Igs as Substrates for Transglutaminase 2: Implications for Autoantibody Production in Celiac Disease

Abstract

Autoantibodies specific for the enzyme transglutaminase 2 (TG2) are a hallmark of the gluten-sensitive enteropathy celiac disease. Production of the Abs is strictly dependent on exposure to dietary gluten proteins, thus raising the question how a foreign Ag (gluten) can induce an autoimmune response. It has been suggested that TG2-reactive B cells are activated by gluten-reactive T cells following receptor-mediated uptake of TG2–gluten complexes. In this study, we propose a revised model that is based on the ability of the BCR to serve as a substrate to TG2 and become cross-linked to gluten-derived peptides. We show that TG2-specific IgD molecules are preferred in the reaction and that binding of TG2 via a common epitope targeted by cells using the IgH variable gene segment (IGHV)5–51 results in more efficient cross-linking. Based on these findings we hypothesize that IgD-expressing B cells using IGHV5–51 are preferentially activated, and we suggest that this property can explain the previously reported low number of somatic mutations as well as the overrepresentation of IGHV5–51 among TG2-specific plasma cells in the celiac lesion. The model also couples gluten peptide uptake by TG2-reactive B cells directly to peptide deamidation, which is necessary for the activation of gluten-reactive T cells. It thereby provides a link between gluten deamidation, T cell activation, and the production of TG2-specific Abs. These are all key events in the development of celiac disease, and by connecting them the model may explain why the same enzyme that catalyzes gluten deamidation is also an autoantigen, something that is hardly coincidental.

Introduction

Celiac disease is an inflammatory disorder of the small intestine caused by a harmful immune response to dietary cereal gluten proteins in genetically susceptible individuals (1). Key players in the immune reaction leading to pathogenic destruction of the intestinal epithelium are CD4⁺ T cells that react specifically with gluten-derived peptides when bound to the predisposing MHC class II molecules HLA-DQ2 (particularly the DQ2.5 variant) and HLA-DQ8. It has been demonstrated that the T cell response generally does not target gluten peptides in their native form but rather their deamidated counterparts in which certain glutamine residues have been converted to glutamic acid, leading to improved binding to the disease-associated HLA molecules (2, 3). The deamidation reaction is mediated by the enzyme transglutaminase 2 (TG2), which is present abundantly in the extracellular matrix beneath the intestinal epithelium (2). Deamidation is one of two Ca²⁺-dependent reactions catalyzed by this enzyme. The other one, termed transamidation, is the covalent cross-linking of two polypeptides through the formation of an isopeptide bond between the side chain carbonyl of a target glutamine and the amino group of a lysine residue. Alternatively, a small-molecule amine can substitute for the lysine (4).

In addition to catalyzing gluten peptide deamidation, TG2 is involved in celiac disease as an autoantigen (5). Production of TG2-specific serum Abs depends on a gluten-containing diet as well as HLA type, as the Abs disappear from the circulation within months after commencement of a gluten-free diet (6, 7) and are only found in individuals who express HLA-DQ2 or HLA-DQ8 (8). TG2-specific serum Abs have proven to be very sensitive and specific markers and are widely used in diagnostic tests (9). IgA Abs are primarily monitored for this purpose, but TG2-specific IgM and IgG Abs are also produced in patients (9–11). Recently, we have shown that, on average, 10% of IgA plasma cells in the small intestinal mucosa of celiac disease patients produce TG2-specific Abs (11). Cloning of the V regions of individual TG2-specific cells revealed a repertoire that had surprisingly few somatic mutations and appeared restricted in the use of V region gene segments. The cells chiefly employed κ L chains, and the H chain repertoire was skewed toward usage of the IgH variable gene segment (IGHV)5–51. The Abs were recombinantly expressed as human IgG1, making up a panel of TG2-specific mAbs. Interestingly, none of the 47 mAbs tested had an inhibitory effect on TG2-mediated deamidation and transamidation, suggesting selective activation of TG2-reactive B cells that bind the enzyme in a way that preserves its catalytic activity (11).

It has been hypothesized that the gluten dependence of the Ab response can be explained by a hapten–carrier-like mechanism (12, 13), in which gluten peptides can serve as a carrier when bound in a complex with TG2, thereby allowing TG2-specific B cells to receive activation help from gluten-specific, HLA-DQ–restricted CD4⁺ T cells (13). Such TG2–gluten complexes could be taken up by B cells either in the form of an enzyme-substrate intermediate or through TG2-mediated cross-linking of gluten peptides to the enzyme itself (14, 15). The model neatly explains the gluten and HLA dependence of anti-TG2 autoantibody production, but it may not explain some of the newly identified features of the response. In this study, we propose an alternative model to explain the gluten-dependent production of TG2-specific Abs in celiac disease based on the ability of TG2-specific Ig molecules to serve as substrates to TG2 and become cross-linked to gluten-derived peptides. As B cell activation in this model depends on active BCR-bound TG2, the inability of TG2-specific Abs to inhibit TG2 enzymatic activity is accounted for. The model also considers the low mutation levels and IGHV5–51 preference among TG2-specific plasma cells, two key observations that were recently confirmed using a high-throughput sequencing approach (16). At the same time, the new model directly couples gluten peptide uptake by B cells to deamidation and presentation to T cells and places TG2-specific B cells in the center of events that drive the pathogenesis.

Materials and Methods

Recombinant proteins

Human TG2 containing an N-terminal His-tag was expressed in Escherichia coli and purified by nickel affinity chromatography as previously described (17). According to the standard protocol, purified TG2 was dialyzed against buffer containing 1 mM DTT before storage. In one set of experiments the effect of the reducing agent was tested using TG2 depleted from DTT by size-exclusion chromatography. A TG2 variant with a BirA biotinylation sequence introduced after the His-tag (BirA-TG2) was obtained by PCR amplification followed by ligation into the BglII site of the baculovirus transfer vector pAcAB3 (BD Pharmingen). The protein was produced in Sf9 insect cells as earlier described for soluble MHC class II molecules (18) except that BirA-TG2 was extracted from the cell pellet and purified on Ni-NTA resin (Novagen). Site-specific biotinylation with the BirA biotin-protein ligase (Avidity) was carried out in vitro according to the instructions from the manufacturer. mAbs were expressed in HEK293 cells as previously described (11, 19). The mAbs were originally generated as IgG1 molecules (11). DNA encoding additional human H chain isotypes was obtained as synthetic genes (GenScript) and used to exchange the sequence of the IgG1 H chain. Ab sequence modifications, including introduction of mutations and exchange of Ab V regions, were done by PCR SOEing followed by insertion of the fragments between the AgeI and HindIII sites in the expression vector. The modified constructs were controlled by sequencing (GATC Biotech), and preserved reactivity of mutated Abs was confirmed in a TG2-specific ELISA.

TG2-mediated cross-linking of Abs

In a typical cross-linking experiment, 0.1 mg/ml recombinant human TG2 was preincubated with 0.2 mg/ml Ab in Tris buffer (50 mM Tris-HCl, pH 7.4) for 30 min at 37°C followed by addition of an equal volume of buffer supplemented with 10 mM CaCl₂ and various concentrations (0–250 μM) of FITC-labeled DQ2.5-glia-α2(EQ) peptide (FITC-Ahx-PQPELPYPQPQLPY, obtained from GL Biochem). The reaction was stopped at the indicated time points by addition of 4× Laemmli sample buffer. The samples were heated in the presence of 2-ME, and the proteins were separated by SDS-PAGE. Fluorescent protein bands were subsequently detected using a Kodak Image Station 4000MM Pro instrument, and the amount of Ab–peptide cross-links was determined by correlating the band intensities to the intensity obtained with a sample of BSA that had been prelabeled with FITC (Sigma-Aldrich) and contained a known concentration of protein-bound fluorescein. The amount of cross-links generated within the first 60 s after addition of CaCl₂ and peptide was used as a measure of the initial reaction rate. Ab incorporated into high molecular mass complexes was detected by Western blotting using the following pooled Abs specific to human H and L chains: rabbit anti-human IgA (Dako), rabbit anti-human IgD (Dako), rabbit anti-human IgG (Dako), rabbit anti-human κ L chains (Dako), goat anti-human Ig (SouthernBiotech), and goat anti-human IgE (Sigma-Aldrich). HRP-conjugated goat anti-rabbit IgG (SouthernBiotech) and rabbit anti-goat IgG (SouthernBiotech) were used consecutively as secondary Abs. To study TG2-mediated cross-linking of Abs attached to beads, IgD engineered to contain a BirA target sequence in the C-terminal end of the H chain was biotinylated and preincubated with streptavidin-coated Dynabeads (Invitrogen) for 30 min at room temperature. The streptavidin-bound IgD (1 μg preincubated with 10 μl beads) was incubated with 0.5 μg TG2 in 50 μl Tris buffer for 30 min at 37°C before the reaction was started by addition of CaCl₂ and peptide as described above. The samples were analyzed after removing the supernatant and eluting the proteins from the beads with 10 μl sample buffer.

Identification of target lysine residues

To identify Ab lysine residues targeted by TG2, cross-linking was carried out as described above for 30 min in the presence of 1 mM of a biotinylated, glutamine-containing peptide (biotin-QLPR, produced by solid-phase peptide synthesis employing Fmoc chemistry [MultiSynTech]). After SDS-PAGE, the Coomassie-stained H chain bands were excised and in-gel digested with trypsin or chymotrypsin overnight at 37°C. The proteolytic fragments were extracted from the gel, and biotinylated peptides were isolated with streptavidin-coated Dynabeads as previously described (20). The cross-linked fragments were detected and identified by MALDI-TOF and MALDI-TOF/TOF mass spectrometry using an Ultraflex II instrument (Bruker Daltonics).

TG2-mediated cleavage of IgD–peptide cross-links

Cross-links between Ab and FITC-labeled peptide were generated as described above using 5 μg IgD and 5 μg biotinylated BirA-TG2. The cross-linking reaction was stopped after 5 min by adding EDTA to 10 mM. The reaction mixture was then added to 30 μl streptavidin-agarose (Sigma-Aldrich) in 50 μl Tris buffer and incubated 30 min at 37°C. The supernatant containing the free peptide was subsequently collected and analyzed by capillary electrophoresis (Agilent Technologies) as previously described (11, 21) using 80 mM sodium borate (pH 9.3) as running buffer and laser-induced fluorescence detection (Picometrics). The agarose beads were washed three times with TBST and four times with water before the bound Ab was eluted by incubating the beads with 20 mM HCl for 30 min. The pH of the eluate was raised to either 5 or 7.4 with 0.2 M Tris buffer. Half of the eluate was then analyzed directly by capillary eletrophoresis, whereas the other half was incubated with 1 μg TG2 and 5 mM CaCl₂ for 30 min at 37°C prior to analysis.

Generation of retrovirally transduced B and T cell lines

Murine A20 B lymphoma cells expressing HLA-DQ2.5 alone or in combination with a TG2-specific or nonspecific BCR of either the IgD or the IgA1 isotype were generated as previously described (11). In brief, codon-optimized synthetic DNA (GenScript) encoding the α-chain (DQA1*05:01) and β-chain (DQB1*02:01) of the HLA-DQ2.5 protein linked with a P2A peptide sequence was cloned into the pMIG-II-eGFP retroviral plasmid (a gift from Dr. Dario Vignali, Pittsburgh, PA) (22) between the BglII and XhoI sites. The internal ribosome entry site and fluorescence tag were removed by digestion with XhoI and SalI followed by religation. For expression of BCRs, DNA (GenScript) encoding κ L chain and transmembrane IgH linked by a furin cleavage site (RRRR) and a T2A peptide sequence was cloned into the pMIG-II plasmid. After retroviral transduction, A20 cells were sorted based on positive staining with mAbs specific to HLA-DQ2.5 (clone 2.12.E11 conjugated to FITC [Sigma-Aldrich] prior to use), human IgD (PerCP-Cy5.5–conjugated IA6-2 [BD Biosciences]), or human IgA (biotinylated B3506B4 [SouthernBiotech] followed by PerCP-Cy5.5–conjugated streptavidin [BD Biosciences]) using a FACSAria flow cytometric cell sorter (BD Biosciences). For flow cytometric analysis, 2 × 10⁵ A20 cells were stained for 30 min in 2% (v/v) FCS/PBS. Binding to TG2 was assessed using TG2 multimers prepared by preincubation of biotinylated BirA-TG2 (final concentration 5 μg/ml) with allophycocyanin-conjugated streptavidin (Prozyme) at a 6:1 molar ratio in 2% FCS/PBS for 3 h. The murine T cell hybridoma line BW58α⁻β⁻ devoid of endogenous TCR was engineered to express human CD4 and the TCR derived from a DQ2.5-glia-α2–reactive CD4⁺ T cell clone (TCC364.1.0.14) as previously described (23).

In vitro T cell activation assay

T cell activation was used as a measure of T–B cell collaboration and was assessed by measuring release of IL-2 after culturing of transduced A20 B cells with transduced hybridoma T cells. Transduced A20 cells were incubated at 10 million cells/ml either with a combination of 1 μg/ml TG2, 2 mM CaCl₂ and various concentrations of native 33-mer peptide containing three overlapping copies of the DQ2.5-glia-α2 epitope (LQLQPFPQPQLPYPQPQLPYPQPQLPYPQPQPF, where underlined residues denote glutamine residues targeted by TG2, obtained from GL Biochem), or with preformed TG2–33-mer complexes for 1 h at room temperature in RPMI 1640. The complexes were generated by incubating 0.6 mg/ml TG2 with 0.1 mM 33-mer peptide in TBS supplemented with 5 mM CaCl₂ for 20 min at room temperature. The reaction was stopped, and TG2 was irreversibly inactivated by adding iodoacetamide to a final concentration of 10 mM followed by incubation for 30 min. The complexes were isolated by size-exclusion chromatography, and the buffer was changed to RPMI 1640. After incubation with Ag, the A20 cells were washed with RPMI 1640 and resuspended at 2 million cells/ml in 5% (v/v) FCS/RPMI 1640 followed by incubation 3 h at 37°C. Alternatively, A20 cells that had not been exposed to Ag were mixed with a deamidated version of the 33-mer prior to incubation at 37°C without removal of the peptide. A total of 25,000 hybridoma T cells specific to the deamidated DQ2.5-glia-α2 epitope were added to 200,000 Ag-pulsed A20 cells or 50,000 A20 cells incubated in the presence of the deamidated peptide, and the cells were cultured together for 18 h at 37°C. Culture supernatants were then collected and assayed for murine IL-2 secretion by ELISA as previously described (23).

TG2-mediated cross-linking of BCRs

To study TG2-mediated cross-linking at the cell surface, ∼5 × 10⁷ transduced A20 cells were incubated with 3 μg TG2 in 1 ml RPMI 1640 supplemented with 5 mM CaCl₂ and 20 μM FITC-labeled peptide for 1 h on ice. The reaction was stopped by washing the cells in cold PBS, and the cells were lysed in 700 μl lysis buffer (1% [w/v] n-dodecyl-β-d-maltoside, 150 mM NaCl, 50 mM Tris-HCl [pH 8.0]) supplemented with cOmplete protease inhibitor mixture (Roche) for 1 h on ice. The lysate was centrifuged for 15 min at 4°C in a microcentrifuge, and the supernatant was either analyzed directly or incubated with 30 μl protein l-agarose (GenScript) for 1 h at 4°C to pull down the BCR molecules. Bound BCR was subsequently eluted from the agarose beads with 10 μl sample buffer and the samples were analyzed by SDS-PAGE followed by detection of fluorescence and Western blotting as described above.

Results

TG2-mediated cross-linking of Ab molecules is dependent on Ab isotype

We have previously seen that TG2-specific mAb can become cross-linked by TG2 when expressed with an IgD or, to a lower extent, IgM H chain C region (11). However, when the same V region was expressed in an IgA1 or IgG1 format, the Ab molecules did not become cross-linked. Bringing individual BCR molecules into close proximity within the plasma membrane is known to stimulate B cell activation (24). TG2-mediated BCR cross-linking could therefore induce selective activation of unmutated cells expressing IgD and IgM on the surface, whereas isotype-switched cells with accumulated mutations would not receive these activation signals. We therefore hypothesized that this mechanism can explain why TG2-specific plasma cells generally seem to have few mutations despite chronic exposure to the triggering Ags gluten and TG2 (11).

We wanted to extend the analysis and therefore engineered a model IGHV5–51 mAb to be expressed with the H chain C region of each of the nine different human Ab isotypes. Again, we saw that Ab molecules were efficiently incorporated into high molecular mass complexes as a result of TG2-mediated cross-linking when expressed as IgD, but only limited cross-linking could be detected with Abs of other isotypes (Fig. 1A). The high molecular mass complexes contained both IgD and TG2 (Fig. 1B). However, a distinct band containing only IgD and not TG2 could be visualized by Western blotting, suggesting that TG2 can generate IgD–IgD cross-links in addition to the likely formation of TG2–TG2 and TG2–IgD complexes.

FIGURE 1.

TG2-mediated cross-linking of different Ab isotypes. A single TG2-specific mAb V region (679-14-E06) was expressed with a κ L chain and the H chain C regions of the different human isotypes. Each Ab was then evaluated as a substrate to TG2. (A) Western blot showing detection of Ab H and L chains after incubation with active TG2 for 10 min at 37°C. (B) Loading of the IgD molecule before and after incubation with TG2 shows that IgD is incorporated into high molecular mass complexes as a result of TG2 activity (left panel). After detection of IgD using rabbit anti-human IgD, the membrane was stripped and TG2 was visualized using mouse anti-TG2 mAb CUB7402 (right panels). The band indicated with an arrow contains cross-linked IgD only, whereas the other high molecular mass bands contain both IgD and TG2. (C) Detection of fluorescent protein gel bands after incubation with active TG2 in the presence of 0.1 mM of a synthetic FITC-labeled gluten peptide (PQPELPYPQPQLPY, corresponding to the known T cell epitope DQ2.5-glia-α2(EQ) with a single TG2-targeted glutamine [underlined]) for 10 min at 37°C. TG2 itself becomes cross-linked to the peptide, but only some Ab isotypes give rise to fluorescent bands. Besides IgG1, IgG3, and IgD, it is possible that IgE serves as a TG2 substrate, but the IgE H chain band overlaps with the TG2 band. IgE, however, is not known to play a role in celiac disease and was disregarded in the further analysis. Some incorporation of the peptide into the κ L chain of the IgE and IgD molecules could also be detected. (D) Quantification of the amount of fluorescence associated with the H chain bands of IgG1, IgG3, and IgD after incubation with TG2 and 0.1 mM peptide for 5 min at 37°C in the presence or absence of 0.5 mM DTT. The values represent the means of three independent experiments. Error bars indicate SD.

To further investigate the ability of the different isotypes to serve as substrates to TG2, cross-linking to a synthetic gluten peptide containing a single target glutamine residue was assessed (Fig. 1C). TG2-mediated cross-linking of gluten-derived peptides to cell-surface BCR molecules could be a route for receptor-mediated uptake of gluten peptides by TG2-specific B cells. After incubating Ab, TG2, and a labeled gluten peptide, the peptide was cross-linked to TG2 itself in agreement with previous reports (14, 15, 25). Additionally, we could detect cross-links between the peptide and the Ab H chains of IgG1, IgG3, and IgD. In contrast to the TG2-mediated formation of cross-links between gluten peptide and the H chain of IgD, the corresponding cross-linking of IgG1 and IgG3 was found to depend on reducing agent being present in the reaction buffer (Fig. 1D). In vivo, Ig molecules are expected to encounter TG2 in an oxidizing environment outside of the cell. Thus, among the tested Ab classes, IgD appears to be the only one that efficiently forms TG2-mediated cross-links with gluten peptide under physiological conditions. IgD therefore seems to possess unique TG2-substrate abilities that allow efficient incorporation of Ab molecules into high molecular mass complexes as well as cross-linking to gluten peptide.

TG2 primarily targets lysine residues in Ab hinge regions

To find out which Ab lysine residues serve as TG2 substrates and form isopeptide bonds with glutamine residues in gluten peptides, we carried out TG2-mediated cross-linking of IgG1, IgG3, and IgD in the presence of a short biotinylated peptide containing a target glutamine residue. After enzymatic digestion of the Ab molecules, biotin-containing fragments were isolated and identified by mass spectrometry. For all three isotypes, this approach picked up lysine residues in the Ab hinge regions as the primary TG2 targets (Supplemental Fig. 1, Table I). For IgG3 and IgD, lysine residues close to the transition between the CH2 and CH3 regions could also be detected as TG2 targets. Additionally, for IgD, a lysine residue in the middle part of the CH2 region (K282) was picked up. Interestingly, this lysine residue is predicted to be in close proximity to the hinge region in the three-dimensional structure of IgD (27).

Table I. Proteolytic Ab peptides containing lysine residues cross-linked by TG2.

Target Lysa	Sequence	Mass (Da)b	Region
IgG1
235	SCDKTH	1468.7	Hinge
235	SCDKTHTCPPCPAPELL	2704.2	Hinge
IgG3
241E / 241T / 241II	CPEPKSCDTPPPCPR	2502.3/2519.4	Hinge
353	ALPAPIEKTISK	1989.2	CH2
IgD
241Z	GGEEKK	1368.8	Hinge
241Z /241AA	GGEEKKK	1496.9	Hinge
241AA/241BB	KKEK	1253.8	Hinge
241BB	KEK	1125.7	Hinge
282	VVGSDLKDAHLTW	2162.3	CH2
367	ALREPAAQAPVKL	2086.3	CH3

Target lysines are shown in bold. In the cases where two lysines were juxtaposed, only one was cross-linked, but the two could not be distinguished. In one peptide, a glutamine residue was found to be deamidated by TG2 (underlined). The identified IgG3 hinge peptide was detected in two forms: with a free N terminus (2519.4 Da) and as a lactam (2502.3 Da), which was produced by spontaneous cyclization, resulting in the loss of ammonia.

^aLysine residues are numbered according to the Kabat numbering scheme (26).

^bMasses include biotin–QLPR peptide cross-linked to TG2-targeted lysine residues (+722.5 Da) and iodoacetamide modifications of cysteines (+57 Da).

The targeting of hinge region lysine residues by TG2 was verified by site-directed mutagenesis of the IgG1 and IgD molecules. Changing the single identified IgG1 target lysine to alanine resulted in complete abolishment of TG2-mediated cross-linking of IgG1 and gluten peptide (Fig. 2A). For IgD, mutation of the eight hinge lysine residues to alanine was sufficient to prevent TG2-mediated formation of high molecular mass complexes made up of IgD (Fig. 2B) but it only partly prevented the formation of cross-links to gluten peptide (Fig. 2A). However, additional mutation of six lysine residues in the IgD C region (Fig. 2C) almost completely blocked cross-linking. In agreement with the effect of mutating hinge lysine residues, deletion of the hinge region also prevented the formation of IgD–IgD cross-links (Fig. 2B). Furthermore, the cross-linking of IgD molecules seems to involve both hinge lysine and hinge glutamine residues, as mutation of the latter into asparagine also prevented cross-linking.

FIGURE 2.

Mutational analysis of TG2-targeted Ab residues. (A and B) Mutations were introduced at specific sites in the C regions of IgG1and IgD, which were expressed with the 679-14-E06 V region. A single lysine was changed into alanine in the IgG1 hinge region. In IgD, all eight hinge lysine residues were replaced with alanine to obtain the mutant H:K. Six additional lysine residues were changed into alanine to obtain the mutant 14xK. In the mutant H:Q the four hinge glutamine residues in IgD were changed into asparagine. Finally, an IgD hinge region deletion mutant (ΔH) was produced. The mutated Abs were compared with their wild-type (WT) counterparts in TG2-mediated cross-linking to FITC-labeled gluten peptide (0.1 mM) by detection of fluorescent protein gel bands (A) and, for the IgD Abs, TG2-mediated incorporation of Ab molecules into high molecular mass complexes as determined by Western blotting using rabbit anti-human IgD Ab (B). The reactions were allowed to proceed for 5 min at 37°C. As shown in Fig. 1B, only the lower of the high molecular mass bands is made up of IgD only, whereas the top bands are made up of both IgD and TG2. (C) Model of an IgD molecule (Protein Data Bank ID code 1ZVO) in which the positions of lysine residues that were replaced in the 14xK mutant are indicated with blue spheres. Residue numbers are indicated for one of the two H chains. The replaced lysine residues were either identified by mass spectrometry analysis or chosen based on their location at exposed positions in the IgD molecule.

TG2-specific IGHV5–51 Abs are cross-linked at a higher rate than other Abs

We have previously seen that many TG2-specific mAbs target common epitopes (epitopes 1–4) and that the epitopes reflect the Ab IGHV usage (28). Thus, most IGHV5–51 mAbs were found to bind to the same region of TG2 (epitope 1).

We now wanted to investigate whether individual mAbs could behave differently as substrates to TG2 as a consequence of different epitopes being targeted. By comparing the initial rate of TG2-mediated cross-linking of gluten peptide to the H chains of eight different TG2-specific IgG1 mAbs, we found that the two included IGHV5–51 mAbs were cross-linked at a markedly higher rate than the rest of the mAbs, which used other IGHVs (Fig. 3A). Thus, the binding of Ab to epitope 1 seems to favor TG2-mediated Ab cross-linking.

FIGURE 3.

Comparison of Abs with different V regions in TG2-mediated cross-linking. (A) Eight different IgG1 mAbs selected from a TG2-specific panel (11) were compared in cross-linking to FITC-labeled gluten peptide. The initial rate of the reaction was measured at different peptide concentrations and the values were fitted to the Michaelis–Menten equation. As indicated, the mAbs have earlier been assigned to different epitopes (28), which reflect the IGHV usage. (B) Two different TG2-specific mAbs (679-14-E06 and 693-1-F06, targeting epitopes 1 and 2, respectively) and a single TG2-negative mAb (693-2-F02, cloned from the TG2-negative population of small intestinal plasma cells from a celiac disease patient) were expressed in both the IgG1 and the IgD format and evaluated as substrates to TG2 by detection of fluorescent protein gel bands (left panel) to asses cross-linking to a FITC-labeled gluten peptide (0.1 mM) and by Western blotting of the same gel (right panel) using rabbit anti-human IgG and anti-human IgD Abs to detect concomitant incorporation of Ab into high molecular mass complexes. Incorporation of Ab molecules into high molecular mass complexes occurred with none of the IgG1 molecules but with all three of IgD molecules, although more efficiently with TG2-specific than TG2-negative Ab. The reactions were allowed to proceed for 5 min at 37°C. (C) Lineweaver–Burk plots showing the Ab–peptide cross-linking rate as a function of gluten peptide concentration measured at different concentrations of TG2-specific IgG1 or IgD. The plotted values represent the initial reaction rates for cross-link formation between Ab H chain and labeled gluten peptide at 37°C with 0.05 mg/ml TG2. At high peptide concentrations the reaction rate to some extent becomes independent of the Ab concentration. This behavior probably reflects limited accessibility of Ab lysine residues to the TG2 active site. The curves look similar for IgG1 and IgD mAbs, but the reaction is ∼10-fold faster with IgD. This difference was observed for both epitope 1 and epitope 2 mAb.

To further investigate the influence of specificity and isotype on TG2-mediated Ab cross-linking, we engineered an additional TG2-specific mAb, targeting a different epitope (epitope 2), to be expressed with the human IgD C region. Thereby, we could compare an epitope 1 and an epitope 2 mAb expressed as IgG1 and IgD. Additionally, a TG2-negative mAb was produced in the same way in both the IgG1 and the IgD format (Fig. 3B). Interestingly, whereas the TG2-specific mAbs were coupled to gluten peptide through TG2-mediated cross-linking both when expressed as IgG1 and IgD, the TG2-negative mAb was only cross-linked efficiently when expressed as IgD, suggesting that Ag binding is necessary to expose the targeted lysine residue in the IgG1 hinge region. For the two TG2-specific mAbs, kinetic analysis of the Ab–gluten peptide cross-linking reaction revealed higher reaction rates when the mAbs were expressed as IgD than as IgG1 (Fig. 3C). For both IgD and IgG1 molecules the reaction rate was higher with the epitope 1–targeting mAb compared with the epitope 2–targeting mAb.

Binding of Ab to epitope 1 allows cross-linking within a single TG2–Ab complex

To find out whether TG2 can cross-link gluten peptide to an Ab while being bound to the Ag-binding site of the same Ab molecule, we preincubated TG2 and IgD to allow formation of immune complexes and carried out cross-linking with gluten peptide in different reaction volumes. We found that cross-linking of TG2-specific IgD targeting epitope 1 was independent of the reaction volume, whereas the reaction rate was decreased by increasing the volume for both epitope 2–targeting and TG2-negative IgD (Fig. 4A). This result indicates that TG2-specific epitope 1–reactive, but not epitope 2–reactive, Abs can become cross-linked in an intraimmune complex reaction where Ab-bound TG2 can act on the same Ab molecule to which it is bound. To confirm this finding, we carried out cross-linking to gluten peptide with the Abs attached to beads (Fig. 4B). Having the Ab on beads did not affect cross-linking of TG2-specific epitope 1 IgD, whereas the cross-linking of both TG2-specific epitope 2 IgD and TG2-negative IgD was greatly reduced, thus demonstrating that the reaction with epitope 1–targeting Ab can take place equally well on a solid phase as in solution.

FIGURE 4.

Cross-linking of IgD molecules under different reaction conditions. (A) The initial rate of Ab cross-linking to FITC-labeled gluten peptide was determined for three mAbs expressed as IgD molecules; two TG2-specific (679-14-E06 and 693-1-F06) and one TG2-negative (693-2-F02). Two different reaction volumes (4 and 10 μl) were used for each peptide concentration. (B) The three IgD mAbs were expressed with a C-terminal extension containing a BirA biotin ligase target sequence, and the mAbs were biotinylated in vitro. Cross-linking to FITC-labeled gluten peptide (0.05 mM) was carried out at 37°C for 3 min either in solution or with the mAbs attached to streptavidin-coated beads, and the amount of fluorescence associated with the H chains was measured. The values represent the means of three independent experiments. Error bars indicate SD.

TG2 can cleave Ab–gluten cross-links leading to gluten deamidation

By following TG2-mediated cross-linking of gluten peptide to Ab H chain over time, we observed that peptide–Ab complexes rapidly reached a maximal amount and then gradually disappeared (Fig. 5A). Furthermore, the rate of fading of the cross-linked H chain signal was increased by addition of extra TG2 enzyme to the reaction mixture. The same overall tendency was seen for both IgG1 and IgD mAbs, but whereas the disappearance of the peptide–IgD H chain signal over time might be explained by incorporation of IgD molecules into high molecular mass complexes, the loss of peptide–IgG1 complexes most likely reflected breaking of the peptide–Ab cross-links (Fig. 5B). The TG2-mediated formation of protein cross-links is known to be dynamic, as the enzyme has the capacity to both create and hydrolyze isopeptide bonds. It was previously demonstrated that TG2 can break apart transamidated products, leading to deamidation of the involved glutamine residues (21, 29, 30). To test whether TG2 can hydrolyze gluten–IgD cross-links, leading to gluten deamidation, we isolated IgD cross-linked to gluten peptide followed by addition of TG2 and monitored the release of peptide by capillary electrophoresis (Fig. 5C). Free deamidated peptide could be detected upon incubation of peptide–IgD complexes with TG2, showing that Ab–gluten isopeptide bonds are, indeed, hydrolyzed in a TG2-mediated reaction leading to gluten deamidation.

FIGURE 5.

TG2-mediated Ab cross-linking followed over time. (A) Two TG2-specific mAbs (679-14-E06 and 693-1-F06) with IgG1 or IgD C regions were cross-linked to FITC-labeled gluten peptide (0.06 mM) at 37°C, and the amount of fluorescence associated with the Ab H chains was quantified at different time points. After 60 min, extra TG2 (1 μg) was added to the samples (indicated by an arrow). (B) Western blot using rabbit anti-human IgG, anti-human IgD, and anti-human κ L chain Abs to detect IgG1 and IgD in samples corresponding to the ones analyzed in (A). IgD but not IgG1 is incorporated into high molecular mass complexes during incubation with active TG2, resulting in gradual disappearance of the monomeric H chain band. (C) The mAb 679-14-E06 IgD was cross-linked to FITC-labeled gluten peptide (0.06 mM) for 5 min at 37°C using a biotinylated TG2 preparation. After pulldown of TG2–IgD immune complexes with streptavidin beads, a mixture of native and deamidated peptide could be detected in the supernatant by capillary electrophoresis (upper panel). After washing, IgD was eluted from the beads by lowering pH, whereas TG2 remained associated with streptavidin. Upon addition of new TG2 to the IgD eluate, free deamidated peptide was released (lower panel).

TG2 can cross-link gluten peptides to IgD BCRs

To investigate whether TG2-mediated cross-linking of Ig molecules and gluten peptide on the surface of TG2-reactive B cells leads to efficient presentation of Ag to gluten-reactive T cells, we generated retrovirally transduced murine A20 B lymphoma cells expressing HLA-DQ2.5 as well as TG2-specific or nonspecific BCRs of either the IgD or IgA1 isotype. All TG2-specific cells bound the Ag, and we did not observe any difference in the TG2-binding capacity between IgD- and IgA1-expressing cells (Supplemental Fig. 2).

Expression of HLA-DQ2.5 in mice is known to be sufficient to allow presentation of HLA-DQ2.5–restricted Ag to T cells (31). Thus, in the presence of free deamidated gluten peptide, all generated HLA-DQ2.5–expressing A20 cell lines were capable of activating gluten-specific T cells (Fig. 6A). However, when the A20 cells were pulsed with a combination of active TG2 and native gluten peptide or preformed, cross-linked TG2–gluten complexes, only cells with a TG2-specific BCR induced T cell activation. As expected, IgD- and IgA1-expressing cells performed equally in taking up and presenting TG2–gluten complexes. Surprisingly, however, no isotype-dependent difference in Ag presentation could be observed after incubation of the cells with active TG2 and free peptide. Because IgD but not IgA1 anti-TG2 mAb served as a TG2 substrate, this finding suggests that, in the employed in vitro assay, peptide is primarily taken up in a complex with TG2, most likely in the form of an enzyme-substrate intermediate, rather than via cross-linking to the BCR. Nevertheless, when the A20 cells were incubated with active TG2 and labeled gluten peptide, we could detect peptide incorporation into the BCR H chain of TG2-specific IgD, but not TG2-negative IgD or TG2-specific IgA1 (Fig. 6B). Additionally, in line with our findings using soluble Ab molecules, TG2-specific IgD was incorporated into high molecular mass complexes by TG2-mediated cross-linking (Fig. 6C). Presumably, B cell specificity for TG2 and the consequent association of the enzyme with the cell surface are required for TG2-mediated IgD BCR cross-linking, because only BCR-bound TG2 will experience a sufficiently high local concentration of membrane-bound IgD molecules. Notably, no cross-links between the peptide and other surface proteins could be detected, nor did we observe TG2–peptide complexes bound to the BCRs. These results suggest that the IgD BCR is a preferred substrate, once TG2 is captured on the cell surface. Conceivably, capture of gluten peptides through covalent attachment to BCRs would allow IgD-expressing, TG2-reactive B cells to accumulate internalized gluten peptides, leading to enhanced presentation to gluten-reactive T cells in vivo.

FIGURE 6.

TG2-mediated cross-linking of BCRs. Murine A20 cells were retrovirally transduced with constructs encoding HLA-DQ2.5 and various BCRs. The TG2-specific mAb 679-14-E06 was expressed as either IgA1 or IgD BCR, whereas TG2-specific 693-1-F06 and TG2-negative 693-2-F02 were expressed as IgD. (A) The transduced A20 cells were incubated with different concentrations of a 33-mer peptide containing the DQ2.5-glia-α2 epitope either in its deamidated (33merE) or native (33merQ) form. Active TG2 was added together with the native peptide. Alternatively, cells were incubated with preformed TG2–33-mer complexes consisting of cross-linked TG2 and peptide. The ability of the cells to present deamidated peptide to transduced DQ2.5-glia-α2–specific hybridoma T cells was assessed by measuring release of IL-2. Notably, A20 cells have been reported to secrete IL-2 upon BCR stimulation (32). However, no IL-2 could be detected in the absence of T cells (data not shown), verifying that IL-2 production in this assay is dependent on T cell activation. The results obtained by overnight incubation with 33merE peptide together with T cells show that all A20 cells are equally capable of presenting deamidated peptide following unspecific uptake or direct binding to surface HLA molecules (33). However, when the cells were pulsed with TG2 and 33merQ peptide or with TG2–33-mer complexes for 1 h prior to incubation with T cells, only cells with a TG2-specific BCR presented the deamidated peptide. T cell activation was more efficient with free TG2 (1 μg/ml) and peptide than with cross-linked TG2-peptide complexes, suggesting that TG2-specific A20 cells incubated with free TG2 and peptide take up the peptide as an enzyme-substrate intermediate rather than as cross-linked TG2–peptide complexes. The figure shows data from two independent experiments. Error bars indicate SD. (B) Cells were incubated with active TG2 in the presence of FITC-labeled gluten peptide (0.02 mM) followed by detection of fluorescent protein gel bands. The cell lysates were either incubated with protein l-agarose to pull down the BCR molecules (left panel) or analyzed directly (right panel). In both cases, TG2-specific IgD BCRs are the only molecules that are coupled efficiently to gluten peptide. (C) Western blot detection of BCR H and L chains (left panel) or TG2 (right panel) using the gel depicted in (B) loaded with pulled-down BCR samples. H and L chains were detected using rabbit anti-human IgA, anti-human IgD, and anti-human κ L chain Abs. The IgD H chain splits up into two distinct bands, which might reflect the full-length transmembrane receptor and a cleaved, GPI-linked form (34, 35). Interestingly, only one of the two appears to be cross-linked efficiently to gluten peptide. After detection of Ig molecules, the membrane was stripped, and TG2 was visualized using the mouse anti-TG2 mAb CUB7402. TG2 is pulled down together with the specific BCRs but is not associated with fluorescent peptide. The background band at ∼40 kDa is possibly derived from endogenous mouse IgG expressed by the A20 cells.

Discussion

In the present study, we demonstrate that Ig molecules can work as substrates for TG2 and take part in cross-link formation. The observed cross-linking efficacy varied greatly between different isotypes, with IgD being the superior TG2 substrate. The Ig molecules could either be cross-linked to themselves or to gluten peptide. Both types of reaction have implications for our understanding of how TG2-reactive autoantibodies are produced in celiac disease.

It has been estimated that ∼20% of mature, naive B cells react with self-antigens (36). Under normal conditions, however, such cells will not become activated and differentiate into autoantibody-producing plasma cells, because cognate, autoreactive T cells are absent. Nevertheless, in celiac disease, gluten ingestion induces the activation of autoreactive B cells. To explain the gluten dependence of the TG2-specific autoantibody response, it has been suggested that TG2-reactive B cells can receive activation help from gluten-reactive T cells (13). This model obviates the need for activation of autoreactive T cells and is supported by the well-documented presence of gluten-specific CD4⁺ T cells in celiac disease (37, 38). The model requires, however, that TG2-specific B cells are capable of efficiently taking up and presenting peptides to gluten-specific T cells. This could happen, as originally suggested, through TG2–gluten complex formation (Fig. 7A, 7B). In support of this model, we have observed that TG2 efficiently cross-links itself into multimeric complexes, which can be decorated with gluten peptides, thereby creating an Ag that can both stimulate TG2-specific B cells and be presented to gluten-specific T cells (39). Alternatively, TG2 bound to a cognate BCR might cross-link gluten peptides to B cell surface proteins. The most obvious substrate candidate for TG2-mediated cross-linking on the B cell surface is the BCR itself, as such a reaction would allow direct uptake and presentation of gluten peptides through receptor-mediated endocytosis (Fig. 7C, 7D). Our results with both soluble Abs and membrane-bound BCRs indicate that cross-linking to gluten peptides occurs efficiently with IgD molecules on the surface of unswitched cells but not with BCRs of isotype-switched cells. According to this model, isotype switching should therefore cause cessation of efficient T–B cell collaboration. As isotype switching happens before or early during T cell–induced somatic hypermutation (40), the model can explain why TG2-specific IgA plasma cells in celiac disease seem to have low levels of mutations (11). Notably, germinal centers do not contain IgD-expressing cells (41). Hence, it is possible that germinal centers are not established, and the TG2-specific response is instead generated extrafollicularly. A primarily extrafollicular response would fit with the observation that TG2-specific serum Abs disappear within months after patients commence a gluten-free diet (6, 7), indicating that long-lived plasma cells are not generated. Germinal center reactions are thought to be essential for the generation of long-lived plasma cells that migrate to the bone marrow, where they continue to secrete high levels of Ab and can sustain an Ag-specific response for decades (42, 43).

FIGURE 7.

Models to explain the gluten-dependent anti-TG2 response in celiac disease. It is assumed that TG2-specific B cells can receive help from T cells specific for deamidated gluten peptides presented in the context of HLA-DQ2.5 (or HLA-DQ8). The uptake of gluten peptides by TG2-specific B cells and the role of the BCR can be envisaged in various ways. (A) The original hapten–carrier model in which B cells take up complexes between TG2 and gluten peptide. The complexes could either be isopeptide linked or consist of TG2 with peptide bound in the active site. (B) The BCR can play an additional role when it becomes cross-linked to neighboring BCRs through transamidation activity of TG2, thereby contributing to B cell activation. Importantly, gluten peptide uptake and presentation is still needed for interaction with T cells. (C and D) The BCR can become cross-linked to gluten peptides by TG2-mediated transamidation and thereby be directly involved in peptide uptake and presentation. The reaction can happen within a single TG2–BCR complex (C) or with a neighboring BCR molecule (D). Other surface proteins might also become cross-linked by BCR-bound TG2 and thereby take part in B cell activation. (E) After receptor-mediated endocytosis of TG2 and BCR cross-linked to a gluten peptide, TG2 can hydrolyze the isopeptide bond linking BCR and peptide, leading to release and deamidation of the peptide. The deamidated peptide is subsequently bound to HLA-DQ and presented on the surface of the cell.

The reason for the extraordinary TG2-substrate properties of IgD appears to be a high number of exposed lysine residues, especially in the hinge region. The IgD hinge is long and flexible and should be easily accessible to TG2. IgG1 and IgG3 were also found to have hinge lysine residues that could serve as TG2 substrates. In contrast to IgD, however, IgG1 and IgG3 were only cross-linked efficiently in the presence of DTT, which is routinely included in the TG2 storage buffer. The explanation for this requirement is presumably that the IgG1 and IgG3 molecules contain disulfide bonds that connect the hinge regions of the two H chains, and the hinge lysine residues are only exposed when these disulfides are broken by a reducing agent.

Apart from being a preferred substrate for TG2-mediated cross-linking to gluten peptides, IgD is the only isotype that was efficiently incorporated into high molecular mass complexes as a result of Ig–Ig cross-link formation. This process might also play a role in the activation of TG2-specific B cells, as BCR aggregation is thought to be fundamental for initial B cell activation prior to interaction with T cells (24). Importantly, however, full activation leading to plasma cell differentiation would still require uptake of gluten peptides and interaction with T cells (Fig. 7B).

Whereas BCR–BCR cross-linking by definition involves several neighboring molecules, each gluten peptide will most likely be cross-linked to a single BCR molecule. This can either be the same molecule to which TG2 is bound (Fig. 7C) or a neighboring one (Fig. 7D). Whether TG2 can cross-link “its own” BCR will depend on the epitope, to which the BCR binds, and the resulting orientation of the enzyme. We have previously reported that there is an overrepresentation of IGHV5–51 among TG2-specific mAbs (11) and that almost all of the mAbs using this gene segment target the same epitope (epitope 1) (28). In the present study, we have shown that TG2 can cross-link IGHV5-51 Ab with faster kinetics than other Abs, because the reaction can take place within each immune complex, possibly leading to an advantage of cells using IGHV5–51 during B cell activation. This finding is in agreement with recent epitope mapping results, which showed that epitope 1 is located on the same plane as the active site on the TG2 surface, indicating that the active site will point toward the BCR upon binding (44).

We observed that TG2 readily cross-links IgD and gluten peptides, but at the same time the enzyme also hydrolyzes previously formed cross-links, leading to peptide deamidation. Thus, as long as glutamine-containing peptides are available, TG2 will generate cross-links, but as soon as the enzyme runs out of target glutamine residues, it will break the isopeptide bonds again. It is therefore possible that peptide–BCR cross-linking can take place on the surface of B cells, but after the BCR complex has been internalized, TG2 will hydrolyze the isopeptide bond that links BCR and gluten peptide (Fig. 7E). The released, deamidated peptide will then be available for binding to HLA molecules and presentation to CD4⁺ T cells. Importantly, TG2-mediated hydrolysis of peptide–IgD complexes could be detected regardless of whether the reaction was carried out at pH 7.4 or pH 5 (data not shown), indicating that TG2 can cleave isopeptide bonds also in the slightly acidic endosomal compartment.

The different models for B cell activation presented in Fig. 7 are not mutually exclusive, and it is possible that they all, to some extent, contribute to the activation of TG2-reactive B cells in celiac disease. Only the models that involve cross-link formation between an IgD receptor and gluten peptide, however, provide a direct explanation for both the low number of mutations and the IGHV5–51 overrepresentation among TG2-specific plasma cells in the gut (Fig. 7C, 7D). At the same time, gluten peptide uptake is directly coupled to subsequent release and deamidation of the peptide (Fig. 7E). Because of this coupling we speculate that B cells are important initiators of the anti-gluten T cell response. That is, if TG2-specific B cells play a central role as APCs, it would explain why gluten-specific T cells are reactive to deamidated rather than native peptides, although both can bind the disease-associated HLA molecules (2, 45). Additionally, intestinal deposits of TG2-specific Abs have been detected prior to the onset of clinical symptoms, suggesting that the activation of TG2-specific B cells is an early event in the development of celiac disease (46).

In summary, we have found that IgD molecules, both in solution and in cell membranes, can serve as TG2 substrates and become incorporated into multimeric complexes or cross-linked to gluten peptide. TG2-mediated coupling of BCR and gluten could represent a mechanism for B cell uptake and presentation of gluten peptide, allowing noncognate collaboration between gluten-reactive T cells and TG2-reactive, IgD-expressing B cells. However, A20 cells expressing a TG2-reactive IgA1 BCR were also able to activate gluten-reactive T cells, indicating that gluten peptide can be taken up without being cross-linked to the BCR. Importantly, however, B cells in vivo will most likely not be exposed to high concentrations of free TG2 and peptide. TG2-specific B cells will presumably pick up the enzyme from their surroundings where it is bound to the extracellular matrix or attached to cell surfaces (47–49). To get T cell help they must then be able to capture gluten peptides from a continuous flow of Ags. The employed in vitro assay is therefore not optimal for studying the interaction between TG2-reactive B cells and gluten-reactive T cells, and to properly address the role of the T–B cell collaboration new animal models are needed.

Disclosures

The authors have no financial conflicts of interest.