Mechanosensitivity of the Cardiac Muscarinic Potassium Channel

Muscarinic potassium channels are heterotetramers of Kir3.1 and other Kir3 channel subunits and play major roles in regulating membrane excitability in cardiac atrial, neuronal, and neuroendocrine tissues. We report here that rabbit atrial muscarinic potassium channels are rapidly and reversibly inhibited by membrane stretch, possibly serving as a mechanoelectrical feedback pathway. To probe the molecular basis for this phenomenon, we heterologously expressed heteromeric Kir3.1/Kir3.4 channels in Xenopusoocytes and found that they possess similar mechanosensitivity in response to hypo-osmolar stress. This could be attributed in part, if not exclusively, to the Kir3.4 subunit, which reproduced the mechanosensitivity of the heteromeric channel when expressed as a homomeric channel in oocytes. Kir3.4 is the first stretch-inactivated potassium channel to be identified molecularly. Physiologically, this feature may be important in atrial volume-sensing and other responses to stretch.

Muscarinic potassium channels are heterotetramers of Kir3.1 and other Kir3 channel subunits and play major roles in regulating membrane excitability in cardiac atrial, neuronal, and neuroendocrine tissues. We report here that rabbit atrial muscarinic potassium channels are rapidly and reversibly inhibited by membrane stretch, possibly serving as a mechanoelectrical feedback pathway. To probe the molecular basis for this phenomenon, we heterologously expressed heteromeric Kir3.1/Kir3.4 channels in Xenopus oocytes and found that they possess similar mechanosensitivity in response to hypo-osmolar stress. This could be attributed in part, if not exclusively, to the Kir3.4 subunit, which reproduced the mechanosensitivity of the heteromeric channel when expressed as a homomeric channel in oocytes. Kir3.4 is the first stretch-inactivated potassium channel to be identified molecularly. Physiologically, this feature may be important in atrial volume-sensing and other responses to stretch.
Although excitation-contraction coupling is the major mechanism regulating cardiac function, mechanoelectrical feedback plays important modulatory roles (1). Mechanoelectrical feedback is particularly essential in the atria of the heart, which regulate vascular volume through secretion of atrial natriuretic peptides when atrial myocytes are stretched. A number of mechano-sensitive ion channels have been identified in atrial tissue, including stretch-activated potassium, chloride, and nonselective cation channels (2)(3)(4)(5)(6) and stretch-inactivated potassium channels (7). However, the molecular identities of these channels are currently unknown. Since cardiac muscarinic potassium channels (K ACh ) 1 (8,9) regulated by G ␤␥ proteins (10) are preferentially expressed in atrial tissues (11), they seemed likely candidates to examine for mechano-sensitive properties. Moreover, they have been characterized at the molecular level as heterotetramers of Kir3.1 (GIRK1) and Kir3.4 (GIRK4) proteins (12).
cRNA Synthesis and Current Recording from Oocytes-Full-length cDNA encoding the Kir3.4 protein from a rat brain library (confirmed by sequencing) was subcloned into pBlueScript (Stratagene, San Diego, CA), and cRNA was made using standardized in vitro methods (Ambion, Austin, TX). The coding region was subcloned such that the 5Ј end had a Kozak sequence and the 3Ј end had a poly(A) tail. Xenopus laevis oocytes were isolated by enzymatic digestion (2 mg/ml collagenase). Stage IV-V oocytes were used for injection. Current usually became detectable 24 h after injection, and experiments were carried out between 24 and 96 h afterward. Whole-oocyte currents were recorded at room temperature with the two-electrode voltage clamp technique (15) using a Dagan (Minneapolis, MN) CA-1 oocyte clamp amplifier, a TL-1 DMA interface for data acquisition and pCLAMP software (Axon Instruments, Foster City, CA). Recording electrodes were pulled from borosilicate pipette glass (A-M Systems Inc., Seattle, WA) and filled with 3 M KCl. Capacitance and leak currents were subtracted after blocking K ϩ currents with 5 mM BaCl 2 . The standard bath solution contained 98 mM KCl, 2 mM KOH, 1.8 mM CaCl 2 , 1.0 mM MgCl 2 , and 5.0 mM HEPES, pH 7.2. For the experiments measuring reversal potentials, KCl was replaced isotonically by NaCl. Giant cell-attached patches were formed on de-vitellinized oocytes (16,17), and currents were recorded under voltage clamp conditions as described previously (17). The bath and pipette solutions (room temperature) contained 98 mM KCl, 2 mM KOH, 1.8 mM CaCl 2 , 1.0 mM MgCl 2 , and 5.0 mM HEPES, pH 7.2.
Hypo-osmotic Challenge-Currents were first recorded using the two-electrode voltage clamp technique in oocytes superfused with bath solution containing 50 mM KCl and 100 mM sucrose. Hypo-osmotic challenge (50%) was induced by removal of sucrose for 15-30 min. Leak current was subtracted after blocking K ϩ current with 5 mM BaCl 2 .

RESULTS
Isolated rabbit atrial myocytes were patch-clamped in the whole-cell configuration, and whole-cell currents were either recorded at a steady holding potential of Ϫ100 mV (Fig. 1a) or during periodic voltage ramps from Ϫ100 to Ϫ20 mV (0.1 mV/ ms) with Na ϩ and Ca 2ϩ currents blocked (Fig. 1b). After potentiating the inward potassium current by exposing the myocyte to 10 M carbachol, 10 cm of H 2 O positive pressure was applied to the patch pipette to stretch the cell membrane. Positive pressure caused a rapid (within 500 msec) decrease in the carbachol-sensitive current, averaging 15 Ϯ 3% at Ϫ100 mV in 5 myocytes (p Ͻ 0.05) (Fig. 1d). Upon withdrawal of positive pressure, the current recovered rapidly and completely. Both before and after the application of positive pressure, current was fully blocked by 5 mM external Ba 2ϩ . Addi-* Supported by National Institutes of Health Grants RO1 HL36729 and Specialized Center of Research in Sudden Cardiac Death P50 HL52319, by American Heart Association, Greater Los Angeles Affiliate, Grants-in-aid 1110-GI1 and 1126-GI1 and Initial Investigator Award 1136-FI1, and by the Laubisch and Kawata Endowments. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
tionally, positive pressure did not induce any comparable changes in current before carbachol or in the presence of 5 mM extracellular Ba 2ϩ (Fig. 1c), and current-voltage relationship was not shifted, only reduced in amplitude (Fig. 1b). These data suggest that the affected current under these conditions was predominantly I K,ACh and unlikely to be due to an artifact or an endogenous mechano-sensitive current. The inactivation was rapid (within 500 ms), and the current recovered completely and similarly fast upon the removal of positive pressure after 5 s. Decreases in I K,ACh were also observed with application of as little as 2 cm of H 2 O positive pressure but developed more slowly and were not fully reversible.
Since atrial G-protein-regulated potassium channels are known to be heteromeric proteins composed of Kir3.1 and Kir3.4 subunits (12), we determined whether Kir3.1/3.4 currents exhibited mechanosensitivity when heterologously coexpressed in Xenopus oocytes. To maximally activate the channels, Kir3.1 and Kir3.4 were coexpressed with an excess of G ␤␥ subunits (by injecting cRNA for G ␤1 and G ␥2 at a 23:1 excess), which produced a 12-fold increase in current compared with expression of Kir3.1/3.4 without G ␤␥ subunits (data not shown) when measured with the two-electrode voltage clamp tech-nique. To test for mechanosensitivity, we examined the effects of reducing osmotic pressure of the bath perfusate on the amplitude of whole-oocyte Kir3.1/3.4 currents. A 50% reduction of osmotic strength of the bath solution to induce oocyte swelling and membrane stretch reversibly decreased heteromeric Kir3.1/3.4 currents measured at Ϫ100 mV by 18 Ϯ 6% (n ϭ 4) (Fig. 2, a and b). The full decrease took about 10 min to develop, comparable to the time course of osmotic swelling documented in previous studies in oocytes (18) and was reversible over a similar time course upon restoring normal osmolarity (Fig. 2, a  and b). Under both normal and hypo-osmolar conditions, the recorded currents were fully blocked by 5 mM extracellular Ba 2ϩ (data not shown), ruling out artifact from contamination by endogenous mechanosensitive channels previously described in oocytes (19,20). In addition, noninjected oocytes showed no currents of comparable magnitude under normal or hypo-osmolar conditions (data not shown). As additional controls, we also tested the effects of hypo-osmolar challenge in oocytes expressing either Kir1.1 (ROMK1) or Kir2.1 (IRK1) channels. In neither case did the magnitude of current change in response to hypo-osmolar challenge (Fig. 2, c-e), suggesting that the cell swelling-induced inhibition of Kir3.1/Kir3.4 channels is specific to Kir3 versus other families of Kir channels.
To determine which Kir3 subunit was responsible for confer-FIG. 1. Mechanosensitivity of I K,ACh in isolated rabbit atrial myocytes. a, effects of applying 10 cm of H 2 O positive pressure to the patch pipette on whole cell current recorded at Ϫ100 mV in an isolated rabbit atrial myocyte. Positive pressure had no effect on the holding current before exposure to carbachol (Carb) but rapidly and reversibly decreased the carbachol-potentiated current by 15%. Application of external 5 mM Ba 2ϩ completely blocked the current, after which reapplication of positive pressure had no effect. b, superimposed current (carbachol-sensitive)-voltage relations obtained from voltage ramps (0.1 mV/ms) from Ϫ100 to Ϫ20 mV before (Carb) ring mechanosensitivity to the heteromeric channel, we attempted to express homomeric Kir3.1 or Kir3.4 channels. As reported previously (21), expression of Kir3.1 channels with or without G ␤␥ subunits produced only small currents, probably representing heteromeric channels formed by Kir3.1 combining with endogenous Kir3.5 (XIR) subunits present in the oocytes (22). Most previous studies have also reported that Kir3.4 channels form homomeric channels only poorly (11,12,23,24). However, by modifying the 5Ј-and 3Ј-untranslated regions (see "Materials and Methods"), we consistently measured large currents, typically ranging from Ϫ20 to Ϫ40 A at Ϫ100 mV with 100 mM potassium in the bath perfusate from Xenopus oocytes expressing Kir3.4 alone, as assayed by the two-electrode voltage clamp (Fig. 3, a and b). The magnitude of the current increased linearly with the amount of cRNA injected in batches of oocytes from the same frog on the same day (Fig. 3c). The failure of the current magnitude to saturate as well as its large magnitude compared with small endogenous currents in uninjected oocytes (Ϫ0.13 Ϯ 0.23 A at Ϫ100 mV, n ϭ 8) argues strongly against the possibility that Kir3.4 proteins were combining with an endogenous Xenopus protein such as Kir3.5 (XIR) (22) to form heteromeric channels. With 100 mM potassium in the bath solution, the large inward currents showed no or mild relaxation, and outward currents were minimal at potentials positive to the potassium equilibrium potential, typical for strong inward rectifier potassium channels (25) (Fig. 3,  a and b). The current was highly selective for potassium over sodium and blocked in a voltage-dependent manner by extracellular cesium and barium, with K d values at Ϫ60 mV of 61 Ϯ 3 M (n ϭ 4) and 92 Ϯ 13 M (n ϭ 4) respectively (data not shown). Unitary current amplitudes at different voltages obtained from all-points histogram analysis revealed a single channel conductance of 33.2 Ϯ 0.3 picosiemens (n ϭ 4) with 100 mM K ϩ in the pipette solution (Fig. 3d), similar to an earlier estimate (12). In contrast to the previous studies in which single channel openings were flickery and more short-lived (11,12), openings lasting 5-50 ms were commonly observed. Like native K ACh channels (26), currents through homomeric Kir3.4 channels increased by an average of 76 Ϯ 17% at Ϫ100 mV (n ϭ 7) in response to 10 M carbachol when coexpressed with the m 2 muscarinic receptor (Fig. 3, e and f), as has been noted previously (12,24). The effect was maximal within 1-2 min and then gradually lessened, probably due to desensitization. Coexpression of Kir3.4 with G ␤␥ subunits (at 23:1 excess of injected G ␤␥ cRNA) also boosted Kir3.4 currents by an average of Ͼ100-fold compared with oocytes from the same batch injected with the same amount of Kir3.4 cRNA alone (n ϭ 4) (Fig. 3f). These results confirm that homomeric Kir3.4 channels are also classic G-protein-regulated potassium channels.
In attempting to measure unitary currents through homomeric Kir3.4 channels, we noted that channels ran down very rapidly after formation of a cell-attached patch (Ͼ50 patches). This rapid rundown precluded single channel analysis from standard patches (electrode tip, 1-3 m) but could be quantified in giant patches (electrode tip, 20 -30 m), with the mean time constant of rundown averaging 2.3 Ϯ 0.6 min (n ϭ 5 giant patches). Single channel events could often be resolved when only a few active channels were left in the giant patch (Fig. 3d). Patch excision always led to an immediate disappearance of channel activity. Since the formation of a gigaseal (or patch excision) subjects the underlying membrane to considerable mechanical forces, these observations suggested that homomeric Kir3.4 channels might be sensitive to membrane stretch. To test directly for mechanosensitivity, we examined the effects of cell swelling induced by hypo-osmotic challenge on wholeoocyte homomeric Kir3.4 currents measured with the two-electrode voltage clamp. A 50% reduction of osmotic strength of the bath solution reversibly caused a 27 Ϯ 4% reduction in Kir3.4 current at Ϫ100 mV (n ϭ 6) (Fig. 4). This finding indicates that the Kir3.4 subunit is responsible in part, if not exclusively, for conferring mechanosensitivity to heteromeric Kir3.1/3.4 channels and to the cardiac G-protein-regulated potassium channel. Whether the Kir3.1 subunit shares similar mechanosensitive properties is uncertain at this point.
To examine the mechanism underlying the mechanosensitivity of Kir3.4 channels, we investigated whether membrane stretch inhibited Kir3.4 currents indirectly by an effect on G-protein signaling. First, we tested the effects of hypo-osmolar challenge on Kir3.4 currents that had been maximally stimulated with carbachol in oocytes coexpressing Kir3.4 and the m 2 receptor. The carbachol-stimulated Kir3.4 currents demonstrated a similar decrease in response to hypo-osmolar challenge as under basal conditions (averaging 31 Ϯ 4%, n ϭ 4) . Voltage was clamped for 60 ms to membrane potentials from Ϫ100 to ϩ80 mV (20 mV increments) from a holding potential of Ϫ20 mV. The dotted line is zero current level. b, I-V relationship for a. c, current amplitudes (at Ϫ100 mV) as a function of the amount of Kir3.4 cRNA injected into batches of oocytes obtained on the same day (n ϭ 6 -9 oocytes/point). d, single channel currents from a giant cellattached patch at the voltages indicated after rundown left only a few active channels. The single channel conductance was 33 pA. e, stimulation of Kir3.4 current by the muscarinic agonist carbachol in an oocyte coexpressing Kir3.4 and the m 2 muscarinic receptor. The control current (Con) recorded during a voltage ramp (0.1 mV/ms) from Ϫ80 to ϩ40 mV was increased by 158% when the oocyte was exposed to 10 M carbachol (Carb). f, mean Ϯ S.E. of the stimulatory effect of carbachol on whole-oocyte currents at Ϫ80 mV in 7 oocytes expressing Kir3.4 and the m 2 muscarinic receptor before and after exposure to 10 M carbachol (cross-bars) and in oocytes expressing Kir3.4 alone (n ϭ 4) or Kir3.4 with an excess of G ␤␥ subunits (23:1 excess of injected G ␤␥ cRNA, n ϭ 4) (solid bars). The larger basal Kir3.4 current level on the left (cross-bar) versus right (solid bar) was due to the greater amount (1:47 ratio) of Kir3.4 cRNA injected. (Fig. 4d). We further examined the effects of hypo-osmotic challenge in oocytes in which Kir3.4 homomeric channels were coexpressed with an excess of G ␤␥ subunits (23:1 excess of injected G ␤␥ cRNA). Hypo-osmotic challenge decreased current at Ϫ100 mV by 18 Ϯ 3% (n ϭ 5) (Fig. 4d). These findings show that the mechanosensitivity of Kir3.4 currents remains intact over a wide range of ambient G ␤␥ levels, including a presumably saturating range. This observation makes it unlikely that fluctuations in the level of G ␤␥ subunits induced by membrane stretch cause the inhibition of current. DISCUSSION Our results demonstrate for the first time that K ACh channels in the atrium are mechano-sensitive, consistent with their participation in the volume-sensing role of this organ. Physiologically, stretch-induced inactivation of K ACh channels during atrial distension would facilitate membrane depolarization and enhance excitability and could potentially contribute to a variety of stretch-induced responses, including contraction-excitation coupling, atrial natriuretic peptide release, stretch-induced arrhythmias, and/or hypertrophic gene programming.
By demonstrating that both heteromeric Kir3.1/3.4 channels and homomeric Kir3.4 channels exhibit similar mechanosensitivity as native rabbit atrial K ACh channels, we provide the first molecular identity of a mammalian stretch-inactivated potassium channel, which will permit structure-function studies to characterize the molecular mechanisms involved. Interestingly, the predicted overall topological structure of Kir3 channels is similar to nonmammalian mechano-sensitive ion channels cloned from Escherichia coli and Caenorhabditis elegans (27,28), suggesting a common structural motif for these mechano-sensitive ion channels.
The mechanism responsible for mechanosensitivity in these channels is unclear at this point. Our findings argue against a mass action effect of stretch on G ␤␥ subunits as the underlying mechanism, since the mechano-sensitive response remained intact and of comparable magnitude over a wide range of ambient G ␤␥ levels (Fig. 4d). For the case in which oocytes were co-injected with Kir3.4 and G ␤␥ cRNA at a 1:23 ratio, we presume that this includes a saturating range of G ␤␥ subunits relative to Kir3.4 molecules, although the ratio of protein molecules cannot be assumed to be the same as the ratio of cRNA injected. Even if a mass action effect is unlikely, G-protein signaling might be involved if membrane stretch inhibited the ability of G ␤␥ subunits to activate the channels by an allosteric, rather than mass action, effect. By constructing chimeric proteins between Kir3.4 and non-G-protein-regulated Kir proteins, it may be possible to resolve this question. Alternatively, the mechanism of mechanosensitivity may not directly involve G-protein signaling. A recent report has demonstrated inhibition of Kir3 currents by protein kinase C (29). It is possible that membrane stretch-induced activation of phospholipase C (30,31) could in turn activate protein kinase C to inhibit the channels. Finally, a direct interaction between Kir3.4 and cytoskeletal elements or direct sensitivity of the channel to membrane curvature are possible mechanisms for mechanosensitivity (32). Actin has been implicated as a cytoskeletal transducer of mechanical force for other mechano-sensitive channels (33) and has been shown to regulate the function of a number of ion channels such as epithelial sodium channels (34). Also, Kir2 channels such as Kir2.1 have been shown to link to the actin cytoskeleton as a means of spatially localizing them at specific regions in the cell (35), although no similar consensus linkage sites have been identified in Kir3 channels. These actin-binding protein sites do not confer mechanosensitivity to Kir2.1 when expressed in oocytes, however, as shown in Fig. 2d. In preliminary experiments, we were also unable to restore Kir3.4 channel activity in excised giant inside-out patches by adding F-actin to the cytoplasmic surface of the patch. Further studies, perhaps involving chimeric constructs between Kir3.4 and other Kir family members, will be required to unravel the molecular basis for stretch-induced inactivation of these channels.