Article Text

Original article
Cell polarity-determining proteins Par-3 and PP-1 are involved in epithelial tight junction defects in coeliac disease
  1. Michael Schumann1,
  2. Dorothee Günzel2,
  3. Nataly Buergel1,
  4. Jan F Richter2,
  5. Hanno Troeger1,
  6. Claudia May1,
  7. Anja Fromm1,2,
  8. Detlef Sorgenfrei2,
  9. Severin Daum1,
  10. Christian Bojarski1,
  11. Martine Heyman3,
  12. Martin Zeitz1,
  13. Michael Fromm2,
  14. Joerg-Dieter Schulzke1,4
  1. 1Department of Gastroenterology, Infectious Diseases and Rheumatology, Campus Benjamin Franklin, Charité Berlin, Berlin, Germany
  2. 2Institute of Clinical Physiology, Campus Benjamin Franklin, Charité Berlin, Berlin, Germany
  3. 3INSERM, U989, Interactions of the Intestinal Epithelium with the Immune System, Université Paris Descartes, Paris Cedex 15, France
  4. 4Department of General Medicine, Campus Benjamin Franklin, Charité Berlin, Berlin, Germany
  1. Correspondence to Professor Dr Joerg-Dieter Schulzke, Charité – Universitätsmedizin Berlin, Campus Benjamin Franklin, Medizinische Klinik für Gastroenterologie, Infektiologie und Rheumatologie, 12200 Berlin, Germany; joerg.schulzke{at}


Background Epithelial barrier defects are well known in coeliac disease, but the mechanisms are only poorly defined. It is unclear, whether barrier disturbance reflects upregulated epithelial transcytosis or paracellular leakage.

Objective To characterise the molecular structure and function of the epithelial tight junction (TJ) and mechanisms of its dysregulation.

Methods Molecular analysis of proteins involved in TJ assembly and their regulation was performed by western blotting and confocal microscopy correlated to electrophysiology.

Results A complex alteration of the composition of epithelial TJ proteins (with more pore-forming claudins like claudin-2 and a reduction in tightening claudins like claudin-3, -5 and -7) was found for protein expression and subcellular localisation, responsible for an increase in paracellular biotin-NHS uptake. In contrast, epithelial apoptosis was only moderately elevated (accounting for a minor portion of barrier defects) and epithelial gross lesions—for example, at cell extrusion zones, were absent. This TJ alteration was linked to an altered localisation/expression of proteins regulating TJ assembly, the polarity complex protein Par-3 and the serine-/threonine phosphatase PP-1.

Conclusions Changes in cell polarity proteins Par-3 and PP-1 are associated with altered expression and assembly of TJ proteins claudin-2, -3, -5 and -7 and ZO-1, causing paracellular leakage in active coeliac disease.

  • Celiac disease
  • epithelial barrier
  • Par-3
  • tight junction
  • apoptosis
  • gastrointestinal lymphoma
  • gamma delta t cells
  • lamina proprial lymphocytes
  • coeliac disease
  • enteropathy
  • endoscopy
  • new imaging technologies
  • confocal laser endomicroscopy
  • chromoendoscopy
  • endoscopic retrograde pancreatography
  • endoscopic ultrasonography
  • endoscopic polypectomy
  • adenoma
  • endoscopic procedures
  • gut immunology
  • epithelial permeability
  • epithelial transport
  • Crohn's disease
  • inflammatory bowel disease
  • mucosal immunology
  • HIV-related gastrointestinal disease
  • Helicobacter pylori
  • acid-related diseases
  • non-ulcer dyspepsia
  • genetic polymorphisms
  • gastric neoplasia
  • inflammation
  • IBD
  • 5-aminosalicylic acid (5-ASA)
  • inflammatory mechanisms
  • radiation enteritis
  • apoptosis

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Significance of this study

What is already known about this subject?

  • The epithelial barrier in coeliac disease (CD) is defective.

  • The fraction of epithelial apoptotic cells is increased.

  • The epithelial polarisation determinant Par-3 is genetically linked to CD.

What are the new findings?

  • Altered expression and subcellular localisation of tight junctional claudins contribute to coeliac epithelial barrier defect.

  • Dysregulated serine/threonine phosphatase PP-1 and Par-3 are linked to intracellular localisation of claudin proteins, the latter pointing to a defect in the polarisation of coeliac epithelial cells.

  • Increased epithelial apoptosis is only a minor contributor to the epithelial barrier defect.

  • Even in high-resolution scanning, lesions like epithelial micro-erosions do not contribute to the barrier disturbance.

How might it impact on clinical practice in the foreseeable future?

  • Treatments designed to strengthen the epithelial barrier in CD need to focus also on the expression and assembly of structural tight junction proteins.

  • Therapeutic strategies influencing the epithelial-to-mesenchymal transformation need to be considered.


In coeliac disease (CD), epithelial barrier disturbance is a prominent feature which perpetuates active disease stages via activation of the mucosal immune response.1 There is an continuing debate about whether this is mainly due to endo-/transcytosis or also due to paracellular defects. The paracellular pathway is sealed by tight junction (TJ) proteins, the molecular composition and distribution of which are still not fully characterised in CD. Furthermore, other mechanisms might account for barrier defects in CD—namely, apoptotic leaks, epithelial gaps and erosive gross lesions, the functional relevance of which is unknown.2–5

TJs are composed of a network of sealing strands with the integral membrane proteins occludin, junctional adhesion molecule, the claudin protein family and tricellulin.6 Earlier studies of our group and others showed the TJ structure to be altered in CD—namely, in freeze fracture electron microscopy analysis.4 7 Subsequently, molecular studies have already elucidated changes in TJ proteins including claudin-2 and ZO-1 phosphorylation,6 8–12 although a complete description of all functionally relevant TJ proteins and their subcellular distribution is still lacking.

Importantly, a recent candidate gene study has identified a link between Par-3 (partition defective-3), playing a crucial role in establishing epithelial cell polarity and TJ assembly, and CD pathogenesis.13 However, TJ assembly is not only determined by the Par-3 protein complex and protein kinase C but also negatively regulated by serine/threonine phosphatases PP-1 and PP-2A.14 15

As far as signalling is concerned, regulation of TJ-associated proteins has been detected in response to cytokines elevated in CD and inflammatory bowel disease.6 8–12 Moreover, Fasano's group has identified prehaptoglobin A (zonulin) as regulating coeliac TJs.16 However, further regulatory influences of the TJ in CD have to be considered to explain the data of the gene association study mentioned above.

Thus, this study aimed to uncover the molecular structure of the TJ in CD and to yield direct evidence for a functional relevance of these changes in paracellular barrier properties. Also, we wanted to consider the influence of other paracellular features such as epithelial apoptosis and erosions and to characterise regulatory pathways contributing to TJ alterations in CD beyond Th1-cytokines and zonulin. Hereby, a contribution of the polarisation complex Par-3/PP-1 was identified.

Materials and methods

A more elaborate version of the ‘Material and methods’ section can be found in the online Supplements.

Patients and healthy controls

We recruited two study groups: (1) healthy controls without any CD-typical clinical symptoms and a normal duodenal mucosal architecture receiving a gluten-containing diet; (2) patients with CD receiving a gluten-containing diet with villous atrophy (duodenal biopsies) and a positive tissue-transglutaminase- or endomysium-IgA serology (online supplemental table). The study was approved by the local ethic committee (Charité Berlin, #227-44) and written consent was obtained from all patients in our study.

Impedance spectroscopy

Endoscopic biopsy specimens of human duodenal mucosa were analysed for one-path impedance spectroscopy as described earlier.17 Owing to the frequency-dependent electrical characteristics of the capacitor, transmural resistance (Rt) was obtained at low frequencies and subepithelial resistance (Rsub) at high frequencies. The epithelial resistance (Re) was obtained from Re = Rt−Rsub.

Mannitol flux measurements

After mounting the duodenal biopsy specimens into Ussing chambers, paracellular permeability was determined from [3H]mannitol fluxes (mucosal-to-serosal), as previously published.18

Ca++ switch experiments

Caco-2 cells stably transfected with claudin-5 and HT-29/B6 cells were seeded on PCF filters and mounted in Ussing chambers for continuous monitoring of transepithelial resistance (Rt).19 A calcium switch assay was carried out as previously described using EGTA (2 mM) and CaCl2 (3 mM).15 Cells were immunostained for claudin-5 and ZO-1.

Conductance scanning

Experiments were carried out as described earlier for identifying epithelial regions with increased conductance.17 20

Quantification of epithelial apoptosis by TUNEL assay

Terminal deoxynucleotidyl transferase-mediated deoxyuridine triphosphate nick-end labelling (TUNEL) assay was performed as described previously.11

Analysis of duodenal surface area

Haematoxylin/eosin-stained sections (3 μm) of duodenal biopsies were analysed with a light microscope in a modification of previous studies.11 18 The ratio of mucosal-to-serosal surface area was determined with the processing software ImageJ (NIH, Bethesda, Maryland, USA).

Western blot analysis

TJ protein expression was determined from membrane extracts of duodenal biopsy specimens, as described previously.21 Primary antibodies were anti-claudin and anti-occludin antibodies and secondary peroxidase-conjugated anti-rabbit IgG and anti-mouse IgG.

Reverse transcription and real-time PCR

Total RNA from duodenal mucosal biopsy specimens was transcribed to cDNA. Real-time PCR with Applied Biosystems' TaqMan Gene Expression Assays no. Hs00267568_m1 (PP-1, α isozyme) and Hs00204426_m1 (PP-2, regulatory subunit A-alpha) was performed. Relative quantification referred to human GAPDH as an endogenous control allowing normalisation of threshold cycle (Ct) values of genes of interest.

Surface biotinylation

Human duodenal mucosa was mounted in Ussing chambers. Sulfo-NHS-biotin (molecular weight 443 Da, final concentration 0.8 mM) was added to the mucosal compartment (30 min, 4°C). Paraformaldehyde-fixed tissue was counterstained with E-cadherin and Dy560-streptavidin and analysed with confocal laser scanning microscopy (LSM).

Immunostaining and confocal laser scanning microscopy

Staining and LSM visualisation was performed as reported earlier.3 Immunostaining was done using polyclonal anti-rabbit and monoclonal anti-mouse antibodies, with AlexaFluor594 anti-rabbit-IgG and AlexaFluor488 anti-mouse IgG secondary antibodies. Relative quantification of claudin proteins along the lateral epithelial membrane was performed using the profiling tool of ImageJ.


All data are means ± SEM. The statistical evaluation is based on multivariant analysis (one-way analysis of variance) or two-sided Student t test for unpaired data or Mann–Whitney U test, as appropriate. A value of p<0.05 was considered significant. To validate the significance of Rt in the calcium switch experiments data were analysed with a curve-fitting algorithm (GraphPadPrism software).


Expression of tight junction proteins

For a molecular characterisation immunoblots of TJ proteins were analysed in duodenal biopsies (figure 1). This study revealed a complex dysregulation involving several claudins and occludin. Pore-forming claudins-2 and -15 were upregulated and the tightening claudins-3, -5 and -7 and occludin were downregulated in CD. The effect on claudin-2 appears to be post-transcriptional, since measurement of claudin-2 mRNA level showed no significant increase in patients with CD (supplemental figure 1).

Figure 1

Western blotting for tight junction (TJ) proteins. Duodenal biopsy samples of patients with coeliac disease were compared with controls. (A) Representative western blots for claudins and occludin. β-Actin was used as a loading control. (B) Densitometric analysis of the TJ protein level by western blotting. *p<0.05, n=7.

Subcellular localisation of structural TJ proteins

Mucosal biotin-NHS exposure revealed inhomogeneous paracellular barrier defects with focal paracellular leaks in CD (figure 2A). To identify local alterations in TJ protein composition and to study their subcellular distribution, coeliac duodenum was immunostained for TJ proteins and visualised by confocal LSM (figure 3). Claudin-2 staining was more intense in CD and still exclusively in TJs, predominantly in crypt TJs. Coeliac mucosae had reduced claudin-3 and did not show the apical enhancement and ZO-1 co-localisation compared with controls (illustrated by the apical-to-basal LSM signal quantification in supplemental figure 2). Claudin-5 and -15 were only partially localised to coeliac TJs. A substantial portion of these proteins was localised to intracellular vesicles and therefore did not participate in paracellular barrier function. The histological pattern of claudin-7 appeared less intense and inhomogeneous in CD with a remarkably differing staining level between different crypts and even in adjacent cells of the same CD mucosa (figure 3, CD detail). Intriguingly, this claudin-7 staining did not show co-localisation with ZO-1 in controls and CD and was strikingly discontinuous along the lateral membrane as revealed by the saw tooth-like pattern in the apical-to-basal LSM quantification tool (supplemental figure 2). The staining patterns of claudin-1, claudin-4 and occludin were not altered in CD (data not shown). To our surprise, staining for the pdz-scaffolder ZO-1 showed remarkable changes in CD. When imaged with optimal saturation and resolution (figure 4), relative ZO-1 quantification along the lateral membrane revealed a partial mislocalisation to more basal regions of the lateral membrane in CD.

Figure 2

Functional barrier studies on coeliac duodenal mucosa. (A) Surface biotinylation. Human duodenal control specimens and acute coeliac biopsy specimens were mounted in Ussing chambers. Biotin-NHS was added to the mucosal compartment for 30 min at 4°C to inhibit endocytotic biotin uptake. Samples were fixed and stained with Dy560-streptavidin (red). Counterstaining was carried out with E-cadherin as an epithelial cell marker (green) and DAPI as a nuclear marker (blue). Arrows mark exclusive apical streptavidin staining (upper figure) or apical and lateral staining (lower figure), arrowheads mark biotin depots in the lamina propria. Bars 10 μm. (B) Conductance scanning of coeliac duodenal mucosae. Conductance scanning was carried out as described in the ‘Methods’ section on specimens of duodenal coeliac mucosa. Local conductivity G was determined at several locations along a scanning line either between residual villi (squares) or along villi (circles). The roughly horizontal lines indicate the absence of significant focal leaks. Data are from three independent experiments.

Figure 3

Subcellular localisation of tight junction (TJ) proteins. Merged pictures of ZO-1 (green), the respective claudin (red) and nuclei (blue, DAPI) as obtained by confocal laser-scanning microscopy of duodenal biopsy specimens from control subjects and patients with acute coeliac disease. Figures show duodenal crypts with the TJ localisation indicated by ZO-1. (A) Claudin-2 was restricted to TJs of crypt cells in acute coeliac disease (CD) and was not detectable in controls. (B) Claudin-3 showed an intense staining in the lateral membrane and partial co-localisation with ZO-1, but was reduced in acute coeliac disease. (C) Claudin-5 was localised to the TJ in both groups, revealing an intracellular vesicular pool only in acute CD. (D) Claudin-7 showed a homogeneous expression pattern along the lateral membrane of crypt and villous cells in the control group, whereas it stained inhomogeneously in acute CD. (E) Claudin-15 was localised to the TJ in control and in acute CD, but was also found in intracellular vesicles in acute CD. Size bars, 20 μm.

Figure 4

Expression of Par-3 and ZO-1 along the lateral cell membrane. (A) High-sensitivity analysis of confocal laser scanning microscopy 12 bit-recordings of ZO-1 and Par-3 fluorescence signals in control and coeliac disease (CD) mucosae. While ZO-1 and Par-3 signals were confined to the most apical portion of the lateral membrane in control crypt cells (small arrows), they were spread along the lateral membrane (small arrows middle panel) or to intracellular vesicles in CD (small arrows lower panel). (B) The diagrams present apical-to-basal signal profiles along lateral membranes as indicated by the long arrow in (A) (five people in each group, 25 lateral membranes per patient, means ± SEM; *p<0.05). Size bars, 10 μm. (C) Par-3 immunofluorescence of a Marsh I duodenum showing a subcellular Par-3 distribution as found in higher Marsh grade histologies (arrows).

Epithelial barrier in CD

To correlate alterations of structural TJ proteins or epithelial apoptoses with the CD-associated barrier defect, barrier function was quantified in duodenal mucosae. One-path impedance spectroscopy and the permeability for 180 Da mannitol was determined (table 1). Epithelial resistance (Re) was significantly reduced by 48% in CD, corresponding to a doubled mannitol permeability.

Table 1

One-path impedance analysis, permeability for mannitol and apoptosis

Dyspolarisation of Par-3

To elucidate mechanisms responsible for the intracellular dyslocalisation of structural TJ proteins, we studied signalling via the TJ-regulator Par-3 and its opponents, the serine/threonine protein phosphatases-1 and -2.13 In the process of epithelial polarisation Par-3 moves to the apical membrane and contributes to the organisation of primordial junctions to belt-like TJs.14 The Par-3 protein level was not changed in CD (figure 5A,B). Analysed for subcellular distribution, Par-3 immunostaining showed an apical membrane localisation at crypts (figure 4) and an apical and lateral membrane distribution at villi in control duodenum (data not shown). In CD, epithelial crypt cells displayed an enhanced basolateral and partially intracellular Par-3 staining closely resembling published data on dyspolarised Drosophila cells expressing malfunctioning mutants of the Par-3 homologue Bazooka (figure 4A,B).22 Interestingly, the Par-3 phenotype was also found in the only patient on a gluten-containing diet with Marsh I histology included in our study, suggesting that earlier stages of CD also display epithelial dyspolarisation (figure 4C). Further support for a dysfunction of Par-associated epithelial cell polarisation came from the analysis of atypical protein kinases-C (aPKCs), which are known to be involved in this process. Immunostaining of aPKC-ξ, but not aPKC-λ, disclosed a distorted cellular distribution, lacking tight junctional PKC-ξ in comparison with controls (supplemental figure 3). The pathogenetic cause of the altered Par-associated cellular polarisation machinery, might be due to mutations in polarisation-associated genes (eg, Par-3) or, possibly also, an induction by cytokines, especially transforming growth factor β (TGFβ).

Figure 5

Regulation of Par-3 and serine-/threonine protein phosphatases (A) Immunoblots of Par-3, PP-1 and PP-2A in control and coeliac duodenum. (B) Densitometric analysis of immunoblots in A (Mann–Whitney U test; *p<0.05 vs control, n=9) (black line within data points represents the median). (C) Compartmentalisation of protein phosphatases and Par-3 in human duodenal mucosa immunostaining for PP-1 and PP-2A. Analysis with confocal laser scanning microscopy (LSM). PP-1 and PP-2A expression is mainly restricted to the epithelial layer in human duodenal mucosa. (D) Transepithelial resistance (Rt) of Caco-2 cells and Caco-2 cells stably expressing claudin-5 (p<0.001, n=6). (E) Analysis of tight junction assembly in Caco-2 cells expressing claudin-5 (calcium switch assay). Cells were pretreated overnight with okadaic acid (OA, 4 nM), tautomycin (tauto, 100 nM) or buffer; Rt was monitored during EGTA treatment (2 mM) and calcium replacement (CaCl2, 3 mM). Single curves were analysed and fitted with GraphPadPrism, n=9, p<0.0001 for tauto versus control and OA versus control. (F) Confocal LSM of Caco-2 cells processed as shown in (E), 30 min after calcium rescue (claudin-5, red; ZO-1, green).

Preincubation of Caco-2 intestinal epithelial cells with TGFβ1 and TGFβ2 retarded the repolarisation after EGTA-induced dyspolarisation (calcium switch assay) (figure 6A). Confocal LSM linked the diminished recovery of TGF-treated cells to leaks in the epithelial layer that showed a disorganised Par-3 pattern with intracellular rather than junctional Par-3 and ZO-1 (figure 6A). Biotinylation of Caco-2 cells which were in the process of polarisation (ie, after calcium addition in a switch experiment) allowed allocation of the functional barrier defect to those TJs that also showed reduced junctional Par-3 and claudin-5 (supplemental figure 4).

Figure 6

Transforming growth factor β (TGF-β) retards epithelial polarisation in Caco-2 cells (A) Caco-2 cells preincubated with TGF-β1 or TGF-β2 (20 ng/ml) were analysed by calcium switch assay. Rt monitoring was carried out as described in figure 5. *p<0.05 versus control. Caco-2 filters were unmounted from the Ussing chamber and stained for Par-3 (red) and ZO-1 (green). TGF-β pretreatment causes the focal loss of belt-like ZO-1 and Par-3-positive tight junction (TJ) structures (arrowheads) in combination with the appearance of intracellular vesicles containing Par-3 and ZO-1 (arrows). (B) Scheme for signalling pathways targeting TJ assembly in coeliac disease.

Expression and function of serine/threonine protein phosphatases

Mucosal serine/threonine phosphatase PP-1 protein expression was of epithelial origin and was increased sixfold in CD, while PP-2A remained unchanged (figure 5A–C). Interestingly, the elevated PP-1 protein level probably results from post-transcriptional regulation, since only a statistically non-significant tendency towards a higher PP-1 mRNA level was found in CD mucosa using real-time PCR (supplemental figure 1). Immunostaining localised PP-1 to the junctional complex of epithelial cells (supplemental figure 5) with only minimal staining in the subepithelial compartment. To elucidate the functional importance of this finding for TJ protein assembly, we studied the impact of PPase activity on barrier function in conjunction to subcellular claudin-5 distribution by PPase inhibition in an epithelial calcium switch assay (figure 5E). As previously published, Caco-2 cells stably overexpressing claudin-5 displayed a 2.1-fold higher Rt than native Caco-2 cells displaying only minimal claudin-5 levels (figure 5D).19 After addition of the calcium-chelator EGTA, Rt rapidly decreased to 22% of the pre-EGTA-Rt. Calcium replenishment increased Rt back to 55% of the pre-EGTA-Rt at 30 min. Overnight preincubation of cells with protein phosphatase inhibitors caused the EGTA-associated reduction in Rt to be less pronounced (tautomycin 31%, okadaic acid 49%) and the calcium-associated rescue of Rt to be stronger (tautomycin 63%, okadaic acid 95%). On confocal LSM Caco-2 cells strongly expressed claudin-5 at the TJ. After addition of EGTA claudin-5 was partially localised to intracellular vesicles (figure 5F). Replenishment of calcium caused claudin-5 to recycle back to the TJ rather than to intracellular vesicles in cells preincubated with the PPase inhibitor tautomycin. To support the validity of these data the results were confirmed with HT-29/B6 cells (supplemental figure 4).

Epithelial apoptosis and gross lesions

Another cause for coeliac barrier defects is epithelial cell apoptosis. To shed light on its relative impact on barrier dysfunction we determined the fraction of apoptotic enterocytes in duodenal mucosae by TUNEL assay (table 1, supplemental figure 6A), which was 2.8% in CD compared with 1.0% in controls. If compared with our previous apoptosis data in HT-29/B6 cells (supplemental figure 6B), where the impact of the apoptotic fraction on electrical conductivity G was determined using camptothecin as selective apoptosis inductor, we can deduce that a 1.8% increase in apoptosis causes only a small increase in electrical conductivity of 4%. This corresponds to only about 0.7 Ω·cm2 in CD mucosae.23

To exclude other relevant epithelial leaks we spatially analysed the CD mucosal barrier by conductance scanning (figure 2B). This revealed a lower electrical conductance at the sites of residual villi than over surface epithelium, suggesting even tighter rather than more leaky villus extrusion zones and no focal electrical conductance leaks in CD (as, for example, seen in ulcerative colitis as positive control).


Molecular structure of the epithelial TJ in CD

From what is already known in the literature the epithelial TJ in CD can be considered to be disorganised.7 24 However, data on the molecular composition and the subcellular distribution of TJ proteins in CD are only rare and partly contradictory. Szakal and coworkers studied the expression level of a subset of TJ proteins—namely, claudin-2, -3 and -4, and showed an increased expression for the pore-forming TJ protein claudin-2 as well as an elevation of the tightening TJ protein claudin-3, thereby leaving the question more or less unresolved, whether there is a contribution of claudins to the CD epithelial barrier defect.9 A study by Ciccocioppo and coworkers evaluated the expression level of another TJ-strand associated protein, ZO-1, which is considered to be of regulatory importance and identified a reduced phosphorylation status of ZO-1 in CD, while the ZO-1 expression level was unaltered.8 Additionally, occludin levels were detected to be lower in CD, which was interpreted as the result of the change in ZO-1 phosphorylation status. On the other hand, two other studies suggested transcriptional downregulation of ZO-1 in CD.25 26 This might reflect differences in the severity of CD which can vary from only distinct changes with high IEL levels (Marsh I) to total villous atrophy (Marsh IIIc).

In our study, patients with severe CD were mainly considered (Marsh III). Also, our study is the most complete analysis of TJ proteins in CD so far and provides additional essential information on their subcellular distribution by confocal LSM, since a measurement of the overall expression level in the cell itself as obtained in western blots (whether altered or not) misses information on the abundance of the protein in the TJ strands. However, an overall quantification is the first essential step and a prerequisite for a subsequent analysis of the subcellular distribution of the strand-forming TJ proteins. This analysis of protein levels identified two pore-forming claudins (claudin-2 and -15) as upregulated and a number of tightening TJ proteins (occludin and claudin-3, -5 and -7) as downregulated. As mentioned above, a prominent role for claudin-2 has been suggested previously,9 a finding which is now completed by the documentation of an increased tight junctional localisation. Interestingly, claudin-2 transfection into L-fibroblasts results in discontinuous TJ strands (pearl string-appearance) which have indeed been observed in CD-TJs by freeze fracture electron microscopy.4 27 28 Such an increase in claudin-2 expression is, for example, induced by tumour necrosis factor α (also upregulated in CD).11 12

Sealing claudins with reduced expression in CD comprise claudins-3, -5 and -7. Claudin-3 has been shown to be a barrier forming protein for ions and larger solutes.29 It is noteworthy that claudin-3 did not localise to TJ at all in CD, which is functionally even more relevant. Claudin-5 has a sealing function for molecules >500 Da and is defective in Crohn's disease.11 19 30 Its expression was reduced in CD and perhaps even more important it was less abundant in TJ strands. Instead, it showed an intracellular vesicle-like staining in CD which could reflect a defective intracellular trafficking either by increased endocytotic uptake or a disturbed membrane flow to the plasma membrane. However, direct data on TJ endocytosis or defective TJ assembly in CD are lacking, although endocytosis of TJ proteins has been reported to be an important mechanism in TJ regulation via actin-dependent endocytosis into caveolae in response to interferon-γ (which is elevated in CD).31 Claudin-7—having only a very faint expression at the TJ—is assumed to affect the intestinal architecture rather than barrier function,32–34 since it is described to interact with the epithelial differentiation antigen EpCAM, which (i) is modulated through the Wnt/tcf pathway and (ii) is mutated in the rare hereditary tufts enteropathy.35 36

Functionally, claudin-15 is a pore-forming TJ protein with a selective permeability for cations and a pivotal role in intestinal glucose absorption.37 Although its expression is upregulated in CD, it shows an intracellular vesicle-type staining pattern rather than a TJ localisation. Thus, the reduction in junctional claudin-15 might contribute to coeliac glucose malabsorption rather than to barrier dysfunction. Moreover, intestinal epithelia lacking claudin-15 were shown to evolve into a megaintestine with hypertrophic villi, enabling speculation about a role for claudin-15 in the pathogenesis of coeliac mucosal architecture changes.37

Of special interest, by quantifying the confocal LSM signal along the lateral plasma membrane the TJ-associated pdz-domain protein ZO-1 was shown to be released from its strong TJ association in CD and to spread out to the basal portion of the basolateral enterocyte membrane. This is in line with a report of Hamada et al, in which a methotrexate-associated rearrangement of ZO-1 was associated with its phosphorylation status.38 These data therefore suggest that diminished tyrosine phosphorylation leads to ZO-1 redistribution with a consequent weakening of the epithelial barrier.8 Studies on ZO-1/-2-deficient epithelial cells have shown that reintroduction of ZO-1 determines the location of TJ assembly with the plasma membrane, thereby highlighting the functional importance of ZO-1 association with distinct parts of the cell membrane.39 Notably, it is also known from wound healing experiments that ZO-1 is one of the first proteins detectable in the primordial spot-like junctions. Taken together, our findings on the localisation of structural TJ strand proteins and TJ-associated proteins in CD, in conjunction with published data on the function of ZO-1, suggest that redistribution of ZO-1 might be an early dysregulatory event that could account for the mislocalisation of structural TJ proteins of the claudin family.

TJ regulation by Par-3 and PP-1 in CD

These complex alterations suggest a principal dysregulation of TJs in CD. Having this in mind and considering the report on a genetic association of the TJ regulator Par-3 with CD,13 we examined the protein expression and cellular distribution of Par-3 and aPKCs and its functional opponents, the protein phosphatases (PP)-1 and -2A. It is known that Par-3 can form a cell polarity driving complex with atypical PKC and Par-6.14 Dysfunction of Par-3 is known to cause a dyspolarisation of epithelial cells in various cell models.14 22 Par-3 and aPKC knock downs have been shown to inhibit the proper recruitment of structural TJ proteins to primordial spot-like junctions and the assembly of a belt-like TJ, respectively, rather than to stimulate disassembly of mature TJs.14 40 41 This is in accordance with our finding that changes in subcellular Par-3 expression between coeliac and control epithelial cells were more pronounced in crypt than surface epithelial cells, where maturation of TJs is already complete. In contrast, the serine/threonine protein phosphatases PP-1 and PP-2A dephosphorylate structural TJ proteins, thereby impairing epithelial barrier function.15 As a result of these studies we suggest that the cause for the severely altered assembly of TJs in CD is an upregulated PP-1 expression and Par-3 localised to the basolateral rather than to the apical membrane of the enterocytes. This was brought to a functional level by Ca2+ switch experiments—an established model for epithelial polarisation—which revealed an intracellular claudin-5 pattern resembling that in CD with PPase activity dependence.

In addition to a genetic background of cell polarisation regulation by Par-3 in CD, TGFβ—known to be secreted by gluten-specific T cells42—also regulates the Par-complex within a process called epithelial-to-mesenchymal transformation (EMT).43 By downregulating proteins of the apical junctional complex and upregulating mesenchymal proteins, EMT induces a ‘dyspolarisation’ of originally polarised epithelial cells. Even if our understanding of this regulation is at a preliminary stage, TGFβ which can retard the polarisation of intestinal epithelial cells seems to be of relevance in CD (scheme in figure 6B). It is difficult to determine, if epithelial dyspolarisation is secondary or primary in CD pathogenesis. However, since Par-3 was found to be genetically associated with CD and the single patient in our study having Marsh I histology also revealed a basolateral Par-3 localisation, it seems reasonable to hypothesise that this process starts early in disease. On the other hand, the fact that it is induced by cytokines that are secreted by gluten-specific T cells might also be compatible with a role in an amplification loop, where barrier defects develop secondarily and further increase the antigenic load of the subepithelial compartment. However, TGFβ derived from ‘regulatory’ CD8+TCRγδ+NKG2A+ intraepithelial T cells has also been proposed to be associated with a mucosa-protective function in patients with treated CD, counteracting the cytotoxic effects of interleukin 15.44

Functional outcome of altered TJs in CD and barrier-depressing processes

The functional consequence of TJ alterations in CD has been questioned in the past 2–3 years by numerous findings of a pronounced transcellular passage via endocytotic uptake of tracers of either size including gliadin and gliadin cleavage products.2 3 45 These studies also uncovered a disturbed degradation of proteins during transcytosis in CD as well as altered transcytotic transport through the enterocytes with consequent antigen presentation by epithelial cells.2 45 Does this necessarily mean that paracellular permeability does not contribute to the barrier disturbance in CD? Our study yields clear-cut evidence for a paracellular penetration of intermediate-sized molecules by means of mannitol and biotin-NHS tracer studies, which resulted in the detection of functional leaks through the TJ between adjacent enterocytes, at least in advanced stages of CD considered here in our analysis (Marsh III). Thus, both routes, paracellular as well as transcellular, appear to be relevant in CD. However, a cut-off-size—defining the macromolecular size that still passes the coeliac epithelial barrier in either process—as well as the disease stage-dependency of this process remain to be defined. So far, it seems clear that gliadin fragments of several kilodaltons have been shown to pass the barrier via transcytosis and thus paracellular leakage may be more relevant for smaller antigens and ions which contribute to leak flux diarrhoea in CD.46 Another process—namely apoptosis, previously discussed to be abundant in CD makes only a minor contribution to the overall epithelial barrier defect as indicated by a correlation with the apoptotic fraction in HT-29/B6 monolayers.5 23 47 Thus, the quantitative impact of epithelial apoptoses in CD might have been overestimated in the past. A more refined technique, conductance scanning, was applied to search for leaks/gaps or erosive lesions arising from extrusion zones in CD. For this purpose, extrusion zones were scanned and the regional conductivity compared with that of surface regions of CD mucosae. However, such gross lesions were not detected in CD.

In summary, we have presented a detailed molecular analysis of the disorganised TJ in CD with claudin-2 to be upregulated and downregulation of claudin-3, -5 and -7 together with a refined description of the subcellular TJ protein distribution. The latter showed that ZO-1 as well as claudin-3 and -5 was distributed off the TJ in CD. Beyond this structural analysis, functional data were provided in direct support of a paracellular barrier breakdown. In contrast, erosive lesions were not detected and epithelial apoptosis only slightly contributed to the barrier disturbance in CD. Finally, the complex tight junctional phenotype in CD is functionally connected to the regulation of TJ assembly including stimulating (Par-3) and blocking factors (PP-1).


Supplementary materials


  • Funding Supported by Deutsche Forschungsgemeinschaft (Schu 2389/1-1 and FOR721/2 TP2).

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval Ethic committee Charité Benjamin Franklin (#227-44).

  • Provenance and peer review Not commissioned; externally peer reviewed.