Claudin 4 knockout mice: normal physiological phenotype with increased susceptibility to lung injury

Abstract

Claudins are tight junction proteins that regulate paracellular ion permeability of epithelium and endothelium. Claudin 4 has been reported to function as a paracellular sodium barrier and is one of three major claudins expressed in lung alveolar epithelial cells (AEC). To directly assess the role of claudin 4 in regulation of alveolar epithelial barrier function and fluid homeostasis in vivo, we generated claudin 4 knockout (Cldn4 KO) mice. Unexpectedly, Cldn4 KO mice exhibited normal physiological phenotype although increased permeability to 5-carboxyfluorescein and decreased alveolar fluid clearance were noted. Cldn4 KO AEC monolayers exhibited unchanged ion permeability, higher solute permeability, and lower short-circuit current compared with monolayers from wild-type mice. Claudin 3 and 18 expression was similar between wild-type and Cldn4 KO alveolar epithelial type II cells. In response to either ventilator-induced lung injury or hyperoxia, claudin 4 expression was markedly upregulated in wild-type mice, whereas Cldn4 KO mice showed greater degrees of lung injury. RNA sequencing, in conjunction with differential expression and upstream analysis after ventilator-induced lung injury, suggested Egr1, Tnf, and Il1b as potential mediators of increased lung injury in Cldn4 KO mice. These results demonstrate that claudin 4 has little effect on normal lung physiology but may function to protect against acute lung injury.

alveolar epithelial cells (AEC) line the gas exchange surface of the lung and form a tight barrier to maintain fluid and solute homeostasis. Cell-cell contacts are maintained by intercellular junctions, which include tight junctions, adherens junctions, gap junctions, and desmosomes. Among these, tight junctions are localized at the most apical aspect of the epithelial/endothelial barrier and are known to have the most prominent role in regulating barrier integrity (58). Breakdown of alveolo-capillary barrier function leads to pulmonary edema and is implicated in the pathogenesis of acute respiratory distress syndrome (ARDS) (5).

Tight junctions are formed by transmembrane proteins, comprised of claudin family members, occludin, tricellulin, junctional adhesion molecules, and associated scaffolding proteins (e.g., zonula occludens-1) that connect transmembrane tight junction proteins to the actin cytoskeleton. Claudins comprise a family of 27 known members in mammals (51) and are generally thought to regulate paracellular permeability to ions and/or solutes. In AEC, claudins 3, 4, and 18 are highly expressed, with lower expression of claudins 5 and 7. Claudin 4 is also expressed in airway epithelial cells but not in lung endothelial cells or other distal lung cells (8, 37, 41, 57, 74). Among other organs, kidney, colon, pancreas, and submandibular gland also express claudin 4 (32, 36, 50, 57, 59).

With a few exceptions, claudin 4 is generally considered to function as a paracellular sodium barrier (10, 11, 72). Conflicting results suggesting that claudin 4 functions as a paracellular chloride channel in kidney collecting duct cell lines have been explained by cell type-dependent differences in the partner claudins with which claudin 4 interacts (35, 36). Claudin 4 has been shown to interact with another claudin 4 molecule in adjacent cells but not with claudins 1, 3, or 5 (13), whereas possible interactions with claudin 18 are unknown. In rat primary AEC, treatment with epidermal growth factor resulted in higher transepithelial resistance (R_t), short-circuit current (I_sc), and Na,K-ATPase activity, which correlated with lower claudin 3 and higher claudin 4 protein expression (8). Overexpression of claudin 4 in rat primary AEC resulted in higher R_t and no changes in I_sc, ionic permeability ratio of sodium to chloride (P_Na/P_Cl), and permeabilities to 0.6- and 10-kDa solutes (52). Conversely, suppression of both claudins 3 and 4 by Clostridum perfringens enterotoxin (CPE) binding domain or claudin 4 knockdown with siRNA resulted in lower R_t of primary cultured AEC monolayers, whereas permeability to a 40-kDa molecule remained unchanged (80). Suppression of claudins 3 and 4 by CPE binding domain in vivo did not result in changes in lung water (80) but instead correlated with lower alveolar fluid clearance (AFC) and higher ventilator-induced lung injury (VILI) in mice. Mechanisms underlying these associations are unknown (61, 80).

To directly evaluate the role of claudin 4 in regulating alveolar epithelial barrier properties and fluid homeostasis in vivo, we generated claudin 4 knockout (Cldn4 KO) mice. Cldn4 KO mice showed surprisingly minimal physiological lung impairment, whereas increased susceptibility to experimental models of acute lung injury was noted. These results demonstrate that claudin 4 has unexpectedly little effect on normal alveolar homeostasis but may confer protection against acute lung injury.

MATERIALS AND METHODS

Generation of Cldn4 KO mice.

All animal protocols were approved by the Institutional Animal Care and Use Committee of the University of Southern California. A conditional (floxed) targeting vector for the claudin 4 gene (Cldn4) was generated using bacterial recombineering as described (45, 48) (also refer to http://web.ncifcrf.gov/research/brb/recombineeringInformation.aspx). Briefly, an 11.7-kb genomic fragment encompassing Cldn4 was retrieved from a bacterial artificial chromosome (BAC) clone including the Cldn4 locus [BAC clone derived from RPCI-22 (129S6/SvEvTac) mouse BAC library, Children's Hospital Oakland Research Institute, Oakland, CA]. LoxP sites were introduced on both sides of the single exon Cldn4 gene at −526 and +2,014 bp relative to the transcription start site of Cldn4 (Fig. 1). A flippase recognition target (FRT)-flanked phosphoglycerokinase promoter-neomycin resistance (PGK-Neo^R) cassette was coinserted with the loxP site located downstream of Cldn4 to enable positive selection of embryonic stem (ES) cell clones, whereas a Herpes simplex virus thymidine kinase (HSV-tk) cassette for negative selection was placed at the 3′ flank of the genomic fragment. The targeting vector was linearized and introduced into W4 ES cells by electroporation, and ES cells were selected with G418 and gancyclovir. Positive clones were identified by Southern blot using flanking probes, and normal karyotypes were verified before host blastocyst injections. Mating of germ-line transmitting chimeras to FLPeR females [strain 129S4/SvJaeSor-Gt(ROSA)26Sor^tm1(FLP1)Dym/J, stock no. 003946, Jackson Laboratories, Bar Harbor, ME] removed the FRT-flanked PGK-Neo^R selection marker by Flp/FRT recombination (20). Offspring harboring the floxed Cldn4 allele without PGK-Neo^R were bred to CMV-cre mice [strain BALB/c-Tg(CMV-cre)1Cgn/J, stock no. 003465, Jackson Laboratories] to delete the Cldn4 gene by Cre/loxP recombination (63). Mice harboring the Cldn4 KO allele were bred once to 129S6/SvEvTac mice to select away the CMV-cre transgene. Heterozygous mice were intercrossed to generate a ubiquitous Cldn4 KO line. A parallel wild-type (WT) line that served as control with the same mixed genetic background was also generated.

Fig. 1.

Generation and confirmation of claudin 4 knockout mice (Cldn4 KO). A: strategy for Cldn4 gene targeting. A loxP site was introduced 5′ of the single exon of Cldn4, followed by introduction of another loxP site plus a flippase recognition target (FRT)-flanked PGK-Neo^R-positive selection marker 3′ of the gene. After homologous recombination in embryonic stem (ES) cells and generation of germ-line-transmitting chimeras, PGK-Neo^R was removed in vivo using Flp/FRT recombination. The Cldn4 gene was deleted by Cre/loxP recombination to generate mice with complete knockout of Cldn4. B: deletion of Cldn4 was verified by RT-PCR using cDNA from total lung RNA. The expected 270-bp PCR product was detected in wild-type (WT) lung RNA samples incubated with reverse transcriptase (+RT) but not without (-RT), whereas no product was observed in any samples from Cldn4 KO lungs. Neg is no template control. C: hematoxylin and eosin staining of lung sections did not show morphological differences or evidence of lung edema. Scale bars = 50 μm.

Histological analysis.

Lungs were cleared of blood by cardiac perfusion with PBS, inflated to 20 cmH₂O with 4% paraformaldehyde, fixed overnight at 4°C, embedded in paraffin, cut into 5-μm sections, and stained with hematoxylin and eosin (H&E). Lung injury was scored according to the official American Thoracic Society Workshop Report (49).

Plasma Na and K measurement.

Plasma Na and K ion concentrations were measured according to published methods (56).

AFC.

Previously described methods (30, 54) were slightly modified. Mice were sedated with 5 mg/kg of intraperitoneal diazepam (Hospira, Lake Forest, IL) and anesthetized with 75 mg/kg of pentobarbital sodium (Virbac Animal Health, Fort Worth, TX). A 20-gauge angiocatheter was inserted into the trachea and connected to the rodent ventilator Inspira ASV (Harvard Apparatus, Holliston, MA). Ventilator settings were as follows: tidal volume 10 ml/kg, respiratory rate 150 breaths/min, and positive end-expiratory pressure (PEEP) 3 cmH₂O. Mice were given 100% oxygen and monitored with a MouseOx pulse oximeter (Starr Life Sciences, Oakmont, PA). Mice were paralyzed with 0.1 mg/ml intraperitoneal pancuronium (Sigma-Aldrich, St. Louis, MO) and 5 min later instilled intratracheally with 12.5 μl of PBS per gram of body weight with 5% BSA and 0.25 mg/ml of BSA-Alexa Fluor 594 conjugate (Life Technologies, Carlsbad, CA). Mice were kept on the ventilator for 30 min, after which alveolar fluid was aspirated. Alveolar fluid fluorescence was measured at excitation/emission wavelengths of 590/622 nm (SpectraMax M2; Molecular Devices, Sunnyvale, CA), and AFC was calculated as %/h: AFC (%/h) = [V_i − (V_i × F_i/F_a)]/V_i × 100 × 2 = (1 − F_i/F_a) × 100 × 2, where V_i is the instilled volume and F_i and F_a are fluorescence in the instillate and aspirate, respectively.

Lung alveolo-capillary permeability.

We estimated lung alveolo-capillary solute permeability using a procedure modified from previously described methods (7, 27, 29, 69). 5-Carboxyfluorescein (5-CF; 5 mg/kg; 376 Da; Life Technologies) was injected through the tail vein. Bronchoalveolar lavage (BAL) fluid and serum were collected, and serum was diluted 100-fold in PBS. Fluorescence of 5-CF was measured at excitation/emission wavelengths of 492/517 nm, and concentrations were determined using a standard curve constructed from known concentrations of 5-CF. A time-course study showed peak fluorescence of 5-CF at 10 min in both BAL fluid and serum, and fluorescence was measured at 10 min for subsequent in vivo 5-CF permeability experiments. Tetramethylrhodamine-isothiocyanate-dextran (TRITC-dextran; 155 kDa; Sigma-Aldrich) was injected intraperitoneally, and samples were processed as above. Fluorescence of TRITC-dextran was measured at 555/580 nm. Fluorescence peaked at 2 h in both BAL fluid and serum and declined rapidly after 4 h; therefore, fluorescence was measured at 2 h for subsequent in vivo TRITC-dextran permeability experiments.

Wet-to-dry lung weight ratio measurements.

Mice were euthanized with intraperitoneal pentobarbital sodium. Lung weights were measured immediately after harvest and again after drying at 65°C for 48 h.

Mouse alveolar epithelial type II cell isolation and preparation of AEC monolayers.

Mouse alveolar epithelial type II (AT2) cells were isolated from WT and Cldn4 KO mice as previously described (15). AEC monolayers were prepared as previously described (15), except 10% (instead of 2%) newborn bovine serum (Thermo Scientific, Waltham, MA) was added to Complete Mouse Medium, and 750,000 cells (instead of 1 million cells) were seeded per 1.1-cm² filter.

Measurement of bioelectric properties of AEC monolayers.

R_t (kΩ cm²) and spontaneous potential difference (PD, mV) were measured 4, 5, and 6 days after plating cells using an epithelial voltohmmeter (World Precision Instruments, Sarasota, FL). Equivalent short-circuit current (I_eq, μA/cm²) was calculated from the following relation: I_eq = PD/R_t. In some experiments, R_t and I_sc (μA/cm²) were measured 5 or 6 days after plating by the mounting of AEC monolayers in modified Ussing chambers as previously described (15). Dilution potential was measured after replacing both apical and basolateral solutions with K-free Ringer's solution (147 mM NaCl, 0.78 mM NaH₂PO₄, 1.8 mM CaCl₂, 0.8 mM MgSO₄, 5.6 mM glucose, 15 mM HEPES, and 0.075 mM BSA at pH 7.4 and 37°C) and then switching the apical solution to Ringer's solution containing 15 mM NaCl and 264 mM mannitol instead of 147 mM NaCl. Bi-ionic potential was measured after replacing the K-free Ringer's solution in the apical compartment with a KCl Ringer's solution containing 147 mM KCl and 0.78 mM KH₂PO₄ instead of 147 mM NaCl and 0.78 mM NaH₂PO₄ (all reagents from Sigma-Aldrich). Both dilution potential and bi-ionic potential were corrected for liquid junction potentials existing at agar bridge tips, which were calculated from the Henderson-Planck equation (3, 82). Permeability ratio of Na ion to Cl ion (P_Na/P_Cl) was estimated from the dilution potential using the Goldman-Hodgkin-Katz equation (14, 26, 82):

$$P_{\text{N}\text{a}}/P_{\text{C}\text{l}}=\frac{r-e^{\frac{V\prime F}{R\text{T}}}}{r{e}^{\frac{V\prime F}{R\text{T}}}-1}$$

where r is activity ratio of NaCl in apical compared with basolateral compartment, e is natural exponential, V′ is the dilution potential observed with apical side as reference, F is Faraday's constant, R is gas constant, and T is absolute temperature. Similarly, permeability ratio of Na ion to K ion (P_Na/P_K) was calculated as the inverse of P_K/P_Na, which was estimated from the bi-ionic potential:

$$P_{\text{K}}/P_{\text{N}\text{a}}=(1+\beta )e^{\frac{V″F}{\mathit{\text{R}}\text{T}}}-\beta$$

where β is permeability ratio of chloride ion to sodium ion (P_Cl/P_Na) and V″ is the bi-ionic potential observed with apical side as reference. The absolute permeability of Na ion (P_Na) was calculated using the Kimizuka-Koketsu equation:

$$P_{\text{N}\text{a}}=\frac{R\text{T}}{F^2}\cdot \frac{G}{a(1+\beta )}$$

where a is the activity of Na at 147 mM (estimated as 113 mM) and G is AEC monolayer conductance observed with K-free Ringer's solution bathing both apical and basolateral sides (40, 82). P_Cl and P_K were then calculated using P_Na/P_Cl and P_Na/P_K, respectively.

In vitro dye permeability assay.

We modified previously described methods (21, 39). Six days after the plating, 5-CF or TRITC-dextran was added to the apical compartment of AEC monolayers at a final concentration of 0.01 and 1 mg/ml, respectively. Basolateral fluorescence of 5-CF and TRITC-dextran were measured at 2, 4, and 6 h and at 6 and 24 h, respectively, by removing 1 ml of basolateral medium and immediately adding 1 ml of fresh medium. As quality control, R_t and respective solute fluorescence of apical fluid were measured at the beginning and end of each solute permeability experiment and were confirmed to have not changed. Amount of translocated fluorescent solute was analyzed by plotting solute mass (pg) as a function of time (s). Flux (J) was calculated as the slope of the regression line per unit surface area (pg·cm⁻²·s⁻¹). Apparent permeability (P_app, cm/s) was estimated as J/C₀, where C₀ is the solute concentration (pg/cm³) in apical fluid at t = 0.

Na,K-ATPase activity of whole membrane samples.

Na,K-ATPase activity was measured as release of inorganic phosphate (Pi) after addition of ATP to whole lung membrane samples with modification of previously described methods (18, 19, 68). To obtain whole membrane samples, lungs harvested from WT and Cldn4 KO mice were homogenized in 2 ml of homogenization buffer [5% sorbitol, 1 mM EGTA, 30 mM histidine, and Proteinase Inhibitor Cocktail Set III (EMD Millipore, Billerica, MA) at pH 7.4] using a Dounce homogenizer. Cell debris was removed by pelleting at 10,000 g at 4°C for 20 min, and the supernatant was ultracentrifuged (TL-100; Beckman Coulter, Brea, CA) twice, each at 100,000 g for 15 min at 4°C. Pellets were resuspended in 200 μl of homogenization buffer, and protein concentration was measured using DC Protein Assay kit (Bio-Rad Laboratories, Hercules, CA). Whole membrane protein (100 μg) was resuspended in a preincubation buffer (final concentrations: 10 mM MgCl₂, 1.3 mM EGTA, and 30 mM histidine at pH 7.4) with or without 10 mM ouabain and incubated in a 37°C water bath for 1 h, followed by addition of assay buffer (final concentrations: 5 mM ATP, 125 mM NaCl, 12.5 mM KCl, and 30 mM histidine at pH 7.4) to start ATP hydrolysis. Every 2 min, 200-μl samples were removed, and 40 μl of ice-cold 30% trichloroacetic acid was added to stop the ATP hydrolysis reaction. Samples were then incubated on ice for 30 min and pelleted at 18,000 g for 3 min, and 100 μl of supernatant was collected. Ammonium molybdate (300 μl, 2 mM) and Reducing Solution (50 μl) consisting of sodium sulfite, sodium bisulfite, and 1-amino-2-napthol-4-sulfonic acid were added sequentially to the supernatant. After 15 min of incubation at room temperature, absorbance at 660 nm was measured. The rate of Pi release was analyzed by plotting the amount of phosphate (micromoles) as a function of time (h). Na,K-ATPase activity was calculated as the difference in the slopes of regression lines obtained in the presence and absence of ouabain, corrected for initial protein amount in whole membrane samples (μmol·mg protein⁻¹·h⁻¹). All reagents were from Sigma-Aldrich unless otherwise noted.

RNA isolation and qRT-PCR.

Total RNA was extracted from freshly isolated AT2 cells or from whole lungs immediately after injury using TRIzol (Life Technologies). RNA quality was assessed by the ratio of absorbance measured at 260 nm to 280 nm (DU-640B, Beckman Coulter) and agarose gel electrophoresis. Complementary DNA was synthesized using ThermoScript RT-PCR System (Life Technologies) after DNase treatment of RNA samples. Reverse transcription reactions without reverse transcriptase were performed to control for possible contamination of genomic DNA. Polr2a was selected as the reference gene after confirming that the threshold cycles (Ct) were similar between WT and Cldn4 KO samples. Selected genes were amplified with SYBR-Green reagent (Life Technologies) in a 7900HT Fast Real-Time PCR System (Life Technologies) and quantified using the ΔΔCt method. For primer pairs, see Table 1. Claudin primer sequences were from Holmes et al. (34), whereas primers for Na,K-ATPase subunits, Egr1, Tnf, and Il1b were designed using Primer-BLAST (www.ncbi.nlm.nih.gov/tools/primer-blast/) or Vector NTI software (Life Technologies).

Table 1.

Primer sequences

Gene	Forward Primer	Reverse Primer
Polr2a	5′-GGCAAGGTCCCACAACCA-3′	5′-ACAATTGATGTGTCCAGGTATGATG-3′
Atp1a1	5′-TCAAGTCTTGGAGCTCGGAACT-3′	5′-ACGTCTGCATCCCCACATG-3′
Atp1b1	5′-TTCATCGGGACCATCCAAGT-3′	5′-TCCTGGTATGTGGGCTTCAGT-3′
Atp1b3	5′-GCCGAGTGGAAGCTGTTCAT-3′	5′-GGTGCGCCCCAGAAACT-3′
Cldn3	5′-AAGCCGAATGGACAAAGAA-3′	5′-CTGGCAAGTAGCTGCAGTG-3′
Cldn4	5′-CGCTACTCTTGCCATTACG-3′	5′-ACTCAGCACACCATGACTTG-3′
Cldn5	5′-GTGGAACGCTCAGATTTCAT-3′	5′-TGGACATTAAGGCAGCATCT-3′
Cldn7	5′-AGGGTCTGCTCTGGTCCTT-3′	5′-GTACGCAGCTTTGCTTTCA-3′
Cldn18	5′-GACCGTTCAGACCAGGTACA-3′	5′-GCGATGCACATCATCACTC-3′
Egr1	5′-GCAGCAGCGCCTTCAATCCT-3′	5′-TCGTCTCCACCATCGCCTTCT-3′
Tnf	5′-GGCCACCACGCTCTTCTGTCTA-3′	5′-TGAGAGGGAGGCCATTTGGG-3′
Il1b	5′-ACTCAACTGTGAAATGCCACC-3′	5′-GACAGCCCAGGTCAAAGGTT-3′

Protein extraction and Western blot.

AT2 cells and lungs were homogenized with a Polytron homogenizer in 2% SDS lysis buffer (2% SDS, 10% glycerol, and 62.5 mM Tris, pH 6.8) with Proteinase Inhibitor Cocktail Set III (EMD Millipore). After lysis for 10 min at room temperature and incubation on ice for 30 min, DNA was sheared by passing samples five times through a 25-gauge needle. Debris was removed by centrifugation at 18,000 g for 5 min at 4°C. Protein concentration was measured using DC Protein Assay kit (Bio-Rad). For Western analysis, equal amounts of protein were resolved by SDS-PAGE under reducing conditions and transferred to Immobilon-P membranes (Millipore). Membranes were blocked in 5% nonfat dry milk and incubated overnight with antibodies to claudin 3 (Life Technologies), claudin 4 (Life Technologies or Santa Cruz Biotechnology, Santa Cruz, CA), claudin 18 (Life Technologies), Na,K-ATPase α1-subunit (EMD Millipore), Na,K-ATPase β1-subunit (GeneTex, Irvine, CA), eukaryotic initiation factor-2α (Santa Cruz Biotechnology), or lamin A/C (Santa Cruz Biotechnology). After being washed, membranes were incubated with a species-specific secondary antibody conjugated to horseradish peroxidase for 60 min at room temperature and visualized by chemiluminescence (Pierce, Rockford, IL) using the FluorChem Imaging System (Model 8900; Alpha Innotech, San Leandro, CA), which was also used for quantification with densitometric analyses.

VILI.

We modified previously described methods (4, 38). Mice were anesthetized with 100 mg/kg of intraperitoneal ketamine (Phoenix Pharmaceuticals, Burlingame, CA) and 20 mg/kg of xylazine (Lloyd Laboratories, Shenandoah, IA). A 20-gauge angiocatheter was inserted into the trachea and connected to an Inspira ASV ventilator. Mice in the low-peak pressure group were ventilated with the following settings: peak inspiratory pressure of 20 cmH₂O, respiratory rate of 70 breaths per minute, and no PEEP. Mice in the high-peak pressure group were ventilated with the following settings: peak inspiratory pressure of 40 cmH₂O, respiratory rate of 25 breaths per minute, and no PEEP. Both groups were ventilated with ambient air. These settings were similar to those used by Belperio et al. (4), who reported that 20 cmH₂O resulted in a tidal volume of 12 ml/kg and 40 cmH₂O resulted in a tidal volume of 24 ml/kg. Peripheral capillary oxygen saturation (SpO₂) levels were monitored with MouseOx pulse oximeter, and 50 mg/kg intraperitoneal ketamine was administered as needed for adequate sedation. During preliminary experiments, SpO₂ levels rapidly declined after 2 h in some Cldn4 KO mice in the high-peak pressure group; therefore, mechanical ventilation was terminated at 2 h for subsequent experiments. For BAL fluid collection, 30 μl of PBS per gram of body weight was administered intratracheally three times. Retrieved fluids were combined and centrifuged at 600 g for 10 min, and protein concentration of supernatant was measured as above.

Hyperoxia exposure.

We modified previously published methods (30). Mice were kept in a hyperoxia chamber (Terra Universal, Fullerton, CA) for 48 or 65 h with free access to water and food. Sodium bicarbonate powder (Sigma-Aldrich) was placed in the chamber to absorb carbon dioxide. Oxygen concentration was kept above 95% (MiniOX I Oxygen Analyzer; Ohio Medical, Gurnee, IL) at a flow of 15 l/h (Visi-Float Flowmeter VFA; Dwyer Instruments, Michigan City, IN). SpO₂ levels were measured immediately after exposure, and values were recorded after reaching a plateau.

RNA sequencing.

RNA was extracted from whole lungs of two mice for each condition. Lungs were analyzed from WT and Cldn4 KO naïve mice, as well as WT and Cldn4 KO mice that underwent VILI. As described further in results, the response to injurious VILI was variable in Cldn4 KO mice, with injury in some Cldn4 KO mice similar to that in WT mice (Cldn4 KOlow) and injury in some Cldn4 KO mice higher than that in WT mice (Cldn4 KOhigh). We therefore analyzed RNA expression of Cldn4 KOlow lungs and Cldn4 KOhigh lungs separately. One microgram of total RNA was processed with TruSeq RNA Sample Preparation Kit v3.0 (Illumina, San Diego, CA) and sequenced using HiSeq 2000 (Illumina) at the Southern California Genotyping Consortium, University of California, Los Angeles (single read, 100 bases). Reads were trimmed and filtered, then aligned to UCSC mm9 RefSeq Genes using Bowtie (43) on the Galaxy platform (http://galaxyproject.org/) (6, 25). A table of read counts was built using the Bioconductor GenomicRanges package, counts were normalized using EDAseq, and differential expression was analyzed using EdgeR (60) in R (version 3.0.1). The false discovery rate (FDR) was controlled using the Benjamini-Hochberg correction. Datasets are deposited in the public Gene Expression Omnibus (GEO) database (GEO accession no. GSE50927). We took advantage of the observation that some Cldn4 KO mice had similar injury to WT, whereas some had greater injury, and we 1) assessed the genes differentially expressed between WT and Cldn4 KOhigh (245 genes), 2) assessed the genes differentially expressed between WT and Cldn4 KOlow (163 genes), and 3) subtracted the intersect of 1) and 2) from 1) (149 genes left). This allowed us to eliminate genes that were likely differentially expressed due to deletion of claudin 4 but are not involved in greater injury. Potential upstream regulators of observed gene expression changes were predicted using the Upstream Analysis function of Ingenuity Pathway Analysis software (IPA; Ingenuity Systems, Redwood City, CA; www.ingenuity.com).

Statistics.

Data are shown as means ± SE. Two group means were compared using unpaired t-tests. Three or more group means were compared by one-way or two-way ANOVA, as appropriate, with Holm-Sidak post hoc analysis. Statistical analyses were performed using SigmaPlot 12 (Systat, San Jose, CA).

RESULTS

Generation and confirmation of Cldn4 KO mice.

To target the Cldn4 gene, we generated a conditional (floxed) allele, enabling gene deletion by Cre/loxP recombination. LoxP sites were inserted on both sides of the single Cldn4 exon to remove the entire coding sequence (Fig. 1A). Total Cldn4 KO mice were generated by crossing floxed Cldn4 mice to CMV-cre mice (63). Successful ubiquitous gene deletion in Cldn4 KO mice was verified by RT-PCR (Fig. 1B). Western analysis using WT and Cldn4 KO tissues showed similar weak bands using three different anti-claudin 4 antibodies (data not shown), likely indicating cross-reactivity with other claudins. One of the anti-claudin 4 antibodies utilized was the same as that used in another claudin 4 knockout study (23), but it failed to specifically detect claudin 4 in lung. This lack of specificity may be due to expression of other claudins cross-reacting with the antibody and/or different posttranslational modifications of claudin 4 in lung vs. kidney, which may have affected antibody affinity. Cldn4 KO mice produced litters of normal size and did not feature an apparent abnormal phenotype at baseline. H&E staining of WT and Cldn4 KO lung sections showed similar histology without evidence of edema (Fig. 1C). No animal died during our observation period (maximum of 8 mo), consistent with results reported by Fujita et al. (23). Because claudin 4 is also expressed in the kidney, we measured serum levels of Na and K, which were unchanged (WT vs. Cldn4 KO: Na 134 ± 2 vs. 137 ± 1 mEq/l, K 4.8 ± 0.1 vs. 4.7 ± 0.2 mEq/l), also consistent with Fujita et al. (23).

Cldn4 KO mice show lower AFC and Na,K-ATPase activity, higher solute permeability, and unchanged wet-to-dry lung weight ratio compared with WT mice.

AFC was 22% lower in Cldn4 KO vs. WT mice (27.0 ± 1.4 vs. 21.1 ± 1.7%/h, P = 0.03, Fig. 2A). Measurements of alveolo-capillary solute permeability using two fluorescently labeled solutes of different sizes (376-Da 5-CF or 155-kDa TRITC-dextran) revealed that 5-CF permeability was 41% higher in Cldn4 KO (BAL fluid concentration of 5-CF: 189 ± 15 ng/ml in WT, 267 ± 22 ng/ml in Cldn4 KO, P = 0.01, Fig. 2B), but 155-kDa TRITC-dextran permeability was unchanged (Fig. 2C). Na,K-ATPase activity was ∼50% lower in Cldn4 KO mice (4.7 ± 0.8 vs. 2.3 ± 0.4 μmol Pi·mg protein⁻¹·h⁻¹, P = 0.03, Fig. 2D), at least partially explaining the reduction in AFC in Cldn4 KO mice. However, despite lower AFC and higher permeability, no apparent lung edema was seen, as wet-to-dry lung weight ratio was unchanged (4.1 ± 0.1 vs. 4.0 ± 0.1, Fig. 2E).

Fig. 2.

Cldn4 KO mice exhibit lower alveolar fluid clearance (AFC), increased 5-carboxyfluorescein (5-CF) permeability and decreased Na,K-ATPase activity. A: Cldn4 KO mice showed 22% lower AFC compared with WT mice at baseline (n = 6, *P = 0.027). B: concentrations of 5-CF in bronchoalveolar lavage (BAL) fluid after tail vein injection were 41% higher in Cldn4 KO mice (n ≥ 5, *P = 0.013). C: concentrations of TRITC-dextran in BAL fluid after intraperitoneal injection were similar between genotypes. D: whole membrane fractions of lungs from Cldn4 KO mice showed 51% lower ouabain-inhibitable ATPase activity compared with WT mice (n = 6, *P = 0.030). E: wet-to-dry lung weight ratios were similar between WT and Cldn4 KO mice (n ≥ 10).

Bioelectric properties of AEC monolayers from Cldn4 KO vs. WT mice.

R_t measured using a voltohmmeter was unchanged between genotypes (least square means over 3 days, 1.47 ± 0.09 vs. 1.40 ± 0.09 kΩ cm²), whereas I_eq was 13% lower in AEC monolayers from Cldn4 KO mice compared with WT (least square means over 3 days, 7.2 ± 0.2 vs. 6.3 ± 0.2 μA/cm², P = 0.006) (Fig. 3, A and B). These results were confirmed by mounting AEC monolayers on modified Ussing chambers; R_t was unchanged between genotypes and I_sc was 26% lower in Cldn4 KO mice (R_t: 1.45 ± 0.12 vs. 1.53 ± 0.24 kΩ cm², I_sc: 5.2 ± 0.4 vs. 3.9 ± 0.5 μA/cm², P < 0.05) (Fig. 3, C and D). Dilution potential (−3.0 ± 0.6 vs. −2.3 ± 0.7 mV) and bi-ionic potential (3.4 ± 0.1 vs. 3.7 ± 0.1 mV) were unchanged between genotypes (Fig. 3, E and F). P_Na/P_Cl and P_Na/P_K, calculated from the Goldman-Hodgkin-Katz equation, and P_Na, P_K, and P_Cl calculated from the Kimizuka-Koketsu equation, were unchanged between genotypes (Table 2), indicating that ion permselectivity was not altered as a result of absence of claudin 4.

Fig. 3.

Ion permeability, equivalent short-circuit current (I_eq), short-circuit current (I_sc), dilution potential, and bi-ionic potential in WT and Cldn4 KO alveolar epithelial cell (AEC) monolayers. A and C: WT and Cldn4 KO AEC monolayers showed similar transepithelial resistance (R_t) measured with a voltohmmeter (A, n ≥ 7) and in modified Ussing chambers (C, n ≥ 11). B and D: Cldn4 KO AEC monolayers showed 13% lower I_eq (B, n ≥ 7, **P = 0.006 for all days) and 26% lower I_sc compared with WT AEC monolayers (D, n ≥ 11, *P < 0.05). E and F: WT and Cldn4 KO AEC monolayers showed similar dilution potential (E) and bi-ionic potential (F) measured in modified Ussing chambers (n ≥ 10).

Table 2.

Permeability to Na, Cl, and K ions across AEC monolayers from WT and Cldn4 KO mice

	PNa/PCl	PNa/PK	PNa, 10−6 cm/s	PCl, 10−6 cm/s	PK, 10−6 cm/s
Number of samples	n ≥ 10	n = 11	n ≥ 10	n ≥ 10	n = 11
WT	1.16 ± 0.03	0.79 ± 0.01	0.82 ± 0.05	0.71 ± 0.04	1.03 ± 0.07
Cldn4 KO	1.12 ± 0.04	0.78 ± 0.004	0.85 ± 0.11	0.79 ± 0.10	1.09 ± 0.14

AEC, alveolar epithelial cell; WT, wild-type; KO, knockout; P, permeability; Cldn4, claudin 4.

Solute permeability of AEC monolayers from Cldn4 KO vs. WT mice.

Permeability to both 376-Da 5-CF (1.5 ± 0.1 vs. 3.1 ± 0.4 × 10⁻⁷ cm/s, P = 0.003) and 155-kDa TRITC-dextran (1.2 ± 0.2 vs. 3.2 ± 0.5 × 10⁻⁸ cm/s, P = 0.003) was increased in Cldn4 KO AEC monolayers compared with WT (Fig. 4, A and B).

Fig. 4.

Cldn4 KO AEC monolayers exhibit increased permeability to solutes. A: Cldn4 KO AEC monolayers showed 2.1-fold higher permeability to 376-Da 5-CF than WT monolayers (n ≥ 9, **P = 0.001). B: Cldn4 KO AEC monolayers showed 2.6-fold higher permeability to 155-kDa TRITC-dextran than WT monolayers (n = 9, **P = 0.003).

Absence of claudin 4 does not change AT2 cell expression of other claudins or Na,K-ATPase subunits.

mRNA levels of those claudins known to be expressed in lung (Cldn3, 5, 7, and 18) were unchanged in AT2 cells obtained from WT and Cldn4 KO mice (Fig. 5A). Protein levels of claudin 3 and claudin 18 were also unchanged in AT2 cells of WT and Cldn4 KO mice (Fig. 5, B and C). qRT-PCR analysis of Na,K-ATPase subunits Atp1a1, Atp1b1, and Atp1b3 all showed similar mRNA expression in AT2 cells of WT and Cldn4 KO mice (Fig. 5D). Western analysis revealed similar expression of Na,K-ATPase α1- and β1-subunits in AT2 cells of WT and Cldn4 KO mice (Fig. 5, E and F).

Fig. 5.

Alveolar epithelial type II (AT2) cells from Cldn4 KO and WT mice showed similar expression of other claudins and Na,K-ATPase subunits. A: AT2 cells from WT and Cldn4 KO mice showed similar Cldn3, 5, 7, and 18 mRNA expression (n = 3). B and C: AT2 cells from WT and Cldn4 KO mice showed similar Cldn3 (B) and 18 (C) protein expression. D: AT2 cells from WT and Cldn4 KO mice showed similar Atp1a1, Atp1b1, and Atp1b3 mRNA expression (n = 3). E and F: AT2 cells from WT and Cldn4 KO mice showed similar Na,K-ATPase α1-subunit (E) and β1-subunit (F) protein expression. eIF2a, eukaryotic initiation factor-2α.

Cldn4 KO mice are more susceptible to lung injury after experimental VILI compared with WT mice.

qRT-PCR showed that lung Cldn4 mRNA expression in WT mice was sixfold higher after 20 cmH₂O ventilation (P < 0.001) and 36-fold higher after 40 cmH₂O ventilation compared with nonventilated (no VILI) controls (P < 0.001 for 20 cmH₂O vs. 40 cmH₂O, Fig. 6A). BAL protein concentration was higher in Cldn4 KO vs. WT mice only in the 40 cmH₂O ventilation group (0.77 ± 0.18 vs. 0.41 ± 0.06 mg/ml, Fig. 6B). Among Cldn4 KO mice, BAL protein concentration was higher in the 40 cmH₂O VILI group compared with no VILI controls (0.77 ± 0.18 vs. 0.11 ± 0.01 mg/ml, Fig. 6B). No differences were seen in all other pairwise comparisons after two-way ANOVA. As measured by BAL protein concentration in Cldn4 KO mice in the 40 cmH₂O ventilation group, three were highly injured compared with WT mice (Cldn4 KOhigh, 1.46 ± 0.18 mg/ml and all >1.2 mg/ml, P < 0.001 vs. WT), and six were similarly injured compared with WT mice (Cldn4 KOlow, 0.43 ± 0.05 mg/ml and all 0.2–0.6 mg/ml, P < 0.001 vs. Cldn4 KOhigh). SpO₂ of WT, Cldn4 KOlow, and Cldn4 KOhigh mice were 93 ± 3%, 98 ± 0.4%, and 50 ± 12%, respectively (P < 0.001 for WT vs. Cldn4 KOhigh and for Cldn4 KOlow vs. Cldn4 KOhigh). This difference within the Cldn4 KO mouse population may be because they were on a mixed 129S6/SvEvTac, C57BL/6, and BALB/c background. No increase in BAL neutrophils was noted (>95% macrophages for both genotypes), consistent with previous results showing that cellular inflammation is not yet seen at 2 h of mechanical ventilation (77, 81).

Fig. 6.

Cldn4 is upregulated during ventilator-induced lung injury (VILI) in WT mice, and Cldn4 KO mice are more susceptible to VILI at 2 h. A: in WT mice, Cldn4 mRNA expression was 6 times higher during noninjurious 20 cmH₂O ventilation (n = 3, ***P < 0.001) and 36 times higher during injurious 40 cmH₂O ventilation (n = 8, ***P < 0.001) compared with nonventilated controls (no VILI). B: BAL protein concentration was significantly higher in Cldn4 KO vs. WT mice only after 40 cmH₂O ventilation (n = 9, *P = 0.013). BAL protein concentration was significantly higher in the 40 cmH₂O ventilation group compared with the no VILI group in Cldn4 KO mice (n ≥ 3, ***P < 0.001) but not in WT mice. No significant differences were seen in all other comparisons.

Cldn4 KO mice are more susceptible to lung injury after hyperoxia exposure compared with WT mice.

After 48 h of exposure to hyperoxia (>95%), no increase in wet-to-dry lung weight ratio was seen in WT mice, indicating that lung edema was not yet evident. Cldn4 mRNA was increased 14-fold in WT mice compared with room air controls (P < 0.001), whereas Cldn3 and Cldn18 levels remained unchanged (Fig. 7A). After 65 h of hyperoxia, Cldn4 KO mice showed lower SpO₂ (43 ± 6 vs. 82 ± 9%, P = 0.005, Fig. 7B), higher wet-to-dry lung weight ratio (6.6 ± 0.1 vs. 5.4 ± 0.2, P = 0.005, Fig. 7C), and higher BAL cell count (164,600 ± 16,131 cells/ml vs. 96,200 ± 20,887, P = 0.03, Fig. 7D), without significant changes in differential cell counts compared with WT mice. BAL protein (WT: 6.4 ± 1.4 vs. Cldn4 KO: 6.3 ± 0.9 mg/ml) and ATS Lung Injury Scores on H&E sections (WT: 0.33 ± 0.03 vs. Cldn4 KO: 0.37 ± 0.03) were similar between genotypes.

Fig. 7.

Cldn4 is upregulated after hyperoxia, and Cldn4 KO mice are more susceptible to hyperoxic lung injury. A: Cldn4 mRNA expression was 14 times higher after 48 h of hyperoxia (n ≥ 3, ***P < 0.001), whereas Cldn3 and Cldn18 were unchanged, in Cldn4 KO vs. WT mice. B: after 65 h of hyperoxia, SpO₂ was lower in Cldn4 KO vs. WT mice (n = 5–6, **P = 0.005). C: wet-to-dry lung weight ratio after 65 h of hyperoxia was 21% higher in Cldn4 KO vs. WT mice (n = 3, ***P < 0.001). D: BAL cell count after 65 h of hyperoxia was 71% higher in Cldn4 KO vs. WT mice (n ≥ 4, *P = 0.03).

Gene expression analysis through RNAseq reveals that Egr1, Tnf, and Il1b may be important mediators of injury in Cldn4 KO lungs during VILI.

In nonventilated lungs, we confirmed through RNAseq that there was no expression of Cldn4 mRNA in Cldn4 KO lungs, and expression of all other claudin genes and all α- and β-subunit genes of Na,K-ATPase were unchanged between genotypes. In ventilator-injured mice, Cldn4 was increased 16-fold in WT lungs, whereas Cldn3 and Cldn18 were unchanged in both WT and Cldn4 KO lungs. As described in materials and methods and results, we obtained a list of differentially expressed genes that were likely involved in increased susceptibility to lung injury in Cldn4 KO mice by reclassifying Cldn4 KO lungs as Cldn4 KOlow and Cldn4 KOhigh. A starburst plot revealing fold changes of Cldn4 KOhigh vs. WT against Cldn4 KOlow vs. WT is shown in Fig. 8A. One hundred forty-nine genes were differentially expressed (FDR cutoff: 0.05, fold change cutoff: 2-fold) between WT and Cldn4 KOhigh but not between WT and Cldn4 KOlow, and 30 genes with the lowest FDR-adjusted P values are shown in Table 3. Genes that were upregulated during VILI in WT mice and were significantly further upregulated in Cldn4 KOhigh are indicated as “Yes”; genes that were not upregulated during VILI in WT mice but were significantly upregulated in Cldn4 KOhigh are indicated as “No”. Many genes on this list were chemokines and immediate early genes. Of note, Egr1 is known to contribute to higher injury during VILI (55) and was the second most highly significantly changed gene. Serpine1, Tnf, Hmox1, and Areg, which have been reported as differentially expressed in several VILI microarray studies (1, 12, 17, 24, 28), were also all significantly upregulated. In addition to the top 30 genes, Il1b, Cyr61, and Il6, shown to be upregulated in VILI microarray studies, were upregulated, as was Ptges, a known downstream effector of Egr1 that has been implicated in VILI (55). We also performed upstream analysis using IPA software, which examines the entire list of differentially expressed genes and predicts upstream regulators of the observed changes in gene expression. Upstream analysis predicted Tnf and Il1b as the top two upstream regulators of the differentially expressed genes and thus the top potential mediators of increased injury (Fig. 8B). All datasets are deposited in the GEO database (GEO accession no. GEO50927). We validated RNAseq results with qRT-PCR in which Egr1 was increased 2.3-fold (P = 0.004) and Il1b was increased 2.6-fold (P = 0.01, Fig. 8C), with a trend toward increased Tnf expression (1.6-fold increase).

Fig. 8.

Gene expression analysis through whole transcriptome sequencing of RNA from lungs after VILI. A: significant changes in gene expression comparing Cldn4 KO mice with higher levels of injury than WT mice (Cldn4 KOhigh) vs. Cldn4 KO mice with levels of injury similar to WT mice (Cldn4 KOlow) vs. WT. Both axes show log₁₀ (FDR-adjusted P value), with upregulation shown as positive values and downregulation shown as negative values. Red (upregulated) and green (downregulated) show genes with more than 2-fold change and FDR-adjusted P < 0.05 between WT and Cldn4 KOhigh but P ≥ 0.05 between WT and Cldn4 KOlow. B: Ingenuity Pathway Analysis software-generated upstream analysis of RNAseq data revealed Tnf and Il1b as the top 2 predicted upstream mediators of increased injury after VILI. C: Egr1 and Il1b mRNA expression is 2.6 and 2.0 times higher in Cldn4 KOhigh (C4KOhigh) compared with WT mice, respectively (n ≥ 3, *P = 0.004 and 0.01, respectively). Egr1 and Il1b mRNA expression was not different between WT and Cldn4 KOlow (C4KOlow) mice or between Cldn4 KOlow and Cldn4 KOhigh mice.

Table 3.

Top 30 genes differentially expressed between WT vs. Cldn4 KOhigh but not WT vs. Cldn4 KOlow during VILI

Gene Symbol	Gene Name	FDR-Adjusted P Value	log2 (fold change)	Upregulated During VILI in WT mice
Ccl2	chemokine (C-C motif) ligand 2	3.6e-24	3.0	Yes
Egr1	early growth response 1	3.3e-17	2.1	Yes
Ccl3	chemokine (C-C motif) ligand 3	3.3e-17	2.3	Yes
Fosb	FBJ osteosarcoma oncogene B	6.2e-16	2.2	Yes
Junb	Jun-B oncogene	8.7e-16	1.9	Yes
Fos	FBJ osteosarcoma oncogene	1.1e-13	1.9	No
Gm13889	predicted gene 13889	1.6e-13	2.1	No
Ier2	immediate early response 2	2.3e-13	1.7	No
Egr3	early growth response 3	5.3e-13	1.9	No
Nr4a1	nuclear receptor subfamily 4, group A, member 1	5.2e-12	1.6	No
Ccl7	chemokine (C-C motif) ligand 7	7.5e-12	2.6	Yes
Fosl1	Fos-like antigen 1	1.1e-11	2.1	Yes
Egr2	early growth response 2	2.4e-11	1.6	Yes
Zfp36	zinc finger protein 36	1.4e-10	1.8	Yes
Pim1	proviral integration site 1	1.9e-10	1.8	No
Serpine1	serine (or cysteine) peptidase inhibitor, clade E, member 1	2.3e-10	1.6	Yes
Thbs1	thrombospondin 1	3.5e-10	1.4	Yes
Btg2	B cell translocation gene 2, anti-proliferative	8.9e-10	1.4	No
Sertad1	SERTA domain containing 1	1.1e-08	1.4	No
Ltf	Lactotransferrin	2.1e-08	1.9	Yes
Ch25h	cholesterol 25-hydroxylase	2.4e-08	2.1	Yes
Tnf	tumor necrosis factor	3.8e-08	2.0	Yes
Alas2	aminolevulinic acid synthase 2, erythroid	8.1e-08	2.3	No
Reg3g	regenerating islet-derived 3 γ	1.1e-07	3.9	No
Atf3	activating transcription factor 3	1.3e-07	1.3	Yes
Hpgd	hydroxyprostaglandin dehydrogenase 15 (NAD)	1.5e-07	−1.4	No
Csrnp1	cysteine-serine-rich nuclear protein 1	1.5e-07	1.4	Yes
Hmox1	heme oxygenase (decycling) 1	1.9e-07	1.3	Yes
Areg	Amphiregulin	2.0e-07	1.3	Yes
Angptl7	angiopoietin-like 7	2.5e-07	−2.3	Yes

Cldn4 KOhigh, Cldn4 KO mice with higher injury than WT mice; Cldn4 KOlow, Cldn4 KO mice with similar injury to WT mice; FDR, false discovery rate; VILI, ventilator-induced lung injury.

DISCUSSION

Our analyses of Cldn4 KO mice led to the unexpected finding that the junctional protein claudin 4 is not key for normal alveolar fluid homeostasis although it has been reported that Cldn4 KO mice succumb to lethal hydronephrosis through uroepithelial hyperplasia after the age of 12 mo (23). Our Cldn4 KO mice were viable, showed no obvious physiological phenotype, and had similar mortality to WT mice up to the age of 8 mo. Subclinically, Cldn4 KO mice exhibited lower AFC (possibly due to lower Na,K-ATPase activity), higher alveolo-capillary permeability to small solutes, and no apparent pulmonary edema. In vitro, primary AEC monolayers from Cldn4 KO mice did not show changes in ion permeability as assessed by R_t, dilution potential, and bi-ionic potential but revealed small decreases in I_eq and I_sc (consistent with lower Na,K-ATPase activity) and higher permeability to solutes. Following both VILI and hyperoxia exposure, claudin 4 is markedly upregulated in WT mice, and Cldn4 KO mice are more susceptible to lung injury. Gene expression analysis in conjunction with IPA revealed Egr1, Tnf, and Il1b as potential mediators of increased injury in Cldn4 KO mice.

Unexpectedly, deletion of claudin 4 led to minimal physiological effects despite decreased AFC and increased solute permeability. This is similar to findings in β₁ and β₂ adrenergic receptor knockout mice, which showed 44% lower AFC with minimal physiological impairment (54). Lower AFC in Cldn4 KO mice appears to be attributable, at least in part, to decreased Na,K-ATPase activity although underlying mechanisms are unclear. Claudin 3 overexpression in rat AEC led to higher I_sc (52), claudin 7 knockout resulted in higher α1-Na,K-ATPase subunit protein expression in kidneys (71), and claudin 18 knockout resulted in higher β1-Na,K-ATPase subunit protein expression in AT2 cells (47), suggesting that claudins may have a general role in regulating Na,K-ATPase activity. In Cldn4 KO mice, lower Na,K-ATPase activity in lung and decreased I_sc in AEC monolayers were not accompanied by changes in α1- or β1-Na,K-ATPase subunit expression. Decreased Na,K-ATPase activity may be due to decreased membrane insertion of α1- and/or β1-Na,K-ATPase subunits and/or decreased enzymatically active Na,K-ATPase from changes in phosphorylation status or FXYD protein expression. Precise elucidation of the mechanisms underlying claudin 4 regulation of Na,K-ATPase activity will require further study.

Our in vivo results showed increased permeability to 376-Da 5-CF but not 155-kDa TRITC-dextran, whereas in vitro permeability was increased to both. This is likely due to redistribution or resorption of TRITC-dextran in vivo but not in vitro, consistent with our observation of a decline in serum and BAL fluid fluorescence from TRITC-dextran after 2 h in vivo, precluding measurements at 24 h as done for in vitro experiments, where apical fluid fluorescence showed little change over time. The discrepancy between ion and solute permeability is consistent with prior reports on AEC (9, 52) and on claudin 4 function (44, 46, 72). It has been proposed that intercellular contacts have two distinct pathways: high-capacity pores through which only small ions pass and low-capacity pathways through which larger molecules can also pass (64, 73). Conforming to this model, claudin 4 may not affect the high-capacity smaller pores but may affect larger low-capacity pathways.

Claudin 4 has been reported to function primarily as a paracellular Na barrier (11, 52, 72, 80). Our observation that deletion of claudin 4 did not increase ion permeability in AEC monolayers, as measured by R_t, contrasts with these reports. This is likely due to our investigation using genetically engineered mice, allowing compensation for the absence of claudin 4, compared with more acute in vitro overexpression, knockdown, or pharmacological methods used in these studies. In primary AEC monolayers, claudin 3 is reported to increase ion permeability (decrease R_t) (52) and claudin 18 decreases ion permeability (increase R_t) (47). Because we did not detect changes in mRNA or protein expression of other claudins, physiological expression levels of claudin 18 may be sufficient to maintain ion barrier integrity of AEC if suppression of claudin 4 is not acute. Alternatively, claudin 4 may enhance barrier function when the cell monolayer is leakier but not when it is tight, as prior studies report lower R_t than in our experiments.

In contrast to minimal physiological impairment of Cldn4 KO mice at baseline, a significant role of claudin 4 was revealed when lungs were injured. Cldn4 KO mice had higher BAL protein concentrations after ventilation at 40 cmH₂O for 2 h, and lower SpO₂, higher wet-to-dry lung weight ratio, and higher BAL cell count after 65 h of exposure to >95% oxygen. In addition, Cldn4 mRNA expression, but not Cldn3 or 18, was markedly increased during VILI and hyperoxia before any apparent lung injury was noted. The mechanisms involved in greater injury to Cldn4 KO mice likely involve increased solute permeability (as observed by increased BAL protein concentrations after VILI), possibly leading to lung edema (as observed by higher wet-to-dry lung weight ratio after hyperoxia). Lower Na,K-ATPase activity and decreased AFC may have led to decreased capacity to compensate for increased permeability, thereby also contributing to lung edema.

We utilized RNAseq to determine genes and pathways potentially involved in exacerbation of lung injury in Cldn4 KO compared with WT mice. Elimination of genes that are differentially expressed between WT and Cldn4 KO mice with similar injury allowed us to focus on genes likely accounting for more severe injury. We identified several up- or downregulated genes known to be involved in VILI, including Egr1, Tnf, and Il1b. RT-PCR confirmed increases in Egr1 and Il1b with a trend toward increased Tnf. The list of highly significant genes was enriched for chemokines and early response genes, including the Egr family, components of the activator protein-1 transcription factor (Fosb, Junb, Fos, and Fosl1), Nr4a1, and Zfp36. These results are consistent with previous microarray experiments in the course of VILI (1, 12, 17, 24, 28). Among these, Egr1, Tnf, and Il1b may be central regulators during the rapid, neutrophil-independent, stretch-induced pulmonary edema phase of VILI. EGR1 protein, a transcription factor, can be expressed within 2 h of VILI and has been shown to exacerbate VILI (33, 53, 55). EGR1 can bind Ptges, Tnf, and Vegfa promoters to induce gene expression, which in turn can increase permeability (55, 65–67, 70, 75). TNF-α, well known as a proinflammatory cytokine, can signal through the p55 receptor to promote edema or the p75 receptor to protect from edema independent of inflammation (79). IL-1β, also known as a proinflammatory cytokine, can also increase permeability and worsen lung edema independent of inflammation (22). It is striking that many of the upregulated genes are known proinflammatory genes, as VILI-associated inflammation is only seen after 3–4 h of ventilation (77, 78, 81). Had Cldn4 KO mice not succumbed to rapid deterioration of oxygenation by 2 h, Cldn4 KO lungs may have shown increased inflammation in addition to increased permeability by 3–4 h. Changes in inflammatory status have also been noted in mice with deletion of occludin, claudin 7, stomach-specific claudin 18, or junctional adhesion molecule A, possibly through altered matrix metalloprotease activity, integrin expression, or transmigration of leukocytes (16, 31, 42, 62). These results implicate a role for claudins in regulation of proinflammatory pathways in epithelia although underlying mechanisms are unclear.

Overall, our investigation reveals that deletion of claudin 4 is associated with physiologically normal mice despite decreased AFC and increased solute permeability. When subjected to lung injury, claudin 4 expression markedly increased in WT mice. Absence of claudin 4 led to higher permeability or increased lung edema during VILI or exposure to hyperoxia, respectively, possibly through Egr1, Tnf, and/or Il1b, indicating a potential role for claudin 4 in protection against acute lung injury. These findings suggest that modulation of claudin 4 expression (2, 76, 80) may be of value for prevention and/or treatment of acute lung injury.

GRANTS

This research was supported by the Hastings Foundation, Whittier Foundation, and NIH research grants R01ES017034, R01HL056590, R37HL062569, R01HL095349, U01HL108364, and R01HL112638. C. Marconett was supported by ACS/Canary postdoctoral fellowship no. PFTED-10-207-01-SIED. Cell and Tissue Imaging Core of the USC Research Center for Liver Diseases was supported by NIH P30 DK048522 and S10 RR022508. E. Crandall is Hastings Professor and Norris Chair of Medicine. Z. Borok is Edgington Chair in Medicine.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

Author contributions: H.K., P.F., K.-J.K., E.D.C., and Z.B. conception and design of research; H.K., P.F., D.G., Y.H.K., and L.D. performed experiments; H.K., P.F., D.G., Y.H.K., C.N.M., L.D., K.-J.K., and Z.B. analyzed data; H.K., P.F., C.N.M., K.-J.K., E.D.C., and Z.B. interpreted results of experiments; H.K. prepared figures; H.K. drafted manuscript; H.K., P.F., Y.H.K., C.N.M., K.-J.K., E.D.C., and Z.B. edited and revised manuscript; H.K., P.F., D.G., Y.H.K., C.N.M., L.D., K.-J.K., E.D.C., and Z.B. approved final version of manuscript.