Report

Evolution of Autoantibody Responses via Somatic Hypermutation Outside of Germinal Centers

See allHide authors and affiliations

Science  20 Sep 2002:
Vol. 297, Issue 5589, pp. 2066-2070
DOI: 10.1126/science.1073924

Abstract

Somatically mutated high-affinity autoantibodies are a hallmark of some autoimmune diseases, including systemic lupus erythematosus. It has long been presumed that germinal centers (GCs) are critical in autoantibody production, because they are the only sites currently believed to sustain a high rate of somatic hypermutation. Contrary to this idea, we found that splenic autoreactive B cells in autoimmune MRL.Faslpr mice proliferated and underwent active somatic hypermutation at the T zone–red pulp border rather than in GCs. Our results implicate this region as an important site for hypermutation and the loss of B cell self-tolerance.

The activation of autoreactive B cells is pivotal for the development of systemic autoimmune diseases, because these lymphocytes secrete pathogenic autoantibodies and promote the activation of pathogenic autoreactive T cells (1). Because affinity-enhancing somatic mutations are prevalent in these autoantibodies, it has long been hypothesized that the GC provides the critical signals for B cell activation in auto-immune disease. This theory, however, has never been proven, and the steps leading to the production of autoantibodies in systemic autoimmune diseases remain elusive.

The rheumatoid factor (RF) autoantibody response was studied by crossing onto the MRL.Faslpr (MRL/lpr) lupus-prone mouse strain a transgene (Tg) encoding an immunoglobulin (Ig) Vh (H chain) that is derived from an RF monoclonal autoantibody (AM14) specific for IgG2a of the “a” allotype (IgG2aa) (2, 3). The resulting mice have a diverse endogenous light chain repertoire along with an increased precursor frequency of an IgG2aa-specific autoantibody that is recreated by the pairing of the AM14 H chain with a rearranged Vκ8 gene product. B cells expressing this autoantibody can be traced using an anti-idiotype (Id) monoclonal antibody (4-44) raised against the original AM14 hybridoma.

We recently showed that in aged, AM14 H Tg, Fas-deficient (lpr), autoimmune-prone mice, numerous RF antibody–secreting cells can be observed (4). Because RF- and other autoantibody–producing B cells in B6/lpr and MRL/lpr mice are somatically mutated and isotype-switched (5,6), it was assumed that the autoimmune response of AM14 Id+ MRL/lpr B cells was generated in GCs. To test this, we stained sections of spleens from a large number of such mice, at a variety of ages, with markers for RF B cells, GCs, and T cells (Fig. 1) (7). Unexpectedly, we observed few RF Id+ GCs in these spleens. Instead, RF/Id+ B cells were predominantly observed at the T zone–red pulp border (Fig. 1) in areas lacking a follicular dendritic cell (FDC) network (Fig. 1, D and E). Id+ cells in these areas were actively proliferating, as indicated by the rapid labeling of a substantial proportion of Id+ cells after bromodeoxyuridine (BrdU) injection (Fig. 1J). As measured by fluorescence-activated cell sorting (FACS) analysis, 15% of splenic RF Id+ B cells were labeled within 2 hours of a single injection of BrdU, indicating a high proliferative rate (8). RF Id+ cells appeared in clusters ranging from tens to thousands of cells, interdigitated with T cells (Fig. 1, A, B, and K), including large numbers of dark-staining antibody-forming cells (Fig. 1) (9). Furthermore, this process is autoantigen-specific, in that it was never observed in IgHb congenic mice that lacked the IgG2aa autoantigen (Fig. 1F), nor was it observed in non–autoimmune-prone mice (4, 10).

Figure 1

RF B cells are at the border of the splenic T zone and the red pulp, not in GCs. [(A) to (E)] Immunohistochemical analysis of adjacent spleen sections from an AM14 MRL/lpr mouse. (A and B) Anti-Id staining (red) identifies cells expressing the AM14 H chain and certain Vκ8-family L chains, which together confer RF specificity. CD3 (blue) identifies T cell zones. (C) GCs, identified by PNA (red), are present (arrow) but do not colocalize with Id+ areas (blue). (D andE) Id+ cells are outside of follicles, identified by monoclonal antibody to complement receptor 1 (CR1), a marker for FDCs. (F) Id+ and CD3 staining of the spleen of an AM14 MRL/lpr IgHb mouse. Id+ cells do not accumulate in these mice because of the absence of autoantigen. (G) Id+ (blue) GCs (red) are easily detectable after immunization of an AM14 H chain MRL/lpr IgHb mouse with IgG2a-containing immune complexes. (H) Identically stained section showing an Id GC from an alum-immunized MRL/lpr IgHb mouse. (I) An Id GC arose spontaneously in an H Tg MRL/lpr mouse, similar to the GC in (B). As expected, microdissection of approximately 80% of this GC followed by PCR amplification yielded no product. (J) Id+ cells in MRL/lpr mice proliferate at the T zone–red pulp border. Red, BrdU; green, Id; blue, 4',6'-diamidino-2-phenylindole (DAPI), which stains the nuclei of all cells; and orange, Thy1.2, a T cell marker. Mice were pulsed with BrdU 2 hours before death (8). Proliferating cells will incorporate BrdU in their nuclei. Arrows show some of the BrdU+/Id+ cells (pink nuclei and green cytoplasm). (K) Spleen section from another AM14 MRL/lpr mouse, demonstrating the close interaction of Id+ cells (green) with CD11c+ DCs (red) and Thy1.2+ T cells (blue).

Cellular interactions that drive autoreactive B cell proliferation outside of GCs would be substantially different from those within a GC. For example, FDCs, which are thought to regulate B cell proliferation in GCs, are completely absent in the RF+ clusters (Fig. 1, D and E). However, CD11c+ dendritic cells (DCs) are known to be prevalent at sites similar to those where AM14 B cells proliferate (although rare in GCs) (11). Indeed, staining for anti-Id and CD11c, a DC marker, revealed many DC-RF B cell interactions in these clusters (Fig. 1K). It is believed that DCs provide unique survival and differentiation signals to B cells and plasmablasts (11).

RF B cell proliferation almost exclusively outside of GCs in the spleen was unexpected, given that autoantibody-secreting B cells from MRL/lpr and other autoimmune-prone mice and humans are heavily somatically mutated. To resolve this issue, we hypothesized that a high rate of hypermutation was taking place at these extrafollicular sites of RF B cell proliferation. To test this hypothesis, Id+ cells at these sites were microdissected and analyzed (Fig. 2, A and B) by polymerase chain reaction (PCR) amplification using primers specific for genes encoding Vκ8 (12). We created 45 libraries of bacterial clones from independent PCR products. We then sequenced the Vκ gene from multiple colonies derived from each library (Table 1and fig. S1). Nearly all sequences were Vκ8-19 or Vκ8-28 genes joined to Jκ4 and occasionally Jκ5; these two V gene segments are highly homologous to the original Vκ8 gene expressed by the AM14 hybridoma. In most cases, the V genes contained a substantial number of somatic mutations, established by comparison to known germline V and J gene segments (13). Consensus sequences and extensive sequencing of germline DNA (fig. S1) confirmed this conclusion and failed to identify additional germline genes.

Figure 2

Genealogical trees generated from sequences of picks from mice 2540 (A and C) and 7983 (B and D), demonstrating ongoing somatic V region diversification. [(A) and (B)] Blue, CD3; red, Id. Areas outlined in black were microdissected, PCR-amplified, and sequenced. Picks 11g1 and 11g2 had approximately 30 and 20 Id+ cells, respectively, and picks 17a3, 17a4, and 17a5 had approximately 5, 10, and 50 Id+ cells. [(C) and (D)] Individual clones were assigned to a particular genealogical tree on the basis of shared and unique mutations. Each open circle represents one or more clones with the same sequence. The numbers of mutated codons are written at the sides of branches. Regions affected are in parentheses. Amino acid changes are written below the codon. Silent mutations are in italics. Mutations that have been seen repeatedly in independent clones, considered hotspots, are indicated by asterisks. Inferred intermediates in clonal evolution are represented by dark circles. Picks 11f3 and 11f4 (C) were taken from the same cluster as 11g1 and 11g2, in an adjacent section. All clones from this same region have a unique germline sequence resembling a combination of the Vκ8-19 and Vκ8-28 gene loci (see fig. S1). Circles at the base of trees represent germline sequences. Sequences for mouse 7983 were isolated from picks taken from three different foci on the same T cell zone, yet each focus has a unique set of mutations. A mutation in tree G produces a stop codon.

Table 1

Summary of all microdissections and sequences along with the distribution of mutations in those sequences. A PCR product was retrieved in ∼75% of microdissections. Sequences recovered from eight mice were grouped by pick and assigned to a lettered genealogical tree. Mutations were classified into three categories based on their positions in these trees. Identical sequences in a pick were conservatively treated as one clone, instead of as the products of multiple cells, and were assigned to one unique sequence. Trunk mutations from picks that have no branches in the genealogy are in bold (for example, Fig. 2D, pick 17a3).

View this table:

In many cases, independent sequences from the same microdissected area differed by one or more mutations, yet shared several other mutations. This hierarchy of shared and unique mutations could be described by genealogical trees, a hallmark of ongoing somatic hypermutation (Fig. 2, C and D). We estimated the light chain mutation rate by mathematical modeling at 0.3 mutations per gene per generation, a rate comparable to that estimated by the same method for hypermutation in GCs (14). Moreover, the pattern of V gene mutations shows clear signs of selection by antigen (table S2).

The unique mutations observed were not the result of mutation elsewhere followed by migration, as sequences from the same or sometimes nearby picks share common mutations and VJ junctions whereas sequences derived from distant sites share a different set of mutations (Table 1 and Fig. 2). Had these unique mutations occurred at distant sites, clonal relatedness would not be confined to local areas. It is still possible that the shared mutations (at the base of the trees) occurred during an initial passage through a GC before further mutation and expansion at the T zone–red pulp border. However, we do not think this is a requirement for mutation outside of GCs, because we have isolated unmutated sequences in the same small picks as mutated sequences (for example, Fig. 2D, tree G).

To confirm the absence of mutating Vκ8 cells in Id GCs, sections spanning the entire spleen of an H Tg MRL/lpr mouse were stained with peanut agglutinin (PNA) and anti-Id. Of the seven GCs found, all contained few if any visible Id+ cells (Fig. 1I). The distinction between Id and Id+ GCs is clearly visible histologically (Fig. 1, C and G). Vκ8 PCR amplification of microdissected areas of these GCs yielded a PCR product in only 1 out of 7 GCs, whereas 10 of 10 control Id+ GCs were PCR-positive (table S1). The Vκ8 sequences from the lone positive GC of mouse 5281 were unmutated and used a different Jκ segment from the mutated sequences recovered from Id+ clusters in nearby and distant T zone–red pulp areas of the same spleen, further arguing against mutation in GCs followed by migration to an Id+ site. In some spleens that have ongoing mutation, no GCs of any type are seen (Table 1 and fig. S2). Somatic hypermutation outside of GCs in RF Tg Fas-deficient mice differs fundamentally from mutation that has been documented in severely immunodeficient mice defective in GC formation. For example, in CD40L- and LTα-deficient mice, the mutation process occurs at unidentified sites (perhaps residual GCs), and mutations themselves are infrequent (15, 16). However, AM14 Tg B cells undergo a high rate of somatic hypermutation at a clearly defined histologic site. Furthermore, Fas-deficient mice are immunocompetent and do have normal GC formation after immunization (17, 18), although this can be reduced in older mice.

The facts that Fas-deficient mice can form GCs that support normal somatic hypermutation (17, 18) and that RF B cells are clearly capable of entering GCs (Fig. 1, G and H) (9) raise the question of why RF B cell mutation in the spleen occurs in extrafollicular sites. Most likely, there is an unusually long duration of RF B cell proliferation at the T zone–red pulp border in AM14 Tg MRL/lpr mice (9). We therefore propose that somatic hypermutation is induced whenever B cells are stimulated to undergo a substantial number of cell cycles under conditions where antigen- and T cell–derived signals required for mutation are present (19). This hypothesis would explain why mutation occurs in GCs, in vitro under appropriate conditions (19), and at extrafollicular sites in chronic autoimmune responses.

RF B cells may persist in MRL/lpr mice because of lack of Fas-mediated apoptosis. However, during foreign protein immune responses, apoptosis and involution of proliferative foci at the T zone–red pulp border occur normally in Fas-deficient mice (20). Therefore, we favor the idea that chronic exposure to antigen as observed in AM14 spleens (fig. S3) and/or the form of the antigen perpetuates the response at the T zone–red pulp border. Autoantibodies commonly target macromolecules with repeating determinants, such as immune complexes and chromatin, that could provide a sustained and strong antigen signal. Another unique facet of the RF response is the potential of chromatin-containing immune complexes to co-stimulate AM14 RF B cells via Toll-like receptor 9, as recently demonstrated in vitro by Leadbetter et al. (21). Such signals could lead to altered localization and/or differentiation of somatic hypermutating B cells. Leadbetter et al. have suggested that the ability to co-signal through Toll-like receptors may be a unifying feature of dominant autoantigens (21). If so, then mutation outside of GCs may be common in autoantibody responses.

Because somatic mutations can create autoreactive B cells from innocuous ones, mechanisms that censor autoreactive mutants are likely to be important in the GC (22–24). These protective mechanisms may be the reason why mutation is normally restricted to the GC. However, when the antigenic stimulus is chronic and the antigens involved may stimulate through unique pathways, B cells may mutate elsewhere (25) and thereby escape the mechanisms that normally censor autoreactive B cells in the GC environment.

Supporting Online Material

www.sciencemag.org/cgi/content/full/297/5589/2066/DC1

Materials and Methods

Figs. S1 to S3

Tables S1 and S2

Reference S1

  • * To whom correspondence should be addressed. E-mail: mark.shlomchik{at}yale.edu

REFERENCES AND NOTES

View Abstract

Navigate This Article