It has long been recognized that antibodies induced experimentally to Tg interact with multiple, widely diverse epitopes on the large (600 kDa) dimeric Tg molecule, whereas human Tg autoantibodies interact with a restricted number of epitopes (Figure 1) (20–23). Similar observations have been made for TPO (24, 25). Panels of human monoclonal TPO and Tg autoantibodies isolated from combinatorial immunoglobulin libraries and expressed as recombinant Fab confirmed restricted epitope recognition [for example (26, 27)]. Recombinant Fab provided the tools to characterize the IDRs on TPO and Tg recognized by antibodies in patients and some euthyroid controls with subclinical disease (Table 1). Characterization of the epitopes recognized by TPO and Tg autoantibodies has also been performed by competition assays using mouse monoclonal antibodies generated (for example) to Tg polypeptides (28), to purified TPO (25), or to TPO peptides (25, 29).

It should be emphasized that binding by the majority of patients’ autoantibodies to TPO is decreased following protein denaturation (47, 48), demonstrating that most TPO autoantibodies interact with epitopes on conformationally intact protein. However, some studies have shown interaction between serum autoantibodies and reduced and denatured TPO [for example, Ref. (49, 50)], suggesting that not all TPO epitopes are conformational. Further, several linear epitopes (51, 52) and polypeptide fragments (53–56) are recognized by some patients’ TPO autoantibodies. Similarly, Tg autoantibodies predominantly recognize native protein [for example, Ref. (57)]. However, recognition of peptide fragments by some Tg autoantibodies has been reported [for example, Ref. (38, 58)].

In this review, because of their dominance in the patient repertoire, we will focus on autoantibodies to TPO and Tg that interact with conformational epitopes.

TPOAb epitopic “fingerprints” of the IDR are stable over time, that is without intramolecular epitope spreading, for example, in families in which the proband has juvenile Hashimoto’s thyroiditis (30), women who develop postpartum thyroiditis (35), and for TgAb in most patients on iodine supplementation (39) (Table 1). Similarly, after iodine-131 treatment in Graves’ disease, there was no evidence of TgAb intramolecular antibody spreading (44). There is some evidence that TPOAb epitopic patterns are inherited in families (36, 37).

Changes are rare in the epitopes recognized by TSHR antibodies, namely thyroid-stimulating antibodies (TSAbs), which are responsible for hyperthyroidism in Graves’ disease, and TSH-blocking antibodies (TBAbs), which cause hypothyroidism. In rare patients, TSHRAb switching from TSAb to TBAb (or vice versa) has been observed [reviewed in Ref. (59)]. The derivation of monoclonal TSAb and TBAb from one blood sample of an unusual patient who alternated between hyperthyroidism and hypothyroidism (60) demonstrated that the contrasting serum biological activities were due to two distinct antibodies.

It is of interest that injecting mice with TPO expressed together with MHC class II on a fibroblast line induced TPOAb that resembled autoantibodies from Hashimoto or Graves’ patients in terms of a high affinity for TPO and recognition of an IDR (46). In the same mouse strain, conventional immunization with TPO protein and adjuvant induced antibodies with lower affinity that recognized diverse epitopes on TPO (Table 1).

Autoantibodies interact with epitopes in an IDR in other autoantibody-mediated diseases [reviewed in Ref. (61)]. For example, in myasthenia gravis, autoantibodies to the acetylcholine receptor are restricted to a major IDR on the α1 subunit (62). Similarly, in the skin blistering diseases such as pemphigus vulgaris and pemphigus foliaceous, autoantibodies interact with an IDR on desmoglein (63).

Overall, the stability of human thyroid autoantibody recognition of an IDR is suggestive of “original antigenic sin.” It is possible that this concept may also apply to other human autoantibodies directed to an IDR on their respective autoantigens.

The mechanisms responsible for original antigenic sin are not fully understood. However, because of the problems caused for vaccination against novel viral strains, approaches have been used to overcome original antigenic sin. These studies provide insight into the basis for this phenomenon, at least from the perspective of T cell epitopes.

One approach involves eliciting cross-reactive responses by immunization with multiple peptide variants (68) or injecting yeasts carrying diverse virus-like particles (69). It has also been suggested that, at least for CD8+ T cell responses, original antigen sin can be overcome by treatment with neutralizing interleukin 10 together with linked T cell epitopes (70). A different strategy involves immunization with adjuvants (such as Bordetella pertussis toxin) that activate dendritic cells (71). The latter approach is of interest in the context of comments by Kim et al. (4) concerning viral antibodies, namely that antigen presentation by B cells favors the activation of memory B cells specific for the first virus rather than naive B cells specific for the subsequent cross-reacting virus. Finally, a simple mechanistic explanation proposes that original antigenic sin occurs because regulatory T cells induced by the first antigen decrease the amount of the second antigen on dendritic cells that activate naive B cells (72).

The most likely explanation for original antigenic sin in thyroid autoantibodies appears to involve antigen presentation by B cells or (in the case of TPO) thyroid cells, rather than dendritic cells. It should be noted that a restricted antibody response focused on the TPO IDR (as in humans) occurs in AKR/J-mice injected with cells expressing human TPO but not in mice of the same strain immunized with human TPO protein and adjuvant (46). The restricted response is likely related to the much lower concentration of cell-associated TPO than TPO injected with adjuvant. Because of their efficiency as specific antigen-presenting cells [for example, Ref. (13)], B cells are likely the antigen-presenting cells in spontaneously arising autoantibodies to Tg and the TSHR A-subunit. Via their specific immunoglobulin receptors, B cells may also capture conformationally intact thyroid autoantigens, a critical factor for the induction of pathogenic TSHR antibodies [reviewed in Ref. (73)].

In addition to suggesting the likely cells involved in autoantigen presentation, IDR recognition simplifies what would otherwise be a gargantuan task, namely exploring the regions recognized by autoantibodies on large, glycosylated thyroid autoantigens: (i)

Tg, the largest thyroid autoantigen comprising a homodimer of 300 kDa molecules, poses a major challenge. Iodination of Tg alters recognition by human autoantibodies (74), possibly by denaturing the antigen. However, epitope mapping using Tg fragments (fusion proteins or digestion products) suggests that some Tg autoantibody epitopes are located in the central region (75) or at the C-terminal end of the molecule (76, 77).

Induced and spontaneous intermolecular spreading for thyroid autoantibodies has been demonstrated for thyroid autoimmunity (Table 2). For example, rabbits immunized with human Tg and complete Freund’s adjuvant develop antibodies to Tg, as expected, as well as antibodies to TPO and antibodies that bind to both Tg and TPO (85, 86). It should be noted that searching for previously postulated bispecific human autoantibodies that recognize both Tg and TPO (“TgPO antibodies”) [for example, Ref. (87)] in a phage display immunoglobulin gene combinatorial, constructed from thyroid-infiltrating B cells of a patient with library serum TgPO-like autoantibody activity, led to multiple antibodies specific for either Tg or TPO but none had TgPOAb activity (88).

Returning to intermolecular epitope spreading, transgenic HLA-DR3 mice immunized with hTSHR DNA develop TSHR antibodies and, in a few mice, mild thyroiditis in association with antibodies to Tg (89). Transgenic BALB/c mice expressing low intrathyroidal levels of the human TSHR A-subunit, depleted of regulatory T cells (CD25 positive) before immunization with human TSHR-A-subunit adenovirus, developed TSHR antibodies and (unexpectedly) TgAb- and TPOAb-associated with massive thyroiditis and hypothyroidism (90). Incidentally, anti-CD25-treated BALB/c mice transplanted with TUBO tumor cells developed antitumor responses together with antithyroid immunity (98).

Of particular interest are observations of spontaneous antibody epitope spreading. NOD.H2^h4 mice develop thyroid autoimmunity, a process that is enhanced by iodide in the drinking water. The first autoantibodies to appear are directed against Tg (99–101), and subsequently, TPO antibodies appear (91). A similar pattern was observed in siblings of probands with juvenile Hashimoto thyroiditis (91). Moreover, TSHR antibodies develop spontaneously in NOD.H2^h4 mice transgenically expressing the human TSHR A-subunit (65).

Combining published data from published studies (45, 102) studies, the development of autoantibodies in female hTSHR/NOD.H2^h4 mice permits comparing the appearance of thyroid autoantibodies over time: TgAbs are present in some 4-month-old mice, and TSHRAb and TPOAb are detectable after the age of 6 months, and all three thyroid autoantibodies were present in more mice after 10 months (Figure 3A). Turning to humans, in a study of prediagnostic markers in Graves’ patients, TPOAb and TgAb were detectable several years before TSHR antibodies (92). Importantly, although in humans the maximum percentage of positive TgAb and TPOAb in humans did not approach 100% as in NOD.H2^h4 mice, the time sequence of antibody reactivity to Tg, TPO, and the TSHR in transgenic NOD.H2^h4 mice (Figure 3A) resembles that in humans (Figure 3B).

There is less extensive evidence for intramolecular thyroid autoantibody epitope spreading (Table 2): Latrofa and colleagues demonstrated that iodine prophylaxis revealed recognition of a previously cryptic TgAb epitope, but this phenomenon may involve the iodination of Tg, thereby generating a neo-antigen (97). In addition, as already mentioned, there are rare examples of TSHR antibody switching from TBAb to TSAb associated with thyroxine therapy, or the reverse, namely TBAb to switching to TSAb [reviewed in Ref. (59)]. As would be expected for antibodies with differing functional effects, the epitopes on the TSHR ectodomain recognized by TSAb and TBAb are different (103, 104), although they interact with closely overlapping portions of the amino terminus of the TSHR A-subunit (105).

The apparent antibody epitope spreading in maternally transferred TSHR antibodies from initial TBAb to TSAb is due to progressive dilution of the high titer TBAb overpowering the lower concentrations of TSAb as human IgG is metabolized [reviewed in Ref. (59)].