The use of epistasis to construct genetic pathways

The ways that geneticists think about epistasis are a little bit different and more informative than the standard modifications to the 9:3:3:1 ratio that are discussed in the chapter. In controlled laboratory experiments with well-studied organisms, the inheritance patterns of genes are usually known before epistasis is encountered, so the phenotype ratios are not something that needs to be worked out. In fact, the two genes used in the crosses are usually chosen because the investigator is specifically interested in their interaction. Thus, epistasis is an expected outcome of the experiment, and inferences from epistasis provide a useful way of understanding how the genes function together to affect a biological process. While this will be encountered more fully in Chapter 15, a brief introduction here will set the stage.

Recall that the two genes being investigated for a possible interaction are inherited by the established principles of Mendelian genetics. Their interaction is deduced from the interpretation of their phenotypes. Many different kinds of phenotypes can be used to look at gene interactions, including the effects of having mutations in two different genes and the patterns of expression of one or both genes in wild-type and mutant individuals. Let’s look at two examples, one that uses one mutant phenotype and an expression pattern phenotype, and a second one that uses two mutant phenotypes.

Using a mutant and an expression pattern

We discussed microarrays in Section 2.5 as a method to examine gene transcription patterns on a genome-wide scale. In wild-type individuals, we extract RNA and hybridize the labeled RNA (or corresponding cDNA) to a microarray to determine which genes are being transcribed and at what levels. The expression profile that we observe from a microarray makes a reasonable phenotype, not only for one gene, but also for many genes.

Now suppose that we do the same microarray experiment using a mutation that eliminates the function of a particular gene, called gene A, that is, we extract RNA from a population of individuals who are mutant for that gene (a/a) and perform the same microarray analysis as we did for wild-type individuals. We show some of these in Figure A. For many of the transcripts that we consider on the microarray, the pattern will be the same in the wild-type and the mutant individuals. We conclude that the transcription of those genes does not depend on the activity of gene A.

Figure A. Gene interactions using molecular assays. Interactions between two genes can be detected by using a mutant version of one gene and another phenotypic assay for the other gene. The assay compares the phenotype of the other gene when the first gene is wild-type with when it is mutant. In this example, the interactions of Gene A with four other genes is shown, using a microarray as the phenotype for Genes 1, 2, 3, and 4. When Gene A is wildtype, Genes 1, 2, and 3 are transcribed while Gene 4 is not. A microarray is then done for the same four genes when Gene A is mutant, shown here as a/a. The transcription of Gene 1 has decreased, as indicated by the lighter gray signal, indicating that functional Gene A is needed for high levels of Gene 1 expression. The results with the other genes are summarized below the figure. Note that this assay does not mean that Gene A directly regulates the expression of Genes 1, 3, and 4; other genes may be involved, and the effect of Gene A could be indirect.

For another set of transcripts on our microarray, the pattern of transcription is different in wild-type and a/a mutants. Maybe some of the genes are transcribed at much lower levels, or possibly not transcribed at all, in the a/a mutant. We infer that these transcripts depend on the normal function of gene A because, when the activity of gene A is eliminated in the a/a mutant, the level of expression of these other genes is decreased. Thus, the normal function of gene A somehow activates or turns on the transcription of these genes. Notice a very important point. We cannot conclude that gene A directly controls the transcription of these genes; the effect of gene A might be exerted through any number of intermediate processes which we have not defined. But we can say that, in some way, the transcription of these other genes depends on the normal function of gene A.

Let’s illustrate this with a simple analogy. Suppose that you are reading in your room one evening when suddenly your lamp goes out. You can reasonably conclude that the activity of your lamp has been changed and that something that your lamp needs for its activity has malfunctioned. There might be  malfunction in some component with a very direct effect—the bulb in your lamp has burned out, for example. But there might also be a malfunction in some component far upstream of your lamp—a tree limb has fallen on a power line,

for example, so that electricity to an entire neighborhood has been cut off. The transcription of a gene as assayed by the microarray is similar to the reading lamp in this analogy; it has changed, but we don’t know if the malfunction (the mutation in gene A) occurred directly upstream of the transcript or in some process far upstream. The dependence of the transcript on the function of gene A shows the strategy or the logic of the overall program by which gene A regulates other processes, but it does not provide any of the details.

Before leaving this example, we should consider the transcription pattern of some other genes in our hypothetical microarray. There will undoubtedly be genes that will be transcribed at a higher level in the a/a mutant than in the wild-type. Thus, our inference is that normal function of gene A is somehow to repress or turn off the transcription of these genes. Again, the effect of gene A could be direct or indirect. The change in transcription is showing the overall program by which gene A affects this process, but not any of the specifics.

We have illustrated this interaction with the changes in transcription, as assayed by a microarray, but this can be generalized to any assay for any process. We could have an assay that looks at changes in splicing pattern, in protein expression, in protein localization, or in any of a number of other processes. The overall strategy for interpreting the results is the same, regardless of the assay or process used.

Using two mutant phenotypes

Sometimes we do not have an easy assay for the transcription or expression of a gene, but we still want to understand how two genes might interact to affect the phenotype. A similar logic of epistasis can still be applied, but we make the inference based on the mutant phenotypes of the two genes, rather than an assay for expression. We can use coat color in Labrador retrievers for this explanation, since it involves a particularly well-studied pair of genes with familiar and recognizable phenotypes.

There are numerous genes that are responsible for coat color variation in mammals, including the well-studied agouti locus discussed in the chapter. In any particular purebred (that is, inbred) breed of dogs or other mammals, some of these genes exhibit no variation, and all purebred dogs of the same breed are homozygous for the same allele. Thus, genes that contribute to the variation in one breed may or may not be responsible for some of the variation in another breed. This is an important concept that will arise again in Chapters 10 and 16; the gene might have a very important function in coat color, but, because there is only one common allele in a particular breed, it does not contribute to coat color variation in this breed.

In Labrador retrievers, three coat colors are recognized by kennel clubs and dog breeders: yellow, black, and chocolate brown, as shown in Figure B. Yellow dogs are homozygous recessive for a trait known as the Extension or E gene; they are ee. The actual E gene is now known to be the melanocortin 1 receptor gene MC1R, which also contributes to pigmentation differences in other mammals by regulating the deposition of pigment granules in skin and hair. In humans, one of the most common forms of red hair arises from mutations in the MC1R gene, and mutations in MC1R were found in the Neanderthal genome as well. Because Labrador retrievers are homozygous for alleles in other genes that regulate or interact with MC1R (including the agouti locus), mutations that eliminate the function of MC1R produce yellow fur.

Figure B. Epistasis using color variation as the mutant phenotypes. Epistasis or gene interactions can also be evaluated using visual phenotypes such as coat color. In dogs, a gene known as E is needed for the deposition of pigment, while a gene known as B is responsible for the synthesis of black or brown pigment. Thus, black Labradors must have the genotype of E_B_; the B allele is responsible for the black pigment, while the E allele allows the pigment to be deposited in the hair. Chocolate Labradors have the genotype E_bb, since the absence of the B allele means the pigment is brown. Yellow Labradors do not deposit the pigment. They therefore have the genotype ee, but the pigment genotype cannot be inferred because no pigment is deposited in the coat. The E gene is defined as being epistatic to the B gene, since the genotype at the E gene affects the phenotype controlled by the B locus. The E gene is considered to act upstream of the B gene. Other genes also affect coat color in dogs, but these are not variable in Labrador retrievers, so their effects are not evident.

Dogs with an E_ genotype have a functional E gene and deposit pigment granules in the skin and hair. These pigment granules in Labrador retrievers can be either black or brown, depending on the function of the TYRP1 gene, which encodes an enzyme responsible for the production of eumelanin, the dark pigment in mammalian hair. In standard coat color genetics, this is often symbolized as the B gene because of its phenotype; it is the same gene responsible for the black or brown rabbits discussed in the chapter. Alleles that encode functional enzymes produce black pigmentation, represented as B_. Three different molecular lesions found in dog breeds, including Labrador retrievers, eliminate the function of the gene and produce a brown pigment instead; all three mutations, although different at the molecular level, produce similar hair color phenotypes, at least in Labrador retrievers, which is known as chocolate to dog breeders. Thus, bb is chocolate brown and B_ is black.

How does these genes interact with each other? Matings between dogs with different coat colors has shown that the E gene is epistatic to the B gene, so that a dog that is ee is yellow; no matter which pigment granules are produced by the activity of the B gene, those granules are not being deposited in the hair because of the mutant form of the E gene. Thus, the E gene acts upstream of the B gene, although the cellular and molecular effect is indirect. 

Let’s talk through this example using our desk lamp again. If the power goes out throughout the neighborhood—that is, we have a malfunction in the upstream gene (ee)—then it makes no difference if the light bulb is functional (B_) or burned out (bb). We will only see the effect of the upstream malfunction even if the bulb is working fine, and the lamp will be out. As before, the interaction between the two genes is showing the overall strategy or program by which they act, but not the specifics. With the E and the B genes, we can also make an analogy to having a functional delivery truck. If the delivery truck is broken down (that is, an ee genotype), it makes no difference if it is delivering books or television sets; no home delivery will occur.

In either analogy, as in epistasis, the interaction between the upstream and downstream genes could be direct or indirect, and the outcomes would be the same. That is a limitation of using epistasis to build molecular, cellular, or biochemical pathways; there could be many intervening steps between the two being tested. On the other hand, the tremendous advantage of using epistasis to build a pathway is that it provides a logical or functional pathway, even when the biochemical and molecular interactions are not known. Many, if not most, of the cellular and biochemical pathways you may be familiar with were originally constructed using epistasis. The molecular and biochemical interactions were worked out after the logical or functional order was determined.