Listening to genetic background noise.

Since the early days of modern genetics, researchers have largely shut their ears to the “background noise” of genetic modifiers that modulate the expression of mendelian traits. Because modifier genes complicate regular patterns of inheritance and because their identification can be difficult, their importance has been recognized, but they have rarely been the focus of genetic studies. However, attention is beginning to focus on these genetic background effects, in part because they are the entrée into the genetics and systems biology of more complex and common diseases, and in part because they have great potential as powerful and effective ways to treat and perhaps prevent disease. Keeping things simple was the key to discovering the rules of inheritance. With insight and luck, Mendel carefully selected traits, emphasizing those that showed a virtually invariant phenotype. His crosses provided segregation ratios for these traits, from which the fundamental and universal rules of inheritance were inferred. Mendel’s landmark discovery would have been impossible if he, like others before him, had selected traits that show more complex patterns of inheritance. And yet, remarkably few traits are truly mendelian. Most traits vary, sometimes in simple ways and sometimes in profound ways. Perhaps the most important discovery from the study of spontaneous, engineered, and chemically induced genetic variants in model organisms such as laboratory mice is that their phenotypes depend strongly on the genetic background. For example, a specific genetic variant may lead to embryonic lethality on some genetic backgrounds but to full viability and no obvious phenotype on other backgrounds. Background genes that can suppress or exacerbate detrimental phenotypes are ubiquitous and highly polymorphic. They are easiest to detect when a single genetic variant is the target of modification, but they are probably involved in both genetically simple and complex traits. In general, phenotypic noise may result from allelic heterogeneity, variable environments, or stochastic effects, as well as from modifier genes. In humans, distinguishing among these sources of variability can be difficult. With model organisms, however, defined crosses can be made, so detecting modifiers is relatively easy. Moreover, because the repertoire of genes in humans is extraordinarily similar to that in other mammals and because the chromosomal arrangement of these genes has been strongly conserved during mammalian evolution, the identity and location of candidate modifier genes in humans can be reliably predicted from studies in model organisms. Modifier genes are now leading to pioneering discoveries about the genetics and biology of hearing. In a study in this issue of the Journal, Schultz et al.1 show that five members of a family with autosomal recessive sensorineural hearing loss were homozygous for a missense mutation at a conserved site in CDH23, the gene that encodes cadherin 23, which mediates calcium-dependent cellto-cell interactions. Mutations in this gene have previously been shown to cause a similar form of hearing loss in waltzer mice.2 The authors also note that three of the five affected persons had severe-to-profound hearing loss, whereas the other two had only high-frequency sensorineural hearing loss. This variability suggested the action of a modifier gene. Again, studies in mice pointed the way. Variants of the Atp2b2 gene, which encodes a calcium pump in the plasma membrane, modulate the severity of hearing loss in waltzer mice. Schultz et al. report that heterozygosity for a missense mutation in ATP2B2, the human homologue of Atp2b2, was associated with the severity of hearing loss in these five persons. During the study, they speculated that variation in the activity of the calcium pump influences cellular interactions that depend on calcium availability. They then tested whether this ATP2B2 variant modifies other forms of hearing loss and found that it affected hearing loss resulting from other genetic causes. Interestingly, heterozygosity for the ATP2B2 variant was not sufficient to cause hearing loss — an observation that is typical of most modifier genes: in the absence of their target gene variants, they generally do not have detectable effects on the trait. These discoveries about the genetic basis of sensorineural hearing loss in humans would have been much more difficult had there not been corresponding discoveries in mice to guide the way. This cross-talk between studies in humans and those in model organisms is probably essential for rapid progress. Modifier genes complicate the genetics of sim-