Each cell division produces two daughter cells, and these then divide and so on, generating an expanding tree of cellular descent from the single fertilized egg cell with which you and I began life. Because of the exotic dance of differentiation, signaling among and within cells leads to differential use of subsets of genes, and trees of cellular descent that separately form our adult tissues, like lungs, kidneys, the gut and so on. This is a tree whose branching 'shape' bears much resemblance to the divergent nature of evolutionary trees, in the sense that from one or a few starter cells, our organ systems accumulate hierarchies of DNA changes due to mutations that arise independently in cells in these trees of organ-system descent. In that rough sense different organs are like different species or isolated populations of the same species. Mutations arising in lineages leading to the gut are different from those in lineages leading to the brain, or muscles, or other systems. These mutations are called somatic because they arise and are transmitted to daughter body cells in the individual but not to the next generation. The latter happens from cells in the germ line, that produces sperm or egg cells; the germ line is another tissue system that accumulates its own inter-cellular hierarchical mutation heterogeneity, but whose cells can be transmitted to the next generation.
|Tree of descent; Darwin, Origin of Species|
Mutations individually usually do nothing, whether they arise in the germ line and are inherited or arise somatically. Of those that do something, mostly that would be at least slightly harmful to the cell (at least, that is what has been estimated in various ways by comparing species DNA sequences and in other ways). The reason is that evolution has over aeons established functional relationships among genomic functions so that mutations are likely to interrupt something that's been working. Still, a small fraction of mutational changes may help the cell.
During life, tissues are a hierarchical mosaic of these somatic mutations and any effects they may have. Both in somatic and germ line lineages, mutations that help may lead to proliferation of the cells that acquire them or, in the latter case, to individuals who can compete better than they would without the mutation. That competition is 'natural selection' as Darwin envisioned it, but somatic competition, among body cells and also among germ line cells may do a lot of purging of harmful mutations because if the variants don't cooperate with each other, that is, interact properly, the cell dies. That is a form of selection, though it's not 'Darwinian' in the sense of being about competition among individual organisms.
Most evolutionary and biomedical geneticists treat individuals as having one genome, a diploid set of chromosomes with their particular nucleotide variations, that is identified by sequencing a cheek swab or blood sample. This sequence is referred to as the individual's 'constitutive' genome, as if all his/her cells had that sequence, and evolutionary genetics is largely about the competition among these genomes. But that ignores the intra-individual somatic variation that mutations produce as if it were unimportant. That is sometimes, perhaps often, a mistake, indeed a culpable mistake, because we know better.
The numbers game
If there have been around 160 billion humans, descended from a small number of founders somewhere in northeast Africa, the variation that has arisen among them has a hierarchical pattern. Of mutations whose descendant copies have successfully have survived to the present, if they are common today they are necessarily old. They arose early in the 'tree' of human descent history. They are typically found in all the world's populations, because the mutation had occurred in Africa before the human emergence both out of and within Africa. If they are more recent, but common in some region such as Europe or the Pacific islands, they are likely to have been present in the human expansion into that area. Indeed, one can say that the reason humans are a global species is that most of our genomic characteristics were in place before our expansion.
More recent mutations are less common. The reason is obvious. A new mutation arises in but a single instance, and our species' reproduction is so slow that many generations of successful transmission must occur before there can be many, geographically dispersed descendant copies. Until recent millennia, are ancestors were born, lived, and died locally. If a mutation is harmful, it is likely not to have proliferated or dispersed very much. In any case, because our species is so widespread and so numerous and so rapidly expanded after agriculture was discovered, most mutations are brand-new or at least very recent--and very rare in their respective population. And if a trait is built during development (and later life) by the contributions of many genes, each of those will have the general characteristics just described.
For practical as well as wishful-thinking reasons, the biomedical genomics community would prefer to see common disease-related variants, because they're easy to find and commercially worth doing something about (like developing a drug). But most causally relevant variants are rare, for the kinds of reasons just given. This is by far the typical finding of genomewide searches for variation affecting complex, late-onset diseases or other traits.
When one thinks of somatic mutation the temptation is to say that, yes, they happen, but mostly they just kill the cell or, because most cells just recently divided, they will be rare in the body. That would likely be true, just as most germ line variants are rare. But we have as many or more early-life cell divisions, with big expansions, as the human species had early individuals with big geographic expansion. So there is no reason to assume that somatic mutation only affects small patches of local cells any more than there is reason to think that human variants are all rare.
Recombination and cellular selection may temper this, and just as a population with, say, 10% diabetics, say, can function, so perhaps can a stomach with 10% dysfunctional cells. But if enough cells malfunction for somatic mutational reasons, the 'public health' of an organ can reflect this as disease that is not mappable by the usual GWAS-like approaches. Further, just as each diabetic may have a different combination of individually rather rare inherited variants across his/her genome, s/he may also have organs that are a comparable mosaic of somatic-mutationally different cells. But all of that will be invisible to the kind of genomic analysis, based on constitutive genomes, that is being done today.
One thing that is important is what I have elsewhere called phenotype amplification. That refers to the need for the somatic effects to involve enough cells that the individual, as a whole, manifests them. Cancer is the obvious case, because the tumor starts in only a single cell but grows to uncontrolled amounts of cells. Epilepsy may be another as I hypothesized in Trends in Genetics a few years ago, because firing neurons make apparently lasting connections to other neurons and anomalously firing neurons may entrain enough others to lead to seizures. There has also been evidence that other traits, including some aspects of neurological aging is due to accumulating mutations in mitochondrial DNA in brain cells.
How important is somatic variation?
Who knows how many other traits there might be in which GWAS-like somatic complexity may cause a given instance of the trait? Could somatic variation account for the typical non-mappability of most instances of most traits or the weak predictability based on constitutive genotypes? Can somatic mutation be dismissed because phenotype amplification is weak?
There are at least two major differences between constitutive variation in populations of individuals and somatic variation in populations of cells that may be relevant to this question. First, as mentioned earlier if a mutant cell dies because it malfunctions for its context, it is not noticed or tallied the way the death of a person from a specific disease is, for example. Secondly, while there is mitotic recombination, it appears only to involve the two homologs in the given cell, thus not spreading variation over cell generations the way that recombination, independent assortment, and random mating do in a population of individuals.
How important the latter difference is, is unclear. It may be enough to undermine the major effects somatic mutation by not spreading variation among as many cells in a tissue as regular recombination does among individuals in a population. This would apply especially to late, local somatic mutation in a tissue. But combinations of rare mutations have major effects in random-mating populations, so this might not be as relevant as it seems since the same can occur in somatic cells. And the lack of dispersal through recombination may actually concentrate damaging mutational combinations in many genome regions, without dispersing but also without diluting those effects in a larger set of cells--that is, the epistatic effects (interactions among genome regions) will not go away because of independent assortment and recombination. Indeed, it may not be that all cells in, say, a liver contain bad combinations of somatic mutations, but a large enough patch within the liver may lead to disease. This is not that different from many other diseases that affect only enough of an organ or system to generate pathology that gets the person to the doc's office.
Somatic mutation is a potential sleeping tiger, mainly recognized because of its devastating effect when it leads to cancer, but largely unrecognized because it is inconvenient, to say the least, to take it seriously: one might have to biopsy impossibly large numbers of tissues on a regular basis to provide the kind of somatic 'GWAS' that current population-based studies have sought, to great frustration.
And finally, each person has his/her own unique mix of inherited and somatic mutation. We are not yet very good at determining far enough in advance when that mix becomes toxic. But just blinding ourselves to what may likely be as least as important an aspect of genomic effects is not a very good way to do science, either.
(Note: this has been edited from the first-posted version of earlier today, just to fix some sloppy phrasing)