Research ArticleCeliac Disease

Comprehensive, Quantitative Mapping of T Cell Epitopes in Gluten in Celiac Disease

See allHide authors and affiliations

Science Translational Medicine  21 Jul 2010:
Vol. 2, Issue 41, pp. 41ra51
DOI: 10.1126/scitranslmed.3001012


Celiac disease is a genetic condition that results in a debilitating immune reaction in the gut to antigens in grain. The antigenic peptides recognized by the T cells that cause this disease are incompletely defined. Our understanding of the epitopes of pathogenic CD4+ T cells is based primarily on responses shown by intestinal T-cells in vitro to hydrolysates or polypeptides of gluten, the causative antigen. A protease-resistant 33-amino acid peptide from wheat α-gliadin is the immunodominant antigen, but little is known about the spectrum of T cell epitopes in rye and barley or the hierarchy of immunodominance and consistency of recognition of T-cell epitopes in vivo. We induced polyclonal gluten-specific T cells in the peripheral blood of celiac patients by feeding them cereal and performed a comprehensive, unbiased analysis of responses to all celiac toxic prolamins, a class of plant storage protein. The peptides that stimulated T cells were the same among patients who ate the same cereal, but were different after wheat, barley and rye ingestion. Unexpectedly, a sequence from ω-gliadin (wheat) and C-hordein (barley) but not α-gliadin was immunodominant regardless of the grain consumed. Furthermore, T cells specific for just three peptides accounted for the majority of gluten-specific T cells, and their recognition of gluten peptides was highly redundant. Our findings show that pathogenic T cells in celiac disease show limited diversity, and therefore suggest that peptide-based therapeutics for this disease and potentially other strongly HLA-restricted immune diseases should be possible.


CD4+ T cell recognition of peptides derived from dietary wheat gluten and related prolamins from barley and rye is fundamental to the pathogenesis of celiac disease (CD) (1). Intestinal gluten-specific T cells, which confer susceptibility to CD, are restricted by the major histocompatibility complex molecules human leukocyte antigen (HLA)–DQ2 or HLA-DQ8 (2, 3). Most known gluten-derived epitopes are relatively resistant to digestive proteases and have been selectively deamidated by tissue transglutaminase (tTG), an enzyme whose activity is increased in the inflamed intestinal mucosa (46). Major clinical problems in the management of CD are that the diagnostics are suboptimal and invasive and that patients must rely on a complex, costly, and lifelong therapy—strict dietary gluten exclusion (7). Even trace dietary contamination by gluten is injurious, and full recovery of intestinal histology is achieved in fewer than half of adults (8, 9). There is a demand for effective therapeutic adjuncts or alternatives to the gluten-free diet; several are under development and their design is predicated on clear understanding of the immunotoxic gluten peptides. One such therapeutic candidate that is effective in various murine models of T cell–mediated disease requires repeated dosing with peptides corresponding to immunodominant T cell epitopes (10).

Designing a peptide immunotherapy would be straightforward if gluten-specific T cells from all CD patients were specific for a limited number of gluten peptides. Indeed, the importance of HLA-DQ2 and HLA-DQ8 in determining susceptibility suggests that the pathogenic T cell response to gluten is highly focused. But despite important advances, two persistent problems have prevented a comprehensive understanding of the pathogenic T cell response in CD. First, T cells specific for gluten are so rare in intestinal tissue that they cannot be directly evaluated. Second, there are hundreds of wheat gluten proteins and many others in rye and barley that are potentially toxic in CD; thus, there are too many candidate epitopes in gluten to simultaneously assess them all in a single unbiased T cell screening assay.

Epitope mapping studies have overcome the scarcity of gluten-reactive T cells in patient samples with T cells derived from intestinal tissue that have been expanded to measurable numbers over days or weeks by periodically restimulating with wheat gluten and mitogens (1118). When screened against a protease-digested recombinant α2-gliadin polypeptide preincubated with tTG, intestinal T cell lines (TCLs) selectively recognize three distinct, overlapping epitopes in tandem repeat (DQ2-α-I, PFPQPELPY; DQ2-α-II, PQPELPYPQ; and DQ2-α-III, PYPQPELPY), which are present within a single protease-resistant 33–amino acid peptide (4). This 33–amino acid oligopeptide is usually the most bioactive gluten peptide recognized by gluten-specific TCLs from HLA-DQ2+ CD donors. Occasional discrepant findings among studies of gluten epitopes have been attributed to subtly different experimental protocols, the use of material from children rather than adults, and the use of gluten from different sources prepared in various ways (15, 17).

Distinct immunoreactive peptides have been successfully isolated from crude hydrolysates of gluten and several pure recombinant α- and γ-gliadins, but gluten is highly insoluble and, despite the fact that all wheat gluten fractions are immunotoxic, there have been no systematic searches for epitopes derived from the aqueous insoluble glutenins or from ω-gliadins, nor have there been systematic epitope mapping studies of prolamins in barley (hordeins) or rye (secalins). With recent improvements in and reduced cost of peptide synthesis, epitope mapping studies have used synthetic gluten peptides spanning an α-gliadin and a γ-gliadin polypeptide (19) and also peptides corresponding to certain sequences predicted to be susceptible to deamidation by tTG and likely to bind efficiently to HLA-DQ2 (20). Systematic searches for T cell epitopes in the many hundreds of complementary DNA–derived wheat, barley, and rye prolamins have been limited to predictive studies; many candidates have been proposed but there is no sense of their relative contribution to the gluten-specific T cell response nor whether they represent distinct epitopes (11).

The only known source of fresh polyclonal gluten-specific T cells at sufficiently high frequency to enumerate without expansion in vitro is peripheral blood from CD donors directly after short-term oral gluten challenge (21). T cells specific for gluten are maximal in blood 6 days after volunteers commence oral gluten challenge and can be enumerated by overnight enzyme-linked immunospot (ELISpot) assay. Such T cells are CD4+, secrete interferon-γ (IFN-γ), preferentially recognize deamidated gluten, and express the α4β7 integrin associated with homing to the intestinal lamina propria (21, 22). A deamidated 17–amino acid oligopeptide encompassing the overlapping DQ2-α-I and DQ2-α-II epitopes is the only peptide in A-gliadin recognized by peripheral blood T cells that have been induced by oral gluten challenge in HLA-DQ2+ CD volunteers.

To test whether the immunotoxicity of prolamins from wheat, barley, and rye is truly limited to just a few critical peptides, we have exploited oral challenge with wheat, barley, or rye to enable antigen presentation to occur in vivo and the sampling of peripheral blood T cells induced in >200 HLA-DQ2+ CD patients. By developing a high-throughput IFN-γ ELISpot assay with peripheral blood mononuclear cells (PBMCs) collected 6 days after commencing oral grain challenge (GC), we have screened customized peptide libraries encompassing >16,000 unique candidate 12–amino acid sequences from all toxic prolamin fractions of wheat, barley, and rye. T cell clones (TCCs) were used to confirm minimal epitopes and to determine the promiscuity of peptide recognition. Unbiased screening of these freshly collected T cells revealed that the immunotoxicity of gluten is both consistent and reducible to three highly immunogenic peptides. Unbiased screening also reveals that promiscuous recognition of gliadin-, hordein-, and secalin-derived peptides by T cells whose optimal epitope is an ω-gliadin/C-hordein–derived sequence, rather than epitopes present in α-gliadin, accounts for the immunotoxicity common to wheat, barley, and rye.


Quantitative ex vivo assessment of patients’ T cells specific for published epitopes

Gluten peptides recognized by TCCs and TCLs were assessed by IFN-γ ELISpot with PBMCs isolated from nine HLA-DQ2+ CD donors 6 days after they began a 3-day wheat GC. IFN-γ ELISpot responses to deamidated gliadin were detected in eight donors [median, 23; range, 13 to 153 spot-forming units (SFUs) per million PBMCs]; among seven of these eight donors, the α-gliadin 17–amino acid oligopeptide QLQPFPQPELPYPQPQP (5 μM) and the related 33–amino acid oligopeptide QLQPFPQPELPYPQPELPYPQPELPYPQPQPF also elicited detectable responses of a similar magnitude (Fig. 1A). Yet, in contrast, DQ2-γ-IV (SQPEQEFPQ) was the only one of 10 other reported gluten T cell epitopes to produce a detectable response, and this was relatively weak and seen only in one donor. We concluded that although T cells specific for DQ2-α-I and DQ2-α-II make a substantial contribution to the T cell population induced by wheat GC, the specificity of many other gluten-specific T cells induced by GC could not be explained by previously reported epitopes.

Fig. 1

IFN-γ ELISpot response after GC defines a hierarchy of peptide responses. Six days after commencing 3-day wheat, barley, or rye challenge, PBMCs from HLA-DQ2+ CD donors were assessed for IFN-γ ELISpot response to previously characterized epitopes and comprehensive peptide libraries. (A) PBMCs from nine donors were assessed after wheat challenge. Dashed lines reflect the cutoff threshold for considering responses “positive,” four times the response to medium alone, and >10 SFUs per well adjusted to SFUs per million PBMCs. (B) Three-day wheat, barley, or rye challenge was undertaken in a total of 86 CD volunteers. PBMCs from donors were used to screen one of five first-round gluten peptide libraries that included 2723 20–amino acid oligopeptides encompassing 16,838 unique 12–amino acid oligopeptides in 313 GenBank entries for gliadins, LMW glutenins, and HMW glutenins (T. aestivum) (after wheat challenge), hordeins (H. vulgare) (after barley challenge), and secalins (S. cerale) (after rye challenge). (C) PBMCs from CD donors were used to screen three second-round gluten peptide libraries that included 441 16–amino acid oligopeptides encompassing reactive 20–amino acid oligopeptides from first-round wheat gliadin and LMW and HMW glutenin libraries (28 donors after wheat challenge), 80 16–amino acid oligopeptides encompassing reactive 20–amino acid oligopeptides from the first-round barley hordein library (10 donors after barley challenge), and 53 16–amino acid oligopeptides encompassing reactive 20–amino acid oligopeptides from the first-round rye secalin library (10 donors after rye challenge). The line indicates minimum score of 5 required to be considered a confirmed stimulatory peptide.

Comprehensive screening of wheat gluten, rye secalin, and barley hordein peptides

Conventionally, CD4+ T cell epitopes are mapped with overlapping 15- to 20–amino acid oligopeptides encompassing all antigen-derived 10- to 12–amino acid oligopeptide linear sequences. This traditional approach is impractical to map all T cell epitopes in gluten; even in 2001 when this study commenced, there were several hundred candidate prolamin genes isolated from wheat, rye, and barley. Instead, we took advantage of the close homology between protein sequences derived from each subgroup of gluten proteins and developed an iterative algorithm to define all unique 12–amino acid oligopeptides encoded by gliadin and glutenin genes in Triticum aestivum (wheat), hordein genes in Hordein vulgare (barley), and secalin genes in Secale cerale (rye) and then to design a minimal number of 20–amino acid oligopeptides encompassing all of these 12–amino acid oligopeptides (23).

Because intestinal and fresh peripheral blood T cells induced by oral GC in CD patients preferentially recognize deamidated prolamins, individual 20–amino acid oligopeptides (25 μg/ml) in each of five “comprehensive” gliadin, low–molecular weight (LMW) glutenin, high–molecular weight (HMW) glutenin, hordein, and secalin libraries (table S1) were pretreated with tTG and then screened in overnight IFN-γ ELISpot assays with fresh PBMCs from a total of 86 HLA-DQ2+ CD volunteers after GC with wheat (to assess the gliadin and half the LMW glutenin library or the HMW glutenin and the second half of the LMW glutenin library), barley (to assess the hordein library), or rye (to assess the secalin library).

A hierarchy of stimulatory peptides was clearly demonstrated for each first-round library (Fig. 1B). For each 20–amino acid oligopeptide, a score (out of 100) was calculated that was equal to the average relative frequency of 20–amino acid oligopeptide specific T cells present in blood (see Materials and Methods). Overall, 61% (1666 of 2723) of 20–amino acid oligopeptides had a score of 0, 13% (343) had a score between 0 and 1, 18% (495) had a score between 1 and 5, 5% (125) had a score between 5 and 10, and 2% (66) had a score between 10 and 30, whereas only 1% (28) had a score >30. Of the first-round tTG-treated 20–amino acid oligopeptides, 219 had scores >5, representing 5% (125 of 2152) of wheat gliadin and glutenin 20–amino acid oligopeptides, 14% (59 of 416) of hordein 20–amino acid oligopeptides, and 23% (35 of 155) of secalin 20–amino acid oligopeptides.

Peptides eliciting at least 5% of the IFN-γ ELISpot response of the most active 20–amino acid oligopeptide in each individual donor were fine-mapped with 16–amino acid oligopeptides (as described in Materials and Methods). Design of second-round peptides predicted that epitopes fell within 12–amino acid sequences having glutamine residues deamidated by tTG to create anchors for 9–amino acid epitopes at positions 4, 6, or 7, facilitating binding to HLA-DQ2; such 12–amino acid sequences had glutamine at position 7 with proline at position 9 but not 8 or 10, and/or a bulky hydrophobic residue at position 10. The 12–amino acid sequence candidate was then flanked at positions −1 and 13 with native residues, and then at positions −2 and 14 with glycine residues. The final 16–amino acid oligopeptide was then synthesized with glutamate or glutamine at positions predicted to be susceptible to deamidation by tTG. For first-round 20–amino acid oligopeptides that lacked glutamine residues predicted to be deamidated by tTG, two 16–amino acid oligopeptides overlapping by 12 residues were synthesized. Second-round libraries consisted of 554 wheat gliadin- and glutenin-derived 16–amino acid oligopeptides, 89 hordein-derived 16–amino acid oligopeptides, and 64 secalin-derived 16–amino acid oligopeptides; they were screened by IFN-γ ELISpot with PBMCs from a further cohort of HLA-DQ2+ CD volunteers after wheat (n = 28), barley (n = 10), or rye (n = 10) GC. A hierarchy of stimulatory peptides was again clearly demonstrated for each second-round library (Fig. 1C).

We elected to focus on the most active candidate deamidated 12–amino acid oligopeptide or wild-type 16–amino acid sequences represented among peptides with a score of ≥5 in both the first-round and the second-round screening libraries (Fig. 1C). Overall, 37 wheat gluten–, 30 barley hordein–, and 29 rye secalin–derived 12- or 16–amino acid sequences were confirmed as distinct T cell–stimulatory sequences (Table 1). These T cell–stimulatory gluten peptides rarely elicited IFN-γ ELISpot responses when incubated with PBMCs collected from CD donors before GC (table S2) and did not elicit responses in PBMCs collected from healthy HLA-DQ2+ volunteers who had eaten a gluten-free diet for 4 weeks and then had wheat GC (table S3).

Table 1

Hierarchy of peptides after wheat, barley, or rye challenge in vivo and recognition by TCCs in vitro. First-round library 20–amino acid oligopeptides with a score of >5 confirmed by second-round 16–amino acid oligopeptides (core 12–amino acid sequence in bold) with score of >5 using day 6 PBMCs after gluten challenge. Recognition determined by TCCs raised to cognate ligand and incubated with second-round peptides and verification library (25 mg/ml). Epitopes: α-I PFPQPELPY, α-II PQPELPYPQ, ω-I PFPQPEQPF, ω-II PQPEQPFPW, and Hor-I PIPEQPQPY. W, wheat gliadin (unless LMW or HMW indicated); B, barley; R, rye. Abbreviations for the amino acids are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.

View this table:

Peptides were called dominant for a particular CD donor if they elicited at least 70% of the response of the most active peptide in each library for that donor (Table 1). In wheat, 20 tTG-treated second-round 16–amino acid oligopeptides were dominant in at least one donor, but only 7 of these were dominant in >10% of donors. Eleven hordein 16–amino acid oligopeptides and seven secalin 16–amino acid oligopeptides elicited dominant responses in >10% of donors after barley and rye challenge, respectively. The highest-scoring tTG-treated second-round wheat gluten-, hordein-, and secalin-derived sequences (W01, LPYPQPQLPYPQ; B01, QPFPQPQQPFPW; and R01, QPFPQPQQPIPQ) were dominant in >50% of donors, and each was recognized by >80% of donors.

Fifty-two of the 96 immunogenic second-round peptides included known or predicted T cell epitopes (11, 15, 16). Included within the 44 newly described T cell–stimulatory gluten peptides identified were four closely related HMW gluten 16–amino acid oligopeptides that resembled the HLA-DQ8–restricted epitope present in HMW glutenin (DQ8-GLT-I: QGYYPTSPQ) (24). The most active of these HMW glutenin peptides (W21: QGQQGYYPISPQQSGQ) was recognized by PBMCs from fewer than one-quarter of donors after wheat GC but was occasionally dominant.

Seven epitopes previously defined with TCCs or TCLs in vitro were represented among the 96 confirmed immunogenic peptides: DQ2-α-I, DQ2-α-II, DQ2-α-III, DQ2-γ-III, DQ2-γ-IV, DQ2-γ-VI, and DQ2-γ-VII. Direct comparison of the relative size of the T cell populations specific for these individual epitopes is complex because many of the immunogenic second-round peptides do or are likely to encompass more than one distinct epitope. The second-round peptides that included the DQ2-α-II and DQ2-α-I or DQ2-α-III epitopes achieved the highest scores after wheat GC (74 and 75, respectively), but the maximal scores of 16–amino acid oligopeptides that included other epitopes were substantially less (DQ2-γ-III: score, 9; DQ2-γ-IV: score, 20; DQ2-γ-VI: score, 6; and DQ2-γ-VII: score, 13). After barley or rye GC, 16–amino acid oligopeptides including DQ2-γ-VI were almost half as active as the most active peptides, and 16–amino acid oligopeptides including DQ2-γ-VII were less than one-sixth as active as the most active peptides.

All but the four HMW glutenin peptides and one gliadin peptide included glutamine at a position predicted to be deamidated by tTG and to enhance binding to HLA-DQ2.

Despite its logistical complexity and the requirement for a large number of CD volunteers, this large-scale, high-throughput peptide library screen exploiting fresh peripheral blood T cells was highly informative; each prolamin fraction of wheat, barley, and rye contained a defined hierarchy of immunogenic peptides.

Dependence of the frequency of peptide-specific T cells in blood on the cereal ingested

The most active T cell–stimulatory second-round library peptides were synthesized to high purity and characterized in greater detail with PBMCs from further HLA-DQ2+ CD donors after wheat (n = 14), barley (n = 11), or rye (n = 8) GC. The dominance of stimulatory peptides was remarkably different after wheat, barley, or rye GC (Table 2). Peptides derived from wheat α-gliadin, including the DQ2-α-I, DQ2-α-II, or DQ2-α-III epitopes, were dominant only after wheat GC. Other frequently dominant peptides were almost exclusively dominant after the ingestion of just one grain. For example, the C-hordein 16–amino acid oligopeptide PQQPIPEQPQPYPQQP (B08-E7) was active only after barley challenge, and the ω-secalin sequence PFPQQPEQIIPQ (R11-E7) only after rye challenge (Fig. 2, B to D). For other peptides such as those including the motif QPFP(W,L,Y,V,I)QPEQPFPQ and also the protease-resistant γ-gliadin 26–amino acid oligopeptide FLQPEQPFPEQPEQPYPEQPEQPFPQ (11), T cell responses were relatively stronger after barley or rye challenge but still could be detected after wheat challenge (Table 2).

Table 2

Grain specificity of select stimulatory sequences after in vivo wheat, barley, or rye challenge, and recognition by T cells in vitro.

View this table:
Fig. 2

The hierarchy of immunodominance is dependent on the cereal grain ingested. (A to D) Four immunodominant stimulatory sequences were tested after wheat (n = 10), barley (n = 8), or rye (n = 6) challenge to examine grain specificity. SFUs were expressed as a percentage of the most active deamidated prolamin peptide tested for each donor. The median is shown.

In contrast, peptides sharing the sequence QPFPQPEQP(F,I)P(W,L,Y,Q)(Q,S) were equally active and frequently dominant after wheat, barley, or rye challenge. The ω-gliadin/C-hordein–derived sequence QPFPQPEQPFPW (W03-E7) was consistently the most active of this peptide family (Fig. 2A). PBMCs isolated from CD donors after GC allowed detailed characterization of the extended ω-gliadin sequence, including W03 [AAG17702(81–102)], with lysine-substituted variants and overlapping, glutamate-substituted 15–amino acid oligopeptides; the sequence PFPQPEQPFPW with a single deamidated glutamine was critical for recognition by T cells induced by wheat, barley, or rye GC (fig. S1). These findings in Australian CD volunteers corroborated earlier observations when a pilot 652-member 20–amino acid oligopeptide gliadin peptide library (table S1) was screened with PBMCs from English HLA-DQ2+ CD donors after either wheat (n = 13) or purified rye challenge (n = 6) (see Materials and Methods and table S4).

Collectively, these unanticipated findings show that the specificities and relative importance of T cell responses generated in vivo depend on the cereal ingested. This unbiased analysis showed that it is an ω-gliadin/C-hordein–derived sequence that is the most consistently stimulatory gluten peptide after wheat, barley, and rye ingestion in vivo.

Distinct families of gluten peptides recognized by TCCs specific for dominant peptides

TCCs were isolated from intestinal lamina propria mononuclear cells (LPMCs) or PBMCs cultured with deamidated gliadin or peptides showing distinct patterns of immunodominance after cereal GCs: wheat α-gliadin (W02-E7), wheat ω-gliadin/barley hordein (W03-E7), barley C-hordein (B08-E2E7), and rye ω-secalin (R11-E4E7). All 12 TCCs were HLA-DQ2–restricted and had T helper 1 (TH1) or TH0 cytokine profiles (table S5). Epitopes recognized by TCCs were determined with lysine-substituted variants of the parent peptide. As expected, the DQ2-α-I or DQ2-α-II epitopes were the core 9–amino acid sequences recognized by the five TCCs raised against the α-gliadin–derived peptide W02-E7. Two of the four TCCs raised against the ω-gliadin/C-hordein peptide W03-E7 recognized the core 9–amino acid sequence PFPQPEQPF (herein named DQ2-ω-I) and the other two recognized the overlapping 9–amino acid sequence PQPEQPFPW (herein named DQ2-ω-II). The single TCC raised against the C-hordein peptide B08-E2E7 recognized the 9–amino acid sequence PIPEQPQPY (herein named DQ2-Hor-I), whereas the 9–amino acid core epitope for the solitary TCC raised to the ω-secalin peptide R11-E4E7 was not determined but named here as DQ2-Sec-I. When screened against the second-round and verification libraries, the TCC raised against deamidated gliadin was found to be specific for the γ-gliadin–derived peptide W11-E7 (9–amino acid core sequence not determined).

Next, we determined whether TCCs were highly specific or degenerate in their recognition of the other immunostimulatory gluten peptides that we had previously identified with fresh polyclonal PBMCs from CD donors. TCCs were screened against 697 tTG-treated 16–amino acid oligopeptides in the second-round library and also against 3028 18–amino acid oligopeptides in a freshly synthesized verification library encompassing all unique deamidated and wild-type 10–amino acid sequences from gliadin, hordein, and secalin (table S1). TCCs were screened by IFN-γ ELISpot with a concentration of peptide equivalent to that previously used to screen the first- and second-round libraries with PBMCs from CD donors. There was little cross-recognition of the four dominant peptides (W02-E7, W03-E7, B08-E2E7, and R11-E4E7) by TCCs but substantial redundancy of peptide recognition for many of the subdominant gluten peptides (Tables 1 and 2 and table S5). Remarkably, 11 clones specific for six epitopes (DQ2-α-I, DQ2-α-II, DQ2-ω-I, DQ2-ω-II, DQ2-Hor-I, and DQ2-Sec-I) recognized 22 of 32 gliadin, 26 of 30 hordein, and 22 of 29 secalin sequences that we had previously defined as T cell–stimulatory peptides for polyclonal T cells from CD donors (Fig. 3, A and B). TCCs specific for ω-gliadin/C-hordein–derived W03-E7 were the most cross-reactive and recognized 54 of the 91 gliadin/hordein/secalin T cell–stimulatory peptides. Among the 54 peptides recognized by TCCs specific for W03-E7 were 18 peptides that are also recognized by TCCs specific for other dominant peptides (W02-E7, B08-E2,E7, or R11-E4E7). In contrast, there was almost no overlap between peptides recognized by TCCs specific for W02-E7, B08-E2,E7, or R11-E4E7 (Fig. 3B). Peptides recognized by TCCs specific for W02-E7, B08-E2,E7, or R11-E4E7 were typically derived from the same cereal as the cognate ligand; for example, TCCs specific for DQ2-α-I or DQ2-α-II selectively recognized wheat gliadin-derived peptides (Tables 1 and 2).

Fig. 3

Immunostimulatory peptides for PBMCs stimulated most TCCs, but TCC cross-reactivity with other dominant peptides was restricted to clones specific for ω-gliadin (W03). Recognition of confirmed stimulatory sequences by 11 TCCs raised to W02-E7 (n = 5), W03-E7 (n = 4), B08-E2E7 (n = 1), and R11-E4E7 (n = 1). (A) In vivo GC followed by IFN-γ ELISpot characterized 96 immunogenic gluten sequences. TCCs were tested against 91 sequences and found to proliferate or secrete IFN-γ in response to 70. (B) The relation between the TCCs and the 70 peptides they recognize is depicted. Five TCCs recognizing W02 reacted to a total of eight sequences. Of these, five of eight sequences were also recognized by TCCs raised to W03. The four TCCs specific for W03 recognized a total of 54 sequences, of which 13 were also recognized by TCCs raised to B08 and/or R11. The TCCs specific for B08 and R11 responded to 9 and 18 sequences, respectively, with only 1 sequence common to both.

Hence, 70 of 96 T cell–stimulatory gluten peptides identified from PBMCs from HLA-DQ2 CD donors after oral GC were recognized by TCCs raised against four peptides.

Defining the major immunodominant peptides in CD

If a few dominant immunostimulatory peptides are the preferred ligands for most gluten-specific T cells in vivo, it would be expected that a mixture of these peptides would elicit a substantial proportion of the response to gluten and also that responses to individual peptides would be additive if assessed with fresh polyclonal T cells from CD donors after oral GC. We assessed IFN-γ ELISpot responses to mixtures of up to 12 peptides by using PBMCs from CD donors; the peptides included the four dominant peptides from wheat, barley, and rye (W02-E7, W03-E7, B08-E2E7, and R11-E4E7). Other peptides were selected because they were occasionally dominant and from diverse gluten fractions including wheat glutenins and gliadins, or oat avenins. Because T cells specific for several peptides were elicited only after wheat, barley, or rye GC, the oral GC was modified to include equal amounts of wheat, barley, and rye. IFN-γ ELISpot responses elicited by the equimolar mixture of the three dominant peptides encompassing DQ2-α-I, DQ2-α-II, DQ2-ω-I, DQ2-ω-II, and DQ2-Hor-I epitopes [W02-E7, W03-E7, and B08-E2E7; cocktail 2 (C2)] were comparable to expanded cocktails of 6 or 12 peptides, but more than twice the mixture of the dominant α- and ω-gliadin wheat peptides (W02-E7 and W03-E7) and eight times greater than the single peptide encompassing DQ2-α-I and DQ2-α-II (W02-E7) (Fig. 4A). When assessed with PBMCs collected from CD donors after single GC, IFN-γ ELISpot responses to C2 (50 μM) were equivalent to as much as 90% of that elicited by optimal concentrations of tTG-treated wheat gliadin, barley hordein, or the most immunogenic secalin fraction (ω-secalin) (320 μg/ml), respectively (Fig. 4, B to D).

Fig. 4

A minimal three-peptide mixture can stimulate an optimal IFN-γ ELISpot response. IFN-γ ELISpot was used to compare the frequencies of T cells from CD donors specific for individual dominant gluten peptides (50 μg/ml) or mixtures of peptides (constituent peptides each 50 μg/ml) after combined challenge with equal amounts of wheat, barley, and rye (n = 13) (A). Each donor’s responses to peptides and mixtures of up to 12 peptides were normalized against their response to the 12-peptide mixture. Maximal T cell responses were observed to the mixture of three peptides (C2). (B to D) Next, IFN-γ ELISpot was used to compare frequencies of T cells from CD donors after wheat (n = 10) (B), barley (n = 5) (C), or rye GC (n = 6) (D) to the mixture of three peptides (C2) (50 μg/ml) to the respective deamidated wheat gliadin, barley hordein, or rye ω-secalin fraction (320 μg/ml). Responses to C2 were equivalent to as much as 90% of the response to prolamin. Peptides assessed were capped by an N-pyroglutamyl residue (pE) and C-amidated. The peptides in C1 were capped variants of α-gliadin W02-E7 (pELQPFPQPELPYPQPQ-NH2) and ω-gliadin/C-hordein W03-E7 (pEQPFPQPEQPFPWQP-NH2); C2 was expanded to also include hordein B08-E2E7 (pEPEQPIPEQPQPYPQQ-NH2); C3 was further expanded to include secalin R11-E4E7 (pEQPFPEQPEQIIPQQP-NH2), homolog of oat avenin–derived T cell–stimulatory peptide (pEYQPYPEQEQPILQQ-NH2), and γ-gliadin W36 (pEYEVIRSLVLRTLPN-NH2); and C4 was expanded to 12 peptides with the addition of HMW glutenin W21 (pEGQQGYYPISPQQSGQ-NH2), LMW T cell epitope reported previously (pEQPPFSEQEQPVLPQ-NH2), γ-gliadin W11-E7 (pEQAFPQPEQTFPHQP-NH2), LMW glutenin W15-E8 (pEGLERPWQEQPLPPQ-NH2), γ-gliadin W17-E6E9 (pEPFPQPEQPELPFPQ-NH2), and α-gliadin W09-E7 (pEPQPFLPELPYPQP-NH2).

Immunostimulatory sequences: Focused in α-gliadin but distributed among other prolamins

There has long been interest from the food industry in naturally nontoxic cereals and gluten proteins or in designing modified gluten proteins devoid of toxicity in CD. Furthermore, it would be ideal for food tests to be able to detect each class of toxic protein in CD. Physical maps were constructed for prolamins in wheat, rye, and barley; those that contained the most immunogenic sequences identified in the second-round screening libraries are shown in Fig. 5. The α-gliadins, historically the focus of efforts to map the gluten peptides that are toxic in CD, were unusual because they had a single, polymorphic region with potent T cell–stimulatory activity. Our findings are consistent with those of others (25): The α-gliadin 33–amino acid oligopeptide, currently ascribed a central role in HLA-DQ2–associated CD, is the longest of a variety of polymorphic T cell–stimulatory sequences; many sequences but not all have DQ2-α-I, some include DQ2-α-II, and a few such as the 33–amino acid oligopeptide also include DQ2-α-III in tandem repeats. TCCs specific for DQ2-α-I and DQ2-α-II rarely recognized sequences in γ- and ω-gliadins and did not recognize peptides from other prolamin families in wheat, barley, or rye. Most other strongly immunogenic γ-gliadin, ω-gliadin, secalin, and hordein polypeptides contained groups of overlapping T cell–stimulatory sequences within several discrete regions. T cell–stimulatory regions in polypeptides from α-gliadin, γ-gliadin, ω-gliadin, B-hordein, C-hordein, 75-kD secalin, and ω-secalin generally included sequences recognized by TCCs specific for DQ2-ω-I and DQ2-ω-II (Fig. 5). T cell–stimulatory sequences shown in Fig. 5 localize to regions rich in glutamine and proline and are predicted by the ExPASy peptide cleavage algorithm to be resistant to peptidases (table S6).

Fig. 5

The distribution of immunodominant sequences is restricted in α-gliadins but not in γ-gliadin, ω-gliadin, hordein, or secalin. Shaded boxes indicate the location and T cell stimulation score (vertical height) of peptide sequences derived from selected immunogenic gluten proteins in wheat, barley, and rye (dotted line indicates signal sequence). The box color indicates peptide recognition by TCCs specific for dominant wheat α-gliadin epitopes (DQ2-α-I or DQ2-α-II) (red), wheat ω-gliadin/barley hordein epitopes (DQ2-ω-I or DQ2-ω-II) (blue), B/C-hordein epitope (DQ2-Hor-I) (yellow), rye ω-secalin epitope (DQ2-Sec-I) (brown), and γ-gliadin epitope contained in W11-E7 (green). Colored hatched boxes indicate sequences recognized by TCCs specific for more than one of these peptides. Black/white hatched boxes indicate sequences not recognized by any of these clones. The number in brackets after GenBank accession numbers indicates the number of other polypeptides found in GenBank that share similar sequences and possess the same T cell–stimulatory peptides in the same relative positions.


Since Dicke first described gluten from wheat and gluten-related fractions from barley and rye as the toxic factor in CD (26, 27), the ultimate goal of many researchers’ efforts has been the definition of the toxic components of gluten. The complexity of gluten and the difficulty in developing disease-relevant, high-throughput bioassays have prevented comprehensive definition of gluten toxicity, now understood to be largely but not exclusively mediated by CD4+ T cells (28). More than 100 gluten peptides have been predicted in silico to be epitopes on the basis of their expected resistance to proteolysis, susceptibility to deamidation by tTG, and conforming to the HLA-DQ2–binding motif (11).

We have now comprehensively characterized the T cell response to all toxic prolamins in adults with CD by using peripheral blood collected from donors (who were normally on strict long-term, gluten-free diet) after oral challenge with grain to screen all potential T cell–stimulatory 12–amino acid sequences from wheat gliadins and glutenins, barley hordeins, and rye secalins. Our algorithmic approach to designing comprehensive peptide libraries of a size that is manageable in an optimized high-throughput IFN-γ ELISpot assay was a critical technical advance (23). However, progress in sequencing prolamin genes has meant that the number of proteins initially screened was less than the number included in GenBank by the end of the study. Ultimately, we were able to use TCCs and distinct epitope hierarchies after wheat, barley, and rye challenge to define a discrete set of immunodominant, nonredundant peptides consistently capable of recapitulating much of the T cell–stimulatory capacity of gluten.

Three key observations from our study reshape our understanding of gluten toxicity in CD. First, the contribution made by T cells specific for many nonredundant, immunodominant gluten peptides, including α-gliadin peptides encompassing DQ2-α-I and DQ2-α-II, the C-hordein peptide encompassing DQ2-Hor-I, and the ω-secalin peptide including DQ2-Sec-I, is critically determined by the grains consumed by a CD donor. Therefore, mapping epitopes for T cells raised against prolamins from a single cereal will provide an incomplete assessment of relevant T cell–stimulatory peptides.

Second, many gluten peptides stimulate T cells. We confirmed 96 T cell–stimulatory peptides and found that 38 distinct peptides were dominant for at least one donor, but the T cell population activated by a single dominant peptide is capable of recognizing and responding to a large number of related gluten sequences. In vitro, we showed that TCCs specific for four dominant peptides recognized 74% of all T cell–stimulatory peptides and 68% of dominant peptides. The diversity of sequences cross-reactive with the most important T cell–stimulatory peptides results partly from having epitopes overlapping within a single 11–amino acid sequence and also because some T cells, especially those specific for the ω-gliadin/C-hordein–derived sequence PFPQPEQPFPW, are highly redundant in their recognition of gluten peptides. Although degenerate recognition by a single T cell receptor (TCR) could explain the diversity of responses from T cells specific for a single dominant peptide, the close sequence homology of reactive peptides indicates that epitope cross-reactivity is a more likely explanation.

Third, although α-gliadin–derived peptides encompassing DQ2-α-I/III and DQ2-α-II are immunodominant in wheat gluten after wheat challenge, our comprehensive screen of prolamins from wheat, barley, and rye indicates that the ω-gliadin/C-hordein–derived peptide encompassing DQ2-ω-I and DQ2-ω-II is recognized by most T cells induced by oral challenge with a mixture of wheat, barley, and rye. Hence, it is an ω-gliadin/C-hordein–derived peptide that might be considered the canonical dominant T cell–stimulatory peptide in HLA-DQ2–associated CD. The immunogenicity of DQ2-ω-I, in particular, may relate to its overlapping and repeated presence throughout ω-gliadin and through corresponding homologous sequences in barley hordein and rye secalin (29), plus its unusual conformationally mobile secondary structure (30, 31).

Although the ω-gliadin fraction of wheat gluten is toxic in CD (32), it is not surprising that the importance of DQ2-ω-I and DQ2-ω-II in CD has been overlooked. There has been a historical focus on wheat gluten and α-gliadins in CD immunotoxicity studies. The A-gliadin component of α-gliadin was the first fully sequenced wheat prolamin (33), and when we commenced our study, there was just one complete wheat ω-gliadin sequence in GenBank compared to 60 for α-gliadins. Previous studies have also used TCLs raised against wheat gluten and gliadin rather than against barley hordein or rye secalin. It might be speculated that TCLs against hordein or secalin would have led to the DQ2-α-I or DQ2-α-II epitopes being overlooked, whereas the importance of DQ2-ω-I and DQ2-ω-II, and also DQ2-Hor-I and DQ2-Sec-I epitopes would have been appreciated earlier. Indeed, the historical emphasis on wheat gluten has arisen because wheat is the more frequently consumed cereal in most Western diets today, yet CD is as common in Scandinavia where relatively larger amounts of rye and barley are consumed.

Our findings with fresh polyclonal T cells from blood after individual GCs and isolated clones support both DQ2-ω-I and DQ2-α-I, and DQ2-ω-II and DQ2-α-II, as four distinct epitopes despite their close structural similarities. This further supports our recent findings that some intestinal TCLs recognize the ω-gliadin sequence encompassing DQ2-ω-I and DQ2-ω-II but not DQ2-α-I or DQ2-α-II (25). These observations contrast with those of Vader et al. (14), who considered DQ2-ω-I a functional homolog of DQ2-α-I (named Horα9/Secα9) and, similarly, that the sequence PQPEPFPQ (named Horα2/Secα2), closely related to DQ2-ω-II, was a functional homolog of DQ2-α-II. Vader et al. attributed the toxicity of barley hordeins and rye secalins to the presence of Horα9/Secα9 and Horα2/Secα2 that were recognized by T cells specific for DQ2-α-I and DQ2-α-II; our findings provide an alternate explanation for the toxicity of rye and barley.

Our findings generally support established principles of epitope selection in HLA-DQ2–associated CD. For instance, the importance of protease resistance in shaping the repertoire of immunogenic gluten peptides was supported by our data, which showed that cleavage sites for trypsin, chymotrypsin, and pepsin are largely restricted to immunologically silent regions of the prolamin polypeptides. However, the explanation for the dominance of just a few peptides rather than the many others that share similar chemical properties that predict epitope selection is unclear. More detailed studies will be needed to measure HLA-DQ2 binding, susceptibility to deamidation by tTG, and protease resistance.

The epitopes DQ2-α-I/III, DQ2-α-II, DQ2-γ-IV, DQ2-γ-VI, and DQ2-γ-VII reported for TCCs were confirmed with PBMCs in the IFN-γ ELISpot, but many others were not. This should not be interpreted as meaning that previously defined epitopes such as DQ2-γ-I are not recognized by relevant T cells, but simply that such T cells are not sufficiently frequent in blood or consistently recognized by CD donors to reach the threshold applied in this study (~7% of the response elicited after wheat challenge by peptides encompassing DQ2-α-I/III and DQ2-α-II). Furthermore, we assigned a hierarchy to peptides according to their recognition by polyclonal T cells; it was not practical to distinguish between peptides containing two or more epitopes and those with a single epitope; hence, dominance was assigned to peptides rather than epitopes.

This study supports a central role in the immunopathogenesis of CD for T cells specific for two overlapping epitopes found in ω-gliadin. TCCs raised to these epitopes recognize sequences derived from virtually all prolamin fractions in wheat, barley, and rye (we were unable to assess glutenins for cross-reactivity). In our hands, gluten peptide recognition by clones specific for dominant T cell–stimulatory sequences derived from α-gliadin, C-hordein, or ω-secalin was far less cross-reactive than clones specific for ω-gliadin. It will also be important to determine whether deamidation of the ω-gliadin peptide encompassing DQ2-ω-I and DQ2-ω-II epitopes enhances both the affinity of binding to HLA-DQ2 and the diversity of cognate TCRs, suggested as a pivotal event in the genesis of gluten immunity in CD (34).

Repeating this study in children would be impractical, because the volume of blood required to screen all potential gluten epitopes is too large, but it will be possible to test whether peptides immunodominant in adults also make a substantial contribution to the gluten-specific T cell response in childhood CD. Such studies in children would require them to follow a strict gluten-free diet and then undergo short-term GC. Whether the specificity of T cell responses to gluten begins focused and then diversifies, or vice versa, is unlikely to be revealed with oral GC in children with established CD because immune responses evolve over weeks rather than years. However, we have noted that the specificity of gluten peptide-specific serum immunoglobulin G (IgG) is identical in adults with active CD and in children immediately after tTG IgA seroconversion before they have diagnosed CD. Equally, our data are derived from volunteers with diagnosed CD, rather than from individuals who might be seropositive for tTG IgA and have unrecognized or asymptomatic CD; future studies will need to address each of these clinically distinct disease phenotypes to confirm the consistency of gluten-specific T cell responses.

Detailed understanding of the T cell–stimulatory sequences in CD will facilitate food tests, design of functional foods to reduce the toxicity of gluten, diagnostics to detect CD-specific T cells, and therapeutics to reduce exposure of T cells to toxic peptides. Because whole gluten is a complex mixture of aqueous insoluble proteins that stimulate both innate and acquired immunity (35), it is unattractive for use in the protein-based desensitization therapy that is effective in allergic diseases (36). Thus, the practical application that motivated us to undertake this study was the possibility of a peptide-based immunotherapy.

Peptide-based immunotherapy with altered peptide ligands (APLs) has shown encouraging therapeutic results in preclinical models with clonal T cell populations (3739), but in clinical trials for multiple sclerosis, it has proven ineffective and occasionally associated with a shift in T cell phenotype from TH1 to TH2 (40, 41). We previously explored the possibility of APLs to antagonize TH1 responses of polyclonal peripheral blood T cells to the 17–amino acid α-gliadin oligopeptide encompassing DQ2-α-I and DQ2-α-II. However, this approach was only modestly effective ex vivo (42). More recently, antigen-specific immunotherapy with escalating microgram doses of peptides predicted to be T cell epitopes (for example, from allergens such as cat dander protein Fel-d-1 implicated in asthma), administered by repeated intradermal injection, has shown efficacy in phase II clinical trials (43). It appears that peptide-based immunotherapy of this type depends on peptide uptake and presentation by immature dendritic cells to promote antigen-specific regulatory T cells and spreading of tolerance (44, 45).

Previously, the diversity and lack of consistency of T cell–stimulatory peptides in gluten compromised the design of a peptide-based immunotherapy for CD. Our study now allows the design of a potential immunotherapy with peptides confirmed as immunodominant in a common human immune disease. The lead compound consists of three immunogenic gluten peptides, which are now in phase I clinical development. A critical step toward a peptide-based immunotherapy for CD will be to show that such a compound is bioactive and targets relevant T cells when administered to volunteers with HLA-DQ2–associated CD. Provided these fundamental immunological properties and safety are established, such a compound promises to provide unique insights into the therapeutic potential of peptides confirmed to be disease-specific T cell agonists.

Materials and Methods

Subjects and controls

The study was approved by the Oxfordshire Regional Ethics Committee and by the Melbourne Health Human Research Ethics Committee. Together, 226 oral GCs (wheat, 113; barley, 41; rye, 43; and combined, 29) were undertaken in CD volunteers [median age, 50 years; range, 19 to 70 years; female, 165 (73%); HLA-DQB1*02 homozygous, 68 (30%)] and 10 wheat challenges in healthy volunteers [median age, 47 years; range, 22 to 64 years; female, 5 (50%); HLA-DQB1*02 homozygous, 3 (30%)]. Nineteen CD volunteers were recruited in Oxford (UK) from the Coeliac Clinic at John Radcliffe Hospital, and the remainder were recruited in Melbourne (Australia) by advertisement in the Victorian State Coeliac Society newsletter. Leukocyte-derived DNA from volunteers was genotyped with a panel of sequence-specific primers to determine HLA-DQA and HLA-DQB alleles (Victorian Transplantation Immunology Service, Parkville, Victoria, Australia) (4648). CD and healthy volunteers were included if they had HLA-DQA1*05 and HLA-DQB1*02 (encoding HLA-DQ2) but did not have either DQA1*03 or DQB1*0302 (alleles encoding HLA-DQ8). CD volunteers had biopsy-proven CD conforming to the European Society of Paediatric Gastroenterology and Nutrition diagnostic criteria (49). Healthy volunteers following a normal gluten-containing diet, and also CD volunteers following a strict gluten-free diet, were confirmed to have normal levels of serum transglutaminase IgA (INOVA Diagnostics). At the time GC was undertaken, CD volunteers had been strictly gluten-free for 3 months, and healthy volunteers for 4 weeks.

Grain challenge

Oral challenges extended over 3 days unless otherwise stated. Wheat challenge consisted of two 50-g slices of gluten-containing wheat bread (Oxford: Sainsbury’s standard white bread; Melbourne: Baker’s Delight white bread block loaf cut to toasting size thickness) for breakfast, followed by two slices for lunch. Barley challenge consisted of pearl barley (Ward McKenzie) cooked into a risotto (150 g dry weight daily). Rye challenge involved rye (Oxford: manually sorted rye cultivar Motto and milled on an experimental mill at Long Ashton Research Station; Melbourne: Biodynamic Rye flour, Eden Valley Biodynamic Farm) baked into muffins (100 g dry weight rye flour daily). Participants undertaking the combined cereal challenge consumed two muffins consisting of 25 g of wheat flour (White Wings, Goodman Fielder Australia), 22 g of barley flour (Four Leaf Milling), and 22 g of rye flour (Four Leaf Milling) each day. Blood (50 to 300 ml) was drawn in the morning before (day 0) and/or 6 days after commencing GC (day 6), with the total volume collected on both days not exceeding 300 ml. Participants were allocated randomly to each challenge.


Synthetic peptides (screening grade and >70% purity) were purchased from Research Genetics, Mimotopes, or Pepscan. Unless otherwise stated, peptide libraries were assessed at a concentration of 25 μg/ml. Hordein and secalin fractions were prepared from rye and barley grown in isolation from other grains and hand-milled flour and fractionated according to published methods (50). Deamidation with guinea pig liver tTG (Sigma) was as described previously (21, 22). Gliadin (Sigma) and other prolamins were incubated for 4 hours at 37°C in 10-fold excess with chymotrypsin (Sigma) in ammonium bicarbonate (pH 8) and finally boiled for 15 min. Prolamin protein concentrations were determined by bicinchoninic acid method (Pierce).

Peptide library design

Wheat gliadin sequences present in GenBank in 2001 were aligned by Lasergene MegAlign software (DNAStar) to allow visual selection of 20–amino acid oligopeptides encompassing unique 12–amino acid oligopeptides (Pilot library). Subsequently, wheat, barley, and rye gluten peptide libraries were designed with a customized algorithm, as previously described (23), to entries for gliadins, glutenins, hordeins, and secalins in National Center for Biotechnology Information GenBank in their genome-encoded (wild-type) sequence (comprehensive library) or both wild-type and in silico transglutaminase-deamidated sequence (verification library) according to defined deamidation motifs (20, 51) (table S1).

Second-round wheat, rye, and barley libraries were designed by reducing selected 20–amino acid oligopeptides to nine overlapping 12–amino acid oligopeptides. If any 12–amino acid sequence incorporated glutamine at position 7 and it conformed to the deamidation motif defined for tTG (QX1PX3, or QX1X2[F,Y,W,I,L,V], where X1 and X3 are not proline) (20, 51), then a 16–amino acid oligopeptide was designed whereby the 12–amino acid sequence with glutamine at position 7 was flanked by the native residues at positions −1 and 13 and by glycine at positions −2 and 14. This strategy allowed the central, potentially deamidated glutamine residue to be accommodated at anchor positions 4, 6, or 7 in any potential 9–amino acid sequence HLA-DQ2 peptide-binding sequence, consistent with the HLA-DQ2–binding motif (52, 53). If selected 20–amino acid oligopeptides did not include any 12–amino acid sequences with glutamine at position 7, then two 16–amino acid oligopeptides overlapping 12 residues were synthesized. Some second-round 16–amino acid oligopeptides with a central glutamine residue susceptible to tTG-mediated deamidation were also synthesized with glutamine replaced by glutamate (in silico deamidation).

ELISpot assay

PBMCs were isolated from heparinized whole blood with Ficoll-Paque Plus in 50-ml Leucosep tubes (Greiner Labortechnik). After being washed three times, PBMCs were resuspended in complete RPMI containing 10% heat-inactivated human AB serum. Overnight IFN-γ ELISpot assays (Mabtech) using 96-well plates (MSIP-S45-10; Millipore) were performed by a modification to the manufacturer’s instructions (21, 22). Peptide libraries were assessed with a single well per peptide. Tetanus toxoid (CSL) (10 light-forming units per milliliter) and phytohemagglutinin (5 μg/ml) were used as positive control antigens. SFUs in individual wells were counted with an automated ELISPOT reader (AID ELISPOT Reader System, AID Autoimmun Diagnostika GmbH). Our data show that responses in the IFN-γ ELISpot assay with the 17–amino acid α-gliadin oligopeptide encompassing DQ2-α-I and DQ2-α-II is qualitatively reproducible with PBMCs in the same donor collected on days 6 and 7 after commencing oral wheat GC, and also when the same donor is challenged on two occasions between 6 and 12 months apart (22). Here, PBMCs of six donors challenged with wheat between 6 and 12 months apart to screen the first-round comprehensive library and subsequent second-round library showed that the 20–amino acid oligopeptides and then 16–amino acid oligopeptides encompassing DQ2-α-I/II/III or DQ2-ω-I/II were consistently immunodominant after each challenge in five of six subjects, whereas peptides encompassing the sequence W033 remained subdominant (table S7).

Isolation and characterization of TCCs

PBMCs were isolated as for the ELISpot assay. LPMCs were isolated by treating small intestinal tissue biopsies with 1 mM dithiothreitol in phosphate-buffered saline (PBS), followed by two incubations at 37°C for 30 min in Dispase II (2.4 U/ml) (Roche). Biopsies were minced and incubated at 37°C for 1 hour in Liberase Blendzyme 3 (2 U/ml) (Roche) and RPMI, washed three times in PBS, and mixed with 1.5 × 106 to 3 × 106 autologous PBMCs irradiated at 20 gray (Gy). Cells were stained with 0.1 μM carboxyfluorescein diacetate succinimidyl ester and plated out at 2 × 105 per well in 96-well plates. Peptide and protein antigens were used at 32 and 100 μg/ml, respectively. Antigen-specific CD4+ TCCs were isolated as previously described (54) and expanded with anti-CD3 as described (55). Antigen specificity was determined by [3H]thymidine proliferation assay or IFN-γ ELISpot with irradiated PBMCs (20 Gy) from HLA-DQ2+ HLA-DQ8 donors.

Expanded antigen-specific clones were tested for clonality with the IOTest Beta Mark (Beckman Coulter). Negative clones were confirmed as clonal by polymerase chain reaction of the TCR Vβ chains. HLA restriction was determined with HLA-DR (10 μg/ml, clone L243) and HLA-DQ (10 μg/ml, clone SPVL3) antibodies. Secretion of IFN-γ, interleukin-4 (IL-4), IL-5, IL-10, IL-13, and IL-17 by clones to cognate antigen was determined in ELISpot assays. Lysine scans of peptides W02-E7, W03-E7, and B08-E2E7 were carried out in IFN-γ ELISpot or proliferation assays with clones specific for these peptides.

Statistical analysis

IFN-γ ELISpot responses were considered significant and, unless otherwise stated, included for analysis when SFUs per well were greater than four times the response to medium alone and >10 SFUs per well. Overall, PBMCs from 91 of 113 (81%) donors undergoing wheat challenge, 30 of 41 (73%) undergoing barley challenge, 32 of 43 (74%) undergoing rye challenge, and 13 of 29 (45%) undergoing combined challenge yielded significant IFN-γ ELISpot responses to at least one peptide. Interim analysis of all donor ELISpot responses to first-round libraries (irrespective of donor’s maximal peptide responses) used a customized expectation maximization algorithm described earlier and also a simplified analysis to identify all 20–amino acid oligopeptides eliciting a response equivalent to at least 5% of the most active peptide in each individual donor and in the group as a whole to enable design of second-round libraries (23). A final simplified assessment used to rank reactive peptides used a score between 0 and 100 equal to the mean normalized response of donors who responded to at least one peptide.

Supplementary Material

Fig. S1. Fine mapping the T cell–stimulatory sequence in tTG-treated ω-gliadin AAG17702(81–102) PQQPQQPQQPFPQPQQPFPWQP by IFN-γ ELISpot with PBMCs from HLA-DQ2+ CD donors after 3-day oral wheat challenge.

Table S1. Gluten peptide libraries.

Table S2. IFN-γ ELISpot responses to peptides before and after 3-day wheat gluten challenge.

Table S3. IFN-γ ELISpot SFU/106 PBMC healthy HLA-DQ2 individuals on gluten-free diet 4 weeks before and after 3-day wheat gluten challenge.

Table S4. Gliadin peptide hierarchy after 3-day wheat or rye challenge in English celiac donors.

Table S5. TCC characterization.

Table S6. Cleavage sites localize to nonstimulatory regions of prolamins.

Table S7. IFN-γ ELISpot responses in six celiac donors after consecutive wheat challenges.


  • * These authors contributed equally to this work.

  • Present address: Department of Medical Statistics, University Medicine Goettingen, 37099 Goettingen, Germany.

  • Present address: St. Vincent’s Institute of Medical Research, 41 Victoria Parade, Fitzroy, Victoria 3065, Australia.

  • Citation: J. A. Tye-Din, J. A. Stewart, J. A. Dromey, T. Beissbarth, D. A. van Heel, A. Tatham, K. Henderson, S. I. Mannering, C. Gianfrani, D. P. Jewell, A. V. S. Hill, J. McCluskey, J. Rossjohn, R. P. Anderson, Comprehensive, quantitative mapping of T cell epitopes in gluten in celiac disease.Sci. Transl. Med. 2, 41ra51 (2010).

References and Notes

  1. Acknowledgments: We thank all the volunteers who participated in this study; C. Pizzey and the Coeliac Society of Victoria for the recruitment of volunteers; G. Tanner (Plant Science Division, Commonwealth Scientific and Research Organization, Black Mountain, Canberra, Australia) for the gift of purified hordein; the Victorian Transplantation and Immunogenetics Service for expert HLA typing of patients; and M. Stewart for the technical assistance in performing ELISpot assays. Funding: J.A.T.-D. was supported by a National Health and Medical Research Council (NHMRC) Postgraduate Medical Scholarship and by a grant from the Australian and New Zealand Coeliac Research Fund; T.B. was supported by grant PBF-S19T10 of the German National Genome Research Network by the Federal Ministry of Education and Research; D.A.v.H. was funded by a Wellcome Trust Clinician Scientist Fellowship (GR068094MA); S.I.M. is supported by the Juvenile Diabetes Research Foundation (10-2006-261); J.R. is supported by an Australian Research Council Federation Fellowship; and R.P.A. holds the Ian Mackay Fellowship from the Walter and Eliza Hall Institute and Melbourne Health and also the Lions Cancer Council Fellowship. This work was supported by NHMRC Project grant 406656, Coeliac UK Project grant, the Graham Bird Memorial Fund (Oxford), the Oxford University College Challenge Seed Fund, BTG International plc, Nexpep Pty Ltd., the NHMRC Independent Research Institutes Infrastructure Support Scheme grant 361646, and Victorian State Government Operational Infrastructure Support. Author contributions: J.A.T.-D., J.A.S., and J.A.D.: study design, data collection, analysis, and manuscript preparation; T.B.: study design and analysis; D.A.v.H.: data collection; A.T.: study design and data collection; K.H., S.I.M., and C.G.: data collection and analysis; D.P.J. and A.V.S.H.: study design; J.M. and J.R.: study design and manuscript preparation; R.P.A.: study design, data collection, analysis, and manuscript preparation. Competing interests: J.A.T.-D., J.A.S., J.A.D., T.B., D.A.v.H., D.P.J., A.V.S.H., and R.P.A. are co-inventors of patents pertaining to the use of gluten peptides in therapeutics, diagnostics, and nontoxic gluten. R.P.A. is a substantial shareholder and director of Nexpep Pty Ltd. and Nexgrain Pty Ltd., companies developing peptide-based therapeutics and diagnostics and nontoxic gluten suitable for CD. R.P.A. is also the chief executive of Nexpep Pty Ltd. J.A.T.-D. is a shareholder of Nexpep Pty Ltd. and Nexgrain Pty Ltd. and a consultant to Nexpep Pty Ltd. J.A.S. and J.A.D. were former consultants to Nexpep Pty Ltd. D.A.v.H. and J.M. have consulted for and are shareholders in Nexpep Pty Ltd. Accession number: Clinical trial /NCT00879749.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article