Regulation of epithelial sodium channels by short actin filaments.

Cytoskeletal elements play an important role in the regulation of ion transport in epithelia. We have studied the effects of actin filaments of different length on the α, β, γ-rENaC (rat epithelial Na+ channel) in planar lipid bilayers. We found the following. 1) Short actin filaments caused a 2-fold decrease in unitary conductance and a 2-fold increase in open probability (Po) of α,β,γ-rENaC. 2) α,β,γ-rENaC could be transiently activated by protein kinase A (PKA) plus ATP in the presence, but not in the absence, of actin. 3) ATP in the presence of actin was also able to induce a transitory activation of α,β,γ-rENaC, although with a shortened time course and with a lower magnitude of change in Po. 4) DNase I, an agent known to prohibit elongation of actin filaments, prevented activation of α,β,γ-rENaC by ATP or PKA plus ATP. 5) Cytochalasin D, added after rundown of α,β,γ-rENaC activity following ATP or PKA plus ATP treatment, produced a second transient activation of α,β,γ-rENaC. 6) Gelsolin, a protein that stabilizes polymerization of actin filaments at certain lengths, evoked a sustained activation of α,β,γ-rENaC at actin/gelsolin ratios of <32:1, with a maximal effect at an actin/gelsolin ratio of 2:1. These results suggest that short actin filaments activate α,β,γ-rENaC. PKA-mediated phosphorylation augments activation of this channel by decreasing the rate of elongation of actin filaments. These results are consistent with the hypothesis that cloned α,β,γ-rENaCs form a core conduction unit of epithelial Na+ channels and that interaction of these channels with other associated proteins, such as short actin filaments, confers regulation to channel activity.

Cantiello and co-workers (15,24) have provided strong evidence that the actin component of the membrane cytoskeleton is involved in the regulation of A6 renal epithelial Na ϩ channels. Specifically, they showed, in patch-clamp experiments of apical membranes of A6 epithelial cells grown on glass coverslips, that short actin filaments were capable of activating Na ϩ channels. These investigators also demonstrated that protein kinase A (PKA)-mediated phosphorylation of actin occurred and that only phosphorylated F-actin could activate the channels (25). They postulated that the effects of actin on Na ϩ channel activity occur either by a direct molecular interaction between actin and the channel protein complex itself or through actin-binding proteins associated with the channel.
Because of the availability of rENaC (26), this study was undertaken to test the hypothesis that short actin filaments could modulate the activity of rENaC and that PKA sensitivity could be conferred onto rENaC, but only in the presence of actin. For this study, we used a reconstitution system in which a single ␣,␤,␥-rENaC could be examined (27). This system affords the opportunity of functionally addressing the interaction between cytoskeletal elements and single Na ϩ channels in a well controlled and well defined environment. We found that short actin filaments decreased the conductance of the channel and increased single channel open probability (P o ). Moreover, exposure of rENaC to either ATP alone or ATP in combination with PKA resulted in a transient stimulation of channel activity, consistent with the effects being mediated by short actin filaments. Functional interactions between ion transport proteins and actin may reflect a novel signal mechanism by which the activity of different transporters may be coordinated.

Materials
Monomeric actin purified from rabbit skeletal muscle or commercially available actin from Sigma was diluted to a final concentration of 4 -10 mg/ml with a buffer containing 2 mM Tris, 0.2 mM CaCl 2 , 0.2 mM MgATP, and 0.2 mM ␤-mercaptoethanol, pH 8.0, and used in the concentrations described for each experiment. The purified actin was a kind gift from Dr. Steven S. Rosenfeld. Actin from either source yielded identical results. This buffer alone had no effect on ␣,␤,␥-rENaC properties at the dilutions used. Recombinant plasma gelsolin was dissolved in 100 mM KCl, 10 mM HEPES, pH 7.4, at a concentration of 10 mg/ml and was nominally calcium-free after dialysis against 0.2 mM EGTA buffer. Cytochalasin D was purchased from Calbiochem. The catalytic subunit of cAMP-dependent protein kinase was purified from bovine heart by Dr. Gail Johnson (University of Alabama at Birmingham) and was added to one or both sides of the bilayer (see experimental conditions described below) to a final concentration of 1.85 ng/ml.

Methods
Planar Lipid Bilayer Experiments-Planar lipid bilayers were made as described earlier (27)  In vitro translation of ␣,␤,␥-rENaC proteins and reconstitution into liposomes were accomplished as described earlier (27); reconstituted proteoliposomes were applied with a fire-polished glass rod to the trans side of a preformed bilayer with the membrane held at Ϫ40 mV. Here and elsewhere throughout this report, applied voltage is referred to the trans chamber, which was connected to the current-to-voltage converter with a feedback resistor of 10 gigaohms and therefore was held at virtual ground. Under these experimental conditions, the channels were oriented with their amiloride-sensitive (extracellular) surface exposed to the trans compartment in Ͼ90% of the incorporations.
Data Analysis-Acquisition and analysis of single channel recordings were performed using pCLAMP software and hardware (Axon Instruments, Inc., Foster City, CA). Data were stored digitally and were filtered at 200 Hz with an 8-pole Bessel filter prior to acquisition at 1 ms/point. All analyses were performed for single active Na ϩ channels. The fact that a given membrane contained only a single Na ϩ channel was ascertained in each experiment by imposing a hydrostatic pressure gradient across the bilayer. Because ENaCs are mechanosensitive (28), all channels resident in the membrane could be activated by this maneuver, including those initially closed. In all experiments performed to date (Ͼ5000) reported here and previously (e.g. Ref. 27), ␣,␤,␥-rENaC displayed three equally spaced subconductive states (13,26, and 39 pS under control conditions and 7, 14, and 21 pS subsequent to addition of actin). The imposition of a hydrostatic pressure gradient always revealed this three-core type of channel activity. We have never observed any of these conductance levels independent of the others; thus, a single rENaC is composed of three concerted gated conductance states (27). Therefore, single channel P o (performed for at least 3 min of continuous recording) was computed using Equation 1: where N is total number of channels (always equal to 1 in these experiments), Ī is the mean current over the period of observation, and i is the main (highest observed) unitary current state determined from all-points current amplitude histograms. The mean current (Ī) over the period of observation was calculated using the events list generated by pCLAMP software and Equation 2: where i m is a current event (all levels, including the zero current level), t m is an event dwell time, and M is the total number of events. Statistical Methods-The probability that the difference between two populations of means was significant was determined by computing the t statistic using Student's t test in one-way analyses of variance. Data are expressed as mean Ϯ S.D. for n experiments.

RESULTS
Effect of Actin on Single ␣,␤,␥-rENaC-The interaction of epithelial Na ϩ channels with cytoskeletal proteins in native epithelia prompted us to examine the hypothesis that the activity of cloned ␣,␤,␥-rENaC may be modulated by actin. Gactin was added to the solution bathing a planar lipid bilayer membrane containing a single ␣,␤,␥-rENaC under conditions that promote rapid polymerization into short filaments (28). Fig. 1A shows typical current records of ␣,␤,␥-rENaC activity at a holding potential of ϩ100 mV prior to and after addition of increasing concentrations of actin. Under control conditions with the membrane bathed with symmetrical 100 mM NaCl (top trace), the channel displayed a major conductance state of 39 Ϯ 2 pS; the channel is composed of three equally spaced subconductance states of 13 Ϯ 1 pS each (dotted lines, levels 1-3; see associated all-points histograms). Mean residence times of the channel, calculated from single exponential fits of dwell-time histograms using an event duration of at least 10 ms, were 121 Ϯ 29, 144 Ϯ 34, 35 Ϯ 21, and 39 Ϯ 18 ms for levels 0 (closed state), 1, 2, and 3 (13-, 26-, and 39-pS open states), respectively. The P o , calculated with respect to the highest (39 pS) conductance level, was 0.11 Ϯ 0.02 (n ϭ 19). These channel properties were significantly altered following addition of at least 0.6 M actin, a concentration critical for polymerization of actin (29 -31), to the solution bathing the cis side of ␣,␤,␥-rENaC-containing bilayers (middle traces). The conductance of each level decreased 2-fold, to 7 Ϯ 1 pS (level 1), 14 Ϯ 1 pS (level 2), and 21 Ϯ 2 pS (level 3). Furthermore, channel P o doubled to 0.21 Ϯ 0.03 (n ϭ 5, p Ͻ 0.001). This increase in single channel P o was due to both an increase in the time spent by the channel in level 1 (7-pS state) and a decrease in the time spent in level 0 (closed state). Mean residence times of ␣,␤,␥-rENaC in its different conductance states in the presence of short actin filaments were 59 Ϯ 12, 1251 Ϯ 247, 44 Ϯ 23, and 157 Ϯ 28 ms for levels 0 -3, respectively. Both mean and unitary current (main state) versus applied voltage plots of ␣,␤,␥-rENaC were linear prior to and after addition of actin (Fig. 1B). Moreover, the amiloride sensitivity of the channel did not change in the presence of actin (K i(amil) ϭ 190 Ϯ 20 nM in the presence of actin, and K i(amil) ϭ 175 Ϯ 15 nM in its absence; n ϭ 3, p Ͼ 0.5), nor did the Na ϩ /K ϩ permeability ratio (10:1). The influence of actin on ␣,␤,␥-rENaC had no pronounced dose dependence, but was rather "all-or-nothing" (Fig. 1A, middle traces). Characteristically, this effect was apparent only when the concentration of G-actin in the cis compartment was Ն0.6 M. Addition of actin to the trans compartment at any concentration (up to 6 M) did not affect channel kinetics in any way (n ϭ 9) (data not shown), nor did addition to both compartments compared with addition to the cis compartment only (n ϭ 7) (data not shown).
ENaCs are mechanosensitive (32), and the imposition of a hydrostatic pressure gradient across the bilayer activated all the channels resident in the membrane. This maneuver was performed at the end of each experiment to provide a reliable assay to determine the total number of channels in the bilayer. In addition, the imposition of a hydrostatic pressure gradient performed in the presence of actin (Fig. 1, bottom trace) addressed two other issues: it demonstrated that, in the presence of actin, all three conductance levels of the channel were still operative, albeit with lowered magnitudes of 7, 14, and 21 pS, respectively, and that the mechanosensitivity properties of ␣,␤,␥-rENaC were still intact. This observation underscores the fact that a single rENaC is composed of these three conductance levels, rather than each level representing the activity of an independent channel.
Short Actin Filaments Activate ␣,␤,␥-rENaC-We next designed experiments to address whether the decrease in conductance of ␣,␤,␥-rENaC is referable to monomeric actin only. Sonication and centrifugation of actin solutions under low ionic strength conditions ensure that only monomeric actin is present (33,34). Addition of this supernatant to the cis compart-ment decreased conductance and increased P o as before ( Fig.  2A, top traces). Cis addition of 100 M MgATP resulted in a transient increase in the open probability of ␣,␤,␥-rENaC. Nonetheless, the presence of ATP did not affect the relative distribution of channel subconductance levels, but rather increased the probability of the channel residing in the higher conductance states. This can be more clearly seen in the associated amplitude histograms. This effect was most pronounced between 10 and 15 min following addition of ATP (third trace). Further recording of ␣,␤,␥-rENaC activity revealed a rundown of channel activity, with almost complete inhibition occurring ϳ60 min after addition of the phosphorylating mixture (fourth trace). If actin was allowed to prepolymerize in vitro prior to its addition to the bilayer solution (100 M MgATP, 100 mM NaCl, and 10 mM MOPS, pH 7.5, for 6 h at room temperature), the single channel conductance of ␣,␤,␥-rENaC still decreased by one-half. However, channel P o did not increase, but rather decreased from 0.11 Ϯ 0.03 to 0.05 Ϯ 0.03 (n ϭ 3, p Ͻ 0.001), similar to that observed following addition of ATP (see above). ATP, in concentrations up to 500 M, had no effect whatsoever on ␣,␤,␥-rENaC properties in the absence of actin. ATP was only effective when present with actin at the cis side of the membrane containing ␣,␤,␥-rENaC. The Na ϩ /K ϩ permeability ratio and the sensitivity of the channel to amiloride, examined either at the peak of activity (by a single drug concentration addition) or after run-down of activity (complete dose-response curves), were not different from those measured in the absence or presence of actin alone (n ϭ 3 for each measurement) (data not shown). These results suggest that actin at any length modifies ␣,␤,␥-rENaC conductance, but that only short actin filaments can increase P o .
It was reported that PKA activated ␣,␤,␥-rENaC following transfection into Madin-Darby canine kidney or 3T3 cells (35), but not in oocytes or planar lipid bilayers. 2 To test the hypothesis that these disparate observations may be reconciled by the involvement of the cytoskeleton (24), experiments were performed when PKA and ATP were added to the bilayer bathing solution(s) in the presence of cis-actin. The catalytic subunit of PKA alone had no influence on channel activity when added to either side of the membrane or to both sides (n ϭ 9) (data not shown). effect was significantly delayed compared with that observed in the presence of ATP alone. Moreover, the maximal extent of activation was ϳ2-fold higher than that in the presence of ATP alone. The P o of ␣,␤,␥-rENaC in the experiment shown was 0.64 at 45 min following addition of ATP. Analogous to the situation with ATP alone, extending the time of recording revealed a rundown of channel activity in the presence of PKA ϩ ATP, although at a slower rate (bottom trace).
To represent the activity of ␣,␤,␥-rENaC throughout the course of these experiments, we have calculated P o for each 3-min period of recording (total of 100 -120 min for each experiment). Because of mixing delays, the calculations were initiated 2 min after each addition of reagents. Fig. 4A depicts the summary time course plots of ␣,␤,␥-rENaC P o following addition of ATP alone, ATP ϩ PKA, ATP or ATP ϩ PKA in the presence of an equimolar concentration of DNase I (an agent known to bind to monomeric G-actin and to prohibit elongation of actin filaments), or ATP or ATP ϩ PKA in the presence of cytochalasin D (an agent known to disassemble long actin filaments). Arrow 1 indicates additions of ATP or PKA ϩ ATP for each experimental design, as described in the legend. DNase I, added 5 min before ATP or PKA ϩ ATP addition, prevented activation of ␣,␤,␥-rENaC. Furthermore, cytochalasin D (5 g/ml final concentration), added after the rundown of ␣,␤,␥-rENaC activity (arrow 2) following ATP or PKA ϩ ATP treatment, produced a second peak of channel activation. The conductance of ␣,␤,␥-rENaC in all these experiments remained unaltered compared with the conductance in the presence of actin, i.e. 7, 14, and 21 pS. Moreover, gelsolin, a Ca 2ϩ -activated actin-severing protein known to stabilize actin filaments at a certain length depending on the actin/gelsolin ratio (37), also evoked activation of ␣,␤,␥-rENaC in a dose-dependent manner (Fig. 4B). This activation was not transient, but sustained. These effects of gelsolin were most pronounced when the actin/ gelsolin ratio was 2:1. The single channel conductance of ␣,␤,␥-rENaC in all these experiments was lowered by one-half with respect to that observed in the absence of actin. These results strongly suggest that the length of actin filaments is critical for the regulation of ␣,␤,␥-rENaC.

DISCUSSION
The actin-based cytoskeleton is a dynamic intracellular structure that plays an essential role in the regulation of cellular events, including the stability of cell shape, the onset of cell motility (38), and the distribution of integral membrane proteins (7,10,39,40). The hypothesis that the actin cytoskeleton is directly involved in the regulation of epithelial Na ϩ channels was first suggested by immunocolocalization studies showing that Na ϩ channels are always in close proximity to actin filaments (15,22). While this interaction may serve as a means to control the spatial distribution of ion channels in the apical membrane of epithelial cells, a requirement in polarized epithelia, the possibility also existed that the colocalization of actin filaments with the apical Na ϩ channels would serve a functional role. Modification of actin filament organization with cytochalasin D, an actin filament disrupter (41)(42)(43), induced apical Na ϩ channel activity in the A6 cell (15). Because DNase I, which inhibits actin polymerization, inhibited the cytochalasin D-induced Na ϩ channel activity, the data suggested that "short" actin filaments might be involved in effect- ing ion channel activation. This hypothesis was further supported by the stimulatory effect of short, but not prepolymerized, actin filaments, which also induced and/or increased Na ϩ channel activity in excised inside-out patches (15). Thus, the formation and length of the actin filaments appeared to be relevant parameters for Na ϩ channel activation. This hypothesis was confirmed by addition of actin and gelsolin at ratios consistent with short oligomers (Ͻ8:1, tri-and tetramers) that induced Na ϩ channel activity in excised patches with no spontaneous activity (15). Actin/gelsolin ratios of Ͼ8:1 were ineffectual in activating channels.
Reconstitution of the ␣,␤,␥-rENaC complex in a lipid bilayer provides the conditions to assess a direct functional role of actin in the regulation of this channel complex. Although spontaneous Na ϩ channel activity was observed in almost all reconstituted preparations, actin immediately modified the conductive properties of the channel, which demonstrated a consistent reduction in the magnitude of its various subconductance states. Although actin was required for Na ϩ channel regulation, its presence was not effective in changing the channel properties prior to reaching a critical concentration for its polymerization (Ͼ0.5 M) (29 -31). Thus, this finding was suggestive of a functional interaction between polymerizing actin, and not G-actin, and the epithelial Na ϩ channel. The addition of polymerizing actin, however, had only a functional effect on decreasing the single channel conductance, but not the kinetic properties of the channel, suggesting that further modulation of actin elongation may be relevant to channel regulation. This hypothesis is supported by our previous data on the effect of actin on Na ϩ channel activity in A6 cells (15) and by the fact that inhibition of F-actin formation by addition of the G-actinbinding protein DNase I prevented the changes caused by the addition of critical concentrations of actin.
Several lines of evidence support the contention that the formation of short actin filaments was relevant to changing rENaC kinetic properties. Addition of PKA plus ATP effected a change in the Na ϩ channel open probability, consistent with previous evidence indicating that PKA phosphorylation of actin slowed its ability to polymerize, thus shifting the equilibrium to that of a longer lived concentration of short actin filaments (24). Subsequent addition of cytochalasin D, an F-actin disrupter that facilitates the formation of short actin filaments that cannot elongate, further increased the change in P o following dissipation of the PKA effect. Addition of ATP alone also changed the P o of the rENaC complex. This effect may result from a change in the "mechanical" properties of the actin filament because the presence of ATP modifies the ability of actin filaments to change conformation from a more "rigid" F-actin⅐ATP to a more "flexible" F-actin⅐ADP complex elicited by a slow, albeit significant, actin-mediated ATP hydrolysis (44). Alternatively, the role of an additional protein kinase in the actin⅐rENaC complex cannot be ruled out.
The most dramatic proof for a role for short actin filaments, but not monomeric or long F-actin, in the regulation of the Na ϩ channel comes from a direct titration of channel function in the presence of actin filaments of various known lengths achieved by stoichiometric ratios of actin to gelsolin. Low actin/gelsolin ratios, consistent with longer actin filaments, were devoid of regulatory function; however, higher ratios (in particular, 1:2-4 actin-gelsolin complexes), had a maximal effect, increasing severalfold the P o of the Na ϩ channel complex. The effect of actin on Na ϩ channel activity had a tendency to run down, as expected from its ability to elongate in the assay solution. Channel activity was once again activated after addition of cytochalasin D, consistent with its disruption of long actin filaments. In contrast, the effect of appropriate actin-gelsolin complexes was stable as predicted by the longevity of these complexes. However, once these complexes were further modified with PKA (data not shown), a rundown of the channel was observed. The results therefore indicate that actin is able to interact with epithelial Na ϩ channels in such a fashion that regulation will initiate and continue until a certain length of F-actin is reached, at which point regulation is no longer feasible. Changes in this equilibrium, as indicated by subsequent addition of cytochalasin D, may return a transient activation.
The literature is replete with examples of a given class of specific ion channel type having widely varying biophysical characteristics (45). In bilayers, purified and reconstituted epithelial Na ϩ channels have conductances ranging from 4 to 160 pS, yet these different conductance channels display comparable Na ϩ /K ϩ selectivity (5:1), amiloride sensitivity (K i ϭ 100 -700 nM), and kinetics (46). Patch-clamp measurements in native cells reveal a single channel conductance of 5 pS (47), while reconstitution of apical membrane vesicles into planar bilayers yields a major amiloride-sensitive Na ϩ channel of 160 pS (48). As first discussed by Palmer (49), several lines of evidence indicate that 5-and 9-pS amiloride-sensitive Na ϩ channels may be interconvertible forms of an identical channel protein.
One of the arguments for this proposal is that epithelial Na ϩ channels purified from A6 cells grown on filter supports, a model system that predominantly expresses a 5-pS channel, expressed a 9-pS channel when they were incorporated into liposomes and patched (50). In addition, mRNA isolated from A6 cells grown on plastic, a condition under which the 9-pS channel is most evident, induces the expression of a highly selective 5-pS Na ϩ channel in Xenopus oocytes (51). Our results demonstrate that single channel conductance and open probability are both strongly influenced by channel interactions with the cytoskeleton. As actin is not present in the bilayer reconstitution system, a channel with a higher conductance and shorter mean open and closed times as compared with the same channel in a native cell will be observed. If, however, actin is present in the presumptive cytoplasmic bathing solution, amiloride-sensitive Na ϩ channels become biophysically comparable to those seen in patch-clamp studies in heterologous expression systems (26,27,52).
A functional interaction between actin filament organization and stretch-activated channels has previously been recognized (53)(54)(55). The fact that rENaC directly interacts with cytoskeletal elements also supports its reported mechanosensitivity (23,32). While the response to hydrostatic pressure was not modified by addition of actin and/or its binding proteins, the actin-induced change in conductance was preserved following activation with hydrostatic pressure. Further modulation of the mechanosensitivity of rENaC by actin filament organization will probably require the sequential interaction of actin filaments with other membrane-associated regulatory pathways, as has been shown for volume regulation and ion channel activity in human melanoma cells (36).
The regulatory role of changes in the actin filament organization induced by the stimulatory effect of PKA on the Na ϩ channel suggests that the mechanism linking this regulatory pathway with the channel proteins requires actin. This is consistent with previous data showing that the cAMP-dependent and PKA-mediated activation of Na ϩ channels in the A6 epithelial cell model requires actin (24). Actin is a suitable substrate for specific PKA-mediated phosphorylation; however, in contrast to its native conformation, phosphorylated actin is a poor substrate for sustaining actin polymerization (24). A mixed pool of phosphorylated and native actin monomers will therefore have a decreased ability to polymerize, resulting in a longer lived pool of short actin filaments, the active substrate for inducing ion channel activation (25). Thus, the data with PKA presented here provide for a uniform interpretation concerning the requirement of actin as the functional interface modulating Na ϩ channel regulation by this kinase. The studies with gelsolin provide unequivocal evidence for the mode by which channel activation occurs, namely, interactions with short actin filaments. In summary, this study represents the first evidence for a direct interaction between actin and epithelial Na ϩ channels.