In species such as mice and humans, B cells that have been recruited in an immune response migrate to lymphoid follicles, where, together with antigen-specific T cells and follicular dendritic cells, they form germinal centers. Here B cells replicate at considerable rates (). Moreover, mutations are introduced in the V regions of B cell receptors at a rate 105-106 times higher than background DNA mutation (). The resulting variants are selected on the basis of their affinity for the antigen presented by follicular dendritic cells. The cells that survive this process are recruited for the memory compartment, and have, in general, higher affinity for antigen than germline cells (). Although the mechanism of somatic hypermutation is not known, there is a wealth of knowledge about its sequence specificity (). Moreover, in most systems where somatic hypermutation of B cell receptors has been described (), the sequence specificity seems to follow similar patterns.
Considering that framework regions satisfy mostly a structural role, while the complementarity-determining regions are responsible for binding the antigen, it would clearly be advantageous to target somatic hypermutation to these latter regions. What is not clear, however, is whether this advantage would be large enough to be selectable, given the multiple sources of stochasticity in immune responses. Evidence for diversity-enhancing selection in the evolution of immunoglobulin CDRs has been presented by Tanaka and Nei (1989). They showed that the rate of nonsynonymous substitution in immunoglobulin CDRs is higher than the rate of synonymous substitution. This suggests that selection must be favoring organisms with diverse immunoglobulin CDRs. The effect that I am setting out to investigate is whether CDRs are not only more diverse, but also more prone to diversification under somatic hypermutation. As I will show in the following sections, such evidence is present in individual V region genes not only from mice and humans, but from a variety of species. Moreover, the compositional biases that are responsible for this effect are also present in most of these species. This argues that the mechanism that is responsible for introducing somatic mutations is shared between the species that I studied. In the field of experimental immunology, this issue is currently under debate. At least two different somatic mutation mechanisms are thought exist: one which active in mammals, the other in sharks and frog. Finally, I will analyze the T cell receptors, showing that some, but not all of T cell receptor sequences have compositional features that might be associated with somatic hypermutation.
In the previous chapter, I proposed that germline antibody genes could constitute the substrate of evolutionary learning, such that the antibodies that constitute the naive immune repertoire coarsely map the pathogenic universe. In this chapter I investigate the hypothesis that antibody genes also learn in evolution how to maximize their chances of rapidly producing a highly specific antibody for a given pathogen. I will show that this is realized by differential codon usage between the framework and complementarity-determining regions of the antibody genes. Namely, given a certain functionality of the antibody molecule, which is determined by its amino acid sequence, the codon bias in FRs minimizes the chance of a replacement mutation under somatic hypermutation, the converse being true for CDRs. Numerous studies attempted to demonstrate that the frequency of replacement mutations is higher in CDRs. Only one of these studies (1997) addresses the question of selection for amino acid sequence versus selection for amino acids that are more likely to undergo mutations. However, while this study showed that, overall, there is differential codon usage between FRs and CDRs, this result did not hold for all the antibody gene sets under study. It was not clear whether the lack of generality was due to the low resolution of the statistical tests employed in the study. In my study, I will make use of large artificial sequence sets, which allow me to design statistical tests on individual antibody gene sequences. I will show that antibody genes from a number of species show differential codon usage bias between FRs and CDRs, such that FRs are expected to undergo significantly lower proportion of replacement mutations than CDRs. The artificial data sets that I construct can be used to investigate the level of mutability optimization in different sequences. In somatic mutation studies, one often is confronted with the question of whether a given number of mutations in the sequence is due to its intrinsic tendency to mutate or to some selection pressure for or against mutations. Such questions can be answered using the approach that I introduce here.
Yet another set of questions that can be answered in my framework have to do with using other substrates than the immunoglobulin gene for somatic hypermutation. These studies are designed to investigate what components in the immunoglobulin gene are required for targeting somatic hypermutation to the immunoglobulin locus. Here one is confronted with the question of whether a low number of observed mutations in the non-immunoglobulin gene is due to a lacking regulatory element, or to an intrinsic low tendency to mutate of the gene used as a substrate. This becomes a trivial question once one has an empirical mutability vector, as I will show below.
The power of the approach that I introduce in this chapter is manifested in a number of other areas as well. I will illustrate this by focusing on two other questions. The first comes from comparative immunology: is the somatic hypermutation mechanism shared between all species? I will show that similar differential codon usage between FRs and CDRs characterize species ranging from sharks to humans. This result argues that, at least among these species, the mechanism for somatic hypermutation is likely to be shared. The second question concerns to somatic hypermutation mechanism itself. I will show that codon bias consistent with low propensity for replacement mutations also characterizes non-immunoglobulin sequences. As these sequences do not undergo somatic hypermutation, this association suggests that the somatic hypermutation mechanism uses components from mutation or repair mechanisms that operate with much wider scope across the genome. Thus, non-immunoglobulin sequences that evolve codon bias to minimize their chance of undergoing replacement mutations in evolution also seem less mutable under somatic hypermutation. I will also show that the somatic hypermutation mechanism picks out the A/T content of a gene. That is, high tendency to undergo replacement mutations is correlated with the content of A and T nucleotides in the gene. This result does not give us the key to what the somatic hypermutation mechanism is, but it may prove useful in narrowing the search for this mechanism. Although the phenomenon that immunoglobulin genes undergo somatic mutation was described almost thirty years ago Weigert et al. (1970), the mechanism responsible for introducing these mutations has not been identified.