Patterns of amino acid substitution at MHC loci

Hughes and Nei [1] recognized that these hypotheses could be distinguished by comparing rates of synonymous and non-synonymous substitution at MHC loci. The results are summarized in Table 1. Notice that they distinguished among three functional regions within the protein and calculated statistics separately for each one:

• codons in the antigen recognition site,

• the remaining codons in the extracellular domain involved in presenting the antigen on the cell surface (the $\alpha_1$ and $\alpha_2$ domains), and

• codons in the extracellular domain that are not directly involved in presenting the antigen on the cell surface (the $\alpha_3$ domain).

Table 1: Rates of synonymous and non-synonymous substitution for loci in the MHC complex of humans and mice (modified from [3] and based on [1]). ARS refers to the antigen recognition site. Significant differences between $K_s$ and $K_a$ are denoted as: * ($P < 0.05$), ** ($P < 0.01$), and *** ($P < 0.001$).

	ARS		$\alpha_1$ and $\alpha_2$		$\alpha_3$
Locus	$K_s$	$K_a$	$K_s$	$K_a$	$K_s$	$K_a$
Human
HLA-A	3.5	13.3***	2.5	1.6	9.5	1.6**
HLA-B	7.1	18.1**	6.9	2.4	1.5	0.5
HLA-C	3.8	8.8	10.4	4.8	2.1	1.0
Mean	4.7	14.1***	5.1	2.4	5.8	1.1**
Mouse
H2-K	15.0	22.9	8.7	5.8	2.3	4.0
H2-L	11.4	19.5	8.8	6.8	0.0	2.5**
Mean	13.2	21.2*	8.8	6.3	1.2	3.6**

Hughes and Nei argue that the unusually low value of $K_s$ in the $\alpha_3$ domain of mice is due to interlocus genetic exchange. If we discount that set of data as unreliable, a clear pattern emerges.

• In the part of the MHC molecule that is not directly involved in presenting antigen, $\alpha_3$ in humans, the rate of non-synonymous substitution is significantly lower than the rate of synonymous substitution, i.e., there is selection against amino acid substitutions.

• In the parts of the MHC molecule that presents antigens, $\alpha_1$ and $\alpha_2$, the rate of synonymous and non-synonymous substitution is indistinguishable, except within the antigen recognition site where there are more non-synonymous than synonymous substitutions, i.e., there is selection for amino acid substitutions.

In a later paper Hughes et al. [2] took these observations even further. They subdivided the antigen recognition site into the binding cleft, the T-cell-receptor-directed residues, and the outward-directed residues. They found that the rate of non-synonymous substitution is much higher in the binding cleft than in other parts of the antigen recognition site and that nucleotide substitutions that change the charge of the associated amino acid residue are even more likely to be incorporated than those that are charge-conservative. In short, we have very strong evidence that natural selection is promoting diversity in the antigen binding capacity of MHC molecules.

Notice, however, that this selection for diversity is combined with overall conservatism in amino acid substitutions. Across the protein as a whole, most non-synonymous substitutions are selected against. Of course, it is that small subset of amino acids where non-synonymous substitutions are selected for that are responsible for adaptive responses to new pathogens.