Thursday, July 25, 2013

The extent of what we can't know about gene function is infinite

A paper in Cell reports preliminary results of an effort to systematically mutate each of the genes in a mouse genome, one at a time, to determine the function of every gene ("Genome-wide Generation and Systematic Phenotyping of Knockout Mice Reveals New Roles for Many Genes," White et al.).  It's the Sanger Institute Mouse Genetics Project.

This isn't the first such effort, nor is the mouse the only organism on which this has been done, but the authors write that the advantages of their project are that all mice have a common genetic background (which actually is both good and bad; genetic background is controlled, but results will differ in other strains of mice), they've used a standard protocol for constructing the alleles they introduce, and the lines they produce are readily available to other researchers.  In addition, White et al. note that other mutagenesis projects have focused on a subset of known and previously characterized genes.  This precludes discovery of novel gene function, and can shed no light on previously uncharacterized genes.  

The method used in this project is targeted mutagenesis of embryonic stem cells, to knock out expression of specific genes. The cells are inserted into the embryonic mouse and its viability, fertility and the effect of the knockout "on a broad range of traits" are then assessed.  Many mouse knockout experiments have been reported to show "no phenotype", that is, have no effect on the mouse, but the authors suggest that other researchers didn't look hard enough.

In the paper, White et al. report statistics for the first 900 or so genes they've done.  
We found that hitherto unpublished genes were as likely to reveal phenotypes as known genes, suggesting that novel genes represent a rich resource for investigating the molecular basis of disease. We found many unexpected phenotypes detected only because we screened for them, emphasizing the value of screening all mutants for a wide range of traits. Haploinsufficiency and pleiotropy were both surprisingly common. Forty-two percent of genes were essential for viability, and these were less likely to have a paralog and more likely to contribute to a protein complex than other genes.
Really, not at all surprising and should not be interpreted as such, but it is nice to have this additional documentation.  The finding that unexpected phenotypes will be uncovered if you screen for them brings to mind developmental biologist Lewis Wolpert's query after yet another experiment showed "no phenotype" in a knockout mouse, "But have you taken it to the opera?".  (Is this apocryphal I ask, finding no documentation other than the last time we blogged it.  If he didn't say it, at least someone did, and it's perfect.)

That novel genes (or functional elements in genomes of various types) are still to be found, and are as likely to affect phenotype as well-characterized genes is, again, no surprise.  We've blogged several times (here and here, e.g.) about a craniofacial genetics project we've been working on, particularly the gene mapping phase.  Gene mapping, or the effort to identify a gene or genes within a chromosomal region statistically associated with a trait or disease of interest, sounds straightforward enough.  But, as we said in our posts on the subject, choosing candidate genes within what can be chromosomal regions that contain tens or hundreds of genes, many of them as yet uncharacterized, is often more of an art than a science.  Or, indeed, a crap shoot.  

Mapping problems are exacerbated by the fact that of the subset of genes that have been characterized, most are known only for a single function, or even a disease with which it's associated. And, investigators gravitate to known genes as they contemplate candidates -- the drunk under the lamp post effect.  White et al. very appropriately call attention to this in the paper, but the extent of the problem is even greater than they indicate.  Yes, they are characterizing the effect of gene knockouts on a broad range of traits, but, well, are they taking their mice to the opera?  Changing the environment; food, temperature, lighting, cage conditions, etc.?  

Indeed, no matter how well-intentioned, they can't expose these mice to all possible environments, and so can't observe all possible phenotypes. There is no way to relate many of these effects to what the same mice would be like in the wild, and until many mice of each genotype are studied the amount of variation even under controlled lab conditions can't be known--so it is only tentative to assign trait effects to the mutant strain.  

And as importantly, they can't do their knockouts in every strain of mice, although genetic background is known to be a huge factor affecting gene expression and effects, which may range from none to lethal.  Genetic background is probably a given mutant gene's most proximal environmental factor there is.  

So, the project reported in this paper is interesting as far as it goes, particularly as it makes clear how much we still don't know about the function of most genes, and why.  But because environments are endlessly changing and every genome is unique, no finite set of experiments can do much more than make it clear how much we'll never know, or be able to predict. The danger is that while digesting what was found we not become bemused by mega-studies of this type into thinking that they show more than they really do.


James Goetz said...

So that is 20,000 genetically engineered mice. And if we want to evaluate "bigenic traits," then that is 400,000,000 engineered mice. And how many specimens do we need for each mutational type? And if some genes react differently to different mutations, then how many types do we need to engineer? And then there are multigenic traits more complex then bigenic traits.

I need a big lab for this:-)

Ken Weiss said...

This is today's omics (totally enumerative) approach, a revised way of doing science. Lots of animals will pay the ultimate price for it. Whether we'll grow out of this kind of epistemology to something better, time will tell. Meanwhile, you'll need to expand your lab about 100 fold!