Dynamics of the human antibody repertoire after B cell depletion in systemic sclerosis

Abstract

Systemic sclerosis with pulmonary arterial hypertension (SSc-PAH) is a debilitating and frequently lethal disease of unknown cause lacking effective treatment options. Lymphocyte anomalies and autoantibodies observed in systemic sclerosis have suggested an autoimmune character. We study the clonal structure of the B cell repertoire in SSc-PAH using immunoglobulin heavy chain (IGH) sequencing before and after B cell depletion. We found SSc-PAH to be associated with anomalies in B cell development, namely, altered VDJ rearrangement frequencies (reduced IGHV2-5 segment usage) and an increased somatic mutation-fixation probability in expanded B cell lineages. SSc-PAH was also characterized by anomalies in B cell homeostasis, namely, an expanded immunoglobulin D-positive (IgD+) proportion with reduced mutation loads and an expanded proportion of highly antibody-secreting cells. Disease signatures pertaining to IGHV2-5 segment usage, IgD proportions, and mutation loads were temporarily reversed after B cell depletion. Analyzing the time course of B cell depletion, we find that the kinetics of naïve replenishment are predictable from baseline measurements alone, that release of plasma cells into the periphery can precede naïve replenishment, and that modes of B cell receptor diversity are highly elastic. Our findings reveal humoral immune signatures of SSc-PAH and uncover determinism in the effects of B cell depletion on the antibody repertoire.

Trial registration: ClinicalTrials.gov NCT01086540 NCT02133781 NCT03020498 NCT03022396 NCT03022422.

Conflict of interest statement

Competing interests: The authors declare that they have no competing interests.

Copyright © 2017 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works.

Figures

Fig. 1. Study design and comparison of study groups across time

(A) Study design. Whether infusions were of rituximab or placebo depended on the study arm. (Note that not all the indicated time points were available for all participants.) (B) Number of distinct IGH sequences observed as a function of time, colored by study arm. (C) Fraction of observed sequences that had the IgA or IgG isotype as a function of time, colored by study arm. (D) Fraction of IgD or IgM sequences that displayed zero somatic mutations as a function of time, colored by study arm. (E) Summary of the signatures identified in (B) and (C): fold change in isotype-switched proportion from baseline to depletion versus fold change in sequence diversity from baseline to depletion. Here, “depletion” refers to the week 4 and week 12 visits; diversities and isotype-switched proportions were averaged across these two time points. Numeric data values are also provided in table S1. (F) Phylogenetic distance between each participant’s baseline repertoire and her repertoire at the week 36 visit. Week 36 was chosen as the common second time point because all but two study arm A recipients had recovered more than 10,000 sequences by then and times beyond week 36 had not been sampled for all participants. Numeric data values are also provided in table S2.

Fig. 2. Signatures of SSc-PAH compared with healthy Ig repertoires at baseline

(A) IgD proportion of the IGH sequence repertoire and average mutation load in IgD. (B) Average number of transcripts recorded per distinct IGH sequence (i.e., sum total of all transcripts across all observed IGH sequences divided by the number of observed IGH sequences). Each point corresponds to one study participant. Wilcoxon–Mann-Whitney test, P = 0.0025; n = 26. (Asterisks in the figure correspond to significance codes: *0.01 ≤ P < 0.05; **0.001 ≤ P < 0.01; ***0.0001 ≤ P < 0.001.) (C) Distribution of transcript abundances, separated by isotype compartment. Each curve corresponds to one study participant. Here, “abundance” is the number of transcript copies associated with a sequence, and “frequency” refers to the number of distinct sequences that display a given abundance value. (D) Percentage of IGH sequences that had a relative transcript abundance of at least 0.05% of the total repertoire, separated by isotype compartment. Points straddling the lower edge of the panel correspond to the value 0, which cannot be rendered on the logarithmic scale. Wilcoxon–Mann-Whitney tests: P = 0.026, n = 26 for IgD/IgM; P = 0.00061, n = 26 for IgG/IgA. [Note: When the single high value in SSc-PAH IgD/IgM is removed (participant SR3), the P value for the IgD/IgM comparison changes from 0.026 to 0.049 (<0.05); there are two points at 0 for healthy IgM/IgD]. (E) Proportion of somatic mutations that were considered to have become fixed in their B cell lineage, that is, present in at least 80% of sequences in that lineage. The proportion of fixed mutations was calculated separately for each lineage and then averaged across all lineages in the relevant size bin. Here, lineage size refers to the number of distinct IGH sequences in the lineage. Wilcoxon–Mann-Whitney tests, for lineage size bins from left to right: P = 0.00040, n = 25; P = 0.27, n = 24; P = 0.015, n = 20; P = 0.022, n = 19; P = 0.023, n = 15. (Note: The two highest points for the “15–19” bin, the highest point for the “20–24” bin, and the two highest points for the “25–29” bin correspond to five distinct SSc-PAH participants, namely, SP5, SP4, SP2, SR4, and SR2.) (F) Percentage of lineages that used the IGHV2-5 segment for each isotype compartment. Points straddling the lower edge of the panel correspond to the value 0, which cannot be rendered on the logarithmic scale. Wilcoxon–Mann-Whitney tests: P = 0.00051, n = 26 for IgD/IgM; P = 0.0045, n = 26 for IgG/IgA. [Note: When the single high point for SSc-PAH IgD/IgM is removed (participant SR3), the P value for the IgD/IgM comparison changes from 5.1 × 10−4 to 7.3 × 10−6. When the single point at 0 in SSc-PAH IgG/IgA is removed (participant SP2), the P value for the IgG/IgA comparison changes from 4.4 × 10−3 to 9.6 × 10−3.]

Fig. 3. Time courses of SSc-PAH signatures

Dashed black lines indicate the range of values observed in healthy participants at baseline. Note that the data displayed for study arms A and B in the present figure (longitudinal experiment) were acquired in a separate batch from the data on healthy participants (cross-sectional experiment; already described in Fig. 2) and so may not be directly comparable to the healthy data; we nevertheless include the dashed lines as a guide. (A) IgD proportion. (B) IgD mutation loads. (C) Percentage of lineages using the IGHV2-5 segment (all isotypes combined). (D) Proportion of somatic mutations observed in a lineage that were considered “fixed,” that is, present in at least 80% of the lineage’s sequences. The proportion of fixed mutations was calculated separately for each lineage containing between 5 and 29 distinct sequences and then averaged across all those lineages.

Fig. 4. Quantitative modeling of repertoire replenishment rates

(A) Fit for the naïve diversity M0 (number of nonmutated IgM sequences, Chao1 estimate) as a function of time, after the onset of replenishment. Only the two participants (SR1 and SR6) with the largest numbers of corresponding time points are shown. Numeric data values are also provided in table S3. (B) Fit parameters describing sequence generation (a), mutation (b), class switch (c1), and apoptosis (c2) rates. SR1 to SR6 are labels for individual participants. Participant SR3 is omitted because an insufficient number of time points after the onset of depletion were available. (C) Baseline naïve diversity versus naïve generation rate. R2 value is for the linear fit shown by the dashed line performed with intercept set to 0. Numeric data values are also provided in table S4. (D) Baseline naïve diversity versus time to onset of replenishment (defined as the interval between second visit and the earliest visit exhibiting more than 5000 distinct IGH sequences). Numeric data values are also provided in table S5. PPMC, Pearson product-moment correlation.

Fig. 5. Assessment of repertoire elasticity and compartment replenishment after depletion

(A) Similarity of various repertoire characteristics to their baseline state on a per-participant basis. The top two rows identify participants and time points; the other rows display data as a heat map for each of the indicated variables, compared with baseline using either a correlation coefficient or a fraction as indicated. Naturally, for each participant, the baseline value is 1; values then tend to drop during repertoire depletion and gradually increase again during replenishment. For isotype usage profiles (resolved by subisotype) and VDJ usage profiles, the value shown at each time point is the correlation to the baseline profile. A repertoire’s “mutation load profile” was defined as the number of nucleotide sequences having each mutation load from 0 to 10 (nucleotide) mutations; we display the correlation of such mutation load histograms to the baseline histogram. “Specific mutations” refers to the set of distinct V-segment amino acid mutations observed in a repertoire (each defined by the identity of the V segment, the position being mutated, and the identity of the observed residue). We displayed the fraction of baseline mutations that were observed at each postbaseline time point to indicate to what extent specific mutations observed at baseline return after B cell depletion. (B and C) Number of transcripts observed as a function of time, separated by compartment (naïve or antigen-experienced), relative to the total number of transcripts (from either compartment) observed at baseline. Note that in (B), the antigen-experienced fraction increases well before the naïve fraction, whereas in (C), it does not. Sequences were considered naïve if they were IgD or IgM and nonmutated and considered antigen-experienced if they had at least one mutation or were class-switched.

Fig. 6. Comparison of repertoire distributions to baseline

(A) Frequency of sequences with each subisotype relative to baseline. (B) Frequency of sequences with various mutation loads relative to baseline. (C) Frequency of lineages within various size bins relative to baseline. Here, lineage size refers to the number of distinct sequences in a lineage.