Saturday, November 28, 2009

The gene for ... ?

Just how do they pack a human into 23,000 genes?

An interesting aspect of biology in this "post-genomic" age is how it has frustrated many researchers. Here we had expected genes "for" alcoholism, schizophrenia, autism, and all the other ills that ail us, but the road to find them has been rocky, tortuous, and has, in many cases, petered out to nothing. It turns out that genomes are complicated and don't come with user manuals. They didn't develop over evolutionary time in tidy ways that generate bodies and behaviors linearly from the DNA. Fruit flies have 15,000 genes, yeast cells have 6,000, and bacteria have 3,000. Most human genes are shared with bananas (i.e. the encoded proteins would function interchangeably). 99% of our DNA is shared with chimpanzees. So what makes us different, and oh-so-much better?

For every condition like eye color that can be attributed to simple mendelian variation in one or a couple of genes, there are a multitude of other conditions whose origin is not simple, but arises from the network of interaction of many genes. Coat types in dogs is another example of simple genetics, outlined in a very nice paper recently, where three genes suffice to explain most of the variation in dog hair types, from short/smooth (the wild-type) to curly, long, wavy, and wiry:

Combinations of novel alleles at three genes (FGF5, RSPO2, and KRT71) create seven different coat phenotypes: (A) short hair; (B) wire hair; (C) "curly-wire" hair; (D) long hair; (E) long, soft hair with furnishings; (F) long, curly hair; and (G) long, curly hair with furnishings.

But most aspects of biology are not so simple. Autism is the example I will focus on most, though similar observations apply to cancer, schizophrenia, personality/temperament, and on down the list of interesting and important conditions/traits.

Genes function in complex networks of regulation, both from the upstream direction of receiving regulatory signals, and downstream in the gene product's (a protein, usually) modification by other systems like phosphorylation and degradation, to eventual roles in combination with other gene products. It's all a big mess of interacting effects. What we see as the ultimate phenotype is the end result of complex mechanisms and lengthy development. Just as a cello has only four strings but an infinitude of musical expression, so genes can be played in different tissues, at different times, and in different volumes to accomplish many tasks.

For example, one set of genes, BMP1 to 20, (bone morphogenic protein), is an evolutionarily related (i.e. duplicated and diverged multiple times) family encoding small proteins that induce developmental events, like formation of cartilage and bone, when secreted by nearby cells. BMP4 is used repeatedly through development to induce notochord formation, eye formation, bone, and tooth formation, and pituitary formation, among others. To quote from one resource:
"Defects in BMP4 are the cause of microphthalmia syndromic type 6 (MCOPS6) also known as microphthalmia and pituitary anomalies or microphthalmia with brain and digit developmental anomalies. Microphthalmia is a clinically heterogeneous disorder of eye formation, ranging from small size of a single eye to complete bilateral absence of ocular tissues (anophthalmia). In many cases, microphthalmia/anophthalmia occurs in association with syndromes that include non-ocular abnormalities. MCOPS6 is characterized by microphthalmia/anophthalmia associated with facial, genital, skeletal, neurologic and endocrine anomalies."
Not a simple story, is it? BMP4 is used, reused, and reused again for similar purposes all over the body. A drug that inhibits its action would have devastating effects, though if that drug could closely control the timing and place of its effects, it might be very useful. That is one of the many challenges of drug development today.

Conversely, a single trait can be composed of the work of many genes. Down syndrome results from the duplication of an entire chromosome- many genes with slight increments in amount of product produced seems to cause a wide spectrum of altered traits. Autism seems similarly be be the consequence of the action of many genes, defects in any one of which can have similar effects. Autism spectrum disorder (ASD) has strong heritability (70% to 90% estimated), yet searches for the responsible genes have come up with not one, but scores of genes. I'll focus on one study that contributes to this story: "Efforts to map disease genes using linkage analysis have found evidence for autism loci on 20 different chromosomes." That is quite a statement, considering that we only have 24 chromosomes [a reader helpfully points out we have only 23!].

This paper used high-tech genomics to look for tiny deletions and other genetic variations throughout the genomes of families afflicted with ASD. Out of 195 autism patients and 196 controls, they found variants in 14 patients versus 2 in the control set. All the variants were heterozygous, indicating that, like in Downs, small increments or decrements in gene function may be responsible, with one normal gene copy remaining in each case. Some of the genes were expressed in the brain, while others were known to participate in retardation disorders, and others have little known about them.

Two other observations stand out. First is that most of these small genomic duplications or deletions are novel- they happened recently, and being deleterious, will die out rapidly as well. Autism "runs" in some families, but most cases result from spontaneous defects in a wide variety of genes. Second is the large number of genes estimated to be in this pool of possible ASD causers- at least 29 found so far between this and other studies, with more to come as more families are analyzed with ever more comprehensive methods. This is relevant to the rate of occurrence of this disorder, which is very high for genetic disorders, and possibly rising. It is known that autism rates go up dramatically for children of older parents.

So ASD seems to be the result of rare defects in any of numerous genes, many known to be involved in synapse formation and activity. It might be that an over-arching pathway of early brain development channels many genetic problems into the same syndrome, much as many problems with cars result in the common syndrome of "it won't start". Thus to think of genes "for" such diseases is problematic, given the complicated relations. There appears to be a developmental process that generates the syndrome, driven by many genes and susceptible to many distinct defects. And the case of ASD doesn't even touch on the separate issue of genetic variants that individually have tiny phenotypic effects on a trait, (such as height, for instance), but combine with many others to determine the overall trait quantitatively.

More deeply, for the genome at large, it isn't size that matters. Soybeans have almost three times the number of genes we have (though one-third the overall amount of DNA). It is how you use what you've got, in complex networks of regulation, combination, and reuse that makes a brain out of a bunch of cells. As the old Sun slogan had it, the network is the computer.

  • A relevant review of genome-wide association studies of "disease genes."
  • Climate disasters, a tad overdrawn.
  • The casually callous and obtuse David Brooks does in health care.
  • Does the Economist know economics?


  1. I thought we had 23 chromosomes? :-)

  2. Thanks - can't count! I'll edit that.