Concerns about the histological assessment in a mouse model of human celiac disease

Celiac disease (CD) is a common immune-mediated, gluten-sensitive enteropathy that develops in human leukocyte antigen (HLA)-DQ2 or HLA-DQ8 allele carriers due to the dysregulation of gluten-specific T helper cells. Although ingested gluten from cereals was identified as the disease driver in the 1950s, the ultimate cause (or disease trigger) remains unknown. CD is further characterized by the production by intestinal B cells of autoantibodies against tissue transglutaminase 2, an enzyme that can modify gluten peptides by deamidation, thus rendering gluten more recognizable for celiac patient T cells. The celiac intestinal lesion is hallmarked by crypt hyperplasia and villus atrophy, but also by intraepithelial T cells and lamina propria plasma cell infiltration.^{1, 2}

So far, an accurate CD model in rodents that would unite the features listed above does not exist.³ HLA-DQ2 or HLA-DQ8 transgenic, mouse class II knockout models provide appropriate restriction of T helper cells, valuable for the experimental study of cellular and humoral immune responses directed at gluten. But HLA-transgenic mice do not develop disease on gluten-containing diet, or following various additional challenges, while any histological changes in the intestine remain mild and rudimentary.^4-7 In contrast, Rag1^−/− mice adoptively transferred with gliadin memory T cells have been described as a CD model with an intestinal phenotype that includes villus atrophy, crypt hyperplasia and mucosal mononuclear cell infiltration.⁸ However, the gluten-sensitive enteropathy in this model depends on the quantitative depletion of regulatory T cells. Consequently, negative control mice on gluten-free diet also show mild-to-moderate (autoimmune) enteropathy.

In-depth histological analyses are vital for the development of animal models of human diseases. Since CD primarily affects the proximal small intestine, and is characterized by well-documented histological changes,² thorough histological investigation of at least the entire small intestine, that is, duodenum, jejunum and ileum, is key for the validation of any animal model of human CD.³ Non-validated surrogate biomarkers such as immune mediators or anti-gliadin antibodies cannot replace histological parameters. The assessment of architectural changes in the small intestine forms an important part of the histological analysis in CD and its animal models. However, it is well known that villus/crypt (V/C) ratios are notoriously unreliable as sole parameters of intestinal damage for various reasons; they are affected by diets, the age of the animal, and location in the intestinal compartment, as well as technical issues, for example, sampling consistency,⁹ tissue orientation¹⁰ or compression by luminal content. Intraepithelial lymphocyte (IEL) counts may be used to quantify mucosal cell-mediated immune reactions, but they correlate poorly with villus damage in mice.¹¹ For all these reasons, it is good practice in experimental gastroenterological research to present integrative histopathological scores and to assess the inflammatory infiltrates in detail,^{8, 12, 13} ideally making use of unbiassed quantitative approaches. Finally, intestinal damage is associated with overt degenerative changes, cell loss through apoptosis or necrosis, and/or changes in the mitotic rate. Their assessment can further inform the study.

Abadie et al., in their original article published in 2020 in Nature,¹⁴ proposed a model for human CD in which overexpression of IL-15 in the intestinal epithelium combined with CD4+ T-cell immunity directed against gluten leads to an activated killer cell phenotype of intestinal intraepithelial CD8+ T cells and villus destruction. To test this hypothesis, the authors used mice characterized by IL-15 transgene overexpression in the intestinal mucosa (under both the villin¹⁵ and MHC class I Dd promotors¹⁶). They showed that gluten feeding of HLA-DQ8, Dd-villin-IL-15 transgenic mice leads to the development of an adaptive T helper 1 (Th1) cell-mediated immune response directed against gluten, and upregulation in the ileum of markers of cytotoxic IEL, such as perforin 1, granzyme B and activating natural killer receptor NKG2D. These findings were interesting, because in wild-type mice, feeding of gluten alone does not induce a Th1/cytotoxic immune response or mucosal damage. It is hence even more surprising that Abadie et al.¹⁴ reported ileal damage in their mouse model and claimed it to be dependent on (1) oral gluten feeding, (2) CD-associated HLA-DQ8 co-transgenic expression and (3) tissue transglutaminase activity.

Previous studies have shown that the overexpression of IL-15 in the intestinal epithelium under the non-classical MHC class I T3b promotor leads to the spontaneous development of severe inflammation of the small intestine in adult mice.¹⁷ In addition, overexpression of IL-15 under the Dd promotor, as in the study by Abadie et al.,¹⁶ has been shown to manifest with multiorgan lymphocytic infiltration. These pathological changes occurred in the absence of any additional dietary or genetic manipulations and were therefore considered a direct result of IL-15 overexpression.

To support the sweeping claim that IL-15 transgenic mice represent a model with CD-like histological changes depending on the combination of oral gluten challenge, the HLA-DQ8 risk allele and tissue transglutaminase activity, Abadie et al. relied exclusively on the semiquantitative assessment of IEL counts and V/C ratios in the ileum. They did neither assess inflammatory infiltrates nor did they present histopathological scores, thereby ignoring common practice in the field.¹³ Clinical or functional parameters such as body weight development or intestinal permeability were also not analysed. Moreover, the authors did not provide any rationale for the unusual choice of the ileum as the sole site for the histological assessment. This is surprising not only because the ileum is less frequently affected in CD than the proximal small intestine, but also because the villus length decreases from duodenum to ileum in mice.^{9, 13, 18} Hence, the choice of the mouse ileum, where villi are physiologically short, as the site for the assessment by V/C ratio of treatment- or transgene-induced pathological changes is inappropriate. In addition, it is noted that in several figures the authors showed ileal cross sections where the villi are compressed by intestinal content (most clearly in figures 3g and 4b). These shortcomings invalidate the reported villus length measurements and V/C ratios.

It is a further concern that Abadie et al.¹⁴ did not provide any information on their sampling approach. To distinguish the ileum from the jejunum in cross-sections, mouse pathologists follow strict standard procedures and generally sample the ileum 1 cm proximal to the cecum.¹⁹ The histology panel in figure 1b shows three small intestinal cross sections, with only the middle (gluten) one providing clear evidence of two serosal attachments; this raises doubts that the left and right images show ileum at all. To specify, the terminal ileum has two serosal attachments, the mesentery and the ileocecal fold, whereas the jejunum only has the mesentery.¹⁸ Not only did these inconsistencies render the basis for histological comparisons in the report by Abadie et al. unreliable, the presented differences in ileal IEL counts and V/C ratios between experimental mice and controls are also very modest (figure 1b, 3a,b, 4b). Given the limitations highlighted above, their meaning and relevance remain completely unclear.

Finally, Abadie et al.¹⁴ used MHC class II (I-A^b) competent HLA-DQ8 transgenic mice in their study. Concerning the claim that the described changes in the ileum are dependent on HLA-DQ8 transgene expression, it is premature to conclude that HLA-DQ8 is functional in the presence of mouse MHC class II, and that HLA transgene expression would lead to the recruitment of DQ8-restricted CD4+ T cells in this model. The authors do not provide any evidence of CD4+ T cell HLA-DQ8 restriction.

Following the original article¹⁴ and our letter to Nature raising our concerns that the journal rejected in 2020, the authors have recently published a detailed protocol for the use of the same model in Current Protocols.²⁰ Here, they specifically instruct the reader to (1) empty the bowel before tissue fixation and (2) sample the distal ileum 0.5 cm proximal from the cecum, two standards to which the original publication evidently did not adhere. The protocol also includes a method for BrdU staining of proliferative intestinal crypt cells, not used as a readout in the original publication.¹⁴ To illustrate this staining method, the authors now show images from the proximal small intestine (location not specified; figure 3). And like in their original article, they state prematurely that they observe differences as a result of gluten challenge in the model, this time in the BrdU staining patterns, without providing or having earlier provided for review an appropriate histological analysis of the full intestine.

In summary, the histopathological changes in the ileum of HLA-DQ8, Dd-villin-IL-15 transgenic mice presented by Abadie et al.¹⁴ are mild to moderate at best, without clear evidence of villus atrophy, crypt hyperplasia or inflammation, and without evidence that they do not represent a direct effect of overexpression of IL-15 in the intestine. Crucial information in particular on the involvement of the duodenum and jejunum, and on the ileal dissection and sampling approach is lacking from the original study. The article provides no reliable evidence that the claimed histopathological changes are caused by gluten feeding, would be linked to HLA-DQ8 restriction of CD4+ T cells, and/or depend on tissue transglutaminase activity. The method paper²⁰ does not provide any validation of the original findings. Therefore, both reports by Abadie et al. fall short of their central claims. To convince the reader, these studies would have required an appropriate and much more thorough morphological investigation.