Ezrin, Maspin, Peroxiredoxin 2, and Heat Shock Protein 27: Potential Targets of a Streptococcal-Induced Autoimmune Response in Psoriasis

Psoriasis vulgaris is a common chronic inflammatory skin disease affecting 2–3% of the white population. It is characterized by scaly, erythematous plaques that result from an inflammatory hyperproliferation of the epidermis and may cover large areas of the body (1). Psoriasis represents a multifactorial disorder with a polygenic predisposition, which includes a strong association with HLA-Cw6 and polymorphisms in various genes relevant for inflammation and immune responses, such as IL-23R or the IL-12p40 subunit (2–4).

Selective immunosuppressive therapies targeting T cells or cytokines suggest that the hyperproliferative psoriatic inflammation may result from a T cell-mediated immune response (3, 5). Although the T cell-derived psoriatic cytokine network has become increasingly clear, the precise mechanisms leading to activation of T cells in the skin of psoriasis patients are less well understood. Extensive analysis of the lesional TCR usage revealed dominant T cell clones in the psoriatic T cell infiltrate that persisted within psoriatic skin lesions over years and reappeared in psoriasis relapses, while being absent from uninvolved skin (6–8). Together with a conserved amino acid motif in the third complementarity-determining region (CDR3) of clonally expanded lesional TCR β-chain rearrangements (9), these findings emphasize that the psoriatic T cell response is directed against dominant psoriatic autoantigens that are preserved within individual patients and may be public to psoriasis in general. Accordingly, identification of the psoriatic autoantigens represents an essential clue to the pathogenesis of psoriasis. It may define the targets maintaining disease activity as the result of persistent T cell activation and identify novel, causally determined therapeutic approaches.

Psoriasis onset or relapses are seen in strong association with a prior tonsil infection with group A β-hemolytic streptococci (GAS/Streptococcus pyogenes) (10–14), and HLA-Cw6⁺ patients are particularly sensitive to the streptococcal trigger (12, 13). The simultaneous presence of the same T cell clones in skin lesions and tonsils in patients with streptococcal-driven psoriasis support a pathogenetic link between streptococcal angina and the lesional psoriatic immune response (15).

These observations suggest that GAS infection of the tonsils could induce a T cell-mediated autoimmune response against skin proteins, which may preferentially be presented by HLA-Cw6 (16–18). After an initial priming phase, which might also involve costimulatory signals by streptococcal superantigens (19), this GAS-induced immune response could extend to the skin by molecular mimicry in a similar fashion as the cross-reactive streptococci-specific autoimmune responses in other poststreptococcal sequelae, such as acute rheumatic fever or poststreptococcal glomerulonephritis (20).

To identify potential psoriatic autoantigens resulting from an antistreptococcal immune response, we immunized rabbits with GAS. GAS immunization induced Abs against several clearly defined proteins expressed in keratinocytes. According to their particular humoral and cellular immunogenicity for psoriasis patients, some of these proteins could represent antigenic targets of a polyspecific (21) streptococcal-induced T cell-mediated autoimmune response and contribute to the formation of psoriatic skin lesions.

Rabbits of the strain Deutscher Riese were immunized with S. pyogenes serotype MT12, NCTC 10085 (provided by Dr. Efstratiou, Public Health Laboratory Service, London, U.K.) by six s.c. injections at 2-wk intervals with 100 μl partially sonicated GAS pellets in IFA. Serum samples were collected before the first immunization and 2 wk after the last immunization.

Preparation and culture conditions of human primary keratinocytes from surgical skin samples, the epidermoid carcinoma cell line A431 (American Type Culture Collection, Rockville, MD), or an EBV-transformed human B cell line have been described (22). To induce proteins dependent on stress or differentiation they were heat shocked by incubation at 43°C for 8 h or grown with high Ca²⁺ concentrations for 24 h (1.2 mM).

Western immunoblotting followed standard protocols (23, 24). For two-dimensional gel electrophoresis, human keratinocytes were lysed in 150 mM NaCl, 1% Nonidet P-40, 50 mM Tris (pH 8.0), centrifuged at 10,000 × g for 20 min, and the proteins were lyophilized. Each 50–100 μg lyophilized keratinocyte lysates were applied to Immobiline Dry Strips (pH 3–10 linear gradient, 180 mm length, Amersham Pharmacia Biotech, Uppsala, Sweden). Isoelectric focusing running conditions were 500 Vh at 500 V, then 3,000 Vh at 2,000 V, and 37,500 Vh at 2,500 V. Blots were blocked in PBS, 0.1% Tween 20 (pH 7.4), 4% BSA, 10% FCS (Sigma-Aldrich, St. Louis, MO) and incubated with human sera (diluted 1:500 in PBS, 0.1% Tween 20 [pH 7.4], 10% FCS, 4% BSA) or rabbit sera (1:1000) for 2 h. Ab reactivity was detected by HRP-conjugated protein A (1:1000) and visualized by diaminobenzidine (Sigma-Aldrich) or ECL reagent (Amersham Pharmacia Biotech).

Peptide sequencing using Edman degradation was done at Toplab (Martinsried, Germany) according to standard procedures. Database and homology searches were performed at the ExPASY Molecular Biology Server of the Swiss Institute of Bioinformatics (www.expasy.org) or the National Center for Biotechnology Information (www.ncbi.nlm.nih.gov).

Recombinant heat shock protein (hsp)27 protein and an epidermal cytokeratin preparation were obtained from Sigma-Aldrich or BIOTREND (Cologne, Germany). pGEX-2T-ezrin was a gift of Dr. Monique Arpin (Laboratoire de Morphogenèse et Signalisation Cellulaires, Institut Curie, Paris, France) and was expressed as GST-fusion protein in Escherichia coli strain BL21 (DE3) (Novagen, Madison, WI). Maspin, peroxiredoxin 2 (PRDX2), and keratin 6 were amplified by RT-PCR from human keratinocyte mRNA, cloned into the His-tag pET16b-vector (Novagen), and expressed as His-tag fusion proteins in E. coli strain XL2-Blue (Stratagene, La Jolla, CA). Synthetic peptides were obtained from Biomers.net (Ulm, Germany).

The local ethics committee approved the study. Patients and healthy controls participated voluntarily and gave written informed consent. Psoriasis patients were typed for HLA class I alleles and classified for type I or II psoriasis: type I psoriasis = early onset (<40 y of age) and/or positive family history and/or inheritance of HLA-Cw6, -B57, or -B13. Type II psoriasis = late onset (>40 y), negative family history for psoriasis, and lack of HLA-Cw6 (2).

Preparation and cultivation of PBMCs, T cell stimulation, and generation of Ag-specific T cell lines were done as described (25, 26). For T cell stimulation, 1 × 10⁶/ml PBLs were grown in the presence of protein Ags (5 μg/ml), peptides (10 μg/ml), or PHA (1:100) for 40 h (IFN-γ–release ELISPOT) or 5 d ([³H]thymidine incorporation). The IFN-γ–release ELISPOT assay (Mabtech, Nacka Strand, Sweden) was performed as described (27). Results were expressed as spot-forming cells (SFCs) per 10⁶ PBMCs. For [³H]thymidine incorporation, triplicate samples of 2 × 10⁵ cells were pulsed with 2 μCi = 74 kBq [³H]thymidine/well for 8 h (Amersham-Buchler, Braunschweig, Germany). Incorporated radioactivity was measured as cpm (28). Negative control values (cultivation of PBMCs in the absence of Ag or with preparations from mock-transfected E. coli) were always subtracted from the results obtained by specific Ag stimulation. For statistical analysis, mean cpm or numbers of SFCs were analyzed in f- and t tests. Probability of error (p value) was set to p < 0.05.

TCR β-chain variable (TCRBV)-repertoire analysis, cloning, and sequencing of TCRBV gene rearrangements were done exactly as described (15). TCR rearrangements are given as deduced amino acid sequence of the CDR3.

CD4/CD8 ratios of the T cell lines were determined as reported previously (15). T cell lines were separated into CD4⁺ and CD8⁺ T cells using magnetic beads (Microbeads, Skedsmokorset, Norway).

To identify common epitopes on proteins from GAS and keratinocytes, rabbits (n = 3) were repeatedly immunized with S. pyogenes, serotype M12, which is frequently observed in psoriasis patients (10). Analysis of the reactivity of the streptococci-specific hyperimmune sera by Western immunoblotting and comparison with the preimmune sera (Fig. 1A) demonstrated that streptococcal immunization had induced serum Abs against various keratinocyte proteins but not proteins from other cell types, such as a lymphoblastoid B cell line or A431 cells (Fig. 1B). This reactivity was largely independent from conditions affecting cell differentiation (high calcium concentrations) or heat stress.

When analyzed with keratinocyte protein lysates fractionated by two-dimensional SDS gel electrophoresis, the preimmune rabbit sera (Fig. 2A) stained only a few keratinocyte proteins with a major reactivity against a protein that, according to its position in the Coomassie-stained gel (Fig. 2E), represented actin. Instead, the streptococci-specific rabbit sera reacted with several keratinocyte proteins (Fig. 2B), as did the sera of the psoriasis patients (n = 7) (Fig. 2C). Seven keratinocyte proteins were stained by the streptococci-specific rabbit sera and all psoriasis patients’ sera but not by the preimmune rabbit or the healthy control sera (n = 8) (Fig. 2D). These proteins were considered potential targets of a streptococcal-induced autoimmune response in psoriasis and identified by amino acid sequencing and alignments with the Swissprot protein sequence library. They represented ezrin/cytovillin, maspin/serpin B5, PRDX2, hsp27/β-1 (hsp27), keratin 6, GAPDH, and inorganic pyrophosphatase (Table I).

To determine the potential role of the different proteins as autoantigens of the psoriatic T cell response, PRDX2, ezrin, and maspin were expressed as recombinant full-length proteins, and keratin 6 was expressed as two overlapping proteins corresponding to aa 10–210 and 190–559 of the keratin 6f isoform. Recombinant hsp27 and a cytokeratin preparation from keratinocytes were purchased. Because of their ubiquitous tissue distribution, GAPDH and inorganic pyrophosphatase were considered less likely to be targets of a tissue-specific autoimmune response, and we omitted them from further analysis.

PBMCs of 76 HLA class I-typed patients with chronic plaque psoriasis and 22 healthy individuals without a family history of psoriasis were stimulated in vitro with these proteins. The incidence of streptococcal infection in this patient population was analyzed by Weisenseel et al. (12); an increased prevalence of streptococcal infection was noted in HLA-Cw6⁺ patients. However, because of a mean disease duration of 15 y, a direct correlation between psoriasis onset and streptococcal angina was not possible. T cell stimulation was measured by an ELISPOT assay identifying IFN-γ–producing cells and expressed as the number of SFCs per 1.5 × 10⁵ PBMCs. The results were differentiated according to the main psoriasis risk allele, HLA-Cw6, which was present in 38 of 74 (51.4%) of the psoriasis patients and two of the healthy controls. PHA stimulation served as positive control.

Numbers of SFCs for PHA stimulation and baseline activation (data not shown) were similar in patients and controls. Two of the proteins, PRDX2 (p = 0.0003) and maspin (p = 0.0091), induced a significantly greater number of SFCs in psoriasis patients than in healthy subjects (Fig. 3, Table II). This difference was more pronounced for HLA-Cw6⁺ patients (PRDX2: p = 0.0006; maspin: p = 0.0026) than for HLA-Cw6⁻ patients (PRDX2: p = 0.0301; maspin: p = 0.3768) (Table II) and was reflected by particularly strong responses of individual patients (Fig. 3).

On average, hsp27 and ezrin (Fig. 3) also induced an increased stimulation in psoriasis patients. Because healthy controls responded to a greater extent as well, the difference was significant only for hsp27 in the HLA-Cw6⁻ patients (p = 0.0307; Table II). Stimulation with the cytokeratin preparation reached statistical significance only for HLA-Cw6⁺ patients (p = 0.047). The average T cell stimulation induced by the partial-length keratin 6 proteins was not increased in the group of psoriasis patients versus healthy subjects, although individual patients responded quite strongly (Fig. 3). None of the proteins induced an increased T cell stimulation in the two HLA-Cw6⁺ healthy subjects.

Thus, PRDX2 and ezrin seem particularly immunogenic for the T cell-mediated immune response in the overall psoriasis population: maspin in HLA-Cw6⁺ and hsp27 in HLA-Cw6⁻ psoriasis patients. As partial-length proteins, keratin 6f seems to be immunogenic for select patients.

In vivo and in vitro Ag stimulation of T cells may promote the oligoclonal expansion of Ag-specific T cell populations, which can be identified by a restricted TCR repertoire (29). Moreover, T cell responses to immunodominant antigenic peptides may be characterized by identical or homologous TCR CDR3 motifs in different individuals (30–32). We used these attributes of T cell Ag specificity to further characterize the relevance of the potential autoantigens for the psoriatic immune response. For this purpose, T cell lines were generated in vitro from the PBMCs of a patient with extensive streptococcal-driven type-1 psoriasis by periodic restimulation with the different recombinant proteins or PHA. As determined by FACS analysis, the Ag-specific T cell lines were dominated by CD8⁺ T cells, with a CD4/CD8 ratio of 0.13, 0.32, 0.28, 0.23, 0.43, and 0.92 for the ezrin-, maspin-, PRDX2-, keratin 6-, hsp27-specific, and PHA-stimulated T cell lines, respectively, compared with 0.89 for the corresponding blood T cells.

The TCR usage of the different T cell lines was compared with that of the psoriatic skin lesion and blood lymphocytes by fragment-length spectratyping of the TCR β-chain rearrangements, which had been amplified by 26 PCR reactions specific for the different TCRBV gene families. This approach may identify restricted T cell populations as the result of a biased usage of TCR lengths within a given TCRBV gene family.

TCRBV gene spectratyping of nonstimulated PBLs and the PHA-driven T cell line showed a predominance of quasi-Gaussian β-chain lengths in most TCRBV gene families, which reflected unselected T cell populations. Representative spectratypes are given in Fig. 4. Various TCRBV gene families of the Ag-specific T cell lines and the psoriatic skin lesion displayed strongly biased patterns of fragment lengths, with discrete prominent peaks indicating Ag-driven oligoclonal T cell expansions. Several of these peaks in the Ag-specific T cell lines and the psoriatic skin lesion overlapped, which indicated that, in the different samples, T cells with TCRs of identical lengths had been selected. The rearrangements of these TCRBV gene families were analyzed by cloning and sequencing of the TCR cDNA. They were TCRBV6, BV13.1, BV13.2, and BV14 for the ezrin-specific T cell line; TCRBV3 for the hsp27-specific T cell line; TCRBV8, BV9, BV17, and BV21 for the maspin-specific T cell line; and TCRBV3, BV8, BV13.2, and BV21 for the PRDX2-specific T cell line. None of the TCR fragment-length patterns of the keratin 6-specific T cell lines overlapped with those of the psoriatic skin lesion.

The TCR rearrangements from blood T cells and the PHA-driven T cell line were heterogeneous. In contrast, many of the TCR rearrangements from the Ag-specific T cell lines were highly repetitive and indicated Ag-driven clonal T cell expansions. Clonal TCR rearrangements within a given TCRBV gene family represented up to 93% of the analyzed TCR sequences (Table III). A similar dominance of individual T cell clones was seen within the psoriatic skin lesion, with up to 89% identical TCR rearrangements in a given TCRBV gene family (Table IV), confirming former findings of an oligoclonal psoriatic T cell expansion (7–9).

Homologies in the CDR3 of expanded TCR β-chains may indicate T cell specificity for the same Ag. Comparison of the deduced amino acid sequences showed obvious homologies in the CDR3 of several clonal TCR rearrangements of the Ag-specific T cell lines and the lesional psoriatic infiltrate (Tables III, IV). A clonally expanded CDR3 motif of the ezrin-specific T cell line with the amino acid sequence SSSGS (clone E1; Table III) was observed in two variations, SSSG and LSSG (clones PV20 and 21; Table IV), in the skin lesion. The PRDX2- (clones P2–6), hsp27- (clone H4), and ezrin-specific (clones E2 and E5) T cell lines and the psoriatic skin lesion (clones PV1, 2, 9–13, and 23) contained clonally expanded CDR3-amino acid motifs with variations around a core of four amino acids: SSGTG (Tables III, IV). A clonotypic TCRBV8-CDR3 motif composed of the amino acids VT of the maspin-specific T cell line (clone M1) corresponded to the lesional psoriatic T cell clone PV8. Clonal TCR rearrangements of the maspin-specific (clone M3) and the ezrin-specific (clones E7–9) T cell lines shared a variation with a dominant CDR3 motif: (F/G/P)LAG(G/V) of the psoriatic skin lesion (clones PV4, 7, 14, 16–19, 24, 25, 30, and 31). Up to five amino acids were identical. These CDR3 homologies of the Ag-specific T cell lines and the corresponding psoriatic skin lesion indicate that the proteins, which induced the clonal T cell expansions in vitro, may be relevant for the activation the lesional psoriatic T cell infiltrate.

Only one of these CDR3 motifs, VT, was found in the corresponding blood sample of the patient (B1-3; Table IV). No other clonal selection or bias toward homologous CDR3 motifs was seen in blood, the PHA-driven T cell line, or the negative control T cell lines.

To assess the phenotype of select clonally expanded T cells, the ezrin- and maspin-specific T cell lines were separated into CD4⁺ and CD8⁺ T cells. TCR β-chain cDNA from these T cell subsets was amplified with primer pairs corresponding to the TCRBV9 rearrangement PLAGG (clone M3) of the maspin-specific cell line or the TCRBV13S1 rearrangement FLAGGP (clone E7) of the ezrin-specific T cell line, cloned, and sequenced. Both TCR rearrangements could be assigned exclusively to the CD8⁺ fraction of the respective T cell line and documented that Ag-stimulation had expanded CD8⁺ T cell clones.

The streptococcal-induced autoimmune response against keratinocyte proteins might result from protein regions sharing amino acid sequence homologies. To identify potential mimicry epitopes of streptococci with hsp27, ezrin, maspin, or PRDX2, a direct sequence comparison with the whole genome of GAS was performed. It identified various homologous regions of up to seven identical consecutive amino acids. Several of them referred to streptococcal Ags other than M-proteins, such as Ig-Fc– or fibronectin-binding proteins, RecF, RopA, or proteins from the lantibiotic gene cluster region. Examples are given in Table V.

When used for stimulation of blood lymphocytes, various peptides from the keratinocyte proteins designed according to these homologies (Ezrin-Pep2, PRDX2-Pep2, hsp-Pep2, and Maspin-Pep1) induced a statistically increased T cell activation in psoriasis patients (n = 32) compared with healthy controls (n = 17), as did several peptides from the corresponding streptococcal proteins: RecF, RopA, and Maspin/Strep (Table VI). Differentiation according to HLA-Cw6 showed that the T cell response was particularly strong in HLA-Cw6⁺ patients (Table VI). The differential immunogenicity of select peptides was particular evident for PRDX2: a shift of 3 aa in the peptide sequence (PRDX2-Pep1: aa 24–35 of PRDX2; PRDX2-Pep2: aa 27–38) increased the stimulatory capacity significantly.

No differences between psoriasis patients and healthy controls were observed for stimulation with tetanus toxoid (TT), a widely used tetanus vaccine (33), or PHA.

In a second approach, which also included keratin 6, potential mimicry epitopes were defined according to preferred amino acids at the anchor positions for the HLA-Cw6–binding groove, which are L, I, V, or Y at position 9 or 6 and I, L, F, or M at position 5 (34). Several nona-peptides within keratin 6, maspin, and ezrin exhibited both anchor motifs for HLA-Cw6 and amino acid homologies with streptococcal proteins (Keratin 6-Pep1 to -Pep8, Maspin-Pep2 and -Pep3, and Ezrin-Pep4) (Table VII). Up to seven of nine amino acids were identical. Homologies largely referred to M-proteins but also to other streptococcal proteins, such as serum opacity factor, fibronectin-binding proteins, or streptococcal hsp70. As determined by the IFN-γ–ELISPOT assay in HLA-Cw6⁺ psoriasis patients (n = 18), several peptides with homologies to streptococcal M-proteins (keratin-Pep2, -Pep3, and -Pep5 and Maspin-Pep 2) induced a pronounced T cell activation, which was greater than T cell activation by TT. Unlike the two half-length keratin proteins, defined keratin 6 peptides promoted T cell activation using this approach (Fig. 5).

Several lines of evidence support that the keratinocyte proteins ezrin, maspin, hsp27, PRDX2, and potentially keratin 6 may serve as autoantigens of a pathogenic T cell response in psoriasis induced by streptococcal infection. Streptococci-specific rabbit sera and psoriasis patients’ sera reacted against these proteins. The reactivity of the streptococci-specific rabbit sera was tissue selective against proteins from keratinocytes but not from a lymphoblastoid B cell line or the epidermoid carcinoma cell line A431. The proteins ezrin, maspin, hsp27, and PRDX2 induced significantly increased T cell activation in PBMCs from psoriasis patients. Upon repetitive restimulation, they promoted an oligoclonal expansion of T cells, which used TCR β-chain rearrangements sharing a high degree of CDR3 homology with clonal TCR rearrangements of the corresponding psoriatic skin lesion. The antigenicity of the proteins could be assigned to peptides sharing amino acid sequence homologies with streptococcal proteins.

Three of these proteins were keratinocyte specific (keratin 6) or selective for epithelial surfaces (ezrin and maspin). Keratin 6 had formerly been proposed as a psoriatic autoantigen because of sequence homologies with streptococcal M-protein (16). However, in our approach, overlapping keratin 6f proteins were immunogenic only for select patients or when used as peptides designed according to HLA-Cw6–binding motifs. This could reflect a lack of ability of the proteasome to cleave appropriate peptides within the APCs.

Ezrin, a cytoskeletal linker protein, is a particularly interesting candidate as a psoriatic autoantigen. A high expression in keratinocytes, the apical microvilli of the enteral mucosa, and the anterior myoepithelial layer of the iris could constitute a potential autoantigenic link between psoriasis, inflammatory bowel disease, and autoimmune uveitis, which are clinically associated (35–38).

Maspin is a soluble cytoplasmic, membranous, and secreted extracellular serine protease inhibitor (serpin) and potent tumor suppressor inhibiting cellular invasion, motility, and proliferation. Maspin is restricted to epithelial cell types and is abundantly expressed in epidermis, stratified squamous epithelium of the tonsils, and the mucosal epithelium of the small intestine (39). Its expression is increased in hyperplastic psoriatic epidermis (40).

Hsp27, an evolutionarily highly conserved molecular chaperon, is homogeneously distributed throughout normal epidermis and strongly increased in psoriatic epidermis (41). An immune response against microbial hsp may cross-react against the corresponding autologous hsp because of molecular mimicry and trigger autoimmune pathology (42). Enhanced expression of hsp27 under stress can unveil previously hidden antigenic determinants, perpetuate autoimmune reactivity (43), and, thus, might promote stress-related worsening of psoriasis. Additionally, hsp27-specific T cells could contribute to the increased risk for cardiovascular morbidity seen in psoriasis (44). Increased endothelial hsp27 expression and intimal infiltration of hsp27-reactive T cells are considered important early events in the immune-mediated pathogenesis of arteriosclerosis (45).

PRDX2 is another stress-induced cytoplasmatic protein. It may be upregulated in the skin by IFN-γ, which is abundantly expressed within psoriatic lesions (46).

The particular antigenicity of the different proteins for psoriasis patients was reflected by a significantly increased T cell activation and by select oligoclonal T cell expansions following periodic Ag-specific restimulation of blood T cells from a patient with streptococcal-driven psoriasis. CDR3s of several dominant TCR β-chain clonotypes from these T cell lines shared up to four or five amino acids in sequence with TCR clonotypes from the corresponding psoriatic lesion. Given the 20 different amino acids, the overall likelihood that two TCR rearrangements share four or five identical amino acids is 1:20⁴ = 1:160,000 or 1:20⁵ = 1:3,200,000. Accordingly, the CDR3 homologies of the Ag-specific T cell lines with the lesional psoriatic clonotypes seem to be nonrandom and support that maspin, ezrin, hsp27, and PRDX2 might represent targets of the lesional psoriatic immune response. This conclusion is further supported by the observation that one of the dominant CDR3 motifs, (F/G/P)LAG(G/V), of the T cell lines and the psoriatic skin lesion were previously identified as a conserved clonotypic CDR3 in HLA-Cw6⁺ psoriasis patients (9). Furthermore, several clonotypic TCRs with homologous CDR3s in the Ag-specific T cell lines (E7, E8, and M3) and the psoriatic skin lesion (PV14, 17, 18, and 24) used different TCR Vβ gene families but had rearranged identical Jβ genes. Because only an estimated 10% of HLA-Cw6⁺ individuals develop psoriasis (47), this BV gene variability, but conserved Jβ gene usage, may reflect the private CD8 T cell repertoires and contribute to the different T cell responses of different individuals to the different Ags, as reported for cross-reactive immune responses in heterologous immunity (48).

Sequence alignments of the different proteins with the whole genome of GAS identified various homologous peptides of up to seven identical amino acids. These homologies were not confined to streptococcal M-proteins but involved other streptococcal proteins, such as RopA, RecF, or FcR proteins. Homologous candidate epitopes from the putative autoantigens or the streptococcal proteins were equally effective in inducing T cell responses, supporting an immunologic link between foreign and self-Ags. This wider spectrum of streptococcal correlates may explain why psoriasis, unlike other nonsuppurative streptococcal sequelae, such as rheumatic fever or poststreptococcal glomerulonephritis, which require particular rheumatogenic or nephritogenic M protein serotypes, may be induced by different strains of Lancefield group A, as well as group C and G, streptococci (10, 49). Unfortunately, the T cell lines could not be tested with the different peptides, because they died off after four rounds of Ag-specific restimulation, probably as a result of the predominance of CD8⁺ T cells and the limited life span of effector cells (50).

The antigenicity of the different proteins and their peptides was particularly evident for HLA-Cw6⁺ patients. Several peptide regions within the proteins showed binding capacities for HLA-Cw6 and homologies to streptococcal proteins, and they activated T cells in HLA-Cw6⁺ patients. Accordingly, the association of psoriasis with HLA-Cw6 might reflect a select capacity of this HLA allele to present common epitopes of streptococcal Ags and their skin-associated mimicry correlates. Ag presentation via MHC class I molecules is supported by the predominance of CD8⁺ T cells in the Ag-specific T cell lines and the CD8 phenotype of two dominant TCR clonotypes (M3 and E7) from the ezrin- and maspin-specific T cell lines. The stimulation of CD8⁺ T cells by soluble Ags suggests that, in vivo, the proteins may be subjected to cross-presentation following a release from apoptotic keratinocytes, which may be further augmented by increased cell damage in psoriasis (51). Cross-presentation is crucial for the development of cytotoxic CD8⁺ T cell responses against tumors and pathogens, as well as for the induction of tolerance. Tolerance to autologous proteins presented this way may be broken by their uptake in APCs as immune complexes (52) or when a simultaneous stimulation of APCs via TLRs occurs (53). GAS may engage different TLRs (54, 55) and induce autoantigen-specific Abs, which then may form immune complexes with the different autoantigens. By these mechanisms, streptococcal infection could break tolerance of CD8⁺ T cells and allow their subsequent autoantigen-specific expansion in psoriasis. Within the skin, keratinocytes might also present these proteins to CD8⁺ T cells directly. Analysis of cell suspensions prepared from psoriatic skin lesions documented a direct contact of CD8⁺ T cells with dendritic cells and keratinocytes (S.M. Kim, P. Besgen, J. Nickel, K. Siewert, K. Dornmair, and J.C. Prinz, manuscript in preparation).

Thus, the proteins identified in our study are interesting candidates for streptococcal-induced autoimmunity in psoriasis. They document that psoriasis may involve a complex Ag-specific autoimmune response against several keratinocyte proteins and serve to further characterize the relationship between psoriasis and its comorbidities. The mechanism making them targets of a skin-selective autoimmune response might involve molecular mimicry due to homologous peptide regions with streptococcal Ags, TCR cross-reactivity, a high density of particular subsets of APCs capable of cross-presentation in the skin, and skin-selective homing receptors on the responsive T cells (15), which now can be analyzed in more detail.

We thank Dr. M. Arpin, Laboratoire de Morphogenèse et Signalisation Cellulaires, Institut Curie, Paris, France, for providing the ezrin clone; Dr. A. Efstratiou, Public Health Laboratory Service, London, U.K., for providing S. pyogenes, serotype M12; and Dr. K. Dornmair, Max-Planck-Institute for Neuroimmunology, Martinsried, Germany, for critically reviewing the manuscript.

Disclosures The authors have no financial conflicts of interest.