Defective ENaC Processing and Function in Tissue Kallikrein-deficient Mice*

An inverse relationship exists between urinary tissue kallikrein (TK) excretion and blood pressure in humans and rodents. In the kidney TK is synthesized in large amounts in the connecting tubule and is mainly released into the urinary fluid where its function remains unknown. In the present study mice with no functional gene coding for TK (TK–/–) were used to test whether the enzyme regulates apically expressed sodium transporters. Semiquantitative immunoblotting of the renal cortex revealed an absence of the 70-kDa form of γ-ENaC in TK–/– mice. Urinary Na+ excretion after amiloride injection was blunted in TK–/– mice, consistent with reduced renal ENaC activity. Amiloride-sensitive transepithelial potential difference in the colon, where TK is also expressed, was decreased in TK–/– mice, whereas amiloride-sensitive alveolar fluid clearance in the lung, where TK is not expressed, was unchanged. In mice lacking the B2 receptor for kinins, the abundance of the 70-kDa form of γ-ENaC was increased, indicating that its absence in TK–/– mice is not kinin-mediated. Incubation of membrane proteins from renal cortex of TK–/– mice with TK resulted in the appearance of the 70-kDa band of the γ-ENaC, indicating that TK was able to promote γ-ENaC cleavage in vitro. Finally, in mouse cortical collecting ducts isolated and microperfused in vitro, the addition of TK in the luminal fluid increased significantly intracellular Na+ concentration, consistent with an activation of the luminal entry of the cation. The results demonstrate that TK, like several other proteases, can activate ENaC in the kidney and the colon.

Tissue kallikrein 5 (TK) 6 is a serine protease that generates kinins locally in many organs, including the kidney, colon, and arteries. In the kidney, TK that is synthesized in large amounts by connecting tubule cells (1) is mainly secreted into the urinary fluid and to a lesser extent to the peritubular interstitium. In the renal interstitium it cleaves locally produced kininogen to yield bradykinin that in turn can activate type-2 (B2) bradykinin receptors. Bradykinin-dependent activation of B2 receptor increases sodium excretion by inhibiting sodium reabsorption in the collecting duct (2). Therefore, the renal kallikrein-kinin system is expected to play a role in renal NaCl balance and blood pressure regulation. Patients with essential hypertension have lower kallikrein levels in their urine (3,4), and mutant mice lacking B2 receptor also exhibit salt-sensitive hypertension (5). However, inactivation of the TK gene in the mouse does not alter blood pressure (6) even though the decrease in renal and urinary kallikrein activity in TK-deficient mice reproduces the phenotype that has been repeatedly associated with hypertension in human and rat studies (7)(8)(9). This finding suggests that low urinary kallikrein excretion observed in hypertensive patients is not a primary cause of high blood pressure (HBP) but rather a consequence of hypertension or of HBPassociated renal defects. An alternative explanation is that TKdeficient mice develop compensatory mechanisms to keep blood pressure at normal levels.
The role of the large amount of TK that is secreted into the urinary fluid remains unknown. One possibility would be that this protease acts directly on the different transporters expressed at the apical side of the tubular cells to modulate their activities. Apical sodium reabsorption along the renal tubule is achieved through multiple sodium transporters. Particularly, in the aldosterone-sensitive distal nephron (ASDN, i.e. distal convoluted tubule (DCT2), connecting tubule, and collecting duct), the amiloride-sensitive epithelial Na ϩ channel ENaC, consisting of ␣-, ␤-, and ␥-subunits, mediates Na ϩ uptake across the apical plasma membrane of principal cells and connecting tubule cells (10). In all renal tubule cells, the Na ϩ is extruded on the basolateral side in exchange for K ϩ by the Na ϩ ,K ϩ -ATPase. Although sodium transport occurs throughout the length of the renal tubule, the fine regulation of sodium excretion occurs in the ASDN, mostly through aldosteronedependent regulation of ENaC.
Because TK production localizes to connecting tubule cells, the urinary side of connecting and collecting duct cells, where ENaC is expressed, is exposed to large amounts of the active enzyme. Because a novel mechanism of proteolytic activation of ENaC by locally produced serine protease has been recently proposed (11,12), we hypothesized that TK might be a paracrine regulator acting directly on ENaC within the ASDN. To test this hypothesis, we used a mouse model with TK gene disruption (TK Ϫ/Ϫ ) to study molecular and functional expression of ENaC.

EXPERIMENTAL PROCEDURES
Animals-The TK Ϫ/Ϫ mice were previously generated in our laboratory (6). Type 2-bradykinin receptor knock out mice (B2 Ϫ/Ϫ ) mice were obtained from The Jackson Laboratory (Bar Harbor, Maine) (13). In all experiments controls consisted in wild type littermates (WT). All the experimental procedures were performed in accordance with the French government animal welfare policy (agreement RA024647151FR).
Aldosterone Infusion Study-7 TK Ϫ/Ϫ and 7 WT mice were infused continuously with aldosterone 100 g⅐kg body weight Ϫ1 ⅐day Ϫ1 diluted in 0.9% NaCl and 5% Me 2 SO administrated by osmotic minipump (Alzet model 2004, Durect Corp., Cupertino, CA). In this particular set of experiments standard laboratory diet was supplemented with 3% Na ϩ and 0.4% K ϩ . Control mice received vehicle alone. Infusion of aldosterone or vehicle was continued over a 28-day period.
Physiological Studies-Animals were housed in metabolic cages and were pair-fed. After 3-5 days adaptation, urines were collected daily for electrolyte measurements. Animals were sacrificed with ketamine and xylazine (0.1 and 0.01 mg⅐g of body weight Ϫ1 , respectively). Plasma and urine electrolytes, creatinine, and aldosterone were determined as described (14). Plasma renin concentration was determined by radioimmunoassay of angiotensin I generated by incubation of the plasma at pH 8.5 in the presence of an excess of rat angiotensinogen (15). Plasma atrial natriuretic peptide concentration was measured by radioimmunoassay (Amersham Biosciences).
Membrane Fraction Preparation-At the time of the sacrifice, kidneys were removed and cut into 5-mm slices. The renal cortex was excised under a stereoscopic microscope and placed into icecold isolation buffer (250 mM sucrose, 20 mM Tris-Hepes, pH 7.4) containing protease inhibitors in 4 g/ml aprotinin, 4 g/ml leupeptin, 1.5 g/ml pepstatin A, and 28 g/ml 4-(2-aminoethyl)benzenesulfonyl fluoride. Lungs were also removed, minced, and placed in the same ice-cold isolation buffer. Minced tissues were homogenized in a Dounce homogenizer (pestle A, 5 passes) followed by 10 passes through a Teflon glass homogenizer rotating at 1000 rpm. The homogenate was centrifuged at 1000 ϫ g for 10 min, and the supernatant was centrifuged at 360,000 ϫ g for 40 min at 4°C. The pellet was resuspended in isolation buffer. Protein contents were determined using the Bradford protein assay (microBradford, Bio-Rad).
Exosome Preparation-Urinary exosomes were prepared as previously described by Pisitkun et al. (16). Briefly, urines from TK Ϫ/Ϫ and WT mice housed in metabolic cages were collected daily in tubes containing a protease inhibitor mixture (1 g/ml leupeptin and 100 g/ml 4-(2-aminoethyl)-benzenesulfonyl fluoride) and sodium azide. A pool of 75-90 ml of urine per group was used. Urine samples were extensively vortexed immediately after they thawed. The urines were centrifuged at 17,000 ϫ g for 15 min at 4°C to remove whole cells, large membrane fractions, and other debris. Supernatants were centrifuged at 200,000 ϫ g for 1 h at 4°C to obtain a low density membrane pellet. The exosome-associated proteins isolated from the pooled urine samples were suspended in isolation solution (250 mM sucrose, 10 mM ethanolamine, pH 7.6, containing protease inhibitors in 1 g/ml leupeptin and 100 g/ml 4-(2-aminoethyl)-benzenesulfonyl fluoride). Protein contents were determined using the Bradford protein assay (Bio-Rad). Values were normalized for urine creatinine to ensure adequate protein loading in Western blot experiments, i.e. to allow comparison of the abundance in exosome-associated proteins from urine samples of the same time for each group, as proposed by others (17).
Antibodies-Rabbit polyclonal antibodies to NHE3 and NKCC2 have been characterized previously (18). Rabbit polyclonal antibodies to NaPi-2a was given by J. Biber (Zürich University, Switzerland) and has been characterized previously (19). Chicken anti Na ϩ ,K ϩ -ATPase polyclonal antibody was purchased from Chemicon International Inc. (Temecula, CA). Rabbit polyclonal antibodies directed to ␣-ENaC (amino acids 46 -68), ␤-ENaC (amino acids 617-638), and ␥-ENaC (amino acids 629 -650) were raised by G. Deschenes using exactly the same peptides originally described by Masilamani et al. (20). All these antibodies have been described and characterized previously (21). When necessary, new batches of antibodies were tested in preliminary experiments to verify that the specificity was identical to that of the antibodies from the original batch (not shown).
Immunoblot Analyses-Membrane proteins were solubilized in SDS-loading buffer (62.5 mM Tris HCl, pH 6.8, 2% SDS, 100 mM dithiothreitol, 10% glycerol, and bromphenol blue), incubated at room temperature for 30 min. Electrophoresis was initially performed for all samples on 7.5% polyacrylamide minigels (XCell SureLock Mini-cell, Invitrogen), which were stained with Coomassie Blue to provide quantitative assessment of loading, as previously described (14). For immunoblotting, proteins were transferred electrophoretically (XCell II Blot Module, Invitrogen) for 1.5 h at 4°C from unstained gels to nitrocellulose membranes (Amersham Biosciences) and then stained with 0.5% Ponceau S in acetic acid to check for uniformity of protein transfer onto the nitrocellulose membrane. Membranes were first incubated in 5% nonfat dry milk in phosphatebuffered saline, pH 7.4, for 1 h at room temperature to block nonspecific binding of antibody followed by overnight at 4°C with the primary antibody (anti-NHE3 1:1,000, anti-NaPi2a 1:20,000, anti-NCC 1:50,000, anti-NKCC2 1:5,000, anti-␣-ENaC 1:3,000, anti-␤-ENaC 1:20,000, anti-␥-ENaC 1:2,000, anti-Na ϩ /K ϩ -ATPase 1:20,000) in phosphate-buffered saline, pH 7.4, containing 1% nonfat dry milk. After four 5-min washes in phosphate-buffered saline, pH 7.4, containing 0.1% Tween 20, membranes were incubated with 1:10,000 dilution of goat anti-rabbit IgG conjugated to horseradish peroxidase (Bio-Rad) or peroxidase-conjugated AffiniPure donkey anti-chicken IgY (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) in phosphate-buffered saline, pH 7.4, containing 5% nonfat dry milk for 2 h at room temperature. Blots were washed as above, and luminal-enhanced chemiluminescence (ECL, PerkinElmer Life Science Products) was used to visualize bound antibodies before exposure to Hyperfilm ECL (Amersham Biosciences). The autoradiography was digitized with use of a laser scanner (Epson Perfection 1650, Epson), and quantification of each band was performed by densitometry using NIH Image software. Densitometric values were normalized to the mean for the control group that was defined as 100%, and results were expressed as the mean Ϯ S.E. For deglycosylation experiments protein samples were incubated for 1 h at 37°C with or without peptide N-glycosidase F according to the manufacturer's instructions (Roche Diagnostics) and processed for immunoblotting as described above.
To assess whether or not TK promotes ␥-ENaC cleavage in vitro, renal cortical membrane fractions from TK Ϫ/Ϫ mice were prepared as described above except that proteases inhibitors were omitted. To compare their ability to promote ␥-ENaC cleavage, urines from TK Ϫ/Ϫ and WT mice were first desalted against Tris buffer (0.2 M, pH 8.2) in a Centricon column (Centricon-10, cut-off 10 kDa, Millipore). The final volume of the desalted urine was one-third that of the initial volume of urine introduced in the column. Membrane fractions were incubated with desalted urines for 1.5 h at 37°C or with purified TK from porcine pancreas (Sigma-Aldrich) before SDS-PAGE and immunoblotting with anti-␥-ENaC antibody as described above.
Measurement of Alveolar Fluid Clearance in Mouse-Sodium-driven alveolar fluid clearance (AFC) was measured in vivo using an in situ non-ventilated mouse lung model, as previously described (22,23). This model has been shown to give AFC values similar to those obtained with the ventilated mouse model over a 15-min period. Briefly, male or female WT or TK Ϫ/Ϫ mice aged 2-5 months were euthanized with intraperitoneal pentobarbital (250 mg/kg) and maintained at 37-38°C using a heating pad, an infrared lamp, and an intra-abdominal monitoring thermistor. A 20-gauge venous catheter was inserted in the trachea through a tracheotomy and tightly fixed. The lungs were inflated with 100% O 2 at 7 cm H 2 O continuous positive airway pressure throughout the experiment. Then, 10 ml/kg instillate was delivered to the lungs over 30 s through the tracheal catheter. The instillate consisted of Ringer's lactate, pH 7.4, adjusted to 325 mosmol/kg of H 2 O with NaCl, containing 5% bovine serum albumin, and 0.1 Ci/ml 125 I-labeled albumin (Cis Bio International, Gif-sur Yvette, France) as a labeled alveolar fluid volume tracer. An alveolar fluid sample (50 -100 l) was aspirated 1 min after instillation and at the end of experiment (15 min later). The aspirates were centrifuged at 3000 ϫ g for 10 min, and the radioactivity in supernatants was counted in duplicate. Alveolar fluid clearance (percentage fluid absorption at 15 min) was calculated from the increase in alveolar fluid albumin as AFC (%) ϭ (Cf Ϫ Ci)/Cf ϫ 100, where Ci and Cf represent the initial and final concentrations of 125 I-labeled albumin in the aspirate at 1 and 15 min, respectively, as assessed by radioactivity measurements. In some experiments amiloride (1 mM) was added to the instillate, and AFC was measured at 15 min as described above. ENaC-mediated AFC represents the difference between total AFC values (in the absence of amiloride) and amiloride-insensitive AFC values (in the presence of amiloride). Results are presented as the means Ϯ S.E. One-way variance analyses were performed and, when allowed by the F value, results were compared by the modified least significant difference (Statview software). p Ͻ 0.05 was considered significant.
Measurement of Rectal Transepithelial Potential Difference-The mice were anesthetized with an intraperitoneal injection of Ketalar (Parke-Davis), 75 g/g of body weight, and Rompun (Bayer, Puteaux, France), 2.3 g/g of body weight, and placed on a heated table. A winged needle filled with isotonic saline was placed in the subcutaneous tissue of the back. A doublebarreled pipette was prepared from borosilicate glass capillaries (1.0-mm outer diamater/0.5-mm inner diameter, Hilgerberg, Malsfeld, Germany) and pulled to a ϳ0.2-mm tip diameter. The first barrel was filled with isotonic saline buffered with 10 mM Na ϩ -HEPES, pH 7.2, and the second barrel with the same solution containing 25 M amiloride. The tip of the double-barreled pipette was placed in the rectum about 3-5 mm from the skin margin. The electrical potential difference was measured between the first barrel and the subcutaneous needle, both connected to Ag ϩ /AgCl electrodes by means of plastic tubes filled with 3 M KCl in 2% agar. The rectal potential difference (PD) was monitored continuously by a VCC600 electrometer (Physiologic Instruments, San Diego, CA) connected to a chart-paper recorder. After stabilization of the rectal PD (about 1 min), 0.05 ml of saline solution was injected through the first barrel as a control maneuver, and the PD was recorded for another 30-s period. A similar volume of saline solution containing 25 M amiloride was injected through the second barrel of the pipette, and the PD was recorded for another minute. The amiloridesensitive PD was calculated as the difference between the PD recorded before and after the addition of amiloride.
Evaluation of Apical Na ϩ Entry in Isolated CCDs Microperfused in Vitro-Experiments were performed as previously described by others (24). Briefly, CCD segments were dissected from C57Bl/6 mouse, mounted on concentric pipettes, and perfused in vitro. During intracellular pH measurement experiments, the average tubule length exposed to bath fluid was limited to 300 -350 m to prevent motion of the tubule. The dissection and initial luminal and peritubular solutions were composed of 25 mM NaCl, 119 mM N-methyl-D-glucamine gluconate, 1.2 mM MgSO 4 , 2 mM K 2 HPO 4 , 2 mM CaCl 2 , 5 mM D-glucose, and 10 mM HEPES and was adjusted to a pH of 7.4. Tubules were bathed and perfused with this same solution. For the experiments CCDs were bathed in a modified solution in which NaCl was iso-osmotically replaced with N-methyl-D-glucamine gluconate to achieve an NaCl concentration of 0 mmol/ liter and containing the Na ϩ ionophore monensin (10 Ϫ5 M). This maneuver effectively eliminated the basolateral mem-brane as a barrier for Na ϩ movement. Under these conditions Na ϩ entry across the apical membrane is the only means of altering [Na] i .
To identify principal and intercalated cells, we labeled intercalated cells by adding fluorescein-labeled peanut lectin (Vector Laboratories) to the luminal perfusate for 5 min and observed which cells were fluorescent with excitation and emission wavelengths of 440 and 530 nm, respectively.
[Na] i in CCD cells was assessed with imaging-based, dual-excitation wavelength fluorescence microscopy with use of the fluorescent probe sodium-binding benzofuran isophthalate (SBFI, Molecular Probes). Tubules were loaded with 2 ϫ 10 Ϫ5 M of the acetoxymethyl ester of SBFI added to the luminal perfusate. Intracellular dye was excited alternatively at 340 and 380 nm with a 100-watt xenon lamp and a computer-controlled chopper assembly. Emitted light was collected through a dichroic mirror, passed through a 510-nm filter, and focused onto a CCD camera (ICCD 2525F, Videoscope International, VA) connected to a computer. The measured light intensities were digitized with 8-bit precision (256 gray level scale) for further analysis. For each tubule 2-4 principal cells were analyzed; the mean gray level for each excitation wavelength was calculated with the Starwise Fluo software (Imstar, Paris, France). SBFI fluorescence ratios (340/380 nm) were used as an estimator of [Na] i values.

RESULTS
Evidence for Decreased Sodium Channel Activity in the Kidney and the Distal Colon-TK is secreted into the luminal fluid of the distal part of the ASDN, where ENaC is believed to be the limiting step for Na ϩ reabsorption, and a strong inverse relationship has been described between the amount of TK excreted in the urine and blood pressure in humans (3,4). We speculated that TK might be a factor regulating ENaC activity. To test renal ENaC function in vivo, we first tested the effects of amiloride on urinary sodium excretion in TK Ϫ/Ϫ mice. Fig. 1A shows that a single dose of amiloride (subcutaneous injection of 1.45 mg⅐kg of body weight Ϫ1 ) significantly increased urinary Na ϩ excretion during the next 2 h after the injection in WT mice, whereas the effect of amiloride on Na ϩ excretion was blunted in TK Ϫ/Ϫ mice. This result indicated that renal ENaC activity was decreased in mice lacking TK production.
The distal colon also expresses all three ENaC subunits, mRNA for TK has been found to be highly expressed in the colon, and abundant TK activity is measurable in the feces (6). Thus, we hypothesized that if TK activates ENaC, this effect may be detectable in the distal colon. We tested this hypothesis in vivo by measuring amiloride-sensitive transepithelial rectal potential difference, an index of ENaC-mediated Na ϩ transport, in WT and TK Ϫ/Ϫ mice. As shown in Fig. 1B, the amiloride-sensitive rectal potential difference was significantly decreased in TK Ϫ/Ϫ mice compared with WT mice, indicating that, as in the kidney, TK is an activating factor of ENaC in the distal colon. Table 1 shows that TK Ϫ/Ϫ mice were able to maintain normal Na ϩ balance with no evidence for extracellular fluid volume depletion. Moreover, the animals were pair-fed, and hence, identical urinary excretion of Na ϩ and Cl Ϫ (see Table 1) indi-  cated that they were able to achieve a steady state. Thus, although the TK Ϫ/Ϫ mice are able to excrete the full amount of NaCl taken in the diet and maintain a steady state, they do so with a concomitant decrease in amiloride-sensitive (ENaC-mediated) sodium absorption, implying that amiloride-independent Na ϩ absorption, presumably upstream from the collecting duct system, is increased.
Luminal TK Activates ENaC in Principal Cells of Cortical Collecting Duct Isolated and Microperfused in Vitro-Our preceding experiments indicate that TK Ϫ/Ϫ mice have a decreased in ENaC-mediated Na ϩ transport in kidney. Therefore, to confirm more directly that TK is able to activate ENaC, we next examined the effects of extracellular (luminal) TK on ENaC in CCDs isolated and microperfused in vitro. Intercalated cells were identified by adding fluorescein-labeled peanut lectin to the luminal perfusate for 5 min and observed which cells were fluorescent.
[Na] i in principal cells was then assessed with use of the fluorescent probe sodium-binding benzofuran isophthalate (SBFI), as previously described (24). ENaC is the limiting step for Na ϩ entry into principal cells; therefore, if TK activates ENaC, [Na] i is expected to increase when TK is perfused to the luminal surface of the CCD. Accordingly, Fig. 2 shows that TK was able to increase significantly the ratio of F340/F380 SBFI fluorescence, which reflects an increase in [Na ϩ ] i , whereas no change was observed in CCDs perfused with vehicle only. Moreover, a TK-dependent increase in [Na ϩ ] i was prevented when the tubules were pretreated with 10 Ϫ5 M amiloride, indicating that the effect of TK on [Na] i is a consequence of ENaC activation.
␥-ENaC Processing Is Impaired in TK Ϫ/Ϫ Mice-Because TK is a serine protease, it is possible that TK-induced ENaC activation occurs through ENaC proteolysis. Thus, we next examined ENaC-related polypeptides by Western blot. As shown in Fig. 3, antibodies to the C terminus of ␥-ENaC detect a major band at 85 kDa as well as a minor 80-kDa band. An additional broad band centered around 70 kDa was detected in WT mice but not in TK Ϫ/Ϫ mice. This 70-kDa band is believed to originate from the proteolytic cleavage of the N-terminal end of the 85-kDa form and to be related to aldosterone-dependent activation of ENaC (20). When renal samples from WT mice were deglycosylated with peptide N-glycosidase and immunoblotted using anti-␥ ENaC, the 85-kDa band shifted to 72 kDa, and the broad 70-kDa band shifted to a major 50-kDa band and a weaker 60-kDa band (Fig. 4A). No change in protein abundance of either the ␣ and ␤ subunits of ENaC or other Na ϩ transporter expressed along the nephron was detected (Fig. 3, Table 2).
TK Is Able to Promote the Cleavage of ␥-ENaC in Vitro-If TK is important for the activation of ENaC via its proteolytic properties on ␥-ENaC subunit, TK should be able to promote the cleavage of this protein in vitro. Renal membrane fractions from    TK Ϫ/Ϫ mice, prepared in the absence of protease inhibitors to avoid in vitro inactivation of TK, were incubated with porcine pancreatic TK at 100 or 30 g/ml or with vehicle alone for 15 min at 37°C. Fig. 5A shows that only the 85-kDa band is detected on renal membrane fractions from TK Ϫ/Ϫ mice exposed to the vehicle and that the 70-kDa band of the ␥-ENaC appeared in samples incubated with both concentrations of TK. Densitometric analyses were performed to quantify the appearance of the kallikrein-induced 70-kDa ␥-ENaC (Fig. 5B). Band densities of the 70-kDa band of 4 independent immunoblots were normalized to total densities of the ␥-ENaC. The 70-kDa band was significantly increased by TK at both concentrations over control. At the highest concentration of TK, the total densities of ␥-ENaC was significantly decreased presumably due to nonspecific protein degradation. Porcine pancreatic TK is extracted from pancreas and might therefore contain small amount of other protease, particularly trypsin, which is known to activate ENaC. Thus, in a second set of experiments, desalted urines collected from WT and TK Ϫ/Ϫ mice and a renal membrane fraction from TK Ϫ/Ϫ mice (i.e. naturally devoid of detectable 70-kDa form of ␥-ENaC; see Fig. 3) were used. As shown in Fig. 5C, when renal membrane fractions prepared from a TK Ϫ/Ϫ mouse were incubated with desalted urines from either WT or TK Ϫ/Ϫ mice, the 70-kDa band of ␥-ENaC appeared in samples incubated with urine from WT mice but not with urine from TK Ϫ/Ϫ mice, indicating that urinary kallikrein favors cleavage of ␥-ENaC in vitro. No cleaved band appeared in untreated samples kept in parallel over the period of the experiment, indicating that the 70-kDa band is not the result of nonspecific degradation.
Impaired ␥-ENaC Processing Leads to Decreased Abundance in Urinary Exosomes-A study of urinary exosome-associated proteins provides a means for analysis of changes in apical protein expression of renal tubular cells (16). We looked at changes in the excretion of the three subunits of ENaC in three independent preparations of urinary exosomes from WT and TK Ϫ/Ϫ mice by immunoblotting using anti-N terminus ␣-ENaC antibody and anti-C terminus ␤or ␥-ENaC antibodies. Fig. 6 shows representative immunoblots. Only the ␤ and ␥ subunits of ENaC were detectable in the urinary exosome preparations from WT mice. In human exosome preparations, all three subunits were detected (16). Regarding ␥-ENaC, note that, as previously reported (16) in human exosomes, only the 70-kDa form of the ␥-ENaC was detectable in mouse exosomes, suggesting that this is the only molecular form of ␥-ENaC present at the apical surface of the collecting duct cells. As expected from the results shown in Fig. 3, the ␥-subunit of ENaC was dramatically decreased in urinary exosome preparation from TK Ϫ/Ϫ mice (25 Ϯ 2% of control, p Ͻ 0.001). We also observed a marked decrease in ␤-ENaC protein abundance (24 Ϯ 6% of control, p Ͻ 0.01). As in WT mice, ␣-ENaC was undetectable in urinary exosomes from TK Ϫ/Ϫ mice. Importantly, the protein abundance of NCC, assessed on same preparations, was not altered (102 Ϯ 34% of control, p ϭ 0.96), indicating that the decrease in ␤and ␥-ENaC subunits is specific and not due to incomplete sampling of urine.
Evidence for Unchanged Sodium Channel Activity in Lung-In contrast to the kidney and the colon, lung alveolar epithelial cells express ENaC but do not secrete TK; therefore, we hypothesized that ENaC processing in the lung is TK-independent. Fig. 8A shows that amiloride-sensitive alveolar fluid clearance, used as an estimate of ENaC-mediated Na ϩ transport (23), was not significantly different in TK Ϫ/Ϫ mice compared with WT mice. Fig. 8B shows that the biochemical profile of ␥-ENaC in the lung appears as two major bands, one at 85 kDa, similar to that observed in the kidney, and another one resolving at 75-70 kDa. These bands were no longer detected when anti-␥-ENaC was preincubated with the peptide antigen (data not shown). There was no significant difference in the abundance of the 85-kDa ␥-ENaC (102 Ϯ 5 versus 100 Ϯ 12% in TK Ϫ/Ϫ and WT mice, respectively) and of the 75-70-kDa ␥-ENaC (103 Ϯ 6 versus 100 Ϯ 8% in TK Ϫ/Ϫ and WT mice, respectively). Deglycosylation with peptide N-glycosidase gave a pattern similar to that seen in renal samples from WT mice except that the size of the major fragment was 60 kDa (data not shown).
Adaptation of TK Ϫ/Ϫ Mice to Dietary Na ϩ Restriction and Response to Aldosterone Infusion-As stated above, TK Ϫ/Ϫ mice showed no evidence for altered blood volume. However, it is not unusual that altered sodium transport in transgenic mice leads to limited phenotypes. Despite decreased ENaC activity in the absence of TK production, TK Ϫ/Ϫ mice were able to adapt normally in response to acute dietary sodium restriction. Both genotypes significantly decreased urinary Na ϩ excretion after Na ϩ restriction. No significant difference was observed among groups (data not shown). In addition, urinary aldosterone levels increased similarly in TK Ϫ/Ϫ (11.96 Ϯ 3.35 nmol/mmol creatinine) and WT mice (12.21 Ϯ 1.64 nmol/mmol creatinine) after 3 days of Na ϩ depletion. This result indicated that in response  to Na ϩ restriction, TK Ϫ/Ϫ mice were able restore normal ENaC function. In WT animals undergoing sodium restriction, the broad 70-kDa band was easily detected on immunoblots for ␥-ENaC (Fig. 9A), consistent with previous observations (20,25,26). In TK Ϫ/Ϫ mice, sodium depletion was associated with the appearance of polypeptides of molecular weights ranging from 75 to 70 kDa (compare WT and TK Ϫ/Ϫ lanes in Fig. 7A). The same results were observed in mice treated with aldosterone (Fig. 9B). Deglycosylation with N-glycosidase F of aldosterone-infused WT and TK Ϫ/Ϫ mice gave a similar pattern except that the size of the major fragments was 50 and 60 kDa for WT and TK Ϫ/Ϫ mice, respectively (Fig. 4B). These sets of experiments demonstrate that in the nominal absence of TK, ENaC can still be processed and activated by other serine proteases, presumably from the channel activating protease (CAP) family.

DISCUSSION
The present study demonstrates that TK is a physiological activator of ENaC. This activation is associated with the proteolytic cleavage of the full-length 85-kDa form of ␥-ENaC subunit within its extracellular domain to produce a truncated peptide of 70 kDa corresponding to the C-terminal end of the entire protein. This demonstration is based on several observations; 1) TK Ϫ/Ϫ mice have decreased ENaC activity in the kidney and also the distal colon, two organs that exhibit ENaC-mediated Na ϩ transport and abundantly produce TK; 2) TK increased ENaC-dependent Na ϩ entry when applied to the luminal surface of CCDs isolated and microperfused in vitro; 3) at steady state, TK Ϫ/Ϫ mice have a marked decrease in the protein abundance of the truncated ␥-ENaC peptide in the kidney and in urinary exosomes, a preparation that reflects the pool of protein apically expressed in renal tubular cells; 4) TK is able to promote the cleavage of ␥-ENaC in vitro; 5) in the lung alveolae, an organ devoid of TK expression, TK disruption did not impair ENaC processing and activity.
ENaC is a heteromultimeric protein composed or three related subunits ␣, ␤, and ␥. All subunits have an intracytoplasmic N and C terminus and two membrane spanning domains connected by a large extracellular loop. ENaC activity can be regulated by changes in membrane protein abundance of the channel or by changes in single-channel open probability.
Recently, it has been proposed that proteolytic processing of ␣ and ␥ subunits by several serine proteases increases ENaC open probability (11,12,(27)(28)(29). Conversely, non-processed ENaC have been demonstrated to be almost inactive when expressed in cell cultures. In this model extracellular trypsin stimulates ENaC, whereas extracellular serine protease inhibitors, such as aprotinin and bikunin, have been shown to decrease ENaC activity (12, 29 -32). However, despite this evidence that proteolytic processing of ENaC stimulates channel activity, the molecular identity of the protease that could mediate this effect in vivo is still uncertain. As stated above, trypsin can activate ENaC. Elastase, a serine protease naturally produced by neutrophil polynuclear has also been proposed to activate ENaC in vivo during pulmonary inflammatory states (28). More recently, direct evidence for exogenous cleavage of ␥-ENaC and activation at the cell surface by elastase in Xenopus oocytes experiments has been provided by Harris et al. (33). However, trypsin and elastase are not very good candidates for in vivo physiological regulation of ENaC in the kidney because they are normally not expressed and secreted in this organ. Furin, a proprotein convertase involved in many physiological or pathophysiological processes, including activation of hormones and growth factors to bacterial toxins and viral glycoproteins processing, is also able to cleave and activate ENaC in vitro (11,34). A family of serine proteases has been identified that could cleave and activate ENaC, including CAP1, CAP2, CAP3, and TMPRSS3 (12,23,(35)(36)(37)(38). Importantly, it has been shown that aldosterone modifies urinary excretion of prostasin, the human ortholog of CAP1 (39), suggesting that prostasin could be the link between aldosterone and proteolytic activation of ENaC in the kidney. Unfortunately, gene disruption of furin or CAP1 are lethal, rendering difficult direct assessment of their physiological role in ENaC regulation (40,41). It is also possible that proteolytic activation of ENaC involves more than one effector. This is supported by the observation that the proteolytic processing of ␥-ENaC by furin seems to occur intracellularly, whereas trypsin, elastase, or the CAPs are believed to act extracellularly. Moreover, it has recently been demonstrated that furin or prostasin is able to produce different polypeptides from ␥-ENaC (42).
Here we report that a serine protease promotes ␥-ENaC cleavage in vivo and in vitro (see Figs. 1A and 5) and that this proteolytic modification correlates with the activation of ENaC. Active TK excretion is positively correlated with mineralocorticoid levels (43,44). Interestingly, it has also been shown that the protein abundance of the 70-kDa form of ␥-ENaC (the truncated form which disappears in TK Ϫ/Ϫ mice) also correlates with circulating aldosterone level (20). Therefore, it is tempting to conclude that TK is responsible for part of the aldosterone-dependent processing/activation of the channel. However, it is important to state that it is not possible to conclude that TK is directly responsible for the cleavage itself (i.e. that TK physically interacts with and cleaves ␥-ENaC). Indeed, in our in vitro experiments we used membrane fractions from TK Ϫ/Ϫ mice. This material does not contain TK, but it probably expresses several other proteases, particularly several members of the CAP family that are membrane-bound. Therefore, it is possible that the action of TK on ␥-ENaC is indirect. For exam- FIGURE 9. Effect of TK knock-out on ␥-ENaC protein in mice undergoing dietary Na ؉ restriction or chronic aldosterone infusion. Immunoblots for ␥-ENaC of membrane fractions from kidney cortex of TK Ϫ/Ϫ and wild type mice under Na ϩ restricted diet (A) or after chronic aldosterone administration (B). Each lane was loaded with a protein sample from a different mouse. Equal loading was confirmed by parallel Coomassie-stained gels. The asterisk indicates a protein sample from a mouse that had not correctly received aldosterone.
ple, it is conceivable that TK activates a CAP-dependent signaling cascade, leading ultimately to ENaC processing as proposed by B. Rossier recently (45).
Because the TK Ϫ/Ϫ mice did not exhibit gross alterations in sodium balance or extracellular fluid regulation, it could be also argued that TK does not play a substantial physiological role under normal circumstances. However, because regulation of Na ϩ reabsorption in the kidney is critical for extracellular fluid volume and blood pressure regulation, the mechanisms for regulation of Na ϩ reabsorption are highly redundant. That is, animals can generally compensate for defective regulation at one renal tubule segment by increasing absorption at another renal tubule segment, lowering blood pressure or reducing glomerular filtration rate to come into Na ϩ balance. Thus, in the present study altered regulation at the level of ENaC is likely to be compensated for. One example of this is seen in mice in which the thiazide-sensitive NaCl cotransporter is responsible for 5-10% of sodium reabsorption under normal circumstance (46). These mice appear healthy and are normal with respect to plasma electrolyte concentrations, serum aldosterone levels, and blood pressure. Moreover, NCC Ϫ/Ϫ mice retain Na ϩ as well as wildtype mice when fed a Na ϩ -depleted diet (47). In contrast, these mice show abnormalities in homeostasis of other ions including Ca 2ϩ (47) and K ϩ (48). One other possible explanation is that the TK gene inactivation does not only disrupt TK production but also impairs local bradykinin production. Inhibition of bradykinin production is expected to limit natriuresis, whereas TK disruption is expected to limit ENaC activation. Therefore, the phenotype arising from both effects might be very limited. Finally, in the absence of TK some proteolytic activation of the channel can still occur and produce several different polypeptides that are presumably active (see Fig. 9). The role of furin or CAPs in this compensation requires further investigation. For example, CAP could permit the cleavage of the 85-kDa ␥-ENaC form and, therefore, activation of ENaC in TK Ϫ/Ϫ mice with elevated circulating aldosterone. Based on the size of the fragments obtained after deglycosylation of the cleaved form of ␥-ENaC (60 and 50 kDa), which are recognized by the anti-␥-ENaC antibody directed against the C terminus of ␥-ENaC, the proteolytic cleavage seems to occur at two different sites in the early portion of the extracellular loop of the ␥-ENaC which are approximately 100 amino acids apart. Deglycosylation of renal samples from either control or aldosterone-treated WT mice showed that the 70-kDa protein shifted to a major 50-kDa protein (Fig. 2). Deglycosylation of kidney samples from aldosterone-infused TK Ϫ/Ϫ mice showed that the broad 75-70-kDa protein shifted to a major 60-kDa protein and a weaker 50-kDa protein (Fig. 2). Membrane samples from lung, an organ not expressing TK (6), showed a similar pattern after deglycosylation. These results suggest that the renal 70-kDa ␥-ENaC could result from the proteolytic cleavage of the 85-kDa ␥-ENaC form by the TK itself (or at least dependent on TK), whereas the 75-70-kDa broad protein seen in kidney sample from TK Ϫ/Ϫ mice with elevated circulating aldosterone or lung sample from either WT or TK Ϫ/Ϫ mice could result from the proteolytic cleavage of the 85-kDa ␥-ENaC by another serine protease such as CAPs or furin. We also report that ENaC is functionally normal in lung alveoli of TK Ϫ/Ϫ mice, confirming that ENaC in the lung can be processed by another protease than TK. These observations highlight the fact that TK-dependent activation of ENaC may be tissue-specific.
In conclusion, this study identified an anti-natriuretic effect of TK, and we propose that TK is a physiological activator of ENaC that belongs to a complex cascade of protease that is required for full stimulation of Na ϩ transport within the ASDN. The data from our model of TK inactivation strongly suggest that the decrease in renal TK production that has been found repeatedly to be associated with high blood pressure in humans is rather a consequence than a cause of this disease.