Journal of Molecular Biology
Volume 432, Issue 7, 27 March 2020, Pages 2405-2427
Journal home page for Journal of Molecular Biology

Assembly of Tight Junction Strands: Claudin-10b and Claudin-3 Form Homo-Tetrameric Building Blocks that Polymerise in a Channel-Independent Manner

https://doi.org/10.1016/j.jmb.2020.02.034Get rights and content

Highlights

  • Multistep polymerisation of claudins is driven by mutual stabilisation of cis- and trans-interfaces.

  • Claudin-3 and -10b differ in oligomerization but share similar face-to-face double-row architecture in strands.

  • Strand polymerisation occurs independent of channel versus barrier conformation.

  • Structural keys for claudin assembly determining paracellular permeability were identified.

Abstract

Tight junctions regulate paracellular permeability size and charge selectively. Models have been proposed for the molecular architecture of tight junction strands and paracellular channels. However, they are not fully consistent with experimental and structural data. Here, we analysed the architecture of claudin-based tight junction strands and channels by cellular reconstitution of strands, structure-guided mutagenesis, in silico protein docking and oligomer modelling. Prototypic channel- (Cldn10b) and barrier-forming (Cldn3) claudins were analysed. Förster resonance energy transfer (FRET) assays indicated multistep claudin polymerisation, starting with cis-oligomerization specific to the claudin subtype, followed by trans-interaction-triggered cis-polymerisation. Alternative protomer interfaces were modelled in silico and tested by cysteine-mediated crosslinking, confocal- and freeze fracture EM-based analysis of strand formation. The analysed claudin mutants included also mutations causing the HELIX syndrome. The results indicated that protomers in Cldn10b and Cldn3 strands form similar antiparallel double rows, as has been suggested for Cldn15. Mutually stabilising ‐hydrophilic and hydrophobic ‐ cis- and trans-interfaces were identified that contained novel key residues of extracellular segments ECS1 and ECS2.

Hydrophobic clustering of the flexible ECS1 β1β2 loops together with ECS2–ECS2 trans-interaction is suggested to be the driving force for conjunction of tetrameric building blocks into claudin polymers. Cldn10b and Cldn3 are indicated to share this polymerisation mechanism. However, in the paracellular centre of tetramers, electrostatic repulsion may lead to formation of pores (Cldn10b) and electrostatic attraction to barriers (Cldn3). Combining in vitro data and in silico modelling, this study improves mechanistic understanding of paracellular permeability regulation by elucidating claudin assembly and its pathologic alteration as in HELIX syndrome.

Introduction

Tight junctions (TJs) regulate paracellular passage of solutes and water across epithelia and endothelia. TJs contain a panel of transmembrane proteins as well as cytoplasmic membrane-associated scaffolding and signalling proteins [1]. Strikingly, claudins constitute the backbone of TJ strands [2]. In recent years, several claudin crystal structures have become available [[3], [4], [5], [6]], and this has greatly enhanced our understanding of the structure of TJ proteins. However, the oligomeric architecture of claudin-based strands is still nebulous. Moreover, formation of paracellular barriers and/or pores depends on the claudin subtype, but the underlying structural principles are still unclear. It is assumed that polymeric claudin strands are formed by cis-interactions between claudin protomers in one membrane and trans-interactions between protomers in opposing membranes [[7], [8], [9]]. It was suggested that cis-oligomerisation occurs at least partly before trans-interaction triggers polymerisation into TJ strands [10,11]. On the basis of the Cldn15 crystal structure and experimental data, Suzuki et al. proposed a model for TJ strands containing paracellular channels [12]. The model consists of anti-parallel double rows of claudin protomers in each membrane, which are joined at cell–cell contacts (Joined anti-parallel DoubleRow model, “JDR model”). However, concerns (e.g. C1–C3 below) about the validity of the model have been raised [13,14] due to inconsistencies. C1: The model leads to steric clashes at trans-interfaces, including the turn region of extracellular segment 2 (ECS2) and the loop between the β1- and β2-strand of ECS1 (β1β2loop). C2: The model does not contain cis-interaction between transmembrane helices (TMHs); this is rather surprising for oligomers of transmembrane proteins, especially since it has been suggested that TMHs are involved in claudin dimerisation [15]. For instance, TMH3 residues of Cldn3 and Cldn5 [16] and residues at the TMH1-ECS1 transition of Cldn10a/10b [17,18] have been found to affect TJ-strand formation. C3: Pore-lining residues in the JDR model of Cldn15 do not correspond with the accessibility of residues sensitive to thiol probes within the Cldn2 cation pore [13,19]. One possible explanation for discrepancies C2 and C3 might be that there are differences between Cldn2, Cldn10 and Cldn15 with respect to their protomer arrangement. Although this is conceivable, we hypothesised that the overall protomer arrangement is similar in the classic claudins (including Cldn2, Cldn10 and Cldn15), due to their high sequence homology, which implies mechanistic similarities [20]. Beside the JDR model [12], other potential protomer arrangements have been proposed that could lead to the formation of TJ strands and channels [[21], [22], [23], [24], [25], [26]].

The objective of the present study is to improve our understanding of the molecular architecture of claudin-based TJ strands and paracellular channels, using cellular reconstitution of TJ strands, structure-guided mutagenesis, in silico protein docking and oligomer modelling. We focussed on (i) Cldn10b (with high homology to Cldn15) as a prototypic channel-forming claudin and (ii) Cldn3 as a prototypic barrier-forming claudin [27]. The results indicate that classic claudins share a common overall oligomeric protomer arrangement, which represents a refinement of the JDR model suggested for Cldn15 [12]. Furthermore, we propose novel trans-interfaces, keys for structural compatibilities between claudin subtypes and a mechanism that can lead to formation of either channels or barriers.

Section snippets

Starting models of claudin homo-oligomers

For Cldn3 and Cldn10b protomers, homology models were generated using Cldn15 as template (PDB ID: 4P79 [3]) and sequence alignments (Figure S11). In silico sampling of potential protomer interfaces within Cldn3- or Cldn10b homo-oligomers was performed (e.g. with HADDOCK platform [28]). This suggested different potential cis- and trans-interfaces. However, none of these interfaces could be convincingly combined to cis-/trans-oligomers, which were fully consistent with published experimental

Discussion

The aim of this study was to improve the understanding of the molecular architecture of claudin-based TJ strands and paracellular channels (Figure 1). For Cldn15, a polymer model has been proposed, the joined double-row (JDR) model [12]. Reservations have been expressed about the validity of the JDR model, for instance, with respect to steric interference within the modelled oligomer (see C1 in Introduction). However, recent molecular dynamics (MD) simulation and protein–protein docking studies

Conclusion

We provide experimental and modelling data that enhance our understanding of the structural foundation of claudin-based formation of paracellular channels and barriers. In particular, we provide the following:

  • Evidence for multistep claudin polymerisation, starting with cis-oligomerisation that is specific to the claudin subtype and extends by trans-interaction-triggered cis-polymerisation (Figure 9)

  • Support for an anti-parallel double-row arrangement of protomers in Cldn10b- and Cldn3 strands,

Plasmids

Murine Cldn3-wt EYFP fusion constructs in pEYFP-N1 (Clontech, Mountain View, USA) have been described previously [41]. Cysteine residues were introduced resulting in Cldn3-A33C, -F34C, -I39C, -D67C, -L69C, -L70C, -N140C and R144C by site-directed mutagenesis as reported before [8]. N- and C-terminal CFP/YFP tagged human Cldn3 (in pcDNA3-YFP, pcDNA3-YFP, pEYFP-N1, pECFP-N1) [42], plasmids of pcDNA3-YFP/CFP-huCldn10b and -10a isoforms and corresponding TMH1/ECS1 chimeras -10b_a and -10a_b [18]

CRediT authorship contribution statement

C. Hempel: Investigation, Methodology, Writing ­ original draft, Writing – review & editing. J. Protze:Investigation, Methodology, Visualization, Writing ­ original draft, Writing – review & editing. E. Altun: Investigation. B. Riebe: Investigation. A. Piontek: Investigation, Methodology. A. Fromm: Investigation, Formal analysis. I.M. Lee: Investigation. T. Saleh: Investigation. D. Günzel: Conceptualization, Writing – review & editing. G. Krause: Supervision, Project administration, Funding

Acknowledgments

This research was funded by the Deutsche Forschungsgemeinschaft (DFG PI 837/4-1, DFG KR1273/8-1) and Promotionsabschlussstipendium of the Charité – Universitätsmedizin Berlin for Caroline Hempel. We thank Dr. Michael Fromm for critical reading of the manuscript.

References (44)

  • L. Winkler et al.

    Molecular determinants of the interaction between Clostridium perfringens enterotoxin fragments and claudin-3

    J. Biol. Chem.

    (2009)
  • A. Fiser et al.

    Modeller: generation and refinement of homology-based protein structure models

    Meth. Enzymol.

    (2003)
  • C. Zihni et al.

    Tight junctions: from simple barriers to multifunctional molecular gates

    Nat. Rev. Mol. Cell Biol.

    (2016)
  • M. Furuse et al.

    A single gene product, claudin-1 or − 2, reconstitutes tight junction strands and recruits occludin in fibroblasts

    J. Cell Biol.

    (1998)
  • H. Suzuki et al.

    Crystal structure of a claudin provides insight into the architecture of tight junctions

    Science.

    (2014)
  • Y. Saitoh et al.

    Structural insight into tight junction disassembly by Clostridium perfringens enterotoxin

    Science.

    (2015)
  • T. Shinoda et al.

    Structural basis for disruption of claudin assembly in tight junctions by an enterotoxin

    Sci. Rep.

    (2016)
  • S. Nakamura et al.

    Morphologic determinant of tight junctions revealed by claudin-3 structures

    Nat. Commun.

    (2019)
  • M. Furuse et al.

    Manner of interaction of heterogeneous claudin species within and between tight junction strands

    J. Cell Biol.

    (1999)
  • J. Piontek et al.

    Formation of tight junction: determinants of homophilic interaction between classic claudins

    FASEB J.

    (2008)
  • J. Piontek et al.

    Elucidating the principles of the molecular organization of heteropolymeric tight junction strands

    Cell. Mol. Life Sci.

    (2011)
  • M. Koval

    Differential pathways of claudin oligomerization and integration into tight junctions

    Tissue Barriers.

    (2013)
  • Cited by (25)

    • Defective claudin-10 causes a novel variation of HELIX syndrome through compromised tight junction strand assembly

      2022, Genes and Diseases
      Citation Excerpt :

      As G165 does not line an interface between claudin protomers (Fig. 4G, H), we propose G165A (and, to a lesser extent, G163A in claudin-10a) to indirectly affect claudin assembly by perturbing the transmembrane four-helix bundle. Underlying mechanisms might involve alterations of the cis-interface close by, slight repositioning of W30 and the extracellular β-sheet involved in cis- and trans-interactions, or alterations of the membrane environment between claudin protomers (Fig. 4G, H).13 To analyze claudin oligomerization, a TJ strand reconstitution assay was performed.

    • Multiscale modelling of claudin-based assemblies: A magnifying glass for novel structures of biological interfaces

      2022, Computational and Structural Biotechnology Journal
      Citation Excerpt :

      Remarkably, this dualism between pore-forming (dimer B, D) and non-pore-forming (dimer C, A) configurations agrees with the bimodal “open” and “close” states postulated for the original TJ dynamic model [9]. Although the two aforementioned pore-like arrangements exhibit relevant structural differences, several experimental [59,80,119] and computational [112,115,120,121] studies suggested that Cldns could form multiple binding configurations and, therefore, a possible coexistence of the two architectures cannot be excluded. Post-translational addition of palmitic acid to Cys residues of membrane proteins is a typical modification adopted to favor the dynamic localization and delocalization within the cell membrane [122–124].

    • Highlighting membrane protein structure and function: A celebration of the protein data bank

      2021, Journal of Biological Chemistry
      Citation Excerpt :

      These predictions have provided testable models of CLDN self-organization, pore structure and selectivity, and the cis and/or trans interfaces that enable these assemblies and functions. Although some models or portions of them have been verified in vitro and/or in vivo using cross-linking, mutagenesis, freeze-fracture EM, fluorescence confocal microscopy, and FRET (338, 341, 342), an experimentally derived structural basis for CLDN interactions and assemblies has yet to be elucidated. Recent progress in determining structures of CLDNs has rapidly advanced our understanding of the interconnected structural and functional biology’s of CLDNs and TJs and the mechanisms by which a bacterial enterotoxin gains access to the mammalian gut.

    View all citing articles on Scopus
    View full text