The epithelial Na+ channel (ENaC) that mediates regulated Na+ reabsorption by epithelial cells in the kidney and lungs can be activated by endogenous proteases such as channel activating protease 1 and exogenous proteases such as trypsin and neutrophil elastase (NE). The mechanism by which exogenous proteases activate the channel is unknown. To test the hypothesis that residues on ENaC mediate protease-dependent channel activation wild-type and mutant ENaC were stably expressed in the FRT epithelial cell line using a tripromoter human ENaC construct, and protease-induced short-circuit current activation was measured in aprotinin-treated cells. The amiloride-sensitive short circuit current (INa) was stimulated by aldosterone (1.5-fold) and dexamethasone (8-fold). Dexamethasone-treated cells were used for all subsequent studies. The serum protease inhibitor aprotinin decreased baseline INa by approximately 50% and INa could be restored to baseline control values by the exogenous addition of trypsin, NE, and porcine pancreatic elastase (PE) but not by thrombin. All protease experiments were thus performed after exposure to aprotinin. Because NE recognition of substrates occurs with a preference for binding valines at the active site, several valines in the extracellular loops of α and γ ENaC were sequentially substituted with glycines. This scan yielded two valine residues in γ ENaC at positions 182 and 193 that resulted in inhibited responses to NE when simultaneously changed to other amino acids. The mutations resulted in decreased rates of activation and decreased activated steady-state current levels. There was an ∼20-fold difference in activation efficiency of NE against wild-type ENaC compared to a mutant with glycine substitutions at positions 182 and 193. However, the mutants remain susceptible to activation by trypsin and the related elastase, PE. Alanine is the preferred P1 position residue for PE and substitution of alanine 190 in the γ subunit eliminated INa activation by PE. Further, substitution with a novel thrombin consensus sequence (LVPRG) beginning at residue 186 in the γ subunit (γTh) allowed for INa activation by thrombin, whereas wild-type ENaC was unresponsive. MALDI-TOF mass spectrometric evaluation of proteolytic digests of a 23-mer peptide encompassing the identified residues (T176-S198) showed that hydrolysis occurred between residues V193 and M194 for NE and between A190 and S191 for PE. In vitro translation studies demonstrated thrombin cleaved the γTh but not the wild-type γ subunit. These results demonstrate that γ subunit valines 182 and 193 are critical for channel activation by NE, alanine 190 is critical for channel activation by PE, and that channel activation can be achieved by inserting a novel thrombin consensus sequence. These results support the conclusion that protease binding and perhaps cleavage of the γ subunit results in ENaC activation.
The regulated transepithelial transport of Na+, critical for maintaining body fluid homeostasis in terrestrial organisms, is attributable to the expression of the heteromeric epithelial Na+ channel (ENaC) (Garty and Palmer, 1997). ENaC consists of α, β, and γ subunits (Canessa et al., 1994; McDonald et al., 1994). It was previously observed that exogenous aprotinin, a serine protease inhibitor, inhibited Na+ transport in cortical collecting duct cells from the kidney (Margolius and Chao, 1980; Orce et al., 1980; Vallet et al., 1997; Nakhoul et al., 1998; Vuagniaux et al., 2002) and human bronchial epithelial cells from the lung (Bridges et al., 2001). Epithelial-derived serine proteases such as the channel activating proteases (CAPs, e.g., prostasin) activated ENaC-mediated Na+ transport when coexpressed in Xenopus oocytes (Vallet et al., 1997; Donaldson et al., 2002; Vuagniaux et al., 2002). The ability of some of these proteases, CAP1 and prostasin, to increase Na+ transport in the coexpression studies was inhibited by aprotinin. In addition to the CAPs, which represent endogenous proteases, ENaC-mediated Na+ transport can also be activated by exogenous trypsin, chymotrypsin, neutrophil elastase, and cathepsin G (Chraibi et al., 1998; Caldwell et al., 2005; Harris et al., 2007).
The notion that protease activation of Na+ transport results from cleavage of ENaC arises from the correlation between the appearance of cleaved forms of the α and γ subunits and the measured ENaC activity. First, salt restriction or aldosterone infusion (stimuli for increasing Na+ reabsorption in the kidney) were associated with the appearance of a lower molecular weight γ subunit in the rat cortical collecting duct while control rats had only the heavier γ subunit (Masilamani et al., 1999; Ergonul et al., 2006). Two molecular weight species of the α and γ subunit were found in whole rat kidneys where up-regulation of Na+ transport in collecting duct cells correlated with an increase of a lower molecular weight fragment of the γ subunit and decrease of the higher molecular weight species suggesting a conversion (Ergonul et al., 2006). Aldosterone up-regulation of prostasin in the cortical collecting duct cells (Narikiyo et al., 2002) may generate the γ subunit cleavage. Further, preventing internalization of ENaC in epithelial cells heterologously expressing ENaC increased the lower molecular weight isoforms of the α and γ subunits at the apical membrane and these changes correlated with decreased sensitivity to exogenous trypsin activation (Knight et al., 2006). Second, mutation of potential furin cleavage consensus sequences found in the α and γ subunits prevented endogenous cleavage of these subunits and reduced Na+ transport rates (Hughey et al., 2004a; Bruns et al., 2007).
The single channel studies by Caldwell et al. (2004, 2005) provided evidence that exogenously added serine proteases could activate the channel by direct proteolysis. These results suggest that proteases can directly bind to the channel and induce activation by cleavage. Based on the observations that trypsin and human neutrophil elastase could maximally activate silent channels, we tested the hypothesis that proteases directly interact with ENaC, causing activation of Na+ transport in a model epithelial cell line. Site-directed mutagenesis and analysis of the kinetics of current activation by the related proteases neutrophil elastase and porcine pancreatic elastases as well as thrombin were used to show that specific protease interactions with ENaC occur in a segment of the γ subunit.
Materials And Methods
Construction of Human α, β, γ ENaC Tripromoter Expression Vector.
Standard molecular biology methods were used to construct the ENaC tripromoter expression vector. The human α, β, and γ ENaC subunit cDNAs were isolated from human lung epithelial cells. All three cDNA sequences were identical to the sequence reported by McDonald et al. (1994) (GenBank/EMBL/DDBJaccession nos. L29007, L36592, and L36593). The α cDNA was subcloned into pECFPN1 and a stop codon introduced at the 3′ end of the α subunit to prevent the expression of the CFP. The β and γ subunits were subcloned into the pEGFPN1 vector (CLONTECH Laboratories, Inc.) to obtain pEGFPN1/β and pEGFPN1/γ. Appropriate start and stop codons were inserted to allow for the expression of the β and γ subunits without the C-terminal expression of EGFP. The DNA fragments from pEGFPN1/β and pEGFP/γ vectors comprising the CMV promoters and the subunit cDNA without the GFP sequences were isolated and then subcloned into pECFPN1/α to generate a 13.4-kb tripromoter αβγ hENaC vector (hENaC). The important features of the hENaC vector are that all three ENaC subunits are arranged in tandem and expression of each subunit is under an independent CMV promoter allowing for the separate expression of each subunit without the expression of ECFP or EGFP. Site-directed mutagenesis was performed using Quick Change II XL (Stratagene) for single and double amino acid substitutions. Overlap PCR was performed using Phusion high-fidelity DNA polymerase (Finnzymes) to introduce a human thrombin consensus sequence (LVPRG) by multiple contiguous substitutions in the γ186-190 sequence of hENaC. Overlapping fragments were synthesized by PCR with forward and reverse primers 5′-CTCGTGCCAAGAGGCTCAATGTCATGCACATCGAGTCC-3′ and 5′-GCTCCTCAGCAGAATAGCTCATGTTGATTTTCTTCTCC-3′. A second set of forward and reverse primers, 5′-CTGCAGGCCACCAACATCTTTGCACAGGTGCCACAGC-3′ and 5′-GCCTCTTGGCACCAACATCTTTGCACAGGTGCCACAGC-3′, containing Sbf I and BbvC I restriction sites were used in a second round of PCR to yield the blunt end product that was subcloned into the pCR 4Blunt-TOPO vector (Invitrogen), digested with Sbf I/BbvC I, and ligated to the appropriate Sbf I/BbvC I digest fragment of hENaC. V5 C terminal epitope–tagged γ ENaC was generated in pc DNA 3.1 hygromycin (Invitrogen) using a standard two-step overlapping PCR method. The sequences of all mutations were verified by DNA sequencing.
RNA Isolation and RT-PCR Experiments
Total RNA was prepared from parental or hENaC FRT cells, using the ThermoScript RT-PCR System (Invitrogen) according to manufacturer's instructions. cDNA was then subjected to PCR amplification using Finnzymes' Phusion High-Fidelity DNA Polymerase (New England Biolabs, Inc.). Primers were designed for identical sequences in human and rat γ ENaC. Primer sequences were as follows: forward 5′-GGA GAG AAG ATC AAA GCC AAA ATC-3′, reverse 5′-GCA TCT CAA TAC TGT TGG CTG GGC-3′. PCR products were separated on a 0.9% agarose gel and visualized by ethidium bromide staining. Amplified products were confirmed by DNA sequencing.
Cell Culture and Transfection
Fisher rat thyroid (FRT) cells were grown in sodium bicarbonate–buffered Hams's F-12 media (Sigma-Aldrich) modified with 10% FCS (Hyclone), 100 U/ml penicillin, and 100 U/ml streptomycin at 37°C and 5% CO2 as previously described (Sheppard et al., 1994). The cells were expanded in plastic tissue culture flasks. Confluent monolayers were stably transfected with hENaC or mutant plasmids using Lipofectamine 2000 (Invitrogen) and Opti-MEM media (Invitrogen) as described by the manufacturer's instructions and selected with G418 (500 μg/ml; Invitrogen). The selected cells were fed with media containing 50 μM amiloride. FRT cells were seeded onto permeable tissue culture inserts (Transwell Clear, 0.4 μM pore size, 6 mm diameter; Corning) and fed three times a week with or without supplementation with dexamethasone (30 nM) or aldosterone (30 nM). The cells matured for at least 10 d and amiloride was removed from the feeding media two feeding cycles before short-circuit current measurements.
Short-Circuit Current (ISC) Measurements
Current measurements were performed 10–14 d after seeding. The epithelial cell monolayers were studied by placing the Transwell tissue culture inserts into Costar Ussing chambers with apical and basolateral bathing solution containing (in mM) 120 NaCl, 25 NaHCO3, 3.3 KH2PO4, 0.8 K2HPO4, 1.2 MgCl2, 1.2 CaCl2, and 10 glucose as previously described (Bridges et al., 2001). The apical and basolateral solutions were continuously circulated with a gas lift. The pH of the solution was 7.4 at 37°C when gassed with a mixture of 95% O2:5% CO2. The monolayers were continuously short circuited with a voltage clamp (University of Iowa, Iowa City, IA), and transepithelial resistance was monitored by periodically applying a 4-mV bipolar pulse and calculating resistance from the current change. The amiloride-sensitive ISC (INa) obtained from the difference in current with and without amiloride (50 μM) in the apical bath was taken as a measure of ENaC-mediated electrogenic Na+ transport. The amiloride-insensitive current was minimal (<5% of the baseline ISC). Untransfected FRT cells did not show an INa. The proteases trypsin (Sigma-Aldrich), human neutrophil elastase (NE; EPC), porcine pancreatic elastase (PE; EPC), and human α thrombin (TH; Sigma-Aldrich) were added to the apical bath and mixed by rapid pipetting with a transfer pipette.
Protocol for INa Activation by Proteases
To observe significant and reproducible protease activation it was necessary to pretreat the cells with the serine protease inhibitor aprotinin. Several epithelia appear to have constitutive activation of Na+ transport that can be inhibited by aprotinin in a time-dependent manner and the transport rate rapidly reverts to control levels upon addition of exogenous trypsin (Vallet et al., 1997; Bridges et al., 2001; Adebamiro et al., 2005; Myerburg et al., 2006). The slow time dependence of aprotinin inhibition and rapid reversal by excess protease suggests that in the presence of aprotinin, active channels are continually retrieved or degraded while new inactive channels are inserted. This pool of newly inserted channels can then be activated by exogenous proteases. The time course of aprotinin inhibition of current in mammalian airway cells has a half life of ∼45 min (Bridges et al., 2001). To ensure maximal inhibition for all constructs examined in experiments involving pre-protease inhibition, the protease inhibitor, aprotinin (10 μM; Sigma-Aldrich) or a vehicle control (PBS) was added to the apical surface of the monolayers overnight as previously reported (Vallet et al., 1997) in the modified Ham's F-12 media 1 and remained in the apical bath solution during ISC measurements. Overnight treatment reduced the variability in the inhibition of INa by aprotinin and improved the quantitative analysis of the results.
Data Evaluation and Analysis
The effects of the proteases were assessed in two ways. First, steady-state INa was assessed before protease addition (baseline INa), then after the INa reached a steady-state 20–30 min after addition of NE, PE, or TH (IPR). Trypsin was then added in excess over the aprotinin concentration to obtain ITryp. To compare the effects of NE, PE, and TH to that of trypsin we used the ratio IPR/ITryp and the normalized value ΔIPRNorm where ΔIPRNorm = (IPR − baseline INa)/(ITryp − baseline INa). IPR was measured 20–30 min after the addition of protease (NE, PE, or TH) and ITryp was measured 5 min after the addition of trypsin. Second, following addition of NE, PE, or TH the time course of the ΔIPRNorm in each experiment was analyzed according to two reactions. The first reaction is shown in Eq. 1.
The quantity C is the inactive channel at the membrane, E is the added enzyme concentration, and A is the channel activated by proteases. The rate coefficients kon and koff have their usual meanings. The differential equation from Eq. 1 is
Under the assumption that the enzyme concentration is constant due to being in excess and there is no autolysis, the differential equation is solved to yield exponential functions in time that can be fitted to the whole time course of current activation (see online supplemental material, available at http://www.jgp.org/cgi/content/full/jgp.200709781/DC1). The time courses for individual experiments were fitted using Matlab (The Mathworks) to Eq. 3.
A linear term was added to Eq. 3 to account for a slow decreasing trend in the current records. The reciprocal of the time constant τ was taken as a measure of the apparent relaxation rate constant kobs. It is shown in the supplemental material that the kobs is given by
The second reaction is of the form shown in Eq. 5.
The differential equations from Eq. 5 are
The koff and kon have their usual meanings and kcat is simply used to express a rate-limiting enzyme concentration-independent step. It is shown in the online supplemental material that the time course of activation can also be approximated by Eq. 3 but here the kobs is given by Eq. 8:
Linear and nonlinear regression analysis of the kobs versus enzyme concentration was performed in OriginPro (OriginLab).
Matrix-assisted Laser Desorption Ionization-Time of Flight Analysis (MALDI-TOF)
The sites of cleavage by NE, PE, and TH were searched by MALDI-TOF mass spectrometry on a 23-mer peptide designated peptide T176-S198 (Ac-TGRKRKVGGSIHKASNVMHIES-NH2) and corresponding to the amino acid sequence T176 to S198 of the γ subunit of ENaC. Additional peptides with glycine substitutions at V182, V193, and A190 as well as a peptide substituted with a TH consensus sequence (LVPRG) beginning at residue 186 were evaluated. Enzymatic digestions were performed at 37°C with NE, PE, and TH at final enzyme concentrations of 200 nM and peptide T176-S198 at 200 μM in 200 mM Tris acetate, pH 7.4. Following the incubation period, 100 μl of the reaction was quenched by lowering the pH to 3 with addition of 1.0% trifluoroacetic acid (TFA). The reaction without and with the enzymes were subjected to MALDI-TOF measurements. A ZipTip C18 solid-phase extraction sorbant (Millipore) was washed with 10 μl of acetonitrile/water (1:1, vol/vol) and then equilibrated with aqueous 0.1% TFA. Peptide samples acidified with 5% aqueous TFA to enhance peptide retention on the stationary phase of the C18 sorbant were loaded onto the tip and eluted with 0.1% TFA in acetonitrile/water (1:1, vol/vol). The matrix solution was prepared by dissolving 10 mg of cyano-4-hydroxycinnamic acid in 1 ml of acetonitrile/water (1:1, vol/vol) containing 0.1% TFA. 0.5 μl of the matrix solution was mixed with 0.5 μl of peptide solution. About 0.5 μl of the mixture was deposited onto the MALDI stage and air dried. Positive ion MALDI-TOF Mass spectra were acquired on an Applied Biosystems Voayger DE-Star mass spectrometer operated in reflector mode. Following time-delayed extraction, the ions were accelerated to 20 kV for TOF analysis. A total of 100 laser shots were acquired and signal averaged per spectrum.
In Vitro Translation and Western Blots
cDNAs were transcribed and translated in vitro in the presence of canine pancreatic microsomal membranes using the TnT-coupled reticulocyte lysate systems (Promega). Total DNA templates (0.5 μg) containing 0.25μg αβ and 0.25μg γ-V5 or 0.25 μg of γTh-V5 ENaC subunits were added to an aliquot of the TnT T7 Quick Master mix containing 2 μl microsomal membranes and incubated at 30°C for 60 min. The translation products were collected by centrifuging for 15 min at ∼10,000 g. and subjected to lysis in 1% Triton X-100 buffer (1% Triton X-100, 500 mM NaCl, 5 mM EDTA, 50 mM Tris-HCl, pH 7.5) at 4°C . The lysates were incubated for 5 min at 37°C in the presence or absence of 200 nM human α thrombin (Sigma-Aldrich). All samples in 1× SDS-PAGE sample buffer containing 100 mM DTT were heated for 10 min at 70°C before loading. Proteins were separated on NuPAGE 4–12% Bis-Tris Gel (Invitrogen) and were transferred to Immobilon-P membrane (Millipore). Proteins were detected by Western blot with mouse monoclonal anti-V5 antibody (Invitrogen) targeting a V5 epitope on the C terminus of the γ ENaC subunit. The signal was developed with the Supersignal west femto maximum sensitivity substrate (Pierce Chemical Co.) and detected with X-Omat Blue XB-1 imaging film (Kodak).
Effects of the proteases on steady-state INa for the wild-type hENaC were determined using paired t tests of (1) INa at baseline versus IPR or ITryp and (2) IPR versus ITryp. Comparison of the ratio of ΔIPRNorm and τ across multiple mutants was performed with one-way ANOVA followed by individual comparisons versus wild-type ENaC using the Bonferroni corrected t test in Origin . P < 0.05 was considered significant.
Online Supplemental Material
Expression of ENaC-mediated Na+ Transport in FRT Cells and Regulation by Corticosteroids and Protease
FRT epithelial cells stably transfected with the α, β, and γ human ENaC subunits in a single tripromoter vector (hENaC) had an ISC that was blocked by addition of amiloride (50 μM) to the apical side (Fig. 1, A and B). The magnitude of the amiloride-sensitive ISC (INa) averaged 3.4 ± 0.12 μA/cm2 (n = 11) and is similar to what has been typically observed by transiently transfecting FRT cells simultaneously with the three ENaC subunits in separate plasmids (Snyder, 2000). Feeding the cells in media supplemented with either aldosterone (30 nM) or dexamethasone (30 nM) for 72 h or more before short circuiting significantly increased INa with no increase in the amiloride-insensitive ISC. Aldosterone induced a 1.5-fold (n = 7) increase in INa while dexamethasone induced an eightfold (n = 11) increase in INa (Fig. 1 B). The half maximal inhibitory concentration (Ki) for amiloride was 570 ± 86 nM with a maximal inhibition of 96 ± 1.2%. To determine whether proteases regulate Na+ transport in FRT cells stably expressing hENaC, the effect of exogenously added trypsin and protease inhibitor were examined. The cells were first preincubated either with the protease inhibitor aprotinin (10 μM) or a vehicle control (PBS) on the apical side for 16–24 h. The cells that were preincubated with aprotinin had decreased INa compared with cells preincubated with PBS. During short circuiting, addition of trypsin (15 μM) to the apical bath, in a 5 μM excess over the aprotinin concentration, caused no statistically significant increase in the INa of PBS-preincubated cells (3.9 ± 0.87 to 4.5 ± 1.03 μA/cm2; n = 11) but significantly increased INa in aprotinin-pretreated cells. The baseline INa in the aprotinin-pretreated cells was 3.1 ± 0.68 and the maximal INa after addition of trypsin (ITryp) was 5.9 ± 0.68 μA/cm2 (n = 12). No differences were found in the trypsin-activated current between PBS- and aprotinin-treated cells. Consequently, aprotinin unmasks a portion of INa that can be regulated by extracellular protease activity similar to the previously reported effects of aprotinin and trypsin. The absolute values of the aprotinin and trypsin effects on INa were greatly enhanced in dexamethasone-treated cells but displayed the same qualitative pattern (Fig. 2, A and B). As already noted, dexamethasone treatment greatly increased the INa (37.0 ± 1.19 μA/cm2, n = 18). Trypsin addition had no effect on INa (37.1 ± 1.14 μA/cm2, n = 18) in PBS-pretreated dexamethasone-stimulated cells. Pretreatment with aprotinin caused a 50% decrease in the baseline INa (21.3 ± 0.85 μA/cm2, n = 18) and in aprotinin-inhibited cells trypsin doubled the INa (40.8 ± 2.09 μA/cm2, n = 18) to the value observed in PBS-pretreated cells (Fig. 2, A and B). An excess of trypsin above the aprotinin concentration was required to stimulate INa. In addition, the stimulatory effects of trypsin on INa in aprotinin-treated cells was blocked by soy bean trypsin inhibitor (ΔINa with SBTI 0.7 ± 1.22 μA/cm2, n = 6). Because the dexamethasone-stimulated hENaC-expressing FRT cells gave a more robust and reproducible INa and the protease regulatory pattern persists in dexamethasone-treated cells, dexamethasone-treated cells were used in the remainder of the study. Dexamethasone treatment of untransfected FRT cells did not produce amiloride- or trypsin- sensitive ISC. In untransfected parental cells treated with PBS, the ISC before trypsin, after trypsin, and after amiloride addition were 0.1 ± 0.06, 0.2 ± 0.06, and 0.2 ± 0.07 μA/cm2 (n = 12), respectively. For cells treated with aprotinin, the ISC were 0.7 ± 0.3, 0.5 ± 0.16, and 0.5 ± 0.16 μA/cm2 (n = 12) before trypsin, after trypsin, and after amiloride, respectively. Furthermore, in cells treated with 30 nM dexamethasone that were stably transfected with the α and β hENaC subunits but without the γhENaC subunit there was no INa before trypsin (0.2 ± 0.12) μA/cm2, after trypsin (0.3 ± 0.08 μA/cm2), and amiloride (0.4 ± 0.09 μA/cm2, n = 6). RT-PCR analysis of RNA derived from parental FRT cells without or with dexamethasone treatment did not detect any PCR product for rat γENaC but the expected γ hENaC product was detected with RNA derived from the hENaC-transfected FRT cells (Fig. 3). Thus dexamethasone-stimulated expression of endogenous rat γENaC and complementation with α and β hENaC subunits cannot explain the INa observed in our studies. The importance of this assertion will become apparent in subsequent results.
In addition to modulating sodium transport, aprotinin and trypsin also altered the transepithelial resistance (RT). Aprotinin treatment caused a 50% decrease in RT compared with PBS-treated hENaC-FRT cells (aprotinin 502 ± 47 Ω cm2 vs. PBS 769 ± 45 Ω cm2, n = 6) and this effect was reversed by trypsin (aprotinin plus trypsin 1555 ± 125 Ω cm2 vs. PBS plus trypsin 1215 ± 101 Ω cm2, n = 6). Amiloride further increased RT in both aprotinin- and PBS trypsin–treated cells to 2614 ± 164 Ω cm2 and 3592 ± 289 Ω cm2 (n = 6), respectively. Amiloride, when added before trypsin, did not prevent the increase in RT caused by trypsin in the aprotinin-treated cells (Δ RT aprotinin plus trypsin 1053 Ω cm2 vs. Δ RT aprotinin plus amiloride plus trypsin 1113 Ω cm2, n = 3). The effects of aprotinin and trypsin on RT were also observed in parental untransfected FRT cells (unpublished data) and these effects have been reported by others in different epithelia (Lynch et al., 1995; Liu et al., 2002). It has been suggested these effects are due to changes in the resistance of the tight junction (Meyer et al., 1988; Kemmner and Zaar, 1991). Serosal to mucosal fluxes of the extracellular maker mannitol revealed aprotinin treatment increased the flux by 60% when compared with PBS-treated hENaC FRT cells (aprotinin 1.3 ± 0.04 nMol/h/cm2 vs. PBS 0.8 ± 0.05 nMol/h/cm2, n = 6) and this effect was reversed by trypsin (aprotinin plus trypsin 0.4 ± 0.07 nMol/h/cm2 vs. PBS plus trypsin 0.5 ± 0.05 nMol/h/cm2, n = 6). Because we studied sodium transport under short-circuited conditions and in symmetric solutions these changes in the paracellular pathway resistance are not expected to influence our measurements of INa. Furthermore as shown below NE, PE, and thrombin do not alter RT but do increase INa in aprotinin-treated hENaC-FRT cells.
Submicromolar Neutrophil and Porcine Elastase Activate ENaC in FRT Cells
It was previously shown that, like trypsin, NE can activate ENaC-mediated Na+ current (Caldwell et al., 2005; Harris et al., 2007). We examined the effect of NE and PE, a similar elastase (Baugh and Travis, 1976; Travis et al., 1979), on the ISC in FRT cells expressing hENaC. Elastase is not inhibited by aprotinin (Baugh and Travis, 1976; Ohlsson and Olsson, 1976), therefore an excess over the aprotinin concentration was not required to stimulate INa. Addition of NE (300 nM) or PE (300 nM) to the apical side of FRT cells expressing hENaC did not alter the INa in cells pretreated with PBS (Fig. 4 and Table I). Subsequent addition of trypsin did not further increase the INa in PBS pretreated cells. Similar to the effects of trypsin NE and PE clearly increased the INa in cells that were pretreated with 10 μM aprotinin as shown in Fig. 4 (B and D). The rise in the INa following elastase addition occurred over a 5-min period and increased to a plateau value equal to the baseline INa of uninhibited cells that were preincubated with PBS. Addition of trypsin subsequent to NE or PE produced small further increases in the INa in aprotinin-treated cells (P < 0.01). The irreversible NE inhibitor MeOSuc-Ala-Ala-Pro-Ala-CMK at 4 μM nearly completely prevented the stimulation of INa by NE (300 nM) (ΔINa without inhibitor 24.1 ± 4.21 μA/cm2, n = 6; ΔINa with inhibitor 2.7 ± 0.60 μA/cm2, n = 6) and confirm the results of Harris et al. (2007) using a novel NE inhibitor in Xenopus oocyte studies. Thus, NE or PE increased INa in aprotinin-inhibited cells to the values observed in control PBS-treated cells and only a small additional further stimulation was seen with trypsin addition in the case of NE-stimulated cells (Fig. 4, E and F, and Table I).
Identification of Elastase-specific Sites I: Human Neutrophil Elastase
To determine whether specific sites on ENaC mediate the NE-dependent activation of INa, we performed site-directed mutagenesis of possible NE binding or hydrolytic sites in the extracellular loops of α and γ ENaC subunits. NE hydrolyzes polypeptides with a strong preference for cleavage of V-X bonds where V is at the P1 position and X is an amino acid at the P1' position. The preference is a function of both the binding affinity of the elastase for the cleavage site (KM) and the rate of hydrolysis (kcat). Hydrolysis by NE can also occur with other small hydrophobic amino acids at the P1 position but to a lesser degree (Zimmerman and Ashe, 1977). We tested the effect of mutating individual valines conservatively to glycines in the extracellular loops of α and γ ENaC because these subunits appear to be involved in activation and or processing by endogenous proteases (Masilamani et al., 1999; Hughey et al., 2004a; Ergonul et al., 2006; Knight et al., 2006). The valine mutation sites were chosen based on their proximity to previously implicated cleavage sites (Hughey et al., 2004a). A total of eight V to G mutants were evaluated, three in the α subunit and five in the γ subunits. Stable cell lines expressing these mutants were selected and the effects of aprotinin, NE, and trypsin on INa were measured. These results are summarized in Table I. All eight V to G mutants expressed baseline currents that were similar to wild-type hENaC-expressing cells. Furthermore aprotinin inhibited the baseline current to a similar degree and NE reversed this inhibition of INa (Table I). Only the double mutant αβγV182G;V193G was different from wild-type ENaC (WT). Unlike wild-type ENaC and several mutants where subsequent addition of trypsin had a minor further effect on INa following NE stimulation in aprotinin-treated cells (Table I, Fig. 4 B, and Fig. 5, A–D), αβγV182G;V193G could be substantially further stimulated by trypsin (Table I and Fig. 4, E and F). As shown in the figures, addition of NE to aprotinin-treated FRT cells expressing wild-type ENaC increased INa from 15.6 ± 0.86 μA/cm2 to 27.8 ± 1.40 μA/cm2 (n = 36) within 10 min. Thereafter, addition of trypsin caused only a small further increase to 31.3 ± 1.60 μA/cm2. Thus NE caused an increase that was ∼85% of the trypsin- induced maximal protease increase in wild-type ENaC-expressing cells. In contrast addition of NE to aprotinin-treated FRT cells expressing the αβγV182G;V193G mutant increased INa from 18.3 ± 0.67 to 25.9 ± 1.07 μA/cm2 (n = 39), but the ISC could be further stimulated to 39.9 ± 1.12 μA/cm2 after the addition of trypsin. Thus in the αβγV182G;V193G cells, NE only stimulated ∼36% of the INa stimulated by trypsin.
A closer examination of the current traces revealed that the rate of activation by NE was also affected by the valine to glycine mutations and this effect was investigated in greater detail. The changes in INa following addition of NE (300 nM) to wild-type and mutant ENaCs were monitored with time and reported (Fig. 6 A) as a fraction of the maximal trypsin-stimulated INa. The fractional current increase normalized to the maximum response after trypsin (ΔIPRNorm) was fitted to exponentials as described in Materials and methods to yield time constants. The ΔIPRNorm ∼30 min after NE addition was taken as the measure of maximal NE-stimulated current. The fitted parameters are summarized in Table II for αβγ and the glycine scanning mutations showing that only specific mutations in the γ subunit appreciably change the time course. Representative time courses for wild type, αβγV182G, αβγV193G, and αβγV182G;V193G are shown in Fig. 5 A. In wild-type and αβγV182G-expressing cells the ΔIPRNorm rose with identical rates and to similar plateau values with amplitudes 0.68 ± 0.05 and 0.64 ± 0.04, respectively. Therefore, mutating V182 to G had no effect on NE activation of current. The cells expressing αβγV193G had ΔIPRNorm of 0.53 ± 0.05, which was not significantly decreased from wild type; but they showed a significant elongation of the time constant from the wild-type value of 3.1 ± 0.18 min (n = 10) to 6.3 ± 0.36 min (n = 5). As noted above, cells expressing the double mutant, αβγV182G;V193G, had a marked reduction of ΔIPRNorm (0.22 ± 0.02) as well as a further elongation of the time constant to 11.8 ± 1.49 min (n = 8). Therefore, while mutating the V182 to G182 had no discernible effect on its own, this mutation produced a pronounced effect in the presence of the already inhibitory V193G mutation. To differentiate an effect caused by changing a V from an effect introduced specifically by the G, several amino acids were substituted at positions 182 and 193. The effect of mutating γ V182 and V193 was not specific to glycine but consistent for several amino acid substitutions (S, T, D, E, and Q). Generally, substitutions at γ182 did not result in significant differences in the τ for the residues tested (Fig. 6). At residue 193, all the substitutions except E caused a significant prolongation of τ compared with wild type (Fig. 6 B). Substitutions at both γ182 and γ193 produced significant elongation of τ for all the substitutions compared with wild type (Fig. 6 B). The changes in ΔIPRNorm caused by the mutations were roughly consistent with the changes in τ. Generally, ΔIPRNorm was not affected by substitutions at γ 182 only, except with the S substitution (Fig. 6 C). The ΔIPRNorm was reduced for substitutions at γ 193, except for the G substitution (Fig. 6 C). All double substitutions at 182 and 193 resulted in decreased ΔIPRNorm compared with wild type. These results are more consistent with the conclusion that removal of a valine (as opposed to introduction of a specific residue) was responsible for both prolongation of τ and decreased ΔIPRNorm. Secondarily, the conserved pattern where substitutions at 182 produce no effect; substitutions at 193 increased τ and decreased ΔIPRNorm; and double substitutions further increasing τ and decreasing ΔIPRNorm suggest that the V193 is of primary importance in current activation by NE but V182 also has some effect unmasked in the absence of V193.
Identification of Elastase-specific Sites II: Porcine Pancreatic Elastase
First, we tested if the valines critical for NE were also critical for PE activation of current. ISC following addition of PE (300 nM) was monitored to determine the effect of simultaneous substitutions at residues 182 and 193 on PE-mediated channel activation. The effects of the double substitutions on ΔIPRNorm were small compared with the results for NE. As shown in Fig. 7 (A and B), the time courses for wild type and G substitutions were slightly prolonged. The τ was prolonged from 2.0 ± 0.14 min (n = 11) for wild type to 3.6 ± 0.27 min (n = 11) for the G substitutions. The effect on ΔIPRNorm was not different between wild type and G substitutions at ∼0.7 for both (n = 16 and 23, respectively; Fig. 7 C). The rest of the double substitutions examined (S, Q, E, and T) had small effects on τ and ΔIPRNorm (Fig. 7, D and E). These results show that the valines 182 and 193 were not critical for increasing INa by PE.
Unlike NE, the preferential residue at P1 is A for PE (Hartley and Shotton, 1971). Therefore one explanation for minor effects of the V substitutions is that PE interacts with the channel via its preferred residue somewhere else. We reasoned that γ A190 between the two valines critical for NE was a likely target. Substitution of the γ A190 to G resulted in significant attenuation of the current response to PE (Fig. 7, F–H). While 1,000 nM PE activated current in wild type and the αβγV182G;V193G double mutant ENaC (Fig. 7, F and G), with τ of 1.94 ± 0.12 and 2.17 ± 0.07 min, respectively, the mutant αβγA190G prevented PE activation of current. Because increases in INa were small to nonexistent in the αβγA190G mutant in response to PE, τ could not be resolved for all experiments. For those experiments where τ was resolved, it averaged 25.6 ± 1.09 min (Fig. 7 H). Furthermore, the ΔIPRNorm, which was 0.70 ± 0.04 and 0.89 ± 0.03 for wild type and αβγV182G;V193G, respectively, was reduced to 0.26 ± 0.02 for αβγA190G (Fig. 7 I). These results are consistent with γ A190 being of primary significance for PE current stimulation, whereas the Vs are more important for NE current stimulation. Because the results are consistent with the P1 preferences for PE and NE, they suggest that these enzymes interact with the channel according to their substrate recognition specificities.
Concentration Dependence of Protease Activation
Because the previous results suggest direct protease–ENaC interaction, a concentration dependence of activation was used to delineate two mechanistic possibilities of current activation: (1) that direct binding was sufficient for channel activation or (2) that binding plus a secondary step is necessary. As can be seen from the progress curves of ΔIPRNorm (Fig. 8 A) the rates of current activation as well as the amplitudes of the increase were concentration dependent. Increasing NE concentration from 30 to 1,000 nM reduced τ for cells expressing wild type from 11.3 ± 1.5 min (n = 8) to 1.6 ± 0.04 min (n = 8). The ΔIPRNorm increased with increasing concentration of NE from 0.22 ± 0.02 at 30 nM to 0.75 ± 0.05 at 1,000 nM (Fig. 8 D). The τ and ΔIPRNorm for cells expressing the αβγV182G; V193G mutant were also concentration dependent (Fig. 8, B and D). Time constants decreased from 13.8 ± 3.2 min (n = 8) at 30 nM NE to 5.9 ± 0.3 min (n = 8) at 1,000 nM NE, and ΔIPRNorm increased from −0.02 ± 0.04 to 0.52 ± 0.02. The concentration dependence of current activation suggests a first approximation of the mechanism of activation of current by NE and PE. Because only one exponential component was present in the fitting of the time courses, the reciprocal of the derived time constants were taken as a pseudo-first-order rate coefficient, kobs. For wild-type ENaC, the pseudo-first-order rate constant (kobs) showed saturation behavior with respect to the added enzyme concentration (Fig. 8 C). From the saturation at relatively slow rates, assuming that the enzyme concentration in the chamber was not altered by concentration-dependent autolysis and that the enzyme concentration is in a large excess over its target ENaC, apparent kinetic parameters were derived based on Eq. 8 by nonlinear regression of kobs against the added enzyme concentration.
In cells expressing wild-type ENaC the KM was 570 ± 240 nM NE and the kcat was 14.4 ± 2.11 × 10−3 s−1. When the αβγV182G;V193G was expressed, the kobs also showed a dependence on enzyme concentration (Fig. 8, B–D). However, the dependence of kobs on enzyme concentration, in contrast to wild-type ENaC, did not show an apparent saturation; consequently the parameters kcat and KM could not be calculated. One reason why this arises may be because the concentration of NE used was well below the KM. A linear regression was performed under the assumption that [E] ≪ KM. The slope of the regression, kcat/KM, was 1,300 ± 200 × M−1s−1. This quotient represents the efficiency of NE activation of the γV182G;V193G double mutant. This quotient for wild-type ENaC was calculated from the fitted parameters and was ∼25,000 M−1s−1. Thus NE efficiency for activating wild-type ENaC is ∼20 times that for activating the αβγV182G;V193G mutant. The ΔIPRNorm current plateau values at each NE concentration were also fitted to a saturation curve of the form
where (ΔIPRNorm)max is the maximum increase and K1/2 is the enzyme concentration at half maximal stimulation. For NE on wild-type ENaC the K1/2 was 100 ± 18 nM, and (ΔIPR/ITryp)Max was 0.84 ± 0.04. For NE on the αβγV182G;V193G mutant, K1/2 was 510 ± 162 nM and (ΔIPRNorm)Max was 0.94 ± 0.14. Consequently, the apparent affinity for NE activation of INa is ∼5-fold weaker for the αβγV182G;193G mutant compared with wild-type ENaC.
A concentration dependence of activation was also observed for PE activation of the INa. The same analysis as described above was performed for the PE datasets. For wild-type ENaC, the saturation behavior of the activation rate (Fig. 9 A) had a KM of 204± 140 nM and kcat was 11.4 ± 1.34 × 10−3 s−1. For the mutant αβγV182G;V193G, the dependence of kobs on PE concentration was well defined, in contrast to what was observed for NE. Unlike NE's effect on this mutant, there was saturation with PE demonstrating a KM of 290 ± 51 nM and a kcat of 8.7 ± 0.39 × 10−3 s−1. Therefore, unlike for NE activation of INa these mutations had a minimal effect on PE activation of INa. The parameters for the ΔIPRNorm versus concentration were K1/2 of 77 ± 30 nM and 35 ± 9.2 nM for both wild type and αβγV182G;V19G, respectively, and (ΔIPRNorm)Max were 0.7 ± 0.2 and 1 ± 0.05, respectively. Clearly for both wild type and αβγV182G;V193G, activation saturation occurred at low enzyme concentrations (Fig. 8 B) and no difference could be discerned between the two constructs. In contrast to these minor effects, the concentration dependence of PE activation of INa in the αβγA190G mutant was significantly impaired. The kobs increased slowly with enzyme concentration, showing no apparent saturation (Fig. 9 A). Linear regression gave kcat/KM of ∼550 M−1s−1. Thus PE activation of wild-type ENaC was ∼100-fold more efficient than PE activation of αβγA190G. This loss in efficiency is accompanied by a 15-fold increase of the K1/2 for ΔIPRNorm to 1,140 nM (Fig. 9 B).
NE and PE Cleavage of the γ ENaC Segment
Next NE and PE were evaluated for their ability to cleave the identified region of γ ENaC and the specific proteolytic sites were determined by MALDI-TOF mass spectrometry. The peptide T176-S198 (200 μM), corresponding to the amino acid sequence T176 to S198 of the γ ENaC subunit, was incubated with 200 nM NE or PE in tris-acetate buffer. MALDI-TOF mass spectrometry showed the undigested peptide T176-S198 at m/z of 2546.6 D. Incubation of peptide T176-S198 with NE for 5 min resulted in a new peak at 1950.9 D corresponding to the molecular weight of peptide T176-V193 as well as a new peak at 615.3 D corresponding to the peptide M194-S198 along with undigested full-length peptide at 2547.1 D (Table III). Mass spectrometry of the PE digest showed the appearance of new peaks at 1650.8 D and 915.2 corresponding to peptides T176-A190 and S191-S198, respectively (Table III). These fragments indicated that NE hydrolyzed the peptide between V193 and M194 primarily, whereas PE hydrolyzed the peptide between A190 and S191. Cleavages at the specific sites were detected within 1 min of incubation. After prolonged incubation with NE (30 min), a peak at 886.0 D corresponding to the peptide T176-V182 was also evident but minor, suggesting that cleavage by NE can occur between V182 and G183 but is not efficient. Substitution of V182 and V193 with glycines prevented cleavage of the peptide by NE and glycine substitution of A190 prevented cleavage by PE. Consequently, a peptide with a sequence contained within γ ENaC can be cleaved by NE and PE at the cleavage sites corresponding to the residues that mediate channel activation by the enzymes.
Thrombin Activation of ENaC
To further investigate the idea that channel activation could be mediated by protease ENaC interaction at the sequence between γ182 and γ193, the γ 186 to 190 sequence (IIHKA) was changed to γ 186…LVPRG to give the mutant αβγTh. The LVPRGS sequence matches the specific thrombin consensus sequence of bovine factor XIII (Takagi and Doolittle, 1974) where thrombin cleaves after R and is often used for cleavage of fusion proteins (Jenny et al., 2003). Aprotinin is not known to inhibit thrombin (Scott et al., 1987; Auerswald et al., 1988; Pintigny and Dachary-Prigent, 1992). Human α thrombin (TH), in contrast to NE and PE, did not elicit significant current increases in FRT cells expressing wild-type ENaC, which were aprotinin pretreated (Fig. 10, A and C). The baseline INa (9.8 ± 1.44 μA/cm2) was not significantly increased after 300 nM TH (11.5 ± 0.91 μA/cm2) but subsequent trypsin addition increased INa to 23.7 ± 0.65 μA/cm2 (n = 8). For cells expressing αβγTh, 300 nM TH elicited rapid current increases (Fig. 10, B–D) as did 300 nM NE. On average, TH increased INa from a baseline of 14.3 ± 1.32 to 23.8 ± 1.17 μA/cm2, which was not significantly further increased by trypsin addition (24.9 ± 0.72 μA/cm2; n = 8). FPR-chloromethyl ketone completely prevented the stimulation of INa by thrombin (ΔINa with inhibitor −1.5 ± 1.18 μA/cm2, n = 6). A comparison of the ΔIPRNorm (Fig. 10 D) showed that while NE and PE activated wild-type ENaC (0.6 ± 0.02 and 0.7 ± 0.01, n = 8–16), TH had only a small effect (0.1 ± 0.02, n = 4). On the other hand TH activated αβγTh (ΔIPRNorm = 0.9 ± 0.02, n = 8) as did NE (ΔIPRNorm = 0.7 ± 0.04, n = 4). Interestingly, while NE also activated αβγTh, activation of αβγTh by PE was significantly impaired (0.2 ± 0.04, n = 4). This result was consistent with the change of γ190A to G, which previously led to inhibited responses to PE. The activation of αβγTh by thrombin was similar to activation of wild-type ENaC by NE and PE, showing a concentration dependence of both the kobs and steady-state ΔIPRNorm (Fig. 10, E and F). The kcat was 10.1 ± 0.13 × 10−3 s−1, KM was 340 ± 19.5 nM, and K1/2 was 14.8 ± 1.98 nM. Thrombin did not cleave the wt T176- S198 peptide but did cleave the peptide substituted with a thrombin consensus sequence (Table III). In vitro translation studies of αβ hENaC together with wt γ or γTh revealed thrombin cleavage of γTh but not wt γ (Fig. 11). In vitro translation of γ hENaC yielded two bands, one at 74 kD and one at 82 kD. Treatment of γTh with thrombin resulted in the appearance of a lower mass C-terminal band at 53 kD. The lower mass band at 53 kD is the expected size for cleavage of γTh at R189 in the thrombin consensus sequence. These results demonstrate thrombin can cleave γTh and this cleavage may mediate channel activation. Several attempts were made to detect a thrombin-liberated γ fragment in αβVth-expressing FRT cells. However the results from these studies were negative and we believe due to the low abundance of ENaC expression in our stable FRT cell lines.
The studies reported here were made possible by the development of a tripromoter plasmid expressing all three human ENaC subunits (hENaC). Cell lines stably expressing wild-type and mutant ENaC could be selected using a single antibiotic. FRT cells were chosen because they are readily transfected, do not express any native rat ENaC, and form a moderately tight epithelium (1–2 kΩcm2). Previous studies in FRT cells using three separate ENaC subunit plasmids have shown ENaC is targeted to the apical membrane, allowing one to use standard transepithelial short circuit current (ISC) measurements to study channel activity (Snyder, 2000; Snyder et al., 2004; Knight et al., 2006). Our studies reveal hENaC-FRT cells respond to aldosterone and dexamethasone treatment in a manner similar to native ENaC expressing epithelia (Garty and Palmer, 1997). Dexamethasone-treated hENaC-FRT cells provided robust INa that were aprotinin inhibited and exogenous protease activated. We used dexamethasone treated, aprotinin-inhibited hENaC-FRT cells to investigate the mechanism of channel activation by exogenous protease. The preponderance of evidence, including functional measurements of INa responses to proteases, mutational analysis, kinetic analysis of protease concentration and time-dependent responses, mass spectrometry of peptide cleavage products, and thrombin cleavage of γTh ENaC, support the conclusion that exogenous proteases directly interact with the γ subunit to cause channel activation and that channel activation may involve cleavage of the γ subunit.
Approximately 50% of the INa in the hENaC-FRT cells was inhibited by the serine protease inhibitor aprotinin. A similar aprotinin inhibitory effect on INa of 50–80% has been observed in a number of epithelia that natively express ENaC and includes the following: toad urinary bladder (Margolius and Chao, 1980; Orce et al., 1980) amphibian (Vallet et al., 1997; Adebamiro et al., 2005) and murine (Nakhoul et al., 1998; Liu et al., 2002) renal epithelia, and human bronchial (Bridges et al., 2001; Myerburg et al., 2006) nasal (Donaldson et al., 2002), and alveolar (Planes et al., 2005) epithelia. One may infer from these observations that there exists an endogenous aprotinin-sensitive protease responsible for ENaC activation. Support for this notion comes from the cloning of several membrane proteases that when coexpressed with ENaC cause channel activation (Vallet et al., 1997; Vuagniaux et al., 2002; Andreasen et al., 2006). The inhibitory effects of aprotinin on INa can be reversed by aprotinin washout or by the addition of excess exogenous proteases such as trypsin and elastase and this too was observed in the hENaC-FRT cells. The aprotinin effect is consistent with the notion that activation of channels by proteases at the apical membrane is blocked by aprotinin, producing a time-dependent reduction of the number of active channels (Adebamiro et al., 2005) that results from internalization or degradation of already activated channels. Channels newly arriving at the apical membrane then form a pool that can be activated by exogenous proteases. It was recently shown that a more substantial increase in current by exogenous trypsin could be seen in transiently transfected FRT cells expressing ENaC (Knight et al., 2006) unlike the small increases we observed. The difference may result from our use of stably transfected cells with a long incubation period on filters. In the studies reported here we used the hENaC-FRT cells and site-directed mutagenesis of hENaC together with the amino acid preferences of NE and PE to determine whether there is a direct interaction between ENaC and proteases in the channel activating process.
There are several possible mechanisms whereby the aprotinin inhibition of INa can be reversed by exogenous proteases in FRT cells. The binding of aprotinin by the proteases can be ruled out since the concentrations of proteases that activate INa are much less than the aprotinin concentration and insufficient to form an aprotinin sink. In other systems such as Xenopus oocytes and fibroblasts, activation by trypsin and NE can be seen in the absence of aprotinin. Protease-dependent effects on trafficking are possible but unlikely given the observations that trypsin did not increase cell surface quantities of channel subunits (Chraibi et al., 1998) and trypsin is able to activate current in excised outside-out membrane patches (Caldwell et al., 2004). A role for trafficking is also unlikely based on the mutational studies reported here and discussed below, indicating that the activation requires binding. The activation of ENaC by extracellular proteases may result from protease interaction with a protease responsive receptor which subsequently activates ENaC or from a direct protease interaction with ENaC. A requirement for a diffusible second messenger has not yet been identified for the protease dependent activation and in particular signaling through G proteins via protease-activated receptors did not appear to be necessary (Chraibi et al., 1998; Caldwell et al., 2004; but see Bengrine et al., 2007). Since it is possible that nonaqueous lipids or localized protein–protein interactions may transduce the protease signal, direct and indirect mechanisms remain distinct possibilities explaining channel activation by proteases.
There are several lines of evidence suggesting a direct protease interaction with the channel. The large extracellular domain (70%) of the ENaC/DEG family of ion channels suggests that extracellular ligand and protein–protein interactions may play a role in the physiology of these channels. Such interactions have been identified for the degenerin members of the family (Tavernarakis and Driscoll, 1997). In this report we identified a segment in the γ subunit that is involved in elastase-dependent activation of the channel. This segment is in the extracellular domain just after the first membrane-spanning domain (M1). The post-M1 region has been implicated in endogenous furin cleavage of γ ENaC and exogenous trypsin cleavage of ASIC1a, a member of the ENaC superfamily of ion channels (Hughey et al., 2004a; Vukicevic et al., 2006). The idea that the post-M1 region of γ ENaC is accessible is supported by the ability to label cell surface ENaC with antibodies directed against this region (Firsov et al., 1996). Cleavage of ASIC1a changes the pH dependence of channel gating (Poirot et al., 2004). Furin cleavage and trypsin activation of ENaC appear to modulate Na+ dependence of channel gating (Chraibi and Horisberger, 2002; Sheng et al., 2006). The segment identified is therefore in an accessible part of the channel that is involved in channel regulation. A scanning mutagenesis, changing the likely primary amino acid position based on the specificity of NE identified two residues in γ ENaC critical for NE-dependent activation of ENaC. Substitutions at γV182 and γV192 specifically inhibited the activation of ENaC by NE; because neither PE, a similar protease, nor trypsin activation was significantly affected. Since the substitutions did not inhibit activation by PE and trypsin and did not affect endogenous protease activation (implied by the equivalent control currents and aprotinin sensitivity for wild-type and mutant ENaC) the substitutions did not impinge upon a common protease activation pathway. The specificity provides evidence that NE interacts directly with ENaC since interaction of proteases with a different receptor that subsequently activates ENaC would result in significant cross-interaction by virtue of ENaC being the final common step in such a pathway. The results show that one residue, γA190, in the vicinity of the residues critical for NE subserves PE-dependent activation of ENaC. As observed for NE, the effect is also specific since NE and trypsin activated the γA190G mutant channels. Furthermore, in the context of the αβγTh construct the specificity of γA190 for PE was evident since this construct with intact γV182 and γV193 residues but with a γA190G mutation was activated by NE and TH but PE activation was impaired. The presence of a sequence recognized by the thrombin active site creates a gain of function whereby the channel, not naturally activated by thrombin, becomes a thrombin-activated channel. When the thrombin catalytic site is blocked by FPR-chloromethyl ketone, thrombin no longer activates the channel. The substitutions that impair NE- and PE-dependent activation result from changing of residues that are most likely to be good substrates based on the known primary amino acid specificities of the proteases to less preferred residues. Finally, heterologously expressed rat ENaC (Caldwell et al., 2005) and human ENaC (Harris et al., 2007) can be activated by NE which is consistent with the conservation of γV193 between the two species. Moreover, alignment of human, rat, mouse, and rabbit γ ENaC sequences show V182, V193, and A190 are conserved. These observations concur to implicate direct protease–ENaC interactions in channel activation.
Inferences about the nature of the protease–ENaC interaction can be drawn from the protease-activated current trajectories. From the analysis of the progress curves for current activation, the kobs dependence on enzyme concentration suggests that activation requires at least two steps; an initial reversible step and a secondary rate-limiting activation step. If proteases interact directly with the channel, then the first step is C + E↔CE, a rapid equilibrium from the association of channel (C) and enzyme (E). If the reaction is limited to one step where species CE is the activated channel, the current trajectory would contain relaxation rate constants or kobs according to kon[E] + koff, which represents a linear dependence of kobs on enzyme concentration. Since the measured kobs was not linear but hyperbolic, CE is not the activated channel. At least one additional concentration-independent step is rate limiting, which may correspond to one or more possible physical states that represent the activated channel. One parsimonious interpretation is that the rate-limiting step includes hydrolysis. We found that the efficiency parameter kcat/KM for NE is on the order of the kcat/KM for the highest efficiencies of NE hydrolysis of its best extended synthetic substrates (i.e., substrates encompassing amino and carboxyl ends of the scissile bond; Koehl et al., 2003). It is noteworthy that the ratio of kcat/KM for the change of V residues to G residues for NE, and A residue to G residue for PE were also consistent with kcat/KM ratios for the same P1 residue changes in synthetic substrates (Zimmerman and Ashe, 1977; Nakajima et al., 1979; Koehl et al., 2003). In fact, the KM determined for NE is remarkably similar to the dissociation constant for the NE α proteinase inhibitor complex (Mellet et al., 1998).
There are also examples of TH cleavage of its synthetic peptide substrates such as the protease-activated receptor 1 peptides (Vu et al., 1991) and natural substrates such as protein C (De Cristofaro and Landolfi, 1999) with similar kcat/KM as observed for TH activation of INa in the results presented. Hydrolysis is further suggested by the MALDI-TOF mass spectrometry performed on the synthetic peptide, which confirmed that NE cleavage is located between V193 and M194 while PE cleavage is located between A190 and S191 in the γ subunit. Thrombin cleaved the T176-S198 peptide when substituted with a thrombin consensus sequence as well as the in vitro–translated γTh ENaC subunit. Harris et al. (2007) have recently shown that NE can cleave the γ subunit when expressed in Xenopus oocytes.
The results of the kinetic analysis of the progress curves, peptide T176-S198 digestions, and thrombin cleavage of γTh are consistent with hydrolysis-dependent channel activation but cannot exclude other possibilities. These include the possibility that the kcat represents a rate-limiting intramolecular rearrangement occurring long after cleavage, perhaps related to the Na+ self inhibition of the channel (Chraibi and Horisberger, 2002; Sheng et al., 2006) or removal of an inhibitory domain (Carattino et al., 2006; Bruns et al., 2007). Secondarily, protease binding to the channel is sufficient to create a rate-limiting step including binding-dependent prevention of tonic inhibition by an ENaC cofactor or binding-induced conformational changes that directly activate the channel. Experiments so far do not rule out the latter two, as it can be imagined that a binding-induced conformational change is sufficient to activate the channel; and an accompanying hydrolysis ensues but has no significance. That the results are consistent with hydrolysis is in line with other studies showing that ENaC expressed at the apical membrane can be cleaved by furin and this cleavage correlates with channel activity (Hughey et al., 2004b). Also, the extent of channel cleavage correlates inversely with the trypsin activation of current (Knight et al., 2006), suggesting that once cleaved by endogenous proteases, channels may no longer be activated by trypsin. There is however evidence suggesting that the aprotinin-sensitive, trypsin-activated current may not be related to the cleavage patterns observed. In alveolar epithelial cells expressing CAP1, an ENaC-mediated INa sensitive to aprotinin inhibition and trypsin reversal, a presumably cleaved fragment of the α subunit is expressed at the apical membrane (Planes et al., 2005). The abundance of this fragment was not affected by aprotinin. In A6 cells, aprotinin inhibition of INa did not change the relative abundance of the low molecular weight species of the α subunit, nor have any effect on the γ subunit (Adebamiro et al., 2005). Furthermore, it was recently shown that mouse CAP1 lacking a catalytic triad retained the ability to activate ENaC (Andreasen et al., 2006). Therefore, furin-independent catalytic and noncatalytic mechanisms remain likely possibilities in channel activation by proteases. Further experiments are needed to determine if the catalytic activity of the exogenous proteases is required or if binding is sufficient to lead to activation. If cleavage is a prerequisite for channel activation then one must demonstrate the cleaved fragments form active channels.
The results of this study have important implications for ENaC regulation by proteases. First, extracellular proteases can produce a graded Na+ transport response over a short time scale. If activation is irreversible, then eventual activation of all channels should occur over extended periods. From the results, the plateau current elicited by NE, PE, and TH correlates with the rate of activation. The plateau values may result from a steady state between protease activation and channel inactivation. The turnover time for ENaC in the membrane varies considerably from biochemical measurements from a few minutes to several hours (Weisz et al., 2000; Alvarez de la Rosa et al., 2002; Hanwell et al., 2002). However, functional experiments in FRT cells suggest a turnover time for channels of ∼30 min (Snyder, 2000). If the NE- and PE-activated channels in FRT cells in the present study have the same turnover time as channels in the baseline condition, then the steady-state current would depend on the rate of protease activation. By changing the rate of protease activation, up- and down-regulation of epithelial Na+ transport can be achieved within a 30-min window. Alternatively, the extent of protease activation could be altered by reducing the deactivation of channels. Recent studies showed that protease activation of ENaC could be up-regulated by maneuvers known to impair channel internalization (Knight et al., 2006). A graded response may also result from an equilibrium association of channel and protease. This association would not be a Michealis-Menten complex but involve further complex rearrangements typical of tight binding proteinase–inhibitor complexes (Beynon and Bond, 1996). Second, predictably from the turnover times and kcat, the concentration of protease needed to achieve channel activation can be quite low. This is clearly evident for the thrombin case where the half maximal concentration was ∼15 nM. Consequently, inadvertent activation of ENaC in inflammatory states such as chronic obstructive pulmonary disease and cystic fibrosis could result from the increased proteases in the local environment (Stockley, 2000) and may be involved in exacerbation of obstructive airways disease. Third, it is noteworthy that only one region was found to be critical for NE-mediated activation. However, multiple furin cleavage sequences were found to be important in the α subunit and in the γ subunit (Hughey et al., 2004a; Sheng et al., 2006). Because of a lack of strict adherence of NE to valines at the P1 position in its substrate, the valine substitution approach may miss other important P1 positions for NE interaction. Therefore, under aprotinin-treated conditions, it cannot be ruled out that NE/PE interacts also at the α subunit. On the other hand, with TH there is reduced likelihood of such an interaction and the results would suggest that interaction at one site in the γ subunit was sufficient for activation. Perhaps, under the aprotinin-treated conditions the α subunits are already cleaved and do not require further cleavage for channel activation. It is emphasized that these conclusions arise in the context of an aprotinin-inhibited state that does not represent the nascent channel state. As a result, ENaC may be acted upon by intra- and extracellular proteases, possibly in a serial fashion, and the segment investigated in this study may represent the site of action of an endogenous protease at the apical membrane.
We thank Drs. John P. Johnson and Neil Bradbury for many valuable suggestions. FRT cells were a gift from Dr. Peter Snyder (University of Iowa College of Medicine, Iowa City, IA). We thank Dr. Glucksman and Dr. Yuanda for help with MALDI-TOF analysis.
Studies were supported by National Institutes of Health grant RO1 DK-61639, National Research Service Award F31DK015311, National Center for Research Resources S10RR19325, Health Resources and Services Administration grant C76 HF03610-01-00, United Negro College Fund-Merck graduate research fellowship, and Novartis.
Lawrence G. Palmer served as editor.
A. Adebamiro and Y. Cheng contributed equally to this work.
Abbreviations used in this paper: CAP, channel activating protease; ENaC, epithelial sodium channel; FRT, fisher rat thyroid; NE, human neutrophil elastase; PE, porcine pancreatic elastase; TFA, trifluoroacetic acid; TH, human alpha thrombin.