Citrullinated Protein Antibodies in Rheumatoid Arthritis

Affinity Maturation Drives Epitope Spreading and Generation of Proinflammatory Anti–Citrullinated Protein Antibodies in Rheumatoid Arthritis

Serra E. Elliott, Sarah Kongpachith, Nithya Lingampalli, Julia Z. Adamska, Bryan J. Cannon, Rong Mao, Lisa K. Blum, and William H. Robinson

Objective. Rheumatoid arthritis (RA) is character- were observed (P < 0.01). Shared complementarity-deter- ized by the presence of anti–citrullinated protein antibod- mining region 3 sequence motifs were identified across ies (ACPAs); nevertheless, the origin, specificity, and subjects. A subset of the plasmablast lineages included functional properties of ACPAs remain poorly under- members derived from later time points with divergent stood. The aim of this study was to characterize the evolu- somatic hypermutations that encoded antibodies that tion of ACPAs by sequencing the plasmablast antibody bind an expanded set of citrullinated antigens. Further- repertoire at serial time points in patients with estab- more, these recombinant, differentially mutated plas- lished RA. mablast antibodies formed immune complexes that Methods. Blood samples were obtained at up to 4 stimulated higher macrophage production of tumor serial time points from 8 individuals with established RA necrosis factor (TNF) compared to antibodies represent- who were positive for ACPAs by the anti–cyclic citrulli- ing earlier time point–derived lineage members that were nated peptide test. CD19+CD3ÀIgDÀCD14ÀCD20À less mutated. CD27+CD38++ plasmablasts were isolated by single-cell Conclusion. These findings demonstrate that sorting and costained with citrullinated peptide established RA is characterized by a persistent IgA tetramers to identify ACPA-expressing plasmablasts. ACPA response that exhibits ongoing affinity matura- Cell-specific oligonucleotide barcodes were utilized, fol- tion. This observation suggests the presence of a per- lowed by large-scale sequencing and bioinformatics sistent mucosal antigen that continually promotes the analysis, to obtain error-corrected, paired heavy- and production of IgA plasmablasts and their affinity mat- light-chain antibody gene sequences for each B cell. uration and epitope spreading, thus leading to the gen- Results. Bioinformatics analysis revealed 170 eration of ACPAs that bind additional citrullinated persistent plasmablast lineages in the RA blood, of antigens and more potently stimulate macrophage pro- which 19% included multiple isotypes. Among IgG- duction of TNF. and IgA-expressing plasmablasts, significantly more IgA-expressing than IgG-expressing persistent lineages Rheumatoid arthritis (RA) is characterized by the Supported by the NIH (National Institute of Allergy and generation of autoantibodies, including rheumatoid factor Infectious Diseases grant 5T32-AI-07290-30 to Dr. Elliott and grants and anti–citrullinated protein antibodies (ACPAs) (1). In U19-AI-11049103, U19-AI-110491, and U01-AI-101981 to Dr. Robin- son and National Institute of Arthritis and Musculoskeletal and Skin the present study, we sequenced the plasmablast antibody Diseases grant R01-AR-063676 to Dr. Robinson). repertoire over serial time points in patients with estab- Serra E. Elliott, PhD, Sarah Kongpachith, Nithya Lingampalli, lished RA to gain insights into the evolution of ACPAs, Julia Z. Adamska, BS, Bryan J. Cannon, MSc, Rong Mao, PhD, Lisa K. Blum, PhD, William H. Robinson, MD, PhD: Stanford University, and to further define the role of ACPAs in promoting the Stanford, California, and VA Palo Alto Health Care System, Palo Alto, pathogenesis of RA. California. ACPAs target citrullinated epitopes arising from Dr. Robinson is a founder, member of the Board of Direc- tors, consultant for, and owner of equity in Atreca, Inc. posttranslational modifications of arginine to citrulline by Address correspondence to William H. Robinson, MD, PhD, peptidylarginine deiminase (PAD) (2). Epitope spreading Division of Immunology and Rheumatology, CCSR 4135, 269 Cam- of ACPAs precedes the onset of clinical symptoms in RA pus Drive, Stanford, CA 94305. E-mail: [email protected]. Submitted for publication November 22, 2017; accepted in (3). Previous studies have demonstrated that recombinant revised form June 12, 2018. murine ACPAs can increase the severity of arthritis in

1946 EPITOPE SPREADING IN RHEUMATOID ARTHRITIS 1947

mice (4,5). A therapeutic strategy of B cell depletion using lineages in the RA blood that persisted across serial time rituximab provides clinical benefit in patients with sero- points and revealed shared heavy-chain (HC) and light- positive RA (6), and its efficacy is associated with mild chain (LC) third complementarity-determining region reductions in circulating ACPA levels (7). A mucosal drive (CDR3) motifs across subjects. Using antigen microarrays has been postulated in the pathogenesis of RA (8). Cigar- and enzyme-linked immunosorbent assays (ELISAs), we ette smoking represents a risk factor for ACPA+ RA (9), demonstrated that, compared to plasmablast lineage mem- and ACPAs are detectable in the sputum and/or serum bers derived from earlier time points, a subset of lineage from subjects with early RA and individuals at risk for members from later time points that were differentially developing RA (10). mutated encoded antibodies targeting additional citrulli- ACPAs may contribute to the pathogenesis of RA nated antigens. In vitro macrophage stimulation assays by stimulating immune effector cells, including macro- revealed that these later time point–derived, affinity- phages (11), which produce tumor necrosis factor (TNF) matured, differentially mutated recombinant plasmablast and other cytokines (12). Macrophages can be activated antibodies generated ICs that more potently stimulated by proinflammatory cytokines, immune complexes (ICs), macrophage production of TNF, a mechanism that may be and Toll-like receptor (TLR) agonists (13). Studies from postulated to promote the pathogenesis of RA. our laboratory and others have shown that ICs composed of RA blood–derived ACPAs and citrullinated proteins PATIENTS AND METHODS stimulate macrophages, both from the blood (14) and from the synovial fluid (15), to produce TNF. These Human blood samples. After obtaining written informed ACPA ICs stimulate macrophages via Fcc receptor type II consent from all potential study subjects under protocols c – approved by the Stanford University Institutional Review Board, (Fc RII) (15 18) and TLR-4 (16,17). Nevertheless, the blood samples were collected in heparin tubes at serial time role of ACPAs and the mechanisms by which affinity mat- points, at intervals of ≥2 months apart, from 8 anti-CCP+ uration and the evolving ACPA repertoire contribute to patients with RA (Table 1). All subjects were recruited from the the development and persistence of RA remain unclear. VA Palo Alto Healthcare System (Palo Alto, California). In this study, we utilized our cell-barcoding anti- Patients with RA met the American College of Rheumatology 1987 and 2010 classification criteria (22,23). Monocyte-derived body repertoire sequencing method (19) to investigate the macrophages were generated from blood samples that were evolution of the ACPA B cell response in RA. Previous obtained from the Stanford Blood Center. Peripheral blood studies utilizing this approach have focused on the blood mononuclear cells (PBMCs) were isolated by density-gradient IgG repertoire at a single time point in patients with estab- centrifugation with Ficoll-Paque PLUS (GE Healthcare Life lished RA (19) or the IgA and IgG repertoire at a single Sciences). Single-cell sorting of plasmablasts. CD19+CD3ÀIgDÀ time point in subjects prior to RA development or in those CD14ÀCD20ÀCD27+CD38++ plasmablasts were single-cell with early RA (20). In this study, at up to 4 serial time sorted using a BD FACSAria (BD Biosciences) as described points, we sequenced plasmablasts, which have been shown previously (19,24) (for more details see Supplementary Meth- to produce ACPAs (19,21), from the blood of 8 individuals ods, available on the Arthritis & Rheumatology web site at with established RA who were positive for ACPAs by the http://onlinelibrary.wiley.com/doi/10.1002/art.40587/abstract). – + For staining of PBMCs, we used fluorophore-conjugated antibod- anti cyclic citrullinated peptide test (anti-CCP ). Our find- ies against CD19 (HIB19; BioLegend), CD3 (UCHT1; BD Bio- ings from bioinformatics analysis revealed plasmablast sciences), IgD (IA6-2; BD Biosciences), CD14 (MφP9; BD

Table 1. Demographic and clinical characteristics of the study subjects with rheumatoid arthritis Number of tender and swollen joints* Age at T1, years Sex† T1 T2 T3 T4

Subject 1 63 F 18 14 (4 mos.) 7 (10 mos.) – Subject 2 61 M 5 2 (3 mos.) 3 (6 mos.) – Subject 3 60 M 4 10 (3 mos.) 22 (8 mos.) 18 (10 mos.) Subject 4 67 M 0 0 (4 mos.) –– Subject 5 76 M 0 10 (7 mos.) –– Subject 6 66 M 0 4 (12 mos.) –– Subject 7 83 M 10 1 (11 mos.) 12 (13 mos.) – Subject 8 74 M 5 4 (10 mos.) –– * The number of tender and swollen joints was summed at each time point (T1–T4), and the number of months (mos.) from initial sampling is indicated in parentheses. † The sex distribution at the VA Palo Alto Health Care System, where the samples were obtained, is predominantly male. 1948 ELLIOTT ET AL

Biosciences), CD20 (L27; BD Biosciences), CD27 (CLB-27/1; Life all antibodies were expressed on the human IgG1 Fc domain. Technologies), CD38 (HB7; BD Biosciences), IgA (IS11-8E10; In-house production was done as described previously, using an Miltenyi Biotec), and IgM (MHM-88; BioLegend). To identify Expi293 Expression System (ThermoFisher) with Expi293F cells ACPA-producing plasmablasts, the PBMCs were costained with (36) (see Supplementary Methods, http://onlinelibrary.wiley.com/ pooled citrullinated peptide tetramers comprising 14 citrulli- doi/10.1002/art.40587/abstract). nated peptides (see Supplementary Methods and Supplemen- Characterization of antibody binding specificities. The tary Table 1, http://onlinelibrary.wiley.com/doi/10.1002/art.40587/ QUANTA Lite CCP3.1 IgG/IgA kit (Inova) was used to test the abstract). CCP-binding activity of plasma and recombinant antibodies, in Cell barcode–enabled sequencing of the plasmablast accordance with the manufacturer’s protocol. For recombinant antibody repertoire. Sequencing of immunoglobulin genes from antibodies, an activity cutoff level of 3 SD above the mean value individual plasmablasts was performed using cell barcodes, as for negative control antibodies was used. Recombinant antibod- described previously (19,20) with minor modifications (see Sup- ies were further screened for binding to citrullinated and native plementary Methods, http://onlinelibrary.wiley.com/doi/10.1002/ peptides/proteins by planar and bead-based antigen arrays, as art.40587/abstract). Well-specific oligonucleotide barcodes (Tru- previously described (3,37). Planar arrays (50 lg/ml) and bead- Grade Oligos; IDT) were added to complementary DNA based arrays (25 lg/ml) were used to evaluate antibody binding (cDNA) by template switching, using Maxima Reverse Tran- to ~350 proteins/peptides or 40 proteins/peptides, respectively. scriptase (ThermoFisher). HC and LC genes were amplified Using fluorophore-conjugated anti-human IgG secondary anti- from pooled cDNA using gene-specific polymerase chain reac- bodies to detect binding, the median fluorescence intensities tion (PCR) primers, and paired-end sequencing (2 9 330) was were determined from quadruplicate print spots (planar array) performed using Illumina MiSeq. Samples underwent 2 rounds or >50 beads (bead array). of PCR (20) or 3 rounds of PCR (a list of the primers used is ELISAs were performed using peptides and full- provided in Supplementary Table 2, available on the Arthritis & length proteins citrullinated in vitro with PAD from rabbit Rheumatology web site at http://onlinelibrary.wiley.com/doi/10. skeletal muscle (Sigma) or recombinant PAD-4 (see Supple- 1002/art.40587/abstract). The antibody repertoire sequencing mentary Methods, http://onlinelibrary.wiley.com/doi/10.1002/ data have been deposited in the NCBI Sequence Read Archive art.40587/abstract). ELISA plates were coated with full-length (accession no. SRP150122). proteins (1–5 lg/ml) and peptides (10–15 lg/ml) diluted in Bioinformatics analysis of immunoglobulin sequences. bicarbonate/carbonate buffer (pH 9.5) (as described in Sup- Sequence data were processed, and the sequences of the HC plementary Methods, http://onlinelibrary.wiley.com/doi/10.1002/ V–D–J regions and LC V–J regions were aligned using ImMu- art.40587/abstract), and horseradish peroxidase–based detec- noGeneTics (IMGT) HighV-Quest (25). Thereafter, phyloge- tion was carried out with a 1-Step Ultra TMB-ELISA substrate netic trees were generated using a previously described method (ThermoFisher). The fold change in activity, based on an activ- (19,20,24) (for more details, see Supplementary Methods at ity cutoff level of 3 SD above the mean activity of negative http://onlinelibrary.wiley.com/doi/10.1002/art.40587/abstract). controls, was calculated. IMGT-based analysis of V-region somatic hypermutations (SHMs) Macrophage stimulation assays. Macrophages were was utilized to compare mutation levels across subjects and across isolated and cultured from plated PBMCs by adhesion with time points. Clonal families and persistent lineages were assigned 30 ng/ml of human macrophage colony-stimulating factor (M- on the basis of shared, IMGT-based assignments of HC and LC CSF; PeproTech) as described previously (16,17) (for more V–J–region genes and meeting the threshold of 60% identity details, see Supplementary Methods at http://onlinelibrary. (according to the Levenshtein distance [26]) within the HC and wiley.com/doi/10.1002/art.40587/abstract). This approach is com- LC CDR3 regions, similar to that previously determined for other monly used for culturing macrophages and has been proposed lineages with different identity thresholds (27). as a reference standard (13). To form plate-bound ICs, the To evaluate HC and LC CDR3 sequence motifs across plates were coated with 50 ll of protein (20 lg/ml), washed patients, cluster analysis was performed using CD-HIT (28,29). with phosphate buffered saline (PBS), and blocked with 1% Persistent lineage–derived HC and LC CDR3 sequences were low-endotoxin bovine serum albumin in PBS (150 ll). There- linked by a 4-X spacer and clustered according to a cutoff level after, 50 ll of recombinant, subject-derived antibodies (50 lg/ of 70% identity; results were visualized with igraph in R (30). ml) was added. After further washing, differentiated macro- For clusters containing >1 subject, linked HC and LC CDR3 phages were added (50,000 cells/well) in medium (5% fetal sequences were used to generate sequence logos (WebLogo bovine serum, without human M-CSF). To block FccRII and/ [31]). Plasmablast sequences were aligned to germline V–J– or TLR-4, cells were preincubated (at 37°C) for 1 hour with region genes using Clustal X version 2.0 (32), input to IgTree anti-CD32 (clone IV.3; StemCell Technologies) and/or an (33), and lineage trees were visualized in Graphviz version 2.38 InSolution TLR-4 inhibitor, TAK-242 (EMD Millipore). (34). Insertions and D-region gene segments were included in Lipopolysaccharide (50 ng/ml; Sigma-Aldrich) was used as a this mutation count from germline. Visualization of the align- positive control. After 24 hours of incubation (at 37°C), ments with Geneious version 7.0.3 (35) identified amino acid supernatants were harvested and levels of TNF were mea- differences between lineage members. sured by ELISA (PeproTech). Generation of recombinant monoclonal antibodies. A Statistical analysis. Statistical analysis was performed subset of subject-derived plasmablast antibody sequences rep- in GraphPad Prism (version 7). One-way or two-way analyses resenting clonal families/lineages and/or antibodies binding to of variance followed by Tukey’s test for multiple comparisons the citrullinated peptide tetramers, as well as negative control were used. P values less than 0.05 were considered significant. antibodies, were recombinantly produced in Dr. Robinson’slabo- The relationship between the number of tender and swollen ratory at Stanford University orbystaffatLakePharma(San joints and the mean normalized size of each clonal family was Carlos, CA). To ensure consistency in the characterization assays, determined by Spearman’s rank correlation. EPITOPE SPREADING IN RHEUMATOID ARTHRITIS 1949

lineages and antibodies that were only observed once (sin- RESULTS gletons) (Figure 1A). Profiling the evolution of the blood plasmablast The representative phylogenetic tree shown in antibody repertoire across serial time points in RA. We Figure 1A depicts trends observed in multiple RA sub- used our cell barcode–enabled antibody repertoire se- jects, and these observations are further highlighted in quencing method (19) to sequence the blood CD19+ the inset images. We identified clonal lineages with rep- CD3ÀIgDÀCD14ÀCD20ÀCD27+CD38++ plasmablast resentatives in multiple time points (insets 1 and 2 in antibody repertoires from up to 4 serial time points in Figure 1A), hereafter referred to as persistent lineages. samples from 8 anti-CCP+ patients with established RA Some clonal expansions consisted of plasmablasts (see Table 1 for patient characteristics). Bioinformatics expressing antibodies of a single isotype (insets 2 and 3 analyses assigned sequencing reads and obtained error- in Figure 1A), whereas others included multiple iso- corrected consensus sequences of the HC V–D–J types (insets 1 and 4 in Figure 1A). regions and LC V–J regions of each plasmablast. These By combining bioinformatics analysis of the data sets were used to construct phylogenetic trees that sequence data with flow-based data from staining with provided an overview of the plasmablast antibody reper- the citrullinated peptide tetramers, we identified candi- toire, including identification of clonally expanded families/ date ACPA-expressing clonal expansions and singletons

Figure 1. Plasmablast clonal lineages persist and evolve over time in anti–cyclic citrullinated peptide positive (anti-CCP+) patients with established rheumatoid arthritis (RA). A, The representative phylogenetic tree captures the relationships among heavy-chain (HC) and light-chain (LC) sequences from blood plasmablasts obtained at 2 time points (T1 and T2) from an anti-CCP+ individual with RA. Each leaf (colored according to isotype and time point) represents the concatenated, error-corrected, IMGT-aligned consensus HC and LC sequence for a single CD19+CD3ÀIgDÀCD14ÀCD20ÀCD27+CD38++ plasmablast, and the tree is anchored by HC V-region gene assignment. Branch colors indicate clonally expanded sequences (cyan) and sequences derived from plasmablasts that bound citrullinated peptide tetramers (pink), including singletons and those belonging to clonal expansions. Sequences selected for recombinant expression are indicated by red arrows and lines. Inset images highlight representative clonal families/lineages encoding antibodies comprising ≥1 isotype or present at ≥1 time point. B, Representative chord diagrams depict plasmablast clonal families and persistent lineages observed at serial time points T1–T4, with the number of months (mo) after T1 designated for subsequent time points. Persistent lineages include singleton observations at any 1 time point. Sector width and color indicate the size and mean number of HC mutations, respectively, for the given family/lineage at each time point. Connecting chords between related lineages at different time points are colored by isotype. Lineages shared by ≥3 time points in RA subject 3 are shaded darker to highlight these persistent lineages. 1950 ELLIOTT ET AL

(Figure 1A). The combination of different affinities with a mean of 30.6 mutations in the HC V-region and cell-surface receptor expression among clonal fam- gene among the 7 RA subjects with T1-derived, full- ily members may lead to differential binding and could length persistent lineage sequences (see Supplementary lead to false-negative findings among the sorted plas- Figure 2B, http://onlinelibrary.wiley.com/doi/10.1002/art. mablasts. Conversely, the increased avidity of the citrul- 40587/abstract). Overall, we identified 46 persistent lin- linated peptide tetramers may lead to some nonspecific eages (27%) demonstrating a high level of SHMs, with a staining (false-positive findings) during our flow-based mean of ≥40 nucleotide mutations in the HC V-region analysis. Thus, the flow-based tetramer staining method gene compared to germline, at ≥1 time point. represents a screening tool for ACPAs. Focusing on the persistent lineages, we deter- Our bioinformatics analysis identified clonally ex- mined the mean fold change in the normalized number panded plasmablast lineages, including families observed at of SHMs across different time points, which allowed us a single time point and lineages persisting across multiple to identify subsets of lineages that were either increas- time points, in blood samples obtained serially during the ing (>1.1-fold change), decreasing (<0.9-fold change), course of RA (Figure 1B; see also Supplementary Figure 1, or staying the same (results in Supplementary Figures available on the Arthritis & Rheumatology web site at http:// 2C–E, http://onlinelibrary.wiley.com/doi/10.1002/art.40587/ onlinelibrary.wiley.com/doi/10.1002/art.40587/abstract). abstract). Although most lineages were changing (i.e., Across the 8 patients with established RA, we identified increasing or decreasing) in SHM levels over time, an eval- 170 persistent lineages comprising single and/or multi- uation of each subset separately showed significantly more ple members at different time points. The chord plots lineages maintaining similar levels of SHMs (P < 0.05) (shownforRAsubjects1,2,3,and6inFigure1B) (Figure 2B). Across the individual RA subjects, we ob- highlight several important trends. We observed persis- served that different subjects exhibited varying percentages tent lineages in all RA subjects examined. In RA sub- of lineages belonging to each change category. RA subjects ject 3, select lineages persisted across 4 time points, 4 and 6 had the highest percentage of lineages with a spanning almost 1 year. Most lineages consistently decreasing number of SHMs, and RA subject 4 also had expressed 1 isotype, while 33 lineages (19%) included the second lowest percentage of lineages with an increasing multiple isotypes (Figure 1B; see also Supplementary number of SHMs. Compared to other subjects, RA subject Figure 1, http://onlinelibrary.wiley.com/doi/10.1002/art. 4 also exhibited a lower mutation level in the HC and LC 40587/abstract). Although some sequences were identical V-region genes among persistent lineage–derived se- across time points, certain plasmablast lineages exhibited quences. Interestingly, RA subject 4 showed low levels of dis- changes (e.g., increases or decreases) in the mean number ease, with no tender and swollen joints, at T1 and 4 months of HC V-region gene SHMs between time points (de- later (Table 1). picted as changing sector colors in Figure 1B). We characterized isotype usage among the IgA- and Further bioinformatics analysis of the SHM levels IgG-expressing plasmablasts. We observed a significantly in the RA blood showed that plasmablast sequences higher number of IgA-expressing plasmablasts captured at belonging to lineages shared across multiple time points each time point for each subject (P < 0.01) (results shown possessed a higher SHM level within the HC V-region in Supplementary Figure 2F, http://onlinelibrary.wiley. gene, as compared to SHM levels in the total sequenced com/doi/10.1002/art.40587/abstract). Furthermore, our anal- population of plasmablasts. The number of mutations ysis of persistent lineages revealed that a significantly higher from the aligned germline sequence was normalized by percentage of persistent lineages expressed antibodies of the length of the V-region identified. We observed a the IgA isotype compared to those expressing antibodies of statistically significant increase in the SHM level among the IgG isotype (P < 0.01) (Figure 2C). Although no corre- the persistent lineage–derived HC sequences compared lation with mutation levels was found, we observed an to that in the total population of plasmablasts at ≥12 inverse correlation between the number of tender and swol- months from time point 1 (T1) (n = 2) (Figure 2A). len joints and the normalized mean size of the clonal fami- Without sequence-length normalization, we observed a lies (see Supplementary Figure 2G, http://onlinelibrary. mean of 29.6 mutations in the HC V-region gene among wiley.com/doi/10.1002/art.40587/abstract). Taken together, the T1-derived sequences that belonged to persistent lin- our observations suggest that there is persistent antigen eages, across each subject’s average mutation level stimulation of the plasmablast response taking place in RA, (see Supplementary Figure 2A, available on the Arthritis and these persistent lineages predominantly express IgA. & Rheumatology web site at http://onlinelibrary.wiley. Among all of the persistent lineages identified, we observed com/doi/10.1002/art.40587/abstract). We observed simi- a subset that continued to increase in mean levels of SHMs lar trends when we focused on the full-length sequences, over months in anti-CCP+ patients with established RA. EPITOPE SPREADING IN RHEUMATOID ARTHRITIS 1951

Figure 2. A subset of the persistent lineages in anti-CCP+ RA patients display increased numbers of somatic hypermutations (SHMs) over time, and persistent lineages predominantly express the IgA isotype. A, Sequences from 8 anti-CCP+ RA patients were analyzed for differences in SHM levels among all consensus, IMGT-aligned sequences (HC and LC Total) compared to the persistent lineage–derived subset (HC and LC Shared). *=P < 0.05 by two-way analysis of variance (ANOVA) with Tukey’s correction. B, Persistent lineages were evaluated for SHM levels and catego- rized according to the mean SHM level over time, determined as increasing (Inc.), maintaining similar (Sim.) levels, or decreasing (Dec.). *=P < 0.05 by one-way ANOVA with Tukey’s correction. C, For IgA/IgG-expressing plasmablasts, the percentage of persistent lineages consistently expressing IgA was compared to that utilizing IgG or both IgG and IgA. ** = P < 0.01 by one-way ANOVA with Tukey’s correction. In A–C, sym- bols represent individual patients; bars indicate the mean Æ SEM. D, Persistent lineage–derived sequences were evaluated for similarities in the third complementarity-determining region, and clusters were visualized by igraph. Each dot represents the sequence from a single cell. Dots are colored by subject or by citrullinated peptide tetramer (Cit-tet) staining. Clusters with ≥1 member binding to citrullinated peptide tetramers are circled in red. Motifs of clusters comprising ≥1 subject were visualized using WebLogo version 3.5.0. See Figure 1 for other definitions.

Evidence of convergence across persistent lineages motif generation (cluster 6 in Figure 2D). Within the clus- from different patients. To characterize amino acid ters, we identified sequences produced by plasmablasts CDR3 sequence motifs among the persistent lineages that bound the citrullinated peptide tetramers (Fig- identified in the different RA subjects, we performed clus- ure 2D). Thus, CD-HIT cluster analysis demonstrated ter analysis on the paired amino acid HC and LC CDR3 that individuals with established RA generate persistent B sequences from individual plasmablasts, using CD-HIT cell lineages that exhibit shared amino acid motifs across (28,29). This analysis revealed multiple clusters containing individuals, some of which express ACPAs. shared sequence motifs (70% identity) across subjects Identification of the antigen targets of recombi- within the identified persistent lineages (Figure 2D). nant antibodies derived from RA plasmablast lineages. Sequences of 2 highly related clusters were grouped for We selected 71 plasmablast antibody sequences for recom- 1952 ELLIOTT ET AL

Figure 3. Recombinant antibodies representing clonal lineages bind citrullinated antigens. A and B, A CCP3.1 enzyme-linked immunosorbent assay was used to measure anti-CCP antibody binding in plasma from individual RA subjects (A) and among the expressed recombinant antibodies (B). Using the manufacturer’s protocol, activity units and the cutoff level for plasma activity (blue dotted line) were determined. For recombinant antibodies, the cutoff level for activity (red dotted line) was set to 3 SD above the mean binding activity of negative control (Neg Ctrl) antibodies. Data are the mean Æ SEM from duplicate determinations. C, Recombinant antibodies were evaluated for epitope specificity using an RA antigen microarray that includes ~350 citrullinated (cit) and native proteins/peptides. Microarrays were probed with recombinant antibodies representing clonal families/lineages and singletons derived from plasmablasts binding to citrullinated peptide tetramers. The heatmap depicts median fluorescence intensities (MFIs) from quadruplicate print spots. Gray indicates an antibody–antigen combination for which data could not be obtained. Each antibody is labeled according to the subject (S) and time point (T) from which it is derived. The recombinant antibodies derived from a plasmablast that stained for the citrullinated peptide tetramers during sorting are marked in red. Peptide sequences similar to those used to generate the tetramers are marked with a blue asterisk. Apo A1 = apolipoprotein A-I; cf = filaggrin; FibA = fibrinogen A; Vim = vimentin (see Figure 1 for other definitions). EPITOPE SPREADING IN RHEUMATOID ARTHRITIS 1953

binant expression (a complete list is provided in Supple- peptide binding. We used RA antigen planar arrays (Fig- mentary Table 3, available on the Arthritis & Rheumatology ure 3C) and bead-based arrays (results shown in Supple- web site at http://onlinelibrary.wiley.com/doi/10.1002/art. mentary Figure 3B, http://onlinelibrary.wiley.com/doi/10. 40587/abstract) as representatives of 1) clonal families/per- 1002/art.40587/abstract) to further characterize binding sistent lineages unique to individuals, 2) lineages possess- specificities. These analyses identified clonal family/lineage ing HC and/or LC CDR3 amino acid sequence motifs recombinant antibodies that bound citrullinated epitopes shared across subjects, and 3) singleton antibodies derived in fibrinogen, histones, clusterin, or a-enolase. Impor- from plasmablasts that bound citrullinated peptide tetra- tantly, certain later time point–derived persistent lineage mers. As negative controls, we used antibodies reactive to members (e.g., rAb66 and rAb82) bound an expanded set influenza (38), desmoglein 3 (39), or derived from ongoing of antigen targets compared to those targeted by antibod- projects investigating non-RA diseases. All antibodies ies derived from earlier time points (Figure 3C). were expressed on the human IgG1 Fc domain. Evidence of affinity maturation driving epitope Plasma from the 8 subjects with established RA spreading. In-depth characterization of the persistent lin- (Figure 3A) and all subject-derived recombinant anti- eages from the different RA subjects revealed mutation bodies (Figure 3B; see also Supplementary Figure 3A, patterns that suggested that affinity maturation drives epi- available on the Arthritis & Rheumatology web site at tope spreading. ELISA analysis of antibodies from a http://onlinelibrary.wiley.com/doi/10.1002/art.40587/abstract) lineage obtained from RA subject 1 confirmed that the were assayed using the CCP3.1 ELISA for citrullinated later time point–derived antibody rAb66 bound specific

Figure 4. Alternatively mutated plasmablast recombinant antibodies exhibit differential binding to citrullinated (Cit) antigens. A and B, Binding activity of recombinant antibodies, representing persistent lineages from rheumatoid arthritis (RA) subject 1 (A) and RA subject 8 (B), to full- length proteins and specific epitopes was assessed by enzyme-linked immunosorbent assay (ELISA). The fold change in antibody binding activity relative to an activity cutoff level of 3 SD above the mean value for negative control antibodies (Neg. Ctrls) (dotted line) was calculated. Values are the mean Æ SEM from duplicate determinations. C, Persistent lineage members were subjected to IgTree analysis to characterize the degree of relatedness between members and the distinct somatic hypermutations accumulated in each B cell–encoded antibody over the course of the affinity maturation process. Connectors between nodes of family trees indicate the number of mutations accumulated for each member compared to the germline (GL) heavy-chain (HC) and light-chain (LC) V–J–region gene sequences. Cyan-colored nodes represent lineage members binding to additional epitopes, as identified by ELISA. D, Nucleotide-based, full-length HC and LC antibody sequences were converted to amino acid sequences and aligned to germline sequences (Geneious). Segments corresponding to different complementarity-determining regions are indicated. The observed differences from germline of these evolving anti–citrullinated protein antibodies are highlighted and colored according to hydropho- bicity, with red indicating more hydrophobic residues. 1954 ELLIOTT ET AL

epitopes of H2B and H2A, while the member from an produced significantly higher levels of TNF in response to earlier time point, rAb65, did not bind these epitopes ICs generated with recombinant ACPAs that were derived (Figure 4A). We also observed substantial binding differ- from persistent lineages from 2 different RA subjects as ences for rAb65 and rAb66 in their binding to full-length, compared to that in response to citrullinated antigen in vitro–citrullinated H2A, H2B, and PAD, the latter of alone (P < 0.001) (Figures 5A and B). Importantly, the which can autocitrullinate (40) (see results in Supplemen- negative control antibody did not induce significantly tary Figure 4A, available on the Arthritis & Rheumatology higher macrophage production of TNF as compared to web site at http://onlinelibrary.wiley.com/doi/10.1002/art. that with citrullinated antigen alone (Figures 5A and B). 40587/abstract). Similarly, for a lineage derived from the In addition, citrullinated H2A and H2B significantly blood of RA subject 8, the later time point–derived, increased TNF production by macrophages compared to more-mutated rAb82 bound full-length, citrullinated H2B that in cultures with macrophages alone (P < 0.05) (Fig- and an H2B-derived epitope, whereas rAb81 did not bind ures 5A and B). However, plate-bound PAD, which these epitopes (Figure 4B). included all in vitro citrullination reagents but lacked IgTree-based analysis (33) revealed a bifurcation H2A or H2B, resulted in a level of TNF production simi- from a common parent, with 37 and 21 shared alterations lar to that produced by macrophages alone. for rAb65/rAb66 and rAb81/rAb82, respectively, followed To investigate the role of FccRII and TLR-4, we by continued, divergent evolution (Figure 4C). Direct preincubated cells with blocking reagents prior to incuba- alignment with germline HC and LC V-region sequences tion with immobilized ICs. Blockade of FccRII reduced demonstrated substantial differences between the lineage the levels of TNF produced by macrophages compared to members in regions outside of the CDR3 (e.g., CDR1, that with ICs alone, with statistically significant reductions CDR2, and framework regions) (Figure 4D). observed in cultures with rAb81 and rAb82 complexed Analyses by ELISA revealed that the persistent with citrullinated H2B (P < 0.05) (Figure 5B) and in cul- lineage–derived antibodies rAb71 and rAb72, which were tures with rAb66 complexed with citrullinated H2A or identified in RA subject 3 and displayed differential reac- PAD (Figure 5A). Furthermore, following preincubation tivity by planar antigen array, exhibited similar binding of the cells with the TLR-4 inhibitor, we observed a signif- reactivity to H2A, H2B, and an H2A-derived peptide (see icant reduction in the levels of TNF stimulated by citrulli- Supplementary Figure 4B, http://onlinelibrary.wiley.com/ nated antigens alone or in complex with recombinant doi/10.1002/art.40587/abstract). In the ELISA analyses antibody (each P < 0.001) (Figures 5A and B). As evaluating antibodies with similar CDR3 sequence reported previously (16,17), blockade of TLR-4 exhibited motifs, the findings confirmed that 2 clusters (C38 a dominant effect in the inhibition of TNF production by and C39) contained at least 1 antibody that bound citrullinated antigen-IC–stimulated macrophages. No dif- citrullinated antigens (see Supplementary Figures 4C ferences in TNF levels were observed between cell cul- and D, http://onlinelibrary.wiley.com/doi/10.1002/art. tures with the TLR-4 inhibitor alone and cell cultures 40587/abstract). The later time point–derived member with both the TLR-4 inhibitor and FccRII blocker. rAb133, from RA subject 6, exhibited differential bind- We further evaluated the effect of accumulated ing activity compared to that derived from an earlier SHMs on IC-based macrophage stimulation. Recombi- time point. In antigen microarray and ELISA analyses, nant antibodies observed at later time points and possess- we confirmed that certain later time point–derived per- ing higher levels of SHMs (i.e., rAb66 and rAb82) sistent plasmablast lineage members encoded antibodies generated ICs that stimulated macrophages to produce that differentially bound citrullinated antigens. Further- significantly more TNF compared to less-mutated anti- more, antibodies present at later time points possessed bodies derived from the same lineage at earlier time increased levels of SHMs and/or divergent SHMs. points (P < 0.05) (Figures 5A and B). Thus, individual Taken together, these findings suggest that affinity mat- recombinant ACPAs form ICs that stimulate macrophage uration over time drives epitope spreading of the TNF production. These results reveal that affinity matura- ACPA response. tion results in the generation of ACPAs that form ICs that Increased stimulation of macrophage TNF pro- more potently stimulate macrophage TNF production duction by affinity-matured recombinant antibodies. To through FccRII and TLR-4. examine the ability of representative, persistent plasmablast lineage–derived, recombinant antibodies to stimulate DISCUSSION macrophage production of TNF, we generated plate- bound ICs of these antibodies with citrullinated H2A, In this study, we characterized the evolving ACPA H2B, or PAD. Human monocyte–derived macrophages plasmablast response in individuals with established RA EPITOPE SPREADING IN RHEUMATOID ARTHRITIS 1955

Figure 5. Rheumatoid arthritis (RA) subject–derived recombinant antibody–citrullinated (Cit) antigen immune complexes (ICs) stimulate macrophages to produce tumor necrosis factor (TNF). Plate-bound ICs were formed by coating wells with a citrullinated antigen (H2B, H2A, or peptidylarginine deiminase [PAD]) followed by incubation with recombinant antibodies representing persistent lineages from RA subject 1 (A) or RA subject 8 (B). Monocyte-derived macrophages were added to each well, and after 24 hours, supernatants were harvested and TNF levels were measured using enzyme-linked immunosorbent assay. A portion of macrophages were preincubated with the Toll-like receptor 4 (TLR-4) inhibitor TAK-242 (TLR4B) and Fcc receptor type II (FccRII) inhibitor anti-CD32 clone IV.3 (FcB), prior to addition to the plate. Lipopolysaccharide (LPS) was used as a positive control, and cells alone with buffers served as negative controls (Neg Ctrl). The same controls are displayed in multiple panels for comparison. TNF levels were compared between antibodies observed at different time points (T1 and T2 or T1 and T3) within the same lineage, and TNF levels produced in the presence of inhibitors were compared to that with IC alone (*=P < 0.05; ** = P < 0.001, by two- way analysis of variance [ANOVA] with Tukey’s correction; TNF levels in cells incubated with citrullinated proteins without IC versus TNF levels in cells alone, † = P < 0.05 by one-way ANOVA with Tukey’s correction). Results are the mean Æ SEM. Representative data are shown, and similar trends were observed in 2 independent experiments. 1956 ELLIOTT ET AL

by integrating information on antibody sequences with A significantly higher percentage of the persistent data on the specificity and functional activity of the plasmablast lineages expressed IgA compared to IgG. encoded antibodies. We identified clonal families and per- Since IgA represents the dominant isotype in mucosally sistent lineages expressing ACPAs with the use of citrulli- driven immune responses, this suggests that a persistent nated peptide tetramers, and representative recombinant mucosal antigen stimulates the ongoing IgA plasmablast antibodies were characterized for their binding specificity response in established RA. This is consistent with the and functional properties. Our analysis revealed that findings from previous studies that demonstrated an ele- select divergently mutated plasmablast antibodies derived vated percentage of IgA plasmablasts in anti-CCP+ indi- from later time points exhibited differential binding speci- viduals at risk for RA (20), an association of periodontal ficity and increased potency to stimulate macrophages to disease with RA (45), and the presence of ACPAs in the produce TNF, as compared to earlier time point–derived sputum of individuals with early RA (10). In summary, lineage members that were less mutated. our findings demonstrate a continued, ongoing IgA plas- Our findings suggest that persistent antigen stim- mablast response in established RA, suggesting that a per- ulation of B cell lineages in individuals with RA results sistent mucosal antigen stimulus plays a key role in in continued affinity maturation and high SHM levels. mediating the persistence of RA. Persistent lineages shared IMGT-based assignments of Our results support a mechanism by which evolv- HC and LC V–J–region genes and 60% identity within ing B cell lineages produce ACPAs that contribute to the the HC and LC CDR3 (according to the Levenshtein pathogenesis of RA by stimulating macrophage TNF distance [26]). Persistent lineage–derived sequences production. Previous studies from our laboratory demon- exhibited higher levels of HC V-region gene SHMs as strated that citrullinated fibrinogen or H2B ICs, gener- compared to the total sequenced plasmablast popula- ated using pooled IgG from RA plasma, stimulated tion. Across the 8 subjects with RA, we observed a macrophages to produce TNF (16,17). Building upon mean of 29.6 mutations within the HC V-region gene these previous observations, we expressed the recombi- of persistent lineages at T1, and observed that 27% of nant antibodies on the IgG1 backbone as a research tool the persistent lineages averaged ≥40 mutations in the and demonstrated that individual, RA subject–derived HC V-region gene in at least 1 time point. ACPAs formed ICs capable of stimulating macrophage These mutation levels are substantially higher TNF production. Furthermore, persistent lineage mem- compared to those observed in HC V-region genes of bers observed at later time points, which possessed H1N1 virus–induced B cells (mean ~19 mutations per increased levels of SHMs, encoded antibodies that 1) patient) from 5 subjects (41) or IgG-expressing memory B exhibited differential binding to citrullinated antigens, cells from 3 healthy donors (mean Æ SD 18 Æ 8.1 muta- and 2) more potently stimulated macrophages to produce tions) (42). This high SHM level is likely generated TNF. We note that the antibodies derived from later time through germinal center (GC)–mediated affinity matura- points did not fully share all of the SHMs that were tion and lineage expansion, during which B cells undergo observed at the earlier time point; however, our IgTree iterations of division and mutation in the GC dark zone analysis showed that plasmablasts derived at different and selection in the light zone, with up to 6 divisions time points share a common parent cell. Thus, differen- observed in the dark zone (43). At a mutation rate of tial, continued affinity maturation from this common par- À ~10 3 per basepair per cell division, it is estimated that ent cell results in the generation of antibodies that exhibit there will be ≥1 mutation across HC and LC V-region epitope spreading and more potently promote inflamma- genes per every 2 divisions (44). Thus, we estimate that tion in RA. these highly mutated plasmablast lineages present in Data from this study provide evidence of a common patients with established RA resulted from >80 cell divi- process by which individuals with RA generate persistent B sions and repeated rounds of division/selection in the cell lineages with shared CDR3 sequence motifs. Previous GCs. At later time points, multiple persistent lineages studies provided evidence of convergent evolution of the exhibited increased and differential mutations among lin- antibody response, such as the generation of antibodies in eage members, while other lineages exhibited fewer muta- response to Dengue virus infection (46) or antibodies tions at later time points. Since we only sampled the against the CD4 binding site of HIV-1 (47). In patients repertoire at set time points, these observations may arise with established RA, we demonstrated that certain persis- through 1) recruitment of a common progenitor that tent B cell lineages shared CDR3 amino acid sequence undergoes successive rounds of division, mutation, and motifs across individuals. In 2 of these shared clusters, we selection in the GC, or 2) stimulation of distinct branches confirmed by ELISA that a representative member bound of the lineage tree at different time points. citrullinated antigens, suggesting that these motifs contain EPITOPE SPREADING IN RHEUMATOID ARTHRITIS 1957

key CDR3 residues that mediate citrullinated epitope AUTHOR CONTRIBUTIONS binding. All authors were involved in drafting the article or revising it There are several limitations to this study. First, we critically for important intellectual content, and all authors approved focused on characterizing the ongoing immune response the final version to be published. Dr. Robinson had full access to all of the data in the study and takes responsibility for the integrity of through sequencing of the plasmablast repertoire. We did the data and the accuracy of the data analysis. not sequence memory B cells, which might contain com- Study conception and design. Elliott, Robinson. mon B cell progenitors. Second, we performed single-cell Acquisition of data. Elliott, Kongpachith, Lingampalli, Adamska, Cannon. sequencing with limited depth. Future studies could utilize Analysis and interpretation of data. Elliott, Mao, Blum, Robinson. a droplet-based method to sequence a larger number of B cells. Third, recombinant antibody binding specificity was analyzed using an antigen microarray containing ~350 dis- REFERENCES tinct proteins/peptides, and it is likely that certain antigen 1. Helmick CG, Felson DT, Lawrence RC, Gabriel S, Hirsch R, targets are not included on this array. Fourth, to isolate Kwoh CK, et al. Estimates of the prevalence of arthritis and other rheumatic conditions in the United States: part I. Arthritis ICs for macrophage stimulation, we formed plate-bound Rheum 2008;58:15–25. ICs; however, this approach may lead to structural con- 2. Wang S, Wang Y. Peptidylarginine deiminases in citrullination, straints, potentially reducing the capacity of these ICs to gene regulation, health and pathogenesis. Biochim Biophys Acta 2013;1829:1126–35. stimulate macrophage TNF production as compared to 3. Sokolove J, Bromberg R, Deane KD, Lahey LJ, Derber LA, that in the presence of ICs formed in vivo. Finally, we Chandra PE, et al. Autoantibody epitope spreading in the pre- used 14 different citrullinated peptide tetramers to iden- clinical phase predicts progression to rheumatoid arthritis. PLoS One 2012;7:e35296. tify ACPA-producing plasmablasts; however, future stud- 4. Kuhn KA, Kulik L, Tomooka B, Braschler KJ, Arend WP, ies could more thoroughly characterize ACPA-expressing Robinson WH, et al. Antibodies against citrullinated proteins B cells by including additional citrullinated peptides and/ enhance tissue injury in experimental autoimmune arthritis. J Clin Invest 2006;116:961–73. or proteins. 5. Uysal H, Bockermann R, Nandakumar KS, Sehnert B, Bajtner E, In summary, sequencing blood plasmablasts from Engstrom€ A, et al. Structure and pathogenicity of antibodies patients with established RA over serial time points specific for citrullinated collagen type II in experimental arthritis. J Exp Med 2009;206:449–62. revealed that ACPA-encoding B cell lineages persist and 6. Isaacs JD, Cohen SB, Emery P, Tak PP, Wang J, Lei G, et al. evolve, exhibiting extensive, progressive, and differential Effect of baseline rheumatoid factor and anticitrullinated peptide SHMs. Intriguingly, these persistent lineages predomi- antibody serotype on rituximab clinical response: a meta-analysis. Ann Rheum Dis 2013;72:329–36. nantly expressed IgA, suggesting a continued mucosal 7. Cambridge G, Leandro MJ, Lahey LJ, Fairhead T, Robinson drive in established RA, an observation that may provide WH, Sokolove J. B cell depletion with rituximab in patients with important insight into the mechanisms underlying the per- rheumatoid arthritis: multiplex bead array reveals the kinetics of IgG and IgA antibodies to citrullinated antigens. J Autoimmun sistence of RA. Continued affinity maturation resulted in 2016;70:22–30. epitope spreading such that B cell lineage members 8. Demoruelle MK, Deane KD, Holers VM. When and where does observed at later time points expressed ACPAs with inflammation begin in rheumatoid arthritis? Curr Opin Rheuma- tol 2014;26:64–71. expanded citrullinated antigen reactivities. Importantly, 9. Klareskog L, Stolt P, Lundberg K, K€allberg H, Bengtsson C, differentially evolved ACPAs observed at later time points Grunewald J, et al. A new model for an etiology of rheumatoid formed ICs that more potently stimulated macrophage arthritis: smoking may trigger HLA–DR (shared epitope)– restricted immune reactions to autoantigens modified by citrulli- TNF production, highlighting the potential for an nation. Arthritis Rheum 2006;54:38–46. increased pathogenic role of these affinity-matured 10. Willis VC, Demoruelle MK, Derber LA, Chartier-Logan CJ, ACPAs. The level of TNF produced by IC stimulation of Parish MC, Pedraza IF, et al. Sputum autoantibodies in patients with established rheumatoid arthritis and subjects at risk of future macrophages was reduced by TLR-4 blockade, thus clinically apparent disease. Arthritis Rheum 2013;65:2545–54. demonstrating the potential of TLR-4 as a therapeutic 11. England BR, Thiele GM, Mikuls TR. Anticitrullinated protein target in RA, as has been suggested previously antibodies: origin and role in the pathogenesis of rheumatoid arthritis. Curr Opin Rheumatol 2017;29:57–64. (16,17,48,49). Taken together, our findings demonstrate 12. Kinne RW, Brauer€ R, Stuhlmuller€ B, Palombo-Kinne E, Burmester that affinity maturation plays a critical role in epitope GR. Macrophages in rheumatoid arthritis. Arthritis Res Ther 2000; spreading and generation of ACPAs that more potently 2:189. 13. Murray PJ, Allen JE, Biswas SK, Fisher EA, Gilroy DW, Goerdt drive macrophage TNF production in RA. S, et al. Macrophage activation and polarization: nomenclature and experimental guidelines. Immunity 2014;41:14–20. 14. Clavel C, Ceccato L, Anquetil F, Serre G, Sebbag M. Among human ACKNOWLEDGMENTS macrophages polarised to different phenotypes, the M-CSF- oriented cells present the highest pro-inflammatory response to the We thank the Stanford Functional Genomics Facility rheumatoid arthritis-specific immune complexes containing ACPA. and Stanford Shared FACS Facility for their assistance. Ann Rheum Dis 2016;75:2184–91. 1958 ELLIOTT ET AL

15. Laurent L, Clavel C, Lemaire O, Anquetil F, Cornillet M, Zabraniecki 32. Larkin MA, Blackshields G, Brown NP, Chenna R, McGettigan L, et al. Fcc receptor profile of monocytes and macrophages from PA, McWilliam H, et al. Clustal W and Clustal X version 2.0. rheumatoid arthritis patients and their response to immune complexes Bioinformatics 2007;23:2947–8. formed with autoantibodies to citrullinated proteins. Ann Rheum Dis 33. Barak M, Zuckerman NS, Edelman H, Unger R, Mehr R. IgTree: 2011;70:1052–9. creating immunoglobulin variable region gene lineage trees. 16. Sokolove J, Zhao X, Chandra PE, Robinson WH. Immune com- J Immunol Methods 2008;338:67–74. plexes containing citrullinated fibrinogen co-stimulate macrophages 34. Gansner ER, North SC. An open graph visualization system and via Toll-like receptor 4 and Fcc receptor. Arthritis Rheum 2011;63: its applications to software engineering. Softw Pract Exp 2000; 53–62. 30:1203–33. 17. Sohn DH, Rhodes C, Onuma K, Zhao X, Sharpe O, Gazitt T, 35. Kearse M, Moir R, Wilson A, Stones-Havas S, Cheung M, et al. Local joint inflammation and histone citrullination in a Sturrock S, et al. Geneious Basic: an integrated and extendable murine model of the transition from preclinical autoimmunity to desktop software platform for the organization and analysis of inflammatory arthritis. Arthritis Rheumatol 2015;67:2877–87. sequence data. Bioinformatics 2012;28:1647–9. 18. Clavel C, Nogueira L, Laurent L, Iobagiu C, Vincent C, Sebbag 36. Nair N, Feng N, Blum LK, Sanyal M, Ding S, Jiang B, et al. M, et al. Induction of macrophage secretion of tumor necrosis fac- VP4- and VP7-specific antibodies mediate heterotypic immunity tor a through Fcc receptor IIa engagement by rheumatoid arthritis– to rotavirus in humans. Sci Transl Med 2017;9:eaam5434. specific autoantibodies to citrullinated proteins complexed with fib- 37. Hueber W, Kidd BA, Tomooka BH, Lee BJ, Bruce B, Fries JF, rinogen. Arthritis Rheum 2008;58:678–88. et al. Antigen microarray profiling of autoantibodies in rheuma- 19. Tan YC, Kongpachith S, Blum LK, Ju CH, Lahey LJ, Lu DR, et al. toid arthritis. Arthritis Rheum 2005;52:2645–55. Barcode-enabled sequencing of plasmablast antibody repertoires in 38. Tan YC, Blum LK, Kongpachith S, Ju CH, Cai X, Lindstrom rheumatoid arthritis. Arthritis Rheumatol 2014;66:2706–15. TM, et al. High-throughput sequencing for natively paired anti- 20. Kinslow JD, Blum LK, Deane KD, Demoruelle MK, Okamoto body chains provides evidence for original antigenic sin shaping Y, Parish MC, et al. Elevated IgA plasmablast levels in subjects the antibody response to influenza vaccination. Clin Immunol at risk of developing rheumatoid arthritis. Arthritis Rheumatol 2014;151:55–65. 2016;68:2372–83. 39. Yeh SW, Cavacini LA, Bhol KC, Lin MS, Kumar M, Duval M, 21. Kerkman PF, Rombouts Y, van der Voort EI, Trouw LA, et al. Pathogenic human monoclonal antibody against desmoglein Huizinga TW, Toes RE, et al. Circulating plasmablasts/plasmacells 3. Clin Immunol 2006;120:68–75. as a source of anticitrullinated protein antibodies in patients with 40. Andrade F, Darrah E, Gucek M, Cole RN, Rosen A, Zhu X. rheumatoid arthritis. Ann Rheum Dis 2013;72:1259–63. Autocitrullination of human peptidyl arginine deiminase type 4 22. Arnett FC, Edworthy SM, Bloch DA, Mcshane DJ, Fries JF, regulates protein citrullination during cell activation. Arthritis Cooper NS, et al. The American Rheumatism Association 1987 Rheum 2010;62:1630–40. revised criteria for the classification of rheumatoid arthritis. 41. Wrammert J, Koutsonanos D, Li GM, Edupuganti S, Sui J, Arthritis Rheum 1988;31:315–24. Morrissey M, et al. Broadly cross-reactive antibodies dominate 23. Aletaha D, Neogi T, Silman AJ, Funovits J, Felson DT, Bingham the human B cell response against 2009 pandemic H1N1 influ- CO, et al. 2010 Rheumatoid arthritis classification criteria: an Amer- enza virus infection. J Exp Med 2011;208:181–93. ican College of Rheumatology/European League Against Rheuma- 42. Tiller T, Tsuiji M, Yurasov S, Velinzon K, Nussenzweig MC, tism collaborative initiative. Arthritis Rheum 2010;62:2569–81. Wardemann H. Autoreactivity in human IgG+ memory B cells. 24. Lu DR, Tan YC, Kongpachith S, Cai X, Stein EA, Lindstrom Immunity 2007;26:205–13. TM, et al. Identifying functional anti-staphylococcus aureus anti- 43. Gitlin AD, Shulman Z, Nussenzweig MC. Clonal selection in the bodies by sequencing patient plasmablast antibody repertoires. germinal center by regulated proliferation and hypermutation. Clin Immunol 2014;152:77–89. Nature 2014;509:637–40. 25. Alamyar E, Giudicelli V, Li S, Duroux P, Lefranc MP. IMGT/ 44. Kindt TJ, Goldsby RA, Osborne BA, Kuby J. Kuby Immunology. HighV-QUEST: the IMGT web portal for immunoglobulin (IG) New York: Macmillan; 2007. or antibody and T cell receptor (TR) analysis from NGS high 45. Potikuri D, Dannana KC, Kanchinadam S, Agrawal S, Kancharla A, throughput and deep sequencing. Immunome Res 2012;8:26. Rajasekhar L, et al. Periodontal disease is significantly higher in 26. Levenshtein VI. Binary codes capable of correcting deletions, non-smoking treatment-naive rheumatoid arthritis patients: results insertions and reversals. Sov Phys Dokl 1966;10:707–10. from a case-control study. Ann Rheum Dis 2012;71:1541–44. 27. Gupta NT, Vander Heiden JA, Uduman M, Gadala-Maria D, 46. Parameswaran P, Liu Y, Roskin KM, Jackson KK, Dixit VP, Lee Yaari G, et al. Change-O: a toolkit for analyzing large-scale B cell JY, et al. Convergent antibody signatures in human dengue. Cell immunoglobulin repertoire sequencing data. Bioinformatics 2015; Host Microbe 2013;13:691–700. 31:3356–8. 47. Wu X, Zhou T, Zhu J, Zhang B, Georgiev I, Wang C, et al. 28. Li W, Godzik A. Cd-hit: a fast program for clustering and com- Focused evolution of HIV-1 neutralizing antibodies revealed by paring large sets of protein or nucleotide sequences. Bioinformat- structures and deep sequencing. Science 2011;333:1593–602. ics 2006;22:1658–9. 48. Hatterer E, Shang L, Simonet P, Herren S, Daubeuf B, Teixeira 29. Fu L, Niu B, Zhu Z, Wu S, Li W. CD-HIT: accelerated for clus- S, et al. A specific anti-citrullinated protein antibody profile iden- tering the next-generation sequencing data. Bioinformatics 2012; tifies a group of rheumatoid arthritis patients with a Toll-like 28:3150–2. receptor 4-mediated disease. Arthritis Res Ther 2016;18:224. 30. Csardi G, Nepusz T. The igraph software package for complex 49. Davis ML, LeVan TD, Yu F, Sayles H, Sokolove J, Robinson W, network research. InterJournal 2006:1695. et al. Associations of Toll-like receptor (TLR)-4 single nucleotide 31. Crooks GE, Hon G, Chandonia JM, Brenner SE. WebLogo: a polymorphisms and rheumatoid arthritis disease progression: an sequence logo generator. Genome Res 2004;14:1188–90. observational cohort study. Int Immunopharmacol 2015;24:346–52.