Phasic Oscillations of Extracellular Potassium (Ko) in Pregnant Rat Myometrium

K-sensitive microelectrodes were used to measure K+ within the extracellular space (Ko) of pregnant rat myometrium. Contractile activity was monitored by measuring either force or bioelectrical signals. Single and double-barreled electrodes were used. Double-barreled electrodes allowed monitoring of electrical activity 15 microns from the site of Ko measurement. From double-barreled electrode experiments, the bioelectrical burst started first, and then Ko began to rise 0.6 ± 0.1 seconds later. This delay indicates that K+ leaves the cells in response to local electrical activity rather than vice versa. Four control experiments were performed to assess the influence of electrical artifacts caused by tissue motion on Ko values. When observed, artifacts were negative and transient, and hence would result in an underestimation of Ko rises. Artifacts were minimized when tissue motion was minimized by fixing the tissue at both ends. At 37°C, 7 single barreled experiments and 45 contractions were analyzed. Resting Ko was within 1 mM of bath K+ (5 mM) at the beginning and end of the experiments. Ko rose during the contraction, fell after the completion of the contraction, and normalized before the next contraction began. Peak Ko values observed during force production were 18.8 ± 5.9 mM, a value high enough to modulate tissue-level electrical activity. Ko required 15.7 ± 2.8 seconds to normalize halfway (t50). Six experiments expressing 38 contractions were performed at 24°C. The contraction period was longer at 24°C. Values for peak Ko (26.2 ± 9.9 mM) and t50 (29.8±16.2 sec) were both larger than at 37°C (p<0.0003 for both). The direct relationships between peak Ko, t50 and the contraction period, suggest elevations in Ko may modulate contraction frequency. The myometrial interstitial space appears to be functionally important, and Ko metabolism may participate in cell-cell interactions.


Introduction
In the 1960s Anderson [1] discovered that tissue-level electrical activity is expressed in myometrial tissue strips. It is now generally accepted that tissue-level contractions are caused by the expression of tissue-level electrical activity and excitation-contraction coupling [2,3]. Over the ensuing decades, investigators elucidated many of the mechanisms of myometrial electrical excitability. A large body of work is now available that details the complex interactions of cell-based systems that are necessary to generate contractions, including a mathematical model of excitationcontraction coupling for myocytes [4]. However, despite a deep understanding of cellular excitability, gaps remain in our understanding of how the electrical mechanisms of the myocyte relate to excitability of the tissue [5]. One interesting hypothesis is that electrical excitability at the tissue-level may in part be regulated by metabolic processes [6]. An example of one such mechanism would be if phasic myometrial contractions caused changes of the ionic composition of the extracellular space.
Potassium in the extracellular space (K o ) can be quantitatively measured using K-sensitive microelectrodes, and changes in K o have been reported in other tissues. Small rises in K o are seen in the brain with photostimulation of the retina [7], but very large rises can be found under pathological conditions [8]. The mechanism of the vascular myogenic response [9] involves elevation of K o [10,11], although other mechanisms likely contribute [12]. Exercise causes a large release of K + from skeletal muscle [13]. In cardiac tissue K o rises by 0.5 to 1.5 mM with each contraction [14] and by 3-4 mM in Purkinje tissue [15], although K o accumulates to much larger values when these tissues are artificially paced faster than K o can normalize. In this work, we will for the first time use K-sensitive electrodes to observe phasic rises in K o in contracting pregnant myometrium.

Pregnant rat myometrium
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Committee on the Ethics of Animal Experiments of the University of Vermont (IACUC protocol #07-055AP). All efforts were made to minimize suffering. Timed pregnant rats were purchased from Charles River and used between d 20 and 21 gestations. Rats were euthanized using pentobarbital and decapitation, and myometrial tissue harvested. Full thickness myometrial strips, ,1.5 mm wide and 1 cm long, were cut parallel with the longitudinal muscle.

Horizontal experimental chamber, experimental conditions
Full thickness myometrial strips were mounted horizontally in an experimental chamber and stretched ,30%. Where indicated, force measurements were performed by securing one end of the tissue to the chamber and the other to a Grass FT-03 force transducer.
Some experiments were performed under conditions designed to minimize tissue movement and reduce motion artifacts. Specifically, the bath solution was not flowed through the chamber (2 ml volume) when the tissue was secured at both ends. Eliminating the force transducer eliminated the movements of the spring arm, but made it impossible to directly measure force.
To determine the timing of contractions without measuring force, we measured bioelectrical activity with contact electrodes (see Measuring AC bioelectrical signals, below). Under no-flow conditions bath solutions were (in mM) 135 NaCl, 5 KCl, 1.8 CaCl 2 , 0.5 MgCl 2 , 10 Na-HEPES, 11 glucose, pH 7.4. Experiments were performed at 3761uC and 24 6 1u, and bath temperatures were monitored using a thermocouple. For experiments at 37uC, temperature was elevated by heating the chamber and preheating the bath solutions prior to exchange. Room temperature experiments were performed without heating.

Measuring DC voltages
Voltages were recorded using one or two A-M systems model 3000 high impedance amplifiers set to the DC mode. A silver chloride-coated wire was placed in the bath and used a reference. The bath ground was also connected to system ground of the head stages of the amplifiers. Signals were filtered at 1 kHz, digitized, stored on a personal computer, and analyzed using AD Instruments Chart 5 software.

Construction of K-sensitive electrodes
K-sensitive electrodes were constructed using methods similar to those of Obrocea and Morris [16]. Single-barreled pipettes were pulled from filamented capillary tubes in a manner similar to pulling patch clamp pipettes. Tip openings of single-barreled electrodes were 1-2 mm and, prior to being made K-sensitive, had impedances of 5-10 MV when filled with 200 mM KCl.
The first step in construction of K-sensitive electrodes was silanization of the pipette tip. Initially, pipettes were filled to the tip with 200 mM KCl. Tips were then dipped in chlorodimethylsilane (Sigma-Aldrich) solution and positive pressure was applied with a syringe, which pushed a drop of the aqueous KCl solution into the organic solvent. Suction was then applied for 3-5 seconds, which brought chlorodimethylsilane into the tip. This process was repeated 3 times, ending with application of positive pressure to expel the organic solvent. Finally, potassium-ionophore 1-cocktail A solution (a valinomycin-based K + sensitive exchanger, Sigma-Aldrich) was introduced into 10 to 100 mm into the tip by application of suction for 3-5 seconds. Impedances of the Ksensitive electrodes were typically 22 to 28 Mohm. Prior to use, all electrodes were stabilized in 5 mM KCl solution for at least 30 minutes, zeroed in 2.5, then tested in 5, 10, 20, and 40 mM KCl solution. Voltage drifts in test solutions were less than 0.5 mV/ minute. For each experiment, the response of each electrode to test solutions was used to convert observed voltage changes to changes in K o .
Some single-barreled electrodes were constructed to serve as negative control electrodes, and were prepared by omitting introduction of potassium-ionophore solution. These ''non-Ksensitive'' electrodes were designed to be unresponsive to changing KCl, which was confirmed by testing in the 2.5 to 20 mM KCl test solutions.
K-sensitive electrode measurements were performed at 24u and 37uC. We anticipated a small (4%) difference of electrode response at these temperatures. However, because we were unable to detect measurable differences in individual electrode responses to test solutions at these two temperatures, we did not correct for temperature changes during data analysis.
Fabrication of double-barreled electrode: one K +sensitive, one K + -insensitive Double-barreled electrodes were made from filamented double capillary tubes (Sutter Instruments, 2BF-150-86-10). After pulling, one capillary tube was then shortened by 1-2 cm by first carefully notching one tube with a glass cutter, then breaking the glass connection between the capillary tubes. The end result was a double-barreled pipette with one back length shorter than the other.
The short electrode was back-filled with bath solution and wrapped with parafilm. The longer electrode was back-filled with 200 mM KCl, and placed in a patch-clamp electrode holder. The pressure/suction procedure described above was then used to silanize and tip-fill the long electrode with potassium-ionophore 1cocktail A solution. The tips were maintained in a beaker containing bath solution until used.

Measuring AC bioelectrical signals
Bioelectrical signals were obtained using contact electrodes as previously reported [17]. Contact electrodes were made from 0.7 mm ID glass capillaries containing a silver chloride-coated silver wire. The electrodes filled with bath solution by capillary action when placed into the chamber and onto the tissue. Tissuelevel bioelectrical signals were monitored by using one of the A-M systems model 3000 high impedance amplifiers in the AC mode. Raw data were recorded using high pass filters set at 1 Hz, low pass set at 1 kHz, and a 60 Hz notch filter. The contact electrodes recorded spike-like bioelectrical signals which were associated with the force-producing phase of each contraction. Even though the bioelectrical spikes occurred at frequencies of 1-2 Hz, the spike widths were ,40 msec (25 Hz). To reduce noise and spurious peaks, raw data were routinely digitally filtered (low pass 30 Hz) for data analysis. Using this frequency window there was minimal loss of fidelity of the bioelectrical signal and the experimental time resolution was ,200 msec. When measuring time intervals between the start of bioelectrical activity and rises in K o , raw data were low pass filtered at no less than 50 Hz to improve time resolution to better than 100 msec.
Using K-sensitive electrodes while simultaneously measuring bioelectrical signals with a separate contact electrode As described above, rat myometrial tissue strips were mounted in the chamber and bioelectrical signals recorded with a contact electrode. Using a separate micromanipulator, a K-sensitive electrode was zeroed in the bath (KCl = 5 mM), then placed into the tissue at a distance of 2-3 mm from the contact electrode.
After initial placement, the K-sensitive electrode typically drifted ,3-4 mV every 10 minutes. The amount of drift was confirmed at the conclusion of each experiment by removing the electrode from the tissue and checking the potential of the bath relative to a standard K + solution. Rather than periodically removing the electrode from the tissue and re-zeroing, corrections for drift were made by measuring the voltage changes using the voltage immediately prior to each contraction as reference. In this manner it was possible to obtain recordings of long duration and many contractions without re-zeroing and repositioning the electrode.
Using double-barreled K-sensitive electrodes while simultaneously measuring force Double barreled electrodes were placed in an electrode holder, and the amplifiers of both electrodes were initially placed in the DC mode. After zeroing the potentials in the bath solution, the electrode was placed into the tissue. In some experiments, local bioelectrical activity was assessed by switching the amplifier of the non-K-sensitive electrode to the AC mode.

Statistics
Average values and standard deviations are reported. Comparisons between data sets were made using the 2-tailed t-test for unpaired or paired data, as appropriate.

Voltage responses of the K-sensitive electrode
After zeroing the K-sensitive electrode in 2.5 mM KCl solution, the electrode was placed in test solutions containing 5, 10, 20 and 40 mM KCl at 24uC. Using the negative terminal of the amplifier probe as bath ground, positive voltage changes were observed with increasing KCl (n = 5; Fig. 1A). Voltage responses were plotted against the logarithm of the ratio of the test solution K + and 2.5 (Fig. 1B). The slope (48.9 mV/decade) is slightly lower than that predicted by the Nernst equation (58.9 mV), yet the linearity indicates the electrodes can quantify K + from 5 to 40 mM. In practice, each electrode was tested prior to use in standard KCl solutions from 2.5 to 20 mM to determine the K + sensitivity. We were unable to find any difference in the responses of individual electrodes between 37u and 24uC (see Figure S1), so we did not correct electrode responses for changes of temperature. Non-Ksensitive electrodes, constructed by omitting the potassiumionophore solution, were insensitive to K + .

Single K-sensitive electrode voltages and bioelectrical signals
In this series of experiments we sought to determine if changes in K o were associated with myometrial contractions. To minimize tissue movement, the tissues were fixed at both ends and experiments were performed under no-flow conditions. Because forces were not measured, bioelectrical activity was used as a reporter for the timing of contractile activity. Bioelectrical activity can be observed with contact electrodes, and is a direct measure of the expression of tissue-level action potentials. We have previously demonstrated that rat myometrium expresses burst-like bioelectrical activity that is closely associated with force generation [17]. Thus, while this technique cannot quantify force production, we were able to precisely determine the beginning and end of each contraction with the first and last spikes of the bioelectrical bursts.
A K-sensitive electrode was placed into the bath, the voltage was zeroed, and then the electrode was placed into the tissue. After several minutes, recorded voltages were 065 mV in all cases, indicating that the K + values at the location of the electrode tip were within ,1 mM of bath K + (5 mM).
With continued monitoring, transient voltage rises of the Ksensitive electrode were observed that were closely associated with burst-like bioelectrical activity (Fig. 2). K o began to rise soon after the start of the bioelectrical activity ( Fig. 2 insert), exhibited a pseudo-stable plateau during the expression of bioelectrical activity, then fell relatively slowly after the end of the burst. Although K o returned to near baseline values before the onset of each contraction, K o elevations extended into the period between contractions. Partly because the voltage changes we observed were so large, and partly because there are no prior reports of K o elevations in myometrium, we performed a series of experiments specifically designed to address the possibility that the observed voltage changes were the result of electrical artifacts rather than rises in K o .
Assessing electrical artifacts due to motion of the tissue Even when held isometrically, myometrial strips move when they contract. As an initial test for artifacts, we recorded signals using non-K-sensitive electrodes (5 experiments; 18 contractions). Non-K-sensitive electrodes were prepared using the silanization procedure, but with omission of the K + -ionophore, cocktail A. As before, tissue strips were fixed at both ends, and bioelectrical signals were recorded with a contact electrode to determine the beginning and end of each contraction. In half the contractions, voltage changes associated with bioelectrical activity were not observed. In the other half, small voltage changes of variable magnitude could be identified (Fig. 3). When observed, the voltage changes were negative, transient, and less than 5 mV. These signals began near the onset of each contraction, and lasted less than 20 seconds. Because the voltage responses of the non-Ksensitive electrodes were always negative, we concluded that artifacts due to motion result in an underestimation of K o rather than an overestimation.

Testing for motion artifacts using double-barreled electrodes and a force transducer
To further test for spurious voltages near the tip of the Ksensitive electrode, we used double-barreled electrodes. Using multiple heat-pull cycles on filamented double capillary tubes, we fabricated double-barreled electrodes with both tips similarly shaped to the single barreled electrode, but separated by 14 to 16 mm (Fig. 4). As detailed in Methods, the back-fill end of one of the capillary tubes was shortened and used as an open electrode, and the longer electrode was made K-sensitive. When filled with bath solution, the impedances of the short electrodes were 3.5 to 5.5 Mohm. The long barreled electrodes were silanized and made sensitive to K + by addition of K + ionophore solution to the tip. The impedances of these electrodes were 24 to 28 Mohm. Both amplifiers were initially set to the DC mode. Because electrical artifacts caused by tissue motion would be detected by the open electrode, we were able to measure force and still be confident that the increased tissue motion inherent to using a force transducer did not introduce unanticipated artifacts.
Three experiments were performed on three different tissues and 19 contractions were analyzed. With each contraction, the non-K-sensitive electrodes displayed negative voltage deflections of varying magnitudes (Fig. 5). The smallest deflection (A) was -3 mV, and the largest (B) transiently approached -20 mV. The   defections closely mirrored force production, and rapidly returned to 0 mV near the end of the contraction. As with the single barreled non-K-sensitive electrodes, no positive voltage deflections were observed from the non-K-sensitive half of the doublebarreled electrode.
The K-sensitive halves of the double-barreled electrodes demonstrated voltage changes similar to those observed using the K-sensitive single barreled electrodes. Rises in K o were closely associated with force production, and K o elevations persisted well into the period between contractions.

Testing for artifacts using tissue stretch
To complete our testing for motion-induced electrical artifacts, we mechanically stretched the tissue using the force transducer and the double-barreled electrode. The purpose of this experiment was to determine if the K-sensitive electrode responds to large tissue displacements in the absence of a contraction. Since stretching myometrium can initiate contractions [18], it was necessary to use the force transducer to demonstrate the experimental stretch did not cause the tissue to contract. The non-K-sensitive electrode was used in the AC mode to monitor for local electrical activity.
After electrode placement, we verified each component was functioning by simultaneously observing force, K o rises, and bioelectrical signals. Then, between contractions, we mechanically stretched the tissue by 1 mm over 2-3 seconds using a micrometer drive (Fig. 6, arrow). From the specifications of the FT-03 transducer, we approximated that during the peak of a contraction the spring of the transducer moved approximately 100 mm. We elected to stretch the tissue 1 mm in order to exceed what the tissue would normally experience and ensure the tissue moved by at least 100 mm along most of its length.
In Figure 6, stretch was not associated with an increase of force or bioelectrical activity, indicating a contraction was not initiated  and no local electrical activity was induced (Fig. 6A, C). However, a very large negative voltage deflection of the K-sensitive electrode was associated with tissue movement (Fig. 6B). The deflection recovered to baseline within 3-5 seconds. Importantly, no large positive voltages were observed in the seconds after the completion of the stretch-induced electrical artifact. Within a minute after the stretch, the tissue spontaneously contracted, as revealed by the force tracing. Contraction-associated rises in K o and bioelectrical activity were observed, confirming correct functioning of the double-barreled electrode. As before, these data indicate that in the absence of a contraction or regional electrical activity, movement of the tissue causes a negative voltage deflection of the K-sensitive electrode. These data confirm that artifacts due to tissue motion do not increase our reported values of K o .

Cause-effect relationship between bioelectrical activity and the onset K o rises
The purpose of this series of experiments is to determine if rises in K o initiates electrical activity, or if electrical activity causes K o rises. In Figure 2 we observed that bioelectrical activity started before the onset of K o rises. To optimize the temporal resolution of the experiment, we minimized tissue motion by performing these experiments with both ends of the tissue fixed and without bath flow. We also measured bioelectrical activity with a separate contact electrode, which tends to further reduce tissue motion. We analyzed 45 contractions from 7 experiments and found bioelectrical activity always preceded the onset of K o rises, with an average interval of 0.9 6 0.5 seconds.
However, those data were obtained using surface electrodes placed 2-3 mm distant from the K-sensitive electrode. Because of this separation, the bioelectrical signals may not have precisely reflected the start of the electrical activity at the tip of the Ksensitive electrode. We therefore analyzed the data obtained using the technique described for Figure 6 (force plus double-barreled electrodes with the non-K-sensitive electrode monitored in the AC mode, and under bath flow conditions). Since the tips of the double-barreled electrode are only ,15 microns apart, the observed bioelectrical activity closely reflects the expression of action potentials at the site K o was monitored.
In three tissues, and 16 contractions, the first spike of the bioelectrical activity always occurred before the beginning of the K o rises. The average time delay was 0.660.1 seconds. These data confirm that the local action potential occurs before K o begins to rise. This directly indicates that rises in K o are caused by expression of electrical activity rather than vice versa. Additionally, the relatively short delay between the start of the bioelectrical activity and the start of the K o rises indicates that K + begins to enter the extracellular space very soon after expression of the local the action potential. Since myocytes are the only reasonable source of K + , this implies that outward K + currents are not restricted to the repolarization phase of the action potential.

Quantifying K o
To convert voltages of the K-sensitive electrode to K o values, we used the equation: R e is the response of the electrode to test solutions, and DmV is the observed voltage change above baseline (defined as the voltage immediately prior to the contraction). Since our initial data indicated resting K o is 561 mM (e.g. the bath K + concentration), we assigned the reference (K ref ) value to 5 mM. This method of converting of DmV to K o compensates for electrode drift that occurs with prolonged recordings, provided K ref remained constant.
To determine if there were long-term changes in K ref over time, we removed the K-sensitive electrode from the tissue after the decay of K o was completed, but before the start of another contraction. This maneuver directly compared K ref in the tissue with the 5 mM K + of the bath. In 5 experiments where the tissue had been studied for at least 60 minutes, the voltage changes observed were near the resolution of the amplifier, and none was more than 3 mV. This indicates that large changes in K ref did not occur, even in long experiments, and equation 1 could be used to quantify K o .

Peak values of K o at 37u and 24uC
In this series, we sought to quantify the magnitude of the K o rises that are associated with contractile activity at two temperatures. To minimize artifacts of tissue movement, we secured both ends of the tissue to the chamber, and did not use bath flow. We used single barreled K-sensitive electrodes, and monitored bioelectrical activity with separate contact electrodes to determine the onset and offset of each tissue-level contraction. K o values were calculated using Equation 1, and peak K o is defined as the maximum value of K o observed during each burst of bioelectrical activity. We performed 7 experiments and analyzed 45 contractions at 37uC (Table 1).
Rat myometrium also expresses spontaneous contractions at room temperature. Except for one experiment where we were able to raise the temperature and maintain the signal from the Ksensitive electrode, different tissues were studied at 24uC. We performed 6 experiments and analyzed 38 contractions at 24uC. At 24uC rises in K o associated with bioelectrical activity were also observed, but peak K o values were significantly greater than at 37uC (26.269.9 vs. 18.865.9 mM; p = 0.0003).

Normalization of K o at 37u and 24uC
As demonstrated in Figures 2 and 5, elevations in K o persist well into the period between contractions. We use the term ''normalization'' to refer to the process where K o returns to ,5 mM before the start of the next contraction. Because rises in K o were not observed after completion of the burst, we used the last spike to mark the beginning of the decay, even if the peak K o did not occur at that time. By using this method of analysis, we were able to eliminate confounding effects due to varying burst durations and minimize the effects of the small irregular fluctuations of K o that were often observed during the peak of the contraction.
To quantify the rate of normalization, we converted K o values to changes in K o (DK o = K o -5), and then plotted DK o vs. time (Fig. 7). In one experiment the data were obtained from a single tissue first at 24uC, then at 37uC, without moving the electrode. At 37uC, the decay of DK o began soon after the end of the burst. At 24uC, a moderate delay was observed prior to the decay. Since the decay at 24uC did not appear to be a first-order mechanism, we approximated the normalization rate of K o by the time it took for DK o to fall to half the DK o value observed at the last spike of the bioelectrical activity (t 50 ). Because t 50 was very sensitive to motion artifacts, we performed these experiments without using bath flow.
In 6 experiments at 24uC and 7 experiments at 37uC, using different tissues, t 50 was more than twice as long at 24uC than at 37uC (unpaired t-test analysis,

Contraction frequency at physiological and reduced temperatures
The period is measured as the time between the first spike of the bioelectrical signal of one contraction and the first spike of the next. Additionally, the bioelectrical signals allow separation of the contraction period into two physiologically distinct periods -the duration of the contraction (burst duration), and the time between contractions (time between the last spike of one contraction and the first of the next).
Rat myometrium spontaneously contracts at 24uC, and raising the temperature shortens the period. However, we are unaware of published reports indicating if the period is shortened by the change of burst duration, change of time between contractions, or both. This series of experiments was performed to address that question, and determine the relationships of peak K o and t 50 to burst duration and the time between contractions.
It is difficult to directly compare contraction frequency and burst duration among tissues because different tissue strips express large variations of contraction patterns. To directly determine the effects of changing temperature, we first obtained stable contractions and recorded bioelectrical activity 24uC. We then increased the temperature to 37uC while continuing to record bioelectrical activity. Data were then compared using the paired t-test.
Three experiments were performed, and 5 contractions were analyzed for each tissue at each temperature. As expected, raising the temperature shortened the contraction period. More importantly, we found that both burst duration and the time between contractions were shorter at 37uC. Interestingly, raising the temperature shortened both parameters in approximately the same proportion (Table 1).

Discussion
The primary finding of this work is that large rises in K o occur during myometrial contractions, and K o elevations persist well into the period between contractions. The second finding is that K o begins to rise ,0.6 seconds after the beginning of the action potential. This indicates two things: 1. Electrical activity causes K o rises rather than vice versa; 2. Tissue excitability may be dynamically modulated by phasic rises in K o .
The values for peak K o we observed are, to our knowledge, the largest reported in any muscle tissue under any conditions. Three factors associated with myometrium likely contribute to this. First, most of the calcium responsible for raising intracellular free calcium originates outside the cell, and then enters the cell via transmembrane calcium currents during the depolarization. This is in contrast to cardiac and skeletal muscle systems, which predominately cycle calcium into and out of intracellular stores. Second, intracellular calcium is highly buffered in myometrium [19]. Therefore, to raise intracellular free calcium, enough calcium has to be brought into the cell to overcome the cytosol's buffer capacity. Third, the myocytes of myometrium are closely packed with narrow restricted spaces between cells. In rodent, the myocytes are closely packed in sheets and in human they are closely packed in bundles. Recently Smith et al [20] confirmed by electron microscopy that the space between human myocytes is no greater than 1 mm.
Putting these factors together, we propose the following mechanism for myometrium: To express a contraction, large inward calcium currents must be generated to overcome the intracellular calcium buffering. In order to maintain the membrane potential at or below 0 mV, the inward calcium current must be balanced by an outward current. The outward currents are predominately K + currents, which move moderate amounts of K + from the inside of the cell into the restricted space between cells. A number of K + channels could participate, such as Kv [21], inwardly rectifying K + (Kir) channels [22], or a variety of Ca 2+ -activated K + channels [23]. Because the volume of the restricted space is very small and K + is not buffered, K o rises.  To assess the feasibility of our proposed mechanism, a semiquantitative assessment can be performed. For a first-order approximation, it is necessary to consider the outward currents, the duration of the currents, and the restricted volume. In isolated myocytes [24] measured at 0 mV, outward currents densities are on the order of 10 mA/cm 2 , and we will consider a 30 second contraction duration. We will not consider diffusion of K o away from the plasma membrane at this time because myometrial myocytes are tightly packed, and we will assume that most myocytes are fully sheathed by the 1 mm restricted space.
Considering a square micron at the surface of a myocyte, a current density of 10 mA/cm 2 lasting 30 seconds will carry a total charge of ,3 picocoulombs/ mm 2 . Since the charge is carried into a 1 mm restricted space, the corresponding volume is 1 mm 3 . Converting coulombs to moles and mm 3 to liters, yields 30 mM. This calculation is only a first-order approximation since it does not consider mechanisms that return K o back into the myocytes (which would reduce peak K o ), but it also does not consider that adjacent myocytes simultaneously move K + into the same restricted space (which would increase peak K o ). However, this assessment does demonstrate that significant rises of K o are feasible.
Because of the novelty of our findings and this being the first application of K-sensitive microelectrodes in myometrium, we investigated the possibility that our results are merely artifacts cause by extraneous conditions, such as tissue motion. Four experimental techniques were used to test for artifacts, all of which indicated that the voltage changes we observed were minimally affected by artifacts. First, we fixed both ends of the tissue to minimize movement, then substituted non-K-sensitive electrodes for K-sensitive electrodes (Fig. 3). This resulted in small negative voltage deflections (,5 mV) that quickly recovered by the end of the contraction. Second, we fabricated double-barreled electrodes (one K-sensitive, one open) and simultaneously recorded two electrical channels with force. One electrode was used for sensing K o and the other used in the DC mode to sense potential changes caused by tissue motion, transmembrane currents, or streaming potentials. The key finding was that voltages measured only a few microns from the tip of the K-sensitive electrode did not mimic rises in K o (Fig. 5). Voltages were again observed to be negativegoing, although they tended to be larger than found with the tissues fixed at both ends. The greater magnitude of these voltages (occasionally approaching220 mV) may be due to the greater motion inherent to using a force transducer.
For a third control experiment, we mechanically stretched the tissue 1 mm and found that the artifacts recorded from the Ksensitive electrode were very large, but also negative (Fig. 6). Importantly, the artifacts produced in this manner rapidly decayed to baseline and did not demonstrate a long-lasting positive voltage signal. Experiments performed at reduced temperature provide the fourth control. Comparing data obtained at 37u and 24uC (Table 1), peak K o was significantly increased and t 50 was significantly lengthened, effects that would be difficult to explain if signals from the K-sensitive electrodes were electrical artifacts.
Lastly, there was the possibility that with each contraction, tissue movement across the electrode tip physically disrupts some cells, which release K + , and by that mechanism causes contraction-associated rises in K o . However, as noted above, after grossly moving the tissue (Fig. 6) the K-sensitive electrode did not display a rebound positive deflection that would suggest damage-induced elevations in K o . Finally, the repeatability of the magnitude of signals from contraction-to-contraction (Fig. 2), and the delay between local electrical activity and the start of the K o signal (,600 msec in the double-barreled experiment), strongly suggest the signals from the K-sensitive electrodes are not the result of tissue disruption.
Taken together, our investigations of motion artifacts indicate that tissue movement likely results in an underestimation of K o rises. Our data also suggest that the magnitude of the electrical artifacts is partly related to the magnitude of tissue movement. There were two possible techniques to compensate for motion artifacts. First, was to minimize tissue motion by securing the tissue at both ends and performing the experiments without bath flow. As demonstrated above, we observed ,5 mV artifact with this technique. Additionally, we could use double-barreled electrodes and subtract the DC signal of the non-K-sensitive electrode from the K-sensitive electrode. The second method introduces additional approximations because it is difficult to match the impedances of the two halves of the double-barreled electrodes, and subtracting signals cannot be made without several assumptions and scaling. Because using the double-barreled electrode yields only a small benefit (if any), quantitative measurements of K o were undertaken by fixing the tissues at both ends and using single K-sensitive electrodes.
In this work we have defined the term ''K o '' to mean the concentration of K + in the extracellular space. Within the extracellular space are the interstitial space, the vascular space, and the perivascular space [25]. For myometrium, there are additional spaces in the connective tissue between muscle layers (rat) or fasciculata [26] (human). Thus, it is feasible that the Ksensitive electrode tip could have been placed in any of these spaces.
The lack of pressurization of the vessels in these experiments, and the relatively small volume of the vascular space compared to cellular space, suggest that the K-sensitive electrodes were routinely placed into the smooth muscle component of the tissue. Since the only reasonable source of K + is from the myocytes, electrode placement in the connective tissue between muscle layers on the rat, or in the spaces between fasciculate in human, would not be expected to produce large signals. Because K o began to rise soon after the start of electrical activity, it is likely that a more specific source of K + was from regions where neighboring myocytes expressed electrical activity -e.g. within layers or bundles. Therefore, it is likely that our observations reflect K o in the interstitial space.
To reduce electrical artifact, some experiments were performed without bath flow. Under these conditions, the tissue could have experienced progressive acidosis or other adverse metabolic events with increasing time. However, under both flow and no-flow conditions, transient rises of K o , were observed over many minutes to hours. In experiments lasting more than one hour, K o between contractions was the same as bath K + . This indicates that K o did not build up in the tissue over time. Taken together, or data suggest that rises of K o were not highly dependent on bath flow. These dynamic changes in K o in myometrium are the first to be directly observed without external pacing in any muscle tissue. To our knowledge, the values for peak K o we report here are also the largest reported in any muscle tissue. The magnitude of these values suggests that K o participates in modulating myometrial contractility during normal function.
While we used the temperature dependence data to argue against the significance of artifacts, the primary purpose behind performing experiments at different temperatures was to investigate the associations among K o , contraction duration, and contraction frequency. Responses of K-sensitive electrodes were unchanged by changing between 24u and 37uC (see Figure S1). Compared with 37uC, larger values of peak K o and slower normalization of K o between contractions were found at 24uC ( Table 1). The greater than 2-fold slowing of rate of K o normalization at reduced temperature suggests the mechanism predominately involves a metabolic process, likely Na + , K + exchange [27]. Temperature-dependent inhibition of Na + , K + exchange may also contribute to peak K o at 24uC being larger than at 37uC.
Raising K o changes the K + Nernst potential, depolarizes the tissue and causes smooth muscle to contract. Through this mechanism, the large rises in K o we observe during the forceproducing phase of the action potential may contribute to lengthening the duration of each contraction.
On the other hand, modest elevations of bath K + (6 to 16 mM) [28] causes relaxation of vascular smooth muscle. Values of K o in this range could activate inward rectifying (Kir) potassium channels [10], which would provide hyperpolarizing currents. In myometrium, elevated K o enhances activity of the electrogenic Na + , K + exchanger [11], which also favors hyperpolarization. Therefore, the modest K o elevations that persisted between myometrial contractions would tend to hyperpolarize the tissue and prolong the time between contractions.
We should emphasize that our data were obtained without external stimulation, and that the K o rises we report here were attributable to phasic contractions that reasonably mimic normal physiological conditions. K o rises likely modulate the function of all the cells that share the same extracellular space. In this sense, K o can be seen as contributing to the syncytial behavior of the tissue through a gap junction-independent mechanism. Cells that reside within the interstitial space, such as Cajal-like cells [29], may also be subject to the influence of changing K o . We propose that K o metabolism provides a novel mechanism for moment-tomoment modulation of phasic myometrial contractions. Figure S1 Responses of K-sensitive electrodes at 246 and 376C. At 24uC, three K-sensitive electrodes were zeroed in 2.5 mM KCl solution, and then tested at 5, 10, 20 and 40 mM. Each electrode was immediately transferred to 2.5 mM KCl solution, re-zeroed, and retested at each KCl concentration at 37uC. (TIF)

Supporting Information
Author Contributions