REPLICATIONAL STRESS SELECTS FOR P53 MUTATION

Abstract

p53 is a critical mediator of cellular responses to a variety of stresses. Given the frequency of p53 mutations in human malignancies and that disruption of p53 has been implicated in chemoresistance, understanding the factors that select for p53 disruption is important both for understanding tumor evolution and for designing cancer therapies. While it is widely believed that genotoxic stress selects for p53 mutations, the effects of DNA damaging agents on long-term proliferative potential are usually not affected by p53 status. Previous reports have demonstrated that despite being activated, p53 loss does not prevent cell cycle arrest and senescence in response to high levels of acute replicational stress. In contrast, we recently reported that chronic exposure of non-transformed cells to low, clinically relevant levels of replicational stress induces p53-dependent senescence-like arrest. Disruption of p53 or its target gene p21^CIP1 antagonizes this arrest, leading to a long-term proliferative advantage. However, when replicational stress is associated with substantial DNA strand breaks, the ability of p53 disruption to up-regulate RAD51 dependent homologous recombination becomes important. Replicational stress is induced by many chemotherapeutic treatments and perhaps by some dietary deficiencies, and may be an important factor that selects for p53 mutations during cancer initiation and progression.

The evolution of cancer appears to be mediated by Natural Selection: an oncogenic mutation that provides a cell with a selective advantage leads to the expansion of the mutant clone. This clonal expansion dramatically increases the chances for secondary adaptive mutations to occur in a cell harboring the initial mutation, which can lead to further rounds of clonal expansion. Thus oncogenic mutations serve as the raw material for tumor evolution, and cancer progression is driven by the selection for mutated clones. Yet, despite the profound importance of selection on the progression of cancers ^{1, 2}, the factors that affect the clonogenic selection of oncogenic mutations have enjoyed relatively little experimental attention. Note that we are using the term “oncogenic mutation” broadly here to include mutations in tumor suppressor genes and heritable epigenetic changes.

The selective pressures that shape tumor evolution can be intrinsic to tumor development. For example, inadequate oxygenation experienced by cells in a growing tumor mass leads to selection for mutations that provide resistance to hypoxia and promote increased angiogenesis. However, extrinsic pressures associated with chemotherapy represent a rather different scenario. While by severely inhibiting cellular proliferation and survival, chemotherapeutic treatments impose profound selective pressures for resistance mutations, the effects of chemotherapy are not limited to tumor cells. Instead, cytotoxic conditions imposed by chemotherapies also affect cells in non-tumor tissues, as these normal cells are put under stress as well. We and others have proposed that the inhibition of proliferation and cell survival should select for resistance mutations in normal cells, initiating oncogenic evolution ^3-5. Chemotherapeutic selection for oncogenic mutations that overcome impaired proliferation is consistent with the increased risk of secondary cancers associated with several chemotherapeutic treatments ^{6, 7}.

The “guardian of the genome” TP53 (p53) gene is mutated in almost half of human malignancies and the overall p53 pathway is thought to be disrupted in most human cancers. Oncogenic p53 mutations have been implicated in resistance to cytotoxic therapies ⁸. Since activation of p53 often leads to senescence or apoptosis, inactivating mutations in p53 or its downstream targets might be expected to be selected under conditions that activate p53. However, for a mutation to be selected, it should result in a long-term proliferative survival advantage, and not merely the short-term avoidance of apoptosis or proliferation arrest. Indeed, while many studies have demonstrated the p53-dependence of survival following various genotoxic insults using short-term survival assays, analyses of long-term survival using clonogenic assays rarely reveals p53 dependence for these same insults ⁹, with the notable exception of hematopoietic cells ¹⁰. Still, in solid tissues, gross genotoxic stress frequently does not appear to select for p53 disruption.

Disruption of p53 prevents the permanent proliferative arrest induced by replicational stress

One of the insults that activates p53 is replicational stress, a condition resulting from inhibition of DNA synthesis. Replicational stress can be induced by different insults, such as inhibition of dNTP synthesis, inhibition of replicative polymerases or by DNA damage that impedes progression of replicational forks. All of these insults cause the increased presence of single-stranded DNA, leading to activation of the ATR-CHK1 pathway and cell cycle arrest in S-phase. Although high levels of replicational stress induced by treatment with the ribonucleotide reductase inhibitor hydroxyurea (HU) have been reported to induce p53 protein accumulation and activation ¹¹, cells are arrested in early S-phase independently of p53 status, and prolonged exposure of cells to high levels of replicational stress leads to a p53-independent permanent proliferation arrest ¹². In other words, despite being activated, p53 status does not appear to influence the ultimate fate of cells experiencing high levels of replicational stress. The p53 dependence of cell survival in response to replicational stress might be different in leukemic cells, as blocking ATR-dependent p53 activation in myeloid leukemia cells delays apoptosis induced by acute HU treatment ¹³. However, this study did not examine the impact of p53 status on long-term survival.

We examined the p53 dependence of cellular responses under chronic exposure to mild, clinically relevant levels of replicational stress that slow DNA replication but do not result in immediate proliferation arrest ¹⁴. We found that continuous exposure of non-transformed rodent and human cells to low concentrations of HU, the replicative polymerase inhibitor aphidicolin or the topoisomerase inhibitor etoposide results in progressive inhibition of proliferation leading to complete and permanent proliferation arrest after several population doublings. Surprisingly, and in contrast to the p53-independence of cell cycle arrest in response to high levels of replicational stress, we found that inhibition of p53 allows cells to maintain steady (albeit reduced) proliferation rates for at least several months. We showed that the inhibition of net proliferative expansion in cells with intact p53 was caused by the accumulation of cells that underwent senescence-like permanent proliferation arrest, and disruption of p53 significantly antagonizes this arrest. Although p53 disruption does not abrogate permanent arrest in all cells, a sufficient fraction of cells is capable of maintaining sufficient proliferation to provide for a substantial net expansion of the population. Notably, the ability of p53 disruption to prevent permanent arrest could only be observed within a certain window of replicational stress levels, as proliferative arrest becomes p53 independent under high concentrations of agents inducing replicational stress while sufficiently low stress levels are permissive for continued expansion of wild-type cells (Figure 1). Notably however, this window appears to encompass the levels of replicational stress experienced by cells in vivo. In particular, the concentrations of HU which we find induce p53-dependent senescence arrest coincide with plasma concentrations of HU (100−200 μM) experienced by patients treated for myeloproliferative disorders ¹⁵. While dependence of replicational stress induced arrest on p53 status is not absolute, even when the difference in the induction of senescence per population doubling is relatively minor, this quantitative difference translates into a substantial qualitative advantage over multiple rounds of proliferation.

Figure 1. A window of replicational stress intensity that selects for p53 loss.

At a certain threshold, replicational stress causes wild-type cells to undergo senescent arrest. Disruption of p53 increases this threshold and allows more cells to avoid permanent arrest at stress levels that arrest wild-type cells, thus promoting the expansion of cells with inhibited p53. However, further increases in replicational stress lead to senescent arrest irrespective of p53 status.

p21^CIP1 and RAD51 are critical mediators of permanent arrest induced by HU and etoposide

We found that the CDK inhibitor p21^CIP1 (p21) is a critical p53 target in mediating permanent proliferation arrest in response to chronic mild replicational stress. As previously shown for high level replicational stress ¹¹, low levels of replicational stress resulted in a delayed p53 dependent upregulation of p21 (after three days of HU or aphidicolin treatment). With prolonged replicational stress, disruption of p21 prevented permanent proliferation arrest in a manner indistinguishable from disruption of p53. However, under etoposide treatment, p21 disruption only partially and temporarily avoided senescent arrest.

Stalling of replication forks resulting from replicational stress is expected to increase the chances of replication fork collapse, leading to increased DNA double stranded breaks ^{16, 17}. Given that a single unrepaired DNA double-stranded break is sufficient to kill a cell ¹⁸, the long-term proliferation of p53 mutant cells should require efficient repair of DNA damage, which is even more important given that p53 mutant cells display increased DNA strand breaks under replicational stress ¹⁹. Disruption of p53 upregulates RAD51 activity, leading to dramatic upregulation of homologous recombination ²⁰, a repair pathway that is critical for resolution of replication associated DNA damage ²¹. p53 has also been reported to repress RAD51 transcription ^{22, 23}. We therefore explored the importance of RAD51 in the ability of p53 mutant cells to overcome the arrest induced by replicational stress. Contrary to our expectations, RAD51 upregulation was not required for the ability of p53 mutation to overcome permanent proliferative arrest induced by HU or aphidicolin. In contrast, inhibiting RAD51 expression led to complete and permanent arrest of p53 mutant cells treated with etoposide, which by inhibiting DNA topoisomerase results in replicational stress by induction of DNA single and double stranded breaks ¹⁴. Therefore, p21 is a critical p53 target in mediating senescence in response to replicational stress, and up-regulation of RAD51 by p53 inactivation is essential when replicational stress is associated with increased DNA strand breaks (Figure 2).

Figure 2. Requirements for p21 and RAD51 vary depending on whether replicational stress is associated with DNA strand breaks.

Left: p53 activation during “simple” replicational stress (HU or aphidicolin) results in p21 upregulation, which mediates senescent arrest. DNp53 antagonizes the arrest by blocking p21 upregulation. Right: When replicational stress is associated with substantial DNA strand breaks, such as with etoposide treatment, DNp53 dependent upregulation of RAD51 dependent homologous recombination is also required for resolution of strand breaks and maintenance of long-term proliferation.

While we were able to identify p21 and RAD51 as critical players that mediate p53 dependence of the senescence-like arrest induced by replicational stress, we do not currently understand the mechanism for how prolonged replicational stress leads to senescence nor why a fraction of p53 mutant cells still undergo permanent arrest. A recent publication from the von Zglinici lab suggests a potential mediator of replicational stress induced senescent-like arrest. This group reported that senescence associated by telomere shortening is caused by mitochondrial dysfunction that leads to increased mitochondrial reactive oxygen species (ROS) production ²⁴. Replicational stress has been shown to stimulate ROS production in fission yeast ²⁵ and p53 exhibits pro- or anti-oxidation functions dependent on its activation state ²⁶. Moreover, as upregulation of p21 has been demonstrated to mediate senescence by inducing ROS accumulation ²⁵, and we observe increased ROS levels in HU and aphidicolin treated cells (Marusyk and DeGregori, unpublished observation), further studies are needed to address the role of ROS in replicational induced senescence in mammalian cells.

Notably, p53 disruption does not prevent perturbations of dNTP levels or prevent S-phase slowing (in the short-term) in cells treated with HU or aphidicolin. In contrast, Bcr-Abl expression overcomes the inhibition of proliferation caused by genetic or HU-mediated inhibition of DNA replication in hematopoietic progenitor cells by restoring the rate of S-phase progression ³. Importantly, these conditions of impaired DNA replication dramatically enhance the competitive advantage provided by the expression of Bcr-Abl, leading to markedly increased leukemia rates in mouse models. Thus, while Bcr-Abl appears to alleviate replicational stress, p53 disruption seems to raise the threshold for the induction of senescence with prolonged replicational stress.

Distinctions between replicational stress- and oncogene-induced senescence

It has been proposed that induction of senescence requires acquisition of unrepaired DNA double-stranded breaks, irrespective of the initial stress, and therefore cellular senescence can be regarded as a permanent DNA damage response state ²⁷. Indeed, several labs have recently demonstrated that oncogene-induced senescence is dependent on DNA damage signaling induced by abnormalities in replication, and that inhibition of DNA damage checkpoint signaling can overcome senescence ^28-30. Both oncogene and replicational stress induced senescence involve abnormal DNA replication, with over-replication for the former and impaired replication for the latter. In both cases, p53 inhibition overcomes the senescent arrest, allowing continued proliferation. However, in contrast to oncogene-induced senescence, we found that pharmacological or small hairpin-RNA mediated ablation of checkpoint signaling fails to rescue the arrest, and can even potentiate senescent arrest ¹⁴. This failure to overcome the senescent arrest, together with the absence of detectable γH2A.X foci in most cells permanently arrested by replicational stress, suggests that inhibition of replication associated with replicational stress and the “over-replication” triggered by oncogenic expression induce senescence via different mechanisms.

Replicational stress should select for initiating oncogenic mutations

We believe that the ability of chronic replicational stress to select for oncogenic mutations such as in p53 in non-transformed cells should contribute to cancer initiation in multiple contexts. HU is widely used in the clinic for the treatment of myeloproliferative disorders, AIDS and sickle cell anemia, while etoposide is commonly used to treat cancers. Moreover, any agent that damages DNA during S-phase should also induce replicational stress, and therefore replicational stress is a condition common to almost all chemotherapies. Many chemotherapeutic regiments including etoposide treatment are associated with the induction of secondary malignancies ^{6, 7}. The induction of these secondary malignancies is generally attributed to mutagenic action of DNA damaging agents. However, growth-limiting conditions caused by chemotherapies should also select for initiating oncogenic mutations which confer resistance ^{3-5, 14}. Notably, a diet rich in fruit and vegetables leads to reduced cancer rates ³¹, and a number of micronutrients (including folate) are important for DNA synthesis, suggesting that a poor diet may increase the risk of cancer at least in part by impeding DNA replication (which could select for oncogenic mutations which abrogate permanent arrest). Furthermore, ultraviolet light induced pyrimidine dimers are expected to cause replicational stress, and higher sun exposure in humans has been shown to increase the size and frequency of clones of p53 mutant skin cells ³². The source, duration and intensity of replicational stress, as well as other factors such as the affected cell type, may contribute to the selection for particular oncogenic mutations. Notably, while p53 disruption by itself is not transforming, clonal outgrowth of p53 mutant cells might substantially increase the probability of transformation resulting from oncogenic mutations that stimulate cellular proliferation, such as oncogenic mutations of RAS. p53 imposes a strong tumor suppressive barrier by promoting oncogene-induced senescence ^{28, 29}, and selective outgrowth of a p53 mutant clone would provide fertile soil for oncogenic mutations that would otherwise induce senescence in wild-type cells.

Finally, we wish to highlight the importance of considering biological effects that despite being relatively modest per any given cell generation, lead to dramatic differences over multiple rounds of proliferation. While p53 mutation results in a relatively modest rescue from replicational stress induced senescence per population doubling, over multiple passages this rescue is compounded, leading to dramatic selection for p53 mutation by drugs inducing replicational stress ¹⁴. Although the focus on dramatic short-term effects is certainly experimentally easier, the failure to consider long-term effects on proliferation can lead to misleading conclusions on the selective advantage conferred by an oncogenic mutation. Indeed, analyses of the p53 dependence for short-term survival following genotoxic insults often do not correlate with long-term clonogenic proliferation, the latter of which better correlates with treatment responses in vivo ⁹.

Tumorigenesis is a lengthy evolutionary process that occurs over many rounds of cell divisions. When taking into account that the tumor mass is maintained by only a small fraction of cells and that cell divisions are offset by cell loss, multiple rounds of proliferation should be required for each population doubling. Such extensive proliferation can allow relatively small differences in selective advantage to accumulate, therefore enabling mutations that confer even partial resistance to engender substantial clonal expansion. Chemotherapic treatments span months to years, providing ample time to select for mutations that provide even partial resistance. Moreover, the cytotoxic effects of chemotherapies not only select for resistance mutations, but also substantially increase cell turnover, including in non-tumor tissues. The resulting compensatory proliferation should promote the expansion of clones with mutations conferring partial chemotherapeutic resistance, possibly leading to fixation of an oncogenic mutation in a population of normal cells, which could contribute to the secondary malignancies resulting from chemotherapeutic and radiation treatments.

Despite tremendous progress in recent decades, our current understanding of tumorigenesis is certainly far from complete. While most research on cancer initiation is focused on the characterization of oncogenic mutations and the intracellular pathways disrupted during tumorigenesis, we believe that understanding the selective forces that shape tumor evolution is equally important. Replicational stress may represent a potent, and perhaps common, mechanism that selects for p53 loss and other oncogenic mutations. Moreover, attention to small, qualitative differences that over multiple rounds of proliferation translate into qualitative differences in selective advantage might be essential for figuring out the puzzle of tumor evolution.

Abbreviations

HU

hydroxyurea

ROS

reactive oxygen species