ZnUMBA – a live imaging method to detect local barrier breaches Free

Epithelial tissues separate internal compartments within an organism and serve as barriers to the external environment. Formed by sheets of polarized cells connected to one another by cell–cell junctions, epithelia regulate selective transport of ions and solutes across the tissue – either through the cells themselves (transcellular transport) or through the space between cells (paracellular transport). Tight junctions (TJs), which are comprised primarily of claudin family and occludin transmembrane proteins along with cytoplasmic linker proteins that connect the TJ to the actomyosin cytoskeleton, are responsible for regulating paracellular transport (Quiros and Nusrat, 2014; Turner et al., 2014; Van Itallie and Anderson, 2014; Varadarajan et al., 2019; Zihni et al., 2016). Along the length of bicellular TJs, claudin-based TJ strands interact between neighboring cells, bringing the cell membranes into close apposition (Farquhar and Palade, 1963; Furuse et al., 1998). The extracellular loops of claudins can form size- and charge-selective pores, where some claudins are predominantly pore forming toward cations or anions, while others are predominantly barrier forming (Günzel and Yu, 2013). At tricellular TJs (tTJs), where three cells come together, specialized proteins including tricellulin (also known as MARVELD2) and angulins help to seal the paracellular space (Higashi and Chiba, 2020; Higashi and Miller, 2017).

The size- and ion-selectivity of TJs differs in different epithelial tissues, and can be modulated by both physiological and pathological stimuli. For example, changes in TJ protein expression or post-translational modifications (Bolinger et al., 2016; Günzel and Yu, 2013; Raleigh et al., 2011), TJ strand number or degree of branching (Claude and Goodenough, 1973; Saito et al., 2021), inflammatory stimuli (Ivanov et al., 2010; Luissint et al., 2016), or tissue mechanics (Stephenson et al., 2019; Varadarajan et al., 2022, 2019) can all modulate barrier function. Therefore, researchers often seek functional readouts to measure the permeability characteristics of epithelia. Traditionally, this is achieved with a combination of two key assays: transepithelial electrical resistance (TER) and tracer permeability assays (Matter and Balda, 2003) (Fig. 1A,B). TER measures ion permeability by generating an electrical current across an epithelial monolayer grown on a permeable support and measuring the resistance (Srinivasan et al., 2015). TER is useful for measuring the magnitude of ion flux across an epithelial monolayer through the pores generated by claudins (Shen et al., 2011; Suzuki et al., 2014). Thus, epithelia with a high proportion of barrier-forming claudins or perturbations that promote improved barrier function will have higher TER (i.e. reduced ion permeability). In contrast, tracer permeability assays measure size-selective permeability characteristics of an epithelial monolayer by adding fluorescent tracer macromolecules of different sizes [for example, fluorescein dye (332 Da), non-charged polyethylene glycol (PEG) oligomers (such as PEG₃, which has a radius of 2.8 Å, or PEG₂₅, which has a radius of 7 Å) or fluorescently labeled dextrans (4 kDa, 10 kDa, 40 kDa, 70 kDa or 250 kDa)] to the apical side of an epithelial monolayer grown on a permeable support and measuring how much of the tracer passes to the basal side (Matter and Balda, 2003; Van Itallie et al., 2008). In this way, increased apparent permeability (P_app), measured by increased passage of tracer to the basal side, indicates increased macromolecular permeability.

Together, TER and tracer permeability assays have been the standard approaches to measure the permeability characteristics of cultured epithelial monolayers. However, these methods have a number of limitations: (1) they are global measures that average barrier function across many cells and are not sensitive enough to detect subcellular changes in barrier function within a monolayer; (2) they do not provide spatial information about the location of sites of increased permeability; (3) although they can be carried out over a time course, they do not provide information about rapid, dynamic changes in the barrier; and (4) they are not compatible for use with live developing embryos such as Xenopus, zebrafish or mouse embryos.

Several recent methods have sought to overcome these limitations, particularly with respect to identifying the location of barrier breaches within an epithelial tissue. For example, ‘sandwich’ assays allow visualization of local barrier function at fixed time points (Dubrovskyi et al., 2013; Ghim et al., 2017; Richter et al., 2016) (Fig. 1C). These sandwich assays rely on the high-affinity interaction of avidin and biotin, and they improve upon previous methods using sulfo-N-hydroxysuccinimide-biotin (sulfo-NHS-biotin), which binds to primary amines that can be detected by avidin probes (Merzdorf et al., 1998; Schumann et al., 2012; Tamura et al., 2008). Sandwich assays employ, for example, fluorescently labeled avidin applied to the basal side and fluorescent biotin applied to the apical side – sandwiching the epithelial monolayer. If there is a leak in the barrier, the fluorescent biotin crosses the TJ and is captured by the avidin, and this local fluorescence intensity can be visualized via microscopy at the experimental endpoint. Sequential rounds of washing the monolayer and adding a different fluorescent biotin can be carried out to visualize snapshots in time, thus revealing changes in the barrier over time. Variations on the sandwich assay approach have shown that barrier function to macromolecules is not uniform across the tissue; barrier breaches open and close over time, and changes in barrier function can occur in response to mechanical cues (Dubrovskyi et al., 2013; Ghim et al., 2017; Richter et al., 2016). However, although able to capture local breaches in the barrier, these assays still only reveal snapshots in time. The low temporal resolution of these techniques makes it impossible to investigate the events immediately preceding and following leaks in barrier function. Additionally, multiple fluorescence channels must be used for sequential labeling, reducing the channels available to monitor other fluorescent probes of interest. Of note, a new modification of the sandwich assay uses a Förster resonance energy transfer (FRET)-based approach to allow for live imaging (Richter et al., 2022). Traditional sandwich assays provide an advantage over the standard TER and tracer permeability assays in that they can reveal the location of barrier breaches within an epithelial tissue. However, the lack of an assay that is sensitive enough to detect when and where breaches in barrier function happen, how long they last and whether they repeat at the same sites remains a limitation for researchers investigating dynamics of barrier function.

We previously developed a method that meets this need: the zinc-based ultrasensitive microscopic barrier assay (ZnUMBA) (Fig. 1D). Using ZnUMBA in Xenopus laevis embryos has allowed us to visualize naturally occurring, transient leaks in barrier function and correlate these with local loss of TJ proteins, local RhoA activation and actomyosin-mediated repair of barrier function (Stephenson et al., 2019). Furthermore, we have recently used ZnUMBA to demonstrate that mechanosensitive channel-mediated Ca²⁺ influx is required to maintain barrier integrity by repairing local leaks that occur when junctions elongate (Varadarajan et al., 2022), and to identify the RhoGEF that activates RhoA flares (Chumki et al., 2022). Here, we further demonstrate the usefulness of ZnUMBA in Xenopus embryos (Fig. 2A) and expand applications of ZnUMBA to other systems, including zebrafish embryos (Fig. 2B) and cultured epithelial monolayers (Fig. 2C).

ZnUMBA relies on a commercially available, small, cell-impermeable dye called FluoZin-3 (FZ3, 847 Da; F24194, Thermo Fisher Scientific). FZ3 is fluorogenic, increasing in fluorescence intensity more than 50-fold when bound to Zn²⁺ (65 Da); as such, there is little background signal until the dye is bound to its target. In ZnUMBA, medium containing ZnCl₂ is applied to the apical side of the epithelium, and FZ3 is introduced to the basal side of the epithelium (Fig. 1D). When localized TJ breaches occur, Zn²⁺ and FZ3 interact, resulting in bright, localized increases in FZ3 fluorescence – specifically at the site of TJ damage. Once the leak is repaired and barrier function is reinstated, FZ3 fluorescence returns to background levels. Therefore, ZnUMBA can detect local changes in barrier function and examine the dynamics of barrier function with high temporal and spatial resolution using conventional confocal microscopy.

Using Xenopus embryos, we have previously demonstrated that ZnUMBA can report global changes in barrier function (Stephenson et al., 2019). Moreover, we have shown that ZnUMBA can detect local, short-lived leaks in barrier function, including naturally occurring leaks within elongating junctions as well as leaks induced by laser injury of the junction (Stephenson et al., 2019). In both cases, leaks are followed by robust accumulation of active RhoA and actomyosin, which are required for sustained barrier reinforcement (Stephenson et al., 2019). Interestingly, inhibiting the actomyosin response does not result in leaks of longer duration, but in leaks that repeat at the same location over time (Stephenson et al., 2019). In order to more closely investigate this phenomenon, we injected FZ3 into the blastocoel of gastrula-stage Xenopus embryos, as previously described (Fig. 2A) (Stephenson et al., 2019), but we refined our laser injury technique and used ZnUMBA to analyze the dynamics of barrier restoration and actin dynamics with higher spatiotemporal resolution (Fig. 3).

Following laser injury, we observed a sharp spike in FZ3 fluorescence at the site of the injury as well as less intense signal increase along the length of the junction (Fig. 3A,A′). Examining z-views revealed that the FZ3 fluorescence at the site of the injury was more apical than the signal along the length of the junction (Fig. 3A″), leading us to speculate that the lateral signal represents diffusion of FZ3 from the injury site, rather than disruption of TJ function along the length of the injured junction. Furthermore, we observed that FZ3 intensity peaked 10 s following laser injury and began to diminish 30 s post injury, whereas F-actin only reached ∼15% of its maximum intensity by 30 s and did not peak until ∼75 s after injury (Fig. 3B). Similarly, junction contraction initiated ∼30 s after injury (Fig. 3B). Thus, while actomyosin-mediated contraction appears to be important for sustained barrier restoration (Chumki et al., 2022; Stephenson et al., 2019; Varadarajan et al., 2022), these data indicate that a still unknown mechanism might be able to temporarily seal the barrier prior to full contraction of the junction.

Many, but not all, RhoA flares are accompanied by an apical plasma membrane protrusion originating from one cell and protruding over its neighbor (Stephenson et al., 2019). We speculate that these membrane protrusions are bleb-like expansions of the plasma membrane resulting from loss of membrane–cortex attachment at the site of TJ breaks. Given the above results, we wanted to test whether the membrane protrusion plays an essential role in temporarily sealing the barrier. To this end, we injured junctions by laser ablation and observed the plasma membrane with mCherry–farnesyl along with FZ3. Of 18 injured junctions from seven embryos, only five (27.8%) exhibited a rounded membrane protrusion similar to those observed in naturally occurring RhoA flares. Of these five, three protrusions began expansion following the peak of FZ3 fluorescence (Fig. 3C), whereas the expansion of the other two protrusions coincided with the peak of FZ3 signal (Fig. 3D). The remaining thirteen junctions showed evidence of membrane reorganization as the junction contracted, likely the result of bunching or folding of excess membrane generated by the contraction (Fig. 3E). Additionally, we observed that some FZ3 signal remained confined to a bright spot and appeared to dissipate in vesicles (Fig. 3E) in eight of 18 junctions (44.4%), indicating that damaged junctions might be removed via endocytosis and/or that endocytosis is needed to stabilize the contracted junction (Cavanaugh et al., 2020). The low proportion of bleb-like membrane protrusions induced by laser injury leads us to speculate that membrane protrusions in naturally occurring RhoA flares originate from an imbalance in actomyosin on either side of the damaged junction. Such an imbalance is less likely to be present in junctions randomly selected for laser-induced junction injury than in those damaged during cell shape change. Based on these data, we conclude that the membrane protrusion is unlikely to be required for temporary sealing of the barrier (Fig. 3B).

Claudin strands are dynamic, capable of breaking and reannealing to nearby strands (Sasaki et al., 2003), and this activity is independent of their ability to associate with actomyosin (Van Itallie et al., 2017). Thus, the temporary seal could be the result of simple strand reorganization that creates a weak barrier that is easily broken if not reinforced by actomyosin-mediated contraction (Fig. 3F). Indeed, six of 18 junctions exhibited multiple peaks of FZ3 fluorescence over time (Fig. 3E), indicating breaking of the temporary seal prior to junction reinforcement. Ultimately, super-resolution live imaging of TJ strands would be required to test this hypothesis. Nonetheless, these experiments highlight the utility of ZnUMBA when coupled with high-resolution live imaging, allowing us to observe the dynamic nature of barrier breaches and generate new hypotheses about their repair.

ZnUMBA requires that embryos be incubated in micromolar to millimolar concentrations of Zn²⁺, which has the potential to interfere with normal cellular function and/or development. In order to test how exposure to Zn²⁺ impacts overall X. laevis embryo viability, we exposed gastrula-stage Xenopus embryos to varying concentrations of ZnCl₂ and assessed gross defects at 1, 3 and 24 h post exposure (Fig. S1). After 1 h and 3 h of ZnCl₂ exposure, abnormalities were rare and appeared at rates similar to those in controls. However, 24 h of exposure to ZnCl₂ did substantially impact development and embryo viability, particularly at the high end of the concentration range tested. Thus, the appropriate ZnCl₂ concentration and duration of exposure should be empirically determined for different model systems.

In order to test the applicability of ZnUMBA in other model systems, we turned to zebrafish embryos. First, early zebrafish embryos were exposed to Danieau's medium supplemented with different ZnCl₂ concentrations (Fig. S2). Assessing viability at different stages during early development allowed us to identify an optimal condition of 750 µM ZnCl₂ in which viability largely resembled the control condition (Danieau's medium only) (Fig. S2, red rectangle). Next, we tested the sensitivity and functionality of the ZnUMBA assay in early zebrafish development by artificially inducing barrier breaches through pharmacological or genetic means. We first tested the effect of EGTA addition. Zebrafish embryos were injected with FZ3 into the interstitial fluid, facing the basal side of the outermost epidermal layer (Fig. 2B), and at 4 h post fertilization (hpf), 10–20 mM EGTA (in Ca²⁺-free medium) was added. The majority of embryos started to become leaky within 90 min after EGTA addition, as indicated by the increase in FZ3 intensity normalized to Alexa Fluor 647–dextran (10 kDa) intensity (Fig. 4A,A′ FZ3/Dex ratio). Dextran intensity was used for normalization as a volume marker of interstitial fluid abundance. ZnUMBA, therefore, allows for monitoring of junctional leakage in zebrafish embryos at high spatiotemporal resolution (see propagation of barrier breaches in Movie 1, EGTA lower panel). Next, we sought to induce barrier breaches through a genetic approach, utilizing the poky mutant, in which maternal-zygotic (MZ) mutant embryos fail to form a functional epidermal barrier due to continued cell cycle and lack of differentiation (Fukazawa et al., 2010). Indeed, when we performed ZnUMBA in MZpoky mutant embryos, the FZ3:dextran fluorescence ratio increased within 80 min after starting the assay (Fig. 4B,B′), indicating that the MZpoky mutants did not have a functional epidermal barrier, consistent with previous reports (Fukazawa et al., 2010), and the embryos gradually started lysing ∼5–6 hpf. Taken together, these results demonstrate that ZnUMBA can successfully detect barrier breaches in zebrafish embryos due to acute (EGTA) as well as gradually acquired (MZpoky mutant) junctional leakages.

Finally, we developed an approach to utilize ZnUMBA with monolayers of cultured epithelial cells (Higashi et al., 2023) (Fig. 2C). For this purpose, we used Madin–Darby canine kidney (MDCK) II cells, which are one of the most commonly used epithelial cell lines for evaluating the structure and function of the TJ. Since wild-type (WT) MDCK II cells express cation-permeable claudin-2 (Cldn2) (Amasheh et al., 2002; Furuse et al., 2001; Tokuda and Furuse, 2015), we used Cldn2-knockout (KO) MDCK II cells (Saito et al., 2021) (Figs S3 and S4). To observe cell sheets with a confocal microscope, the cells were cultured on the bottom surface of a Transwell filter (Fig. 2C). The bottom-surface culture setup did not inhibit barrier formation compared to the normal top-surface culture setup, although the cultures took 1–2 d longer to reach the full barrier function, and the TER value at plateau was slightly higher than that in the top-surface culture conditions (Fig. S5). After the cells formed a cell sheet, apical and basal media were replaced with ZnCl₂- and FZ3-containing solutions, respectively (Fig. 2C). Exposure of the Cldn2-KO MDCK II cell monolayer to 2 mM ZnCl₂ from the apical side did not significantly affect the TER value within 1 h (Fig. S5C).

To test whether ZnUMBA can detect paracellular permeation of ions, WT MDCK II cells were mixed with Cldn2-KO cells (Fig. 5A). The TER values of WT MDCK II cells and Cldn2-KO cells were ∼70 Ω•cm² and ∼2500 Ω•cm², respectively (Fig. S4). To distinguish WT cells from Cldn2-KO cells, nuclear localization signal-conjugated green fluorescent protein (GFP–nls) was stably expressed in the WT cells (Fig. S4). Before the addition of FZ3, GFP–nls alone was visualized in the green channel (Fig. 5A). Upon addition of FZ3, the FZ3 intensity at the cell–cell junctions between WT cells labeled with GFP–nls drastically increased, whereas the signal between Cldn2-KO cells remained modest (Fig. 5A), indicating that ZnUMBA can detect global changes in barrier function in MDCK II cells. Next, to evaluate whether ZnUMBA can be used to detect local breaches, we used angulin-1 (Ang1; also known as LSR)-KO cells. Ang1 localizes specifically at tTJs in MDCK II cells (Higashi and Chiba, 2020; Higashi and Miller, 2017), and loss of Ang1 results in impaired barrier function at tTJs (Masuda et al., 2011; Sugawara et al., 2021). We established Ang1-KO cells from the Cldn2-KO clone and labeled them with GFP–nls (Figs S3 and S4). When the Ang1/Cldn2-double KO (dKO) cells were mixed with Cldn2-KO cells, FZ3 signal appeared preferentially at tricellular contacts of Ang1/Cldn2-dKO cells compared with those of Cldn2-KO cells, where the signal remained modest (Fig. 5B). Similar results were obtained when Ang1/Cldn2-dKO cells were mixed with Cldn2-KO cells labeled with nuclear localization signal-conjugated mCherry (mCherry–nls), which made it easier to see the preferential increase in FZ3 signal at tricellular contacts formed by cells with Ang1-KO, as there was no mCherry–nls signal in those cells (Fig. 5C). These data indicate that ZnUMBA can detect local TJ barrier dysfunction in cultured epithelial monolayers.

We also tested ZnUMBA in combination with live-imaging of MDCK II cells. In the Cldn2-KO cell sheet, ZnUMBA could detect naturally occurring leaks at cell–cell boundaries (Fig. 6A). In contrast with the spot-like FZ3 signals detected at naturally occurring leaks in Xenopus embryos (Stephenson et al., 2019), the signal appeared to spread along the cell–cell boundaries across the vertices in MDCK II cell sheets. This difference might be because the Zn²⁺ ions can diffuse and spread in intercellular spaces more easily in MDCK II cell sheets than in intact Xenopus embryos. When Rhodamine B isothiocyanate (RITC)–dextran (10 kDa) was added to the basal medium, no obvious increase was observed in the RITC–dextran signal at the sites where FZ3 signal was detected (Fig. 6B; Movie 2), indicating that the increase in FZ3 signal was not caused by the widening of paracellular space. The locally increased FZ3 signals disappeared within minutes, suggesting that there is a molecular mechanism to detect and repair TJ breaks in MDCK II cells. Taken together, these data indicate that ZnUMBA is applicable for use in cultured epithelial cells and can detect leaks that are either naturally occurring or caused by loss of TJ components.

Here, we build upon recent publications utilizing the ZnUMBA method we developed (Chan et al., 2019; Chumki et al., 2022; Higashi et al., 2023; Stephenson et al., 2019; Varadarajan et al., 2022) to further explain ZnUMBA, demonstrate its capabilities and provide expanded examples of its usefulness. In gastrula-stage Xenopus embryos, we show that ZnUMBA can be used, in combination with laser injury of the junction, to measure the dynamics of barrier restoration. Additionally, we demonstrate that by quantifying the FZ3:dextran ratio, ZnUMBA can be used in zebrafish embryos to reveal barrier breaches with high spatiotemporal precision and different dynamic patterns; rapid barrier breaches were detected by global application of EGTA, and gradual leakage was monitored by genetically disrupted barrier function in MZpoky mutants. Finally, in cultured MDCK II epithelial cells, we demonstrate that Cldn2-KO MDCK II cells (Saito et al., 2021) are suitable for detecting transient local barrier leaks with ZnUMBA, whereas the WT MDCK II cell line is quite leaky, causing significant background signal. Furthermore, we show that leaks at tTJs are detected in an Ang1/Cldn2-dKO MDCK II cell line. Taken together, these data underscore the utility and capabilities of ZnUMBA for detecting dynamic barrier breaches in a variety of systems.

The existing methods to analyze barrier function generally involve analysis of cultured epithelial cells, since they allow for easy perturbation and quantification of junctional properties; however, aspects of TJ structure and remodeling in cultured cells might not be fully representative of intact in vivo systems. Recently, variations on tracer permeability assays have been applied in order to monitor barrier function in vivo, aiming to measure local and/or dynamic TJ permeability in mouse models or even in human patients with intestinal barrier function diseases. For example, in mice orally gavaged with fluorescein isothiocyanate (FITC)–dextran then later sacrificed, serum FITC fluorescence can be measured as a global readout; moreover, overlays of FITC fluorescence signal in a colon section and Hematoxylin and Eosin staining of the same section can reveal information about local sites of barrier damage (Cario et al., 2007; Furuta et al., 2001). In a more precise but technically challenging in vivo murine method called exteriorized intestinal loop (iLoop), a well-vascularized exteriorized intestinal segment is made, the iLoop lumen is then injected with FITC–dextran, and intestinal permeability is assessed after an incubation period by quantifying fluorescence in blood serum (Boerner et al., 2021). iLoop offers an advantage over the oral gavage delivery of fluorescent tracer in that it allows researchers to study the permeability properties of specific localized areas in the intestine (terminal ileum or proximal colon) that are commonly involved in inflammatory bowel disease (IBD). Additionally, various reagents can be injected into the iLoop lumen, such as chemokines, cytokines, bacterial pathogens, toxins, antibodies and therapeutics, to examine how these affect barrier function (Boerner et al., 2021). Perhaps in the future, iLoop could be combined with ZnUMBA to measure local and dynamic changes in TJ permeability. In human patients with IBD, confocal laser endomicroscopy utilizes intravenous fluorescein injection and a confocal fluorescence microscope incorporated in an endoscope in order to visualize the leakiness of intestinal epithelia. This technique has shown that there are particularly leaky areas of the intestinal epithelium in vivo, and these correlate with disease severity (Lim et al., 2014; Rasmussen et al., 2015).

In developing Xenopus, zebrafish or other externally developing model organism embryos, simple epithelia coat the surface of the embryo and are easily accessible for microscopy, making them well suited for live imaging of barrier function. In Xenopus, we have attempted to visualize the penetration of fluorescent molecules such as Alexa Fluor 488–dextran (3 kDa) (Reyes et al., 2014) or fluorescein (332 Da) (Higashi et al., 2016) across TJs by live imaging. Reyes et al. (2014) found that Anillin knockdown results in increased depth of penetration of Alexa Fluor 488–dextran; however, this likely reflects cell shape change (apical doming) caused by disruption of junctional actomyosin rather than increased permeability, because dextran did not detectably penetrate to the basal compartment (Reyes et al., 2014). Higashi et al. (2016) examined epithelial barrier function during cytokinesis by directly imaging fluorescein applied to the apical surface of Xenopus embryos. Under control conditions, it was not possible to detect the tracer beyond the TJ, even at the contractile ring, a site that promotes a major cell shape change, challenging junction integrity (Hatte et al., 2018; Higashi et al., 2016).

The variations on tracer permeability assays used for monitoring barrier function in vivo discussed above all suffer from a lack of specific signal increase at sites of local barrier leaks. Thus, a local, transient leak in the barrier would be difficult to detect against the background signal. Therefore, with ZnUMBA, we sought to develop a more sensitive barrier assay with minimal background, in which a breach of the TJ results in strongly increased fluorescence. Advantages of ZnUMBA with respect to existing techniques include: (1) FZ3 is a fluorogenic dye, and its fluorescence intensity increases ∼50-fold specifically at barrier breach sites where FZ3 and Zn²⁺ come in contact; (2) ZnUMBA is a live microscopy-based assay that can provide quantitative information about barrier function both at the tissue scale and subcellular level; (3) ZnUMBA is capable of revealing localized, transiently-occurring barrier leaks, thus identifying spatially where barrier breaches occur with respect to cell morphology or tissue architecture; and (4) ZnUMBA provides high temporal resolution, detecting changes in barrier function over time.

Despite clear advantages, ZnUMBA also presents several limitations. First, embryo viability can be affected by high concentrations of ZnCl₂; however, we show that at lower concentrations of ZnCl₂, the assay works effectively to detect barrier breaches and does not cause phenotypic effects. Furthermore, exposing epithelia acutely (1 h or less) to Zn²⁺ when performing ZnUMBA reduces the possibility that Zn²⁺ could affect barrier function, as different studies have reported that Zn²⁺ can either enhance or diminish epithelial barrier function (Shao et al., 2017; Wang et al., 2013; Xiao et al., 2018). As ZnUMBA is adapted for use in other systems, it will be important to empirically determine an appropriate concentration range of ZnCl₂ and expose epithelial cells to the lowest dose of ZnCl₂ for the shortest amount of time possible. Second, ZnUMBA can suffer from steadily increasing background FZ3 fluorescence over time. In Xenopus embryos, we found that decreasing the ZnCl₂ concentration (1–2 mM), increasing the image acquisition exposure time and adding Ca²⁺/EDTA along with FZ3 to sequester endogenous Zn²⁺ in the blastocoel improved background fluorescence and reduced noise. Nevertheless, the assay is still subject to gradually increasing FZ3 fluorescence over time; therefore, a method of normalizing signal to background (such as using a fluorescently labeled dextran) is required for appropriate data interpretation.

Another limitation of ZnUMBA is the potential impact of the expression of cation-specific pore-forming claudins on the assay. Some claudins, including claudin-2, are pore-forming claudins and allow a high volume of cations to cross the barrier, while restricting anions and other molecules. In the case of gastrula-stage Xenopus embryos, the most highly expressed claudins are claudin-6, claudin-7 and claudin-4 (Session et al., 2016), none of which are known cation pores (Günzel and Yu, 2013). Interestingly, in the teleost fish group – especially in Takifugu rubripes, but also in zebrafish – there has been extensive expansion of the claudin gene family, and several claudin genes have been identified that have no mammalian ortholog (Loh et al., 2004). This enrichment of claudin genes might have been due to an increased necessity for osmoregulation in fish species (Loh et al., 2004; Siddiqui et al., 2010). In gastrula-stage zebrafish embryos, the majority of highly expressed claudin genes (see zfin.org) encode barrier-forming claudins, such as Claudin-8.2 (also known as Claudin 8-like; mainly barrier-forming), Claudin-b (orthologous to mammalian claudin-4; barrier-forming), and Claudin-e (also known as Cb84) and Claudin-f (most closely related to claudin-3 and claudin-4; mainly barrier-forming), with a smaller number of predominantly pore-forming claudins (such as Claudin-7b) (Günzel and Yu, 2013; Kollmar et al., 2001; Siddiqui et al., 2010).

As we found with MDCK II cells, expression of cation-selective pore-forming claudins can lead to background signal that impedes detection of local breaches. There are two strains of MDCK cells that have similar TJ morphology but differ in the profile of claudins they express (Furuse et al., 2001). MDCK I cells do not express claudin-2, whereas MDCK II cells do (Furuse et al., 2001; Tokuda and Furuse, 2015). MDCK II cells exhibit fewer clonal differences in junctional properties from one population to another, so they are generally preferred by TJ researchers (Matter and Balda, 2003). We found that WT MDCK II cells are quite leaky and not amenable for use with ZnUMBA. However, our data reveal that knocking out claudin-2 in MDCK II cells makes them suitable for analysis by ZnUMBA. It is expected that ZnUMBA would be applicable to MDCK I or other cultured cell lines that do not express cation-selective pore-forming claudins without any genetic engineering.

The prevalence of cation-selective pore-forming claudins will result in a progressive global increase in FZ3 signal via the pore pathway. Global increase in signal might also be indicative of global defects in TJs, such as those caused by EGTA addition or genetic perturbations. We speculate that localized, transient increases in FZ3 fluorescence intensity are the result of the leak pathway (Monaco et al., 2021), the non-specific flux of molecules across the TJ barrier due to transient breaks in TJ strands. However, some claudin pores can open and close, similar to other ion channels (Weber and Turner, 2017). Thus, it is possible that local increases in FZ3 signal represent the regulated opening of cation pores rather than disruption of the strand network itself. As ZnUMBA cannot definitively distinguish between the leak and pore pathways, researchers should use knowledge of their model system and experimental context when interpreting ZnUMBA results.

Currently, ZnUMBA uses a commercially available dye, FZ3, applied to the basal side and ZnCl₂ applied to the apical side of the barrier. These widely available reagents and compatibility with conventional confocal microscopy make this assay accessible for potential use in diverse systems including developing model organisms, organ-specific setups like exteriorized intestinal loops, 3D organoids, and 2D cell epithelial or endothelial cell culture. Additionally, because changes in TJ permeability underlie multiple disease pathologies, and bacterial pathogens often target TJs, ZnUMBA could be used to better understand the mechanisms underlying these pathologies and to aid in drug discovery research by identifying compounds that modulate subcellular dynamics of barrier function. In the future, variations on the ZnUMBA method might be developed to expand the fluorescent dye color palette available or to measure size-specific permeability dynamics. Other fluorogenic dyes that increase significantly in fluorescence upon binding their target could be used instead of FZ3. For example, fluorogenic dyes that are ligands for self-labeling tags such as Halo tags (Zeng et al., 2019) would be a good option. In this case, the fluorogenic Halo dye would be added to the apical side of the epithelium, and a Halo-tagged basolateral transmembrane protein would be expressed in the cells, or recombinant Halo-tagged protein would be added to the basal side.

The development of ZnUMBA and its application in several systems including Xenopus embryos (Chumki et al., 2022; Stephenson et al., 2019; Varadarajan et al., 2022), mouse embryos (Chan et al., 2019), zebrafish embryos (this paper) and cultured epithelial monolayers (this paper) (Higashi et al., 2023) represents a powerful tool for detecting global and local dynamic changes in barrier function with spatiotemporal precision. We recommend that, going forward, researchers should not only use traditional global, averaged measures of barrier function (TER, tracer permeability assays), but also measure local, dynamic changes in barrier function that can be detected with ZnUMBA. ZnUMBA has already revealed new information about subcellular dynamics of barrier function. New variations of ZnUMBA to meet different experimental needs should allow for broad application of this live imaging barrier assay to various epithelial contexts and will open the door to future discoveries.

Adult WT or albino X. laevis frogs were purchased from Nasco or Xenopus 1. Female frogs were injected with human chorionic gonadotropin (HCG) to induce them to lay eggs (Wlizla et al., 2018). Male frogs were used for acquisition of testes for sperm preparations (Wlizla et al., 2018). Frogs were housed in a recirculating tank system (Tecniplast), which monitors water quality parameters (temperature, pH and conductivity) to ensure safe and consistent water quality for an optimal environment for frog health. Daily health and maintenance checks were performed by Animal Care Staff. All studies strictly adhered to the compliance standards of the US Department of Health and Human Services Guide for the Care and Use of Laboratory Animals, and were approved by the University of Michigan's Institutional Animal Care and Use Committee. A board-certified laboratory animal veterinarian oversees our animal facility.

Laser injury experiments were performed on WT or albino X. laevis embryos that had been microinjected with mRNA for Lifeact–mRFP (Bement et al., 2015), mCherry–farnesyl (Reyes et al., 2014), or Lifeact–miRFP703 (Yamamoto et al., 2021) at the 1-, 2- or 4-cell stage and were allowed to develop to gastrula stage at 15°C. Prior to imaging, 10 nl of 1 mM FluoZin-3 (FZ3, 847 Da; F24194, Thermo Fisher Scientific), 100 mM CaCl₂, and 100 mM EDTA were microinjected into the blastocoel of Nieuwkoop and Faber stage 10–11 X. laevis embryos. EDTA was used to reduce baseline levels of FZ3 fluorescence from endogenous Zn²⁺, and equimolar Ca²⁺ was added to offset the potential effects of Ca²⁺ chelation by EDTA. Following microinjection, embryos were allowed to heal for a minimum of 5 min before being mounted in a slide containing 1 mM ZnCl₂ in 0.1×MMR (Wlizla et al., 2018). Note: while we find it beneficial to reduce the time embryos are exposed to ZnCl₂, we have found that embryos can be microinjected with FZ3 and stored for several hours prior to imaging without negative consequences.

Embryos were imaged using an Olympus FV1000 scanning confocal microscope with a 60× PlanApo objective (NA 1.4) and mFV10-ASW software. Six z-slices with a step size of 0.6 µm were collected every 3.5 s. A small circular region of interest (ROI) approximately the width of the junction (0.4–0.6 µm diameter) was placed roughly midway between two vertices, and a minimum of ten pre-injury frames were collected to establish a baseline. Injury was initiated manually by the user initiating the SIM scanner and 405 nm laser (100% power) and was stopped automatically after 5 s by the software. A maximum of three junction injuries were performed per embryo.

Quantification was performed in FIJI (https://imagej.net/software/fiji/) on summed z-projections with the assistance of custom macros (available upon request) and the LOI Interpolater tool in the Timelapse plugin. Every two to ten frames, a 1 µm wide segmented line was drawn over the junction from vertex to vertex, and the LOI Interpolater tool was used to draw lines in the remaining frames. Junction length, as well as mean intensity over the line of interest, were measured for FZ3 and Lifeact–mRFP channels. FZ3 and Lifeact–mRFP intensities were normalized by dividing the mean intensities of an injured junction by the mean intensities of a reference junction from the same movie. Junction length changes were normalized by subtracting the average junction length of the ten frames prior to the injury from each junction length measurement.

Uninjected WT Xenopus embryos were allowed to develop to early gastrula stage at 15°C. Groups of 18–20 healthy embryos were incubated in 0.1×MMR plus the indicated concentration of ZnCl₂ at room temperature (RT). Embryos were photographed prior to ZnCl₂ exposure and after 1, 3 and 24 h of ZnCl₂ exposure. Abnormalities seen at gastrula stage (i.e. 1 h and 3 h) included small patches of lysing cells and exogastrulation. At the 24 h time point, embryos were classified as abnormal if they were notably developmentally delayed or had developed abnormal growths. Embryos were classified as dead/lysing if they had cell material outside their body or if development had stopped and they were whitish in color.

Embryo collection from Danio rerio and staging of embryos were carried out as previously described (Kimmel et al., 1995; Westerfield, 2007). Embryos from AB strain were used as a control for MZpoky mutant experiments, and AB or TL strains were used for EGTA experiments. For experiments, poky homozygous mutant fish (Fukazawa et al., 2010) were incrossed to gain maternal-zygotic poky mutants. Breeding of fish was performed at the Institute of Science and Technology Austria (IST Austria) zebrafish facility in line with local regulations and with the approval of the Ethic Committee of IST Austria regulating animal care and usage.

For mRNA transcription, SP6 mMessage mMachine Kit (Ambion) was used and mRNAs were injected via glass capillaries (30-0020, Harvard Apparatus; pulled by a needle puller P-97, Sutter Instruments) while mounted on a microinjection system (PV820, World Precision Instruments). Injections of 50 pg membrane-RFP (Iioka et al., 2004) with 0.2% Phenol Red at 1-cell stage were performed as previously described (Westerfield, 2007). For FZ3 injections, embryos were injected into the interstitial fluid with 100 µM FZ3 together with 0.25 mg/ml Alexa Fluor 647–dextran 10,000 MW (Invitrogen, #D22914) at 3 hpf (1k-cell stage).

Embryos were dechorionated. For live imaging of MZpoky mutant experiments, embryos were mounted in 0.3% low-melting-point agarose (Invitrogen) on glass-bottom dishes (MatTek).

For live imaging of EGTA treatments, embryos were embedded into 0.3% low-melting-point agarose already containing 10–20 mM EGTA in calcium-free medium (isotonic Ringer's solution without calcium: 116 mM NaCl, 2.9 mM KCl, 5 mM HEPES, pH 7.2) (Westerfield, 2007) and mounted in four-chamber glass-bottom dishes (MatTek) for imaging different conditions in parallel, namely, EGTA addition versus Danieau's medium [58 mM NaCl, 0.7 mM KCl, 0.4 mM MgSO₄, 0.6 mM Ca(NO₃)₂ and 5 mM HEPES (pH 7.2)] as a control.

High-resolution spinning disk confocal imaging was performed on a Zeiss Axio Observer Z1 microscope equipped with a 40×/1.2 W objective (C-APOCHROMAT, Korr UV-VIS-IR). Imaging was started at 4 hpf and within 30 min after addition of EGTA. The most superficial epidermal layer (enveloping layer, EVL) and deep cell interface were imaged by taking z-stacks between 10–25 µm at z-steps of 0.5–1 µm at acquisition times of 0.5–6 min.

FZ3 fluorescence signal was normalized to the interstitial fluid abundance via Alexa Fluor 647–dextran labeling. Maximum intensity projections of the EVL and deep cell interface were generated to monitor the region in proximity to the barrier breach. Then, a ROI with the size of 40×40 µm was selected, and FZ3 mean intensity was divided by Alexa Fluor 647–dextran mean intensity and plotted over time.

In order to determine an appropriate concentration of ZnCl₂, embryos were collected and exposed to different concentrations of ZnCl₂ starting from ∼3–4 hpf until 1–1.5 d post fertilization in either dechorionated or chorionated conditions. ZnCl₂ concentrations were 750 µM, 1 mM or 2 mM. At higher concentrations (2 mM ZnCl₂), a small fraction of embryos showed defects such as lysis or abnormal development, whereas at lower concentrations (750 µM ZnCl₂), embryos mimicked WT control viability rate.

Rat anti-claudin-2 monoclonal antibody (mAb, clone 2D7) (Saito et al., 2021) was previously described. Rat anti-occludin mAb (clone MOC37) (Saitou et al., 1997) and rabbit anti-angulin-1 polyclonal antibody (pAb) (Oda et al., 2020) were kindly provided by Professor Mikio Furuse (National Institute for Physiological Sciences, Okazaki, Japan). Rabbit anti-ZO-1 pAb (#61-7300), rabbit anti-claudin-2 pAb (#51-6100), rabbit anti-claudin-3 pAb (#34-1700), mouse anti-claudin-4 mAb (#32-9400), rabbit anti-tricellulin mAb (clone 54H19L38, #700191) and rabbit anti-ZO-2 pAb (#71-1400) were purchased from Thermo Fisher Scientific (MA, USA). Rat anti-ZO-1 (alpha+) mAb (clone R40.76; sc-33725), mouse anti-cingulin mAb (clone G-6; sc-365264) and goat anti-ZO-3 pAb (C-17; sc-11478) were from Santa Cruz Biotechnology (CA, USA). Rabbit anti-claudin-1 pAb (#18815) and rabbit anti-claudin-7 pAb (#18875) were from Immuno-Biological Laboratories (Gumma, Japan). Rabbit anti-occludin pAb (#LS-B2187) was from Lifespan (RI, USA). Rabbit anti-E-cadherin mAb (clone 24E10; #3195T) was from Cell Signaling Technology (MA, USA). Mouse anti-β-actin mAb (clone AC-15; #A1978) was obtained from Sigma-Aldrich (MO, USA). Rabbit anti-GFP pAb was from MBL (Nagoya, Japan). These primary antibodies were used at 1:500 dilution for immunofluorescence staining and at 1:5000 dilution for immunoblotting except for rat anti-claudin-2 and anti-occludin mAbs, which were used at 1:10 dilution of hybridoma cell culture supernatant.

For the secondary antibodies used in the immunofluorescence staining, Cy3-conjugated donkey anti-rat IgG pAb (#712-165-153), Alexa Fluor 488-conjugated donkey anti-rabbit IgG pAb (#711-545-152), and Alexa Fluor 647-conjugated donkey anti-rat IgG pAb (#712-605-153) were purchased from Jackson ImmunoResearch Laboratories (PA, USA) and used at 1:500 dilution. For immunoblotting, horseradish peroxidase (HRP)-linked sheep anti-mouse IgG pAb (#NA931V; GE Healthcare, CT, USA), HRP-linked goat anti-rabbit IgG pAb (#7074P; Cell Signaling Technology), HRP-linked goat anti-rat IgG pAb (#NA935V; GE Healthcare) and HRP-linked rabbit anti-goat IgG pAb (DAKO #P0449; Agilent, CA, USA) were used at 1:5000 dilution.

MDCK II cells were kindly provided by Professor Mikio Furuse (National Institute for Physiological Sciences, Okazaki, Japan) and were maintained in DMEM (Sigma-Aldrich) supplemented with 5% fetal bovine serum (FBS; Sigma-Aldrich) at 37°C in a 5% CO₂ incubator. Mycoplasma contamination was tested by PCR with a primer set of 5′-ACACCATGGAGCTGGTAAT-3′ and 5′-CTTC(A/T)TCGACTT(C/T)CAGACCCAAGGCAT-3′ using GoTaq DNA polymerase (Promega, WI, USA), and all cell clones were confirmed negative.

MDCK II cells cultured on coverslips or Transwell filters with 0.4 µm pore size (#3401; Corning, NY, USA) were fixed with 1% formaldehyde at RT for 15 min. The cell membrane was permeabilized with 0.2% Triton X-100 in phosphate-buffered saline (PBS) at RT for 10 min. Then, the cells were blocked with 2% bovine serum albumin (BSA) in PBS at RT for 30 min, and were incubated with primary antibodies in PBS containing 0.2% BSA at RT for 1 h, followed by secondary antibodies in PBS containing 0.2% BSA at RT for 30 min. The coverslips or filters were mounted with FLUORO-GEL II with DAPI (Electron Microscopy Sciences, PA, USA) and observed with an inverted laser scanning confocal microscope (FV1000; Olympus, Tokyo, Japan) with a 60× oil-immersion objective lens (UPlanSApo 60×; Olympus) at laser wavelengths of 405, 488 and 559 nm. Images were acquired with Fluoview ver. 4.2b software (Olympus) and processed with ImageJ 1.53a (National Institutes of Health, Bethesda, MD, USA) and Photoshop 2020 (Adobe).

MDCK II cells were lysed with sodium dodecyl sulfate (SDS) sample buffer [62.5 mM Tris-HCl, pH 6.8, 2% (w/v) SDS, 10% (v/v) glycerol, 5% (w/v) β-mercaptoethanol, 0.005% (w/v) Bromophenol Blue], and proteins were separated in SDS-polyacrylamide gels (Fujifilm WAKO, Osaka, Japan). The proteins were transferred to PVDF membranes (Immobilon; Merck, Darmstadt, Germany) and blocked with 5% (w/v) non-fat dry milk in Tris-buffered saline containing 0.1% (v/v) Tween-20 (TBST) at RT for 30 min. Then, the membrane was incubated with primary antibody diluted in TBST at 4°C overnight. The next day, the membrane was washed with TBST, incubated with secondary antibody at RT for 1 h, and developed using the enhanced chemiluminescence method (ECL prime; GE Healthcare). The images were recorded using LAS4000 (GE Healthcare) and processed with Photoshop. Full blots are shown in Fig. S6.

Cldn2-KO MDCK II cells were established as described previously (Saito et al., 2021). Specifically, 5×10⁴ MDCK II cells were transiently transfected with pSpCas9(BB)-2A-Puro (PX459) plasmids (plasmid 62988; Addgene, MA, USA) (Ran et al., 2013) encoding Cas9 and gRNAs for Cldn2 (sequences are shown in Fig. S3) using PEI-max (Polysciences, PA, USA). The next day, the cells were treated with 3 µg/ml puromycin (Sigma-Aldrich) for 1 d, and cells were sparsely reseeded onto a 10 cm dish. After 7–10 d, cell clones were picked, and aliquots of the cells were analyzed by genomic PCR using GoTaq DNA polymerase and specific primers (F_C2, 5′-GATGCCTTCTTGAGCCTGCTTGTGG-3′; R_C2, 5′-AGCACCTTCTGACATGATACAGTGC-3′). The amplicons were cloned into pGEM-T-easy vector (Promega) using the TA-cloning method, and the DNA sequences were analyzed (Macrogen Japan, Kyoto, Japan). The cell clones where both alleles of the claudin-2 gene locus were knocked out were chosen and further validated by immunofluorescence staining and immunoblotting. Likewise, Ang1/Cldn2-dKO cells were established using Cldn2-KO MDCK II cells as a parental cell line (gRNA sequences are shown in Fig. S3) and were screened using specific primers for the angulin-1 (LSR) gene locus (F1_A1, 5′-TGCTCCTCGTCCACGTTATTTCC-3′; F2_A1, 5′-CCTCTTTCTCAGCACCTTGTGCGC-3′; R1_A1, 5′-GGATCAGCAGATCCCGGCCTC-3′).

5×10⁴ MDCK II cells or Ang1/Cldn2-dKO cells were transfected using PEI-max with a plasmid (pCAG-nGFP) encoding GFP conjugated with 3× nuclear localization signals (DPKKKRKVRS) (Saito et al., 2021). The cell clones were selected with 200 µg/ml of G418 (Sigma-Aldrich) for ∼10 d and were screened by fluorescence microscopy. The vector encoding nuclear mCherry (mCherry–nls) was constructed by cloning the DNA fragments encoding IRES (internal ribosome entry site), mCherry and the 3× nuclear localization signals into the pCAG vector, in which the neomycin resistance gene was replaced with a puromycin resistance gene. 5×10⁴ Cldn2-KO cells were transfected with the vector using PEI-max and selected with 3 μg/ml of puromycin (Sigma-Aldrich) for 1 d. The cells were then screened by fluorescence microscopy.

To culture the cells on the bottom of the Transwell filters (#3401, Corning, NY, USA), the filters were placed upside-down in a clean container, and 3.3×10⁵ cells resuspended in 300 µl of DMEM (5% FBS) were placed on the top of the Transwell to cover the entire surface of the filter. Then, the container was placed in a CO₂ incubator and incubated at 37°C overnight. The next day, the filters were put into the wells of a 12-well plate filled with 2 ml of DMEM (5% FBS), and 500 µl of DMEM (5% FBS) was added into the filters. The medium was changed every day. When the TER values reached a plateau, the filters were washed with Z-medium [30% Hanks' balanced salt solution with Ca²⁺ and Mg²⁺ (HBSS, Gibco #14025-092; Thermo Fisher Scientific), 65% DMEM without Phenol Red (D1145; Sigma-Aldrich) and 5% FBS]. A glass-bottom dish was placed on an oil-immersion 60× lens (UPlanSApo) of an inverted laser-scanning confocal microscope (FV1000) and 200 µl of Z-medium containing 2 mM ZnCl₂ was put on the center of the dish. Then, 200 µl of 10 µM FZ3 solution in Z-medium containing 1 µM CaCl₂-EDTA was added into the filter, and the filter was placed on the glass bottom dish. Ca²⁺-EDTA chelates the excess Zn²⁺ in the basal compartment without sequestering Ca²⁺, which is required for the maintenance of cell–cell junctions. To evaluate the shape of paracellular space, 10 µg/ml Rhodamine B isothiocyanate (RITC)–dextran (10 kDa; R8881, Sigma-Aldrich) could be added to the basal compartment. The fluorescence signal of FZ3 was detected using a 488 nm laser and the filter set for GFP. The images were recorded using Fluoview ver. 4.2b software and processed with ImageJ and Photoshop.

MDCK II cells were cultured on the top or bottom surface of Transwell filters. The electric resistance between the apical and basal compartments was measured each day using a volt–ohm meter Millicell ERS-2 (EMD Millipore, MA, USA). The electric resistance of a blank filter was subtracted from each measurement, and then the culture area of the Transwell filter (1 cm²) was multiplied to calculate the unit area resistance.

The authors thank their respective lab members for feedback and helpful discussions. We thank the bioimaging and zebrafish facilities of IST Austria for their support.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.260668.reviewer-comments.pdf.