Tight junction regulation, intestinal permeability, and mucosal immunity in gastrointestinal health and disease

INTRODUCTION

Disease-associated intestinal barrier loss has been recognized for decades. In many cases this reflects mucosal injury that, in disorders targeting the gastrointestinal tract, can be a cause or consequence of disease. However, an important clue related to disease development flows from the 1986 observation that intestinal permeability is increased in up to 10% of healthy, first-degree relatives of Crohn’s disease (CD) patients [1,2]. After initial controversy [3,4], this has become widely accepted [5] and has led many to postulate that the relatives with increased permeability were at greater risk of developing CD, There were, however, no supporting data until 2020, when it was shown that the first-degree relatives with increased permeability had a 3-fold increase in risk of developing IBD compared to relatives without increased permeability [6]. In that study, increased permeability, which was determined based on a single measurement in each subject, could be detected up to 3 years prior to IBD diagnosis. Despite an older report that linked increased permeability in first-degree relatives to the NOD2^3020insC mutation [7], the 2020 study did not identify significant associations with known barrier-related or IBD genes [8]. The mechanism of increased permeability in first-degree relatives therefore remains to be determined. Nevertheless, the data support the conclusion that modest permeability increases are associated with increased risk of developing IBD in at-risk populations, i.e., first-degree relatives. Similarly, Il10^−/− mice display increased intestinal permeability prior to development of clinically evident disease [9]. Although compelling, these data do not prove that there is cause-and-effect relationship between barrier loss and disease.

Further clinical data supporting the idea that increased permeability is a marker of disease risk comes from a trio of studies showing that, among CD patients in remission, those with increased permeability relapse at substantially greater rates [10-12]. Differences in relapse rates were apparent up to 20 months after a single permeability measurement, suggesting that the measurements were not made just as a flare was about to occur. Similarly, an in vivo confocal endomicroscopy study illustrated that, during remission, increased permeability was a marker of increased relapse risk [13]. However, the observation that increased fecal calprotectin, a marker of mucosal neutrophil infiltration, predicts relapse supports the hypothesis that increased permeability is secondary to active mucosal inflammation [5]. As discussed below, animal models have been used to dissect the relationship between increased intestinal permeability, mucosal immune activation, and disease.

CYTOSKELETAL MECHANISMS OF BARRIER REGULATION

In the absence of mucosal damage, intestinal permeability primarily reflects paracellular flux, i.e., through the space between adjacent cells. The primary determinant of paracellular flux is the tight junction, which form a selectively-permeable seal just below the brush border. Transcellular transport, e.g., by nutrient transporters, is not a component of intestinal permeability but can initiate signaling events that regulate tight junction permeability and create transepithelial gradients that drive passive transport across tight junctions. For example, transcellular Na⁺-nutrient cotransport triggers events that lead to increased tight junction permeability [14] and creates an osmotic gradient that drives passive absorption of water, ions, and small macromolecules, including nutrients [15,16]. Subsequent trans-tight junction efflux of absorbed Na⁺ back into the lumen supports ongoing Na⁺-dependent transport, which would otherwise cease once luminal Na⁺ was depleted. Analyses of this normal physiological process have provided critical insight into mechanisms of tight junction regulation.

Ex vivo studies of rodent mucosae showed that Na⁺-nutrient cotransport led to condensation of perijunctional actomyosin, as seen by transmission electron microscopy [17]. It was subsequently shown that this reflected activation of myosin light chain kinase (MLCK), which phosphorylated myosin II regulatory light chain (MLC) to effect actomyosin contraction in cultured epithelial monolayers as well as rodent and human intestinal mucosae [18,19]. This first identification of signaling events that regulate tight junction permeability led investigators to ask if the same events could contribute to permeability increases induced by immune stimuli. In vitro analyses showed that apoptosis-independent barrier loss induced by tumor necrosis factor (TNF) could be reversed by MLCK inhibition [20]. In vivo analyses of the acute small intestinal diarrhea that follows systemic T cell activation showed that this was also associated with perijunctional actomyosin condensation, MLCK activation, MLC phosphorylation, and increased tight junction permeability [21]. Moreover, mucosal application of an enzymatic MLCK inhibitor or genetic knockout of the long MLCK expressed in epithelial cells prevented MLC phosphorylation, preserved tight junction barrier function, and blocked diarrhea [21]. Despite in vitro studies suggesting that activation of rho kinases can also increase MLC phosphorylation, and tight junction permeability, rho kinase inhibition did not prevent MLCK activation, MLC phosphorylation, or barrier loss. Subsequent analyses of mucosal biopsies showed that MLCK expression and MLC phosphorylation, a measure of MLCK activity, were increased in IBD patients and that the magnitude of these increases correlated with degree of histological inflammation. MLCK is, therefore, a critical regulator of in vivo tight junction permeability following physiological and immunological stimuli and is associated with disease activity in IBD patient biopsies.

DISTINCT ROUTES OF TIGHT JUNCTION FLUX

Most in vitro studies rely on transepithelial electrical resistance (TER), a measure of paracellular ion flux, to assess tight junction permeability. Some studies used larger, macromolecular probes, e.g., mannitol or fluorescent dextrans, whose paracellular flux typically correlates with TER. This led to the conclusion that ions and macromolecular probes cross tight junctions in the same manner. Although this might not explain the charge-selectivity of tight junctions, which preferentially allow cation flux in the gut, and some authors had postulated that tight junctions contained a variety of channel sizes [22,23], the single channel type model remained dogma for decades.

More recently, two in vitro studies marked the death of the single channel model. The first study [24] showed that expression of the recently-discovered tight junction protein claudin-2 reduced TER and increased paracellular Na⁺ ion conductance [25,26]. When coupled with flux analysis of a mixture of differently sized polyethylene glycols, it became clear that claudin-2 specifically increased permeability of a high-capacity pathway with a pore diameter of ~0.8 nm [24]. The data also indicated the presence of a separate, low-capacity pathway that allowed flux of larger polyethylene glycols [24]. The second key study compared the effects of IL-13, which increases claudin-2 expression and reduces TER of cultured epithelial monolayers [27], to TNF, which, as discussed above, reduces TER by activating MLCK. Although both cytokines reduced TER similarly, only TNF increased paracellular dextran permeability. Conversely, IL-13, but not TNF, selectively increased paracellular Na⁺ permeability [28]. These data led both groups to conclude that are were two distinct pathways across the tight junction. The first, a high-conductance, exquisitely size-selective (~0.8 nm diameter) and charge-selective route was termed the pore pathway [29,30]. The second, low-capacity leak pathway was size-selective, but allowed flux of much larger molecules, up to ~12 nm diameter, and was not charge-selective [31]. Thus, the immune system can differentially regulate enhance pore or leak pathway permeabilities via IL-13 or TNF, respectively (Fig. 1).

FIGURE.

Intestinal paracellular permeability reflects flux across three distinct transport pathways: the high-conductance, charge- and size-selective pore pathway (green), the low-conductance, size-selective leak pathway (blue), and the non-selective unrestricted pathway (red). Unlike the pore and leak pathways, the unrestricted pathway is tight junction independent.

Claudin-2 (green) is transcriptionally regulated during development and inflammatory disorders by signals that include IL-6, IL-13, and IL-22. This allows paracellular Na⁺ and water flux that can, by undefined mechanisms, potentiate immune-mediated disease.

Non-muscle long myosin light chain kinase (yellow) is the prototypic regulator of the leak pathway, which accommodates somewhat larger molecules and can be activated by physiological and inflammatory stimuli, including Na⁺-nutrient cotransport, TNF, LIGHT, IL-1β, and lipopolysaccharide. Recruitment of a specific long myosin light chain kinase splice variant, MLCK1, to the perijunctional actomyosin ring leads to perijunctional MLC phosphorylation, endocytosis of the tight junction protein occludin (blue), and increased leak pathway permeability.

Microbial-derived molecules, but not intact microbes, can cross the leak pathway and trigger mucosal immune responses that, in turn, cause further increases in leak pathway permeability. This self-amplifying cycle has been proposed as a contributor to disease pathogenesis.

The unrestricted pathway occurs at sites of tissue damage, where a complete loss of epithelial barrier function allows small and large molecules, including intact microbes, to access the lamina propria.

PAMP, pathogen-associated molecular patterns; DAMP, damage-associated molecular patterns; MF/DC, macrophage or dendritic cell; ILC, innate lymphoid cell; LPS, lipopolysaccharide; LIGHT, lymphotoxin-like inducible protein that competes with glycoprotein D for herpesvirus entry mediator on T cells.

IN VIVO ANALYSES OF PORE AND LEAK PATHWAYS

Although the in vitro data demonstrate the existence of two distinct tight junction flux pathways in vitro, all commonly used probes of in vivo permeability, including mannitol, were too large to traverse the pore pathway. The ideal probe would be inert to human or bacterial enzymes, incapable of transcellular transport, and small enough to crosse tight junction pores. Creatinine (~0.6 nm diameter) was identified as a viable pore pathway probe [32,33] and used together with 4 kD (~2.8 nm) and 70 kD dextrans (~12 nm) to assess leak pathway and unrestricted pathway (damage) permeabilities, respectively [34].

Application of the three-probe assay to C. rodentium colitis showed that creatinine permeability increased within 2 days of infection [33]. This was followed, sequentially, by increases in 4 kD and 70 kD dextran permeabilities. Although creatinine and 4 kD dextran permeabilities remained elevated, the increased 70 kD dextran permeability caused the ratios of creatinine or 4 kD dextran permeabilities to 70 kD dextran permeability to returned to pre-infection levels. Thus, once mucosal damage occurred, as measured by 70 kD dextran permeability, there were no longer specific increases in pore or leak pathway permeabilities. The persistent increases in creatinine and 4 kD dextran permeabilities were due to flux across the unrestricted pathway. Use of the three-probe cocktail to assess changes induced by acute T cell activation showed increased creatinine and 4 kD dextran, but not 70 kD dextran, permeabilities, indicating leak pathway activation [34]. Further, the comparable increases in creatinine and 4 kD dextran flux suggest that the pore was not affected following acute T cell activation. The permeabilities of all three probes increased in DSS colitis, consistent with damage being the predominant mechanism of barrier loss in that model. More widespread use of this three-probe approach is likely to provide greater insight into mechanisms of barrier loss in preclinical disease models. Although creatinine permeability has been assessed in human subjects [32], dextrans cannot be used. The probes used most frequently in human subjects, lactulose and mannitol, are too large to traverse the pore pathway and too close in size to discriminate between leak and unrestricted pathways. Thus, the detailed analysis needed for individualized therapeutic approaches will require development of new intestinal permeability probes.

IN VIVO CONTRIBUTIONS OF THE PORE PATHWAY TO DISEASE

Although not been measured in human subjects, the marked upregulation of claudin-2 in ulcerative colitis, Crohn’s disease, and a wide range of inflammatory conditions affecting the intestines, including HIV, celiac disease, and gut perforation-induced sepsis [27,35-37], suggest increased pore pathway permeability. Consistent with this, creatinine permeability increases in wild type (WT), but not claudin-2 knockout (Cldn2^KO), mice after cecal ligation and puncture-induced sepsis model [37]. Similarly, ex vivo analyses showed that administration of exogenous IL-13 selectively increased claudin-2 expression and pore pathway permeability in WT, but not Cldn2^KO mice [33]. Conversely, mice with transgenic claudin-2 overexpression (Cldn2^Tg) mice demonstrated replicated IL-13-induced pore pathway permeability increases without IL-13 treatment [33]. Claudin-2 is, therefore, both necessary and sufficient for IL-13-induced increases in pore pathway permeability. Together, these human and mouse data support the hypothesis that claudin-2 is a driver of diarrheal disease [36,38].

In vivo, IL-22-dependent claudin-2 upregulation increases pore pathway permeability within 2 days of C. rodentium infection [33]. Moreover, mucosal colonization, duration of shedding, and disease severity were amplified in Cldn2^KO, and attenuated in Cldn2^Tg, mice relative to WT [33]. Cldn2^KO mice could be rescued by inducing osmotic diarrhea, indicating that water efflux across claudin-2 channels is the mechanism by which claudin-2 upregulation limits [33]. Similarly, Cldn2^Tg mice are protected from DSS-induced colitis, likely as a result DSS dilution [39]. These data suggest that claudin-2-mediated Na⁺ and water efflux is a primitive defense mechanism.

In contrast to enteric infection and chemical injury, analyses of gut perforation-induced sepsis and experimental immune-mediated IBD suggest that claudin-2 upregulation exacerbates disease [37,40]. In the cecal ligation and puncture sepsis model, survival of Cldn2^KO mice was 4-fold greater than WT mice [37]. Cldn2^KO also impacted sepsis-induced dysbiosis, and microbial transfer studies suggested that this was, at least partially, responsible for their improved survival.

Similar to sepsis, claudin-2 overexpression exacerbated, and knockout attenuated, experimental immune-mediated IBD [40]. Further, pharmacological claudin-2 channel inhibition was protective in WT, but not Cldn2^KO, mice suggesting targeted claudin-2 inactivation could be a viable experimental approach. Recent advances in molecular dynamics simulations offer hope that this approach will guide development of claudin-2 channel inhibitors [41,42].

The contrasting effects of claudin-2 upregulation on infectious and immune-mediated disease may, initially, be difficult to understand. These can be reconciled by the hypothesis that claudin-2 promotes inflammatory responses that enhance pathogen clearance but can also contribute to tissue damage and disease progression.

IN VIVO CONTRIBUTIONS OF THE LEAK PATHWAY TO DISEASE

The permeability probes that have been used in human subjects measure leak or unrestricted pathway permeability. Thus, the increased permeability in some healthy, first-degree CD relatives must be due to leak or unrestricted pathway upregulation. The quantitatively modest degree of permeability increases in these healthy individuals was modeled by expressing a constitutively-active MLCK catalytic domain within the intestinal epithelium in mice (CA-MLCK^Tg). In addition to increased leak pathway permeability, these mice displayed central and peripheral nervous system alterations, behavioral abnormalities, and mucosal immune activation but did not develop spontaneous disease [43-45]. Similarly, mucosal immune activation occurs in mice lacking the JAM-A (encoded by F11r) or with functional disruption of E-cadherin, all of which have marked leak and unrestricted pathway barrier loss [46-50]. Although the extent of epithelial damage makes it difficult to link DSS-induced colitis to tight junctions, the observation that global F11r^−/−, but not endothelial-specific F11r^−/−, mice are hypersensitive to DSS-induced colitis suggests that the phenotype is linked to epithelial JAM-A loss [47].

CA-MLCK^Tg are hypersensitive to experimental, immune-mediated IBD. This is particularly evident at early times during disease progression [43]. Conversely, mice lacking the intestinal epithelial long MLCK isoform of Mylk are protected from experimental, immune-mediated IBD, but only early in the course of disease [51]. In both cases, epithelial damage caused barrier loss due and accelerated pathogenesis to equalize severities of WT, CA-MLCK^Tg, and long MLCK knockout mice [43,51]. In contrast, the protection afforded by long MLCK knockout was more durable in a graft-versus-host disease model, where epithelial damage is less severe than experimental, immune-mediated IBD [52,53]. Together with the observations that MLCK expression and activity as well as intestinal permeability are increased in GVHD patients [52,54], these data suggest that a targeted inhibitor of MLCK-dependent leak pathway permeability increases could be effective in disorders beyond IBD.

Long MLCK knockout mice demonstrate the therapeutic potential of enzymatic MLCK inhibition. However, MYLK, the gene that encodes long, non-muscle, MLCK also encodes short, smooth muscle, MLCK. Long MLCK transcription and translation are initiated upstream of short MLCK such that the two enzymes share common C-terminal catalytic and calmodulin regulatory domains, but long MLCK includes unique N-terminal sequences [55,56]. Thus, any enzymatic inhibitor of long MLCK would also inhibit short MLCK and have unacceptable toxicities, as illustrated by the hypotension, aperistalsis, and death that follow smooth muscle-specific Mylk deletion [57]. Further conservation between smooth, skeletal, and cardiac muscle MLCK catalytic domains would likely prevent development of inhibitors that are selective for non-muscle/smooth muscle MLCK.

The observation that two distinct long MLCK splice variants, MLCK1 and MLCK2, are expressed in intestinal epithelium but that only MLCK1 localizes to the perijunctional actomyosin ring and regulates barrier function [58] provided an alternative to enzymatic inhibition. Perijunctional MLCK1 recruitment depends on a short exon that completes an immunoglobulin-like domain but is spliced out of MLCK2 transcripts. This domain is required for perijunctional MLCK1 recruitment at steady-state and in response to inflammatory stimuli, e.g., TNF [59]. A small molecule that binds the unique MLCK1 immunoglobulin-like domain diverts MLCK1 away from the tight junction at steady-state [60]. This molecule, termed divertin, also blocks TNF-induced MLCK1 perijunctional recruitment and prevents acute TNF-induced, MLCK-dependent MLC phosphorylation and barrier loss without apparent toxicities. In immune-mediated experimental IBD, divertin was equal or superior to anti-TNF by all measures [60]. Thus, although divertin itself is not suitable for further development, it demonstrates the therapeutic potential of blocking MLCK1 recruitment as a means of preventing or reversing inflammation-induced barrier loss in vivo.

PORE-LEAK PATHWAY CROSSTALK

Although pore and leak pathways are considered independently in the discussion above, it is clear that they interact with one another directly and via immune signaling. For example, increased mucosal IL-13 in CA-MLCK^Tg mice upregulates claudin-2 expression [28]. Conversely, claudin-2 upregulation in experimental IBD is attenuated in long MLCK knockout mice [51]. Beyond immunity, intracellular events that may link pore and leak pathways include p38 MAPK/AP-1 signaling [61-65], NF-κB signaling [64,66-68], and autophagy [69-72].

CONCLUSIONS AND FUTURE PERSPECTIVES

The potential contributions of intestinal barrier loss to disease have been discussed for more than 60 years, preceding even the first anatomical description of the tight junction [73-75]. The 1986 discovery of increased permeability in healthy relatives of Crohn’s disease patients in was a major breakthrough that led to the hypothesis that barrier loss might be a cause, as well as a consequence, of disease [1]. The simultaneous identification of the first tight junction protein, ZO-1 [76], and recognition, one year later, that tight junction permeability could be regulated by physiological stimuli provided new molecular insight as well as support for the then novel idea that the intestinal barrier was dynamic [14,17]. The subsequent identification of the first first transmembrane tight junction protein, occludin [77], the claudin protein family [78], and the critical roles of MLCK in physiological and pathobiological processes [17,18,20,21,51,60]. Thus, in only a few decades, have brought the field from discovery of the tight junction itself to elucidation of regulatory mechanisms and development of pre-clinical agents that limit disease by modifying tight junction permeability.

The tremendous progress made over the last two decades might lead some to conclude that the discovery phase has reached an endpoint. In reality, however, much remains to be done. For example, current understanding of the nature and regulation of tight junction protein interactions with one another, lipids, and the cytoskeleton is relatively cursory. Further, we have only begun to understand the mechanisms by which physiological and pathobiological stimuli modify tight junction structure and permeability, their intersection with non-barrier tight junction functions, mucosal and systemic immunity, and the gut microbiome. Finally, it is critical to define specific causes of barrier loss, including modulation of pore, leak, and unrestricted pathway permeabilities, and the molecular mechanisms that drive these in both preclinical models and human subjects. These and other advances areas needed to support development and application of therapeutic approaches that target the epithelial barrier and, potentially, advance efficacy beyond the limits of currently available agents [79].

Purpose of review

The contributions of intestinal barrier loss, i.e., increased permeability, to multiple disorders, including inflammatory bowel disease (IBD), have been a topic of speculation for many years, and the literature is replete with conclusions based on correlation and speculation. The goal of this article is to critically review recent advances in mechanistic understanding of barrier regulation and the evidence for and against contributions of intestinal barrier loss to disease pathogenesis.

Recent findings

It is now recognized that intestinal permeability reflects the combined effects of two distinct routes across tight junctions, which form selectively-permeable seals between adjacent epithelial cells, and mucosal damage that leads to nonselective barrier loss. These are referred to as pore and leak pathways across the tight junction and an unrestricted pathway at sites of damage. Despite advances in phenotypic and mechanistic characterization of three distinct permeability pathways, development of experimental agents that specifically target these pathways, and demonstrated efficacy in preclinical models, no therapeutics have been tested in clinical trials.

Summary

After decades of speculation, therapeutic interventions that target the intestinal barrier are nearly within reach. More widespread use of available tools and development of new tools that discriminate between pore, leak, and unrestricted pathway permeabilities and underlying regulatory mechanisms will be essential to understanding the local and systemic consequences of intestinal barrier loss.

KEY POINTS

• Increased intestinal permeability is present in a subset of healthy, first-degree relatives of Crohn’s disease patients and is a risk factor for subsequent inflammatory bowel disease development. In Crohn’s disease patients, increased intestinal permeability is associated with relapse.

• Intestinal permeability reflects flux across tight junctions, which seal the paracellular space, by two distinct routes and tight junction-independent, unrestricted flux at sites of tissue damage. Immune and other stimuli differentially regulate these permeability pathways by activating divergent intracellular processes. Understanding which pathway is effected is, therefore, a prerequisite to development and use of molecularly targeted barrier-restorative therapies.

• Studies of preclinical mouse models indicate that tight junction-specific barrier loss promotes enteric pathogen clearance, but that barrier restoration limits experimental inflammatory bowel disease, graft-versus-host disease, and sepsis.

• Tight junction barrier restoration may provide a non-immunosuppressive/immunomodulatory approach that for both active inflammatory bowel disease and maintenance of remission.

Abbreviations

CD

Crohn’s disease

IBD

inflammatory bowel disease

MLCK

myosin light chain kinase

MLC

myosin regulatory light chain

TER

transepithelial electrical resistance