Convergent mutation trajectories convert functional self-tolerance in IGHV4-34 B cells to genetic tolerance encoded in the antibody Authors: Christopher J. Jara, Sherin Zachariah, Katherine J. L. Jackson, Timothy J. Peters, Etienne Masle-Farquhar, Deborah L. Burnett, Eric Lam, Megan Faulks, Amanda Russell, Lisa Miosge, Stewart Smith, David C. Tscharke, Robert Brink, Christopher C. Goodnow, Joanne H. Reed Published: 2025 ABSTRACT SignificanceAnalysis of mice expressing a human autoantibody explains why many B cells making antibodies against self-antigens are not eliminated but circulate in a functionally tolerant state of anergy. Upon binding to a virus their progeny follow rapid, reproducible, convergent antibody mutation trajectories so that their antibodies become genetically specified not to bind self and only bind the virus. The results illuminate how antibodies normally gain their exquisite specificity. Many B cells circulate in a non-antibody-secreting, reversible state of functional self-tolerance called anergy. Why are self-reactive B cells retained in an anergic state in secondary lymphoid tissues instead of being eliminated in the bone marrow? This question has remained unresolved since the discovery in transgenic mice that B cells making surface IgM against self-lysozyme are not deleted but are actively prevented from proliferating and forming autoantibody-secreting plasma cells through “clonal anergy” involving a cell autonomous series of checkpoints (1–3). B cell anergy results from downregulation of surface IgM and increased production of surface IgD (4), dampening of surface Ig signaling by SHP1 tyrosine phosphatase to allow the cells to accumulate (5), inhibition of surface Ig signaling by SHIP and PTEN (6, 7), and chronic low concentration calcium signaling by surface Ig to induce transcription factors EGR1, EGR2, EGR3, and Nur77/NR4A1, which in turn suppress plasma cell differentiation and induce a set of “tolerance-response” mRNAs that encodes ZFP318, IgD, and known B cell negative regulators (4, 8–11). The majority of mature, naïve follicular B cells in mice display the cardinal features of anergy to varying degrees: surface IgM downregulation and induction of NR4A1 and EGR2 (2, 10, 11). Most mature B cells in humans also display low surface IgM and high IgD, but the complexity of immunoglobulins in both humans and mice obscures a clear view of the self-antigens provoking the anergy response. One important class of human autoantibodies reduces this complexity and provides a unique opportunity to address why anergic B cells are kept. Autoantibodies against the i-antigen, a linear repeat of the disaccharide N-acetyllactosamine (Galβ1-4GlcNAc) or its branched variant, I-antigen, displayed by erythrocyte and lymphocyte surface proteins (12, 13), represent one of the first demonstrations that the body has the potential to make antibodies against self (14–19). When secreted in the serum as IgM these autoantibodies can agglutinate erythrocytes at low temperatures in vitro, leading to their initial discovery as “cold agglutinins” that complicated blood typing. In some people, these autoantibodies are secreted in sufficient quantities to agglutinate and fix complement to erythrocytes in chilled parts of the body, leading to cold agglutinin disease and hemolytic anemia (14–16, 20, 21). Lymphocytes preferentially display i-antigen and can be agglutinated by anti-i autoantibodies (14, 15). B lymphocytes in particular have a high density of i-antigen, enabling binding of these autoantibodies to be measured more directly by flow cytometry (22–25) and enabling i-antigen bearing surface proteins to be analyzed immunochemically (26–28). Autoantibodies to i/I-antigen are unusually homogeneous because binding is determined principally by a single heavy chain variable element, IGHV4-34 (29–31). Ninety-four percent of IGHV4-34 IgM antibodies demonstrate i- or I-antigen binding in sensitive erythrocyte agglutination assays despite varying light chains and heavy chain CDR3 sequences (32, 33). This explains a “public idiotype” shared by most autoantibodies against i- or I-antigen, best defined by the monoclonal anti-idiotype antibody 9G4 (34, 35). IGHV4-34 is productively rearranged and expressed as mRNA by 5 to 10% of circulating naïve B cells, and consequently the 9G4 idiotype is present on a mean of 6% of follicular mantle zone B cells in human lymph nodes and tonsil but only 0.2% of serum IgM (35, 36). In addition to not forming plasma cells, additional evidence that 9G4+ follicular B cells expressing IGHV4-34 are anergic is their low surface IgM but high IgD and blunted intracellular calcium responses when these receptors are crosslinked (37). Not all human B cells making IGHV4-34 antibodies are blocked from responding to foreign antigens. Hypermutated IGHV4-34 IgG and IgA antibodies are readily detected in human tonsil (38) and by deep IgG repertoire mRNA sequencing of blood memory B cells from healthy donors (39). Hypermutated IGHV4-34 antibodies are found among IgG antibodies elicited by immunization against foreign antigens such as Rhesus D (RhD), rotavirus, tetanus toxoid, or vaccinia virus (39). Reversion of the somatic mutations in three of these IgG antibodies, one against RhD and two against vaccinia virus, revealed the unmutated ancestor of each bound to i-antigen but the V(D)J mutations had selectively eliminated the autoantibody specificity (40). These results are consistent with the process of “clonal redemption” from anergic B cell precursors defined in mouse antibody responses to lysozyme proteins (41, 42). However, it remains to be determined whether anergic IGHV4-34 B cells can actually be recruited into responses to foreign antigens, and with what efficiency and mutation trajectories do their progeny lose self-binding. Here, we addressed these questions by generating knock-in transgenic mice expressing a representative human IGHV4-34 antibody on a large percentage of the nascent B cell repertoire. We focused on the unmutated ancestor of antibody 589 (43), a highly mutated IGHV4-34 antibody against A14L, a nonglycosylated major protein of the vaccinia virus inner membrane (44). Antibody 589 was isolated from a healthy adult who had received vaccinia virus during the global vaccination effort that led to smallpox eradication in the 1970s. Three decades later the donor was boosted with vaccinia virus and after 14 d individual plasma cells were sorted from the blood and antibody heavy and light chain mRNAs sequenced and expressed (43). While 589 IgG does not bind i-antigen when measured directly by flow cytometry on human B cells, upon reverting the 19 heavy chain and 6 light chain substitutions the ancestral pre589 antibody is dual-reactive: binding self i-antigen and foreign vaccinia virus (40). By tracing B cells expressing this dual-reactive human autoantibody in mice, before and after vaccination with vaccinia virus, the experiments reveal an extraordinarily rapid and efficient process converting functionally tolerant, anergic IGHV4-34 B cells into genetically tolerant progeny cells making virus-specific antibodies that have lost binding to self. Results Ig-Gene Targeted Mice Expressing a Self-Reactive IGHV4-34 Ancestor of Antibody Against Vaccinia Virus. To develop an experimental model for analyzing B cells expressing human IGHV4-34 i-antigen binding antibodies, we focused on the unmutated ancestor of antibody 589, a highly mutated IGHV4-34 antibody against the A14L protein in the vaccinia virus membrane that was isolated from a single circulating plasma cell in a healthy individual vaccinated and boosted with vaccinia virus over a 30-y period (43). To test whether the unmutated ancestor (pre589) was reactive with mouse B cells, as we previously showed for human B cells (40), we expressed pre589 and 589 as recombinant human IgG1 and tested their binding to mouse bone marrow and spleen cells by flow cytometry. The unmutated ancestor pre589 IgG bound immature and mature mouse B cells intensely whereas the hypermutated 589 IgG had completely lost B cell binding (Fig. 1). Other cells in bone marrow or spleen were weakly stained by pre589 IgG, and this was also abolished in the mutated 589 antibody. These results are consistent with previous evidence that IGHV4-34 antibodies recognize O-linked, linear repeats of N-acetyllactosamine i-antigen on mouse and human B cells (23, 26, 27). The human IGHV4-34 antibody pre589 is also an autoantibody to mouse cells. Flow cytometric staining of the indicated mouse cell subsets by recombinant human IgG1 corresponding to the mutated 589 antibody isolated from a vaccinated donor (orange), the unmutated ancestor pre589 (red), or no recombinant IgG1 as a control (gray). Binding was detected using secondary antibody to human IgG. Results are representative of two independent experiments. Gene targeting with CRISPR Cas9 was used to produce transgenic C57BL/6 mice expressing pre589 immunoglobulin from the endogenous heavy and light chain genes. Promoter, leader, and VDJ exons encoding the rearranged unmutated ancestor pre589 VDJH (IGHV4-34*01, IGHD2-2*02, and IGHJ4*02) were integrated between the mouse joining segment (JH) clusters and the heavy chain intron enhancer, switch region elements, and mouse IgM constant region exons (Fig. 2A). Similarly, a promoter, leader, and rearranged pre589 VJK (IGKV1-39*01 and IGKJ1*01) exons were inserted between the native mouse joining (JK) elements and the kappa light chain intron enhancer and mouse kappa constant region exon (Fig. 2A). This leaves the possibility for upstream V(D)J rearrangements on the targeted Igh and Igk alleles, although we aimed to diminish mRNA expression from any such rearrangements by including a transcription termination and polyadenylation sequence between the upstream J elements and the integrated promoter and pre589 V(D)J exons (Fig. 2A). Mice with the targeted pre589-VDJH Igh and pre589-VJK Igk genes were crossed to produce heterozygous transgenic progeny expressing both pre589-IgH and pre589-IgK, termed pre589-tg mice. Pre589 heavy and light chains expressed from the mouse Igh and Igk loci support B cell maturation. (A) Schematic of targeted Igh and Igk loci. L, leader sequence exon; V, variable segment; D, diversity segment; J, joining segment; enh, enhancer; Cμ,Cδ, IgM, IgD constant region exons; Cκ, Kappa constant region exon; pA, polyadenylation site, pA. (B) Representative flow cytometric analysis of bone marrow CD19+ B220+ B-lineage cells stained with 9G4 antibody to IGHV4-34 and antibody to mouse kappa light chain. (C and D) Frequency of 9G4 and kappa positive or negative immature (C) or mature B cells (D) among bone marrow B-lineage cells. Each datapoint is from an individual pre589-tg (closed circle, n= 10) or wild-type (WT) (open circle, n= 9) mouse, pooled from three independent experiments. Results of Mann–Whitney U tests: P > 0.05 (ns), P < 0.05 (*), P < 0.01(**), P < 0.001 (***), P < 0.0001 (****). (E) Representative flow cytometric analysis of B-lineage cells in bone marrow or spleen from pre589-tg, pre589-tg Rag1−/−, and WT mice, stained with 9G4 antibody to IGHV4-34 and antibody to mouse kappa or lambda light chains. Flow cytometry analysis of B220+ cells in bone marrow revealed that immature CD93+ CD23− B cells expressing surface kappa light chains were twice as frequent in pre589-tg mice as in nontransgenic WT control mice, consistent with expression of the pre-rearranged targeted pre589 kappa chain. Three quarters of these nascent IgK+ cells expressed the pre589 H-chain detected with the 9G4 antibody to IGHV4-34 (Fig. 2 B and C). Most mature B cells also expressed the pre589 H-chain paired with kappa light chains, although ~25% of IgK+ immature and mature B cells lacked 9G4 staining (Fig. 2D). These 9G4− IgK+ cells must express a productive rearrangement of an endogenous mouse VDJH element instead, because they failed to develop in pre589-tg Rag1−/− mice (Fig. 2E). Very few of the 9G4+ cells were kappa chain-negative, with only 0.9% and 0.3% of B cells pairing the pre589 H-chain with lambda light chains in bone marrow or spleen, respectively (Fig. 2E). As expected, these were absent in pre589-tg Rag1−/− mice. It was not possible to determine the fraction of 9G4+ IgK+ cells expressing the pre589 kappa light chain since we lacked an appropriate idiotypic antibody for the pre589 light chain. However, in pre589-tg Rag1−/− mice 98% of mature splenic B cells were 9G4+ IgK+, establishing that the targeted pre589-H and -K chains support efficient B cell maturation in the absence of endogenous Ig rearrangement. Self-Tolerance in B Cells Expressing Pre-589. Despite the high frequency of B cells expressing the IGHV4-34 pre589 IgM on their cell surface, serum of pre589-tg and pre589-tg Rag1−/− mice lacked detectable secreted pre589 IgM when measured in the flow cytometric assay for B cell binding (SI Appendix, Fig. S1). This result parallels the observation in humans that 6% of B cells but only 0.2% of serum IgM carries the IGHV4-34 idiotype detected by monoclonal antibody 9G4, suggesting the 9G4+ cells are actively prevented from secreting cold agglutinin autoantibodies (35, 36). The absence of serum pre589 IgM parallels the earlier observation in double transgenic mice expressing hen egg lysozyme (HEL) as a self-antigen with anti-HEL B cells. High frequencies of B cells expressing cell surface anti-HEL IgM and IgD develop despite binding to self HEL protein, but in response to self-HEL binding the B cells make a tolerance response that actively inhibits secretion of anti-HEL IgM autoantibodies (1, 3, 11). To test for cellular evidence that the pre589-expressing B cells actively make a tolerance response, we assessed surface immunoglobulin expression of 9G4+ B cells in the periphery in pre589-tg mice. Active downregulation of surface IgM (45) is a cardinal feature of anergic B cells chronically binding autoantigen above a threshold occupancy of surface IgM and IgD receptors (1, 2). In wildtype mice with a polyclonal B cell repertoire, surface IgM is downregulated on individual follicular B cells across a range of 100-fold, apparently proportional to their self-reactivity (2, 4, 10, 11). Human follicular B cells exhibit a similarly broad distribution of surface IgM downregulation, with 9G4+ B cells having less surface IgM than the median 9G4− follicular cell (37). In the spleen of pre589-tg mice, 9G4+ transitional 1 (CD93+ CD23−) B cells had much lower surface IgM than 9G4− transitional cells that expressed other heavy chains (Fig. 3 A and B). Among mature follicular (CD93− CD23+) B cells, most 9G4-negative cells in pre589-tg mice and follicular B cells in WT control mice had downregulated IgM relative to transitional B cells, resulting in a broad range of surface IgM, and had increased surface IgD homogenously (Fig. 3 A and B). 9G4+ follicular B cells showed a narrower and greater degree of IgM downregulation, corresponding to 9G4-negative cells in the lower half of the surface IgM distribution, and had less surface IgD than their 9G4− counterparts. Comparable downregulation of surface IgM and IgD occurred on 9G4+ transitional and follicular B cells in pre589-tg Rag1−/− mice (Fig. 3A). Downregulation of pre589 surface IgM. Representative flow cytometric analysis of CD19+ B220+ B-lineage cells in the spleen from pre589-tg, pre589-tg Rag1−/−, and WT mice, stained with 9G4 antibody to IGHV4-34 and antibodies to CD93, CD23, and CD21. (A) Gating on 9G4+ or 9G4− B cells showing the % transitional (CD93+ CD23−) or follicular (CD93− CD23+) and showing surface IgM and IgD on these subsets. (B) Distribution of surface IgM expression among 9G4+ (red) and 9G4− (gray) transitional or mature follicular B cells in the spleen of five independent pre589-tg mice and among 9G4− cells from WT mice. (C and D) Frequency of 9G4+ or 9G4− B cells with the phenotype of follicular (CD23+ CD21+) or marginal zone (CD23− CD21++) B cells. Results of Mann–Whitney U tests: P < 0.0001 (****). Exclusion of B cells from the splenic marginal zone subset is another consequence of binding self-antigen and making a tolerance response in HEL/anti-HEL Ig-transgenic mice (46). In the spleen of pre589-tg mice, 9G4+ mature B cells accounted for the majority of follicular B cells (CD21low CD23+) but were completely absent from the marginal zone subset (CD21+ CD23low, Fig. 3 C and D). The latter was composed entirely of 9G4-negative B cells that expressed endogenous Ig-gene rearrangements, since no marginal zone B cells developed in pre589-tg Rag1−/− mice. Analysis of lymphocytes from the peritoneal cavity of pre589-tg mice revealed that 9G4+ B cells were excluded from the B1 (CD19+ CD23−) subset and only comprised B2 (CD19+ CD23+) cells with low surface IgM relative to 9G4− B2 cells (Fig. 4A). All the peritoneal B1 cells were 9G4-negative, as were a subset of B2 cells with a broad range of surface IgM expression (Fig. 4A). As part of a separate study comparing gene expression between B1 and B2 subsets, we sorted CD19+ CD23+ peritoneal B2 cells from pre589-tg mice without exposing the cells to 9G4 anti-idiotype antibody, and analyzed them by single cell global and V(D)J mRNA sequencing. V(D)J mRNA sequence analysis resolved the CD23+ B2 cells into 31% expressing the pre589 human IGHV4-34 H chain and 57% expressing diverse endogenous mouse IGHV rearrangements (Dataset S1). The pre589+ and pre589− cells nevertheless formed a single cluster by uniform manifold approximation and projection (UMAP) of their overall gene expression profile (Fig. 4B). Differential mRNA expression between pre589+ and pre589− CD23+ B2 cells. (A) Representative flow cytometry of CD19+ peritoneal cavity lymphocytes from pre589-tg or WT mice, showing the % of 9G4+ or 9G4− B cells that are CD23− (B1 subset) or CD23+ (B2 subset). (B–D) CD19+ CD23+ B2 cells were sorted (Left panel) and analyzed by 10X Chromium 5’VDJ and global single cell RNA sequencing. Single cells were split into those expressing human IGHV4-34 (pre589+) and those not expressing the targeted IGHV4-34 cassette (pre589−). These two subsets formed a single cluster by UMAP of global mRNA composition (Right panel). (C) Gene set enrichment analysis (GSEA) of: (Left) the set of mRNAs increased in tolerant MD4 transgenic CD23+ B cells from double transgenic mice expressing HEL as a self-protein compared to nontolerant CD23+ cells from MD4 single transgenic mice (4); (Right) the set of mRNAs decreased in CD23+ B cells with targeted inactivation of the Egr2 and Egr3 genes compared to CD23+ B cells from WT mice (11). (D) Volcano plot of differentially expressed genes between pre589+ and pre589− CD23+ B cells. Labeled are the differentially expressed mRNAs with the strongest evidence of increase in pre589+ cells belonging to the gene sets in (C). Differentially expressed genes (GEO Accession No. GSE308256) and gene sets for GSEA are detailed in Datasets S2 and S3, respectively. GSEA (Fig. 4C) revealed one of the most different sets (family-wise error rate, FWER < 0.001) between CD23+ B2 cells expressing pre589 and CD23+ B2 cells with other Igh sequences comprised a set of mRNAs increased in anergic CD23+ splenic B cells from double transgenic mice expressing anti-HEL (MD4) Ig receptors and soluble HEL self-antigen compared to nontolerant CD23+ cells from anti-HEL (MD4) Ig control animals lacking self antigen (4). Globally, 583 genes were differentially expressed between pre589+ and pre589− B2 cells (FWER < 0.05) (Fig. 4D). Of 171 genes significantly increased in pre589+ B2 cells, 139 (81%) were previously measured by Sabouri et al. (4) and 40 out of 139 genes (29%) were also increased in HEL tolerant B cells from double transgenic mice (Fig. 4D and SI Appendix, Table S1). These tolerance-induced mRNAs included Cd72 and Fcmr encoding inhibitory receptors on B cells, Zfp318 responsible for increased Ighd and IgD expression to dampen the tolerance response to self-antigen, and Sdc1 encoding the cell surface protein Syndecan1, which is induced proportionally to surface IgM downregulation (4). Chronic binding of self-antigen to HEL-tolerant follicular B cells repetitively stimulates an intracellular calcium-calcineurin-NFAT signaling pathway (8) to induce expression of the EGR2 and EGR3 transcription factors (4, 9). EGR2 is induced proportionally to surface IgM downregulation in normal mouse CD23+ B cells, and deficiency of EGR2 and its paralogue EGR3 in these cells diminishes surface IgM downregulation and diminishes expression of a set of tolerance-response genes with EGR2/EGR3-binding sites (11). GSEA revealed this EGR2/EGR3-dependent gene set was also enriched for genes with increased expression in pre589+ compared to pre589− CD23+ B2 cells (FWER < 0.001; Fig. 4C). Of 166 genes significantly increased in pre589+ cells and analyzed by Masle-Farquhar et al. (11), 34 genes (21%) are EGR2/EGR3 dependent. This set included HEL-tolerance-response genes defined by Sabouri et al. (4), notably Zfp318, Il4i1, Rgl1, Ndrg1, and Scimp (Fig. 4D). Indeed, the majority of genes with the strongest increase in pre589+ B cells were members of the gene sets that are increased in HEL-tolerant CD23+ B cells and/or are decreased in EGR2/EGR3 deficient CD23+ B cells (Fig. 4D). These results reveal an active tolerance response to self-antigen in B cells expressing IGHV4-34 pre589 surface immunoglobulin. Vaccinia Virus Triggers pre589 B Cells into Germinal Center Hypermutation Reactions. The hypermutated 589 antibody was isolated from a circulating plasma cell 2 wk after a vaccinia virus booster given to a healthy person who had been vaccinated against smallpox multiple times over a period of 30 y (43). Given the evidence above that the unmutated ancestor pre589 antibody is self-reactive and triggers a clonal anergy response in mature B cells, we tested whether vaccinia virus infection overcomes anergy to stimulate pre589 B cells to proliferate and hypermutate in germinal centers. Because of the high frequency of pre589-expressing B cells in pre589-tg Rag1−/− mice, we diluted a small number of these homogeneous cells into a normal B cell repertoire by adoptive transfer (SI Appendix, Fig. S2). Wildtype C57BL/6 CD45.1 congenic mice received 5 × 105 splenocytes from pre589-tg Rag1−/− C57BL/6 mice carrying the CD45.2 allotypic marker. After 24 h, the recipient mice received an intraperitoneal injection of live vaccinia virus. Splenic B cells from recipient mice were evaluated at 10, 16, and 24 d postinfection. CD45.2+ pre589 donor-derived B cells represented 0.1% of total B cells in the recipient mice on days 10 and 16 but were increased among Fashi CD38low germinal center B cells to a mean of 0.9%, 1.9%, and 0.5% of cells on days 10, 16, and 24, respectively (SI Appendix, Fig. S2). V(D)J mRNA sequencing of Igh and Igk transcripts in index sorted single CD45.2+ germinal center B cells showed that pre589 B cells undergo extensive hypermutation in vaccinia infected mice (Fig. 5A and SI Appendix, Fig. S3). Ten days postinfection, pre589 B cells acquired a mean of 1.8 nonsynonymous Igh mutations (range = 0 to 11, n = 388 sequences), increasing to a mean of 5.2 on day 16 (n= 341) and 8.3 on day 24 (n= 172). On day 10, a mean of 44% of Igh mutations were nonsynonymous, rising to 68% on day 16 and 72% on day 24. The mean number of synonymous Igh mutations increased less steeply. Fewer nonsynonymous mutations accumulated in the pre589 light chain, reaching a mean of 3 per sequenced cell on day 24 (SI Appendix, Fig. S4). Recurring VDJH mutation trajectories in CD45.2+ germinal center cells first reduce i-antigen binding and then increase binding to vaccinia virus. Single CD45.2+ germinal center B cells were sorted from recipient mice on days 10 (n= 10 recipient mice), 16 (n= 4 mice), and 24 (n= 6 mice) after vaccination as in Fig. 5, and the heavy and light chain mRNA was sequenced from each cell. Data are pooled from two independent experiments analyzed on days 10 and 24 and one experiment on day 16. (A) Number (Left) and percentage (Right) of VDJH mutations that are nonsynonymous or synonymous in individual cells. Each datapoint represents an individual B cell; box plots show mean and interquartile range; whiskers denote 5 to 95% range. Statistical significance was determined using Wilcox difference in means with Bonferroni correction. P> 0.05 (ns), P< 0.05 (*), P< 0.01(**), P< 0.001 (***), P< 0.0001 (****). (B) Number of sequenced cells with the five most recurrent VDJH mutations, either singly or in the indicated combinations. Filled dots convey the presence of the indicated mutation, with lines denoting specific combinations of mutations. (C) Amino acid sequence of pre589 VDJH, and location of recurring mutations tested in recombinant IgG antibodies. (D) Representative flow cytometric profiles of binding to mouse B cells by recombinant pre589 IgG with indicated VDJH mutations. (E) Binding of recombinant pre589 IgG with indicated VDJH mutations to mouse B cells measured by flow cytometry and to vaccinia virus by ELISA, compared to fully ancestral pre589 and fully mutated 589 IgG. Dashed lines indicate binding of pre589 and of 589 h/pre589k (with all 19 h-chain mutations but none of the k-chain mutations). Area of circles represents the frequency of cells with the indicated mutation combinations among sequenced cells at each timepoint. Convergent Mutation Trajectories Start with Mutations that Decrease Self-Binding. The preponderance of nonsynonymous heavy chain mutations induced by vaccinia virus in pre589 germinal center B cells focused on five amino acid substitutions that recurred in different cells and independent mice at each of the three timepoints (Fig. 5 B and C and SI Appendix, Fig. S3). Two were in CDR2H: S64R, acquired by 72% of sequenced cells on day 24; and S59R, acquired by 29% of cells on day 24. Three were in CDR3H: Y112.2F, in 74% of cells on day 24; S111.1G, in 67% of cells on day 24; and Y110F, in 23% of cells on day 24. With the exception of S59R, each of these mutations were acquired in the 589 antibody originally isolated from the vaccinia boosted human donor (43). On day 10 after vaccination in the mice, these 5 mutations had mostly been acquired as singletons in different pre589 germinal center B cells, with S64R, S111.1G, and Y112.2F most frequent (Fig. 5B). By day 16, 99% of the pre589 GC cells had at least one of the five mutations, 30% had paired two of these mutations, and 7.2% had a trio of these mutations. The most frequent pairs of mutations were S64R paired with S111.1G or with Y112.2F, present in 15% of pre589 GC cells on day 16. By day 24, 28% of pre589 descendants had acquired S64R, S111.1G, and Y112.2F as a trio, and 6% had acquired all four mutations present in the original human 589 antibody (Fig. 5B). These patterns recurred in 10 independent vaccinia-infected mice on day 10, four mice on day 16, and 6 mice on day 24. Thus, in the progeny of many individual pre589 B cells, frequently recurring mutation trajectories started in five different ways but converged within 24 d on combinations of mutations recapitulating those acquired in a person vaccinated repeatedly over 30 y. Of the 19 heavy chain nonsynonymous mutations in the original 589 plasma cell, 7 were found recurrently in pre589-descendent cells in the vaccinia infected mice, comprising the four described above plus S40T in FR2H (23% of cells on day 24), T65S in CDR2H (11% on day 24), and S92R in FR3H (8% on day 24). In addition to S59R, two other recurrent heavy chain mutations in the vaccinia infected mice were not present in the original 589 human antibody: I78M (23% of cells on day 24), and F87L (20% on day 24). By contrast, in the pre589 light chain only one nonsynonymous mutation recurred frequently: Y38S in CDR1H was present in 14% of sequenced cells on day 24 and corresponds to one of the seven substitutions in the original human 589 antibody (SI Appendix, Figs. S3 and S4). To determine the impact of recurring heavy chain mutations on binding to i-antigen and to vaccinia virus, recombinant pre589 IgG antibodies were synthesized as human IgG1 with the most common heavy chain mutations as singletons, pairs, trios, and quartets. Mutated heavy chains were all paired with the same unmutated pre589 IgK light chain to control for any light chain impact on binding. Expressed in parallel was the 589 heavy chain with all 19 nonsynonymous mutations paired with the unmutated pre589 light chain. Binding to i-antigen was measured using the flow cytometric assay for mouse B cell binding described in Fig. 1. Vaccinia binding was measured by ELISA with immobilized inactivated vaccinia virus. Binding in each assay was compared to pre589 IgG and the fully mutated 589 heavy chain paired with the pre589 light chain (589 h/pre589k), noting as above that the light chain acquired fewer mutations overall and only one recurrent mutation (SI Appendix, Fig S4). Each of the recurring single point mutations decreased binding to mouse B cells compared to the ancestral pre589 antibody (Fig. 5D). Y112.2F, which was the most frequent single mutation by day 16 (Fig. 5B), had the largest impact on decreasing self-reactivity to mouse B cells, whereas Y110F which rarely occurred as a single mutation had the smallest impact. The single mutations had negligible impact on binding to vaccinia. The most frequent pairs of mutations on day 16, S64R paired with S111.1G or with Y112.2F, almost fully abolished binding to mouse B cells, comparable to fully mutated 589 H chain paired with pre589 light chain. These pairs of mutations also increased binding to vaccinia virus, and the trio of S64R, S111.1G, and Y112.2F—which dominated responses at day 24—further improved vaccinia binding. Similar effects on B cell and vaccinia binding characterized the parallel mutation trajectory involving S59R, a substitution not found in the original 589 plasma cell. S59R was most frequently paired with S111.1G on day 16, and this mutation pair abolished B cell binding and increased vaccinia binding comparably to the more frequent S64R + S111.1G pair. None of the mutation trios nor quartets attained vaccinia virus binding equal to the 589 heavy chain + pre589k (589 h/pre589k) IgG control, which in turn bound vaccinia less than the fully mutated 589 antibody. Thus, the most frequently recurring mutation trajectories started with one of five single mutations that diminished self-reactivity with i-antigen, and then converged on pairs and trios of these mutations that abolished self-reactivity and progressively increased virus binding. Discussion The experiments above demonstrate how human antibody specificity is sculpted in the progeny of anergic B cells to yield antibodies that bind a virus but not self. The results explain why B cells with the potential to secrete IGHV4-34 autoantibody against the blood group i/I-antigens are retained, comprising 5 to 10% of the circulating human naïve B cell repertoire, instead of being deleted in the bone marrow. Using CRISPR Cas9 gene targeting to engineer mice where most of the B cells express a representative human IGHV4-34 antibody that binds to self i-antigen glycans on human and mouse B cells, the experiments demonstrate that B cells displaying this antibody as membrane IgM and IgD actively make a clonal anergy response, including downregulation of surface IgM and induction of EGR2-target genes, IgD, and other tolerance-response mRNAs. While these i-antigen reactive B cells are functionally silenced as measured by serum autoantibody secretion, they are not completely anergic in the sense originally defined by Nossal wherein cells lose the capacity to respond to antigen or mitogens (47, 48). The pre589 B cells responded to vaccinia virus by proliferating and hypermutating in germinal centers despite competition with a large excess of other B cells. The progeny of anergic IGHV4-34 B cells efficiently and reproducibly followed a set of mutation trajectories to convert a dual-reactive antibody against self and vaccinia virus into a virus-specific antibody. This process begins with rapid selection within 10 d for five alternative mutations that diminish binding to self i-antigen but preserve binding to vaccinia virus, acquired by 99% of the daughter cells by day 16 after vaccination. These different starting trajectories rapidly converge, with 30% of daughter cells on day 16 combining 2 or 3 of the five starting mutations to abolish self-reactivity and confer higher binding to the infecting vaccinia virus. Remarkably, these specificity maturation trajectories converge on a quartet of mutations acquired in the human 589 plasma cell in the original smallpox-vaccinated donor, underscoring how well the humanized mouse experiments match human infection itself. The findings here provide experimental evidence confirming and extending previous observations that IGHV4-34 B cells are anergic. Stevenson previously showed that B cells with the 9G4 idiotype are frequent in the follicular mantle zone of human lymph node and spleen but make a 30-times lower contribution to serum IgM (35, 36), and these conclusions were reinforced by Pugh-Bernard et al. (49) and Cappione et al. (37). Similar to HEL-tolerant B cells in transgenic mice (8, 50), Cappione et al. (37) showed human 9G4+ B cells made a diminished intracellular calcium elevation upon stimulation of surface IgD and IgM, although it was not possible to exclude the possibility that prior engagement by the 9G4 antibody had diminished the calcium response. The induction of calcium-responsive, calcineurin-sensitive B cell tolerance-response genes and EGR2/EGR3 target genes in pre589 (IGHV4-34+) B cells revealed here, using single cell mRNA sequencing that obviates the need to stain with anti-idiotype antibody, supports the conclusion that i-antigen binding B cells are regulated by modulating the amplitude and frequency of intracellular calcium responses to surface IgM and IgD (8, 9, 51) to disfavor plasma cell responses and favor germinal center reactions (4, 11, 39). The similarities in the active B cell anergy response in IGHV4-34 B cells to lysozyme-tolerant B cells is informative for human primary cold agglutinin disease and B cell lymphoproliferative disease. The anergy response in lysozyme tolerant B cells selectively dampens surface immunoglobulin signaling through CARD11 to NF-κB (8, 52). Escaping the dampening of this pathway in anergic IGHV4-34 B cells would explain why CARD11 gain-of-function mutations are acquired in 30% of the clonal expansions of IGHV4-34 IgM B cells secreting i/I autoantibodies in primary cold agglutinin disease but are rarely found in IgM monoclonal gammopathies secreting other antibody specificities (18, 20, 21, 53). The human antibody mutation trajectories followed by pre589 B cells in the germinal center of vaccinia infected mice have similarities and differences with those previously described in germinal center B cells undergoing clonal redemption and specificity maturation of a mouse antibody reacting to self and foreign lysozyme variants. In the lysozyme studies, a bispecific antibody that began with equal affinity against self-HEL3X protein and foreign duck egg lysozyme (DEL) or foreign flexible HELR101D rapidly evolved into monospecific antibodies against the foreign antigen with 5,000- or 1,900-fold difference, respectively, in binding affinity for foreign over self (41, 42). As here, the recurrent trajectories started with mutations that diminished affinity for self but had negligible effect on affinity for foreign. However, the large number of alternative starting points used here contrasts with the single dominant starting point in specificity maturation between self-HEL3X protein and foreign DEL. The two lysozymes are rigid and structurally very similar, such that only a single structural change in the conformation of CDR1H of the antibody provided a way to diminish binding to self while preserving binding to foreign (41). The self and foreign targets of the pre589 antibody represent the opposite extreme: i-antigen is a polysaccharide whereas vaccinia virus A14L is a nonglycosylated integral membrane protein (44). The large chemical difference between the two antigens presumably explains how five different single mutations in CDR2H or CDR3H diminish binding to the self-polysaccharide but preserve binding to the viral protein. It is notable that an FR1H hydrophobic patch mutation that diminishes i-antigen binding and was present in the human 589 antibody, Q6E, was not a prevalent starting point for the trajectories found in the mice. We previously observed Q6E also diminishes binding to vaccinia virus (40) and this backward step may disfavor it relative to the other starting points, as observed for backward step mutations in the lysozyme experiments (41, 42). The finding that heavy chain mutations in FR1 were not used in pre589 specificity maturation and that Q6E diminishes binding to vaccinia virus implies that the unique IGHV4-34 FR1H sequence contributes to binding both antigens, as also observed from analyzing Q6E and other FR1H mutations for IGHV4-34 antibody binding to i-antigen and the erythrocyte integral membrane protein RhD (54). Instead of FR1H hydrophobic patch mutations, selective loss of i-antigen binding during pre589 specificity maturation was achieved with mutations in the heavy chain CDR2 and CDR3, consistent with previous evidence that sequence variation in CDR3H modulates i/I-antigen binding by IGHV4-34 cold agglutinins (55). Interestingly, the two CDR2H mutations that diminish i-antigen binding by pre589 introduce an arginine, a residue that can increase or decrease antibody specificity in a context-dependent manner (56). Identical CDR2H Ser>Arg substitutions were present in 4 of 20 unique IGHV4-34 IgG or IgA sequences from human tonsil (38). Beyond virus infection, the results here also provide insights into the antibody response in people undergoing alloimmunization to mismatched erythrocytes. RhD-negative individuals inherit a null allele for the twelve-spanning transmembrane protein on erythrocytes, RhD. When they are exposed to RhD+ erythrocytes, the foreign RhD protein triggers antibodies that specifically agglutinate RhD+ erythrocytes and cause erythroblastosis fetalis in RhD- mothers carrying an RhD+ fetus. RhD alloimmunization is not reported to induce appreciable anti-i/I autoantibodies. Nevertheless, 70% of the elicited anti-RhD IgM B cells use IGHV4-34, 90% of these paired with a lambda light chain, and 94% of these IGHV4-34 antibodies also bind i-antigen in sensitive agglutination tests (32, 33). These antibodies displayed between 1 and 7 heavy chain mutations consistent with postgerminal center unswitched memory B cells (57). The dominance of IGHV4-34 anti-i antibodies is consistent with the findings here that i-antigen binding anergic B cells are not unresponsive but proliferate and hypermutate if they also bind an immunogenic foreign protein, whether that be on a virus or on mismatched human erythrocytes. In contrast to the diverse mutation trajectories available with vaccinia virus A14L, targeting RhD constrains mutation trajectories to eliminate i-antigen binding (54). RhD-null individuals still display a closely related self-protein on their erythrocytes, RhCE. The foreign RhD and self RhCE proteins differ by only 6-10 surface exposed amino acids, constraining how antibodies can bind specifically to RhD and not RhCE. This constraint is met by IGHV4-34 lambda antibodies elicited by RhD alloimmunization, as they do not agglutinate RhD- RHCE+ erythrocytes. But to bind RhD in this way, the IGHV4-34 lambda paratope depends on VH residues that also confer binding to self i-antigen (54), blocking mutation trajectories to eliminate i-antigen binding. B cells making high affinity anti-RhD IgG in alloimmunized people rarely use IGHV4-34 and primarily use an unrelated variable element, IGHV3-33 (57). A rare exception is an IGHV4-34 kappa IgG hybridoma isolated from an alloimmunized person, which had 18 heavy chain mutations and bound RhD but not i-antigen (58). Like the IGHV4-34 lambda clones, the unmutated ancestor of the IGHV4-34 kappa clone bound i-antigen (40) but its’ different RhD-binding paratope involving a kappa V element opened up a mutation trajectory to selectively eliminate i-antigen binding but preserve RhD binding, allowing subsequent progeny to accumulate more mutations and acquire high affinity against RhD (58). Anergic B cells displaying surface IgM and IgD against the gp120/gp41 envelope protein of HIV start from more profound functional tolerance than pre589 B cells, and follow more complex mutation trajectories during HIV infection. Self reactivity in the unmutated common ancestor of the human antibody CH103, which targets the CD4 binding site of HIV gp120, caused greater surface IgM downregulation and fewer mature B cells in Ig gene knock-in mice than observed here for pre589 B cells. CH103 ancestor cells were nevertheless stimulated to hypermutate during HIV infection in the original human donor, paradoxically increasing binding to a broad panel of self proteins during acquisition of mutations conferring broad virus-neutralizing activity despite separate mutations conferring each activity (59, 60). Functional tolerance was even more profound—with no accumulation of mature B cells—in Ig-gene knock-in mice where anergic B cells displayed the unmutated ancestor of human antibody 2F5, targeting the HIV gp41 membrane proximal external region. Cellular progeny nevertheless acquired antibody mutations during HIV infection that diminished but did not abolish the strong self-reactivity, possibly because retaining the capacity to insert into self-membrane lipids is necessary for virus-neutralizing activity (61). Lack of available mutation trajectories to convert B cell functional tolerance by clonal anergy into genetic tolerance within the antibody sequence itself is likely to explain cold agglutinin autoantibodies arising in people with atypical pneumonia. By contrast with other lung infections, 69% of people with atypical pneumonia from Mycoplasma pneumoniae are triggered to make cold agglutinins peaking in their serum 11 to 24 d after symptom onset, with high titers in 8% resulting in hemolytic anemia and correlating with more extensive lung involvement (62, 63). Sialylated linear or branched poly-N-acetyllactosamine chains are the host cell target of receptors on the attachment organelle of M. pneumoniae, enabling the bacteria to adhere and glide along the surface of human cells (64–66). Consequently, the immunogenic bacterial cells become decorated with self i/I-antigen (67), which will break anergy in IGHV4-34 B cells based on the results here with vaccinia infection. Unlike the 99% of daughter cells responding to vaccinia infection here, hypermutating IGHV4-34 progeny in people with atypical pneumonia will find no mutation trajectories to lose binding to self i/I-antigen while retaining binding to i/I-antigen on Mycoplasma particles, since the latter is the self i/I-antigen. In the lysozyme model, when immunogenic foreign erythrocytes are decorated with lysozymes that are structurally identical to those on the surface of the mouse’s own cells, there was no pathway for specificity maturation and the mutating germinal center progeny transiently followed doomed mutation trajectories that eliminated antigen binding altogether (41, 42). Given a high burden of M pneumoniae bacteria in the lungs, and high frequency of i/I-antigen binding IGHV4-34 B cells in the circulation, transient production of anti-i/I autoantibodies in most people with this infection is likely to reflect the reversible nature of B cell anergy (39, 68) coupled with the lack of mutation trajectories to eliminate self-reactivity genetically. Lack of mutation trajectories to convert B cell functional tolerance to genetic tolerance in the antibody itself may explain another public idiotype autoantibody, cold-precipitating serum IgM rheumatoid factors against IgG encoded by IGHV1-69, which are produced when self IgG decorates circulating virus particles in 50% of people with chronic hepatitis c virus infection (69). Functional and genetic tolerance in B cells making i/I autoantibodies—one of the first autoantibodies to be discovered—may thus serve as a key reference point for how other human autoantibodies are normally regulated and how that breaks down in autoimmune disease. Materials and Methods IGHV4-34 Antibody Sequences. The mutated immunoglobulin sequence of antibody 589, originally identified in a healthy individual after vaccinia virus immunization (43), was obtained from the NCBI database. Using the International Immunogenetics Information System (IMGT) V-quest junction analysis tool (70, 71), the 589 heavy and light chain sequences were reverted to its inferred unmutated germline ancestor, “pre589” (40). Generation of pre589-tg and pre589-tg Rag1−/− Mice. The pre589-tg and pre589-tg Rag1−/− genetically modified mouse lines were produced at the Mouse Engineering Garvan/ABR (MEGA) Facility (Moss Vale and Sydney, Australia) using CRISPR/Cas9-mediated gene targeting in C57BL/6J single-cell embryos (72). To target the Ig heavy chain gene and produce the pre589-IgH transgenic line, a homologous recombination plasmid was prepared that contained a 1,464 bp fragment made up of the following sequences (5′ to 3′): i) bGH polyadenylation site (208 bp), ii) promoter from mouse C57BL/6J Ighv3-8 (676 bp), iii) the germline reverted “pre589-IgH” (IGHV4-34*01), IGHD2-2*02, and IGHJ4*02 of 589 (580 bp). This fragment was inserted 69 bp 3′ of the mouse Ighj4 gene using an sgRNA based on the Cas9 target site TGAGCATTGCAGACTAATCTTGG (PAM site underlined and italicized). To target the Ig kappa chain gene and produce the pre589-IgK transgenic line, a homologous recombination plasmid was prepared that contained a 1,420 bp fragment made up of the following sequences (5′ to 3′): i) bGH polyadenylation site (208 bp), ii) promoter from mouse C57BL/6J Igkv5-45 (637 bp), iii) the reverted “pre589-IgL” (IGKV1-39*01) and IGKJ1*01 of 589 (575 bp). This fragment was inserted 111 bp 3′ of the mouse Igkj5 gene using an sgRNA based on the Cas9 target site GGGCTCATTATCAGTTGACGTGG (PAM site underlined and italicized). Founder mice for the pre589-IgH and pre589-IgK lines were identified by PCR, bred with WT C57BL/6J mice, and then crossed to produce pre589-tg mice that were heterozygous for the pre589-IgH and pre589-IgL alleles. Pre589-tg mice were crossed with homozygous Rag1 (Recombination Activating Gene-1) deficient (Rag1−/−) mice to prevent the recombination of endogenous Ig variable regions and ensure all B cells expressed the pre589 BCR sequence. C57BL/6J mice were used as WT controls in all flow cytometry experiments. All mice used in this study were 8 to 14 wk old, bred and cared for at the Australian BioResources (ABR) facility in pathogen-free environments. The Animal Ethics Committees at the Garvan Institute of Medical Research and Australian National University approved all protocols and procedures. Flow Cytometry Sample Preparation and Immunophenotyping with 9G4. 9G4 antibody (purified rat anti-human IGHV4-34*01; IgM Biosciences) was conjugated to Alexa Fluor 647 (AF647) using an antibody labeling kit (Life Technologies), according to the manufacturer’s instructions. Mouse bone marrow and spleen single cell suspensions were washed and kept in FACS buffer (PBS/5% FCS/0.02% sodium azide). All antibody incubation steps were completed on ice and in the dark, with centrifugations at 2,000 rpm for 3 min at 4 °C. Cells were incubated with anti-CD16/32 (ebioscience) Fc block for 15 min, washed twice incubated for 30 to 45 min, with a cocktail of flow cytometry antibodies (including 9G4) in PBS. After two washes, cells were fixed for 15 min in PBS/5% formalin (2% formaldehyde; Sigma) and resuspended in FACS buffer, for acquisition on the Symphony (BD) flow cytometer. Single-Cell RNA Sequencing of Peritoneal Cavity Residing B2 Cells. A total of 27,000 and 28,000 peritoneal cavity B2 cells were bulk sorted into 1.5 mL tubes containing sterile PBS/10% FCS from n= 2 sex- and age-matched pre589-tg mice. A total of 10,000 cells were captured in the 5′ V(D)J + GEX 10× reaction using the Chromium Single-Cell 5′ Gel Beads and v2 Library kits (10X Genomics). Quality control of libraries was assessed with an Agilent BioAnalyzer High Sensitivity DNA chip, pooled, and quantified using qPCR (KAPA Library Quantification Kit). Denatured libraries were loaded onto an Illumina NovaSeq 6000 and sequenced using a 150-cycle High-Output Kit. Raw base call (BCL) files generated during sequencing were demultiplexed into sample-specific FASTQ files and processed using Cell Ranger vdj pipeline (version 7.1.0). Sequences were aligned to the knocked in, germline reverted human pre589 IGH and IGK variable (V/D/J) genes alongside mouse Igh and Igk constant region genes using IgBLAST (version 1.19.0). Gene expression count matrices were exported by cell barcode using R. Single cell gene expression counts were normalized using Sanity (73) with default values. Single cells were excluded if mitochondrial reads accounted for more than 20% of the reads. Differentially expressed genes between Pre589+ and Pre589− B2 cells were determined using limma (74) on log transcription quotients from Sanity. After Bonferroni correction, genes with a FWER < 0.05 were called as significantly differentially expressed. GSEA, (75) was performed on Hallmark (mh) and Cell Signature (m8) ontologies, and terms with FWER < 0.05 were defined as significantly enriched. Vaccinia Virus Production and Infection. Vaccinia virus WR (equivalent to ATCC #VR1354) was a kind gift of Bernard Moss, NIH, Bethesda, MD. Virus stocks were grown on 143B osteosarcoma, or BHK-21 cells and purified by centrifugation through a 36% sucrose cushion before titration on BSC-1 cells using standard methods. Vaccinia virus was used to infect mice and for use as foreign antigen in ELISAs. Adoptive Transfer of pre589-tg CD45.2+ Cells into Vaccinia Virus Immunized WT CD45.1+ Mice and Immunization with Vaccinia Virus. 5 × 105 splenocytes from pre589-tg Rag1−/− (CD45.2+) mice were resuspended in 100 µL of sterile PBS and intravenously injected into CD45.1+ recipient mice which were then infected with 1 × 106 PFU vaccinia virus by intraperitoneal injection 24 h later. Recipient mice were killed at three timepoints: 10, 16, and 24 d post vaccinia virus immunization and spleens were harvested. Live CD45.2+ germinal center B cells (7AAD− CD19+ CD45.1− CD45.2+ FAS+ CD38low) were single-cell index sorted into sterile 96-well plates using BD FACS Aria II and III sorters. Immediately after sorting, the 96-well plates were sealed with DNAse/RNAse-free adhesive PCR plate seals (Sarstedt), centrifuged at 1,500 rpm for 1 min at 4 °C, then snap frozen on dry ice and stored at −80 °C. Single Cell Sequencing of CD45.2+ Germinal Center B Cell Receptors. Single cell B cell receptor (BCR) sequencing was performed by modifying an existing protocol originally developed to sequence the murine heavy and light chain immunoglobulin genes of C57BL/6 mice (76). Since the pre589-tg mice express human variable heavy and light chain sequences with murine constant regions, a mix of published primers (76, 77) and primers designed using SnapGene software (Version 5.3.1) were used in PCRs. Forward primer sequences bound in leader regions and reverse primers in constant regions to ensure mutated sequences were detected. All primers were synthesized by Integrated DNA Technologies (IDT). Reverse transcription PCR and targeted amplification of pre589 IgM and IgG heavy and light chain sequences was carried out using the Qiagen OneStep kit, according to the manufacturer’s recommendations. A nested PCR was performed using 2.5µL of the cDNA primary PCR product, separating the heavy and light chain nested PCR in separate 96-well plates using NEB Taq polymerase. Wells containing PCR products, determined by agarose (1%) gel electrophoresis, were Sanger sequenced by Genewiz (South Plainfield, New Jersey). MixCR (version 3.0.9) was used to assemble the sequences. Recombinant Antibodies. Pre589 antibodies expressing each of the five recurrent mutations in the heavy chain (S59R, S64R, Y110F, S111.1G, Y112.2F) as single point mutations, pairs, trios, or quartets were synthesized as human IgG1 by Genscript. All antibodies produced were paired with the germline pre589 IgK light chain, including a 589 heavy chain paired with a pre589 IgK (589 h pre589k). Monoclonal IgG1 antibodies that were >0.01 mg/mL and >80% purity, as assessed by Nanodrop and SDS-PAGE, were used in downstream binding assays. Recombinant Antibody Binding to Mouse Lymphocytes. Spleen single-cell suspensions from C57BL/6 mice were incubated with individual recombinant human IGHV4-34 IgG autoantibody variants (n= 25) at 15 µg/mL for 30 min on ice. After two washes with FACS buffer, cells were incubated for 45 min with the following anti-mouse antibodies in PBS (1X): FAS FITC, CD93 PE, CD21 BV421, CD38 BV510, CD19 BV605, CD23 BV711, IgD BUV395, B220 BUV737, and anti-human IgG PECY7. 7AAD PERCPCY5.5 (Biolegend) was used to exclude dead cells. Approximately 5 × 105 events were acquired per sample on the BD FACSymphony. Recombinant Antibody Binding to Vaccinia Virus. Recombinant human IGHV4-34 antibody binding to vaccinia virus was evaluated by ELISA as described (40) with the following modifications. Vaccinia virus was inactivated with 4′ aminomethyltrioxsalen hydrochloride and long wave ultraviolet (UV-C) light and the equivalent of 1 × 107 plaque forming units per mL were used to coat the wells of a 384-well ELISA plate. Binding of IGHV4-34 recombinant antibodies to vaccinia virus was quantified using a ClarioSTAR plus microplate reader, by measuring optical density at 405 nm. Statistical Analysis. To assess statistical significance between two groups with normally distributed data, a Student’s t test was used and the Mann–Whitney U test was applied for two groups with nonnormally distributed data. A one-way ANOVA was used when a single independent variable was compared across more than two groups with normally distributed data. A two-way ANOVA was applied for comparisons involving two independent variables with normally distributed data. For nonnormally distributed data involving more than two groups, the Kruskal–Wallis test was performed. Significance is represented accordingly: P ≤ 0.05 (*), P ≤ 0.01 (**), P ≤ 0.001 (***), P ≤ 0.0001 (****). Nonsignificant results were labeled as “ns” when P ≥ 0.5. Graphpad Prism software (V10.0) was used for statistical data analysis and figure generation. Supplementary Material Appendix 01 (PDF) Dataset S01 (XLSX) Dataset S02 (XLSX) Dataset S03 (XLSX)