Oncogene (2003) 22, 7028–7033 & 2003 Nature Publishing Group All rights reserved 0950-9232/03 $25.00 www.nature.com/onc

Radiation-induced bystander effects, and models

Carmel Mothersill*,w,1 and Colin Seymour1,2

1Radiation and Environmental Science Center, Dublin Institute of Technology, Kevin Street, Dublin 8, Ireland; 2Saint Luke’s Institute of Research, Highfield Road, Rathgar, Dublin 6, Ireland

Implications for carcinogenesis of -induced by- ability, result in a being ‘initiated’ due to at stander effects are both mechanistic and practical. They a critical site. This can be followed by promotion and include induction of second , perturbations to progression of an essentially clonal population. Radia- tissue social control and induction of genomic instability tion is assumed only to be involved in initiation. The new and delayed or immediate in areas not receiving position that many suggest is necessary in order to a direct deposition of energy. Bystander effects have accommodate both bystander effects and genomic consequences for DNA damage-mutation-cancer initiation instability is that the multiple-step carcinogenic process paradigms of radiation carcinogenesis that provide the is mediated from the start and throughout, by epigenetic mechanistic justification for low-dose risk estimates. If and indirect effects of radiation. A major pillar of this carcinogenesis does not result from directly induced DNA proposed paradigm shift is that bystander effects induced mutations, then the carcinogenic initiation process may by radiation in nonexposed cells actually drive the not simply relate to radiation dose. Modification of the process of genomic instability and that this genomic preclonal state through genetic and epigenetic mechanisms instability both precedes and facilitates the evolution of may occur. To deal with the complexity of these clonal (potentially carcinogenic) mutations. interactions, a ‘chaotic’ or ‘bifurcation’ model invoking In the target theory paradigm, all radiation events autopoietic theory is proposed that could accommodate were contained in the ‘hit’ cell. There was a direct energy both beneficial (hormetic) and harmful effects of radiation deposition leading to DNA damage. In this paradigm, at comparable doses. Carcinogenesis may then be thought only the cell directly exposed to radiation could suffer of as the result of a disturbance of the genetic/epigenetic any damage. The changing paradigm, by including balance occurring within the organ. Ultimate clonality downstream effects occurring in cells not directly may reflect domination due to selection processes rather exposed to radiation, but receiving damage signals from than the initiating damage. irradiated cells, allows cell–cell and cell–matrix commu- Oncogene (2003) 22, 7028–7033. doi:10.1038/sj.onc.1206882 nication to be involved. Implicit in this changing paradigm is a change in target size, making large Keywords: radiation; carcinogenesis; bystander effects; interaction distances and longer times of interaction chaos models; social evolution; autopoietic theory relevant. While the bystander factor(s) has not yet been identified, the existence of this signalling mechanism, at least theoretically, allows the damage signal production resulting from dose deposition to be distinguished from Introduction the cellular response. This introduces a multidimen- sional context to radiation carcinogenesis where spatial Radiation-induced bystander effects have been widely and temporal parameters, and responses of recipient reviewed in the literature (Mothersill and Seymour, 2001; cells are as important as dose. Goldberg and Lehnert, 2002; Lorimore and Wright, The measured end points of cellular bystander 2003; Morgan 2003a, b; Mothersill and Seymour, 2003), response have been mutation, gene induction, micro- and refer to effects detected in cells that were not directly nuclei formation, cell transformation and cell killing (see ‘hit’ by an ionizing radiation track. This paper is reviews cited above). These are similar to those intended to discuss the implications of these effects in measured for genomic instability and again demonstrate the carcinogenic process and to discuss whether they the close association between the two phenomena. The require a paradigm shift in the radiation carcinogenesis demonstrated existence of a high frequency of nonclonal field from the existing position involving, essentially, the mutations in postirradiation populations of cells, which hit (target) theory, where DNA damage is seen as is characteristic of genomic instability alters the resulting from random hits to the chemical structure of established single-cell, multiple-step paradigm of carci- DNA which can, with a definable dose-related prob- nogenesis towards a model postulating tissue interac- tions and responses interfacing with instability to facilitate cancer formation. This is a vastly more *Correspondence: Dr C Mothersill; E-mail: [email protected] w - Current address: McMaster University, Dept Medical Physics and complex situation than the simple ‘dose DNA Applied Radiation Science, Hamilton, Ontario, Canada break-mutation’ theory. Bystander effects and carcinogenesis C Mothersill and C Seymour 7029 The single cell mutation theory of radiation-induced 1992; Loeb, 2001). Thus it is reasonable to postulate cancer suggests that the tumour arises through stages of that the initial change is epigenetically induced and is initiation, promotion and progression. The initial sustained by cell–cell communication or environmental mutation caused by radiation is a rare and stochastic factors. Epigenetic effects can be defined as causing event. In contrast, a genomic instability theory would mitotically heritable changes in gene expression that are suggest a common event cause, with gene mutation not caused by an alteration of the DNA sequence. occurring frequently, but being controlled through cell– Methylation changes are an example and there is cell communication-mediated processes involving by- evidence of methylation changes following exposure of stander signals and responses. Only if the epigenetic cells to bystander signals (Grandjean et al., submitted). control breaks down does progression to cancer occur. A distinction between an epigenetic mechanism and a This model also allows a reversibility after radiation genetic mechanism is that epigenetic changes are exposure, not available in the initiation, promotion and chemically reversible. These epigenetic influences can progression model. occur from different cells or tissues, or can occur after a Bystander effects and genomic instability are both lapse in time from the original radiation event. induced at very low doses. To use environmental In Figure 1, the options that a cell population exposed toxicology jargon, no doses corresponding to no either to direct irradiation or bystander signals could observable effect level (NOEL) or lowest observable face are shown schematically. Rather than adhere to an effect level (LOEL) have yet been defined. The lowest initiation, promotion and progression model which is so doses used (one track to one cell in a strongly identified with the DNA damage-centred population or 2 mGy gamma radiation to a population paradigm, we have conceptualized induction, fixation of cells) caused the same amount of genomic instability and expression stages in the evolution of a cancer where or other bystander end point as doses that were orders tissue processes can influence outcome both before and of magnitude higher (Seymour and Mothersill, 2000; after fixation of DNA damage. Kadhim et al., 2001; Ostreicher et al., 2003) Tradition- There is a wealth of evidence related to the fact that ally, radiation protection authorities have combined many tumours do have a clonal origin (or perhaps more target theory with the single cell initiation, promotion accurately an early clonal mutation) and a distinct and progression model to predict a linear reduction of identifiable genetic fingerprint (Noguchi et al., 1995; carcinogenic effect with decreasing radiation dose. Bedi et al., 1996; Werness et al., 1997). However, it is Bystander and genomic instability effects introduce a generally impossible to know what level of variability nonlinearity into the low-dose area, where the response existed before the emergence of clonality and whether of the tissue/cell population rather than the doses will this is a secondary consequence of selection events determine the fate of the cell and ultimately, the occurring very early in the evolution of the altered state. organism. Given the complexity and variety of cellular It is our hypothesis that once a clonal event has interactions and the randomness of radiation damage, occurred, that is, a direct heritable DNA mutation, whether direct or bystander-induced, we have consi- then it is ‘committed’ or ‘fixed’ and cannot be reversed dered whether this response is predictable or whether it through epigenetic mechanisms. In terms of a general is essentially chaotic. chaotic model, the myriad of constant chemical inter- actions within a cell are essentially chaotic. The change in DNA status (through mutation) could then be

Clonality of tumours, bystander effects and genomic instability Options and choices for exposed cell populations

There is evidence that bystander signals can induce Affected cells I induction genomic instability both in vivo (Watson et al., 2000; Live with Lorimore et al., 2001) and in vitro (Seymour and die damage Mothersill, 1997; Lorimore et al., 1998). This instability recover recover is very frequent and nonclonal, which means that a badly well daughter cell can show a mutation not transmitted by perpetuate LIFE the parent cell. This can be distinguished from other die damage II fixation forms of nonclonal mutation or mutator phenotype by lethal non-lethal the lack of any common predisposing mutation within mutations mutations the lineage. For example, mutations in ATM, p53 or life shortening transforming components of DNA repair produce nonclonal muta- DEATH mutations mutations tions and chromosome aberrations, but these offspring killed evade continue to express their original inherited mutations, by surveillance surveillance III expression and so remain consistent with a clonal initiation, senesce altered state promotion and progression model. By contrast, radia- of cancer tion-induced genomic instability occurs at too high a Figure 1 Schematic representation of the options and choices frequency to result from a mutator phenotype mutation available to an irradiated or otherwise damaged cell at each of directly induced by radiation (Seymour and Mothersill, several proposed decision points postexposure

Oncogene Bystander effects and carcinogenesis C Mothersill and C Seymour 7030 considered a bifurcation point, after which the funda- probability of survival? In a temporal context, impor- mental dynamics of the system have altered. Before this tant questions for the whole organism concern life span bifurcation, there is a possibility that cellular interac- after exposure and survival to reproductive age. tions could have different outcomes. Clonality could, At the level where cancer initiation occurs, tissue therefore, also fix and commit responses to bystander context is clearly important in determining whether and other environmental signals. damaged cells remain capable of reproducing or whether they are directed into one or more ‘stop’ pathways such as , terminal differentiation or necrosis, which Bystander effects, carcinogenesis and the meaning of put them ‘beyond reach’ for inclusion in carcinogenic ‘dose’ processes. It is not clear whether there is one or many bystander Since bystander effects have been shown to be indepen- factors, but data from the authors’ laboratory suggest a dent of dose and to be widely induced in the population common initial response mediated by a calcium pulse to of cells exposed to very low doses, this would appear to the signal from the irradiated cell population and that remove any possibility of a simple linear relationship the downstream transduction of this pulse determines between the amount of radiation received by a system which way the bystander cells respond (Lyng et al., and the probability of cancer arising at low radiation 2000, 2002). Examination of populations of cells doses. Rather, it appears that ‘outcome’ is determined receiving culture media from irradiated cells suggests by the response to a signal that radiation is in the system. that all cells in the population produce the calcium Clearly therefore, the outcome is determined by many pulse, but not all go on to die (Maguire, Lyng, factors including underlying genetic predisposition, Mothersill, Seymour, manuscript in preparation). A temporal and positional factors (such as cell cycle, key area of research now is to determine the mechanisms differentiation state, age, distance of the cell from blood by which the live or die choice is made and to quantify supply, etc.) and lifestyle factors (such as nutritional numbers of cells taking different pathways at a range of status, smoking, antioxidants, chemical pollutants) that points where choice is possible. It is suggested that dose modulate response to damage. Depending on the or, more accurately, damage thresholds will exist where hierarchical level under consideration, these modulating the predominant choices suddenly change. factors may be more or less relevant. Possible major bifurcation points at different organizational levels from cell to tumour could include the following – a cell can The model live or die, a tissue can function or not, an organism can survive or not. There are different questions and The model being developed assumes that the basic priorities at each level. Possible questions of relevance cellular state is chaotic. Chaotic models have been to radiation carcinogenesis that might feed data into a proposed in many areas (see e.g. Anderson et al., 2001; bifurcation model could include for example questions Rambihar et al., 2000; Sedivy et al., 2000;). Mathema- concerning spatial context such as what are the tical chaos models are complex and difficult to use, but neighbouring cells doing and what is their relative the theory of chaos can be used to gain insights amount of damage? Is accessible information action- into complex behaviour. Within chaos theory, the same able? Cellular energy budgets and societal responsibil- cause will not always produce the same effect. The chaos ities and possibilities would be important here. In the in the system may be dissipated through a bifurcation context of time scale over which radiation responses event, which effectively changes the state of that system. occur, normal life span of the cell/tissue pre-exposure As the chaos in the system increases, the amount of and in the potentially changed life span in the new disturbance needed to cause a bifurcation decreases. postexposure situation are important factors. Can the The amount of chaos within the system is constantly tissue structure be maintained pending repopulation, changing (primarily through cell–cell communication), after which damaged cells could die without compro- until a bifurcation event occurs. In the last few mising tissue function? Is the necessary energy for repair sentences, ‘genomic instability’ might well be substituted available before cell division becomes essential? Regard- for the word ‘chaos’. A schematic to illustrate this idea ing the context of a cell in its environment at the issue is shown in Figure 2. A prediction of this would be level, the important factors would be whether the that reducing chaos in the system through any damage was widespread and dangerous, whether it is means will reduce effects (activity), increasing it will structural or functional and what the minimum cell increase effects (good or bad) and that the ability to number is to retain necessary tissue function and to modify amount of chaos may be important. Regarding repopulate. Questions about prioritizing energy use and the importance of growth arrest, apoptosis (and of sharing metabolites at the tissue level could impact here. course, proliferation and escape from growth control), At the organism level the questions might be quite which are well-known effects of exposure to low doses, different. For example, is it possible to live with reduced these would be regarded in the model as major function of this organ/tissue? Can a duplicate organ or consequences of bifurcation choices. The point being different organ compensate? Can enzyme production/ that because of the complexity of the factors governing immune response or other systemic systems aid recov- the choice, that choice is only predictable at the ery? Can diet or external (lifestyle) factors change the population or statistical probability level and not at

Oncogene Bystander effects and carcinogenesis C Mothersill and C Seymour 7031 Basic chaos model of low dose radiation action A chaotic model also implies a nonlinearity of effect with dose, with the possibility of small disturbances causing large effects. The amount of disturbance a cell could tolerate would depend not only on how chaotic

Cell communicatio Cell the individual cell was but also on how chaotic its Area of surrounding environment was. A small disturbance iterative could modify not only an individual cell but also change in the system the in change the surrounding tissue. This ‘epigenetic instability’ could then cause a ripple effect, possibly changing the states of Area of other cells. This bifurcation model does not entirely bifurcation eliminate any dose–effect relationships. All it states is

n that at low doses (of radiation or other carcinogens), the effect is unpredictable, whereas at higher doses it

becomes more predictable. Actual evidence that early Amount of chaos chaos of Amount effects of low-dose radiation act chaotically will require accumulation of data demonstrating the probability of Amount of disturbance multiple different end points occurring in exposed cells Figure 2 Interplay of proposed factors involved in a chaotic or cells signalled by exposed cells. Every reaction within model; of cellular function postirradiation the cells will in turn affect other reactions, and it will be deterministic, but will be too complex to model in other ways. Current methods for measuring radiation effects the individual cell or organism level (much like the at low doses may be too insensitive to demonstrate ‘heads or tails’ choice). chaotic patterns that tend to get buried in statistical This model allows for feedback loops or iteration analysis, but the application of proteomics or positional between the genetic and epigenetic constituents of the analyses using a combination of microbeam technology cell, and may be generally described by autopoietic and confocal lambda stacking offer exciting new theory (Maturana, 1975; Varela, 1979; Maturana and possibilities. Varela, 1980). Although this theory was first developed to describe social system interactions, it can be adapted to biological systems. Autopoiesis (from the Greek, Bystander effects and radiation risk auto, meaning ‘self’ and ‘poiesis’ meaning ‘creation’) effectively describes a system as occupying a specific The implications of a bifurcation model for the topological domain and continually regenerating itself carcinogenic risk of radiation are relevant at low from the network of processes that produced it. radiation doses. Essentially the model predicts a wide Essential to the theory is that the cell is constantly variety of cellular responses. Not all of these will be redefining itself as circumstances change, and so there- detrimental to the cell or organism. They may be fore constant communication between the nucleus and beneficial or have no effect. Equally they may be cytoplasm, and other cells, are involved in this model. detrimental and/or carcinogenic. They may have one This would, for example, be similar to the changes outcome at one hierarchical level (for example a cell may occurring during proliferation, differentiation or apop- die), but quite another at a different level (the organism tosis. Both a healthy life and disease progression are avoids death by cancer). Current experimental ap- viewed as a constantly reiterative reaction between proaches in toxicological sciences tend to measure one genetic and epigenetic forces in the tissue. This model end point in a defined assay and determine ‘dose would allow modification of the initial carcinogenic response’ this way. Experiments also tend to measure process through both genetic and epigenetic mechan- damage or harm rather than beneficial or absent effects. isms. The general concept of this model can be regarded Since the proposed model predicts a chaotic response, as simplistic given the complexity of the cell, but it does the effect may not be directly proportionate to the dose allow for epigenetic effects to be continually modified by and may vary due to factors not attributable to continuing intra- and extracellular events. ‘experimental error’. Risk of carcinogenesis at low This theory is, therefore, similar to the ‘stroma–tissue radiation doses is therefore very difficult to estimate. interactions’ model proposed by Barcellos-Hoff and At a certain radiation dose (the threshold of tolerance Brooks (2001), which suggests a role for TGFb in the for the cell or system), the energy or substance entering transduction of the bystander effect. The usefulness of the cell or system will start to cause a predictable their model is that it distinguishes between two types of disturbance (i.e. death). At higher doses, the chaos is communication. One type of communication is an essentially over-ridden and the system becomes more exchange of information only; the other type of predictably deterministic. This is not to imply that communication results in a change to both the sender nontargeted effects have no relevance at high doses, and the recipient. If this concept is adopted into the simply that as the nontargeted effects are dose bifurcation or chaos model, then some cellular interac- independent, their relative importance as a proportion tions will alter the chaos within the system, while others of the total effect tends to decrease as the dose increases. will cause bifurcation and change the system. The point where both bystander-induced death and

Oncogene Bystander effects and carcinogenesis C Mothersill and C Seymour 7032 directly induced death contribute equally to total cell itself determined by the complex interplay of genetic and death as been shown in one study to be around 2 Gy epigenetic factors, bystander effectively (Seymour and Mothersill, 2000). Clearly, more studies precludes the estimation of risk by extrapolation from with different cell lines and using end points more high doses. They also make use of epidemiology from appropriate for carcinogenesis would be very useful. other population cohorts of dubious value, unless these However, based on existing data it appears that the are genetically and epigenetically identical and exposed net result is that at low doses the carcinogenic effect of to similar radiation situations. radiation is determined by the response of the cell to Finally, it is important to consider why bystander and injury or perceived and not by the injury (dose) per se. delayed effects occur postirradiation. Since cell death is Carcinogenesis thus becomes the result of a disturbance a very common delayed response (Limoli et al., 1998), of the genetic/epigenetic balance occurring within the they might be indicative of some evolutionary mechan- responding organ. While much of this is conjecture, ism for minimizing adverse consequences to damage by there is enough evidence at least to put this forward as a coordinating or inducing responses at the tissue level. speculative theory. For this model it does not matter The minority of exposed individuals that develop cancer whether the direct or indirect effects of radiation are might then be indicative of polymorphic variation in considered. The bystander effect widens the area damage response pathways. Whatever the reason for previously thought to be exposed to radiation. It also bystander effects, they support models that would increases the range of responses and end points that postulate a threshold for radiation effects below which need to be considered and the time scale over which they prediction of outcome was not possible. may be relevant.

Conclusions Bystander effects and individual susceptibility to cancer The authors are aware that it is not yet established The bystander effect will effectively increase the size of whether there is one bystander factor or several, and the ‘at risk’ population of cells for radiation risks, but that the initial response may vary from cell to cell and the bifurcation model predicts that the risk will not organ to organ. The attractiveness of the suggested automatically increase with dose, and that below a ‘bifurcation’ approach is, however, that it allows the threshold the effect may equally well be beneficial or apparent contradictory effects of low radiation doses to harmful. The consequences of these approaches both for be accommodated. It also provides a way of incorpora- carcinogenesis and radiation risk management are that ting bystander effects and genomic instability into a new at low doses, the response to radiation becomes model. This model reflects a true paradigm shift from important, in terms of how much bystander factor is DNA-centred radiobiology (and cancer biology) to a produced, how far it travels and the effect it has. This more organic system where final outcome is not just a means that individual susceptibility (whether genetic or function of initial damage, but encompasses the whole due to lifestyle) becomes more relevant than the dose rich variety of available responses, which a system (cell, received. We already know that there is a variation tissue or organism) can use to achieve the best between humans and that there are probably genetic compromise between short-term survival and long-term factors involved in the production of bystander signals sustainability. (Watson et al., 2000; Mothersill et al., 2001). Variation in the extent and type of bystander effect has also been Acknowledgements seen in cell lines (Mothersill et al., 2002). This again We acknowledge grant support from the Science Foundation suggests that response and not dose is the dominant of Ireland, Saint Luke’s Hospital and from Cancer Research determinant of outcome at low doses. As response is Ireland.

References

Anderson GR, Stoler DL and Brenner BM. (2001). Bioessays, Limoli CL, Hartmann A, Shephard L, Yang CR, Boothman 23, 1037–1046. DA, Bartholomew J and Morgan WF. (1998). Cancer Res., Barcellos-Hoff MH and Brooks AL. (2001). Radiat. Res., 156, 58, 3712–3718. 618–627. Loeb LA. (2001). Cancer Res., 61, 3230–3239. Bedi GC, Westra WH, Gabrielson E, Koch W and Sidransky Lorimore SA, Kadhim MA, Pocock DA, Papworth D, Stevens D. (1996). Cancer Res., 56, 2484–2487. DL, Goodhead DT and Wright EG. (1998). Proc. Natl. Goldberg Z and Lehnert BE. (2002). Int. J. Oncol., 21, Acad. Sci. USA, 95, 5730–5733. 337–349. Lorimore SA, Coates PJ, Scobie GE, Milne G and Wright EG. Grandjean V, Mukherjee R, Kapoor A, Keyes E, Seymour (2001). Oncogene, 20, 7085–7095. CB, Mothersill C and Schofield PN: Radiat. Res. submitted. Lorimore SA and Wright EG. (2003). Int. J. Radiat. Biol., 79, Kadhim MA, Marsden SJ, Malcolmson AM, Folkard M, 15–25. Goodhead DT, Prise KM and Michael BD. (2001). Radiat. Lyng FM, Seymour CB and Mothersill C. (2000). Br. J. Res., 155, 122–126. Cancer, 83, 1223–1230.

Oncogene Bystander effects and carcinogenesis C Mothersill and C Seymour 7033 Lyng FM, Seymour CB and Mothersill C. (2002). Radiat. Res., Rambihar VS and Baum M. (2000). A mathematical (chaos 157, 365–370. and complexity)theory of medicine, health and disease: Maturana HR. (1975). Int. J. Mach. Stud., 7, 3–34. refiguring medical thought. In: A New Chaos Based Medicine Maturana HR and Varela FJ. (1980). Autopoiesis and Beyond 2000; The Response to Evidence. Vashna Publica- Cognition: The Realization of the Living, Vol. 42. Boston tions: Toronto. Studies in the Philosophy of Science, Dordrecht. Sedivy R, Thurner S, Kastner J and Maurer G. (2000). Wien. Morgan WF. (2003a). Radiat. Res., 159, 567–580. Klin. Wochenschr., 112, 177–183. Morgan WF. (2003b). Radiat. Res., 159, 581–596. Seymour C and Mothersill C. (1992). Mutat. Res., 267, Mothersill C and Seymour CB. (2001). Radiat. Res., 155, 19–30. 759–767. Seymour CB and Mothersill C. (1997). Radiat. Oncol. Invest., Mothersill C and Seymour CB. (2003). J. Exp. Haematol.. 5, 106–110. Mothersill C, Seymour CB and Joiner MC. (2002). Radiat. Seymour CB and Mothersill C. (2000). Radiat. Res., 153, Res., 157, 526–532. 508–511. Mothersill C, Wright EG, Rea DM, Murphy D, Lorimore S, Varela FJ. (1979). Principles of Biological Autonomy. Elsevier: Seymour CB and O’Malley K. (2001). Carcinogenesis, 22, New York. 1465–1471. Watson GE, Lorimore SA, MacDonald DA and Wright EG. Noguchi S, Aihara T, Koyama H, Motomura K, Inaji H and (2000). Cancer Res., 60, 5608–5611. Imaoka S. (1995). Cancer Lett., 90, 57–63. Werness BA, DiCioccio RA and Piver MS. (1997). Hum. Ostreicher J, Prise KM, Michael BD, Vogt J, Butz T and Pathol., 28, 626–630. Tanner JM. (2003). Strahlenther Onkol., 179, 69–77.

Oncogene