mode ( Figure 6 B). The finding has many implications to understand the gain of function of p53 mutation and accumulations of p53 mutants in the cancer cell. Changing of binding modes balance might selectively disable certain pathways while maintaining some of wild-type functions, switch wild-type functions to different functions and activate response elements which are not wild-type p53REs.
Comparison with latest electron microscopy of p53–DNA complex
After submitting this work, the quaternary structure of p53–DNA complex in solution was solved by a combination of small-angle X-ray scattering (SAXS), NMR and electron microscopy (EM) ( 48 ). Even though the low resolution (25–30 Å) cannot provide atomic details, their experiments provided new information to compare with our computational predictions. Here we highlight three features described in the quaternary structure work ( 48 ). (i) Multiple binding modes have been observed. Two leading conformations were identified by the SAXS. Even though it was suggested that the two conformations differ only by the position of the tetramerization domain with respect to the core domains–DNA complex, it is very likely that variation of core domain–DNA-binding modes leads to the two conformations observed experimentally. (ii) EM images have shown that tetramerization domain connects core domains in the p53–DNA complex via only ‘two’ well-defined linkers which correspond to quarter sites one and four (upper panel, Figure 7 ). This EM feature strongly suggests that it may take only two monomers in p53 tetramer to tightly bind DNA, confirming H14 interaction pattern. (iii) p53 dimer–dimer interface has to be highly twisted to fit EM image. With large twist angle, p53 tetramer cannot use its four monomers to bind p53RE without insertion in the Q1234 mode. Instead, we found our p53–DNA supercoil complex in H14 mode fits EM image nicely ( Figure 7 ). The EM experiments did not provide DNA position in the p53–DNA complex. Therefore, the see-through channel in the EM map imply the possible DNA position ( 48 ). As can be seen in the Figure 8, our predicted DNA position and highly bended shape match the see-through channel in the EM map. Combined all these features together, our computational results agree with the observed quaternary structure of p53–DNA complex.
Molecular dynamics simulations
We focus on p53 core domain–DNA interactions. Tetrameric structural models were assembled, based on available X-ray structures of trimeric p53–DNA (PDB code 1tsr) ( 8 ) and dimeric p53–DNA ( 9 , 10 ) complexes. The supercoiled DNA was modeled using the X-ray structure of the nucleosome NP147 centered around THY50 (PDB code: 1kx5) ( 14 ), with nucleic acids mutated to the corresponding sequence of p53REs. The Holliday junction was modeled using the crystal structure of the Flpe–Holliday junction (PDB code: 1m6x) ( 15 ).
The p53 tetramer–DNA complex was solvated with a TIP3P water box with a margin of at least 10 Å from any edge of the water box (typical size of 130 Å × 110 Å × 90 Å) to any protein or DNA atom. The sodium ions were pre-equilibrated in a small water box used in construction of solvation box containing p53 tetramer complex. The initial box had more ions than needed, and the sodium ions too close to solute (5 Å) were removed first, and then sodium ions far from solute were also removed to make overall system neutral. MD simulations were performed using the NAMD package ( 16 ) and the Charmm 27 force field ( 17 ), with constant pressure ensembles (NPT) at 1 atm and temperature at 300 K°. The time step was 2 fs with a SHAKE constraint on all bonds with hydrogen atoms. Productive MD runs were performed after 5000 steps of minimizations and three 150 ps heating and equilibration runs. In the first 10 ns, electrostatic interactions had a cutoff distance of 10 Å. For the simulations after 10 ns, long-range electrostatic interactions were calculated with the PME method ( 18 ).
Three cysteine residues coordinated with Zinc were deprotonated. The bond distances of the three Zn–S (Zn-Cys) and Zn–N (Zn-His) bonds were fixed during the simulations. The charge and vdW parameters of the Zn and deprotonated Cys were taken from Maynard and Covell ( 19 ). We selectively protonated five histidine residues, following the suggestion of Wright et al. ( 20 ). The only two histidines left in the deprotonated state were His179 (coordinated to zinc) and His214 (close to R174).
DNA bending angles were estimated using angles between normal vector pairs calculated with FREEHELIX98 ( 21 ). Protein can cause DNA to bend in different behaviors, including kink, writhe or continuous curvature. A clear definition of bending angle is only valid in the case of DNA kink. The matrix of the angles between all normal vector pairs may characterize DNA bending ( 21 ). We use the difference between the average normal vector angles in the first halfsite and the second halfsite of p53 RE as DNA bending angle, based on the average conformation during 13–14 ns MD simulation. Statistical errors are estimated using 8, 9, 10, 11 and 12 bp in each halfsite.
Free energy landscape and Monte Carlo (MC) simulations
MC simulations were used to estimate the binding mode probabilities based on conformations obtained in the last 5 ns of MD simulations. Using the CHARMM 27 force field ( 17 ) and Generalized Born using molecular volume (GBMV) ( 22 ) implemented in the CHARMM package and, each conformer was first minimized 1000 cycles and then the conformation energy was evaluated by grid-based GBMV. The minimization does not change conformations obtained from MD simulation and only relaxed local geometries due to thermal fluctuation occurred in MD simulations. In the GBMV calculation, no distance cutoff was used, the dielectric constant of water was set to 80 and the Debye-Huckel ionic term was 0.2 to reflect the salt effect.
Following the same procedure, we also evaluated the binding probability of p53 R273H mutant using the exact conformation ensemble obtained from wild-type p53–DNA complex and only relaxing local structural changes due to mutation. Each binding mode had 500 conformations initially. We divided the mutant complexes into four groups. In one mutant monomer group, each of the monomers in the p53 tetramer was mutated in turn, and the other three monomers were unchanged leading to 8000 combinations used in binding probability evaluation. In two mutant monomer groups, two of the monomers in the p53 tetramer were mutated in turn and the other two monomers are unchanged. This combination led to 3000 conformations for each binding mode, and 12 000 conformations were used to evaluate the binding probabilities. The binding probabilities in the groups of three mutants and four mutants were similarly evaluated. Even though current approach allows one to test conformational energy changes due to mutation for a given conformation ensemble, it overlooked the mutation effects to change conformational ensemble. However the large combinational possibility prevented us to run additional MD simulations for each possible mutant complex.
The combination of explicit water simulation with subsequent conformational energy analysis using implicit solvation takes the advantage of proper conformation sampling in the molecular dynamics simulation and efficient energy evaluation with MM-GB approach ( 23 ). Still the possible shortcomings existed in each of the methods (MD simulations in explicit water, GBMV energy evaluation, MC population estimation and mutation effects) may not be fixed easily. Proper electrostatics calculations and conformation sampling are coupled. Ewald summation introduces an artificial periodicity into non-periodic molecular system in solution. Even though the artifacts maybe not obvious for polyalanine, for charged molecule and ionic solvation the consequences should be addressed ( 24 ). The solvation effects in GBMV energy calculations used in the energy landscape calculations are also approximate, and the conformations used in GBMV analysis are taken from explicit water simulations. For systems such as those explored here, simulations may be expected to largely sample conformations reachable from the starting conformations. The energy barriers may be too high to sample conformations away from these states, which prevents a single long MD simulation to sample all relevant binding modes automatically. Thus, the GBMV energy evaluations and subsequence MC simulations do not overcome the shortcoming inherited from early simulations. The advantages of using the MC simulation to estimate binding mode probability rely on that (i) the MC simulation has good numerical stability and (ii) MC simulations allow transition probabilities among several binding modes to be controlled. Even though all modes are allowed to freely change in the current simulation, one may define transition pathway in future study. Nevertheless, the binding mode probabilities need to be carefully checked. Here we seek to check the correlations of the binding mode probabilities from molecular simulations and DNA sequence analyses. A correlation of two totally different methods would indicate the molecular simulations provided essential structural basis of p53 tetramer to read the information encoded in DNA sequences.
Sequence-dependent quarter-site couplings
Definition of quarter-site coupling (A) and five corresponding binding modes. (B) Q1234 is the mode of fully occupied quarter-sites with half-site palindrome. (C) T13 mode. (D) T24 mode. (E) The H14 mode uses quarter-sites 1 and 4 to tightly bind p53, and (F) H23 where quarter-sites 2 and 3 are fully occupied. Starting models assembled from existing crystal structures for four binding modes (Q1234, T13, T24 and H14) are illustrated. The initial structures are subjected to refinements using molecular dynamics simulations.
Definition of quarter-site coupling (A) and five corresponding binding modes. (B) Q1234 is the mode of fully occupied quarter-sites with half-site palindrome. (C) T13 mode. (D) T24 mode. (E) The H14 mode uses quarter-sites 1 and 4 to tightly bind p53, and (F) H23 where quarter-sites 2 and 3 are fully occupied. Starting models assembled from existing crystal structures for four binding modes (Q1234, T13, T24 and H14) are illustrated. The initial structures are subjected to refinements using molecular dynamics simulations.
For each potential p53RE sequence in the human genome ( 25 ), the numbers of bases involved in the Q-coupling, H-coupling and T-coupling are calculated according to the definitions in Equations ( 2–4 ), respectively. The overall coupling counts in the dataset are the summation of the individual counts of coupled bases in each position in each p53RE.
Mechanisms that maintain robust homeostasis in genetic and biochemical networks are essential for the fitness of organisms in a changing and challenging environment (1). Many physiologically important variables are under tight homeostatic control, where internal concentrations or fluxes are maintained at well-defined levels despite environmental perturbations. Such perfect adaptation/homeostasis (2) has been found, for example, in bacterial chemotaxis (3𠄶), photoreceptor responses (7), and MAP-kinase regulation (8). Drengstig etਊl. (11) have recently shown how perfect adaptation motifs may be identified in reaction kinetic networks.
Although perfect homeostasis can be related to the control-theoretic concepts of integral feedback or integral control (12,13), it has recently been shown that, in reaction kinetic terms, perfect homeostasis is closely connected to the presence of a zero-order flux (14), which controls another controlling agent (i.e., control of the controller). The latter is responsible for the removal or synthesis of a homeostatically regulated intermediate. Fig.ਁ shows two controller motifs from Ni etਊl. (14), in which intermediate A is homeostatically regulated. Fig.ਁ a presents an inflow controller, where the control mechanism can compensate for large in-flow perturbations of A, and Fig.ਁ b presents an outflow-controller, where A shows homeostasis when A is subject to large fluctuations in its removal. It should be noted that these control schemes will generally fail, when large outflows occur in inflow controllers or large inflows occur in outflow controllers (14).
Schemes of inflow (a) and outflow (b) homeostatic controllers in which component A shows robust homeostasis against environmentally uncontrolled perturbations in the inflow and outflow of A (14). Eadapt represents an enzyme important in the adaptation/homeostasis of A, Etr represents one or several enzymes important in transforming/removing A, and ksynth is a rate constant associated with the synthesis of A. Thick solid arrows with kpert indicate where in the controller inflow or outflow perturbations occur. For a more detailed discussion of these schemes, see Ni etਊl. (14).
Here, we demonstrate that the two homeostatic controllers in Fig.ਁ can show damped or practically undamped large amplitude harmonic oscillations. The degree of damping depends on the binding characteristics between the controller Eadapt and A, as well as on the synthesis and removal of the homeostatically controlled intermediate A. To our knowledge, this is the first example that describes large amplitude harmonic oscillations in a biochemical oscillator model (see the recent review on design principles of biochemical oscillators (15)).
Interestingly, the controller in Fig.ਁ a shows high similarity to the feedback control of p53 by Mdm2, when A is taken as p53, Eadapt as Mdm2, and Etr as the class of Mdm2-independent proteasomal degradation reactions of p53 (16). In the presence of DNA damage, p53 is upregulated by slowing down its various degradation reactions, but still requires a tight control to avoid premature apoptosis by high levels of p53 (22,23). We propose the idea that this control is mediated by Mdm2 and related factors by means of a homeostatic inflow mechanism, which maintains a level of p53 in a state of indecisiveness, until a final decision between cell cycle arrest/DNA repair and apoptosis is made (24). Oscillations in p53/Mdm2 (25,26) may participate in making this decision. In the proposed inflow model, harmonic oscillations in p53 and Mdm2 can occur when p53 binds strongly to the Mdm2-induced degradation machinery, where p53 oscillates around the level defined by the homeostatic controller. Due to the harmonic character of the oscillation, rapid molecular noise leads to large variations in the p53/Mdm2 amplitude whereas the period is only little affected𠅊 behavior that has been experimentally observed (27) but which is difficult to reproduce by deterministic limit-cycle models (27,28). Large fluctuating amplitudes in the p53/Mdm2 oscillations seem to be of importance in determining cell fate (26,27), as will be discussed in more detail below. Thus, a homeostatic inflow model provides an integrative view on the negative feedback regulation of p53 and the appearance of oscillations. Such a view may also provide new insights into the origin and role of oscillations observed in homeostatically controlled molecular networks.
Harmonic oscillations in perfect controllers
A possible kinetic representation for the inflow-controller scheme of Fig.ਁ a can be given by
where V max E i = k cat E i E i tot with k cat E i and Ei tot is the turnover number and total concentration of enzyme species i, respectively. The n is the reaction order with respect to A in the removal of A by Eadapt. With respect to the discussion that will follow below, it may be noted that zero-order kinetics with respect to A (n = 0 in Eq. 1) may be obtained by
when K M E adapt ≪ A . In terms of a rapid equilibrium model of the Michaelis-Menten equation, small K M E adapt values can be interpreted as a strong affinity between substrate A and Eadapt.
The set-point for homeostatic regulation in A is determined by setting Eq. 2 to zero and demanding that the controller Eadapt is removed by another control species (Eset) under zero-order conditions. This requires that K M E set ≪ E adapt , which gives the homeostatic set-point Aset for A at steady-state conditions (14):
A is robustly regulated as long as the right-hand term of Eq. remains practically constant and as long the degradation in A is not dominating with respect to the influxes kpert and ksynth. In Eq. 2, the removal of Eadapt by Eset using Michaelis-Menten kinetics also ensures that, even at low K M E set values, Eadapt does not become negative, as it sometimes would if the Aset term ( V max E set E adapt ) / ( K M E set + E adapt ) were to be replaced by a true constant (14).
An interesting aspect is that oscillations emerge in the controller when the reaction order becomes zero with respect to A. Fig.ਂ illustrates this behavior by applying a stepwise change in kpert (from 1.0 to 2.0) when the system is initially at a steady state at first and zero reaction orders with respect to A. By using the rate constant values described in Fig.ਂ such that the term V max E tr A / ( K M E tr + A ) in Eq. 1 can be neglected and assuming zero-order kinetics with respect to A, we can approximate Eqs. 1 and 2 by Eq. 5,
which leads to undamped harmonic oscillations in A and Eadapt with a period length P = 2π/(k · kadapt).
Generation of harmonic oscillations for the homeostatic controller in Fig.ਁ a by decreasing reaction order n with respect to A. Rate constant values are kpert = 1.0, kadapt = 3.0, k cat E set = 6 × 10 6 , K M E set = 1 × 10 − 6 , ksynth = 1.0, k cat E tr = 1 × 10 2 , and K M E tr = 1 × 10 2 with Aset = 1.0. The reaction orders n with respect to A are (a) 1.0 (b) 1 × 10 𢄡 (c) 1 × 10 𢄢 and (d) 0.0. At time t = 10.0 a.u., kpert is increased from 1.0 to 2.0 and the system approaches a new steady state. Note that A shows robust homeostasis with Aset = 1.0. With decreasing n values harmonic oscillations are emerging where A oscillates around Aset with a peak amplitude approaching Aset as nਊpproaches zero.
A kinetic representation of the outflow control scheme of Fig.ਁ b can be described as
In this formulation, the controller shows an oscillatory response in A and Eadapt for moderate ksynth values and for low values in K M E tr and K M E set , i.e., having a zero-order degradation of A in Eq. 6 and a first-order degradation rate of Eadapt with respect to A in Eq. 7. In the following we will focus on the inflow controller scheme of Fig.ਁ a as a simple model for the p53 regulatory system and its oscillatory behavior.
Regulation of p53
The p53 system is one of the most complex regulatory networks known (22,24,29). It is involved in the control of cell cycle, senescence, DNA repair, apoptosis, and the prevention of tumor development. More than half of all human tumors contain mutations of the p53 gene and in almost all tumors the p53 regulatory circuit is nonfunctional (31,32). Normally (i.e., in the absence of DNA damaging conditions), p53 levels are low due to a rapid degradation by ubiquitin-dependent and ubiquitin-independent pathways with an approximate p53 half-life between 6 and 30 min (16,36,37). An important regulator of p53 is Mdm2, an E3 (ubiquitin) ligase for p53 and other tumor suppressors (38,39). p53 activates the transcription of Mdm2, which binds p53 (40), ubiquitinates it, and thus initiates the proteasomal degradation of p53 both in the nucleus and cytosol (41). This is the central autoregulatory (negative) feedback loop of p53 (29,32). In the presence of DNA damage or oxidative stress, p53 is upregulated by several mechanisms that inhibit Mdm2 activity (42), increase Mdm2 autodegradation (43), and inhibit p53 degradation (44,45). This leads either to cell cycle arrest and DNA repair at lower DNA damage, or to the induction of programmed cell death (apoptosis) at higher DNA damage (24,46,47).
Interestingly, in the presence of high DNA damage, p53 and Mdm2 have been found to oscillate (25,48). The origin and purpose of these oscillations is little understood, but may be of considerable interest (26,50,51).
We became interested in the feedback regulation of the p53/Mdm2 system and its oscillatory response because it shows a close analogy to the inflow homeostatic control scheme shown in Fig.ਁ a with A ≡ p53 and Eadapt ≡ Mdm2. Theontrol scheme suggests that, under DNA-damaging conditions, p53 is homeostatically regulated to a certain upper level defined by Mdm2 (and other factors), at which it decides on the essential cellular functions mentioned above. This view is supported by the fact that transgenic mice, which lack both Mdm2 and p53, grow up normally, whereas mice lacking only Mdm2 die as embryos, possibly due to the uncontrolled apoptotic activity of p53 (22,23).
Once p53 is regulated to a high level, harmonic oscillations can occur when p53 binds strongly to ternary or multiprotein complexes/scaffolds containing Mdm2 (52), which are involved in the (proteasomal) degradation of p53. In the presence of rapidly fluctuating molecular noise, the harmonic character of the p53/Mdm2 oscillations leads to a large variability in their amplitudes but not in their frequency, as will be shown below. This property is difficult to simulate by limit cycle models (28). At normal conditions, i.e., in the absence of DNA damage, p53 is rapidly degraded by ubiquitin-dependent and ubiquitin-independent processes, keeping p53 levels well below its upper limits.
Fig.ਃ a shows an outline of a simple inflow regulatory circuit for the p53-Mdm2 system. A kinetic representation of this model can be given by the following equations:
(a) The p53-Mdm2 negative feedback system as a homeostatic inflow control model. Reactions outlined in black occur in the absence of DNA damage. Under latter conditions, p53 is considered to be rapidly removed through Mdm2 and through Mdm2-independent proteasomal degradation. The Mdm2-independent degradation processes are represented in the model by Ed with Michaelis-Menten parameters K M E d and k cat E d . Eset Mdm2 is an enzyme or a class of enzymes involved in the degradation of Mdm2. When this degradation becomes zero-order with respect to Mdm2, then p53 shows robust homeostatic regulation to the set-point p53set = ( k cat E set Mdm 2 E set , tot Mdm 2 ) / k s Mdm 2 . However, due to rapid p53 degradation at normal conditions, p53 levels are well below p53set. In the presence of DNA damage the degradation of p53 is slowed down and p53 is stabilized. One of the stabilizing mechanisms involve upregulation of NQO1 (16,19,20). Due to the zero-order degradation of Mdm2 by Eset Mdm2 , p53 levels are limited by the set-value p53set (see Fig. , a and b). When the removal of p53 induced by Mdm2 becomes zero-order with respect to p53, harmonic oscillations in p53 and Mdm2 are generated (see Fig. , c and d). p53 ∗ and Mdm2 ∗ represent posttranslational modification species of p53 and Mdm2, respectively. There is evidence that the modified forms p53 ∗ and Mdm2 ∗ do interact much less (30,35). In the model, p53 ∗ and Mdm2 ∗ are assumed to be in rapid equilibrium with p53 and Mdm2, respectively. (b) Molecular mechanism in the Mdm2-mediated degradation which can lead to zero-order kinetics with respect to p53 and first-order kinetics with respect to Mdm2. p53 and Mdm2 bind to a protein complex/scaffold C, which leads to the ubiquitination and degradation in p53. A strong binding of p53 to the complex (small KA and KBA values) lead to zero-order kinetics with respect to p53, whereas a relative weak binding of Mdm2 lead to first-order kinetics with respect to Mdm2. For details, see main text and the Supporting Material.
The Mdm2-mediated degradation term in Eq. 8 is based on a rapid equilibrium among p53, Mdm2, and a protein complex/scaffold C as illustrated in Fig.ਃ b (and described in more detail in the Supporting Material). C0 denotes the total concentration of C, and KA, KB, KAB, and KBA are dissociation (KM) constants. Due to the “Principle of Detailed Balance” (55), we have KA · KAB = KB · KBA. Low values in the Ki values indicate strong binding and stable complexes. Zero-order kinetics in p53 can be achieved by low KA and KBA values, whereas first-order kinetics with respect to Mdm2 is obtained for relative large KAB values. Applying these conditions, we get
In Fig.ਃ a, the outline in black shows the functioning of the system in the absence of DNA damage. p53 is held at low levels due to degradation through a Mdm2-mediated ubiquitin-dependence and Mdm2-independent proteasomal degradations (16,36,37).
Under DNA-damaging conditions, p53 is upregulated and posttranslationally modified. One of the processes that lead toਊn increase in p53 is the Mdm2-independent upregulation of NADH quinone oxidoreductase 1 (NQO1) (16). NQO1 binds to p53 and thereby stabilizes it (56). Both p53 and Mdm2 undergo posttranslational modifications (22,57), where phosphorylated and acetylated forms are indicated in the model by p53 ∗ and Mdm2 ∗ . These forms interact much less than unmodified forms of Mdm2 and p53 and thus cause a stabilization of p53. Due to the decrease in the Mdm2-independent degradation of p53 under DNA damaging conditions but due to the presence of the (still operative) zero-order kinetic degradation of unmodified p53 by unmodified Mdm2, harmonic oscillations of p53 and Mdm2 are initiated, and subsequently propagated to the posttranslationally modified forms p53 ∗ and Mdm2 ∗ . In addition, MdmX has been shown to bind to both Mdm2 and p53, which stabilizes each of these species (58). In our model, the stabilization of Mdm2 and p53 by MdmX is lumped together with the formation of the Mdm2 ∗ and p53 ∗ species. However, it should be noted that MdmX is also present in undamaged cells and considered to maintain transcriptionally inactive p53 in the nucleus of these cells (58).
Fig. shows concentration profiles for p53 and Mdm2 using the model Eqs. 8 with decreasing rates in the ubiquitin-independent degradation of p53 when the Mdm2-mediated degradation of p53 is zero-order with respect to p53. Large degradation rates in p53 through Ed lead to p53 levels well below p53set ( Fig. a),
whereas p53 levels become homeostatically regulated when the Ed-induced degradation becomes sufficiently low ( Fig. b). At even lower Ed-mediated degradation of p53, damped oscillations appear ( Fig. c), where p53 oscillates around p53set with a peak amplitude, which (in the absence of noise) cannot exceed p53set ( Fig. d).
Generation of harmonic oscillations for the homeostatic inflow controller of the p53-Mdm2 system by upregulating p53, i.e., by successively decreasing the k cat E d value of the Mdm2-independent degradation of p53 (see Fig.ਃ and Eqs. 8). Rate constant values (in a.u.) are as follows: ks p53 = 3.5, KA · KAB = 1.0 × 10 𢄤 , KAB = 1.0 × 10 2 , KBA = 1.0 × 10 𢄧 , k′ · C0 = 40.0, ks Mdm2 = 3.0, k cat E set Mdm 2 = 6.0 × 10 6 , K M E set Mdm 2 = 1.0 × 10 − 6 , K M E d = 1.0 × 10 4 , kr p53lowast = 50.0, kd p53lowast = 0.0, kr Mdm2lowast = 1.0 × 10 2 , kd Mdm2lowast = 0.0, Eset,tot Mdm2 = 5.0 × 10 𢄧 , Ed,tot = 0.1, and p53set = 1.0. (a) High values of k cat E d lead to p53 steady-state values well below its homeostatic set-value p53set. At t = 20 time units k cat E d is decreased from 1.0 × 10 7 to 1.0 × 10 6 with ks p53lowast = ks Mdm2lowast = 0.0, which leads to an increase in the p53 and Mdm2 steady-state levels. (b) At t = 20 time units, k cat E d is decreased from 1.0 × 10 6 to 1.0 × 10 5 with ks p53lowast = ks Mdm2lowast = 0.1. Note that p53ਊttains now its homeostatic regulated set-value. (c) At t = 20 time units, k cat E d is decreased from 1.0 × 10 6 to 1.0 × 10 4 with ks p53lowast = ks Mdm2lowast = 0.5. Damped harmonic oscillations in p53 start to emerge around the homeostatic set-value. (d) At t = 20 time units, k cat E d is decreased from 1.0 × 10 6 to 1.0 × 10 2 with ks p53lowast = ks Mdm2lowast = 1.0. Much less damped harmonic oscillations in p53, p53 ∗ , Mdm2, and Mdm2 ∗ are generated (data for p53 ∗ and Mdm2 ∗ not shown).
Amplitude/frequency behavior and influence of noise
The damping of the oscillations given by Eqs. 8 depends on several parameters. A strong damping or no oscillatory response is observed when p53 or the posttranslationally modified species Mdm2 ∗ or p53 ∗ are rapidly degraded, i.e., when rate constants k cat E d , k d p 53 ∗ , or k d Mdm 2 ∗ are large compared to the influx of p53. On the other hand, sustained oscillations are observed when k cat E d , k d p 53 ∗ , or k d Mdm 2 ∗ are much lower than the influx of p53 into the controller. When k s p 53 ∗ and k s Mdm 2 ∗ are zero, the system oscillates with the period 2π/(k · ks Mdm2 ). When k s p 53 ∗ , and k s Mdm 2 ∗ are nonzero, sustained oscillations are also observed when k d p 53 ∗ and k d Mdm 2 ∗ are zero and a rapid equilibrium between p53 ∗ and p53 as well as Mdm2 ∗ and Mdm2 is established. The period increases as the rapid equilibrium is shifted more to the p53 ∗ and/or Mdm2 ∗ side. If the equilibrium between posttranslationally modified p53 ∗ /Mdm2 ∗ and unmodified p53/Mdm2 is slow compared with the influx of p53, oscillations become damped.
Fig.ਅ shows trajectories of (practically) undamped oscillations in the p53-Mdm2 phase plane with different initial concentrations. Because the system is harmonic (conservative), no limit cycle is observed, but parallel trajectories emerge. In the case in which the trajectories reach the ordinate (Mdm2 axis when p53 = 0), Mdm2 concentrations decrease until an oscillator with maximum peak amplitude equal to the p53 set-value emerges (trajectory 5 in Fig.ਅ ). One may consider such a behavior as a filtering of large excursions in p53 down to a maximum peak level determined by p53set.
Phase plane trajectories of p53-Mdm2 harmonic oscillations going through three cycles. Rate constants as described in the legend of Fig. . To illustrate that ks p53 can be chosen without affecting the p53 oscillations around p53set, ks p53 was set to 11.0, and k cat E d = 1.0 × 10 2 . For the sake of simplicity, all ks,r,d p53lowast and ks,r,d Mdm2lowast rate constants are set to zero. Dots show different p53 and Mdm2 start concentrations. Because the system is conservative, parallel trajectories 1𠄷 emerge from each of the starting points. Trajectories 1𠄴 which lie outside of trajectory 5 (which is tangenting the ordinate at p53 = 0) will hit the ordinate at low p53 levels and Mdm2 concentrations will decrease until the system emerges as trajectory 5 oscillations, which have the largest peak amplitude equal to the p53 set-value. Trajectories 6 and 7 which start inside of trajectory 5 will not be altered, and the system oscillates with peak amplitudes lower than the p53 set-value.
The period is not affected by the remaining rate constants. Note that k cat E d needs to be sufficiently small and K M E d needs to be sufficiently large to get oscillations, but the period of the oscillations is not dependent on those values.
Due to the large amplitude variations found for experimentally recorded p53/Mdm2 oscillations (27), we became interested in the effect of fluctuations on the model. For this purpose, rate parameters were allowed to vary randomly and rapidly within a certain range by using the Fortran routine RAN1 (59). Fig.ਆ , a and b, shows the variations in ks p53 and k cat E d as a function of time (see Supporting Material for an overview of all rate parameter variations). Fig.ਆ c shows the behavior of the model compared to experimental data ( Fig.ਆ d). The computations show, in agreement with the experimental observations, that the amplitude of the oscillations is subject to considerable variation, whereas the period and the phase relationship between p53 and Mdm2 are little affected. However, it should be noted that changes in the average values of k (Eq. 12) or ks Mdm2 will lead to period changes, because these two parameters determine period length (compare with Eq. 5).
Rapid fluctuations in rate parameters lead to variations in the amplitude of the p53/Mdm2 oscillations, but preserve their period. For the sake of simplicity, all ks, r,d p53lowast and ks,r,d Mdm2lowast rate constants are set to zero. (aਊnd b) Variations for ks p53 and k cat E d , respectively see also Supporting Material. (c) Resulting oscillations in p53 and Mdm2 levels when applying rapid fluctuations for all rate parameters within the ranges indicated by Table S1 in the Supporting Material. Rate constant values of k and ks Mdm2 have been adjusted such that the period of the harmonic unperturbed oscillations is close to the experimental value of 5.5 h (27). (d) Observed p53 (solid line) and Mdm2 (dashed line) oscillations in single cells. (Replotted from upper left of Fig.ਁ B in Geva-Zatorsky etਊl. (27).)
Stress and Behavior
Heribert Hofer , Marion L. East , in Advances in the Study of Behavior , 1998
A. The Cellular Stress Response
The “ cellular stress response ” is the activation of a class of proteins known as “heat-shock” proteins (HSP) when a cell is submitted to a transient rise in temperature, low pH, low oxygen level, or other physical or chemical treatments ( Morimoto et al., 1990 Nover, 1991 van Eden and Young, 1996 ). These treatments lead to the accumulation of denatured and/or aggregated proteins inside the cell. HSPs repair the damage resulting from these conditions in a variety of ways ( Watson, 1990 Burel et al., 1992 Parsell et al., 1994 ). Hence, HSPs are often called stress proteins and an increase in their production is described as a “cellular stress response.” HSPs are present in all cells of all prokaryotic and eukaryotic organisms and the amino acid sequences of these proteins are highly conserved in all groups of organisms ( Morimoto et al., 1990 Nover, 1991 van Eden and Young, 1996 ). These characteristics indicate an evolutionary continuity of selection pressures exerted by stress since the earliest life forms and suggest that the cellular response may provide the molecular basis of universal components of the organismal stress response. Links between molecular processes on a cellular level and the organismal response are increasingly recognized ( van Eden and Young, 1996 ). For instance, the capacity to cope with infections can be linked to the state of repair of cells of the immune system ( Macario, 1995 ). Thus, for an evolutionary understanding of stress, the cellular stress response is of considerable interest.
In order to organize knowledge of the p53 interactome into a coherent framework, a logical model of the p53 system was constructed ( Figure 1 , Table S1 in File S1). In this model, nodes represent genes or associated proteins that interact with p53, and edges represent the interactions between them. Two types of interacting processes are considered: activating or inhibiting. In an activating interaction, the result is an induction of activity of target node(s), and in an inhibitory interaction, the result is a repression of activity of target node(s) . For example, the induction of p53 stimulates the expression of MDM2 (Mdm2, p53 E3 ubiquitin protein ligase homolog (mouse)) , which is represented by an activating interaction from p53 to MDM2. At the same time, MDM2 activation leads to the down-regulation of p53, which is represented by an inhibiting interaction from MDM2 to p53 .
Java interface programs were created to extract p53 interactions from the STRING database. We then manually curated the data and used Gene Ontology annotations to connect the network to DNA damage input and apoptosis output. CellNetAnalyzer was used for analysis and simulations, and the results were validated using literature surveys and experimental approaches including western blotting and microarray analysis.
Although there are numerous databases recording genetic and protein-protein interactions, few record the effect the interaction has on the target node. A notable exception is the STRING (Search Tool for the Retrieval of Interacting Genes/Proteins) database , which distinguishes between different modes of action, including activation, inhibition and binding. Interaction records of the human p53 interactome were first retrieved automatically from the STRING database (see Material and Methods). The interactions were filtered by retaining only high confidence scores as defined by STRING (more than 0.7). However, because of the limitations of current text mining methods in identifying modes of action, even the group of high-confidence interactions was found to contain some errors. To avoid incorrect data being included into the model, all interaction records were thus manually curated by surveying the associated literature and searching for additional evidence. Examples of the types of errors found and details of interactions that were corrected following the manual curation process are provided in Table 2 and Figures S1–S4 in File S1.
|Scenario name||Input signal||Model type||Percentage of determined nodes|
|Scenario 1||DNA damage ON||P53 wild-type||87.9%|
|Scenario 2||DNA damage OFF||P53 wild-type||88.3%|
|Scenario 3||DNA damage ON||P53 knock-out||45.9%|
|Scenario 4||DNA damage OFF||P53 knock-out||46.3%|
A recurrent question in the construction of in silico models is to define the boundaries of the system. In order to obtain a complete coverage of the p53 interactome, yet keep the size of the system within acceptable limits for simulation, we included all high-confidence interactions with genes/proteins interacting directly with p53, and added all interactions between these genes/proteins that do not involve p53 directly. This process ensured that regulatory feedback loops were included in the model. In a few cases, different proteins were combined into a single node reflecting the fact that earlier literature did not distinguish between them: this was the case for HRAS (v-Ha-ras, Harvey rat sarcoma viral oncogene homolog), KRAS (v-Ki-ras2, Kirsten rat sarcoma viral oncogene homolog), NRAS (neuroblastoma RAS viral (v-ras) oncogene homolog) and RASD1 (RAS, dexamethasone-induced 1), represented as a single node RAS CCNA1 (cyclin A1) and CCNA2 (cyclin A2) represented as CCNA CSNK2A1 (casein kinase 2, alpha 1 polypeptide) and CSNK2A2 (casein kinase 2, alpha prime polypeptide) represented as CSNK2.
Cells respond to numerous stress stimuli including ionizing and UV (ultraviolet) radiation, oncogene activation, heat shock, hypoxia, etc . The DNA damage response mediated by p53 is well studied and most clinically relevant as the majority of cancer treatment strategies involve DNA damage pathways. Therefore, DNA damage was added as an input signal by connecting the network to a single input node representing DNA damage. Similarly, apoptosis and cellular senescence were selected as the best studied and most clinically relevant outputs among numerous other possibilities including cell cycle arrest, DNA repair and angiogenesis. Thus, the network was connected to two output nodes representing apoptosis and senescence. Links from DNA damage and towards apoptosis and senescence were curated using Gene Ontology terms (Tables S3–S5 in File S1) as well as additional manual curation. The resulting model, named PKT206, comprised 203 gene/protein nodes, an input node (DNA damage), two output nodes (apoptosis and senescence) and 738 interactions. Complete lists of genes/proteins and interactions with references to literature based evidence are provided in Tables S1 and S3–S5 in File S1.
Structure of the p53 logical model
The p53 node is connected to 202 genes or proteins in the network and participates in 225 interactions ( Figure 2 ). Five layers can be distinguished in the network according to the relationship of nodes to p53: the input signal, DNA damage upstream nodes of p53 p53 itself and MDM2 downstream nodes of p53 and the outputs, apoptosis and senescence. It was found that 67 nodes functioned as upstream nodes of p53. For example, ATM functions as a DNA damage inducible node upstream of p53  it activates p53 directly as well as through CHEK2 (checkpoint kinase 2) up-regulation –. 146 nodes functioned as p53 target genes, including well studied pro apoptotic genes such as BAX  and CDKN1A that controls cell cycle arrest . 11 genes functioned both as upstream and downstream nodes of p53 and were involved in two step feedback loops.
The PKT206 model represented by Cytoscape includes 203 gene/protein nodes, an input node (DNA damage), two output nodes (apoptosis and cellular senescence) and 738 edges. Activation and inhibition connections are represented by blue and red arrows, respectively. The input node was marked by green the nodes upstream of p53 were marked by yellow p53 and MDM2 were marked by red, the nodes downstream of p53 were marked by light blue and the output nodes were marked by orange.
We calculated the connectivity degree of the 206 nodes in the network ( Figure 3 ). The connectivity degree of a gene indicates the number of interactions for this gene. The most connected gene was p53, which participated in 225 interactions in the PKT206 model. There were 30 genes with connectivity degree between 10 and 100 and the remaining genes were involved in less than 10 interactions.
The degree distribution of 206 nodes in the model was obtained by the NetworkAnalyzer plugin for Cytoscape both axes in the figure are in logarithmic scale.
The network contains 30 two-step feedback loops in total, with 14 involving p53. Some of them play a significant role in p53 regulation for example, the feedback loops involving p53, MDM2 and MDM4 (Mdm4 p53 binding protein homolog (mouse)), which include five interactions: p53 activates MDM2 MDM2 inhibits p53 MDM2 inhibits MDM4 MDM4 activates MDM2 and MDM4 inhibits MDM2 . Feedback loops play a crucial role in p53 regulation and are thought to increase the robustness of the system in response to perturbations .
P53 has been implicated in numerous cellular responses to stress including IR (ionizing radiation), UV, oncogene activation, and hypoxia. For this model to be able to predict cellular fate in response to stress, we linked 20 nodes to the input signal DNA damage (Table S3 in File S1). Most of the links from DNA damage are activations and only 3 are inhibitions (DNA damage inhibits PTTG1 (pituitary tumour-transforming 1), MYC (v-myc, myelocytomatosis viral oncogene homolog (avian)) and AURKA (aurora kinase A). Similarly, p53 controls numerous cellular responses to stress such as cell cycle arrest, DNA damage repair, senescence and apoptosis. We found 95 links between downstream gene nodes and apoptosis and 77 nodes interact with the apoptosis node. Among them, 18 nodes both promoted and prevented apoptosis, 38 nodes only induced apoptosis and 21 nodes only had anti-apoptotic function. We found 52 genes connected to senescence by 61 links, among which 28 promote and 33 prevent senescence.
Analysis of dependencies in the p53 model
Logical dependencies between genes/proteins are represented by the dependency matrix , which represents the effects between all pairs of nodes in the model. Six types of effects are defined by CellNetAnalyzer based on the existence (or not) of positive and negative paths between two nodes: no effect, ambivalent factor, weak inhibitor, weak activator, strong inhibitor, and strong activator (see Methods for details). There are 42,436 (206) elements in the dependency matrix, of which 23,468 correspond to interactions having no effect 16,540 are ambivalent factors 1100 are weak inhibitors 1240 are weak activators 33 are strong inhibitors and 55 are strong activators (Table S6 in File S1). The majority of dependency matrix elements are no effect or ambivalent factors. The large number of ambivalent factors is due to the complexity of regulatory effects between nodes, which are affected by both positive and negative feedback loops and pathways. For example, there are both positive and negative paths from ATM to CHEK2: the positive path is a direct activation of CHEK2 by ATM, while the negative path is an indirect inhibition, as ATM activates p53, p53 inhibits MYC, MYC activates E2F1 (E2F transcription factor 1), and E2F1 activates CHEK2. As a result, the interaction between these two nodes is determined by opposing activating and inhibiting effects, resulting in it being classified as ambivalent (Figure S5 in File S1).
In silico simulation of mutation effects
In order to evaluate the capacity of the PKT206 model to predict perturbation effects, we performed in silico knock-out tests, in which a particular node was removed from the network thus mimicking in vivo mutation effects. As 85% of genes or proteins in the PKT206 model were poorly connected, p53 and those 30 genes with more than 10 interactions were selected to perform in silico knock-out tests. For instance, we simulated a p53 knock-out by removing the p53 node from the network and analyzed the effects of this perturbation. By comparing the dependency matrix after the p53 node was removed with the wild-type case, changes in matrix elements revealed how relationships between nodes were affected by the deletion. 11,785 out of the 42,025 (205) elements in the matrix changed as a result of p53 removal ( Figure 4A ). Major changes are listed in Table S7 in File S1. The most significant changes were from ambivalent factors to activators or inhibitors, reflecting the fact that p53 plays a major role in modulating the system's effects. 11 out of 31 in silico knock-out tests had major changes in the new dependency matrix when a certain node was removed (Table S6 in File S1). 63 potential predictions of major changes in dependency cells were obtained from those 11 in silico knock-out tests ( Table 1 ). There were no major effect changes found in the other 20 in silico knock-out tests.
(A) Distribution of changes in the dependency matrix of the p53 in silico knock-out compared to the wild-type. The gray cycle represents no effect elements, the orange circle represents ambivalent factors, the light green circle represents weak activators, the pink circle represent weak inhibitors, the dark red circle represents strong inhibitors, and the dark green circle represents strong activators the direction of the arrow represents the direction of changes in the knock-out. (B) Chk1 (CHEK1) activation is increased in p53 negative background. U2OS cells that have functional p53 and SAOS2 cells that lack functional p53 were treated with 10 µM etoposide for 16 hours. Cell extracts were analyzed by SDS PAGE and western blot analysis using antibodies against total Chk1, ATR and ATM. ATM and ATR phosphorylated Chk1 at Ser 345.
|Gene deleted||Activated node||Reported effects from literature||References||Predictions||Verified status|
|p53||DNA damage||Expression level of Fas enhanced||||DNA damage promoted upregulation of FAS||Verified by literature|
|p53||LATS2||Cell death enhanced||||LATS2 induced apoptosis||Verified by literature|
|p53||DNA damage||Expression level of CHEK1 enhanced||||DNA damage promoted upregulation of CHEK1||Verified by literature|
|P53||KLF4||CCNB1 reduced||||KLF4 reduced expression of CCNB1||Verified by literature|
|P53||ATM||ATM enhanced CHEK1||Verified in this publication|
|P53||ATR||ATR enhanced CHEK1||Verified in this publication|
|P53||MAPK14||Stimulation of BAX||||BAX enhanced||Consistent with prediction|
|VEGFA||SERPINB5||Apoptosis enhanced in the presence of MMP3 and MMP9 inhibition||||Apoptosis enhanced||Consistent with prediction|
|MDM2||ATM||DYRK2 induced in the presence and absence of DNA damage||||DYRK2 enhanced||Consistent with prediction|
|MDM2||ATR||DYRK2 induced in the presence and absence of DNA damage||||DYRK2 enhanced||Consistent with prediction|
|CDK2||CDKN1A||Apoptosis decreased but not confirmed directly||, ||Apoptosis reduced||Consistent with prediction|
|CDK2||CDKN1A||CDK2 regulates senescence suppression by MYC||||Cellular senescence increased||Consistent with prediction|
|P53||SGK||Cellular senescence decreased||PNP|
|P53||MAPK14||Cellular senescence decreased||PNP|
|P53||LATS2||Cellular senescence decreased||PNP|
|VEGFA||FOXM1||Cellular senescence decreased||PNP|
|P53||DNA damage||CDK4 reduced||PNP|
|P53||DNA damage||FGF2 reduced||PNP|
|TGFB1||DNA damage||MAPK8 enhanced||PNP|
|E2F1||DNA damage||AATF enhanced||PNP|
|E2F1||DNA damage||CHEK2 enhanced||PNP|
|E2F1||DNA damage||CDK5 enhanced||PNP|
|P53||IFNA1||TLR3 reduced||||TLR3 enhanced||Opposite to prediction|
We confirmed 4 out of these 63 predictions through literature searches, focusing on major changes caused by the p53 deletion which were expected to have stronger experimental effects. For example, the effect of DNA damage onto FAS (Fas (TNF receptor superfamily, member 6)) changed from an ambivalent factor in the p53 wild-type model to a strong activator when p53 was removed. The effect of DNA damage onto FAS was classified as ambivalent in the wild-type cells because there are potential negative paths from DNA damage to FAS through MYC and PTTG1, in addition to a direct positive path from DNA damage to FAS. When p53 is deleted, only the positive path subsists. Manna et al. have determined that in p53 minus cells, Fas protein levels are elevated under DNA damage compared to p53 wild-type cells, which is in agreement with our prediction . Similarly to FAS, the effect of LATS2 (LATS, large tumour suppressor, homolog 2 (Drosophila)) onto apoptosis was changed from an ambivalent factor in the p53 wild-type model to a strong activator when p53 was removed. It was found that in both p53 wild-type (A549) and p53 minus cells (H1299), LATS2 was able to induce apoptosis and that apoptosis is slightly increased in H1299 as measured by PARP and caspase 9 cleavage . We observed that the effect of DNA damage onto CHEK1 (checkpoint kinase 1) changed from an ambivalent factor in the p53 wild-type to a strong activator when p53 was removed. CHEK1 protein levels were found to be higher in p53 −/− cells than in p53 +/+ HCT116 colorectal cancer cells treated by daunorubicin , which also matches our predictions ( Table 1 ). It was reported that KLF4 (Kruppel-like factor 4(gut)) caused more reduction of CCNB1 (cyclin B1) expression in p53 −/− HCT116 than in p53 +/+ HCT116 cells  and it matched our model prediction. However, one prediction out of those 63 predictions was found opposite to the literature evidence. The prediction pointed out that IFNA1 (interferon, alpha 1) enhanced TLR3 (toll-like receptor 3) in p53 mutant cells compared to p53 wild type cells. But this was opposite to the fact reported by Taura et al. that IFNA1 exposed to the DNA damaging drug 5-fluoro-uracil(5-FU) reduced the expression of TLR3 in p53 −/− HCT116 cell compared to p53 +/+ HCT116 cells .
In addition to literature based validation, we obtained in vitro based experimental evidence to support novel predictions of the model. The model predicted that in the absence of functional p53, the effects of ATM and ATR (ataxia telangiectasia and Rad3 related) onto CHEK1 would both change from ambivalent factors to strong activators. A western blot analysis of U2OS human osteosarcoma cells that have wild-type p53, and of SAOS2 cells that have mutant non-functional p53, demonstrated that CHEK1 is activated to a higher extent in the p53 mutant background than in the p53 wild-type background ( Figure 4B ) validating this prediction. Furthermore, higher levels and potential activation of ATM and ATR kinases was observed in p53 minus cells than in p53 positive cells. According to the model, there are both positive and negative paths between ATM, ATR, CHEK1, and p53 in the wild type cells, and therefore in p53 mutant cells this balance is disturbed ( Figure 5 ). This confirms the predictive capability of our modelling approach and has consequences for treatment of p53 negative tumours.
(A) Positive and negative pathways from ATM/ATR to CHEK1 in p53 wild type cells as known from literature survey (B) Positive and negative pathways from ATM/ATR to CHEK1 in p53 minus cells. ARF is cyclin-dependent kinase inhibitor 2A. PPM1D is protein phosphatase 1D. pRB is retinoblastoma 1.
Logical steady state analysis
The p53 protein is known to maintain genomic stability and the absence of p53 leads to cellular proliferation in response to DNA damage . The absence of genetic stability triggers the accumulation of mutations in normal cells and causes cancer . In order to investigate how such loss of stability could be captured by our model, we carried out a comparative logical steady state analysis in the p53 wild-type model and in silico p53 knock-out.
In a logical steady state (LSS), the state of each node remains the same over time . Each node can have three different states: inactivated (𠇀”), activated (𠇁”) or undetermined (“NaN”). We investigated four scenarios for logical steady state analysis: (1) DNA damage is activated in p53 wild-type background (2) DNA damage is not activated in p53 wild-type background (3) DNA damage is activated in p53 knock-out background (4) DNA damage is not activated in p53 knock-out background ( Figure 6 , Table 2 and Table S8 in File S1). The comparison of logical steady states in different scenarios revealed that a large number of node states did not change with the change of input signal. This result is explained by the large number of ambivalent effects between nodes and feedback loops in the network, which make the model robust to input signal perturbations. The proportion of determined states was 181 out of 206 nodes (87.9%) in scenario (1), 182 out of 206 nodes (88.3%) in scenario (2), 94 out of 205 nodes (45.9%) in scenario (3) and 95 out of 205 nodes (46.3%) in scenario (4) ( Table 2 ). These numbers show that almost half of the nodes whose state is determined in the wild-type, become undetermined in the in silico p53 knock-out.
The nodes with state 𠇁” were represented in green, the nodes with state ”NaN” (un determined) were represented in orange, and the nodes with state 𠇀” were represented in red. (A) P53 wild type when DNA damage was ”ON” (B) P53 wild type when DNA damage was ”OFF” (C) P53 mutant when DNA damage was ”ON” (D) P53 mutant when DNA damage was ”OFF”.
Comparing the state of 202 genes which interact with p53 in p53 wild type cells in the presence of DNA damage and those in p53 mutant cells in the presence of DNA damage, we found that only 29 genes were up-regulated, 113 genes did not change and 60 genes were down regulated ( Table 3 ). The change of FEN1 (flap structure-specific endonuclease 1) state was moreover experimentally verified by Christmann et al, through finding that FEN1 was repressed in p53 null cells under DNA damage . TLR3 was found to be down-regulated in p53 mutant cells under DNA damage .
|Source scenario||Target scenario||Number of up-regulated genes||Number of unchanged genes||Number of down-regulated genes|
|P53 wild type with DNA damage||P53 mutant with DNA damage||29(14%)||113(56%)||60(30%)|
|P53 wild type without DNA damage||P53 mutant with DNA damage||30(15%)||112(55%)||60(30%)|
|P53 wild type with DNA damage||P53 wild type without DNA damage||5(2%)||185(92%)||12(6%)|
|P53 mutant with DNA damage||P53 mutant without DNA damage||7(3%)||181(90%)||14(7%)|
Comparing the state of these 202 genes in p53 wild type cells in the absence of DNA damage and those in p53 mutant cells in the absence of DNA damage ( Table 3 ), we found that 30 genes were up-regulated, 112 genes remained the same and 60 genes were down-regulated in p53 wild type cells in the absence of DNA damage. The change of 6 nodes were verified by O'Prey et al. . 4 nodes were demonstrated as correct predictions: the expression levels of FAS (TNF receptor superfamily, member 6), TNFRSF10B (tumour necrosis factor receptor superfamily, member 10b), PERP (PERP, TP53 apoptosis effector) and p53AIP1 (tumour protein p53 regulated apoptosis inducing protein 1), were down-regulated from p53 wild type cells without DNA damage to p53 mutant cells without DNA damage, whereas the other 2 nodes MDM2 and CDKN1A were predicted as unchanged by model simulation. However, O'Prey et al. observed their down-regulation from p53 wild type cells without DNA damage to p53 mutant cells without DNA damage . According to the criteria defined in the methods section, four predictions were correct and the other two were small error predictions.
Comparing the state of these 202 nodes in p53 wild type cells in the presence of DNA damage and those nodes in p53 wild type cells in the absence of DNA damage ( Table 3 ), we found that only 5 genes were up-regulated, 185 genes were not changed and 12 genes were down-regulated in p53 wild type cell induced by DNA damage.
Comparing the state of these 202 nodes in p53 mutant cells in the presence of DNA damage and those nodes in p53 mutant cells in the absence of DNA damage ( Table 3 ), we found that 7 genes were up-regulated, 181 genes remained the same and 14 genes were down-regulated in p53 mutant cells induced by DNA damage. Together these above results reflect the fact that p53 helps to stabilize the system.
The changes in the state of anti-apoptotic and anti-senescence genes are shown in Table S9 in File S1 and those of pro-apoptotic and pro-senescence genes are listed in Table S10 in File S1. This distribution illustrates the reason why the apoptosis output was also activated in p53 mutant cells. The majority of those 56 pro-apoptotic genes and 39 anti-apoptotic genes were not changed in the same type of cells treated by DNA damage. The absence of p53 caused obvious changes of both pro-apoptotic and anti-apoptotic genes once the cells were treated with DNA damage. The number of pro-apoptotic and anti-apoptotic genes which were up-regulated or down-regulated increased with the depletion of p53. Among those 56 pro-apoptotic genes, FAS and p53AIP1 were up-regulated in p53 mutant cells when treated by DNA damage. FGF2 (fibroblast growth factor 2(basic)) had both pro-apoptotic and anti-apoptotic function in the PKT206 model and it was down-regulated in p53 wild type cells or p53 mutant cells in the presence of DNA damage. Notably, IGF1R (insulin-like growth factor 1 receptor) and PDGFRB (platelet-derived growth factor receptor, beta polypeptide) were upregulated in p53 minus scenarios, which together with FGF2 changes highlighted growth factor mediated signalling pathways as important factor contributing to survival of these tumours. Approaches that will decrease expression of antiapoptotic genes and increase expression of proapoptotic genes would improve cancer therapy and therefore these genes represent potential therapeutic targets. Two anti-senescence genes were upregulated in the absence of p53, in the presence or absence of DNA damage (CDK4 and FGF2). If DNA damage was applied to either wild type or p53 minus cells, seven anti-senescence genes were increased (Table S9 in File S1). Only three pro-senescence genes increased when DNA damage was applied to wild type or p53 minus cells, and 12 pro-senescence genes decreased in the same conditions (Table S10 in File S1).
Genome-wide experimental validation
In order to evaluate the predictive capability of our logical model on a genome-wide level, predictions of logical steady state analysis in the in silico p53 knock-out were compared with gene expression profiles from microarray analysis. The simulation results of our model were compared with microarray data from 4 different cell types. For this purpose U2OS human osteosarcoma cells that are p53 wild-type and SAOS2 cells that lack functional p53 were treated with the clinically used drug etoposide that causes DNA damage and activates p53. Moreover, we utilized microarray experimental data sets obtained from HCT116 cell lines that have wild type and mutant p53 not treated by DNA damage from <"type":"entrez-geo","attrs":<"text":"GSE10795","term_id":"10795">> GSE10795 .
In order to compare both sets of values and evaluate the performance of our model, we used the approach presented by Christensen et al . The predicted change of gene state between p53 wild-type and in silico knockout was quantified by a variable Emod which could take one of three values: 𢄡, 0 or 1 (see Materials and Methods for details). The experimentally observed change of gene state was represented by a variable Eexp which could take the same three values. For both Emod and Eexp, a value of 𢄡 meant significantly decreased expression, 0 meant no significant change, and 1 meant significantly increased expression.
Using the results of logical steady state analysis, we calculated the value Emod of those 3 types of different cell lines. We extracted relevant genes in the microarray data whose gene names matched those in our logical model. A threshold θ was considered to determine whether a gene was significantly up-regulated, down-regulated or unchanged. In a similar way, we calculated the value Eexp of each gene and listed the number of genes with different changes in Table S11 in File S1.
We then defined the difference between model predictions and experimental results as |Emod – Eexp|. This difference can take three possible values: 0, 1 or 2. Here, a value of 0 meant that the simulation prediction matched the experimental result 1 meant that there was a small error between the prediction and the experimental result 2 meant that there was a large error between the prediction and the experimental result. The distribution of values was calculated and listed in Table 4 . Comparing the changes of gene states between different scenarios with experimental microarray data, 5 scenarios were analyzed. The correct prediction rate ranged from 52% to 71% depending on the cell type, with highly significant p-values compared to random predictions. The percentage of small error predictions ranged from 28% to 42%, and large error predictions were obtained for less than 6% of the genes depending on the cell type. Remarkably, growth factors and receptors FGF2 and IGF1R were identified as common, and PDGFR and TGFA as specific factors contributing to U2OS human osteosarcoma and HCT116 colon cancer cells growth, respectively (Table S11 in File S1). For example, IGF1R is an anti-apoptotic gene upregulated in SAOS2 cells when compared to U2OS cells, whereas FGF2 that can be both pro and antiapoptotic gene is upregulated in SAOS2 cells. In HCT116 cells, with the mutant p53 similarly to SAOS2, there is upregulation of antiapoptotic IGF1R, but PDGFRB and TGFA (transforming growth factor, alpha) are also upregulated, and FGF2 does not change in these cells (Table S11 in File S1), indicating that both general (IGF1R) and cell type specific (PDGFRB and TGFA) pathways were uncovered by the model. Two anti-senescence genes (DDIT4 and DKK1) were upregulated and three (RRM2B, FGF2, FHL2) were down-regulated in SAOS2 cells in the presence of DNA damage, whereas in the absence of DNA damage S100A6 and DKK1 were upregulated and only FGF2 was downregulated. Interestingly, DDIT4 was downregulated in U2OS cells exposed to DNA damage, but upregulated in SAOS2 exposed to DNA damage. There were more changes among pro-senescence genes, where CDKN1A (p21) featured as a major regulator of cell senescence, amongst growth factors and DNA repair genes (Table S11 in File S1).
|Experiment source scenario||Experiment target scenario||Model LSSA simulation||Total number of genes||Number of true predictions||p-value of true predictions||Number of small error predictions||Number of large error predictions|
|U2OS cells under DNA damage||SAOS2 cells under DNA damage||P53 wt with DNA damage ON vs p53 null with DNA damage ON||200||109(54.5%)||2.6휐||80(40%)||11(5.5%)|
|U2OS cells without DNA damage||SAOS2 cells without DNA damage||P53 wt with DNA damage OFF vs p53 null with DNA damage OFF||200||111(55.5%)||4.1휐||77(38.5%)||12(6%)|
|U2OS cells without DNA damage||U2OS cells under DNA damage||P53 wt with DNA damage ON vs p53 wtwith DNA damage OFF||200||142(71%)||<㰐||56(28%)||2(1%)|
|SAOS2 cells without DNA damage||SAOS2 cells under DNA damage||P53 null with DNA damage ON vs p53 null with DNA damage OFF||200||131(65.5%)||<㰐||65(32.5%)||4(2%)|
|HCT116 cells p53+/+ without DNA damage||HCT116 cells p53−/−without DNA damage||P53 null with DNA damage OFF vs p53 wt with DNA damage OFF||169||88(52.1%)||1.8휐 𢄧||72(42.6%)||9(5.3%)|
Brazil’s cancer curse
The startling discovery that hundreds of thousands of Brazilians have a genetic mutation that undermines their ability to resist cancer is helping labs worldwide in their search for new treatments for the disease. Sue Armstrong reports.
Listen to the free audio version
Share this story
Pedro Gomez is a short, powerfully built man in his 60s with the ruddy face and sun-tanned arms of an outdoor worker. He is wearing a short-sleeved black shirt, jeans and a baseball cap. Gomez is worried about a small lump on his finger, he tells the doctor, cancer geneticist Maria Isabel Achatz. Taking his hand in hers to get a closer look, Achatz talks to him gently then leans forward to inspect another small lesion behind his ear.
Gomez is one of Achatz’s regular patients at the A C Camargo Cancer Center in São Paulo, Brazil. He is extraordinarily susceptible to cancer. So too are many members of his extended family cancer is so common among them – and premature death so painfully familiar – that until they learned very recently of the cause, some believed their family was cursed.
Gomez’s is not the only family affected. The ‘curse’ afflicts hundreds of thousands of people in Brazil. One of the highest-profile was José Alencar, the country’s popular and charismatic Vice President under Luíz Ignácio ‘Lula’ da Silva. Alencar died in 2011 after being diagnosed with cancer in 1997. Over the years, as tumours spread relentlessly throughout his body, he underwent more and more operations in Brazil and the USA, having a kidney, most of his stomach and large chunks of his bowel removed. The Vice President talked candidly about his disease and used his own experience to advocate for early detection of cancer.
What Gomez, Alencar and the other Brazilians have in common is a single change in their DNA – a mutation in a gene called p53 that undermines their ability to resist cancer.
p53 has turned out to be the most important single gene in cancer, and has been one of the most popular areas of study in the history of molecular biology. The gene was discovered in 1979 by David Lane, working at the Imperial Cancer Research Fund in London, and coincidentally at exactly the same time by three other groups working independently in the USA and France, and led by Arnold Levine, Lloyd Old and Pierre May.
p53 is a tumour suppressor. Its job is to protect us from cancer by making sure that, when our cells divide as part of routine growth and maintenance of our bodies, they do so without making dangerous mistakes. If the DNA – the cell’s operating instructions – is damaged or not copied faithfully as it divides to produce new ‘daughter’ cells, p53 stops the cell in its tracks and sends in the repair team before allowing the dividing cell to proceed. If the DNA damage is irreparable, p53 puts the cell into a state of ‘replicative senescence’ so it cannot divide again or else it instructs the cell to commit suicide lest it run amok.
When one reflects that, over an average lifespan, a person will experience some 10,000 trillion cell divisions, and that it takes just one rogue cell to start a tumour, the importance of this gene becomes clear. Because of its vital role in quality control, David Lane nicknamed p53 ‘the guardian of the genome’. The gene itself is disabled by mutation or some other faulty mechanism in almost every case of human cancer.
Most often this corruption of p53 occurs spontaneously in individual cells or tissues that have sustained some damage in the rough and tumble of living, and this can set them on the path to cancer – a risk that increases the longer one lives. But some people are born with corrupted p53 in every cell of their body, and are extremely vulnerable to cancer from their earliest days.
Life with Li–Fraumeni syndrome
Sue Armstrong meets Pan Pantziarka, whose son George had Li–Fraumeni syndrome and lived with cancer from early childhood.
© Aart-Jan Venema
As a young woman, Achatz left her family home in Rio to study art in Paris. But a trip to India with fellow students during a college break was a life-changing experience. While visiting a leprosy colony in a remote desert location on the Kashmir border, she met the missionary who ran it: Mother Teresa. “It was an amazing encounter, and I thought, ‘Well, I just have to go back and do something [more worthwhile],’” says Achatz. On her return to Brazil she studied medicine, opting eventually to specialise in genetics.
Among the first patients she saw in the clinic were a number of people who had already suffered multiple bouts of cancer, often starting in childhood, and their tumours were typical of the cancers most commonly seen in people with Li–Fraumeni syndrome. What’s more, when she drew up detailed family trees with her patients – a routine practice in genetic counselling with certain diseases – she uncovered trails of cancer among their relatives, often reaching back generations. They had all the hallmarks of Li–Fraumeni, but Achatz was perplexed: “It really struck me because this was considered to be a very rare syndrome around the world. There were only 280 families described in the medical literature at that time, and I had 30. So I thought, ‘Either I’m over-diagnosing or something unique is happening here.’”
Her colleagues in Brazil were as intrigued as she by what she was seeing in her clinic and encouraged her to take her story to a cancer conference in France in 2002. There Achatz caught the attention of Pierre Hainaut, a tall, bespectacled Belgian who worked at the World Health Organization’s International Agency for Cancer Research in Lyon. Hainaut was custodian of a database of all the different mutations in the p53 gene recorded in the medical literature, and the types of cancer with which each mutation was associated. Aware from his records of the extreme rarity of Li–Fraumeni syndrome, he was fascinated by Achatz’s case notes. He persuaded the young doctor to return to France with blood samples from her Brazilian patients and work with him on identifying exactly what was wrong with their p53 genes.
The two researchers were in for some surprises. Very few of the patients had ‘classic’ mutations in p53 associated with Li–Fraumeni cases elsewhere in the world Achatz’s initial conclusion was that she had over-diagnosed the syndrome. But closer inspection revealed that many of her patients had a p53 mutation that was outside any of the hotspots on the gene known to be most vulnerable to corruption. What’s more, all the patients with this unique mutation carried the exact same copy of the gene.
Some 1,200 km to the south of São Paulo, Patricia Prolla – a fellow cancer geneticist working in Porto Alegre – was also seeing an unusual number of patients with Li–Fraumeni syndrome. And when these turned out to have the same p53 mutation as Achatz’s patients, Prolla and Hainaut resolved to find out how prevalent it might be in the general population. They tested blood from a large sample of apparently healthy women enrolled in a preventive breast screening programme at the Porto Alegre clinic and found, remarkably, that nearly one in 300 had the faulty p53 gene. This startling result was confirmed by a screening programme among nearly 200,000 newborn babies in the nearby state of Parana, where doctors had been finding especially high rates of adrenal gland cancer in small children. Again, it was linked to the same p53 mutation.
“That means that the population of south and south-eastern Brazil has an immense number of Li–Fraumeni carriers, probably more than 300,000,” says Achatz. “People are just not aware of this, so probably many cancers that are occurring in the population in general are due to this mutation and people just don’t realise.”
And it’s not just Brazil. Very recently, the same mutation in p53 has also been found in neighbouring Paraguay, where geneticists randomly tested 10,000 samples of blood from newborn babies. The results suggest that here too several thousand people could be living with Li–Fraumeni syndrome.
© Aart-Jan Venema
If thousands of people share an identical genetic mutation, it’s not due to coincidence. There must have been a ‘founder’, one man (so the thinking goes) with Li–Fraumeni syndrome who passed his mutant gene on to his offspring, setting the ball rolling down the generations.
We don’t know the name of this original carrier, the common ancestor of all today’s carriers, nor where he came from – he may have been an immigrant from Europe. The rogue gene is believed to have travelled along the routes opened from the coast to the interior by early explorers, settlers and military men. One appealing notion is that the founder was a tropeiro, one of a band of travelling traders who journeyed by mule between the scattered settlements, carrying goods and gossip and mail in the 17th and 18th centuries. Away from home most of the time, a tropeiro would likely have had a string of girlfriends along the road, offering an ideal opportunity for passing on his genes. One of Achatz’s biggest Li–Fraumeni families can trace its history back to tropeiro ancestors.
But Hainaut believes a more likely candidate for this ‘patient zero’ would be either a military man or a bandeirante – one of the ruthless adventurers who raided inland for native slaves to trade and to search for precious minerals. When gold was discovered in the 17th century, the rush was on to claim the territory for Portugal before the Spanish could do so. Both the bandeirantes and government servants set to with feverish intent, carving out routes to the interior and creating new settlements along the way. A distribution map of the founder mutation corresponds closely with these routes.
If the founder had been carrying one of the p53 mutations of classic Li–Fraumeni syndrome, it’s unlikely his genes would have spread so far. The lifetime risk of developing cancer for people with such mutations is around 90 per cent, and people born with such pernicious genes have a much-reduced chance of raising a family. (This is one reason why so few cases had been recorded in the medical literature when Achatz first started to see the syndrome in her clinic.) The lifetime risk of cancer for those with the Brazilian mutation is 50 to 70 per cent and, paradoxically, it is this milder character that has enabled it to spread so far and affect such huge numbers of people. Most carriers survive long enough to pass the gene on to their children, and some never develop cancer at all.
© Aart-Jan Venema
The A C Camargo Cancer Center is in a run-down neighbourhood of São Paulo, with narrow streets, small shops and open-fronted diners. In its modern laboratories, which dominate the skyline, is stored the biggest collection of tumour samples in the region – 30,000 scraps of tissue preserved in paraffin wax blocks, meticulously labelled and filed in cabinets.
By studying these tumour samples, Achatz and her colleagues are trying to understand how p53 works in people rather than lab dishes or mice, and how cancer develops when the gene stops functioning properly. For example, one of Achatz’s patients is a woman who, by the age of 18, had developed 14 different tumours. Samples have been taken from many of these tumours, and now the researchers can examine the differences between the DNA of the cancerous tissue and that of the woman’s normal cells.
Meanwhile, Achatz’s colleague at A C Camargo, Fernanda Fortes, wants to know why children with the Brazilian p53 mutation have at least a ten-fold higher risk of adrenal gland cancer than the general population. And, as not all children with the mutation develop this cancer, what tips the balance in those that do? By analysing as many tissue samples as possible from these children, Fortes is hoping to find out. She already knows that the acidity in their tumour cells is higher than normal. And she knows that this is significant – but how significant and in what way? Is this higher acidity in general a cause or a consequence of malignancy?
This is part of a much bigger topic that is exciting the p53 research community right now: the role of the metabolism in cancer, for it turns out the tumour suppressor is a major player in this arena too.
© Aart-Jan Venema
That the metabolism in cancer cells is highly abnormal is not a new discovery. In the 1920s, the German biologist and medic Otto Warburg noticed that cancer cells consume glucose at an enormous rate. He found that whereas most normal cells break down glucose and shunt its products into the mitochondria – the powerhouses of the cell – where they are burned in the furnace to produce energy, tumour cells partially suppress the activity of the mitochondria and use much of the glucose to create the building blocks of new cells. This metabolic process, known as aerobic glycolysis, takes nearly 20 times as much glucose as mitochondrial respiration to produce the energy that cells need – hence cancer cells’ voracious appetite for glucose.
Warburg believed this altered metabolism was the cause of cancer, and said so in a 1956 paper. But his provocative theory was soon overshadowed by the molecular biology revolution, as excited scientists began to look for the causes of everything in our DNA. The excessive appetite for glucose (the so-called Warburg effect), they said, was a consequence of malignant transformation of cells, not a driving force. But now evidence is mounting that metabolism does play an active part in tumour development after all. Recent work on p53 in particular, says Hainaut, points to the fact that metabolic factors are “absolutely fundamental to the biology of cancer”.
There had been clues around since the 1990s that p53 is involved in metabolism, but it wasn’t at all clear how this fitted the picture of the gene as a tumour suppressor. In 2005, however, scientists at the US National Institutes of Health compared the endurance of normal mice with ones whose p53 gene had been deleted. The mice were put into a bucket of water, and those lacking p53 went under much more quickly than the normal ones: clearly they were having difficulty generating enough energy to keep afloat. So, what was going on?
At her lab at Glasgow’s Beatson Institute, Karen Vousden and her fellow researchers have discovered that, in the normal course of events, p53 plays a subtle role behind the scenes. It’s not just watching and waiting to stop or kill potentially dangerous cells, but is actually helping cells to avoid or survive things that might damage them – that is, things that might trigger its anti-tumour response. In other words, p53 is playing a double game: it promotes survival under some conditions, but when it senses things are getting out of control, it calls in the death squad.
The way that p53 promotes survival, explains Vousden, is as a regulator of metabolism, by helping cells cope with fluctuations in the fuel supply. “This might be something that happens all the time, and you wouldn’t necessarily want to kill every cell that just transiently doesn’t have enough glucose. So in those situations, it’s pretty clear p53 helps cells survive. And it does so by allowing the cell to reorganise its metabolism.”
As a basic regulator of metabolism, p53 helps cells resist the glucose-guzzling, inefficient Warburg effect except in emergencies. It also helps clear away free radicals – the corrosive by-products of burning sugar for energy in the mitochondria – thus encouraging the survival of cells by limiting the damage these particles can do to DNA. But if the tumour suppressor isn’t working, harmful free radicals can proliferate, and corrupted cells are free to hijack the metabolic machinery and switch over to glycolysis, which enormously boosts their ability to replicate. This is cancer in the making.
This line of research into the metabolic abnormalities of cancer offers some tantalising prospects for patients. For example, what if we could raid the medicine cabinet for drugs that already exist for metabolic diseases and repurpose them as new treatments for cancer? “You wouldn’t even need to do clinical trials for safety,” points out Vousden, “because these drugs have already been used in millions of humans for years.”
It’s an idea many labs around the world, including her own and Hainaut’s in France, are already exploring with metformin, the most widely prescribed drug for diabetes, which targets faulty glucose metabolism. People with diabetes are usually at increased risk of cancer, but doctors started noticing that the cancer risk in long-term users of metformin seemed to be lower even than that of the non-diabetic population. Could the drug be having a protective effect? Experiments in the lab showed that it is indeed toxic to cancer cells.
“There are good and bad points,” cautions Hainaut. “Metformin will be easy to introduce into cancer treatment because it’s already on the market and there’s a lot of experience of giving it to patients: it’s safe, proven, easy to administer. It has all the characteristics to make a quick hit in cancer treatment if it has a positive effect. But in terms of addressing the glucose weakness of cancer cells, it’s not that strong.”
Metformin is already being tested beyond the lab, with clinical trials of cancer patients in many centres around the world, and Hainaut is encouraging Achatz to try it with some of her patients too. But doctors and scientists alike are acutely aware of the sensitivity of their research among Brazil’s Li–Fraumeni families, and the danger of exciting premature hopes in people desperate for breakthroughs.
© Aart-Jan Venema
Since the identification of the mutant p53 gene in so many members of Pedro Gomez’s large family, they have been struggling – each in their own way – to cope with the implications for themselves and their loved ones. Gomez’s brother, mayor of a small town on the outskirts of São Paolo, took the blood test but balked at learning the results. It was only when his daughter was diagnosed with breast cancer on the eve of her wedding day that he realised he could not hide from the truth. The wedding was postponed while she recovered from a double mastectomy, and today she bullies her father into going along with her for their annual screenings at A C Camargo.
Two of the mayor’s nieces are also carriers of the mutant gene. One says of her diagnosis, “It changed my life for ever it has made me really crazy.” She dreads her annual check-ups, which are time-consuming, invasive and stressful as she waits for the results, always expecting bad news having lost her mother to breast cancer. She frets for her young son, whom she has not yet had tested, and frets too over the morality of having more children, which she and her new husband badly want, and about the possibility of losing her ovaries, womb or breasts to cancer before she can do so. Her cousin, who also wants children, is more philosophical: what will be will be, she shrugs. When she received the news that she had the mutation, the shock she might have felt for herself and her father, who received his test results at the same time, was overwhelmed by concern about her mother’s intense distress for her family.
Achatz is acutely aware of the emotional struggles of the Li–Fraumeni families she sees every day in her clinic. “It’s very clear to me that I’m in science to treat my patients,” she says. “Everything I do goes back to how it affects them.”
So, what of the prospects for the diabetes drug? “Between the proof of principle that metformin works in humans and knowing how to deliver it in the right conditions there are still a lot of steps,” cautions Pierre Hainaut. “But I am seriously hopeful it will work – at least for the Brazilians.”