Aaargh! Physicists! Again!

A while back, two physicists, Paul Davies and Charles Lineweaver, announced their explanation for cancer with a novel theory, which is theirs, that cancers are atavisms recapitulating in a Haeckelian reverse double backflip their premetazoan ancestry. They seemed very proud of their idea.

I was aghast, as you might guess. They even claimed that human embryos go through a fish/amphibian stage with gills, webbed feet, and tails in a pattern of Haeckelian development. They do not understand evolution, development, or cancer, facts that were apparent even in the absence of their admission that they had no prior knowledge, and it was freaking embarrassing to see two smart guys with a measure of legitimate prestige in their own specialties charging off into another discipline with such crackpot notions.

Now they've done it again, repeating the same claims all over again. And worse, they've now published it in the journal Physical Biology, under the title "Cancer tumors as Metazoa 1.0: tapping genes of ancient ancestors".

The genes of cellular cooperation that evolved with multicellularity about a billion years ago are the same genes that malfunction to cause cancer. We hypothesize that cancer is an atavistic condition that occurs when genetic or epigenetic malfunction unlocks an ancient ‘toolkit’ of pre-existing adaptations, re-establishing the dominance of an earlier layer of genes that controlled loose-knit colonies of only partially differentiated cells, similar to tumors. The existence of such a toolkit implies that the progress of the neoplasm in the host organism differs distinctively from normal Darwinian evolution. Comparative genomics and the phylogeny of basal metazoans, opisthokonta and basal multicellular eukaryotes should help identify the relevant genes and yield the order in which they evolved. This order will be a rough guide to the reverse order in which cancer develops, as mutations disrupt the genes of cellular cooperation. Our proposal is consistent with current understanding of cancer and explains the paradoxical rapidity with which cancer acquires a suite of mutually-supportive complex abilities. Finally we make several predictions and suggest ways to test this model.

Oh. My. Gob. So…much…wrongness.

I read the paper. I cringed. This blogger read the summaries. He cringed. I think anyone who is the least bit informed about biology will feel the same way…but this crap will continue to get published in legitimate journals because…because…hey, physicist! They must know something we don't! (Well, they do know things we don't, but we know things they don't, and knowledge of cosmology doesn't translate into knowledge of molecular biology. Or vice versa.)

The conceits of the paper are 1) a lack of knowledge about cellular evolution that allows them to posit evidence-free scenarios; 2) peculiar notions about molecular biology that allow them to imagine whole invisible networks of primeval genes lurking as atavisms beneath the polished exteriors of urbane and civilized modern cells; and 3) bizarre misconceptions about cancer causing mutations that they can't back up with a single specific example.

So let's start with the first objection, the weird evolutionary history enshrined in the title. They postulate a step in evolution just above the colonial stage, with groups of cells having marginal specializations — which is reasonable, up to a point.

The transition from unicellular to complex multicellular organisms took place over an extended period starting at least 1 billion years ago (Hedges and Kumar 2009). Importantly, ‘advanced’ metazoan life of the form we now know, i.e. organisms with cell specialization and organ differentiation, was preceded by colonies of eukaryotic cells in which cellular cooperation was fairly rudimentary, consisting of networks of adhering cells exchanging information chemically, and forming self-organized assemblages with only a moderate division of labor. These proto-metazoans were effectively small, loosely-knit ecosystems that fell short of the complex organization and regulation we associate with most modern metazoans. In short, proto-metazoans, which we dub Metazoans 1.0, were tumor-like neoplasms.

Aaargh. That last sentence. It's like announcing a theory of human evolution which has as a premise that chimpanzees are mentally retarded people — it manages to insult and mischaracterize both chimps and people with mental handicaps. No, early metazoans were healthy, functional, successful organisms. Neoplasms are abnormal masses of tissue produced by excessive reproduction of cells with broken regulatory machinery. You don't get to declare the relevance of your analogy by simply declaring both sides equivalent; there is no reasonable, evidence-based rationale for declaring early metazoans to be a bunch of tumors.

Davies and Lineweaver don't provide any evidence for their evolutionary history, either. It's really a simplistic just-so story in which they claim metazoan precursors were just like cancers so that they can claim that cancer uncovers a lurking atavism which proves that metazoan precursors were just like cancers. It's so circular I got dizzy.

Which leads us to their problem #2: 'atavisms'. They've got this vision of wedging poor old Darwin into supporting a hierarchy. Like Haeckel's version of the story, we're built of accumulated layers of new traits, where old ones are buried and fossilized beneath new genes, and all it takes is a little erosion to expose well-preserved relics. It's all terminal addition, with a little nod to some modification of earlier stages of our evolution.

It is, however, in the nature of Darwinian evolution that life builds opportunistically on what has gone before. The genetic apparatus of the new Metazoa 2.0 was overlain on the old genetic apparatus of Metazoa 1.0. The genes of Metazoa 1.0 were tinkered with where possible, and suppressed where necessary. But many are still there, constituting a robust toolkit for the survival, maintenance and propagation of non- differentiated or weakly-differentiated cells—‘tumors’—and when things go wrong (often in senescence of the organism) with the nuanced overlay that characterizes Metazoa 2.0, the system may revert to the ancient, more robust way of building multicellular assemblages—Metazoa 1.0.

Aaargh. No. Genes that are suppressed decay and are lost, not lurking. Genes that remain and are tinkered with by mutations acquire new functions, lose old ones, and are assembled into new, coevolving networks. If they're still there, they're still part of the current regulatory pathways; they're not in a separate 'layer', they're not waiting, unmodified, to blithely re-enact ancestral states. No, really, you can't take one of your cells, switch off a few genes, and set it free in the ocean to swim off and follow its primitive lifestyle. Cancers are not dreaming of protistan independence.

Again, they're playing the same trick: postulating a condition that would make their model work, because it would make the model work, not because they actually have evidence that this is how gene networks function.

Along the way, they try to argue with standard models of cancer formation which assume, for instance that cell proliferation is a product of broken regulatory brakes, not activation of a sleeper program to reinvoke an ancestral state. Davies and Lineweaver point out that cancer cells actually have patterns of integration and communication that promote survival — how could that be if they were merely rogue cells rampaging madly through the body? I wanted to scream at them, "because they are slightly modified metazoan cells still, you blinkered crackpots!" Cancer cells can execute growth-promoting activities like secreting signals that encourage a blood supply to infiltrate the tumor because they are intrinsically multicellular, not because they are reverting to a cunning barbarism.

They also object to a model they call "internal Darwinism": that cancer cells have the suite of traits that they do not because of some internal coherent plan, but because they experience random mutations, many of which kill the cancer cells, but we only see the successful combinations of traits that produce proliferative, invasive tumors. And they claim to have evidence against that model. They don't.

Our explanation of cancer as an atavism that short- circuits the Metazoa 2.0 regulatory system and unleashes the suppressed Metazoa 1.0 system receives support from the amazing pleiotropy of some enzymes, and, as has been realized recently, some micro-RNAs (miRNAs). Thus the enzyme COX-2 and the miRNA known as miR-31 have been found to control not just one, but a collection of tumorigenic factors. Such remarkable efficiency and economy would be deeply puzzling if it arose from a few decades of internal Darwinism, but makes perfect sense if it had been honed by evolution over an extended period of time to form an optimized package that constitutes a type of on–off switch for a set of previously adaptive traits.

But…but…pleiotropy is an essentially universal property of regulatory genes! We expect this kind of widely interacting behavior. And why would you consider expression of an interaction between two components of a regulatory system to be a property of "previously adaptive traits" rather than a property of currently functional and adaptive traits? This kind of argument resolves nothing in their model.

Let's now consider that third problem I mentioned, that they don't seem to be able to discuss how known cancer mutations work, and don't give any specific examples of atavistic gene circuitry exposed by a modified oncogene. I'll be specific; let's use Rb as an example.

Rb is a known oncogene, that is, a gene that when modified or broken can cause one step in the path towards cancer. It's normal role in healthy cells is as a regulator of the cell cycle.

Dividing cells follow a cycle. Most cells are in G₁ (Gap 1), doing what cells do, and then under control of clock-like changes in specific genes, they can enter the S (synthesis) phase, when their DNA is replicated, followed by a G₂ phase (gap 2), and than an actively dividing mitotic or M phase. Each of these phases has a checkpoint where a battery of proteins survey the state of the cell and either permit the process to proceed, or block it if there are problems. In extreme cases, the checkpoint proteins can determine that the cell is so irreparably damaged that the only option is suicide, and the cell will self destruct.

This is a process that cancer needs to disrupt if it is to continue; cancer cells typically have damaged DNA or aberrant signals flying everywhere that ought to be triggering all kinds of alarms in the checkpoint system, and either stopping cell division immediately, or activating repair mechanisms that fix the damage, or just killing the corrupted cell immediately.

One of the most critical points in this cycle is called the R or Restriction point. Prior to this point, the cell is sensitive to external signals that can induce cell division; after this point, the cell no longer pays attention to those signals, because it is on a rigidly programmed track towards completing cell division. This is that last fateful moment of decision before the cell commits to dividing.

Standing at this point is an essential guardian of the cell cycle, pRb. This protein is an inhibitor of cell division, acting as a tumor suppressor gene. It's the guard at the gate, and it must be satisfied that all is well in the cell before it will allow division to continue. It's default mode is to stop cell division, but it recieves signals from a wide array of pathways that can tell it to stand down and let the process continue. Trust me, control of this gene is complicated because it is so essential to well-regulated cell division: look at it here, standing sentry just above the yellow R point, with all these other pathways talking to it:

I think you can see how this gene can contribute to cancer when it's defective. Shoot the guard, open the gate wide, and allow cell divisions to proceed unchecked.

What I don't see, and what doesn't fit the Davies/Lineweaver model, is where the emergent atavisms are. The components of the cell cycle that are disinhibited by loss of this protein are not hidden away and are not unused, primitive versions of cell cycle regulators; they are simply the contemporary regulators with an essential limiting factor removed.

This is not to say that the evolutionary history of the gene is unimportant. There have been several phylogenetic studies of Rb; it's present in most eukaryotes (and some homologous sequences have been identified in the Archaea!), single-celled as well as multicellular, but it does show increasing complexity in its function in multicellular organisms. It is simply not the case, however, that you can find "layers" of Rb function that allow you to peel away modern functions to expose an antique Rb that makes it revert to an ancient pattern of behavior.

Let's let Davies and Lineweaver get even further from science. Why should we like their model? Because it's more optimistic than those poopyhead biologists' ideas!

Given cancer’s formidable complexity and diversity, how might one make progress toward controlling it? If the atavism hypothesis is correct, there are new reasons for optimism. The postulated toolkit of Metazoa 2.0, although admittedly complex, is nevertheless a fixed and finite feature of multicellular life. The number of tools in the kit is not infinite. What one cancer learns cannot be passed on to the next generation of cancers in other patients. Cancer is not going anywhere evolutionarily; it just starts up all over again in the next patient. Although cancer may seem like a perpetually moving target, a given cancer has a strictly limited set of atavistic possibilities open to it. Thus, cancer is a limited and ultimately predictable adversary. This understanding of cancer as a limited atavism should engender optimism among oncologists. The anticipated precision of personalized drug therapies will not be infinite. This view contrasts sharply with the open-ended possibilities for cancer implied by the ‘internal Darwinism’ model.

Scientists shouldn't be looking for optimism, they should be searching for the truth, which is sometimes going to be grim. So what if the 'internal Darwinism' model implies an intimidating number of possibilities? It is what it is. Every cancer is different, every cancer is complicated, and waving your hands and pretending that there is an ancestrally-derived unity of mechanisms under every single cancer doesn't make it so. Davies and Lineweaver are going to have to do better than parading decades-old evolutionary misconceptions at me to persuade me of their case.

Cao L, Peng B, Yao L, Zhang X, Sun K, Yang X, Yu L (2010) The ancient function of RB-E2F pathway: insights from its evolutionary history. Biol Direct 5:55.

Davies PCW, Lineweaver CH (2012) Cancer tumors as Metazoa 1.0: tapping genes of ancient ancestors. Phys. Biol. 8 015001-015008.

Takemura M (2005) Evolutionary history of the retinoblastoma gene from archaea to eukarya. Biosystems 82(3):266-72.