Role of gut epithelial-cell derived transglutaminase 2 in the formation of celiac disease autoantibodies

Reviewer #1 (Public review):

The manuscript provides compelling evidence that IEC-derived TG2 can drive anti-TG2 autoimmunity in the presence of gluten-specific T-cell help. However, several issues should be addressed.

(1) First, the physiological relevance of the TG2-DGP fusion construct should be discussed in greater depth, as the permanently linked fusion protein may not fully recapitulate naturally occurring TG2-gluten complexes.

(2) Second, the proposed mechanism of antigen uptake into Peyer's patches remains largely inferential, and evidence supporting involvement of M cells or other antigen-sampling pathways would strengthen the conclusions. The authors conclude that Peyer's patches are the primary site of priming and propose that luminal TG2-gluten antigen is sampled from the intestinal lumen. However, the study does not directly demonstrate the route of antigen entry (for example, via M cells, dendritic cells, or other pathways). Additional discussion or experimental evidence supporting the proposed antigen trafficking mechanism would strengthen the conclusions. Given the emphasis on Peyer's patches, direct assessment of M-cell involvement would be particularly valuable.

Suggested experiment: RANKL-induced M-cell expansion (or M-cell depletion) to test whether M-cell-mediated sampling is required for induction of TG2 autoimmunity.

Minor Points:

(1) Villin is known to be expressed throughout the whole villus axis; however, the transgenic model described here seems to express the fusion protein in a specific cell type only; the authors should explain the reason for this. It would be better if the authors specified the region used in immunohistochemistry.

(2) The authors showed that CD4 proliferation holds high until a dilution of 1:24; however, B cells show until 1:2 dilution. In the co-culturing, the author used CD4: B cells approximately in the ratio of 1:10. The reason for this should be explained and justified.

(3) The authors showed that TG2-specific B cells were enriched in the PP, accompanied by higher gluten-specific CD4 cells. Recent studies showed that M cells express TGM2 and can indeed present the gliadin peptide to T cells independent of APCs. (Wang et al. 2025 Nature). The authors should discuss this in light of their findings.

(4) Regarding mice, mentioning the gender and age of the mice is important. Please include this information.

(5) It would be better if the authors specified the region of tissue used in immunohistochemistry.

(6) Regarding the lysate preparation, the authors should mention whether they used the whole small tissue or a segment. It would be helpful if the author provides the IEC lysate in terms of protein concentration rather than dilution.

Reviewer #2 (Public review):

Summary:

The manuscript investigates whether transglutaminase 2 (TG2) released from intestinal epithelial cells (IECs) can trigger anti-TG2 autoantibody production by activating TG2-specific B cells that recognize the TG2-gluten complex, with assistance from gluten-specific T cells.

Strengths:

The main study is well designed and mechanistically important, providing direct evidence for a longstanding hypothesis in celiac disease pathogenesis. The mouse model, which uses an inducible TG2 fusion to a deamidated gluten peptide sequence (TG2-DGP fusion) in IECs, is elegant and directly tests a mechanism that could previously only be inferred indirectly. The Sollid group previously showed TG2 is present in the gut lumen (Meling et al., 2024) and can facilitate B-T cell collaboration in vitro (Iversen et al., 2020). This paper closes the loop by showing IEC-restricted TG2-DGP expression drives autoantibody formation in vivo. The experiments include appropriate controls, including the use of a Tgm2-/- B cell donor to avoid confounding by TG2 shed during cell isolation. The paper also shows that B cell expansion occurs in Peyer's patches (PP), whereas class-switching involves mesenteric lymph nodes (MLN).

Weaknesses:

Some suggestions for revision:

(1) Physiological relevance of antigen dose and kinetics. The dox-induced pulse of TG2-DGP expression is a bolus-like event, whereas endogenous IEC shedding and TG2 release presumably occur continuously at low levels. The author should discuss more how the kinetics and magnitude of the induced signal compare with estimated physiological TG2 derived from IEC turnover, and whether the strong immunogenicity observed here could overstate the potency of the natural process.

(2) The translational value of the model may hinge on epitope relevance. Celiac anti-TG2 antibodies preferentially target a small number of conformational epitopes on open TG2, and this restriction is tied to how TG2 is captured and presented. A central claim of the paper is that IEC-derived TG2 is the relevant source for autoantibody elicitation, rather than extracellular matrix-associated TG2 or other pools. If IEC-derived TG2 only activates non-human-relevant epitopes, that would undermine the claim that this is the mechanism responsible for the actual patient autoantibody repertoire. Whether IEC-derived TG2-DGP would recruit a broader range of TG2 epitopes seen in human disease, or only the subset compatible with this conformational epitope, is worth examining.

Reviewer #3 (Public review):

Summary:

This manuscript from Løberg and coworkers tests a previously suggested hypothesis for how coeliac disease is triggered. The disease is characterized by T cells responding to deamidated gluten peptides that are invariably associated with the production of IgA antibodies against the self-protein transglutaminase 2 (TG2), which is responsible for the deamidation. According to their hypothesis, intracellular TG2 is released into the gut when intestinal epithelial cells (IEC) are shed into the lumen and degraded, where it forms transient complexes with gluten while deamidating it. As complexes are taken up into gut-associated lymphoid tissues, the complexes will function as hapten-carrier antigens for B-T cell interactions, driving B cell responses against TG2 and T cell responses against deamidated gluten. They have previously created a mouse strain carrying human antibodies specific for TG2, TCR specific for deamidated gluten and HLA molecules that can efficiently present the gluten peptides. Here, they have now created a novel mouse model where they can induce expression of TG2 fused to deamidated gluten peptides in IEC. Using this together with previous mouse strains, they show that IEC lysates from induced animals contain antigens that will activate antigen-specific B and T cells in vitro, and that the fusion protein can be detected in the gut lumen between 12 hours and 3 days after induction of expression. Furthermore, when TG2- and deamidated gluten-specific B and T cells are transferred into the mice that are subsequently induced, TG2-specific IgG and IgA antibodies are produced and there is strong activation of germinal centre responses of the B cells and Tfh responses by the T cells. However, for this response to happen, the transferred T cells need to be orally primed with deamidated peptides and the strong mucosal adjuvant CT before induction of the TG2 fusion protein that will activate B cells.

Strengths:

There is a lack of animal models that mimic celiac disease, at least partly because it is a complex condition involving both classical hypersensitivity and autoimmunity. While the major T cell antigens responsible are gluten peptides that have been deamidated through the action of transglutaminase 2 (TG2) and are presented on specific HLA molecules, the disease is also invariably associated with an autoantibody response against the enzyme, TG2, that modifies the peptides, usually consisting of IgA antibodies. The antibody response is so strongly correlated with the disease that it is now used for diagnosis in the clinic. The group responsible for the current work has previously developed mouse models in efforts to study the disease, including strains expressing disease-associated human HLA, TCR and antibodies. Although they have so far not been able to fully replicate the disease in mice, these have been useful tools to study how the disease may be triggered in the gut.

Where TG2 modifies gluten peptides and where it is encountered by B cells has been a bit of a puzzle. Most agree that TG2 is primarily an intracellular protein, but that significant amounts can also be found extracellularly, attached to cell surfaces or the extracellular matrix. Gluten peptides are mainly located extracellularly or in endosomes, and for this reason extracellular TG2 in the gut have been thought to be most important for deamidation. However, where TG2 is present in the gut has not been definitively answered. In addition, an attractive model that would explain the immune responses is that extracellular TG2 bind gluten peptides, that these are taken up by TG2-specific B cells, and deamidated peptides are subsequently presented on MHCII for T cells. This would require TG2-peptide complexes to be present extracellularly where B and T cells are both present. The authors have previously hypothesized that TG2 may be produced in IEC and released into the gut lumen, where it would form complexes with gluten peptides and then be taken up by M cells positioned over gut-associated lymphoid tissues, in particular Peyer´s patches. Previously, they have demonstrated that B and T cells can be forced to interact in Peyer´s patches following an oral immunization with antibodies specifically recognizing that B cell receptor that are linked to a gluten peptide in the presence of a strong mucosal adjuvant in the form of cholera toxin.

Here, the authors now specifically address whether expression of TG2 in IEC is sufficient to trigger autoreactive B and T cells to respond. They show that induced expression of a TG2-gluten peptide fusion protein specifically in IECs is sufficient to trigger the production of IgG and IgA antibodies against TG2 when antigen-specific B and T cells are present. Following induction, they also demonstrate activation of B and T cells in Peyer´s patches, resulting in autoantibody production of IgA and IgG antibodies. The use of an inductive system to express the antigen is an important control, with transferred cells only activated when the antigen is produced. In addition, the fact that activation of cells mainly occurs in Peyer´s patches further supports that it is dependent on TG2 fusion protein secreted into the gut lumen that triggers the response, as the main function of Peyer´s patches is to initiate responses to antigen in the lumen.

Overall, the authors present convincing data that proteins primarily present intracellularly in IEC can trigger antigen-specific responses in Peyer´s patches, given that antigen-specific B and T cells are present.

Weaknesses:

This paper is still some way from establishing a mouse model that mimics celiac disease, and the authors do not make any such claims. It demonstrates a situation that could be similar to when celiac disease is first triggered in humans. However, the fact that a TG2-specific IgA antibody response can be triggered in the model does not necessarily show that it can be triggered in a more natural situation (when the antigen would need to be a TG2 - gluten peptide complex) or that it is the only way for such a response to be triggered. Hence, it does not really prove that the model describes the situation in people with coeliac disease, only that a similar response can be triggered in the mouse model.

There are also many remaining questions regarding the response that could have been further developed. Some of which should probably have been addressed already in this manuscript. First, given the importance of the expression of the antigen in IECs into the lumen and the strong responses in Peyer´s patches, the authors should have provided additional information about the expression of the TG2-fusion protein in epithelial cells in the follicular-associated epithelium covering and other epithelia surrounding the structure. Is it possible that even M cells express the antigen, and that it may not need to be in the lumen to reach the Peyer´s patches? Second, it is not addressed whether antigen-specific pre-activation of T cells using peptide and CT is strictly needed or if CT alone could have created an environment where antigen-specific responses are triggered by the TG2 fusion protein alone (i.e. using CT as an adjuvant for a response to an antigen expressed in IEC). Or even if other inflammatory gut responses, not necessarily triggered by CT but instead by virus, pathogenic bacteria or oral dextran, would have sufficed. Although T cell cross-reactivity to external antigens could play a role in humans developing disease, it would be easier to envision a situation where any gut inflammation, in the presence of antigen-specific B and T cells and a susceptible HLA environment, could trigger a response. Finally, the role of dendritic cells and/or other APCs present in the gut and/or Peyer´s patches in the priming of T cells should probably have been discussed, i.e., whether presentation of antigen by B cells is sufficient or not. If inflammation or CT alone is sufficient for T cell priming (see above), this indirectly supports that direct presentation from B cells may play an important role in T cell activation.

Other themes that could have been further developed are more general, involving the system and its correlation with celiac disease. For example, what would happen if the expression of the TG2 fusion protein is maintained over time using dox; will damage be observed to the gut mucosa that is similar to the one seen in celiac patients? What would happen when the antigen is no longer expressed? Will the mice maintain memory B and T cells that can rapidly respond upon re-expression of the antigen? Or would the constant presence of endogenous TG2 eventually lead to tolerance development? It would be particularly interesting to see if the cells are first triggered/primed by the expression of the GP2 fusion antigen in IEC, and then determine if memory cells can be activated, or active responses maintained, if gluten is given orally to the mice and its interaction with endogenous TG2 is then sufficient.

Author response:

We thank you and the three anonymous reviewers for your careful evaluation of our manuscript. We here provide a provisional author response to the reviewer’s comments.

Reviewer #1:

Major point 1.

First, the physiological relevance of the TG2-DGP fusion construct should be discussed in greater depth, as the permanently linked fusion protein may not fully recapitulate naturally occurring TG2-gluten complexes.

We thank the reviewer for this important point and agree. The TG2-DGP fusion protein is an artificial antigen and enforces a fixed linkage between TG2 and a deamidated gluten peptide, whereas naturally occurring TG2-gluten complexes are transient and their abundance and availability to TG2-specific B cells remains uncertain. We will discuss this limitation in greater detail in our revised manuscript.

Major point 2.

Second, the proposed mechanism of antigen uptake into Peyer's patches remains largely inferential, and evidence supporting involvement of M cells or other antigen-sampling pathways would strengthen the conclusions. The authors conclude that Peyer's patches are the primary site of priming and propose that luminal TG2-gluten antigen is sampled from the intestinal lumen. However, the study does not directly demonstrate the route of antigen entry (for example, via M cells, dendritic cells, or other pathways). Additional discussion or experimental evidence supporting the proposed antigen trafficking mechanism would strengthen the conclusions. Given the emphasis on Peyer's patches, direct assessment of M-cell involvement would be particularly valuable.

Suggested experiment: RANKL-induced M-cell expansion (or M-cell depletion) to test whether M-cell-mediated sampling is required for induction of TG2 autoimmunity.

We thank the reviewer for this excellent suggestion. Assessing M-cell involvement would indeed be very interesting, and M-cell depletion would be a logical way to test whether M-cell-mediated uptake is required for TG2 autoimmunity in our model. These experiments however require breeding our mice and validating M-cell depletion in our HLA-DQ2.5 Tgm2-ω18merEE x Villin1-rtTA transgenic mice, which is a time-consuming endeavor and would be best suited in a follow-up study. To address the reviewer’s concern in the current manuscript, we will discuss that our data identify the Peyer’s patch as an important site for B-cell priming. We will discuss entry routes via M cells as one plausible mechanism alongside other mechanisms such as uptake by dendritic cells.

Minor point 1.  

Villin is known to be expressed throughout the whole villus axis; however, the transgenic model described here seems to express the fusion protein in a specific cell type only; the authors should explain the reason for this. It would be better if the authors specified the region used in immunohistochemistry.

We will address this point and provide additional data on the expression of the TG2-DGP fusion protein in epithelial cells the revised manuscript, as is also recommended by reviewer #3.

Minor point 2.

The authors showed that CD4 proliferation holds high until a dilution of 1:24; however, B cells show until 1:2 dilution. In the co-culturing, the author used CD4: B cells approximately in the ratio of 1:10. The reason for this should be explained and justified.

The CD4 T cell : B cell ratio used here was not specifically optimized for this study, and the choice of ratio, if any, was primarily a result of relatively poor CD4 T cell yields. We agree that the T cell and B cell readouts differ in sensitivity, consistent with our previous publication (PMID 34731200) in which we observed efficient T-cell proliferation in response to recombinant TG2-DGP fusion protein at much lower antigen concentrations then those required to detect B-cell proliferation. Because B-cell proliferation depends on T-cell help, we acknowledge that the ratio can affect the B-cell readout, particularly at low antigen doses. Our conclusion, however, does not concern the antigen threshold required to activate either cell type, but rather that the epithelial cell lysate engages both T cells and B cells.

Minor point 3.

The authors showed that TG2-specific B cells were enriched in the PP, accompanied by higher gluten-specific CD4 cells. Recent studies showed that M cells express TGM2 and can indeed present the gliadin peptide to T cells independent of APCs. (Wang et al. 2025 Nature). The authors should discuss this in light of their findings.

The paper the reviewer refers to demonstrates that human M cells can present antigens to T cells, but importantly the same paper shows that mouse M cells do not have this capacity. This species difference infers that this mechanism does not affect the interpretation of our mouse model. It is also not directly relevant to the question our study addresses as M-cell presentation to T cells does not explain how TG2-specific B cells acquire TG2-gluten antigen, which is the step that drives the autoantibody response in our model.

Minor point 4.

Regarding mice, mentioning the gender and age of the mice is important. Please include this information.

This information will be included in the revised manuscript.

Minor point 5.

It would be better if the authors specified the region of tissue used in immunohistochemistry.

See response to minor point 1.

Minor point 6.

Regarding the lysate preparation, the authors should mention whether they used the whole small tissue or a segment. It would be helpful if the author provides the IEC lysate in terms of protein concentration rather than dilution.

This information will be included in the revised manuscript.

Reviewer #2:

Major point 1.

Physiological relevance of antigen dose and kinetics. The dox-induced pulse of TG2-DGP expression is a bolus-like event, whereas endogenous IEC shedding and TG2 release presumably occur continuously at low levels. The author should discuss more how the kinetics and magnitude of the induced signal compare with estimated physiological TG2 derived from IEC turnover, and whether the strong immunogenicity observed here could overstate the potency of the natural process.

The reviewer raises a very important and valid point that also relates to major point 1 raised by reviewer #1. We will discuss how our artificial model relates to endogenous TG2-gluten complex formation in greater depth in the revised manuscript.  

Major point 2.

The translational value of the model may hinge on epitope relevance. Celiac anti-TG2 antibodies preferentially target a small number of conformational epitopes on open TG2, and this restriction is tied to how TG2 is captured and presented. A central claim of the paper is that IEC-derived TG2 is the relevant source for autoantibody elicitation, rather than extracellular matrix-associated TG2 or other pools. If IEC-derived TG2 only activates non-human-relevant epitopes, that would undermine the claim that this is the mechanism responsible for the actual patient autoantibody repertoire. Whether IEC-derived TG2-DGP would recruit a broader range of TG2 epitopes seen in human disease, or only the subset compatible with this conformational epitope, is worth examining.

The anti-TG2 B cells in our model carry a B-cell receptor derived from a celiac disease patient, which targets epitope 1 of TG2, a dominant epitope of celiac autoantibodies. This antibody binds mouse and human TG2 with equal affinity (PMID ID 31727780). The response elicited in our model is therefore directed at a human disease-relevant epitope, an epitope that is also present on mouse TG2. We will clarify this in the revised manuscript.

Reviewer #3:

Major point 1.

First, given the importance of the expression of the antigen in IECs into the lumen and the strong responses in Peyer´s patches, the authors should have provided additional information about the expression of the TG2-fusion protein in epithelial cells in the follicular-associated epithelium covering and other epithelia surrounding the structure. Is it possible that even M cells express the antigen, and that it may not need to be in the lumen to reach the Peyer´s patches?

We will address this point and provide additional data on the expression of the TG2-DGP fusion protein in epithelial cells of the follicular-associated epithelium in the revised manuscript. 

Major point 2.

Second, it is not addressed whether antigen-specific pre-activation of T cells using peptide and CT is strictly needed or if CT alone could have created an environment where antigen-specific responses are triggered by the TG2 fusion protein alone (i.e. using CT as an adjuvant for a response to an antigen expressed in IEC). Or even if other inflammatory gut responses, not necessarily triggered by CT but instead by virus, pathogenic bacteria or oral dextran, would have sufficed. Although T cell cross-reactivity to external antigens could play a role in humans developing disease, it would be easier to envision a situation where any gut inflammation, in the presence of antigen-specific B and T cells and a susceptible HLA environment, could trigger a response.

We thank the reviewer for this important point. We agree that it is relevant to assess whether an inflammatory environment could trigger antigen-specific responses to the TG2-gluten fusion protein. To address this, we will attempt to include new experimental data in the revised manuscript in which the pre-activation of gluten-specific T cells is omitted, and cholera toxin as adjuvant is administered at the time of TG2-gluten protein induction. Testing other inflammatory triggers such as viral or bacterial infection is beyond the scope of the study, but we will discuss it as an important question in the revised version.

Major point 3.

Finally, the role of dendritic cells and/or other APCs present in the gut and/or Peyer´s patches in the priming of T cells should probably have been discussed, i.e., whether presentation of antigen by B cells is sufficient or not. If inflammation or CT alone is sufficient for T cell priming (see above), this indirectly supports that direct presentation from B cells may play an important role in T cell activation.

We will expand the discussion on priming by B cells, dendritic cells and other antigen-presenting cells in light of the new data from Major point 2. We note that these experiments will indicate whether antibody responses can occur without pre-activation of T cells with peptide and cholera toxin, but they will not by themselves identify B cells as the antigen-presenting cells involved in priming of the T cells.