Monday, January 11, 2016

Food-Fight Alert!! Is cancer bad luck or environment? Part I: the basic issues

Not long ago Vogelstein and Tomasetti stirred the pot by suggesting that most cancer was due to the bad luck of inherent mutational events in cell duplication, rather than to exposure to environmental agents.  We wrote a pair of posts on this at the time. Of course, we know that many environmental factors, such as ionizing radiation and smoking, contribute causally to cancer because (1) they are known mutagens, and (2) there are dose or exposure relationships with subsequent cancer incidence. However, most known or suspected environmental exposures do not change cancer risk very much or if they do it is difficult to estimate or even prove the effect.  For the purposes of this post we'll simplify things and assume that what transforms normal cells into cancer cells is genetic mutations; though causation isn't always so straightforward, that won't change our basic storyline here.

Vogelstein and Tomasetti upset the environmental epidemiologists' apple cart by using some statistical analysis of cancer risks related, essentially, to the number of cells at risk, their normal time of renewal by cell division, and age (time as correlated with number of cell divisions).  Again simplifying, the number of at-risk actively dividing cells is correlated with the risk of cancer, as a function of age (reflecting time for cell mutational events), and with a couple of major exceptions like smoking, this result did not require including data on exposure to known mutagens.  V and T suggested that the inherently imperfect process of DNA replication in cell division could, in itself, account for the age- and tissue-specific patterns of cancer.  V and T estimated that except for the clear cases like smoking, a large fraction of cancers were not 'environmental' in the primary causal sense, but were just due, as they said, to bad luck: the wrong set of mutations occurring in some line of body cells due to inherent mutation when DNA is copied before cell division, and not detected or corrected by the cell.  Their point was that, excepting some clear-cut environmental risks such as ionizing and ultraviolet radiation and smoking, cancer can't be prevented by life-style changes, because its occurrence is largely due to the inherent mutations arising from imperfect DNA replication.

Boy, did this cause a stink among environmental epidemiologists!  Now one we think undeniable factor in this food fight is that environmental epidemologists and the schools of public health that support them (or, more accurately, that the epidemiologists support with their grants) would be put out of business if their very long, very large, and very expensive studies of environmental risk (and the huge percent of additional overhead that pays the schools' members meal-tickets) were undercut--and not funded and the money went elsewhere.  In a sense of lost pride, which is always a factor in science because it's run by humans, all that epidemiological work would go to waste, to the chagrin of many, if it was based on misunderstanding the basic nature of the mutagenic and hence carcinogenic processes.

So naturally the V and T explanation has been heavily criticized from within the industry.  But they will also raise the point, and it's a valid one, that we clearly are exposed to many different agents and chemicals that are the result of our culture and not inevitable and are known to cause mutations in cell culture, and these certainly must contribute to cancer risk.  The environmentalists naturally want the bulk of causation to be due to such lifestyle factors because (1) they do exist, and (2) they are preventable at least in principle.  They don't in principle object to the reality that inherent mutations do arise and can contribute to cancer risk, but they assert that most cancer is due to bad behavior rather than bad luck and hence we should concentrate on changing our behavior.

Now in response, a paper in Nature ("Substantial contribution of extrinsic risk factors to cancer development," Wu et al.) provides a statistical analysis of cancer data that is a rebuttal to V and T's assertions.  The authors present various arguments to rebut V and T's assertion that most cancer can be attributed to inherent mutation, and argue instead that external factors account for 70 to 90% of risk.  So there!

In fact, these are a variety of technical arguments, and you can judge which seem more persuasive (many blog and other commentaries are also available as this question hits home to important issues--including vested interests).  But nobody can credibly deny that both environment and inherent DNA replication errors are involved.  DNA replication is demonstrably subject to uncorrected mutational change, and that (for example) is what has largely driven evolution--unless epidemiologists want to argue that for all species in history, lifestyle factors were the major mutagens, which is plausible but very hard to prove in any credible sense.  

At the same time, environmental agents do include mutational effects of various sorts and higher doses generally mean more mutations and higher risk.  So the gist of the legitimate argument (besides professional pride or territoriality and preservation of public health's mega-studies) is really the relative importance of environment vs inherent processes.  The territoriality component of this is reminiscent of the angry assertion among geneticists, about 30 years ago, that environmental epidemiologists and their very expensive studies were soaking up all the money so geneticists couldn't get much of it.  That is one reason geneticists were so delighted when cheap genome sequencing and genetic epidemiological studies (like GWAS) came along, promising to solve problems that environmental epidemiology wasn't answering--to show that it's all in the genes (and so that's where the funding should go).  

But back to basic biology 
Cells in each of our tissues have their own life history.  Many or most tissues are comprised of specialized stem cells that divide and one of the daughter cells differentiates into a mature cell of that tissue type.  This is how, for example, the actively secreting or absorbing cells in the gut are produced and replaced during life.  Various circumstances inherent and environmentally derived can affect the rate of such cell division. Stimulating division is not the same as being a direct mutagen, but there is a confounding because more cell division means more inherent mutational accumulation.  That is, an environmental component can increase risk without being a mutagen and the mutation is due to inherent DNA replication error.  Cell division rates among our different tissues vary quite a lot, as some tissues are continually renewing during life, others less so, some renew under specific circumstances (e.g., pregnancy or hormonal cycles), and so on.

As we age, cell divisions slow down, also in patterned ways.  So mutations will accumulate more slowly and they may be less likely to cause an affected cell to divide rapidly.  After menopause, breast cells slow or stop dividing.  Other cells, as in the gut or other organs, may still divide, but less often.  Since mutation, whether caused by bad luck or by mutagenic agents, affects cells when they divide and copy their DNA, mutation rates and hence cancer rates often slow with advancing age.  So the rate of cancer incidence is age-specific as well as related to the size of organs and lifestyle stimulates to growth or mutation.  These are at least a general characteristics of cancer epidemiology.

It would be very surprising if there were no age-related aspect to cancer (as there is with most degenerative disease).  The absolute risk might diminish with lower exposure to environmental mutagens or mitogens, but the replicability and international consistency of basic patterns suggests inherent cytological etiology.  It does not, of course, in any sense rule out environmental factors working in concert with normal tissue activity, so that as noted above it's not easy to isolate environment from inherent causes.

Wu et al.'s analysis makes many assumptions, the data (on exposures and cell-counts) are suspect in many ways, and it is difficult to accept that any particular analysis is definitive.  And in any case, since both types of causation are clearly at work, where is the importance of the particular percentages of risk due to each?  Clearly strong avoidable risks should be avoided, but clearly we should not chase down every miniscule risk or complex unavoidable lifestyle aspect, when we know inherent mutations arise and we have a lot of important diseases to try to understand better, not just cancer.

Given this, and without discussing the fine points of the statistical arguments, the obvious bottom line that both camps agree on is that both inherent and environmental mutagenic factors contribute to cancer risk. However, having summarized these points generally, we would like to make a more subtle point about this, that in a sense shows how senseless the argument is (except for the money that's at stake). As we've noted before, if you take into account the age-dependency of risk of diseases of this sort, and the competing causes that are there to take us away, both sides in this food fight come away with egg on their face.  We'll explain what we mean, tomorrow.


Peter said...

As far as I can see, this is a semantic debate and not a scientific one!

Vogelstein and Tomasetti ask the question of why some tissues become cancerous and other don't, and find that it's in large part due to the number of stem cell divisions. Therefore, most types of cancer are due to random factors. The other papers ask the question of why some people get cancer while others don't, and find that it's in large part due to extrinsic risk factors. Therefore most cases of cancer are due to environmental risk.

To illustrate the difference, let's imagine a model animal with three organs: the squish, the blart and the gronk. Cancers of the squish and blart arise entirely at random during cell division. The blart undergoes 10x more divisions than the squish, and so squish/blart cancers have a prevalence of 0.1% and 1% respectively. Gronk cancer is different, and is caused by eating cheese. 10% of the population eats cheese, and so 10% of the population gets cancer of the gronk.

In this model, it is true to say that 2/3 of all types of cancer arise purely by chance. It is also true to say that ~90% of all cases of cancer are caused by environmental factors.

Fundamentally, these two studies asked different questions. The first question is mainly interesting for the purpose of understanding the molecular biology of cell division. The second is mainly interesting for the purpose of screening for and treating or preventing cancer cases. Setting it up as some kind of massive argument is mainly interesting for the purpose of selling newspapers.

Ken Weiss said...

Good points. We have more to say about this tomorrow, on somewhat different aspects, but nothing that undermines what you say. This is a territorial dispute to a great extent, with semantics at its root. There are others who argue that cancer is not a mutational disease but due to other factors. The either-or nature of the arguments in a way also demonstrate your point. In a sense, also, people I think want external and hence avoidable causes of mortality, with the idea that they can become immortal. There is also the question of 'blame'. Purely random 'causes' assign no blame but do relieve industry etc. of culpability. Genes assign blame to peoples' inherent worth. Environmental causes put the blame on others. All very human, in a sense.

Peter said...

I think the two studies are actually beautifully complementary. The Vogelstein study tells you which types of cancer have a strong environmental component, and therefore what cancers it's worth doing an environmental epidemiological study for.

Since Pancreatic islet cancer and osteosarcoma came up very low in Vogelstein's aERS plot (i.e. risk is no greater than expected from the number of stem cell divisions), it's probably futile to look for environmental triggers. However, since lung cancer and colorectal cancer have high aERS, then it's work working out what the additional risk factors are and how we can reduce them.

Mind you, I think Vogelstein's also wrong in characterising the "risk per cell division" as "random". A huge proportion of carcinogens _work_ by increasing the error rate per cell division, i.e. it's an environmental factor, but which will also preferentially affect the cells that undergo most cell divisions.

Ken Weiss said...

Even mitogens will increase mutation via inherent replication errors, and hence will be environmental but not mutagenic. Vogelstein would probably argue that the location of the mutation is randomly located (which is technically inaccurate but at least in regard to function probably OK to say). One recent hypothesis about retinoblastoma is that the RB mutation leads to errors in methylation and hence gene expression--I have not seen any testing of that idea. I also think, though I have not attempted to scour the papers or the many pro and con comments about the details, V and T were found to have used faulty statistical methods and there are questions about the number of at-risk cell divisions, etc. In that sense, the argument is about the statistical methods as much as the biology. I think that the life history (age patterns) of cancers show relationships to cell division, and in that sense mutagens are intertwined with inherent processes. So calling them 'complementary' may be quite appropriate.