Human peripheral blood antibodies with long HCDR3s are established primarily at original recombination using a limited subset of germline genes

Abstract

A number of antibodies that efficiently neutralize microbial targets contain long heavy chain complementarity determining region 3 (HCDR3) loops. For HIV, several of the most broad and potently neutralizing antibodies have exceptionally long HCDR3s. Two broad potently neutralizing HIV-specific antibodies, PG9 and PG16, exhibit secondary structure. Two other long HCDR3 antibodies, 2F5 and 4E10, protect against mucosal challenge with SHIV. Induction of such long HCDR3 antibodies may be critical to the design of an effective vaccine strategy for HIV and other pathogens, however it is unclear at present how to induce such antibodies. Here, we present genetic evidence that human peripheral blood antibodies containing long HCDR3s are not primarily generated by insertions introduced during the somatic hypermutation process. Instead, they are typically formed by processes occurring as part of the original recombination event. Thus, the response of B cells encoding antibodies with long HCDR3s results from selection of unusual clones from the naïve repertoire rather than through accumulation of insertions. These antibodies typically use a small subset of D and J gene segments that are particularly suited to encoding long HCDR3s, resulting in the incorporation of highly conserved genetic elements in the majority of antibody sequences encoding long HCDR3s.

Figure 1. Increased HCDR3 length does not correlate with affinity maturation events.

(A) Peripheral blood antibody sequences were grouped by mutation frequency and the percent of sequences in each group that contained codon-length (non-frameshift) insertions was calculated for each donor. All values for healthy donors from Group 1 (n = 4) and Group 2 (n = 3) are shown in the left panel, with the mean percentage ± SEM shown for each mutation value. In the right panel, sequences from Group 1 healthy donors were segregated by B cell subset, and the best-fit linear regression for each subset is shown. (B) Peripheral blood antibody sequences were grouped by HCDR3 length (in amino acids) and the mean number of mutations for each HCDR3 length group was calculated for each donor. As in Figure 1A, the left panel shows the mean ± SEM for all donors in Group 1 and Group 2. The right panel shows the best-fit linear regression of each B cell subset for Group 1 donors. The percent of sequences within each HCDR3 length group containing non-frameshift insertions also was calculated. In the left panel, the mean percentage ± SEM for all donors in Group 1 and Group 2 is shown. In the right panel, the best-fit linear regression of each B cell subset is shown for Group 1 donors. (C) Peripheral blood antibody sequences from Group 1 healthy donors were grouped by donor into naïve and memory subsets and the percent of sequences containing long HCDR3s (≥24 amino acids in length, or two standard deviations above the mean HCDR3 length). The percentages for each donor are shown, with the mean ± SEM. (D) The donor groups from Figure 1C were analyzed for the frequency of very long HCDR3s (≥28 amino acids in length, i.e., three standard deviations above the mean HCDR3 length). The percentages for each donor are shown, with the mean ± SEM. The p values were determined using a one-way ANOVA. All statistically significant differences are indicated. * = p<0.05, ** = p<0.01, *** = p<0.001.

Figure 2. Skewed germline gene usage in antibodies containing long or very long HCDR3s.

Peripheral blood antibody sequences from Group 1 (n = 4) and Group 2 (n = 3) healthy donors were assembled into the following three groups by HCDR3 length: (1) all HCDR3s, which contains all sequences of any HCDR3 length; (2) long HCDR3s, which contains only sequences with a HCDR3 length ≥24 amino acids; and (3) very long HCDR3s, which contains only sequences with a HCDR3 length ≥28 amino acids. The frequency of each germline variable gene family, diversity gene family, and joining gene was determined for each HCDR3 length group. The mean frequency ± SEM is shown. All HCDR3 lengths were calculated using the IMGT numbering system. The p values were determined using a two-way ANOVA with Bonferroni post-tests. All statistically significant differences are indicated. * = p<0.05, ** = p<0.01, *** = p<0.001.

Figure 3. Long HCDR3s correlate with N-addition, P-addition and germline gene usage.

For all figure sections, the leftmost panel corresponds to peripheral blood antibody sequences from Group 1 (n = 4) and Group 2 (n = 3) healthy donors. The middle panel corresponds to antibody sequences from Group 1 donors, segregated by B cell subset. The rightmost panel corresponds to peripheral blood antibody sequences from HIV-infected donors (n = 4). (A) Peripheral blood antibody sequences were grouped by HCDR3 length (in amino acids) and the average N-addition length and P-addition length (both in nucleotides) was calculated for each HCDR3 length group. The mean length ± SEM is shown. Regression analysis of N-addition length produced a non-linear, exponential curve of best fit. Regression analysis of P-addition length produced a linear best fit. (B) Peripheral blood antibody sequences were grouped by HCDR3 length and the frequency of sequences encoding either diversity gene family 2 (D2) or diversity gene family 3 (D3) was calculated for each HCDR3 length group. The mean frequency of D2/D3 gene family use ± SEM for each HCDR3 length group is shown. Regression analysis produced a non-linear, sigmoidal curve of best fit. (C) Peripheral blood antibody sequences were grouped by HCDR3 length and the frequency of sequences encoding joining gene 6 (J_H6) was calculated for each HCDR3 length group. The mean frequency of D2/D3 gene family use ± SEM for each HCDR3 length group is shown. Regression analysis produced a non-linear, sigmoidal curve of best fit. (D) The frequency of sequences encoding both the J_H6 germline gene and D2/D3 germline gene family members was determined for each HCDR3 length group. The mean frequency of J_H6/D2/D3 gene family use ± SEM for each HCDR3 length group is shown. Non-linear regression analysis produced a sigmoidal curve of best fit.

Figure 4. Long HCDR3s preferentially use reading frames (RF) that result in reduced hydrophobicity.

(A) Peripheral blood antibody sequences from Group 1 and Group 2 healthy donors (left panel) or HIV-infected donors (right panel) were assembled into three HCDR3 length groups: (1) all HCDR3s; (2) HCDR3s of at least 24 amino acids; and (3) HCDR3s of at least 28 amino acids. (B) The percentage of sequences within each HCDR3 group using reading frame 2 of the diversity gene (RF2) was calculated for each HCDR3 length group. Non-linear regression analysis produced a sigmoidal curve of best fit (r² = 0.84). (C) Sequences that do not encode the joining gene J_H6 (top panel) or do encode J_H6 (bottom panel) were grouped by HCDR3 length and RF2 use within each HCDR3 length group was determined. The mean frequency ± SEM is shown. Non-linear regression analysis produced a sigmoidal curve of best fit. (D) The percentage of hydrophobic residues was calculated for each reading frame of every functional (lacking stop codons) diversity gene in the D2 and D3 germline gene families. The mean percentage ± SEM is shown for each reading frame. The RF2 hydrophobicity of the diversity genes which were shown to be increased in long HCDR3s are indicated by filled circles. The p values were determined using a Student’s two-tailed t-test. (E) The grand average of hydropathicity (GRAVY) was calculated for each functional reading frame of each D2 and D3 gene. A positive GRAVY score indicates hydrophobicity, and a negative GRAVY score indicates hydrophilicity. The mean GRAVY score ± SEM is shown for each reading frame. The RF2 GRAVY scores of the diversity genes that were shown to be increased in long HCDR3s are indicated by filled circles. The p values were determined using Student’s two-tailed t-test. All statistically significant differences are indicated. * = p<0.05, ** = p<0.01, *** = p<0.001.

Figure 5. Amino acid residues in J_H6 and RF2 of D3-3 germline gene segments are critical to binding and neutralization of HIV by long HCDR3-containing antibodies PG9 and PG16.

The HCDR3 amino acid sequences of HIV-specific mAbs PG9 and PG16 are shown aligned to the amino acid sequences of germline D3-3 and J_H6 genes. Dashes in the alignments indicate conservation with the respective PG antibody. Residues shown to be critical to binding or neutralization of HIV, defined as ≥10-fold decrease in either binding or neutralization when mutagenized , are indicated by filled circles.