Predominant Autoantibody Production by Early Human B Cell Precursors

See allHide authors and affiliations

Science  05 Sep 2003:
Vol. 301, Issue 5638, pp. 1374-1377
DOI: 10.1126/science.1086907


During B lymphocyte development, antibodies are assembled by random gene segment reassortment to produce a vast number of specificities. A potential disadvantage of this process is that some of the antibodies produced are self-reactive. We determined the prevalence of self-reactive antibody formation and its regulation in human B cells. A majority (55 to 75%) of all antibodies expressed by early immature B cells displayed self-reactivity, including polyreactive and anti-nuclear specificities. Most of these autoantibodies were removed from the population at two discrete checkpoints during B cell development. Inefficient checkpoint regulation would lead to substantial increases in circulating autoantibodies.

Self-reactivity has concerned immunologists since the time of Ehrlich, who referred to this potential problem as leading to “horror auto-toxicus” (1). Landsteiner's finding that the immune repertoire is highly diverse focused the problem on how such diversity could be generated while avoiding self-reactivity (2). We now know that antibody diversity is produced by V(D)J recombination, and experiments using transgenic mice have shown that at least three mechanisms account for silencing of self-reactive antibodies during B cell development: receptor editing, deletion, and anergy (36). However, the actual number of self-reactive antibodies that arise during B cell development has not been accurately determined, nor is it known precisely when such antibodies are removed from the repertoire under physiologic circumstances.

To examine the development and silencing of autoreactive B cells, we cloned antibodies from single B cells derived from the bone marrow and blood of three healthy human donors and tested them for reactivity against cell lysates and a panel of defined antigens (fig. S1 and tables S1 to S10) (7, 8). Precursor B cells with the surface phenotype of pre-B cells that expressed functional Igκ or Igλ chain transcripts were designated as early immature B cells. These were distinguished from pre-B cells that did not express functional light chains and from immature B cells that expressed cell surface immunoglobulin M (IgM) (fig. S1 and tables S1 to S6).

Although amino acid sequence alone cannot predict whether an antibody will be self-reactive, long Ig heavy chain complementarity-determining regions 3 (IgH CDR3) have been associated with self-reactive or polyreactive antibodies (912). Analysis of human antibodies cloned from pools of developing B cells showed that IgH CDR3s from progenitor B cells are significantly longer on average than those from peripheral B cells (13, 14). A second feature associated with self-reactivity is the presence of positively charged amino acids within IgH CDR3 (15, 16). We found that antibodies with long and/or highly positively charged IgH CDR3s were enriched in pre-B cells and early immature B cells (Fig. 1). These features were selectively lost from the repertoire as B cells progressed through development (Fig. 1). We found no significant differences in Ig light chain (IgL) CDR3 length or charge, nor in IgH or IgL V/J usage, between B cell fractions (fig. S1). Thus, IgH CDR3 features associated with self-reactive antibodies appear to be removed from the B cell repertoire as B cells progress through development.

Fig. 1.

Counterselection of long and positively charged IgH CDR3 regions during human B cell development. (A) Frequencies of IgH CDR3s with ≤9, 10 to 14, 15 to 19, and ≥20 amino acids. The average IgH CDR3 length in pre-B cells was 15.6 amino acids, which was not significantly different from 16.1 amino acids in early immature B cells (P = 0.453). In mature naïve B cells, the average length of the IgH CDR3 was significantly decreased to 13.5 amino acids (pre-B + early immature versus naïve B, P = 0.0001). (B) Pie charts show proportion of IgH CDR3s with ≥3, 2, 1, or 0 positive charges (pre-B + early immature versus naïve B, P = 0.017; pre-B versus early immature B, P = 0.24). Absolute numbers of sequences analyzed in each B cell compartment for both length and charge are given below the pie charts.

To determine whether individual antibodies cloned from different B cell subsets were self-reactive, we expressed them in 293A cells (7). We initially tested 248 antibodies for binding to nuclear and cytoplasmic antigens (HEp-2 cell line extract) used in standard clinical assays for anti-nuclear antibodies (ANAs). We found that 75.9% of all antibodies cloned from early immature B cells showed high levels of reactivity in HEp-2 enzyme-linked immunosorbent assays (ELISAs) (7) (Fig. 2, A and B). The frequency of reactive clones decreased to 43.1% in the immature stage but remained at 40.7% in newly emigrated B cells in the blood; an additional drop in reactivity in HEp-2 ELISAs was seen between the new emigrant and naïve B cell compartments (Fig. 2, A and B). The relatively high percentage of HEp-2 lysate–reactive antibodies found in naïve B cells is consistent with findings from other studies that used supernatants from Epstein-Barr virus–transformed peripheral blood B cells (17, 18). From these observations, we conclude that during normal human B cell development, large numbers of self-reactive antibodies are removed from the repertoire during the immature B cell stage in the bone marrow, as well as during the transition from the new emigrant to the mature naïve B cell stage in the periphery.

Fig. 2.

HEp-2 ELISA and immunofluorescence results for three healthy donors: (A) First, (B) second and third donor. Reactive antibodies are indicated by clone number followed by IgL designation. Dashed lines show ED38 positive control (7). Horizontal line shows cutoff OD405 (optical density at 405 nm) for positive reactivity (7); n = number of antibodies shown in each graph. Percentages represent frequency of self-reactive antibodies in each B cell fraction averaged across the three donors (7). (C) Typical nuclear, subnuclear plus cytoplasmic, and cytoplasmic staining patterns found in IFAs. (D) Frequency of self-reactive antibodies in each B cell compartment with nuclear, nuclear plus cytoplasmic, and cytoplasmic staining patterns, and frequency of nonreactive antibodies. Data fromthree donors were pooled.

Antibodies that are reactive in HEp-2 ELISAs include true ANAs that may be pathogenic and are associated with autoimmune diseases such as systemic lupus erythematosus (19). Other antibodies are directed against cytosolic antigens and show no clear disease association (19). To distinguish between ANAs and antibodies that bind cytosolic components, we performed indirect immunofluorescence assays (IFAs) on fixed HEp-2 cells (Fig. 2, C and D). We found that 58.6% of the antibodies expressed by early immature B cells showed either a nuclear or nuclear/cytoplasmic staining pattern, and 13.8% stained the cytoplasm (Fig. 2D). Antibodies with nuclear or nuclear/cytoplasmic ANA staining patterns were removed from the repertoire at the same developmental checkpoints as HEp-2 ELISA reactive antibodies, whereas antibodies reactive with cytoplasmic components were not removed from the B cell repertoire efficiently (Fig. 2, A, B, and D).

As an additional measure of self-reactivity, we tested antibodies for binding to a defined set of antigens including single-stranded and double-stranded DNA (ssDNA and dsDNA), insulin, and lipopolysaccharide (LPS) by ELISA (Fig. 3). Antibodies that bind to more than one of these antigens are considered polyreactive. We found that 55.2% of the antibodies cloned from early immature B cells were polyreactive; in contrast, small numbers of antibodies cloned from immature, new emigrant, and mature naïve B cells displayed polyreactivity (Fig. 3). These results show that more than half of the antibodies expressed in early B cell compartments are polyreactive and suggest that nearly 90% of the polyreactive antibodies are counterselected in the immature B cell stage.

Fig. 3.

Polyreactive antibodies in B cell development. Data shown are from ELISAs for reactivity with ssDNA, dsDNA, insulin, and LPS. Percentages represent frequency of polyreactive antibodies. Dashed lines show ED38 positive control (7). Reactive antibodies are indicated by clone number followed by IgL designation; n = number of antibodies tested for each fraction. Note that the polyreactive antibodies fromnaïve B cells showed lower levels of reactivity on average than those from early immature B cells.

To determine whether self-reactivity is correlated with specific Ig sequence features, we compared the CDR3s from HEp-2 ELISA reactive and nonreactive antibodies (Fig. 4) (figs. S1 and S2). Only two features were significantly enriched in the self-reactive antibodies: long IgH CDR3s (Fig. 4A) and positively charged amino acids (Fig. 4B).

Fig. 4.

Self-reactive B cells display long and positively charged IgH CDR3s. Data shown are for IgH CDR3 sequences from HEp-2 ELISA reactive (n = 65) and nonreactive (n = 77) antibodies from pooled immature and new emigrant B cells. (A) Proportion of sequences with IgH CDR3 lengths of ≤9, 10 to 14, 15 to 19, and ≥20 amino acids. IgH CDR3 regions of self-reactive antibodies are significantly longer than those of non–self-reactive antibodies (P = 0.03). (B) Frequency of IgH CDR3 regions with the indicated number of positive charges. The frequency of positively charged IgH CDR3 regions is significantly increased in self-reactive versus non–self-reactive antibodies (P = 0.02).

Antibody selection is initiated upon light chain gene expression in early immature B cells (20). These cells continue to express the recombinase activating genes RAG1 and RAG2 (2123) and can replace self-reactive or otherwise defective antibodies by continued recombination or “receptor editing” (3, 4, 24, 25). In the mouse, 25% of the antibody repertoire and 50% of all Igλ antibodies appear to be products of receptor editing, but the amount of editing that is directly due to self-reactivity is not known (26, 27). As an initial assessment of whether loss of self-reactivity was associated with receptor editing in humans, we compared downstream Jκ usage (increased downstream Jκ usage is a signature of receptor editing) between early immature B cells and naïve B cells (28). We found no significant differences in Jκ usage in the more mature B cells (fig. S1). However, this analysis does not allow reliable evaluation of the importance of editing or deletion in antibody selection in humans because it does not take into account editing by Igλ and the prominent role of Igλ in the human antibody repertoire.

Immature B cells are highly susceptible to deletion by receptor cross-linking (29). The number of human B cells deleted during development is not known, but in the mouse it has been estimated that 80 to 90% of all newly produced B cells are deleted before they enter the mature B cell compartment (30, 31). Our observation that a large number of self-reactive antibodies are lost during B cell development may in part account for this phenomenon. We have established that developing B cells in humans express large numbers of self-reactive antibodies, and that in the bone marrow the immature B cell stage is one of two important checkpoints for selection against autoantibodies.

The second checkpoint for selection against self-reactive antibodies in humans is at the transition between new emigrant and mature B cells in the periphery (30, 32). Nonetheless, 20% of all antibodies produced by mature human B cells show HEp-2 reactivity, and 4.3% are polyreactive. The HEp-2 ELISA reactive antibodies typically recognize cytosolic components that may not be encountered by developing B cells under physiologic circumstances. Polyreactivity, including self-reactivity, is associated with protective natural antibodies (3335). Therefore, a potential benefit of these B cells is that they may contribute to natural antibody production, but the persistence of autoantibody-producing B cells may also explain why autoantibodies are frequently detected in diseases that involve tissue destruction.

Ig variable, diversity, and joining gene fragments have been selected by evolution for 200 million years, and in that time highly self-reactive germline variable genes have been removed from the antibody repertoire. Nonetheless, we find that 55 to 75% of all antibodies emerging in the bone marrow are self-reactive. The primary cause of autoantibody formation appears to be random nucleotide addition and deletion in IgH CDR3s, which also yield much of the diversity in the heavy chain repertoire. The cost of this additional diversity appears to be that at least half of the initial antibody repertoire must be removed. Our results establish the initial extent of autoantibody production in healthy individuals and suggest that autoantibody regulation requires two distinct B cell developmental checkpoints. Little is understood about how tolerance is broken in autoimmune diseases in humans, but the finding that large numbers of autoantibodies are produced under physiologic circumstances suggests that even small changes in the efficiency of autoantibody regulation would lead to increased susceptibility to autoimmunity.

Supporting Online Material

Materials and Methods

Figs. S1 and S2

Tables S1 to S11


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article