Introduction
Reactive oxygen species (ROS) serve many cellular functions; for example, second messenger, anti-bacterial agent, mutagen, aging-accelerant and growth stimulant. With regard to neoplasia, the view has generally been that ROS cause cancer through a number of mechanisms, including the induction of DNA damage and alteration of intracellular signaling1. A provocative new report from Takahashi et al. on page 1291 of this issue2, however, suggests that ROS have an unexpected role in inducing and maintaining senescence-induced tumour suppression.
Senescence is a specialized form of terminal differentiation that is usually irreversible and induced by a number of stimuli associated with neoplastic growth, such as oncogene activation and telomere dysfunction. In particular, ROS is known to induce senescence in a number of systems, although the in vivo significance is unclear and the mechanism is not well understood. Although well-described in cultured cells for over four decades, the role of senescence in intact organisms had, until recently, been controversial. Several reports (reviewed in ref. 3), have demonstrated that senescence is not merely an artifact of cell culture but occurs in vivo to limit tumour growth, even in very young mice and humans. Perhaps the most compelling demonstration of this occurs in dysplastic naevi — a pigmented skin lesion thought, in some cases, to be a precursor of malignant melanoma. Unexpectedly, a high frequency of naevi contain signature oncogenic mutations associated with melanoma, yet such naevi do not proliferate and instead exhibit classic features of senescence4, 5. Therefore, although most of us have these common cutaneous lesions, even as children, only an unlucky few will ever develop melanoma because these would-be cancers are usually checked by senescence.
A vexing problem for the field, however, has been determining how senescence, which occurs in many distinct cell types in response to a wide variety of cues, is different in biochemical terms from reversible forms of growth arrest. For example, although senescence is closely associated with activation of the p16INK4a tumour suppressor gene, it has never been clear why other cell-cycle inhibitors, such as p18INK4c, are not. One model proposes that senescence results from a prolonged and potent anti-proliferative signal (for example, as conveyed by p16INK4a or p53 activation) coupled with other ill-defined senescence-promoting stimuli. The nature of these additional stimuli has been unclear, but has been suggested to be forms of 'cellular stress', including ROS. It is this question that the study by Takahashi et al. addresses.
To understand the mechanics of senescence, the authors used an immortalized human fibroblast line containing a temperature-sensitive mutant of the SV40 large T-antigen that functions by binding and inactivating both the Rb and p53 tumour suppressor proteins at permissive temperature. Normally, Rb binds to E2F transcription factors to prevent the induction of genes required for the completion of the cell cycle, whereas the p53 transcription factor induces a number of genes that block proliferation. The activation of p53 and/or Rb is required for senescence in this system, and therefore growth at the permissive temperature blocks the induction of senescence. This system also allows for the reactivation of these anti-proliferative pathways when the cells are switched to the restrictive temperature, which inactivates the large T-antigen, activates Rb–p53 and results in senescence. The authors showed that when these cells were grown under standard conditions, including media with high-serum supplementation, they senesced rapidly after being switched to the restrictive temperature. Moreover, once senescence was established, these cells were not able to re-enter the cell cycle when switched to the permissive temperature. Therefore, in agreement with previously published data6, 7, established senescence of these human cells does not require persistent p53 or Rb activity, and is extremely difficult, if not impossible, to reverse.
The authors next examined what features of cell culture were able to cooperate with Rb–p53 activation to induce senescence. When the large T-antigen was inactivated in cells cultured in media with reduced serum supplementation, cells ceased proliferation but did not become senescent. Instead, these cells re-entered the cell cycle normally when they were switched back to the permissive temperature. This result suggests that senescence requires both a potent anti-proliferative signal (for example, p16INK4a and Rb activation), and the presumably strong mitogenic signals associated with culture in high serum media (Fig. 1). This finding is consistent with a previous study that showed that activation of the ERK–MAPK pathway, a strong proliferative signal in many cell types, is necessary for the establishment of senescence in a related cell-culture model8. Establishing senescence also requires that this period of conflict lasts at least a few days6, during which time widespread and characteristic changes in chromatin architecture occur, associated with the production of senescence-associated heterochromatic foci9.
Figure 1: Senescence requires a mitogenic signal under conditions of cellular stress.
During low-stress conditions, mitogens inactivate Rb and therefore activate E2F to induce proliferation. The authors suggest that as part of its S-phase promoting activity, E2F activation decreases ROS levels by regulating genes involved in ROS production and metabolism. Therefore, under these settings, mitogenic signals are matched by low levels of ROS and proliferation ensues. In conditions of high cellular stress, however, tumour suppressor genes such as p53 and p16INK4a are activated, leading to inhibition of E2F, through Rb. In this setting, mitogenic stimulation is not accompanied by E2F activation and ROS levels accumulate to senescence-promoting levels. The authors suggest that high levels of ROS also induce a PKC-mediated block of cytokinesis as part of the senescence programme.
Full size image (24 KB)Because of prior work linking ROS and senescence in human fibroblasts10, the authors went on to examine the level of ROS in proliferating and senescent cells. As expected, they found that ROS levels were more elevated in cells grown in high serum than in low serum media. Surprisingly, growth arrest resulting from changing to the restrictive temperature markedly increased the levels of ROS. Moreover, fully senescent cells demonstrated even higher levels of ROS, and these increased levels persisted after returning the cells to the permissive temperature. Importantly, the authors could block the establishment of senescence by treating the cells with an oxygen radical scavenger, suggesting that ROS promotes senescence. The authors make a particularly tantalizing speculation as to how this process may work: Rb, through E2F transcription factors, regulates enzymes (for example, MnSOD, GPX and catalase) involved in ROS production and metabolism. Therefore, mitogenic stimulation induces a growth-related production of ROS, while concomitant Rb activation induces an enzymatic production of ROS and impedes ROS degradation, suggesting these cellular signals converge to produce sharply elevated ROS levels (Fig. 1). In agreement with this hypothesis, the authors showed that p16INK4a expression in normal fibroblasts, which presumably does little more than activate Rb, induced ROS levels to almost the same levels as the strong proliferative signals associated with the ectopic expression of oncogenic Ras. Because of convincing data linking ROS and aging, many experts in the field have wondered whether ROS promotes senescence by inducing p16INK4a expression and Rb activation. Although this may be, the authors show that, in fact, p16INK4a–Rb activation can induce ROS.
To understand how ROS may help to enforce the senescent state, the authors focused on protein kinase C. In agreement with previous work11, the authors showed that increased levels of ROS were associated with the formation of the activated, catalytic fragment of PKC
, which, in turn, seems to lead to the further production of ROS, thus establishing a self-sustaining activation loop. As PKC activation has been linked to G2–M cell-cycle arrest12, the authors astutely realized that this finding may explain a long-known feature of senescent cells: on mitogen stimulation, senescent cells will synthesize a modest amount of new DNA, but will not traverse the cell cycle6, 7. Through a variety of genetic and pharmacologic approaches, the authors show that PKC activation induces a sharp reduction in WARTS protein expression (a kinase required for cytokinesis) through an unknown mechanism. Taken together, these data suggest that the PKC-mediated depletion of WARTS is an important component of the senescence program. Although these experiments are convincing, one can imagine other mechanisms whereby high levels of ROS may contribute to senescence, and it could be argued that other likely effects of elevated levels of ROS in senescent cells are not considered by this model.
The findings of Takahashi et al. seem to define another function for ROS — enforcer of senescence. Given the evidence that ROS promote mammalian aging, this conclusion will be satisfying to researchers in gerontology who believe senescence contributes to aging, as this work better explains the known links between ROS and senescence. The widespread belief that anti-oxidants may retard mammalian aging is not challenged by these data. However, this is not the case for the belief that scavengers of ROS (for example, anti-oxidants) should prevent cancer. In contrast to the well-described tumour-promoting activities of ROS, this work suggests a means whereby ROS significantly participate in an anti-cancer mechanism of undisputed importance. Human clinical-trial data suggests the authors may be on to something: two large, blind, randomized trials of supplementation with beta-carotene, a moderate anti-oxidant, have been conducted in patients at increased risk for cancer (current and former smokers), and both yielded the same result — beta-carotene supplementation increased lung cancer risk13, 14. Although there are other tenable explanations for these findings15, the notion that ROS contributes to tumour suppression as suggested by Takahashi et al. may be a potential reason as to why extended anti-oxidant supplementation does not seem effective in cancer chemoprevention.
