Challenging the prevailing thought style

Frederick Grinnell


Nobel Laureate Albert Szent-Györgyi’s prescription for discovery was seeing what everybody else has seen and thinking what nobody else has thought. René Magritte’s 1936 oil painting Perspicacity shows a seated artist staring at a solitary egg on a draped table. On his canvas, he paints a bird in full flight. As expressed in Szent-Györgyi’s prescription and Magritte’s painting, discovery frequently requires unconventional thinking. The more novel a discovery claim, the greater its potential to refashion the thought style and affect subsequent research in the field. At the same time, novelty also challenges intersubjectivity and can come into conflict with the prevailing thought style. As a result, highly novel discovery claims sometimes are received with considerable skepticism by the research community.

I learned firsthand the difficulty of challenging the prevailing thought style when I was a postdoctoral fellow with Paul Srere. Srere was engaged in a dispute with many in the research community concerning how best to understand enzyme regulation. His focus was on the enzymes of the citric acid cycle. Many of these enzymes, he emphasized, were present in cells at much higher concentrations compared to those typically studied by biochemists. “But a cell is not a test tube!” Srere argued that understanding enzyme regulation required analysis of enzymes in the context of their supramolecular organization within cells.

Frustrated by the resistance of conventional thinking in the research community to his new ideas, Srere would begin his seminars by describing the I Ching. The I Ching is a Chinese philosophical system attributed to the Emperor Fu Hsi (circa 2800 B.C.E.) that is used to understand the past and predict the future. Paul explained to his audience (mockingly) that he was skeptical about the 5,000-year-old method until he examined the trigram symbols arranged in a circle. He was amazed to learn that Fu Hsi had predicted the citric acid cycle well before the cycle was discovered by Sir Hans Krebs. Then Srere would show the front cover of the Beatles 1969 album Abbey Road. He would point out some of the clues on the cover indicating that Paul McCartney had died, a popular conspiracy theory at the time. Srere told his listeners that he was fascinated “by the evidence, logic and inevitability of the conclusion that McCartney died.…It was a compelling story. The only flaw was that he was alive.” And then Srere reached his timeless conclusion: “Given a large mass of data, we can by judicious selection construct perfectly plausible unassailable theories—all of which, some of which, or none of which may be right”.

While the skepticism of conventional thinking can delay acceptance of new ideas, if the ideas are correct, then the research community eventually catches up. One sign of the research community catching up with Paul Srere was the establishment in 1987 of an ongoing research conference titled Enzyme Organization and Cell Function (later renamed Macromolecular Organization and Cell Function). In recognition of his contributions to the field, Srere’s fellow biochemists established the “Paul Srere Memorial Lecture” beginning with the 2004 conference.

The history of Nobel Prizes includes many examples of novel discoveries that were either ignored or disputed for years, for example, tumor viruses (Nobel Prize in 1966), chemiosmotic theory (Nobel Prize in 1978), transposable genetic elements (Nobel Prize in 1983), and catalytic ribonucleic acid (Nobel Prize in 1989). The presentation speech for Krebs when he won the 1953 Nobel Prize for the discovery of the citric acid cycle reminded the audience that “in the beginning Krebs was quite alone with his idea, and when he first presented it, it was criticized by many”. The presentation speech for Stanley Prusiner when he won the 1997 Nobel Prize for the discovery of prions contained similar remarks: The hypothesis that prions are able to replicate without a genome and to cause disease violated all conventional conceptions and during the 1980s was severely criticized. For more than 10 years, Stanley Prusiner fought an uneven battle against overwhelming opposition.

In the end, accommodation of the thought style to novelty frequently follows the path described by William James: “First, you know, a new theory is attacked as absurd; then it is admitted to be true, but obvious and insignificant; finally, it is seen to be so important that its adversaries claim that they themselves discovered it”.

Frederick Grinnell - Credibility: Validating Discovery Claims

Read more

When catastrophe strikes a cancer cell

Jose M. C. Tubio & Xavier Estivill

How do the mutations that lead to cancer come about? The traditional view is that a gradual process involving continual acquisition of heritable genetic changes by cells causes cancer. There is, however, an alternative view that single catastrophic events can lead to multiple mutations. In a paper published in Cell, Stephens et al. provide evidence for the concept of catastrophism in cancer.

The authors coin the term chromothripsis (from ‘chromo’, for chromosome; and ‘thripsis’, for breaking into small pieces) to designate this phenomenon.

Stephens and colleagues combined next-generation sequencing and single nucleotide polymorphism (SNP) array data to analyse the patterns of somatic (non-germline) genomic rearrangements in tumours. Intriguingly, they found that in some cases the changes consist of tens to hundreds of rearrangements, confined to one or a few chromosomes.

Rearrangements are common in cancer, so how can the authors tell whether such mutations were caused by a single event? They argue that the final configuration of the rearrangements they observed could be explained only by a single catastrophic episode, rather than by a series of independent events. Rearrangements due to chromothripsis are usually restricted to a few chromosomes, within which breakpoints show a non-random distribution.

The potential implications of chromothripsis as a cause and/or mediator of cancer are evident. As Stephens et al. point out, the generation of so many rearrangements in a single genomic crisis makes it likely that more than one cancer-causing lesion would occur.

Read more in Nature

Common Disease: Are Causative Alleles Common or Rare?

Robert Shields from PLoS Biology

It has been said that a week is a long time in politics. But in human disease gene mapping, 10 years can seem a very short time indeed. It once seemed so simple: find a family with a number of affected individuals and narrow down regions of the genome shared by affected individuals but not their unaffected siblings. This process (family linkage analysis) was lengthy but had notable success with some diseases, including hereditary breast cancers caused by the BRCA1 and BRCA2 genes. Yet, many diseases known to have a genetic component (because they tend to run in families or siblings show a high concordance) do not follow a simple Mendelian pattern of inheritance and cannot be dissected in this way. Instead, researchers tried an “association” approach, starting with a large number of unrelated individuals, to find gene variants, or alleles, that are more common in affected than in unaffected controls. For such a strategy to work, the diseases must be influenced by variants that are quite common in the populations.

Some readers might recall the heated debates about the common disease, common variant (CD-CV) hypothesis. Using arguments based on population genetics (such as the rate of creation and purging of deleterious alleles, the genetic bottleneck in the human population and subsequent population expansion), the CD-CV hypothesis proposed that in common diseases with a genetic component, some predisposing alleles are relatively common and a combination of alleles or environmental effects was required before disease occurred, much like being dealt a bad hand from a common deck of cards. Under this hypothesis, disease-associated alleles might be found by using common gene variants, such as single nucleotide polymorphisms (SNPs), as a guide and comparing affected individuals with controls. Others cast doubt on this idea and suggested that common diseases are unlikely to be caused by common alleles and more likely to be caused by rarer ones; they too deployed arguments based on population genetics and suggested that association studies using common genetic variants might not be successful. As with all scientific debates, there seemed only one way out: collect the data and see. Well, 10 years and several millions of dollars later we have a lot of data, but are we any the wiser? Do we understand the allelic spectrum of disease any better than we did 10 years ago?

There have now been over 700 genome-wide association studies (GWAS) published linking many variants to over a hundred diseases. Many of these results are robust in that they can be replicated in several populations, leaving little doubt that common variants can contribute to common diseases. The problem is that the effect of these variants on disease is often rather modest, so that people with the disease-predisposing alleles are only slightly more likely to get the disease than those without. Larger and larger studies reveal more disease genes, usually with smaller and smaller effect on overall disease risk. The “missing heritability” problem then arises because, even in aggregate, these loci typically fall somewhat short of explaining the entire genetic component of disease risk. So where are the genes accounting for major predisposition to disease? One possible explanation is that GWAS do not directly reveal the disease-causative DNA variant, but rather a common DNA variant (usually an SNP) that is close enough to be genetically linked to it (almost always inherited together) and common enough to be on the genotyping microarrays. This has spurred more effort (and more expense) to find rarer and rarer SNPs by sequencing more genomes and make even larger arrays in the hope that the new SNPs may be in even closer linkage with the causative allele. Alternatively, it's possible that the disease-predisposing variants are not SNPs at all, but other changes in the genome, such as a duplicated or deleted gene or region—a so-called copy number variant (CNV)—or a result of epigenetic marks in the chromatin; neither of these would show up using the current generation of microarrays that look just at SNPs.

A paper published recently in PLoS Biology from the lab of David Goldstein put the cat amongst the CD-CV pigeons by suggesting that rather than common diseases being caused by common alleles, maybe rare alleles each with a large effect on disease might be creating “synthetic associations” in the GWAS signal by occurring, by chance, more often with one common allele than another. The paper used statistical reasoning to suggest that such synthetic associations are possible—but are they likely? Given how much time and money has been invested in surveying SNPs and attempting to match them up to diseases, the relative importance of such synthetic associations would have important implications on the direction of future research. The paper got a lot of publicity, even making the New York Times.

Now, some might say that no one likes the implication that they have been barking up the wrong scientific tree, still less perhaps that such a critique garnered a lot of publicity. But the issue is best settled by discussion—and data—which is why in this issue of PLoS Biology we publish two critiques of the original article together with a response from the original authors. The critiques argue that although rare variants could in theory create synthetic associations, this is not a likely explanation for the missing heritability. Perhaps with further advances in ever cheaper sequencing technologies and the ability to sequence whole genomes from affected individuals we will, before the next 10 years are up, finally have a better understanding of the missing pieces of the genetic causes of common disease.

Protein Bias

From The Scientist:

New data show that protein research is stuck on a small set of molecules that was hot in the 1990s

At the dawn of the 21st century, scientists completed the first draft of the human genome and reported that human cells encode more than 20,000 proteins. But most of the protein research performed since has focused on about 2,000 proteins (mostly enzymes) that were already known and whose functions were well studied, according to an analysis published this week in Nature.

Why do you think researchers are less likely to study previously unexplored systems? The first reason is that scientists are sometimes just like dogs with a bone: They just love their problems. And they love going deeper and deeper, and it's the richness and the complexity of their problem that drives them.

The other thing is that you can get funding for proteins for which there's a preexisting community. The third thing is that when you rationalize your grant, we get rewarded for the elegance for which we weave a tale. You need to have context. When you go to publish a paper, if you work on an unknown protein, you're less likely to capture the imagination of the peer reviewers and they'll probably ask for a lot more work. To study an unknown protein, it takes longer, and that's what nobody has in this modern world -- time.

And then the last thing is research tools. There is no molecular biologist now who would not prefer to use a genome wide knockdown set of RNAis, for example, as opposed to the single RNAi. That's a better experiment, but no such tools exist for proteins. If you have a new protein, you have to make a knockout cell, you have to make an antibody, an inhibitor or mutant. It's a lot of work to create the infrastructure to even do the experiment.

Evolution at Two Levels: On Genes and Form

In their classic paper “Evolution at Two Levels in Humans and Chimpanzees", published exactly 30 years ago, Mary-Claire King and Allan Wilson described the great similarity between many proteins of chimpanzees and humans. They concluded that the small degree of molecular divergence observed could not account for the anatomical or behavioral differences between chimps and humans. Rather, they proposed that evolutionary changes in anatomy and way of life are more often based on changes in the mechanisms controlling the expression of genes than on sequence changes in proteins.

This article was a milestone in three respects. First, because it was the first comparison of a large set of proteins between closely related species, it may be considered one of the first contributions to “comparative genomics” (although no such discipline existed for another two decades). Second, because it extrapolated from molecular data to make inferences about the evolution of form, it may also be considered a pioneering study in evolutionary developmental biology. And third, its focus on the question of human evolution and human capabilities, relative to our losest living relative, marked the beginning of the quest to understand the genetic basis of the origins of human traits.

(...)

From the outset of this review, I make the sharp distinction between the evolution of anatomy and the evolution of physiology. Changing the size, shape, number, or color patterns of physical traits is fundamentally different from changing the chemistry of physiological processes. There is ample evidence from studies of the evolution of proteins directly involved in animal vision, respiration, digestive metabolism, and host defense that the evolution of coding sequences plays a key role in some (but not all) important physiological differences between species. In contrast, the relative contribution of coding or regulatory sequence evolution to the evolution of anatomy stands as the more open question.

Sean B. Carroll - Evolution at Two Levels: On Genes and Form

Heterotopy

Changes in the spatial regulation of toolkit genes and the genes they regulate are associated with morphological divergence.

A vast body of data has accumulated that has linked differences in where toolkit genes are expressed, or where the genes they regulate are expressed, with morphological differences between animals at various taxonomic levels. Most studies have analyzed situations in which the spatial location of a developmental process (for example, the making of a limb, the formation of a pigmentation pattern, the development of epithelial appendages, etc.) has been altered. The classical term for such spatial changes in development is heterotopy (changes in the timing of a process are known as heterochrony). The close correspondence between heterotopic shifts in gene expression, development, and morphology, combined with the known roles of these genes in model taxa, have provided compelling evidence that changes in morphology generally result from changes in the spatiotemporal regulation of gene expression during development.

Explanations for the evolution of anatomy have thus focused on the genetic and molecular mechanisms underlying the evolution of spatial gene regulation. And the keys to understanding spatial gene regulation are the architectures of gene-regulatory regions and transcriptional networks.


Sean B. Carroll. Cell 134, July 11, 2008.

Functional Equivalence of Distant Orthologs and Paralogs

Many animal toolkit proteins, despite over 1 billion years of independent evolution in different lineages, often exhibit functionally equivalent activities in vivo when substituted for one another. These observations indicate that the biochemical properties of these proteins and their interactions with receptors, cofactors, etc. have diverged little over vast expanses of time.

What makes these results so surprising and notable is that the function of any transcription factor or ligand is dependent upon interaction with other endogenous proteins—transcription factors, coactivators, corepressors, and parts of the transcriptional machinery in the case of transcription factors, cell surface receptors in the case of ligands. One might expect that genetic drift or the coadaptation of proteins within lineages would lead to functional incompatibilities among proteins from different taxa, especially those separated by over 1 billion years of independent evolution. Yet, many proteins are so conserved that they can interact and function together with other proteins from long-diverged taxa. These results have been more the rule than the exception for transcription factors. Numerous studies of eukaryotic transcription demonstrate that the basic transcriptional machinery, many coactivators, corepressors, chromatin-remodeling complexes, and the protein motifs through which transcription factors interact with them, are often very well conserved.

Sean B. Carroll. Cell 134, July 11, 2008.

Evolved genetic guesses

Once it is recognized that not all mutations are random mistakes, the way one sees the relationship between physiological and developmental adaptation and evolutionary adaptation begins to change. We are used to thinking of them as very different: physiological and developmental changes involve instruction - what happens in cell or organisms is controlled by internal or external regulatory signals; evolutionary changes involve selection - some heritable variants are preferred to others. In the jargon of philosophers of biology, the physiological and developmental processes that underlie a phenotype are "proximate causes", while evolutionary processes - natural selection and whatever else has constructed the phenotype during evolutionary history - are "ultimate causes".

Yet, if the generation of some heritable variation is under physiological or developmental control, how distinct are the two types of causes? Seeing evolution purely in terms of selection acting on randomly generated variation is wrong, because it involves instructive processes too. As we see it, the dichotomy between physiology/development and evolution, and between proximate and ultimate causes, it is not as absolute as we have been led to believe. They grade into one another. At one extreme there are purely selective processes, acting on chance variation, while at the other there are purely instructive processes,which are totally physiological or developmental and do not involve any selection. Between these extremes we find the majority of processes in the real world, which are to varying degrees both instructive and selective. Some developmental changes, such as those occurring during the development of the immune system, also involve selection, whereas some evolutionary changes, particularly in bacteria and plants, may have instructive components. In other words, Darwiniam evolution can include Lamarckian processes, because the heritable variation on which selection acts is not entirely blind to function; some of it is induced or "acquired" in response to the conditions of life.

From: Evolution in four dimensions by Eva jablonka and Marion J. Lamb

Time to have fun with a bad project

Gene swap key to evolution

Horizontal gene transfer accounts for the majority of prokaryotic protein evolution

Microbes evolve predominantly by acquiring genes from other microbes, new research suggests, challenging previous theories that gene duplication is the primary driver of protein evolution in prokaryotes.

The finding, published in PLoS Genetics, could change the way scientists study and model biological networks and protein evolution.

"Even at a meeting last summer, there were those that thought that bacteria genomes expanded mostly through duplications and others that argued that it was due to gene acquisition," wrote Howard Ochman, an evolutionary biologist at Yale University who was not involved in the research, in an Email to The Scientist. "Now we all have a paper to point to that does a very good job of answering this question," he said. "Their conclusions are really robust."

Prokaryotes, including bacteria and archaea, thrive in diverse conditions thanks to their ability to rapidly modify their repertoire of proteins. This is achieved in two ways: by receiving genes from other prokaryotes, called horizontal gene transfer -- the nefarious way that bacteria acquire antibiotic resistance -- or by gene duplication, in which an existing gene is copied, taking on a new or enhanced function as mutations accumulate.

Past analyses using few, distantly related genomes estimated that horizontal gene transfer contributes to, at best, 25 percent of the expansion of protein families -- that is, the addition of proteins with novel functions or structures. But the recent availability of numerous, closely related prokaryotic genomes tempted Todd Treangen and Eduardo Rocha at the Institut Pasteur in Paris to more accurately test which biological process is the main driver of prokaryote protein evolution. "The genomic data was finally there to do a more in depth study," said Treangen, now a postdoc at the University of Maryland.

The duo analyzed 110 genomes of varying size from 8 clades of prokaryotes, focusing in on 3,190 defined protein families. The results were unambiguous: 80 to 90 percent of protein families had expanded through horizontal gene transfer. In addition, the researchers found that the two processes have different evolutionary roles: transferred genes persist longer in populations while duplicated genes are transient but more highly expressed.

"Overall, the role of gene transfer in protein diversification has been underestimated," said Treangen. Still, he noted, they analyzed only a tiny fraction of the microbes that exist in the world, and further research should be done as more genomes become available.

It would be nice to study the same two processes in eukaryotes, said Patrick Keeling, a molecular evolutionary biologist at the University of British Columbia who was not involved in the research. Yet despite numerous documented cases of horizontal gene transfer in eukaryotes, including plants, it would be hard to test because of the lack of genomic data from enough closely related eukaryotes (which have significantly larger, less manageable genomes than prokaryotes).

Still, "it raises some really fascinating questions about whether [eukaryotes] evolve in the same way," said Keeling.

Treangen, T.J. et al., "Horizontal Transfer, Not Duplication, Drives the Expansion of Protein Families in Prokaryotes," PLoS Genetics, 7:e1101284, 2011.

The Scientist