B cell variable genes have evolved their codon usage to focus the targeted patterns of somatic mutation on the complementarity determining regions

Abstract

The exceptional ability of B cells to diversify through somatic mutation and improve affinity of the repertoire toward the antigens is the cornerstone of adaptive immunity. Somatic mutation is not evenly distributed and exhibits certain micro-sequence specificities. We show here that the combination of somatic mutation targeting and the codon usage in human B cell receptor (BCR) Variable (V) genes create expected patterns of mutation and post mutation changes that are focused on their complementarity determining regions (CDR). T cell V genes are also skewed in targeting mutations but to a lesser extent and are lacking the codon usage bias observed in BCRs. This suggests that the observed skew in T cell receptors is due to their amino acid usage, which is similar to that of BCRs. The mutation targeting and the codon bias allow B cell CDRs to diversify by specifically accumulating nonconservative changes. We counted the distribution of mutations to CDR in 4 different human datasets. In all four cases we found that the number of actual mutations in the CDR correlated significantly with the V gene mutation biases to the CDR predicted by our models. Finally, it appears that the mutation bias in V genes indeed relates to their long-term survival in actual human repertoires. We observed that resting repertoires of B cells overexpressed V genes that were especially biased toward focused mutation and change in the CDR. This bias in V gene usage was somewhat relaxed at the height of the immune response to a vaccine, presumably because of the need for a wider diversity in a primary response. However, older patients did not retain this flexibility and were biased toward using only highly skewed V genes at all stages of their response.

Keywords: Affinity maturation; B cells; Codon bias; Codon usage; Evolution; Somatic hypermutation.

Figure 1

Expected likelihood of mutation or change of amino acid, following mutation being in the CDR, of BCR VH, Vλ, Vκ, and TCR Vα and Vβ genes, under a targeted model of mutation: Expected likelihood that given a mutation occurs it will occur (a) in the CDR of a given V gene (b) cause an amino acid change in the CDR and (c) cause a non conservative amino acid change in the CDR. Given the length of the V sequences, the fraction of positions in CDR is ~0.27 marked by the dashed line. The black star represents the actual CDR fraction of that family and the green star is the mean fraction of each family.

Figure 2

The distribution of Z scores when comparing the original germline CDR fraction of each V gene to the mean and standard deviation of its related simulated dataset (a) comparing Z scores of the expected fraction of mutations in CDR; (b) the expected fraction of amino acid change in the CDR; and (c) the expected fraction of non conservative amino acid changes in the CDR. The distribution means falling in the shaded gray region are not statistically distinct in their codon bias from the simulated distribution.

Figure 2

The distribution of Z scores when comparing the original germline CDR fraction of each V gene to the mean and standard deviation of its related simulated dataset (a) comparing Z scores of the expected fraction of mutations in CDR; (b) the expected fraction of amino acid change in the CDR; and (c) the expected fraction of non conservative amino acid changes in the CDR. The distribution means falling in the shaded gray region are not statistically distinct in their codon bias from the simulated distribution.

Figure 2

The distribution of Z scores when comparing the original germline CDR fraction of each V gene to the mean and standard deviation of its related simulated dataset (a) comparing Z scores of the expected fraction of mutations in CDR; (b) the expected fraction of amino acid change in the CDR; and (c) the expected fraction of non conservative amino acid changes in the CDR. The distribution means falling in the shaded gray region are not statistically distinct in their codon bias from the simulated distribution.

Figure 3

CDF of the average mutation fraction (see Results section 3.3) per position compared to a uniform distribution of mutation fractions across the V genes - BCR V_H (black), Vλ (orange), Vκ (green), TCR Vβ (yellow) and Vα (blue). BCR V genes were not distinct from each other but were significantly distinct from TCR Vα. TCR Vβ was intermediate as it was not significantly distinct from either TCR α V or the BCR V genes.

Figure 4

The average, by positions scores for BCR V_H, Vλ, Vκ TCR Vβ and Vα in CDR (purple) and FR (blue) for (a) Mscore (b) Rscore and (c) Tscore, under a targeted model of mutation.

Figure 4

The average, by positions scores for BCR V_H, Vλ, Vκ TCR Vβ and Vα in CDR (purple) and FR (blue) for (a) Mscore (b) Rscore and (c) Tscore, under a targeted model of mutation.

Figure 4

The average, by positions scores for BCR V_H, Vλ, Vκ TCR Vβ and Vα in CDR (purple) and FR (blue) for (a) Mscore (b) Rscore and (c) Tscore, under a targeted model of mutation.

Figure 5

The germline based model predictions of expected mutation fraction are predictive of the observed mutation in CDR segments of the gene: A plot of the correlation of the predicted CDR fraction of nonconservative changes to observed fraction nonconservative mutations in the CDR, (blue line - ρ = 0.458, p=1.526e-03). The same for nonsynonymous changes in CDR, (red line - ρ = 0.526, p=1.817e-04) and for synonymous change in the 4 fold redundant amino acids [“A”, “G”, “P”, “T”, “V” where no selection is expected to occur] (green line ρ=0.760024, p=5.80e-10).

Figure 6

Comparison of V_H gene usage post immunization and their distribution amongst different levels of bias for change in the CDR. V_H gene distribution is shown in young (top) and old (bottom) populations at day 0, 7 and 28 post immunization. Greater spread represents higher usage of the respective V_H gene with the specific expected fraction of (a) mutations in the CDR, (b) amino acid changes following mutation in the CDR and (c) nonconservative changes in the CDR. The dashed line represents the actual CDR fraction. The red star is the mean if V_H genes were all used equally and black star is the mean when V_H gene usage is taken into account.

Figure 6

Comparison of V_H gene usage post immunization and their distribution amongst different levels of bias for change in the CDR. V_H gene distribution is shown in young (top) and old (bottom) populations at day 0, 7 and 28 post immunization. Greater spread represents higher usage of the respective V_H gene with the specific expected fraction of (a) mutations in the CDR, (b) amino acid changes following mutation in the CDR and (c) nonconservative changes in the CDR. The dashed line represents the actual CDR fraction. The red star is the mean if V_H genes were all used equally and black star is the mean when V_H gene usage is taken into account.