I want to continue with my theme that the patterning node is not just genotype, and that the genotype is not just genes. In particular, I want to talk about gene interaction, and why the genotype is not the sum of the genes. This is an area in which I am an active researcher (in fact as I write this I have a computer churning away on a rather large analysis), which can be a problem. When you are too close to a subject often it is difficult to write anything that anybody but you can understand. I will try to keep this general, but if I fail let me know and I will do what I can to clean things up!
First, off it is obvious that genes interact. Genes themselves do not interact (don’t argue with me about modern molecular findings, I am taking a last millennium view of the gene). Rather, genes produce products such as enzymes that work in enzymatic pathways. This means that the gene products inevitably interact. In a standard enzymatic pathway you might find that some chemical, Chemical X (with apologies to the Powerpuff Girls), is acted on by an enzyme, enzyme A, producing a product, chemical Y, which is then acted on by enzyme B to produce chemical Z. (for some very obscure reason this image is not going into the post well. click on it if you need to see it better.)
The rate at which enzyme B converts Y into Z will depend to a large extent on the relative amount of Y and Z in the system. If enzyme A is very efficient there will be lots of Y available, and enzyme B will work quickly. If different individuals have variants of B that work at different rates these rate differences will show up in the rate at which Z is produced. If, on the other hand enzyme A is very inefficient the rate at which B works will be a function of how much of chemical Y is available. Variation in enzyme B will not be expressed since even very inefficient forms of B are using up Y as quickly as it is made. This type of interaction, that is an interaction between two loci that affects the phenotype, is termed epistasis.
There are lots of examples of epistasis. One of my favorites is coat color in Labrador retrievers. In Labradors, dogs with a BB or Bb genotype are black, whereas those with a bb genotype are chocolate; that is, unless they have the ee genotype at a second locus. Regardless of the genotype at the B locus ee dogs are yellow.
So, do we say that the B locus codes for coat color or that it has no effect? The answer depends on the genotype at a second locus. This is the crux of the biscuit (again apologies to Frank Zappa). The effect of an allele on the phenotype depends on the background it is found in.
There are lots of potential interactions among two loci with two alleles, and even more with additional alleles or loci. In the Labrador example we can turn the colors into numbers by assigning a 2 to black, a 1 to chocolate, and a 0 to yellow. Then the table becomes:
The numbers are values for the phenotypes (often paradoxically called genotypic values), and in principle, for any pair of loci with two alleles we can fill in the appropriate phenotypes.
What I have just outlined has been called “physiological epistasis”. Physiological epistasis can be thought of as the interaction among loci measured on an absolute scale. Using physiological epistatic measures it is clear that the B and E loci interact epistatically. It is important to contrast this with “statistical epistasis” which is whether or not there are epistatic interactions in a population. In the above example, if the population was fixed for the EE (or ee) genotype we would say there were no epistatic interactions. This is because only the first column would be observed in the population, and we would say that the B locus was a simple dominant locus. Statistical epistasis is only seen when both alleles at both loci are present. From a more nuanced perspective the amount of epistasis will depend on the gene frequencies at both loci. In general the amount of epistasis will be greatest when both loci have a gene frequency of 0.5, and it will diminish as frequencies diverge from that intermediate value.
It is useful to consider an analogy to height here. I am 5 feet 5 inches tall, and this is true whether I am measured in the Netherlands or Guatemala. That is, my height is a fixed attribute that is independent of context. Although my height is fixed, whether I am tall or short is not fixed, and depends upon where the comparison is done. In the Netherlands the average male is about 6 feet 1 inch tall, and as a consequence I would be considered short. In contrast in Guatemala the average male is 5 feet 2 inches tall, and I would be considered tall. So, am I tall or am I short? It depends on where I am being measured. (height data from http://www.disabled-world.com/artman/publish/height-chart.shtml)
Just as it is not possible to make statements about being tall or short outside of the context of the groups being compared, it is not possible to make statements about the effect of a gene on the phenotype independent of the genetic background in which it is measured. If we like black dogs, and we are using a population fixed for the EE genotype we can say that the B gene is a “good gene” because it causes dogs to be black. However, if we are using a population fixed for the ee genotype the B gene has no effect on the phenotype, and it would be considered a neutral gene.